You are on page 1of 101

Near-Surface Slip Flow and Hindered Colloidal

Diffusion at the Nano-Scale
by
Pong-Yu (Peter) Huang
MFL TR 2006-01
Microfluidics Laboratory
Division of Engineering
Brown University
Providence, RI 02912
October 2006
This report is the PhD thesis of Peter Huang, supervised by Prof. Kenneth Breuer in the Division
of Engineering, and submitted to the Graduate School at Brown University in September 2006.
For more information contact: kbreuer@brown.edu
i
Abstract
Aqueous boundary solution slip at a solid surface and hindered diffusion of suspended near-
surface particles are experimentally investigated with a three-dimensional total internal reflection
velocimetry (3D-TIRV) technique. The sub-micrometer penetration depth of evanescent wave offers
low-background-noise images where a 10-nm displacement can be accurately measured. At shear
rates less than 1800 sec
−1
, a shear-induced slip length of less than 100 nanometers is observed
at a glass surface of sub-nanometer roughness. Surface hydrophobicity is also found to aid slip.
However, the notion that electrokinetic and electrostatic forces acting on charged tracer particles
lead to slip length measurement inaccuracy is disputed by experiments conducted with aqueous
solutions consisting of electrolytes, which report slip length of similar magnitudes. Anisotropic
hindered diffusion of suspended near-surface particles is demonstrated with simultaneous three-
dimensional measurements. Hindered diffusion coefficients of 1.5-µm radius particles within 300
nm from the solid wall are shown to be in close agreement with previously published theories.
Lastly, effects of hindered diffusion and wall presence on the accuracies of TIRV and other near-
surface measurement methods are studied via Brownian dynamics simulations. The results reveal
that hindered diffusion causes a minor bias toward values obtained at the wall if a small interval is
allowed between successive image acquisitions, while particle drop-outs lead to over-estimation of
fluidic velocity values if the time interval between image acquisitions is large. It is also found that
the apparent velocities of near-wall particles are not time-invariant. Still, true fluidic velocity values
can be obtained from the particle apparent velocities with proper scaling. Such scaling correction is
applicable to particle tracking velocimetry measurements, but not to correlation based velocimetry
measurements for their lack of accounting of particle drop-outs. Shear effect on near-wall particle
translation in a finite imaging depth results in an asymmetric apparent velocity distribution, which
is another potential source of error for correlation based velocimetries. These findings lead to a
conclusion that particle tracking based velocimetry methods, such as TIRV, are more superior than
correlation based velocimetry methods are in making near-surface fluidic measurements.
Copyright, c _, Pong-Yu (Peter) Huang, 2006
i
Pong-Yu (Peter) Huang was born in Taipei, Taiwan and attended schools there until the
age of fifteen. After graduating high school from International School Manila in 1996,
he attended Cornell University where he received a Bachelor of Arts degree in physics in
2000. Upon graduation, he enrolled at the Division of Engineering, Brown University,
to pursue a graduate degree under the guidance of Professor Kenneth Breuer. In 2002,
he completed a Master of Science degree in engineering, and continued to pursue a
Doctor of Philosophy degree since 2003. He currently resides in the New England region
with his wife and plans to pursue an academic career. Mr. Huang enjoys travelling and
anything related to baseball (except sabermetrics), and holds great interests in economic
perspectives of events in the world. He is a proud fan of the New York Mets since 1999.
ii
This dissertation consists of experimental studies of boundary slip and hindered diffusion in
the nano-scale. Included are an overall review and summary of the whole dissertation in the
first chapter, and a collection of journal papers, each as a subsequent chapter, with details of
experimental methods and results. The contents of the chapters are:
Chapter 1. Introduction. Including basic concepts of slip flow, hindered diffusion and total
internal reflection velocimetry, reviews of previously published works, and discussions of the main
results obtained.
Chapter 2. Paper 1. “Near-surface velocimetry using evanescent wave illumination” by Song-
wan Jin, Peter Huang, Jinil Park, J Y. Yoo and Kenneth S. Breuer. Experiments in Fluids, Vol.
37, pp. 825-833, 2004.
Chapter 3. Paper 2. “Direct measurement of slip velocities using three-dimensional total
internal reflection velocimetry” by Peter Huang, Jeffrey S. Guasto and Kenneth S. Breuer. Journal
of Fluid Mechanics, Vol. 566, pp. 447-464, 2006.
Chapter 4. Paper 3. “Direct measurement of slip length in electrolyte solutions” by Peter
Huang and Kenneth S. Breuer. To be submitted to Physics of Fluids, 2006.
Chapter 5. Paper 4. “Direct measurement of anisotropic near-wall hindered diffusion using
total internal reflection velocimetry” by Peter Huang and Kenneth S. Breuer. To be submitted to
Physical Review E, 2006.
Chapter 6. Paper 5. “Simulations of hindered diffusion in shear flow and its implications for
near-wall velocimetry” by Peter Huang and Kenneth S. Breuer. To be submitted to Physics of
Fluids, 2006.
Chapter 7. Concluding remarks and suggested studies.
iii
I would like to first thank my mentor, Professor Kenneth Breuer, for his guidance and support
through my graduate study. Without his encouragement, constructive criticisms and inexhaustible
knowledge, the projects described in this dissertation would not have been fruitful. I also would
like to thank my dissertation readers, Professor Eric Lauga and Professor Anubhav Tripathi, for
their insightful comments during dissertation revisions. Numerous discussions with Professor Tri-
pathi and Professor Thomas Powers during the years on chemistry and statistical mechanics are
acknowledged and very much appreciated.
A special thank goes to my fellow graduate student and my friend Jeffrey Guasto. Without
his creative ideas and brilliant suggestions many subtleties of experimental methods and results
would not have been found. I also would like to thank Dr. Songwan Jin, for his introduction of the
total internal reflection fluorescent microscopy setup. In addition, I am very grateful to Professor
MinJun Kim, Professor Sylvain Cloutier, Jahn Torres, Teng-Fang Kuo, Brian Burke, Jinkee Lee
and Matt Kerby, for their generous help in micro-fabrication, optical system design and setup,
nano-scale measurements and sharing of lab equipments.
As a lowly ranked graduate student, the administrative and research supports of Mr. Jeffrey
Brown, Mr. Michael Jibitsky, Mr. Brian Corkum, Mr. Charles Vicker and Ms. Virginia Novak
are especially appreciated. Without their kind assistance many research projects would not have
moved forward as smoothly as they did.
Finally I would like to thank my parents, my sister and my extended family for their love,
support and best wishes over the years. My greatest gratitude goes to my wife, for her love,
patience and encouragement throughout the years of graduate study, and for putting up with me
during the ups and downs of experiments.
iv
Contents
1 Introduction 1
1.1 BOUNDARY SLIP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 HINDERED DIFFUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 TOTAL INTERNAL REFLECTION VELOCIMETRY . . . . . . . . . . . . . . . . 4
1.4 SUBTLETIES OF NEAR-WALL SHEAR FLOW . . . . . . . . . . . . . . . . . . . . 6
2 Paper 1: Near-surface velocimetry using evanescent wave illumination 8
2.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 EXPERIMENTAL PROCEDURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Cross-stream Velocity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.2 Streamwise Velocity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.3 Comparisons with Numerical Simulations . . . . . . . . . . . . . . . . . . . . 18
2.4 CONCLUDING REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Paper 2: Direct measurement of slip velocities using three-dimensional total
internal reflection velocimetry 22
3.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 THEORETICAL CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 Total Internal Reflection Microscopy . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.2 Emission Intensity of Fluorescent Particles . . . . . . . . . . . . . . . . . . . 24
3.2.3 Near-Surface Shear Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 EXPERIMENTAL PROCEDURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.1 Materials and Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.2 Velocimetry Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.1 Validation of Intensity Calibration Curve . . . . . . . . . . . . . . . . . . . . 31
3.4.2 Hindered Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.3 Velocity Distributions of Particles . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.4 Measurements of Apparent Slip Velocities . . . . . . . . . . . . . . . . . . . . 34
3.5 CONCLUDING REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4 Paper 3: Direct measurement of slip length in electrolyte solutions 39
5 Paper 4: Direct measurement of anisotropic near-wall hindered diffusion using
total internal reflection velocimetry 43
5.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
v
5.2 THEORY OF HINDERED DIFFUSION . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3 EXPERIMENTAL PROCEDURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4 RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.5 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6 Paper 5: Simulations of hindered diffusion in shear flow and its implications for
near-wall velocimetry 50
6.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.2 THEORIES AND COMPUTATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.2.1 The Langevin Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.2.2 Effects of Shear on Particle Velocities . . . . . . . . . . . . . . . . . . . . . . 53
6.2.3 Hindered Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2.4 Implementation of Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.3 RESULTS AND DISCUSSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.3.1 Sedimentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.3.2 Particle Displacement Due to Hindered Diffusion . . . . . . . . . . . . . . . . 55
6.3.3 Particle Drop-in’s and Drop-out’s . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3.4 Horizontal Apparent Velocity Distributions . . . . . . . . . . . . . . . . . . . 60
6.3.5 Time Evolution of Apparent Velocity Distributions . . . . . . . . . . . . . . . 62
6.4 CONCLUDING REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
7 Concluding Remarks and Suggested Studies 68
A Calibration of beam incident angle 70
B Shear-induced lift force (or more precisely, lack of ) on near-wall submicron-
particles 73
C Evanescent wave image of micron-sized fluorescent particles 75
vi
List of Figures
1.1 Schematic of total internal reflection fluorescence microscopy . . . . . . . . . . . . . 5
2.1 Schematic of objective-based TIRFM . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Schematic of the experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Particle images of different illumination methods: widefield (direct, flood) illumination 12
2.4 Particle images of different illumination methods: near-wall TIRFM image . . . . . . 13
2.5 Schematic of near-wall particles moving near the surface illustrating the observation
range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Distribution of particle velocity vectors of 200-nm particles . . . . . . . . . . . . . . 14
2.7 Cross-stream Brownian motion of particles in an infinite medium and near wall . . . 15
2.8 Experimentally-measured distribution of streamwise velocities of 200-nm diameter
particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.9 Apparent particle velocity vs. shear rate . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.10 Streamwise velocity distribution of 200-nm particles determined from Monte Carlo
simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.11 Streamwise apparent velocity predicted by simulation . . . . . . . . . . . . . . . . . 20
3.1 Objective-based total internal reflection fluorescence microscopy . . . . . . . . . . . . 24
3.2 The emission intensity distribution of particles in a uniform concentration field . . . 25
3.3 The predicted position distribution of particles with 0.5 < I
e
/I
e
0
< 1 in a uniform
concentration field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 The ratio of statistical apparent velocity of particles and mean fluid velocity under
no slip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5 Image acquisition system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.6 Fluorescent particle intensity as a function of its distance to the glass surface . . . . 30
3.7 Observed fluorescent particle intensity distribution and its probability density func-
tion (PDF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.8 Ratio of hindered diffusion coefficients (D
exp
) for Brownian motion parallel to a surface 32
3.9 Distribution of observed particle streamwise velocities under various shear rates . . . 33
3.10 Distribution of observed particle streamwise velocities after scaling . . . . . . . . . . 34
3.11 Experimental apparent velocity of particles in a shear flow over a hydrophilic surface 35
3.12 Experimental apparent velocity of particles in a shear flow over a hydrophobic surface 36
3.13 The additional slip length due to surface hydrophobicity . . . . . . . . . . . . . . . . 36
4.1 Objective-based total internal reflection velocimetry (TIRV) system . . . . . . . . . 40
4.2 Measured slip lengths of aqueous solutions . . . . . . . . . . . . . . . . . . . . . . . . 42
5.1 Schematic of total internal reflection velocimetry . . . . . . . . . . . . . . . . . . . . 45
vii
5.2 Intensity calibration of 1.5-µm radius fluorescent particles in evanescent field . . . . 46
5.3 Hindered diffusion correction vs. particle/glass gap size . . . . . . . . . . . . . . . . 47
6.1 A schematic of the simulation geometry . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.2 Sedimentation of large particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.3 Sedimentation of small particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.4 Particle displacement distribution due to hindered diffusion . . . . . . . . . . . . . . 58
6.5 Spatial distribution of small particles due to hindered diffusion . . . . . . . . . . . . 59
6.6 Spatial distribution of particles from a finite layer at various times . . . . . . . . . . 59
6.7 Percentage of particles remaining in imaging range . . . . . . . . . . . . . . . . . . . 60
6.8 Apparent velocity distribution of particles remaining in a finite imaging range . . . . 61
6.9 Normalized apparent velocity distribution of particles in a finite imaging range . . . 61
6.10 Apparent velocity distribution of particles from various imaging ranges . . . . . . . . 62
6.11 Collapsed apparent velocity distribution of particles at various imaging range . . . . 63
6.12 Time evolution of apparent velocity distribution . . . . . . . . . . . . . . . . . . . . 63
6.13 Time evolution of mean particle apparent velocity . . . . . . . . . . . . . . . . . . . 64
6.14 Rescaled time evolution of mean particle apparent velocity . . . . . . . . . . . . . . . 65
6.15 Schematic of potential paths for particle translation . . . . . . . . . . . . . . . . . . 66
A.1 Schematic of laser beam refraction at the glass/air interface . . . . . . . . . . . . . . 71
A.2 Photo and schematic of beam incident angle measurement . . . . . . . . . . . . . . . 71
A.3 Plot of laser beam angle vs. converging lens position . . . . . . . . . . . . . . . . . . 72
C.1 Schematic of COMSOL simulation geometry . . . . . . . . . . . . . . . . . . . . . . . 76
C.2 COMSOL simulation of total internal reflection . . . . . . . . . . . . . . . . . . . . . 76
C.3 Intensity decay of evanescent field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4 COMSOL simulation of a particle with d/λ = 0.39 in evanescent field . . . . . . . . 77
C.5 COMSOL simulation of a particle with d/λ = 6 in evanescent field . . . . . . . . . . 78
C.6 COMSOL simulation of a particle with d/λ = 12 in evanescent field . . . . . . . . . 78
C.7 Overall evanescent energy in the suspended particles . . . . . . . . . . . . . . . . . . 79
C.8 Cross-sectional intensity profile of a 3-µm particle (d/λ = 6) . . . . . . . . . . . . . . 80
C.9 Digital image of a 3-µm particle (d/λ = 6) in contact with glass surface . . . . . . . 81
C.10 Cross-sectional intensity profile of a 6-µm particle (d/λ = 12) . . . . . . . . . . . . . 82
C.11 Digital image of a 6-µm particle (d/λ = 12) in contact with glass surface . . . . . . . 82
viii
List of Tables
6.1 Sample values of the Peclet number and the sedimentation coefficient . . . . . . . . 53
6.2 Representative values of the non-dimensional time between consecutive image acqui-
sition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
ix
Chapter 1
Introduction
Since the introduction of the first microfluidic device, these miniaturized fluidic manipulation sys-
tems have been regarded as one of the most promising technologies of the late twentieth century.
In particular, investigations into its application in biotechnology has been the most intense. Exam-
ples of such applications include immunosensors [1], reagent mixing [2], content sorter [3] and drug
delivery [4]. Microfluidic devices are very attractive in biotechnology over conventional technology
because they require small sample volume and produce rapid results. Indeed, for chemical reac-
tions whose reaction sites are at solid surfaces, the high surface-area-to-volume-ratio characteristic
of microfluidics offers a much higher efficiency [5]. On the other hand, the high surface-area-to-
volume-ratio also means that near-surface phenomena will have a much larger influence on the bulk
of the fluid content, for example the viscous drag on the channel walls. In both cases, understand-
ing of interactions between the fluid content and the solid boundary is critical in designing and
analyzing microfluidic devices.
Under most circumstances, the solid boundary is viewed as rigid and inert such that physical
and structural changes due to fluidic forces are nonexistent. Thus the majority of interesting
surface-induced physical phenomena occur in the near-surface region of the fluid phase and can be
categorized into two groups: (1) changes of the fluid mechanical characteristics due to the presence
of the solid surface; (2) interactions between the dissolved molecules, suspended particulates and
the solid surface. Examples of physical phenomena in the former group include electrokinetic flow
[6], slip flow [7] and surface chemistry directed flow [8, 9], while particle or cell adhesion [10] and
detachment [11], increased hydrodynamic drag [12], electrostatic interactions [13] and effects of
depletion layers [14] are effects of the latter group.
1.1 BOUNDARY SLIP
One of the fluid mechanics questions that have remained unanswered over centuries is the slip flow
and its origin. Unlike polymer solutions whose slip behaviors have been thoroughly investigated
[15, 16, 17], the existence of boundary slip in a Newtonian liquid has been a controversial topic. Even
though the idea of a slip flow has been considered by scientists as early as Newton [18], empirical
data up to the late twentieth century had deemed it negligible. The strong experimental support
of such assumption could be due to the fact that the magnitude of slip velocity was much smaller
than the measuring instrument accuracy and thus remained undetectable. Another possibility is
the relatively insignificant impact of slip on the fluid bulk flow in the macroscopic scale. Westin et
al. [19] calculated that the ratio of slip velocity to bulk velocity becomes very small if the channel
height is more than a few hundred micrometers. Therefore when macroscopic flow measurements
1
2
were conducted slip effect could be hardly noticeable.
In the past decade slip flow began to receive more attention again, because some earlier ex-
periments [20, 21] and molecular dynamics simulations [22, 23, 24] have reported slip boundary
conditions in Newtonian liquid. Since then many experimental attempts were made to confirm the
existence and the magnitude of slip velocities, with varying degrees of accuracy and uncertainty. In
some of these experiments, slip velocity was indirectly inferred from measurement of other physical
quantities. For instance, the relationship between applied pressure drops and measured flow rates
was used by Choi et al. [25] to infer slip velocity, while the forces required to move two surfaces
separated by a thin film of the test fluid were employed Zhu & Granick [26], Neto et al. [27] and
Cottin-Bizonne et al. [28] to calculate slip velocities. Others took a direct approach of studying
slip by measuring near-surface flow velocities. Examples of the direct measurements include fluo-
rescence recovery after photobleaching (FRAP) by Pit et al. [29] and micro-scale particle image
velocimetry (µPIV) by Tretheway & Meinhart [30] and Joseph & Tabeling [31].
These experimental studies produced various magnitudes and dependencies of boundary slip.
The commonly used formulation to characterize boundary slip is the slip length, δ, which is based
on the Navier hypothesis that the velocity of a fluid at a solid surface is proportional to the shear
stress. That is,
δ =
U
slip
˙ γ
s
, (1.1)
where U
slip
is the slip velocity and ˙ γ
s
is the fluid shear rate at the solid surface. Experimentally
measured slip lengths were reported to range from micrometers [20, 30, 32], hundreds of nanometers
[29, 33] to less than 100 nanometers [25, 26, 27, 28, 31]. Disagreement also exist on whether the
slip behavior is shear-dependent [21, 25, 26, 27] or shear-independent [20, 28, 29, 31, 32]. Of
particular interest to microfluidics is aqueous solution slip over different surface conditions as water
is the most common solvent in biological and chemical applications. Although it is generally agreed
among research findings that surface hydrophobicity aids slip effect [25, 30], again the magnitude
of such an effect remains controversial.
The ability of total internal reflection velocimetry (TIRV, to be introduced later) in performing
near-surface measurements makes it a perfect technique for direct observation of slip velocities in a
shear flow. In Paper 1: Near-surface velocimetry using evanescent wave illumination (chapter 2), we
report on our first attempt to measure slip velocities of deionized water using the TIRV technique.
In this report, it is shown that velocities of fluid elements within a few hundred nanometers from
a solid surface can be accurately measured, demonstrating the potential of the TIRV technique in
making direct measurements of slip velocities. Although a quantitative conclusion of slip length
magnitude could not be drawn due to the uncertainty in determination of the exact imaging range,
a qualitative conclusion is reached that surface slip length of an aqueous solution would be small
and surface chemistry-dependent, and a slip length above a few hundred nanometers is unlikely.
The accuracy of slip measurements is further improved with the implementation of three-
dimensional tracking in the TIRV technique. In Paper 2: Direct measurement of slip velocities
using three-dimensional total internal reflection velocimetry (chapter 3), we report on a slip length
of deionized water being less than 100 nm, and an additional slip length of 16 nm is found at-
tributable to surface hydrophobicity. In conducting this experiment, several physical factors that
are suspected of inducing slip are also examined. With measurements conducted at glass sur-
faces of sub-nanometer surface roughness, the obtained slip length values agree with the surface-
roughness dependency previously reported by Granick et al. [34] The issue of apparent slip caused
by nano-bubbles at the glass surface is additionally disputed with experiments conducted with
sub-nanometer surface roughness and degassed test fluids.
3
Still, questions remained on whether other physical forces, particularly the electrostatic and
electrokinetic effects proposed by Lauga et al. [7, 35], would lead to false slip results in TIRV mea-
surements. We address such a concern in Paper 3: Direct measurement of slip length in electrolyte
solutions (chapter 4), by repeating the slip measurements with electrolyte solutions whose ionic
concentrations have been predicted to reduce the apparent slip effect to sub-molecular level. It is
observed that the presence of electrolytes has no effect on the measured slip lengths, suggesting
that the observed slip velocities are most likely not due to electrostatic and electrokinetic effects,
but are consequences of true boundary slip.
1.2 HINDERED DIFFUSION
Another near-surface phenomenon with great fluid mechanical and mass transport implications is
hindered diffusion of colloidal particles. The concept of near-wall hindered diffusion stems from an
increased frictional coefficient on a near-wall object. Einstein proposed that the diffusion coefficient
of a colloidal object is the ratio of the fluid thermal energy to the object’s frictional coefficient in
Stokes flow. Thus if a particle experiences an increase in fluidic friction as it approaches a solid
wall, its diffusive translation or Brownian motion will be hindered. Brenner [36] and Goldman et al.
[37, 38] were the first to analytically solve the increased frictional coefficients of a near-wall sphere
with a lubrication formulation. Their results were later re-confirmed by Chaoui & Feuillebois [12]
with an spherical harmonics expansion method.
Besides being a fluid mechanics theory, hindered diffusion has additional implications in chemical
engineering and drug delivery systems since diffusion is the dominant mass transport mechanism
in microfluidics. Because of its significance, advancement of near-surface measurement techniques
had lead to several experimental studies of hindered diffusion. Employed experimental techniques
include evanescent light scattering off freely suspended particles by Bevan & Prieve [39], evanescent
light-scattering spectroscopy by Hosoda et al. [40], catch-and-release video microscopy with optical
tweezers by Lin et al. [41], evanescent light scattering coupled with optical radiation pressure
by Oetama & Walz [42] and three-dimensional ratiometric total internal reflection fluorescence
microscopy with nanoparticles by Banerjee & Kihm [43]. Most of these studies have reported
hindered diffusion coefficients agreeing to values derived from the theories of Brenner and Goldman
et al.
In the theories of Brenner and Goldman et al., the increased friction factor is a function of
particle radius and gap size between the particle and the solid wall. Unlike in the fluid bulk,
the fluidic friction also becomes anisotropic in the near-wall region. In other words, the hindered
diffusion coefficient is different for directions normal and parallel to the solid surface. Few studies
have attempted to measure such anisotropicity simultaneously due to experimental limitations. Lin
et al. [41] demonstrated such anisotropicity for particles whose gap-size-to-radius ratio is greater
than one, and found hindered diffusion coefficient values agreeing with theories. Another study
that reported anisotropic hindered diffusion coefficients for gap-size-to-radius ratio greater than
one was conducted by Banerjee & Kihm [44]. However, their study reported that the measured
diffusion coefficient in the direction normal to the solid wall agreed with the theoretical values only
when large particles were tested.
By applying the 3D-TIRV technique to freely suspended micron-sized fluorescent particles, we
aim to simultaneously observe the three-dimensional anisotropic hindered diffusion for gap-size-to-
radius ratio much less than one. The theory under this condition is vastly different from that of the
gap-size-to-radius ratio greater than one, and based on our literature survey no such experimen-
tal confirmation of the theory has been reported. In Paper 4: Direct measurement of anisotropic
4
near-wall hindered diffusion using total internal reflection velocimetry (chapter 5), we demonstrate
that the three-dimensional tracking technique of TIRV can be adopted into 3D displacement mea-
surements of freely suspended 1.5-µm radius particles. The displacement measurements reveal that
the hindered diffusion coefficients are in close agreement with the theoretical values predicted by
the asymptotic solutions of Brenner and Goldman et al. for gap-size-to-radius ratio much less than
one, and hindered diffusion anisotropicity is simultaneously demonstrated in all data sets.
1.3 TOTAL INTERNAL REFLECTION VELOCIMETRY
With so much interest in near-surface phenomena, researchers have developed various techniques
to study them. Optical microscopy has been widely used to observe interactions in the micrometer
scale. However, as fabrication technology advances, the definition of “near-surface” has also evolved
from microscopic and to the nano-scale. Traditional optical techniques are no longer sufficient now
because the visible wavelength is the physical limit of optical resolution (∼ 0.5 µm). A frequently
used technique to overcome this obstacle is evanescent wave imaging [45]. In this technique, a beam
of collimated laser light is brought to a solid/liquid interface through the solid phase at a large
incident angle, as shown in figure 1.1. Since solids typically have higher indices of refraction than
fluids do, total internal reflection occurs at the interface if the incident angle is greater than the
critical angle predicted by Snell’s Law. In fact, at the nano-scale this reflection is not “total”, but
a small fraction of energy, called evanescent waves, extends into the fluid phase. In the fluid phase
the evanescent wave energy or intensity, I, decays exponentially with distance, z, away from the
two-medium interface by
I(z) = I
0
e
−z/p
, (1.2)
where I
0
is the intensity at the interface and p is known as the evanescent wave penetration depth.
The penetration depth, which characterizes the length scale of the evanescent field, can be calculated
from
p =
λ
0

