INTRODUCTION TO MICROFLUIDICS

Introduction to Microfluidics
Patrick Tabeling
École Supérieure de Physique et de Chimie Industrielles, (ESPCI) Paris,
Paris Sciences et Lettres (PSL) University.
 Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Patrick Tabeling 2023
The moral rights of the author have been asserted
First published in hardback in 2005
First published in paperback in 2010
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2023907882
ISBN 978–0–19–284530–6
DOI: 10.1093/oso/9780192845306.001.0001
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
To Isa
Preface

Microfluidics has progressed considerably over the last twenty years, and the time has
come to envisage a serious update of the first edition of Introduction to Microfluidics,
published in 2005. In fact, this second edition is more than an update. Compared
to the first one, it keeps the same structure, the same spirit, the same attempts to
explain things, on a physical basis, in depth and simply, whenever possible, but it is not
reducible to an update. The present edition results from complete rewriting of the first
one, nurtured by the considerable amount of information collected in the field over the
last two decades. So much information has been gathered in twenty years. So many
revisions of the visions of the field have been made. Things that looked impossible
in the 1990s gave rise to an important industry, ten years later. This is the case of
next generation sequencing (NGS). Things that looked revolutionary turned out to
be disappointing. The history of microfluidics is full of dreams that became real and
appealing evidence that became wrong. Let us return to the turn of the century. At that
time, the microfluidic market (i.e. without ink jet printing) was small, and scepticism
was floating around concerning the potential of the technology to find its feet in a
market, even though it was regularly announced, here and there, that microfluidics
would revolutionarize the twenty-first century. Common sense led to the theory, in fact
wrong, that driving flows through tiny channels, at an industrial scale, without leaks,
clogging, bubbles, or uncontrolled adsorption, was impossible. The opposite viewpoint
believing that it is straightforward to create a complex, functional, microfluidic device
was unrealistic. Still, successful microfluidic products emerged, while, in the meantime,
the technology penetrated into an increasing number of new domains. The market
steadily increased at a two-digit rate, reaching, today, seventeen billions dollars. At
the moment, hundreds of millions of devices are sold every year. For example, 1.2
million Illumina microfluidic flow cells, for gene sequencing, are shipped every single
year. In the meantime, fundamental phenomena, such as capillarity, wetting, slippage,
and nanofluidic transport have been better understood or, in a number of puzzling
cases, just understood. Over the years, the early vision of the domain, based on a
strict analogy with microelectronics, gradually shifted to a new paradigm, in which the
microfluidic tool box is no longer restricted to MOS-FET surrogates, but incorporates
a much broader palette of materials and mechanisms.
Since the first edition, twenty books on microfluidics have been published. Many are
good or very good. But some look more like galleries of cartoons exhibiting systems
supposed to work, with no data showing that they do. This is a style. I tried to avoid
this type of presentation. Other books, often engineering-oriented, describe basic phe-
nomena without much depth. In fact, caricaturing subtle mechanisms is a difficult
exercise. Sometimes, errors are made. Perhaps, we could content ourselves with this
viii Preface

level of description. After all, we can use a computer with no idea how a transistor
works. As P. Anderson stated, science is organized in hierarchical levels, built on top
of each other, the upper using concepts elaborated by the lower. Both operate inde-
pendently, elaborating their own laws and paradigms. Then, as just said, computing
researchers do not need to know what a transistor is, and there is no necessity for
fluid mechanicists to understand kinetic theory of gases. In the 1990s, it was thought
that soon, microfluidic researchers would no longer need fluid mechanics. Engineers
would design microfluidic machines, without noticing that fluids are running in them.
Perhaps, one day, microfluidics will reach this stage. But microfluidics has not reached
this level yet. And it is still valuable and probably important, to acquire a minimal
knowledge on the subtleties that the field relies upon, at least to know the limitations
that they induce. I attempted to provide this type of knowledge throughout this book.
Many calculations developed in the present book are elementary. Still, they allow us
to rationalize most of the physics involved in microfluidic flow and transport phenom-
ena. Physico–chemical processes, developing on surfaces or in bulks, so important in
microfluidics, are explained on an intuitive basis, without using the powerful machin-
ery of thermodynamics, whose exquisite concepts sometimes obscure the underpin-
ning physics. Emphasis is placed on modern applications, in which biology plays an
important part. Also, as in the first edition, a forty-page introduction to microfluidic
technologies is given. The book includes enough material to build up courses of various
formats, from a few to twenty hours or so.

Fig. 0.1: Ant playing microguitar (From M. Seiffert (2003)).

Acknowledgements

First, I would like to thank C. M. Ho, who hosted me in 2000, at the University of
California at Los Angeles (UCLA), at the time when the ’golden years’ of microfluidics
were just launching. Thanks to him, I discovered the field. On my return to Paris,
one year later, P. G. De Gennes offered me to create a microfluidic laboratory at
Ecole Supérieure de Chimie Industrielle (ESPCI), which I did with enthusiasm. Since
then, for twenty years, I have had the chance to meet extraordinary people, students,
postdocs, and colleagues, and to build strong friendships. Writing the second edition
is a sort of noetic journey through these exciting years.
Microfluidics is a broad field, and it is impossible to become, all of a sudden, a world
class expert on all the subjects it embraces. I was fortunate to receive help from many
colleagues, who kindly exchanged correspondance and, in many cases, corrected errors
and misconceptions. I acknowledge J. Ottino, who made numerous remarks on the
chaotic mixing section. In the strange atmosphere of the COVID confinement days, we
had long exchanges over the internet, evoking, sometimes, the years when we pertained
to the same community. This left the impression that writing a book is also a look
into the past, sometimes exposing the author to vertigo. I had the same feelings with
my friends Y. Pomeau, and J.P. Bouchaud, for their remarks concerning Brownian
motion and chaos, and K. Moffatt for his comments on Batchelor’s approach to fluid
mechanics. I had discussions with G. Whitesides and A. Manz concerning microfluidics,
its past and future. J. Eijkels and A. van den Berg expressed their enlightening vision
of nanofluidics which I refer to in the book. D.Lohse updated the poor knowledge I had
on nanobubbles, H. Stone made a detailed reading of my presentation of the Navier-
Stokes equations, F. Mugele gave precious remarks on electrowetting and J. Cossy on
surfactant chemistry. With my friend S. Quake, I had a long semantic dispute about
the word ‘vesicle’, not ended yet. Z. Z. Li, a former student, now professor at Beijing
Institute of Technology, performed experiments to obtain the useful (unpublished)
images shown in Chap 4, related to step emulsification technique. I also benefited
from great remarks from D. Quere, J. O. Fossum, T. Lecuit, V. Hessel, J. L Viovy,
J. Eggers, M. Bazant, R. Ismagilov, P. Doyle, H. Bruus, A. Strooke, F. Wyart, E.
Clement, A. Skjeltorp, E. Raphael, M. Tatoulian, Y. Tran, K. Jensen, S. Wereley, J.
Bibette, A. Griffiths and D. Weitz . I also acknowledge J. Mouly and C. Midelet, from
Yole Development, for their explanations of the microfluidic market. Deep thanks to C.
Rollard, for her description of the realm of spiders. I am also indebted to J.O. Fossum,
E. Torino, P.A. Netti, for great discussions and for their invitations to sabbatical
periods at Nordheim University and the Italian Institute of Technology, in Napoli. I
would also like to acknowledge A. Libchaber, from whom I learned a lot. Discussions
with my close colleagues, P. Nghe and J. Mc Graw were inspiring and exchanges in
my group provided a source of thinking and learning: thanks to P. Garneret, E. Coz,
E. Martin, U. Soysal, P. Nieckele, M. Russo, and I. Maimouni. I am also grateful to
M. Dhunnoo for her help and L. Dehove, for her appealing cover and the great figures
shown in this book.
Contents

1 Introduction 1
1.1 Astonishing microfluidic systems in nature 1
1.2 Exquisite microfluidic control in the human body 3
1.3 MEMS, the mother of microfluidics 6
1.4 The birth of microfluidics 9
1.5 The advent of soft technology 14
1.6 Diversification of the technology and broadening of the applica-
tions 16
1.7 MicroReaction Technology (MRT) 21
1.8 Nanofluidics 22
1.9 The microfluidic market 26
1.10 Future of microfluidics 30
1.11 Reviews and books 38
1.12 Organization of the book 39
References 40
2 Physics at the microscale 48
2.1 The scales of small things 48
2.2 Intermolecular Forces - basics 58
2.3 Nano-Micro and Millifluidics 71
2.4 The physics of miniaturization 73
2.5 Scaling laws in nature 77
2.6 Miniaturization of electrostatic systems 87
2.7 Miniaturization of electromagnetic systems 91
2.8 Miniaturization of mechanical systems - the vibrating microbeam 93
2.9 Miniaturization of thermal systems 95
2.10 Sampling and throughput 97
References 100
3 Hydrodynamics of microfluidics 1: channels 103
3.1 The flow equations and the boundary conditions 103
3.2 Slippage in gases 118
3.3 Slippage in liquids 121
3.4 Microfluidics at small Reynolds numbers 127
3.5 Resistances and capacitances in microfluidics 141
3.6 Inertial microfluidics and millifluidics 150
References 158
xii Contents

4 Hydrodynamics of microfluidics 2: droplets 162

4.1 Liquid–vapour interfaces 162
4.2 Laplace’s law 169
4.3 Surfactants 181
4.4 Wetting 188
4.5 Droplets advancing on a surface 198
4.6 The governing equations and the capillary number 202
4.7 The Landau–Levich and Bretherton films 205
4.8 The Rayleigh–Plateau instability 208
4.9 Washburn law and paper microfluidics 212
4.10 Production of microfluidic droplets and bubbles 215
4.11 Characteristics of microfluidic droplets and bubbles 225
References 240
5 Transport in microfluidics 245
5.1 The microscopic origin of diffusion 245
5.2 Advection-diffusion equation and its properties 253
5.3 Analysis of diffusion phenomena 257
5.4 Analysis of dispersion phenomena 263
5.5 Brief introduction to chaos and chaotic mixing 269
5.6 Mixing in microfluidic devices 276
5.7 Four applications of transport of matter in microfluidics 286
5.8 Transport of matter across interfaces 290
5.9 Particles and microfluidics 299
5.10 Particles in inertial regimes 307
5.11 Adsorption 308
5.12 Chromatography 312
5.13 Thermal transport by conduction 318
5.14 Convection-diffusion heat equation and properties 324
5.15 Heat transfer in the presence of a flow in microsystems 329
5.16 Evaporation and drying 336
5.17 Microexchangers for electronic components 340
References 344
6 Electrokinetics 350
6.1 Introduction 350
6.2 Basic notions of electrostatics of macroscopic media 350
6.3 The electrokinetic equations 356
6.4 The electrical double layer 359
6.5 Electro-osmosis 372
6.6 Electrophoresis 381
6.7 Microfluidic electrokinetic separation 393
6.8 Dielectrophoresis 400
6.9 Three illustrations/applications of dielectrophoresis 404
6.10 Electrowetting 406
 Contents xiii

