See discussions, stats, and author profiles for this publication at: [Link]

net/publication/33683303

Design, Microfabrication and Characterization of MHD Pumps and their

Applications in NMR Environments

Article · January 2006

Source: OAI

CITATIONS READS

7 876

1 author:

Alexandra Homsy
HES-SO University of Applied Sciences and Arts Western Switzerland
57 PUBLICATIONS 1,133 CITATIONS

SEE PROFILE

All content following this page was uploaded by Alexandra Homsy on 05 June 2015.

The user has requested enhancement of the downloaded file.

 Design, Microfabrication, and
Characterization of MHD Pumps
and their Applications in NMR
Environments
Dissertation submitted to the faculty of Sciences of the University of
Neuchâtel, in fulfillment of the requirements for the degree of
"Docteur ès Sciences"

by

Alexandra Homsy
Dipl. Ing. Phys. EPFL

Institute of Microtechnology
University of Neuchâtel
Rue Jaquet-Droz 1, CH-2002 Neuchâtel
Switzerland

2006
2
 3

Ce travail est dédié

A mes Parents

A ma soeur
4
Résumé

Contrôler les petits volumes avec une grande précision est d’une impor-
tance cruciale pour la microfluidique et le développement de "systèmes
d’analyse intégrés" (µTAS). Dans le cas des puces microfluidiques intégrant
l’analyse par résonnance magnétique nucléaire (RMN), un débit volumique
faible et précis est même impératif. La RMN est une technique d’analyse
chimique très précise et très populaire qui s’effectue dans un environnement
à très fort champ magnétique (B0 ). Cette thèse vous présente une pompe
à l’échelle microfluidique qui utilise la magnétohydrodynamique (MHD)
pour manipuler les liquides dans un environnement RMN.

Nous allons commencer par expliquer les phénomènes physiques se cachant

derrière la magnétohydrodynamique. Ce phénomène a permis de dévelop-
per une technique nouvelle permettant de bouger des espèces ioniques en
présence à la fois de champs magnétiques et électriques. Les micropom-
pes MHD contrôlent le flux à l’intérieur du système microfluidique en
générant une différence de pression le long du microcanal, proportionnelle
à l’intensité du courant électrique et au champ magnétique. Pomper des
solutions électrolytiques avec la MHD implique la génération de courants
électriques continus (DC) dans de tels environnements. Bien que peu de
recherche ait été conduite à ce sujet, nos études préliminaires sur un éven-
tail de géométries MHD très simples ont conduit à un design de microp-
ompe MHD idéal pour le pompage en mode DC. Le débit volumique de
la micropompe est proportionnel à la densité de courant et à la densité du
flux magnétique. La micropompe génère une haute densité de courant DC

i
ii RÉSUMÉ

à travers son canal principal grâce à un alignement de canaux subsidiaires

(de chaque côté du canal principal) ne faisant que 100 nm de profond. La
génération de bulles dans les canaux due à l’électrolyse a été évitée, les
électrodes responsables du phénomènes étant placées dans des réservoirs
extérieurs en connexion fluidique avec le système.

En partenariat avec le groupe de design de microsystèmes (IMM - LMIS3)

à l’EPFL, nous avons développé un système microfluidique compatible
avec la détection RMN. Nous avons démontré qu’il était possible de dé-
placer avec précision des solutions électrolytiques à l’aide de la magnétohy-
drodynamique dans les environnements RMN. Le débit maximum généré
par notre pompe MHD est de 1.5 µL·min−1 dans l’aimant supraconducteur
RMN.

Mots clés
MicroTAS, Laboratoire sur puce, Microfluidique, Micropompe, Flux mag-
netohydrodynamique, MHD DC, Analyse RMN
Abstract

Precise control of small volumes of sample is of great interest in microflu-

idics and for developing portable, micro total analysis systems (µTAS).
More specifically, low and precise flow rates are imperative for the ma-
nipulation of fluids within an on-chip nuclear magnetic resonance (NMR)
system. NMR is a powerful and popular molecular analysis technique that
takes place in high static magnetic field environments (B0 ). In this thesis,
we present a microfluidics-based micropump using magnetohydrodynam-
ics for the manipulation of fluids in an NMR environment.

The physics behind the magnetohydrodynamics (MHD) phenomenon will

be discussed first. MHD introduces a novel way to move ionic species in a
microfluidic system combining both magnetic and electric fields. Next, the
idea of an MHD-based micropump exploiting B0 to precisely control and
manipulate fluids will be presented. Pumping electrolytic solutions with
MHD involves the generation of an electric current in such environments.
Although not much research has been done on MHD pumping of elec-
trolytic solutions in microchannels, preliminary studies with simple MHD
pump designs has led to an ideal microchannel geometry and DC MHD
micropump. It is shown that MHD flow rate is directly proportional to both
current density and magnetic field intensity. The micropump generates a
high DC current density across the pumping channel with the help of an
array of nanometer deep side-channels. Bubble generation in the channels
due to water electrolysis was avoided by placing the electrodes in outer
reservoirs in fluidic connection with the system.

iii
iv ABSTRACT

In partnership with the microsystems design group (IMM - LMIS3)

at EPFL, we developed a chip-based microfluidic system compatible with
NMR detection. We have been able to demonstrate that micropumping
of electrolytes based on MHD was possible in NMR environments. The
maximum flow rate generated by our MHD pump was 1.5 µL·min−1 in the
NMR superconductive magnet.

Keywords
MicroTAS, Lab on a chip, Microfluidics, Micropump, Magnetohydrody-
namic flow, DC MHD, NMR analysis
Contents

Résumé i

Abstract iii

1 Introduction 1
1.1 Micro total analysis systems (µTAS) . . . . . . . . . . . . 1
1.2 Microfluidics . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Micropumps . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Fabrication methods . . . . . . . . . . . . . . . . . . . . . 5
1.5 Nuclear Magnetic Resonance (NMR) . . . . . . . . . . . 6
1.6 Research goals of this thesis . . . . . . . . . . . . . . . . 9

2 MHD pumping of electrolytes 19

2.1 The origins of magnetohydrodynamics (MHD) . . . . . . 19
2.2 MHD theory . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.1 From Maxwell’s equations to the Lorentz Force . . 20
2.2.2 Application to poor conductors . . . . . . . . . . 21
2.3 Pumping electrolytes with MHD in the DC mode . . . . . 25
2.3.1 High DC current generation in electrolytes . . . . 25
2.3.2 Ion transport and electrokinetic flows . . . . . . . 28
2.4 State of the art in MHD micropumps . . . . . . . . . . . . 33
2.5 Advantages and limitations of MHD pumping . . . . . . . 35

v
vi CONTENTS

3 DC MHD pump developments 41

3.1 Experimental conditions . . . . . . . . . . . . . . . . . . 42
3.1.1 Electrolyte properties . . . . . . . . . . . . . . . . 42
3.1.2 Magnet properties . . . . . . . . . . . . . . . . . 43
3.1.3 Experimental methods . . . . . . . . . . . . . . . 43
3.2 MHD in a PDMS T-Channel . . . . . . . . . . . . . . . . 45
3.3 PDMS channel on top of platinum electrodes . . . . . . . 51
3.4 Glass channel with electrodes on the inner wall . . . . . . 54
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4 A high current density DC MHD micropump 65

4.1 New design concept . . . . . . . . . . . . . . . . . . . . . 66
4.2 Experimental Methods . . . . . . . . . . . . . . . . . . . 68
4.2.1 Micropump fabrication . . . . . . . . . . . . . . 68
4.2.2 Pumping experiments . . . . . . . . . . . . . . . 70
4.3 Results and discussion . . . . . . . . . . . . . . . . . . . 70
4.3.1 Flow limitation due to back pressure . . . . . . . . 74
4.3.2 Joule heating and power dissipation . . . . . . . . 74
4.3.3 MHD flow dependence on buffer ionic strength and
pH . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 78

5 Electrical description of the pump 81

5.1 Model description . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Resistance measurements . . . . . . . . . . . . . . . . . . 86
5.3 Flow observations . . . . . . . . . . . . . . . . . . . . . . 88
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6 MHD pumping in NMR environments 95

6.1 Experimental methods . . . . . . . . . . . . . . . . . . . 96
6.1.1 Micropump fabrication . . . . . . . . . . . . . . . 96
6.1.2 Experimental setup . . . . . . . . . . . . . . . . . 98
6.1.3 Flow measurement methods . . . . . . . . . . . . 102
6.2 Results and discussion . . . . . . . . . . . . . . . . . . . 105
6.2.1 Electrical characterization . . . . . . . . . . . . . 105
CONTENTS vii

6.2.2 Characterization of the new flow rate measurement

method . . . . . . . . . . . . . . . . . . . . . . . 109
6.2.3 Measurements with permanent magnets . . . . . . 112
6.2.4 Measurements in NMR superconductive magnet . 114
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 117

7 Integration of NMR detection 121

7.1 Fabrication of microchannels compatible with NMR . . . 121
7.2 Combining MHD pumping with NMR detection . . . . . . 122
7.3 Perspectives on MHD-NMR applications . . . . . . . . . 129
7.3.1 Chemical reaction monitoring . . . . . . . . . . . 129
7.3.2 MHD flow monitoring by MRI . . . . . . . . . . . 130
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 130

8 Summary and Outlook 135

Appendix 141

A Glossary 141

B MHD pump details 145

C Classification of Magnetic Materials 149

D Citrate-Phosphate buffer preparation 153

Acknowledgements 155

Publications 159
viii CONTENTS
List of Figures

2.1 Flow profiles at low and large Hartmann numbers. . . . . . 22

2.2 Parallel magnetic Lorenz forces on cations and anions mi-
grating towards the negative and the positive electrodes, re-
spectively. . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.1 Geometrical distribution of magnetic field . . . . . . . . . 44

3.2 View of the experimental setup . . . . . . . . . . . . . . . 46
3.3 Principle of use of the first MHD design . . . . . . . . . . 47
3.4 Measured current as function of channel width and applied
voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5 Flow directions in presence of electric and magnetic field . 50
3.6 Principle of use of the second MHD design . . . . . . . . 52
3.7 Measured current as function of electrolyte nature and ap-
plied voltage . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.8 Top view of a glass channel with the Pt electrodes on the
walls of the channel . . . . . . . . . . . . . . . . . . . . . 55
3.9 Schematic overview of the fabrication of the third genera-
tion device . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.10 eSEM pictures of a channel cross-section . . . . . . . . . 58
3.11 Current frequency spectra . . . . . . . . . . . . . . . . . . 59

4.1 Schematic diagram comparing electrode location between

previous designs and new concept . . . . . . . . . . . . . 67
4.2 Schematic cross-sectional view of the pumping area. . . . 69

ix
x LIST OF FIGURES

4.3 View of the chip with PDMS on top for the inlet/outlet
reservoirs as well as electrode reservoirs. . . . . . . . . . . 71
4.4 Maximum velocity of 6 µm carboxylate-modified polystyrene
beads as a function of applied current density . . . . . . . 73
4.5 Current-Voltage behavior of the chip during the velocity ex-
periments depicted in Fig. 4.4. . . . . . . . . . . . . . . . 76

5.1 Description of pump electrical model . . . . . . . . . . . . 83

5.2 Measured current as function of buffer ionic strength . . . 87
5.3 Vectors representing eo, ep and MHD movement. . . . . . 91

6.1 Microfabrication steps for the micropump . . . . . . . . . 97

6.2 Three dimensional drawing of the MHD pump . . . . . . . 99
6.3 View of the encapsulated chip . . . . . . . . . . . . . . . 100
6.4 View of the chip holder for measurement on top of perma-
nent magnets . . . . . . . . . . . . . . . . . . . . . . . . 103
6.5 View of the chip holder for measurement inside the bulky
superconductive magnet . . . . . . . . . . . . . . . . . . 104
6.6 Current-voltage curves of the new chip . . . . . . . . . . . 107
6.7 Evolution of the plug position . . . . . . . . . . . . . . . 110
6.8 MHD Flow mean velocity as a function of current density . 112
6.9 MHD Flow rate as a function of J . . . . . . . . . . . . . . 115

7.1 Cross-sectional eSEM view of channels of symmetric cross-

section . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.2 Planar NMR probe integrated on top of microchannels . . 123
7.3 View of the chip holder for NMR measurements inside the
bulky superconductive magnet . . . . . . . . . . . . . . . 125
7.4 NMR spectrum of anisole . . . . . . . . . . . . . . . . . . 126

B.1 Three dimensional drawing of the MHD pump . . . . . . . 145

B.2 Three dimensional drawing of the holder . . . . . . . . . . 146
B.3 Technical drawing of the top part of holder with detailed
dimensions and views. . . . . . . . . . . . . . . . . . . . 147
LIST OF FIGURES xi

B.4 Technical drawing of the bottom part of holder with de-

tailed dimensions and views. . . . . . . . . . . . . . . . . 148

C.1 Periodic table of elements . . . . . . . . . . . . . . . . . . 152

D.1 Table for the preparation of the constant ionic strength buffer.
This table was taken from P.J. Elving et al. . . . . . . . . 154
xii LIST OF FIGURES
Chapter 1

Introduction

A µTAS (micro total analysis system) is defined as a platform, not bigger

than a compact disc, which contains all the components necessary to per-
form the complete analysis of any raw chemical species. In this chapter, we
will introduce the concept of µTAS, and explain why there has been many
recent advances in the field. The transport mechanisms (namely microflu-
idics) involved in those systems will be described as well as the technology
necessary to build such small and complex devices. The last paragraph
of this chapter will be devoted to the scope of the thesis, which was to
build a µTAS for NMR (nuclear magnetic resonance) analysis of chemi-
cal/biological samples.

1.1 Micro total analysis systems (µTAS)

Chemical analysis are affected by the volumes used, in the sense that smaller
volumes will often decrease the response time of an analysis as diffusion
times are shortened, mixing is improved, and reaction conditions can be
better controlled. Over the past decade, a need for miniaturization in lab-
oratories for low cost, fast and sensitive measurements has developed in
many fields, including chemical production, DNA analysis, drug discovery,
and environmental analysis [1–6].

1
2 CHAPTER 1. INTRODUCTION

The concept of µTAS was first introduced by Andreas Manz and H.

Michael Widmer [7–10]. The area of µTAS is growing rapidly, and the
idea is to place all the functions carried out in a laboratory room onto a
chip (Lab-on-a-chip). Here, all the components necessary for the analy-
sis are assembled into a small device, which performs the whole analysis
process in an autonomous fashion. However, instead of assembling the sys-
tem from discreet components, it is integrated onto a planar, microchip-like
platform [11–17]. Those microchips are made of networks of microchan-
nels with dimensions in the range from hundreds down to few micrometers
(see section 1.2), which are fabricated in substrates like glass, quartz, sil-
icon, or polymers [18–21]. Miniaturization firstly leads to the reduction
of reagent consumption, greater separation efficiencies, and reduced anal-
ysis time. On the other hand, the applicability of µTAS for biochemical
analysis is also determined by considerations such as improved analytical
performance, component integration, increased throughput and automation.

1.2 Microfluidics
Microfluidics refers to devices and methods for controlling and manipulat-
ing fluid flow with length scales less than a millimeter. The first microflu-
idic devices fabricated using microelectronics technology were investigated
in the 70’s [22–25]. Microfluidic devices are used for various applications
such as active/passive valves and micropumps [26, 27], reactors and mix-
ers [28–30], inkjet [31] and pressure flow sensors [32].

In microfluidic systems, the theory of fluid mechanics for macroscopic

pipes can be used to calculate liquid flow parameters such as the pres-
sure drop. However, due to the dimensions of the microsystems, not all
of macroscopic fluid mechanics can be applied to microfluidic systems
[33–36]. Because of its viscous properties, the flow of a fluid inside a
channel will be classified either as laminar or turbulent [37, 38]. This will
depend on various parameters such as the physical properties, flow rate and
geometry of the system. In a laminar flow, the streamlines never intersect.
In a turbulent flow, velocity and pressure show random variation with line
1.2. MICROFLUIDICS 3

and space coordinates. The transition to turbulence is when the flow ceases
to be laminar to become turbulent.

The parameter which describes the transition is the Reynolds number,

Re, which is defined by:

vDh ρ
Re = (1.1)
µ

where v is the average stream velocity [m·s−1 ], ρ is the fluid density [kg·m−3 ],
µ is the absolute viscosity [Pa·s], and Dh is the hydraulic diameter. Dh is
equal to 4A/Pw where A is the cross sectional area of the channel and Pw is
the wetted perimeter of the channel.

Re can be regarded as a ratio of the inertial force to the viscous forces and
is a dimensionless parameter. To determine whether a given flow is lami-
nar or turbulent in macroscopic systems, the Re is evaluated and compared
to the transitional Reynolds number, Ret , 2300 [39]. In the early 90’s, it
was demonstrated that the Ret for macroscopic pipe systems cannot always
be used for microfluidic systems to determine the flow regime [33]. How-
ever, alternative values or the transitional number were introduced which
depend on the ratio between the length of the channel, L, and the hydraulic
diameter. According to [33], for 1 < L/Dh < 70 the transition number is
estimated as:

L
Ret  30 , 1 < L/Dh < 70 (1.2)
Dh

Due to the small dimensions of microchannels, the Re is usually much less

than 100, often less than 1. In this Re regime, flow is completely laminar.
The transition to turbulent flow generally occurs at Re larger than 2000.
Laminar flow provides a means by which molecules can be transported in a
relatively predictable manner through microchannels. We will see in details
in section 2.2.2 how this property of flow will simplify our description of
the pressure drop in microchannels.
4 CHAPTER 1. INTRODUCTION

Table 1.1: Performance comparison for a selection of micropumps. Qmax is

the value of maximum flow rate achieved by such pumps.
Category Driver Working Fluid Qmax [µL·min−1 ] Ref.
Displacement Glass diaphragm Water 8 [40]
Displacement PDMS diaphragm Water 0.1 [41]
Dynamic EHD Ethanol 14·103 [42]
Dynamic EHD Water 2 [43]
Dynamic EO Tetraborate Buffer 0.2 [44]
Dynamic EO Borate Buffer 33·103 [45]
Dynamic AC MHD 1M Saline solution 18 [46]
Dynamic DC MHD Seawater 63 [47]
Dynamic Accoustic Water 0.255 [48]

1.3 Micropumps
In order to pump and precisely control small volumes of sample, microp-
umps need to be integrated into µTAS. A complete review about microp-
umps in microscale devices was recently published by D.J. Laser and J.G.
Santiago [27]. Micro-fabricated pumps are generally falling into one of
two main categories. The displacement pumps exert pressure forces on the
working fluid through one or more moving boundaries. The dynamic pumps
continuously add energy to the working fluid in a manner that directly in-
creases either its momentum or its pressure. Table 1.1 compares the perfor-
mance of a selected sample of micropumps. The magnitude of the gener-
ated flow rates is independent of the category, and vary from nL·min−1 to
mL·min−1 .

Displacement pumps
Most of the displacement pumps operate in a periodic manner, incorporat-
ing some means of rectifying periodic fluid motion to produce net flow.
Such periodic displacement pumps can be further broken down into pumps
that are based on reciprocating motion, as of a piston or a diaphragm [41,
1.4. FABRICATION METHODS 5

49], and pumps that are based on rotary elements such as gears or valves.
The majority are reciprocating displacement pumps [40] in which the mov-
ing surface is a diaphragm. These are sometimes called membrane pumps
or diaphragm pumps. Another subcategory of displacement pumps are
aperiodic displacement pumps, the operation of which does not inherently
depend on periodic movement of the pressure-exerting boundary. Aperi-
odic displacement pumps typically pump only a limited volume of working
fluid; a syringe pump is a common macroscale example.

Dynamic pumps
Dynamic pumps are based on a direct transformation of non-mechanical or
mechanical energy into a continuous fluid movement. A first example are
the centrifugal pumps [50]. Those are typically ineffective at low Reynolds
numbers and have only been miniaturized to a limited extent. Electro-
hydrodynamic (EHD) [42, 51], electro-osmotic (EO) [52–57], and mag-
netohydrodynamic pumps (MHD) [46, 47] use pumping mechanisms in
which an electromagnetic field interacts directly with the working fluid [3]
to produce pressure and flow. Another example is the acoustic-wave mi-
cropump [48].

1.4 Fabrication methods for microfluidic devices

Microfabrication [58] comprises the use of a set of manufacturing tools
based on batch processing of thin and thick film commonly used in the elec-
tronics industry. The majority of microfluidic devices presented to date rely
on such planar fabrication technology which is based on three processes:
layer formation, pattern definition and layer removal [59]. Such processes
are very critical and necessitate to be handled in a controlled dust-free en-
vironment, called cleanroom. At the beginning of microfluidics research
in the early 1990s, these techniques were readily available in the field of
microfabrication, and were therefore directly adopted for the fabrication of
microchannels. Substrates of choice were mono-crystalline silicon, low-
alkali glass and fused silica in the form of polished wafers. The wafer sur-
6 CHAPTER 1. INTRODUCTION

face is covered with an inert material layer, which is subsequently patterned

by photolithography. The underlying bulk material can then be etched to
form channels, reaction chambers, filters, and the like. After removal of
the protective coating used during etching, the still-open microchannels are
sealed with a coverplate to form a closed channel network. In most cases,
the coverplate has access holes that match up with the ends of the channels
to allow filling of the networks with reagents. Since the mid-90s, poly-
mer microfluidic devices have received increasing attention, both because
of their low cost and simple fabrication. The development has been driven
by the need for inexpensive, rapid prototyping, and disposable microchips
for mass markets. On the research end, the quest for rapid-prototyping
techniques for laboratory experiments has driven this development. Poly-
mer microfabrication by hot embossing or injection molding allows mass
production using a variety of materials with highest precision [60]. Casting
and molding of elastomers, especially poly(dimethylsiloxane) (PDMS), has
been welcomed by the research community, as it allows rapid replication
of microfluidic devices without the requirement of a semiconductor-style
cleanroom and hazardous chemistry [61].

1.5 Nuclear Magnetic Resonance (NMR)

The Nuclear Magnetic Resonance (NMR) phenomenon, discovered in 1946,
became a standard spectroscopy method in chemistry, physics, and biology.
In the early 1970s this physical phenomenon was also used to develop an
imaging method for medicine. The aim of this section is to briefly describe
the basic phenomena that form the basis of NMR spectroscopy and imag-
ing to help the reader understand the future applications of our µTAS. In
addition to this section, the reader is strongly advised to have a look at
references [62] and [63].

NMR Spectroscopy
NMR spectrometry exploits the fundamental property of the atomic nucleus
that is the nuclear spin In . The nuclear magnetic moment m [A·m2 ] of a
1.5. NUCLEAR MAGNETIC RESONANCE (NMR) 7

nucleus is related to its angular momentum L [J·s−1 ] by

m = γL (1.3)

The ratio γ is called the gyromagnetic ratio [rad·T−1 ·s−1 ] and has a dif-
ferent value for each type of nucleus. Due to their natural abundance and
large γ, 1 H nuclei (In = 1/2) are extensively used in NMR spectroscopy
experiments.

When placed in a static magnetic field B0 , 1 H nuclei can adopt one of

two spin states corresponding to two possible orientations of the magnetic
moment m (parallel or anti-parallel to B0 ). These two states are separated
by an energy barrier
hγB0
∆E = (1.4)
2π
For detection of this transition energy, radiation given by ∆E = hν must be
applied, and combining this with Equ. 1.4 gives the fundamental resonance
condition for all NMR experiments:

γB0
ν0 = (1.5)
2π

The motion of spins in an externally applied magnetic field B0 is a pre-

cession of their magnetic moment m around B0 at an angular frequency
ω0 = −γB0 called the Larmor frequency [s−1 ]. The Larmor frequency cor-
responds to the resonance frequency of Equ. 1.5. Magnetic moments m
of nuclei add up to a macroscopic magnetization M [A·m−1 ]. At equilib-
rium, the population difference between spin states corresponds to a bulk
magnetization M0 aligned along B0 .