_
n
2
s
sin
2
θ −n
2
l
_

1
2
, (1.3)
where λ
0
is the wavelength of the incident light, θ is the beam incident angle and n
s
and n
l
are
the refractive indices of the solid and liquid, respectively. For visible light, the evanescent wave
energy decays to zero within a few hundred nanometers. Thus by using the evanescent wave as
an illumination source, one can observe dynamics of fluids and suspended particulates within a
few hundred nanometers from the solid surface, while contents outside of this layer would remain
invisible. Consequently near-surface phenomena can be observed with clarity and without the
interfering noise contributed from objects in the fluid bulk.
Since the 1970’s, biologists have combined fluorescence and evanescent wave imaging in studying
near-surface biological events [46, 47]. Termed “total internal reflection fluorescence microscopy”
(TIRFM), this technique allows observation of fluorescently labelled samples placed in the evanes-
cent field and under a high magnification microscope. Fluorescence is a quantum optical effect
of a unique group of molecules called fluorophores, whose quantum states can be excited by a
specific range of visible light and release photons of a longer wavelength after initial excitation.
Their molecular sizes are excellent for probing near-surface regions at the nano-scale. Examples of
TIRFM include the previously mentioned fluorescent recovery after photobleaching (FRAP) and
fluorescence resonance energy transfer (FRET). In FRAP, the fluorophores of a chosen spot in-
side the microscope field of view is purposely photobleached with a high power laser beam. The
light-emitting fluorophores in the surrounding area will diffuse into this “dark” spot immediately
following the photobleaching. By measuring the diffusivity of these light-emitting fluorophores, one
5
θ
z
liquid
solid
~ penetration depth
Figure 1.1: Schematic of total internal reflection fluorescence microscopy (TIRFM). A fluorescent
particle suspended in water was placed in an evanescent field. If the illumination beam incident
angle, θ, is greater than the critical angle predicted by Snell’s law, total internal reflection occurs
at the solid/liquid interface. The evanescent energy then illuminates the encapsulated fluorophores
inside the particle.
can study the targeted molecule mobility. For instance, FRAP has been reported in measurement
of protein mobility and activity in living cells [48]. FRET, on the other hand, takes advantages
of a quantum energy transfer that occurs between matched fluorescent molecules of two different
wavelengths [49]. Because such energy transfer can only occur if the two molecules are less than
10 nm apart, they offer an opportunity to examine interactions in the molecular level.
Surprisingly, a long time had passed before physical scientists finally caught up with the merits
of evanescent wave imaging. At first, evanescent energy scattered off non-fluorescent suspended
particles was used in quantitative measurements of near-surface dynamics. Such experimental
studies include measurement of surface charge densities on suspended particles by von Grunberg
et al. [50], quantifying double-layer repulsion between a colloidal particle and a glass substrate by
Flicker et al. [51], effects of particle size on near-wall particle depletion by Kok et al. [14], charged
particle adsorption by Kun & Fendler [52], and hindered diffusion by Bevan & Prieve [39]. Only in
the past decade were fluorescent tracer particles used in microscopic dynamics measurements.
Particle-based velocimetry has long been used in fluid visualization and measurement [53]. It
is based on an intuitive and for most part correct assumption that the seeding tracer particles are
carried by the fluid surrounding them, and therefore their translational velocities must be that of the
local fluid elements. Apparent velocities of the tracer particles are calculated based on displacements
of the tracer particles and the time between successive particle imaging. When particle-based
velocimetry methods were adopted to study microfluidics, sub-micron fluorescent tracer particles
were used to minimize light scattering and imaging noise while attaining spatial resolutions of
tens of nanometers [54]. Santiago et al. [55] were the first to demonstrate a correlation-based
microscopic particle image velocimetry (µPIV) to map out a velocity field around a Hele-Shaw cell.
In their experiment, cross-correlation analysis was performed on successive images of fluorescent
tracer particles to identify the most probable ensemble displacements. Since their report, µPIV
has been the work horse among diagnostic techniques of microfluidics.
When applying to near-surface measurements, µPIV does suffer a major drawback. Its resolu-
tion in the direction perpendicular to the imaging focal plane is limited by the focal depth of the
6
microscope objective lens and is at least 0.5 µm [31]. Consequently, µPIV could not distinguish
physical features within 0.5 µm from a solid surface. Zettner & Yoda [56] overcame µPIV’s defi-
ciency by combining it with evanescent wave illumination. The short effective range of evanescent
wave allowed them to measure fluid velocities that were within 280 nm from a solid surface. Li
et al. [57] further extended this nano-PIV technique into a multilayer correlation analysis, where
quantitative measurement can be performed on events that occur within 100 nm from the surface.
Zettner & Yoda’s nano-PIV is undoubtedly a novel technique. However it still has two significant
limitations. Firstly, rapid Brownian motions of nanoparticles could lead to a large amount of particle
“drop-in” and “drop-out”, or events where particles diffuse in and out of the evanescent field between
consecutive frames. Since the statistical analysis of nano-PIV does not make an attempt to treat the
drop-in’s and drop-out’s, significant measurement errors could persist. Secondly, because diffusion
is treated as measurement noise in the cross-correlation analysis of the nano-PIV, it is unsuitable
if diffusive dynamics is of interest.
In Paper 1: Near-surface velocimetry using evanescent wave illumination (chapter 2), we present
a technique, called “total internal reflection velocimetry” (TIRV), as an alternative to nano-PIV
in near-surface measurements. By combining TIRFM with tracking of individual tracer particle,
Brownian motions of the particle ensemble can be monitored and statistically analyzed, leading
to a more accurate accounting of all contributing particles. In Paper 2: Direct measurement of
slip velocities using three-dimensional total internal reflection velocimetry (chapter 3), we extend
the TIRV technique to three-dimensional, further improving its accuracy and applicability. It is
observed that the peak intensity of a fluorescent particle decays exponentially in the same length
scale as the evanescent wave energy, and thus a particle’s peak intensity can be used to infer its
distance from the solid surface.
As mentioned previously, we use the TIRV technique to measure aqueous solution boundary
slip and near-surface hindered diffusion of spherical particles. In Paper 2: Direct measurement
of slip velocities using three-dimensional total internal reflection velocimetry (chapter 3) and in
Paper 3: Direct measurement of slip length in electrolyte solutions (chapter 4), 100- and 200-nm
radius fluorescent particles are employed in 3D-TIRV experiments to measure slip lengths over
hydrophilic and hydrophobic surfaces. The analysis of the image data reveals that a statistical
approach is needed in making near-surface velocimetry measurement even for particles whose size
variation is as small as a few percent, owing to the sensitivity of evanescent wave imaging. In Paper
4: Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection
velocimetry (chapter 5), the TIRV technique is extended to observing anisotropic near-wall hindered
diffusion of 1.5-µm radius fluorescent particles. Under evanescent wave imaging, the peak intensity
of these large particles follow an exponential decay relation that is also identical to the exponential
decay of evanescent energy, and thus the same three-dimensional particle tracking principle applies.
Both slip and hindered diffusion studies demonstrate the applicability and versatility of the TIRV
technique in making near-wall particle velocity and displacement measurements.
1.4 SUBTLETIES OF NEAR-WALL SHEAR FLOW
As previous discussion have emphasized, Brownian motion of particles in the near-surface region is a
non-negligible mechanism and has significant implications in the accuracy of near-wall velocimetry.
Many of the issues that need to be considered, however, could not be easily assessed through
experimental data. Thus numerical approaches offer insightful alternatives to explore the subtleties
of near-wall flow.
The main concern in near-wall velocimetry is the ensemble behavior of tracer particles in the
7
time interval between consecutive image acquisitions, and more importantly, how such ensemble
behavior affects the accuracy of velocity measurements. Two equally viable numerical approaches
have been proposed to study Brownian motion-related problems [58]. In the Fokker-Planck ap-
proach, the time evolution of the particle ensemble phase space configuration function is computed
through numerical solution of the momentum-position partial differential equations. The ensemble
behaviors are then obtained from probability analysis using the phase space configuration function.
Recently, Sadr et al. [59] had taken the Fokker-Planck approach to study the diffusion-induced bias
in near-wall velocimetry.
The other approach takes advantage of the Langevin equation in which the displacement of
a particle consists of a deterministic component (such as displacement of the suspending fluid)
and a stochastic component (such as Brownian motion). With a proper time step size, initial and
boundary conditions, ensemble behavior of particles can be obtained through statistical analysis of
a large number of repeated simulations. Also called Brownian Dynamics (BD) simulations for its
direct emulation of Brownian motions, this approach has been used in studying the binding rates
between attached molecules [60], colloidal particle deposition in a microchannel flow [61], accuracy
of potential energy profiles determined by evanescent wave scattering [62] and the accuracy of
nano-PIV [63]. Its intuitive resemblance to a physical velocimetry experiment has gained some
popularity in colloidal studies.
We are most interested in the ensemble behaviors of near-surface particles in a shear flow, and
the effects of the wall presence and hindered diffusion on the measured velocities because these
are issues that we have noticed during our TIRV measurements of boundary slip and anisotropic
hindered diffusion. In Paper 5: Simulations of hindered diffusion in shear flow and its implications
for near-wall velocimetry (chapter 6), the relative strengths of near-wall shear and diffusive forces
are explored through Brownian dynamics simulations. It is observed that the shear force dominates
over diffusion at Pe > 3. In addition, asymmetric shape and width of apparent velocity distributions
are attributed to the shear effect, and we believe it can lead to significant inaccuracy in nano-PIV
if not treated appropriately.
The same Brownian dynamics simulation also provides a tool to assess the effects of hindered
diffusion and wall presence on various near-wall velocimetry techniques. With diffusive time and
particle radius as time and length scales, respectively, it is found that if the time interval between
successive particle image acquisitions is short, hindered diffusion causes an under-estimation of
fluid velocity due to biased sampling of tracer particles very close to the wall. However, if the
time interval between successive particle image acquisitions is long, a significant portion of the
tracer particles would sample planes of higher velocities, resulting in an over-estimation of overall
fluid velocity. Particle drop-in’s and drop-out’s, are found to be a potential source of measurement
error for nano-PIV because of the rapid Brownian motion of nanoparticles. The combined effect of
the observations described above is that near-wall particle-based velocimetry measurement is not
time-invariant. Still, the results found in the Brownian dynamics simulation offer a way where the
true fluidic velocities can be inferred from particle apparent velocities through proper scaling. This
scaling method, however, is available only for TIRV but not for nano-PIV, making TIRV the more
superior method in near-surface measurements.
Chapter 2
Paper 1: Near-surface velocimetry
using evanescent wave illumination
Total Internal Reflection Velocimetry (TIRV) is used to measure particle motion in the near-wall
region of a microfluidic system. TIRV images are illuminated with the evanescent field of an
incident laser pulse and contain only particles that are very close to channel surface. Sub-micron
sized fluorescent particles suspended in water are used as seed particles and their images are analyzed
with a PTV algorithm to extract information about apparent slip velocity. At relatively low shear
rates (less than 2500 sec
−1
), a velocity, proportional to the shear rate was observed. The statistical
difference between velocities measured over hydrophilic and hydrophobic surfaces was found to be
minimal. The results suggest that the slip length, if present, is less than 25 nm, but uncertainty
regarding the exact character of the illumination field prevents a more accurate measurement at this
time. Numerical simulations are presented to help understand the results and to provide insight into
the mechanisms that result in the experimentally observed distributions. Issues associated with the
accuracy of the experimental technique and the interpretations of the experimental results are also
discussed.
2.1 INTRODUCTION
Over the past few years, studies regarding near-wall fluid motion have intensified because of the
emergence of micro-electro-mechanical-systems (MEMS), where the fluid-surface interaction is im-
portant to the design, fabrication and performance of these devices. In particular, the no-slip
boundary condition, which has long been experimentally verified in macroscopic flow, now receives
a great amount of attention in microfluidics. Moreover, many biological applications of microflu-
idic devices require surface coatings that change the wetting ability of aqueous or biological fluids.
Thus, a better understanding of fluid-surface interaction becomes necessary in designing applicable
devices.
Many experimental techniques have been applied to study the effect of slip on both hydrophilic
and hydrophobic surfaces. In general, slip measurement techniques can be classified into indirect
and direct methods. In the indirect methods, slip velocity is inferred from other measurements, for
example, the flow rate and pressure within the flow duct [20, 21, 25], or from forces required to move
the surfaces [64]. Examples of the direct measurement methods include hot film anemometry [32],
total internal reflection-fluorescence recovery after photobleaching (FRAP) [29] and micro-PIV [30].
There are also other groups of researchers who attempted to understand the slip phenomenon from
8
9
a theoretical perspective through molecular dynamics simulations [22, 23, 24]. The results from
these studies are quite diverse, and “apparent slip” lengths ranging from nanometers to microns
have been reported. Similarly, the dependence of the slip on the applied shear rate is also a matter
of debate with some results suggesting a constant slip length, while others reporting a slip length
dependent on shear rate and/or surface hydrophobicity.
Recently, a novel technique termed total internal reflection fluorescence microscopy (TIRFM)
has been applied to near-wall fluid measurements. The TIRFM technique has long been used in
biological studies, such as cell-substrate contacts, vesicle fusion and single-molecule observation
[65, 66, 67]. It takes advantage of the fact that total internal reflection occurs when light travels
from a dense medium (such as a glass cover slip) with refractive index n
1
into a less dense medium
(such as the sample fluid) with refractive index n
2
. Total internal reflection occurs if the incident
angle is larger than the critical angle θ
c
= sin
−1
(n
2
/n
1
), which can be derived from Snell’s Law.
When total internal reflection occurs, some fraction of the light energy penetrates into the less dense
medium and propagates parallel to the surface. This parallel light wave is known as the “evanescent
wave”. One characteristic of the evanescent wave is that its intensity, I, decays exponentially with
respect to the distance from the two-media interface. Thus, images taken under evanescent wave
illumination have low background noise, showing clear features of sample structure that is near the
interface. There have also been several recent experimental studies of fluid mechanics by applying
evanescent wave illumination [29, 56, 68]. In particular, Zettner and Yoda [56] used the PIV
technique combined with evanescent wave illumination to measure near-surface velocity field in
rotating Couette flow.
In this paper, we describe the Total Internal Reflection Velocimetry (TIRV) technique, in which
the TIRFM technique is applied to observe particle motion in the close vicinity of microchannel
walls with different surface wetting properties. A particle tracking velocimetry (PTV) technique is
applied to track particle displacements and flow velocities near the wall.
2.2 EXPERIMENTAL PROCEDURES
TIRFM requires the creation of total internal reflection of a laser beam at the glass-sample interface,
and the presence of fluorescent molecules or particles suspended in the sample fluid for imaging
and measurement. Two different optical configurations are commonly used to create an evanescent
field close to the glass-sample interface. The first method utilizes a prism to guide the laser beam
into the glass cover slip, which then serves as a waveguide, allowing total internal reflection to
occur. The other method, which is prismless, has the illuminating laser beam directed through a
high numerical aperture (NA) microscope objective at an angle that creates total internal reflection
at the glass-sample interface, as shown in Fig. 2.1. The maximum incident angle which can be
achieved by using an objective lens is given by
θ
max
= sin
−1
(NA/n
glass
) . (2.1)
As mentioned before, the intensity of an evanescent wave decays exponentially with distance z
away from the two-media interface
I(z) = I
0
e
−z/d
, (2.2)
where I
0
is the light intensity at the interface and d is known as the evanescent wave penetration
depth. The penetration depth, which characterizes the evanescent field, can be calculated as:
d =
λ
0

_
n
2
1
sin
2
θ −n
2
2
_

1
2
, (2.3)
10
~ Penetration depth
Fluorescent particle solution
Objective lens
Immersion oil
Glass (n = 1.51)
θ
Water (n = 1.33)
z
x
U
Figure 2.1: Schematic of objective-based TIRFM.
where λ
0
is the wavelength of the incident light and n
1
and n
2
are the indices of refraction for glass
and water, respectively. Based on these calculations, one can choose appropriate laser wavelength
and microscope objective or prism to achieve desired penetration depth.
In the present experiment, illustrated in Fig. 2.2, we have adopted the prismless TIRFM
method. A dual head Quantel Brilliant Q-Switched Nd:YAG laser was used to generate a pair
of 532 nm, 5 nsec laser pulses to illuminate the flow. The laser heads were timed to fire with a
pre-determined pulse separation (1.1 to 3.3 ms, depending on shear rate), illuminating two images
for use in a Particle Tracking Velocimetry (PTV) analysis. The beams were focused off-axis at the
back focal plane of a Nikon PL Apo NA 1.4 100X oil immersion objective lens mounted on a Nikon
Eclipse TE200 inverted epi-fluorescent microscope. Because the laser beam is focused off-axis, the
beam incident angle at the glass-fluid interface can be controlled by changing the radial location
of the laser beam. With water as the sample fluid, the experimental setup had a critical angle
θ
c
= 61.9

. At incident angles equal to or larger than θ
c
, the laser beam could be totally reflected
off the water-glass interface, creating an evanescent wave field. With a water-glass interface, λ
0
=
532 nm and NA = 1.4, it can be calculated that the penetration depth is approximately 150 nm
when the incident angle is θ = 63.4

.
The test cell consisted of a precision glass cover slip on which a polydimethylsiloxane (PDMS)
microchannel was attached. The PDMS channels were fabricated using standard “soft lithography”
techniques [69] and measured 31±1 µm deep, 250±1 µm wide and 25 mm long. They were bonded
to the test surface after exposure to an oxygen plasma which renders both the glass and PDMS
surfaces highly hydrophilic for a short time during which they bond on contact. Following bonding,
the channels were immersed in deionized-water to maintain their hydrophilic surface characteristics.
To create hydrophobic microchannels, it is necessary to coat the glass surface with a transparent
and hydrophobic substance because glass is naturally hydrophilic. We coated glass by immersing
glass into a 5-mM octadecyltrichorosilane (OTS) solution (dissolved in hexadecane) at room tem-
perature for several hours. A self-assembled monolayer (SAM) of OTS formed on the glass surface,
and was made permanent after the glass was rinsed in hexane and ethanol, dried in nitrogen, and
baked at 100

C on a hotplate for a few hours. The coating process was performed in a nitrogen gas
chamber to avoid bulk polymerization. The coated glass was bonded to a cleaned, patterned PDMS
channel to create a hydrophobic microchannel. It has been reported that surface roughness can
11
ICCD
Nd-YAG laser
532 nm
Syringe pump
100X, NA = 1.4
Objective lens
Dichroic mirror and
Barrier filter
PDMS channel
Convex lens
Pulse generator Computer
1360 x 1036 x 12 bit
532 nm, 5 nsec pulse
1.1 - 3.3 ms separation
Figure 2.2: Schematic of the experimental setup.
strongly inhibit boundary slip by transferring momentum if liquid molecules are trapped inside the
holes of the monolayer [29]. Thus, an atomic force microscope (AFM) measurement of glass surface
roughness was also performed. The polished Pyrex glass wafer was measured to have a surface
RMS roughness of 2 nm, while the same wafer had a surface roughness of 4 nm after OTS-coating.
Monodisperse Fluorescent Polymer Microspheres (Duke Scientific) with diameters of 200 nm
(±5%) and 300 nm (±3%), suspended in pure water at 0.02% and 0.04% volume fractions respec-
tively, were used as seed particles in this experiment. These microspheres have a peak absorption
wavelength of 542 nm and emit a red visible light at 612 nm. Particle solutions were flowed through
the test channels using a Harvard Apparatus 22 syringe pump, at flow rates ranging from 0.6 to
6.0 µL/min (±0.35%). An air gap between the piston and the fluid in the syringe and long tubes
were used to dampen pulsations from the syringe motion. Near-wall particle images were projected
onto a Q-Imaging Intensified Retiga CCD camera, which has a capacity of recording 1360-pixels-
by-1036-pixels 12-bit images. Under 100X magnification, each image pixel corresponds to 64.3 nm
in the flow channel and the diffraction-limited spot size is 30.9 µm [70]. For 200- and 300-nm
particles, the image diameters are 36.8 and 43.1 µm, respectively [55]. Typically 200 image pairs
were recorded for each flow condition. The recorded image pair was finally analyzed with a PTV
algorithm written in our group using the MATLAB programming environment. The choice of Par-
ticle Tracking Velocimetry (PTV), rather than Particle Image Velocimetry (PIV) is an important
one and this will be re-addressed later on.
Measurements were taken at several shear rates. For the results presented here, the shear
rate was estimated from the applied volume flow rate, Q, assuming steady laminar flow of an
incompressible fluid between two parallel plates without wall slip
˙ γ
wall
=
_
du
dy
_
wall
=
6Q
h
2
w
, (2.4)
12
Figure 2.3: Particle images of different illumination methods: widefield (direct, flood) illumination.
where h is the channel depth and w is the channel width. If there is significant slip at the wall, the
shear rate could be slightly lower than is estimated here [25]. However, for the current experiments,
any possible slip velocity is so small that this uncertainty can be neglected (this presumption will
be confirmed by the results in the following section).
2.3 RESULTS AND DISCUSSION
Figures 2.3 and 2.4 compare particle images of widefield (direct, flood) illumination and evanescent
field illumination. A large amount of background noise is observed in the case of direct illumina-
tion. Such background noise is attributed to the light emitting from the out-of-focus particles in
the bulk of the fluid and such images are common in micro-PIV measurements but lead to poor
cross-correlations during PIV analysis. Evanescent field illumination, on the other hand, eliminates
much of the background noise because the evanescent light penetration depth is restricted to ap-
proximately 150 nanometers, and particles in the bulk fluid are not illuminated. This characteristic
allows easy detection of only particles that are close to the channel surface.
To measure particle velocities near the channel surface, several hundred image pairs at various
low shear rates were taken for subsequent PTV image analysis. Bright particles with intensities
above a pre-determined threshold value were identified in each image. In this study, the intensity
threshold was set at 50% of the brightest particle intensity (usually about 1500-2000 counts from
a 12 bit CCD camera). This threshold value allowed us to detect almost every particle which can
be identified with the naked eye and restricts the analysis to particles that are well within the
evanescent field and in focus. The number of particles identified in one image is not very sensitive
to the exact threshold value. Typically about 10-20 particles were identified in each image, although
approximately 50% of these were not tracked between images due to out-of-plane diffusion which
causes particles to enter and leave the observation region.
It is important to point out that the determination of an intensity threshold sets an observation
range, as illustrated in Fig. 2.5. The lower bound of the observation range is the particle radius,
13
Figure 2.4: Particle images of different illumination methods: near-wall TIRFM image.
Glass Cover Slip
Fluid Velocity
Observation
Range
Figure 2.5: Schematic of near-wall particles moving near the surface illustrating the observation
range.
representing a particle in contact with the channel surface. As the particle moves farther from the
wall, and hence into a region of lower illumination intensity, the emitted intensity also falls. Thus,
the intensity threshold chosen sets an upper bound on the observation range which is different from
the penetration depth of the evanescent field. The exact size of the observation range was not
known in these experiments and this uncertainty turned out to be the major source of uncertainty
in these experiments and is discussed later on.
Although tracking the particle displacements in the x-y plane (parallel to the surface) was
straightforward, tracking the particles motion in the z -axis (normal to the surface) is not easy.
Ovryn [71] developed a method to determine the three-dimensional position of a spherical particle,
based on resolving the spatial variations in the scattering pattern. However in the evanescent field,
illumination intensity varies exponentially with distance from the wall (as shown in Eq. 2.2), and
consequently it is not easy to directly quantitatively associate the fluorescent scattering of a particle
in a non-uniform illumination field with the particle’s distance from the wall. For this reason we
14
100 50 0 50 100
40
20
0
20
40
60
80
100
120
140
Cross-stream velocity (μm/sec)
S
t
r
e
a
m
w
i
s
e

v
e
o
c
i
t
y

(
μ
m
/
s
e
c
)

Figure 2.6: Distribution of particle velocity vectors of 200-nm particles with ˙ γ = 469 sec
−1
.
did not attempt to track the particle’s motion normal to the surface. There is no reason why this
should not be possible in the future.
The intensities of pixels surrounding the peak intensity were fitted to a two-dimensional Gaus-
sian curve [72, 73, 74]. Because the particles are known to be spherical with diameter standard
deviation less than 5%, a two-dimensional Gaussian curve fit closely approximates the Lommel func-
tion, and allowed us to locate the particle center coordinates with sub-pixel resolution. Particles
were matched between images using a custom particle tracking algorithm and displacement vectors
were calculated from the displacements of particle centers. A threshold displacement was imposed
to detect error vectors, which were defined as displacements larger than deemed reasonable. These
large displacement vectors usually result from matching two different particles mistakenly. At each
testing condition, the total number of validated displacement vectors ranged from 700 to 1500. Us-
ing the geometric scale and the applied pulse separation, a velocity vector can be calculated from
each displacement vector. Figure 2.6 shows an example of a collection of velocity vectors obtained
for a single shear rate. The Brownian motion is particularly strong due to the particle’s small size.
2.3.1 Cross-stream Velocity Analysis
Analysis of the displacements of particles in the cross-stream direction was used as to check the
consistency of the experiment results. In this direction, in the absence of any mean flow, one would
expect that the particle motion would be purely driven by thermal fluctuations. Indeed, we found
that the distributions of spanwise velocities had a Gaussian distribution with mean and skewness
close to zero. The width of the particle displacement distribution, σ, was used to compute the local
diffusion coefficient, D:
D =
σ
2
∆t
2
, (2.5)
15
0 500 1000 1500 2000 2500
0
0.5
1
1.5
2
2.5
x 10
13
Shear rate (1/sec)
D
i
f
f
.