References 412
7 An introduction to microfabrication 415
7.1 Introduction 415
7.2 Current situation of microtechnologies 415
7.3 The environment of microfabrication 417
7.4 Photolithography 419
7.5 Direct writing or maskless photolithography 424
7.6 Microfabrication methods for silicon and glass devices 425
7.7 PDMS-based moulding – soft lithography 437
7.8 Computer Numerical Control (CNC) Micromilling 449
7.9 3D Printing or Additive Manufacturing (AM) 449
7.10 Paper microfluidics 451
7.11 Other technologies 453
References 459
Index 462
1
Introduction

In the 1970s, it became possible to miniaturize electromechanical systems, down to

the micrometric scale. This gave rise to a new field, called Micro-ElectroMechanical
Systems (MEMS). Later, in the 1990s, the field expanded, creating all sorts of mi-
crodevices, in which fluids, driven under control, gave rise to new functionalities. This
prompted the birth of a new field – microfluidics – the central subject of this book.
Microfluidics can be defined as the science of manipulation of fluids in systems of
micrometric size. Fluids can be gases, liquids, Newtonian or not, mono or multiphasic
(e.g. oil and water). Systems can be devices with channels, patterned surfaces, or
paper sheets. Micrometric is an order of magnitude. In practice, microfluidic scales
range from 100 nm to 1 mm. The definition proposed here is currently used in the
field1 and we will adopt it throughout the book.2

1.1 Astonishing microfluidic systems in nature

Obviously, nature manipulates, with exquisite control, fluids at the microscale. Oth-
erwise no life would be possible. The tree is an example. In the tree of Fig. 1.1, tens
of thousands of leaves are nourished by a network containing thousands of capillaries
of diameters on the order of tens of micrometres (in the trunk and the branches) and
billions of pores of several tens of nanometres (in the mesophyll cell system, in the

1 In an influential paper [1], titled ‘The origins and the future of microfluidics’, G. Whitesides
defined microfluidics as ‘the science and technology of systems that process or manipulate small
(10−9 to 10−18 liters) amounts of fluids, using channels with dimensions of tens to hundreds of
micrometers’. The definition, although more restrictive about the scales, is essentially is the same as
ours.
2 Two remarks must be made at this level:
- Microfluidics should not be confused with microhydrodynamics. Microhydrodynamics is the study
of creeping flows, i.e. flows at low Reynolds numbers. In the definition of microhydrodynamics, ‘micro’
refers to the Reynolds number, not the system size. In microfluidics, ‘micro’ refers to the system size,
not the Reynolds number. One can have microhydrodynamic flows of decimetric sizes (e.g. honey
poured from a spoon) and microfluidic flows operating at substantially high Reynolds numbers (e.g.
inertial micromixers).
- ‘Microfluid’ is a word that sometimes appears in the literature. It is not a physical concept.
‘Fluid’ refers to a state of matter defined microscopically. In the bulk, gases and liquids circulating in
microchannels possess exactly the same microscopic structure as in large containers. There is no new
phase. In extremely confined systems, for instance, in carbon nanotubes, the liquid structure can be
affected by the walls. However, this concerns only nanometric scales.
2 Introduction

Fig. 1.1: This tree possesses a complex network of capillaries (xylem and phloem) that
supplies sap homogeneously to the tens of thousands of leaves that it carries on its branches,
and redistribute carbohydrates and other organic compounds between leaves, roots, and fruits
(credit: Johannes Plenio) [2].

leaf).3 In the pores, sap evaporates, through a process called transpiration, creating
interfaces that pull the sap from the roots to the top [3, 4]. Despite the complexity of
the network, the supply of sap is stable in time and homogeneous in space. The hy-
drodynamics of the tree is extremely subtle. For instance, large trees drive the sap at
negative pressures [5,6]. No hydraulic system made by humans functions like this. The
reason is that when the pressure is negative, bubbles nucleate and the flow is unstable
and becomes out of control. In trees, bubbles grow, but several mechanisms, including
mechanical, block their propagation, thereby preventing embolism and death [4].
We will come back to the trees in Chapter 4, when capillary phenomena are discussed.
Here, it is enough to observe that the tree provides an example of exquisite microflu-
idic control.. This control is achieved not for the pleasure of realizing a technological
performance, but to ensure survival.
A similar problematic holds for the spider. The spider produces4 long silken threads, a
few dozen micrometres in diameter, forming a complex pattern -the spider web-, each
thread developing a resistance to rupture that is twice as great as that of steel [7, 8].
How does the spider manage to produce this material? To make a long story short,
5
the silk solution is contained in the glands, which are highlighted in Fig. 1.2. In
most spiders, six glands produce different types of silks. The solution is driven in small
capillaries, several tens of μm in diameter. By actuating valves, or selecting the glands,

3 The mesophyll part is much more resistive than the other parts of the tree. This explains why
the tree distributes the sap among the leaves, in a remarkably homogeneous manner.
4 All spiders produce silk but only a fraction of them produce webs. Those which do not produce
webs use the silk wires to build cages protecting their eggs, or to pass from one leaf or one branch to
another to move more swiftly.
5 An excellent, description of the spider silk production is given by the Museum of Australia, in an
article named: ‘Silk: the spider’s success story’.
 Exquisite microfluidic control in the human body 3

the spider chooses the silk it wants to produce, for one part of the web or another.
Each gland works as part of a pair. In the capillary, one type of silk occupies the
central part of the channel, and the other the periphery. Thus, they form a concentric
system in which the two silks flow side by side. In their journeys, the silk solutions
deshydrate and, combined with the effect of the large deformation rate to which they
are subjected,6 tend to harden. At the end of these journeys, the soft material is
extruded at the back of the abdomen, through tens of submicrometric nozzles; it then
evaporates and solidifies. The spider thus controls a complex annular flow structure,
the dehydratation process along the capillary, and the final extrusion through many
nozzles. This technological tour de force allows it to catch insects to eat.

1.2: Two glands produce the silk of

the spider web, each gland secreting a
particular type of silk. The silk is in a
liquid state in the glands. During its
journey towards the outlet, the solu-
tion hardens under the effect of large
deformations, dehydrates, turns into
a soft material, exits the spider, evap-
orates, and solidifies.

1.2 Exquisite microfluidic control in the human body

Humans control flows at the microscale with a high level of precision, and for the
same reasons as animals and trees: to be viable. The examples are numerous and they
concern almost all parts of the body. Here, just a few are mentioned.
Blood circulation The blood network is formed by vessels of different types: arteries,
veins, and capillaries. Each of these plays a specific role in the circulation process.
The total length of the vessels is impressive: 100,000 km. Their diameters range
from 25 mm (the aorta) to 8 μm, i.e. the size of a red blood cell.7 Most of the vessels
are less than 100 μm in diameter, i.e. they have a microfluidic size. Fig.1.3 (A)
shows the vessel network of a lymph node. Its geometrical complexity is evident.
Blood itself is a complex fluid: it comprises a plasma, which includes a large variety
of solutes, and several types of cells -mostly red blood cells (erythrocytes)-. A
number of phenomena, described in the literature, alter the homogeneity of the
suspension. The Fahraeus effect [9] whereby, along microchannels whose diameter
decreases in the streamwise direction, the cell concentration decreases. The effect
is linked to the particular interaction that the cells develop with the walls. In
the blood network, different hydrodynamic regimes take place, depending on the
ratio of vessel diameter to blood cell diameter. Despite this complexity, the blood
circulation is precisely controlled. The averaged flow rate is tuned so as to be
6 It is considered that the silk solution behaves as a shear-thickening fluid.
7 Red blood cells are biconcave disks, 8 μm in diameter average
4 Introduction

Fig. 1.3: (A) Representation of a network of blood vessels. Gases and nutrients are exchanged
between the blood and surrounding tissue through the permeable walls of capillaries, the
smallest blood vessels(from Design cell, 2013). (B) Sketch of the gas transfers taking place
in the lungs. (C) Migration of a megakaryocyte in the bone tissue, penetration in the blood
vessel, and the subsequent platelet formation.

compatible with the metabolism. In the network, no clogging occurs, and the
pressure drops are kept small. Driving a suspension, under precise control, in such
a complex microfluidic system, represents a formidable challenge that evolution
has managed to meet.

Gas exchanges in lungs Lungs take in oxygen and eliminate carbon dioxide. The
gas transfer between blood and air, at the alveola level, is schematized in Fig.
1.3 (B). The membrane (made of cells) separating the alveola and the surround-
ing capillaries is one cell thick, i.e. about 1 μm. With such a small dimension,
gases pass quickly through the membrane into the blood. Oxygen-deficient, car-
bon dioxide-rich blood return to the heart. Although capillaries host only one
cell in their cross sections, no clogging occurs. Lungs typically contain 480 mil-
lion alveoli. The respiratory system provides an example where humans exert an
exquisite control of the blood circulation and gas exchanges, at the micron scale.

Platelet production : Platelets are cells measuring 2-3 μm and without nucleus.
They prevent haemorrhages. Without them, we would not survive. To carry out
this function, platelets, initially spherical, take the form of a star, adhere and
aggregate at a wound and trigger thrombin generation and fibrin formation to
create a clot, further strengthened by the accumulation of white blood cells [10].
Fig. 1.3(C) sketches the process of formation. Megakariocytes (30 μm large), are
produced in the bone marrow, then travel through the bone tissue and enter
the blood circulation, by migrating through the vessel wall. In the blood stream,
being tethered to the wall, they are stretched out by the shear flow and break up
to small vesicular blobs of micrometre size, called platelets [11]. The lifespan of
platelets is limited, and we must produce billions of functional platelets every day
to renew them. If the blood circulation were too fast, turbulence would develop
and the breakup process would become out of control. If the blood circulation
were too slow, the megakariocyte would not break up and no platelet would be
 Exquisite microfluidic control in the human body 5

produced. The process is thus finely controlled.

(a) (b) (c) Water channel

Evaporation
Tear film:
Mucous layer
Watery layer
+ Cell membrane
Oily layer Cell membrane
Skin +

Cornea Conjunctive
Sweat gland

Fig. 1.4: A - Sketch of a skin pore with its eccrine gland; B - Eye with its film: C - Aquaporin,
across which water molecules circulate.

Sweating Sweating cools down the body, during exercise, when the temperature is
hot or in case of fever [12]. Fig. 1.4 shows an eccrine gland, which produces sweat
and drives it through the microfluidic sudoral channel (50 μm in diameter), up
to the skin pore. The flow rate is determined by the kinetics of secretion of the
gland. Once at the skin level, sweat forms a film or a droplet, which evaporates
and, in turn, according to thermodynamics, cools down the blood circulating in
the region and thereby the body. Typical energy losses per unit of time are 350
watts, several times the basal metabolism, i.e. the energy, per unit of time, used
by a human at rest, as will be seen in Chap. 2. Should the sweat flow be too
large, the skin would be covered by a thick film, and, in the opposite case, there
would be no thermal regulation. Thereby, again in this case, evolution has led to
a fine control of microfluidic sweat flows, which plays an instrumental role in our
thermal stability.

