REVIEWS OF MODERN PHYSICS, VOLUME 77, JULY 2005

Microfluidics: Fluid physics at the nanoliter scale

Todd M. Squires*
Departments of Physics and Applied & Computational Mathematics, California Institute of
Technology, Pasadena, California 91125, USA

Stephen R. Quake†
Departments of Applied Physics and Physics, California Institute of Technology,
Pasadena, California 91125, USA
共Published 6 October 2005兲

Microfabricated integrated circuits revolutionized computation by vastly reducing the space, labor,
and time required for calculations. Microfluidic systems hold similar promise for the large-scale
automation of chemistry and biology, suggesting the possibility of numerous experiments performed
rapidly and in parallel, while consuming little reagent. While it is too early to tell whether such a vision
will be realized, significant progress has been achieved, and various applications of significant scientific
and practical interest have been developed. Here a review of the physics of small volumes 共nanoliters兲
of fluids is presented, as parametrized by a series of dimensionless numbers expressing the relative
importance of various physical phenomena. Specifically, this review explores the Reynolds number
Re, addressing inertial effects; the Péclet number Pe, which concerns convective and diffusive
transport; the capillary number Ca expressing the importance of interfacial tension; the Deborah,
Weissenberg, and elasticity numbers De, Wi, and El, describing elastic effects due to deformable
microstructural elements like polymers; the Grashof and Rayleigh numbers Gr and Ra, describing
density-driven flows; and the Knudsen number, describing the importance of noncontinuum molecular
effects. Furthermore, the long-range nature of viscous flows and the small device dimensions inherent
in microfluidics mean that the influence of boundaries is typically significant. A variety of strategies
have been developed to manipulate fluids by exploiting boundary effects; among these are
electrokinetic effects, acoustic streaming, and fluid-structure interactions. The goal is to describe the
physics behind the rich variety of fluid phenomena occurring on the nanoliter scale using simple
scaling arguments, with the hopes of developing an intuitive sense for this occasionally
counterintuitive world.

CONTENTS 2. Fluid control using patterned surfaces 990

3. Fluid manipulation with capillary forces 991
a. Solid-liquid interfacial energy manipulation 991
I. Introduction 978
b. Surfaces with wettability gradients 992
II. Dimensionless Numbers in Microfluidics 979
i. Thermal gradients 关Fig. 19共a兲兴 992
A. The Reynolds number: Inertial irrelevance 981
ii. Reactive fluids 关Fig. 19共b兲兴 992
B. The Péclet number and the problem of mixing 982
iii. Self-propelling liquid bi-slugs 关Fig. 19共c兲兴 992
1. Sensing, filtering, and fabricating with
iv. Optowetting 关Fig. 19共d兲兴 992
parallel laminar flows 983
v. Electrowetting 993
a. T sensor 983
vi. Dielectrophoresis 994
b. Filtration without membranes 983
c. Liquid-liquid surface tension manipulation 994
c. Fabricating and interrogating using multiple
D. The Weissenberg and Deborah numbers: Let the
laminar streams 984
mountains flow 995
2. Fighting diffusion: Particle separations 985
1. Microfluidic single-polymer studies 998
3. Beating diffusion: When mixing matters 985 2. Elastic instabilities 998
a. Taylor dispersion 986 3. Controlled polymer deformation for
b. Rotary mixer 987 separations 1000
c. Mixing by chaotic advection 988 E. The Grashof and Rayleigh numbers: Tracer-driven
C. The capillary number: Free-surface deformations 989 flows 1001
1. Droplet formation in two-phase flows 989 1. Protein crystallization 1002
2. Polymerase chain reaction 共PCR兲 1004
3. Electrohydrodynamic instabilities 1004
F. The Knudsen number: When molecules matter 1005
*Present address: Department of Chemical Engineering,
1. Molecular effects in gases 1005
University of California, Santa Barbara, CA 93106. Electronic
address: [email protected] 2. Fluid slip over nonwetting surfaces 1006
† III. Driving Flows at the Boundaries 1007
Present address: Department of Bioengineering, Stanford
University, Stanford, CA 94305. Electronic address: A. Electrokinetic effects 1007
[email protected] 1. Electrostatics in ionic solutions 1007

0034-6861/2005/77共3兲/977共50兲/$50.00 977 ©2005 The American Physical Society

978 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

2. Electrokinetics: Electro-osmosis and

electrophoresis 1008
3. Recent variations on classical electrokinetic
techniques 1008
a. Patterned surfaces with nonuniform charge 1009
b. Active control of surface charge 1009
c. Induced-charge electrokinetic phenomena 1009
B. Acoustic streaming 1011
1. Quartz wind 1011
2. Boundary-induced streaming 1012
3. Cavitation microstreaming 1013
C. Fluid-structure interactions 1014
1. Driving flows with soft walls 1014
2. Intelligent walls made of responsive
hydrogel 1015
FIG. 1. 共Color兲 Sophisticated networks of fluidic microchan-
Acknowledgments 1016
nels allow for unprecedented automation and parallelization of
List of Symbols 1016
biological and chemical systems. Pictured is a large-scale, inte-
References 1016 grated chip to measure protein interactions. The colored con-
trol lines manipulate valves that function as gates and pumps,
which drive reagent solutions through the uncolored channels
I. INTRODUCTION
to be stored in circular chambers for subsequent reaction. Cir-
The integrated circuit revolution has allowed large- cular chambers are 250 ␮m diameter. 共Figure courtesy of S.
scale automation of computation, fundamentally alter- Maerkl.兲
ing the way we live and do science. A natural question
to ask is whether it is possible to automate biology and ultimately incorporate some sort of microfluidic plumb-
chemistry to a similar extent, and if so, would the con- ing.
sequences be as dramatic? The industrial approach to One important manner in which microfluidics differs
addressing biological integration has come in the form from microelectronics is that the fundamental physics
of enormous robotic fluidic workstations that take up changes more rapidly as the size scale is decreased. De-
entire laboratories and require considerable expense, spite decades of ever smaller transistors and higher den-
space, and labor. This approach is reminiscent of the sities in semiconductor electronics, the industry has yet
macroscopic approach to circuits consisting of massive to reach the length scale where scaling has caused a
vacuum-tube–based arrays in the early twentieth cen- qualitative change in physical phenomena, e.g., from the
tury. Like the vacuum-tube computing revolution, such classical to the quantum regime. Single electron transis-
biological automation has already been highly produc- tors and other mesoscopic devices have been developed,
tive and has led to whole genome sequencing and yet remain exotic objects of laboratory exploration. By
analysis—a new revolution in biology. Microfluidics, contrast, fluidic systems can rather quickly reach length
which is essentially a field dedicated to miniaturized scales where the fundamental fluid physics changes dra-
plumbing and fluidic manipulation, offers the possibility matically. One of the best-known examples is that mass
of solving outstanding system integration issues for biol- transport in microfluidic devices is generally dominated
ogy and chemistry by allowing automation to proceed to by viscous dissipation, and inertial effects are generally
scales that will rival electronic integrated circuits 共Fig. negligible. Since inertia provides the nonlinearity that is
1兲. responsible for numerous instabilities and for turbulence
Although most current effort in microfluidics con- itself, its loss might seem to render microfluidic flows
cerns devices with applications in chemistry, biology, and uninteresting.
medicine, there are also applications in the physical sci- In fact, we argue that the opposite is true. Microfluidic
ences for control systems and heat management 关e.g., physics is quite rich, drawing from much of classical
Zhang et al. 共2002兲兴, energy generation 关e.g., Ferrigno et physics and chemistry—fluid mechanics, electrostatics,
al. 共2002兲 and Choban et al. 共2004兲兴, and display technol- thermodynamics, statistical mechanics, elasticity, and
ogy 关e.g., Hayes and Feenstra 共2003兲兴. There is a long polymer physics, to name a few. Notably, the small di-
history of using fluidics as control systems, ranging from mensions that quell the inertial nonlinearity bring other
logic devices to thrust reversers in aircraft; this program physical phenomena, less familiar on our macroscale, to
ultimately foundered in part because scaling properties prominence.
of the fluid physics prevented miniaturization 共Hum- Following the invention of the transistor, some con-
phrey and Tarumoto, 1965; Foster and Parker, 1970; cerned themselves primarily with designing and building
Joyce, 1983兲. Liquid crystal displays and ink jet printers circuits to perform various computational tasks, whereas
关e.g., Bassous et al. 共1977兲兴 are ubiquitous consumer others continued to study and explore the science under-
products that can be thought of as microfluidic devices, lying semiconductors and microcircuits. Similarly in mi-
and have had enormous industrial impact. As fuel cells crofluidics some focus primarily on developing microflu-
become more widely deployed, it is likely that they will idic devices and tools during which, of course, scientific

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 979

issues arise and are explored, whereas others focus on thus indicate an important change in physics going from
exploring and understanding the physics of these sys- everyday fluid volumes to the nanoliter scale. In other
tems. cases, the numbers are independent of length scale yet
In this spirit, the focus of this review lies explicitly still have important implications and applications.
with the physics underlying microfluidic devices, empha- We begin our review of each number with an intuitive,
sizing the variety of physical phenomena and the man- paradigmatic model example to illustrate the basic phys-
ner in which they have been exploited. Perspective ics at hand with simple scaling arguments. Following
pieces, both early 共Manz et al., 1993; Ramsey et al., 1995兲 each model example, we discuss various devices and sys-
and more recent 共Ramsey, 1999; Figeys and Pinto, 2000; tems that have been designed to utilize or probe the
Stone and Kim, 2001; Meldrum and Holl, 2002; Hong physics at hand. In the latter stages of the review we
and Quake, 2003兲, describe the promise, progress, and consider another set of phenomena that are characteris-
challenges for the field. “Lab on a chip” or “micro total tic of microfluidic devices—the importance of boundary
analysis systems” 共Manz et al., 1993兲 have recently been conditions, and in particular the dominant role played
reviewed in a pair of companion articles, one focusing by surface effects. Such boundary-driven effects are of
on the history and overall approach 共Reyes et al., 2002兲, particular importance to microfludic devices because the
and the other on standard operations in analytical chem- characteristic geometries of microfluidic devices nearly
istry and their microfluidic applications 共Auroux et al., always have large surface-area-to-volume ratios. Specifi-
2002兲. Reviews of more specific applications for micro- cally, we discuss electrokinetic effects, acoustic stream-
fluidics include protein crystallization 共Hansen and ing, and fluid-structure interactions.
Quake, 2003兲, proteomics 共Dolnik, 1997, 1999; Dolnik The past decade has seen a tremendous effort in mi-
and Hutterer, 2001; Figeys and Pinto, 2001; Hancock et crofluidic research, with many thousands of research ar-
al., 2002; Lion et al., 2003兲, microfluidics based on micro- ticles and several dedicated journals. In reviewing the
mechanical systems 共MEMS兲 共Ho and Tai, 1998; Ver- literature, we have sought to include references from the
poorte and De Rooij, 2003兲, micropumps 共Nguyen et al., archival scientific literature which are, to the best of our
2002; Laser and Santiago, 2004兲, clinical and forensic ap- knowledge, first chronologically or which possess signifi-
plications 共Thormann et al., 2001; Verpoorte, 2002兲, cant pedagogical value with respect to the physics in-
sample pretreatment 共Lichtenberg et al., 2002兲, and mo- volved. There are other useful references which in some
lecular diagnostics 共Landers, 2003; Auroux et al., 2004兲. cases predate the archival publications to be found in
The advantages and applications of soft 共rather than sili- various yearly conference proceedings such as
con兲 microfluidics have been reviewed 共McDonald et al., MicroTAS and Hilton Head Sensors and Actuators
2000; Quake and Scherer, 2000; Whitesides and Stroock, which are not indexed in the Web of Science; for reasons
2001; Whitesides et al., 2001; Ng et al., 2002; Sia and of scope and accessibility they are not included in this
Whitesides, 2003兲. Finally, separation techniques, par- review.
ticularly for nucleic acids, occupy a central role in many Microfluidics is an enormous field, and we have been
microfluidic devices. Specific reviews include DNA sepa- ambitious in terms of the scope of physical phenomena
ration and analysis methods in the microfluidic context treated here. Many, even most, of these phenomena
共Ugaz et al., 2004兲, DNA separation in microfabricated could merit their own review. We have inevitably over-
devices 共Tegenfeldt et al., 2004兲, the theory of polymer looked interesting, important, and deserving work in
separation 共Viovy, 2000; Slater et al., 2002, 2003; Ashton preparing this review, and we apologize in advance for
et al., 2003; Sartori et al., 2003兲, and microchip-based such oversights. This field is as interdisciplinary as any,
polymer electrophoresis 共Bousse et al., 2000; Bruin, and the numerous disciplines involved have and will
2000; Dolnik et al., 2000兲. continue to overlook and reinvent each others’ work.
Rather than duplicate these efforts, the present re- Our hope in this review is to help researchers think in-
view emphasizes the physics of the field over its specific tuitively about the diverse physics of these systems, to
applications, and should complement other physically attempt to bridge the various communities involved, and
minded reviews 共Brody et al., 1996; Beebe et al., 2002; to convey the interest, elegance, and variety of physical
Stone et al., 2004兲. Specifically, we consider fluid physics phenomena that manifest themselves on the nanoliter
that occurs within channels with typically ⬃100 ␮m scale.
transverse dimension—for reference, 1 nl= 共100 ␮m兲3.
We note, however, that the physics explored here is also II. DIMENSIONLESS NUMBERS IN MICROFLUIDICS
relevant for complex fluids and colloids, porous media,
and biological systems, to name a few. A wide variety of physical phenomena occur in micro-
The essential fluid physics of a system is dictated by a fluidic devices, the importance of which must be judged
competition between various phenomena, which is cap- against competing phenomena. Dimensionless numbers
tured by a series of dimensionless numbers expressing expressing the ratio of these phenomena give a sense for
their relative importance. These dimensionless numbers where a system sits in fluidic parameter space. Moti-
form a sort of parameter space for microfluidic physics, vated by this picture, we review this parameter space,
which we set out to explore. We organized this review one dimension at a time, by exploring the dimensionless
around those numbers—in some cases, they scale with a numbers listed in Table I: the Reynolds number Re, re-
characteristic system length such as channel size and lating inertial forces to viscous forces; the Péclet number

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

980 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

TABLE I. Dimensionless numbers used in this review.

Re Reynolds ␳ U 0L 0 inertial/viscous Eq. 共5兲

␩
Pe Péclet U 0L 0 convection/diffusion Eq. 共7兲
D
Ca capillary ␩U0 viscous/interfacial Eq. 共19兲
␥
Wi Weissenberg ␶p␥˙ polymer relaxation time/ shear rate time Eq. 共24兲

De Deborah ␶p polymer relaxation time/ flow time Eq. 共25兲

␶flow
El elasticity ␶ p␩ elastic effects/ inertial effects Eq. 共26兲
␳h2
Gr Grashof ␳ U bL 0 Re for buoyant flow Eq. 共30兲
␩
Ra Rayleigh U bL 0 Pe for buoyant flow Eq. 共29兲
D
Kn Knudsen ␤ slip length/ macroscopic length Eqs. 共36兲 and 共38兲
L0

Pe, relating convection to diffusion; the capillary num- continuum version of F = ma on a per unit volume basis:

冉 冊
ber Ca, relating viscous forces to surface tension; the
Deborah, Weissenberg, and elasticity numbers De, Wi, ⳵u
␳ J + f = − ⵜp + ␩ⵜ2u + f,
+ u·ⵜu = ⵜ·␴ 共1兲
and El, expressing elastic effects; the Grashof and Ray- ⳵t
leigh numbers Gr and Ra, relating transport mecha-
nisms in buoyancy-driven flows; and the Knudsen num- where inertial acceleration terms appear on the left and
ber Kn, relating microscopic to macroscopic length forces on the right. Here f represents body force densi-
scales. Throughout this review, we will make reference ties, which will include many of the effects discussed in
to the standard microchannels shown in Fig. 2. this review. When inertial forces are small compared to
Before beginning, however, we very briefly review ba- viscous forces, which is usually the case in microfluidic
sic fluid properties and the dimensional parameters that devices, the nonlinear term can be neglected, leaving the
characterize them. Fluids are continuum materials, and Stokes equation
so discrete quantities like mass and force give way to
continuous fields like density ␳ and force density f that ⳵u
␳ = ⵜ·␴ + f = − ⵜp + ␩ⵜ2u + f. 共2兲
are defined per unit volume. The concept of a small fluid ⳵t
element is often invoked by analogy with discrete me-
chanics. Forces on such elements arise from fluid In both cases, mass conservation requires
stresses ␴J 共forces per unit area兲 exerted on the element
⳵␳
surfaces, in addition to externally applied body forces f + ⵜ·共␳u兲 = 0, 共3兲
exerted on the bulk of the element. ⳵t
The velocity field for a Newtonian fluid obeys the giving the incompressibility condition ⵜ·u = 0 for slowly
Navier-Stokes equations, which essentially represent the flowing fluids with nearly constant density like water.
Physically, the tensor stress ␴
J consists of normal and
tangential components. Pressure, which is an isotropic
force per unit area pn̂ exerted normal to any surface, is
the most familiar normal stress. Viscous stresses ␴ Jv
⬃ ␩ ⵜ u contain both normal and shear components. Ad-
ditional stresses, both normal and shear, arise in com-
plex fluids with deformable microstructure: polymer so-
FIG. 2. 共Color in online edition兲 Model 共a兲 rectangular and 共b兲 lutions, discussed in Sec. II.D, provide one example.
circular microchannels, through which fluid flows with charac- Interfacial 共capillary兲 stresses ␴c ⬃ ␥ / R, where ␥ is sur-
teristic velocity scale U0. Channel length will be denoted l, face tension and R is surface curvature, are exerted nor-
width 共or radius兲 w, and height 共shortest dimension兲 h. The mal to free fluid surfaces, whereas surface tension gradi-
coordinate z points downstream, y spans the width, and x ents give Marangoni stresses ␴m ⬃ ⵜ␥ that are exerted
spans the height. along the surface.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 981

TABLE II. Physical properties of water at 20 °C and 1 atm

pressure 共Lide, 1999兲.

Density ␳ 1.0 g / cm3

Shear viscosity ␩ 1.0⫻ 10−2 g / cm s
Kinematic viscosity ␯ 1.0⫻ 10−2 cm2 / s
Surface tension 共in air兲 ␥ 73 g / s2
Thermal variation of ␥ d␥ / dT −0.15 g / s2 °C
Thermal diffusivity ␬H 1.4⫻ 10−3 cm2 / s
FIG. 3. 共Color in online edition兲 Inertial forces exerted by
Compressibility ␣␳ −4.6⫻ 10−11 / ba
accelerating fluid elements. 共a兲 A small fluid element rounding
Coefficient of thermal expansion ␣T 2.1⫻ 10−4 / °C a corner loses ␳U0 of momentum during the time ␶0 ⬃ w / U0
required to turn the corner, giving a centrifugal force density
of order fi ⬃ ␳U20 / w. 共b兲 A fluid element flowing through a con-
It is important to remember that fluids are made of traction with velocity gradient du / dz ⬃ U0 / l gains momentum
molecules, and this discreteness reveals itself in the phe- density at a rate fi ⬃ ␳U20 / l. In both cases, the inertial force
nomenon of diffusion. Both the molecules of the fluid exerted on the fluid is equal and opposite to the force required
and any “tracers” or additives move diffusively through to accelerate each fluid element, and is the same for flows in
the fluid. A diffusivity arises for each tracer—the ther- either direction.
mal diffusivity ␬H for heat, the diffusivity D
⬃ kBT / 6␲␩a for spherical solute molecules of size a, and The force exerted on the rest of the fluid is equal and
the kinematic viscosity ␯ = ␩ / ␳ for fluid vorticity. Repre- opposite to the force required to accelerate each fluid
sentative values for these and other fluid properties are element. Thus the force on the fluid due to a curved
given in Table II. streamline points outwards centrifugally, and the inertial
force in an expansion or contraction points towards the
A. The Reynolds number: Inertial irrelevance wide end of the channel, regardless of the flow direction.
An oscillatory flow through an asymmetric channel can
Of all dimensionless numbers, the Reynolds number drive a rectified, steady fluid motion which forms the
共Re兲 is the one most often mentioned in connection with basis for valveless diffuser pumps 共Stemme and Stemme,
microfluidics. Ironically, it may also be the least interest- 1993; Olsson et al., 1999兲.
ing number for microfluidics: after all, almost without In both cases, the magnitude of inertial and viscous
exception, microfluidic devices employ fluids with Rey- force densities must be compared. Viscous force densi-
nolds numbers that are small enough for inertial effects ties result from gradients in viscous stress, and thus scale
to be irrelevant. Nevertheless, because low Reynolds as fv ⬃ ␩U0 / L20, where L0 is a typical length scale. The
number flows contradict day-to-day human experience ratio of these two force densities,
with fluids, one must unlearn a lifetime’s worth of
high-Re intuition in order to effectively think about mi- f i ␳ U 0L 0
crofluidics. Purcell’s description of this counterintuitive = ⬅ Re, 共5兲
fv ␩
world, Life at low Reynolds number 共Purcell, 1977兲 pro-
vides an excellent overview, and Brody et al. 共1996兲 ex- is known as the Reynolds number.
plore its biotechnological analog. More detailed ac- Reynolds numbers for common microfluidic devices
counts are given in conference proceedings 共Batchelor, can be estimated as follows. With water as the typical
1977; Hinch, 1988兲 and textbooks 共Happel and Brenner, working fluid, typical velocities of 1 ␮m / s – 1 cm/ s, and
1983; Kim and Karilla, 1991; Leal, 1992; Pozrikidis, 1992; typical channel radii of 1 – 100 ␮m, the Reynolds num-
Deen, 1998兲. bers range between O共10−6兲 and O共10兲. These low values
For the physicist more familiar with Newton’s laws of Re affirm that viscous forces typically overwhelm in-
than with the convective derivative ␳u · ⵜ u of the ertial forces, and the resulting flows are linear. The loss
Navier-Stokes equations, Fig. 3 depicts two simple illus- of nonlinearity at low Re led to the demise of the re-
trations. The first involves a fluid that flows with velocity search effort devoted to miniaturized fluidic computa-
U0 through a microchannel of width w that makes a sud- tion, whose fundamental elements relied upon inertial
den right turn. During the turn time ␶0 ⬃ w / U0, a fluid effects and thus could not scale down like their solid-
element rounding the corner loses momentum density state electronic competition 共Humphrey and Tarumoto,
␳U0 by exerting an inertial centrifugal force density fi 1965; Foster and Parker, 1970; Joyce, 1983兲.
⬃ ␳U0 / ␶0 = ␳U20 / w. The other involves a fluid flowing in a When Re is very small, the nonlinear terms in Eq. 共1兲
channel that contracts over a length l. By mass conser- disappear, resulting in linear and predictable Stokes flow
vation, the velocity increases as u ⬃ U0共1 + z / l兲, causing a 关Eq. 共2兲兴. As Re increases, the first features of inertia
fluid element to gain momentum at a rate become apparent. For example, in flow through a circu-
lar channel that is slightly curved with radius of curva-
du du ␳U20 ture R much larger than the channel radius w, centrifu-
fi ⬃ ␳ = ␳U0 ⬃ . 共4兲
dt dz l gal forces on fluid elements drive a secondary flow. The

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

982 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

from rapidly oscillating flows can give rise to the steady

streaming flow discussed in Sec. III.B.
Inertia rarely plays a significant role in microfluidic
systems, and as systems are made ever smaller, it be-
comes even less relevant. Without the inertial nonlinear-
ity, straightforward microfluidic systems have regular,
deterministic flow. There do exist, however, many physi-
cal processes exploited in microfluidics—capillary effects
at free surfaces, viscoelasticity in polymer solutions, and
electrokinetic effects, to name a few—whose nonlineari-
ties may even increase as device dimensions diminish,
and which give rise to a rich variety of microfluidic phe-
FIG. 4. 共Color in online edition兲 A secondary Dean flow re- nomena. The remainder of this review explores this
sults from inertial forces on fluid flowing 共into the page兲 world.
through a slightly curved channel. Fluid elements are forced
centrifugally towards the outside of the turn, and shaded con-
tours show the magnitude of the centrifugal force density
共greatest in the center, where the primary flow is fastest兲. This B. The Péclet number and the problem of mixing
inhomogeneous forcing drives the secondary flow shown in the
In the everyday high-Re world, random eddies con-
plane of the page.
tinuously churn the fluids in which we live—and in the
process, stretch and fold fluid elements chaotically. As a
no-slip boundary condition and inhomogeneous flow result of this so-called turbulent mixing, gradients are
profile, u ⬃ U0关1 − 共r / w兲2兴, cause inhomogeneous cen- significantly enhanced and time scales for mixing are
trifugal forcing, fi ⬃ U20关1 − 共r / w兲2兴2 / R, and give the sec- dramatically reduced. In fact, without turbulent mixing
ondary “Dean flow” UD ⬃ Re共w / R兲U0 shown in Fig. 4 or thermal convection, it would take about a year to
共Dean, 1927, 1928; McConalogue and Srivastava, 1968; smell your feet after taking off your shoes! The laminar
Leal, 1992兲. Secondary flow in twisted pipes have served fluid flows that naturally arise in the low-Re world of
as a model system for mixing studies 共Jones et al., 1989; microfluidics, however, force mixing to occur by diffu-
Sharp et al., 1991; Jones and Young, 1994兲, and have sion alone, which can result in unacceptably long mixing
recently been applied to create a passive, three- times of order minutes or more.
dimensional “serpentine” micromixer 共Liu et al., 2000; Purely diffusive mixing can be desirable or not, de-
Rush et al., 2002, albeit with rectangular channels and pending on the application. Microfluidic chemical reac-
appreciable Re兲. As Re gets still higher, the nonlinear tors require different solutions to be brought together
inertial term destabilizes the flow, resulting in unpredict- and mixed rapidly, allowing the dynamics of the reac-
able, irregular turbulent flow. The standard example for tions to be probed, rather than the diffusive dynamics of
the transition to turbulence involves flow through a the molecules themselves. The opposite problem is
straight circular pipe, where the transition occurs for Re faced, however, in sorting and analyzing the products of
between 2000 and 3000. Such Reynolds numbers are sig- those same reactions: the faster the mixing, the harder
nificantly higher than those encountered in microfluidic the separation. Controlling dispersion in microfluidic de-
devices, and microfluidic flows generally fall safely vices, then, is often of paramount importance.
within the laminar flow regime. For example, consider a T junction in which two fluids
Thus far we have focused on steady inertial forcing are injected to flow alongside each other, as in Fig. 5共a兲.
due to the convective derivative ␳u · ⵜ u, which is almost How far down the channel must the fluids flow before
always unimportant for microfluidic flows because its the channel is homogenized? A simple estimate requires
time dependence arises from U0. The linear unsteady the particles or molecules to diffuse across the entire
term ␳ ⳵ u / ⳵t, on the other hand, sets the inertial time channel, giving a time ␶D ⬃ w2 / D, where w is the width
scale ␶i required to establish steady flows. This time of the channel. During this time, the stripe will have
scale can be estimated by balancing the unsteady inertial moved a distance Z ⬃ U0w2 / D down the channel, so that
force density ⬃␳U0 / ␶i with the viscous force density the number of channel widths required for complete
␩U0 / L20, giving mixing would be of order
Z U 0w
⬃ ⬅ Pe. 共7兲
␳L20 w D
␶i ⬃ . 共6兲
␩ The dimensionless number on the right is known as the
Péclet number 共Pe兲, which expresses the relative impor-
The inertial time scale can be interpreted as the time tance of convection to diffusion. In this example, the
required for vorticity to diffuse a distance L0, with dif- number of channel widths required for full mixing varies
fusivity ␯ = ␩ / ␳. This time scale is typically small, and linearly with Pe. Using the diffusivities in Table III, we
pressure-driven flow requires ␶i ⬃ 10 ms to reach steady see that even a small protein flowing with the fluid
state in rigid 100-␮m channels. However, inertial effects through a 100-␮m channel at 100 ␮m / s requires Pe

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 983

FIG. 5. 共Color in online edition兲 共a兲 The microfluidic T sensor

共Kamholz et al., 1999兲. Different fluids are brought together at
a T junction to flow alongside each other down the channel. A
FIG. 6. 共Color in online edition兲 The membraneless H filter
simple estimate suggests that the interdiffusion zone spreads
exploits the different rates at which tracers with different dif-
diffusively, with the square root of time 共or downstream dis-
fusivities 共and thus Pe兲 spread across a channel. The length l is
tance兲, although 共b兲–共d兲 show this naive argument to break
chosen so that large waste products do not have time to diffuse
down near the “floor” and “ceiling” of the channel. Confocal
across the channel, and thus remain confined to their initial
microscopy reveals the three-dimensional nature of the
stream, whereas smaller molecules of interest diffuse across
spreading of the interface in the T sensor 共Ismagilov et al.,
the channel into the neighboring stream. At the outlet, the
2000兲. 共b兲 Fluorescent tracers mark reactions occurring in the
second stream contains the more mobile species almost exclu-
interdiffusion zone, here seen from above. 共c兲, 共d兲 The no-slip
sively 共Brody et al., 1996; Brody and Yager, 1997兲.
nature of the top and bottom walls of the channel affect the
flow profile, so that tracer molecules near the boundaries dif-
fuse and spread with z1/3, rather than z1/2. Reprinted with per- reactions introduce additional time scales and complex-
mission from Ismagilov et al., 2000. ©2000, AIP. ity into the system.兲 T sensors have be used to measure
analyte concentration 共Weigl and Yager, 1999兲 and ana-
⬃ 250 channel widths 共approximately 2.5 cm and 4 min兲 lyte diffusivities and reaction kinetics 共Kamholz et al.,
to completely mix. 1999, 2001; Baroud et al., 2003兲. Finally, competitive im-
munoassays have been performed by injecting an anti-
body solution alongside a solution of known, labeled an-
1. Sensing, filtering, and fabricating with parallel laminar
tigen. Antigen-antibody binding is evidenced by marker
flows
accumulation in the interdiffusion zone, and an un-
The competition between convection and diffusion, marked antigen can be detected differentially, as compe-
embodied in the Péclet number, forms the basis for a tition for antibody binding alters the marker profile
number of techniques for sensing and separating flow measured downstream 共Hatch et al., 2001兲.
ingredients. The following devices are designed to oper- The naive picture presented above does not tell the
ate at intermediate Pe, where differences in solute diffu- whole story about the T sensor, however. Ignoring the
sion rates play the key role. channel top and bottom leads to the expectation that the
width of the interdiffusion zone should grow with z1/2, as
a. T sensor in Eq. 共7兲. In fact, experiment and analysis have shown
the front to behave in a more complicated fashion. Con-
One device that employs adjacent laminar streams is focal microscopy of the three-dimensional front 共Fig. 5兲
the T sensor, shown in Fig. 5共a兲 共Kamholz et al., 1999; determined the spreading near the top and bottom walls
Weigl and Yager, 1999兲. Two fluid streams are brought to
to vary with z1/3 and with z1/2 near the middle 共Ismagilov
flow alongside each other down a channel, and solute
et al., 2000兲, which occurs because fluid near the top and
molecules in each stream diffuse into the other, forming
bottom moves more slowly than the middle, so that trac-
an interdiffusion zone whose boundary is measured,
ers do not move as far downstream as they diffuse across
typically with a fluorescent marker. 共Note, however, that
the streamlines. The near-wall interfacial dynamics can
thus be understood in terms of the classic Lévêque prob-
TABLE III. Typical diffusivities for various tracers in water at lem of diffusion in a shear flow 共Lévêque, 1928; Deen,
room temperature. 1998; Ismagilov et al., 2000; Kamholz and Yager, 2001兲.
Further analysis near the inlet reveals z1/3 spreading
Characteristic diffusivities near the walls and z1/2 in the channel center, which then
Particle Typical size Diffusion constant catches up downstream 共Kamholz and Yager, 2002兲.
Solute ion 10−1 nm 2 ⫻ 103 ␮m2 / s
Small protein 5 nm 40 ␮m2 / s b. Filtration without membranes
Virus 100 nm 2 ␮m2 / s Figure 6 depicts the H filter, a simple device that fil-
Bacterium 1 ␮m 0.2 ␮m2 / s ters particles by size without a membrane 共Brody et al.,
Mammalian/human cell 10 ␮m 0.02 ␮m2 / s 1996; Brody and Yager, 1997兲. As in the T sensor, two
different streams are brought together to flow alongside