C
o
e
f
f
.

x

R
a
d
i
u
s

(
c
m
3
/
s
e
c
)
Figure 2.7: Cross-stream Brownian motion of particles in an infinite medium and near wall (ex-
periment). Solid symbol represents 200-nm particles and open symbol represents 300-nm particles
(, hydrophilic surface; ♦, hydrophobic surface). The dashed line denotes particle diffusion in an
infinite medium calculated on the basis of the Stokes-Einstein relation.
and Fig. 2.7 shows the cross-stream diffusion coefficient obtained for 200-nm (solid symbol) and
300-nm (open symbol) particles at a variety of different streamwise shear rates over both hydrophilic
and hydrophobic surfaces. The diffusion coefficient is scaled by the particle radius, r, and the dotted
lines represent scaled particle diffusion in an infinite medium calculated based on the Stokes-Einstein
relation:
D
0
r =
k
B
T
6πη
, (2.6)
where D
0
is the Stokes-Einstein diffusion coefficient, k
B
is the Boltzmann constant, T is the medium
temperature, and η is the fluid viscosity. As expected, the measured spanwise diffusion coefficients
are independent of the streamwise shear rate and the diffusion coefficient, when appropriately
scaled, is independent of particle radius. However, as indicated by the dotted line in Fig. 2.7,
the measured values for D are approximately 30% that of the bulk diffusion expected. This is
confirmation of the presence of hindered diffusion of particles near a solid surface [37, 38]. If we use
the observed value of D with the theory for hindered diffusion [37, 38] to calculate the ”apparent
particle distance” from the wall, defined as the distance between the particle center and the wall, we
find it to be 110 nm for the 200-nm particles and 165 nm for the 300-nm particles. This is consistent
with the high intensity threshold chosen which defines a small observation range. Unfortunately,
since we did not carefully monitor the observation range, a more precise estimate is not available
for this data.
2.3.2 Streamwise Velocity Analysis
Figure 2.8 shows an example of the distribution of the streamwise velocities, found by projecting the
velocity distributions onto the mean flow direction. Several features are worth noting. Firstly, the
location of the peak of the distribution increases with increasing shear rate, as one would expect.
16
200 0 200 400 600 800
0
0.01
0.02
Velocity ( μm/sec)
P
D
F
234 (1/sec)
1172 (1/sec)
2344 (1/sec)
Figure 2.8: Experimentally-measured distribution of streamwise velocities of 200-nm diameter par-
ticles over a range of shear rates.
Less expected, and in contrast to the distributions for the spanwise velocities, is the observation
that the distribution widens with increasing shear rate, and that the skewness clearly shifts from
0.15 at the low shear rate (barely noticeable) to a clearly observable positive value of 0.85 (a
stronger positive tail) at the highest shear rate presented. In this case the most-probable velocity
is quite different from the average velocity, raising a fundamental question about the definition of
the actual fluid velocity in terms of the observed particle velocity distribution. Figure 2.9 shows
the variation of the mean velocities as a function of shear rate, the error bars depict the standard
deviation about the mean. The velocities fall on a straight line through the origin, and there are no
discernible differences between the hydrophilic and hydrophobic surfaces and, more interestingly,
between the 200- and 300-nm diameter particles.
The reason for the observed increase in the distribution width and skewness can be explained
in terms of the observation range from which particle motions are recorded. At each shear plane,
the true velocity distribution for tracer particles has a Gaussian shape with a mean equal to the
flow velocity at that plane, and standard deviation determined by the temperature and particle
size. However, any micro-PTV or PIV system, including the current TIRV integrates through the
observation range and thus samples particle motion from a range of shear planes over which the
mean velocity changes. Thus, for a fixed observation range, the range of mean velocities sampled
increases as the shear rate increases, leading to the observed broadening of the velocity distribution
for the streamwise velocity. A similar argument can be used to explain the increasing skewness
at higher shear rates. Here, we note that, due to the presence of the wall, a particle is more
likely to move upwards (and thus experience a higher local velocity) than to move downwards (and
experience a lower local velocity). Thus, higher velocity fluctuations are more probable than lower
ones in the vicinity near the wall (a similar argument is used to explain positive skewness in the
streamwise velocity distribution in a turbulent wall-bounded flow).
The lack of any differences between the mean velocities obtained for 200-nm and 300-nm par-
ticles is more troubling. One would expect that the 300-nm particles would sample shear planes
at higher local velocities than those sampled by 200-nm particles, and thus the average velocities
17
0 500 1000 1500 2000 2500
0
100
200
300
400
Shear rate (1/sec)
A
p
p
a
r
e
n
t

v
e
l
o
c
i
t
y
(
μ
m
/
s
e
c
)
Figure 2.9: Apparent particle velocity vs. shear rate (♦, 300-nm particle, hydrophilic surface; ,
300-nm particle, hydrophilic surface; , 200-nm particle, hydrophilic surface; ◦, 200-nm particle,
hydrophilic surface).
should be larger for the larger tracer particles. The fact that this is not observed remains a some-
what puzzling result and one that needs to be resolved in future measurements. However, there are
several factors that are not yet well understood that could contribute to this counter-intuitive ob-
servation, including uncertainty in the measurement of the illumination observation range, particle
motions in addition to pure translation (such as rolling motion), and other forces on the particles
(e.g. colloidal).
The drift velocities measured can result from two effects: (i) the finite thickness of the observa-
tion range and (ii) the possible existence of a slip velocity at the surface. We can assess the relative
contributions of these effects using a very simple theoretical model. If we assume a constant shear
very near the surface, we can write the local velocity profile as:
u(y) = u
s
+ ˙ γy, (2.7)
where u
s
is the slip velocity. If we use a Navier representation for the slip velocity, u
s
= l

˙ γ, (where
l

is the slip length) and compute the average velocity in the mean depth of the observation range,
δ, we find that the average, or ”apparent slip” velocity is
¯ u = (l

+δ) ˙ γ, (2.8)
The slope of the data in Fig. 2.9 is approximately 100-nm, representing this combination of
both the penetration depth and the slip length. Given that the observation range is estimated to be
approximately 100 - 150 nm, this suggests that the slip length, if present, is likely no more than 10
nm, consistent with the results of Choi et al. [25] for these low shear rates. At the present, our lack
of quantitative knowledge about the evanescent wave penetration depth (which probably varied
from experiment to experiment) and the observation range makes a more precise estimate for the
slip length impossible at this time. However, with the appropriate characterization and consistent
control of the observation range, this measurement will be possible and such experiments are
currently underway. Finally, we note that a slip length of order 1000 nm (as proposed by Tretheway
18
and Meinhart [30], and Zhu and Granick [64]) is completely unsupported by these data, as it would
predict apparent slip velocities at least an order of magnitude larger than observed, and would be
independent of any reasonable estimate for the observation range.
2.3.3 Comparisons with Numerical Simulations
The experimental results were compared with a numerical simulation of near-wall Brownian motion
in shear flow. The motivation for this was to generate an intuitive model with which to help
understand the experimental results. This simple two-dimensional numerical model was developed
to simulate the Brownian motion of particles with radius, r, in a no-slip steady-state Couette flow
between parallel infinite plates separated by a distance h, the experimental channel depth. At
the beginning of the simulation, 10,000 particles were uniformly distributed between r and δ, the
assumed upper bound of the observation range. At each time step, each particle was moved by
a linear combination of the mean shear flow (evaluated at the particle’s geometric center) and
Brownian motion (sampled from a normal distribution with a zero mean and a standard deviation
of

2D∆t). Other effects such as particle rotation, etc were not considered at this time.
The Brownian motion was modified to account for the hindered diffusion both parallel and
perpendicular to the wall [36, 37, 38]. For the Brownian motion normal to the wall [36], the
modified diffusion constant, D

, is such that
D

D
0
=
_
4
3
sinh α

n=1
n(n + 1)
(2n −1) (2n + 3)
_
2 sinh (2n + 1) α + (2n + 1) sinh 2α
4 sinh
2
_
n +
1
2
_
α −(2n + 1)
2
sinh
2
α
−1
__
−1
,
(2.9)
where D
0
is the Stokes-Einstein diffusion coefficient, α = cosh
−1
(z/r) and z is the distance between
the geometrical center of a particle and the wall. For the Brownian motion parallel to the wall, the
modified diffusion constant, D

, is such that [37]
D

D
0
=1 −
9
16
_
h
r
_
−1
+
1
8
_
h
r
_
−3

45
256
_
h
r
_
−4

1
16
_
h
r
_
−5
+O
_
h
r
_
−6
,
(2.10)
by the method of reflection [37, 38] for (z −r)/r > 1, and
D

D
0
= −
2
_
ln
_
h−r
r
_
−0.9543
¸
_
ln
_
h−r
r

2
−4.325 ln
_
h−r
r
_
+ 1.591
, (2.11)
which is the asymptotic solution [37, 38] for (z −r)/r < 1.
After 1000 time steps, particles that remained within the observation range were counted and
their displacements from their initial positions were calculated to obtain a particle displacement
distribution, similar to that obtained in the physical experiments. To emulate the experiment,
plate separation distance, particle radii, flow shear rate and the total time (1000 time steps) were
chosen to be equal to measurement conditions, and the assumed observation range was chosen to
be 100 - 175 nm from the wall.
19
200 0 200 400 600 800
0
0.005
0.01
0.015
0.02
Velocity(μm/sec)
P
D
F
234 (1/sec)
1172 (1/sec)
2344 (1/sec)
Figure 2.10: Streamwise velocity distribution of 200-nm particles determined from Monte Carlo
simulation at observation depth = 175 nm.
Problems can arise when the simulated particle is predicted to pass through the solid surface
during the finite time step. In the present simulation two different wall boundary conditions were
tested to deal with this situation. The first condition applied a specular reflection, in which particle
was simply reflected off the wall and back into the flow. The second condition emulated near-wall
lubrication condition and simply allowed the particle to remain at the wall until Brownian motion
perpendicular to the wall transported it back into the bulk flow, one or more time steps later. Only
a few particles needed to be treated with these heuristic boundary conditions, and it was found
that the two models gave almost indistinguishable results, with less than 1% statistical difference
in the mean and standard deviation of the resultant particle distributions. Figure 2.10 shows
the streamwise velocity distributions of 200-nm particles at various shear rates, while Fig. 2.11
shows the variation of the mean velocities and their standard deviations, denoted as error bars, as
a function of shear rate. The rather simplistic nature of the simulation is such that one should
not make direct, quantitative, comparisons with the experimental results (Fig. 2.8 and 2.9). In
addition, the choice of observation range for the computation has not yet been optimized to match
the simulation results with those of the experiments. Nevertheless, the similarities between the two
figures are striking, showing similar increases in the mean, standard deviation and (less obvious,
but present) skewness of the distributions as the shear rate increases.
Unlike the experimental results in Fig. 2.9 which did not show any difference in measuring
velocity using different particle size, there is a clear difference in the estimated velocity using
different particle size in Fig. 2.11. As discussed earlier this difference due to particle size is to
be expected for a fixed observation range, and this result adds weight to our suspicion that the
observation range in the experiments were not consistent, and that future experiments should pay
more attention to this important issue.
20
0 500 1000 1500 2000 2500
0
100
200
300
400
500
Shear rate (sec
1
)
A
p
p
a
r
e
n
t

v
e
l
o
c
i
t
y

(
μ
m
/
s
e
c
)
Figure 2.11: Streamwise apparent velocity predicted by simulation (, 200-nm particles; ◦, 300-nm
particles) at the observation depth of 175 nm.
2.4 CONCLUDING REMARKS
We have demonstrated that Total Internal Reflection Velocimetry (TIRV) can be a very powerful
technique for the study of near-wall flows in microfluidic systems. TIRV images have very low
background noise that allows accurate PIV and PTV measurements of flow velocities within a few
hundred nanometers of the surface. By using sub-micron seed particles, we have demonstrated
the ability to measure both velocity distributions and mean velocities in the plane parallel to the
surface.
The determination of velocities in the streamwise and spanwise directions is extremely accurate,
and comparable to ”standard” microPIV techniques, and is limited by the same constraints - the
accuracy of particle centroid location, diffraction-limited imaging of sub-micron particles, etc [70].
However, the technique also promises the ability to use intensity information to obtain wall-normal
displacements as well as precise estimates for the slip velocity at the surface. The present imple-
mentation did not (unfortunately) control with sufficient accuracy or consistency the characteristics
of the evanescent wave penetration depth which is very dependent on the incident beam angle. For
this reason, we were not able to measure vertical particle motion and our current resolution of the
slip length measurement is limited such that we can only state that a slip length, if present, is of
order 10 nm or less. However, the estimate for the slip length is limited only by the accuracy with
which we can measure the observation range and so we expect to be able to make measurements
with uncertainties less than 5 nm in the near future.
The accuracy of the TIRV system can still be improved by several means that are subject to
future investigation. First, by using a shorter illumination wavelength, the penetration depth of
evanescent field can be decreased. Secondly, smaller seed particles can move closer to the channel
surface than larger particles, although they also suffer from higher Brownian diffusion and so the
particle drop-out will increase. Both changes will allow direct measurements of flow velocities that
are closer to the channel surface, where slip might contribute more significantly to the local flow
velocity.
21
Finally, and most importantly, the technique does raise fundamental questions regarding the
use of ”passive” tracer particles to estimate the velocity field in microfluidic systems. Our results
reveal several features of particle motion in the near vicinity of the surface, including non-Gaussian
Brownian motion (due to the finite thickness of the observation range) and hindered diffusion due
to near-wall lubrication effects. There are also effects other than the translational hydrodynamic
drag that can contribute to the particle motion such that the velocity measurements might be
compromised. One is the possibility of particle rotation; a second is the possibility of particle-
wall colloidal interactions and the relative roles of gravity, thermal noise and electrostatic effects
(both the bead and the surface have surface charges). This is an area that has been studied by
a number of researchers, including Flicker et al. [51], Crocker and Grier [75] and Behrens and
Grier [76] who have identified colloidal forces that are highly dependant on bead diameter, surface
charge, ionic strength (and for pure water such as used in these experiments, the rate at which
ions leak from the glass into solution). A full treatment of these issues is beyond the scope of the
current paper, and indeed, the experimental techniques described here to can be used in further
studies of such phenomena. Perhaps the most important consequence of this is that techniques that
treat ensembles of particles (such as PIV which looks at the collective motion of particles in an
interrogation region, and often averages cross-correlations before peak detection) might be suspect
close to walls due to particle interference effects which cannot be neglected and where the particle
motion is not represented by Gaussian statistics (and hence the average particle velocity is different
from the most probably particle velocity). For this reason, tracking each particle independently
and then computing statistics from the individual particle motions might yield a more accurate
view of the near-wall velocity field.
Chapter 3
Paper 2: Direct measurement of slip
velocities using three-dimensional
total internal reflection velocimetry
The existence and magnitude of slip velocities between deionized water and a smooth glass surface
is studied experimentally. Sub-micron fluorescent particles are suspended in water and imaged
using Total Internal Reflection Velocimetry (TIRV). For water flowing over a hydrophilic surface,
the measurements are in agreement with previous experiments and indicate that slip, if present, is
minimal at low shear rates, but increases slightly as the shear rate increases. Surface hydrophobicity
is observed to induce a small slip velocity, with the slip length reaching a maximum of 96 nm at a
shear rate of 1800 s
−1
. Issues associated with the experimental technique and the interpretation of
results are also discussed.
3.1 INTRODUCTION
The century-old assumption of the no-slip boundary condition between a liquid and a solid has been
challenged by recent experimental results and molecular dynamic simulations and has been the
subject of many recent investigations exploring flows over both wetting and non-wetting surfaces.
Although there is considerable disagreement regarding the existence of slip over a hydrophilic
surface, it is generally believed that surface hydrophobicity aids slip, while the exact mechanism
is not yet understood (for a review of this topic, see [7]). Experimental studies have reported
a wide range of slip lengths, ranging from micrometers [20, 30, 32, 64], hundreds of nanometers
[29, 33], to tens of nanometers or smaller (including no-slip) [25, 26, 27, 28, 31]. Molecular dynamics
simulations, on the other hand, suggest small slip lengths, mostly less than 100 nm [22, 23, 24, 77,
78, 79].
The experiments have been conducted under various direct and indirect measurement techniques
with varying accuracy and uncertainty. In indirect measurements, slip velocity is inferred from
other flow quantities, such as the relationship between flow rate and pressure drop [25] or the
measurement of the forces required to move a pair of crossed cylindrical surfaces separated by a
thin film of the test fluid [26, 27, 28, 64]. In contrast, alternative techniques such as micro-particle
image velocimetry (PIV) have resulted in direct measurements of near-surface fluid velocity using
tracer particles [30, 31].
Although the direct measurements have a strong appeal, there are two major drawbacks inher-
22
23
ent to conventional micro-PIV. The first is that the velocity is determined from the motion of a
collection of particles in an interrogation area (IA) and, in order to obtain a reasonable estimate of
the velocity, the IA must be measuring several hundred nanometers in size [30]. Hence the determi-
nation of phenomena which might measure a few nanometers is very difficult to resolve. A second
problem is that the size of measurement volume along the optical axis for flood-illuminated micro-
PIV techniques is defined by the focal depth of the optical system, which is typically of the order
of 0.5 to 1 micron [30, 31]. Although Joseph & Tabeling [31] reported a slip length measurement
with 100-nm resolution, micro-PIV would still be insufficient to provide an affirmative conclusion
if the true slip length is below 100 nm.
One attractive approach that solves both of these problems is the use of Total Internal Reflection
Velocimetry (TIRV) [80] which uses total internal reflection of an incident laser pulse to generate a
highly localized illumination of the surface flow, and relies on tracking the motion of single tracer
particles rather than the collective particle motion that is used in PIV. Total internal reflection
fluorescence microscopy (TIRFM) has long been used by biologists to study near-surface dynamics
[65, 66, 67]. Zettner & Yoda [56] combined micro-PIV and TIRFM to study near-surface Couette
flow while Pit et al. [29] measured slip velocity of hexadecane over a sapphire substrate using
TIRFM and fluorescence recovery after photobleaching. We first reported on the TIRV technique
in the context of slip velocities for near-wall microflows [80], but were unable to make a conclusive
determination of the size or character of any slip lengths due to difficulties in the determination
the location of tracer particles within the narrow evanescent illumination region.
The present manuscript reports on experiments in which the TIRV technique has been sub-
stantially improved to the point that quantitative statements can be made regarding the nature of
slip velocities over hydrophilic and hydrophobic surfaces. By measuring the intensity of tracer par-
ticles’ fluorescence, and in conjunction with a statistical model for the optical and hydrodynamic
behaviour of small particles near a surface, we are able to track their motion within a narrow region
above the solid surface and to compare that motion to the motion predicted for different levels of
boundary slip.
3.2 THEORETICAL CONSIDERATIONS
3.2.1 Total Internal Reflection Microscopy
The evanescent field and its application to fluorescence microscopy has been thoroughly documented
in literature [e.g. 40, 44, 47, 56, 63, 80]. In summary, an evanescent field can be created near a
solid-liquid interface where total internal reflection occurs, as illustrated in figure 4.1. The field
intensity, I, decays exponentially with distance, z, away from the two-medium interface by
I(z) = I
0
e
−z/p
, (3.1)
where I
0
is the intensity at the interface and p is known as the evanescent wave penetration depth.
The penetration depth, which characterizes the length scale of the evanescent field, can be calculated
from
p =
λ
0

_
n
2
1
sin
2
θ −n
2
2
_

1
2
, (3.2)
where λ
0
is the wavelength of the incident light, θ is the beam incident angle and n
1
and n
2
are
the indices of refraction of the solid and liquid, respectively.
24
Figure 3.1: Objective-based total internal reflection fluorescence microscopy.
3.2.2 Emission Intensity of Fluorescent Particles
For spherical tracer particles with a uniform volumetric fluorophore distribution in an evanescent
field, Kihm et al. [44] calculated the particles’ emission intensities as an exponential function
of their distances to a substrate surface. Through experience we have found, however, that the
intensity of fluorescence emission from a large number of nominally identical particles can vary due
to a variety of factors, including statistical variations in illumination intensity, quantum efficiency of
the imaging system and, most importantly, a distribution in the physical size of the particles. If we
parameterize all of these variations by an “effective emission radius”, r, we can model the particle’s
probability density function (PDF) of its effective emission radius with a Gaussian distribution
P(r) = P
0
exp
_

_
r
a
−1
_
2
σ
2
r
_
, (3.3)
where a is the mean effective emission radius of the batch, σ
r
is the characteristic variation (standard
deviation) and P
0
is a normalization constant. Based on the calculation of Kihm et al. [44] and
the assumption that emission intensity of a fluorescent particle is proportional to the number of
fluorophores within its volume, the emission intensity, I
e
, of a particle in an evanescent field is
I
e
(r, h) =
_
r
a
_
3
_
I
e
0
e
−(h−a)/d
_
, (3.4)
where h is the distance from the particle’s centre to the substrate surface at which total internal
reflection occurs, I
e
0
is the emitted intensity of a single particle with r = a located at h = a (ie.
touching the substrate surface), and d is an intensity decay length that need not be the same
as the evanescent penetration depth. The intensity decay length can be obtained experimentally
by statistically measuring intensities of particles as a function of particles’ distances away from
the substrate. Combining (3.3) and (3.4), one gets the joint PDF of the emitted intensity and a
particle’s distance to the substrate as
P(I
e
, h) = αexp
_
¸
¸
¸
_
¸
¸
¸
_

_
3
_
_
I
e
I
e
0
_
e
(h−a)/d
−1
_
2
σ
2
r
_
¸
¸
¸
_
¸
¸
¸
_
. (3.5)
25
0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
Normalized Intensity
P
D
F
σ
r
= 0.05
σ
r
= 0.1
σ
r
= 0.15
σ
r
= 0.2
Figure 3.2: The emission intensity distribution of particles in a uniform concentration field. The
normalized intensity in the horizontal axis is defined as I
e
/I
e
0
, and the calculation is performed
with a normalized observation range of 0 < h/a < 3 and a normalized intensity decay length of
d/a = 1.8.
When an ensemble of tracer particles are imaged, they would certainly be located at a range of
distances from the substrate surface. The lower limit of this imaging range is h = a, representing a
particle touching the substrate surface. The upper limit of the imaging range is determined by the
sensitivity of the recording medium, such as an intensified CCD camera used in this study. For a
given imaging range of h
1
< h < h
2
, the PDF of a tracer particle’s intensity is given by
P(I
e
, h
1
< h < h
2
) =
_
h
2
h
1
P(I
e
, h)c(h) dh, (3.6)
where c(h) is the concentration profile of particles in the fluid. A sample plot of this distribution
(3.6) is shown in figure 3.2 for a range of values of σ
r
. For a low variation in effective emission
radius (σ
r
small), we see that the particles in the defined imaging range are sharply defined in
intensity, with the brightest particles located at the surface (I
e
/I
e
0
= 1) and the faintest particles
at the outer edge of the evanescent field (I
e
/I
e
0
= 0.2). However, as the variation in particle radius
increases, the transitions become more blurred, and some larger particles close the wall contribute
to the long tail at high observed intensities, while small particles close to the edge of the imaging
range are represented by the smooth transition close to I
e
/I
e
0
= 0.
When performing image analysis, however, one need not analyse all imaged particles. Instead,
using particle intensity as a guide, one can choose to identify and analyse only the particles within
a certain desirable intensity range, ignoring particles that appear too bright or too dim. Equation
(3.5) could also describe the particle position distribution within a normalized intensity range of
α < I
e
/I
e
0
< β by
P (h, α < I
e
/I
e
0
< β) = c(h)
_
β
α
P(I
e
, h) d
_
I
e
I
e
0
_
. (3.7)
26
100 150 200 250 300 350
0
2
4
6
8
x 10
−3
Particle Center Distance to Wall (nm)
P
D
F
Figure 3.3: The predicted position distribution of particles with 0.5 < I
e
/I
e
0
< 1 in a uniform
concentration field. In this prediction, a = 100 nm, d = 181 nm and σ
r
= 0.18.
An example of this position distribution (3.7) is shown in figure 3.3 for a normalized intensity
range of 0.5 < I
e
/I
e
0
< 1. The most obvious feature of figure 3.3 is that the distribution is non-
uniform with the majority of particles contributing to this range of intensities coming from close
to the substrate surface (< 250 nm). In addition, one can observe that the distribution has a long
positive tail that gradually decays to zero at far away into the fluid bulk. Because of the large
particle effective radius variation in this example, a large particle far away (for example, > 300
nm) can exhibit in the same the same fluorescence intensity as that of smaller particles close to
the substrate surface. These observations reinforce the fact that a particle’s intensity cannot be
monotonically related to its distance to the solid surface, and attention should be given to any
intensity-based quantitative analysis by recognizing the non-uniform position distribution and the
contribution of particles far away from wall.
3.2.3 Near-Surface Shear Flow
Although we are interested in determining the velocity of a pure fluid near a solid surface, we
actually measure the velocity of small spheres near the surface, and it is critical to recognize the
differences between these two quantities. It is well known that shear and near-surface hydrodynamic
effects can cause a tracer particle to rotate and translate at a velocity lower than the local velocity
of the fluid in the same shear plane [12, 38]. Goldman et al. [38] proposed that a particle’s
translational velocity, U, in a shear flow with local shear rate, S, is given by
U
hS
· 1 −
5
16
_
h
r
_
−3
, (3.8)
valid for large h/r, and
U
hS
·
0.7431
0.6376 −0.2 ln
_
h
r
−1
_, (3.9)
27
1 1.5 2 2.5 3 3.5 4
0.4
0.5
0.6
0.7
0.8
0.9
1
h
2
/r
¯
U
¯
V
h
1
/r = 1
h
1
/r = 1.5
Figure 3.4: The ratio of statistical apparent velocity of particles and mean fluid velocity under no
slip in an imaging range of h
1
< h < h
2
, which is non-dimensionalized by particle radius.
¯
V is
computed with linear fluid velocity profile, which is a good approximation near a surface. In the
computation of
¯
U, c(h) is assumed to be uniform.
valid for small h/r. Although there exists no analytical approximation for intermediate values of
h/r, Pierres et al. [60] proposed a cubic approximation to numerical values presented by Goldman
et al. [38]:
U
hS
·
_
r
h
_
exp
_
0.68902 + 0.54756
_
ln
_
h
r
−1
__
+ 0.072332
_
ln
_
h
r
−1
__
2
+0.0037644
_
ln
_
h
r
−1
__
3
_
.
(3.10)
The apparent velocity,
¯
U, of a large ensemble particles chosen from a normalized intensity range of
α < I
e
/I
e
0
< β and located in an imaging range of h
1
< h < h
2
is given by the average of the local
velocity integrated over the imaging range:
¯
U =
1
h
2
−h
1
_
h
2
h
1
U (h, a, S (h)) P(h, α < I
e
/I
e
0
< β) dh. (3.11)
This average apparent velocity,
¯
U, can be compared to the mean velocity, of the pure fluid,
¯
V ,
calculated for the same imaging range. The ratio of the two is shown in figure 3.4, from which
we see that the average apparent velocity of particles is, not surprisingly, significantly lower than
that of pure fluid velocity when the imaging range is very close to the substrate surface. This
difference must be incorporated into the data analysis, and failure to recognize this effect can lead
to misinterpretation of the measured data and underestimation of slip lengths.
It should be noted that (3.11) is based on a no-slip condition at the fluid/solid interface. If
there exist a slip velocity, U
slip
, at the solid boundary, the apparent velocity of the same ensemble
28
of particles would be
¯
U
app
= U
slip
+
¯
U = δS
w
+
¯
U, (3.12)
where the slip velocity is characterized as the product of the slip length, δ, and the wall shear rate,
S
w
.
3.3 EXPERIMENTAL PROCEDURES
3.3.1 Materials and Setup
A schematic of the image acquisition system is shown in figure 3.5. Fluorescent particles were
illuminated with an objective-based TIRFM method, as detailed in [80]. A pair of 532-nm, 5-
nsec laser pulses (Quantel Twins Q-Switched Double Pulse Nd:YAG Laser) were directed through
a Nikon PL Apo NA 1.45 100X TIRF oil immersion objective at an angle that created total
internal reflection at a glass-water interface, thus illuminating the near-surface region of water with
an evanescent field (figure 4.1). Fluorescent images of near-surface particles were projected onto
a Q-Imaging Intensified Retiga CCD camera (ICCD), capable of recording 1360-pixels-by-1036-
pixels 12-bit images. Under 100X magnification, each image pixel corresponds to 64.3 nm in the
flow channel and the diffraction-limited spot size is 19.7 µm (calculated based on [70]). A TTL
pulse generator (Berkeley Nucleonics Model 555) was used to synchronize laser pulses and image
acquisitions. The energy of each illuminating laser pulse was recorded simultaneously with each
image acquisition and this was used to adjust the resultant images and thus account for pulse-to-
pulse variations in illumination intensity.
Test channels were fabricated using a polydimethylsiloxane (PDMS) (Dow Corning Sylgard 184)
molding technique [69] and bonded onto polished glass wafers. The dimensions of the test channels
were 50 ± 1 µm deep, 250 ± 1 µm wide and 15 mm long. To create hydrophilic microchannels,
the pre-bonding PDMS surface was exposed to an oxygen plasma and immediately immersed in
deionized-water after bonding to a glass wafer. Hydrophobic channels, on the other hand, were
fabricated by bonding untreated PDMS channels onto octadecyltrichorosilane(OTS)-coated glass
wafers. Detailed coating procedures can be found in [80]. It has been reported that water has a
contact angle of 120 degrees with a similarly prepared OTS-coating [30, 81]. The roughness of the
test surface was measured using an atomic force microscope prior to bonding. The hydrophilic and
hydrophobic surfaces had an RMS surface roughness of 0.47 nm and 0.35 nm respectively.
Monodisperse Fluorescent Polystyrene Microspheres (Duke Scientific) with diameters of 200 nm
(± 5%) suspended in purified water (Fluka) at 0.04% volume fraction were used as tracer particles
in this experiment. These microspheres have a peak absorption wavelength of 542 nm and emit a
red light at 612 nm. There are three reasons that 200-nm diameter particles were selected as probes
in this experiment. First, smaller size fluorescent particles have much larger size variation, making
relation of particle center distance from the solid wall and their fluorescent intensities less reliable.
Secondly, smaller particles do not contain enough fluorophores to provide significant signals with
a 5-nsec laser illumination. Lastly, 200-nm particles do not diffuse too far during the inter-frame
time, with diffusion out of imaging range being our major concern.
A Harvard Apparatus 22 syringe pump was used to maintain particle solution flow rates, ranging
from 1 to 10 µL/min (±0.35%). To maintain steady pumping motions of the Harvard Apparatus
stepper motor, we purposely introduced a large volume of air between the piston and the fluid in
the syringe. This volume of air acted as a pressure shock absorber to damp out the oscillatory
forcing of the syringe pump screw. In addition, a long tubing (> 1 m long) was used to connect
29
ICCD
Nd-YAG laser
532 nm
Syringe pump
100X, NA = 1.45
Objective lens
Dichroic mirror and
Barrier filter
PDMS channel
Convex lens
Pulse generator
Computer
1360 x 1036
x 12 bit
532 nm, 5 nsec pulse
2 - 5 ms separation
Power
Attenuator
Half-wave
plate NI_DAQ
Mirror
Convex Lens
10 micron pinhole
Convex Lens
Polarizing
Beamsplitter
Pulse Energy
Meter
Figure 3.5: Image acquisition system.
the syringe and the channel inlet. Particle flow motion was observed to be very steady under this
setup.
1200 images were captured at each flow rate, as well as in quiescent fluid. Once all necessary
images of particle motion were captured, the test channel was filled with rhodamine-B solution.
Images of the rhodamine-B solution were subsequently captured under the same evanescent illumi-
nation, and averaged to obtain an image characterizing the spatial distribution of the illuminating
laser beam.
It has been suggested that nanobubbles and a gas bubble layer on hydrophobic surfaces could
be alternative sources of apparent slip [7, 34, 82]. Prior to each experiment, the fluid was degassed
by placing the solution in vacuum for at least 30 minutes. This procedure has been reported to
significantly eliminate nanobubble formation [83].
It is intuitive to assume that one can infer the location of a particle in the evanescent illumination
field based on the fluorescence intensity. Kihm et al. [44] used a ratiometric fluorescence intensity
to track particle motions three-dimensionally. Although this is theoretically feasible, in practice,
successful use of this technique requires precise knowledge of the illumination beam incident angle
and a solution of Maxwell’s equation for an evanescent field in a three media system (glass, water
and polystyrene), which can be difficult to express explicitly. Thus, an experimental method was
devised to obtain a ratiometric relation between particle emission intensity and its distance to the
glass surface. We attached individual fluorescent particles to fine tips of glass micropipettes, which
were translated perpendicularly to the glass substrate in the evanescent field with a 0.4-nm precision
stage (MadCity Nano-OP25). At each distance to the surface, 30 images of the attached particles
were captured. The pixel values of the imaged particles were then averaged, and subsequently fitted
to a two-dimensional Gaussian function to find their centre intensities. The process was repeated
30
0 50 100 150 200
0
0.2
0.4
0.6
0.8
1
h−r, (nm)
N
o
r
m
a
l
i
z
e
d