Tears Tear film is essential for clear vision and eye health [13]. It protects the ocu-
lar surface with moisture, transports waste away, and provides a smooth optical
surface. This film has a micrometric thickness. After each blink, it re-forms. De-
ficiencies in the tear film causes blurred vision, burning, foreign body sensation,
and tearing. Flow control is instrumental to avoid these problems.

Aquaporin The preceding examples were concerned with micrometric scales. Hu-
mans also control flows at the nanometric scale. An example is aquaporin. The
function of this molecule is to transport water across cell membranes in response
to osmotic gradients [14]. Without cells could blow, as red blood cells do when
they are suddenly immersed in a salty solution. Typically, cells contain up to 30
aquaporins. As illustrated in Fig. 1.4(C), in the presence of an osmotic pressure
gradient, water molecules are forced to circulate through the aquaporin, in one
direction or the other, to cancel it. Aquaporins are extremely selective: only water
molecules pass through them. They never clog, due to the action of an anti-fouling
mechanism that is not yet understood. Their permeability is sufficiently high [15]
to enable fast equilibration, since if the equilibration process were too slow, cells
would be unstable. It has been suggested that the hourglass geometry of the
6 Introduction

molecule plays a role in this property. Aquaporin illustrates well the exquisite
control that humans have developed at the nanoscale to ensure their viability.

1.3 MEMS, the mother of microfluidics

As evoked in the first lines of the introduction, man has begun to create, in the seven-
ties, machines of micrometric sizes. These systems were called Micro ElectroMechanical
Systems (MEMS). MEMS sizes currently range between 1 μm and 500 μm. An excel-
lent example of a MEMS, hilglighted in [16], is shown in Fig. 1.5. This MEMS is a
microgear whose overall size is 200 μm. It is held by an ant. The figure illustrates the
intrusion of a man-made machine into the small animal realm. The entry into micro-
metric scales is not a new feat, however. Since the invention of the optical microscope
in the sixteenth century, the micrometric world has been scrutinized in detail. The mi-
croscope has allowed countless scientific discoveries to be made. However, before 1970,
humans did not act at a micrometric scale, which is precisely what MEMS allows. As
a result, when MEMS appeared on the scene, their potential of applications was clear.
The first example concerns airbags for cars.

1.5: Ant holding a nickel micro-gear,

made by LIGA technology (German
for ‘lithographie, galvanofomung, ab-
fomung’). This ant was metallized
and placed in a vacuum in order
to be photographed by electron mi-
croscopy. This image was provided by
the Karlsruhe group (Germany) [17].

MEMS for airbags, which first appeared in the 1980s, consist of an integrated system
on a silicon wafer that is just a few millimetres long. This is shown in Fig. 1.6. The
heart of the chip is made of two combs, one fixed and the other mobile; the capacitance
of these combs varies under the effect of an impact. As we will see in Chapter 2, the
miniaturization of the capacitor element allows the creation of a highly sensitive and
rapid detector. The industrial success of MEMS for airbag is not solely due to the
improvement in sensor response and sensitivity, but also to the ability to integrate
detection, information analysis, and signal processing all on one single chip. Just as
with integrated circuits, this chip can be reproduced by the million. The cost, which is
critical in the field of automobile manufacturing, becomes very advantageous compared
to traditional systems. For this reason, all modern automobiles now use MEMS for
their airbags, and tens of millions of these devices are produced each year.
The history of MEMS is interesting. The year 1959 is often considered to be the
beginning of the history of micro- and nanotechnology. In December of that year,
 MEMS, the mother of microfluidics 7

Fig. 1.6: A - SEM image of a part of a MEMS - accelerometer for airbag (Reprinted from Ref.
[18], with permission from IOP Publishing, Ltd, Copyright 2022); B principle of functioning
of the accelerometer.

a visionary speech was given by Richard P. Feynman during the American Physical
Society (APS) meeting at Caltech. This speech was entitled ‘There is plenty of room
at the bottom’ [19]. An early part of the speech is as follows:

I would like to describe a field, in which little has been done, but in which an enormous
amount can be done in principle. This field is not quite the same as the others in that it will
not tell us much of fundamental physics (in the sense of ‘What are the strange particles?’)
but it is more like solid-state physics in the sense that it might tell us much of great interest
about the strange phenomena that occur in complex situations. Furthermore, a point that is
most important is that it would have an enormous number of technical applications.

Feynman saw no physical reason why the 50 volumes of Encyclopedia Britannica could
not be inscribed on the head of a needle. One letter would only need to consist of less
than a dozen or so molecules. Confronted with the difficulty of working at micro-
metric scales, he suggested that we ‘train ants how to teach mites’ how to construct
miniaturized machines’.

How many times when you are working on something frustratingly tiny like your wife’s wrist
watch, have you said to yourself, ‘If I could only train an ant to do this!’ What I would
like to suggest is the possibility of training an ant to train a mite to do this. What are the
possibilities of small but movable machines? They may or may not be useful, but they surely
would be fun to make.

These suggestions or predictions did not remain just a fantasy; since three decades
later, in 1990, the word ‘IBM’ was spelled out with 35 xenon atoms [20] (see Fig.1.7).

1.7: IBM spelled out with 35 xenon

atoms, deposited on a cold (4K)
Nickel (110) surface. The atoms were
transferred from the sharp Scan-
ning Tunnelling Microscope (STM)
tip and imaged with the STM [20].
This image received a large echo.
8 Introduction

The first MEMS devices were created two decades after Feynman’s speech [21]. The
first microbeam was reported in 1982 (Fig. 1.8A), and the first microspring in 1988.
The first micromotor was created in 1988 [22,23] (Fig. 1.8B)8 . It consisted of an electro-
static motor, where the rotating electric field was generated by electrodes evaporated
onto polysilicon. One major difficulty was that stiction (the combined phenomena of
adhesion and friction), which tended to block the rotor was exacerbated by minia-
turization. The solution to this problem consisted of reducing the surface area of the
rotor/substrate contact, but this made microfabrication of this machine more difficult.

Fig. 1.8: (A) First microbeam (1982). (B) First micromotor, made at UC Berkeley by Tai
and Muller in 1989. This motor has been placed next to a human hair whose diameter is on
the order of 200 μm. (Courtesy of Professor Richard S. Muller, Berkeley Sensor & Actuator
Center, University of California, Berkeley).

Other examples are a microgripper, a hot-wire rake [21], and an astonishing microgui-
tar, shown in Chapter 7, with nanostrings 30 nm in size, vibrating at MHz frequencies.9
In addition, some unsubstantiated concepts were proposed: for instance, a MEMS con-
sisting of inclinable mirrors that permit communication between the ground and an
airborne micro-engine.10 The project failed, but the concept was appealing. Not all of
these objects were practical, but all of them stimulated the imagination.
Today, the MEMS market is estimated to be worth between US 10 and US 13 billions,
depending on the agencies. Examples of companies are ARM Holdings. Bosch, Cisco
Systems Inc., InvenSense, Knowles Electronics, MediaTek Inc. Microchip Technology
Inc., Samsung Developers, STMicroelectronics, and, Texas Instruments. MEMS pro-
duction includes a variety of products like MEMS for airbags, microgyroscopes for
mobile phones, micromirrors for digital projectors (Fig. 1.9), pressure sensors, to men-

8 We will see in Chapter 2 that this micromotor can comprise the base element of a microturbine
that converts chemical energy to electrical energy. It is also interesting to note that microgears,
fabricated using MEMS technology, are often used today in clock making.
9 The guitar, 10 μm long, with six 30 nm wide cords, will be shown in Chapter 7. It was realized
at Cornell, in 1997, in the group led by Prof Craighead.
10 This concerned a project, written by Kristofer S. J. Pister, Joe Kahn, and Bernhard Boser, the
University of California, Berkeley, in 1997, whose objective was to build wireless sensor nodes with a
volume of one cubic millimeter.
 The birth of microfluidics 9

Fig. 1.9: (A) Optical MEMS (MOEMS) made by Texas Instrument, used in digital projectors.
(B) An array of micro-mirrors used in a video-projector. The size of each mirror is 10 μm
[24](Creative Commons Attribution 4.0 Unported License).

tion a few. The applications are impressive, but there is still room at the bottom. For
instance, no MEMS is capable of flying and act, for instance, like a mosquito. Creating
a MEMS-mosquiito would necessitate developing an extremely lightweight and power-
ful energy source, that does not exist yet, and integrating an extraordinarily efficient
motor, along with ultraminiaturized micro-pumps. All this, today, looks impossible.

1.4 The birth of microfluidics

We now concentrate on microfluidics. As in the case in many fields, the birth of mi-
crofluidics, defined as ‘the science of manipulation of fluids at the microscale’, as said
above, has been preceded by precursors and even precursors of precursors. People have
succeeded, in many circumstances, in manipulating fluids at the micrometric scale, in
a controlled manner, with only their hands. To do this, they exploited hydrodynamical
laws that naturally provide a control of the micrometric scale. An example is painting.
Painters deposit micrometric layers on walls, whose thickness is controlled by the speed
of the brush. Another example is soap bubbles. The films, stabilized by surfactants,
are sub-micrometric. These examples show that without microtechnological toolbox,
the micrometric scale, in a number of cases, can be controlled.
Of much greater technological significance is a device producing submillimetric droplets.
It was invented in 1964 by R. G. Sweet [25, 26] (see Fig. 1.10).

~
18

16
14 12
24
HV+ 10
30 20
22
8
28
26

32
4

HV–

4 6 1.10: Sketch of the system used by R. G.

2
Sweet [25].
10 Introduction

The invention, with others, played an instrumental role in the field of ink-jet printing,
a major application of microfluidics. Here, droplets were emitted and deposited on a
moving sheet. A key point was the electrostatic control, which allowed large quantities
of nanoliter droplets to be produced, in a reproducible manner, opening an avenue
towards the fabrication of ink-jet printers.

1.11: First microfluidic sys-

tem (1975) [28]: a miniaturized
gas chromatographer, including
an injector, a 1.5 m long mi-
crochannel and a thermal detec-
tor.(Reprinted with the permis-
sion of IEEE, copyright 2022.)