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

984 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

FIG. 7. A variant of the H filter 共Fig. 6兲 that separates motile

from nonmotile sperm. While the traditional H filter relies
FIG. 8. A three-electrode system fabricated within existing mi-
upon particle diffusivity differences for differentiation, this de-
crochannels. 共a兲 Initially, a gold strip runs perpendicular to a
vice exploits the fact that motile sperm disperse across and
microchannel. 共b兲 Three high-Pe fluid streams flow down the
homogenize the channel much more rapidly than nonmotile
channel, of which the central stream contains a chemical that
ones, which spread via diffusion alone. Adapted and reprinted
etches away the underlying gold and separates the two elec-
with permission from Cho, Schuster, et al., 2003. ©2003,
trodes. 共c兲 Finally, a reaction at the interface of two high-Pe
American Chemical Society.
fluid streams causes silver to precipitate and form a thin wire
on the floor of the channel. Reprinted with permission from
Kenis et al., 1999. ©1999, AAAS.
each other down a channel. One stream is a dilute solu-
tion of different-sized molecules or particles, each of
which has its own diffusivity and Péclet number. Using c. Fabricating and interrogating using multiple laminar
the simple argument above, each Pe determines the streams
channel length required for that component to diffuse
The above devices work in the intermediate-Pe re-
across the channel width. The H filter works when the
gime: the T sensor requires the interface to spread dif-
length l or flow velocity U0 is chosen so that one Pe is
fusively on appropriate time and length scales, and the
small and the other is large. Low-Pe small solute par- H filter is optimal when Pe is small for the component to
ticles diffuse to fill the channel before exiting, whereas be extracted, and large for the “waste.” We now discuss
high-Pe large particles remain confined to their half of devices and techniques that are optimal in the large-Pe
the channel. A reverse T sensor then reseparates the regime, wherein multiple fluid streams can flow along-
flow into two streams, of which one contains mostly side each other over long distances with minimal mixing.
small particles, and the other contains both species. Nonmixing, high-Pe flows can be used to fabricate and
The H filter requires only that the components to be etch structures within existing microchannels 共Takayama
separated spread at different rates transverse to the flow, et al., 1999, 2001; Kenis et al., 2000兲. For example, Fig. 8
and works regardless of the cause of the difference in shows the fabrication of a three-electrode system within
dispersivity. Figure 7 shows a variant on the H filter, a microchannel 共Kenis et al., 1999兲, in which parallel
wherein motile sperm are separated from nonmotile streams selectively etch areas of an electrode, after
ones 共Cho, Schuster, et al., 2003兲. Rather than differing which a silver wire is formed by a precipitation reaction
in diffusivity, motile sperm rapidly and randomly swim between two streams. Other fabrication techniques in-
to fill the channel, as compared with nonmotile sperm, clude selectively patterning hydrophobic stripes along a
which spread via diffusion alone. Here as above, the spe- channel 共Zhao et al., 2001兲, cutting trenches into channel
cies that spreads quickly is extracted. walls using etchants, selectively coating portions of
A device by Schilling et al. 共2002兲 demonstrates the channel walls with molecules or cells, and forming linear
deposits of crystals, polymeric lines, and thin wires
type of analysis that can be performed with these prin-
within an existing channel 共Kenis et al., 1999兲.
ciples. An H filter brings a stream of cells into contact
By growing or placing living cells within microchan-
with a chemical agent that lyses the cell membrane, al- nels, cellular behavior and response can be probed lo-
lowing certain intracellular molecules to escape. One cally if different laminar streams affect different parts of
output of the H filter contains both waste cells and in- the cell. For example, the stream to the left of the
tracellular molecules. The other contains mostly the dashed line in Fig. 9 contains trypsin, which disrupts cell-
molecules of interest, which are then brought to flow substrate adhesions and causes a bovine capillary endo-
alongside a stream of detection molecules in a T sensor thelial cell to lose its local footing and move to the right
to probe various properties, e.g., identification or con- 共Takayama et al., 2003兲. Similarly, locally stimulating re-
centration. ceptors allows the study of cell signaling 共Sawano et al.,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 985

FIG. 9. 共Color in online edition兲 Local cell manipulation with

parallel laminar streams. The stream to the left of the dashed
line contains trypsin, which disrupts cell adhesion and causes a FIG. 10. 共Color in online edition兲 Sorting by diffusion. To
bovine capillary endothelial cell to recoil towards the “inert” leading order, particles driven through the array follow stream-
stream on the right. 共a兲 Trypsin starts, 共b兲 1 min; 共c兲 3.5 min. lines, and travel across streamlines through diffusion alone. 共a兲
Reprinted with permission from Takayama et al., 2003. ©2003, When the array has a broken spatial symmetry, the hopping
Elsevier. rates are biased in one direction. Reprinted with permission
from Duke and Austin, 1998. 共b兲 Smaller DNA fragments hop
more frequently and thus follow trajectories that are more
2002兲, selectively disrupting actin filaments probes local transversely directed than larger particles. Reprinted with per-
changes in cytoskeletal structure, and selectively label- mission from Chou et al., 2000. ©2000, Wiley.
ing mitochondria at opposite ends of a cell allows their
motion to be traced 共Takayama et al., 2001, 2003兲. 1992; Magnasco, 1993; Astumian and Bier, 1994; Prost et
al., 1994; Bier and Astumian, 1996兲. Figure 10 shows mi-
crofabricated arrays of asymmetric obstacles, through
2. Fighting diffusion: Particle separations
which particles are driven electrophoretically or with
Molecular separation techniques in analytical chemis- pressure-driven flow. In the absence of diffusion 共i.e.,
try were among the earliest applications of microfluidic PeⰇ 1兲, suspended particles follow field or flow lines. At
devices 共Harrison et al., 1992, 1993; Manz et al., 1993; moderate Pe, however, a particle occasionally diffuses
Reyes et al., 2002兲. Many techniques exist to separate far enough to cross over into a neighboring trajectory in
the components of solution for analysis, typically by size a direction biased by the array’s broken symmetry
or electric charge 关see, e.g., Giddings 共1991兲兴. For ex- 共Duke and Austin, 1998; Ertas, 1998兲. The hopping rate
ample, measurements in molecular biology and genom- of each species increases as ’its Pe decreases, so that
ics often require DNA fragments to be sorted by size, smaller 共higher-D, lower-Pe兲 particles make lateral hops
which is typically done electrophoretically. Different- more frequently, and thus separate by size. However,
sized fragments travel with relative velocities rates ⌬U Austin et al. 共2002兲 and Huang, Silberzan, et al. 共2002兲
and spread diffusively with diffusivity D. 共Note that D argued that this simple picture was insufficient for sepa-
reflects molecular diffusivity as well as Taylor dispersiv- ration, and one of two criteria must be met: 共i兲 field
ity, as discussed below.兲 A mixed collection of fragments and/or flow lines must penetrate the obstacles, or 共ii兲
evolves into “bands” separated by a distance ⌬Z ⬃ ⌬Ut, particles must be comparable in size to the gaps between
and whose widths grow as ⌬W ⬃ 共Dt兲1/2. Assuming the obstacles. The latter allows for separations even in the
different bands to travel with velocity scale U0, an esti- absence of diffusion 共Pe→ ⬁; Huang et al., 2004兲. The
mate for effective separation can be obtained. To clearly former has been achieved in electrophoretic separations
resolve the bands, the separation between bands must by using slightly conducting posts 共Huang, Silberzan, et
greatly exceed the width of each, which requires al., 2002兲, and in “optical fractionization” wherein fo-

冉 冊 1/2
cused laser arrays impede particles but not fluid flow
⌬Ut Ⰷ 冑Dt →
U 0Z 共MacDonald et al., 2003; Gluckstad, 2004; Ladavac et al.,
Ⰷ 1. 共8兲
D 2004兲.
Here we have assumed ⌬U ⬇ O共U0兲 and converted time
3. Beating diffusion: When mixing matters
into downstream distance via Z = U0t. Another Péclet
number, Pe= U0Z / D, thus appears, which is called the Thus far, we have concentrated on parallel laminar
theoretical number of plates in the separation literature streams, where interfaces between fluids of different
共Giddings, 1991兲. Since U0 ⬀ E in electrophoretic separa- content are largely parallel to the fluid velocity. This
tions, Eq. 共8兲 indicates that strong electric fields and long strategy was chosen because the mixing of the two
channels are best for separation. streams occurs slowly enough to be monitored, manipu-
We close our discussion of intermediate-Pe effects lated, measured, or even ignored. As such, fluid motion
with a final example for the separation of suspended plays a minimal role in the evolution of the interface.
particles by size 共Duke and Austin, 1998; Ertas, 1998; The first hint that we have seen of the role convection
Chou et al., 1999; Austin et al., 2002; Cabodi et al.,’ 2002; plays in interface evolution appeared in Fig. 5, where the
Huang, Silberzan, et al., 2002兲, related to theoretical parabolic velocity profile modified the scaling of the in-
ideas involving “ratchet” potentials 共Ajdari and Prost, terface spreading from Pe1/2 to Pe1/3 near the T junction.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

986 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

In some applications involving inhomogeneous fluids,

however, the side-by-side configuration of parallel lami-
nar streams is not possible. This occurs when molecules
or particles are separated by size, since separation bands
are typically oriented perpendicular to the flow or field
direction. In these cases, convection invariably acts to
fight the separation process.
Many applications, moreover, require exactly the op-
posite: a method to rapidly mix an inhomogeneous solu-
tion. To probe reaction kinetics or protein folding, for
example, distinct fluids must be mixed rapidly enough
that the system is reaction limited, rather than diffusion
limited. Since mixing times or distances can be very
long, strategies are required to enhance mixing in mi-
crodevices. All mixing—whether in turbulent or laminar
flows—ultimately occurs due to molecular diffusion.
Fluid stirring can be used to stretch and fold inhomoge-
neous fluid blobs until mixing 共diffusive migration across FIG. 11. 共Color in online edition兲 Taylor dispersion. 共a兲 An
streamlines兲 occurs 关following the terminology of Eckart infinitesimally thin tracer stripe in a circular channel of radius
共1948兲 and Aref and Balachandar 共1986兲兴. The basic idea w. 共b兲 Under a pressure-driven flow, the stripe is convectively
behind mixers is to use stirring motions to reduce the stretched into a parabolic shape whose width ⌬W increases
distance over which mixing must occur. linearly with time. 共c兲 After a characteristic time scale ␶D
A simple and elegant method for enhancing mixing ⬃ w2 / D for tracers to diffuse across the channel, the parabola
rates by reducing mixing distances involves flow focusing is “smoothed” into a plug of width WTD ⬃ U0w2 / D. 共d兲 The
共as in Fig. 14, but without capillary effects兲, in which a plug can be broken into many thin stripes, each of which sub-
sequently undergoes processes 共a兲–共c兲 to be smoothed into
central stream is narrowed to a width w using an outer
plugs of width WTD on a time scale ␶D. A tracer thus takes a
flow. In so doing, diffusive penetration of molecules random step of size ⱗWTD on each time step ␶D, causing the
from the outer flow into the central stream occurs more stripe to evolve as a Gaussian spreading with effective diffu-
rapidly 共Knight et al., 1998; Pabit and Hagen, 2002兲. sivity Dz ⬃ U20w2 / D.
Such an approach has been used to study protein folding
on submillisecond time scales 共Pollack et al., 1999, 2001;
Kauffmann et al., 2001; Pabit and Hagen, 2002; Lipman a. Taylor dispersion
et al., 2003兲; see Eaton et al. 共2000兲 for an early review. A Taylor dispersion provides a classic example of the
related approach reduces the diffusion length by forcing role of convection in dispersing inhomogeneous flows
the solution through a packed bed of microbeads 共Seong 共Taylor, 1953, 1954兲. A thin stripe of tracers initially
and Crooks, 2002兲. spans a circular channel of radius w, as in Fig. 11共a兲. In
More elaborate methods are required, however, to ho- the absence of diffusion, pressure-driven 共Poiseuille兲
mogenize the contents of a microfluidic channel or flow with profile uz = U0共1 − r2 / w2兲 stretches the stripe
chamber. The classic text of Ottino 共1989兲 provides an into a parabola, where the tracer in the center leads the
excellent introduction to mixing science. Various strate- tracer at the walls by a distance U0t. In a reference
gies are being developed to enhance microfluidic mixing, frame that moves with the mean flow, the tracer stripe is
and a recent issue of Philosophical Transactions has stretched to ± ⬃ U0t 关Fig. 11共b兲兴. However, there is a time
been dedicated entirely to microfluidic mixing. For this scale ␶D ⬃ w2 / D at which molecular diffusion across the
issue, Ottino and Wiggins 共2004兲 provide an overview, channel smears the parabolic stripe into a plug of width
Wiggins and Ottino 共2004兲 explore the mathematical WTD ⬃ U0w2 / D 关Fig. 11共c兲兴. Thus a tracer within the ini-
foundations of mixing, Stremler et al. 共2004兲 discuss de- tial stripe effectively takes a random step of size ⱗWTD
sign principles, and Bringer et al. 共2004兲, Bottausci et al. each time step ␶D. Each stripe within the plug goes
共2004兲, Campbell and Grzybowski 共2004兲, Darhuber et through the same process: each is convectively stretched
al. 共2004兲, Mensing et al. 共2004兲, Stroock and McGraw and then diffusively smeared to a width WTD after a time
共2004兲, and Tabeling et al. 共2004兲 explore different mix- ␶D. After N steps, the initially thin stripe evolves into a
ing systems in depth. Gaussian with width
It is of great importance to understand the role of
fluid convection in the dispersion of tracers, by which we
mean any object that is transported by, and diffuses in,
the fluid 共examples include dyes, analyte molecules, pro-
teins, cells, salt, or even heat兲. A pattern will emerge:
具W2典1/2 ⬃ N1/2WTD ⬃ 冉 冊
U20w2
D
t
1/2
. 共9兲

tracer dispersion is initially dominated by convective

stretching with the fluid, which is then followed some The tracer distribution thus grows diffusively 共as t1/2兲
time later by diffusive homogenization. with an effective long-time axial diffusivity,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 987

U20w2
Dz ⬃ ⬃ Pe2D, 共10兲
D
that occurs in addition to the molecular diffusivity D.
Detailed analyses for specific geometries yields prefac-
tors, as presented in textbooks such as Brenner and Ed-
wards 共1993兲 and Deen 共1998兲. Important examples in-
clude flow in a circular channel with radius w, for which
Dz = Ū2w2 / 48D; flow in shallow rectangular channel of
height h, for which Dz = Ū2h2 / 210D; and simple shear
flow between plates separated by h, for which Dz FIG. 12. 共Color in online edition兲 Rotary mixer 共a兲, in which
= 2Ū2h2 / 15D. valves 关Figs. 45共a兲 and 45共b兲兴 pump peristaltically around the
Several important points are in order. First, the axial ring. Reprinted from Chou et al., 2001, with kind permission of
Taylor dispersivity Dz 关Eq. 共10兲兴 is only valid on long Springer Science and Business Media, ©2001. In the convec-
time scales t Ⰷ w2 / D, or after a downstream distance L tively stirred regime, an initially vertical stripe 共b兲 deforms un-
Ⰷ Pe w. This is sometimes impractically long for micro- der the parabolic flow 共c兲 to make N = U0␶s / 2␲R revolutions,
fluidic applications—Taylor’s original experiments were reducing the effective distance between branches of the stripe
to heff ⬃ h / 2N. The stripe continues to roll up, and heff gets
conducted using a 500-␮m capillary and lasted four
increasingly smaller, until 共d兲 molecular diffusion across heff is
hours! On shorter time scales, tracer blobs are stretched
fast enough to homogenize the channel. This requires a time
convectively, and grow linearly in time. Only after trac- ␶con ⬃ Pe−2/3␶D, where ␶D ⬃ h2 / D.
ers have had enough time to diffuse across the channel
does a blob spread diffusively with dispersivity Dz 关Eq.
共10兲兴. Taylor dispersion acts in the direction of the flow, dients are sufficiently steep or time scales sufficiently
and is thus not observed in T sensors and H filters, long.
where gradients in tracer concentration are perpendicu- As a paradigmatic example, we investigate the rotary
lar to the flow direction. Second, Taylor dispersion may mixer shown in Fig. 12共a兲 共Chou et al., 2001兲. For sim-
appear to dramatically enhance dispersion for small sol- plicity, we use physical arguments and treat a two-
utes, since Dz ⬃ D−1. Without Taylor dispersion, how- dimensional model, similar to mixing in a slit channel
ever, convection would stretch tracer blobs indefinitely 共Handique and Burns, 2001兲. A more mathematical
with time; Taylor dispersion in fact cuts off the convec- treatment is provided by Gleeson et al. 共2004兲. Fluid is
tive growth and reduces it to diffusive 共t1/2兲 growth. Thus pumped with average velocity U0 around a closed, circu-
stripes in pressure-driven flow evolve into Gaussian lar channel of radius R and height h. How quickly the
bands moving at the mean flow speed and spreading fluid mixes depends on how quickly the fluid flows.
with an effective diffusivity D + Dz. In fact, in sufficiently There are three distinct regimes: 共i兲 diffusion dominated,
slow flows and/or thin channels for which Peⱗ 1, tracers 共ii兲 Taylor dispersion mediated, and 共iii兲 convectively
in pressure-driven flow will evolve as though the flow stirred.
were uniform 共e.g., like electro-osmotic “plug” flow, Sec. (i) Diffusion-dominated mixing. When the pumping
III.A兲. velocity U0 is extremely slow 共Pe= U0h / D Ⰶ 1兲, molecu-
A vast literature has developed around Taylor disper- lar diffusion plays the dominant role in tracer dispersion,
sion, with over 2000 citations to date. Following Taylor’s beating even Taylor dispersion 共D Ⰷ Dz兲. In this case,
original exploration of the subject 共Taylor, 1953, 1954兲, mixing is achieved only after tracers diffuse around the
Taylor’s picture was generalized using a method of mo- circumference of the ring, requiring a time ␶R,

冉 冊
ments that allows for arbitrary channel shapes and flows
共Aris, 1956, 1959兲, and generalized Taylor dispersion 共2␲R兲2 2␲R 2
␶R ⬃ = ␶D . 共11兲
theory was developed 共Brenner and Edwards, 1993兲. D h
Note the mixing time does not depend on U0 or there-
fore on Pe.
b. Rotary mixer (ii) Taylor dispersion-mediated mixing. If the flow rate
is increased so that PeⰇ 1, Taylor dispersion assumes the
Above we developed the basic picture of Taylor dis-
dominant role in axial dispersion. In this case, axial
persion, wherein initially convective stretching of tracers
spreading increases diffusively with Taylor dispersivity
eventually gave way to diffusive dispersion, with en-
Dz 关Eq. 共10兲兴, resulting in a mixing time
hanced axial dispersivity. We now examine how convec-
tive stretching can be exploited to enhance mixing of 共2␲R兲2 D共2␲R兲2
inhomogeneous fluids in microfluidic devices. The gen- ␶TD ⬃ ⬃ ⬃ Pe−2␶R . 共12兲
Dz U20h2
eral theme to keep in mind is that there are typically two
steps: 共i兲 tracer blobs are initially stretched convectively In order for Taylor dispersion to dominate, molecules
by the flow, and 共ii兲 molecular diffusion of the tracers must diffuse across the channel 共requiring ␶D ⬃ h2 / D兲
across fluid streamlines homogenizes the flow once gra- before convective stretching folds a stripe into itself 共re-

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

988 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

quiring ␶ ⬃ 2␲R / U0兲. Taylor-dispersion-mediated mixing

thus occurs for Pe in the range 1 Ⰶ PeⰆ 2␲R / h.
(iii) Convectively-stirred mixing. Finally, when the flow
rate is sufficiently rapid 共PeⰇ 2␲R / h兲, a tracer stripe is
folded into itself multiple times before molecules diffuse
across the channel. After a stirring time ␶s, fluid in the
center makes N = U0␶s / 2␲R full revolutions. A vertical
stripe is stretched by the flow into a parabola, which
folds over itself N times, and the distance between adja-
cent stripes decreases as heff ⬃ h / 2N 关Fig. 12共b兲兴. For full
mixing to occur, molecules in the stripe need only diffuse
across the reduced distance heff, which requires a time
␶h ⬃ heff
2
/ D. Stirring proceeds until mixing wins, which
occurs when the two time scales roughly coincide: ␶s
⬃ ␶h = ␶con, FIG. 13. 共Color in online edition兲 Continuous-flow staggered

Ncon ⬃ 冉 冊
h
␲R
1/3
Pe1/3 , 共13a兲
herringbone mixer, in which grooved channel walls drive alter-
nating, asymmetric helical secondary flows that chaotically stir
the fluid. Each cycle cuts the distance between stripes in half,

冉 冊 冉 冊
so that the distance between stripes decreases exponentially
␲R 2/3
h 4/3
with the number of cycles. Diffusive mixing occurs when the
␶con ⬃ Pe−2/3 ␶D ⬃ Pe−2/3 ␶R . 共13b兲 tracer can diffuse from one stripe to the next before another
h ␲R
cycle has occurred, giving a mixing time that depends logarith-
Before continuing, we note a few points about the mically on Pe. Thus the channel cross section is rapidly mixed,
rotary mixer. First, mixing or dispersion is only en- and Taylor dispersion along the channel is significantly re-
hanced across flow gradients, just as with Taylor disper- duced. Reprinted with permission from Stroock, Dertinger, Aj-
sion. Therefore, the rotary mixer is most effective in dari, et al., 2002. ©2002, AAAS.
batch processes, in which a homogeneous plug is mixed
thoroughly into a fluid stored in the ring. Examples
where batch process are mixed in this manner include fluid elements. After N cycles requiring ␶cyc ⬃ NLcyc / U,
protein solubility studies 共Hansen et al., 2004兲 and where Lcyc is the cycle length, stripes are separated by a
single-cell DNA extraction and amplification 共Hong et distance heff ⬃ h / 2N. Following the above reasoning,
al., 2004兲. Second, in the convectively stirred mixing re- mixing occurs when the time to diffuse between stripes
gime 共PeⰇ R / h兲, the number of cycles required for mix-
2
heff / D is comparable to the cycle time ␶cyc, which re-
ing increases algebraically with Pe, because fluid ele- quires
ments are stretched linearly in time. Below we will see Nchaotic ⬃ ln Pe, 共14a兲
the number of cycles required for chaotic mixers, which
stretch fluid elements exponentially in time, increases
Lcyc ln Pe
logarithmically with Pe. In all cases, the time per cycle ␶chaotic ⬃ ␶D . 共14b兲
varies like Pe⫺1, giving a mixing time that decreases al- h Pe
gebraically with Pe. Indeed, the number of cycles 共or mixing length兲 mea-
sured in the staggered herringbone mixer depends loga-
rithmically on Pe over six decades 共Stroock, Dertinger,
c. Mixing by chaotic advection Ajdari, et al., 2002兲.
When fluid elements move chaotically, they are Another interesting fact about chaotic mixing in such
stretched and folded exponentially, enhancing mixing. continuous-flow systems is that Taylor dispersion is dra-
Turbulent mixing works on this principle. Although matically reduced 共Stroock, Dertinger, Ajdari, et al.,
Stokes flows are inherently linear, even simple Stokes 2002兲. Taylor dispersion, as described above, cuts off
flows can have chaotic streamlines that exponentially convective stretching when fluid elements have sampled
stretch and fold. This forms the basis for Aref’s classic all cross-sectional positions and, therefore, all axial ve-
mixing strategy, chaotic advection 共Aref, 1984, 2002兲. locities. In the standard picture of Taylor dispersion, this
The staggered herringbone mixer is a chaotic mixer for “sampling” occurs due to diffusion alone. In the chaotic
continuous flow systems that does not depend on inertia mixer described above, however, tracers explore the
共Stroock, Dertinger, Ajdari, et al., 2002; Stroock and channel cross section chaotically, quickly cutting off con-
McGraw, 2004兲. In it asymmetric grooves in the channel vective spreading and dramatically reducing axial disper-
walls induce an axially modulated secondary flow sion. Such dispersion reduction was studied in the
共Stroock, Dertinger, Whitesides, et al., 2002兲 in the form twisted pipe 共Jones and Young, 1994; Castelain et al.,
of an asymmetric set of counter-rotating fluid rolls 共Fig. 2000; Rush et al., 2002兲 and measured in the herringbone
13兲. The asymmetry of the rolls is periodically reversed, mixer 共Stroock, Dertinger, Ajdari, et al., 2002兲.
so that the distance between stripes halves with each Steady, incompressible two-dimensional flows are in-
cycle, leading to exponential stretching and folding of tegrable and cannot exhibit chaotic trajectories 关Ottino

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 989

共1989兲, Aref 共2002兲, Wiggins and Ottino 共2004兲, and ref-

erences therein兴. However, steady three-dimensional
flows can have chaotic streamlines 共Bajer and Moffatt,
1990; Stone et al., 1991; Kroujiline and Stone, 1999; Aj-
dari, Stroock, Dertingeret al., 2002; Stroock, Dertinger,
Whitesides, et al., 2002兲, as can unsteady two-
dimensional flows 共Aref, 1984; Ottino, 1989兲. “Topo-
logic” chaotic mixing uses interlocking helical channels
designed to replicate the chaotic Baker’s transformation
共Chen and Meiners, 2004兲. Oscillatory flows driven FIG. 14. Capillary instabilities in a microfluidic two-phase
across a main channel stretch and fold fluid in the pri- flow. A stream of water flows between streams of oil and is
mary flow into itself, significantly enhancing mixing geometrically focused into a narrow cylindrical jet. The jet is
共Glasgow and Aubry, 2003; Bottausci et al., 2004; Okkels destabilized by the Rayleigh-plateau instability and forms
small, monodisperse droplets. Reprinted with permission from
and Tabeling, 2004; Tabeling et al., 2004兲. Chaotic flows
Anna et al., 2003. ©2003, AIP.
often contain ordered regions that hamper mixing. The
barriers they present can be overcome by driving
singularity-induced diffusion to achieve uniform mixing ample, Fig. 14 共Anna et al., 2003兲 shows a thin central
in time-dependent, three-dimensional laminar flows 共So- stream of water breaking into drops due to the
lomon and Mezic, 2003兲. Rayleigh-Plateau instability 共Rayleigh, 1879; Chan-
The above ideas employed continuous chaotic flows. drasekhar, 1981; de Gennes, Brouchard-Wyart, and
Bounded Stokes flows may also be chaotic, since their Quere, 2004兲. Clearly, surface tension can play an impor-
streamlines need not be closed 共Bajer and Moffatt, 1990; tant role in microfluidic flows when immiscible free sur-
Stone et al., 1991; Kroujiline and Stone, 1999兲. This al- faces are present.
lows chaotic mixing to occur within drops, where a typi-
cal technique involves superposing two simple flow
fields. A sedimenting drop in a shear flow 共Bryden and 1. Droplet formation in two-phase flows
Brenner, 1999兲, a drop driven along a turning micro- Thorsen et al. 共2001兲 demonstrated that microfluidic
channel 共Song, Bringer, et al., 2003; Song and Ismagilov, devices could be used to create controllable droplet
2003; Song, Tice, et al., 2003; Tice et al., 2003; Bringer et emulsions in immiscible fluids, by injecting water into a
al., 2004兲, and oscillating drops 共Angilella and Brancher, stream of oil at a T junction 共Fig. 15兲. Were there no
2003兲 all can exhibit chaotic mixing. An applied electric interfacial tension between water and oil, the two
field induces an electrohydrodynamic quadrupolar flow streams would flow alongside one other, as in the T sen-
in the droplet 共Taylor, 1966; Melcher and Taylor, 1969; sors. However, competing stresses drive the interface:
Torza et al., 1971兲 that, when combined with the dipolar surface tension acts to reduce the interfacial area, and
flow induced by drop sedimentation, can be chaotic 共Lee viscous stresses act to extend and drag the interface
et al., 2000; Ward and Homsy, 2001, 2003; Im and Kang, downstream. These stresses destabilize the interface and
2003兲. cause droplets of radius R to form 共Rallison, 1984;
In principle, any technique for manipulating fluid flow Stone, 1994兲. An estimate for the size of the droplets can
can be used to promote rapid mixing. Examples include be obtained by balancing the two stresses on the inter-
electrowetting 共Paik et al., 2003兲 nonlinear electrokinetic
effects 共Takhistov et al., 2003; Bazant and Squires, 2004兲,
electro-osmosis in inhomogeneously charged channels
共Qian and Bau, 2002兲, acoustic streaming 共Yang et al.,
2000, 2001兲, and others.
Lastly, a few microfluidic mixers have been developed
by direct analogy with turbulence. Rather than employ a
deterministic but chaotic flow, these micromixers desta-
bilize the flow using noninertial means. Electrohydrody-
namic instabilities 共Oddy et al., 2001; El Moctar et al.,
2003; Lin et al., 2004; Fig. 36兲 and elastic instabilities of
dilute polymer solutions 共Groisman and Steinberg, 2001,
2004兲 have proven effective.

FIG. 15. Monodisperse microdroplet generation in a simple

C. The capillary number: Free-surface deformations microfluidic device. The interface between flowing immiscible
fluids is driven by competing stresses: viscous shear stresses
Thus far, we have assumed that fluids are miscible: tend to extend and drag the interface, whereas surface tension
parallel streams were assumed to flow alongside each tends to reduce the interfacial area. The competition between
other, and tracers diffused freely from one stream to the the two leads to drop sizes 共scaled by the channel height h兲 of
other. Between immiscible fluids, however, a surface ten- order R ⬃ Ca−1. Adapted with permission from Thorsen et al.,
sion ␥ affects the dynamics of the free surface. For ex- 2001.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

990 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

a curving microchannel—as with the chaotic mixers de-

scribed in Sec. II.B.3, the mixing time for these droplets
was measured to depend logarithmically on Pe 共Song,
Bringer, et al., 2003兲.

2. Fluid control using patterned surfaces

The large surface-to-volume ratios of microfluidic de-
vices render surface effects increasingly important, par-
ticularly when free fluid surfaces are present. Surface
FIG. 16. 共Color in online edition兲 Chaotic advection in a
tensions can exert significant stress that results in free
pumped droplet. A droplet driven down a channel experiences
a dipolar, circulating flow; when the channel changes direction,
surface deformations and/or bulk liquid motion.
the axis of the dipole changes as well. Alternating dipolar flows Most simply, capillary forces tend to draw fluid into
drive rapid mixing by chaotic advection, with mixing lengths wetting microchannels wherein the solid-liquid interfa-
that vary logarithmically with Pe. Reprinted with permission cial energy ␥sl is lower than for the solid-gas interface
from Song, Bringer, et al., 2003. ©2003, AIP. ␥sg. 共Note that we consider the second fluid to be a gas
for simplicity of notation; it could also be a second liq-
uid.兲 The dynamics of surface tension-driven fluid intru-
face 共Taylor, 1934兲. Capillary stresses of magnitude ␥ / R sion into a pipe of radius w follow from a balance of
balance viscous stresses ␩U0 / h, giving a characteristic capillary and viscous forces 共Washburn, 1921兲. The
droplet size curved meniscus at the fluid-gas interface gives rise to a
␥ h Laplace pressure ⌬p ⬃ ⌬␥ / w, where ⌬␥ = ␥sl − ␥sg repre-
R⬃ h= . 共15兲 sents the net energy 共per area兲 gained or lost as the fluid
␩U0 Ca interface advances. This pressure drives the fluid column
Here we have introduced the capillary number of length z into the channel via Poiseuille flow with ve-
locity scale
␩U0
Ca = , 共16兲 ⌬pw2 ⌬␥w
␥ u⬃ ⬃ . 共17兲
␩z ␩z
a dimensionless parameter found whenever interfacial
stresses compete with viscous stresses. As expected, the capillary number determines the dy-
An advantage to this strategy is that one can produce namics 共Ca⬃ w / z兲. Because the column length z
monodisperse droplets, due to the deterministic nature changes as the front moves 共u = ż兲, fluid invades at an
of microfluidic flows. Further work on droplet-forming ever slowing rate, according to the Washburn equation
devices includes studies of the rich variety of droplet
patterns that forms in channels 共Thorsen et al., 2001;
Dreyfus et al., 2003兲, and studies of microchannel emul-
z⬃ 冉 冊
⌬␥w
␩
t
1/2
. 共18兲

sification 共Sugiura et al., 2000, 2001, 2002a, 2002b兲. One subtle point about wetting-driven scaling is in or-
Smaller droplets can be formed using flow focusing, ei- der. Usually, the dynamics of such systems are found by
ther by increasing shear gradients or by drawing the assuming that the energy gained by wetting is dissipated
stream into a thin jet that breaks up by the Rayleigh- viscously—presumably in the bulk of the fluid. However,
Plateau instability 共Fig. 14; Anna et al., 2003兲. Notably, fluid dissipation diverges near the contact line, requiring
the dynamics of flow-focusing bubble-forming systems is a cutoff length scale from microscopic physics 共Dussan,
largely independent of Ca for both low-Re 共Garstecki et 1979; de Gennes, 1985; de Gennes, Brochard-Wyart, and
al., 2004兲 and high-Re 共Ganan-Calvo, 1998; Ganan- Quere, 2004; Eggers and Stone, 2004兲. Depending on the
Calvo and Gordillo, 2001; Gordillo et al., 2001兲 flows. amount of flowing fluid, either bulk or contact line dis-
Geometric techniques have been used to create droplets sipation can determine the dynamics. In this review, we
of varying sizes and size distributions 共Link et al., 2004兲. will always assume bulk dissipation to dominate, al-
Introducing colloidal spheres allows the production of though this need not be the case.
monodisperse tetrahedral photonic balls for potential Drops can be repeatedly drawn into a capillary pump
use as building blocks for photonic materials 共Yi, Jeon, consisting of multiple, parallel, wettable channels, where
et al., 2003; Yi, Thorsen, et al., 2003兲. a narrow capillary retention valve brakes the motion
This strategy has been used to study chemical reaction 共Juncker et al., 2002兲. Handique et al. 共1997兲 use nonwet-
kinetics on millisecond time scales by injecting reactive ting channel walls to control capillary motion—for ex-
chemicals together to form drops 共Song, Bringer, et al., ample, a nonwetting section arrests capillary intrusion.
2003; Song and Ismagilov, 2003; Song, Tice, et al., 2003; Furthermore, fluid does not enter nonwetting side chan-
Tice et al., 2003; Bringer et al., 2004兲, and in developing a nels, which have thus been used as vents for droplets
platform for protein crystallization 共Zheng et al., 2003兲. driven by pressurized gas—once the droplet moves past
Droplets are advantageous for this sort of study because the vent, the gas escapes through the side channel rather
reagents are not dispersed beyond the boundary of the than continue to drive the droplet. Last, gas forced
drop. Figure 16 depicts mixing in droplets pumped down through nonwetting side channels have been used to

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 991

FIG. 17. 共Color in online edition兲 Pressure-sensitive valves based on patterned surfaces with differing wetting properties. The
central stripe is hydrophilic, the lower channel is moderately hydrophobic, and the upper channel is most hydrophobic. 共a兲 Virtual
walls contain the fluid so long as the applied pressure is below a critical value, and water is pumped along the central stripe. 共b兲
When the applied pressure exceeds a threshold value, however, the fluid is driven through one virtual wall and onto the moder-
ately hydrophobic surface. 共c兲 Finally, above a second threshold, the fluid is driven onto the upper 共most hydrophobic兲 surface, and
flows into all channels. Reprinted with permission from Zhao et al., 2001. ©2001, AAAS.

split and meter droplets at precise locations. This me- fying the solid-liquid surface tension ␥sl, or by inducing a
tered injection system has been used, for example, in an gradient in the liquid-gas surface tension ␥. Examples of
integrated DNA analysis device 共Burns et al., 1998兲. the former include flows due to electrowetting, surface
Hydrophobic stripes have been used to confine fluids gradients, and reactive flows. The latter include ther-
共Handique et al., 1997兲, and to create “wall-less” chan- mocapillary, electrocapillary, and solutocapillary motion.
nels 共Gau et al., 1999; Kataoka and Troian, 1999; Zhao et
al., 2001兲. The static structures of such confined liquids a. Solid-liquid interfacial energy manipulation
has been studied 共Gau et al., 1999; Darhuber et al., 2000兲, Several strategies for exploiting solid-liquid surface
and capillary spreading along hydrophilic stripes has tension have been demonstrated for microfluidic ma-
been shown theoretically and experimentally to vary nipulation. See Bain 共2001兲 for a brief review. Figure 18
with the square root of time 共Darhuber et al., 2001兲. presents one of the simplest examples: a long droplet of
Spreading along a hydrophilic stripe of width w re- length L in a channel of radius w with an inhomoge-
sembles Washburn’s analysis, except that the height of neous surface: hydrophobic 共with ␥slL兲 for z ⬍ 0 and hy-
the fluid stripe varies as h ⬃ w2 / Rr, where Rr is the ra- drophilic 共␥slR ⬎ ␥slL兲 for z ⬎ 0 共Raphael, 1988; Ondarçuhu
dius of curvature of the inlet reservoir drop, leading to and Veyssie, 1991; Weislogel, 1997兲. Energetically, the
z2 ⬃ 共␥w4 / ␩Rr3兲t 共Darhuber et al., 2001兲. Spreading along droplet wants to move onto the hydrophilic surface, and
capillary microstripes must occur slowly enough that moving with velocity U decreases the stored interfacial
CaⰆ 1; otherwise, capillary forces are not strong enough energy at a rate ⬃⌬␥wU, where ⌬␥ = ␥slL − ␥slR. This en-
to keep the fluid confined to the hydrophilic stripe. ergy is lost to viscous dissipation, which consumes a
Patterned surfaces have been employed in closed mi- power ␩ 兰 共⳵u / ⳵z兲2dV ⬃ ␩U2L when dissipation is domi-
crochannels to manipulate multiple immiscible fluids nated by viscous shear in the bulk. Assuming the capil-
within a single microchannel. Fluids can be driven sim- lary energy released to be balanced by viscous dissipa-
ply using an applied pressure low enough that capillary tion gives
forces keep the interface pinned to the edge of the pat-
terned stripe 共that is, ⌬p ⱗ ␥ / w兲 共Zhao, Moore, and ␩U w
⬅ Ca = . 共19兲
Beebe, 2002; Zhao, Viernes, et al., 2002兲. If the fluid ex- ⌬␥ L
ceeds this critical pressure, it is forced onto the hydro-
Again we see that the capillary number arises naturally,
phobic region. This effect forms the basis for the
because capillary stresses are balanced by viscous
pressure-sensitive gates in Fig. 17 共Zhao et al., 2001,
2003兲, in which liquid expands into nonwetting channels
once a critical pressure has been exceeded. With mul-
tiple fluids in the channel, reaction kinetics can be
probed at their large shared interface 共Zhao et al., 2001兲,
and immiscible streams can be driven in opposite direc-
tions to enable highly efficient countercurrent exchange
of tracers like gas molecules or heat across the interface
共Zhao, Viernes, et al., 2002兲.