C
e
n
t
e
r

I
n
t
e
n
s
i
t
y
Figure 3.6: Fluorescent particle intensity as a function of its distance to the glass surface. The
particles used here are 200 nm in diameter. The solid line is a least-square exponential fit to the
data whose decay length, λ, is 181 nm.
several times with different particles. The particle intensities are shown in figure 3.6, along with a
exponential fit to the measured data:
I
peak
= Ae
−λ(h−r)
+B, (3.13)
where A, B and λ are constants determined by a least-squares fit. The fitted curve predicts an
intensity decay depth of 181 nm, very close to the evanescent penetration depth that we would
predict based on the imaging lens, wavelength and refractive indices of the system.
3.3.2 Velocimetry Analysis
The analysis of particle image pairs was performed using a custom particle-tracking program written
in a MATLAB environment. Prior to velocimetry analysis, images of particles stuck on the glass
substrate had to be removed from each image pair to prevent undesirable detection by the Particle
Tracking Velocimetry (PTV) algorithm. Although the presence of these stuck particles could distort
the velocity field, they generally had a very low density and were usually placed at the edge of an
image and far away from where the velocimetry measurements were performed. For each image, the
signals due to stuck particles and background noise were removed by subtracting a reference image
formed from the rolling average of 10 previous and 10 subsequent image pairs. To account for laser
pulse energy fluctuation, every image was scaled by the measured laser illumination energy.
Velocimetry analysis started with finding the positions and peak intensities of all particles
via threshold-identification and Gaussian fitting of the intensities of the pixels surrounding the
peak. Subsequently each particle’s displacement was obtained with a limited-range nearest-neighbor
matching, and the displacement was decomposed into streamwise and cross-stream components
(for details of the tracking algorithm, see [80]). The particle seeding density was low enough such
31
0 0.5 1 1.5 2
0
0.5
1
1.5
2
P
D
F
Intensity Ratio (I
e
/I
e
0
)
Figure 3.7: Observed fluorescent particle intensity distribution and its probability density function
(PDF). The solid line represents the theoretical prediction (3.6) for an imaging range of 100 to 400
nm and a 15% intensity statistical variation with a uniform concentration distribution of particles.
that this tracking was unambiguous. Intensities of the “trackable” particles were converted to a
normalized intensity ratio obtained by dividing the raw intensity counts by the average of all the
stuck particles’ initial intensity counts. Finally, particles within a desired normalized intensity range
were selected for each flow rate and their mean streamwise velocity was defined as the apparent
velocity of the ensemble. For the hindered diffusion analysis, the same analysis steps were followed
except that diffusion coefficients of particles in quiescent fluid were obtained from the variance of
displacements instead of velocities.
3.4 RESULTS AND DISCUSSION
3.4.1 Validation of Intensity Calibration Curve
Figure 3.7 shows the distribution of particle intensities recorded during a typical data acquisition
(in this case, in quiescent fluid). The accompanying solid line is the theoretical prediction (3.6),
assuming an imaging range of 100 - 400 nm, and an effective emission radius variation of σ
r
= 0.15.
The good agreement between the observed and predicted intensity variation confirms our calibration
results and the statistical theory of the observed particle intensities discussed earlier.
3.4.2 Hindered Diffusion
Near-surface tracer particles are known to exhibit anisotropic hindered diffusion due to hydrody-
namic effects, and the hindered diffusion coefficient in the direction parallel to the substrate surface
32
0 0.5 1 1.5 2
0
0.2
0.4
0.6
0.8
1
(h−a)/a
D
e
x
p
/
D
0
Data 1
Data 2
Method of Reflection
Figure 3.8: Ratio of hindered diffusion coefficients (D
exp
) for Brownian motion parallel to a surface.
D

is [37]
D

D
0
= 1 −
9
16
_
h
r
_
−1
+
1
8
_
h
r
_
−3

45
256
_
h
r
_
−4

1
16
_
h
r
_
−5
+O
_
h
r
_
−6
, (3.14)
where D
0
is the Stokes-Einstein diffusivity of a spherical particle in the fluid bulk. This “Method
of Reflection” approximation is generally valid for (h − r)/r > 1. For (h − r)/r < 1, Goldman et
al. [38] proposed an asymptotic solution,
D

D
0
= −
2
_
ln
_
h−r
r
_
−0.9543
¸
_
ln
_
h−r
r

2
−4.325 ln
_
h−r
r
_
+ 1.591
. (3.15)
The combination of our particle-tracking algorithm and the intensity calibration curve was
applied to analyse near-surface particle Brownian motion parallel to the glass substrate. The
results are compared to the Methods of Reflection calculation (3.14) and shown in figure 3.8. The
experimental data shows reasonably good agreement with the theory, including the fall in the
effective diffusion coefficient close to the surface. The results further improve our confidence in the
overall experimental method and subsequent analysis.
3.4.3 Velocity Distributions of Particles
Figure 3.9 shows distributions of particle velocities at three different shear rates. Note that the
streamwise velocity distributions widen as shear rate increases, and their asymmetric shapes show
an increasing skewness as the shear rate increases. The increasing distribution width and skewness
at high shear rates are both results of sampling particle displacements from an imaging range with
finite thickness [80]. Physically, at higher shear rates, the finite-thickness imaging range contains
particles that are translated by a wider range of local velocities, hence the increasing width of the
33
−200 0 200 400 600 800 1000
0
0.005
0.01
0.015
0.02
Particle Velocity (µm/sec)
P
D
F
S = 183 sec
−1
S = 732 sec
−1
S = 1465 sec
−1
Figure 3.9: Distribution of observed particle streamwise velocities under various shear rates. Data
include measurements of flow over a hydrophobic surface.
apparent velocity distributions. In addition, because the particles are confined by the solid wall,
they are more likely to move away from the wall and into a faster moving shear plane. Thus the
distribution exhibits a more pronounced positive skewness at the higher shear rates.
The effect of shear with a finite-thickness imaging range can be characterized if the streamwise
particle velocity in figure 3.9, V , is scaled by
˜
V =
V −¸V )
¸V )
, (3.16)
where
˜
V is the scaled streamwise particle velocity and ¸V ) is the mean value of each velocity
distribution. The scaled velocity distributions are shown in figure 3.10. At each shear rate, the
relatively contributions of velocities from shear flow and Brownian motion can be characterized
by the Peclet number, Pe. Here the Peclet number is defined as the ratio of the shear-induced
displacement, L, to the diffusive displacement, s, during the interframe time period, ∆t,
Pe =
L
s
=
¸V ) ∆t

2D
0
∆t
. (3.17)
Because ¸V ) is shear-dependent only, the shear-induced effects in all velocity distributions in
figure 3.10 are scaled to equal magnitudes. It can be observed that small Peclet numbers are
associated with greater distribution widths while the distributions collapse onto a single skewed
distribution at high Peclet numbers. The explanation is the following: at low Peclet numbers the
distributions widths scale inversely to shear, suggesting a diffusion dominance and thus the large
widths of the velocity distributions are due to Brownian motion. As the Peclet number increases, the
effect of shear flow on particle motion increases and when Pe > 3, particle motions are completely
shear-dominant, causing the distributions to overlap. It should be noted that the overlapping
34
−1.5 −1 −0.5 0 0.5 1 1.5
0
0.5
1
1.5
Scaled Particle Velocity
P
D
F
Pe = 1.20
Pe = 1.93
Pe = 3.28
Pe = 4.16
Pe = 4.43
Pe = 5.52
Figure 3.10: Distribution of observed particle streamwise velocities after scaling the velocities
distributions in figure 3.9. The velocities are scaled according to (3.17). Pe of 1.20, 3.28 and 4.43
correspond to the shear rates of 183, 732 and 1465 sec
−1
in figure 3.9, respectively.
distributions have the same distribution width, asymmetric shape and the same positive tail, which
are absent in small Peclet number cases. These observations confirm the explanation that in figure
3.9, the increasing distribution width and skewness at high shear rates are both effects of shear.
3.4.4 Measurements of Apparent Slip Velocities
For each flow case, the intensity distribution of all “trackable” particles were compared to the
theoretical prediction (3.6) to identify the unknown parameters such as h
1
, h
2
and σ
r
through
optimization, as demonstrated in section 3.4.1 and figure 3.7. With these parameters known,
we calculated the theoretical apparent velocities of a particle ensemble, chosen from a normalized
intensity range, by using equation (3.12) and varying the slip length conditions. Subsequently, these
velocities were compared to the experimental apparent velocities (mean values of the streamwise
velocity distribution) of particles from the same normalized intensity range to extract experimental
slip velocities and slip lengths.
Figures 3.11 and 3.12 show the experimental and theoretical apparent velocities as a function
of wall shear rate over hydrophilic and hydrophobic surfaces, respectively. In both figures, the
solid lines represent the theoretical apparent velocities assuming zero slip (equation (3.11)). The
dashed and dotted lines are the expected apparent velocities if one were to assume a 50-nm and
a 100-nm slip length, respectively (calculated using (3.12)). In the case of the hydrophilic surface
(figure 3.11), the measured slip length ranges from 26 to 57 nm with a slightly increasing trend as
wall shear rate increases. The data obtained for flow over the hydrophobic surface (figure 3.12),
on the other hand, suggests a larger slip length ranging from 37 to 96 nm, with a more observable
increasing trend of wall shear rate dependence at higher shear rates. A quantitative comparison of
the two cases shows a slip length attributed to surface hydrophobicity ranging from -7 nm at low
35
0 500 1000 1500 2000
0
100
200
300
400
500
600
700
Wall Shear Rate (1/sec)
A
p
p
a
r
e
n
t

V
e
l
o
c
i
t
y

(
µ
m
/
s
e
c
)
No Slip (Theoretical)
Slip length = 50 nm
Slip length = 100 nm
Exp. data
Figure 3.11: Experimental apparent velocity of particles in a shear flow over a hydrophilic surface
(0.21 < I
e
/I
e
0
< 0.78), and its comparison to slopes of expected apparent velocities if there exists
a 0, 50 or 100-nm slip length. The error bar of each experimental data point represents a 95%
confidence interval. In all cases the 95% confidence intervals are within ±3% of the reported mean
velocity values.
shear rates to 54 nm at the highest tested shear rate, with an average value of 16 nm in between
(figure 3.13). The additional slip caused by surface hydrophobicity is in agreement with many
experimental results [25, 26, 28, 31], but is in sharp disagreement with others [30, 33, 64].
There are three possible sources of measurement uncertainty in figure 3.11 and 3.12: determina-
tion of (a) the wall shear rate, (b) the apparent velocities of particle ensembles and (c) the imaging
range. Because wall shear rates are calculated based on microchannel dimensions and flow rates
which are controlled by the syringe pump, we estimate that the uncertainty in the wall shear rates
is less than 3%. If there is significant slip at the wall, the shear rates could be slightly lowered than
the reported values. However, the largest observed slip length is less than 1% of the channel height,
and thus the slip effect on wall shear rate calculation can be safely neglected. The determination of
apparent velocities of particle ensembles also has extremely low uncertainly because central limit
theorem constrains the statistical mean of a large sample (a few thousand trackings) to a very low
uncertainty (< ±3% of the reported velocity values for 95% confidence intervals). In addition, with
subpixel resolution of the particle centre detection, our PTV algorithm is able to track particle
displacements with high accuracy (< 0.1 pixel, [84]). The largest source of uncertainty is the deter-
mination of the imaging range. The imaging range is determined statistically based on the particle
intensity decay length obtained from figure 3.6 where intensity measurements were performed in
increments of 20 nm, which could serve as the upper limit of imaging range uncertainty. However,
it should be noted that, with an uncertainty of 20 nm taken into consideration, the largest mea-
sured slip length still will not exceed 150 nm – much smaller than other experimental studies which
reported slip lengths greater than a few hundred nanometers.
One should note that the corrections due to shear-induced rotation (figures 3.11 and 3.12) do
36
0 500 1000 1500 2000
0
100
200
300
400
500
600
Wall Shear Rate (1/sec)
A
p
p
a
r
e
n
t

V
e
l
o
c
i
t
y

(
µ
m
/
s
e
c
)
No Slip (Theoretical)
Slip length = 50 nm
Slip length = 100 nm
Exp. data
Figure 3.12: Experimental apparent velocity of particles in a shear flow over a hydrophobic surface
(0.5 < I
e
/I
e
0
< 1.0), and its comparison to slopes of expected apparent velocities if there exists
a 0, 50 or 100-nm slip length. The error bar of each experimental data point represents a 95%
confidence interval. In all cases the 95% confidence intervals are within ±3% of the reported mean
velocity values.
0 500 1000 1500 2000
−20
0
20
40
60
Wall Shear Rate (1/sec)
A
d
d
i
t
i
o
n
a
l

S
l
i
p

L
e
n
g
t
h

(
n
m
)
Figure 3.13: The additional slip length due to surface hydrophobicity.
37
not account for the possibility of slip between the fluid and the tracer particle, and we are unaware
of any work that extends the theory of Goldman et al. [38] to include the effect of surface slip.
Physically, surface slip is unlikely to be an issue, because the tracer particles have relatively rough
surfaces. Nevertheless, we can qualitatively estimate its effect. Slip between the fluid and the
tracer particle will modify the particle velocity in two ways. Firstly, the reduced shear force will
result in the particle velocity lagging the local fluid velocity. Secondly, however, slip will reduce
the shear-induced rotation of the particle. Since these two mechanisms act in opposition to each
other, we believe that, as a first approximation, we can ignore this effect.
Granick et al. [34] and Lauga et al. [7] both pointed out that large surface roughness can
significantly decrease slip. The slip length we measured with sub-nanometer surface roughness,
however, agrees well with the surface roughness dependency reported by Zhu & Granick [26]. It
should also be noted that Joseph & Tabeling [31], with flow experiments conducted under similar
surface roughness, hydrophobic coating and shear rates, reported slip length very close to the
present results. It has also been suggested that electrokinetic effects could significantly affect
velocimetry measurements of suspended particles close to a surface. Lauga [35] reported that a
streaming potential can induce electrophoretic motion of particles if particle surfaces are charged,
and applying this theory to the current conditions results in an “apparent” slip length of at most
18 nm.
Electrostatic interaction between tracer particles and surface could also affect accuracy of ve-
locimetry measurements [7]. If the tracer particles and the substrate surface are similarly charged,
electrostatic repulsion can create a depletion layer near the substrate surface that effectively shifts
the imaging range. Because the slip length calculations in figure 3.11 and 3.12 are based on the
assumption that particle concentration is uniform near surface, a depletion layer could be a source
of “apparent slip” if it is not recognized. Lumma et al. [33] suggested that the depletion layer
thickness can be as large as 0.9 µm for a pure water and glass combination. Such a thickness seems
unreasonably large as it is much bigger than the evanescent field penetration depth and conse-
quently few fluorescent particles would be observed. In contrast, Flicker et al. [51] experimentally
measured the depletion layer thickness to be in the order of 10 to 30 nm. For our experimental
setup, calculations suggest that, based on the overall particle seeding density, if the particles are
uniformly distributed in the fluid one should be able to see around 19 particles in a 26 µm-by-26
µm image field. This is in close agreement with our experimentally-derived images, and we can
conclude that the depletion layer thickness must be much smaller than our imaging range and
thus effects due to a depletion layer thickness of this magnitude would not be detectable with our
current imaging technique. However, recognizing the possibility of such depletion layer’s existence,
we re-analysed the theoretical curves in figure 3.11 and 3.12 by assuming an imaging range farther
away from the substrate surface. We found such a shift would result in higher theoretical apparent
velocities of particle ensembles and therefore lower slip lengths for all shear rates, suggesting that
our reported slip lengths are most likely the upper bound of the true slip length value.
The electrostatic and electrokinetic interactions are an interesting pair as their significance de-
creases with increasing ionic strength of the test fluid. Perhaps a more accurate direct measurement
of the true surface slip can be obtained if aqueous solutions of moderately high ionic strengths were
used.
3.5 CONCLUDING REMARKS
The results presented here represent direct measurements of local velocities in the near wall region
of a fluid-solid boundary. Despite the relative simplicity of the experiment, several subtle issues
38
need to be carefully addressed, including the spatial non-uniformity of the illumination region, the
effects of the local shear on the dynamics of seeding particles, and the statistical variations in the
effective visibility of the seeding particles. The current experiment confirms previous measurements
in our own group [25] as well as some other recent measurements [26, 31] that there is minimal slip
over hydrophilic surfaces, and that hydrophobic surfaces do appear to introduce a discernable, but
small boundary slip.
Although the present technique does allow for measurement of particle motions very close to the
solid surface, it does suffer from some limitations and there is room for improvement. Most notably,
the tracking in the wall-normal direction is currently limited to an accuracy of approximately 10%
of the evanescent penetration depth (about 20 nm in the current case). If particles with a higher
degree of monodispersity can be reliably found, and if the imaging system noise can be reduced
(both of which are technical improvements on the immediate horizon), we believe that the technique
can be used to track particles with an order of magnitude improvement in accuracy. However, even
with the current levels of uncertainty, there is no possibility that the flow exhibits a slip length
more than 150 nm (figure 3.12) over the range of shear rates tested.
While the debate of slip versus no-slip will likely continue in the foreseeable future, we believe
that a consensus is forming from more recent experiments that the slip effect is not as extreme
as some studies might have suggested. As many authors have pointed out, measurement and
interpretation of slip data in the nanoscale can be very tricky because other physical phenomena,
such as electrokinetics and nanobubbles, might lead to observations of an “apparent slip” caused by
other complicating factors. All of these suggest that close attention must be to paid to documenting
the experimental conditions, and that this might well reveal the reasons behind the range of reported
apparent slip phenomena.
Chapter 4
Paper 3: Direct measurement of slip
length in electrolyte solutions
The emergence of microfluidics over the past decade has brought out increased research interests
in liquid boundary slip, which for centuries has been regarded as negligible. Advancements in
flow measurement techniques have resulted in many recent experiments reporting aqueous slip
length with a wide range of values and various degrees of dependency on applied shear rate, surface
hydrophobicity and surface smoothness [20, 21, 25, 26, 28, 30, 31, 32, 33]. A total internal reflection
velocimetry (TIRV) technique was recently reported to be an effective method in measuring liquid
velocities in the region of less than 300 nm from solid surfaces [80]. By conducting a three-
dimensional TIRV (3D-TIRV) experiment, Huang et al. reported that slip length can be obtained
within a 20-nm uncertainty [85]. They reported a small but non-negligible slip length for de-ionized
aqueous solutions in shear flows over both hydrophilic and hydrophobic surfaces. However, some
questions remained as to whether electrostatic and/or electrokinetic forces acting on the tracer
particles could lead to false slip observation [7, 35]. In this letter, slip length measurement of ionic
aqueous solutions using TIRV is presented and compared to the previous results of Huang et al. as
an attempt to answer these open questions.
The TIRV technique uses total internal reflection of an incident laser beam to generate a highly
localized illumination of the near-boundary liquid phase, and relies on tracking motions of individual
tracer particles to determine fluid velocity vectors in the planes parallel to a solid surface (as
demonstrated in figure 4.1). The exponentially decaying evanescent field leads to determination
of tracer particles’ positions in the direction normal to the solid surface based on their fluorescent
intensities. Subsequently slip velocities and slip length can be inferred from the measured apparent
velocity vectors by applying a statistical model for optical and hydrodynamic behaviors of small
particles near a solid/liquid interface. A detailed description of the 3D-TIRV theoretical work,
including a proof of concept and a discussion of its subtleties, was presented by Huang et al. [85].
A brief summary is presented in this letter to facilitate the readers’ understanding and resultant
discussion.
When a fluorescent particle of radius a is illuminated by an evanescent field of incident wave-
length λ
0
(a < λ
0
), its emission intensity, I
e
, exhibits the same exponentially decaying characteristic
as the evanescent field energy does, namely,
I
e
(z) = I
e
0
e
−h/p
, (4.1)
where h is the distance between the particle center and the solid phase, I
e
0
is the emission intensity
of the particle when it is in contact with the solid phase (h = a), and p is the evanescent wave
39
40
Figure 4.1: Objective-based total internal reflection velocimetry (TIRV) system.
penetration depth. The penetration depth, which characterizes the length scale of the evanescent
field, can be calculated from
p =
λ
0

_
n
2
1
sin
2
θ −n
2
2
_

1
2
, (4.2)
where n
1
and n
2
are the indices of refraction of the solid boundary and liquid, respectively. However,
a variety of factors, including statistical variations in illumination intensity, quantum efficiency of
the imaging system and, most importantly, non-uniformity of the particle sizes, can lead to varia-
tions in the observed fluorescent intensities of an ensemble of nominally identical particles. Thus,
a statistical analysis is carried out by characterizing these variations through a single parameter
r, or effective emission radius, with which a joint probability density function (PDF), P (h, I
e
), of
the particle distance to the solid surface and emitted fluorescent intensity can be obtained.
If near-wall velocities of a pure fluid are to be measured by tracking motions of tracer particles,
it is important to recognize that shear and near-surface hydrodynamic effects can cause a tracer
particle to rotate and translate at a velocity, U, lower than the velocity of the local fluid element in
the same shear plane [12, 38, 60]. Factors that determine the particle translation velocity include
particle radius, a, particle distance to the solid surface, h, and the local fluid shear rate, S (h).
Therefore corrections need to be made when one convert fluid velocities into particle velocities. It
should be noted that in the presented analysis, the corrected particle translation velocity does not
account for the possibility of slip between the fluid and the tracer particle, and thus any calculation
based on U represents a no-slip case.
Because tracer particles are used to probe near-surface shear flow, assurance is needed that the
particle ensemble uniformly samples all shear layers. It has been suggested that shear-induced lift
can be a source of particle migration away from the solid wall [86]. However, based on the theory
presented by Cherukat & McLaughlin [87], the shear-induced lift under the presented experimental
conditions is insignificant, and thus alleviating this concern (see Appendix B).
The ensemble-averaged velocity,
¯
U, of a large number of uniformly distributed particles emitting
in an intensity range of α < I
e
< β and located in an imaging range of h
1
< h < h
2
, can be
calculated by
¯
U =
1
h
2
−h
1
_
h
2
h
1
U (h, a, S (h)) P(h, α < I
e
< β) dh. (4.3)
Because equation (4.3) is based on a no-slip boundary condition, if there exists a slip velocity, U
slip
,
41
at the liquid/solid substrate interface, the same particle ensemble will exhibit an apparent velocity
of
¯
U
app
= U
slip
+
¯
U = δ ˙ γ
w
+
¯
U, (4.4)
where U
slip
is characterized as the product of the slip length, δ, and the wall shear rate, ˙ γ
w
. It
should be noted that
¯
U
app
can be experimentally measured by finding an ensemble-averaged velocity
of “trackable” tracer particles while ˙ γ
w
is computed from flow rate and channel geometry. With
¯
U
calculated by using equation (4.3), one slip length can be obtained for each wall shear rate.
Monodispersed Fluorescent Polystyrene Microspheres (Duke Scientific) with diameters of 100
(±10%) and 200 nm (±5%) suspended in purified water (Fluka) at 0.005-0.04% volume fraction
were used as tracer particles in this experiment. These microspheres have a peak absorption
wavelength of 542 nm and emit at 612 nm. Prior to each experiment, the fluid was degassed
by placing the solution in vacuum for at least 30 minutes. This procedure has been reported to
significantly eliminate nanobubble formation [83], which could be an alternative source of apparent
slip [7, 34, 82].
Test channels were fabricated using a polydimethylsiloxane (PDMS) (Dow Corning Sylgard 184)
molding technique [69] and bonded onto polished glass wafers. The dimensions of the test channels
were 50 ± 1 µm deep, 250 ± 1 µm wide and 15 mm long. Hydrophilic and hydrophobic microchan-
nels were created by oxygen plasma treatment of the PDMS mold and by bonding untreated PDMS
channels onto octadecyltrichorosilane(OTS)-coated glass wafers, respectively. The hydrophilic and
hydrophobic glass surfaces had an RMS surface roughness of 0.47 nm and 0.35 nm respectively.
A continuous-wave 514-nm Argon-ion laser beam (Coherent) was directed through a Nikon PL
Apo NA 1.45 100X TIRF oil immersion objective at an angle that created total internal reflection at
a glass-water interface, thus illuminating the near-surface region of water with an evanescent field
(figure 4.1). Fluorescent images of near-surface particles were captured by a Q-Imaging Intensified
Retiga CCD camera (ICCD). Approximately 1000 images were captured at each flow rate. The
flows were driven by a Harvard Apparatus 22 syringe pump, with special attention paid to removing
pulsations and maintaining steady pumping motions. Detailed description and schematics of the
experimental setup have been presented in [85].
The analysis of particle image pairs was performed using a custom particle-tracking algorithm.
In summary, velocimetry analysis started with finding the positions and peak intensities of all
particles via threshold-identification and Gaussian fitting of the intensities of the pixels surrounding
the peak. Subsequently each particle’s displacement was obtained with a limited-range nearest-
neighbor matching. Finally, particles within a desired intensity range were selected for each flow
rate and their mean streamwise velocity was defined as the apparent velocity of the ensemble and
compared to theoretical no-slip values. Slip lengths were then obtained based on equation (4.4).
Figure 4.2 shows the measured slip lengths of aqueous solutions under various experimental
parameters. The first observation to be made is that all measurements report slip lengths of
approximately 100 nm or less. These results further confirm previously reported measurements
[25, 26, 31, 85] that there is minimal slip over smooth solid surfaces. Slip on a hydrophobic surface
appears to be slightly higher than on a hydrophilic surface. A quantitative comparison of all
measurements shows that the slip length attributed to surface hydrophobicity averages to 22 nm at
the tested shear rates. The additional slip caused by surface hydrophobicity is in agreement with
many experimental results [25, 26, 28, 31], but is in sharp disagreement with some others [30, 33].
Electrokinetic effects [35] and electrostatic repulsion between tracer particles and glass surface
[7] have both been proposed as sources that might cause an apparent slip and could affect the
accuracy of slip velocity measurements. Their significance, however, is expected to decrease with
increasing ionic strength of test fluids. Lauga [35] suggested that for an aqueous solution with 1
42
0 500 1000 1500 2000
0
50
100
150
Wall Shear Rate (1/sec)
A
p
p
a
r
e
n
t