A very small alternating magnetic field with angular frequency ω = ω0

is capable of significantly altering the direction of magnetization. This ef-
fect is the Nuclear Magnetic Resonance (NMR) phenomenon. The excita-
tion of the spins is given by an applied alternating field B1x (t) = 2B1 cos(ωt)
perpendicular to B0 at frequency ω = ω0 . The field B1 causes a precession
8 CHAPTER 1. INTRODUCTION

of the magnetization so that after time t, M is at an angle θ with respect to

B0 given by:
θ(t) = γB1t (1.6)

After a pulse duration τ, the magnetization will return to its equilibrium

position aligned along B0 . The NMR signal is obtained by measuring the
voltage induced in the detection coil by the rotating magnetization while
it returns to its equilibrium position along the axis of B0 . This signal is
measured in the time domain as an oscillating, decaying voltage induced
in the receiver coil by the magnetization in free precession. This signal
is known as the Free Induction Decay (FID), and has a maximum initial
amplitude
S0 = ω0 M0 B1Vs (1.7)

where Vs is the sample volume with uniformly distributed B1 (⊥ B0 ) and

M0 (same direction as B0 ). In most NMR experiments, the time-domain
FID is transposed in the frequency domain by way of a Fourier transform.
Each peak in the frequency domain will refer to a specific property of the
nucleus and its environment, and its position is known as the chemical shift
δ (in ppm). The chemical shift is a molecular parameter, and depends only
on the sample conditions (solvent, concentration, temperature) and not on
the spectrometer frequency.

Magnetic Resonance Imaging (MRI)

The imaging technique by magnetic resonance adds spatial encoding to the

signal already described for NMR spectroscopy. The spatial encoding of
the sample volume is done by the superposition of a linear magnetic field
gradient G to the static magnetic field B0 . Nuclear spins at position r will
hence resonate at the local Larmor frequency

ω(r) = γ(B0 + G · r) (1.8)

The signal contribution from spins occupying a small element of the sample
volume dV at position r is proportional to the local spin density ρspin (r).
1.6. RESEARCH GOALS OF THIS THESIS 9

The MRI image is obtained by a combination of gradient pulses, RF excita-

tion pulses and data acquisition periods, which together with the order and
timing they are applied is known as the pulse sequence.

Advantages of NMR analysis

NMR is one of the most powerful chemical analysis techniques. It enables,
for example, the identification of components of complex pharmaceutical
and natural product mixtures and may also be important in studies of racem-
ization of drug metabolites, a process that often occurs in vivo. The largest
challenge to NMR methodology lies in its sensitivity. Compared with other
techniques, such as mass spectroscopy, NMR methods for screening mix-
tures are limited by their relative insensitivity. A boost in sensitivity and
screening capacity of the NMR technique is possible by the integration of
microcoils in microfluidic devices.

1.6 Research goals of this thesis: an NMR µTAS

This thesis is part of a common FNS project with the group of Prof. Popovic
(EPFL-IMM-LMIS3), which is specialized in integrated magnetic sensors.
The goal of this collaboration is to explore the integration of Nuclear Mag-
netic Resonance (NMR) spectroscopy in nL volumes with a microfluidic
system for chemical applications. Activities at the EPFL - LMIS focused on
the modeling, design, fabrication and evaluation of micro-fabricated probes
for NMR detection in µTAS. Our activities at UNINE - SAMLAB focused
on the fabrication and optimization of a microfluidic chip-based system
compatible with NMR detection.

The research goals of this thesis are to fabricate a microfluidic chip

compatible with NMR, to take advantage of the high magnetic field inher-
ent to NMR spectroscopy to pump fluids, and then to choose a chemical
demonstrator to evaluate the integration of NMR with the fluidic microsys-
tem.
10 CHAPTER 1. INTRODUCTION
Bibliography

[1] S.C. Jakeway, A.J. de Mello, and E.L. Russell. Miniaturized total
analysis systems for biological analysis. Fresenius Journal of Analyt-
ical Chemistry, 366:525–539, 2000.

[2] D. Figeys and D. Pinto. Lab-on-a-chip: A revolution in biological and

medical sciences. Analytical Chemistry, 72:330A–335A, 2000.

[3] G.J.M. Bruin. Recent developments in electrokinetically driven analy-

sis on microfabricated devices. Electrophoresis, 21:3931–3951, 2000.

[4] J. Khandurina and A. Guttman. Bioanalysis in microfluidic devices.

Journal of Chromatography A, 943:159–183, 2002.

[5] E. Verpoorte. Microfluidic chips for clinical and forensic analysis.

Electrophoresis, 23:677–712, 2002.

[6] B.H. Weigl, R.L. Bardell, and C.R. Cabrera. Lab-on-a-chip for drug
development. Advanced Drug Delivery Reviews, 55:349–377, 2003.

[7] A. Manz, N. Graber, and H.M. Widmer. Miniaturized total chemical-

analysis systems - a novel concept for chemical sensing. Sensors and
Actuators B, 1:244–248, 1990.

[8] D.R. Reyes, D. Iossifidis, P.-A. Auroux, and A. Manz. Micro Total
Analysis Systems. 1. Introduction, Theory, and Technology. Analyti-
cal Chemistry, 74:2623–2636, 2002.

11
12 BIBLIOGRAPHY

[9] P.-A. Auroux, D. Iossifidis, D.R. Reyes, and A. Manz. Micro Total
Analysis Systems. 2. Analytical standard operations and Applications.
Analytical Chemistry, 74:2637–2652, 2002.

[10] T. Vilkner, D Janasek, and A. Manz. Micro Total Analysis Systems.

Recent developments. Analytical Chemistry, 76:3373–3386, 2004.

[11] A. Manz, D.J. Harrison, E.M.J. Verpoorte, J.C. Fettiger, A. Paulus,

H. Pauli, and H.M. Widmer. Planar chips technology for miniaturiza-
tion and integration of separation techniques into monitoring systems
- capillary electrophoresis on a chip. Journal of Chromatography,
593:253–258, 1992.

[12] K. Seiler, D.J. Harrison, and A. Manz. Planar glass chips for capil-
lary electrophoresis - repetitive sample injection, quantification, and
separation efficiency. Analytical Chemistry, 65:1481–1488, 1993.

[13] D.J. Harrison, A. Manz, Z.H. Fan, H. Ludi, and H.M. Widmer. Cap-
illary electrophoresis and sample injection systems integrated on a
planar glass chip. Analytical Chemistry, 64:1926–1932, 1992.

[14] D.J. Harrison, K. Fluri, K. Seiler, Z.H. Fan, C.S. Effenhauser, and
A. Manz. Micromachining a miniaturized capillary electrophoresis-
based chemical analysis system on a chip. Science, 261:895–897,
1993.

[15] C.S. Effenhauser, A. Manz, and H.M. Widmer. Glass chips for high-
speed capillary electrophoresis separations with submicrometer plate
heights. Analytical Chemistry, 65:2637–2642, 1993.

[16] S.C. Jacobson, R. Hergenroder, L.B. Koutny, and J.M. Ramsey. High-
speed separations on a chip. Analytical Chemistry, 66:1114–1118,
1994.

[17] E. Verpoorte and N.F. de Rooij. Microfluidics meets MEMS. Pro-

ceedings of the IEEE, 91(6):930–953, 2003.
BIBLIOGRAPHY 13

[18] H. Becker, K. Lowack, and A. Manz. Planar quartz chips with sub-
micron channels for two-dimensional capillary electrophoresis appli-
cations. Journal of Micromechanics and Microengineering, 8:24–28,
1998.

[19] M. Stjernstrom and J. Roeraade. Method for fabrication of microflu-

idic systems in glass. Journal of Micromechanics and Microengineer-
ing, 8:33–38, 1998.

[20] Y. Fintschenko and A. van den Berg. Silicon microtechnology and

microstructures in separation science. Journal of Chromatography A,
819:3–12, 1998.

[21] K. Hosokawa, T. Fujii, and I. Endo. Handling of picoliter liquid sam-

ple in a poly(dimethysiloxane)-based microfluidic device. Analytical
Chemistry, 71(20):4781–4785, 1999.

[22] J.E. Lilja and S.E. Lennart. Apparatus for sampling, mixing the sam-
ple with a reagent and making particularly optical analysis. US patent
no 4 088 448, filed 1976.

[23] R. Columbus. Liquid transport device and method. US patent no 4

233 029, filed 1979.

[24] S.C. Terry, J.H. Jerman, and J.B. Angel. A gas chromatographic air
analyzer fabricated on a silicon wafer. IEEE Trans. Electron. Devices,
26:1880–1886, 1979.

[25] E. Bassous, H.H. Taub, and L. Kuhn. Ink jet printing nozzle arrays
etched in silicon. Applied Physics Letters, 31:135–137, 1977.

[26] W.K. Schomburg, J. Fahrenberg, D. Maas, and R. Rapp. Active valves

and pumps for microfluidics. Journal of Micromechanics and Micro-
engineering, 3:216–218, 1993.

[27] D.J. Laser and J. G. Santiago. A review of micropumps. Journal of

Micromechanics and Microengineering, 14:R35–R64, 2004.
14 BIBLIOGRAPHY

[28] C. Erbacher, F.G. Bessoth, M. Busch, E. Verpoorte, and A. Manz.

Towards integrated continuous-flow chemical reactors. Mikrochimica
Acta, 131:19–24, 1999.

[29] F.G. Bessoth, A.J. deMello, and A. Manz. Microstructure for efficient
continuous flow mixing. Analytical Communications, 36:213–215,
1999.

[30] N. Lion, T.C. Rohner, L. Dayon, I.L. Arnaud, E. Damoc, N. Youh-

novski, Z.-Y. Wu, C. Roussel, J. Josserand, H. Jensen, J.S. Rossier,
M. Przybylski, and H.H. Girault. Microfluidics systems in pro-
teomics. Electrophoresis, 24:3533–3562, 2003.

[31] C.D. Meinhart and H. Zhang. The flow structure inside a microfab-
ricated inkjet printhead. Journal of Microelectromechanical Systems,
9:67–75, 2000.

[32] M. Boillat, B. van der Schoot, P. Arquint, and N.F. de Rooij. Con-
trolled liquid dosing in micro-instruments. In SPIE Conference, ed-
itor, Microfluidic Devices and Systems II, Santa Clara - California,
1999.

[33] P. Gravesen, O.S. Jensen, and J. Branebjerg. Microfluidics - A review.

Journal of Micromechanics and Microengineering, 3:168–182, 1993.

[34] P. Tabeling. Introduction à la microfluidique. Belin, Paris, 2003.

[35] H.A. Stone, A.D. Stroock, and A. Ajdari. Engineering flows in small
devices. Annual Review of Fluid Mechanics, 36:381–411, 2004.

[36] D. Sinton. Microscale flow visualization. Microfluidics and Nanoflu-

idics, 1:2–21, 2004.

[37] M.C. Potter and D.C. Wiggert. Mechanics of Fluids. Second edition,
1997.

[38] A.L. Lydersen. Fluid Flow and Heat Transfer. John Wiley and Sons
Ltd., 1979.
BIBLIOGRAPHY 15

[39] F.M. White. Fluid Mechanics, 4th ed. McGraw-Hill, New York, 1999.

[40] H.T.G. van Lintel, F.C.M. van de Pol, and S. Bouwstra. A piezo-
electric micropump based on micromachining of silicon. Sensors and
Actuators, 15:153–167, 1988.

[41] M.A. Unger, H.-P. Chou, T. Thorsen, A. Scherer, and S.R. Quake.
Monolithic microfabricated valves and pumps by multilayer soft
lithography. Science, 288:113–116, 2000.

[42] A. Richter, K.A. Hofmann, A. Plettner, and H. Sandmaier. The elec-

trohydrodynamic micro flow meter. IEEE, pages 935–938, 1991.

[43] G. Fuhr, T. Schnelle T, and B. Wagner. Traveling wave-driven mi-

crofabricated electrohydrodynamic pumps for liquids. Journal of Mi-
cromechanics and Microengineering, 4:217–226, 1994.

[44] P.H. Paul, D.W. Arnold, and D.J. Rakestraw. Electrokinetic generation
of high pressures using porous microstructures. In Micro-TAS Con-
ference, editor, Micro Total Analysis Systems 1998, Banff - Canada,
1998.

[45] S.H. Yao, D.E. Hertzog, S.L. Zeng, J.C. Mikkelsen, and J.G. Santiago.
Porous glass electroosmotic pumps: design and experiments. Journal
of Colloid Interface Science, 268:143–153, 2003.

[46] A.V. Lemoff and A.P. Lee. An AC magnetohydrodynamic microp-

ump. Sensors and Actuators B, 63:178–185, 2000.

[47] J. Jang and S.S. Lee. Theoretical and experimental study of MHD
magnetohydrodynamic micropump. Sensors and Actuators A, 80:84–
89, 2000.

[48] P. Luginbuhl. Microfabricated lamb wave device based on pzt sol-

gel thin film for mechanical transport of solid particles and liquids.
Journal of Microelectromechanical Systems, 6:337–346, 1997.
16 BIBLIOGRAPHY

[49] T. Thorsen, S.J. Maerkl, and S.R. Quake. Microfluidic large-scale

integration. Science, 298:580–584, 2002.

[50] P. Cooper. Centrifugal pump theory Pump Handbook. McGraw-Hill,

New York, 2000.

[51] A. Richter and H. Sandmaier. An electrohydrodynamic micropump.

IEEE, pages 99–104, 1990.

[52] G. Ocvirk, M. Munroe, T. Tang, R. Oleschuk, K. Westra, and

D.J. Harrison. Electrokinetic control of fluid flow in native
poly(dimethylsiloxane) capillary electrophoresis devices. Elec-
trophoresis, 21:107–115, 2000.

[53] A.D. Stroock, S.K. Dertinger, G.M. Whitesides, and A. Ajdari. Pat-
terning flows using grooved surfaces. Analytical Chemistry, 74:5306–
5312, 2002.

[54] V. Linder. Polymeric Microfluidic Platforms for Immunodetection.

phd, University of Neuchatel, 2002.

[55] A.D. Stroock and G.M Whitesides. Controlling flows in microchan-

nels with patterned surface charge and topography. Accounts of Chem-
ical Research, 36:597–604, 2003.

[56] I. Gitlin, A.D. Stroock, and G.M. Whitesides. Pumping based on

transverse electrokinetic effects. Applied Physics Letters, 83(7):1486,
2003.

[57] B.J. Kirby and E.F. Hasselbrink Jr. Zeta potential of microfluidic sub-
strates: 1. theory, experimental techniques, and effects on separations.
Electrophoresis, 25:187–202, 2004.

[58] M.J. Madou. Fundamentals of microfabrication: the science of minia-

turisation. CRC Press, Boca Raton, Florida, 1997.

[59] J. Lichtenberg. A Microchip-based Ion Analysis System. Phd, Univer-

sity of Neuchatel, 2002.
BIBLIOGRAPHY 17

[60] H. Becker and C. Gärtner. Polymer microfabrication methods for mi-

crolfuidic analytical applications. Electrophoresis, 21:12–26, 2000.

[61] J.C. McDonald, D.C. Duffy, J.R. Anderson, D.T. Chiu, H. Wu, O.J.A.
Schueller, and G.M. Whitesides. Fabrication of microfluidic systems
in poly(dimethylsiloxane). Electrophoresis, 21:27–40, 2000.

[62] C. Massin. Microfabricated Planar Coils in Nuclear Magnetic Reson-

nance. Phd, Ecole Polytechnique Fédérale de Lausanne, 2004.

[63] R.J. Abraham, J. Fisher, and P. Loftus. Introduction to NMR spec-

troscopy. John Wiley and Sons Ltd., 1988.
18 BIBLIOGRAPHY
Chapter 2

MHD pumping of electrolytes

In the first section of this chapter, the physics behind the magnetohydro-
dynamic phenomenon will first be described, then applied to our particular
experimental conditions. We will derive from that result that it is crucial
to generate a current density of sufficient strength in our microchannels to
achieve high pumping efficiency. Next, we will describe how it is possible
to generate MHD in microchannels with a DC current in buffered solutions.
This chapter will end with a review of the state-of-the-art of MHD microp-
umps and a discussion about the advantages and limiting factors of MHD
pumping.

2.1 The origins of magnetohydrodynamics (MHD)

Magnetohydrodynamics (MHD) is the study of the interaction between
magnetic fields and moving, conducting fluids. Magnetic fields influence
many natural and man-made flows. They are routinely used in industry to
heat, pump, stir and levitate liquid metals. There are three types of mag-
netic fields; a terrestrial magnetic field which is maintained by fluid motion
in the earth’s core, a solar magnetic field, which generates sunspots and
solar flares, and a galactic field which influences the formation of stars.

The MHD phenomenon was first observed by W. Richie [1] in 1833. The

19
20 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

first experiments about magnetic field effects on electrochemical systems

date back more than a century ago and have been credited to M. Faraday [2].
However, the development of MHD as an independent scientific discipline
started at the first half of the 20th century when astrophysicists realized
how ubiquitous magnetic fields and plasmas are throughout the universe.
In 1940, H. Alfvén formulated the main principles of MHD, and in 1970
was awarded the Nobel Prize for his pioneering work in MHD. Scientists
in several fields are interested in MHD, namely for plasma physics, geo-
physics, or liquid metal flow engineering [3].

2.2 MHD theory

The governing equations of MHD come from the combination of the theory
of electrodynamics and fluid mechanics. They consist simply of Newton’s
law of motion and the pre-Maxwell forms of the laws of electrodynamics.

2.2.1 From Maxwell’s equations to the Lorentz Force

Governing equations of Electrodynamics
We are interested in conducting, non-magnetic, and incompressible liquids.
The electromagnetic equations governing MHD for such material proper-
ties are described as follows (see [4] for details):

Charge conservation:
→
− → − ∂
∇ · J = − ρe = 0 (2.1)
∂t
where J is the current density [A·m-2 ], and ρe the charge density [C·m−3 ].

Lorentz force in presence of both electrical current and magnetic field:

→
− → − → −
FL = J × B ·VD (2.2)
where FL is the Lorentz force [N], B the magnetic flux density [T], and VD
the volume exposed to both J and B. The induced Lorentz force will be
maximum when B is orthogonal to J.
2.2. MHD THEORY 21

J is derived upon Ohm’s law:

→
− →
− − → −
J = σ( E + →
v × B) (2.3)

where σ is the electrical conductivity of the material [S·m-1 ], E the electric

field [V·m-1 ], and v the velocity of the charged particle [m·s-1 ].

Governing equations of Fluid Mechanics

In the absence of Lorentz forces, the equation of motion of a small element
of fluid as it progresses through a flow field is stated by the Navier-Stokes
equation. It comes from the application of the second law of Newton to a
small element of fluid of volume δV. The fluid is supposed to be incom-
pressible1 and Newtonian2 .

Continuity equation:
→
− →
∇ ·−
v =0 (2.4)
Momentum equation:
D → 1→−
(−
v ) = − ∇ P + η∇2 →
−
v (2.5)
Dt ρ

where P is the pressure [Pa], ρ is the density [kg·m-3 ], η = µ/ρ is the kine-
matic viscosity, µ being the absolute viscosity of the fluid [Pa·s]. The ex-
pression D(.)/Dt represents the convective derivative3 . The boundary con-
dition on v corresponding to equation 2.5 is that v = 0 on any stationary
solid surface (called also the no-slip condition).

2.2.2 Application to poor conductors

The relative strength of the induced magnetic field is characterized by a
dimensionless number [4] called the Magnetic Reynolds number (Rm):
1 Assumption: stream velocity  speed of sound
2 The viscosity of the fluid is constant, and does not depend on the shear rate.
3 It is the rate of change of a quantity associated with a given element of fluid. Applied
→
− −
to any vector → − D →
a , it will have the following form: Dt (−
a ) = ∂t∂ (→
−
a ) + (→
−
v · ∇ )(→
a ).
22 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

Figure 2.1: Flow profiles at low and large Hartmann numbers.

Rm = vl/λ  1 (2.6)
where l is a characteristic length scale of the flow [m], and λ the magnetic
diffusivity [m2 s−1 ]. Compared to metals or plasma fluids, electrolytes are
considered as poor conductors. For such conditions, the Magnetic Reynolds
number (Rm) of the medium is typically much smaller than 1. In this sit-
uation of low Rm, the magnetic field B will exert a major influence on the
velocity of the charged particles v, but v will not significantly perturb B. In
other words, the magnetic field associated with induced currents is negligi-
ble compared to the imposed magnetic field.

MHD flow in ducts

When a constant current is injected into a fluid under the influence of a
magnetic field, the resulting Lorentz force will, in general, produce motion.
Electromagnetic pumps were one of the earliest applications of MHD, and
were routinely used in metallurgical industries [3]. Those devices consisted
of a current perpendicular to a magnetic field, with both fields being perpen-
dicular to the axis of the duct. Assuming the duct filled with a conducting
material, the resulting Lorentz force provided the pumping action all along
the axis of the duct. MHD flow in ducts was first studied by the engineer
→
− → −
J. Hartmann. His work focused on the effects of J × B on boundary lay-
ers. He demonstrated that the flow velocity increases rapidly over a short
2.2. MHD THEORY 23

distance from the wall (see Fig. 2.1). This boundary layer, which has thick-
ness ∼ δ (see 4 ), is called a Hartmann layer [4]. The Hartmann number, for
a channel of radius r is defined by

Ha = r/δ = B · r · (σ/ρη)1/2 (2.7)

We see that Ha is directly proportional to the conductivity of the fluid.

For Ha close to zero, Hartman flow has a parabolic profile (see Fig. 2.1).
Because electrolytes are poor conductors, their corresponding Ha will be
close to zero. The MHD flow in a cylindrical pipe is expected to have a
parabolic profile.

MHD pumping in microchannels

Due to the small dimensions of microchannels, the Reynolds number (Re)

will most always be below the transitional Reynolds number (Ret ). In
our particular case, we will typically encounter mean flow velocities of
1 mm·sec−1 for aqueous buffer solutions in 20 mm long and 75 µm deep
semi-circular microchannels. The values for Re and Ret , defined in section
1.2, were calculated to be Re = 0.15 with Ret = 400. Re  Ret shows that
the flow in our microchannels will always be laminar.

The left part of the fluid mechanics equation 2.5 has the following form:

D → →
− −
(−
v)= →
∂−
v + (→
−
v · ∇ )(→
v) (2.8)
Dt ∂t

Because of the laminar flow regime (low Re), inertia terms can be neglected.
→
− −
We will have (→ −v · ∇ )(→
v ) = 0. Equation 2.5 for a low Re steady flow
( ∂t∂ →
−
v = 0) is then:

→
−
∇ P = µ∇2 →
−
v (2.9)
4 δ = (ρη/σB2 )1/2
24 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

The solution of equation 2.9 with the no-slip condition at the walls gives a
parabolic flow distribution with a maximum for the velocity at the very mid-
dle of the channel. Such flows were studied by G. Hagen and J.L. Poiseuille
in 1940 [5], both of whom established the 1-D pressure-drop law:

∆P = Q · Rhy (2.10)
where ∆P is the pressure drop in the channel [Pa], Q is the volumetric flow
rate [m3 · s−1 ], and Rhy is the hydraulic resistance of the channel [N·s·m−5 ].
A channel of rectangular cross-section will have the following expression
for the hydraulic resistance [6, 7]:

8µL(w + h)2
Rhy = (2.11)
w3 h3
for channels of half-circular cross-section [8]:

128µL
Rhy = (2.12)
3πh4
and for capillaries of circular cross-section:

8µL
Rhy = (2.13)
πh4
where w is the channel width, h is the channel height, and L is the total
channel length.

MHD flow in a microchannel has a parabolic-shaped profile. The Lorentz

force acts on the whole body of the fluid, and generates a body force (which
can be seen as a pressure drop) ∆P in the microchannel. This ∆P acting on
the cross-sectional area A of the channel by the Lorentz force (Equ. 2.2) is
then

FL J · B ·VD J · B · A · l →
− → −
∆P = = = = J · B · l (if J ⊥ B ) (2.14)
A A A
2.3. PUMPING ELECTROLYTES WITH MHD IN THE DC MODE 25

where l is the channel length on which both perpendicular current and mag-
netic field are applied5 . We then equate the expressions 2.10 and 2.14. If
no additional pressure gradient is imposed on the channel (e.g. by gravity),
we can obtain the following relationships for the volumetric flow rate and
average linear flow velocity induced by MHD in microchannels:

J ·B·l
Q= (2.15)
Rhy
J ·B·l
v= (2.16)
A · Rhy

As we see in equations 2.15 and 2.16, the flow rate and average flow
velocity are proportional to the applied current density and magnetic field,
as well as to a geometric constant for the channel.

2.3 Pumping electrolytes with MHD in the DC mode

MHD flow rate is proportional to magnetic field strength and current den-
sity. The magnetic field in our case is DC and will have discrete values,
depending on the strength of the permanent magnets or on the voltage ap-
plied to the NMR coil. The magnetic field will not interact with the ions
in the solution if those are not moving (i.e. no current). When a DC cur-
rent is generated in an ionic solution it will be accompanied by chemical
changes at the electrode location. In this section we will describe the chem-
ical and transport processes accompanying a high DC current generation in
electrolyte solutions.