3. Fluid manipulation with capillary forces

FIG. 18. 共Color in online edition兲 Droplet motion due to a
Capillary forces can also be used to manipulate and gradient in solid-liquid interfacial energy. A droplet straddling
transport fluids with free surfaces. In equilibrium, free hydrophobic and hydrophilic surfaces can release stored inter-
surfaces are determined by a balance of interfacial facial energy by moving further into the hydrophilic region. A
forces. Actively modifying interfacial energies, however, droplet moving with velocity U releases ⌬␥wU energy per unit
disrupts the balance and causes motion. time; this energy goes into viscous dissipation, which consumes
Broadly speaking, capillary manipulation of microflu- a power ␩U2L. As a result, the droplet moves with velocity
ids has been achieved via two strategies: either by modi- U ⬃ ⌬␥w / 共␩L兲.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

992 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

stresses. In this case, the droplet moves with velocity

scale U ⬃ ⌬␥w / ␩L.
Below, we discuss specific realizations: surfaces with
static wettability gradients, thermal gradients, reactive
fluids, self-propelling bi-slugs, optowetting, and 共most
popular兲 electrowetting. Although the examples we dis-
cuss below may employ alternate geometries, the basic
physics is the same for droplets moving through a chan-
nel, two-dimensional 共2D兲/3D droplets moving on a sub-
strate, or 2D/3D drops moving between two plates. In
principle, all can be used in microfluidic systems.

b. Surfaces with wettability gradients

The motion of a small fluid droplet on a surface with a
wettability 共interfacial energy兲 gradient was originally
examined theoretically as part of a study of cell spread-
ing 共Greenspan, 1978兲. Further treatments followed, in-
cluding fluid spreading on a patchy surface 共Raphael,
1988兲. Brochard 共1989兲 provided an elegant and compre-
hensive study of the motion of large or small droplets
关compared with the capillary length 共␥ / ␳g兲1/2兴 due to FIG. 19. 共Color in online edition兲 Driving droplet motion with
both solid surfaces with wettability gradients, and for gradients in solid-liquid interfacial energy driven by 共a兲 ther-
liquid-liquid tension gradients. Experimentally, fluids mal gradients, 共b兲 droplets that contain a chemical that reacts
straddling a wetting discontinuity were observed to to decrease surface wettability, 共c兲 liquid bi-slugs that leave a
move to the more highly wetting side 共Ondarçuhu and coating film that lowers the overall surface energy, and 共d兲
Veyssie, 1991; Weislogel, 1997兲, and a fluid droplet was light-induced reactions that create wettability gradients.
shown to travel uphill on a surface with wettability gra-
dient 共Chaudhury and Whitesides, 1992兲. A more recent Gennes, 1998; Lee and Laibinis, 2000; Lee, Kwok, and
manifestation involves the rapid motion of water drop- Laibinis 2002兲. Such techniques have not yet been used
lets condensing on a surface with a static wettability gra- explicitly for microfluidics, but could be useful for one-
dient. In this case, condensation energy establishes a shot applications.
thermal gradient, further driving motion 共Daniel et al.,
2001; Wasan et al., 2001兲.
iii. Self-propelling liquid bi-slugs [Fig. 19(c)]
i. Thermal gradients [Fig. 19(a)] A compound liquid bi-slug consisting of two indi-
While more attention has been paid to temperature vidual slugs of different liquids in a capillary can un-
variations in fluid-fluid surface tensions, and the Ma- dergo a capillary-driven self-propelled motion 共Bico and
rangoni flows that result, the solid-liquid interfacial ten- Quere, 2000, 2002兲. Moving the bi-slug creates interfa-
sion depends on temperature as well. A slug of fluid in a cial area behind the slug, and destroys it in front. Similar
microchannel moves to reduce its total interfacial en- to the reactive fluids described above 共Bain et al., 1994;
ergy, as does a droplet on a solid surface with a thermal Dos Santos and Ondarçuhu, 1995兲, this surface energy
gradient. The latter, however, can also move due to Ma- alteration provides a driving force for droplet motion.
rangoni stresses that drive the interface itself 共Brochard, The dependence of bi-slug velocity upon wetting prop-
1989; Brzoska et al., 1993兲. Burns et al. employ these erties, viscosity differences, and bi-slug length has been
effects in DNA analysis devices 共Burns et al., 1996, 1998; explored theoretically and experimentally 共Bico and
Sammarco and Burns, 1999, 2000兲, although their term, Quere, 2002兲. Like reactive fluids, bi-slugs alter the sur-
“thermocapillary,” has traditionally referred to interfa- face they have visited and stop after exhausting their
cial motion due to surface tension gradients 共Young et driving fuel.
al., 1959兲. By analogy with electrowetting, perhaps ther-
mowetting would be more appropriate. iv. Optowetting [Fig. 19(d)]
Light can drive droplet motion when the surface is
ii. Reactive fluids [Fig. 19(b)] coated with a photosensitive material. Static optowet-
The above examples involve constant, but inhomoge- ting, i.e., a change in contact angle change due to uni-
neous, interfacial energies ␥sl. Another technique for form light irradiation, has been reported 共Wang et al.,
driving fluid motion involves reactive fluids: a droplet 1997; Abbott et al., 1999; Shin and Abbott, 1999; Miyau-
that contains a chemical that decreases the wettability of chi et al., 2002; Huang et al., 2003兲. Furthermore, droplet
a solid surface spontaneously moves, seeking to avoid motion has been driven using gradients in light intensity
any areas it has visited 共Bain et al., 1994; Dos Santos and 共Ichimura et al., 2000; Oh et al., 2002兲. UV light of vary-
Ondarçuhu, 1995; Shanahan and de Gennes, 1997; de ing intensity causes a wettability gradient that drives

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 993

共per unit length兲 −cV2⌬z / 2, so that a force 共per unit

length兲 cV2 / 2 pulls the line out. This electrowetting ten-
sion gives a new force balance,

冉
␥sg − ␥sl −
cV2
2
冊
− ␥ cos ␪EW = 0, 共21兲

and the contact angle changes according to the Lipp-

mann equation 共Lippmann, 1875兲
cV2
cos ␪EW = cos ␪eq + . 共22兲
2␥
Similar physics occurs at electrified fluid-fluid interfaces,
giving electrocapillarity.
Initially, electrowetting was performed on bare elec-
trodes, in which case the capacitance of the water-
electrode surface is given by the capacitance of the ionic
double layer within the water, c ⬃ ⑀w / ␭D, where ␭D is the
FIG. 20. 共Color in online edition兲 Wetting and electrowetting. Debye screening length, typically of order nanometers
共a兲 In standard equilibrium wetting, each interface has an as- 共Sec. III.A兲. The large capacitance of the double layer
sociated tension, and the sum of forces on the three-phase allows large contact angle changes for relatively small
contact line must balance, leading to an equilibrium contact
applied voltages. However, electrochemical reactions oc-
angle ␪eq. 共b兲 When a potential difference V is applied between
cur at electrode-water interfaces when the applied po-
the drop and the solid surface it touches, an effective parallel-
plate capacitor c is charged at the interface. The charged ca-
tential exceeds a critical 共system-dependent兲 value typi-
pacitor lowers the solid-liquid interfacial energy by an amount cally of order 1 V. Such reactions fundamentally alter
cV2 / 2, which lowers the steady-state contact angle ␪EW. the physics of the system, e.g., by creating gaseous oxy-
gen and hydrogen, by injecting metallic ions into solu-
tion, or by changing the solution pH.
droplet motion, which can be reversed by blue-light ir- More recently, electrowetting using electrodes coated
radiation. Velocities of 36 ␮m / s were reported for 2-␮l with a thin dielectric layer, called EWOD 共electrowet-
olive oil drops. ting on dielectric兲 or EICE 共electrowetting on insulator
coated electrodes兲, has been probed 共Pollack et al., 2000;
Quilliet and Berge, 2001; Saeki et al., 2001; Seyrat and
v. Electrowetting Hayes, 2001; Lee, Moon, et al., 2002; Moon et al., 2002兲.
Finally, we explore electrowetting, which is probably Contact angle changes of 40° can be driven with as low
the most popular means of driving microfluidic motion as 15-V applied potentials 共Moon et al., 2002兲. Several
by modifying ␥sl 共Quilliet and Berge, 2001兲. The basic advantages stand out: 共i兲 higher potentials can be ap-
physics of electrowetting can be understood within the plied without causing undesired electrochemical reac-
picture of equilibrium wetting 关Fig. 20共a兲兴. For reviews, tions; 共ii兲 clean surfaces can be prepared, reducing sur-
see Dussan 共1979兲, de Gennes 共1985兲, de Gennes, face defects that pin the contact line and reduce
Brochard-Wyart, and Quere 共2004兲, and Eggers and reversibility; and 共iii兲 highly nonwetting surfaces can be
Stone 共2004兲. A tension 共energy/area, or force/length兲 is made that enhance the changes due to electrowetting.
associated with each interface: solid-liquid ␥sl, solid-gas Drawbacks include 共i兲 lower capacitance requires a
␥sg, liquid-gas ␥lg ⬅ ␥. Each exerts a force on the three- stronger applied potential to achieve a given contact
phase contact line, and a balance of forces gives Young’s angle change, and 共ii兲 biomolecules tend to bind nonspe-
equation, cifically to the hydrophobic surfaces typically used in
such studies 共Yoon and Garrell, 2003兲.
␥sg − ␥sl − ␥ cos ␪eq = 0, 共20兲 A common observation is that the contact angle in-
creases according to Eq. 共22兲 until some critical voltage
defining the equilibrium contact angle. Vc, at which point electrowetting saturates. 关Recently,
When a potential difference V is applied between the however, Krupenkin et al. 共2004兲 demonstrated that a
drop and the substrate, the solid-liquid interface be- superhydrophobic nanostructured surface becomes com-
haves as a capacitor, with capacitance per unit area c pletely wet with only 22 V.兴 More generally, the satura-
关Fig. 20共b兲兴. The larger the solid-liquid interface, the tion mechanism remains unknown 共Quilliet and Berge,
greater the total capacitance, which decreases the total 2001; Moon et al., 2002兲, although various mechanisms
electrostatic energy and drives the drop to spread. The have been proposed: vanishing ␥sl for potentials exceed-
free energy per unit interfacial area is lowered by an ing a critical value 共Peykov et al., 2000兲, ionization near
amount cV2 / 2, since cV2 / 2 is stored in the capacitor and the wedge-shaped contact line 共Vallet et al., 1999兲,
−cV2 is done by the battery 共Feynman et al., 1989兲. Mov- charge trapping in the dielectric layer 共Verheijen and
ing the contact line a distance ⌬z requires a virtual work Prins, 1999兲, charge injection into a polymeric layer

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

994 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Several variants on the theme of electrowetting have

been demonstrated. Surfaces have been developed that,
under a sufficiently strong applied potential, undergo an
electrochemical reaction that abruptly changes the con-
tact angle 共Sondag-Huethorst and Fokkink, 1994兲. Simi-
FIG. 21. 共Color in online edition兲 A three-electrode system larly, monolayers have been demonstrated to have elec-
splits a 200-nl drop of deionized water in 8 / 30 s using elec- trically driven conformational changes, allowing
trowetting. Scale bar is 1.4 mm. Reprinted with permission reversible wettability control 共Lahann et al., 2003兲. Last,
from Cho, Moon, and Kim, 2003. ©2003, IEEE. optoelectrowetting employs a photoconducting dielec-
tric layer that switches from dielectric to conductor
when illuminated. In this way, light triggers an enhance-
共Janocha et al., 2000兲, dielectric breakdown 共Blake et al., ment of electrowetting by effectively reducing the paral-
2000; Seyrat and Hayes, 2001兲, surface roughness 共Jan- lel plate capacitor separation 共Chiou et al., 2003兲.
ocha et al., 2000兲, water molecule adsorption
共Schneemilch et al., 2000兲, and vertical forces on the con- vi. Dielectrophoresis
tact line 共Kang, 2002兲.
Finally, we briefly discuss dielectrophoresis, in which a
A clear advantage to electrocapillary or electrowet-
dielectric body is forced either up or down a gradient in
ting is the speed with which potentials can be applied
electric field strength in order to minimize total electro-
共e.g., compared to thermal transport兲. For example, an
static energy 共Pohl, 1978兲. This phenomenon is perhaps
electrowetting display was demonstrated to have 12-ms
more familiar to physicists from optical tweezers, which
switching times, which is fast enough for video 共Hayes
represents a higher-frequency 共optical兲 regime. Because
and Feenstra, 2003兲, and a fluid column was shown to
the dielectric constant of a body is frequency dependent,
rise 20 mm over about 1 s 共Welters and Fokkink, 1998兲.
particles can be separated and manipulated using ac
Droplet manipulation is likewise fast—velocities on
fields of various frequencies 关reviewed by Pohl 共1978兲;
solid substrates of order 1 mm/ s – 10 cm/ s have been
Green, Ramos, and Morgan 共2000兲; Gascoyne and Vyk-
demonstrated 共Washizu, 1998; Pollack et al., 2000, 2002;
oukal 共2004兲兴, and drive droplet-based microanalysis
Cho, Moon, and Kim, 2003兲 and drops have been split in
platforms 共Jones et al., 2001; Gascoyne et al., 2002; Gas-
less than a second 共Cho, Moon, and Kim, 2003兲 共Fig. 21兲.
coyne and Vykoukal, 2004; Schwartz et al., 2004兲. Al-
Early work suggesting displays based on electrowet- though droplet dielectrophoresis and electrowetting are
ting 共Beni and Hackwood, 1981; Beni and Tenan, 1981兲, often discussed together, the two represent distinct
has progessed to video-speed “electronic paper” 共Hayes physical mechanisms 关see, e.g., Jones 共2002兲 and Jones et
and Feenstra, 2003; Roques-Carmes et al., 2004兲. Micro- al. 共2003兲兴.
droplets act as optical lenses, for which electrowetting
allows precise alignment and focal-length tunability
共Berge and Peseux, 2000; Krupenkin et al., 2003; Yang et c. Liquid-liquid surface tension manipulation
al., 2003兲. Furthermore, dynamically tunable optical fi- The above examples involved a droplet moving down
ber devices, switches, and waveguides have been devel- a gradient in total interfacial energy, with a speed set
oped based on electrowetting 共Jackel et al., 1982, 1983; dynamically—the energy released as the droplet moved
Mach et al., 2002; Acharya et al., 2003; Hsieh et al., 2003兲. was lost to viscous dissipation. Here, we discuss a final,
Specific microfluidic examples include electrowetting- more subtle capillary phenomenon. We consider micro-
based micropumps 共Colgate and Matsumoto, 1990; Jun fluidic motions established by the motion of fluid-fluid
and Kim, 1998; Yun et al., 2002兲, creation, manipulation, interfaces, which move in response to unbalanced
and metering of fluid droplets 共Lee and Kim, 2000; Pol- stresses. A fluid-fluid interfacial tension depends upon
lack et al., 2000; Lee, Moon, et al., 2002; Cho, Moon, and temperature, electrostatic potential, and surfactant con-
Kim, 2003; Kuo et al., 2003; Ren et al., 2004兲, and chemi- centration, and surface tension gradients can be estab-
cal reactors that merge and mix droplets 共presumably兲 lished by externally inducing a gradient in such proper-
by electrowetting 共Taniguchi et al., 2002兲. Electrowetting ties. Gradients in interfacial tension along a free surface
in multiple parallel channels gives rise to an electrocap- set the interface itself into motion, giving a fluid flow
illary pressure that can be used to drive fluid motion generally known as Marangoni flow.
共Prins et al., 2001兲, and “designer” droplet shapes can be Droplets suspended in a fluid with a background gra-
manipulated and maintained 共Mugele and Herminghaus, dient of temperature, electric potential, or surfactant
2002; Bienia et al., 2003兲. In biofluidic chips, biomol- concentration will experience thermocapillary 共Young et
ecules have a tendency to passively adsorb onto hydro- al., 1959; Chabbra and De Kee, 1992兲, electrocapillary
phobic surfaces 共Absolom et al., 1987; van der Vegt et al., 共Newman, 1991兲, or solutocapillary 共Levich, 1962; Levich
1996; Makohliso et al., 1998兲, reducing reversibility in and Kuznetsov, 1962; Levich and Krylov, 1969兲 motion
electrowetting devices. Voltage strategies have been de- 共Fig. 22兲. Below, we discuss microfluidic applications of
veloped to reduce this surface fouling 共Yoon and Gar- these free-surface flows. As above, viscous stresses bal-
rell, 2003兲. Finally, electrowetting increases the loading ance capillary stresses, and thus their motion is set by
efficiency of biological microcantilever arrays 共Belaubre the capillary number Ca⬃ 1, or U ⬃ ⌬␥ / ␩. For ther-
et al., 2003兲. mocapillary motion, ⌬␥ ⬃ 共d␥ / dT兲共ⵜT兲R, for solutocap-

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 995

The most popular microfluidic Marangoni flow is ther-

mocapillary. Homsy and co-workers completed experi-
mental 共Lajeunesse and Homsy, 2003兲 and theoretical
共Mazouchi and Homsy, 2000, 2001兲 studies of ther-
mocapillary bubble motion through capillaries which
built upon two classic dynamic free-surface problems:
the Bretherton problem involving the motion of a
bubble through a pipe 共Bretherton, 1961兲 and the
Landau-Levich problem 共Landau and Levich, 1942兲 in-
volving the wetting film left on a plate pulled out of a
liquid bath.
Many microfluidic thermocapillary flows occur on
solid walls, where temperature variations in both ␥sl and
␥ drive fluidic motion 共Brochard, 1989; Brzoska et al.,
1993兲. In addition to droplet motion, thermocapillary
motion has been used to drive thin liquid films to spread;
classic references are Levich 共1962兲 and Ludviksson and
Lightfoot 共1971兲. Troian and co-workers have experi-
FIG. 22. 共Color in online edition兲 Thermocapillary, electrocap- mentally, theoretically, and numerically explored ther-
illary, and solutocapillary droplet motion occurs due to surface mocapillary microfluidic flows along patterned hydro-
tension gradients arising from 共a兲 thermal gradients in the philic stripes 共Kataoka and Troian, 1999; Darhuber,
background solution that drive droplets to warmer tempera-
Davis, et al., 2003兲 and the mixing that accompanies
tures; 共b兲 electric fields that drive droplets away from the simi-
these flows 共Darhuber et al., 2004兲, complementing stud-
larly charged electrode; and 共c兲 background surfactant gradi-
ies of thermocapillary contact line instabilities and their
ents that lead droplets to move towards regions of greater
surfactant concentration.
control 共Kataoka and Troian, 1997, 1998兲. An optical
means 共via absorption兲 to drive thermocapillary motion
was recently demonstrated, allowing feedback control
illary motion, ⌬␥ ⬃ 共d␥ / d␳c兲共ⵜ␳c兲R, and for electrocapil- methods to suppress these instabilities 共Garnier et al.,
lary motion, ⌬␥ ⬃ 共d␥ / dV兲共ⵜV兲R ⬃ 共d␥ / dV兲qER, where 2003兲. As with thermocapillary droplet motion, capillary
q is the interfacial charge. forces balance viscous forces in motion along stripes, al-
though an additional length scale arises 共the droplet
Electrocapillary motion 共Frumkin and Levich, 1945;
height h depends on stripe width w, but may be set dy-
Newman, 1991兲, and the essentially identical continuous
namically and thus depend on Ca兲. Various scaling rela-
electrowetting 共Beni et al., 1982兲 have been little used in
tions can be derived, and velocities up to 600 ␮m / s were
microfluidics other than for droplet manipulation 共Lee
observed for liquid poly共dimethysiloxane兲 共PDMS兲
and Kim, 2000兲 and in a pump design 共Yun et al., 2002兲.
drops 共Darhuber, Davis, et al., 2003兲. High-resolution
Likewise, solutocapillary flows 共Levich, 1962; Levich
temperature distributions 共Darhuber et al., 2002兲 have
and Kuznetsov, 1962; Levich and Krylov, 1969兲 have enabled thermocapillary manipulation of droplets, in-
seen little use in microfluidic devices, in part because cluding droplet transport and splitting 共Fig. 23; Darhu-
maintaining a surfactant concentration gradient is not ber, Valentino, et al., 2003兲.
straightforward. However, Abbott and co-workers have Finally, we note that surface tension gradients due to
explored molecules that can be electrochemically thermal gradients form the basis for Marangoni-Benard
switched between surfactant and nonsurfactant states. In convection 共Davis, 1987兲. Recently, a ratchet pump was
effect, electrochemical reactions create surfactant mol- demonstrated by driving Marangoni-Benard convection
ecules at one electrode and destroy them at the other, above an asymmetrically patterned substrate 共Stroock et
which set up the necessary concentration gradients to al., 2003兲.
drive Marangoni flows 共Gallardo et al., 1995, 1996; Ben-
nett et al., 1996; Gallardo and Abbott, 1997兲. This tech-
D. The Weissenberg and Deborah numbers: Let the
nique can be used in closed channels by using a gel-fluid
mountains flow
free surface rather than the fluid-air interface of open
channels 共Lee et al., 1999兲. Such electrochemical ma- Thus far, we have considered the influences of inertia,
nipulation allows droplet and particle transport as well diffusion, and surface tension on microfluidic flows. Dis-
as thin-film patterning 共Gallardo et al., 1999兲. Similar solved polymers add an elastic component to the fluid
techniques involve solution-based reducing agents 共Jong that further enriches flow behavior. Microfluidic devices
and Abbott, 1998兲 or photosensitive surfactants 共Shin have proved useful for such systems in two senses: first,
and Abbott, 1999兲 rather than electrochemical reactions. to study the effects of fluid flow on polymer behavior,
Last, electrochemically switchable surfactants have been and second, to study the effects of polymers on the be-
used to selectively solubilize and deposit compounds on havior of fluid flows. In both cases, microfluidic devices
electrodes, providing a technique for particle separation simplify the experiments by eliminating inertial effects
共Rosslee and Abbott, 2001兲. that compete or mask the physics of interest. Further-

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

996 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

= u0 + ėz. The bead separation Rp changes due to three

physical effects: 共i兲 Brownian motion drives beads apart
with average speed RpṘp ⬃ D ⬃ kBT / ␰, 共ii兲 the spring
pulls the beads together with velocity Ṙp ⬃ −kHRp / ␰, and
共iii兲 the extensional flow advects the forward bead more
quickly than the rear bead, driving them apart with ve-
locity Ṙp ⬃ ėRp 共Bird, Hassager, et al., 1987; Larson,
1999兲. Each effect has its own characteristic time scale:
␶D = R20 / D for the beads to diffusively explore a length
scale R0, ␶p = ␰ / kH for the polymer spring to relax, and
␶e = ė−1 is a time scale associated with the extensional
flow.
While this example helps in developing physical in-
sight and intuition for microfluidics, its applicability is
generally limited. It assumes polymers do not interact,
which requires the solution to be dilute. The solution
FIG. 23. Using embedded microheaters to establish thermal stress is assumed to be dominated by the viscous solvent,
gradients, a dodecane drop of width⬃ 1 mm 共a兲–共e兲 is split in rather than the polymers, constituting a so-called Boger
8.5 s, 共f兲–共i兲 travels ⬃6 mm in 104 s, and 共j兲–共l兲 turns a corner in fluid 共Boger, 1977; Boger and Yeow, 1992兲. Polymer de-
164 s. Reprinted with permission from Darhuber, Valentino, et formations are also assumed to be small since we have
al., 2003. ©2003, AIP. ignored the nonlinear elasticity for large deformations.
Furthermore, the simplest form for the polymer-solvent
interaction has been assumed, whereas in general this
more, significant shear rates can be readily obtained ow-
interaction is conformation dependent.
ing to the small dimensions of microfluidic devices,
Despite these simplifications, several important phe-
which provides a strong but controllable stretching force
nomena are illustrated by the above simple example.
to distort polymer molecules and probe their elastic
qualities. The Weissenberg number. In equilibrium ė = 0, spring
First, however, we turn to a simple picture of polymer forces balance Brownian forces to give a characteristic
dynamics to illustrate the basic physics at hand. The sim- polymer size R0 ⬃ 共kBT / kH兲1/2 ⬃ 共Nb2兲1/2. By contrast, an
plest model system 共Kuhn, 1934兲 treats a polymer as a extensional flow acts to drive the beads apart 共like a
dumbbell with two beads each of hydrodynamic resis- negative spring兲, and alters the steady polymer size via
tance ␰ connected by a Hookean 共entropic兲 spring of
stiffness kH. The spring constant kH can be obtained for
a freely jointed chain with N steps of length b as kH
Rp共ė兲 ⬃ 冉 k BT
kH − ė␰
冊 1/2
⬃
R0
共1 − Wi兲1/2
. 共23兲

= 3kBT / Nb2. For simplicity, we confine the polymer to Here we have introduced the Weissenberg number
the 共one-dimensional兲 line along the center of a con-
Wi = ␶pė or ␶p␥˙ , 共24兲
tracting microchannel, as in Fig. 24. Due to the contrac-
tion, the local 共extensional兲 flow felt by the polymer in- which relates the polymer relaxation time to the flow
creases as it moves down the channel, roughly like uz deformation time, either inverse extension rate ė−1 or
shear rate ␥˙ −1. When Wi is small, the polymer relaxes
before the flow deforms it significantly, and perturba-
tions to equilibrium are small. As Wi approaches 1, the
polymer does not have time to relax and is deformed
significantly.
The Deborah number. Another relevant time scale
␶flow characteristic of the flow geometry may also exist.
For example, a channel that contracts over a length L0
introduces a geometric time scale ␶flow = L0 / U0 required
FIG. 24. The effect of a nonuniform flow on a model polymer, for a polymer to traverse it. Likewise, an oscillatory flow
represented as two beads separated by a distance R on a introduces an oscillation time ␶flow ⬃ ␻−1. The flow time
spring. 共a兲 The center of mass of the polymer moves with ve-
scale ␶flow can be long or short compared with the poly-
locity UCM, whereas the two beads experience relative motion
UR ⬃ ėRp − kHRp / ␰ due to external flow gradients and spring
mer relaxation time ␶p, resulting in a dimensionless ratio
forces. Brownian forces balance spring forces to give a steady- known as the Deborah number
state polymer size. When Wiⲏ O共1兲, the extensional flow over- De = ␶p/␶flow . 共25兲
whelms the spring, and the polymer unravels via a coil-stretch
transition. 共b兲 As the beads are pulled toward the spring, each Amusingly, the Deborah number is named for one of the
exerts a force back on the fluid, resulting in a force-dipole flow earliest fluid mechanics, the prophetess Deborah
and an anisotropic contribution to the normal stress. 共Reiner, 1964兲, who observed that “…the mountains

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 997

关For a review of contraction flows of viscoelastic fluids,

see Boger 共1987兲.兴
The coil-stretch transition. As shear rates increase and
Wi becomes O共1兲, fluid forces overcome the restoring
forces of the entropic spring, causing the polymer to un-
ravel to a stretched state 共de Gennes, 1974; Hinch, 1977;
Bird, Hassager, et al., 1987; Larson and Magda, 1989;
Schroeder et al., 2003兲. In fact, the model Hookean
spring used here extends indefinitely, which is clearly un-
physical. Various remedies are employed in flow simula-
tions, such as the finitely extensible, nonlinearly elastic
共FENE兲 dumbbell 共Bird, Hassager, et al., 1987; Larson,
1988, 1999; Rallison and Hinch, 1988兲. Nonlinear elastic
effects follow from many polymer models; single-
molecule experiments indicate that the wormlike chain
captures DNA elasticity particularly well 共Smith et al.,
1992; Bustamante et al., 1994; Vologodskii, 1994; Marko
FIG. 25. 共Color in online edition兲 Microfluidic devices allow
and Siggia, 1995; Meiners and Quake, 2000兲.
new regimes of Re-De space to be explored for non-
Normal stress differences. Polymers introduce an an-
Newtonian flows 共Rodd et al., 2004兲. Typical flow behaviors are
isotropic component to the normal stress: an extended
shown for two different semidilute PEO solutions flowing into
a 400– 25-␮m contraction at two different elasticity numbers polymer exerts a tension along its axis of extension, so
共El= 9 and 89兲. Experiments with increasing flow rates thus that the stress on a surface perpendicular to the exten-
correspond to straight line trajectories in Re-De space. Previ- sion is lower than that on the surfaces parallel. Shear
ous studies were limited to El extremes—highly elastic Boger flows are similar but more complicated, since there exist
fluids 共ElⰇ 1兲 and shear-thinning viscous fluids 共ElⰆ 1兲. Figure three different normal stresses: in the flow direction, in
courtesy of L. Rodd. the gradient direction, and in the vorticity direction 共per-
pendicular to both兲, giving two normal stress differences.
These normal stress differences are responsible for
flowed before the Lord…” 共Judges 5:5兲. Implicit in her many of the remarkable properties of non-Newtonian
statement is the recognition that the rheological re- fluids 共Bird, Armstrong, and Hassager, 1987兲.
sponse of a material depends on the time scale over Flow modifications. Neglecting Brownian motion,
which it is observed: mountains flow on geological time each bead is advected with the flow and driven by the
scales 共DeⰆ 1兲, but are rigid on human time scales 共De restoring spring force. Advection involves moving with
Ⰷ 1兲. the local fluid environment, and thus exerts no force on
Note that the usage of De and Wi can vary: some the fluid. By contrast, the spring must exert a force to
references use Wi exclusively to describe steady shear pull the bead through its environment. Thus in the far
flows and use De for the general case, whereas others field, each deformed polymer contributes a force-dipole
use Wi for local flow time scales due to a local shear rate stresslet flow to the background. These stresslets form
and De for global flow time scales due to a residence the basis for a flow computational technique 共Phillips,
time in the flow. 2003兲. In the dilute, weakly elastic limit, this perturba-
The elasticity number. As the flow velocity U0 in- tion is small. As elastic effects become stronger, how-
creases, elastic effects become stronger and De and Wi ever, these perturbations become significant as well, and
increase. However, the Reynolds number Re increases the coupling between polymeric stress and fluid flow be-
in the same way, so that inertial effects become more comes strongly nonlinear.
important as well. The elasticity number Flow memory. Flows strain suspended polymers,
De ␶p␩ which exert an elastic stress back on the fluid. The finite
El = = , 共26兲 polymer relaxation time ␶p gives memory to the fluid, as
Re ␳h2
the conformation of each polymer depends on the his-
where h is the shortest dimension setting the shear rate, tory of the flow. This stands in contrast with Newtonian
expresses the relative importance of elastic to inertial Stokes flows, which are uniquely determined at any in-
effects. Significantly, El depends only on geometry and stant by boundary conditions and external forcing.
material properties of the fluid, and is independent of Constitutive equation for macroscale flows. The sus-
flow rate. The small dimensions h inherent in microflu- pension of Hookean dumbbells in a viscous solvent is
idic devices allow high-El flows to be established more one of very few models for polymeric microstructure
easily; comparably elastic flows in macroscopic experi- that yields a closed-form macroscopic constitutive equa-
ments would require significantly more viscous solvents. tion, the Oldroyd-B equation 共Oldroyd, 1950; Bird, Has-
As a result, microfluidic devices have opened up hereto- sager, et al., 1987; Larson, 1999兲. This equation accounts
fore unexplored regions in Re-De space for viscoelastic for the coupled evolution of the fluid flow and the poly-
flows. For example, Rodd et al. 共2004兲 explore new re- meric stress tensor, and has been called the hydrogen
gimes in viscoelastic contraction flows, shown in Fig. 25. atom of elastic fluids. Although shortcomings certainly

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

998 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

exist such as the unbounded high-Wi extension de-

scribed above, the Oldroyd-B equation provides the
simplest description of elastic fluids and captures many
features of elastic flow behaviors qualitatively and semi-
quantitatively. However, note that when device dimen-
sions get very small on the order of the length of
stretched polymers, such continuum models are not
likely to be appropriate to describe such flows.