S
l
i
p

L
e
n
g
t
h

(
n
m
)
200−nm, DI−water, Hydrophilic
200−nm, DI−water, Hydrophobic
200−nm, 0.1mM NaCl, Hydrophobic
200−nm, 1mM NaCl, Hydrophobic
100−nm, DI−water, Hydrophobic
Figure 4.2: Measured slip lengths of aqueous solutions. The error bar on each data point repre-
sents a 95% confidence interval of each ensemble calculation. The uncertainty of each slip length
measurement (not drawn for plot clarity) is less than 20 nm, estimated based on penetration depth
calibration and particle radius variation [85]. Data of “200-nm, DI-water, Hydrophilic” and “200-
nm, DI-water, Hydrophobic” have been previously reported by Huang et al.
mM NaCl concentration, the apparent slip length would become sub-molecular. As shown in figure
4.2, variations in aqueous ionic concentration do not change measured slip lengths significantly.
This observation suggests that the measured slip lengths are most likely not due to ionic effects,
but are consequences of true boundary slip.
Another important observation is that the slip lengths do not vary significantly over the tested
range of shear rates. This observation agrees with experimental studies conducted under similar
ranges of moderate shear rates [28, 29, 31]. Still, this result does not suggest that slip length is
independent of shear rates, as the conducted measurements span only one order of magnitude of
shear rates. Zhu & Granick [26], Neto et al. [27] and Choi et al. [25] all tested over larger ranges
of shear rates (> 2 orders of magnitude) and reported shear-dependent slip lengths. It is physically
plausible that the shear dependence of slip length is nonlinear and the shear effect is significant
only at higher shear rates. To assess this, a TIRV system equipped with a very sensitive high-speed
camera would be needed for measurements at high shear rates.
Chapter 5
Paper 4: Direct measurement of
anisotropic near-wall hindered
diffusion using total internal reflection
velocimetry
By applying the three-dimensional total internal reflection velocimetry (3D-TIRV) technique to
freely suspended micron-sized fluorescent particles, we aim to simultaneously observe the three-
dimensional anisotropic hindered diffusion for gap-size-to-radius ratio much less than one. We
demonstrate that the three-dimensional tracking technique of TIRV used for nanoparticles can be
adopted into 3D displacement measurements of freely suspended 1.5-µm radius particles. The dis-
placement measurements reveal that the hindered diffusion coefficients are in close agreement with
the theoretical values predicted by the asymptotic solutions of Brenner [36] and Goldman et al. [37]
for gap-size-to-radius ratio much less than one, and hindered diffusion anisotropicity is simultane-
ously observed in all data sets.
5.1 INTRODUCTION
Advancement in colloidal sciences have lead to many applications of micro- and nano-particles in
engineering research (eg. particle imaging velocimetry) and biotechnology (eg. drug delivery). At
these length scales Brownian motions are significant and understanding these thermal agitations
are essential in making practical use of these tiny particles. Inside a fluid bulk the diffusivity of
an isolated particle follows the Stokes-Einstein relation, which is the ratio of fluid thermal energy
to the particle’s hydrodynamic mobility. In the presence of a nearby solid boundary, however, a
particle would experience an increased, anisotropic drag which hinders its mobility. Brenner [36]
and Goldman et al. [37] were the first to developed drag force correction factors for near-wall
spheres under no inertial effects and a no-slip boundary condition. These correction factors have
since evolved into theories of hindered diffusion [88].
Following these works, many experimental studies have been conducted to observe hindered dif-
fusion and to verify the correction factors. Demonstrated experimental techniques include evanes-
cent light-scattering [39, 40, 42], total internal reflection fluorescence microscopy (TIRFM) [43, 44]
and combined optical tweezers and digital video microscopy [41]. In most of the above mentioned
studies, simultaneous three-dimensional measurements of the anisotropic diffusivity could not be
43
44
conducted due to experimental limitations. One exception is a ratiometric-TIRFM measurement
conducted by Banerjee & Kihm [43], who reported anisotropic diffusivities that only partially agree
with the theories, probably due an inaccurate accounting of particle positions. Another exception
is a study reported by Lin et al. [41], who demonstrated that the anisotropic hindered diffusion
could extend far away from the wall.
In the theories of Brenner and Goldman et al. the hindered mobility of a near-wall particle is
a function of the particle’s radius, a, and the gap size between the particle and the wall, h. They
proposed analytical solutions of the hindered mobility that are vastly different for h/a 1 and
for h/a ¸ 1. Based on our literature survey, no study has demonstrated the anisotropic hindered
diffusion for h/a ¸ 1 with a simultaneous three-dimensional measurement as studies of Banerjee
& Kihm and Lin et al. both attempted to measure hindered diffusion at h/a 1. In this paper we
present our experimental study aimed to validate the anisotropic mobility correction coefficients for
h/a ¸1 through a direct measurement of hindered diffusion using three-dimensional total internal
reflection velocimetry (3D-TIRV).
5.2 THEORY OF HINDERED DIFFUSION
When an isolated particle is at the vicinity of a solid boundary, its Brownian motion is hindered
due to an increase in hydrodynamic drag. The presence of the solid wall also breaks the symmetry
of particle dynamics, resulting in anisotropic Brownian motion. Brenner [36] successfully solved
the lubrication equation of particle motion normal to the solid wall and proposed an infinite series
solution of the drag correction factor, β

. He reported that the correction factor is a function of
the particle radius, a, and the gap size between the particle and the solid wall, h. Later Bevan &
Prieve [39], using a regression method, reported that
β


D

D
0

6h
2
+ 2ah
6h
2
+ 9ah + 2a
2
(5.1)
is a close approximation to Brenner’s infinite series solution. For the diffusivity correction factor in
the direction parallel to the solid wall, β

, an exact analytical solution was never found. Instead,
Goldman et al. [37] offered an asymptotic solution,
β


D

D
0
= −
2
_
ln
_
h
a
_
−0.9543
¸
_
ln
_
h
a

2
−4.325 ln
_
h
a
_
+ 1.591
, (5.2)
for particles very close to the wall, defined by h/a ¸1.
5.3 EXPERIMENTAL PROCEDURES
To experimentally verify equations (5.1) and (5.2), diffusion of micron-sized fluorescent particles in
an evanescent field was tracked with digital imaging microscopy. The technique, termed 3D-TIRV,
have been demonstrated by Huang et al. in measuring motions of near-wall nanoparticles [89].
Unlike evanescent light-scattering experiment of non-fluorescent particles, TIRV allows particle
centers to be identified on the captured fluorescent images, and consequently particle motions
can be tracked via the displacements of particle centers. In addition, fluorescent intensity of the
particles can be used to infer the instantaneous separation distance between the particle and the
two-media interface, as commonly done in traditional TIRFM experiments [44, 90].
45
θ
z
x
water (n = 1.33)
glass (n = 1.51)
~ penetration depth
Figure 5.1: Schematic of TIRV. A red fluorescent microparticle suspended in water is placed in
a green evanescent field created within the focal depth of a high numerical aperture microscopy
objective. A collimated laser beam through the objective is used as the illumination light source.
At the illumination beam incident angle, θ, greater than 61.9

, total internal reflection occurs at the
glass/water interface. Because the particle radius is much larger than the evanescent penetration
depth, the evanescent energy illuminates only the lower portion of the encapsulated fluorophores.
Our 3D-TIRV setup was created on an inverted epi-fluorescent microscope (Nikon TE-2000). A
continuous-wave, collimated 514-nm Argon-ion laser beam was directed through a Nikon PL Apo
NA 1.45 100X TIRF oil immersion objective at an angle that created total internal reflection at
a glass/water interface. The water phase was an aqueous solution consisted of 10mM NaCl and
individual 1.5µm-radius (±5%) fluorescent polystyrene particles (refractive index 1.59, Spherotec
Inc.) suspended at 0.02% volume fraction. The low volume fraction, and thus low particle number
density, ensured the isolated-particle assumption of the theories was met. Before each measurement
the solution was sonicated in intervals of 30 seconds for 10 minutes to disperse coagulated particles.
For image acquisition, 200 µL of the solution was injected into a closed reservoir formed by a piece
of carved polydimethylsiloxane (PDMS) sandwiched between glass coverslips. Subsequently TIRV
imaging of hindered particle Brownian motions was performed at one of the glass/solution interface.
The fluorescent images of the particles were projected onto a Q-Imaging Intensified Retiga CCD
camera (ICCD), and recorded at a 10-Hz frequency. Overall, 2000 to 2500 sample images were
captured for each experiment for reliable statistics.
A particle’s 3D position was determined through threshold-identification and Gaussian fitting of
the intensities of the pixels surrounding the peak. This method has been previously demonstrated
in TIRV with nanoparticles. Still, a calibration experiment was necessary to verify its validity for
1.5-µm particles, especially in determining the relationship between the particle/wall gap size and
the particle’s peak fluorescent intensity. To do so, individual 1.5-µm radius fluorescent polystyrene
particles were attached to polished conical tips of thin graphite rods. By mounting the assembly to
a 0.4-nm precision translation stage, the fluorescent particles were traversed through the evanescent
field and their fluorescent images were digitally captured at various distances to the glass surface.
The result of this calibration is shown in figure 5.2. The close agreement between particle intensity’s
exponential decay length and the evanescent field penetration depth suggests that particle-glass gap
size can be reliably inferred from the particle peak intensity. We also found that the precision of
center identification was approximately 0.5 pixels, or 32 nm.
46
0 50 100 150 200 250 300
0
0.2
0.4
0.6
0.8
1
Gap size, h (nm)
I
/
I
0
Particle 1
Particle 2
Particle 3
Exp. fit (d = 142 nm)
Figure 5.2: Gaussian-fitted peak intensities, I, of 1.5-µm radius fluorescent particles in evanescent
field, taken at 0 to 275 nm away from the glass surface. I
0
is the fitted intensity at h = 0. Each data
point represents the particle peak intensity averaged over 10 images, while the length of the error
bar spans two standard deviations. The laser incident angle is 64

and the resulting evanescent
field penetration depth is 148 nm. All data series exhibit an exponential decay tendency versus
gap size. With an exponential decay length of 142 nm, the best fit of the intensity data shows close
agreement to the penetration depth.
Once positions of all imaged particles were determined, particle displacements were calculated
between consecutive frames. The hindered diffusion coefficient in the direction parallel to glass
surface, D

, was calculated from the equation
¸∆R)
2
= πD

∆t (5.3)
where ∆R is the averaged radial displacement of particles, ∆t is the time between consecutive
image acquisitions (= 100 millisecond), and ¸) represents ensemble average. Similarly, the hindered
diffusion coefficient in the direction normal to glass surface, D

, was calculated from
¸(∆z)
2
) −¸∆z)
2
= 2D

∆t (5.4)
where ∆z is the particle displacement normal to the glass surface. Finally, to obtain the hindered
diffusion correction factors, β

and β

, diffusion coefficient of the same 1.5-µm particles were
measured in the fluid bulk, using regular fluorescent digital microscopy.
5.4 RESULTS AND DISCUSSION
Shown in figure 5.3 are hindered diffusion correction factors measured by 3D-TIRV. In addition to
experimental results, hindered diffusion coefficients from a Brownian dynamics simulation [89] are
also plotted for comparison. These results are compared to the asymptotic solution of Goldman et
47
0 0.05 0.1 0.15 0.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
h/a
D
/
D
0
GCB
MOR
Bevan
D
X
(Sim.)
D
Z
(Sim.)
D
X
(Exp.#1)
D
Z
(Exp.#1)
D
X
(Exp.#2)
D
Z
(Exp.#2)
Figure 5.3: Hindered diffusion correction factor (D
x
/D
0
, D
z
/D
0
) vs. non-dimensional gap size
between particles and glass surface (h/a). D
0
is the diffusion coefficient measured in the fluid
bulk while D
x
= D

and D
z
= D

. Each data point is obtained from an ensemble of particles
found within ±0.005 or ±0.01 of the targeted gap size values. “GCB”, “MOR” and “Bevan” repre-
sent asymptotic solution of Goldman et al., the Bevan approximation and “Method of Reflection”
solution, respectively. “Exp.” represents experimental data while “Sim.” means data obtained
from Brownian dynamics simulation. Each error bar represents the 95% confidence interval of
measurement.
al. (equation (5.2)), Bevan’s approximation (equation (5.1)) and a “Method of Reflection” (MOR)
solution [37]. Theoretically equation (5.2) is more appropriate for the measured gap sizes of h/a ¸1
while the method of reflection solution is more accurate at h/a > 1, but nevertheless they are both
plotted for a quantitative comparison. The correction factor in the direction parallel to glass surface,
β

≡ D
x
/D
0
, increases from approximately 0.2 at h/a ≈ 0 to 0.4 at h/a ≈ 1.3, while the correction
factor in the direction normal to glass surface, β

= D
z
/D
0
, increases from 0 to 0.1 in the same
h/a range, demonstrating the anisotropicity of near-wall hindered diffusion. It is observed that the
measured β

values at h/a < 0.05 are very close to that predicted by equation (5.2) but significantly
deviate from the MOR solution, confirming the validity of the asymptotic solution at h/a ¸ 1.
The transition from the asymptotic solution to the method of reflection solution is demonstrated
at h/a > 0.1 as the measured β

values fall between the two solutions. The measured β

agrees
well with Bevan’s approximation, except at 0.05 < h/a < 0.1 where measured β

is slight larger
than the predicted values possibly due to insufficient or slightly biased particle sampling. A close
agreement between the measured and the predicted β

values at extremely small h/a has also been
reported by Oetama & Walz [42], who conducted one-dimensional hindered diffusion measurement
in the range h/a < 0.025. Finally, the experimental data also agree well with the simulation results,
confirming the validity of our measurement technique.
As in most near-wall colloidal measurements, we realize the importance of recognizing other
acting physical forces that could lead to measurement inaccuracy. One of these forces is gravitational
48
pull or sedimentation of the particles. At a first glance, it is perceivable that sedimentation could
lead to bias in diffusion measurement normal to the glass boundary. In the fluid bulk of density ρ
f
and dynamic viscosity µ, the terminal settling velocity, w
s
, of a particle with radius a and density
ρ
p
is
w
s
=
2 (ρ
p
−ρ
f
) ga
2

, (5.5)
where g is the gravitational acceleration. For a 1.5-µm radius polystyrene particle of density 1050
kg/m
3
, this terminal settling velocity would be 245 nm/sec, equivalent to a settling displacement
of 24.5 nm during the imaging inter-frame time of 0.1 seconds. However, when the particle ap-
proaches the solid boundary, its terminal velocity is expected to decrease with a correction factor
identical to β

for the same hydrodynamic reason of decreased mobility. Based on equation (5.1),
within the evanescent penetration depth the particles would settle for a distance of less 2.45 nm
in the inter-frame time of 0.1 seconds, or equivalently a non-dimensional gap size of 0.0016 for
our measurement gap size of h/a < 0.14. Therefore in the time scale of our Brownian motion
measurements, sedimentation can be considered as a negligible factor.
Besides hydrodynamic interaction, electrostatic repulsion can also exist between the polystyrene
particles and the glass surface as they both carry negative surface charges when immersed in
aqueous solution [51]. Indeed when the same fluorescent particles were suspended in de-ionized
water and observed under evanescent wave imaging, we found few particles came close to glass
surface. However the length scale of electrostatic repulsion can be easily decreased by increasing
the electrolyte concentration of the suspending aqueous solution. By adding 10 mM of NaCl to the
solution, we decreased the electric double layer thickness surrounding the particles and the glass
surface to a Debye length ∼ 3 nm, or h/a ∼ 0.002, and thus significantly reduced the effect of
electrostatic repulsion.
Another colloidal force that could potentially affect measurement accuracy is van der Waals
force between the particles and the glass. However, for a particle/glass gap size of 10 nm and
greater, van der Waals force scales with h
−3
[91], making its effective range even shorter than that
of electrostatic repulsion. Thus it is safe to assume that van der Waals force is also insignificant
for the range of h/a under consideration.
The hydrodynamic mobility correction factors proposed by Brenner and Goldman et al. assume
a no-slip boundary condition at particle surfaces. If fluid slip exists at the particle surfaces, it
is expected that the theories of Brenner and Goldman et al. would under-estimate near-surface
particle mobility and the hindered diffusion coefficients. Several recent theoretical and experimental
studies, however, offer insights into whether slip effect should be taken into consideration. First,
by conducting slip measurements of water flowing over hydrophobic surfaces, Zhu & Granick [26]
concluded that apparent slip lengths are only a few nanometers if the rms surface roughness is
3.5 nm or larger. Because the polystyrene microspheres used in the current hindered diffusion
experiment are hydrophilic and have surface roughness of at least 10 nm [92], the effective slip
length at particle surfaces, δ, is expected to be of only a few nanometers in magnitude, if not sub-
nanometer. Thus under the current experimental conditions, h/δ ¸ 10. Consequently, based on
the theory of particle mobility presented by Vinogradova [93], slip of such magnitude would have an
insignificant effect on the particle mobility for the particle diameter and particle-wall gap size under
consideration. Therefore the no-slip assumption is considered valid in the current experiment, and
is additionally verified by the agreement between experimental data and theoretical values shown
in figure 5.3.
49
5.5 SUMMARY
A direct measurement of hindered diffusion was conducted with total internal reflection velocimetry.
We believe that this is the first simultaneous three-dimensional measurement of hindered diffusion
coefficients for h/a ¸ 1. The anisotropic characteristic of hindered diffusion is experimentally
confirmed and the correction factors are found to agree with the theories of Brenner and Goldman
et al. The results confirm the increase of hydrodynamic drag when a particle approaches a solid
boundary, and such correction shall be applied to not only diffusion but also other translational
motion of particles where the drag force is of concern.
Chapter 6
Paper 5: Simulations of hindered
diffusion in shear flow and its
implications for near-wall velocimetry
The ensemble behaviors of near-surface particles in a shear flow, and the effects of the wall presence
and hindered diffusion on the measured apparent velocities are explored through Brownian dynamics
simulations with diffusive time and particle radius as time and length scales, respectively. It is
observed that the shear force dominates over diffusion at Pe > 3, and asymmetric shape and width
of apparent velocity distributions are attributed to the shear effect. It is also found that if the
time interval between successive particle image acquisitions is short, an under-estimation of fluid
velocity can result from biased sampling of tracer particles very close to the wall. However, if the
time interval between successive particle image acquisitions is long, a significant portion of the
tracer particles would sample planes of higher velocities, resulting in an over-estimation of overall
fluid velocity. Particle drop-in’s and drop-out’s, are found to be a potential source of measurement
error for nano-PIV because of the rapid Brownian motion of nanoparticles. The combined effect
of the observations described above is that near-wall particle-based velocimetry measurement is not
time-invariant. Still, the results found in the Brownian dynamics simulation offer a way where the
true fluidic velocities can be inferred from particle apparent velocities through proper scaling. This
scaling method, however, is available only for TIRV but not for nano-PIV, making TIRV the more
superior method in near-surface measurements.
6.1 INTRODUCTION
Colloidal particle-based image velocimetry has long been used as an experimental method in mea-
suring fluid velocity profile and other physical quantities. Its accuracy heavily relies on the assump-
tion that tracer particles will accurately trace local fluid velocities. In the past decade, particle
image velocimetry (PIV) has been extended to the micro- [55] and nano-scale [56] where size of
tracer particles has also decreased from microns to nanometers. At these physical length scales,
Brownian motion of tracer particles can be quite significant, thus violating the assumption of par-
ticles travelling at velocities equal to that of the surrounding fluids. Consequently, conducting an
accurate velocimetry experiment requires a good understanding and a careful treatment of Brow-
nian motion via either a thorough error analysis [55], an ensemble correlation averaging technique
[54], a statistical analysis of the particle ensemble [80, 85] or a method of statistical tracking [94].
50
51
In the vicinity of a solid wall, a colloidal particle will undergo a “hindered” Brownian motion
due to an lubricational increase of its hydrodynamic drag. This is known as hindered diffusion and
has been reported both theoretically [12, 36, 37, 38] and experimentally [39, 41, 42, 43, 95]. In most
cases, such a spatially dependent Brownian motion would lead to greater difficulties in accessing
measurement uncertainties. Only a few researchers managed to turn the disadvantages into merits
by proposing usage of the extent of hindering to reconstruct near-wall shear flow velocity profile
[96] and to probe the no-slip boundary condition [97].
To study the ensemble behavior of hindered and non-hindered Brownian motions, there are
two equally appropriate approaches [58]. One is the Fokker-Planck approach, which solves the
momentum-space partial differential equations to obtain a time evolution of spatial configuration
function of the particle ensemble. Recently, Sadr et al. [59] had taken the Fokker-Planck approach
to study the hindered diffusion of particles in the direction normal to the wall and its effect on
the bias of velocimetry. The other approach is a Brownian dynamics (BD) simulation, which is
to simulate motion of each particle in a large ensemble through a stochastic differential equation.
In this approach, both deterministic (such as fluid flow and sedimentation) and stochastic (such
as Brownian motion) processes contribute to particle displacements. The grand simulation then
yields a spatial distribution of the particle ensemble after each time step. Details provided by this
method are not only intuitive in understanding ensemble behavior but also bear a close resemblance
to physical velocimetry experiments.
The usefulness of BD simulations has been demonstrated in studying the impact of hindered
Brownian diffusion on the accuracy of nano-PIV [63], on the colloidal particle deposition in a
microchannel flow [61], and on the accuracy of potential energy profiles determined via total internal
reflection microscopy [62]. What has yet to be examined is the effect of hindered diffusion and the
presence of the wall on the accuracy of velocimetry. In addition, it would be of interest to examine
how the relative strength of hindered diffusion and shear flow leads to skewed near-wall apparent
velocity distributions first reported by Jin et al. in their total internal reflection velocimetry (TIRV)
experiment [80], and the relationship between particle diffusion, time between image acquisitions
and the accuracy of velocity measurements based on ensemble averages. In this paper the theories
of near-wall particle dynamics will first be presented and followed by a detailed description of
the BD simulation algorithm. We then discuss the simulation results and their implications, and
conclude with a summary and our thoughts on the accuracies of various velocimetry methods.
6.2 THEORIES AND COMPUTATIONS
6.2.1 The Langevin Equation
In Brownian dynamics simulation, particle displacements are computed based on a stochastic
Langevin equation [58]. For a particle in a shear flow as shown in figure 6.1, its displacement
between simulation time step i and step i + 1 with a step size δt is,
x

i+1
−x

i
= Uδt +
dD
x
dx

δt +
D
x
k
B
Θ
F
x
δt +N
_
0,
_
2D
x
δt
_
, (6.1a)
z

i+1
−z

i
=
dD
z
dz

δt +
D
z
k
B
Θ
F
z
δt +N
_
0,
_
2D
z
δt
_
, (6.1b)
where (x

, z

) is the particle’s center position, k
B
is the Boltzmann constant and Θ is the fluid
temperature. D
x
and D
z
are the diffusion coefficients in the directions parallel and normal to the
solid surface, respectively. In the fluid bulk, both D
x
and D
z
would be equal to the Stoke-Einstein
52
Solid
z
x
a
z’
x’
S
U
Fluid
Figure 6.1: A schematic of the simulation geometry. A colloidal particle of radius a is freely
suspended in a fluid but near a solid boundary. The fluid is undergoing a shear flow of shear rate
S, and thus the particle translates due to both fluid flow and its Brownian motion. A no-slip
boundary condition is assumed at the fluid/solid interface. Even though existence of slip has been
reported [85], its presence would not alter the simulation results and is thus neglected.
diffusivity, D
0
. As it has been mentioned earlier, near-wall diffusion is anisotropic and hindered,
and thus D
x
< D
0
and D
z
< D
0
. It should be noted that because D
x
is only a function of z

,
dD
x
/dx

= 0 in equation (6.1a). The z

-dependency of D
x
and D
z
is discussed in section 6.2.3.
F
x
and F
z
represent external forces acting on the particle in the x- and z-directions, respectively,
and N
_
0,

2Dδt
_
denotes the stochastic displacement of Brownian motion randomly sampled
from a normal distribution with a zero mean and a standard deviation

2Dδt. In the current
BD simulation scheme there exists no force in the x-direction and F
x
= 0. Density mismatch
between tracer particles and the suspending fluid, on the other hand, leads to either sedimentation
or floatation of the particles. The gravitation sinking force acting on the particles, F
g
, is thus the
only external force in the z-direction and
F
g
=

3
a
3
∆ρg, (6.2)
where ∆ρ is the density difference between the tracer particles and the suspending fluid, and g is
the gravitational acceleration.
The equations of motion (6.1a) and (6.1b) are non-dimensionalized by choosing the particle
radius, a, as the length scale and time required for the particle to diffuse a distance of one radius,
a
2
/D
0
, as the time scale. Therefore equations (6.1a) and (6.1b) become
X

i+1
= X

i
+F
_
Z

i
_
Pe Z

i
δT +N
_
0,
_

x
(Z

i
) δT
_
(6.3a)
and
Z

i+1
= Z

i
+

z
dZ

¸
¸
¸
¸
Z

i
δT +G β
z
_
Z

i
_
δT +N
_
0,
_

z
(Z

i
) δT
_
(6.3b)
53
S = 10
0
sec
−1
10
1
sec
−1
10
2
sec
−1
10
3
sec
−1
G =
a = 100 nm Pe = 0.0046 0.0463 0.4630 4.630 −5.042 10
−5
a = 1 µm 4.630 46.30 463.0 4630 -0.5042
Table 6.1: Sample values of the Peclet number, Pe, and sedimentation coefficient, G.
∆t = 0.1 msec ∆t = 1 msec ∆t = 10 msec ∆t = 100 msec
a = 100 nm 0.022 0.22 2.2 22
a = 1 µm 0.000022 0.00022 0.0022 0.022
Table 6.2: Representative values of the non-dimensional time between consecutive image acquisi-
tion, ∆T.
where X

≡ x

/a, Z

≡ z

/a, and T ≡ D
0
t/a
2
are the non-dimensional x

, z

and t, respectively and
F
_
Z

_

U
z

S
, β
x
_
Z

_

D
x
(z

)
D
0
, β
z
_
Z

_

D
z
(z

)
D
0
as described in sections 6.2.2 and 6.2.3 while
Pe ≡
Sa
2
D
0
, G ≡
4πa
4
g∆ρ
3k
B
Θ
.
The Peclet number, Pe, represents the relative strength of shear force and diffusion, and the
sedimentation coefficient, G, characterizes the gravitational pull on the particles. Representative
values of Pe and G are shown in table 6.1.
In image-based velocimetry measurements, another important control parameter is the time
between consecutive image acquisitions, ∆t. Some sample values of non-dimensional time between
consecutive image acquisition, ∆T, for 1-µm and 100-nm particles are shown in table 6.2.
6.2.2 Effects of Shear on Particle Velocities
It is well known that shear and near-surface hydrodynamic effects can cause a tracer particle to
rotate and translate at a velocity lower than the local fluid velocity in the same shear plane [12, 38].
Goldman et al. [38] proposed that the translational velocity, U, of a particle with radius a in a
shear flow of a local shear rate S is given by
U
z

S
≡ F
_
Z

_
· 1 −
5
16
_
Z

_
−3
, (6.4)
which is valid for large Z

, and
U
z

S
≡ F
_
Z

_
·
0.7431
0.6376 −0.2 ln (Z

−1)
(6.5)
for small Z

. Although there exists no analytical approximation for intermediate values of Z

,
Pierres et al. [60] proposed a cubic approximation to numerical values presented by Goldman et
al. in [38]:
U
z