2.3.1 High DC current generation in electrolytes

Electrolysis
High DC current generation (i.e. I ≥ 0.1 mA) in a ionic aqueous solu-
tion is always associated with electrolysis. The application of a potential
5l , L, the total length of the channel
26 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

difference across electrodes in an electrolytic medium will drive chemi-

cal changes at the electrode locations. The ions in the solution will carry
charges between the electrodes.

When a DC voltage greater than the electrode potential (E p ) is applied

across platinum electrodes immersed in an aqueous NaCl solution, the fol-
lowing electrochemical reactions will occur [9, 10]:

Anode:
2H2 O O2 + 4H + + 4e− (2.17)

Cathode:
2H2 O + 2e− H2 + 2OH − (2.18)

The positive electrode, called anode, will oxidize water and produce oxy-
gen gas. The negative electrode, called cathode, will reduce water and
produce hydrogen gas. The term electrolysis, as well as electrode, anode,
cathode, and electrolyte were introduced by M. Faraday in 1835 [11]. Other
reactions may occur at the electrodes as well, depending on the composi-
tion of the electrolyte and the magnitude of the applied voltage. In our case,
the oxidation of chloride (2Cl − Cl2 + 2e− ) will also take place at the an-
ode. The medium around the positive electrode will become more and more
acidic, whereas the one around the negative electrode will become alkaline.
Gas production at the electrodes by electrolysis is the phenomenon respon-
sible for the formation of bubbles. The rate of bubble generation will be
related to the current density.

pH control
The change in pH at the electrode location is a phenomenon that can be
controlled. The rate at which the pH changes will depend on the amount
of hydroxide and hydronium ions produced at both electrode locations. M.
Faraday showed that the number of moles Nx of ion X produced after a time
t at a current I is:
I ·t
Nx = (2.19)
a·F
2.3. PUMPING ELECTROLYTES WITH MHD IN THE DC MODE 27

where F = 9.6487 ·104 is the Faraday constant [C·mol−1 ], and a is the

valence of ion X.

In order to regulate the concentration of hydroxide and hydronium ion con-

centrations, buffer solutions are used. Buffers are defined as substances
which by their presence in solution increase the amount of acid or alkali
that must be added to cause unit change in pH [12]. The intensity of buffer
action, that is, the effectiveness with which a solution resists changes in hy-
drogen ion concentration, is called the buffer capacity [13]. In our case, a
high buffer capacity is essential to maintain a stable pH in the pump during
the course of an experiment.

Choice of electrode

The physical and chemical properties of the electrode material are also of
crucial importance, as it will have a great influence on the electrochemi-
cal processes at the electrode/solution interface. With some electrodes, a
particular electrochemical reaction may occur slowly or not at all; but with
another type, the rate of reaction may be increased under the same set of
conditions. One of the most widely used electrode material is platinum; this
is mostly due to its stable and well characterized electrochemical behavior.
For that reason, we decided to use platinum for the work described herein.

Choice of electrolyte solutions

Any conductive solution may be used to generate the current necessary

for generating a MHD flow. However, a particular choice of electrolyte
solution may be necessary if the µTAS system is sensitive to the side effects
of electrolysis. If gas bubble formation has to be avoided, a non-aqueous
solution with redox species [14–17] may be chosen. Buffers with high
capacity are also of great interest in MHD pumping if one wants to avoid
large pH gradients in between the electrodes.
28 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

Heat generation

When the temperature rises in the system, the heat will influence, among
other things, the sample stability, buffer viscosity, chemical equilibria, pH
and the resulting migration time for any given species. Joule heating refers
to the increase in temperature of a conductor as a result of resistance to
an electrical current flowing through it. In capillary electrophoresis sys-
tems, the control of this phenomenon is of crucial interest and has been
extensively studied. It was found that the magnitude of the capillary tem-
perature, as well as the temperature gradients, are proportional to the power
dissipation [18]:
UI
Qd = =E I (2.20)
L
where Qd is the power dissipation [W·m−1 ], U is the applied voltage [V], I
is the current [A], and L is the overall length [m].

In microfluidic systems, Joule heating becomes problematic for Qd values

[19] larger than 3 W·m−1 (instead of 1 W·m−1 for standard capillaries).

2.3.2 Ion transport and electrokinetic flows

Ion transport

The rate of any transport process is characterized by its flux, which is the
amount of substance which passes through a unit area during 1 second. In
the case of the displacement of charges, this flux is the electrical current per
unit area, and is called the current density. Many phenomena can be at the
origin of the displacement of charges.

• Charges can move by diffusion, under the influence of a charge den-

sity gradient. This process is governed by Fick’s law:
−−→ →
−
JFick = −D ∇C (2.21)

where C is the concentration of the ions in the solution [kg·m−3 ] and

2.3. PUMPING ELECTROLYTES WITH MHD IN THE DC MODE 29

D is the diffusion coefficient6 [m2 ·s−1 ] or diffusivity. D specifies

the diffusion rate of the given substance. The minus sign shows that
the diffusion flux is oriented from higher to lower concentrations.
Adapted to an electrical environment, this law has the following form
(both sides of equ. 2.21 were divided by q/m):
→
− →
−
JD = −D ∇ ρe (2.22)

where JD is the current density originating from diffusion, and ρe is

the charge density [C·m−3 ].

• Charge movement can also be provoked by a body translation (for

example a flow). The induced current is then expressed by:
→
−
JT = ρe →
−
v (2.23)

where v is the velocity of charge translation.

• Charges can also move under the influence of the electric field. In
this case, we have the relationship:
→
− →
−
Je = σ E (2.24)

The total flux of charged substances, or current density J is expressed as

the sum of all three contributions we just enumerated which is:
→
− →
− →
−
J = −D ∇ ρe + ρe →
−
v +σE (2.25)

Substituted into Navier-Stokes equation describing the flux of a charged

fluid, this expression and the charge conservation equation will give the
electrokinetic equation of the charge movement. As it is not the main topic
of this thesis, it will not be depicted here.
6 Stokes-Einstein’s law describes this coefficient by the theory of Brownian motion, D =
kT
6πrµ where k is Boltzmann’s constant, T is the absolute temperature [K], r is the ion Stoke’s
radius [m]
30 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

Electrokinetic flow

The transport mechanisms related to the effects of the electric field on a

fluid can be divided into three fundamental components: electrophoretic
(ep), electro-osmotic (eo), and dielectrophoretic effects [20–23]. Elec-
trophoresis is defined as the motion of charged particles under an applied
electric field in a fluid at rest. Electro-osmosis is the induction of motion
of an ionic liquid under the effect of an electric field. Dielectrophoresis
is the motion of polarizable particles under the influence of a non-uniform
electric field. Dielectrophoresis phenomena will not be observable in this
work, and will therefore not be treated in this section.

A charged particle in an insulating medium under the influence of the

electric field E is subjected to the Coulomb force F = qE. In absence of
Brownian motion, this particle obeys to equation of motion:

→
− → − d−
q E − Fv = m → v =0 (2.26)
dt
where q [C] is the charge of the ion in the solution. Fv is the viscous force
experienced by the moving ion and is given by Stokes law
→
−
Fv = 6πrµ→
−
v (2.27)

where r [m] is the ion Stoke’s radius. The inertia term in the preceding
equation was voluntarily neglected; this is related to Stokes approximation
for moving particles at low Reynolds numbers. From those last two equa-
tions we deduce the electrophoretic velocity, which is proportional to µep ,
the electrophoretic mobility of the ion. µep depends on the chemical nature
of the buffer.
− →
− q →
−
v→
ep = µep · E = ·E (2.28)
6πµr
Each ion in the buffer will have a different electrophoretic mobility, depend-
ing on both their electric charge and Stokes radius in a solvated medium.
Table 2.1 presents examples of radii values [10, 24].
2.3. PUMPING ELECTROLYTES WITH MHD IN THE DC MODE 31

Table 2.1: Stokes radii of selected solvated ions

Ion radius [Å]

K+ , Cl− 3
OH− 3.5
HPO− 4 4
Na+ 4.5
H+ 9

When a dielectric (like glass) is immersed in an electrolyte, a surface

charge appears spontaneously. The mobile charges of the liquid reorganize
and form a double layer7 near the immersed wall. When the electric field is
applied in a parallel fashion to the glass capillary, a bulk flow is generated.
This electro-osmotic flow is due to the Coulomb force on the mobile part
of the electrolyte double layer. In a glass capillary, this layer consists of
positive ions which are attracted to the negatively charged capillary wall.
The flow of these cations towards the negative electrode causes the whole
capillary volume to migrate as a consequence of the viscous drag caused
by the movement of the mobile layer. The eo velocity is proportional to
µeo f , the eo mobility of the solution in the capillary. µeo f depends on the
surface properties of the capillary walls, and on the chemical properties of
the solution (buffer pH, ionic strength, etc...).

− →
− εζ →
−
v−
→
eo f = µeo f · E = ·E (2.29)
µ
where ε [C·V−1 ·m−1 ] is the dielectric constant of the medium, and ζ [V] is
the Zeta potential8 created at the plane of shear at the channel’s wall.

We saw that both eof and ep flows in the microscale are proportional to
7 calledthe Stern layer
8 The 4πρe δ
Zeta potential is the electrical potential at the edge of the Stern layer: ζ = ε ,
where δ is the double layer thickness.
32 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

the electric field without any relation to the geometry of the channel (depth
of channels for example).

Effect of the Lorentz force on individual moving charges

The expression for the electromagnetic force (the Lorentz force, FLi ) on
individual moving charge i is expressed by:
−
→ →
−
FLi = qi →
−
vi × B (2.30)

where qi and vi are the charge and velocity of the moving particle i. During
electrolysis cations and anions move in opposite directions, so that the signs
of q+ v+ and q− v− are both positive by convention (subscripted "+" and "-
" denote cations and anions, respectively). As a result, both cations and
anions experience their Lorentz forces in the same direction (see Fig. 2.2).
The total Lorentz force FL acting upon a defined volume VD containing n
charge couples (n = n+ + n− ) is the sum of the Lorentz forces acting upon
all moving ions contained in this volume, and is given by:

FL = (n+ q+ v+ + n− q− v− ) × B ·VD (2.31)

n+ and n− are the numbers of cations and anions, respectively. The quantity
(n+ q+ v+ + n− q− v− ) is the current density J. Therefore, the equation that

Figure 2.2: Parallel magnetic Lorenz forces on cations and anions migrat-
ing towards the negative and the positive electrodes, respectively.
2.4. STATE OF THE ART IN MHD MICROPUMPS 33

define FL is identical to that depicted with the macroscopic treatment (see

equation 2.2).

2.4 State of the art in MHD micropumps

One of the first demonstrations of MHD pumping of saline solutions (sea-
water) in microchannels was presented by Jang and Lee in 2000 [25]. The
microchannels, made in silicon, were closed, the applied current was DC,
and MHD pumping was induced using a permanent DC magnet. The pump
was able to produce a maximum flow rate of 63 µL·min−1 . However, elec-
trolysis of water led to the formation of gas bubbles in the channel, dis-
rupting the flow. Bau et al.. [26, 30–32] studied DC MHD of saline so-
lutions in microchannels made of ceramic tape. Due to gas bubble for-
mation, the microchannels had to be used in an unsealed state9 . It was
also observed that the electrodes, made of gold paste, were consumed very
quickly by electrolysis. Redox-based MHD pumping was studied by Lev-
entis et al. [14] and more recently by Clark et al. [33]. Other groups studied
MHD pumping of saline solutions with combined AC magnetic fields and
an AC current at high frequencies (≥ 1 kHz) to avoid bubble formation
at the electrodes [27–29, 34]. However, MHD pumping in the AC mode
leads to low flow rates as a result of the low intensity of the AC magnets.
This is because the operation of electromagnets under high frequency and
current conditions is limited by induction heating [29] and Eddy current
formation [34].

Table 2.2 compares experimental conditions and main properties for

a consistent sample of all MHD micropumps published so far. The com-
pared micropumps had channel depths of 3 mm or less, and the pumped
solutions were electrolytic solutions. The pumps that had the best perfor-
mance in term of flow rate and velocity were the ones with maximized J, B
and depth of channel (h), and this was achieved by maximizing the chan-
nel cross-sectional area (A). On the other hand, the best performances in
9 The device was also left open to facilitate the introduction of fluids and flow visualiza-

tion, as the tape was not optically transparent

CHAPTER 2. MHD PUMPING OF ELECTROLYTES

U I A AJ l J B ∆PMHD vMHD QMHD

(V) (mA) (mm2 ) (mm2 ) (mm) (A·m−2 ) (T) (Pa) (mm·s−1 ) (µL·min−1 )
Jang et al. [25] 30 DC 1.8 0.4 30 30 60 0.44 1 2.6∗ 63∗
Leventis et al. [14] >1.3 DC 35 18 225 75 155 1.35 16 0.4 450·103
Bau et al. [26] 4 DC 15 1.9 292 172 51 0.4 3.5 0.4 45
Lemoff et al. [27] 6.6 AC 140 0.2 1.5 4 92105 0.013 5 1.5 18
West et al. [28] 5 AC 90 0.2 5 28 17684 0.011 5.5 0.24 3
Eijkel et al. [29] 4 AC 40 6·10−3 2 63 21100 0.1 133 0.04 14·10−3
Table 2.2: Performance comparison of previously published MHD pumps. All values for voltage (U), current (I),
channel cross-sectional area (A), total length of electrodes along the pumping channel (l), MHD flow mean velocity in the pumping
channel (vMHD ) and MHD flow rate (QMHD ) were experimental data, and were taken from references [14, 25–29]. Most of values
for the electrode cross-sectional area (AJ ) and current density (J) across the pumping channel had to be calculated. The body force
(∆PMHD ) generated by the pumps was calculated using relation 2.14. ∗ Both values were taken from experimental measurements. If
calculated with relations 2.16 and 2.15, the predicted velocity and flow rate would be 0.16 mm·s−1 and 4 µL·min−1 respectively.
34
2.5. ADVANTAGES AND LIMITATIONS OF MHD PUMPING 35

terms of generated body force were the pumps with maximized J, B and
electrode length (l), and this was achieved by minimizing the electrodes
cross-sectional area (AJ ).

Applications for MHD pumping reported in the literature include the

study of mixing [26, 32, 35], polymerase chain reaction (PCR) [28], sepa-
ration methods [29, 34], chaotic flow [36], and trace metals analysis [16].

2.5 Advantages of MHD pumping compared to other

mechanisms, and limitations
Advantages
Magnetohydrodynamic pumps operate without moving parts, and require
only low voltages (less than 10 V) for their actuation. Furthermore, liquid
flow does not depend on the chemical nature of the capillary surface unlike
in electro-osmotic flow. Forward and reverse flows are possible, with a flow
profile similar to that observed for pressure-driven flows. In addition, it can
manipulate unprocessed biofluids with wide ranges of properties as long
as the solution is at least slightly conductive. For all these reasons, MHD
pumping has attracted a growing interest for the handling of conductive
fluids in microchannels over the past few years.

Limitations
One of the limitations of MHD pumps are the geometrical constraint as-
sociated with the vectorial origin of the Lorentz force (FL ): this force is
the body force generating the flow in the MHD pumping channel. For the
pumping to occur, FL is perpendicular to both J and B at the same time, and
has to be in the direction of the pumping channel. In an NMR environment,
the chip will have to be held horizontally for FL to be in the plane of the chip
(along the pumping channel) with J across the pumping channel. Another
limitation is the drastic decrease of the MHD flow rate with the solute ionic
concentration. Finally, the change in pH very near to the electrodes, as a
36 CHAPTER 2. MHD PUMPING OF ELECTROLYTES

consequence of hydrolysis, could also be considered as a limiting factor for

MHD pumping as biological specimens are quite sensitive to pH.
Bibliography

[1] W. Ritchie. Experimental researches in electro-magnetism and

magneto-electricity. Philosophical Transactions of the Royal Society
of London, 123:313–321, 1833.

[2] M. Faraday. Diary, volume IV 288. London, 1833.

[3] R.S. Tessier and M.-J. Baker. Handbook of electromagnetic pump

technology. Elsevier Science Publishing Co., 1987.

[4] P.A. Davidson. An introduction to MHD. Cambridge University Press,

2001.

[5] F.M. White. Fluid Mechanics, 4th ed. McGraw-Hill, New York, 1999.

[6] W.J. Beek and K.M.K. Muttzall. Transport Phenomena. John Wiley
and Sons Ltd., 1975.

[7] P. Gravesen, O.S. Jensen, and J. Branebjerg. Microfluidics - A review.

Journal of Micromechanics and Microengineering, 3:168–182, 1993.

[8] E. Oosterbroek. Modeling, design and realization of microfluidics

components. PhD thesis, University of Twente, 1999.

[9] J. Koryta. Ions, Electrodes, and Membranes. John Wiley and Sons
Ltd., 1982.

[10] J. Koryta. Principles of electrochemistry. John Wiley and Sons Ltd.,

1987.

37
38 BIBLIOGRAPHY

[11] M. Faraday. Experimental Researches in Electricity, volume 7th Se-

ries. London, 1834.

[12] D.D. Van Slyke. On the measurement of buffer values and on the
relationship of buffer value to the dissociation constant of the buffer
and the concentration and reaction of the buffer solution. Journal of
Biological Chemistry, 52:525–570, 1922.

[13] R.G. Bates. determination of pH. John Wiley and Sons Ltd., 1973.

[14] N. Leventis and X. Gao. Magnetohydrodynamic electrochemistry in

the field of Nd-Fe-B magnets. Theory, experiment, and application in
self-powered flow delivery systems. Analytical Chemistry, 73:3981–
3992, 2001.

[15] E.A. Clark, P.U. Arumugam, and I. Fritsch. Control of fluid flow
using redox magnetohydrodynamics for chip-based microfluidics and
analytical applications. In Inc. The Electrochemical Society, editor,
206th meeting, 2004.

[16] E.A. Clark and I. Fritsch. Anodic stripping voltametry enhancement

by redox magnetohydrodynamics. Analytical Chemistry, 76:2415–
2418, 2004.

[17] J.S. Wilkes, M.L. Williams, and R.L. Musselman. Electrochemical

magnetohydrodynamic pumping of ionic liquids. Electrochemistry,
8:742–744, 2005.

[18] R.J. Nelson, A. Paulus, A.S. Cohen, A. Guttman, and B.L. Karger.
Use of Peltier thermoelectric devices to control column temperature
in high-performance capillary electrophoresis. Journal of Chromatog-
raphy, 480:111–127, 1989.

[19] A.M. Garcýa Campana, W.R.G. Baeyens, H.Y. Aboul-Enein, and

X. Zhang. Miniaturization of capillary electrophoresis systems using
micromachining techniques. Journal of Microcolumn Separations,
10(4):339–355, 1998.
BIBLIOGRAPHY 39

[20] P.D. Grossman and J.C. Colburn. Capillary Electrophoresis "Theory

and Practice". Academic Press, San Diego, CA, 1992.

[21] A. Manz, C.S. Effenhauser, N. Burggraft, D.J. Harrison, K. Seiler, and

K. Fluri. Electroosmotic pumping and electrophoretic separations for
miniaturized chemical analysis systems. Journal of Micromechanics
and Microengineering, 4:257–265, 1994.

[22] Baker. Capillary Electrophoresis. 1995.

[23] H. Girault. Electrochimie physique et analytique. Presses Polytech-

niques et Universitaires Romandes, 2001.

[24] J.M. Bockris and A.K.N. Reddy. Modern electrochemistry, volume 1.

Plenum Press, 1970.

[25] J. Jang and S.S. Lee. Theoretical and experimental study of MHD
magnetohydrodynamic micropump. Sensors and Actuators A, 80:84–
89, 2000.

[26] H.H. Bau, J. Zhong, and M. Yi. A minute magnetohydrodynamic

(MHD) mixer. Sensors and Actuators B, 79:207–215, 2001.

[27] A.V. Lemoff and A.P. Lee. An AC magnetohydrodynamic microp-

ump. Sensors and Actuators B, 63:178–185, 2000.

[28] J. West, B. Karamata, B. Lillis, J.P. Gleeson, J. Alderman, J.K.

Collins, W. Lane, A. Mathewson, and H. Berney. Application of mag-
netohydrodynamic actuation to continuous flow chemistry. Lab On a
Chip, 2:224–230, 2002.

[29] J.C.T. Eijkel, C. Dalton, C.J. Hayden, J.P.H. Burt, and A. Manz. A cir-
cular ac magnetohydrodynamic micropump for chromatographic ap-
plications. Sensors and Actuators B, 92:215–221, 2003.

[30] J. Zhong, M. Yi, and H.H. Bau. Magneto hydrodynamic (MHD) pump
fabricated with ceramic tapes. Sensors and Actuators A, 96:59–66,
2002.
40 BIBLIOGRAPHY

[31] H.H. Bau, J. Zhu, S. Qian, and Y. Xiang. A magneto-

hydrodynamically controlled fluidic network. Sensors and Actuators
B, 88:205–216, 2003.

[32] S. Qian and H.H. Bau. Magneto-hydrodynamic stirrer for stationary

and moving fluids. Sensors and Actuators B, 106:859–870, 2005.

[33] P.U. Arumugam, E.A. Clark, and I. Fritsch. Use of paired, bonded
NdFeB magnets in redox magnetohydrodynamics. Analytical Chem-
istry, 77:1167–1171, 2005.

[34] J.B. Bao and D.J. Harrison. Fabrication of microchips for running
liquid chomatography by magnetohydrodynamic flow. In 7th Interna-
tional Conference on Miniaturized Chemical and Biochemical Analy-
sis Systems, pages 407–410, Squaw Valley, California, USA, 2003.

[35] J. West, J.P. Gleeson, J. Alderman, J.K. Collins, and H. Berney. Struc-
turing laminar flows using annular magnetohydrodynamic actuation.
Sensors and Actuators B, 96:190–199, 2003.

[36] Y. Xiang and H.H. Bau. Complex magnetohydrodynamic low-

reynolds-number flows. Physical Review E, 68:016312, 2003.
Chapter 3

DC MHD pump developments

The evolution of our experimental research in DC MHD pumping will

be presented in this chapter. Experimental conditions are described first.
These conditions changed as we gained a better understanding of the many
phenomena associated with DC MHD in electrolyte environments. Three
generations of microfluidic chips have been designed and tested. The MHD
phenomenon was studied in the devices with electrolytic solutions of high
ionic strength. The electrode and channel layout improved with each gen-
eration. The common parameters for all experiments included a static mag-
netic field (permanent magnet) and a DC voltage applied between two elec-
trodes. The main parameters responsible for MHD flow are the magnetic
field strength and the electrical current between the electrodes (expressed
in equation 2.16). In our case, the magnetic field is a permanent magnet
with a fixed field strength (between 0.2 and 0.4T). Therefore, the current is
the only parameter that can be modulated in our experiments. We focused
on designing microfluidic channels with integrated electrodes which could
generate a high current density inside the channel.

41
42 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

3.1 Experimental conditions

3.1.1 Electrolyte properties

Saline solutions

The electrolytes used for the first experiments (described in the following
sections) were NaCl and KCl in DI water. Their concentration varied from
0.01 M to 1 M. The most frequently used concentration was 0.1 M. The
resistivity of these solutions were measured to be 5 · 104 Ω·m for DI water,
0.69 Ω·m for 0.1 M KCl, and 0.12 Ω·m for 1 M NaCl. The initial pH of
saline solutions was also measured. The pH of the solutions was varied
by adding drops of 0.1 M NaOH. This resulted in pH ranges around 9 -
10. The values for the absolute viscosity were taken from [1]. They are
based on the values of aqueous solutions of NaCl at 20˚C. The absolute
viscosity is 1.095·10−3 N·m-2 ·s for 1 M NaCl, and 1.008·10−3 N·m-2 ·s for
0.1 M NaCl. Polystyrene beads are diluted in the saline solution for the
observation of the flow . Three different kinds of beads were used: the
2 to 6 µm diameter carboxylate-modified beads and 1 µm diameter amine-
modified beads came from Polysciences Europe GMBH (Germany) and the
2.5 µm diameter AlignFlow beads came from Invitrogen AG (Switzerland).