1. Microfluidic single-polymer studies

The low Reynolds numbers inherent in microfluidic
devices allow one to study the dynamics of individual
polymer molecules in precisely defined flows. This per-
mits direct tests of models for polymer dynamics, and FIG. 26. 共Color in online edition兲 Single fluorescently labeled
complements parallel studies using Brownian dynamics. DNA molecules subjected to a microfluidic elongational flow
Furthermore, single-molecule studies 共experimental and allow a direct visualization of the molecular individualism of
theoretical兲 reveal significant fluctuations that would be polymers, which unravel in a number of characteristic modes.
difficult or impossible to measure from bulk studies. A Four such modes are shown evolving left to right. Inset: exten-
central goal of research into complex fluids involves re- sional flow established at a microchannel cross junction. Re-
lating macroscale behavior to microscale structure, and printed with permission from Perkins et al., 1997. ©1997,
microfluidic devices show considerable potential to- AAAS.
wards this end.
A variety of single-molecule polymer studies have energetically similar conformations shown in Fig. 26 共de
been performed in microfluidic devices. Early single- Gennes, 1997; Perkins et al., 1997兲. The coil-stretch tran-
polymer studies did not involve a flow so much as a sition was predicted to be hysteretic because the friction
polymer held fixed in a moving bath. Using these tech- ␰ increases as the polymer unravels and thus ␶p in-
niques, the elasticity 共Smith et al., 1992兲 and relaxation creases; thus a lower extension rate ė is required to
times of individual DNA molecules were measured 共Per- maintain a stretched polymer 共de Gennes, 1974; Hinch,
kins et al., 1994, 1995兲, with complementary simulations 1977兲. On the other hand, the strong increase of kH with
共Quake, 1994; Larson et al., 1997兲. These experiments polymer extension shortens the relaxation time, which
demonstrated that internal hydrodynamic interactions could mask or overwhelm any hysteresis in most practi-
remain significant even at large extensions. Experiments cal situations 共Meiners and Quake, 2000兲. However, hys-
on single DNA molecules in shear flows revealed an av- teresis was observed in single-molecule experiments in
erage polymer shape that was approximately ellipsoidal, shallow channels 共Schroeder et al., 2003兲.
as expected. Individual polymers, however, showed mo-
lecular individualism 共de Gennes, 1997兲: large, aperiodic
2. Elastic instabilities
extension and tumbling motions, seen both in experi-
ments 共Smith et al., 1999; Doyle et al., 2000; Ladoux and Not surprisingly, non-Newtonian fluids display a host
Doyle, 2000兲 and in Brownian dynamics simulations of flow instabilities, even in the absence of inertia 共Pear-
共Liu, 1989; Hur et al., 2000兲. Likewise, the relation be- son, 1976; Petrie and Denn, 1976; Larson, 1992兲. Histori-
tween microscopic and macroscopic polymeric fluid be- cally, quantitative understanding has been difficult, in
havior was investigated in a microfluidic shear cell by part because typical non-Newtonian fluids have a shear-
fluorescently labeling a small number of probe polymers rate dependent viscosity 共most frequently shear thin-
in a solution of identical, invisible polymers 共Babcock et ning兲. That shear thinning would exacerbate instabilities
al., 2000兲. Individual DNA molecules in Poiseuille flow makes sense: locally increasing the shear rate lowers the
in a small capillary 共Manneville et al., 1996兲 unravel in a local viscosity, which makes the fluid easier to shear, and
manner consistent with the stem and flower theory for so on. It is thus frequently difficult to tease apart the
polymer extension 共Brochard-Wyart, 1995兲. Microfluidic effects of shear thinning and elasticity. However, the dis-
flows have been employed to stretch DNA and tether it covery of highly elastic, but constant-viscosity, Boger flu-
at both ends, allowing the transverse fluctuations to be ids 共Boger, 1977; Boger and Yeow, 1992兲 has facilitated
studied 共Crut et al., 2003; Gueroui et al., 2003兲. the study of purely elastic effects.
Polymer dynamics in extensional flows were studied Instabilities have been discovered that are purely elas-
using crossed microchannels, as shown in Fig. 26 共Per- tic in nature, occurring at a critical De or We, even as
kins et al., 1997; Smith and Chu, 1998; Schroeder et al., Re→ 0 共McKinley et al., 1996; Shaqfeh, 1996兲. Such elas-
2003兲 or flow focusing 共Wong et al., 2003兲, and were tic instabilities were originally identified in Taylor-
complemented by Brownian dynamics studies 共Larson et Couette flow 共Muller et al., 1989; Larson et al., 1990,
al., 1999; Maroja et al., 2001; Hur et al., 2002兲. These 1994; Shaqfeh et al., 1992; Baumert and Muller, 1995;
studies revealed a more subtle coil-stretch transition Graham, 1998兲, and occur additionally in Dean flow 共Joo
than originally envisioned, due in part to the multiple and Shaqfeh, 1991, 1992兲, lid-driven cavity flow 共Pakdel

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 999

and McKinley, 1996兲, cone-plate and coaxial disk flows

共Phan-Thien, 1983, 1985; McKinley et al., 1991, 1995;
Olagunju, 1995兲, rotating parallel plates 共McKinley et al.,
1991; Byars et al., 1994兲, in the wake behind a confined
circular cylinders 共McKinley et al., 1993兲, and in free sur-
faces that buckle 共Kumar and Graham, 2000; Graham,
2003兲. The influence of viscous heating on this instability
introduces an additional dimensionless parameter 共the
Nahme number兲, whose influence has been examined
theoretically and experimentally 共Al-Mubaiyedh et al.,
1999; White and Muller, 2000, 2003; Rothstein and
McKinley, 2001兲.
Curved streamlines are an almost ubiquitous ingredi-
ent of elastic instabilities. Generically, polymers that
have been extended along curved streamlines are taken
by fluctuations across either shear-rate gradients or ten-
sion gradients in the base state. This couples the hoop
stress ␴␪␪ acting along the curved streamline to the ra-
dial and axial flows, which can amplify the perturbation.
Such elastic instabilities in a saturated nonlinear state
give rise to irregular fluctuations and elastic turbulence
共Groisman and Steinberg, 2000, 2004; Larson, 2000兲,
which has been exploited to enhance microfluidic mixing
共Groisman and Steinberg, 2001兲.
Only recently have such elastic nonlinearities been ex-
ploited in microfluidic flows. Efforts in decades past to
design and construct nonlinear fluidic logic elements re-
sulted in such common items as pulsating shower heads,
windshield wipers, and sprinkler systems 共Humphrey FIG. 27. 共Color in online edition兲 Nonlinear viscoelastic flow
and Tarumoto, 1965; Foster and Parker, 1970; Joyce, resistor. While a low-Re Newtonian fluid would exhibit a linear
1983兲. Attempts to build miniaturized fluidic computers, pressure-flow relationship, elastic stresses in dilute polymer so-
however, foundered, since inertial nonlinearities vanish lutions introduce significant nonlinearities, even at very small
as devices were scaled down and Re→ 0. Elastic insta- Re. 共a兲 Elastic nonlinearities cause an increasingly complex
bilities occur at vanishingly small Reynolds number, and flow as pressure increases. A vortex forms on the upstream
can reintroduce flow nonlinearities in microfluidic de- side of the contraction between applied pressures 24 and 36
vices. With microfluidic computing in mind, several non- Pa, which corresponds to the elbow in the flow rate vs pressure
linear microfluidic flow elements have been designed plot 共b兲. The contraction vortex extends upstream and in-
creases in size as pressure increases. 共b兲 The flux vs pressure
that exploit elastic nonlinearities: a nonlinear fluid resis-
profile is linear for low applied pressures, but flattens for ap-
tor 共Fig. 27兲, a bistable flip-flop memory element 共Fig.
plied pressures above a critical value 共where Wi⬎ 1兲, creating
28兲, and a flow rectifier 共Fig. 29; Groisman et al., 2003; an approximately constant current source. Reprinted with per-
Groisman and Quake, 2004兲. mission from Groisman et al., 2003. ©2003, AAAS.
We consider first the nonlinear flow resistor 共Fig. 27兲.
Newtonian low-Re flows vary linearly with pressure, re-
gardless of channel geometry. By contrast, dilute poly-
mer solutions behave linearly only for small applied
pressures, for which WiⰆ 1 and polymers deform only
slightly. Above a threshold pressure, however, the non-
linear flow resistor passes a nearly constant flow of elas-
tic fluid over nearly a decade increase in applied pres-
sure. This constant flow regime occurs once Wi⬎ 1,
which can be understood qualitatively from energy argu-
ments. Under such strong flows, additional power input
that would normally increase the flow rate is instead
consumed by further deformation of the polymers FIG. 28. Nonlinear viscoelastic flip-flop memory device. Dilute
共Groisman et al., 2003兲. This deformation is stored en- polymer solutions 共one dark, one light兲 that enter a junction
tropically but is not returned to the flow due to the hys- from opposite sides choose an exit channel, contrasting with
teretic differences between coil stretching and relax- Newtonian fluids that split to flow through both. The fluids
ation. flow stably through the exit they have chosen, but switch when
The fluidic flip-flop memory element is bistable, with a transient 共50 ms兲 pressure pulse is applied. Reprinted with
oppositely directed fluids into a four-channel junction permission from Groisman et al., 2003. ©2003, AAAS.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1000 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

FIG. 29. 共Color in online edition兲 Viscoelastic fluidic rectifier. Low-Re Newtonian fluid flows are reversible and indifferent to the
flow direction. By contrast, elastic stresses in dilute polymeric solutions partially rectify such flows, leading to quantitative and
qualitative differences. 共a兲–共d兲 show dilute polymer solution flows in the forward direction, and 共e兲–共h兲 in the reverse direction. 共i兲
Above a critical pressure, flow rates in the forward direction begin to exceed those in reverse, giving a flux ratio 共inset兲 that can
be as high as 2. Reprinted with permission from Groisman and Quake, 2004.

共Fig. 28兲. Macroscopic flows through a related crossed ences needed to drive a given flow rate forward and
geometry were studied by Cochrane et al. 共1981兲, albeit backward differ by as much as a factor of 2.
without mention of bistability or applications. If the flu-
ids were Newtonian, each stream would simply split and
3. Controlled polymer deformation for separations
flow evenly into both outlet channels. Non-Newtonian
fluid streams, on the other hand, break symmetry and Separating biopolymers like DNA and proteins is of
choose one of the two. The junction remembers this utmost importance for many applications and disci-
state due to the stability of the flow, but can be switched plines, including biology, biochemistry, forensic science,
using a transient pressure shock 共Groisman et al., 2003兲. drug design, genome sequencing, proteomics, and so on.
This effect is likely related to the “tubeless siphon” phe- As such, many techniques have been developed to sepa-
nomenon exhibited by polymeric fluids 共Bird, Arm- rate polymers by size, of which various electrophoretic
strong, and Hassager, 1987兲, in which polymers stretched techniques are standard. While we discuss electrokinetic
and aligned with the flow exert an elastic stress that effects in more detail in Sec. III.A, we note here an
tends to pull neighboring molecules in the flow direc- important and rather surprising property of electro-
tion. phoresis: the electrophoretic mobility ␮E of an object
Last, we examine the flow rectifier shown in Fig. 29. 共relating its velocity to applied field strength, u = ␮EE兲 is
Linear, low-Re Newtonian flows behave identically independent of its size. Among many counterintuitive
when forced forward or backward through a channel. consequences is that electrophoresis in aqueous solution
Polymeric flows, on the other hand, can be rectified by can separate objects by surface charge but not size.
elastic stresses, whereupon an applied pressure ⌬p As a result, electrophoretic polymer separations are
drives a different flow rate in one direction than the typically performed in gels, wherein the resistance to
other. The channel asymmetry causes advecting poly- polymer motion increases for longer polymers. Viovy
mers to experience a different flow history for forward 共2000兲 and Slater et al. 共2002, 2003兲 provide excellent
and backward flows, which changes the molecular re- reviews of the physics behind polymer electrophoresis,
sponse and macroscopic stress. This leads to a partial Ugaz et al. 共2004兲 provides a nice overview of various
rectification in the device whereby the pressure differ- techniques in a microfluidic context, and Tegenfeldt et

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1001

FIG. 30. DNA molecules of different lengths naturally follow different trajectories through a pulsed-field electrophoretic DNA
prism. 共A兲 An applied electric field extends polymer molecules in the direction of the field at time t0. When an electric field is
applied at an oblique angle 共t1兲, shorter polymers encounter fewer obstacles and travel more easily along the second field direction
than longer polymers. 共B兲–共D兲 Separation of 共1兲 61 kb, 共2兲 114 kb, 共3兲 158 kb, 共4兲 209 kb DNA with fields of varying strength. 共B兲
250-ms pulses with field strengths 32 and 20 V / cm are strong enough to separate the DNA into two bands. 共C兲 Stronger fields 共240
and 150 V / cm, 40-ms pulses兲 separate the DNA into four distinct bands. 共D兲 Microfludic channels at the outlets collect the
separated DNA. Reprinted with permission from Huang, Tegenfeldt, et al., 2002. ©2002, NPG.

al. 共2004兲 describes separation efforts using microfabri- electrophoretic DNA prism 共Huang, Tegenfeldt, et al.,
cated matrices. We do not attempt to cover the tremen- 2002兲. Because smaller fragments travel further along
dous effort towards microfluidic separations, but rather the secondary field direction than longer ones, the angle
refer the reader to existing reviews focusing on such de- of travel of a polymer depends on its length 共analogous
vices 共Bousse et al., 2000; Bruin, 2000; Auroux et al., to light in a prism兲, allowing fragments to be sorted by
2002兲, on electrophoretic separation techniques for length into an array of microchannels. Theoretical ef-
DNA 共Dolnik et al., 2000; Viovy, 2000; Slater et al., 2002; forts include the Taylor dispersion of DNA in such sys-
Ashton et al., 2003; Barbier and Viovy, 2003; Landers, tems 共Dorfman and Viovy, 2004兲 and Brownian dynam-
2003; Sartori et al., 2003兲, and for proteins 共Dolnik, 1997, ics simulations. The latter suggest that randomly placed
1999; Dolnik and Hutterer, 2001; Figeys and Pinto, 2001; posts are better suited for polymer separations than
Hancock et al., 2002; Lion et al., 2003兲. regular arrays, in the dilute limit 共Patel and Shaqfeh,
Different electrophoretic strategies utilize differing 2003兲. These and other techniques 关such as entropic
degrees of polymer deformation and relaxation, and can trapping and separation 共Turner et al., 2002兲兴 are re-
thus be connected to Wi and De. Polymers with size R0 viewed by Slater et al. 共2002兲.
that are small relative to the gel’s pore size bg do not Finally, we note that these microfabricated array stud-
deform significantly and flow freely without separating, ies complement single-molecule studies of DNA under
whereas collisions and entanglement between the poly- flow, because holding a polymer in place against electro-
mer and gel depend strongly on polymer length when phoretic driving mimics holding a polymer in place
R0 ⬃ bg, causing strong separation. However, when R0 against a background flow 共Long et al., 1996; Bakajin et
al., 1998兲. Furthermore, thin “slit” channels allow the
Ⰷ bg, the polymer must reptate in a highly extended
effect of molecular confinement on polymers to be stud-
state to move, and once again its mobility is insensitive
ied, both experimentally, and theoretically 共Brochard
to length. Periodically changing the field direction
and de Gennes, 1977; Bakajin et al., 1998; Dimalanta et
共pulsed-field electrophoresis兲 exploits the length-
al., 2003; Jendrejack, Jendrejack, Schwartz, et al., 2003;
dependent drag when such highly extended polymers
Jendrejack et al., 2004; Woo et al., 2004a, 2004b兲.
are driven perpendicular to their extension.
Through microfabrication, a well-characterized set of
obstacles can be created to allow a more systematic E. The Grashof and Rayleigh numbers: Tracer-driven flows
study of the physics of polymer electrophoresis, with the
aim of rational and systematic improvement of separa- Thus far, gravity has played no role in our discussion.
tion systems 共Volkmuth and Austin, 1992; Austin and If fluids of different densities are brought together, how-
Volkmuth, 1993; Chou et al., 2000; Tegenfeldt et al., ever, gravity may play a role. In order to understand,
2004兲. Systems have been developed to exploit con- design, and utilize such systems, one must understand
trolled polymer deformation about microfabricated the effects that density differences have on fluid behav-
posts 共Volkmuth and Austin, 1992; Austin and Volk- ior. To explore these effects, consider the channel shown
muth, 1993; Volkmuth et al., 1994; Duke et al., 1996, in Fig. 31. The fluid in one half of the channel contains
1997; Chou et al., 2000; Bakajin et al., 2001; Huang, dense solute molecules at a concentration c, whereas the
Tegenfeldt, et al., 2002兲 or self-assembling posts of mag- fluid in the other half is clear. We assume solution den-
netic beads 共Doyle et al., 2002兲. For example, Fig. 30 sity differences vary linearly with solute concentration
depicts a microfabricated array used as a pulsed-field differences via

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1002 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

One might complain that the Rayleigh and Grashof

numbers are little more than Reynolds and Péclet num-
bers in slight disguise. The criticism is valid, since the
two pairs of dimensionless numbers compare the same
fundamental effects. The distinction lies in the fact that
in buoyancy-driven flows, the tracers are themselves the
agents responsible for driving the flow. Systems contain-
ing truly passive tracers whose behavior is determined
FIG. 31. 共Color in online edition兲 Free-interface diffusion in by Pe, differ significantly from those containing buoyant
protein crystallization. 共a兲 Initial configuration. The interface or dense tracers. In other words, a high Pe flow looks
evolution for varying regimes of buoyant flow: 共b兲 RaⰆ 1, Gr very different than a high Ra flow.
Ⰶ 1, where diffusion dominates and the interface spreads dif- In fact, a similarly artificial distinction was made with
fusively; 共c兲 RaⰇ 1, GrⰆ 1, where advective transport domi- less fanfare between the Péclet number and the Rey-
nates and tongues of each fluid are driven into the other; and nolds number itself. After all, the Reynolds number can
共d兲 RaⰇ 1, GrⰇ 1, where inertia drives the gravity currents to
be viewed as a Péclet number for vorticity: at low Re,
splash.
Stokes flow solves the vorticity diffusion equation,
whereas advection of vorticity becomes more significant
⌬␳ = ␳0␣c⌬c, 共27兲 as Re increases. With Re, just as with Ra and Gr, the
distinction with Pe is relevant precisely because the
where ␳0 is the density of the clear liquid. An analogous tracer, whether vorticity or density, acts back on the flow.
system involving thermal 共rather than solute兲 gradients For this reason, these numbers are related,
is obtained when a temperature field ⌬T replaces the
Pe Ra ␯
solute concentration field ⌬c, in which case ␣c = ␣T is the ⬅ = , 共31兲
coefficient of thermal expansion. Re Gr D
The system in Fig. 31 is clearly unstable, as buoyant and their ratios reflect the ratios of diffusivities for the
forces drive the dense fluid downhill into the less dense relevant tracers. The ratio of vorticity diffusivity ␯ to
fluid, and vice versa. The system could evolve in several solute diffusivity D is called the Schmidt number Sc
ways, depending on channel height h and fluid proper- = ␩ / D, whereas the ratio of ␯ to thermal diffusivity ␬T is
ties. At issue is the fact that buoyancy-driven fluid ve- the Prandtl number Pr= ␯ / ␬T.
locities are not known at the outset, but are instead de- In what follows, we explore a few examples where
termined by the distribution of buoyant forces acting on microfluidics exploits or allows one to neglect the effects
the fluid. To obtain a scaling estimate, we assume a qua- of buoyancy.
sisteady force balance, where viscous forces 共fv
⬃ ␩Ub / h2兲 balance buoyant forces 共fb ⬃ ␣c␳0⌬cg兲 to give
a buoyant velocity scale 1. Protein crystallization

␣c␳0⌬cgh2 Protein crystallization is an application of tremendous

Ub ⬃ . 共28兲 importance to structural biology and biotechnology 关see,
␩ e.g., McPherson 共1999兲兴, and one in which density gradi-
With the buoyant velocity scale Ub, we can determine ents in solution can play a significant role. Although
whether convection or diffusion dominates the solute DNA sequencing reveals the amino acid sequence of
transport. The ratio of convective to diffusive flux is proteins, little to nothing can be said about the protein’s
given by actual function unless its folded structure is known. Un-
fortunately, despite decades of research, predicting the
convective flux Ubh ␣c␳0⌬cgh3 structure of the protein from its sequence remains noto-
= ⬃ ⬅ Ra, 共29兲
diffusive flux D ␩D riously difficult, even for the smallest of proteins. At
present, the most popular method for resolving such
and is termed the Rayleigh number Ra. When RaⰆ 1,
structures involves x-ray crystallography, which requires
diffusion dominates over advection and the interface
large, high-quality protein crystals to be grown. This,
spreads diffusively, as shown in Fig. 31共b兲. If, on the
too, has proven difficult, in part because the parameter
other hand, the flow is fast and outcompetes diffusion
space is so vast 共temperature, protein concentration,
共RaⰇ 1兲, then a gravity current of each fluid penetrates
precipitant identity and concentration, pH, etc.兲. Protein
the other, as shown in Fig. 31共c兲. Figure 31共d兲 shows an
crystallization is typically attempted using an intuitively
even faster flow, where inertia causes the fingers to
guided but massive sweep through parameter space—
splash. For this to occur, inertial forces must be signifi-
essentially trial and error. Such extensive parameter
cant relative to viscous forces, which requires the
sweeps consume large quantities of time, effort, and
Grashof number Gr, defined by
material—which may be expensive, difficult, or even im-
fi ␳0Ubh ⌬c␣c␳20gh3 possible to acquire.
⬃ ⬃ ⬅ Gr, 共30兲 Microfluidics is ideally suited for such a task, as it al-
fv ␩ ␩2
lows automated, parallel sweeps through parameter
to be large. space using far less material than conventional

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1003

FIG. 33. Interferometric images of crystal melting. 共a兲 Dense

fluid is formed in the vicinity of a dissolving crystal and falls in
a wake, indicating high Gr. Crystal growth suffers the opposite
problem: buoyant fluid forms around the crystal as dense mol-
ecules are locally removed from solution, establishing a con-
vective flow that adversely affects crystal quality. 共b兲 Dissolu-
tion occurs more isotropically when the crystal is held in a
0.1% 共weight/volume兲 agarose gel, which suppresses convec-
FIG. 32. 共Color in online edition兲 Low Gr protein- tion. Reprinted with permission from Garcia-Ruiz et al., 2001.
crystallization chip 共Hansen et al., 2002兲. Protein and buffer ©2001, Elsevier.
solutions fill opposing sets of chambers, which are initially iso-
lated by a central valve. Here the valve is opened, and the
interface evolves diffusively due to the low Gr and Ra in the While all of these methods have been useful in under-
system. The slow evolution of the resulting solutions 共spanning standing the basic physics of protein crystallization, as of
a range in parameter space兲 allows protein crystals to nucleate this writing, only the microfluidic format has been suc-
and grow. Figure courtesy of C. Hansen. cessfully adapted to high-throughput screening.
An additional factor complicating the growth of pro-
tein crystals is that the molecules themselves tend to be
共macro兲techniques 共Hansen and Quake, 2003; Zheng et larger, more complicated, and softer than atoms found in
al., 2003兲. For example, the microfluidic device of solid-state crystals familiar to most physicists. It is there-
Hansen et al. 共2004兲 rapidly and systematically explores fore comparatively easier to disrupt protein crystalliza-
the phase space for protein solubility with minimal re- tion with defects, impurities, or even slight mismatches
agent consumption. Furthermore, the small dimensions in molecular orientation. One factor that is believed to
of microfluidic devices allows for different fluid physics adversely affect protein crystal quality is fluid convec-
than are found in macroscopic treatments. This can play tion. Since solute protein is depleted near the crystal,
an important role in crystallization techniques such as solution near the crystal can drive a buoyant flow if Gr
free-interface diffusion 共Salemme, 1972; Ng et al., 2003兲 and Ra are not low enough, which introduces a number
and can be used to manipulate the kinetics of nucleation of effects involving impurities and inhomogeneous
and growth. growth 共Grant and Saville, 1991; Vekilov et al., 1998; Lee
Figure 32 shows a microfluidic device designed to and Chernov, 2002兲. 共The inverse effect, wherein the
crystallize proteins by free-interface diffusion 共Hansen et dense layer of fluid that forms around a dissolving crys-
al., 2002兲. Two sets of chambers of varying volume are tal falls in a wake, is shown in Fig. 33.兲 For this and the
separated by long, narrow channels separated by valves. other reasons enumerated above, microfluidic devices
One set of chambers is filled with a dense protein solu- are expected to have a decisive advantage and to be-
tion of known concentration, and the other is filled by a come the standard in large-scale structural biology ef-
solution with a concentration of crystallizing agents. The forts.
valves connecting each pair of chambers have been While spatially inhomogeneous growth due to buoy-
opened, and the interface between the two solutions ant flow is undesirable for protein crystals, an analogous
evolves, as in Fig. 31. Gradually, the contents of each mechanism has been exploited to easily fabricate sharp
chamber make their way into the other, and a swath of optical probe tips, shown in Fig. 34 共Lambelet et al.,
trajectories through 共cp , cs兲 parameter space is explored. 1998; Unger et al., 1998; Stockle et al., 1999兲. An etchant
Free-interface diffusion therefore allows many condi- solution eats away at an optical fiber confined to a tube.
tions to be explored in a single experiment, and also The dense product-laden/etchant-sparse fluid falls down
traces out a path for crystals to nucleate and grow. This along the tube center, and product-free/etchant-rich
is achieved by ensuring that the interface spreads diffu- fluid moves upwards towards the fiber edges. The flow
sively 关Fig. 31共b兲兴, which requires both Ra and Gr to be drives a preferential etching of the fiber edges, providing
small. The small dimensions of microfluidic devices al- positive feedback that further reinforces the flow. This
low the RaⰆ 1, GrⰆ 1 regime to be easily reached. Pre- technique requires large Ra 共due to the concentration of
viously, this was achieved by taking samples into space dense product兲 to drive the flow, but small enough Gr
to decrease g 共DeLucas et al., 1989, 2002; McPherson et for the flow to be regular. The etchant is advected with,
al., 1999; Lee and Chernov, 2002兲, which does not allow and diffuses through, the buoyancy-driven flow set up by
for the highly parallel studies required for de novo crys- the etch products. Thus Pe for the etchant must be large;
tallization, or using an agarose gel to suppress convec- otherwise, diffusion would homogenize the etchant con-
tion 共Fig. 33; Garcia-Ruiz et al., 2001; Ng et al., 2003兲. centration, decoupling the etch process from the flow,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1004 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

FIG. 34. 共Color in online edition兲 A phenomenon similar to

that in Fig. 33共a兲 is exploited in tube etching, a technique for
fabricating sharp optical fiber tips. 共a兲 The dense products of
an etching reaction fall along the tube axis. 共b兲 This drives a
fluid flow that delivers “fresh” etchant to the fiber edges, con-
centrates the etched products towards the center, and rein-
forces the fluid flow. Reprinted with permission from Stockle et
al., 1999. ©1999, AIP.

and no sharp tip would be formed. Finally, we note that FIG. 35. 共Color in online edition兲 Polymerase chain reaction
phenomena similar to Figs. 33 and 34 are observed in 共PCR兲 in a steady temperature field. 共a兲 Solution is driven
electrodeposition 共Huth et al., 1995兲. The analogy be- along a channel that winds through temperature regions 共i兲,
tween electrohydrodynamics and buoyancy-driven flows 共ii兲, and 共iii兲 designed to cause DNA melting, extension, and
will be pursued shortly. annealing. Thus the temperature profile seen by the solution
matches a standard PCR cycle. Reprinted with permission
from Kopp et al., 1998. ©1998 AAAS. 共b兲 A PCR reactor in
2. Polymerase chain reaction „PCR… which temperature gradients drive a convective fluid flow that
takes suspended DNA molecules through a temperature pro-
Polymerse chain reaction 共PCR兲 is a process used to file designed to resemble that of PCR. Reprinted with permis-
exponentially replicate double-stranded DNA, allowing sion from Braun et al., 2003.
even a very small amount of DNA to be amplified into a
sufficient amount for sophisticated analysis. PCR in-
volves a three-step thermal cycle in the presence of a DNA molecules 共PeⰇ 1兲 through the variable tempera-
reagent soup: 共i兲 heating the solution to melt the DNA ture profile in the fluid. By properly designing the ex-
by separating each ds-DNA into two single strands; 共ii兲 periment, the temperature profile experienced by DNA
cooling so that DNA primers and DNA polymerase en- can be made to resemble that of conventional PCR, thus
zymes bind each strand 共annealing兲; and 共iii兲 warming allowing the chain reaction to proceed. The underlying
slightly to promote the base-by-base DNA replication convective flows have been established in two ways: 共i兲
by the polymerase 共extension兲. Ideally, each PCR cycle uniformly heating the bottom plate and cooling the top
doubles the number of double-stranded DNA mol- plate enough to drive Rayleigh-Benard convection at
ecules. Ra⬃ 106 共Krishnan et al., 2002兲, or 共ii兲 heating the fluid
Integrating PCR into microfluidic devices has been inhomogeneously 关Fig. 35共b兲兴 at Ra⬃ 104 to establish
achieved by many groups, typically by cycling the tem- laminar convective flow 共Braun et al., 2003兲. Finally,
perature of a microfluidic sample to replicate the stan- DNA molecules advect with the flow along the bottom
dard macroscopic PCR 共Wilding et al., 1994; Burns et al., of the cell towards the center of the roll, but also move
1996; Cheng et al., 1996; Woolley et al., 1996; Schmalzing outwards via thermophoresis 共itself poorly understood兲,
et al., 1997; Belgrader et al., 1999; Khandurina et al., leading to trapping in a ring so long as PeⰇ 1 共Braun and
2000; Chiou et al., 2001; Hong et al., 2001; Lagally, Em- Libchaber, 2002兲.
rich, and Mathies, 2001; Lagally, Medintz, and Mathies,
2001; Auroux et al., 2004兲. An alternate strategy involves
3. Electrohydrodynamic instabilities
pumping solution through various temperature zones to
mimic PCR 关Fig. 35共a兲兴 共Kopp et al., 1998; Liu, Enzel- We have presented two cases where Ra and Gr are
berger, and Quake, 2002兲, whose benefit is that cycle relevant for describing the resulting flow. Solution den-
time no longer depends on the time required to heat or sity gradients 共however they arise兲 couple to an external
cool the solution and its surroundings. Another ap- field 共gravity兲 to drive a fluid flow that in turn advects
proach exploits high-Ra buoyant flows to perform PCR the density gradients. We now discuss an interesting
in a steady temperature profile without an external variant on this picture, involving electrical Rayleigh and
pump 共Krishnan et al., 2002; Braun et al., 2003兲. The Grashof numbers. Conductivity gradients in solution
basic idea involves establishing a large convective flow couple to electric fields to drive a fluid flow that, in turn,
whose roll fills the experimental cell. This flow advects advects those same gradients. Physically, the 共local兲 con-

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1005

共Oddy et al., 2001兲. They examined this instability ex-

perimentally, analytically, and numerically 共Lin et al.,
2004; Chen et al., 2005兲, in the context of related EHD
instabilities in lower-conductivity solvents 共Melcher and
Schwarz, 1968; Melcher and Taylor, 1969; Hoburg and
Melcher, 1976, 1977; Saville, 1997; Baygents and Baldes-
sari, 1998兲. Controlling and avoiding this EHD instabil-
ity allowed Jung et al. 共2003兲 to achieve a thousandfold
amplication of solute concentration using field-amplified
sample stacking, an order of magnitude greater than
previous microfluidic devices 共Jacobson and Ramsey,
1995; Lichtenberg et al., 2001; Yang and Chien, 2001兲.
FIG. 36. 共Color in online edition兲 Mixing due to a high-Rae We note finally that electrical Rayleigh systems are
electrohydrodynamic instability. When an electric field is ap- richer in some ways than their buoyant counterparts.
plied along a channel containing solutions of different conduc- While the electrical and buoyant systems described
tivities flowing side by side, charge separation at the interface above have certain similarities, important distinctions do
introduces non-neutral fluid elements. When forcing is strong exist. Significantly, the charged species themselves can
enough that the interface cannot relax by diffusion, the non- 共and do兲 affect the external field strongly, whereas the
neutral fluid elements are driven by the field and amplify per- gravitational interaction between buoyant or dense sol-
turbations to cause instability. 共a兲 An experimental time series;
ute molecules is incredibly small. Such systems remain
共b兲 corresponding simulations 共Lin et al., 2004兲. Figure courtesy
largely unexplored.
of J. Santiago.