In the 1970s, silicon-based MEMS technology was well advanced, many clean rooms
were available, and there were no difficulty in etching grooves on silicon wafers to
create microchannels. In this context, the first microfluidic device, invented by S.C.
Terry [27], appeared on the scene in 1975. Unlike Sweet’s invention, the manufacturing
and design of the device prefigured the silicon-based microfluidic devices that would
later develop, giving birth to microfluidics. The device made by Terry [27, 28] was a
miniaturized gas chromatographer. It is shown in Fig. 1.11.
Terry’s device circulated gas through a 1.5 m microchannel etched in a silicon wafer,
bonded to a glass plate. The system included a miniaturized electromagnetic injector
and a thermal detector. Separations of gaseous hydrocarbon mixtures were performed
in less than 10 s, without compromising the efficiency, which was an impressive feat
at that time.
Nonetheless, although the device was industrialized several years later, the techno-
logical jump remained isolated, because, during these years, the separation-science
community was not ready to adopt silicon technology [29]. It was only after 1990 that
the advantages of miniaturization were thrust into the spotlight, for its application to
electrokinetic liquid chromatography [30–32].
A seminal paper appeared in 1990 [30]. By reasoning on scaling laws, A. Manz et
al. argued that miniaturization enables the creation of performing separation systems,
combining portability, low cost, and high speed, without compromising the efficiency.11
We will describe these systems in Chapter 6. The paper introduced a new acronym,

11 The conclusion of the paper was: ‘A basic theory of hydrodynamics and diffusion indicates faster
and more efficient chromatographic separations, faster electrophoretic separations and shorter trans-
port times for a miniaturized TAS. The consumption of carrier, reagent or mobile phase is dramat-
ically smaller. A multi-channel device would allow the simultaneous performance of a large number
of measurements (under the same conditions).’
 The birth of microfluidics 11

1.12: First electrokinetic separation mi-

crosystem (1992). Overall dlmensions are
14.8 cm x 3.9 cm x 1 cm. (Reprinted rom
Ref. [31], with permission of Americal Chem-
ical Society. Copyright 2022.)

μTAS, which stood for Micro Total Analysis System. The term looks awkward. What
does it mean? A ‘Total Analysis System’ (TAS) refers to a system in which analyti-
cal instruments are transported on a cart, to perform the total analysis of a sample
(sampling, sample transport, chemical reactions, detection) in the field. Thanks to
miniaturization, no cart was needed, and the equivalent system was called μTAS. The
ideas were substantiated two years later, with the realization of the first electrokinetic
separation microsystem [30].12 This demonstrated performances in line with expecta-
tions, i.e. high speed of separation, and excellent efficiency, transportability, and low
cost. Microfluidics was born.

Later, all sorts of microfluidic systems were fabricated : electrophoretic separation as-
says [33, 34], electro-osmotic pumps [35], diffusive separation systems [36], micromix-
ers [37–42], DNA amplification systems [43–49], cytometers [50, 51], DNA separation
assays [98–102], centrifugal microfluidics (see review [103]) and chemical microreac-
tors [52–56], to cite a few examples. A number of these inventions were to play an
important role, a decade later, in the ramping up of the microfluidic market.

1.13: Cover of the pro-

ceedings of the first μTAS
conference organized by
A. van den Berg and
P. Bergveld in 1994, in
Enschede (the Nether-
lands). (Reproduced
with permission from the
Chemical and Biological
Microsystems Society
(CBMS), copyright 2022
CBMS.)

Up to 1994, the young microfluidic community gathered in different meetings, in par-

ticular in MEMS conferences (launched six years before, in 1988), in which one or
two sessions were devoted to microfluidics. The situation was not optimal. Controlling

12 In fact, the device was manufactured in 1988.

12 Introduction

fluid flows in microdevices required discussion of subtle hydrodynamic and physico-

chemical phenomena, and these were far from the focus of these meetings. Therefore, a
desire to organize specialized seminars emerged, in order to focus on these phenomena,
and, in the meantime, take the specific technological context into account, most often
overlooked in traditional academic meetings. The first μTAS conference was organized
by A. van den Berg and P. Bergveld in 1994, in Enschede (the Netherlands). The cover
of the proceedings is shown in Fig. 1.13. A total of 160 participants attended. Today,
μTAS conferences bring together 1,500 participants.13

In these early days, a ‘paradigm’, i.e. a set of concepts and practices that define a sci-
entific discipline at any particular period of time’, or, more simply, ‘what the members
of a scientific community, and they alone, share’ [57], emerged. It was thought that the
objective of microfluidics was to create basic microfluidic functions, or ‘bricks’, and
assemble them, in a way similar to microelectronics, so as to generate complex func-
tionalities, that could respond to unmet needs in biology and chemistry. The notion of
lab-on-a-chip was frequently put forward, and stunning cartoons designed to illustrate
the concept, such as those of Fig. 1.1414 , appeared in many broad-readership (for in-
stance Ref. [58]), augmenting the visibility of the field. But why did the community
not fabricate complex microfluidic systems, comparable to microelectronic processors,
as they envisioned? One bottleneck was valve integration. It appeared impossible, at
that time, with silicon, plastics or glass, to integrate more than a few valves on a
device. There was no chance to compete with the millions of transistors integrated, at
that time, on central processing units (CPUs).

In the same period, hydrodynamic experiments revealed unexpected phenomena; for

instance, large slippages, which apparently contradicted the no-slip dogma of hydro-
dynamics [59, 60]. In the meantime, microfluidic technology revealed the mechanical
behavior of single DNA molecules. The experiment was performed by Chu et al. [61]
in 1993, in cross-flow microchannels. This work laid down the foundations of a new
domain: the study of single molecules. It also showed that microfluidics could provide
new tools for investigating fundamental questions. Around that time, a number of
fundamental contributions allowed the physical role of the confinement to be clarified.
This concerned, for instance, electrohydrodynamics, electrowetting, chaotic mixing,
polymer dynamics, gas flows, or fluid interfaces. The list is long and we will return to
these subjects later. Molecular dynamics simulations, on the other hand, enlightened

13 The second meeting, μTAS 96, was held in Basel with 275 participants. The first two meetings were
held as informal workshops. By the time of the third workshop, μTAS 1998 (420 participants), held
in Banff, the workshop had become a worldwide conference. The number of participants continued
to increase in μTAS 2000 (about 500 participants) held in Enschede and μTAS 2001 (about 700
participants) held in Monterey. The number of submitted papers also dramatically increased in this
period from 130 in 1998, to 230 in 2000, to nearly 400 in 2001. From 2001, μTAS became an annual
symposium (text of presentation of μTAS 2002, held in Nara (Japan), and written by Yoshinobu
Baba and Shuichi Shoji).
14 Fig. 1.14(A) is stunning, but it is misleading in the sense that, in practice, the functionalities
shown on the figure are much more difficult to integrate than suggested. Perhaps inspired by this
unrealistic illustration, or understanding the analogy between microelectronics and microfluidics in a
too literal sense, several companies wasted much time and energy developing complex lab-on-a-chip
that had no chance of working
 The birth of microfluidics 13

Fig. 1.14: (A) Cartoon published in 1998 [58], showing an imaginary microfluidic device in-
tegrating several functionalities: electrophoretic separation, heating, driving, mixing, extrac-
tion, and polymerase chain reaction (PCR) amplification. (B) Image elaborated by Caliper,
around 2000, that illustrates well the lab-on-a-chip concept.

a number of phenomena, sometimes in conflict with the experiment. This concerned,

for instance, slippage, nanobubble dynamics, or electrowetting. Again, we will return
to these subjects later.

By the turn of the twenty-first century, many microfluidic devices were created in the
labs, some patented (one hundred or so) awaiting commercial developments, many
not. Before 2000, the global microfluidic market was a nano-market, worth less than
US$100 M.15 A number of microfluidic products, although elegant, did not meet any
market. This was the case of the microfluidic fountain pen [62]. Others did not function
with sufficient reliability: this was the case of the glucose watch, dedicated to monitor
glucose by sampling interstitial fluids. It will take years before this product finds its
place in the market [63]. Other inventions met considerable success: one developed
by i-Stat (created in 1983), of extremely simple construction, dedicated to glucose
measurements. In the 1990s, i-Stat company shipped several million cartridges every
year to hospitals. Two commercial products, whose production volumes were not as
important, but which promised an interesting future, must be mentioned: the Bioana-
lyzer, dedicated to DNA separation, made by Caliper Agilent, and the SmartCycler R

System, a Point of Care (POC) molecular system, developed by Cepheid (founded

in 1996). The two systems marked the entry of microfluidics in molecular biology.
Cepheid would become, years later, a major player in this field. In the meantime,
several microfluidic foundries, such as Micronit (founded in 1999), or ChipChop, and
later on, Dolomite (2005) for microfabrication, and Fluigent (2006) and Elvesys (2011)
for instrumentation, were created. They played an important role in the development
of microfluidics.

15 Although inkjet printing is a microfluidic system, by convention, it is not counted as part of this
market.
14 Introduction

Who first used the word microfluidics ?

Before 1993, the word microfluidics’ was essentially absent from the vocabulary of
the microsystem community It did not appear anywhere in the 857 pages of the
Microsystem technology 90 Proceedings [67], an important conference of the young
MEMS community. Instead, the words ‘micro liquid-handling’, ‘micro-hydraulics’,
‘micro fluid’, with or without hyphen, or micro-liquid flows’ were preferred [68]. ‘Mi-
crofluidics’ was also absent from Terry’s [28], and Manz’s [30,31] papers, published in
1979, 1990, and 1992. Why was this the case ? We may hypothesize that researchers
of that time, involved in what we now call ‘microfluidics’, were reluctant to use a word
that looked esoteric. The situation seemed to change in 1993, after the publication,
by P.Gravesen, of a survey titled ‘Microfluidics: A Review’ [65]. The paper acquired a
significant visibility. Although no etymologic study has been done, and probably will
never be done, one may hypothesize that the review strongly contributed to, if not
initiated, the spread of the word. Later on, although invisible in the MicroTas94 Pro-
ceedings, the word ‘microfluidics’ increasingly appeared in the literature. The word
was suitable since it represented, as pointed out by R. Zengerle [66], a ‘headline’ cov-
ering all types of actions involving fluids, and performed in microsystems. After 1997,
it frequently appeared and eventually acquired a prominent place in the language.
Finally, the community working on fluid manipulations at the microscale adopted it
to give a name to their field.

1.5 The advent of soft technology

An important step was taken in 1998, with the development, by G. Whitesides, of

soft technology [69–71].16 With soft technology, devices were no longer made in glass
or silicon, but in soft materials. From a technological perspective, this represented a
considerable shift. Two images, extracted from [69] and [70], are shown in Fig.1.15.

Fig. 1.15: (A) Scanning electron micrographs of a PDMS honeycomb structure, created by
molding the polymer against a photoresist mould (Reprinted from Ref. [69] with permission
from Wiley and Sons, copyright 2022 Wiley and Sons). (B) Double-T section of the network of
channels. The roughness in the side walls of the PDMS channels arises from the limited mask
resolution (Reprinted with permission from Ref. [70], copyright 2022 American Chemical
Society.).