S
≡ F
_
Z

_
·
_
1
Z

_
exp
_
0.68902 + 0.54756
_
ln
_
Z

−1

+0.072332
_
ln
_
Z

−1

2
+ 0.0037644
_
ln
_
Z

−1

3
_
.
(6.6)
54
Because a particle’s translational velocity depends on the shear rate and the distance to the wall,
it needs to be updated in equation (6.1a) at every time step of the simulation.
It has also been suggested that shear-induced lift can be a source of particle migration away
from the wall [86]. However, based on the theory presented by Cherukat & McLaughlin [87], the
shear-induced lift is insignificant for micro- and nano-particles in a low shear flow regime, and thus
will be neglected (for details, see Appendix B).
6.2.3 Hindered Diffusion
Near-wall tracer particles are known to exhibit anisotropic hindered Brownian motion due to hy-
drodynamic effects, and their hindered diffusion coefficient in the direction parallel to the solid
surface, D
x
, is [37]
D
x
D
0
≡ β
x
_
Z

_
= 1 −
9
16
_
Z

_
−1
+
1
8
_
Z

_
−3

45
256
_
Z

_
−4

1
16
_
Z

_
−5
+O
_
Z

_
−6
, (6.7)
where D
0
is the Stokes-Einstein diffusivity of an isolated spherical particle in the fluid bulk. This
“Method of Reflection” approximation is more accurate for Z

> 2. For Z

< 2, Goldman et al.
[38] proposed an asymptotic solution,
D
x
D
0
≡ β
x
_
Z

_
= −
2 [ln (Z

−1) −0.9543]
[ln (Z

−1)]
2
−4.325 ln (Z

−1) + 1.591
. (6.8)
In the direction normal to the wall, the modified diffusion constant, D
z
, is such that [36]
D
z
D
0
≡ β
z
_
Z

_
=
_
4
3
sinh α

n=1
n(n + 1)
(2n −1) (2n + 3)
_
2 sinh (2n + 1) α + (2n + 1) sinh 2α
4 sinh
2
_
n +
1
2
_
α −(2n + 1)
2
sinh
2
α
−1
__
−1
,
(6.9)
where α = cosh
−1
(Z

). Bevan & Prieve [39] reported that the equation
β
z
_
Z

_
=
6 (Z

−1)
2
+ 2 (Z

−1)
6 (Z

−1)
2
+ 9 (Z

−1) + 2
(6.10)
well approximates equation (6.9) and is much less computationally intensive than the infinite series.
Like the particle translational velocity, the dependence of these hindered diffusion coefficients on
Z

requires them to be updated at every time step of the simulation.
6.2.4 Implementation of Simulation
Since the experiments are conducted using dilute suspensions, the BD simulations were conducted
under an assumption of no particle-particle hydrodynamic interactions. At the beginning of each
simulation, a particle was situated at X

= 0 and randomly placed in a pre-determined range of
1 < Z

< 10. The simulation then progressed for a total of ∆T/δT steps as prescribed by equations
(6.3a) and (6.3b), with F and β’s updated and the position (X

, Z

) of the particle recorded after
each time step. Because the solid wall was located at Z = 0, the smallest Z

value a particle could
have was Z

= 1, where the particle would be in contact with the wall. A boundary condition
was needed in the event that a particle attempted to enter the solid wall during a simulation step.
55
Peters & Barenbrug [98, 99] have studied the efficiencies of different boundary conditions for BD
simulations. Here we chose a simple and yet effective specular reflection to prevent a particle from
entering the wall. With the time step taken small enough, the use of the boundary condition was
seldom triggered and less than 0.001% of the simulated displacements required applications of the
boundary condition.
The single particle simulation was then repeated 10
5
times to obtain a large ensemble. It was
also repeated for various values of parameters Pe and G to study their effects. The initial positions
of particles and the seeding of our random number generator were made identical for all runs of
simulations to ensure that the differences in results found between runs were the consequences of
simulation parameters only.
Finally, it is important to mention the selection of the size of the computational time step,
δT, which has two physical constraints [58]. First, the time step must be much greater than the
particle relaxation time, mD
0
/k
B
Θ, where m is mass of one particle. In nondimensional terms it
is equivalent to
δT ¸
mD
0
2
k
B
Θa
2
∼ O
_
10
−6
_
(6.11)
for a > 100 nm. Secondly, numerical accuracy requires that the time step must be short enough
such that the diffusion coefficients and their gradients are essentially constant during the time step
(that is, δT ¸ 1). Therefore δT was chosen to be 10
−4
, which satisfies both constraints and was
numerically efficient.
6.3 RESULTS AND DISCUSSIONS
6.3.1 Sedimentation
Accurate particle velocimetry measurements require unbiased sampling of fluid velocities by particle
ensembles. A common bias of this sort is sedimentation of particles between image acquisitions.
Shown in figures 6.2 and 6.3 are BD simulations of particle distribution for 1.5-µm and 100-nm
radius particles, respectively, sampled after various times ∆T. In the case of 1.5-µm particles, it can
be observed that the distribution started out uniformly, but evolves with time and eventually settle
into a steady, non-uniform spatial distribution after a transient period. The 100-nm particles,
on the other hand, remain dispersed through the volume at all times. In both case the final
spatial distribution agrees with the Boltzmann distribution prediction [91]. These results are not
surprising, but do serve to validate our simulation. With sedimentation being negligible for small
particles, we concentrate on presenting results of sub-micron particle simulations, as these particles
are commonly used in micro-PIV, nano-PIV and TIRV as tracers.
6.3.2 Particle Displacement Due to Hindered Diffusion
The first observation made is the displacement probability density function (PDF) of hindered
Brownian motion in the direction perpendicular to the wall, or the z-direction. It is well known
that the one-dimensional displacement PDF of an isolated particle in the fluid bulk is a Normal
distribution N
_
0,

2D
0
∆t
_
. However, with the presence of the wall hindered diffusion is expected
to break the symmetry of the displacement PDF, and is demonstrated in figure 6.4. Particles that
start off farther away from the wall can diffuse a longer distance within a given amount of time,
as demonstrated by the wider displacement distribution widths. In addition, these displacement
PDF’s are more symmetric and collapse onto each other. The displacement PDF’s of near-wall
56
1 2 3 4 5 6 7 8 9 10
0
0.2
0.4
0.6
0.8
1
Z
P
D
F
(a)
1 2 3 4 5 6 7 8 9 10
0
0.5
1
1.5
2
2.5
Z
P
D
F
(b)
Δ T = 0
Δ T = 0.5
Δ T = 1
Δ T = 2
Δ T = 3
Δ T = 5
Δ T = 5
Δ T = 6
Δ T = 7
Δ T = 8
Δ T = 9
Δ T = 10
Boltzmann Distribution
Figure 6.2: Sedimentation of particles at G = −2.5, which is equivalent to a 1.5µm-radius
polystyrene particle (density 1050 kg/m
3
) suspended in pure water. The particles are uniformly
distributed in the volume (1 < Z

< 10) at T = 0 and allowed to diffuse only within this volume.
PDF stands for probability density function.
57
1 2 3 4 5 6 7 8 9 10
0.04
0.06
0.08
0.1
0.12
Z
P
D
F
(a)
1 2 3 4 5 6 7 8 9 10
0.04
0.06
0.08
0.1
0.12
Z
P
D
F
(b)
T = 0
T = 0.5
T = 1
T = 2
T = 3
T = 5
T = 5
T = 6
T = 7
T = 8
T = 9
T = 10
Boltzmann Distribution
Figure 6.3: Sedimentation of particles at G = −5 10
−5
, which is equivalent to a 100nm-radius
polystyrene particle (density 1050 kg/m
3
) suspended in pure water. The particles are uniformly
distributed in the volume (1 < Z

< 10) and allowed to diffuse only within this volume.
58
−2 −1 0 1 2
0
0.5
1
1.5
2
Z − Z
0
P
D
F
Z
0
= 1.5
Z
0
= 2
Z
0
= 6
Z
0
= 11
Z
0
= 51
Z
0
= 101
Figure 6.4: Displacement distribution of small particles from various initial positions, or Z
0
. All
particles are allowed to diffuse for an identical time ∆T = 0.1, during which time no particle is
able to reach the wall.
particles, on the other hand, are significantly skewed with a more pronounced tails in the positive
values and are narrower. This observation suggests that hindered diffusion could lead to a bias of
measurements toward values obtained in the near-wall region.
Further proof of the hindered diffusion effect is shown in figure 6.5. Particles start to diffuse
away from the initial position of Z

= 2 as time evolves, with the spatial distribution skewed toward
the wall (Z = 0), which is a direct consequence of hindered diffusion. After long periods of time
when the particles are allowed to reach the wall through diffusion (∆T > 0.3), the skewness of the
spatial PDF is even more pronounced, leading to a high concentration of particles at the wall. Such
time-dependence of the particle spatial distribution could lead to time-dependent apparent velocity
measurements and is further discussed in section 6.3.5.
6.3.3 Particle Drop-in’s and Drop-out’s
The accuracy of velocimetry measurements relies on the ability to detect the same particles in
consecutive image acquisitions. This is especially difficult in near-wall microscopy measurements
because near-wall sub-micron particles can diffuse away from the near-wall region quickly. One
such situation is demonstrated in figure 6.6. Suppose the imaging range of near-wall microscopy is
1 < Z

< 5, which is a good estimate for evanescent wave imaging using 100-nm radius particles.
The particles, initially uniformly distributed within the imaging range, will start to leave the
imaging range as time evolves. The portion of the particles that are not present in the imaging
range at the second image acquisition are said to have “dropped out”. Clearly, more particles
would drop out with longer inter-acquisition time. For the overall concentration of the particles to
remain uniform, many “outside” particles would have to “drop in” to take over the space. In PIV,
images of the dropping in particles at the second image acquisition would lead to additional errors
59
1 2 3 4 5
0
0.5
1
1.5
Z
P
D
F
Δ T = 0
Δ T = 0.1
Δ T = 0.3
Δ T = 0.5
Δ T = 1
Δ T = 5
Figure 6.5: Spatial distribution of small particles at various times due to hindered diffusion. All
particles are located at Z

= 2 at T = 0.
0 5 10 15
0
0.05
0.1
0.15
0.2
0.25
Z’
P
D
F
Δ T = 0
Δ T = 0.0001
Δ T = 0.001
Δ T = 0.01
Δ T = 0.1
Δ T = 1
Δ T = 10
Figure 6.6: Spatial distribution of particles located within the range of 1 < Z

< 5 at T = 0. The
distributions are taken after various inter-acquisition times, ∆T.
60
0 2 4 6 8 10
30
40
50
60
70
80
90
100
Δ T
%

R
e
m
a
i
n
i
n
g

i
n

I
m
a
g
i
n
g

R
a
n
g
e 1.0 < Z
0
< 3.0
1.0 < Z
0
< 5.0
3.0 < Z
0
< 5.0
Figure 6.7: Percentage of particles in the imaging range for different inter-acquisition time. The
imaging range is the same as the initial particle distribution range, Z
0
.
in measurements.
As mentioned previously, the major concern for PIV is the percentage of particle drop-in’s and
drop-out’s because all imageable particles at the time of image acquisition are analyzed without
any effort to distinguish the entering and the leaving. The percentage of particles that do not
drop out for various inter-acquisition time is shown in 6.7, which is in agreement with a separate
numerical study conducted by Sadr et al. [59] Not surprisingly, a longer times between consecutive
acquisition results in a lower percentage of “good” particles, which are particles that do not drop
out. In dimensional terms, more than 90% of 100-nm particles remain in the imaging range of
100 < z < 300 nm (10% drop-out) if image acquisitions are taken at 1 msec apart. However, this
percentage drops to 65% if the inter-acquisition time is 10 msec. Based on this result, near-wall PIV
measurement with sub-micron particles should be conducted with a inter-acquisition time as short
as possible to ensure accuracy. If an acceptable percentage of drop-out’s is 95%, the acquisitions
should be taken with ∆T < 0.2, or ∆t = 0.5 msec for 100-nm radius tracer particles. The drop-
out’s, however, present less of a problem for Particle Tracking Velocimetry (PTV) as the technique
makes an effort to distinguish the particle drop-in’s and drop-out’s.
6.3.4 Horizontal Apparent Velocity Distributions
Subsequent analysis concentrates on horizontal motion of particles that do not drop out, emulating
physical TIRV measurements. In particular, the focus is on the effect of shear on apparent velocity
measurements. Shown in figure 6.8 is the distribution of non-dimensional apparent velocity, V =
∆X

/∆T, for particles under various Peclet number, Pe, or equivalently the shear rates. It is
clearly observable that the apparent velocity PDF’s widen and skew as shear rate increases. This
agrees well with experimentally measured PDF’s reported by Jin et al. [80] and Huang et al. [85].
Jin et al. suggested that the increasing skewness and widening of the apparent velocity distribution
61
0 10 20 30 40 50
0
0.2
0.4
0.6
0.8
V
P
D
F
Pe = 0.5
Pe = 1
Pe = 3
Pe = 5
Pe = 10
Figure 6.8: Apparent velocity distribution of particles remaining in the imaging range 1 < Z < 3
after ∆T = 10. Larger Pe leads to an apparent velocity distribution with larger mean, width, and
skewness.
−3 −2 −1 0 1 2 3
0
0.5
1
1.5
(V−V
avg
)/V
avg
P
D
F
Pe = 0.3
Pe = 0.5
Pe = 1
Pe = 3
Pe = 5
Pe = 10
Pe = 30
Figure 6.9: Normalized apparent velocity distribution of particles remaining in the imaging range
1 < Z < 3 after ∆T = 10. V
avg
is the mean of each apparent velocity PDF. The normalized
PDF’s collapse onto each other at Pe > 3. In general, larger Pe has a narrower but more skewed
normalized PDF.
62
0 50 100 150
0
0.01
0.02
0.03
0.04
0.05
V
P
D
F
1 < Z
0
< 2
2 < Z
0
< 3
3 < Z
0
< 4
4 < Z
0
< 5
5 < Z
0
< 6
Figure 6.10: Apparent velocity distribution of particles at various imaging ranges. All apparent
velocity distributions are measured at ∆T = 10 and Pe = 10. Particles that start off farther away
from the surface move faster because they are carried by fluids at higher velocity planes, and their
distributions are more symmetric due to less influence of the wall and hindered diffusion.
violates the symmetric PDF assumption of PIV analysis and could lead to measurement errors.
Upon further analysis, it is observed that the apparent velocity PDF’s can collapse onto a single
distribution for Pe > 3, as shown in figure 6.9. It is noted that the collapsed PDF’s for Pe > 3
have a smaller distribution width but a larger skewness, while the distributions of Pe < 3 are wider
but more symmetric. The explanation is that at low shear rates, the horizontal displacement of
the particle ensemble is dominated by diffusion, leading to a wider normalized PDF. At high shear
rates, on the other hand, the finite thickness of the imaging range and shear effect give rise to
the skewness and distribution widening. This physical explanation is supported by the fact that
both skewness and distribution width are scalable by the shear rate only at Pe > 3. The same
shear-dominance effect at Pe > 3 has been reported experimentally by Huang et al. [85], and the
agreement of BD simulations further validates the presented conclusion.
Another source of apparent velocity distribution skewness is the hindered diffusion and the
presence of the wall. One can observe in figure 6.10 that the apparent velocity PDF’s of imaging
ranges farther away from the wall are wider and less skewed than that of imaging ranges closer to
the wall. The difference is even more obvious if each PDF is re-scaled by its distribution mean,
V
avg
, as plotted in figure 6.11. The combined effect of the hindered diffusion and the shear-induced
rotational slow-down in the near-wall region causes the apparent velocity PDF’s to skew, while at
far away the wall presence is not felt and the symmetric distributions again show scaling similarity.
6.3.5 Time Evolution of Apparent Velocity Distributions
Finally, the time evolution of apparent velocity distributions was examined and shown in figure
6.12. For a fixed Pe, it is observed that diffusion is quite dominant at small ∆T, leading to apparent
63
−100 −50 0 50 100
0
0.01
0.02
0.03
0.04
0.05
V − V
avg
P
D
F
1 < Z
0
< 2
2 < Z
0
< 3
3 < Z
0
< 4
4 < Z
0
< 5
5 < Z
0
< 6
6 < Z
0
< 7
7 < Z
0
< 8
Figure 6.11: Collapsed apparent velocity distribution shown in figure 6.10. The PDF’s are scaled
by subtracting their corresponding distribution mean, V
avg
.
−300 −200 −100 0 100 200 300 400
0
0.005
0.01
0.015
0.02
0.025
V
P
D
F
Δ T = 0.0001
Δ T = 0.01
Δ T = 0.01
Δ T = 0.1
Δ T = 1
Δ T = 10
Figure 6.12: Time evolution of apparent velocity distribution. All apparent velocity PDF’s are
obtained at Pe = 30 and with particles within the imaging range of 1 < Z

< 3 at T = 0. Note
that the apparent velocity distribution narrows and skews with increasing inter-acquisition time,
∆T.
64
0 2 4 6 8 10
0.9
0.95
1
1.05
1.1
1.15
Δ T
V
a
v
g
/
V
s
h
e
a
r
Pe = 0.3
Pe = 0.5
Pe = 1
Pe = 3
Pe = 5
Pe = 10
Pe = 30
Figure 6.13: Time evolution of mean particle apparent velocity. Each data point represents the
mean value of one apparent velocity distribution shown in figure 6.12. All mean apparent velocities
are obtained from particles that remained at imaging range of 1 < Z < 3. V
shear
is the calcu-
lated velocity of the particles in a near-wall shear flow if there is no Brownian motion, and is the
target quantity that velocimetry is trying to measure. Thus V
avg
/V
shear
can be interpreted as the
accuracy of velocimetry measurement. Note that data of all Pe show a minimum velocity ratio at
approximately ∆T = 1.
65
10
−3
10
−2
10
−1
10
0
10
1
0.9
1
1.1
1.2
1.3
1.4
1.5
ΔT/W
2
V
a
v
g
/
V
s
h
e
a
r
W = 2
W = 4
W = 6
W = 8
W = 10
Figure 6.14: Rescaled time evolution of mean particle apparent velocity. For each data series, the
imaging range is set at 1 < Z

< W and the non-dimensional time is rescaled with 1/W
2
. Note that
the velocity ratios, V
avg
/V
shear
, collapse onto each other under this rescaled time scheme, except
near the minimum. Minimal velocity ratios for all W’s occur near an alternative non-dimensional
time scale of ∆T/W
2
≈ 10
−1
. All data series have Pe = 30.
velocity distributions that are significantly wider than that of at large ∆T. Due to increased shear
dominance, the apparent velocity distribution narrows as time increases. Such change can be
explained by the fact diffusive velocity distribution wideth scales with ∆T
−1/2
,
¸∆X) ∼

∆T ⇒ ¸V ) =
¸∆X)
∆T

1

∆T
, (6.12)
while the shear-induced flow velocity distribution width scales linearly with ∆T. Thus diffusion
accounts for a large fraction of the measured apparent velocity at short inter-acquisition time,
while the majority of the measure velocity is due to the shear flow at large ∆T. This explains
the narrowing of the apparent velocity distributions in figure 6.12 as ∆T increases. Clearly, if the
apparent velocity distributions evolve with time, it would be of interest to see if the mean value of
the velocity distributions evolve as a function of time as well.
The goal of vecimetry measurements is to accurately estimate fluid velocities using the tracer
particle apparent velocity as an estimator. If V
shear
is defined as the mean velocity of all fluidic
planes inside the imaging range and V
avg
is the mean value of an apparent velocity distribution such
as the ones shown in figure 6.12, then V
avg
/V
shear
can be interpreted as velocimetry measurement
accuracy with V
avg
/V
shear
= 1 being the ideal measurement. The time dependence of the velocity
ratio V
avg
/V
shear
is shown in figure 6.13. It can be observed that all V
avg
/V
shear
ratios exhibit the
form of a concave-up function of ∆T, with a minimum occurring at approximately ∆T = 1. The
velocity ratio would keep increasing as ∆T increases. Figures 6.12 and 6.13 demonstrate that both
the shape and the mean of an apparent velocity distribution are not time-invariant.
With further analysis, it can be observed that the values of ∆T at which the minima of
V
avg
/V
shear
ratio occur depend on the thickness of the imaging range. In their paper Sadr et al. [59]
suggested an alternative time scale, ∆T/W
2
, if the imaging range is defined at 1 < Z

< W. This
alternative time scale, therefore, can be interpreted as the time it takes for a particle in contact
66
path 2
path 1
Imaging
range
Figure 6.15: Schematic of 2 potential paths for particle translation. Between image acquisitions,
particles that have taken different paths could have travelled different distances while sampling
different velocities along the way. Two potential paths of a particle are shown in this figure. If
the particle takes path 1, it samples the velocities of the lower planes and thus translates a short
horizontal distance. On the other hand if the particle takes path 2, it samples the higher velocities
of planes that are out of the imaging range, and thus translates a larger horizontal distance.
with the wall to diffuse out of imaging range due to Brownian motion. Using this non-dimensional
time scale, figure 6.14 demonstrates that the velocity ratio V
avg
/V
shear
also has a scaling similarity.
In addition, it should be noted that in both figures 6.9 and 6.14, V
avg
/V
shear
→ 0 at ∆T → 0 as
the instantaneous apparent velocities of particles would be the closest to the local fluidic velocities.
So why is the velocity ratio minimal at ∆T/W
2
≈ 10
−1
, and increases monotonically afterward?
The answer to this question lies in figure 6.5 and is illustrated in figure 6.15. For small ∆T, the
skewed distribution of particles toward the wall suggests that the imaged particles are more likely
to move closer to the wall. That is, path 1 in figure 6.15 is more likely to be taken than path
2. Since drop-out’s are less significant for small ∆T, the consequence is that the tracer particles
sample toward the lower velocity planes, and thus the mean apparent velocity decreases. On the
other hand for large ∆T, some of the particles that have left the imaging range are given enough
time to return at the second acquisition, such path 2 illustrated in figure 6.15. The larger the ∆T,
the higher probability that particles can sample farther away from the surface and still manage to
return to the imaging range at the time of second image acquisition. Because they have sampled
the fluidic velocities of the higher shear planes, these contributions of fluidic velocities outside of
the imaging range are reflected in the higher values of mean particle apparent velocities.
Clearly, the physical reasoning described in the previous paragraph suggests that the thickness of
the imaging range would be an important parameter, which justifies the scaling of non-dimensional
time ∆T with 1/W
2
. Indeed Sadr et al. used the Fokker-Planck approach to study the PDF
of fluidic planes that an ensemble of particles would have sampled between image acquisitions in
nano-PIV. They obtained a ratio of averaged Z

value sampled by the particles to the mid-distance
of the imaging range, (W + 1) /2. The time dependence of this ratio is in striking resemblance to
that shown in figure 6.14. In both cases, the turning point in the relative contribution of the two
groups of particles (the ones that move closer to the surface and the ones that sample the higher
shear plane and return) occurs at ∆T/W
2
≈ 10
−1
. This suggests that the percentage of particle
drop-out and the positions whose fluidic velocities are sampled by the particles play a crucial role
in the accuracy of velocimetry measurements.
Figures 6.13 and 6.14 can also serve as guides for choosing the appropriate ∆T for near-wall
particle tracking velocimetry. First, one can observe in figure 6.13 that the velocity ratio is Pe-
67
invariant except near the minimum where fluctuations exist. Thus the selection of ∆T such that
the velocity ratio would be minimum should be avoided. Such avoidance is further reinforced by the
observation that the velocity ratio in figure 6.14 is not W-invariant in regions near the minimum.
It can be clearly observed that the “dip” in velocity ratio depends on the values of W, where large
W values lead to a lower minimal velocity ratios, while the velocity ratio shows no dependence on
W at both the higher and lower ends of the time scale ∆T/W
2
. These observations suggest that
choosing an inter-acquisition time of ∆T/W
2
∼ O
_
10
−1
_
would most likely yield measurement
inaccuracy.
6.4 CONCLUDING REMARKS
Brownian dynamics simulation offers an attractive approach to the study of near-wall particle dy-
namics. We demonstrate that wall-induced hindered diffusion and translational slow-down of a
colloidal particle in a simple shear flow can be accurately captured by Brownian dynamics sim-
ulations, with results confirmed by experimental data. In near-wall colloidal dynamics, physical
length and time scales are particle radius and diffusive time, and the problem can be mathematically
closed by specifying a Peclet number and a sedimentation constant. In the direction perpendic-
ular to the wall, hindered diffusion skews the displacement distributions toward to wall. In the
direction parallel to the wall, colloidal particle motion is dominated by shear when Peclet number
is greater than 3, with hindered diffusion also contributing to changes in the velocity distribution
width and its skewness. Furthermore, velocity distributions evolve as a function of time, reporting
mean velocities that are time-dependent with a minimum at ∆T/W
2
≈ 10
−1
.
This study also provides an assessment tool to the relative accuracy of near-wall PIV and PTV.
Because of a large amount of drop-in’s and drop-outs when nanoparticles are used, the inaccuracy
of PIV in near-wall measurement could be significant. PIV will remain accurate as long as the
time between acquisition is much smaller than the diffusion time of tracer particles. As tracer
particles get smaller and their diffusion become greater, the incapability of imaging equipments to
make consecutive image acquisitions with an extremely short inter-acquisition time would limit the
applicability of PIV in near-wall measurements.
On the other hand, PTV and SPTV (Statistical Particle Tracking Velocimetry) [94] are much
more reliable measurements because PTV attempts to exclude particle drop-in’s and drop-out’s
while SPTV treats the drop-in’s and drop-out’s as statistical noise. The potential error of both
methods would stem from biased sampling of the velocities at various shear planes. However, figure
6.14 offers a way to correct for this bias. By using it as a scaling guide, the correct shear-induced
velocity of the particles can be retrieved from the measured mean velocity values. An alternative
method is to calculate which inter-acquisition time would result in the measured mean velocity
being equal to the shear-induced velocity before experiments are conducted. Neither correction
method can be applied to PIV measurements.
Chapter 7
Concluding Remarks and Suggested
Studies
In this dissertation we have presented our experimental studies of aqueous solution boundary slip
and anisotropic hindered diffusion, and described the total internal reflection velocimetry (TIRV)
method used in these experiments. A Brownian dynamics simulation study is also presented to
compare the effectiveness and accuracy of TIRV and nano-PIV. Furthermore, these studies open
up opportunities for future experimental works in the areas of fluid mechanics and colloidal science.
First of all in pursuing a better understanding of boundary slip, water is by no means the
only test fluid of interest. Water and similar solvents such as alcohols have a polar molecular
configuration, and their chemical properties are vastly different from the non-polar group of solvents,
such as hexane and toluene. It would certainly be of scientific interest to see if the solvent molecular
polarity would have an effect on boundary slip, and if the solid surface charges would interact with
the weak charges carried by the solve molecules.
The TIRV technique can certainly be extended to near-surface measurements in a gas phase.
Although the existence of boundary slip at a gas/solid interface has been well documented, the
TIRV technique offers a method to make direct measurement of the gas slip velocities. In addition,
TIRV can also be applied to experimentally investigate flow characteristics within the gas boundary
layer. Of course, fluorescent aerosols would be required as tracer particles, and we believe such
technology is possible in the near future.
Returning to the liquid phase, particle or cell adhesion and desorption are both areas of great
interests in colloidal science and biology, and the TIRV technique is ideal for studying these phe-
nomena. The area of interest include both specific and non-specific binding of suspensions and
the solid boundary. To investigate specific chemical binding, the particles and the solid boundary
surface can be coated with molecules that bind correspondingly, such as ligands and receptors to
study blood cell adhesion. Besides binding, kinetics of surface-based chemical reactions can also be
studied with the same mechanism.
Understandably, advancement of technology would soon deem particles of 100 nanometer in-
sufficient in probing the near-wall region as scientists would like to look even closer to the solid
surface. Thus a new breed of tracer particles or molecules in the order of 1 nanometer would be
needed. In fact, quantum dots have been viewed as the next generation of tracer probes, and their
usage under evanescent wave imaging settings have been demonstrated by Pouya et al. [100] and
Guasto et al. [94]. These quantum dots are small (a few nanometer in diameter), more uniform in
size than polystyrene fluorescent particles, and available in water or organic solvent soluble forms,
thus making them an ideal candidate to probe regions that are even closer to the wall with a variety
68
69
of liquids. Perhaps a more accurate measurement of the slip velocities can be achieved by using
quantum dots in TIRV.
The Brownian motion of these quantum dots is undoubtedly much more significant and the par-
ticle tracking method in the TIRV will be incapable of dealing with the large proportion of particle
drop-in’s and drop-out’s between image acquisitions. An additional limitation of the particle-
tracking based TIRV is its requirement of low particle seeding density to avoid tracking ambiguity.
In some experimental situations, however, a high seeding density might be desirable to preserve
the particle-particle interactions and/or to increase measurement efficiency. The statistical particle
tracking velocimetry (SPTV) developed by Guasto et al. [94] is a perfect solution to these prob-
lems. They proposed that the true statistics of the desired physical displacement can be extracted
by matching each detected particle to all others, provided that experimental parameters such as
camera noise, imaging depth, particle size and diffusivity, and seeding density are known. The
SPTV thus eliminates the needs to keep track of particle drop-in’s and drop-out’s and to unam-
biguously matching detected particles. It is certainly more advantageous than other particle-based
velocimetries, especially in near-surface measurements.
Appendix A
Calibration of beam incident angle
In our TIRV setup, the total internal reflection was created at the glass/water interface by directing
a laser beam through a microscope objective lens. Knowing the beam incident angle is critical in
determining evanescent field penetration depth. In this appendix, the measurement method of the
illumination beam incident angle is presented.
As demonstrated in equation (2.3), it would be ideal if one can measure the beam incident angle
directly when determining the penetration depth. In practice, because the total internal reflection
bends the reflected beam back into the objective lens, as shown in figure 2.1, such measurement is
extremely difficult.
The illumination laser beam is directed into the objective lens by passing through the a con-
verging lens. The bending of the beam is achieved by translating the converging lens [47]. Thus
a geometric relation should exist between the the lens translation distance and the beam incident
angle, and this relation should be true for all incident angles despite of total internal reflection. To
determine this relation, the laser beam is directed through the objective as used in a non-TIRF
setup. The beam is then allowed to refract at a glass/air interface, as shown in figure A.1. Snell’s
law determines the refraction angle by
n
1
sin φ = n
2
sin θ, (A.1)
where n
1
and n
2
are refractive indices of glass and air, respectively. The refracted beam in air is
then projected to a far-away wall or ceiling where the refractive angle can be measured accurately
(shown in figure A.2).
By examining the data, we found a linear relationship between the lens position, x, and sin θ.
Using equation (A.1), an empirical linear relation between x and sin φ was found (shown in figure
A.3). The best fit line was then extended to φ greater than critical angle to determine incident
angle. The empirical equation was found to accurately predict the lens position of critical angle.
70
71
Glass
Air
φ
θ
Figure A.1: Schematic of laser beam refraction at the glass/air interface.
θ
Figure A.2: Photo and schematic of beam incident angle measurement. The refraction angle θ is
the same as that of figure A.1.
72
8.5 9 9.5 10 10.5 11
−0.2
0
0.2
0.4
0.6
0.8
x (mm)
s
i
n
(
φ
)