ductivity of a solution depends on the local concentra- F. The Knudsen number: When molecules matter
tion of charge-carrying ions, which evolve via diffusion,
advection, or due to electric forces. Throughout this review, we have always assumed that
As with buoyancy-driven flows, the electrohydrody- the fluid can be treated as a continuum. As we ponder
namic 共EHD兲 fluid velocity is not known at the outset, and probe fluids on ever smaller length scales, a natural
but depends on the distribution of charges, conductivi- question arises as to when the continuum approximation
ties, and fields. By analogy with buoyancy-driven flows, loses its validity.
however, one can imagine that diffusion can outcompete
slow flows so that the system evolves smoothly. On the
1. Molecular effects in gases
other hand, convection or forced motion may be stron-
ger than diffusion, in which case positive feedback and We have implicitly focused on liquids thus far, and
instabilities can occur. have not explicitly discussed microfluidic gas flows. De-
Continuing with this analogy, one can derive electrical spite their importance in numerous industrial and tech-
Grashof and Rayleigh numbers 共Baygents and Baldes- nological applications, such flows seem to have been
sari, 1998; Lin et al., 2004兲. An EHD velocity scale is largely overlooked by the physics community, owing per-
obtained under the assumption that charges separate un- haps to the overwhelming importance of liquids in biol-
til the fields they set up is of order the applied field E0, ogy and analytical chemistry. Various issues are intro-
giving a charge density scale ␳e ⬃ ⑀wE0 / L0. The electrical duced when the working fluid is a gas rather than a
force on the fluid fE = ␳eE0 ⬃ ⑀wE20 / L0 is balanced by vis- liquid 共Ho and Tai, 1998; Gad-el Hak, 1999兲, and here
cous forces f␯ ⬃ ␩U / L20, giving an EHD velocity scale we focus on two primary effects.
First, the molecular-level distinction between liquids
⑀wE20L0 and gases can have important ramifications for microflu-
UEHD ⬃ . 共32兲
␩ idic flows. While liquid molecules are in constant colli-
sion, gas molecules move ballistically and collide only
From this velocity scale, an electrical Rayleigh number rarely. Using the kinetic theory of gases, one can calcu-
UEHDL0 ⑀wE20L20 late the mean free path between collisions to be
Rae = = 共33兲
D ␩D 1
␭f ⬃ , 共35兲
and electrical Grashof number na2

UEHDL0 ⑀wE20L20 where n is the number density of molecules with radius

Gre = = 共34兲 a. For example, an ideal gas at 1 atm and 25 °C has a
␯ ␩␯
mean free path ␭f ⬃ 70 nm that increases at lower pres-
follow 共Baygents and Baldessari, 1998; Lin et al., 2004兲. sures or higher temperatures.
Recently, Santiago and co-workers have exploited As device dimensions get smaller, the mean free path
conductivity gradients in microfluidic devices to various occupies an increasingly significant portion of the flow—
ends. Figure 36 demonstrates the rapid electrokinetic and thus plays an increasingly important role. The
mixing that occurs when an EHD instability is driven Knudsen number

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1006 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

␭f Bizonne et al., 2002兲 and colloid-tipped atomic force mi-

Kn = 共36兲 croscope cantilevers 共Craig et al., 2001; Bonaccurso et al.,
L 2002; Sun et al., 2002; Neto et al., 2003; Vinogradova and
Yakubov, 2003兲, pressure and/or flow rate relations were
expresses the ratio of the mean free path 共the length measured in capillaries 共Schnell, 1956; Churaev et al.,
scale on which molecules matter兲 to a macroscopic 1984; Watanabe et al., 1999; Cheng and Giordano, 2002;
length scale L. The latter is typically a length scale rep- Choi et al., 2003兲 and through nanopores 共Cheikh and
resentative of the device, but could also be given by the Koper, 2003兲, fluid velocity profiles were measured using
length scale for temperature, pressure, or density gradi- micron-resolution particle image velocimetry 共Tre-
ents. theway and Meinhart, 2002兲, double-focus fluorescence
Noncontinuum effects play an increasing role as Kn cross correlation 共Lumma et al., 2003兲, and fluorescence
increases. Roughly speaking, molecules located farther recovery after photobleaching 共Pit et al., 2000兲. A gener-
than ␭f from a solid wall do not see the wall, whereas ally 共but not universally兲 consistent picture has emerged
closer molecules can collide with the wall rather than that wetting 共hydrophilic兲 surfaces obey the no-slip
other molecules. This implies that the fluid behaves like boundary condition as expected, whereas clean nonwet-
a continuum up to a distance ␭f from the wall, and influ- ting 共hydrophobic兲 surfaces exhibit apparent slip—often
ences the boundary conditions obeyed by the fluid. Max- to a surprisingly large extent. In fact, apparent slip
well 共1879兲, at the suggestion of a referee, predicted the lengths on the order of microns were measured—about
no-slip boundary condition to be violated for diffuse a thousand times longer than the molecular size or mean
wall collisions, yielding instead the slip condition free path one might expect. Some reports indicate a slip

u0 = ␤ 冏 冏
du
dn 0
, 共37兲
length that is shear-rate dependent, suggesting a yield-
stress slip boundary condition 共Spikes and Granick,
2003兲.
where ␤ is a slip length of order ␭f. This represents the A fundamental question that arises is whether fluid
leading-order correction due to finite Kn; second-order truly slips over solid surfaces, invalidating the no-slip
effects may play a role as Kn increases 共Arkilic et al., boundary condition, or whether the experiments reflect
2001; Maurer et al., 2003兲. an apparent slip that arises from surface inhomogene-
Second, the density of a gas typically depends much ities or a complex interface with additional physics.
more strongly on temperature and pressure than that of Molecular-dynamics simulations, whose physical ingredi-
a liquid. Therefore, compressibility can play a much ents are known, have indeed shown fluid to slip 共albeit
more important role in microfluidic gas flows, particu- with much shorter slip length兲, particularly when the at-
larly when significant differences in T or p exist. Both traction between molecules is stronger than the wall/
fluid density and mean free path ␭f are affected, and Kn molecule attraction 共Thompson and Robbins, 1990;
can vary appreciably along the channel. Three distinct Thompson and Troian, 1997; Barrat and Bocquet, 1999a,
Kn regimes have been measured in pressure-driven flow: 1999b; Cieplak et al., 2001; Cottin-Bizonne et al., 2003兲.
KnⰆ 1, where the gas behaves as a no-slip fluid, Kn⬃ 1, Although the shear rates employed in such simulations
where the gas behaves as a continuum but slips at the are much higher than in experiments, the reported slip
boundaries, and KnⰇ 1, where the continuum approxi- varies linearly with shear rate, suggesting linear re-
mation breaks down completely 共Gad-el Hak, 1999兲. sponse behavior and a constant slip length.
Various explanations have been proposed to account
for apparent slip effects 共Vinogradova, 1999兲. A per-
2. Fluid slip over nonwetting surfaces
fectly slipping but rough surface obeys a macroscopic
Liquid molecules, on the other hand, are in constant boundary condition with slip length of order the rough-
contact, resulting in significantly higher incompressibil- ness length scale 共Richardson, 1973; Jansons, 1988兲.
ity and making the concept of ␭f less meaningful. As From this standpoint, any macroscopic slip length 共real
such, noncontinuum effects appear to play a role only or apparent兲 will be confined to the roughness scale. In-
when the fluids themselves are confined to molecular deed, measured slip does decrease as surfaces are made
length scales. Various experiments indicate that water smoother 共Zhu and Granick, 2002a; Bonaccurso et al.,
and other liquids behave as continuum fluids until they 2003兲. However, even nonwetting, molecularly smooth
are confined to a few 共e.g., of order 10兲 molecular layers surfaces have exhibited large apparent slip lengths. A
关e.g., Israelachvili 共1986兲, Gee et al. 共1990兲, and Raviv et gas layer at the interface, which would lubricate and al-
al. 共2001兲兴. ter the fluid flow, has been invoked for an explanation
Recently, however, measurements have indicated that 共Ruckenstein and Rajora, 1983; Ruckenstein and
the no-slip boundary condition long assumed to hold for Churaev, 1991; Vinogradova, 1995, 1996; de Gennes,
liquids may be violated 共Vinogradova, 1999; Granick et 2002; Andrienko et al., 2003; Lauga and Stone, 2003;
al., 2003兲. A variety of experimental techniques have Yang et al., 2003; Lauga and Brenner, 2004; Tretheway
been used to probe the solid-liquid boundary condition and Meinhart, 2004兲. The idea is that dissolved gas mol-
for various liquids: drainage forces were measured using ecules are drawn out of solution to nucleate a gas layer
the surface force apparatus 共Horn et al., 2000; Baudry et at the solid-liquid interface and lower the surface energy
al., 2001; Zhu and Granick, 2001, 2002a, 2002b; Cottin- 共Lum et al., 1999; de Gennes, 2002; Attard, 2003; Gran-

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1007

ick et al., 2003兲. This hypothesis is bolstered by the ob-

servation of nanobubbles using tapping-mode atomic
force microscopy 共Ishida et al., 2000; Tyrrell and Attard,
2001, 2002兲, and further supported by the observation
that apparent slip decreases when the fluid is degassed
共Boehnke et al., 1999; Granick et al., 2003兲.
The no-slip boundary condition never enjoyed solid
theoretical backing, but rather was accepted due to its
apparent experimental success 共Goldstein, 1938兲. While
no consensus has been reached on the physical origin of
the measured slip, its effects will be felt as microfluidic FIG. 37. 共Color in online edition兲 共a兲 An ionic screening cloud
devices are made ever smaller. An understanding of the of width⬃ ␭D forms around a charged solid surface in an elec-
apparent slip might allow surfaces to be designed spe- trolytic solution. An external electric field forces the mobile
cifically to slip, e.g., to reduce hydrodynamic resistance ions in the screening cloud to result in electro-osmotic flow
and dispersion 共Cottin-Bizonne et al., 2003; Granick et 关Eq. 共46兲兴. 共b兲 Under an applied electric field, a freely sus-
al., 2003兲. In this case, an analogous Knudsen number pended charged particle in an electrolyte moves via electro-
phoresis, with velocity equal in magnitude and opposite in di-
␤ rection to Eq. 共46兲. Reprinted with permission from Squires
Kn = 共38兲
L and Bazant, 2004. ©2004, Cambridge University Press.

will be significant in the description of such flows.

nature of the flow. Large voltages are typically required,
III. DRIVING FLOWS AT THE BOUNDARIES
reducing the portability of the device. Electrochemical
reactions must occur at electrodes in order to maintain
Owing to the small dimensions of microfluidic devices an electric field in solution, which give rise to a host of
and to the long-range nature of viscous flows, bound- adverse effects, including metallic ion injection, water
aries are never very far in microfluidics. Therefore, tech- electrolysis and its associated bubble formation, and the
niques that exploit boundary effects can be quite effec- introduction of pH or solute gradients.
tive in microfluidic manipulation. In what follows, we Here, we focus on the basic physics of electrokinetics,
describe three such phenomena: electrokinetic effects, and explore recent variants on classic techniques. More
acoustic streaming, and fluid-structure interactions. detailed accounts of the fundamentals are given in text-
books and reviews 共Levich, 1962; Saville, 1977; Ander-
son, 1989; Russel et al., 1989; Lyklema, 1991; Probstein,
A. Electrokinetic effects
1994; Hunter, 2000; Ghosal, 2004兲.
Of all nonmechanical fluid manipulation techniques in
microfluidic contexts, electrokinetic techniques are al-
most certainly the most popular. Historically, early mi-
1. Electrostatics in ionic solutions
crofluidic devices were designed for capillary electro-
phoretic separations, and it was natural to use electro- Aqueous solutions are conductive due to the ubiqui-
osmotic flow 共EOF兲 to pump as well as to separate. 关See tous presence of dissolved ions 共e.g., from dissolved
Ghosal 共2004兲 for a review of the physics of EOF.兴 EOF salts, ionic groups on surfaces, or dissociated water mol-
in perfectly uniform, straight channels has a flat velocity ecules兲. Solid-liquid interfaces tend to develop surface
profile, implying less convective dispersion than for
charge, which attracts oppositely charged counterions
pressure-driven flow. Note, however, that Taylor disper-
and repels similarly charged co-ions. The resulting ionic
sion reduces convective dispersion—so that in suffi-
double layer screens the surface charge over a character-
ciently thin channels 共for which Peⱗ 1兲, tracer distribu-
istic Debye length ␭D. With the exception of these
tions in EOF and pressure-driven flow will evolve
similarly. Finally, EOF velocities do not depend on the charged double layers, the fluid is charge neutral.
microchannel dimensions, in striking contrast with The diffuse part of the double layer 关Fig. 37共a兲兴 is es-
pressure-driven flows, whose velocity decreases with the tablished when diffusive transport tending to smooth
second power of channel size. ion gradients balances electrostatic transport driving
However, electrokinetic techniques exhibit significant counterions towards the interface. 关A compact layer also
drawbacks. Electro-osmotic flows are not very robust, as exists 共Russel et al., 1989; Lyklema, 1991; Hunter, 2000兲,
they depend sensitively on the physicochemical proper- but we neglect it here.兴 We describe the diffuse layer for
ties of the solution and channel walls. Electro-osmotic a simple electrolyte of monovalent ions; results for mul-
flows depend on the walls’ charge density, which varies tiple, multivalent ions follow naturally and can be found
with solution pH, ionic strength, and 共uncontrollably兲 in the above-mentioned textbooks.
when solute molecules adsorb onto the walls. Inhomo- Ions in solution with number density n± and charge ±e
geneities in surface charge density—intended or not— set up an electrostatic field ␾ obeying Poisson’s equa-
give rise to stray pressure gradients that disrupt the plug tion,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1008 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

TABLE IV. Ion densities and screening lengths for aqueous q0

electrolytic solutions of varying salt concentration at 20 °C. ␨= 共44兲
⑀ w␬
Properties of aqueous electrolytic solutions for weak potentials. Nonlinear screening for strong po-
Concentration Ion density 共ions/ nm3兲 ␭D tentials gives a double-layer capacitance that depends
exponentially on ␨ 共Russel et al., 1989; Lyklema, 1991;
1M 0.6 0.3 nm Hunter, 2000兲.
0.1M 0.06 1 nm
1 mM 6 ⫻ 10−4 10 nm
0.1 ␮M 共pure, pH 7兲 6 ⫻ 10−8 1 ␮m 2. Electrokinetics: Electro-osmosis and electrophoresis
When an external electric field E储 is applied along a
charged surface, the nonzero charge density in the dif-
␳e e共n+ − n−兲 fuse layer gives rise to an electric body force f = ␳eE储ẑ,
ⵜ 2␾ = − =− , 共39兲 also tangent to the surface. In this case, the Stokes equa-
⑀w ⑀w tions 关Eq. 共2兲兴 can be integrated to give
where ⑀w is the permittivity of the solution. Due to their ⑀ w␨
small size, the ions are highly mobile and respond u= E储共1 − e−␬y兲, 共45兲
␩
quickly to the local electric field with potential ␺. In
equilibrium, each ion is statistically distributed accord- representing a fluid velocity that exponentially ap-
ing to the Boltzmann distribution proaches Smoluchowski’s constant slip velocity,

冉 冊
⑀ w␨
e␺ us = E储 , 共46兲
n± = n0 exp ⫿ , 共40兲 ␩
k BT
outside of the diffuse layer. When channel walls com-
where n0 is the bulk concentration of ions. The fields are prise the solid-fluid interface, electro-osmotic flow
related by the mean-field assumption: the electrostatic 共EOF兲 is driven 关Fig. 37共a兲兴 with a velocity that varies
potential ␾ set up by the ions 关Eq. 共39兲兴 is assumed to be linearly with the applied field 关Eq. 共46兲兴.
the same as the field ␺ to which they respond 关Eq. 共40兲兴. When the charged solid-fluid interface is a freely sus-
Combining Eqs. 共39兲 and 共40兲, we arrive at the nonlinear pended particle, electro-osmotic slip causes the particle
Poisson-Boltzmann equations itself to move electrophoretically 关Fig. 37共b兲兴 with veloc-
ity

ⵜ2 冉 冊
e␾
k BT
= k2 sinh
e␾
k BT
,冉 冊 共41兲 u=
⑀ w␨
␩
E 储 ⬅ ␮ EE 储 . 共47兲

which can be linearized for small potentials 共␾ Ⰶ kBT / e Several rather surprising results hold in the limit where
⬇ 26 mV兲, giving the double-layer ␭D is thin compared to the particle size:
the electrophoretic mobility ␮E for an object with fixed ␨
ⵜ 2␾ = ␬ 2␾ . 共42兲 is independent of particle size, shape, or number; mul-
tiple particles of the same zeta potential 共but possibly
Here ␭D = ␬−1 is the screening length, different shape and size兲 experience identically zero in-

冉 冊
teraction; and the fluid flow is parallel everywhere to
⑀ wk BT 1/2
electric field lines 共Morrison, 1970; Anderson, 1989兲. Fi-
␬−1 ⬅ ␭D = ⬅ 共8␲n0␭B兲−1/2 , 共43兲
2n0e2 nally, the electrophoretic mobility of a particle is inde-
pendent of its dielectric constant 共O’Brien and White,
where ␭B = e2 / 4␲⑀wkBT 共⬇0.7 nm in water兲 is the Bjer- 1978兲.
rum length, reflecting the distance that must be main-
tained between oppositely charged ions to prevent ag-
gregation. Water dissolves salts so well because its small, 3. Recent variations on classical electrokinetic
polar molecules form hydration shells around ions that techniques
enforce this separation; nonpolar solvents do not, and We have discussed the classic electrokinetic picture, in
thus have much lower conductivities 共Morrison, 1993兲. which homogeneous surfaces possess a constant surface
Table IV provides typical screening lengths. Notably, charge–zeta potential. Interesting and useful phenom-
even pure pH-7 water has only a 1-␮m screening length. ena can arise in systems with inhomogeneous zeta po-
Typical aqueous screening lengths are ⬃1 nm—much tentials. Here we explore three variations on this theme:
smaller than typical device dimensions. 共i兲 systems with patterned surface charge, 共ii兲 systems
A charged planar surface has a screening cloud obey- whose surface charge can be actively controlled, and 共iii兲
ing Eq. 共42兲, giving ␾ ⬃ ␨e−␬y. The potential drop across nonlinear induced-charge electrokinetic flows that occur
the charge cloud, called the zeta potential ␨, is related to when applied fields act on the charge clouds that they
the surface charge density q0 via induce around polarizable surfaces.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1009

Related work includes investigations into the three-

dimensional EOF generated by nonuniform ␨ distribu-
tions 共Long et al., 1999; Erickson and Li, 2002b; Ghosal,
2002b兲, EOF in inhomogeneous microchannels with fi-
nite double layer 共Ren and Li, 2001兲, and EOF in a mi-
crochannel containing two different fluids and thus a
step change in ␨ that displace one another 共Herr et al.,
2000; Ren et al., 2001兲. Microchannels with nonuniform ␨
or nonplanar surface topography provide a natural way
to mix fluids 共Stroock et al., 2000; Erickson and Li,
2002a; Johnson and Locascio, 2002; Johnson et al., 2002;
Qian and Bau, 2002兲. However, nonuniform zeta poten-
tials 共whether intentional or not兲 and channel turns in-
duce additional dispersion and band broadening in cap-
illary electrophoresis 共Long et al., 1999; Ghosal, 2002a,
FIG. 38. 共Color in online edition兲 Electro-osmotic flow over 2002b, 2003, 2004兲. Local ␨ variations 共Johnson et al.,
inhomogeneously charged surfaces. 共a兲 Tracer particle trajec- 2001; Henry et al., 2002; Qiao and Aluru, 2003兲 or spe-
tories demonstrate the cellular EOF flow over a surface with cial channel profiles 共Culbertson et al., 2000; Griffiths
alternating surface charge density. Adapted with permission and Nilson, 2000, 2002; Paegel et al., 2000; Zubritsky,
from Stroock et al., 2000. 共b兲 Patterning both surface charge 2000; Dutta and Leighton, 2001, 2002; Molho et al., 2001;
density and surface shape allows a net flow to be driven even Fiechtner and Cummings, 2003, 2004兲 have been de-
for a surface with no net charge. Reprinted with permission signed to minimize such effects.
from Ajdari, 1995.

b. Active control of surface charge

a. Patterned surfaces with nonuniform charge
Because conventional electro-osmotic flow varies lin-
The above discussion 共and most studies in general兲 early with the zeta potential of the wall, actively modi-
assumes constant ␨. Somewhat recently, electrokinetic fying the zeta potential of the wall allows active control
effects involving inhomogeneous ␨ have been investi- of the electro-osmotic flow. Such ideas were applied for
gated, with sometimes surprising results. Teubner 共1982兲 EOF control: by analogy with the microelectronic field-
considered electro-osmotic flow over surfaces with inho- effect transistor 共FET兲, microfluidic flow FETs have
mogeneous zeta potential, but did not focus on its pecu- been developed 共Ghowsi and Gale, 1991; Gajar and
liarities. Anderson and Idol 共1985兲 highlighted a rather Geis, 1992; Schasfoort et al., 1999; Buch et al., 2001兲.
counterintuitive result: a particle’s charge cannot be de- Flow FETs employ an additional electrode embedded
termined from its electrophoretic mobility alone as is behind the insulating channel walls; maintaining a non-
often assumed. Anderson 共1985兲 studied EOF through zero potential difference between this auxillary elec-
an inhomogeneously charged channel and found that trode and the fluid modifies the effective zeta potential,
while the average flow is proportional to the average ␨, which in turn modifies the EOF. Another method for
the flow itself is highly inhomogeneous and contains re- controlling surface charge densities involves coated
circulating rolls. Related work explored the electro- walls with TiO2, whose ␨ changes under UV light 共Moor-
phoresis of inhomogeneously charged particles of vari- thy et al., 2001兲.
ous geometries 共Fair and Anderson, 1989, 1990, 1992;
Fair, 1990; Solomentsev and Anderson, 1995; Long and
Ajdari, 1996, 1998兲. For example, Long and Ajdari c. Induced-charge electrokinetic phenomena
共1998兲 found cylindrical shapes with special cross section Above, we have discussed the electrokinetics of
and charge distributions whose electrophoretic velocity charges with inhomogeneous zeta potentials due to pat-
is always perpendicular to the applied field. terned surfaces or to active charge control. In this final
Ajdari 共1995兲 furthered these ideas and explored their section, we discuss cases where such inhomogeneities
use in microfluidic systems, with an emphasis on the rich naturally occur due to charge clouds that are induced
phenomena that occur when spatial symmetries are bro- around conducting or polarizable surfaces. These give
ken. For example, a net electro-osmotic pumping can rise to various nonlinear electrokinetic phenomena that
occur over a surface with no average charge when both have been discovered and explored in recent years, in-
the shape and the surface charge density are allowed to cluding “ac electro-osmosis” 共Ramos et al., 1999; Ajdari,
vary 关Fig. 38共b兲兴, and electro-osmotic flows can be driven 2000兲, ac colloidal self-assembly near electrodes 共Trau et
either transverse or parallel to the applied field by al., 1997; Yeh et al., 1997; Gong et al., 2002, 2003; Nadal,
breaking appropriate charge and shape symmetries 共Aj- Argoul, Hanusse, et al., 2002; Ristenpart et al., 2003兲,
dari, 1995, 1996, 2002; Stroock et al., 2000; Gitlin et al., and induced-charge electrokinetic phenomena around
2003兲. An advantage to transverse electrokinetic pump- conducting colloids and posts 共Murtsovkin, 1996; Bazant
ing is that modest potentials applied across narrow chan- and Squires, 2004; Squires and Bazant, 2004兲. The ingre-
nels give strong fields and flows 共Gitlin et al., 2003兲. dient that unifies all of these phenomena is a charge

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1010 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

cloud that is induced by the external field, and subse-

quently driven by that same field, so that U ⬃ E20. The
term induced-charge electrokinetic 共ICEK兲 phenomena
emphasizes the central role played by the induced
charges. Two broad classes of ICEK flows have been
studied: 共i兲 those whose inducing surfaces are the elec-
trodes themselves, and 共ii兲 those whose inducing surface
is independent of the electrodes. The former rely on
transient fields and charge clouds, and require ac fields
of a suitable frequency, whereas the latter can persist
with either dc or ac driving.
ICEK phenomena were initially explored in the col-
loid community. Levich 共1962兲 considered the charge
cloud induced around a conducting metal sphere in a
constant electric field in order to demonstrate that the
electrophoretic mobility of such a particle is the same as
if it were an insulator. The actual flows around such col-
loids are more difficult to measure than electrophoretic
velocities, and are of secondary importance to colloid
science. As a result, the flows received little attention
until Gamayunov, Murtsovkin, Dukhin, and others in-
vestigated them theoretically 共Dukhin, 1986; Dukhin
and Murtsovkin, 1986; Gamayunov et al., 1986; Murts-
ovkin, 1996兲 and experimentally 共Murtsovkin and Man- FIG. 39. 共Color in online edition兲 共a兲 A suddenly applied elec-
trov, 1991; Gamayunov et al., 1992兲. tric field intersects a conducting cylinder at right angles to
Similar ideas have recently been developed for micro- maintain an equipotential surface. A current is set up, how-
fluidic devices. We illustrate with perhaps the simplest ever, since mobile ions are driven along the field. These ions
example of induced-charge electrokinetic phenomena, accumulate in an induced charge double layer adjacent to the
conducting surface. 共b兲 Steady state is achieved when the in-
illustrated in Fig. 39 共Murtsovkin, 1996; Bazant and
duced double layer has developed sufficiently that no field
Squires, 2004; Squires and Bazant, 2004兲. An electric
lines penetrate the double layer. 共c兲 The steady-state ICEO
field that is suddenly applied across an electrolytic solu-
flow established around a conducting cylinder. The quadrupo-
tion containing a conducting post initially intersects the lar nature of the electro-osmotic slip velocity can be under-
post at right angles to satisfy the equipotential condition, stood from 共b兲, as the product of a 共dipolar兲 field driving a
as in Fig. 39共a兲. However, mobile ions in solution are 共dipolar兲 induced charge cloud. This flow persists in an ac field,
driven along field lines and accumulate at the liquid- since reversing the field direction reverses the sign of the in-
solid boundary, causing the field itself to evolve. A duced charge. Therefore the flow, which involves the product
dipolar-induced charge cloud develops—negatively of the two, is unchanged. Reprinted with permission from
charged on one side, and positively on the other—that Squires and Bazant, 2004. ©2004, Cambridge University Press.
reaches steady state when no field lines penetrate the
charge cloud, as in Fig. 39共b兲. In steady state, the in- ing the resulting flow 共which depends on the product of
duced zeta potential is dipolar, with magnitude ␨i the two兲 unchanged. ac electrokinetic effects circumvent
⬇ E0R representing the potential drop of the field E0 many of the problems that plague dc electrokinetics,
over the post size R. Using Eq. 共46兲, an induced-charge arising from the electrochemical reactions required to
electro-osmotic slip velocity is established with magni- maintain a steady electric field. In a recent ICEK vari-
tude ant, Lastochkin et al. 共2004兲 drive such reactions deliber-
ately 共which depend exponentially on voltage兲 and
⑀wE20R achieve 10 times stronger flows than with the capacitive,
UICEO ⬇ . 共48兲 ideally polarized electrode techniques described above.
␩
ac fields can be established without electrochemical
Not surprisingly, the ICEO velocity scale is the same as reactions, so long as the frequency is low enough that
the EHD velocity scale 关Eq. 共32兲兴. The ICEO flow is induced charge clouds have time to form, but fast
generically quadrupolar, with fluid that is drawn towards enough that electrodes and the fields they establish are
the body along the axis of the field, and is ejected from not screened. The time scale required for an induced
the equator 关Fig. 39共c兲兴. Such flows have been seen in charge cloud to form around a conducting body is given
microfluidic 共Levitan et al., 2005兲 and colloidal 共Gama- by
yunov et al., 1992兲 experiments. ␭ DR
An important characteristic of ICEK flows is that they ␶c ⬇ , 共49兲
Di
persist in ac fields, unlike their classic fixed-charge coun-
terparts which average to zero. Reversing the applied where Di is the diffusivity of the ions. This represents
field reverses the sign of the induced charge cloud, leav- the RC time for an equivalent circuit consisting of a

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1011

tener, et al., 2002兲. The symmetric systems produce

steady, counter-rotating fluid rolls 共Ramos et al., 1999;
Gonzalez et al., 2000; Green, Ramos, Gonzalez, et al.,
2000; Green et al., 2002; Nadal, Argoul, Kestener, et al.,
2002兲. Asymmetric electrodes pump fluid in the direc-
tion of broken symmetry 共Ajdari, 2000; Ramos et al.,
2003兲, as was verified experimentally 共Brown et al., 2001;
Studer et al., 2002; Mpholo et al., 2003兲. A pump exploit-
ing this effect establishes flow velocities of 450 ␮m / s us-
ing 2-V, 10-kHz rms applied potentials 共Mpholo et al.,
2003兲.
ICEK phenomena are not limited to conducting sur-
faces, as induced charge clouds form around any polar-
izable 共dielectric兲 surface, although they are strongest
around clean conducting surfaces 共Squires and Bazant,
2004兲. Thamida and Chang 共2002兲 observed a nonlinear
electrokinetic jet directed away from a corner that can
be understood with the above picture. A variety of
physical processes enrich such systems: Takhistov et al.
FIG. 40. ac electro-osmosis over coplanar electrodes. 共a兲 When
共2003兲 observe vortices at sharp channel corners, and
co-planar electrodes are driven near a particular frequency, an
argue that they occur due to electrolyte depletion under
inhomogeneous charge cloud is induced over each electrode,
which both modifies the external field and is driven by it. 共b兲
the strong field near the corner. Ben and Chang 共2002兲
The rectified, steady ac electro-osmotic flow, with measured and Lastochkin et al. 共2004兲 have investigated related
streaklines on the left and calculated streamlines on the right. systems under higher applied potentials, in which elec-
If the driving frequency is too small, charge clouds fully screen trochemical Faradaic reactions occur at electrodes,
each electrode, and no external field exists to drive the flow. yielding stronger microfluidic flows.
On the other hand, if the frequency is too fast, induced charge
clouds do not have time to form. Therefore, ac electro-osmosis
B. Acoustic streaming
is strongest at a given 共charging兲 frequency. Adapted with per-
mission from from Green et al., 2002.
Little explored thus far in microfluidics, acoustic
streaming represents one of very few inertial phenom-
charge cloud capacitor and a bulk resistor 共MacDonald, ena that may actually play a significant role in microflu-
idic devices. As discussed in Sec. II.A, small feature sizes
1970; Simonov and Shilov, 1977; Bazant and Squires,
typically prevent flow velocities from being high enough
2004; Squires and Bazant, 2004兲. Similarly, the time scale
to yield high numbers. High-frequency acoustic waves,
over which the electrodes themselves are screened is
however, can circumvent such difficulties. At first, peri-
given by ␶e ⬇ ␭DL / Di, where L is the separation between
odic wave motion might seem to be of little use for fluid
electrodes 共Ajdari, 2000; Gonzalez et al., 2000; Holl-
manipulation. However, the inertial nonlinearity can
ingsworth and Saville, 2003; Bazant et al., 2004兲. Thus,
rectify oscillatory fluid motion to give a time-averaged
ICEK applications allow closely spaced electrodes to be flow called steady or acoustic streaming, as shown by
used, so that higher fields can be established using low Rayleigh 共1884兲. A basic introduction is given in Chap.
applied voltages. 5.13 of Batchelor 共2001兲, more comprehensive treat-
This picture allows an easier understanding of the first ments can be found in Chap. 8.4 of Lighthill 共2001兲 and
work on microfluidic induced-charge electrokinetic phe- Lighthill 共1978兲, and detailed reviews are given by Stuart
nomena 共Ramos et al., 1999; Ajdari, 2000兲. In these sys- 共1963兲, Nyborg 共1998兲, and Riley 共2001兲. Here, we dis-
tems, induced charge clouds form inhomogeneously cuss three broad classes of steady streaming: quartz
around the electrodes themselves, so that the electrode wind, boundary-induced streaming, and cavitation mi-
screening time ␶el and the charge cloud formation time crostreaming.
␶c coincide 共Ramos et al., 1999; Ajdari, 2000; Gonzalez et
al., 2000兲. Transient fields drive transient charge clouds
1. Quartz wind
to establish these flows, and thus are limited to frequen-
cies around ␻ ⬃ ␶−1el . Three such systems have been inves- The simplest manifestion of acoustic streaming is the
tigated: 共i兲 a pair of coplanar electrodes deposited on a quartz wind, named for the wind observed to blow away
single channel wall as in Fig. 40 共Ramos et al., 1999; from oscillating quartz crystals. Dissipation causes
Gonzalez et al., 2000; Green, Ramos, Gonzalez, et al., acoustic waves to decay over an attenuation length
2000; Green et al., 2002兲; 共ii兲 an asymmetric array of co- ␣−1共␻兲, resulting in an oscillatory velocity field u␻
planar electrodes embedded in a channel wall 共Ajdari, ⬃ U0e−␣zei␻tẑ with zero time average. However, the os-
2000; Brown et al., 2001; Studer et al., 2002; Mpholo et cillation amplitude of a fluid particle decreases in one
al., 2003; Ramos et al., 2003兲; and 共iii兲 a dielectric half cycle as it moves away from the source; conversely,
“stripe” deposited on an electrode 共Nadal, Argoul, Kes- its amplitude increases as it moves toward the source.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1012 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Oscillating fluid particles thus constantly impart momen-

tum to the fluid, driving a secondary streaming flow
away from the acoustic source 共Lighthill, 1978兲. This rec-
tified forcing comes from the nonlinear inertial term that
we have so far neglected. Time averaging Eq. 共1兲 yields a
steady body force
fQW = − ␳具u␻ · ⵜu␻典 ⬃ ␳␣U20e−2␣zẑ, 共50兲
which drives the wind felt in front of stereo loudspeak-
ers.
The quartz wind can be used to drive microfluidic
flows by sending an acoustic plane wave along a micro-
channel of length L ⲏ ␣−1. The body force gives a pres-
sure drop ⌬PQW = 兰fQWdz ⬃ ␳U20 that drives a Poiseuille
FIG. 41. Acoustic streaming around a cylinder. The inertial
flow with flow rate
nonlinearity rectifies the oscillatory flow, leading to the steady
␳U20w4 streaming flow shown. Outside of the oscillatory boundary
QQW ⬃ , 共51兲 layer, the steady streaming flow is directed towards the stagna-
␩L
tion points at the top and bottom. The streaming flow has
which suffers from the w4 scaling typical of pressure- small streaming Reynolds number Res and decays with dis-
driven flows. For appreciable forcing, the channel length tance as r−1. Figure courtesy of Tatsuno; experiments described
should have dimensions comparable to or larger than in Tatsuno, 1980.
the attenuation length ␣−1. Since ␣−1 ⬇ 8.3 mm in water
at 50 MHz and decreases like ␻−2 共Rife et al., 2000兲, a
3 d
natural size limit is imposed on quartz-wind-driven mi- us ⬃ − U0 U0 , 共54兲
crofluidic devices, which are typically centimeter sized 4␻ dx储
共Zhu and Kim, 1998; Rife et al., 2000; Yang et al., 2000, and flows in the direction of decreasing free-stream ve-
2001兲. locity 共e.g., toward stagnation points兲.
Boundary-induced acoustic streaming is reminiscent
2. Boundary-induced streaming of electro-osmotic flow 共Sec. III.A兲, in that body forces
localized near a surface give rise to an effective slip ve-
While the quartz wind occurs in the bulk of the fluid, locity. However, EOF increases monotonically to Eq.
steady streaming flows occur around solid boundaries as 共46兲, whereas streaming flows reverse direction within
well, with an amplitude that is 共remarkably兲 indepen- the Stokes layer, causing recirculating rolls 共Fig. 41兲.
dent of viscosity. The flow outside of the Stokes layer obeys the Navier-
An oscillatory flow 共with amplitude U0兲 over a solid Stokes equations 关Eq. 共1兲兴 with a slip velocity given by
surface must vanish on the surface due to the no-slip Eq. 共54兲. The external flow due to this slip velocity has
condition. In unsteady viscous flows, vorticity generated an associated streaming Reynolds number,
at the wall spreads diffusively 共with diffusivity ␯兲, so that
wall effects are confined to the distance that vorticity usR U20
Res = = , 共55兲
can diffuse in one cycle before it is annihilated. An os- ␯ ␻␯
cillatory boundary layer 共called a Stokes layer兲 of size

冉冊
giving two regimes of acoustic streaming, for large or
␯ 1/2
small Res 共Stuart, 1963, 1966兲. Early work on steady
␦⬃ 共52兲
␻ streaming flows almost universally assumed Res Ⰶ 1, a
regime called RNW streaming 共Rayleigh, 1884; Nyborg,
contains all of the vorticity in the flow, with irrotational
1953; Westervelt, 1953兲. In the macroscopic world, large-
flow outside of the Stokes layer.
Res Stuart streaming flows are more common, although
If the amplitude of oscillatory flow U0 above the
more challenging theoretically. Speaking broadly, RNW
Stokes layer varies with length scale R along the bound- streaming involves long-ranged Stokes flows, whereas
ary, the nonlinear inertial term ␳u · ⵜ u rectifies the oscil- Stuart streaming flows are confined to an outer bound-
latory flow to give a steady inertial forcing, as with the
ary layer of size Res−1/2R, along with isolated jets and
quartz wind. The steady inertial force 共fi ⬃ ␳U20 / R兲 is bal- potential flows.
anced by steady viscous forces fv ⬃ ␩us / ␦ 2 ⬃ ␳␻us that The free-streaming velocity can change for several
are, ironically, independent of viscosity. This results in a reasons:
velocity scale for steady boundary-driven streaming, Standing acoustic waves. Historically the first theoret-
U20 ical treatment, Rayleigh 共1884兲 calculated the secondary
us ⬃ . 共53兲 streaming due to a standing acoustic wave in a tube. As
␻R
expected, flow is towards the nodes of the standing wave
The acoustic boundary-driven slip velocity outside the pattern.
Stokes layer is given by Acoustic attenuation. Acoustic attenuation gives a

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1013

FIG. 43. 共Color in online edition兲 共a,b兲 Secondary streaming

eddies in an oscillatory flow that arise due to 共a兲 a localized
bump in the channel profile 共Hall, 1974兲, and 共b兲 a sinusoidally
varying channel profile 共Hall, 1974兲. 共c兲 Streaming flow arising
from oscillatory flow in a slightly curved channel 共Lyne, 1971兲.
FIG. 42. Acoustic streaming and chemical capture around a Flow in the central region of the pipe resembles Dean flow
microelectrode post. An electrochemical reaction releases a 共Fig. 4兲, but flows in the opposite direction.
marker into the fluid that is captured within the Stokes layer.
The left series represents theory and the right represents ex-
periments. 共a兲, 共b兲, and 共c兲 Reynolds number Re= a2␻ / ␩ = 500, centrifugally forced兲. In the low-frequency limit, the
200, and 100, respectively. Adapted and reprinted with permis- Stokes layer becomes larger than the pipe itself, and
sion from Lutz et al., 2003. ©2003, National Academy of Sci- Dean flow is recovered.
ences.