16 Whitesides’ papers were preceded, in 1997, by two pioneering works [72, 73].
 The advent of soft technology 15

Figure 1.15(A) is taken from Ref. [69]. The paper explained the concept of soft technol-
ogy, and its two main facets, microprinting and micromoulding, which we will discuss
in Chapter 7.
Figure 1.15(B) shows microchannels made in PolyDiMethylSiloxane (PDMS) [70]. We
will hear much about PDMS in the book. It is almost a miraculous material. It has
properties that no other material possesses: deformable, transparent in the visible
range, insulating, hydrophobic, sticking to glass in a reversible manner. We will see, in
Chapter 7, that all of these properties make soft lithography possible. In Fig. 1.15(B),
the aqueous sample is introduced in the double T, then driven by electroosmotic
forces into a long microchannel (not shown), along which electrokinetic separation is
performed, in a manner similar to Manz’s work [31]. The walls are rough, due to the low
mask resolution. The authors achieved ionic separation in the system, with a resolution
and a speed comparable to silicon devices. The paper thus suggested, that, even though
surface chemistry is not as controlled as in glass or silicon, electrokinetic separation,
the major application of microfluidics at that time, is feasible. For chromatographic
experts, this was quite a surprise.
Nonetheless, the message received by many researchers was not about separation. It
was that PDMS microdevices were easy to create.17 Once the mould was fabricated,
the rest of the technological process could be made outside a clean room, without
specialized skills. A master student could learn it in one day. The simplification of the
technological process gave rise to a surge of activity. Many laboratories, interested in
microfluidics, but with limited access to clean rooms, came to the field and started
investigating new directions. One could compare this period to the transition from
centralized informatics to laptop computing.
Why did all the community not rush out to use soft technology, abandoning silicon?
Two reasons can be provided: the first is that PDMS surfaces are not stable, meaning
that performing accurate separations with such a material was impossible. For the
community of that time, composed, mostly, of analytical chemists, this represented a
serious limitation. Secondly, PDMS devices cannot be produced in large quantities.
Scaling up is not possible, while with silicon it is. These two arguments led most of
the microfluidic community of that time to keep working with silicon and glass. The
arguments were not unreasonable. Today, half of the hundreds of millions of devices
produced in the microfluidic market are made with glass and silicon.
Most researchers using soft technology were thus newcomers. In the same period, press-
ing needs appeared: examples are DNA sequencing, cell sorting, molecule screening,
single cell analysis, and proteomics. As mentioned earlier for the particular case of
DNA sequencing, all these applications conveyed big numbers: five billion cells to sort
for isolating 1-10 circular tumor cells (CTC) [76], hundred of thousands of compounds
for drug discovery [80], millions of fluid manipulations for performing gene sequenc-
ing [75]. How so many experiments could be carried out rapidly and in parallel, while
consuming little reagent [74]? This was the new challenge faced by microfluidics. The
17 The discovery of SU8, by IBM, in the 1990s, allowed thick moulds to be made in one photolitho-
graphic step, facilitating the fabrication of PDMS devices.
16 Introduction

challenge was not only intellectual. Today, the aforementioned domains represent the
largest share of the microfluidic market.18
In this context, a surge of innovations emerged. One landmark was the demonstration,
in 2002, that, by exploiting the deformability of PDMS [77], thousands of valves could
be integrated on the same device [78]. This is shown in Fig. 1.16. The valve problem,
raised in the previous decade, seemed to be solved. The technology gave birth to a
company, Fluidigm, whose valuation would soon reach US$1bn. With Cepheid and
Fluidigm, two microfluidic unicorns were born in the years 2000-2010.19

1.16: The device, whose channels are visu-

alized with dyes, contains 2056 microvalves.
The system performs distinct assays in 256
nl subreaction chambers. Each of them is
individually addressed, using a multiplexed
valve system,. With it, only 20 valves are
needed to control the assay. (Reprinted
from [78], with permission from the Ameri-
can Association for the Advancement of Sci-
ence. Copyright 2022.)

The device shown in Fig. 1.16 contains 2,056 microvalves. Each of the 256 chambers
on the chip is individually addressed. A multiplexer allows for reducing the number of
connections, from N (N being the number of chambers) to 2 log2 N [78]. It is impossible
to realize such a system with glass or silicon. For microfluidics, a new period opened.

1.6 Diversification of the technology and broadening of the

applications
In the years 2000-08, thanks to soft lithography, the microfluidic community had an
easier access to technology. The number of applications, including those involving large
numbers, increased, and growing public support raised momentum in universities. In
such conditions, the microfluidic community grew substantially, reaching thousands
of researchers worldwide. The stimulating atmosphere of that time was perceptible in
the μTAS conferences.
Droplet-based microfluidics appeared in the period 2000-2002. We will present the
technology in Chapter 4. At first glance, producing microdroplets under control did not
appear new. For instance, ink-jet printers and fluorescent active sorters (FACS) already
18 Before the advent of the COVID-19 pandemic.
19 At about the same time, Theranos, which promised to perform a hundred different tests with a
single drop of blood, reached much higher valuations. Ten years later, the company collapsed. Because
of the absence of peer-reviewed publications issued by the company, links between the microfluidic
community and this company were inexistent.
 Diversification of the technology and broadening of the applications 17

performed the task. The new idea was that, by playing with the device, droplets could
be filled with reagents, and thus operate as microreactors, being moreover well mixed,
thanks to vigorous recirculations developing inside them, as we will see in Chapter 4.
Millions of chemical reactions per hour could be run, using minute amounts of reagents
(less than one nanolitre or so per droplet). These performances allowed researchers to
work with large numbers, responding to the needs appearing at the turn of the century,
as mentioned earlier.

1.17: Two reagents A and B are

introduced in a T junction, in
a flux of oil, and form droplets.
By varying the flow rates, each
droplet can acquire different reagent
concentrations. This system allows
high throughput screening (kHz) of
biochemical reactions, using min-
imal quantities (nl) of reagents.
(Reprinted from Ref. [79], with per-
mission of John Wiley and Sons,
copyright 2022.)

Fig. 1.17, published in 2003 [79], illustrates well the situation. In short (this will be
explained in more detail in Chapter 4), the entry was composed of three channels: one
for reagent A, another for reagent B, and, in between a buffer. By varying flow rates,
the amount of reagent could be varied in each droplet. Then it was possible to screen
a chemical reaction, as a function of the phase ratio. With droplets being emitted
at kHz, the screening was orders of magnitude faster than could be achieved by any
robot.
In about the same period, it was demonstrated that more complex droplets, in the form
of double or multiple emulsions, could also be produced under control. This topics was
developed by D.Weitz (Harvard). This was an important step, because it established
the link between microfluidics and colloidal and material sciences.
Fig. 1.18 is extracted from a paper published in 2005 [84]. The device consists of three
glass tubes, one with a square, the others with circular cross sections, placed inside
each other. Along the tubes three liquids, immiscible or immiscible by pairs, are driven,
forming droplets inside droplets, i.e. double emulsions. The structures of the emulsions
depend on the flow rates, the nature of the phases and the formulation. The device
stood against the technological wind of the time: neither soft nor hard technology were
used.20 The work opened a route towards creating objects difficult or impossible to
make. Implications in various domains such as cosmetics, oil or food industries were
numerous.

20 Double emulsions have indeed also been created in microchannels, made with soft or hard tech-
nologies (see, for instance, [86]). The advantage of glass tube technique is its capacity to obtain well
defined hydrophilic and hydrophobic regions. Disadvantages are size limitations (double emulsions
below 30 μ are difficult to make), delicateness of the geometrical adjustment of the tube noses, and
limited parallelisation.
18 Introduction

Fig. 1.18: Microcapillary geometry for generating double emulsions from coaxial jets [84].
(A) Schematic of the coaxial microcapillary fluidic device. The geometry requires the outer
fluid to be immiscible with the middle fluid and the middle fluid to be in turn immiscible
with the inner fluid. (B to E) Double emulsions containing only one internal droplet. (F
and G) Double emulsions containing many internal drops with different size and number
distributions. (H) Double emulsion drops, each containing a single internal droplet, flowing
in the collection tube. The devices used to generate these double emulsions had different
geometries. (Reprinted from Ref [84], with permission of the American Association for the
Advancement of Science, copyright 2022).

Still in the same period, i.e. around the years 2000, digital microfluidics appeared
[87, 88]. The technology exploited a technique, called electroWetting on dielectrics
(EWOD), developed first by Berge [89]. EWOD consists of placing a conducting droplet
(such as salted water), polarized at some tension, over a thin hydrophobic insulating
layer, deposited on a metallic electrode. The electrical tension changes the apparent
contact angle. With this, the solid surface can switch, in a fast time (sub-millisecond)
from hydrophobic to hydrophilic state. Digital microfluidics exploits this functional-
ity, by displacing droplets over a surface patterned with addressable electrodes. In
the period 2000-10, several groups [88, 90–95] demonstrated impressive realizations:
rotating droplets, droplets moving on substrates in a complicated manner, interact-
ing together, mixing reagents, initiating chemical reactions and cutting droplets. The
technology substantiated well the concept of ‘lab-on-a-chip’. Some papers, written by
2010, promised the advent of another revolution [96].
Let us pause for a moment to consider the genome sequencing and its link to mi-
crofluidics. Sequencing illustrates two notions: big numbers and complexity without
valves. In the years 1990-2005, the human genome programme, in full development,
was exclusively based on Sanger sequencing [97]. Massive use was made of restriction
enzymes, which fragment the genome into small single-stranded pieces. This is where
big numbers came into play. Typically, ten fractionations per kilobase were needed
to determine primary sequences [104]. For the human genome, this represented no
fewer than ten million separations -a huge number-. The community thus invented
new separation technologies that were faster, more integrated, more automatized [58],
 Diversification of the technology and broadening of the applications 19

and more parallelized [98]. Impressive achievements were made. However, the sequenc-
ing community did not use this work. The most popular sequencer, ABI Prism 3700,
which, historically, sequenced the human genome, used a 96 parallel capillary configu-
ration [104]. There was not a single microfluidic chip in the machine. The penetration of
microfluidics came later, with the advent of Next Generation Sequencing (NGS). The
first NGS machine, the GS20, appeared in 2005 [104], and it could decipher 20 million
bases in 5.5 hours. The success of GS20 was due, in part, to the use of large microflu-
idic assays. Years later, methods, such as nanopore sequencing (Oxfore Nanopore) or
sequencing by synthesis (Illumina) also made full use of micro/nanofluidic technology.
This technology has clearly become core to genome sequencing.

These accomplishments demonstrated that a microfluidic device can perform millions

of different tasks without using a single valve. How to do that? By relying on molec-
ular actuators: enzymes (polymerases, recombinase, reverse transcriptase, helicase,
etc.), molecular motors, selective affinities (such as antibodies), and self assembly
processes (which spontaneously pattern molecular assemblies). In NGS microfluidic
devices, these processes, working synergistically, allowed extremely complex function-
alities to be developed. The same ideas will be used, later, in other fields of application,
such as the single cell.

Paper microfluidics came later, in the years 2005-07 [106]. The idea was to replace
standard microfluidic substrates (glass, silicon, plastics, or PDMS) with paper, much
cheaper (its cost is a few hundredths of a cent per sheet), and easier to source. Be-
ing burnable, it reduces the risk of contamination. This material is thus particularly
suitable for diagnostic applications in developing countries. Paper had been used for
decades for diagnostics, but the novelty brought by G. Whitesides’ group was paper
patterning [107]. Paper microfluidics appeared as a new branch of microfluidics, hold-
ing potential in several areas, while using a cheaper and more available substrate. We
will consider this further in Chapters 4 and 7.