Buffer solutions

In the experiments described in chapters 4 and 6, buffered solutions were

preferred, as they enabled a better control on the pH of the solution between
the electrodes [2]. Citrate-phosphate buffer was chosen for the experiments,
because it works in a wide range of pH (from 2 to 8) with the possibility
of adjusting the ionic strength between 0.25 M to 1 M. Buffer solutions
of varying pH but constant ionic strength have been described by Elving
et al. [3]. The buffer compositions were selected from the McIlvaine for-
mulas [4], and the desired ionic strength is maintained by the addition of
potassium chloride. Details about the proportion of all three specific species
as a function of the desired pH and ionic strength is found in Appendix D.
3.1. EXPERIMENTAL CONDITIONS 43

3.1.2 Magnet properties

For all experiments in our laboratory we used rare earth NdFeB perma-
nent magnets. They come from two different providers, namely Maurer
Magnetic AG (Switzerland) and Forcefield (CO, USA). With the help of
K. Ehrmann and C. Massin from the LMIS3 laboratory at EPFL, we first
performed an estimation of the intensity of the magnetic field (B) and its di-
rection in the microchannel using the finite element software Maxwell 3D
(Ansoft corp, PA, USA). An example of the magnitude and shape of B, as
a function of the channel’s distance to the center of the magnet, is shown
in Fig. 3.1. As a result, we could predict the variation of the intensity of
the field (from 0.25 to 0.4 T) as a function of the distance between the mi-
crochannel and the surface of the magnet (from 500 µm to 5 mm). The
dimensions of the magnets varied in thickness (from 5 mm to 28 mm) and
in width (from 10 to 50 mm diameter). The direction of the magnetic field
was very uniform in the center of the magnet’s surface (see Fig. 3.1). In
the experiments, magnets with the wider diameters1 were preferred as they
provided the most uniform magnetic field perpendicular to the plane of the
chip.

The intensity of the field for each magnet was confirmed by measurements
with a 3-axis magnetic-field transducer (model 3MR-1C, Sentron AG, Switzer-
land). Special care was taken to position the Hall sensor at the exact height
corresponding to the channel position where pumping occurred. This mea-
surement of the accurate magnetic field intensity in the channel enabled us
to compare our measurements with the expected velocity calculated from
theory (see equation 2.16).

3.1.3 Experimental methods

An example for a typical DC MHD experimental setup is shown in Fig.
3.2. In the first set of experiments, a HP E3617A (Duebendorf, Switzer-
land) voltage source was used for the experiments in the DC mode. The
current compliance of the source was limited to 100 mA. The source was
1 Diameter of the magnets bigger than the outer dimensions of the chip.
44 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Figure 3.1: (Top) Magnet shape and magnetic field direction. (Bottom)
Magnitude of magnetic field B along the z axis (from center of magnet).
3.2. MHD IN A PDMS T-CHANNEL 45

able to generate DC voltages up to 60 V, which is large enough for our

experiments. The current generated in the channel was measured in series
with a high impedance ampmeter from Keithley Instruments S.A. (Zurich,
Switzerland). A constant voltage (0 - 60 V) was applied across two elec-
trodes. Most of the experiments used pure platinum electrodes for the
generation of ionic current, but some measurements were also taken with
silver/silver-chloride electrodes. The constant voltage generated a constant
current across the electrodes. The current range was set between 0.6 to
5 mA, depending on the experiment. Note that the current was not com-
pletely constant, as it depended also on the chemical environment around
the electrodes. However, the standard deviation of the current generated
across the channel was also measured. In the later experiments (described
in Ch. 6), a constant current was applied across two Pt electrodes with a
Keithley 2400 source meter, the compliance of the source was set to 4 mA,
and the corresponding voltage didn’t exceed 10 V.

We characterized the MHD flow by observing the movement of polystyrene

beads in the solution. We used a stereoscopic microscope (Leica AG, Glat-
tbrugg, Switzerland) which was connected to a JVC digital camera (Spitzer
Electronics AG, Oberwill, Switzerland) for visual observation. The linear
flow velocity was determined by measuring the time required for beads to
travel a distance between two defined points in the pumping channel. For
our measurements, we selected the fastest moving beads to determine ex-
perimentally the maximum velocity for each condition. For a parabolic flow
profile, those beads will be located in the middle of the channel. The max-
imum velocity was calculated by averaging the velocity of the 10 fastest
moving beads from a population of 50 beads. Errors for the measurement
were based on the standard deviation calculated for the 10 data points.

3.2 MHD in a PDMS T-Channel

Our first MHD flow study consisted of a simple microchannel design, fab-
ricated by rapid-prototyping of poly(dimethylsiloxane) (PDMS) elastomer
material. Design considerations included: 1) ease of chip manufacture 2)
46 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Figure 3.2: View of the experimental setup

use of a constant, homogeneous magnetic field 3) placement of the elec-

trodes in the outer reservoirs of the device. For ease of chip manufacturing,
we used PDMS. The use of a constant magnetic field ensured experimental
conditions resembling those inside the strong magnet used for the NMR
analysis at EPFL. The generation of gas bubbles in the microchannel was
prevented by placing the electrodes in the outer reservoirs.

Fabrication
Channels were fabricated in PDMS using a silicon master, and sealed with
a glass wafer. We applied a DC voltage along the side-channel, which was
filled with a conducting electrolyte and perpendicular to the main chan-
nel (see Fig. 3.3). The ionic transport lead to the formation of a current.
Therefore, in the presence of a magnetic field, we expect to observe an
3.2. MHD IN A PDMS T-CHANNEL 47

MHD-related flow in the main channel.

Experiments
Channel dimensions varied from 50 to 260 µm in depth, and from 50 to 500
µm in width. The length of the side channel set the distance separating the
electrodes and was constant, with a value of 8 mm. The length of the main
channel varied from 8 mm to 8 cm. Note that standard channels dimensions
for all the final experiments were 260 µm in depth, 300 µm in width, and
both 8 mm length. DC voltage varied from 0 to 30 V, with a source-limited
current of 0.1 A. Current was generated by platinum (Pt) wire electrodes
placed in the reservoirs at the entrance of the side-channel. The two elec-
trodes were always separated by a 8 mm long channel. The electrode area,
a parameter necessary for current density calculations, is (for this specific
geometry) the cross-sectional area of the side-channel, which is 0.08 mm2 .
For MHD observations, the magnet was positioned below the channels, in
contact with the glass coverplate. The 500 µm thick coverplate fixed the
distance between the magnet and the ions in the solution. In case of 260
µm deep channels, the distance between the surface of the magnet and the
middle of the channel was 630 µm, leading to a magnetic field intensity of
0.4 T in the microchannel.

Figure 3.3: Principle of use of the first MHD design

48 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Current generation in the side-channel

We characterized the ohmic behavior of the solution by measuring the cur-

rent generated across the side-channel as a function of the DC voltage in-
tensity (see Fig. 3.4 for an example). The most stable and high current
intensities were obtained for voltages of 20 V. The standard current value
was 0.5 mA for 0.1 M KCl and 20 V. This corresponds to a calculated cur-
rent density of 6410 A·m-2 for side-channels with 0.08 mm2 cross-section.

For saline solutions, we found that the higher the ionic strength, the higher
the recorded current for a given voltage. However, at the same time, we
observed that the higher the ionic strength, the more bubbles were gener-
ated for the same given voltage. We found that saline solutions of 0.1 M
had the most stable electrochemical behavior; they achieved the optimum
current/voltage quality ratio with minimum gas bubble formation.

MHD observations

The fluorescence microscope pictures in Fig. 3.5 show the movement of

polystyrene beads under the influence of electric and magnetic fields. A
clear change of the flow profile is observed when the magnetic field was
present. The vertical channel in Fig. 3.5 is the side-channel over which the
voltage is applied. The horizontal channel is the main channel on which the
Lorentz Force should act. Note that no movement in the main channel is
observed when only the magnetic field is applied.

The measurement of bead velocity under the influence of MHD was made
difficult by the presence of the strong electrokinetic flow in the side-channel.
We made visual estimations of bead velocity, that was 0.1 mm · s-1 for 0.1
M KCl with 0.3 × 0.26 × 8 mm3 channels with a current density of 6410
A·m-2 (I = 0.5 mA).

We calculated the MHD velocity corresponding to our specific experimen-

tal conditions and chip geometry with the help of equation 2.16 and the
expression of the hydraulic resistance for rectangular channels:
3.2. MHD IN A PDMS T-CHANNEL 49

Figure 3.4: Measured current as function of channel width and applied

voltage. The channels were 50 µm deep with 1 M NaCl solution.
50 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

w3 h2
vMHD = J · B · (3.1)
8µL(w + h)2

where w (= l) and h are the channel and side channel’s width and height,
L is the total length of the channel, and µ is the absolute viscosity of the
solution in the channel. The calculated MHD speed for such experimental
conditions, using Equation 3.1 was 0.2 mm·s-1 . We see that the measured
MHD velocity is half the calculated one. This is due to a MHD flow per-
turbation by the strong electrokinetic flow in the side-channel.

The ionic current generates unwanted high and uncontrollable electroki-

netic flow from the side channel to the main channel, thereby not allowing
a clear visualization of the MHD pumping. Therefore, we designed a new
microfluidic device to avoid such unwanted and uncontrollable flows that
arise from the side channels.

Figure 3.5: (Left) Flow directions in presence of electric field alone. Beads
are seen to move from top to bottom, following the electric field lines.
(Right) Flow directions in presence of both electric and magnetic fields.
Beads are seen to move also from right to left, showing a change in behav-
ior when the magnetic field is added to the electric field.
3.3. PDMS CHANNEL ON TOP OF PLATINUM ELECTRODES 51

3.3 PDMS channel on top of platinum electrodes

This second generation of microfluidic devices was designed and fabricated
without side channels to avoid the unwanted electrokinetic flow that dis-
turbed the MHD flow in our previous chip.

Fabrication
Top and side views of the device are shown in Fig. 3.6. We used PDMS
material for the channels, and we patterned the electrodes directly on the
glass wafer that was used for PDMS sealing. We used platinum instead
of silver/silver chloride for the micromachined electrodes, as a standard
procedure to deposit silver was not available in the cleanroom. We encoun-
tered problems aligning the PDMS channels to the electrodes. We observed
a change in the dimensions of the PDMS channel (clear decrease) after it
was peeled from the silicon master. Only straight channels that were twice
as wide (200 µm) as the space between the electrodes (100 µm) could be
aligned. Another design with circular electrodes and channels could not be
aligned properly due to this change in PDMS dimensions (compared with
electrode design) and was thus not evaluated.

Experiments
The channels were 50 µm deep and 200 µm wide. The total length of the
channels was 20 mm. The electrodes were made of platinum on top of an
adhesion layer of titanium. Electrodes were 100 nm thick, 100 µm wide
each, 100 µm apart, for a total length of 20 mm. Since the channel was
wider than the space between the electrodes, it left a significant electrode
area in contact with the solution on both sides. This area is the product of
the mean electrode width inside the channel (50 µm) by the total length of
the electrodes (20 mm): that is 1 mm2 . DC voltage varied from 0 to 4 V.
Gas bubbles were generated within the channel above 2.5 V. Those bubbles
forced the solution out of the microchannel, not allowing higher voltages.
As for the first experiment, the DC magnet was placed below the channels
in contact with the coverplate. The intensity of the magnetic field in the
52 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Figure 3.6: (Top) Top and cross-sectional sketches of the PDMS channel on
top of the Pt electrodes. (Bottom) Top view of the PDMS channel aligned
on the electrodes.
3.3. PDMS CHANNEL ON TOP OF PLATINUM ELECTRODES 53

channel was 0.4 T. Solutions of 1 M NaCl and 100 mM NaOH were used
for the experiments.

Current generation

We characterized the ohmic behavior of the solutions, by observation of the

current generated as a function of the applied DC voltage (see Fig. 3.7).
The most stable and high current intensities were obtained for voltages of
2.5V.
The standard current value was 47 µA for 1 M NaCl and 2.5 V. This cor-
responded to a calculated current density of 47 A·m-2 . We measured the
current as a function of the chemical composition of the electrolyte inside
the channel. The current measured is clearly dependent on the nature of the
electrolyte between the electrodes (see Fig. 3.7).

MHD observations

The expected MHD velocity was calculated from Equation 2.16. For such
experimental conditions, the MHD velocity follows the relation:

l w2 h2
vMHD = J · B · (3.2)
L 8µ(w + h)2
where w and h are the channel’s width and height (resp.), l and L are the
total length of the electrodes and channel (resp.). Using equation 3.2, the
expected velocity was calculated to be 3.8 µm·s-1 for applied current density
of 47 A·m-2 , 0.4 T magnetic field, and 1 M NaCl.

We did not observe any MHD-related movement from the first MHD exper-
iments with this chip. No bead movement due to MHD could be observed.
This may be from the lack of uniformity in the electrode area, which is due
to the difficulty to align the electrodes with the channels.

We designed and fabricated a third device based on the 2nd generation. We

integrated the channels and electrodes on the same Pyrex wafer to avoid the
54 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

alignment problem. Pyrex was chosen as the substrate, because it will be

the material needed for the final experiments in the NMR setup.

3.4 Glass channel with electrodes on the inner wall

We used glass as the material for the third device generation. We integrated
the electrodes in the channels. Because the current generated is propor-
tional to the electrode surface, we designed a device with a large electrode
area (see Fig. 3.8).

Figure 3.7: Measured current as function of electrolyte nature and applied

voltage. In this case, three electrolytes are compared, namely 1 M NaCl,
100 mM NaOH, and DI water.
3.4. GLASS CHANNEL WITH ELECTRODES ON THE INNER WALL55

Figure 3.8: Top view of a glass channel with the Pt electrodes on the walls
of the channel. The channels presented here are 150 µm wide.
56 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Fabrication

To fabricate this microfluidic device, the channels were first etched in glass.
Then the platinum electrodes were deposited on the wall of the channels by
a lift-off approach (see Fig. 3.9 for details about the fabrication process).

Experiments

The channels were 75 µm deep. Two channels widths were fabricated,

namely 150 and 1000 µm. The total length of the channels were 40 mm
and 120 mm. The electrodes were made of platinum (on top of an adhesion
layer of titanium), and followed the channel’s walls (see Fig. 3.10 for a
close-up view of the inside of the channel). Electrodes were 20 mm long,
60 µm wide, and about 80 µm to 1000 µm apart (depending on the width
of the channel). Electrode area was 1.2 mm2 . DC voltage varied from 0 to
2V, as gas bubbles were generated after 2 V. As for the other experiments,
the DC magnet was placed below the channels in contact with the wafer.
The expected intensity of the magnetic field was also around 0.4 T. The
electrolyte solution chosen for the experiments was 0.1 M KCl in DI water.

Current generation

The maximum measured current in the channel, before bubble generation,

was 10.5 µA for UDC = 2 V. This corresponds to a calculated current density
of 9 A·m-2 . This very low value for the generated current intensity is due
to polarization effects at the electrode-electrolyte interface. If the voltage
applied exceeded a certain threshold, water electrolysis occurred between
the electrodes, leading to bubble formation inside the channel.

MHD observations

We calculated the expected MHD velocity with Equation 2.16. For such
experimental conditions and chip geometry (half-circular glass channel),
the expected MHD velocity follows the relation:
3.4. GLASS CHANNEL WITH ELECTRODES ON THE INNER WALL57

Figure 3.9: Schematic overview of the fabrication of the third generation

device: 1) Start with Pyrex-glass wafer with 400 nm of polysilicon. 2)
Photolithography, RIE etching of polysilicon mask, and deep etching with
20% HF to form the glass channels. 3) Removal of polysilicon, spray depo-
sition of photoresist, and photolithography using the alignment marks. 4)
Evaporation of metals (200 Å Ti + 1000 Å Pt) on the substrate, followed by
lift-off of the metals with acetone and ultra-sound.
58 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

l 3h2
vMHD = J · B · (3.3)
L 64µ

where h is the channel height, l and L are the total length of the electrodes
and channel, respectively. With the help of Equation 3.3 for an applied
current density of 9 A·m-2 , 0.4 T magnetic field, and 0.1 M KCl a maximum
flow velocity of 0.5 µm·s-1 is expected, which is too low to be observed.

Water electrolysis is an unwanted electrochemical phenomenon. It will

deteriorate the electrode surfaces, and the generated bubbles will force the
liquid out of the channels. Due to this electrolysis problem, it was not
possible to generate a current of sufficient strength to actually observe any
flow due to MHD in those channels.

Figure 3.10: eSEM pictures of a channel cross-section

3.4. GLASS CHANNEL WITH ELECTRODES ON THE INNER WALL59

AC Current generation

In order to obtain a better understanding of the device, we switched to AC

voltage, which is typically used in electrochemistry to avoid polarization ef-
fects and bubble formation. We found that the optimum working frequency
was 7-10 kHz (see Fig 3.11).

Figure 3.11: Current frequency spectra for an applied RMS voltage of 2 V

(0.1 M KCl, 1 mm wide channels).

We could achieve a 150 fold higher current density JRMS = 1333 A·m-2 (IRMS
= 1.6 mA at URMS = 2 V and f = 10 kHz) between the electrodes inside the
channel compared with the DC mode for the same voltage. The maximum
AC current density was 6667 A·m-2 (IRMS = 8 mA for 0.1M KCl, URMS =
14 V and f = 7 kHz). With the help of Equation 3.3, the expected veloci-
ties were: vMHD = 70 µm·s-1 for applied current density of 1333 A·m-2 and
vMHD = 0.35 mm·s-1 for applied current density of 6667 A·m-2 for a 0.4 T
60 CHAPTER 3. DC MHD PUMP DEVELOPMENTS

Table 3.1: Comparison of current density and MHD flow velocity for the 3
different designs at maximum voltage and B = 0.4 T (the AC current was
measured at f = 10 kHz). The expected body force (∆PMHD ) generated by
the pumps was calculated thanks to relation 2.14.

1st Design 2nd Design 3rd Design

DC DC DC AC
Max. Voltage [V] 20 2.5 2 14
Electrode distance [µm] 8000 100 80 80
Electrode area [mm2 ] 0.08 1 1.2 1.2
KCl concentration [M] 0.1 1 0.1 0.1
Max. Current [µA] 500 47 10.5 8000
Max. Current density [A·m-2 ] 6410 47 9 6667
Length of electrodes (l) [mm] 0.3 20 20 20
Expected vMHD [mm·s-1 ] 0.2 4 · 10−3 5 · 10−4 0.35
Measured vMHD [mm·s-1 ] 0.1 na na na
Expected ∆PMHD [Pa] 0.8 0.4 0.1 53

magnetic field strength, and 0.1 M KCl electrolyte solution.

We didn’t perform MHD pumping experiments in the AC mode, since it

was not the final aim of this thesis. But this design is seen to be perfect to
achieve AC MHD pumping with microchannels made in glass.

3.5 Summary

This study of MHD pumps with electrodes located in direct contact with
the pumping channel has enabled us to have a preliminary understanding
of the MHD phenomenon in an electrolyte environment. Table 3.1 shows a
comparison of the electrical properties and expected MHD flow velocities
for each of them.
3.5. SUMMARY 61

Experimental observations of the MHD phenomenon in electrolyte envi-

ronments was possible only with the first generation of MHD chip (see sec-
tion 3.2). However, we encountered difficulties to observe a proper MHD
pumping in a single managed direction due to the channel geometry (cross
intersection). The second and third designs demonstrated the possibility to
integrate electrodes with a large area inside the pumping channel, but the
threshold voltage before the generation of electrolysis bubbles at the elec-
trode/electrolyte interface limited their applicability. The maximum current
density generated by the chips was not sufficient to observe any MHD re-
lated flow.

All three studies demonstrated that high current densities (500 A·m-2
or more) in the DC mode required voltages greater than 2V, that is above
the threshold for electrolysis of aqueous solutions. The following chapter
shows how it is possible to generate an MHD flow with voltages leading
to electrolysis, but without any perturbation due to electrokinetic flow from
the side channel or bubble formation.
62 CHAPTER 3. DC MHD PUMP DEVELOPMENTS
Bibliography

[1] D.R. Lide. C.R.C. Handbook of Chemistry and Physics. 76th edition,
1995.

[2] R.G. Bates. determination of pH. John Wiley and Sons Ltd., 1973.

[3] P.J. Elving, J.M. Markowitz, and I. Rosenthal. Preparation of buffer

systems of constant ionic strength. Analytical Chemistry, 28:1179–
1180, 1956.

[4] T.C. McIlvaine. A buffer solution for colorimetric comparison. Journal

of Biological Chemistry, 49:183, 1921.

63
64 BIBLIOGRAPHY
Chapter 4

A high current density DC

MHD micropump

This chapter is adapted from:

A high current density DC magnetohydrodynamic (MHD) micropump

Alexandra Homsy, Sander Koster, Jan C.T. Eijkel, Albert van den Berg, F.
Lucklum, E. Verpoorte, and Nico F. de Rooij, Lab On a Chip, Volume 5,
2005, pp. 466-471.

A high flow rate DC magnetohydrodynamic (MHD) micropump

Alexandra Homsy, Sander Koster, Jan C.T. Eijkel, Albert van den Berg, E.
Verpoorte, and Nico F. de Rooij, Presented at the Micro-TAS 2004 Confer-
ence, Malmo, Sweden, Proceedings pp. 372-374, Sept. 26-30, 2004.

65
66CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

This chapter describes the working principle of a DC magnetohydrody-

namic (MHD) micropump that can be operated at high DC current densities
(J) in 75-µm-deep microfluidic channels without introducing gas bubbles
into the pumping channel. The main design feature for current generation
is a micromachined frit-like structure that connects the pumping channel
to side reservoirs, where platinum electrodes are located. Current densities
up to 4000 A·m−2 could be obtained without noticeable Joule heating in
the system. The pump performance was studied as a function of current
density and magnetic field intensity, as well as buffer ionic strength and
pH. Bead maximum velocities of up to 1 mm·s−1 (0.3 µL·min−1 ) were ob-
served in buffered solutions using a 0.4 T NdFeB permanent magnet, at an
applied current density of 4000 A·m−2 . This pump is intended for transport
of electrolyte solutions having a relatively high ionic strength (0.5 - 1 M) in
a DC magnetic field environment. The MHD micropump presented here is
made in glass, and is operated in the DC mode. Gas bubble formation due
to DC operation does not influence the performance of the pump since the
bubbles are generated outside the pumping channel. This pump is intended
to be integrated together with planar NMR coils in a microfluidic system
for NMR analysis [1]. Ultimately, the pump will take advantage of the high
DC magnetic field (7 T) necessary for NMR analysis to pump and manip-
ulate the fluids in the chip. In the NMR environment, a minimum 16-fold
increase in volumetric flow rate for a given applied current density is ex-
pected. The application of this pump for the study of biological samples in
a miniaturized total analysis system (µTAS) with integrated NMR detection
is foreseen.

4.1 New design concept

The first generation of MHD chips (see Section 3.2) enabled the generation
of high DC current across electrodes connected by one deep side-channel
without introducing gas bubbles in the pumping channel. However, the
electrokinetic flow of buffer between the electrodes significantly disturbed
the MHD flow itself. All the designs of MHD micropumps presented in
the literature (and in previous chapters) have integrated the electrodes di-
4.1. NEW DESIGN CONCEPT 67

Figure 4.1: Schematic diagram comparing electrode location and chip de-
sign for previous MHD pump designs (left) and the new concept (right).

rectly into the pumping channel for current generation. When aqueous so-
lutions are actuated with a DC voltage source, bubble formation around
the electrodes (due to water electrolysis) is unavoidable1 . In order to avoid
disruption of flow due to bubble formation in DC operation, we designed
two micromachined frit-like structures which connect the pumping chan-
nel to adjacent side reservoirs. The electrodes are actually located in these
side reservoirs, and are thus physically separated from the main channel
in which flow is generated. In this way, bubbles formed at the electrodes
never enter the main channel, escaping instead through open reservoirs to
the atmosphere. A comparison of the present design concept with that of
previously published MHD pumps is shown schematically in Fig. 4.1.
The idea for the adopted design is based on [2], where a similar ge-
ometry was used to perform free-flow electrophoresis on silicon chips. A
high ionic current with a very low volumetric flow rate can be generated
over the main channel across a network of very shallow frit-like structures
(these will be referred to as "side channels" from now on). When this ionic
current is coupled to a perpendicular magnetic field, a body force is gen-
erated in the perpendicular direction, all along the pumping channel. With
this new design concept, we have been able to generate a current density
in the pumping channel that is 50 times higher, in channels with heights 4
times smaller, than that reported by others in closed channels [3].