3. Cavitation microstreaming
length scale R ⬃ ␣ over which the free streaming veloc-
−1
We discuss one last streaming mechanism that has
ity changes. Thus two types of streaming occur when been exploited for microfluidic use. Bubbles subjected
attenuating acoustic waves travel down a channel: 共i兲 the to acoustic waves expand and contract as the local pres-
quartz wind drives a parabolic flow, as discussed above, sure varies. Bubble shape and fluid oscillations are rec-
and 共ii兲 boundary-driven streaming establishes a plug tified to give so-called cavitation microstreaming 共Elder,
flow. As channels get smaller, boundary-driven stream- 1959兲, and bubbles that both translate and pulsate give
ing dominates 共Frampton et al., 2003兲: a local slip veloc- rise to a dipolar streaming flow that is longer ranged for
ity is established with magnitude us ⬃ ␣U20 / spherical bubbles 共us ⬃ r−1兲 than for solid spheres 共us
␻ exp共−2␣z兲, that gives an average flow rate Q ⬃ r−2兲 共Longuet-Higgins, 1998兲. Bubble translations and
⬃ ␣U20w2 / ␻ down a long channel, whose w2 scaling can oscillations occur simultaneously when a bubble is at-
exceed the w4 quartz-wind scaling 关Eq. 共51兲兴 for small tached to a solid wall, since volume conservation re-
enough w. quires the bubble center to move away from the wall as
Curved boundaries. Steady streaming occurs around it expands. The streaming flow around an attached
vibrating spheres and cylinders or 共equivalently兲 due to bubble is shown in Fig. 44, and has been used to lyse
oscillatory flow past cylinders and spheres 关Riley 共2001兲, vesicles 共Marmottant and Hilgenfeldt, 2003兲 and to en-
and references therein兴. From Eq. 共54兲, the streaming hance mixing in cell capture 共Liu, Yang, et al., 2002兲 and
slip velocity 共just outside the Stokes layer兲 is directed DNA hybridization applications 共Liu, Lenigk, et al.,
towards the decreasing free-stream velocity, i.e., towards
the stagnation points at the two poles. The resulting flow
is thus quadrupolar, and decays with distance as r−1 for
cylinders 共r−2 for spheres兲. Flow reversal in the Stokes
layer gives rise to a set of recirculating eddies 共Fig. 41兲
whose size varies as ␯␻−1/2 关Eq. 共52兲兴. Recently, such ed-
dies have been exploited to make a microchemical trap,
shown in Fig. 42, that stores chemicals released at the
cylinder surface 共Lutz et al., 2003兲.
Examples that may find use for mixing and pumping
in microfluidics include streaming due to oscillatory
flows in curved pipes 关Fig. 43共c兲兴 共Lyne, 1971兲, tapering
channels 共Grotberg, 1984; Gaver and Grotberg, 1986兲,
and channels of variable cross section 关Figs. 43共a兲 and
43共b兲兴 共Hall, 1974兲. The streaming flow in a curved pipe
关Fig. 43共c兲兴 resembles Dean flow 共Fig. 4兲, suggesting a FIG. 44. Cavitation microstreaming. The streaming flow
zero-Re analog of the serpentine mixer 共Liu et al., 2000兲 around a pulsating and translationally oscillating bubble gives
that uses acoustic waves, rather than the primary shear stresses that can be used to lyse vesicles. Measured par-
pressure-driven flow, to introduce the stirring flow. Note ticle trajectories are shown with calculated streamlines. Re-
that the flow in Fig. 43共c兲 is directed opposite to Dean printed with permission from Marmottant and Hilgenfeldt,
flow 共except within the Stokes layer, where the flow is 2003. ©2003, NPG.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1014 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

2003兲. A further rectification occurs when the bubble is

held adjacent to a solid protrusion, in which case a net
pumping occurs in the direction of broken symmetry
共Marmottant and Hilgenfeldt, 2004兲.

C. Fluid-structure interactions

Our final example of boundary-driven flows is perhaps

the simplest: moving the walls themselves in order to
physically force the fluid. This strategy is practical when FIG. 45. 共Color in online edition兲 Flexible fluidic valves. 共a兲
the walls are sufficiently compliant to allow significant An actively controlled valve results when a soft 共PDMS兲 mem-
deformation. Soft polymeric materials have tunable elas- brane separates a fluidic channel from a gas-filled actuation
channel. 共b兲 Pressurizing the gas in the actuation channel
tic moduli that are typically many orders of magnitude
pushes the membrane upwards to seal off the fluid channel. A
lower than crystalline solids like silicon, and thus have
peristaltic pump can be made when three such valves are
proven popular and successful for such applications placed in a row and driven cyclically 共Fig. 46兲. Adapted with
共Quake and Scherer, 2000; Whitesides et al., 2001兲. permission from Studer et al., 2004. ©2004, AIP. 共c,d兲 A flex-
We discuss here two broad methods of controlling the ible flap that is more easily bent in the forward direction than
fluid by deforming the structure that encloses it. First, in reverse forms a passive valve that allows fluid flow in one
externally applied pressure can deform membranes and direction only. Reprinted from Jeon et al., 2002, with kind per-
displace fluid, creating valves and pumps. Second, active mission of Springer Science and Business Media.
and self-regulating fluid control elements can be built
with intelligent hydrogels that swell or collapse as envi-
ronmental conditions change. pressurized, but close when the fluid in front is pressur-
ized 共Beebe et al., 2000; Jeon et al., 2002兲.
Three or more active valves arranged along a channel
1. Driving flows with soft walls 共Fig. 46兲 pump fluid peristaltically when the valves are
Mechanically deforming channel walls to physically opened and closed cyclically 共Unger et al., 2000; Chou et
force fluid motion is by nature a robust technique, since al., 2001兲. A variety of devices have been designed using
it is insensitive to almost all physicochemical properties such valves and pumps 共Hong and Quake, 2003兲, such as
of the solution—in marked contrast with most other the rotary mixer of Fig. 12 共Chou et al., 2001兲, the pro-
strategies. Furthermore, local fluid control is attained in tein crystallization chips of Fig. 32 共Hansen et al., 2002兲,
straightforward fashion using defined membrane defor- and protein solubility chip 共Hansen et al., 2004兲, flow
mations. injection devices 共Leach et al., 2003兲, and cell analyzers
These techniques require a material that is soft and sorters 共Fu et al., 2002; Wheeler et al., 2003兲. Bor-
enough to be deformed using reasonable applied pres- rowing from microelectronic large-scale integration
sures, and the material of choice is typically cross-linked strategies, Nc fluid chambers can be addressed with a
polymeric PDMS. Early fabrication techniques using number Nl of control lines that increases only logarith-
PDMS 共Wilbur et al., 1994; Kim et al., 1995兲 and PDMS- mically 共Nl ⬃ lnNc兲 共Thorsen et al., 2002兲. Using this and
based microfluidic applications 共Delamarche et al., 1997; other ideas for integration, a device has been fabricated
Effenhauser et al., 1997兲 have led to an explosion of in- that performs 400 PCR reactions using only 41 pipetting
terest in soft lithography 共Xia and Whitesides, 1998; Mc- steps 共as compared to 1200 steps for conventional pro-
Donald et al., 2000; Quake and Scherer, 2000; Unger et cesses兲 共Liu, Hansen, and Quake, 2003兲. This represents
al., 2000; Whitesides and Stroock, 2001; Whitesides et al., significant progress towards achieving the promise of mi-
2001; Ng et al., 2002; Sia and Whitesides, 2003兲. Advan- crofluidic “labs on a chip,” as the device performs many
tages include a tunable elastic modulus, a gas permeabil- automated experiments simultaneously while consuming
ity that allows dead-end channels to be easily filled and little reagent.
allows an oxygen supply enabling cell cultures 共Walker et
al., 2002; Leclerc et al., 2003兲, and a remarkable ease,
speed, and low fabrication cost compared to silicon-
based devices 共Quake and Scherer, 2000; Whitesides and
Stroock, 2001兲.
Figure 45共a兲 shows a simple mechanical valve fabri-
cated using multilayer soft lithography, in which a fluid
channel is separated from a control channel by a thin
membrane. When the control line is pressurized 关Fig. FIG. 46. 共Color in online edition兲 Three active valves 共Fig. 45兲
45共b兲兴, the membrane is pushed up to close the fluid placed in a row and driven cyclically form a versatile peristaltic
channel 共Unger et al., 2000; Ismagilov et al., 2001; Studer pump whose operation is insensitive to almost all physico-
et al., 2004兲. Soft passive valves are shown in Figs. 45共c兲 chemical properties of the solution. Reprinted with permission
and 45共d兲, which open when the fluid behind the valve is from Unger et al., 2000. ©2000, AAAS.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1015

The above examples highlight the deliberate, driven

deformation of soft channel elements to manipulate flu-
ids. Because it is soft, PDMS deforms under applied
pressures, whether or not such deformations are in-
tended. This may be particularly relevant in light of the
high pressures required to drive fluid flows through
small channels. The influence of channel deformations
on diffusion and dispersion has been examined 共Holden
et al., 2003兲, and soft diffraction gratings on channel
walls have been used as local pressure sensors
共Hosokawa et al., 2002兲.
More generally, the coupling between local fluid stress
and elastic deformation of channel walls has been little
studied in the microfluidic context. Compliance causes
walls to expand and store extra fluid under applied pres-
sure. Thus compliant walls act like fluidic capacitors, by
analogy with electronic circuits. Compliant channel sys-
tems thus intrinsically exhibit distributed capacitance FIG. 47. 共Color in online edition兲 Smart hydrogel-based flow
and resistance, leading to RC times and low-pass filter- control structures. 共a兲,共b兲 A pH-responsive valve is formed
ing. A low-frequency oscillatory applied pressure de- when pH-sensitive hydrogel posts span a microchannel. 共a兲
forms the walls to a steady state 共fully charging the ca- The valve open at one pH, and 共b兲 the valve closed at another
pH. 共c兲–共f兲 A fluid pH sorter. 共c兲 Two hydrogels are used, one
pacitor兲 and drives a fluid current through the channel.
that expands at low pH and contracts at high pH, and one that
High-frequency pressures, on the other hand, drive fluid
does the opposite. 共d兲 At pH ⬃ 4 – 5, the left post swells to
into and out of the space created by deforming channel
block the channel, whereas the right shrinks to allow fluid to
walls, and drive little appreciable flow. pass. 共e兲 At intermediate 共⬃7兲 pH, both posts block the chan-
Fluid-structure interactions have long been studied in nels, and 共f兲 at high pH 共⬃8兲, fluid flows is blocked to the right
the context of biofluid dynamics, as many organisms but flows to the left. Adapted and reprinted with permission
transport fluid through compliant or collapsible tubes from Beebe et al., 2000. ©2000, NPG.
共Grotberg and Jensen, 2004兲. Although much of this
work concerns higher-Re flows, the coupling between
the fluid flow and the shape of its surrounding vessel swelling process depend on the collective diffusion of
reveals a host of instabilities and nonlinearities that solution into the gel, and thus vary with the square of gel
might be exploited for microfluidic flows. For example, a dimension 共Matsuo and Tanaka, 1988; De et al., 2002兲.
nonlinear fluidic limiter has been designed with compli- While slow for macroscopic gels, the smaller dimensions
ant elastic walls 共King and Toner, 2003兲. inherent in microfluidic applications yield faster re-
sponses, that can be even further diminished without
loss of functionality by coating microfabricated solid
2. Intelligent walls made of responsive hydrogel structures with hydrogel 共Beebe et al., 2000兲.
The soft microfluidic elements discussed above de- An intriguing and powerful aspect of such smart
form in response to applied pressures. Interesting and valves is that they allow self-regulating fluidic systems.
useful phenomena arise when channel walls and solid Figure 47 shows two flow regulation devices: the first
structures contain materials that are themselves active. contains hydrogel-coated posts that swell to block a
Intelligent hydrogels have been developed that swell or channel when pH values are too extreme, and the sec-
collapse in response to a wide range of stimuli, e.g., tem- ond sorts fluid by pH into two channels. One channel
perature, solvent composition, pH, salinity, electric contains a hydrogel post that swells at high pH to allow
fields, specific antigens, and light, to list a few 共Tanaka, only low-pH fluid, and vice versa 共Beebe et al., 2000兲.
1978; Tanaka et al., 1980, 1982; Suzuki and Tanaka, 1990; Furthermore, a self-regulating system has been designed
Miyata et al., 1999兲. An overview of intelligent gels is that automatically maintains a constant pH 共Beebe et al.,
given by Osada and Rossmurphy 共1993兲, and discussions 2000; Eddington et al., 2001; Lee et al., 2003兲. While
of their application in the microfluidic context are given these examples all involve pH-sensitive hydrogels, simi-
by Jager et al. 共2000兲, Moorthy and Beebe 共2003兲, and lar devices could be designed that are responsive to
Roy and Gupta 共2003兲. other stimuli. Biology uses similar flow control tech-
Hydrogels can be understood in terms of a polymer niques; for example, plant xylem 共hollow tubes used to
gel whose solvent quality changes in response to an en- transport water from the roots to the leaves兲 employ
vironmental change, causing dramatic swelling or hydrogels to alter hydrodynamic resistance based on the
deswelling. A more rigorous understanding follows from water’s ionic strength 共Zwieniecki et al., 2001兲.
Flory-Huggins theory, which accounts for entropy and Microfluidic valves have been fabricated from ther-
enthalpy of mixing, rubber elasticity, chain-chain inter- moresponsive hydrogels 共Peters et al., 1997; Harmon et
actions, and hydrogen ion dissociation 共Flory, 1953; al., 2003; Luo et al., 2003; Richter et al., 2003; Selvaga-
Tanaka, 1978; Tanaka et al., 1982兲. The dynamics of the napathy et al., 2003; Yu et al., 2003兲, and are typically

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1016 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

driven using embedded resistive microheaters. Valves kH polymer spring constant

have also been made from pH-sensitive hydrogels. kB Boltzmann constant
共Beebe et al., 2000; Eddington et al., 2001; Yu et al., 2001; l channel length
Liu, Yu, and Beebe, 2002兲, and one-shot pumps and n number density
valves have been demonstrated using expandable micro- p pressure
spheres 共Griss et al., 2002兲. Furthermore, hydrogels have r radial coordinate
been used as reactors and detectors 共Moorthy and u fluid velocity field
Beebe, 2002; Zhan et al., 2002兲. us slip velocity
We close this section with an interesting phenomenon w channel width
that has yet to be utilized microfluidically, but which x coordinate across channel width
opens intriguing possibilities. Chemical networks whose y coordinate spanning channel height
reactions oscillate spontaneously, known as chemical os-
z downstream coordinate
cillators, can be systematically designed 共Rabai et al.,
␣ attenuation length
1990兲, and key elements of these networks can be incor-
␣␳ compressibility
porated into the hydrogel polymer itself. The result is a
structure whose volume oscillates periodically 共Yoshida
␣T coefficient of thermal expansion
et al., 1996, 2003; Crook et al., 2002; Sakai and Yoshida, ␣c density coefficient for species c
2004兲. Such self-oscillating structures might find interest- ␤ slip length
ing and useful applications in microfluidic devices. ␦ oscillatory boundary layer thickness
⑀w dielectric constant of water
␨ zeta potential
ACKNOWLEDGMENTS
␩ shear viscosity
We are grateful to Armand Ajdari, W. French Ander- ␥˙ fluid shear rate
son, Martin Bazant, John Brady, Gary Leal, Gareth ␥ liquid-liquid surface tension
McKinley, and Howard Stone for critical readings of the ␥sl solid-liquid surface tension
manuscript, and their many detailed comments and sug- ␥sg solid-gas surface tension
gestions. Helpful conversations with Jian Gong, Carl ␬H thermal diffusivity
Hansen, Eric Lauga, and Aaron Wheeler are gratefully ␭D Debye screening length
acknowledged. T.M.S. gratefully acknowledges support ␭B Bjerrum length
from a Lee A. Dubridge Prize Postdoctoral Fellowship ␮E electrophoretic mobility
at Caltech and an NSF Mathematical Sciences Postdoc- ␯=␩/␳ kinematic viscosity
toral Fellowship. ␳ fluid density
␳e charge density
LIST OF SYMBOLS ␳c surfactant concentration
␴
J ,␴ stress tensor
D diffusivity ␶D diffusion time
Dz Taylor dispersivity ␶flow flow time scale
E electric-field strength ␶i inertial time scale
L0 generic length scale ␶p polymer relaxation time
R curvature scale ␰ hydrodynamic resistance of bead
R p, R 0 polymer size ␻ oscillation frequency
T temperature
U0 velocity scale
REFERENCES
⌬U interspecies velocity difference
V potential difference Abbott, S., J. Ralston, G. Reynolds, and R. Hayes, 1999, Lang-
⌬W bandwidth muir 15, 8923.
Z downstream distance scale Absolom, D. R., W. Zingg, and A. W. Neumann, 1987, J.
⌬Z band separation Biomed. Mater. Res. 21, 161.
a molecular/particulate radius Acharya, B. R., T. Krupenkin, S. Ramachandran, Z. Wang, C.
b freely jointed chain step length C. Huang, and J. A. Rogers, 2003, Appl. Phys. Lett. 83, 4912.
c capacitance per unit area Ajdari, A., 1995, Phys. Rev. Lett. 75, 755.
⌬c 共dense兲 solute concentration field Ajdari, A., 1996, Phys. Rev. E 53, 4996.
Ajdari, A., 2000, Phys. Rev. E 61, R45.
ė extension rate
Ajdari, A., 2002, Phys. Rev. E 65, 016301.
e electron charge
Ajdari, A., and J. Prost, 1992, C. R. Acad. Sci., Ser. IIb: Mec.,
f, f body force density on fluid Phys., Chim., Astron. 315, 1635.
fv viscous force density Al-Mubaiyedh, U. A., R. Sureshkumar, and B. Khomami,
fi inertial force density 1999, Phys. Fluids 11, 3217.
g gravitational acceleration Anderson, J. L., 1985, J. Colloid Interface Sci. 105, 45.
h channel height Anderson, J. L., 1989, Annu. Rev. Fluid Mech. 21, 61.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1017

Anderson, J. L., and W. K. Idol, 1985, Chem. Eng. Commun. erre, A. Dagkessamanskaia, E. Trevisiol, J. M. Francois, and
38, 93. C. Bergaud, 2003, Appl. Phys. Lett. 82, 3122.
Andrienko, D., B. Dunweg, and O. I. Vinogradova, 2003, J. Belgrader, P., W. Benett, D. Hadley, J. Richards, P. Stratton, R.
Chem. Phys. 119, 13106. Mariella, and F. Milanovich, 1999, Science 284, 449.
Angilella, J. R., and J. P. Brancher, 2003, Phys. Fluids 15, 261. Ben, Y., and H. C. Chang, 2002, J. Fluid Mech. 461, 229.
Anna, S. L., N. Bontoux, and H. A. Stone, 2003, Appl. Phys. Beni, G., and S. Hackwood, 1981, Appl. Phys. Lett. 38, 207.
Lett. 82, 364. Beni, G., S. Hackwood, and J. L. Jackel, 1982, Appl. Phys.
Aref, H., 1984, J. Fluid Mech. 143, 1. Lett. 40, 912.
Aref, H., 2002, Phys. Fluids 14, 1315. Beni, G., and M. A. Tenan, 1981, J. Appl. Phys. 52, 6011.
Aref, H., and S. Balachandar, 1986, Phys. Fluids 29, 3515. Bennett, D. E., B. S. Gallardo, and N. L. Abbott, 1996, J. Am.
Aris, R., 1956, Proc. R. Soc. London, Ser. A 235, 67. Chem. Soc. 118, 6499.
Aris, R., 1959, Proc. R. Soc. London, Ser. A 252, 538. Berge, B., and J. Peseux, 2000, Eur. Phys. J. E 3, 159.
Arkilic, E. B., K. S. Breuer, and M. A. Schmidt, 2001, J. Fluid Bico, J., and D. Quere, 2000, Europhys. Lett. 51, 546.
Mech. 437, 29. Bico, J., and D. Quere, 2002, J. Fluid Mech. 467, 101.
Ashton, R., C. Padala, and R. S. Kane, 2003, Curr. Opin. Bio- Bienia, M., C. Quilliet, and M. Vallade, 2003, Langmuir 19,
technol. 14, 497. 9328.
Astumian, R. D., and M. Bier, 1994, Phys. Rev. Lett. 72, 1766. Bier, M., and R. D. Astumian, 1996, Phys. Rev. Lett. 76, 4277.
Attard, P., 2003, Adv. Colloid Interface Sci. 104, 75. Bird, R. B., R. C. Armstrong, and O. Hassager, 1987, Dynam-
Auroux, P. A., D. Iossifidis, D. R. Reyes, and A. Manz, 2002, ics of Polymeric Liquids, Fluid Mechanics No. 1 共Wiley, New
Anal. Chem. 74, 2637. York兲.
Auroux, P.-A., Y. Koc, A. deMello, A. Manz, and P. J. R. Day, Bird, R. B., O. Hassager, R. C. Armstrong, and C. F. Curtiss,
2004, Lab Chip 4, 534. 1987, Dynamics of Polymeric Liquids, Kinetic Theory No. 2
Austin, R. H., N. Darnton, R. Huang, J. Sturm, O. Bakajin, 共Wiley, New York兲.
and T. Duke, 2002, Appl. Phys. A: Mater. Sci. Process. 75, Blake, T. D., A. Clarke, and E. H. Stattersfield, 2000,
279. Langmuir 16, 2928.
Austin, R. H., and W. D. Volkmuth, 1993, Analusis 21, 235. Boehnke, U. C., T. Remmler, H. Motschmann, S. Wurlitzer, J.
Babcock, H. P., D. E. Smith, J. S. Hur, E. S. G. Shaqfeh, and S. Hauwede, and T. M. Fischer, 1999, J. Colloid Interface Sci.
Chu, 2000, Phys. Rev. Lett. 85, 2018. 211, 243.
Bain, C. D., 2001, ChemPhysChem 2, 580. Boger, D. V., 1977, J. Non-Newtonian Fluid Mech. 3, 87.
Bain, C. D., G. D. Burnetthall, and R. R. Montgomerie, 1994, Boger, D. V., 1987, Annu. Rev. Fluid Mech. 19, 157.
Nature 共London兲 372, 414. Boger, D. V., and Y. L. Yeow, 1992, Exp. Therm. Fluid Sci. 5,
Bajer, K., and H. K. Moffatt, 1990, J. Fluid Mech. 212, 337. 633.
Bakajin, O., T. A. J. Duke, J. Tegenfeldt, C. F. Chou, S. S. Bonaccurso, E., H. J. Butt, and V. S. J. Craig, 2003, Phys. Rev.
Chan, R. H. Austin, and E. C. Cox, 2001, Anal. Chem. 73, Lett. 90, 144501.
6053. Bonaccurso, E., M. Kappl, and H. J. Butt, 2002, Phys. Rev.
Bakajin, O. B., T. A. J. Duke, C. F. Chou, S. S. Chan, R. H. Lett. 88, 076103.
Austin, and E. C. Cox, 1998, Phys. Rev. Lett. 80, 2737. Bottausci, F., I. Mezic, C. D. Meinhart, and C. Cardonne, 2004,
Barbier, V., and J. L. Viovy, 2003, Curr. Opin. Biotechnol. 14, Philos. Trans. R. Soc. London, Ser. A 362, 1001.
51. Bousse, L., C. Cohen, T. Nikiforov, A. Chow, A. R. Kopf-Sill,
Baroud, C. N., F. Okkels, L. Menetrier, and P. Tabeling, 2003, R. Dubrow, and J. W. Parce, 2000, Annu. Rev. Biophys. Bio-
Phys. Rev. E 67, 060104. mol. Struct. 29, 155.
Barrat, J. L., and L. Bocquet, 1999a, Faraday Discuss. 112, 119. Braun, D., N. L. Goddard, and A. Libchaber, 2003, Phys. Rev.
Barrat, J. L., and L. Bocquet, 1999b, Phys. Rev. Lett. 82, 4671. Lett. 91, 158103.
Bassous, E., H. H. Taub, and L. Kuhn, 1977, Appl. Phys. Lett. Braun, D., and A. Libchaber, 2002, Phys. Rev. Lett. 89, 188103.
31, 135. Brenner, H., and D. A. Edwards, 1993, Macrotransport Pro-
Batchelor, G. K., 1977, “Developments in microhydrodynam- cesses 共Butterworth-Heinemann, Boston兲.
ics,” Theoretical and Applied Mechanics, edited by W. T. Koi- Bretherton, F. P., 1961, J. Fluid Mech. 10, 166.
ter 共North-Holland, Amsterdam兲. Bringer, M. R., C. J. Gerdts, H. Song, J. D. Tice, and R. F.
Batchelor, G. K., 2001, An Introduction to Fluid Mechanics Ismagilov, 2004, Philos. Trans. R. Soc. London, Ser. A 362,
共Cambridge University Press, Cambridge, England兲. 1087.
Baudry, J., E. Charlaix, A. Tonck, and D. Mazuyer, 2001, Brochard, F., 1989, Langmuir 5, 432.
Langmuir 17, 5232. Brochard, F., and P. G. de Gennes, 1977, J. Chem. Phys. 67, 52.
Baumert, B. M., and S. J. Muller, 1995, Rheol. Acta 34, 147. Brochard-Wyart, F., 1995, Europhys. Lett. 30, 387.
Baygents, J. C., and F. Baldessari, 1998, Phys. Fluids 10, 301. Brody, J. P., and P. Yager, 1997, Sens. Actuators, A 58, 13.
Bazant, M. Z., and T. M. Squires, 2004, Phys. Rev. Lett. 92, Brody, J. P., P. Yager, R. E. Goldstein, and R. H. Austin, 1996,
066101. Biophys. J. 71, 3430.
Bazant, M. Z., K. Thornton, and A. Ajdari, 2004, Phys. Rev. E Brown, A. B. D., C. G. Smith, and A. R. Rennie, 2001, Phys.
70, 021506. Rev. E 63, 016305.
Beebe, D. J., G. A. Mensing, and G. M. Walker, 2002, Annu. Bruin, G. J. M., 2000, Electrophoresis 21, 3931.
Rev. Biomed. Eng. 4, 261. Bryden, M. D., and H. Brenner, 1999, J. Fluid Mech. 379, 319.
Beebe, D. J., J. S. Moore, J. M. Bauer, Q. Yu, R. H. Liu, C. Brzoska, J. B., F. Brochard-Wyart, and F. Rondelez, 1993,
Devadoss, and B. H. Jo, 2000, Nature 共London兲 404, 588. Langmuir 9, 2220.
Belaubre, P., M. Guirardel, G. Garcia, J. B. Pourciel, V. Leb- Buch, J. S., P. C. Wang, D. L. DeVoe, and C. S. Lee, 2001,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1018 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Electrophoresis 22, 3902. Cottin-Bizonne, C., S. Jurine, J. Baudry, J. Crassous, F. Re-