sin(φ) = − 0.33 x + 3.5
Data
Linear fit
Figure A.3: Plot of laser beam angle vs. converging lens position.
Appendix B
Shear-induced lift force (or more
precisely, lack of ) on near-wall
submicron-particles
It has been suggested that a shear-induced lift force can act on suspended particles in a linear
shear flow, making them more likely to move away from a solid boundary. This would result in a
non-uniform distribution of particles near our glass solid surface, where our total internal reflection
velocimetry (TIRV) measurements take place. Because our statistical analysis is based on the
assumption of uniform particle distribution, the shear-induced lift force would lead to significant
inaccuracies. However, we believe that such a lift force is insignificant in our flow regime, and in
this appendix, an established theory is presented to support our argument.
The subject of lift forces acting on a small sphere in a wall-bounded linear shear flow has been
thoroughly studied by Cherukat & McLaughlin [87]. Here we will present only the theory that
applies to the flow conditions under consideration. Suppose that a free-rotating rigid sphere of
radius a is in a Newtonian incompressible fluid of kinematic viscosity ν and is in the vicinity of a
solid wall. In the presence of a linear shear flow, the free-rotating sphere can can travel at a velocity
U
sph
that is different from the fluid velocity, U
G
, of the shear plane located at its center [38]. We
can define a characteristic Reynolds number based on the velocity difference, U
s
= U
sph
−U
G
, with
Re
s
=
U
s
a
ν
. (B.1)
Another characteristic Reynolds number based on shear rate can be defined as
Re
G
=
Ga
2
ν
, (B.2)
where G is the wall shear rate. In this geometry, the wall can be considered as located in the ”inner
region” of flow around the particle if Re
s
¸κ and Re
G
¸κ
2
, where κ ≡ a/h and h is the distance
between the particle’s center and the wall. For near-wall particle velocimetry described in chapters
3 and 4, Re
s
∼ Re
G
∼ 10
−7
while κ ∼ O(1), and thus the inner region theory of lift force applies.
For a flat wall located in the inner region of flow around a free-rotating particle, the lift force,
F
L
, which is perpendicular to the wall, is scaled by [87]
F
L
∼ Re
s
I, (B.3)
73
74
where I is a coefficient that can be numerically estimated by
I =
_
1.7631 + 0.3561κ −1.1837κ
2
+ 0.845163κ
3
¸

_
3.21439
κ
+ 2.6760 + 0.8248κ −0.4616κ
2
_ _
Re
G
Re
s
_
+
_
1.8081 + 0.879585κ −1.9009κ
2
+ 0.98149κ
3
¸
_
Re
G
Re
s
_
2
.
(B.4)
Again for the velocimetry conditions described in chapters 3 and 4, I ∼ O
_
10
2
_
. Therefore
F
L
∼ Re
s
I
_
10
−7
_ _
10
2
_
¸1, (B.5)
and the lift force acting on near-wall particles is insignificant and can be neglected for all practical
purposes.
Appendix C
Evanescent wave image of
micron-sized fluorescent particles
Imaging of fluorescent nanoparticles under TIRFM have been well reported by many groups, partic-
ularly in the exponential decay relation between particle intensity and its distance to the substrate
surface [44, 85, 90]. But for micron-sized fluorescent particles in evanescent field, there has been no
report on their imaging characteristics. In this appendix, numerical simulations were performed to
investigate some of the imaging characteristics associated with fluorescent particles under evanes-
cent wave imaging.
The numerical simulation was conducted in the commercially packaged COMSOL Multiphysics
software environment. The simulation geometry is simplified from three-dimensional to two-
dimensional but is still a representative of near-surface imaging, and shown in figure C.1. Polystyrene
particles (refractive index 1.59) of three different diameters, d/λ = 0.39, 6 and 12 where λ is the
incident beam wavelength, are immersed in the water phase (refractive index 1.33) and placed in
the vicinity of the glass phase (refractive index 1.51). A Gaussian beam enters from the lower-left
boundary and reflects off the water/glass interface at a 64.54

angle, while scattering boundary
conditions are applied to all other external boundaries. Simulations were repeated for various gap
sizes, h, between the particle and the glass, with h/λ ranging from 0 to 2. The effectiveness of the
COMSOL Multiphysics package in modelling total internal reflection and evanescent field was first
demonstrated in the absence of a polystyrene particle, and is shown in figure C.2. The simulation
model perfectly captures total internal reflection and an exponential decay of evanescent energy
density in the water phase is found to be in agreement with theories (shown in figure C.3).
Time-averaged energy densities of particles in evanescent field are shown in figures C.4, C.5 and
C.6, respectively. In figure C.4, it can be observed that the presence of a particle with diameter
d/λ = 0.39 does not distort the energy field. In contrast, larger particles such as d/λ = 6 and
12 can distort the energy field quite significantly. Their presence can be seen to create a shadow
in the reflected beam. These lost energy leaks into the interior of the large particles through
a “frustrated-total internal reflection mechanism”, illuminating regions beyond the penetration
depth of evanescent wave. These energy continue to reflect and refract at the polystyrene/water
interface, resulting in a high energy density in the lower-right side of the particle, while some
additional energy escapes into the right half of the water phase.
Using the subdomain integration function of COMSOL, one can find out the overall energies
inside the polystyrene particles. This quantity is important because in TIRFM experiments, only
the light emitted by the fluorophores inside the polystyrene particles is imageable, and the fluo-
rophores are excited only by the light energy that enters the particles. The overall energies that
75
76
Incident Gaussian Beam
Glass Phase
Water Phase
Polystyrene Particle
Figure C.1: Schematic of COMSOL simulation geometry. A polystyrene particle is suspended
in the water phase and located in the vicinity of glass’interface. Scattering (or floating) boundary
condition is applied to all external boundaries at the water and glass phases, except at the boundary
inside the glass phase where the incident Gaussian beam enters. Because the incident angle of the
Gaussian beam is 64.54

, a total internal reflection occurs at the glass/water interface. The particle
shown in the figure has a diameter of d/λ = 6.
Figure C.2: COMSOL simulation of total internal reflection. Plotted in the figure is time-averaged
total energy density of total internal reflection at a glass/water interface. The geometry of this
simulation is identical to that of figure C.1, except the presence of a suspended particle.
77
0 0.5 1 1.5 2
0
0.2
0.4
0.6
0.8
1
h/λ
N
o
r
m
a
l
i
z
e
d

I
n
t
e
n
s
i
t
y


Simulation
Theory
Figure C.3: The exponential intensity decay of evanescent field in figure C.2. Close agreement
between simulation results and theoretical calculations is obtained.
Particle
Figure C.4: COMSOL simulation of a particle with d/λ = 0.39 in evanescent field. Plotted in the
figure is time-averaged total energy density, and the geometry of this simulation is identical to that
of figure C.1, except that the particle diameter is with d/λ = 0.39 and the gap size between the
particle and the glass surface is h/λ = 0.1.
78
Figure C.5: COMSOL simulation of a particle with d/λ = 6 in evanescent field. Plotted in the
figure is time-averaged total energy density, and the geometry of this simulation is identical to that
of figure C.1, except that the gap size between the particle and the glass surface is h/λ = 0.1.
Figure C.6: COMSOL simulation of a particle with d/λ = 12 in evanescent field. Plotted in the
figure is time-averaged total energy density, and the geometry of this simulation is identical to that
of figure C.1, except that the particle diameter is d/λ = 12 and the gap size between the particle
and the glass surface is h/λ = 0.1.
79
0 0.5 1 1.5 2
0
0.2
0.4
0.6
0.8
1
h/λ
N
o
r
m
a
l
i
z
e
d

I
n
t
e
n
s
i
t
y


d/λ = 0.39
d/λ = 6
d/λ = 12
Evanescent Field
Figure C.7: Overall evanescent energy in the suspended particles at various gap sizes. The in-
tensities are normalized by the intensity observed when the particle is in contact with the glass
surface. The evanescent field intensity is obtained from the cross-sectional time-averaged total
energy density profile of figure C.2.
enter the particles at gap sizes h/λ ranging from 0 to 2 is shown in figure C.7. As it has been
demonstrated by Huang et al. [85] and Sarkar et al. [90], the overall intensity of a small particle
(d/λ = 0.39) follows an exponential decay function that is identical to the evanescent energy decay.
This forms the basis of using the overall intensity of nanoparticles for 3D displacement tracking.
The overall intensity of larger particles, however, does not follow the same exponential decay func-
tion of the evanescent field, particularly at small gap sizes. This is probably due to the fact that
a larger radius of curvature provides more surface area for light to leak into the particles through
the frustrated-total internal reflection mechanism. Still, a near-exponential dependency of large
particle intensity on the gap size is observed in figure C.7.
A more relative question is what will a fluorescent particle with diameter greater than the
illumination wavelength look like under a high numerical aperture (NA), high magnitude objective
whose focal depth, δ, is roughly δ/λ = 2. One of the ways to find out is to convolute the energy
density inside the particles with diffraction equation and point spread function. However, such a
convolution is computationally complicated and expensive. A simpler way is to look at the intensity
profile when the fluorescent energy is projected onto the focal plane, with which one could still gain
some qualitative understanding of the image. For a point light source that is a distance z off the
focal plane, it’s intensity is [101]
I (z) =
_
sin (u/4)
u/4
_
2
I
0
(C.1)
where I
0
is the point source emission intensity and
u =

λ
_
a
f
_
2
z. (C.2)
Here λ is the wavelength, a is the radius of the objective lens and f is the objective focal length. In
80
−4 −2 0 2 4
0
0.2
0.4
0.6
0.8
1
I
/
I
0
x/λ
Figure C.8: Cross-sectional intensity profile of a 3-µm particle (d/λ = 6). The blue squares
represent the cross-section intensity profile indicated as a red line in figure C.9. The red dotted
line, however, represents the projected intensity profile based on results of COMSOL simulation of
a particle of the same size and equation (C.3). I
0
is the peak intensity and x/λ = 0 corresponds to
the particle geometric center. The direction of the evanescent wave propagation is toward positive
x/λ.
getting equation (C.1), it is assumed that both a and f are much larger than λ. a/f is commonly
referred to as the f-number of lens and for a microscope objective, it is equal to the numerical
aperture [73]. Therefore, for a particle in contact of the wall and its center is located at x = 0 and
z = a, its projected intensity at point x = x
0
on the z = z
0
plane is [101]
I
p
(x
o
) =
_
2a−b
b
I (x
0
, z) zdz (C.3)
where b = a−
_
a
2
−x
2
0
and I (x
0
, z) is calculated using equation (C.1). This projection is obviously
a simplified calculation as it assumes negligible diffraction. This leads to physical inaccuracy as we
will demonstrate subsequently.
The intensity projection of a particle with d/λ = 6 is shown in figure C.8 as a red dotted line,
where the focal plane is set at the glass surface. In the projection, the intensity exhibits a peak value
near the particle center, while a longer tail is observed in the direction of the laser beam propagation.
A TIRFM image of a 3-µm particle (equivalently d/λ = 6) is shown in figure C.9. A peak intensity
near the center can be clearly observed in the image. For comparison, a cross-sectional intensity
profile of figure C.9 is also shown in figure C.8 as blue squares. This intensity profile is similar in
shape when compared to that of red dotted line from simulation. A peak intensity is found near the
particle center, while a slightly longer tail is found in the direction of evanescent wave propagation.
However, the intensity profile from actual imaging of a particle appears to be wider than that of the
simulated result. This is due to the fact that the computation leading to the projection (red dotted
line) in figure C.8 neglects diffraction of light, which is important in microscopy. Nevertheless, the
81
1 µm
Figure C.9: Digital image of a 3-µm particle (d/λ = 6) in contact with glass surface. The intensity
profile along the red cross-section shown is plotted in figure C.8.
resemblance of of the two profiles is striking and the profiles are in qualitative agreement.
The same imaging projection is also performed for a 6-µm particle (d/λ = 12) and shown in
figure C.10. With a particle of this size, we can again see that there is a bright peak near its center,
while the asymmetric tail on the right side is even more pronounced. A similar observation can be
made on an actual image of a 6-µm particle (figure C.11) as well as on a cross-sectional intensity
profile (figure C.10, blue squares). A large bright center is clearly visible while the right half of
the particle is significantly brighter. In addition, toward the right tip, there are small, alternating
regions of bright spots, which can be also be observed in the right tail of the red simulation intensity
profile in figure C.10. Again, the modelling results and the actual image show qualitative agreement
despite the simplicity of the integration model in equations (C.1) and (C.3). However, as observed in
figure C.8, the intensity profile obtained from the simulation results and equation (C.3) is narrower
than that of an actual image because of the neglected diffraction in numerical integration.
82
−6 −4 −2 0 2 4 6
0
0.2
0.4
0.6
0.8
1
x/λ
I
/
I
0
Figure C.10: Cross-sectional intensity profile of a 6-µm particle (d/λ = 12). The blue squares
represent the cross-section intensity profile indicated as a red line in figure C.11. The red line,
however, represents the projected intensity profile based on results of COMSOL simulation of a
particle of the same size and equation (C.3). I
0
is the peak intensity and x/λ = 0 corresponds to
the particle geometric center. The direction of the evanescent wave propagation is toward positive
x/λ.
2 µm
Figure C.11: Digital image of a 6-µm particle (d/λ = 12) in contact with glass surface. The
intensity profile along the red cross-section shown is plotted in figure C.10.
Bibliography
[1] Adam Bange, H. Brian Halsall, and William R. Heineman. Microfluidic immunosensor sys-
tems. Biosensors and Bioelectronics, 20:2488–2503, 2005.
[2] Samuel K. Sia and George M. Whitesides. Microfluidic devices fabricated in
poly(dimethylsiloxane) for biological studies. Electrophoresis, 24:3563–3576, 2003.
[3] Jan Kruger, Kirat Singh, Alan O’neill, Carl Jackson, Alan Morrison, and Peter O’Brien.
Development of a microfluidic device for fluorescence activated cell sorting. Journal of Mi-
cromechanics and Microengineering, 12:486–494, 2002.
[4] W. Mark Saltzman and William L. Olbricht. Building drug delivery into tissue engineering.
Nature Reviews, 1:177–186, 2002.
[5] David J. Beebe, Glennys A. Mensing, and Glenn M. Walker. Physics and applications of
microfluidics in biology. Annual Review of Biomedical Engineering, 4:261–286, 2002.
[6] George E. Karniadakis and Ali Beskok. Micro flows: fundamentals and simulation. Springer,
2002.
[7] Eric Lauga, Michael P. Brenner, and Howard A. Stone. Microfluidics: The no-slip boundary
condition. In J. Foss anc C. Tropea and A. Yarin, editors, Handbook of Experimental Fluid
Dynamics, chapter 15. Springer, New York, 2006.
[8] Bin Zhao, Jeffrey S. Moore, and David J. Beebe. Surface-directed liquid flow inside mi-
crochannels. Science, 291:1023–1026, 2001.
[9] Bin Zhao, Jeffrey S. Moore, and David J. Beebe. Principles of surface-directed liquid flow in
microfluidic channels. Analytical Chemistry, 74:4259–4268, 2002.
[10] Zbigniew Adamczyk, Katarzyna Jaszczolt, Barbara Siwek, and Pawel Weronski. Irreversible
adsorption of particles at random-site surfaces. Journal of Chemical Physics, 120:11155–
11162, 2004.
[11] Kai-Chien Chang and Daniel A. Hammer. Influence of direction and type of applied force on
the detachment of macromolecularly-bound particles from surfaces. Langmuir, 12:2271–2282,
1996.
[12] M. Chaoui and F. Feuillebois. Creeping flow around a sphere in a shear flow close to a wall.
Quarterly Journal of Mechanics and Applied Mathematics, 56:381–410, 2003.
83
84
[13] Poppo J. Wit, Albert Poortinga, Jaap Noordmans, Henry C. van der Mei, and Henk J.
Busscher. Deposition of polystyrene particles in a parallel plate flow chamber under attractive
and repulsive electrostatic conditions. Langmuir, 15:2620–2626, 1999.
[14] P. J. A. Hartman Kok, S. G. Kazarian, B. J. Briscoe, and C. J. Lawrence. Effects of particle
size on near-wall depletion in mono-dispersed colloidal suspensions. Journal of Colloid and
Interface Science, 280:511–517, 2004.
[15] Remco Tuinier and Takashi Taniguchi. Polymer depletion-induced slip near an interface.
Journal of Physics: Condensed Matter, 17:L9–L14, 2005.
[16] Yoram Cohen and A.B. Metzner. Apparent slip flow of polymer solutions. Journal of Rheology,
29:67–102, 1985.
[17] Howard A. Barnes. A review of the slip (wall depletion) of polymer solutions, emulsions and
particle suspensions in viscometers: its cause character and cure. Journal of Non-Newtonian
Fluid Mechanics, 56:221–251, 1995.
[18] Sydney Goldstein. Fluid mechanics in the first half of this century. Annual Review of Fluid
Mechanics, 1:1–29, 1969.
[19] K. Johan A. Westin, Kenneth S. Breuer, Chang-Hwan Choi, Peter Huang, Zhiqiang Cao,
Bruce Caswell, Peter D. Richardson, and Merwin Sibulkin. Liquid transport properties in sub-
micron channel flows. In Proceedings of 2001 ASME International Mechanical Engineering
Congress and Exposition, 2001.
[20] Erhard Schnell. Slippage of water over nonwettable surfaces. Journal of Applied Physics,
27:1149–1152, 1956.
[21] N. V. Churaev, V. D. Sobolev, and A. N. Somov. Slippage of liquids over lyophobic solid
surfaces. Journal of Colloid and Interface Science, 97:574–581, 1984.
[22] Peter A. Thompson and Sandra M. Troian. A general boundary condition for liquid flow at
solid surfaces. Nature, 389:360–362, 1997.
[23] Jean-Louis Barrat and Lyderic Bocquet. Large slip effect at a nonwetting fluid-solid interface.
Physical Review Letters, 82:4671–4674, 1999.
[24] Marek Cieplak, Joel Koplik, and Jayanth R. Banavar. Boundary conditions at a fluid-solid
surface. Physical Review Letters, 86:803–806, 2001.
[25] Chang-Hwan Choi, Johan A. Westin, and Kenneth S. Breuer. Apparent slip flows in hy-
drophilic and hydrophobic microchannels. Physics of Fluids, 15:2897–2902, 2003.
[26] YingXi Zhu and Steve Granick. Limites of the hydrodynamic no-slip boundary condition.
Physical Review Letters, 88:106102, 2002.
[27] C. Neto, V. S. J. Craig, and D. R. M. Williams. Evidence of shear-dependent boundary slip
in newtonian liquids. The European Physical Journal E, 12:S71–S74, 2003.
[28] C. Cottin-Bizonne, B. Cross, A. Steinberger, and E. Charlaiz. Boundary slip on smooth hy-
drophobic surfaces: intrinsic effects and possible artifacts. Physical Review Letters, 94:056102,
2005.
85
[29] R. Pit, H. Hervet, and L. Leger. Direct experimental evidence of slip in hexadecane: solid
interface. Physical Review Letters, 85:980–983, 2000.
[30] Derek C. Tretheway and Carl D. Meinhart. Apparent fluid slip at hydrophobic microchannel
walls. Physics of Fluids, 14:L9–L12, 2002.
[31] Pierre Joseph and Patrick Tabeling. Direct measurement of the apparent slip length. Physical
Review E, 71:035303(R), 2005.
[32] Keizo Watanabe, Yanuar, and Hiroshi Mizunuma. Slip of newtonian fluids at solid boundary.
JSME International Journal, 41:525–529, 1998.
[33] D. Lumma, A. Best, A. Gansen, F. Feuillebois, J. O. Radler, and O. I. Vinogradova. Flow
profile near a wall measured by double-focus fluorescence cross-correlation. Physical Review
E, 67:056313, 2003.
[34] Steve Granick, Yingxi Zhu, and Hyunjung Lee. Slippery questions about complex fluids
flowing past solids. Nature Materials, 2:221–227, 2003.
[35] Eric Lauga. Apparent slip due to the motion of suspended particles in flows of electrolyte
solutions. Langmuir, 20:8924–8930, 2004.
[36] Howard Brenner. The slow motion of a sphere through a viscous fluid towards a plane wall.
Chemical Engineering Science, 16:242–251, 1961.
[37] A. J. Goldman, R. G. Cox, and H. Brenner. Slow viscous motion of a sphere parallel to a
plane wall - i: motion through a quiescent fluid. Chemical Engineering Science, 22:637–651,
1967.
[38] A. J. Goldman, R. G. Cox, and H. Brenner. Slow viscous motion of a sphere parallel to a
plane wall - ii: Couette flow. Chemical Engineering Science, 22:653–660, 1967.
[39] Michael A. Bevan and Dennis C. Prieve. Hindered diffusion of colloidal particles very near
to a wall: revisited. Journal of Chemical Physics, 113:1228–1236, 2000.
[40] M. Hosoda, K. Sakai, and K. Takagi. Measurement of anisotropic brownian motion near an
interface by evanescent light-scattering spectroscopy. Physical Review E, 58:6275–6280, 1998.
[41] Binhua Lin, Jonathan Yu, and Stuart A. Rice. Direct measurements of constrained brownian
motion of an isolated sphere between two walls. Physical Review E, 62:3909–3919, 2000.
[42] Ratna J. Oetama and John Y. Walz. Simultaneous investigation of sedimentation and
diffusion of a single colloidal particle near an interfce. The Journal of Chemical Physics,
124:164713, 2006.
[43] Arindam Banerjee and Kenneth D. Kihm. Experimental verification of near-wall hindered
diffusion for the brownian motion of nanoparticles using evanescent wave microscopy. Physical
Review E, 72:042101, 2005.
[44] K. D. Kihm, A. Banerjee, C. K. Choi, and T. Takagi. Near-wall hindered brownian diffusion of
nanoparticles examined by three-dimensional ratiometric total internal reflection fluorescence
microscopy (3-d r-tirfm). Experiments in Fluids, 37:811–824, 2004.
86
[45] Daniel Axelrod. Total internal reflection fluorescence microscopy. In Methods in Cell Biology,
volume 30, chapter 9, pages 245–270. Academic Press, Inc., 1989.
[46] Chris Rowe Tiatt, George P. Anderson, and Frances S. Ligler. Evanescent wave fluorescence
biosensors. Biosensors and Bioelectronics, 20:2470–2487, 2005.
[47] Daniel Axelrod. Total internal reflection fluorescence microscopy in cell biology. Traffic,
2:764–774, 2001.
[48] Eric A. J. Reits and Jacques J. Neefjes. From fixed to frap: measuring protein mobility and
activity in living cells. Nature Cell Biology, 3:E145–E147, 2001.
[49] Taekjip Ha. Single-molecule fluorescence resonance energy transfer. Methods, 25:78–86, 2001.
[50] H. H. von Grunberg, L. Helden, P. Leiderer, and C. Bechinger. Measurement of surface
charge densities on brownian particles using total internal reflection microscopy. Journal of
Chemical Physics, 114:10094–10104, 2001.
[51] Scott G. Flicker, Jennifer L. Tipa, and Stacy G. Bike. Quantifying double-layer repulsion
between a colloidal sphere and a glass plate using total internal reflection microscopy. Journal
of Colloid and Interface Science, 158:317–325, 1993.
[52] Robert Kun and Janos H. Fendler. Use of attenuated total internal reflection-fourier transform
infrared spectroscopy to investigate the adsorption of and interactions between charged latex
particles. Journal of Physical Chemistry, 108:3462–3468, 2004.
[53] P. Buchhave. Particle image velocimetry. In Lars Lading, Graham Wigley, and Preben
Buchhave, editors, Optical diagnostics for flow processes, pages 247–270. Plenum Press, New
York, 1994.
[54] S. T. Wereley and C. D. Meinhart. Micron-resolution particle image velocimetry. In K. Breuer,
editor, Microscale Diagnostic Techniques, pages 51–112. Springer, 2005.
[55] J. G. Santiago, S. T. Wereley, C. D. Meinhart, D. J. Beebe, and R. J. Adrian. A particle
image velocimetry system for microfluidics. Experiments in Fluids, 25:316–319, 1998.
[56] C. M. Zettner and M. Yoda. Particle velocity field measurements in a near-wall flow using
evanescent wave illumination. Experiments in Fluids, 34:115–121, 2003.
[57] Haifeng Li, Reza Sadr, and Minami Yoda. Multilayer nano-particle image velocimetry. Ex-
periments in Fluids, 41:185–194, 2006.
[58] Donald L. Ermak and J.A. McCammon. Brownian dynamics with hydrodynamic interactions.
Journal of Chemical Physics, 69:1352–1360, 1978.
[59] Reza Sadr, Christel Hohenegger, Haifeng Li, Peter J. Mucha, and Minami Yoda. Diffusion-
induced bias in near-wall velocimetry. Submitted to Journal of Fluid Mechanics, 2006.
[60] Anne Pierres, Anne-Marie Benoliel, Cheng Zhu, and Pierre Bongrand. Diffusion of micro-
spheres in shear flow near a wall: use to measure binding rates between attached molecules.
Biophysical Journal, 81:25–42, 2001.
87
[61] H. N. Unni and C. Yang. Brownian dynamics simulation and experimental study of colloidal
particle deposition in a microchannel flow. Journal of Colloid and Interface Science, 291:28–
36, 2005.
[62] David S. Sholl, Michael K. Fenwick, Edward Atman, and Dennis C. Prieve. Brownian dy-
namics simulation of the motion of a rigid sphere in a viscous fluid very near a wall. Journal
of Chemical Physics, 113:9268–9278, 2000.
[63] Reza Sadr, Haifeng Li, and Minami Yoda. Impact of hindered brownian diffusion on the
accuracy of particle-image velocimetry using evanescent-wave illumination. Experiments in
Fluids, 38:90–98, 2005.
[64] YingXi Zhu and Steve Granick. Rate-dependent slip of newtonian liquid at smooth surfaces.
Physical Review Letters, 87:096105, 2001.
[65] N. L. Thompson and B. C. Langerholm. Total internal reflection fluorescence: applications
in cellular biophysics. Current Opinion in Biotechnology, 8:58–64, 1997.
[66] Jeffrey S. Burmeister, Lauri A. Olivier, W. M. Reichert, and George A. Truskey. Application
of total internal reflection fluorescence microscopy to study cell adhesion to biomaterials.
Biomaterials, 19:307–325, 1998.
[67] Derek Toomre and Dietmar J. Manstein. Lighting up the cell surface with evanescent wave
microscopy. Trends in Cell Biology, 11:298–303, 2001.
[68] J. Yamada. Evanescent wave doppler velocimetry for a wall’s near field. Applied Physics
Letters, 75:1805–1806, 1999.
[69] David C. Duffy, J. Cooper McDonald, Olivier J. A. Schueller, and George M. Whitesides.
Rapid prototyping of microfluidic systems in poly(dimethylsiloxane). Analytical Chemistry,
70:4974–4984, 1998.
[70] Carl D. Meinhart and Steve T. Wereley. The theory of diffraction-limited resolution in
microparticles image velocimetry. Measurement Science and Technology, 14:1047–1053, 2003.
[71] B. Ovryn. Three-dimensional forward scattering particle image velocimetry applied to a
microscopic field-of-view. Experiments in Fluids, 29:S175–S184, 2000.
[72] R. J. Adrian and C. S. Yao. Pulsed laser technique application to liquid and gaseous flows
and the scattering power of seed materials. Applied Optics, 24:44–52, 1985.
[73] M. G. Olsen and R. J. Adrian. Out-of-focus effects on particle image visibility and correlation
in microscopic particle image velocimetry. Experiments in Fluids, 29:S166–S174, 2000.
[74] C. J. Bourdon, M. G. Olsen, and A. D. Gorby. Validation of an analytic solution for depth of
correlation in microscopic particle image velocimetry. Measurement Science and Technology,
15:318–327, 2004.
[75] J. C. Crocker and D. G. Grier. When like charges attract: The effects of geometrical confine-
ment on long-range colloidal interactions. Physical Review Letters, 77:1897–1900, 1996.
[76] S. H. Behrens and D. G. Grier. The pair interaction of charged colloidal spheres near a
charged wall. Physical Review E, 64:050401, 2001.
88
[77] T. M. Galea and Phil Attard. Molecular dynamics study of the effect of atomic roughness on
the slip length at the fluid-solid boundary during shear flow. Langmuir, 20:3477–3482, 2004.
[78] Gyoko Nagayama and Ping Cheng. Effects of interface wettability on microscale flow by
molecular dynamics simulation. International Journal of Heat and Mass Transfer, 47:501–
513, 2004.
[79] Cecile Cottin-Bizonne, Jean-Louis Barrat, Lyderic Bocquet, and Elisabeth Charlaiz. Low-
friction flows of liquid at nanopattened interfaces. Nature Materials, 2:237–240, 2005.
[80] S. Jin, P. Huang, J. Park, J. Y. Yoo, and K. S. Breuer. Near-surface velocimetry using
evanescent wave illumination. Experiments in Fluids, 37:825–833, 2004.
[81] Chang-Hwan Choi, K. Johan A. Westin, and Kenneth S. Breuer. To slip or not to slip - water
flows in hydrophilic and hydrophobic microchannels. In Proceedings of IMECE 2002, pages
IMECE2002–33707, 2002.
[82] Eric Lauga and Michael P. Brenner. Dynamic mechanisms for apparent slip on hydrophobic
surfaces. Physical Review E, 70:026311, 2004.
[83] Xue H. Zhang, Xiao D. Zhang, Shi T. Lou, Zhi X. Zhang, Jie. L. Sun, and Jun Hu. De-
gassing and temperature effects on the formation of nanobubbles at the mica/water interface.
Langmuir, 20:3813–3815, 2004.
[84] Michael K. Cheezum, William F. Walker, and William H. Guilford. Quantitative comparison
of algorithms for tracking single fluorescent particles. Biophysical Journal, 81:2378–2388,
2001.
[85] Peter Huang, Jeffrey S. Guasto, and Kenneth S. Breuer. Direct measurement of slip velocities
using three-dimensional total internal reflection velocimetry. Journal of Fluid Mechanics,
566:447–464, 2006.
[86] Michael R. King and David T. Leighton Jr. Measurement of the inertial lift on a moving
sphere in contact with a plane wall in a shear flow. Physics of Fluids, 9:1248–1255, 1997.
[87] Pradeep Cherukat and John B. McLaughlin. The inertial lift on a rigid sphere in a linear
shear flow field near a flat wall. Journal of Fluid Mechanics, 263:1–18, 1994.
[88] Nasser A. Frej and Dennis C. Prieve. Hindered diffusion of a single sphere very near a wall
in a nonuniform force field. Journal of Chemical Physics, 98:7552–7564, 1993.
[89] Peter Huang and Kenneth S. Breuer. Simulations of hindered diffusion in shear flow and its
implications for near-wall velocimetry. In preparation, 2006.
[90] Atom Sarkar, Ragan B. Robertson, and Julio M. Fernandez. Simultaneous atomic force
microscope and fluorescence measurements of protein unfolding using a calibrated evanescent
wave. Proceedings of the National Academy of Sciences, 101:12882–12886, 2004.
[91] Richard A. L. Jones. Soft Condensed Matter. Oxford University Press, Oxford, England,
2002.
89
[92] Lakkapragada Suresh and John Y. Walz. Direct measurement of the effect of surface roughness
on the colloidal forces between a particle and flat plate. Journal of Colloid and Interface
Science, 196:177–190, 1997.
[93] Olga L. Vinogradova. Slippage of water over hydrophobic surfaces. Internal Journal of
Mineral Processing, 56:31–60, 1999.
[94] Jeffrey S. Guasto, Peter Huang, and Kenneth S. Breuer. Statistical particle tracking ve-
locimetry using molecular and quantum dot tracer particles. in press, 2006.
[95] Peter Huang and Kenneth S. Breuer. Direct measurement of anisotropic near-wall hindered
diffusion using total internal reflection velocimetry. In preparation, 2006.
[96] Christel Hohenegger and Peter J. Mucha. Statistical reconstruction of velocity profiles for
nano particle image velocimetry. To be published, 2006.
[97] Eric Lauga and Todd M. Squires. Brownian motion near a partial-slip boundary: A local
probe of the no-slip condition. Physics of Fluids, 17:103102, 2005.
[98] E. A. J. F. Peters and Th. M. A. O. M. Barenbrug. Efficient brownian dynamics simulation
of particles near wall. i. reflecting and absorbing walls. Physical Review E, 66:056701, 2002.
[99] E. A. J. F. Peters and Th. M. A. O. M. Barenbrug. Efficient brownian dynamics simulation
of particles near wall. i. sticky walls. Physical Review E, 66:056702, 2002.
[100] Shahram Pouya, Manoochehr Koochesfahani, Preston Snee, Moungi Bawendi, and Daniel
Nocera. Single quantum dot (qd) imaging of fluid flow near surfaces. Experiments in Fluids,
39:784–786, 2005.
[101] Max Born and Emil Wolf. Principles of Optics: Electromagnetic theory of propagation inter-
ference and diffraction of light. Cambridge University Press, 6th edition, 1980.