1 The only solution to avoid bubble generation at the electrode surface in these cases
would be to choose an appropriate selection of electrolytes and electrode material (redox
species) yielding a non-gaseous product in the electrode reaction.
68CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

4.2 Experimental Methods

4.2.1 Micropump fabrication

Pyrex 7740 wafers (Sensor Prep Services Inc.) were used as substrates for
the fabrication of the MHD pump. As the MHD flow rate is proportional
to the current density (J), the chip was designed to generate a high J across
the channel pumping area, without introducing bubbles into the channel.
Details of the chip design can be seen in Fig. 4.2. Along both sides of
the main channel (22 mm x 150 µm x 75 µm), arrays of 800 perpendicular
side channels (10 µm x 10 µm x 100 nm) were etched. These channels
connected the main channel to the side-reservoirs (16 mm x 2.5 mm x 75
µm), in which the platinum electrodes were inserted.
Both main channel and side channels are processed on the same Pyrex
wafer by a two-step photolithography process in our cleanroom facility.
The 75-µm-deep main channel and reservoirs were etched first with 20 %
HF, using polysilicon (2 depositions at 2000 Å each) as an etch mask. Then,
polysilicon was removed in a 60˚C KOH etch bath, and AZ4562 photoresist
was sprayed (Airbrush Rotring CH) on top of the wafer. A photolithogra-
phy mask with the side channel openings was then aligned to the wafer with
the help of a mask aligner (Electronic Vision AL6-2, Schaerding, Austria).
After UV exposure and development, the photoresist served as an etch mask
for the buffered HF etch (BE 7-1 NC1 from Laporte Electronics) performed
to obtain the shallow 100-nm-deep side channels. The channels were then
sealed by fusion bonding to a Pyrex wafer coverplate with drilled holes
to introduce fluids to the main channel and side reservoirs. A piece of 5-
mm-thick polydimethylsiloxane (PDMS - Sylgard 184 from Dow Corning,
Distrelec AG, Switzerland) was reversibly sealed to the top of the chip to
provide the outer liquid reservoirs (see Fig. 4.3). A pair of holes (4 mm
diameter) in the PDMS layer was aligned over the microchannel inlet and
outlet, and platinum wire electrodes were placed in a wide opening (5 x 5
x 5 mm3 ) positioned over the side reservoirs.
 4.2. EXPERIMENTAL METHODS

Figure 4.2: Schematic cross-sectional view of the pumping area. The main channel is 22 mm long, 150
µm wide and 75 µm deep. A 16-mm-long segment of the channel was used as the MHD pump. The side
channels are 100 nm deep, 10 µm wide and 10 µm long.
69
70CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

4.2.2 Pumping experiments

Citrate-phosphate buffer (53 mM citrate, 207 mM phosphate) with a pH
varying from 3 to 7 was chosen for these experiments. KCl was added to the
buffer to increase the ionic strength of the solution to fixed values of 0.5 M
and 1 M. Fluid velocity was observed using a stereoscopic microscope (Le-
ica MZ12) connected to a digital camera (JVC TK-C1380). A 10× magni-
fication was used. Polystyrene beads (Polysciences, carboxylate-modified
microspheres, 2.55% in water) with a diameter of 6 µm were diluted 500
times in the buffer for flow visualization. The linear flow velocity was de-
termined by measuring the time required for beads to travel between two
defined points in the pumping channel. Only beads having a maximum ve-
locity (i.e. beads located in the middle of the channel) were considered.
The error associated with the measurement was calculated for 10 beads in
a total sample of 50 beads per observation. A constant voltage (0 - 60 V)
was applied across two platinum electrodes by a DC power supply (HP
E3617A). A multimeter (Keithley 175) was used to monitor the current.
All measurements were performed under conditions of constant current,
which ranged between 0.6 mA and 5 mA, depending on the experiment. A
NdFeB rare earth permanent magnet (Forcefield, USA) was placed under
the device. Three magnets, having three different magnetic field intensities
(namely 0.22, 0.32 and 0.42 T) were used during the measurements. The
dimensions of the magnets (50 mm diameter) were larger than the outer
dimensions of the chip. The magnet was placed so that it was almost in
contact with the chip, in order to assure good homogeneity of the magnetic
field in the channel. The magnetic field intensities were measured with a 3-
axis magnetic-field transducer (model 3MR-1C, Sentron AG Switzerland).

4.3 Results and discussion

The MHD device presented in Fig. 4.3 has a total of 4 reservoirs. Two are
the side reservoirs, where the Pt electrodes were positioned. The two other
reservoirs, perpendicular to the sidechannel reservoirs, were connected to
the inlet and outlet of the main channel where the pumping was induced.
4.3. RESULTS AND DISCUSSION 71

Filling of the device with buffer began by filling the main channel, followed
by the side reservoirs. The side channels were filled automatically by cap-
illary action when liquid was introduced to the main channel. No problems
related to the trapping of air bubbles were encountered when side channels
were filled. For flow visualization, the main channel was filled with the
buffer containing 6 µm polystyrene beads.
When the DC voltage was applied between the platinum electrodes,
a net electric current in the mA range was measured, increasing linearly
with the voltage. We used a buffered solution to avoid large pH changes
around the electrodes. Gas bubbles that formed around the platinum elec-

Figure 4.3: View of the chip with PDMS on top for the inlet/outlet reser-
voirs as well as electrode reservoirs. The Pyrex coverplate bonded on top
of the processed wafer contains a standard grid of holes for generic use in
microdevice fabrication. In our case, chip operation relies on only 10 of
these holes.
72CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

trodes could easily escape through the large holes in the PDMS. Gas bub-
bles were never seen entering the side reservoirs (and therefore the side
channels). Bead movement was not observed when only an electrical field
or a magnetic field was applied at a buffer ionic strength between 0.5 and
1 M. However, when the current was applied across the channel together
with a perpendicularly applied magnetic field, beads began to move above
a current density threshold of 500 A·m−2 (corresponding to a voltage ≥ 3.5
V). Beads located in the center of the channel were seen to move faster than
the beads on the sides close to the channel walls, which is typical behavior
for flows with a parabolic profile. Fig. 4.4 shows the maximum bead ve-
locity in the main channel as a function of J for 3 different magnetic field
intensities. The bead velocity was seen to increase with B. Measurements
with more values of B would be needed to confirm the linear behavior with
increasing B. This observation is of interest, as we intend to integrate this
pumping method into a chip for integrated NMR analysis using a 7 T mag-
net [1]. The expected velocity at 7 T is therefore also shown in Fig. 4.4.
Assuming that flow velocity varies linearly with B (as predicted by Equ.
2.16), we expect a minimum 16-fold increase in bead velocity for a given
current density value in a 7 T NMR environment. In other words, much
lower actuation voltage would be required to obtain a given velocity.

The bead velocity was also seen to increase with J. Maximum observed
velocities were 1 mm·s−1 for a magnetic field intensity of 0.42 T and a
current density of 4000 A·m−2 . Measured velocities were on the same or-
der of magnitude as theory. However, measured velocities were less than
expected from theory for low J values, while in the higher current den-
sity range, beads exhibited velocities higher than theory. The main sources
for these deviations are believed to come from the back flow induced by
hydrostatic pressure (at low J), and Joule heating (at high J). The relative
contributions from these effects will depend on the total pressure induced
by MHD and on the power dissipation capability of the chip.
4.3. RESULTS AND DISCUSSION 73

Figure 4.4: Maximum velocity of 6 µm carboxylate-modified polystyrene

beads as a function of applied current density for 3 different magnetic field
intensities in a citrate phosphate buffer of pH = 5 (207 mM Phosphate, 53
mM Citrate), with 744 mM KCl added to increase the ionic strength of the
buffer. Error bars indicate the standard deviation for 10 beads of maximum
velocity in a total sample of 50 beads per measurement. The theoretical
curves were calculated using Equation 2.16.
74CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

4.3.1 Flow limitation due to back pressure

Equation 2.14 describes the theoretical pressure generated by MHD. A flow
in the direction opposite to MHD was often observed when measurements
were stopped. Back pressure resulting from this flow will depend on the
total measurement time, as this determines the total liquid volume trans-
ported from one reservoir to the other. Height differences of liquid between
inlet and outlet reservoirs can result when significant amounts of liquid have
been pumped, with more liquid in the outlet reservoir than in the inlet reser-
voir. This can result in a hydrostatic pressure difference, or back pressure,
acting against the MHD flow. The back pressure was measured at the end
of each MHD experiment by observing the velocity of the beads traveling
in the opposite direction after a total measurement time, t. The pressure of
the backflow was estimated with the help of Equations 2.10, 2.14 and 2.12.
It is obvious that a more accurate method to determine the back pressure
would involve integration of a pressure sensor into the chip. As velocities
are additive, the measured bead velocity will be the velocity induced by
MHD minus the velocity gradually induced by the back pressure. In terms
of pressure, the total measured pressure (∆PT OT ) will be equal to the pres-
sure induced only by MHD (∆PMHD ) minus the back pressure (∆PBAK ):

∆PT OT = ∆PMHD − ∆PBAK (4.1)

Table 4.1 compares the back pressure with the measured pressure, and
the pressure expected from MHD pumping (for a 1 M buffer solution of
pH = 5, and magnet field strength of 0.4 T). It can be concluded that back
pressure plays a significant role in the total pressure of the system (up to
22% of total measured pressure in this specific case), but it is not large
enough to stop MHD flow under a reasonable total measurement time.

4.3.2 Joule heating and power dissipation

At high current densities, the flow velocity was observed to be higher than
that calculated. One explanation is that the temperature in the main channel
began to rise because of Joule heating occurring in the side channel area.
4.3. RESULTS AND DISCUSSION 75

Table 4.1: Comparison of the pressure generated and of the calculated

power dissipation Qd in the pump as a function of applied current den-
sity. Each measurement was performed for a time, t. The pumped solution
was a 1 M citrate-phosphate buffer of pH = 5; magnetic field intensity of
0.4 T.
U J t ∆PTOT ∆PBAK ∆PMHD ∆PMHD THEO Qd
(V) (A·m−2 ) (min) (Pa) (Pa) (Pa) (Pa) (W·m−1 )
5 494 28 5 0 5 3 0.1
10 1725 25 8 1 9 11 1.0
15 2549 21 15 3 19 16 2.5
20 3620 10 28 1 29 23 4.9

Above a certain threshold, the temperature will decrease the viscosity of

the pumped solution, leading to an increase in flow velocity. Figure 4.5
shows the Ohm plot measured under conditions corresponding to the mea-
surements of Fig. 4.4. The current density begins to deviate from linearity
after 2500 A·m−2 . This shows that Joule heating may affect our measure-
ments of MHD bead velocities above this current density threshold for an
applied voltage of 12 V with such a buffer solution. The last column of
Table 4.1 shows the power dissipation (Qd ) along the side channels arrays
for the same experimental conditions as in Fig. 4.4.

Power dissipation is widely used as an indicator for temperature increases

due to Joule heating (see section 2.3.1) in discussions about efficiency in
electrophoresis separations [4]. Applied to our specific geometry, this indi-
cator is estimated as follows:

(U − 2) · I
Qd = (4.2)
lcap
where lcap is the total length of the side channels2 [m], U is the voltage ap-
2 This value is the sum of the lengths (10 µm) of all the 800 sidechannels located on both

sides of the main channel.

76CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

Figure 4.5: Current-Voltage behavior of the chip during the velocity exper-
iments depicted in Fig. 4.4.
4.3. RESULTS AND DISCUSSION 77

plied across the system [V], and I is the current [A] across the side channels.
In microfluidic systems, Joule heating was observed to occur for Qd val-
ues [5] larger than 3 W·m−1 (instead of 1 W·m−1 for standard capillaries).
With our specific geometry, the power dissipation threshold corresponding
to the beginning of Joule heating is seen at 2.5 W·m−1 . This threshold is
reached if the applied voltage is higher than 12 V in 1 M buffer solutions.

4.3.3 MHD flow dependence on buffer ionic strength and pH

Table 4.2 shows values of the measured maximum bead velocities at a mid-
range current density for various pH and ionic strengths of the buffer. The
last column of the table shows the theoretical maximum velocity (calcu-
lated from Equation 2.16) for the corresponding experimental parameters.
A strong dependence of bead velocity on both current density and magnetic
flux density is observed for all buffer conditions. Comparing values at a
fixed ionic strength but different B shows in fact that there is very little, if
any, variation of bead velocity as a function of pH at a given ionic strength
and B. The beads used for the velocity measurements were carboxylate-
coated. The pKa of carboxyl groups has a mean value of 4.9 [6]. The beads
which gave the most reproducible measurements were the ones handled at
pH ≤ pKa. By comparing the theoretical and measured velocity values, we
see that the results are of the same order of magnitude, and follow the same
main variations as a function of the magnetic field.

At lower ionic strengths (0.01 - 0.1 M), a flow in the J direction was
observed, adversely affecting MHD flow in the main channel. This flow is
most likely electro-osmosis, originating in the side channel arrays3 . Efforts
are currently focusing on improvement of the chip design and implemen-
tation of surface chemistry to reduce electro-osmotic flow, in order to de-
crease these effects. Once the issue of unwanted electroosmotic flow has
been addressed, our MHD pump (in contrast with electroosmotic pumps)
will be suitable for use with solutions of relatively high ionic strengths (≥
3 Those effects lead to a non wanted high volumetric flow across the side channels, which

stops the MHD flow in the perpendicular direction.

78CHAPTER 4. A HIGH CURRENT DENSITY DC MHD MICROPUMP

Table 4.2: Comparison between theoretical maximum MHD flow velocity

and measured maximum bead velocities (mm·s−1 ). For a current density of
2083 A ·m−2 , and citrate phosphate buffer with KCl added for fixed ionic
strength
Ionic Strength Observed velocity Observed velocity Calculated
0.5 M pH=3 pH=5 velocity
0.22T 0.20±0.01 0.19±0.01 0.17
0.42T 0.28±0.02 0.33±0.02 0.34

Ionic Strength Obs. velocity Obs. velocity Obs. velocity Calculated

1M pH=3 pH=5 pH=7 velocity
0.22T 0.10±0.02 0.12±0.01 0.12±0.02 0.17
0.32T 0.20±0.02 0.22±0.02 0.22±0.01 0.26
0.42T 0.31±0.02 0.32±0.03 0.28±0.01 0.34

0.1 M). This could prove advantageous for physiological applications re-
quiring higher buffer concentrations.

4.4 Conclusion
We have demonstrated a novel concept for a DC MHD micropump that al-
lows pumping with a maximum velocity of up to 1 mm·s−1 (0.3 µl·min−1 )
in 75-µm-deep glass microchannels with a 0.42 T permanent magnet. The
chip design permitted the use of DC current without gas bubble perturba-
tion in the pumping channel, by using a micromachined frit-like structure
to provide electrical contact between the main channel and side reservoirs,
where the electrodes were located. Future work with these devices will
include a study of MHD flow in the high DC magnetic field of the NMR
instrumentation, and its integration into a µTAS for on-chip NMR analy-
sis. Such a µTAS will enable online monitoring of chemical reactions, and
analysis of biomolecules in aqueous phase.
Bibliography

[1] C. Massin, F. Vincent, A. Homsy, K. Ehrmann, G. Boero, P.-A. Besse,

A. Daridon, E. Verpoorte, N.F. de Rooij, and R.S. Popovic. Planar
microcoil-based microfluidic NMR probes. Journal of Magnetic Reso-
nance, 164:242–255, 2003.

[2] D.E. Raymond, A. Manz, and H.M. Widmer. Continuous sample pre-
treatment using a free-flow electrophoresis device integrated onto a sil-
icon chip. Analytical Chemistry, 66:2858–2865, 1994.

[3] J. Jang and S.S. Lee. Theoretical and experimental study of MHD
magnetohydrodynamic micropump. Sensors and Actuators A, 80:84–
89, 2000.

[4] R.J. Nelson, A. Paulus, A.S. Cohen, A. Guttman, and B.L. Karger. Use
of Peltier thermoelectric devices to control column temperature in high-
performance capillary electrophoresis. Journal of Chromatography,
480:111–127, 1989.

[5] A.M. Garcýa Campana, W.R.G. Baeyens, H.Y. Aboul-Enein, and

X. Zhang. Miniaturization of capillary electrophoresis systems us-
ing micromachining techniques. Journal of Microcolumn Separations,
10(4):339–355, 1998.

[6] S.H. Behrens and D.V. Grier. The charge of glass and silica surfaces.
Journal of Chemical Physics, 115(14):6716, 2001.

79
80 BIBLIOGRAPHY
Chapter 5

Electrical description of the

pump

This chapter, complementary to Chapter 4, will present an electrical model

specific to our pump geometry. As we used charged beads for flow veloc-
ity measurements, both beads and ions in the solutions were subjected to
MHD, electroosmosis (eo) and electrophoresis (ep). Eo and ep give rise to
a flow perpendicular to MHD. The understanding of all the electrical field
processes is important for the interpretation of our flow observations. The
model we present in this chapter will help us predict and quantify the effect
of the transverse electrokinetic flow. These secondary flows may be the
cause for the perturbation of MHD pumping at low ionic strength and even
for the loss of sample species from the main channel.

5.1 Model description

We used a simple resistor model to explain the electrical behavior of our
pump. When a voltage is applied along a channel filled with a fluid, the
electrical resistance is by definition:

d
R = %· (5.1)
A

81
82 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

where % is the measured resistivity of the solution [Ω·m], d is the length of

the channel [m] with a cross-sectional area A [m2 ]. Figure 5.1 shows the
equivalent electrical scheme of the pump. Assuming that this equivalent
scheme is an approximation, we can use it to describe the impedance of the
pump at low and higher DC voltages. At low voltages (i.e. in the mV range)
the electrical behavior of the pump is capacitive (with a capacitance Cdl )
whereas at higher voltages it is faradic. The high DC current generation
condition for the pump involves application of voltages higher than 5 V. In
this range1 , the electrical behavior of the pump will always be resistive. If
we apply the well-known Kirchhoff rules associated with Ohm’s law to our
circuit, the total voltage U applied to the system will be:

U = Ui −U f = Ui (if U f = 0)
= ∆Ucha + 2∆Ures + 2∆Usid (5.2)
= Rtot · I

where:
Rtot = Rcha + 2Rres + 2Rsid (5.3)
Rsid is the total electrical resistance of the side-channels in parallel, and if
we have n side channels of same resistance Ri :
n
1 1 n
=∑ = (5.4)
Rsid i=1 Ri Ri

The current density J across an area AJ is related to the current:

I
J= (5.5)
AJ
The above expressions allow us to define the desired geometric parameters
of the pump such as the total number of side-channels as well as their length
and width. The MHD flow rate is directly proportional to J in the main
channel. The value for J generated in the main channel should be as high
1 An obvious second condition is the absence of Joule heating (see sections 2.3.1 and

4.3.2)
5.1. MODEL DESCRIPTION 83

Figure 5.1: Schematic top view of the pump and equivalent electrical
scheme. Each channel geometry has a different electrical resistance: Rres
is the resistance of the side-reservoir area, Rcha is the resistance of the main
channel, and Rsid is the resistance of all side-channels in parallel in be-
tween one side-reservoir and the main channel. The electrical resistance
of the buffer Rsol in parallel with the capacitance Cdl represents the elec-
trodes. Cdl , the double layer capacitance, stands for the polarization of the
electrodes at the electrode-electrolyte interface.
84 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

as possible with a corresponding voltage as low as possible. Since the intent

of the side channels is to have a very high hydraulic resistance, the depth
is fixed to be as shallow as possible (100 nm). With such a small depth,
about 98% of the applied voltage will drop on both sides of the pumping
channel, at the location of the side-channels. With the parameters of our
pump geometry, the electrical resistance of one array of side-channels is
expressed as follows:
Ri % ds
Rsid = = · (5.6)
n n hs ws
where ds (10 µm) is the length, ws (10 µm) is the width, and hs (100 nm) is
the depth of the n (800) side-channels. The main channel has a depth h (75
µm), a width d (150 µm), and a length l (1.6 cm, along which the current is
applied).
d
Rcha = % · (5.7)
hl
Rsid ds hl
= · = 100 (5.8)
Rcha n hs ws d
We see that the electrical resistance of the array of side channels is 100
times higher than of the main channel, this confirms that the majority of the
voltage will drop over the array of side-channels.

Electric field
Another important parameter in our system is the intensity of the elec-
tric field at various locations in the pump. Obviously, E across the side-
channels is much higher than in the main channel. A large E value would
be a problem for MHD flow, as it would induce perpendicular flows through
electrokinetics (see section 2.3.2).

The electric field across the main channel (Echa ) is expressed as:

∆U U3 −U2 Rcha · I %
Echa = = = = ·I (5.9)
d d d h·l
The electric field across one array of side-channels (Esid ) is:
5.1. MODEL DESCRIPTION 85

U2 −U1 Rs I % I
Esid = = = · (5.10)
ds ds n hs ws
To compare Esid with Echa , we take the ratio of the two expressions as a
function of I:

Esid hl
= (5.11)
Echa n hs ws
And applied to our specific pump geometry,

Esid 75 · 10−6 · 1.6 · 10−2

= = 1500 (5.12)
Echa 800 · 100 · 10−9 · 10 · 10−6

We notice that the electric field in the side channels is 1500 times more
intense than the one in the main channel.

Electrokinetic flows
The effects of an applied electric field on a fluid are described in paragraph
2.3.2. It is important to note that electrokinetic (ek) flows are proportional
to the electric field despite the geometry of the microchannel (ie. the depth
of side channels). This means that above a certain eof velocity threshold,
the MHD flow in the pumping channel is likely to be perturbed by the
perpendicular ek flows originating from the side channels. The ek flows
that perturb MHD pumping are the ones generated by the electric field in
the main channel (Echa ).

The expression for the MHD velocity is compared with the bulk ek flow
velocity of eof. We first express equation 2.16 as a function of our specific
geometry, and then replace J by 2.24 to extract the dependence on Echa .

3 h2 l 3 h2 l B
vMHD = J ·B = Echa (5.13)
64 µ L 64 µ L %
The viscosity µ and electric field Echa were eliminated by expressing the
ratio of vMHD over the eo flow velocity expressed in chapter 2.3.2:
86 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

vMHD 3 h2 l B 1
= (5.14)
veo 64 L % ε ζ
The ratio vMHD
veo is inversely proportional to ζ and %. The zeta potential [1, 2]
should therefore be minimized in order to avoid perturbations of MHD flow
due to eof. High ionic strength solutions, having smaller resistivity are also
likely to allow the ratio to be higher.

5.2 Resistance measurements

The voltage - current measurements of the pump for 3 different buffer ionic
strengths are shown in Fig. 5.2. Each of the three plots show both linear
and non-linear parts. The boundary between the linear and the non-linear
part depends on the ionic strength of the buffer. In the particular case of
figure 5.2, the voltage threshold between both parts will be 35 V for 0.25
M, 17 V for 0.5 M, and 12 V for 1 M ionic strengths. Note that Ohm’s law
is applicable to the linear part. However, above a certain current-voltage
threshold, a non-linear behavior is observed which is probably caused by a
decreasing electrical resistance due to Joule heating.

The measured electrical resistances of the pump filled with buffer of differ-
ent ionic strengths are compared in Table 5.1. The two last columns com-
pare the calculated total electrical resistance of the pump and the measure-
ments. The resistance for an applied voltage around 10 V was measured,
and this was still in the linear part of the current-voltage plot of the pump.
The ratio of calculated vs. measured electrical resistances of the system is
about the same for all three ionic strengths and equals 1.4. Our electrical
model of the pump gives us an estimation of the total electrical resistance
that is 50% less than what was measured. This is most likely due to imper-
fections of the resultant side-channels geometry after the microfabrication
steps.

Above the ohmic behavior threshold, which is different for each ionic
strength, the measured resistance diminishes with increasing voltage. This
5.2. RESISTANCE MEASUREMENTS 87

Table 5.1: Comparison of the electrical resistance measured for 3 different

buffer ionic strengths with the one calculated from our model (Rcalc , Equ.
5.7). The pumped solution was a citrate-phosphate buffer of pH = 5, the
applied voltage is around 10 V.

U I Rmes Rcalc Rmes

[V] [mA] [kΩ] [kΩ] Rcalc
0.25 M 10 0.2 52.9 35.9 1.5
0.5 M 10.3 1.3 7.9 6.4 1.2
1M 10.9 2.5 4.4 3.2 1.4

Figure 5.2: Measured current as function of citrate-phosphate buffers ionic

strength and applied voltage for the new pump. The straight lines are linear
fits. (pH = 5, KCl was added for increased ionic strength to 0.5 and 1 M)
88 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

phenomenon happens in our pump for voltages higher than 12 V under all
experimental conditions investigated in this work, and is related to Joule
heating.