Burns, M. A., B. N. Johnson, S. N. Brahmasandra, K. Hand- stagno, and E. Charlaix, 2002, Eur. Phys. J. E 9, 47.
ique, J. R. Webster, M. Krishnan, T. S. Sammarco, P. M. Man, Craig, V. S. J., C. Neto, and D. R. M. Williams, 2001, Phys.
D. Jones, D. Heldsinger, C. H. Mastrangelo, and D. T. Burke, Rev. Lett. 87, 054504.
1998, Science 282, 484. Crook, C. J., A. Smith, R. A. L. Jones, and A. J. Ryan, 2002,
Burns, M. A., C. H. Mastrangelo, T. S. Sammarco, F. P. Man, J. Phys. Chem. Chem. Phys. 4, 1367.
R. Webster, B. N. Johnson, B. Foerster, D. Jones, Y. Fields, Crut, A., D. Lasne, J. F. Allemand, M. Dahan, and P. Desbi-
A. R. Kaiser, and D. T. Burke, 1996, Proc. Natl. Acad. Sci. olles, 2003, Phys. Rev. E 67, 051910.
U.S.A. 93, 5556. Culbertson, C. T., S. C. Jacobson, and J. M. Ramsey, 2000,
Bustamante, C., J. F. Marko, E. D. Siggia, and S. Smith, 1994, Anal. Chem. 72, 5814.
Science 265, 1599. Daniel, S., M. K. Chaudhury, and J. C. Chen, 2001, Science
Byars, J. A., A. Oztekin, R. A. Brown, and G. H. McKinley, 291, 633.
1994, J. Fluid Mech. 271, 173. Darhuber, A. A., J. Z. Chen, J. M. Davis, and S. M. Troian,
Cabodi, M., Y. F. Chen, S. W. P. Turner, H. G. Craighead, and 2004, Philos. Trans. R. Soc. London, Ser. A 362, 1037.
R. H. Austin, 2002, Electrophoresis 23, 3496. Darhuber, A. A., J. M. Davis, S. M. Troian, and W. W. Reisner,
Campbell, C. J., and B. A. Grzybowski, 2004, Philos. Trans. R. 2003, Phys. Fluids 15, 1295.
Soc. London, Ser. A 362, 1069. Darhuber, A. A., S. M. Troian, S. M. Miller, and S. Wagner,
Castelain, C., D. Berger, P. Legentilhomme, A. Mokrani, and 2000, J. Appl. Phys. 87, 7768.
H. Peerhossaini, 2000, Int. J. Heat Mass Transfer 43, 3687. Darhuber, A. A., S. M. Troian, and W. W. Reisner, 2001, Phys.
Chabbra, R. P., and D. De Kee, 1992, Transport Processes in Rev. E 64, 031603.
Bubbles, Drops and Particles 共Hemisphere, New York兲. Darhuber, A. A., S. M. Troian, and S. Wagner, 2002, J. Appl.
Chandrasekhar, S., 1981, Hydrodynamic and Hydromagnetic Phys. 91, 5686.
Stability 共Dover, New York兲. Darhuber, A. A., J. P. Valentino, J. M. Davis, S. M. Troian, and
Chaudhury, M. K., and G. M. Whitesides, 1992, Science 256, S. Wagner, 2003, Appl. Phys. Lett. 82, 657.
1539. Davis, S. H., 1987, Annu. Rev. Fluid Mech. 19, 403.
Cheikh, C., and G. Koper, 2003, Phys. Rev. Lett. 91, 156102. De, S. K., N. R. Aluru, B. Johnson, W. C. Crone, D. J. Beebe,
Chen, C.-H., H. Lin, S. K. Lele, and J. G. Santiago, 2005, J. and J. Moore, 2002, J. Microelectromech. Syst. 11, 544.
Fluid Mech. 524, 263. Dean, W. R., 1927, Philos. Mag. 4, 208.
Chen, H., and J. C. Meiners, 2004, Appl. Phys. Lett. 84, 2193. Dean, W. R., 1928, Philos. Mag. 5, 673.
Cheng, J., M. A. Shoffner, G. E. Hvichia, L. J. Kricka, and P. Deen, W. M., 1998, Analysis of Transport Phenomena 共Oxford
Wilding, 1996, Nucleic Acids Res. 24, 380. University Press, New York兲.
Cheng, J. T., and N. Giordano, 2002, Phys. Rev. E 65, 031206. de Gennes, P. G., 1974, J. Chem. Phys. 60, 5030.
Chiou, J., P. Matsudaira, A. Sonin, and D. Ehrlich, 2001, Anal. de Gennes, P. G., 1985, Rev. Mod. Phys. 57, 827.
Chem. 73, 2018. de Gennes, P. G., 1997, Science 276, 1999.
Chiou, P. Y., H. Moon, H. Toshiyoshi, C. J. Kim, and M. C. Wu, de Gennes, P. G., 1998, Physica A 249, 196.
2003, Sens. Actuators, A 104, 222. de Gennes, P. G., 2002, Langmuir 18, 3413.
Cho, B. S., T. G. Schuster, X. Zhu, D. Chang, G. D. Smith, and de Gennes, P. G., F. Brochard-Wyart, and D. Quere, 2004,
S. Takayama, 2003, Anal. Chem. 75, 1671. Capillary and Wetting Phenomena: Drops, Bubbles, Pearls,
Cho, S. K., H. J. Moon, and C. J. Kim, 2003, J. Microelectro- Waves 共Springer-Verlag, New York兲.
mech. Syst. 12, 70. Delamarche, E., A. Bernard, H. Schmid, B. Michel, and H.
Choban, E. R., L. J. Markoski, A. Wieckowski, and P. J. A. Biebuyck, 1997, Science 276, 779.
Kenis, 2004, J. Power Sources 128, 54. DeLucas, L. J., K. M. Moore, M. M. Long, R. Rouleau, T.
Choi, C. H., K. J. A. Westin, and K. S. Breuer, 2003, Phys. Bray, W. Crysel, and L. Weise, 2002, J. Cryst. Growth 237,
Fluids 15, 2897. 1646.
Chou, C. F., R. H. Austin, O. Bakajin, J. O. Tegenfeldt, J. A. DeLucas, L. J., et al., 1989, Science 246, 651.
Castelino, S. S. Chan, E. C. Cox, H. Craighead, N. Darnton, Dolnik, V., 1997, Electrophoresis 18, 2353.
T. Duke, J. Y. Han, and S. Turner, 2000, Electrophoresis 21, Dolnik, V., 1999, Electrophoresis 20, 3106.
81. Dolnik, V., and K. M. Hutterer, 2001, Electrophoresis 22, 4163.
Chou, C. F., O. Bakajin, S. W. P. Turner, T. A. J. Duke, S. S. Dolnik, V., S. R. Liu, and S. Jovanovich, 2000, Electrophoresis
Chan, E. C. Cox, H. G. Craighead, and R. H. Austin, 1999, 21, 41.
Proc. Natl. Acad. Sci. U.S.A. 96, 13762. Dorfman, K. D., and J. L. Viovy, 2004, Phys. Rev. E 69,
Chou, H. P., M. A. Unger, and S. Quake, 2001, Biomed. Mi- 011901.
crodevices 3, 323. Dos Santos, D. F., and T. Ondarçuhu, 1995, Phys. Rev. Lett. 75,
Churaev, N. V., V. D. Sobolev, and A. N. Somov, 1984, J. Col- 2972.
loid Interface Sci. 97, 574. Doyle, P. S., J. Bibette, A. Bancaud, and J. L. Viovy, 2002,
Cieplak, M., J. Koplik, and J. R. Banavar, 2001, Phys. Rev. Science 295, 2237.
Lett. 86, 803. Doyle, P. S., B. Ladoux, and J. L. Viovy, 2000, Phys. Rev. Lett.
Cochrane, T., K. Walters, and M. F. Webster, 1981, Philos. 84, 4769.
Trans. R. Soc. London, Ser. A 301, 163. Dreyfus, R., P. Tabeling, and H. Willaime, 2003, Phys. Rev.
Colgate, E., and H. Matsumoto, 1990, J. Vac. Sci. Technol. A 8, Lett. 90, 144505.
3625. Duke, T., G. Monnelly, R. H. Austin, and E. C. Cox, 1997,
Cottin-Bizonne, C., J. L. Barrat, L. Bocquet, and E. Charlaix, Electrophoresis 18, 17.
2003, Nat. Mater. 2, 237. Duke, T. A. J., and R. H. Austin, 1998, Phys. Rev. Lett. 80,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1019

1552. Gamayunov, N. I., V. A. Murtsovkin, and A. S. Dukhin, 1986,

Duke, T. A. J., R. H. Austin, E. C. Cox, and S. S. Chan, 1996, Colloid J. USSR 48, 197.
Electrophoresis 17, 1075. Ganan-Calvo, A. M., 1998, Phys. Rev. Lett. 80, 285.
Dukhin, A. S., 1986, Colloid J. USSR 48, 376. Ganan-Calvo, A. M., and J. M. Gordillo, 2001, Phys. Rev. Lett.
Dukhin, A. S., and V. A. Murtsovkin, 1986, Colloid J. USSR 87, 274501.
48, 203. Garcia-Ruiz, J. M., M. L. Novella, R. Moreno, and J. A.
Dussan, E. B., 1979, Annu. Rev. Fluid Mech. 11, 371. Gavira, 2001, J. Cryst. Growth 232, 165.
Dutta, D., and D. T. Leighton, 2001, Anal. Chem. 73, 504. Garmer, N., R. O. Grigoriev, and M. F. Schatz, 2003, Phys.
Dutta, D., and D. T. Leighton, 2002, Anal. Chem. 74, 1007. Rev. Lett. 91, 054501.
Eaton, W. A., V. Munoz, S. J. Hagen, G. S. Jas, L. J. Lapidus, Garstecki, P., I. Gitlin, W. DiLuzio, G. M. Whitesides, E. Ku-
E. R. Henry, and J. Hofrichter, 2000, Annu. Rev. Biophys. macheva, and H. A. Stone, 2004, Appl. Phys. Lett. 85, 2649.
Biomol. Struct. 29, 327. Gascoyne, P., C. Mahidol, M. Ruchirawat, J. Satayavivad, P.
Eckart, C., 1948, J. Mar. Res. 7, 265. Watcharasit, and F. F. Becker, 2002, Lab Chip 2, 70.
Eddington, D. T., R. H. Liu, J. S. Moore, and D. J. Beebe, Gascoyne, P. R. C., and J. V. Vykoukal, 2004, Proc. IEEE 92,
2001, Lab Chip 1, 96. 22.
Effenhauser, C. S., G. J. M. Bruin, A. Paulus, and M. Ehrat, Gau, H., S. Herminghaus, P. Lenz, and R. Lipowsky, 1999,
1997, Anal. Chem. 69, 3451. Science 283, 46.
Eggers, J., and H. A. Stone, 2004, J. Fluid Mech. 505, 309. Gaver, D. P., and J. B. Grotberg, 1986, J. Fluid Mech. 172, 47.
El Moctar, A. O., N. Aubry, and J. Batton, 2003, Lab Chip 3, Gee, M. L., P. M. McGuiggan, J. N. Israelachvili, and A. M.
273. Homola, 1990, J. Chem. Phys. 93, 1895.
Elder, S. A., 1959, J. Acoust. Soc. Am. 31, 54. Ghosal, S., 2002a, Anal. Chem. 74, 4198.
Erickson, D., and D. Q. Li, 2002a, Langmuir 18, 1883. Ghosal, S., 2002b, J. Fluid Mech. 459, 103.
Erickson, D., and D. Q. Li, 2002b, Langmuir 18, 8949. Ghosal, S., 2003, J. Fluid Mech. 491, 285.
Ertas, D., 1998, Phys. Rev. Lett. 80, 1548. Ghosal, S., 2004, Electrophoresis 25, 214.
Fair,’M. C., 1990, Int. J. Multiphase Flow 16, 1131. Ghowsi, K., and R. J. Gale, 1991, J. Chromatogr. 559, 95.
Fair, M. C., and J. L. Anderson, 1989, J. Colloid Interface Sci. Giddings, J. C., 1991, Unified Separation Science 共Wiley, New
127, 388. York兲.
Fair, M. C., and J. L. Anderson, 1990, Int. J. Multiphase Flow Gitlin, I., A. D. Stroock, G. M. Whitesides, and A. Ajdari,
16, 663. 2003, Appl. Phys. Lett. 83, 1486.
Fair, M. C., and J. L. Anderson, 1992, Langmuir 8, 2850. Glasgow, I., and N. Aubry, 2003, Lab Chip 3, 114.
Ferrigno, R., A. D. Stroock, T. D. Clark, M. Mayer, and G. M. Gleeson, J. P., O. M. Roche, J. West, and A. Gelb, 2004, SIAM
Whitesides, 2002, J. Am. Chem. Soc. 124, 12930. J. Appl. Math. 64, 1294.
Feynman, R. P., R. B. Leighton, and M. L. Sands, 1989, The Gluckstad, J., 2004, Nat. Mater. 3, 9.
Feynman Lectures on Physics, Vol. 2 共Addison-Wesley, Read- Goldstein, S., 1938, Modern Developments in Fluid Dynamics
ing, MA兲. 共Clarendon Press, Oxford兲.
Fiechtner, G. J., and E. B. Cummings, 2003, Anal. Chem. 75, Gong, T. Y., D. T. Wu, and D. W. M. Marr, 2002, Langmuir 18,
4747. 10064.
Fiechtner, G. J., and E. B. Cummings, 2004, J. Chromatogr., A Gong, T. Y., D. T. Wu, and D. W. M. Marr, 2003, Langmuir 19,
1027, 245. 5967.
Figeys, D., and D. Pinto, 2000, Anal. Chem. 72, 330A. Gonzalez, A., A. Ramos, N. G. Green, A. Castellanos, and H.
Figeys, D., and D. Pinto, 2001, Electrophoresis 22, 208. Morgan, 2000, Phys. Rev. E 61, 4019.
Flory, P. J., 1953, Principles of Polymer Chemistry 共Cornell, Gordillo, J. M., A. M. Ganan-Calvo, and M. Perez-Saborid,
Ithaca兲. 2001, Phys. Fluids 13, 3839.
Foster, K., and G. A. Parker, 1970, Fluidics: Components and Graham, M. D., 1998, J. Fluid Mech. 360, 341.
Circuits 共Wiley Interscience, London兲. Graham, M. D., 2003, Phys. Fluids 15, 1702.
Frampton, K. D., S. E. Martin, and K. Minor, 2003, Appl. Granick, S., Y. X. Zhu, and H. Lee, 2003, Nat. Mater. 2, 221.
Acoust. 64, 681. Grant, M. L., and D. A. Saville, 1991, J. Cryst. Growth 108, 8.
Frumkin, A., and V. Levich, 1945, Acta Physicochim. URSS 20, Green, N. G., A. Ramos, A. Gonzalez, H. Morgan, and A.
769. Castellanos, 2000, Phys. Rev. E 61, 4011.
Fu, A. Y., H. P. Chou, C. Spence, F. H. Arnold, and S. R. Green, N. G., A. Ramos, A. Gonzalez, H. Morgan, and A.
Quake, 2002, Anal. Chem. 74, 2451. Castellanos, 2002, Phys. Rev. E 66, 026305.
Gad-el Hak, M., 1999, J. Fluids Eng. 121, 5. Green, N. G., A. Ramos, and H. Morgan, 2000, J. Phys. D 33,
Gajar, S. A., and M. W. Geis, 1992, J. Electrochem. Soc. 139, 632.
2833. Greenspan, H. P., 1978, J. Fluid Mech. 84, 125.
Gallardo, B. S., and N. L. Abbott, 1997, Langmuir 13, 203. Griffiths, S. K., and R. H. Nilson, 2000, Anal. Chem. 72, 5473.
Gallardo, B. S., V. K. Gupta, F. D. Eagerton, L. I. Jong, V. S. Griffiths, S. K., and R. H. Nilson, 2002, Anal. Chem. 74, 2960.
Craig, R. R. Shah, and N. L. Abbott, 1999, Science 283, 57. Griss, P., H. Andersson, and G. Stemme, 2002, Lab Chip 2,
Gallardo, B. S., M. J. Hwa, and N. L. Abbott, 1995, Langmuir 117.
11, 4209. Groisman, A., M. Enzelberger, and S. R. Quake, 2003, Science
Gallardo, B. S., K. L. Metcalfe, and N. L. Abbott, 1996, 300, 955.
Langmuir 12, 4116. Groisman, A., and S. Quake, 2004, Phys. Rev. Lett. 92, 094501.
Gamayunov, N. I., G. I. Mantrov, and V. A. Murtsovkin, 1992, Groisman, A., and V. Steinberg, 2000, Nature 共London兲 405,
Colloid J. 54, 20. 53.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1020 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Groisman, A., and V. Steinberg, 2001, Nature 共London兲 410, Shen, 2003, Surf. Rev. Lett. 10, 569.
905. Huang, L. R., E. C. Cox, R. H. Austin, and J. C. Sturm, 2004,
Groisman, A., and V. Steinberg, 2004, New J. Phys. 6, 29. Science 304, 987.
Grotberg, J. B., 1984, J. Fluid Mech. 141, 249. Huang, L. R., P. Silberzan, J. O. Tegenfeldt, E. C. Cox, J. C.
Grotberg, J. B., and O. E. Jensen, 2004, Annu. Rev. Fluid Sturm, R. H. Austin, and H. Craighead, 2002, Phys. Rev. Lett.
Mech. 36, 121. 89, 178301.
Gueroui, Z., E. Freyssingeas, C. Place, and B. Berge, 2003, Huang, L. R., J. O. Tegenfeldt, J. J. Kraeft, J. C. Sturm, R. H.
Eur. Phys. J. E 11, 105. Austin, and E. C. Cox, 2002, Nat. Biotechnol. 20, 1048.
Hall, P., 1974, J. Fluid Mech. 64, 209. Humphrey, E. F., and D. H. Tarumoto, 1965, Eds., Fluidics
Hancock, W. S., S.-L. Wu, and P. Shieh, 2002, Proteomics 2, 共Fluid Amplifier Associates, Boston兲.
352. Hunter, R. J., 2000, Foundations of Colloid Science, 2nd ed.
Handique, K., and M. A. Burns, 2001, J. Micromech. 共Oxford University Press, Oxford兲.
Microeng. 11, 548. Hur, J. S., E. S. G. Shaqfeh, H. P. Babcock, and S. Chu, 2002,
Handique, K., B. P. Gogoi, D. T. Burke, C. H. Mastrangelo, Phys. Rev. E 66, 011915.
and M. A. Burns, 1997, Proc. SPIE 3224, 185. Hur, J. S., E. S. G. Shaqfeh, and R. G. Larson, 2000, J. Rheol.
Hansen, C., and S. R. Quake, 2003, Curr. Opin. Struct. Biol. 44, 713.
13, 538. Huth, J. M., H. L. Swinney, W. D. McCormick, A. Kuhn, and
Hansen, C., M. O. A. Sommer, and S. R. Quake, 2004, Proc. F. Argoul, 1995, Phys. Rev. E 51, 3444.
Natl. Acad. Sci. U.S.A. 101, 14431. Ichimura, K., S. K. Oh, and M. Nakagawa, 2000, Science 288,
Hansen, C. L., E. Skordalakes, J. M. Berger, and S. R. Quake, 1624.
2002, Proc. Natl. Acad. Sci. U.S.A. 99, 16531. Im, D. J., and I. S. Kang, 2003, J. Colloid Interface Sci. 266,
Happel, J., and H. Brenner, 1983, Low Reynolds Number Hy- 127.
drodynamics 共Martinus Nijhoff, The Hague兲. Ishida, N., T. Inoue, M. Miyahara, and K. Higashitani, 2000,
Harmon, M. E., M. Tang, and C. W. Frank, 2003, Polymer 44, Langmuir 16, 6377.
4547. Ismagilov, R. F., D. Rosmarin, P. J. A. Kenis, D. T. Chiu, W.
Harrison, D. J., K. Fluri, K. Seiler, Z. H. Fan, C. S. Effen- Zhang, H. A. Stone, and G. M. Whitesides, 2001, Anal.
hauser, and A. Manz, 1993, Science 261, 895. Chem. 73, 4682.
Harrison, D. J., A. Manz, Z. H. Fan, H. Ludi, and H. M. Wid- Ismagilov, R. F., A. D. Stroock, P. J. A. Kenis, G. Whitesides,
mer, 1992, Anal. Chem. 64, 1926. and H. A. Stone, 2000, Appl. Phys. Lett. 76, 2376.
Hatch, A., A. E. Kamholz, K. R. Hawkins, M. S. Munson, E. Israelachvili, J. N., 1986, J. Colloid Interface Sci. 110, 263.
A. Schilling, B. H. Weigl, and P. Yager, 2001, Nat. Biotechnol. Jackel, J. L., S. Hackwood, and G. Beni, 1982, Appl. Phys.
19, 461. Lett. 40, 4.
Hayes, R. A., and B. J. Feenstra, 2003, Nature 共London兲 425, Jackel, J. L., S. Hackwood, J. J. Veselka, and G. Beni, 1983,
383. Appl. Opt. 22, 1765.
Henry, A. C., E. A. Waddell, R. Shreiner, and L. E. Locascio, Jacobson, S. C., and J. M. Ramsey, 1995, Electrophoresis 16,
2002, Electrophoresis 23, 791. 481.
Herr, A. E., J. I. Molho, J. G. Santiago, M. G. Mungal, T. W. Jager, E. W. H., E. Smela, and O. Inganas, 2000, Science 290,
Kenny, and M. G. Garguilo, 2000, Anal. Chem. 72, 1053. 1540.
Hinch, E. J., 1977, Phys. Fluids 20, s22. Janocha, B., H. Bauser, C. Oehr, H. Brunner, and W. Gopel,
Hinch, E. J., 1988, in Disorder and Mixing, edited by E. 2000, Langmuir 16, 3349.
Guyon, J.-P. Nadal, and Y. Pomeau 共Kluwer Academic, Dor- Jansons, K. M., 1988, Phys. Fluids 31, 15.
drecht兲, pp. 43–55. Jendrejack, R. M., E. T. Dimalanta, D. C. Schwartz, M. D.
Ho, C.-M., and Y.-C. Tai, 1998, Annu. Rev. Fluid Mech. 30, Graham, and J. J. de Pablo, 2003, Phys. Rev. Lett. 91, 038102.
579. Jendrejack, R. M., D. C. Schwartz, J. J. de Pablo, and M. D.
Hoburg, J. F., and J. R. Melcher, 1976, J. Fluid Mech. 73, 333. Graham, 2004, J. Chem. Phys. 120, 2513.
Hoburg, J. F., and J. R. Melcher, 1977, Phys. Fluids 20, 903. Jendrejack, R. M., D. C. Schwartz, M. D. Graham, and J. J. de
Holden, M. A., S. Kumar, A. Beskok, and P. S. Cremer, 2003, Pablo, 2003, J. Chem. Phys. 119, 1165.
J. Micromech. Microeng. 13, 412. Jeon, N. L., D. T. Chiu, C. J. Wargo, H. K. Wu, I. S. Choi, J. R.
Hollingsworth, A. D., and D. A. Saville, 2003, J. Colloid Inter- Anderson, and G. M. Whitesides, 2002, Biomed. Microde-
face Sci. 257, 65. vices 4, 117.
Hong, J. W., T. Fujii, M. Seki, T. Yamamoto, and I. Endo, 2001, Johnson, T. J., and L. E. Locascio, 2002, Lab Chip 2, 135.
Electrophoresis 22, 328. Johnson, T. J., D. Ross, M. Gaitan, and L. E. Locascio, 2001,
Hong, J. W., and S. R. Quake, 2003, Nat. Biotechnol. 21, 1179. Anal. Chem. 73, 3656.
Hong, J. W., V. Studer, G. Hang, W. F. Anderson, and S. R. Johnson, T. J., D. Ross, and L. E. Locascio, 2002, Anal. Chem.
Quake, 2004, Nat. Biotechnol. 22, 435. 74, 45.
Horn, R. G., O. I. Vinogradova, M. E. Mackay, and N. Phan- Jones, S. W., O. M. Thomas, and H. Aref, 1989, J. Fluid Mech.
Thien, 2000, J. Chem. Phys. 112, 6424. 209, 335.
Hosokawa, K., K. Hanada, and R. Maeda, 2002, J. Micromech. Jones, S. W., and W. R. Young, 1994, J. Fluid Mech. 280, 149.
Microeng. 12, 1. Jones, T. B., 2002, Langmuir 18, 4437.
Hsieh, J., P. Mach, F. Cattaneo, S. Yang, T. Krupenkine, K. Jones, T. B., J. D. Fowler, Y. S. Chang, and C. J. Kim, 2003,
Baldwin, and J. A. Rogers, 2003, IEEE Photonics Technol. Langmuir 19, 7646.
Lett. 15, 81. Jones, T. B., M. Gunji, M. Washizu, and M. J. Feldman, 2001, J.
Huang, L., Y. Zhang, J. H. Liao, L. N. Xu, N. Gu, and H. J. Appl. Phys. 89, 1441.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1021

Jong, L. I., and N. L. Abbott, 1998, Langmuir 14, 2235. Lagally, E. T., I. Medintz, and R. A. Mathies, 2001, Anal.
Joo, Y. L., and E. S. G. Shaqfeh, 1991, Phys. Fluids A 3, 1691. Chem. 73, 565.
Joo, Y. L., and E. S. G. Shaqfeh, 1992, Phys. Fluids A 4, 524. Lahann, J., S. Mitragotri, T. N. Tran, H. Kaido, J. Sundaram, I.
Joyce, J. W., 1983, Fluidics—Basic Components and Applica- S. Choi, S. Hoffer, G. A. Somorjai, and R. Langer, 2003,
tions, Special Report HDL-SR-83-9, Harry Diamond Labora- Science 299, 371.
tories. Lajeunesse, E., and G. M. Homsy, 2003, Phys. Fluids 15, 308.
Jun, T. K., and C. J. Kim, 1998, J. Appl. Phys. 83, 5658. Lambelet, P., A. Sayah, M. Pfeffer, C. Philipona, and F.
Juncker, D., H. Schmid, U. Drechsler, H. Wolf, M. Wolf, B. Marquis-Weible, 1998, Appl. Opt. 37, 7289.
Michel, N. de Rooij, and E. Delamarche, 2002, Anal. Chem. Landau, L., and V. Levich, 1942, Acta Physicochim. URSS 17,
74, 6139. 42.
Jung, B., R. Bharadwaj, and J. G. Santiago, 2003, Electro- Landers, J. P., 2003, Anal. Chem. 75, 2919.
phoresis 24, 3476. Larson, R. G., 1988, Constitutive Equations for Polymer Melts
Kamholz, A. E., E. A. Schilling, and P. Yager, 2001, Biophys. J. and Solutions 共Butterworths, Boston兲.
80, 1967. Larson, R. G., 1992, Rheol. Acta 31, 213.
Kamholz, A. E., B. H. Weigl, B. A. Finlayson, and P. Yager, Larson, R. G., 1999, The Structure and Rheology of Complex
1999, Anal. Chem. 71, 5340. Fluids 共Oxford University Press, New York兲.
Kamholz, A. E., and P. Yager, 2001, Biophys. J. 80, 155. Larson, R. G., 2000, Nature 共London兲 405, 27.
Kamholz, A. E., and P. Yager, 2002, Sens. Actuators B 82, 117. Larson, R. G., H. Hu, D. E. Smith, and S. Chu, 1999, J. Rheol.
Kang, K. H., 2002, Langmuir 18, 10318. 43, 267.
Kataoka, D. E., and S. M. Troian, 1997, J. Colloid Interface Larson, R. G., and J. J. Magda, 1989, Macromolecules 22, 3004.
Sci. 192, 350. Larson, R. G., S. J. Muller, and E. S. G. Shaqfeh, 1994, J.
Kataoka, D. E., and S. M. Troian, 1998, J. Colloid Interface Non-Newtonian Fluid Mech. 51, 195.
Sci. 203, 335. Larson, R. G., T. T. Perkins, D. E. Smith, and S. Chu, 1997,
Kataoka, D. E., and S. M. Troian, 1999, Nature 共London兲 402, Phys. Rev. E 55, 1794.
794. Larson, R. G., E. S. G. Shaqfeh, and S. J. Muller, 1990, J. Fluid
Kauffmann, E., N. C. Darnton, R. H. Austin, C. Batt, and K. Mech. 218, 573.
Gerwert, 2001, Proc. Natl. Acad. Sci. U.S.A. 98, 6646. Laser, D. J., and J. G. Santiago, 2004, J. Micromech. Microeng.
Kenis, P. J. A., R. F. Ismagilov, S. Takayama, G. M. Whitesides, 14, R35.
S. L. Li, and H. S. White, 2000, Acc. Chem. Res. 33, 841. Lastochkin, D., R. H. Zhou, P. Wang, Y. X. Ben, and H. C.
Kenis, P. J. A., R. F. Ismagilov, and G. M. Whitesides, 1999, Chang, 2004, J. Appl. Phys. 96, 1730.
Science 285, 83. Lauga, E., and M. P. Brenner, 2004, Phys. Rev. E 70, 026311.
Khandurina, J., T. E. McKnight, S. C. Jacobson, L. C. Waters, Lauga, E., and H. A. Stone, 2003, J. Fluid Mech. 489, 55.
R. S. Foote, and J. M. Ramsey, 2000, Anal. Chem. 72, 2995. Leach, A. M., A. R. Wheeler, and R. N. Zare, 2003, Anal.
Kim, E., Y. N. Xia, and G. M. Whitesides, 1995, Nature Chem. 75, 967.
共London兲 376, 581. Leal, L. G., 1992, Laminar Flow and Convective Transport
Kim, S., and S. J. Karilla, 1991, Microhydrodynamics: Prin- Processes: Scaling Principles and Asymptotic Analysis
ciples and Selected Applications 共Butterworth-Heinemann, 共Butterworth-Heinemann, Boston兲.
Boston兲. Leclerc, E., Y. Sakai, and T. Fujii, 2003, Biomed. Microdevices
King, M. R., and M. Toner, 2003, in Proceedings of MuTAS 5, 109.
2003, edited by M. A. Northrup, K. F. Jensen, and D. J. Har- Lee, C. P., and A. A. Chernov, 2002, J. Cryst. Growth 240, 531.
rison 共Transducer Research Foundation, Squaw Valley兲, Vol. Lee, H. J., D. J. Fermin, R. M. Corn, and H. H. Girault, 1999,
1, pp. 49–52. Electrochem. Commun. 1, 190.
Knight, J. B., A. Vishwanath, J. P. Brody, and R. H. Austin, Lee, J., and C. J. Kim, 2000, J. Microelectromech. Syst. 9, 171.
1998, Phys. Rev. Lett. 80, 3863. Lee, J., H. Moon, J. Fowler, T. Schoellhammer, and C. J. Kim,
Kopp, M. U., A. J. de Mello, and A. Manz, 1998, Science 280, 2002, Sens. Actuators, A 95, 259.
1046. Lee, S. H., D. T. Eddington, Y. M. Kim, W. S. Kim, and D. J.
Krishnan, M., V. M. Ugaz, and M. A. Burns, 2002, Science 298, Beebe, 2003, J. Microelectromech. Syst. 12, 848.
793. Lee, S. M., D. J. Im, and I. S. Kang, 2000, Phys. Fluids 12, 1899.
Kroujiline, D., and H. A. Stone, 1999, Physica D 130, 105. Lee, S. W., D. Y. Kwok, and P. E. Laibinis, 2002, Phys. Rev. E
Krupenkin, T., S. Yang, and P. Mach, 2003, Appl. Phys. Lett. 65, 051602.
82, 316. Lee, S. W., and P. E. Laibinis, 2000, J. Am. Chem. Soc. 122,
Krupenkin, T. N., J. A. Taylor, T. M. Schneider, and S. Yang, 5395.
2004, Langmuir 20, 3824. Lévêque, M. A., 1928, Ann. Mines 13, 201.
Kuhn, W., 1934, Kolloid-Z. 68, 2. Levich, V. G., 1962, Physicochemical Hydrodynamics
Kumar, K. A., and M. D. Graham, 2000, J. Non-Newtonian 共Prentice-Hall, Englewood Cliffs, NJ兲.
Fluid Mech. 89, 337. Levich, V. G., and V. S. Krylov, 1969, Annu. Rev. Fluid Mech.
Kuo, J. S., P. Spicar-Mihalic, I. Rodriguez, and D. T. Chiu, 1, 293.
2003, Langmuir 19, 250. Levich, V. G., and A. M. Kuznetsov, 1962, Dokl. Akad. Nauk
Ladavac, K., K. Kasza, and D. G. Grier, 2004, Phys. Rev. E 70, SSSR 146, 145.
010901. Levitan, J. A., S. Devasenathipathy, V. Studer, T. Thorsen, T.
Ladoux, B., and P. S. Doyle, 2000, Europhys. Lett. 52, 511. M. Squires, and M. Z. Bazant, 2005, Colloids Surf., A 267,
Lagally, E. T., C. A. Emrich, and R. A. Mathies, 2001, Lab 122.
Chip 1, 102. Lichtenberg, J., N. F. de Rooij, and E. Verpoorte, 2002, Talanta

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1022 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