1.19: Multiplexed analysis of a

model sample of urine (glucose so-
lution) [108]. The white is the pa-
per, the grey is the wax, defin-
ing channels. Two colorimetric de-
tections are performed in duplicate.
(Reprinted from Ref. [108], with
permission from American Chemical
Society, copyright 2022)

Later, in the period 2008-21, many inventions and discoveries were made. Thousands of
studies were published every year (today, about 5000). Obviously, reviewing them lies
beyond the scope of the book. We may nonetheless mention the appearance of new
technologies (3D printing, continuous flow lithography, Northland optical adhesive
20 Introduction

technology (NOA), hydrogel caging, etc.), raising the microfluidic tool box to a level
where a considerable number of microfabrication problems can be solved.
Special mention should be made to 3D printing, which has already impacted the way
in which microfluidic devices are built. This will be discussed in Chapter 7, but an
example is shown in Fig. 1.20.

1.20: An example of 3D microprinting.

Courtesy of Microlight3D, printed with
SmartPrint UV [109].

Another technology worth mentioning is organ on a chip (OOC). OOC creates systems
that mimick organs, in the hope of performing drug studies on representative models.
It has the potential to impact the pharmaceutical industry, by refining analysis, pro-
ducing large amounts of pertinent data, speeding up trials, accelerating early phases
of regulation processes, and, importantly, suppressing the use of animals. OOC was
the missing link between in vitro and in vivo drug trials. An early example of OOC,
mimicking lungs, is shown in Fig. 1.21, and was reported in 2010 [111].

(b)
Air Capillaries

Endothelium Membrane

Pip
Side chambers Diaphragm

Fig. 1.21: (A) Microfluidic model of a lung [111].(B) Real lung

In Fig. 1.21(A), a membrane, coated with epithelial cells, separates two microchannels:
one is for air circulation and the other filled with a liquid. This structure attempts
to mimick the real lung, shown in Fig. 1.21 (B). The gap is large, but the hope is
that, over the years, it will progressively shrink, reaching levels where drug testing
on animals becomes less representative than OOCs. Other OOCs, realized in the last
decade are intestines, kidneys, hearts, and vascularization networks [110].
 MicroReaction Technology (MRT) 21

In the meantime, fundamental questions such as slippage were clarified, albeit not
solved. After fifteen years of discussion, the nanobubble mystery was resolved in 2015
[85]. Different manners of handling fluids were proposed, sometimes counter-intuitively
(such as working at moderately high Reynolds numbers to reorder particles conveyed
by a flow, giving rise to ‘inertial microfluidics’), and progress in the ability to create
complex functionalities has been made (such as handling cells in droplets, barcoding
and processing them). The examples given are far from exhaustive.
Looking back on thirty years of microfluidics, it appears that most of the founding
microfluidic discoveries have been made in a surprisingly short period of time: be-
tween, roughly, 1998 and 2005. One could thus call this period the ‘golden years’ of
microfluidics.

1.7 MicroReaction Technology (MRT)

Microreaction technology (MRT) is part of a general approach called process intensi-

fication. The idea is that, as will be seen in Chapter 2, miniaturization enhances heat
exchange and facilitates temperature control, which in turn increases the specificity
of the reaction, by avoiding the formation of unwanted chemical species [112]. MRT
focusses on the millimetric and submillimetric scales, where better flow control can
be achieved, thanks to the absence of turbulence. Let us recall that turbulence mixes,
but also creates uncontrolled flow heterogeneities: for instance random eddies in the
corners, or behind the blades of a mixer. These eddies induce, in turn, heterogeneities
in temperature, concentration, and, eventually, product characteristics. The key point
of MRT is to operate below turbulence onset to avoid these phenomena21 and, in the
meantime, achieve throughputs of industrial relevance.
The first conference of MRT, international conference on microreaction technology
(IMRET), took place in 1997, three years after μTAS 1994. Since then, IMRET has
met annually.
From a technological viewpoint, because of high temperatures, and often high pres-
sures, MRT uses hard materials, such as slilicon, glass, ceramics, or metal. High
throughput and resistance to clogging impose larger microchannels than in microflu-
idics, typically in the milimetre range. An example of a MRT device, dedicated to the
chemical synthesis of carbamates, is shown in Fig. 1.22 [113].
With MRT, all kinds of synthesis have been performed over the years, starting with
quantum dot synthesis, which is better controlled with MRT than traditional tech-
niques [114]. Other applications are combinatorial chemistry and high throughput
screening.

21 The turbulence onset strongly depends on the flow geometry and is characterized by a Reynolds
number (see Chapter 3). Because of hysteresis phenomena, turbulence thresholds are difficult to define.
In the microfluidic literature, values ranging between 1000 and 3000 are taken without discussion.
This range is appropriate for most of the geometries of practical interest, such as straight pipes or
channels but does not apply for other geometries, such as converging or diverging flows.
22 Introduction

1.22: Multi-step microfluidic chemical

synthesis of carbamates starting from
aqueous azide and organic acid chlo-
ride using the Curtius rearrangement
reaction. The scheme involves three re-
action steps and two separation steps.
Reprinted from Ref. [113], with permis-
sion of John Wiley and Sons, copyright
2022.

By working in small rather than large microreactors, efficiency is increased and plant
sizes can be reduced. Concerning the throughput, parallelization allows volumes of
industrial interest to be reached. In this approach, safety and reliability are improved.
The first attempts to substantiate these ideas were made in the years 2005-10. An
example is shown in Fig. 1.23. Today, the global MRT market is around US$1bn [116].

1.23: Is it possible to miniatur-

ize a plant ? An example of a mil-
lifluidic plant (Reprinted from
[115], with permission of Wiley
and Sons, copyright 2022).

1.8 Nanofluidics

As will be discussed in Chapter 3, nanofluidics22 concerns the 1-100 nm range. Similarly

to microfluidics, nanofluidics can be defined as the science of manipulation of fluids at
the nanoscale.
Interestingly, humans have manipulated fluids at that scale long before clean rooms
were used. For instance, the study of Newton black soap films, 4 nm thick, traces
back to 1877 [120]. Ultrafiltration membranes designed by engineers have pore sizes

22 The word ‘nanofluidics’ appeared, for the first time, in a review written by H. Craighead [117] in
2000
 Nanofluidics 23

below 10 nm. Equilibrium films, created by A. Sheludko [121] in the 1960s, for mea-
suring disjoining pressures, had nanometric thicknesses [121, 122]. A landmark in the
domain was the invention of the force machine [123], in which fluids were confined in
nanometric (open) chambers for studying liquid behavior at the nanometric scale (a
detailed description of the instrument will be given in Chapter 2). Other examples can
be found in the review by J. Eijkel and A. Van den Berg [126].

In the 1990s, it was possible, with e-beam or optical lithography, to create nanochan-
nels less than 100 nm high, or holes less than 100 nm in diameter. In this context, sev-
eral contributions were made,23 such as the realization of a silicon membrane with 50
nm pore sizes in 1990 [118] and a 88 nm nanofluidic field-effect transistor in 1992 [119].
Later, in 1999, a nanofluidic device, built with the intent to develop a chromato-
graphic functionality, based on an original principle, was published by H. Craighead
et al [127].24 It is shown in Fig. 1.24.

Fig. 1.24: (A) Cell used in Ref. [127] (Reprinted from the preceding reference, with permission
of American Physical Society, courtesy of H. Craighead, copyright 2022.) (B) Numerically
designed single-walled nanotube 1.34 nm long with a diameter of 0.81 nm , solvated in a
water reservoir and simulated for 66 ns. Despite its strongly hydrophobic character, the
initially empty central channel of the nanotube is rapidly filled by water from the surrounding
reservoir, and remains occupied by about five water molecules during the entire simulation,
as shown in the figure [129].

The system was composed of a series of channels of different depths. Large’ channels
were 1 μm deep, and shallow ones had a depth of 90 nm, i.e. smaller than the gyration
radius of λ phage DNAs used in the experiments. Under the action of an applied
electric field, DNA strands periodically escape the chambers to migrate through the
shallow channels. In this process, they stretch out. Arguments based on entropy show
that the probability of escape is an exponential function of the DNA length. This was
the efficient size separation mechanism that the authors were seeking. It turned out
that the amplitude of the effect was much smaller than expected, for reasons explained

23 I am indebted to Jan Eijkel for bringing to my attention a number of contributions, along with
sharing his vision on the history of nanofluidics, consistent with that exposed here.
24 In the description we give, one may mention two studies, in which nanofluidic transport phenom-
ena were studied: DNA separation in a nanopillar array in 1998 [124], and mixing in a flow focusing
geometry in the same year [125]. These studies did not create nanochannels. They used microfluidic
systems to reach, through subtle physical mechanisms, the range of nanoscales
24 Introduction

by them.25 Later, in 2000, the system was modified and efficient DNA separation could
be demonstrated [128].
In about the same period (2001), Hummer et al. [129] simulated a single-walled carbon
nanotube (CNT), 1.34 nm long, 0.81 nm in internal diameter which was solvated in a
water reservoir and tracked for 66 ns. Despite its strongly hydrophobic character, the
nanotube was rapidly filled by water.26 We learned that water molecules circulated in
the CNT in single file, developing pulses with peaks of about 30 water molecules per
nanosecond.27 Soon after, CNTs were shown (numerically), to develop large slippages.
[132] The effect was directly measured twelve years later [134]. The phenomenon is
still not understood today.
The two studies, combined with other experimental [118, 119] (already mentioned)
and numerical [133,135,136] contributions, prompted the development of nanofluidics.
Soon after, nanofluidics swiftly expanded. An extremely short list of examples, far
from exhaustive, includes the observation of polarization concentration induced by
the overlapping of Debye layers [137], the invention of nanofluidic diodes and bipolar
transistors [138], DNA separation through arrays of nanopillars [139], and negative
capillary pressure studies [143]. Fundamental studies dedicated to slippage were also
performed, using surface force apparatus (SFA), nano or micro particle image ve-
locimetry (μ PIV), coupled or not to total internal reflection fluorescence (TIRF) or
pressure measurements in nanochannels and, more recently measurements through
single CNTs [134]. Of particular note is membranes made with aligned CNTs, realized
in 2004 [140] (see Fig. 1.25). By offering large permeabilities [141, 142]28 and high se-
lectivities, due to the extremely small pore sizes, they opened a new area in the field
of membrane technology.