5.3 Flow observations

We used carboxylate coated latex beads in a citrate - phosphate buffer solu-

tion of pH = 5 (see section 4.2.2) to observe the flow in the main channel.
The pump showed two distinct behaviors when voltage only was applied
across the main channel, depending on the ionic strength of the buffer. At
low ionic strength (i.e. 0.25 M or less) the beads were seen to be stacked
against the side-channel walls. This bead migration against the side chan-
nel wall was noticed when only DC voltage was applied. The electroki-
netic flow across the side channels seemed to be high enough to perturb
our pump, as no bead flow due to MHD was observed in the main channel
when the additional magnetic field was applied. At high ionic strength (i.e.
0.5 M or more) no flow perpendicular to the main channel direction was
noticed when only a DC voltage was applied. When B was added, MHD
pumping of beads was observed along the main channel.

Low ionic strength

We performed only few measurements with low ionic strength citrate-phosphate

buffers (Istr ≤ 0.25 M). Under those conditions, MHD observation experi-
ments were not possible. In order to analyze this problem, we measured the
current for a given voltage (Fig. 5.2), as well as conductivity of the buffer.
As discussed in the preceding section, this perturbation is believed to be
caused by ek flows.

The expected velocity for MHD and corresponding perpendicular elec-

troosmotic flow velocity were calculated and compared for a buffer of pH
= 5 under the influence of an applied voltage of 10 V (I = 0.19 mA) and a
5.3. FLOW OBSERVATIONS 89

magnetic field of 0.4 T. The current density in the main channel will be:

I
J= = 167 A · m−1 (5.15)
hl
MHD velocity in a half-circular channel:

3 h2 l
vMHD = J · B = 13.4 µm · s−1 (5.16)
64 µ L
where L (22 mm) is the total length of the main channel, and µ (1.1·10−3
[Pa·s]) is the absolute viscosity of the buffer. Electric field in the main
channel:
%
Echa = · I = 204 V · m−1 (5.17)
hl
The glass channel walls in a 0.25 M buffer environment have an estimated
zeta potential of 20 mV [1–3]. The ratio of MHD to eo velocity is therefore:

vMHD 3 h2 l B 1
= =5 (5.18)
veo 64 L % ε ζ

In a 0.25 M buffer, the eof flow velocity will be 5 times less than MHD.
We see that the expected electrokinetic flow velocity is very close to the
one from MHD. Those calculations are summarized in Table 5.2. The eo
flow under such conditions is believed to be so high that it prevented the
observation of MHD flow with our bead technique.

High ionic strength

At high ionic strength we observed MHD flow in the main channel, and we
did not notice any perturbation from the side channels. In order to com-
plete our understanding of the electrokinetic phenomena, we compared the
expected MHD flow velocities with the eo velocity for every ionic strength
of our buffer in Table 5.2. Under high ionic strength conditions, we noticed
that MHD flow is at least 28 times faster than eo flow.
90 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

Table 5.2: Comparison of calculated eof and MHD flow velocities in the
main channel for an applied voltage of 10 V and 0.42 T magnet strength.
The pumped solution was a citrate-phosphate buffer of pH = 5.
% I Jmain vMHD Emain veo f vMHD
(Ω·m) (mA) (A·m−2 ) (µm·s−1 ) (V·m−1 ) (µm·s−1 ) veo f
0.25 M 1.2 0.2 167 13.4 204 2.6 5
0.5 M 0.2 1.3 1083 87.3 238 3.0 28
1M 0.1 2.5 2083 167.8 229 2.9 55

Influence of transverse electrokinetic flows

In the preceding sections we noticed that the flow originating from eo was
on the same order of magnitude as MHD. At low ionic strength, under
the influence of an electric current, the beads were pulled against the side-
channel wall. At high ionic strength, this behavior was not observed when
only current was applied. Under MHD flow measurements (at high ionic
strength) a proportion of flowing beads stopped their course on the side
channel walls. The beads, which are negatively charged (carboxy groups),
were always seen to be attracted to the side-channels wall on the same side
as the anode, the positive electrode. By labeling the buffer flowing in the
main channel with amaranth (a red dye, see section 6.1.2), we also observed
that the color of the transparent buffer in the side reservoirs (anode side)
began to turn red at the entrance of the side channels within a few minutes.
Since amaranth has a negative charge it also moved towards the anode.
Electroosmotic flow is a bulk flow, and is directed towards the cathode in
glass channels. This observed movement of beads and amaranth towards
the anode is due to the ep mobility of the beads (or dye) being much larger
than eo flow across the side channels.

We believe that our MHD observations were always the superimposition of

electrokinetic flows (eo+ ep) together with MHD (see Fig. 5.3). At low
ionic strength (when MHD ≈ eo + ep) the addition of the three vectors lead
to a velocity vector following the direction of the side channels. At high
5.3. FLOW OBSERVATIONS 91

Figure 5.3: Vectors representing eo, ep and MHD movement and their sum
vector for the cases that: A) MHD  (ep + eo) B) MHD ≈ (ep + eo).

ionic strength MHD velocity will be high enough for the vector sum to be
in the pumping direction. As an example, if we compare our calculations
of velocities at low ionic strength (0.25 M), we see that eo flow is similar
in magnitude to MHD (only 5 times lower than MHD). No specific exper-
iments were conducted to determine what is the concentration threshold at
which MHD pumping is less perturbed by ek flows. But in order to get rid
of eo completely, one could think of treating the surface of the channels
with an uncharged molecule (a sugar molecule for example) as uncharged
molecules should drop the zeta potential at the channel walls (in best cases,
down to zero).

The sample ions present in the main channel will disappear in the reser-
voirs via the connecting side channels, and be replaced by ions from the
opposite side reservoir. Note that if this mixing behavior between the sam-
ple ions in the main channel and the ions in the side reservoirs is unwanted,
one could think of placing the MHD pump downstream of the experiment.
92 CHAPTER 5. ELECTRICAL DESCRIPTION OF THE PUMP

5.4 Summary
We described the electric working principles of our pump with a simple
model that is similar to electrical resistance measurements. This model en-
abled us to also calculate the expected electroosmotic flow induced by an
electric field in the main channel. Eo velocity was compared with the ex-
pected MHD velocity to explore the limitations of our pump in terms of
the inevitable electrokinetic flows that disturbed MHD flow. Treating the
surface of the channel with uncharged molecules should drop the electroos-
motic flow, and placing the MHD pump downstream of the microfluidic
system should prevent any unwanted mixing of ions from the side reser-
voirs with sample ions in the channel.
Bibliography

[1] C. Schwer and E. Kenndler. Electrophoresis in fused-silica capillaries:

The influence of organic solvents on the electroosmotic velocity and
the ZETA potential. Analytical Chemistry, 63:1801 – 1807, 1991. exp.
det. of pK fused silica.

[2] Y. Gu and D. Li. The Zeta-potential of glass surface in contact with

aqueous solutions. Journal of Colloid and Interface Science, 226:328–
339, 2000.

[3] B.J. Kirby and E.F. Hasselbrink Jr. Zeta potential of microfluidic sub-
strates: 1. theory, experimental techniques, and effects on separations.
Electrophoresis, 25:187–202, 2004.

93
94 BIBLIOGRAPHY
Chapter 6

MHD pumping in NMR

environments

In this chapter, we present our DC magnetohydrodynamic (MHD) pump as

a component of a nuclear magnetic resonance (NMR) microfluidic chip.
This is the first time that MHD pumping in an NMR environment has
been observed and demonstrated. The design of the chip was changed
from straight to U-shaped channel (with same total length) to reduce the
overall dimensions of the pump, and thus to facilitate its integration in the
NMR measurement system. This chip generates a maximum flow rate of
1.5 µL·min-1 (2.8 mm·s−1 in the microchannel) for an applied voltage of
19 V (1600 A·m−2 with 1M buffer) in a 7 T NMR superconductive mag-
net. We developed a simple method to measure flow rate inside the bulky
NMR magnet by monitoring the displacement of a liquid-liquid interface of
two immiscible liquids in an off-chip capillary. Flow rate measurements of
beads and liquid-liquid interface as a function of both current and magnetic
field intensity of permanent magnets were compared, and showed equiv-
alent results. This allowed us to characterize and compare the flow rate
generated by the micropump using a permanent magnet (B1 = 0.33 T) with
the NMR superconductive magnet (B0 = 7.05 T). We observed a 21-fold
increase in flow rate corresponding to the ratio of the magnetic field in-
tensities (B0 /B1 = 21) in accordance with the theoretical flow dependence

95
96 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

on the magnetic field intensity. The final aim is to integrate MHD pumps
together with planar NMR coils in a microfluidic system for NMR analy-
sis [1]. The work presented in this section includes the results obtained by
Mr. Frieder Lucklum [2] during his internship under our supervision.

6.1 Experimental methods

The design and working principle of this DC MHD pump, as well as the
standard working buffers, has been described in Chapter 4 and in [3].

6.1.1 Micropump fabrication

In order to reduce the dimensions of the pump, we introduced small modi-
fications to the process described in section 4.2.1. We processed two com-
plementary sets of Pyrex wafers in our cleanroom facility. The first set of
wafers includes the microchannel, and the second the holes for fluidic ac-
cess to the side-reservoirs. The design and fabrication process for the two
sets of wafers are shown in Fig. 6.1. Before step 1, we dipped both sets1
in 50% HF for 25 min. The result was through holes: 700 µm diameter
holes in the first set for bottom access to the microchannel, and rectangu-
lar shaped holes in the second set for fluidic access from the top to the
side-reservoirs. The first set of wafers was then processed as previously
described in section 4.2.1.

The main channel (16 mm × 150 µm × 75 µm) is now U-shaped to make

the overall dimensions of the pump smaller. A total number of 785 side-
channels were etched across the main channel. The side-channels widths
and spacing were 10 µm, the depth 100 nm, and their length varied from
10 to 100 µm. The side-reservoirs were 75 µm deep and their detailed di-
mensions are shown in Fig. 6.2. We initially planned a final step with evap-
oration of platinum electrodes on the bottom of the side reservoirs. This
step was temporarily halted due to two major technology problem related
1 with structured 400 nm polysilicon and 1.8 µm of AZ1518 photoresist as an etch mask
6.1. EXPERIMENTAL METHODS 97

Figure 6.1: Microfabrication steps for the micropump. 1. Start with Pyrex
wafers coated with 400 nm CVD polysilicon. 2. Photolithography followed
by RIE etching to define the channel and the side-reservoirs location. 3.
Deep 20% HF etch: defines the length of the side-channels. Polysilicon
removal with KOH. 4. Photolithography with spray photoresist followed
by BHF etch of the shallow side-channels; alignment of the wafer with a
Pyrex coverplate. 5. Fusion bonding of both Pyrex wafers at 650˚C.
98 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

to electrical contact2 and fusion bonding efficiency.

For pump encapsulation and fluidic connection of the holder with outer
capillaries, we chose non-ferromagnetic material 3 . In other words, the ma-
terial was chosen so that there was no interaction with the strong magnetic
field of the NMR superconductive magnet. We chose PMMA material for
the encapsulation of the pump. Figure 6.3 shows a picture of the holder.
We present a three dimensional view and detailed dimensions of the holder
in appendix B. We employed small O-rings for the fluidic sealing at the
bottom chip-holder interface. We assembled the top and bottom part of
the holder with screws made in copper. We integrated the platinum wire
electrodes on the top part of the holder. A thin sheet of latex4 provided a
proper sealing between the chip and top part of the holder. The top part also
served as a reservoir for ionic current generation. We connected the holder
with outside capillaries with 062 Minstac fittings (The Lee Company, West-
brook, CT, USA).

6.1.2 Experimental setup

We used the same citrate phosphate buffers as those in chapter 4. The
buffers composition were 53 mM citrate (pH = 6), with 207 mM phosphate
for 0.25 M buffer, and with 244 mM or 744 mM KCl in addition to reach
ionic strengths of 0.5 M or 1 M respectively. For each buffer concentration,
we measured the pH and conductivity with electrodes from ThermoOrion
company (Beverly, MA, USA). We saturated the buffer with amaranth, a
red-colored salt, to bring visual contrast for flow observations. Table 6.1
shows the measured properties of the buffer solutions. Polystyrene beads
(carboxylate-modified microspheres, 2.55% in water, from Polysciences
Europe GMBH, Germany) with a diameter of 6 µm were diluted 500 times
2 The metal lines were disrupted at the edge of the channel, where the channel side-wall

is nearly vertical.
3 See appendix C for a discussion about magnetic classification of materials.
4 The sheet of latex was manufactured by pouring first a monolayer of liquid latex

(XUR R liquid latex, Michael Schermons, Berlin, Germany) on a flat surface, letting it
dry, and then cutting a piece having the shape of the parts to seal together.
6.1. EXPERIMENTAL METHODS 99

Figure 6.2: Three dimensional drawing of the MHD pump (Drawing cour-
tesy of F. Lucklum). The dimensions of the chip and side reservoirs were:
a = 4 mm, b = 6 mm, c = 2.45 mm, d = 2.4 mm, e = 8.4 mm, f = 12.65 mm,
g = 8.95 mm, h = 1 mm. We measured the dimensions on a processed chip,
with a precision of ± 0.05 mm.
100 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

Figure 6.3: (a) View of the chip on top of the O-rings on holder. (b) View
of the chip with integrated platinum electrodes on the bottom of the side-
reservoirs (c) Close-up view of a channel with flowing beads inside. (d)
View of the chip with the latex sealing before closing the holder. (e) View
of the encapsulated chip filled with amaranth-colored buffer. The anisole
bubble is the transparent plug in between buffer seen on the outer capillary
in front of the holder.
6.1. EXPERIMENTAL METHODS 101

Table 6.1: Electric and chemical properties of the buffers used [4]

Conductivity Resistivity pH
[mS·cm-1 ] [Ω·m]
0.25 M Buffer 21.4 0.47 6.3
0.5 M Buffer 44.9 0.22 6.0
1 M Buffer 88.3 0.11 5.8
1 M Buffer + Amaranth 87.8 0.11 5.9

in the buffer for flow visualization. A constant current (0 - 5 mA) was ap-
plied across the two Pt electrodes with a 2400 source meter from Keithley
Instruments S.A. (Zurich, Switzerland), the compliance of the source was
set to 100 V.

We utilized NdFeB rare earth magnets (Forcefield, CO, USA) of varied

magnetic field strength for the pumping experiments on top of permanent
magnets. We chose magnet diameters of 50 mm (diameter larger than the
chip dimensions, see Fig. 6.4) and designed the chip holder as thin as possi-
ble to ensure good homogeneity of the magnetic field in the microchannel.
The magnetic field intensities were measured with a 3- axis magnetic-field
transducer (model 3MR-1C, Sentron AG, Switzerland). For those measure-
ments, we positioned the Hall sensor at the exact height corresponding to
the microchannel position. Therefore, we measured two different magnetic
field intensities of B1 = 0.33 and B2 = 0.38 T. We performed the experi-
ments within the NMR superconductive magnet in collaboration with Klaus
Ehrmann of the LMIS3 laboratory in EPFL. This magnet was part of a 300
mHz NMR spectrometer setup (DRX 300, Bruker Biospin AG, Switzer-
land). The intensity of the static magnetic field generated inside the bulky
NMR superconductive magnet was calculated to be 7.05 T. The working
space inside this magnet was limited: it has a cylinder volume of 40 mm
diameter, and experiments had to be conducted at a specific height position
102 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

Table 6.2: Physical properties of anisole and aqueous solutions

Viscosity Relative Solubility in Surface

[mPa·s-1 ] density water tension
[nN·m-1 ]
Anisole 1.06 0.99 poor 35.1
Water 0.89 1 high 71.99
1 M Buffer 1.09 1 high ≤ 71.99

where the field is uniform5 . We also designed a custom holder for the ex-
periments inside the NMR magnet to allow the MHD pump to be positioned
in the homogeneous field region, and to contact the Pt electrodes with the
current source outside of the NMR instrument (see Fig. 6.5).

6.1.3 Flow measurement methods

We conducted the flow measurements in a closed loop fashion by connect-
ing the inlet and the outlet of the PMMA holder with an off-chip PVC
capillary (inner diameter of 1 mm). We measured the flow velocity with
the same bead technique as described in section 3.1.3.

In the bulky B0 , we developed an indirect measurement method to observe

MHD flow rate, as optical characterization was not possible in such an ap-
paratus. We measured the lateral displacement of an Anisole-bubble in-
terface in between the amaranth-colored buffer. We chose anisole for its
immiscible properties with water, and because it has the same viscosity and
density as water (see Table 6.2). The surface tension of anisole6 , however,
differed significantly from aqueous solutions. We measured the anisole
plug displacement in a Teflon capillary (5 cm long, inner diameter of 0.81
mm) connected in between the outer PVC capillaries. The flow rate in the
5 This distance was approx. h = 270 mm from the bottom entrance of the NMR cylinder.
6 Anisole wets the Teflon capillary wall better than water.
6.1. EXPERIMENTAL METHODS 103

Figure 6.4: View of the chip holder for measurement on top of a 40 mm

diameter NdFeB permanent magnet (with 1.5 cm long Teflon flow mea-
surement capillary).
104 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

Figure 6.5: View of the chip holder for measurement inside the bulky super-
conductive magnet (with 1.5 cm long Teflon flow measurement capillary).
A picture of the NMR superconductive magnet is seen on the right. The
holder is inserted into the NMR magnet from the bottom, as the white ar-
row indicates.
6.2. RESULTS AND DISCUSSION 105

whole microfluidic system (Q) is directly proportional to the lateral dis-

placement (dh) of the plug. For a given measurement time dt we will have:
dh 2
Q= · πrCAP (6.1)
dt
where rCAP is the radius of the teflon capillary used for the flow rate mea-
surement. The continuity of the system implies that the flow rate is the same
in the measuring capillary as it is in the microchannel (only the velocities
differ), we used equation 2.15 (for half circular channels7 ) for comparison
with theory. We estimated the error on the displacement measurement to be
half of the smallest graduation of our scale meter; the error was thus 0.25
mm.

6.2 Results and discussion

Fig. 6.4 shows a close-up top view of the encapsulated chip. The top
reservoirs are filled with the buffer. When both side-reservoirs and main
channel of the chip are filled, the ionic current can be generated between
the electrodes.

6.2.1 Electrical characterization

We first characterized the current-voltage behavior of the chip as a function
of the conductivity of the solutions and the geometry of the side-channels.

Ionic strength
The upper panel of Fig. 6.6 shows the current-voltage behavior for a mi-
cropump with 20 µm long side-channels, filled with buffers of selected con-
ductivities. We see that the higher the concentration, the higher the current
for a fixed voltage. The current increases linearly with the voltage until
a certain point where the linearity is lost. Above this threshold, the tem-
perature of the solution begins to rise: Joule heating begins. The voltage
7 Q = 3π l · h4 · J · B
128µ L
106 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

threshold is different for each solution, it is higher for the less concentrated
buffers.

Side-channel length

When a given voltage is applied across the side-channels, the electrical re-
sistance of the pump will depend on the dimensions of the side-channels.
More specifically, the longer the side-channels, the higher the electrical re-
sistance and therefore the less the current intensity in the main channel. We
designed the new pump with 4 different side-channel lengths, namely 10,
20, 50 and 100 µm. The aim of the side-channels is to produce a DC ionic
current as high as possible in the main channel for a given DC voltage with-
out any bubble formation or Joule heating in the main channel. The bottom
panel of Fig. 6.6 shows the current-voltage behavior of micropumps with
3 different side-channel lengths for a buffer solution of 1 M ionic strength.
We see that, for the same buffer conductivity, the shorter the side-channel
length, the higher the current for a given voltage. The voltage threshold
of the linear range is also different for each geometry. The voltage thresh-
old is higher for the longest side-channels geometries. The slope of the
current-voltage curve in the linear range corresponds to the measurement
of the electrical resistance of the pump. As expected from Equation 5.6 (ex-
pressed from our electrical model presented in Chapter 5), the resistance is
proportional to side-channel length. The 10 µm long side-channels should
then show the best electrical behavior. Although the chips with the best
bonding reproducibility were the 100 µm long channels, we performed the
experiments with MHD pumps of 20 and 50 µm side-channels geometry.
The current generated by the 100 µm long side-channels was too small to
achieve any noticeable MHD pumping with the permanent magnets. We
also could not test MHD pumps with 10 µm long side-channels as none of
them did bond properly after the fusion bonding microfabrication step8 .

8 The available surface was too small to allow sufficient adhesion for proper bonding
6.2. RESULTS AND DISCUSSION 107

Figure 6.6: Characterization of the new chip with current-voltage curves.

The deviation on the voltage measurement corresponds to 1% of each data
value. The data linear fit is represented by the dotted lines. [Top] Chip
behavior as a function of the ionic strength of the buffer (chip with 20 µm
long side-channels). [Bottom] Chip behavior as a function of the length of
the side-channels (chip filled with 1 M buffer).
108 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

Table 6.3: Comparison of the Joule heating thresholds as a function: (1) Of

side-channel lengths with 1 M buffer solution in the pumps; (2) Of buffer
ionic strengths in a pump with 20 µm long side-channels .

(1) U Jmax Qdiss

[V] [A·m-2 ] [W·m−1 ]
20 µm 19 1600 2.1
50 µm 23 1600 2.6
100 µm 45.5 818 2.7

(2) U Jmax Qdiss

[V] [A·m-2 ] [W·m−1 ]
0.25 M 47 1000 3.4
0.5 M 28 1200 2.4
1M 19 1600 2.1

Joule heating

In Fig. 6.6 the Joule heating threshold has been determined for all three
side-channel lengths (20, 50, and 100 µm) with a 1 M buffer, as well as
for all three buffer concentrations (0.25, 0.5, and 1 M) in a pump with
20 µm long side-channels. Table 6.3 summarizes those values. The Joule
heating threshold occurs when the voltage - current behavior of the pump
ceases to be linear. At this point, the system begins to heat up, and the
increase in temperature changes the physical behavior of the pumped fluid
(see 2.3.1 for details). For each of these thresholds, we also calculated the
corresponding power dissipation (Qdiss ) using equation 4.2. Corresponding
Qdiss vary from 2.1 to 3.4 W·m−1 . The applied voltage should therefore not
be higher than 19 V for MHD pumping of a 1 M solution with the 20 µm
long side-channel geometry, and 23 V for 1 M solution with 50 µm long
side-channels.
6.2. RESULTS AND DISCUSSION 109

Bubble generation

Another limiting factor that may affect our system in terms of duration of a
single experiment is the rate of bubble generation around the Pt electrodes.
We observed that after a certain time (30-60 min), the bubbles around the
electrodes in the outer reservoir (on top of the holder) do not escape in the
air, but stay in the space between the electrode, the PMMA reservoir and the
buffer solution. This insulating gaseous spacing between the electrode and
the buffer gradually decreases current generation between the electrodes
(for a fixed voltage). This is so far the only time-limiting factor encountered
with our pumping system, and it is easily avoided by decreasing the operat-
ing current density. This time limitation prevented us to perform flow rate
measurements at high current density (> 500 A·m-2 ) for durations longer
than 30 min.

6.2.2 Characterization of the new flow rate measurement method

Inside the superconductive magnet, the flow rate was characterized by track-
ing the displacement of a liquid-liquid interface within a teflon capillary
connected in a closed-loop fashion with the pump. Various interfaces were
tested, the most reproducible results could be achieved by introducing a
"bubble" of Anisole in the buffer solution. Anisole is a derivative of ben-
zene, with similar viscosity and density as water. We conducted the flow
rate measurements as follows. First the anisole plug was introduced in
between the colored buffer in the teflon external capillary. A state of equi-
librium was achieved when no movement of the plug was noticed. This
equilibrium was reached after a mean waiting time of 30 min. Then the
chip was fixed on top of the NMR chip holder (see Fig. 6.5) and inserted
inside the NMR magnet. The measurement time began and ended when
the current source was switched on and off, respectively. The plug posi-
tion was recorded immediately after the current was switched off and the
holder removed from the NMR magnet. As seen in the top panel of Fig.
6.7, the plug’s displacement is a linear function of the measurement time,
as expected by the theory.
110 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

Figure 6.7: Evolution of the plug position. [Top] As a function of the total
measurement time (20 µm long side-channels, 1M Buffer, J = 450 A·m-2 ,
U = 6 V, 5 cm long Teflon capillary). [Bottom] As a function of the waiting
time after the end of measurement (50 µm long side-channels, 1M Buffer,
J = 610 A·m-2 , U = 10 V, 1.5 cm long Teflon capillary).
6.2. RESULTS AND DISCUSSION 111

Backflow

The time gap between the current switch-off and the plug position mea-
surement was less than 5 seconds. This remark is important because we
noticed a movement of the plug in the direction opposite to MHD pump-
ing after each measurement. This reverse movement usually didn’t exceed
10% of the total measurement9 , but in the case of a short measurement cap-
illary (1.5 cm long, as depicted in Fig. 6.5) it could exceed 50%. This
high reverse flow was measured for the 1.5 cm long capillary as a func-
tion of the elapsed time after measurement, and its evolution is seen in the
bottom panel of Fig. 6.7. We see that the backflow is very fast within the
first 5 minutes, and that it levels off after a waiting time of 15 min. This
displacement in the reverse direction was also recorded 15 min after each
measurement.