56, 233. Marko, J. F., and E. D. Siggia, 1995, Macromolecules 28, 8759.
Lichtenberg, J., E. Verpoorte, and N. F. de Rooij, 2001, Marmottant, P., and S. Hilgenfeldt, 2003, Nature 共London兲 423,
Electrophoresis 22, 258. 153.
Lide, D. R., 1999, Ed., CRC Handbook of Chemistry and Phys- Marmottant, P., and S. Hilgenfeldt, 2004, Proc. Natl. Acad. Sci.
ics, 79th ed. 共CRC Press, Boca Raton兲. U.S.A. 101, 9523.
Lighthill, J., 1978, J. Sound Vib. 61, 391. Maroja, A. M., F. A. Oliveira, M. Ciesla, and L. Longa, 2001,
Lighthill, J., 2001, Waves in Fluids 共Cambridge University Phys. Rev. E 63, 061801.
Press, Cambridge, England兲. Matsuo, E. S., and T. Tanaka, 1988, J. Chem. Phys. 89, 1695.
Lin, H., B. D. Storey, M. H. Oddy, C.-H. Chen, and J. G. Maurer, J., P. Tabeling, P. Joseph, and H. Willaime, 2003, Phys.
Santiago, 2004, Phys. Fluids 16, 1922. Fluids 15, 2613.
Link, D. R., S. L. Anna, D. A. Weitz, and H. A. Stone, 2004, Maxwell, J., 1879, Philos. Trans. R. Soc. London 170, 231.
Phys. Rev. Lett. 92, 054503. Mazouchi, A., and G. M. Homsy, 2000, Phys. Fluids 12, 542.
Lion, N., et al., 2003, Electrophoresis 24, 3533. Mazouchi, A., and G. M. Homsy, 2001, Phys. Fluids 13, 1594.
Lipman, E. A., B. Schuler, O. Bakajin, and W. A. Eaton, 2003, McConalogue, D. J., and R. S. Srivastava, 1968, Proc. R. Soc.
Science 301, 1233. London, Ser. A 307, 37.
Lippmann, G., 1875, Ann. Chim. Phys. 5, 494. McDonald, J. C., D. C. Duffy, J. R. Anderson, D. T. Chiu, H.
Liu, J., M. Enzelberger, and S. Quake, 2002, Electrophoresis K. Wu, O. J. A. Schueller, and G. M. Whitesides, 2000,
23, 1531. Electrophoresis 21, 27.
Liu, J., C. Hansen, and S. R. Quake, 2003, Anal. Chem. 75, McKinley, G. H., R. C. Armstrong, and R. A. Brown, 1993,
4718. Philos. Trans. R. Soc. London, Ser. A 344, 265.
Liu, R. H., R. Lenigk, R. L. Druyor-Sanchez, J. Yang, and P. McKinley, G. H., J. A. Byars, R. A. Brown, and R. C. Arm-
Grodzinski, 2003, Anal. Chem. 75, 1911. strong, 1991, J. Non-Newtonian Fluid Mech. 40, 201.
Liu, R. H., M. A. Stremler, K. V. Sharp, M. G. Olsen, J. G. McKinley, G. H., A. Oztekin, J. A. Byars, and R. A. Brown,
Santiago, R. J. Adrian, H. Aref, and D. J. Beebe, 2000, J. 1995, J. Fluid Mech. 285, 123.
Microelectromech. Syst. 9, 190. McKinley, G. H., P. Pakdel, and A. Oztekin, 1996, J. Non-
Liu, R. H., J. Yang, M. Z. Pindera, M. Mathavale, and P. Newtonian Fluid Mech. 67, 19.
Grodzinski, 2002, Lab Chip 2, 151. McPherson, A., 1999, Crystallization of Biological Macromol-
Liu, R. H., Q. Yu, and D. J. Beebe, 2002, J. Microelectromech. ecules 共Cold Spring Harbor Laboratory Press,
Syst. 11, 45. Plainview, NY兲.
Liu, T. W., 1989, J. Chem. Phys. 90, 5826. McPherson, A., A. J. Malkin, Y. G. Kuznetsov, S. Koszelak, M.
Long, D., and A. Ajdari, 1996, Electrophoresis 17, 1161. Wells, G. Jenkins, J. Howard, and G. Lawson, 1999, J. Cryst.
Long, D., and A. Ajdari, 1998, Phys. Rev. Lett. 81, 1529. Growth 196, 572.
Long, D., H. A. Stone, and A. Ajdari, 1999, J. Colloid Inter- Meiners, J. C., and S. R. Quake, 2000, Phys. Rev. Lett. 84,
face Sci. 212, 338. 5014.
Long, D., J. L. Viovy, and A. Ajdari, 1996, Phys. Rev. Lett. 76, Melcher, J. R., and W. J. Schwarz, 1968, Phys. Fluids 11, 2604.
3858. Melcher, J. R., and G. I. Taylor, 1969, Annu. Rev. Fluid Mech.
Longuet-Higgins, M. S., 1998, Proc. R. Soc. London, Ser. A 1, 111.
454, 725. Meldrum, D. R., and M. R. Holl, 2002, Science 297, 1197.
Ludviksson, V., and E. N. Lightfoot, 1971, AIChE J. 17, 1166. Mensing, G. A., T. M. Pearce, M. D. Graham, and D. J. Beebe,
Lum, K., D. Chandler, and J. D. Weeks, 1999, J. Phys. Chem. B 2004, Philos. Trans. R. Soc. London, Ser. A 362, 1059.
103, 4570. Miyata, T., N. Asami, and T. Uragami, 1999, Nature 共London兲
Lumma, D., A. Best, A. Gansen, F. Feuillebois, J. O. Radler, 399, 766.
and O. I. Vinogradova, 2003, Phys. Rev. E 67, 056313. Miyauchi, M., N. Kieda, S. Hishita, T. Mitsuhashi, A. Naka-
Luo, Q. Z., S. Mutlu, Y. B. Gianchandani, F. Svec, and J. M. J. jima, T. Watanabe, and K. Hashimoto, 2002, Surf. Sci. 511,
Frechet, 2003, Electrophoresis 24, 3694. 401.
Lutz, B. R., J. Chen, and D. T. Schwartz, 2003, Proc. Natl. Molho, J. I., A. E. Herr, B. P. Mosier, J. G. Santiago, T. W.
Acad. Sci. U.S.A. 100, 4395. Kenny, R. A. Brennen, G. B. Gordon, and B. Mohammadi,
Lyklema, J., 1991, Fundamentals of Interface and Colloid Sci- 2001, Anal. Chem. 73, 1350.
ence, Vol. 2 共Academic, London兲. Moon, H., S. K. Cho, R. L. Garrell, and C. J. Kim, 2002, J.
Lyne, W. H., 1971, J. Fluid Mech. 45, 13. Appl. Phys. 92, 4080.
MacDonald, J. R., 1970, Trans. Faraday Soc. 66, 943. Moorthy, J., and D. J. Beebe, 2002, Lab Chip 2, 76.
MacDonald, M. P., G. C. Spalding, and K. Dholakia, 2003, Moorthy, J., and D. J. Beebe, 2003, Anal. Chem. 75, 292A.
Nature 共London兲 426, 421. Moorthy, J., C. Khoury, J. S. Moore, and D. J. Beebe, 2001,
Mach, P., T. Krupenkin, S. Yang, and J. A. Rogers, 2002, Appl. Sens. Actuators B 75, 223.
Phys. Lett. 81, 202. Morrison, F. A., 1970, J. Colloid Interface Sci. 34, 210.
Magnasco, M. O., 1993, Phys. Rev. Lett. 71, 1477. Morrison, I. D., 1993, Colloids Surf., A 71, 1.
Makohliso, S. A., L. Giovangrandi, D. Leonard, H. J. Mathieu, Mpholo, M., C. G. Smith, and A. B. D. Brown, 2003, Sens.
M. Ilegems, and P. Aebischer, 1998, Biosens. Bioelectron. 13, Actuators B 92, 262.
1227. Mugele, F., and S. Herminghaus, 2002, Appl. Phys. Lett. 81,
Manneville, S., P. Cluzel, J. L. Viovy, D. Chatenay, and F. Ca- 2303.
ron, 1996, Europhys. Lett. 36, 413. Muller, S. J., R. G. Larson, and E. S. G. Shaqfeh, 1989, Rheol.
Manz, A., D. J. Harrison, E. Verpoorte, and H. M. Widmer, Acta 28, 499.
1993, Adv. Chromatogr. 33, 1. Murtsovkin, V. A., 1996, Colloid J. 58, 341.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1023

Murtsovkin, V. A., and G. I. Mantrov, 1991, Colloid J. USSR Matter in Nonuniform Electric Fields 共Cambridge University
53, 240. Press, Cambridge, England兲.
Nadal, F., F. Argoul, P. Hanusse, B. Pouligny, and A. Ajdari, Pollack, L., M. W. Tate, N. C. Darnton, J. B. Knight, S. M.
2002, Phys. Rev. E 65, 061409. Gruner, W. A. Eaton, and R. H. Austin, 1999, Proc. Natl.
Nadal, F., F. Argoul, P. Kestener, B. Pouligny, C. Ybert, and A. Acad. Sci. U.S.A. 96, 10115.
Ajdari, 2002, Eur. Phys. J. E 9, 387. Pollack, L., M. W. Tate, A. C. Finnefrock, C. Kalidas, S. Trot-
Neto, C., V. S. J. Craig, and D. R. M. Williams, 2003, Eur. Phys. ter, N. C. Darnton, L. Lurio, R. H. Austin, C. A. Batt, S. M.
J. E 12, S71. Gruner, and S. G. J. Mochrie, 2001, Phys. Rev. Lett. 86, 4962.
Newman, J. S., 1991, Electrochemical Systems 共Prentice-Hall, Pollack, M. G., R. B. Fair, and A. D. Shenderov, 2000, Appl.
Englewood Cliffs, NJ兲. Phys. Lett. 77, 1725.
Ng, J. D., J. A. Gavira, and J. M. Garcia-Ruiz, 2003, J. Struct. Pollack, M. G., A. D. Shenderov, and R. B. Fair, 2002, Lab
Biol. 142, 218. Chip 2, 96.
Ng, J. M. K., I. Gitlin, A. D. Stroock, and G. M. Whitesides, Pozrikidis, C., 1992, Boundary Integral and Singularity Meth-
2002, Electrophoresis 23, 3461. ods for Linearized Viscous Flow 共Cambridge University
Nguyen, N., X. Huang, and T. K. Chuan, 2002, J. Fluids Eng. Press, Cambridge, England兲.
124, 384. Prins, M. W. J., W. J. J. Welters, and J. W. Weekamp, 2001,
Nyborg, W. L., 1953, J. Acoust. Soc. Am. 25, 68. Science 291, 277.
Nyborg, W. L., 1998, in Nonlinear Acoustics, edited by M. F. Probstein, R. F., 1994, Physicochemical Hydrodynamics 共Wiley,
Hamilton and D. T. Blackstock 共Academic, San Diego兲, pp. New York兲.
207–231. Prost, J., J. F. Chauwin, L. Peliti, and A. Ajdari, 1994, Phys.
O’Brien, R. W., and L. R. White, 1978, J. Chem. Soc., Faraday Rev. Lett. 72, 2652.
Trans. 2 74, 16071626. Purcell, E. M., 1977, Am. J. Phys. 45, 3.
Oddy, M. H., J. G. Santiago, and J. C. Mikkelsen, 2001, Anal. Qian, S. Z., and H. H. Bau, 2002, Anal. Chem. 74, 3616.
Chem. 73, 5822. Qiao, R., and N. R. Aluru, 2003, Sens. Actuators, A 104, 268.
Oh, S. K., M. Nakagawa, and K. Ichimura, 2002, J. Mater. Quake, S. R., 1994, J. Chem. Phys. 101, 4307.
Chem. 12, 2262. Quake, S. R., and A. Scherer, 2000, Science 290, 1536.
Okkels, F., and P. Tabeling, 2004, Phys. Rev. Lett. 92, 038301. Quilliet, C., and B. Berge, 2001, Curr. Opin. Colloid Interface
Sci. 6, 34.
Olagunju, D. O., 1995, ZAMP 46, 946.
Rabai, G., M. Orban, and I. R. Epstein, 1990, Acc. Chem. Res.
Oldroyd, J. G., 1950, Proc. R. Soc. London, Ser. A 200, 523.
23, 258.
Olsson, A., G. Stemme, and E. Stemme, 1999, J. Micromech.
Rallison, J. M., 1984, Annu. Rev. Fluid Mech. 16, 45.
Microeng. 9, 34.
Rallison, J. M., and E. J. Hinch, 1988, J. Non-Newtonian Fluid
Ondarçuhu, T., and M. Veyssie, 1991, J. Phys. II 1, 75.
Mech. 29, 37.
Osada, Y., and S. B. Rossmurphy, 1993, Sci. Am. 268 共5兲, 82.
Ramos, A., A. Gonzalez, A. Castellanos, N. G. Green, and H.
Ottino, J. M., 1989, The Kinematics of Mixing: Stretching,
Morgan, 2003, Phys. Rev. E 67, 056302.
Chaos, and Transport 共Cambridge University Press, Cam-
Ramos, A., H. Morgan, N. G. Green, and A. Castellanos, 1999,
bridge, England兲.
J. Colloid Interface Sci. 217, 420.
Ottino, J. M., and S. Wiggins, 2004, Philos. Trans. R. Soc. Lon-
Ramsey, J. M., 1999, Nat. Biotechnol. 17, 1061.
don, Ser. A 362, 923.
Ramsey, J. M., S. C. Jacobson, and M. R. Knapp, 1995, Nat.
Pabit, S. A., and S. J. Hagen, 2002, Biophys. J. 83, 2872. Med. 共N.Y.兲 1, 1093.
Paegel, B. M., L. D. Hutt, P. C. Simpson, and R. A. Mathies, Raphael, E., 1988, C. R. Acad. Sci., Ser. IIb: Mec., Phys.,
2000, Anal. Chem. 72, 3030. Chim., Astron. 306, 751.
Paik, P., V. K. Pamula, M. G. Pollack, and R. B. Fair, 2003, Lab Raviv, U., P. Laurat, and J. Klein, 2001, Nature 共London兲 413,
Chip 3, 28. 51.
Pakdel, P., and G. H. McKinley, 1996, Phys. Rev. Lett. 77, 2459. Rayleigh, L., 1879, Proc. London Math. Soc. 10, 4.
Patel, P. D., and E. S. G. Shaqfeh, 2003, J. Chem. Phys. 118, Rayleigh, L., 1884, Philos. Trans. R. Soc. London 175, 1.
2941. Reiner, M., 1964, Phys. Today 17 共1兲, 62.
Pearson, J. R. A., 1976, Annu. Rev. Fluid Mech. 8, 163. Ren, H., R. B. Fair, and M. G. Pollack, 2004, Sens. Actuators
Perkins, T. T., S. R. Quake, D. E. Smith, and S. Chu, 1994, B 98, 319.
Science 264, 822. Ren, L. Q., C. Escobedo, and D. Li, 2001, J. Colloid Interface
Perkins, T. T., D. E. Smith, and S. Chu, 1997, Science 276, 2016. Sci. 242, 264.
Perkins, T. T., D. E. Smith, R. G. Larson, and S. Chu, 1995, Ren, L. Q., and D. Q. Li, 2001, J. Colloid Interface Sci. 243,
Science 268, 83. 255.
Peters, E. C., F. Svec, and J. M. J. Frechet, 1997, Adv. Mater. Reyes, D. R., D. Iossifidis, P. A. Auroux, and A. Manz, 2002,
共Weinheim, Ger.兲 9, 630. Anal. Chem. 74, 2623.
Petrie, C. J. S., and M. M. Denn, 1976, AIChE J. 22, 209. Richardson, S., 1973, J. Fluid Mech. 59, 707.
Peykov, V., A. Quinn, and J. Ralston, 2000, Colloid Polym. Sci. Richter, A., D. Kuckling, S. Howitz, T. Gehring, and K. F.
278, 789. Arndt, 2003, J. Microelectromech. Syst. 12, 748.
Phan-Thien, N., 1983, J. Non-Newtonian Fluid Mech. 13, 325. Rife, J. C., M. I. Bell, J. S. Horwitz, M. N. Kabler, R. C. Y.
Phan-Thien, N., 1985, J. Non-Newtonian Fluid Mech. 17, 37. Aeuyeung, and W. J. Kim, 2000, Sens. Actuators, A 86, 135.
Phillips, R. J., 2003, J. Fluid Mech. 481, 77. Riley, N., 2001, Annu. Rev. Fluid Mech. 33, 43.
Pit, R., H. Hervet, and L. Leger, 2000, Phys. Rev. Lett. 85, 980. Ristenpart, W. D., I. A. Aksay, and D. A. Saville, 2003, Phys.
Pohl, H. A., 1978, Dielectrophoresis: The Behavior of Neutral Rev. Lett. 90, 128303.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1024 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Rodd, L. E., T. P. Scott, D. V. Boger, J. J. Cooper-White, and Kenward, L. C. McCormick, and F. Tessier, 2002,
G. H. McKinley, 2004, XIVth International Congress on Rhe- Electrophoresis 23, 3791.
ology 共The Korean Society of Rheology, Seoul, South Korea兲. Slater, G. W., M. Kenward, L. C. McCormick, and M. G.
Roques-Carmes, T., R. A. Hayes, B. J. Feenstra, and L. J. M. Gauthier, 2003, Curr. Opin. Biotechnol. 14, 58.
Schlangen, 2004, J. Appl. Phys. 95, 4389. Smith, D. E., H. P. Babcock, and S. Chu, 1999, Science 283,
Rosslee, C. A., and N. L. Abbott, 2001, Anal. Chem. 73, 4808. 1724.
Rothstein, J. P., and G. H. McKinley, 2001, Phys. Fluids 13, Smith, D. E., and S. Chu, 1998, Science 281, 1335.
382. Smith, S. B., L. Finzi, and C. Bustamante, 1992, Science 258,
Roy, I., and M. N. Gupta, 2003, Chem. Biol. 10, 1161. 1122.
Ruckenstein, E., and N. Churaev, 1991, J. Colloid Interface Solomentsev, Y., and J. L. Anderson, 1995, Ind. Eng. Chem.
Sci. 147, 535. Res. 34, 3231.
Ruckenstein, E., and P. Rajora, 1983, J. Colloid Interface Sci. Solomon, T. H., and I. Mezic, 2003, Nature 共London兲 425, 376.
96, 488. Sondag-Huethorst, J. A. M., and L. G. J. Fokkink, 1994, Lang-
Rush, B. M., K. D. Dorfman, H. Brenner, and S. Kim, 2002, muir 10, 4380.
Ind. Eng. Chem. Res. 41, 4652. Song, H., M. R. Bringer, J. D. Tice, C. J. Gerdts, and R. F.
Russel, W. B., D. A. Saville, and W. R. Schowalter, 1989, Col- Ismagilov, 2003, Appl. Phys. Lett. 83, 4664.
loidal Dispersions 共Cambridge University Press, Cambridge, Song, H., and R. F. Ismagilov, 2003, J. Am. Chem. Soc. 125,
England兲. 14613.
Saeki, F., J. Baum, H. Moon, J. Y. Yoon, C. J. Kim, and R. L. Song, H., J. D. Tice, and R. F. Ismagilov, 2003, Angew. Chem.,
Garrell, 2001, Polym. Mater. Sci. Eng. 85, 12. Int. Ed. 42, 768.
Sakai, T., and R. Yoshida, 2004, Langmuir 20, 1036. Spikes, H., and S. Granick, 2003, Langmuir 19, 5065.
Salemme, F. R., 1972, Arch. Biochem. Biophys. 151, 533. Squires, T. M., and M. Z. Bazant, 2004, J. Fluid Mech. 509, 217.
Sammarco, T. S., and M. A. Burns, 1999, AIChE J. 45, 350. Stemme, E., and G. Stemme, 1993, Sens. Actuators, A 39, 159.
Sammarco, T. S., and M. A. Burns, 2000, J. Micromech. Stockle, R., C. Fokas, V. Deckert, R. Zenobi, B. Sick, B.
Microeng. 10, 42. Hecht, and U. Wild, 1999, Appl. Phys. Lett. 75, 160.
Sartori, A., V. Barbier, and J. L. Viovy, 2003, Electrophoresis Stone, H. A., 1994, Annu. Rev. Fluid Mech. 26, 65.
24, 421. Stone, H. A., and S. Kim, 2001, AIChE J. 47, 1250.
Saville, D. A., 1977, Annu. Rev. Fluid Mech. 9, 321. Stone, H. A., A. Nadim, and S. H. Strogatz, 1991, J. Fluid
Saville, D. A., 1997, Annu. Rev. Fluid Mech. 29, 27. Mech. 232, 629.
Sawano, A., S. Takayama, M. Matsuda, and A. Miyawaki, Stone, H. A., A. D. Stroock, and A. Ajdari, 2004, Annu. Rev.
2002, Dev. Cell 3, 245. Fluid Mech. 36, 381411.
Schasfoort, R. B. M., S. Schlautmann, L. Hendrikse, and A. Stremler, M. A., F. R. Haselton, and H. Aref, 2004, Philos.
van den Berg, 1999, Science 286, 942. Trans. R. Soc. London, Ser. A 362, 1019.
Schilling, E. A., A. E. Kamholz, and P. Yager, 2002, Anal. Stroock, A. D., S. K. Dertinger, G. M. Whitesides, and A.
Chem. 74, 1798. Ajdari, 2002, Anal. Chem. 74, 5306.
Schmalzing, D., L. Koutny, A. Adourian, P. Belgrader, P. Mat- Stroock, A. D., S. K. W. Dertinger, A. Ajdari, I. Mezic, H. A.
sudaira, and D. Ehrlich, 1997, Proc. Natl. Acad. Sci. U.S.A. Stone, and G. M. Whitesides, 2002, Science 295, 647.
94, 10273. Stroock, A. D., R. F. Ismagilov, H. A. Stone, and G. M.
Schneemilch, M., W. J. J. Welters, R. A. Hayes, and J. Ralston, Whitesides, 2003, Langmuir 19, 4358.
2000, Langmuir 16, 2924. Stroock, A. D., and G. J. McGraw, 2004, Philos. Trans. R. Soc.
Schnell, E., 1956, J. Appl. Phys. 27, 1149. London, Ser. A 362, 971.
Schroeder, C. M., H. P. Babcock, E. S. G. Shaqfeh, and S. Chu, Stroock, A. D., M. Weck, D. T. Chiu, W. T. S. Huck, P. J. A.
2003, Science 301, 1515. Kenis, R. F. Ismagilov, and G. M. Whitesides, 2000, Phys.
Schwartz, J. A., J. V. Vykoukal, and P. R. C. Gascoyne, 2004, Rev. Lett. 84, 3314.
Lab Chip 4, 11. Stuart, J. T., 1963, in Laminar Boundary Layers, edited by L.
Selvaganapathy, P., E. T. Carlen, and C. H. Mastrangelo, 2003, Rosenhead 共Clarendon, Oxford兲, pp. 349–408.
Sens. Actuators, A 104, 275. Stuart, J. T., 1966, J. Fluid Mech. 24, 673.
Seong, G. H., and R. M. Crooks, 2002, J. Am. Chem. Soc. 124, Studer, V., G. Hang, A. Pandolfi, M. Ortiz, W. F. Anderson,
13360. and S. R. Quake, 2004, J. Appl. Phys. 95, 393.
Seyrat, E., and R. A. Hayes, 2001, J. Appl. Phys. 90, 1383. Studer, V., A. Pepin, Y. Chen, and A. Ajdari, 2002, Microelec-
Shanahan, M. E. R., and P. G. de Gennes, 1997, C. R. Acad. tron. Eng. 61-2, 915.
Sci., Ser. IIb: Mec., Phys., Chim., Astron. 324, 261. Sugiura, S., M. Nakajima, S. Iwamoto, and M. Seki, 2001,
Shaqfeh, E. S. G., 1996, Annu. Rev. Fluid Mech. 28, 129. Langmuir 17, 5562.
Shaqfeh, E. S. G., S. J. Muller, and R. G. Larson, 1992, J. Fluid Sugiura, S., M. Nakajima, and M. Seki, 2002a, Langmuir 18,
Mech. 235, 285. 5708.
Sharp, M. K., R. D. Kamm, A. H. Shapiro, E. Kimmel, and G. Sugiura, S., M. Nakajima, and M. Seki, 2002b, Langmuir 18,
E. Karniadakis, 1991, J. Fluid Mech. 223, 537. 3854.
Shin, J. Y., and N. L. Abbott, 1999, Langmuir 15, 4404. Sugiura, S., M. Nakajima, J. H. Tong, H. Nabetani, and M.
Sia, S. K., and G. M. Whitesides, 2003, Electrophoresis 24, Seki, 2000, J. Colloid Interface Sci. 227, 95.
3563. Sun, G. X., E. Bonaccurso, V. Franz, and H. J. Butt, 2002, J.
Simonov, I. N., and V. N. Shilov, 1977, Colloid J. USSR 39, Chem. Phys. 117, 10311.
775. Suzuki, A., and T. Tanaka, 1990, Nature 共London兲 346, 345.
Slater, G. W., S. Guillouzic, M. G. Gauthier, J. F. Mercier, M. Tabeling, P., M. Chabert, A. Dodge, C. Jullien, and F. Okkels,

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale 1025

2004, Philos. Trans. R. Soc. London, Ser. A 362, 987. Vekilov, P. G., B. R. Thomas, and F. Rosenberger, 1998, J.
Takayama, S., J. C. McDonald, E. Ostuni, M. N. Liang, P. J. A. Phys. Chem. B 102, 5208.
Kenis, R. F. Ismagilov, and G. M. Whitesides, 1999, Proc. Verheijen, H. J. J., and M. W. J. Prins, 1999, Langmuir 15,
Natl. Acad. Sci. U.S.A. 96, 5545. 6616.
Takayama, S., E. Ostuni, P. LeDuc, K. Naruse, D. E. Ingber, Verpoorte, E., 2002, Electrophoresis 23, 677.
and G. M. Whitesides, 2001, Nature 共London兲 411, 1016. Verpoorte, E., and N. F. De Rooij, 2003, Proc. IEEE 91, 930.
Takayama, S., E. Ostuni, P. LeDuc, K. Naruse, D. E. Ingber, Vinogradova, O. I., 1995, Langmuir 11, 2213.
and G. M. Whitesides, 2003, Chem. Biol. 10, 123. Vinogradova, O. I., 1996, Langmuir 12, 5963.
Takhistov, P., K. Duginova, and H. C. Chang, 2003, J. Colloid Vinogradova, O. I., 1999, Int. J. Min. Process. 56, 31.
Interface Sci. 263, 133. Vinogradova, O. I., and G. E. Yakubov, 2003, Langmuir 19,
Tanaka, T., 1978, Phys. Rev. Lett. 40, 820. 1227.
Tanaka, T., D. Fillmore, S. T. Sun, I. Nishio, G. Swislow, and Viovy, J. L., 2000, Rev. Mod. Phys. 72, 813.
A. Shah, 1980, Phys. Rev. Lett. 45, 1636. Volkmuth, W. D., and R. H. Austin, 1992, Nature 共London兲
Tanaka, T., I. Nishio, S. T. Sun, and S. Uenonishio, 1982, Sci- 358, 600.
ence 218, 467. Volkmuth, W. D., T. Duke, M. C. Wu, R. H. Austin, and A.
Taniguchi, T., T. Torii, and T. Higuchi, 2002, Lab Chip 2, 19. Szabo, 1994, Phys. Rev. Lett. 72, 2117.
Tatsuno, M., 1980, Rep. Res. Inst. Appl. Mech. 共Kyushu Univ.兲 Vologodskii, A., 1994, Macromolecules 27, 5623.
28, 35. Walker, G. M., M. S. Ozers, and D. J. Beebe, 2002, Biomed.
Taylor, G., 1953, Proc. R. Soc. London, Ser. A 219, 186. Microdevices 4, 161.
Taylor, G., 1954, Proc. R. Soc. London, Ser. A 225, 473. Wang, R., K. Hashimoto, A. Fujishima, M. Chikuni, E.
Taylor, G. I., 1934, Proc. R. Soc. London, Ser. A 146, 501. Kojima, A. Kitamura, M. Shimohigoshi, and T. Watanabe,
Taylor, G. I., 1966, Proc. R. Soc. London, Ser. A 291, 159. 1997, Nature 共London兲 388, 431.
Tegenfeldt, J. O., C. Prinz, H. Cao, R. L. Huang, R. H. Austin, Ward, T., and G. M. Homsy, 2001, Phys. Fluids 13, 3521.
S. Y. Chou, E. C. Cox, and J. C. Sturm, 2004, Anal. Bioanal. Ward, T., and G. M. Homsy, 2003, Phys. Fluids 15, 2987.
Chem. 378, 1678. Wasan, D. T., A. D. Nikolov, and H. Brenner, 2001, Science
Teubner, M., 1982, J. Chem. Phys. 76, 5564. 291, 605.
Thamida, S. K., and H. C. Chang, 2002, Phys. Fluids 14, 4315. Washburn, E. W., 1921, Phys. Rev. 17, 273.
Thompson, P. A., and M. O. Robbins, 1990, Phys. Rev. A 41, Washizu, M., 1998, IEEE Trans. Ind. Appl. 34, 732.
6830. Watanabe, K., Yanuar, and H. Udagawa, 1999, J. Fluid Mech.
Thompson, P. A., and S. M. Troian, 1997, Nature 共London兲 381, 225.
389, 360. Weigl, B. H., and P. Yager, 1999, Science 283, 346.
Thormann, W., I. S. Lurie, B. McCord, U. Marti, B. Cenni, and Weislogel, M. M., 1997, AIChE J. 43, 645.
N. Malik, 2001, Electrophoresis 22, 4216. Welters, W. J. J., and L. G. J. Fokkink, 1998, Langmuir 14,
Thorsen, T., S. J. Maerkl, and S. R. Quake, 2002, Science 298, 1535.
580. Westervelt, P. J., 1953, J. Acoust. Soc. Am. 25, 60.
Thorsen, T., R. W. Roberts, F. H. Arnold, and S. R. Quake, Wheeler, A. R., W. R. Throndset, R. J. Whelan, A. M. Leach,
2001, Phys. Rev. Lett. 86, 4163. R. N. Zare, Y. H. Liao, K. Farrell, I. D. Manger, and A.
Tice, J. D., H. Song, A. D. Lyon, and R. F. Ismagilov, 2003, Daridon, 2003, Anal. Chem. 75, 3581.
Langmuir 19, 9127. White, J. M., and S. J. Muller, 2000, Phys. Rev. Lett. 84, 5130.
Torza, S., R. G. Cox, and S. G. Mason, 1971, Philos. Trans. R. White, J. M., and S. J. Muller, 2003, J. Rheol. 47, 1467.
Soc. London, Ser. A 269, 295. Whitesides, G. M., E. Ostuni, S. Takayama, X. Jiang, and D. E.
Trau, M., D. A. Saville, and I. A. Aksay, 1997, Langmuir 13, Ingber, 2001, Annu. Rev. Biomed. Eng. 3, 335.
6375. Whitesides, G. M., and A. D. Stroock, 2001, Phys. Today 54
Tretheway, D. C., and C. D. Meinhart, 2002, Phys. Fluids 14, 共6兲, 42.
L9. Wiggins, S., and J. M. Ottino, 2004, Philos. Trans. R. Soc. Lon-
Tretheway, D. C., and C. D. Meinhart, 2004, Phys. Fluids 16, don, Ser. A 362, 937.
1509. Wilbur, J. L., A. Kumar, E. Kim, and G. M. Whitesides, 1994,
Turner, S. W. P., M. Cabodi, and H. G. Craighead, 2002, Phys. Adv. Mater. 共Weinheim, Ger.兲 6, 600.
Rev. Lett. 88, 128103. Wilding, P., M. A. Shoffner, and L. J. Kricka, 1994, Clin. Chem.
Tyrrell, J. W. G., and P. Attard, 2001, Phys. Rev. Lett. 87, 40, 1815.
176104. Wong, P. K., Y.-K. Lee, and C.-M. Ho, 2003, J. Fluid Mech.
Tyrrell, J. W. G., and P. Attard, 2002, Langmuir 18, 160. 495, 55.
Ugaz, V. M., R. D. Elms, R. C. Lo, F. A. Shaikh, and M. A. Woo, N. J., E. S. G. Shaqfeh, and B. Khomami, 2004a, J.
Burns, 2004, Philos. Trans. R. Soc. London, Ser. A 362, 1105. Rheol. 48, 299.
Unger, M. A., H. P. Chou, T. Thorsen, A. Scherer, and S. R. Woo, N. J., E. S. G. Shaqfeh, and B. Khomami, 2004b, J.
Quake, 2000, Science 288, 113. Rheol. 48, 281.
Unger, M. A., D. A. Kossakovski, R. Kongovi, J. L. Beau- Woolley, A. T., D. Hadley, P. Landre, A. J. deMello, R. A.
champ, J. D. Baldeschwieler, and D. V. Palanker, 1998, Rev. Mathies, and M. A. Northrup, 1996, Anal. Chem. 68, 4081.
Sci. Instrum. 69, 2988. Xia, Y. N., and G. M. Whitesides, 1998, Annu. Rev. Mater. Sci.
Vallet, M., M. Vallade, and B. Berge, 1999, Eur. Phys. J. B 11, 28, 153.
583. Yang, H., and R. L. Chien, 2001, J. Chromatogr., A 924, 155.
van der Vegt, W., W. Norde, H. C. van der Mei, and H. J. Yang, S., T. N. Krupenkin, P. Mach, and E. A. Chandross, 2003,
Busscher, 1996, J. Colloid Interface Sci. 179, 57. Adv. Mater. 共Weinheim, Ger.兲 15, 940.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005

1026 T. M. Squires and S. R. Quake: Microfluidics: Fluid physics at the nanoliter scale

Yang, Z., H. Goto, M. Matsumoto, and R. Maeda, 2000, Microelectromech. Syst. 11, 454.
Electrophoresis 21, 116. Zhan, W., G. H. Seong, and R. M. Crooks, 2002, Anal. Chem.
Yang, Z., S. Matsumoto, H. Goto, M. Matsumoto, and R. 74, 4647.
Maeda, 2001, Sens. Actuators, A 93, 266. Zhang, L., J. M. Koo, L. Jiang, M. Asheghi, K. E. Goodson, J.
Yeh, S. R., M. Seul, and B. I. Shraiman, 1997, Nature 共London兲 G. Santiago, and T. W. Kenny, 2002, J. Microelectromech.
386, 57. Syst. 11, 12.
Yi, G. R., S. J. Jeon, T. Thorsen, V. N. Manoharan, S. R. Zhao, B., J. S. Moore, and D. J. Beebe, 2001, Science 291,
Quake, D. J. Pine, and S. M. Yang, 2003, Synth. Met. 139, 1023.
803. Zhao, B., J. S. Moore, and D. J. Beebe, 2002, Anal. Chem. 74,
Yi, G. R., T. Thorsen, V. N. Manoharan, M. J. Hwang, S. J.
4259.
Jeon, D. J. Pine, S. R. Quake, and S. M. Yang, 2003, Adv.
Zhao, B., J. S. Moore, and D. J. Beebo, 2003, Langmuir 19,
Mater. 共Weinheim, Ger.兲 15, 1300.
1873.
Yoon, J. Y., and R. L. Garrell, 2003, Anal. Chem. 75, 5097.
Zhao, B., N. O. L. Viernes, J. S. Moore, and D. J. Beebe, 2002,
Yoshida, R., T. Takahashi, T. Yamaguchi, and H. Ichijo, 1996, J.
Am. Chem. Soc. 118, 5134. J. Am. Chem. Soc. 124, 5284.
Yoshida, R., K. Takei, and T. Yamaguchi, 2003, Macromol- Zheng, B., L. S. Roach, and R. F. Ismagilov, 2003, J. Am.
ecules 36, 1759. Chem. Soc. 125, 11170.
Young, N. O., J. S. Goldstein, and M. J. Block, 1959, J. Fluid Zhu, X., and E. S. Kim, 1998, Sens. Actuators, A 66, 355.
Mech. 6, 350. Zhu, Y. X., and S. Granick, 2001, Phys. Rev. Lett. 87, 096105.
Yu, C., S. Mutlu, P. Selvaganapathy, C. H. Mastrangelo, F. Zhu, Y. X., and S. Granick, 2002a, Phys. Rev. Lett. 88, 106102.
Svec, and J. M. J. Frechett, 2003, Anal. Chem. 75, 1958. Zhu, Y. X., and S. Granick, 2002b, Langmuir 18, 10058.
Yu, Q., J. M. Bauer, J. S. Moore, and D. J. Beebe, 2001, Appl. Zubritsky, E., 2000, Anal. Chem. 72, 687A.
Phys. Lett. 78, 2589. Zwieniecki, M. A., P. J. Melcher, and N. M. Holbrook, 2001,
Yun, K. S., I. J. Cho, J. U. Bu, C. J. Kim, and E. Yoon, 2002, J. Science 291, 1059.

Rev. Mod. Phys., Vol. 77, No. 3, July 2005