(a) (b)

50 m

Fig. 1.25: (A) A sketch of a CNT-based membrane structure. A polymer (PS) is embedded
between the CNTs; the pore size is their inner-tube diameter [141,142]. (B) Scanning electron
microscope (SEM) image of a vertically aligned array of CNTs produced using a Fe-catalysed
chemical vapor deposition (CVD) process [141]

25 The reason was [127]: ‘Bigger molecules escape faster simply because more monomers are facing
the thin slit, and are able to form a beachhead for escape’. This effect still led to a dependence of the
DNA mobility on the chain length, but it was much less strong.
26 The phenomenon was analyzed in [130]. It is attributed to liquid structuring effects, which reverses
the effective wettability of the tube.
27 Years later, other ’subcontinuum’ flow regimes will be discovered [131].
28 The permeability for the smallest CNTs (1 nm in inner diameter), is four orders of magnitudes
larger than no-slip hydrodynamics predicts
 Nanofluidics 25

In the last few years, a new era came with the advent of novel 2D and 3D materials:
functionalized CNTs, graphene, boron nitride (h-BN) and molybdenum disulphide
(MoS2) [149–151].

New phenomena were observed in ultra-confined systems (i.e., typically, of sub-2 nm

sizes): large slippages (already mentioned), strong decrease of water dielectric permit-
tivity (dead water’), non linear transport along with unexpected large shifts in the
freezing transitions [152]. These phenomena may call for a quantum description [149].
If confirmed, this would represent a major conceptual breakthrough. More strongly
than in the past, progress made in the domain showed that nanofluidics is not a mere
extension of microfluidics, but a distinct field, built on original physical effects and
using specific materials and techniques of investigation. The period also inspired new
applications, including energy harvesting, high flux membranes, water desalinization,
and nanoscale flow sensors.

Performing nanofluidic experiments is challenging, because several forces of various

origins, such as electrostatic, van der Waals, entropic or hydrophobic, coupled to the
effect of the confinement, come into play together and are uneasy to disentangle [148].
Therefore, when an unexpected phenomenon shows up, it is often difficult to determine
its origin. This may explain, for example, why it took fifteen years to understand the
origin of nanobubble stability (see Chapter 5). In addition, observation needs sophis-
ticated equipment and the dynamics, often in the nanosecond range, is challenging to
track. In this context, it was difficult to envisage the industrialization of a nanofluidic
product. This was nonetheless done by Pacific Bio, with zeroth mode cavities, and
Oxford Nanopore, with arrrays of nanoholes, both dedicated to sequencing DNA at
high speeds and large throughputs (see Fig. 1.26.).

Fig. 1.26: Oxford Nanopore technology. The purified sample is introduced in a chamber con-
taining arrays of nanopores. DNA strands, mobilized by an electric field, cross the arrays and
deliver molecular information on the sequence that they contain. The error rate is larger than
that of Illumina, but the compacity of Oxford Nanopore product is remarkable. (Courtesy of
Oxford Nanopore.)

The Oxford Nanopore system, shown in Fig. 1.26, includes a plate, pierced with
nanoholes, across which an electric field is applied. With the help of enzymes, single-
strand DNA (ssDNA) pass through the hole and each nucleotide is read, thanks to its
26 Introduction

electrical signature.

This section mentions a small fraction of the important work done, in nanofluidics,
since its birth, i.e. by the turn of the century. The reader may refer to reviews [126,144–
149] for detailed information. It should be noted that nanofluidics is part of a broader
world which includes, for example drug delivery, colloid science, membrane technology,
and nanochemistry. These domains, connected to nanofluidics provide opportunities
for developing fruitful synergies.

1.9 The microfluidic market

1.9.1 How to estimate the microfluidic market ?

The delicate task of defining microfluidic products. The microfluidic market inte-
grates microfluidic devices (raw chips), microfluidic products (cartridges with reagents
and packagings), and microfluidic instruments. Analysts often distinguish between two
segments: instruments and products. We will focus on the second one, which is by far
the most important.

How declare that a product pertains or not to the microfluidic market? Let us take the
example of dried blood spots (DBS), which store newborn blood samples. Collection is
achieved by imbibition and storage by drying. From a physical perspective, imbibition
can be viewed as a microfluidic process, because it drives flows, in a controlled man-
ner, through submillimetric pores. Lateral flow tests, such as pregnancy tests, urine
dipsticks, and glucose monitoring sensors [162], whose markets are several tens of bil-
lions of $, are in a similar situation. Should they be considered microfluidic products?
The answer is no. Analysts require a minimal level of technological functionalities, in
particular regarding fluid management, in order to decide whether a product pertains
or not to the microfluidic market. There is a blurred zone around this requirement,
which explains why market estimates may vary from one agency to another. Still, the
figures are, within 30%, consistent. Fig. 1.27 illustrates the discussion, for the case of
the diagnostic market. The figure is adapted from a recent report published by Yole
Development.

The diagnostic market gathers in vitro systems enabling the detection of contagious
diseases. It includes two types of products: molecular and immunoassay. Each category
is in turn divided into centralized (high throughput) and decentralized (point of care
(POC) systems). The centralized segment includes large systems, plates, tubes, cen-
trifuge equipment, etc. The machines manipulate μlitre volumes and they use standard
assays, which do not incorporate any microfluidic component. It is logical to exclude
them from the microfluidic market. By contrast, Cepheid products manipulate fluids
in microfluidic devices. The cartridges thereby pertain to the microfluidic market. Im-
munassays are excluded, for the reasons discussed above. All this leads to Fig. 1.27,
in which selected products are enclosed in oblong boxes.
 The microfluidic market 27

Fig. 1.27: Products of the diagnostic industry, declared belonging (inside oblongs) or not
belonging (outside the oblongs) in the microfluidic market. (Adapted from a recent report
from Yole Development.)

The case of inkjet printing. Ink-jet printers, for paper, are truly microfluidic systems:
droplets emitted by printers are typically 50 μm in diameter. Fig. 1.28 shows the
dynamics leading to the formation of droplets, on their way making contact with the
paper sheet. Paper inkjets exploit MEMS technology, in both versions (bubble or piezo-
injectors [154]) and manipulate microjets with high precision, producing monodisperse
droplets, under high throughput conditions. In fact, ink-jet printers represents one the
most remarkable achievement of microfluidics. An example is shown in Fig. 1.28.
According to several agencies (Mordor Intelligence, Future Market Insight) (FMI),
the overall 2021 market of ink-jet printers is on the order of US$50bn. This is more
than twice the rest of the microfluidic market, as will be seen below. Although it
could be legitimate to add them up for establishing a ‘total’ microfluidic market, it
is conventional to treat the ink-jet printing market separetly. The microfluidic market
that we present here therefore excludes ink-jet printers.

1.9.2 Today’s market

As mentioned above, before 2000, the microfluidic market (i.e. without ink-jet print-
ing) was insignificant and scepticism was floating around concerning the potential of
the technology to find its feet in a market, even though it was regularly announced,
here and there, that microfluidics would revolutionarize the twenty-first century [158].
Common sense led to the belief, in fact wrong, that driving flows through tiny channels,
at the industrial scale, without leaks, clogging, bubbles, nor uncontrolled adsorption,
was impossible. The opposite viewpoint - believing that it is straightforward to create
28 Introduction

(a) (b)
Piezo
tranducer

Nozzle

Droplets

Fig. 1.28: (A) Piezo acoustic inkjet printer of Ref. [156]; (B) Time series of jetted ink droplets,
stroboscopically recorded with single-flash photography. (Left to right) Multiple images of
single droplets with a delay of 3 μs between the individual droplets. Here the opening radius
of the nozzle is 15 μm and the diameter of the droplet 23 μm, which corresponds to a droplet
volume of 11 pL. The final velocity of the droplet is 4 m/s. The figure illustrates the imaging
quality and the absence of motion blur due to the use of the 8-ns iLIF (illumination by
laser-induced fluorescence) technique [155–157].
,

a complex, functional, microfluidic device - was unrealistic. To add to uncertainties,

the remark, occasionally made by prominent researchers, that microfluidic devices are
small, but, with the tubings and the pumps needed to operate the system, they more
often resemble a ‘chip in the lab’ rather than a ‘lab on a chip’, casted doubts about
the possibility of creating microfluidic devices of industrial relevance.
Despite all this, a number of successful microfluidic products emerged. Technical dif-
ficulties could be solved, circumvented, or just found irrelevant. Impressive industrial
challenges, previously considered out of reach, could be met, such as driving ssDNA
through nanoholes and collect genomic information from it. Finally, over the last two
decades, the microfluidic market steadily increased at a two-digit rate. Today, hun-
dreds of millions of devices are sold every year. For example, 400 million Illumina
microfluidic flow cells are shipped every year.
Let us analyse three examples, which, today, generate revenues between US$200mn–US$3bn
(see Fig. 1.29).
Quidel Triage In Fig. 1.29, the device is a microfluidic adaptation of the Elisa
test: the blood sample, in which targeted antigens are fluorescently labelled, is
driven, by capillarity, in a straight microchannel, with antibodies immobilized on
the walls. As in Elisa tests, antigens are captured by the antibodies, and, after
rinsing, fluorescence emission is analysed. With this test, levels of troponin, whose
elevation indicates heart muscle damage, can be measured, with a few nanograms
per litre sensitivity. This is a major application. Sales are on the order of US$200
mn.
Ilumina New Generation Sequencing (NGS) technology, developed by Illumina, is
based, as all NGS methods, on microfluidic technology. The cartridge contains
a patterned cell, with wells, called a flow cell, fabricated over a Complementary
 The microfluidic market 29

Fig. 1.29: Three examples of types of products, based on microfluidics, generating, in 2020,
turn overs in the range US$200mn–US$3bn: (A) Triage for cardiac control (Quidel); (B) DNA
sequencer (illumina HiSeq 2500); (C) Kindle (e-ink technology)
,

metal–oxide–semiconductor (CMOS) chip. In the cell, each well faces a photodi-

ode. In a nutshell, sequencing is performed by hybridizing sampled ssDNA on an-
chored probes, and monitoring, in real time, the light emitted during the process.
Microfluidics plays an instrumental role in this domain. Without microfluidics, no
NGS based sequencing could exist.29 Illumina revenues generated by the prod-
uct reached US$3.2bn in 2020. As noted above, the company ships 400 million
flow-cells each year.

e-ink The Pocketbook Inkpad Lite is based on microfluidic droplets (the technology
will be described in Chapter 4). Its price ranges between $200 -$300. The display
technology, commercialized by E-ink, generated US$530mn in 2021.

Who are the microfluidic players? Fig. 1.30 organizes them in three groups, each
corresponding to a range of revenues (the figure is adapted from Yole-Development).

Market estimates are given by several agencies: Yole Development, Market and Mar-
kets, Grand View Research, uFluidics, Research & Market, and BCC Research. On
average, the microfluidic market is estimated around 17 US$bn with a compound an-
nual growth rate (CAGR) around 15 - 20%. Should 20% be accurate, the market would
reach 100 B$ in 2030. A plot showing the evolution of the market over the years, again
made by Yole Development, is shown in Fig. 1.31.

Fig. 1.31 moreover provides information on the market segments. In decreasing order
(in terms of volume), we have: point of care, tools for pharmateutical and research,
clinical diagnostics, industrial diagnostics, optical actuation (liquid lenses), and man-
ufacturing.