Equivalence with bead measurement technique

In order to validate the new flow measurement technique, we compared the

ratios Qx /Bx at a given current density. Theory says that Qx /Bx should
be constant when J is constant. For J = 650 A·m-2 , we had the following
observations:

QB1 QB2
= 8.48 · 10−2 µL · min−1 · T−1 , = 8.42 · 10−2 µL · min−1 · T−1
B1 B2
(6.2)
QB0
and = 8.58 · 10−2 µL · min−1 T−1 (6.3)
B0
where QB1 and QB2 were flow rates measured with the bead velocity tech-
nique and permanent magnets B1 and B2 , and QB0 was the flow rate mea-
sured with the liquid-liquid interface technique in the NMR magnet of mag-
netic field intensity B0 . We see that all three ratios gave the same value
(within 2 %), showing that both measurement techniques are comparable.

9 for long Teflon measurement capillaries of typically 5 cm long

112 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

6.2.3 Measurements with permanent magnets

In small B environments, i.e. when the MHD pump was characterized with
permanent magnets, the flow rate was measured by means of bead veloc-
ity and anisole-bubble displacement. We did our measurements with MHD
pumps having 20 µm long side-channels, and 1 M buffer as the pumped
solution. The maximum flow rates, corresponding to an applied current
density of 650 A·m-2 , were respectively 0.028 and 0.032 µL·min-1 for mag-
netic field intensities of respectively B1 = 0.33 and B2 0.38 T. The bead
velocity as a function of the applied current density for two different per-
manent magnets can be seen in Fig. 6.8. We see that the measured flow is
proportional to the applied current and that it follows theory.

Figure 6.8: MHD Flow mean velocity as a function of current density (for 2
different magnetic field intensities, 20 µm long side-channels, and without
anisole). Error bars indicate the standard deviation for 10 beads in a total
sample of 50 beads per measurement. The straight line is the theoretical
mean velocity, calculated using equation 2.16.
6.2. RESULTS AND DISCUSSION 113

This measurement was only possible if the fluidic system was closed by
connecting the inlet and outlet with outside capillaries in a closed-loop
fashion. In fact, the flow rate generated by MHD in small B environments
seemed to be too weak to dispense any movement in an open system. This
closed-loop fashion enabled us to get rid of the unnecessary backpressure
due to of hydrostatic and capillary forces. To understand the influence of
hydrostatic pressure on the performances of the pump, we calculate the
maximum body force generated by our pump for a current density of 650
A·m-2 and a permanent magnetic field of 0.38 T we will have (thanks to
equation 2.14):

∆P = J · B · l = 650 × 0.38 × 16 · 10-3 = 4 Pa (6.4)

This expression for the body force generated by our pump does not account
for the influence of gravity. The hydrostatic pressure drop generated by
gravity is proportional to the height difference (∆h) between two points so
that:
∆P = ρ · g · ∆h (6.5)
where ρ is the fluid density, and g (≈ 9.8 m·s−2 ) is the gravity constant. If
we calculate the threshold for ∆h at which the hydrostatic pressure drop of
our buffer equals the maximum MHD body force we obtain:

∆P = 4 = ρ · g · ∆h (6.6)
4
∆h = = 4.1 · 10-4 m = 0.4 mm (6.7)
103 · 9.8
This means that if we try to measure the flow rate of our micropump by
leaving the inlet and outlet open with a height difference higher than 0.4
mm (which is easily achievable) we will be unable to observe any move-
ment induced by the MHD phenomenon only. The closed loop configu-
ration of the system enabled us to cancel out the flow perturbation due to
hydrostatic pressure.

Note that the bead flow measurements were not performed simultane-
ously with anisole bubble displacement. At low current densities (less than
114 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

500 A·m-2 ), the measured flow rates with the anisole interface technique
were at least half those calculated from theory, and the less the current, the
more the gap between theoretical flow rate and measurements. In fact the
anisole bubble seemed to introduce a resistance to the flow for low cur-
rent densities. This resistance was not observed when the beads only were
present. Moreover, the equivalence of those two flow measurement meth-
ods with permanent magnets was only observed for high current densities
(higher than 500 A·m-2 ). Therefore we state that the bead measurement
technique is a suitable technique for low flow rates measurements, whereas
the anisole bubble displacement is best for high flow rates observations.

6.2.4 Measurements in NMR superconductive magnet

Preliminary results of MHD pumping in NMR environments with our pre-

vious MHD pump (described in chapter 4) were conducted by Frieder Luck-
lum. The MHD pumping was successful, and results are described in detail
in his internship report [2].

In the NMR environment we characterized the flow with two different side-
channel lengths for the MHD pump geometry (namely 20 and 50 µm long)
and 1 M buffer as the pumped solution. We measured the displacement of
the anisole plug as a function of the applied current. In any case, we com-
pared the measurements with theory. Every pump gave equivalent results,
namely very good aggreement with theory. An example of the evolution
of the flow rate in the NMR magnet as a function of current density for a
pump with 20 µm long side-channels is shown in Fig. 6.9. We see that the
measured flow rate depends on the applied J and increases with it in a lin-
ear fashion, as for flow with permanent magnet actuation. For low values
of current, the measured flow rate is lower than our calculations, but corre-
sponds again for J ≥ 450 A·m-2 . This phenomenon at low J is believed to
be related to capillary pressure issues. The results presented in expressions
6.2 and 6.3 can also be used to show that the volumetric flow rate scales
6.2. RESULTS AND DISCUSSION 115

Figure 6.9: MHD Flow rate as a function of current density (for 3 different
magnetic field intensities, and 20 µm long side-channels). Error bars in-
dicate the systematic error of the scale-measurement method. The straight
line is the theoretical flow rate, calculated using equation 2.15.
116 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS

with the magnetic field:

QB0 0.605 B0 7.05

= = 21.6 and = = 21.4 (6.8)
QB1 0.028 B1 0.33

QB0 0.605 B0 7.05

= = 18.9 and = = 18.5 (6.9)
QB2 0.032 B2 0.38

The change from the permanent magnets (B1 , B2 ) to the NMR magnet (B0 )
increased the field strength by a factor of 18.5 and 21.4 respectively. The
MHD measurements (at a fixed current density of 650 A·m-2 ) showed an
increase in the flow rate by a factor of 18.9 and 21.6 respectively. As the
flow rates at a fixed current density for the same chip geometry are directly
proportional to the magnetic field strength (see expression 2.15), the equiv-
alence of the ratios proved the flow rate dependance on B.

The maximum body force generated by our pump for a current density of
1600 A·m-2 and a NMR magnetic field of 7.05 T is:

∆P = J · B · l = 1600 × 7.05 × 16 · 10-3 = 180.5 Pa (6.10)

which corresponds to a height difference (see equ. 6.5) of:

180.5
∆h = = 1.8 · 10-2 m = 18 mm (6.11)
103 · 9.8
So far, the only MHD pump that generated a MHD body force of more than
100 Pa was fabricated by Eijkel et al. [5]. Combining an AC magnetic field
of 0.1 T, an AC current density of 21100 A·m-2 , on a total electrode length
of 63 mm, they generated a maximum body force of 133 Pa.

With regard of these results, we see that MHD pumping in an NMR en-
vironment has been successfully achieved. The pumping flow rate is rather
low, with a maximum of 1.5 µL·min-1 for an applied current of 1600 A·m-2
(corresponding voltage of 19 V for 20 µm long side-channels). This low
flow rate may be problematic if one wants to monitor a chemical reaction
6.3. SUMMARY 117

with fast kinetics. In this case, the pump should be integrated together with
the NMR detection coils on one single chip so that the distance between
the pumping/mixing part and the NMR analysis microcoils is minimized.
It was not possible in our case to integrate the pump and the NMR coils on
one single chip because of geometrical issues: the NMR microfluidic chip
has to stand in a plane parallel to B0 , whereas the MHD chip has to stand
perpendicular to B0 .

6.3 Summary
We developed a simple method of flow rate measurement inside the bulky
NMR magnet, and we successfully achieved MHD pumping inside an NMR
environment in accordance with MHD theory. The flow characterization
method with immiscible fluids is suitable for high flow rates observation
(namely for J ≥ 500 A·m-2 ), whereas the bead velocity method is best for
lower flow rate characterization on top of permanent magnets. The pump-
ing flow rate is rather low, with a maximum of 1.5 µL·min-1 (high flow
velocity of 2.8 mm·s−1 in the microchannel) for an applied current of 1600
A·m-2 in the 7 T NMR superconductive magnet. Future work with these
devices include the integration of this MHD pump into a µTAS for NMR
analysis. The envisioned applications are online monitoring of a chemical
reaction with our pump also used as the mixer, and non-destructive MHD
flow characterization by NMR microcoils.
118 CHAPTER 6. MHD PUMPING IN NMR ENVIRONMENTS
Bibliography

[1] C. Massin, F. Vincent, A. Homsy, K. Ehrmann, G. Boero, P.-A. Besse,

A. Daridon, E. Verpoorte, N.F. de Rooij, and R.S. Popovic. Planar
microcoil-based microfluidic NMR probes. Journal of Magnetic Reso-
nance, 164:242–255, 2003.

[2] F. Lucklum. MHD pump integration into NMR environment. Internship

report, University of Neuchatel, 2005.

[3] A. Homsy, S. Koster, J.C.T. Eijkel, A. van den Berg, F. Lucklum,

E. Verpoorte, and N.F. de Rooij. A high current density DC magneto-
hydrodynamic (MHD) micropump. Lab On a Chip, 5:466–471, 2005.

[4] D.R. Lide. C.R.C. Handbook of Chemistry and Physics. 76th edition,
1995.

[5] J.C.T. Eijkel, C. Dalton, C.J. Hayden, J.P.H. Burt, and A. Manz. A cir-
cular ac magnetohydrodynamic micropump for chromatographic appli-
cations. Sensors and Actuators B, 92:215–221, 2003.

119
120 BIBLIOGRAPHY
Chapter 7

Integration of Nuclear
Magnetic Resonance detection
with microfluidics and MHD
pumping

This chapter describes the integration of NMR detection with microfluidics

and MHD pumping. We first present the technological challenges associ-
ated with integrating NMR coils with a glass microfabricated microfluidic
chip. The second section describes preliminary experiments where we com-
bined MHD pumping with NMR detection. The two last sections describe
feasibility studies for applications involving the integration of NMR detec-
tion together with MHD pumping. The NMR phenomenon, basic theory
and applications have been introduced in section 1.5.

7.1 Fabrication of microchannels compatible with NMR

The largest challenge associated with NMR methodology is the sensitivity
of detection. To maximize sensitivity, it is important that the microflu-
idic structures are designed in such a way that optimum homogeneity and

121
122 CHAPTER 7. INTEGRATION OF NMR DETECTION

Figure 7.1: Cross-sectional eSEM view of channels of symmetric cross-

section

strength of the magnetic field is achieved in the detection volume. Both

theoretical considerations and simulations performed at the EPFL for this
project indicate that the magnetic field homogeneity should be better main-
tained in microchannels and cuvettes with symmetrical cross-sections. We
used Pyrex-glass wafers as a substrate for the microfluidics devices. Pyrex
channels were etched isotropically in liquid HF to form a rotated-D-shaped
cross-section. Then, two Pyrex wafers with complementary etched chan-
nels were aligned and bonded to obtain the final symmetric cross-section
(see Fig. 7.1). Next, the NMR probes were microfabricated on top of
the device by our partners at EPFL - LMIS3 [1–3]. Figure 7.2 shows a
schematic view of the fabrication of the chip as well as pictures of the NMR
probe integrated on top of the microchannels.

7.2 Combining MHD pumping with NMR detection

In order to combine MHD pumping with NMR detection, the position
of the channels with respect to the static magnetic field B0 of the NMR
spectroscopy apparatus is of great importance: best pumping efficiency is
achieved when B0 is perpendicular to both the pumping channel and the
current density across the channel. In the NMR spectrometer, B0 is verti-
cal, which means that our chip must be positioned horizontally to generate
7.2. COMBINING MHD PUMPING WITH NMR DETECTION 123

Figure 7.2: Planar NMR probe integrated on top of microchannels. (1)

Simplified fabrication process of planar microcoils on a glass substrate with
integrated channels. (a) Access holes and half channels are etched in the
bottom substrate. (b) Half channels are etched in the top substrate, which is
then fusion bonded to the bottom substrate. The top wafer is thinned down
to a total thickness of 140 µm by polishing (75 µm chamber and channels,
65 µm membrane). (c) The planar coils are microfabricated using a pat-
terned SU-8 layer as a mold for Cu electroplating. (2) Photo of a 5-turn,
500 µm inner diameter, electroplated coil on a glass substrate with inte-
grated microfluidics. The sample chamber underneath the coil has a 500
µm diameter and 150 µm depth, resulting in an approximate sample vol-
ume of 30 nanoliters. (3) Monolithic NMR chip with integrated coils and
microfluidic channels. Sample is injected trough connection tubes attached
on the backside of the chip. The planar coil, which is on the opposite side
of the chip, is electrically contacted using wire bonding. (All sketches and
comments courtesy of C. Massin and F. Vincent, EPFL)
124 CHAPTER 7. INTEGRATION OF NMR DETECTION

MHD pumping in the channel. However, the NMR chips have an integrated
planar coil on top of microchannels, which generates an alternating field B1
perpendicular to the plane of the chip. Therefore, if an NMR experiment
is held with such a chip, the coil must generate a B1 perpendicular to B0
(see section 1.5). This means that the NMR chip has to be held vertically.
This major geometry issue made it challenging to integrate both NMR and
MHD functionalities in one single planar chip. One way to integrate NMR
and MHD into one chip would be to fabricate a planar NMR coil that gen-
erates a B1 in the same plane as the chip. This technology is possible for
planar NMR coils [4] but was not readily available, and such developments
are beyond the scope of this project.

Connecting two microfluidic chips

Since MHD integration with NMR in one single chip was not an attractive
option in our lab, we decided to connect one vertical NMR microfluidic
chip to our horizontal MHD pump in a closed loop fashion with the help of
external PVC capillaries. We connected our pump to one of two existing
NMR microfluidic chips. The first NMR chip was fabricated from Pyrex;
the corresponding sample volume under the NMR coil being 540 nL. The
second NMR chip was made of SU-8, and the sample volume under the
NMR coil was 26 nL. We chose the SU-8 chip as a smaller volume was
more appropriate to maximize the pumping velocity under the coil.

We considered two experimental concepts to characterize MHD pump-

ing by the NMR chip. The first concept involved the introduction of the
buffer/anisole/buffer interface in the NMR channel. The NMR coils moni-
tor the entrance of one liquid into the NMR detection zone and the removal
of the other. If several NMR acquisitions are performed during the liq-
uid exchange, the spectra should show peaks of both liquids with changing
signal strength. The second concept involved NMR monitoring of buffer
alone1 as a function of the applied MHD flow rate. The NMR signal will
1 in the SU-8 channel
7.2. COMBINING MHD PUMPING WITH NMR DETECTION 125

Figure 7.3: View of the chip holder for combined MHD pumping with
NMR measurement inside the bulky superconductive magnet. The two mi-
crofluidic chips were connected together with outside PVC capillaries in a
closed loop fashion. The NMR chip was fabricated from SU-8 channels
and contained integrated NMR coils. The microchannel was 22 mm long,
360 µm wide and 65 µm deep. The coils integrated on top (and on the bot-
tom) of the SU-8 channel covered 1100 µm of the length of the channel,
corresponding to a sample analysis volume of 1100 × 360 × 65 µm3 (26
nL) under the coil. This NMR chip was developed, fabricated and handled
by K. Ehrmann.
126 CHAPTER 7. INTEGRATION OF NMR DETECTION

Figure 7.4: NMR spectrums acquired by K. Ehrmann with the SU-8 NMR
chip connected to our MHD pump. The analyzed substances under the
NMR coil were respectively: a) buffer alone, b) buffer + anisole, and c)
anisole alone.
7.2. COMBINING MHD PUMPING WITH NMR DETECTION 127

be at a maximum if the molecules excited do not leave the volume under the
coil between excitation and detection. The NMR signal will be at a min-
imum if the molecules are fast enough to travel the whole volume under
the coil between excitation and detection, leaving no NMR signal to be de-
tected. A change in the peak intensity of water molecules will be observed
as a function of the applied MHD flow rate if the flow velocity in the NMR
channel varies between those two boundary conditions.

We chose the first concept because the expected maximum flow velocity
under the coil was too slow to lead to any decrease in NMR spectrum inten-
sity. All experiments were done in collaboration with K. Ehrmann (LMIS3
- EPFL). Fig. 7.3 shows a picture of the fluidic connection. The NMR
microfluidic chip was developed by K. Ehrmann [5], and the geometry of
the SU-8 channels was rectangular. Before the pumping experiments, we
measured NMR reference spectrums of both chemicals in the SU-8 chan-
nels. The spectrums of anisole and buffer didn’t overlap at all (see Fig.
7.4). Therefore, our simple technique should show if a movement of the
buffer/anisole/buffer is taking place in the SU-8 microchannels. A body
force of 180.5 Pa was imposed by the pump on the microfluidic system (J
= 1600 A·m−2 ). The flow rate Q in the microfluidic system is related to the
pressure drop ∆P by Poiseuille law:

∆P = Q(Rhy MHD + Rhy NMR ) (7.1)

where Rhy MHD is the hydraulic resistance of the half-circular MHD pump-
ing channel (see Equ. 2.12) and Rhy NMR is the hydraulic resistance of the
rectangular SU-8 channels (see Equ. 2.11). The flow rate generated by
the pump in this system will be 1.3 times less than what is generated by
the pump alone. The flow rate in the system2 is then 1.12 µL·min−1 , cor-
responding to an expected flow velocity in the SU-8 channel around 0.8
mm·s−1 . A liquid at the entrance of the SU-8 channel moving with a ve-
locity of 0.8 mm·s−1 should cross the whole length in a little more than 27
seconds.
2 The flow rate corresponding to 1600 A·m−2 for the pump alone is 1.5 µL·min−1 .
128 CHAPTER 7. INTEGRATION OF NMR DETECTION

We could not measure any change of NMR spectrum between anisole and
buffer when MHD pumping was actuated together with NMR detection.
We discovered that the problem stemmed from the entrance of the SU-8
microchannel. After few fruitless attempts to analyze the problem, we ob-
served that the SU-8 wall collapsed at the location of the fluidic connection.
Since it was the only SU-8 chip available, we could not continue our MHD
flow NMR analysis at this point of time. In fact, this SU-8 NMR chip was
first designed to perform static NMR analysis. We believe that MHD flow
measurements with interconnected NMR chips is possible, but the pump
should be connected either to a new batch of low volume Pyrex NMR chan-
nels or with improved fluidic connection with the SU-8 channels.

Summary

MHD pumping combined with NMR detection has not been validated yet,
but should be feasible by interconnecting two microchips3 . In order to per-
form this measurement, a new cycle of chip designs is envisioned. The fast
(but not so useful) solution would be to improve the fluidic connection be-
tween the chips by choosing tougher materials like Pyrex for the NMR chip
or by adapting a gentle sealing to the inlet of the SU-8 channel. The second
solution would be to develop a planar chip with integrated MHD pump and
NMR coils. The timescale of this last solution would be longer than the
first one, but the envisioned applications are more promising, as developed
in the next section.

3 The hydraulic resistance of the connected NMR chip must be equal of smaller than the

one of the pump.

7.3. PERSPECTIVES ON MHD-NMR APPLICATIONS 129

7.3 Perspectives on MHD-NMR applications

7.3.1 Chemical reaction actuated by MHD and monitored by

NMR
One possible application of MHD-NMR devices is to use the MHD pump
integrated in an NMR microfluidic chip4 to enable dynamic monitoring of a
chemical reaction [6], to monitor the interaction between two molecules [7,
8], or to observe conformation changes of macromolecules [9]. The MHD
pump would play the role of the chemical reaction initiator, by initiating the
mixing of the two chemical species. Two mixing methods are envisioned.

One method would place the MHD pump at the end of the microfluidic net-
work (just before the outlet), two inlets channels leading to a single channel
would allow both chemicals to enter the fluidic network, and to mix by dif-
fusion. This kind of mixing method has already proven to be efficient, and
enabled mixing times from 30 min down to 2 s [6], depending on the flow
rate (defined by an off-chip syringe pump). In their experiments, it was
important to set the intensity of the flow rate not too high (from 0.1 to 4
µL·min-1 ), to make sure the excited volume of the sample wasn’t removed
from the detection region before NMR signal acquisition was completed.
With an MHD pump integrated in the microfluidic network, the flow rate
will be controlled on-chip, and will correspond to the range of flow rate of
interest (the maximum flow rate achieved by our last design of MHD pump
geometry was 1.5 µL·min-1 ).

Another method would use our specific MHD pump geometry as the mixer
for a chemical reaction, or as a method to change the solvent composition
from a solution with suspended macromolecules (like proteins [9]). The
way to use the pump as a mixer is to fill the pumping channel with one
species and to fill the side reservoirs with the other. The mixing efficiency
will depend on the ionic strength of the species (especially in the side reser-

4 The NMR excitation will be achieved by an integrated planar coil that generates a

magnetic field in the same plane as the chip, like in ref [4].
130 CHAPTER 7. INTEGRATION OF NMR DETECTION

voirs). The mixing behavior of our MHD pump hasn’t been studied in de-
tails5 , but it could be another application for our specific MHD design.

7.3.2 On-chip MHD flow monitoring by MRI

The ability to measure a flow by magnetic resonance imaging (MRI) was
studied by Caprihan et al. [10], they demonstrated that it was possible to
obtain a velocity profile of the material under flow. Very recently, Golo-
shevsky et al. [11] observed a two-dimensional flow profile by MRI in a
capillary using two planar biaxial gradient coils incorporated with a planar
Helmholtz RF NMR coil [12]. The envisioned application was to measure
the viscosity of the fluid with the help of a non-invasive and non-destructive
mode of operation. The measured variable is the velocity of the molecules
in the capillary. Goloshevsky et al. measured velocities from 0.45 to 2.57
mm·s−1 (with 0.05 mm·s−1 resolution) for applied flow rates from 47 to
236 µL·min−1 in 1.4 mm diameter capillaries. Attention was paid to the
flow rate (generated by very precise syringe pumps) so that the flow wasn’t
too fast for the excited spins to leave the RF coil before being detected.

With such a measurement method it would be possible to observe and char-

acterize the profile of MHD flows. MHD pumps with their high perfor-
mances at relatively low flow rates could be a candidate of choice of flow
actuation technique for NMR or MRI analysis6 .

7.4 Summary
This chapter was aimed to describe several aspects of the integration of mi-
crofluidics and MHD pumping together with NMR detection. We saw that
substrates like Pyrex glass with HF etched cylindrical microchannels lead to
5 The transient flow was indeed minimized in our developments by using nanometer-
deep side-channels.
6 The flow is slow enough for the excited spins not to leave the volume under the RF coil

before being detected.

7.4. SUMMARY 131

the ideal geometry for the integration of planar NMR coils. Preliminary ex-
periments with interconnected glass MHD pump and SU-8 microchannel7
were interrupted by fluidic connection perturbations due to the fragility of
the SU-8 channel. The description of possible applications involving the
combination of MHD pumping and NMR detection indicate the need in
NMR techniques for low flow rates (in the range of 0.1 to 240 µL·min−1 )
so that the flow wasn’t too fast for the excited spins to leave the RF coil
before being detected. The high performances of MHD pumps at relatively
low flow rates are therefore seen as an asset for NMR and MRI applica-
tions. Both examples of possible use of MHD in NMR devices are feasible
if both systems are integrated in one planar microfabricated chip.