29 NGS, compared to traditional sequencing, based on Sanger’s method, allowed to reduce sequenc-
ing prices along with increase speeds by orders of magnitude (see Chapter 5)
30 Introduction

Fig. 1.30: Microfluidic players organized in three groups, each corresponding to a range of
revenues [153]. (Adapted from Yole Development.)

Fig. 1.31: Microfluidic market and its evolution (Courtesy of Yole Development) [153]

1.10 Future of microfluidics

1.10.1 Microfluidics and Pasteur and Bohr quadrants

In 1997, D.Stokes proposed the ‘quadrant model [163, 164] (see Fig. 1.32). This model
classifies research activity into four quadrants. Three of them are defined in the fol-
lowing manner:
Bohr quadrant -pure basic research : N. Bohr’s efforts were entirely dedicated to
 Future of microfluidics 31

answer fundamental questions of physics. In 1913, he explained the origin of the

spectral lines of hydrogen, by representing the atom as a nucleus orbited by an
electron [159]. The Bohr atom model, which contributed to laying the foundations
of quantum theory, was conceived under the sole momentum of intellectual cu-
riosity, in a manner disconnected from applications. One paradox is that quantum
physics led to the invention of the transistor, and further to computing science,
internet -all inventions that, without exaggeration, revolutionized the world-.

Edison quadrant -applied research : Edison was a prolific inventor, with more
than thousand patents. Examples are the phonograph and the motion picture
camera. The electrical bulb lamp was not invented by him, but he managed to
increase its lifetime considerably, by using carbon filaments. Edison’s approach
was empirical, with no serious attempt to understand the physics involved in his
inventions.

Pasteur quadrant -use-inspired basic research : Pasteur discovered microbiol-

ogy by working on diseases of beer and wine, not understood at that time. His
discovery of the vaccine for rabies, made in 1885, was stirred up by the bite of
a 9 year old child by a rabid dog. Other examples, found when studying his life,
illustrate the synergy between societal/industrial inspiration and fundamental
discoveries.

The unamed fourth quadrant represents research motivated neither by intellectual

curiosity nor by applications. It is difficult to comment on this quadrant.

Pure basic Use-inspired

research basic research
High
Quest for fundamental

BOHR QUADRANT PASTEUR QUADRANT

understanding?

Applied
research

Low

EDISON QUADRANT 1.32: The four sectors of research:

Low High Bohr, Pasteur, Edison, and unnamed
Consideration of use? [163].

Regarding microfluidics, Pasteur’s quadrant is certainly the most relevant: microflu-

idics offers responses to industrial and societal needs, and, in the meantime, allows us
to address basic questions. Since the early days of microfluidics, the community has
worked mostly in the Pasteur quadrant.

However, it pays frequent visits to the Bohr quadrant. Often, along the lines of Feyn-
man’s talk, discoveries emerged from the excitation of realizing something that has
never been done before, or from efforts to understand strange phenomena occurring
at the microscale. For example, a precursor work done on microfluidic droplets was
32 Introduction

motivated solely by curiosity [165–167]. The objective of Ref. [165, 166], was to re-
port the outstanding monodispersivity of droplets produced in microjets, and that of
Ref. [167],was to describe droplet patterns observed in microchannels. No application
was mentioned. Later, it was realized that droplets could play an important role in
colloidal science, chemistry, material science, and biology. Another example is PDMS
soft technology, whose precursor studies were mostly motivated by technological chal-
lenges [168]. More striking is the case of nanofluidics, regarding which some of the
investigators, around the turn of the twenty-first century, had no idea about the ap-
plications that the technology could generate. Later, nanofluidics will become a core
technology for new genome sequencing. Today, the questions raised in this domain,
concerning, for example, the hydrodynamical behavior of carbon nanotubes, clearly
pertain to the Bohr quadrant. Other examples can be mentioned (nanobubbles, inertial
microfluidics), where research is mainly curiosity driven.

1.10.2 Promising areas of application

The advantages of microfluidics, put forward multiple times over the years, are as
follows: small quantities of reagents, small samples, high speeds of analysis, cost re-
duction, high sensitivity of separation, portability, exquisite flow control, functionality
integration, automatization, parallelisation, high level of compartmentalization, and
large information throughput. Disadvantages also exist (cleaning constraints for de-
vice fabrication, difficulty of producing large volumes of product, no standardization,
difficulty of scaling up for PDMS, and, for complex constructions, reliability below
industrial standards), but they are outweighted by the advantages offered by the tech-
nology.

The general vision is that, with such capabilities, microfluidics can resolve bottlenecks
hampering the development, sometimes the existence, of key industrial products (for
instance, NGS), or hindering the realization of industrial or research ambitions (for
instance, performing in-situ analyses on Mars, retrieving information stored in DNA
memories, or building representative models of the kidney). This is what microfluidics
has successfully done over three decades. Concerning the forthcoming years, since fluid
control at the submillimetric scale is increasingly required in a number of fields, and
many domains do not benefit yet from microfluidic technology, or even have no aware-
ness of it, conditions are met for the microfluidic field to keep growing at substantial
rate. This linear vision will be disturbed or disrupted by the emergence of new dis-
coveries in microfluidics or in related fields. Examples include the discovery of new
enzymes, that could simplify and reduce the cost of biological workflows, and the syn-
thesis of new materials, which could impact microfabrication techniques, as PDMS
did in the past. In addition, the priorities of the society, which support a major part
of microfluidic activity, may change. Yesterday, a priority was given to the human
genome programme. Today, energy production and the environment, are subjects of
major importance. All the problems of predicting the future lie in the unpredictable
advent of major discoveries and the extreme difficulty to visualize the evolution of
societal needs.
 Future of microfluidics 33

During the years 2000-05, institutions attempted to sketch the future. They produced
‘roadmaps’, or white papers’ for microfluidics, to structure the young field and to
reassure or stimulate funding agencies and investors, as Moore’s law did for microelec-
tronics. Today, on re-reading these roadmaps, the general impression is that they were
more concerned with platforms than with the future of the field [169]. This probably
shows how difficult it is to predict the future. The scientific literature was more per-
tinent. Reviews offering a vision of the future of microfluidics, or subfields pertaining
to it (see for instance [1, 74, 160, 223]), often pointed in the right directions.30
Here, we will look into the crystal ball by listing, without claim for completeness,
a number of promising areas of application of microfluidics, in a manner similar to
Ref. [161]. This is set out in Table 1.1.

Application Example
Diagnostics Perform low-cost SARS COV2 molecular testing
with paper microfluidics
Single cell Establish the transcriptome, at the single cell level,
of tumour cancer cells
Screening Discover antibiotics and enzymes by screening
large quantities of biochemical reactions
Delivery Produce lipid nanoparticles for nucleic acid delivery
(for example, for RNA-based vaccines)
Organ on a chip Create artificial liver, lung or intestine to test drugs
MRT Produce high-quality materials, benefiting from
high throughput and flow control
Cell culturing Produce, with high throughput and low cost, stem cells of high quality
Material Science Materials with novel functionalities
Membrane Science Membranes for energy harvesting or energy production
Cosmetics Cosmetic products and instrumentation
DNA storage Retrieve information stored in DNA matrices
Display technology Display images with microfluidic entities, such as droplets

Table 1.1 : Non-exhaustive list of promising areas of application for microfluidics and
nanofluidics

The table deserves few comments

Diagnostics: Today, therapeutic drugs are making rapid progress but common sense
says that, in order to cure a disease, one must first identify it. This is where di-
agnostic comes on stage. Tests pertain to two categories: antigen and molecular.
Antigen tests have limited sensitivities, but their costs are low and their usage
simple. To-day, molecular tests are mostly based on polymerase chain reaction
(PCR). In order to reach high sensitivities, these tests perform an extraction be-
fore amplifying the nucleic acids. The technology is thus more sensitive but more
30 It is amazing to note that, in some cases, authors did not predict the advent of technologies that,
later on, they invented [1].
34 Introduction

complicated and more costly than antigen tests. Since 2000, simpler methods of
amplification have been developed. An example is loop amplification (LAMP),
invented in 2000 [171], which operates under isothermal conditions, much simpler
to implement. It it is thus regrettable that PCR keeps dominating the molecular
market. It has been shown that paper microfluidics, described in Chapters 4 and 7,
and based on LAMP, performs molecular diagnostics, with integrated membrane
purification, with sensitivities and specificities comparable to those of PCR, much
lower cost, and with a minimal logistics [170,172–174]. There is no doubt, that, in
the future, paper microfluidics will represent an important technology for molec-
ular point of care (POC) diagnostics. At the moment, the problematic is linked
to business, not to technology.
Delivery: Delivery is about creation of particles, to be swallowed, inhaled, injected
into the blood circulation, or injected intracutaneously. These particles circu-
late or diffuse in the body and then, at the end of their journeys, deliver their
load, with a controlled kinetics. Words often used are ‘targeted drug delivery’
or ‘controlled drug release’. Delivery primarily concerns human health but also
agriculture: for example, pheromones are encapsulated and delivered at certain
periods of the year, in order to disrupt insect reproduction. Delivery methods are
diverse: physico-chemical (for instance hydrogel dissolution in tissues), acoustic
(bubble or droplet bursting) [175], thermal, or chemical. Delivery kinetics ranges
from fraction of seconds to months, and the sizes of the particle range from 100
nm to tens of micrometres. There is a natural coupling between delivery and
microfluidics [176]. Microfluidics allows, by reducing polydispersivity, improving
encapsulaton rates, suppressing of burst phenomena (too much delivery in a short
time, which can be lethal) and, more generally, optimizing delivery kinetics.
Let us pause for a moment to consider lipid nanoparticles (LNPs) (see Fig.
1.33(A)). We will see, in Chapter 5, that these objects gain considerably being
produced in microfluidic or millifluidic devices. LNPs are complex nanoparticles,
50 to 100 nm in diameter, formed through a self-assembly process that includes,
according to Ref. [178], three steps: discoidal cluster formation, aggregation of
clusters into larger patches, and vesicle formation.31 LNPs travel in the blood
circulation, attach to a cell thanks to their affinity, and penetrate into it. Hav-
ing penetrated the cell, LNPs deliver their passengers, designed to interact with
the cell machinery (see Fig. 1.33(B)). Molecules carried by LNPs can be DNA,
mRNA or siRNA. LNPs have been highlighted during the COVID-19 crisis. They
vectorized mRNA into cells and made possible mRNA-based vaccination [177].
There is little doubt that the domain of delivery, coupled to microfluidic technol-
ogy, will expand in the future.
Organ on a chip (OOC): As noted above, OOCs are biomimetic objects enabling
drug testing on representative models. One outcome of the technology is to ac-

31 LNP formulation includes, to control stability, electrostatic charge, and pH, reagents such as 1,2-
dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene
glycol-2000 (DMG-PEG2000), Citric acid, sodium citrate tribasic dehydrate and cholesterol