7 with integrated NMR microcoils on top

132 CHAPTER 7. INTEGRATION OF NMR DETECTION
Bibliography

[1] C. Massin, A. Daridon, F. Vincent, G. Boero, P.-A. Besse, E. Ver-

poorte, N.F. de Rooij, and R.S. Popovic. A microfabricated probe
with integrated coils and channels for on-chip NMR spectroscopy. In
Micro-TAS Conference, editor, Micro Total Analysis Systems 2001,
Monterey - California, 2001.

[2] A. Homsy, J. Lichtenberg, C. Massin, F. Vincent, P.-A. Besse, R.S.

Popovic, N.F. de Rooij, and E. Verpoorte. Fabrication of microfluidic
channels with symmetric cross-sections for integrated NMR analysis.
In Micro-TAS Conference, editor, Micro Total Analysis Systems 2002,
Nara - Japan, 2002.

[3] C. Massin, F. Vincent, A. Homsy, K. Ehrmann, G. Boero, P.-A. Besse,

A. Daridon, E. Verpoorte, N.F. de Rooij, and R.S. Popovic. Planar
microcoil-based microfluidic NMR probes. Journal of Magnetic Res-
onance, 164:242–255, 2003.

[4] K.R. Minard, G.R. Holtom, L.E. Kathmann, P.D. Majors, B.D. Thrall,
and R.A. Wind. Simultaneous 1H PFG-NMR and confocal mi-
croscopy of monolayer cell cultures: Effects of apoptosis and necrosis
on water diffusion and compartmentalization. Magnetic Resonance in
Medicine, 52:495–505, 2004.

[5] K. Ehrmann, N. Saillen, F. Vincent, C. Massin, P.-A. Besse, and R.S.

Popovic. Microfabrication of Helmholtz coils with integrated chan-
nels for NMR spectroscopy. In Roland Zengerle, editor, 19th IEEE

133
134 BIBLIOGRAPHY

International Conference on Micro Electro Mechanical Systems, Is-

tambul, Turkey, 2006.

[6] H. Wensink, F. Benito-Lopez, D.C. Hermes, W. Verboom, H.J.G.E.

Gardeniers, D.N. Reinhoudt, and A. van den Berg. Measuring reac-
tion kinetics in a lab-on-a-chip by microcoil NMR. Lab On a chip,
5:280–284, 2005.

[7] A. Chen and M.J. Shapiro. Affinity NMR. A new drug-screening

tool that probes ligand-receptor interactions. Analytical Chemistry,
71(19):669 A, 1999.

[8] M. Pursch, L.C. Sander, and K. Albert. Understanding reversed-phase

LC with solid-state NMR. Analytical Chemistry, 71(21):773 A, 1999.

[9] M. Kakuta, D.A. Jayawickrama, A.M. Wolters, A. Manz, and J.V.

Sweedler. Micromixer-based time-resolved NMR: Applications to
ubiquitin protein conformation. Analytical Chemistry, 75:956–960,
2003.

[10] A. Caprihan and E. Fukushima. Flow measurements by NMR.

Physics Reports, 198(4):195–235, 1990.

[11] A.G. Goloshevsky, J.H. Walton, M.V. Shutov, J.S. de Ropp, S.D.
Collins, and M.J. McCarthy. Integration of biaxial planar gradient
coils and an RF microcoil for NMR flow imaging. Measurement Sci-
ence and Technology, 16:505–512, 2005.

[12] J.H. Walton, J.S. de Ropp, M.V. Shutovand A.G. Goloshevsky, M.J.
McCarthy, R.L. Smith, and S.D. Collins. A micromachined double-
tuned NMR microprobe. Analytical Chemistry, 75:5030–5036, 2003.
Chapter 8

Summary and Outlook

We have demonstrated a novel concept for a DC MHD micropump, allow-

ing the use of DC current without gas bubble perturbation in the pumping
channel. We used a micromachined frit-like structure to provide electrical
contact between the main channel and side reservoirs where the electrodes
were located. Such a chip design allowed MHD pumping with a flow rate
up to 0.3 µl·min−1 (mean velocity of 0.5 mm·s−1 ) for an applied current
density of 4000 A·m-2 in 75-µm-deep glass microchannels on top of a 0.42
T permanent magnet. We then observed and demonstrated MHD pump-
ing in an NMR environment for the first time using a more compact MHD
pump design. The pumping flow rate in the 7 T NMR superconductive
magnet had a maximum of 1.5 µL·min-1 (mean velocity of 2.8 mm·s−1 in
the microchannel) for an applied current density of 1600 A·m-2 . Table 8.1
shows a comparison of experimental conditions and main properties of our
MHD pump with the state of art of DC and AC MHD micropumps. In this
table, we highlight the parameters affecting the performance of such pump-
ing systems. The pumps that had the best performance in term of flow rate
and velocity are the ones where J, B and depth of channel (h) are max-
imized. In practice, they carried out those performances by maximizing
the channel cross-sectional area (A). On the other hand, the best perfor-
mances in terms of generated body force were the pumps where J, B and
electrode length (l) were maximized. The micropumps that fulfilled this

135
136 CHAPTER 8. SUMMARY AND OUTLOOK

sort of performance had the electrode cross-sectional area (AJ ) minimized.

The body force and flow velocity generated by our pump in NMR environ-
ment has surpassed all previously published MHD pumps this far. In our
system, we achieved such performances by minimizing AJ and maximizing
J. Preliminary experiments with interconnected glass MHD pump and SU-
8 microchannel1 were interrupted by fluidic connection perturbations with
the SU-8 channel. The description of envisioned applications involving the
combination of MHD pumping and NMR detection indicate the need in
NMR techniques for flow rates in the range of 0.1 to 240 µL·min−1 . The
high performances of MHD pumps at relatively low flow rates are therefore
seen as an asset for NMR and MRI applications. Both examples of possible
use of MHD in NMR devices are feasible if both systems are integrated in
one monolithic microfluidic system.

Outlook

Future research about MHD pumping with integrated NMR detection should
first focus on the integration within one monolithic chip of both function-
alities. An NMR coil generating the analyzing magnetic field in the same
plane as the MHD pump [7] is necessary for proper function of the chip.
The performance in terms of generated MHD flow rate should also be raised
by a minimum factor of 10 (and maximum 150) for the pump to be indis-
pensable for NMR analysis applications.

The parasite eof flow originating from the side-channels (in the direction
perpendicular to MHD pumping) was not studied in detail. A meticulous
eof experimental study as a function of the pumped buffer’s ionic strength
and surface chemistry of the microfluidic system could help to better un-
derstand eof side effects on the micropump performances. Changing the
surface chemistry of the glass microchip with a hydrophillic uncharged
molecule (like sugars) should help the user reducing eof, as uncharged sur-
faces should drop the zeta potential at the channel walls (in best cases, down
1 with integrated NMR microcoils on top
 U I A AJ l J B ∆PMHD vMHD QMHD
(V) (mA) (mm2 ) (mm2 ) (mm) (A·m−2 ) (T) (Pa) (mm·s−1 ) (µL·min−1 )

Jang et al. [1] 30 DC 1.8 0.4 30 30 60 0.44 1 2.6∗ 63∗

Leventis et al. [2] >1.3 DC 35 18 225 75 155 1.35 16 0.4 450·103
Bau et al. [3] 4 DC 15 1.9 292 172 51 0.4 3.5 0.4 45
Lemoff et al. [4] 6.6 AC 140 0.2 1.5 4 92105 0.013 5 1.5 18
West et al. [5] 5 AC 90 0.2 5 28 17684 0.011 5.5 0.24 3
Eijkel et al. [6] 4 AC 40 6·10−3 2 63 21100 0.1 133 0.04 14·10−3

Chapter 4 16 DC 4.8 8.8·10−3 1.2 16 4000 0.42 27 0.5 0.3

Chapter 6 19 DC 2 8.8·10−3 1.2 16 1600 7.05 180 2.8 1.5

Table 8.1: Performance comparison of previously published MHD pumps with our MHD pump presented
in Ch 4 and 6. All values for voltage (U), current (I), channel cross-sectional area (A), total length of electrodes along the
pumping channel (l), MHD flow mean velocity in the pumping channel (vMHD ) and MHD flow rate (QMHD ) were experimental
data, and were taken from references [1–6]. Most of the values for the electrode cross-sectional area (AJ ) and current density
(J) across the pumping channel had to be calculated. The body force (∆PMHD ) generated by the pumps was calculated thanks to
relation 2.14. ∗ Both values were taken from experimental measurements. If calculated with relations 2.16 and 2.15, the predicted
velocity and flow rate would be 0.16 mm·s−1 and 4 µL·min−1 respectively.
137
138 CHAPTER 8. SUMMARY AND OUTLOOK

to zero). On the other hand, this eof side-effect from the side channels could
be exploited and amplified to transform the pump into a micro-mixer.
Bibliography

[1] J. Jang and S.S. Lee. Theoretical and experimental study of MHD
magnetohydrodynamic micropump. Sensors and Actuators A, 80:84–
89, 2000.

[2] N. Leventis and X. Gao. Magnetohydrodynamic electrochemistry in

the field of Nd-Fe-B magnets. Theory, experiment, and application in
self-powered flow delivery systems. Analytical Chemistry, 73:3981–
3992, 2001.

[3] H.H. Bau, J. Zhong, and M. Yi. A minute magnetohydrodynamic

(MHD) mixer. Sensors and Actuators B, 79:207–215, 2001.

[4] A.V. Lemoff and A.P. Lee. An AC magnetohydrodynamic micropump.

Sensors and Actuators B, 63:178–185, 2000.

[5] J. West, B. Karamata, B. Lillis, J.P. Gleeson, J. Alderman, J.K. Collins,

W. Lane, A. Mathewson, and H. Berney. Application of magnetohy-
drodynamic actuation to continuous flow chemistry. Lab On a Chip,
2:224–230, 2002.

[6] J.C.T. Eijkel, C. Dalton, C.J. Hayden, J.P.H. Burt, and A. Manz. A cir-
cular ac magnetohydrodynamic micropump for chromatographic appli-
cations. Sensors and Actuators B, 92:215–221, 2003.

[7] K.R. Minard, G.R. Holtom, L.E. Kathmann, P.D. Majors, B.D. Thrall,
and R.A. Wind. Simultaneous 1H PFG-NMR and confocal microscopy
of monolayer cell cultures: Effects of apoptosis and necrosis on water

139
140 BIBLIOGRAPHY

diffusion and compartmentalization. Magnetic Resonance in Medicine,

52:495–505, 2004.
Appendix A

Glossary

Latin Symbols Description SI Units

a Valence of ion
A Channel cross-sectional area [m2 ]
AJ Electrode cross-sectional area [m2 ]
B Magnetic field [T]
C Concentration [kg·m−3 ]
Cdl Double layer capacitance [C·V−1 ]
D Diffusion coefficient [m2 ·s−1 ]
Dh Hydraulic diameter [m]
∆E Transition energy [J]
E Electric field [V·m−1 ]
E0 Redox standard potential [V]
Ep Electrode potential [V]
F Faraday constant [C·mol−1 ]
FL Lorentz force [N]
G Magnetic field gradient [T·m−1 ]
h Planck’s constant [J·s]
h Height of channel [m]

141
142 APPENDIX A. GLOSSARY

Ha Hartmann number
I Electric current [A]
In Nuclear spin
J Current density [A·m−2 ]
k Boltzmann’s constant [J·K−1 ]
L Length of channel [m]
L Angular momentum [J·s−1 ]
l Length of electrodes [m]
m Nuclear moment [A·m2 ]
m Mass [kg]
M Macroscopic magnetization [A·m−1 ]
N Number of moles [mol]
P Pressure [Pa]
q Charge of a particle [C]
Q Volumetric flow rate [m3 ·s−1 ]
Qd Power dissipation [W·m−1 ]
r Radius [m]
R Electrical resistance [Ω]
Re Reynolds number
Ret Transitional Reynolds number
Rm Magnetic Reynolds number
Rhy Hydraulic resistance [N·s·m−5 ]
t Time [s]
T Absolute temperature [K]
U Applied voltage [V]
v Velocity [m·s−1 ]
V Volume [m3 ]
w Width of channel [m]
 143

Greek Symbols Description SI Units

δ Chemical shift (in ppm)

δ Thickness [m]
ε Dielectric constant of the medium [C·V−1 ·m−1 ]
η Kinematic viscosity [m2 ·s−1 ]
γ Gyromagnetic ratio [rad·T−1 ·s−1 ]
λ Magnetic diffusivity [m2 ·s−1 ]
µ Absolute viscosity [Pa·s]
µm Magnetic permittivity [V·s·A−1 ·m−1 ]
ν0 Frequency [Hz]
ω Angular frequency [s−1 ]
ρ Density [kg·m−3 ]
ρe Charge density [C·m−3 ]
% Resistivity [Ω·m]
σ Electrical conductivity [S·m-1 ]
θ Angle [rad]
ζ Zeta potential [V]
144 APPENDIX A. GLOSSARY

Abbreviation Description

1-D One dimensional

AC Alternating current
CVD Chemical vapor deposition
DC Direct current
DI Deionized
ek Electrokinetic
ep Electrophoretic
eo Electro-osmotic
eof Electro-osmotic flow
EPFL Ecole polytechnique fédérale de Lausanne
FID Free induction decay
IMM Institute of microelectronics and microsystems
IMT Institut de microtechnique
LMIS3 Laboratoire de microsystèmes n˚3
MHD Magnetohydrodynamic
MRI Magnetic resonance imaging
NMR Nuclear magnetic resonance
PDMS poly(dimethylsiloxane)
SAMLAB Sensors, actuators, and microsystems laboratory
SU-8 High aspect ratio Epoxy based photoresist
UNINE Université de Neuchâtel
µTAS Micro total analysis system
Appendix B

MHD pump details

Figure B.1: Three dimensional drawing of the MHD pump (Drawing cour-
tesy of F. Lucklum). The dimensions of the side reservoirs openings were:
a = 2.35 mm, b = 5.2 mm, c = 1.3 mm, d = 7.2 mm, e = 8.4 mm, f = 6.35
mm, g = 8.95 mm, h = 1 mm, i = 12.65 mm. We measured the dimensions
on a processed chip, with a precision of ± 0.05 mm.

145
146 APPENDIX B. MHD PUMP DETAILS

Figure B.2: Three dimensional drawing of the holder.

 147

Figure B.3: Technical drawing of the top part of holder with detailed di-
mensions and views.
148 APPENDIX B. MHD PUMP DETAILS

Figure B.4: Technical drawing of the bottom part of holder with detailed
dimensions and views.
Appendix C

Classification of Magnetic
Materials

All materials can be classified in terms of their magnetic behavior falling

into one of five categories depending on their bulk magnetic susceptibility.
The two most common types of magnetism are diamagnetism and param-
agnetism, which account for the magnetic properties of most of the periodic
table of elements at room temperature (see Fig. C.1). These elements are
usually referred to as non-magnetic, whereas those which are referred to as
magnetic are actually classified as ferromagnetic. The only other type of
magnetism observed in pure elements at room temperature is antiferromag-
netism. Finally, magnetic materials can also be classified as ferrimagnetic
although this is not observed in any pure element but can only be found
in compounds, such as the mixed oxides, known as ferrites, from which
ferrimagnetism derives its name.

Diamagnetism

In a diamagnetic material the atoms have no net magnetic moment when

there is no applied field. Under the influence of an applied field (B) the
spinning electrons precess and this motion, produces a magnetisation (M)
in the opposite direction to that of the applied field. All materials have a

149
150 APPENDIX C. CLASSIFICATION OF MAGNETIC MATERIALS

diamagnetic effect, however, it is often the case that the diamagnetic effect
is masked by the larger paramagnetic or ferromagnetic term. The value of
susceptibility is independent of temperature. Most elements in the periodic
table are diamagnetic materials, including copper, silver, gold, glass, or
water.

Paramagnetism

Paramagnetic metals have a small and positive susceptibility to magnetic

fields. These materials are slightly attracted by a magnetic field and the
material does not retain the magnetic properties when the external field is
removed. Paramagnetic properties are due to the presence of some un-
paired electrons and from the realignment of the electron orbits caused by
the external magnetic field. Paramagnetic materials include platinum, mag-
nesium, lithium, and tantalum.

Ferromagnetism

Ferromagnetic materials have a large and positive susceptibility to an exter-

nal magnetic field. They exhibit a strong attraction to magnetic fields and
are able to retain their magnetic properties after the external field has been
removed. Ferromagnetic materials have some unpaired electrons so their
atoms have a net magnetic moment. They get their strong magnetic proper-
ties due to the presence of magnetic domains. In these domains, large num-
bers of atoms moments (1012 to 1015) are aligned parallel so that the mag-
netic force within the domain is strong. When a ferromagnetic material is
in the demagnetized state, the domains are nearly randomly organized and
the net magnetic field for the part as a whole is zero. When a magnetizing
force is applied, the domains become aligned to produce a strong magnetic
field within the part. Iron, nickel, and cobalt are examples of ferromag-
netic materials. Components with these materials are commonly inspected
using the magnetic particle method. Iron, for example, is a ferromagnetic
material.
 151

Antiferromagnetism
In the periodic table the only element exhibiting antiferromagnetism at
room temperature is chromium. Antiferromagnetic materials are very sim-
ilar to ferromagnetic materials but the exchange interaction between neigh-
boring atoms leads to the anti-parallel alignment of the atomic magnetic
moments. Therefore, the magnetic field cancels out and the material ap-
pears to behave in the same way as a paramagnetic material. Like ferro-
magnetic materials these materials become paramagnetic above a transition
temperature, known as the Néel temperature, TN . (Cr: TN = 37o C).

Ferrimagnetism
Ferrimagnetism is only observed in compounds, which have more complex
crystal structures than pure elements. Within these materials the exchange
interactions lead to parallel alignment of atoms in some of the crystal sites
and anti-parallel alignment of others. The material breaks down into mag-
netic domains, just like a ferromagnetic material and the magnetic behavior
is also very similar, although ferrimagnetic materials usually have lower
saturation magnetization.
152 APPENDIX C. CLASSIFICATION OF MAGNETIC MATERIALS

Figure C.1: Periodic table showing the type of magnetic behavior of each
element at room temperature.
Appendix D

Citrate-Phosphate buffer
preparation

The table showed next page was extracted from a paper written by P.J.
Elving et al. in 1956. Elving et al. described the way to fabricate buffer so-
lutions of varying pH but constant ionic strength. The buffer compositions
were selected from the McIlvaine formulas, and the desired ionic strength
is maintained by the addition of potassium chloride.

153
154 APPENDIX D. CITRATE-PHOSPHATE BUFFER PREPARATION

Figure D.1: Table for the preparation of the constant ionic strength buffer.
This table was taken from P.J. Elving et al.
Acknowledgements

My greatest thanks go to Professor Nico de Rooij for giving me the op-

portunity to follow my PhD research in his group, and for those past four
wonderful years. Thank also to Sabeth Verpoorte who accepted me as one
of her PhD students within the µTAS team.

My special thanks go to Milena Koudelka-Hep and Peter van der Wal, who
were always available to answer my "stupid" questions about chemistry
and electrochemistry. Your support has been highly valuable to a humble
physicist.

I also gratefully acknowledge Jan Eijkel, of the lab on a chip group at the
university of Twente in The Netherlands, and Prof. Radivoje Popovic, of
the microsystems design group at EPFL, for kindly agreeing to be the co-
examiners of my PhD thesis.

This work wouldn’t have been possible without the valuable financial sup-
port provided by an EPFL Presidential Fund (Nr. 581.552) and by the Swiss
National Science Foundation (Project Nr. 2100-61549.00).

Great thanks to Sander Koster and Vincent Linder for their support in the
management of my project. And to a great extend to all of my fellow µTAS-
colleagues Jan Lichtenberg, Frieder Lucklum, Anpan Han, Anna-Maria
Spehar, Tobias Kraus, Gian-Luca Lettieri, Arash Dodge, Laura Ceriotti,
Roos Guijt, and Nathan Lacher.

155
156 ACKNOWLEDGEMENTS

Thanks to Charles Massin, Franck Vincent, Klaus Ehrmann, Pierre-Andre

Besse, and Giovanni Boero for their support in the NMR and magnetic part
of the project.

The microfabrication of the devices presented in this thesis would not have
been possible without the help of IMT’s very valuable technical team: Syl-
vain Jeannneret, Gianni Mondin, Edith Milotte, Nicole Hegelbach-Guye,
Giovanni Bergonzi, Pierre-André Clerc, Sabina Jenny, Sylviane Pochon,
and José Vaquera. Your valuable help and great mood was always greatly
appreciated. Thanks also to Claudio Novelli for his patience and great
skills towards all my computer worries... Thanks to Florence Rohrbach
and Karine Frossard for their secretarial work. Mireille Leboeuf and Mas-
soud Dadras have contributed to the beautiful electron microscope images
seen in this work.

I give my deepest thank my friend Laure Aeschimann, who introduced me

to SAMLAB for the first time. This journey wouldn’t have been the same
without your friendship and support all through the years here in Neucha-
tel, but also during our studies in EPFL Lausanne. Thanks also to Luca
Berdondini for our valuable and highly motivating scientific discussions in
the train on our way home to Lausanne. I thank also all of you in the SAM-
LAB group for the unforgettable friendly and efficient team atmosphere.

The daily travel from Lausanne to Neuchatel in the train wouldn’t have
been the same without all the people I met during this journey. Thanks to
all of you for your friendship!

A special thanks to my friend Sumita Pennathur, from the Mechanical En-

gineering Department at Stanford University, who patiently verified and
carefully corrected my english in every chapter of this thesis.

Thank you Remo for your love, your friendship, and unlimited patience
during this whole writing of thesis.
 157

And last but not least, I would like to give my endless thanks to my parents
and to my sister. Without your infinite love and support none of this would
have ever been possible. I love you! Thank you XXX
158 ACKNOWLEDGEMENTS
Publications

List of publications

C. Massin, F. Vincent, A. Homsy, K. Ehrmann, G. Boero, P.-A. Besse, A.

Daridon, E. Verpoorte, N.F. de Rooij, and R.S. Popovic. Planar microcoil-
based microfluidic NMR probes. Journal of Magnetic Resonance 164, 286-
293 (2003).

A. Homsy, S. Koster, J.C.T. Eijkel, A. van den Berg, F. Lucklum, E. Ver-

poorte, and N.F. de Rooij. A high current density DC magnetohydrody-
namic (MHD) micropump. Lab On a Chip 5, 466-471 (2005).

A. Homsy, V. Linder, F. Lucklum, and N.F. de Rooij. Magnetohydrody-

namic Pumping in Nuclear Magnetic Resonance Environments. Sensors
and Actuators B: Chemical (accepted September 2006).

Contribution to Conferences

Oral presentation

A. Homsy, S. Koster, J.C.T. Eijkel, A. van den Berg, E. Verpoorte, and N.F.
de Rooij, A high flow rate DC magnetohydrodynamic (MHD) micropump.
Presented at the Micro-TAS 2004 Conference, Malmo, Sweden, Proceed-

159
160 PUBLICATIONS

ings pp. 372-374, Sept. 26-30 (2004).

Posters

A. Homsy, N.F. de Rooij, E. Verpoorte, C. Massin, F. Vincent, G. Boero,

P.A. Besse, R.S. Popovic. Glass technology for microfluidic channels with
symmetric cross-Sections. 3rd CMI-ComLab revue, Ecole Polytechnique
Fédérale de Lausanne, Lausanne, Switzerland (2002).

A. Homsy, J. Lichtenberg, C. Massin, F. Vincent, P.-A. Besse, R.S. Popovic,

N.F. de Rooij, and E. Verpoorte. Fabrication of microfluidic channels with
symmetric cross-sections for integrated NMR analysis. Micro-TAS 2002
Conference, Nara, Japan, Proceedings pp. 115-117, Nov. 3-7 (2002).

A. Homsy, S. Koster, E. Verpoorte, N.F. de Rooij. Microfluidic system for

integrated NMR analysis and MHD pumping. GRC Physics and Chemistry
of Microfluidics conference, Big Sky, Montana, USA (2003).

A. Homsy, S. Koster, N.F. de Rooij. Glass technology for microfluidic chan-

nels with integrated platinum electrodes. 5th CMI-ComLab revue, Ecole
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland (2004).

A. Homsy, S. Koster, N.F. de Rooij. A high current density magnetohy-

drodynamic (MHD) micropump. 6th CMI-ComLab revue, Ecole Polytech-
nique Fédérale de Lausanne, Lausanne, Switzerland (2005).

A. Homsy, F. Lucklum, V. Linder, N.F. de Rooij. Magnetohydrodynamic

pumping in nuclear magnetic environments. GRC Physics and Chemistry
of Microfluidics conference, Oxford, GB (2005).

View publication stats