This report is the PhD thesis of Peter Huang, supervised by Prof. Kenneth Breuer in the Division of Engineering, and submitted to the Graduate School at Brown University in September 2006. For more information contact: kbreuer@brown.edu

i

Abstract Aqueous boundary solution slip at a solid surface and hindered diffusion of suspended nearsurface particles are experimentally investigated with a three-dimensional total internal reflection velocimetry (3D-TIRV) technique. The sub-micrometer penetration depth of evanescent wave offers low-background-noise images where a 10-nm displacement can be accurately measured. At shear rates less than 1800 sec−1 , a shear-induced slip length of less than 100 nanometers is observed at a glass surface of sub-nanometer roughness. Surface hydrophobicity is also found to aid slip. However, the notion that electrokinetic and electrostatic forces acting on charged tracer particles lead to slip length measurement inaccuracy is disputed by experiments conducted with aqueous solutions consisting of electrolytes, which report slip length of similar magnitudes. Anisotropic hindered diffusion of suspended near-surface particles is demonstrated with simultaneous threedimensional measurements. Hindered diffusion coefficients of 1.5-µm radius particles within 300 nm from the solid wall are shown to be in close agreement with previously published theories. Lastly, effects of hindered diffusion and wall presence on the accuracies of TIRV and other nearsurface measurement methods are studied via Brownian dynamics simulations. The results reveal that hindered diffusion causes a minor bias toward values obtained at the wall if a small interval is allowed between successive image acquisitions, while particle drop-outs lead to over-estimation of fluidic velocity values if the time interval between image acquisitions is large. It is also found that the apparent velocities of near-wall particles are not time-invariant. Still, true fluidic velocity values can be obtained from the particle apparent velocities with proper scaling. Such scaling correction is applicable to particle tracking velocimetry measurements, but not to correlation based velocimetry measurements for their lack of accounting of particle drop-outs. Shear effect on near-wall particle translation in a finite imaging depth results in an asymmetric apparent velocity distribution, which is another potential source of error for correlation based velocimetries. These findings lead to a conclusion that particle tracking based velocimetry methods, such as TIRV, are more superior than correlation based velocimetry methods are in making near-surface fluidic measurements.

Copyright. c . 2006 i . Pong-Yu (Peter) Huang.

He currently resides in the New England region with his wife and plans to pursue an academic career. Taiwan and attended schools there until the age of fifteen. and continued to pursue a Doctor of Philosophy degree since 2003. After graduating high school from International School Manila in 1996. ii . he attended Cornell University where he received a Bachelor of Arts degree in physics in 2000. Huang enjoys travelling and anything related to baseball (except sabermetrics). he completed a Master of Science degree in engineering. In 2002. Upon graduation. and holds great interests in economic perspectives of events in the world. He is a proud fan of the New York Mets since 1999.Pong-Yu (Peter) Huang was born in Taipei. he enrolled at the Division of Engineering. Brown University. Mr. to pursue a graduate degree under the guidance of Professor Kenneth Breuer.

pp. 37. 2006. To be submitted to Physical Review E. Guasto and Kenneth S. Included are an overall review and summary of the whole dissertation in the first chapter. Breuer. 2004. To be submitted to Physics of Fluids. Chapter 7. Vol. Chapter 5. 2006. reviews of previously published works. Concluding remarks and suggested studies. 2006. with details of experimental methods and results. Yoo and Kenneth S. “Near-surface velocimetry using evanescent wave illumination” by Songwan Jin.This dissertation consists of experimental studies of boundary slip and hindered diffusion in the nano-scale. Chapter 2. Experiments in Fluids. hindered diffusion and total internal reflection velocimetry. The contents of the chapters are: Chapter 1. J Y. Journal of Fluid Mechanics. 566. Including basic concepts of slip flow. Jeffrey S. Breuer. and discussions of the main results obtained. Vol. Paper 5. Paper 4. Chapter 6. Breuer. Breuer. 2006. and a collection of journal papers. 447-464. To be submitted to Physics of Fluids. “Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry” by Peter Huang. Introduction. each as a subsequent chapter. Peter Huang. pp. iii . Paper 2. “Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection velocimetry” by Peter Huang and Kenneth S. Paper 1. “Direct measurement of slip length in electrolyte solutions” by Peter Huang and Kenneth S. 825-833. Paper 3. Chapter 3. Breuer. Jinil Park. “Simulations of hindered diffusion in shear flow and its implications for near-wall velocimetry” by Peter Huang and Kenneth S. Chapter 4.

Numerous discussions with Professor Tripathi and Professor Thomas Powers during the years on chemistry and statistical mechanics are acknowledged and very much appreciated. Brian Burke. patience and encouragement throughout the years of graduate study. iv . My greatest gratitude goes to my wife. Without his creative ideas and brilliant suggestions many subtleties of experimental methods and results would not have been found. for his introduction of the total internal reflection fluorescent microscopy setup. for their generous help in micro-fabrication. the administrative and research supports of Mr. Professor Kenneth Breuer. Mr. Professor Eric Lauga and Professor Anubhav Tripathi. Teng-Fang Kuo. I also would like to thank my dissertation readers. Professor Sylvain Cloutier. my sister and my extended family for their love. Without his encouragement. Charles Vicker and Ms. for their insightful comments during dissertation revisions. Mr. As a lowly ranked graduate student. Jeffrey Brown. Michael Jibitsky.I would like to first thank my mentor. the projects described in this dissertation would not have been fruitful. Without their kind assistance many research projects would not have moved forward as smoothly as they did. Mr. Jahn Torres. support and best wishes over the years. Brian Corkum. for his guidance and support through my graduate study. Songwan Jin. and for putting up with me during the ups and downs of experiments. for her love. In addition. optical system design and setup. Finally I would like to thank my parents. Jinkee Lee and Matt Kerby. I also would like to thank Dr. Virginia Novak are especially appreciated. A special thank goes to my fellow graduate student and my friend Jeffrey Guasto. constructive criticisms and inexhaustible knowledge. I am very grateful to Professor MinJun Kim. nano-scale measurements and sharing of lab equipments.

. . . . .2 THEORETICAL CONSIDERATIONS . . . . . .4 CONCLUDING REMARKS . .3. . . . . . . .Contents 1 Introduction 1. . . . . . . . .3 RESULTS AND DISCUSSION . . . . . . . . . . . 2. . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . .1 INTRODUCTION . . . . . . . . . .1 INTRODUCTION . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . 1 1 3 4 6 8 8 9 12 14 15 18 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Velocimetry Analysis . . . 2. . . . . . . . . . . . . . . . . . . . . . . 3. 1. . . . . . . . . . . . . . . . . . . . . .1 INTRODUCTION . . . . . . . . . . 3. . . . . . . . .2 HINDERED DIFFUSION . . . . . .4 Measurements of Apparent Slip Velocities . . . . . . . . . . . . . . . . . . . . .3 EXPERIMENTAL PROCEDURES . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . 2 Paper 1: Near-surface velocimetry using evanescent wave illumination 2. . . . . . . . . . . . . . . . . . . . .1 Materials and Setup . . . . . . . . . . . . . . . . . . . . . .3 Comparisons with Numerical Simulations . . . . .2 Streamwise Velocity Analysis . . . . . .3 Near-Surface Shear Flow . . 4 Paper 3: Direct measurement of slip length in electrolyte solutions total . . . . . . . . . . . . . . . . . . . . . 3. . . . . .2 Emission Intensity of Fluorescent Particles . . . . . . . . . . . . .2. . . 2. . . . . . . . . . . . . .4 RESULTS AND DISCUSSION . . . . . . 3. . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . 2. . .1 Validation of Intensity Calibration Curve . . . . . . . . . . . . . . 3 Paper 2: Direct measurement of slip velocities using three-dimensional internal reflection velocimetry 3. 3. . . . . .4.1 BOUNDARY SLIP . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. .3.4 SUBTLETIES OF NEAR-WALL SHEAR FLOW . . . . . . . . . . . . . . 2. 22 22 23 23 24 26 28 28 30 31 31 31 32 34 37 39 5 Paper 4: Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection velocimetry 43 5. . .1 Cross-stream Velocity Analysis . .4. . . . . . . . . . 3. . . . . . . . .1 Total Internal Reflection Microscopy . .3 Velocity Distributions of Particles . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . 3. . . .2 EXPERIMENTAL PROCEDURES . 3.2. . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 v . . . .5 CONCLUDING REMARKS . . . . . .2 Hindered Diffusion . . . . . . . . . . .3. . . . . . . . .3 TOTAL INTERNAL REFLECTION VELOCIMETRY 1. . . . . . . . . . . . . . .

. . . . .5 THEORY OF HINDERED DIFFUSION EXPERIMENTAL PROCEDURES . 6. . . . . . . .3.3. . . . . . . . . . . . . . . . . . . 6. . . . . . . . 6. . . . . . . .2 Effects of Shear on Particle Velocities . . . . .2. . . . . 6. . .4 5. . . . . .3 5. . .3. . . . . . . .1 INTRODUCTION . . . . . . . .2 THEORIES AND COMPUTATIONS . . . . . . . . 6. . . . lack of ) on near-wall submicronparticles 73 C Evanescent wave image of micron-sized fluorescent particles 75 vi . . . . . . . . . . .4 Implementation of Simulation . . . . . . . . .2 5. . . . . . . . . . . . . . . . . . . . . 6. . . . .3 RESULTS AND DISCUSSIONS .4 CONCLUDING REMARKS . . SUMMARY . . . 7 Concluding Remarks and Suggested Studies A Calibration of beam incident angle and its implications for . . . . . . . . . .2 Particle Displacement Due to Hindered Diffusion . .2. .4 Horizontal Apparent Velocity Distributions . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . RESULTS AND DISCUSSION . . . . . . . . . . . . . . .1 Sedimentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . . . . 6. . . . . . . . . . . . . . 44 44 46 49 6 Paper 5: Simulations of hindered diffusion in shear flow near-wall velocimetry 6. . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . 50 50 51 51 53 54 54 55 55 55 58 60 62 67 68 70 B Shear-induced lift force (or more precisely. .1 The Langevin Equation . . . . . . . . . . . . . . 6. . . . . . . . . . .3. . . . . . . . . .3. . . . . . 6. .3 Particle Drop-in’s and Drop-out’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Hindered Diffusion . . . . . . . . . . .5 Time Evolution of Apparent Velocity Distributions 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . 6. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Distribution of particle velocity vectors of 200-nm particles . . . . . . . . . . . . .12 3. . . . . . . . . . . . . . . . . . . .5 < I e /I0 < 1 in a uniform concentration field . Image acquisition system . . . . Experimental apparent velocity of particles in a shear flow over a hydrophilic surface Experimental apparent velocity of particles in a shear flow over a hydrophobic surface The additional slip length due to surface hydrophobicity . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . Schematic of the experimental setup . . . . . . . . . . . . . .10 Streamwise velocity distribution of 200-nm particles determined from Monte Carlo simulation . . . . . . . .4 2. 2. . . . . . . . . . . . 42 Schematic of total internal reflection velocimetry . . . . . . . . . . Observed fluorescent particle intensity distribution and its probability density function (PDF) . . . .3 3. 45 vii . . . . . . . . .10 3. . . . . . . . . . . . . . 40 Measured slip lengths of aqueous solutions . . . . . . . . . . 2. . . . . . . 3. . . . . . . shear rate . . . . . . . .8 3. . . . . .1 3. . . . . .1 4. . . . . . . . . . . . . The ratio of statistical apparent velocity of particles and mean fluid velocity under no slip . . . . . . . . . .9 3. . . . Particle images of different illumination methods: widefield (direct. . . .5 3. . . . . . . . . . . . .7 Cross-stream Brownian motion of particles in an infinite medium and near wall . . . . . . . . . . . . . . . . . . . . .11 3. .11 Streamwise apparent velocity predicted by simulation . . . . . . . . . . . . . . . . . . . . . Fluorescent particle intensity as a function of its distance to the glass surface . . e The predicted position distribution of particles with 0. . . . . . Ratio of hindered diffusion coefficients (Dexp ) for Brownian motion parallel to a surface Distribution of observed particle streamwise velocities under various shear rates . .1 Objective-based total internal reflection fluorescence microscopy . . Objective-based total internal reflection velocimetry (TIRV) system . . . . . . . . . . .1 2.9 Apparent particle velocity vs. . . . . 2. . . . .13 4. . . . . . . . . . . . . . . flood) illumination Particle images of different illumination methods: near-wall TIRFM image . . . . 2. Distribution of observed particle streamwise velocities after scaling . . . . . . . . . . . . . . . .1 2. . . . . . . . . . . . . . .7 3. . . . . . . . . .8 Experimentally-measured distribution of streamwise velocities of 200-nm diameter particles . . . . . . . . . . . .6 3. . . . . . . . . . . . . . . . . . . . . . The emission intensity distribution of particles in a uniform concentration field .2 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 10 11 12 13 13 14 15 16 17 19 20 24 25 26 27 29 30 31 32 33 34 35 36 36 Schematic of objective-based TIRFM . . . . . . . . . 2. . . . . . .4 3. . . . . . . .2 5. . . . . . Schematic of near-wall particles moving near the surface illustrating the observation range . . .3 2. . . . . . . . . . . . . .2 2. . . . . . .5 Schematic of total internal reflection fluorescence microscopy .List of Figures 1.

3 Plot of laser beam angle vs. . . . . . . .4 COMSOL simulation of a particle with d/λ = 0. . . . . . . . C. . . . .3 6. . Time evolution of apparent velocity distribution . . .15 Intensity calibration of 1. . Collapsed apparent velocity distribution of particles at various imaging range . . .10 Cross-sectional intensity profile of a 6-µm particle (d/λ = 12) . . . . .2 6. . . . . . Particle displacement distribution due to hindered diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . 47 A schematic of the simulation geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12 6. Spatial distribution of particles from a finite layer at various times . . . . . . . . . . . .1 Schematic of laser beam refraction at the glass/air interface . . . 71 A. . . . . . . . . . . . . Sedimentation of small particles . . 72 C.9 Digital image of a 3-µm particle (d/λ = 6) in contact with glass surface C. . . . . .2 5. . . . . . . . C.3 Intensity decay of evanescent field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39 in evanescent field . . . . . . . . .11 6. . . . . . . . . . . C. . . . . . . . . . C. . . . .6 6. . . . . . . . .5 COMSOL simulation of a particle with d/λ = 6 in evanescent field . . . 46 Hindered diffusion correction vs. . . .1 6. . . . . . . . . . . . . .7 6. . . . . . . .14 6. . . . . . . . . . . . . . . . . . . . . .5. . C. . Sedimentation of large particles . . . . . . . . . . . . . . . . . . . . . .5 6.13 6. . . converging lens position . . . . . . . . . .7 Overall evanescent energy in the suspended particles . . . . . . . . . . . . C. . . . . Normalized apparent velocity distribution of particles in a finite imaging range Apparent velocity distribution of particles from various imaging ranges . . Rescaled time evolution of mean particle apparent velocity . . . . .8 6. . particle/glass gap size . . . . . . . . .1 Schematic of COMSOL simulation geometry . . . . . . . . . . . Apparent velocity distribution of particles remaining in a finite imaging range . . . . . Schematic of potential paths for particle translation . . . . . . . . . . . . . . . 76 76 77 77 78 78 79 80 81 82 82 viii . . . . .9 6. . . . . 71 A. .8 Cross-sectional intensity profile of a 3-µm particle (d/λ = 6) . . . . . . . Percentage of particles remaining in imaging range . . . . . . Time evolution of mean particle apparent velocity . . Spatial distribution of small particles due to hindered diffusion .5-µm radius fluorescent particles in evanescent field . C. . . . . . . . . . . .4 6. . .3 6. 52 56 57 58 59 59 60 61 61 62 63 63 64 65 66 A. . . . . .6 COMSOL simulation of a particle with d/λ = 12 in evanescent field . . . . .10 6. . C. . . . . .2 Photo and schematic of beam incident angle measurement . . . . . . . . . . . . . . . . . . . . .2 COMSOL simulation of total internal reflection . . . . . . . . . . . . . . .11 Digital image of a 6-µm particle (d/λ = 12) in contact with glass surface . C. . . . . . . . . .

. . 53 Representative values of the non-dimensional time between consecutive image acquisition . . . . . . . .1 6. . . . .2 Sample values of the Peclet number and the sedimentation coefficient . . . . . . . . . . . . . . . . . . . . . . . . . .List of Tables 6. 53 ix . . . . . . . . . . . . . .

17]. suspended particulates and the solid surface. investigations into its application in biotechnology has been the most intense. [19] calculated that the ratio of slip velocity to bulk velocity becomes very small if the channel height is more than a few hundred micrometers. Even though the idea of a slip flow has been considered by scientists as early as Newton [18]. Thus the majority of interesting surface-induced physical phenomena occur in the near-surface region of the fluid phase and can be categorized into two groups: (1) changes of the fluid mechanical characteristics due to the presence of the solid surface. In particular. slip flow [7] and surface chemistry directed flow [8. (2) interactions between the dissolved molecules. the existence of boundary slip in a Newtonian liquid has been a controversial topic. the high surface-area-tovolume-ratio also means that near-surface phenomena will have a much larger influence on the bulk of the fluid content. while particle or cell adhesion [10] and detachment [11]. The strong experimental support of such assumption could be due to the fact that the magnitude of slip velocity was much smaller than the measuring instrument accuracy and thus remained undetectable. the solid boundary is viewed as rigid and inert such that physical and structural changes due to fluidic forces are nonexistent. 1. for chemical reactions whose reaction sites are at solid surfaces. increased hydrodynamic drag [12]. In both cases. Microfluidic devices are very attractive in biotechnology over conventional technology because they require small sample volume and produce rapid results. Examples of physical phenomena in the former group include electrokinetic flow [6]. empirical data up to the late twentieth century had deemed it negligible.Chapter 1 Introduction Since the introduction of the first microfluidic device. Unlike polymer solutions whose slip behaviors have been thoroughly investigated [15. Under most circumstances. content sorter [3] and drug delivery [4]. Westin et al. these miniaturized fluidic manipulation systems have been regarded as one of the most promising technologies of the late twentieth century. for example the viscous drag on the channel walls. the high surface-area-to-volume-ratio characteristic of microfluidics offers a much higher efficiency [5]. Another possibility is the relatively insignificant impact of slip on the fluid bulk flow in the macroscopic scale. Indeed. On the other hand. electrostatic interactions [13] and effects of depletion layers [14] are effects of the latter group. Therefore when macroscopic flow measurements 1 .1 BOUNDARY SLIP One of the fluid mechanics questions that have remained unanswered over centuries is the slip flow and its origin. 9]. 16. understanding of interactions between the fluid content and the solid boundary is critical in designing and analyzing microfluidic devices. reagent mixing [2]. Examples of such applications include immunosensors [1].

(1. In some of these experiments. 32]. demonstrating the potential of the TIRV technique in making direct measurements of slip velocities. 31]. With measurements conducted at glass surfaces of sub-nanometer surface roughness. 23. which is based on the Navier hypothesis that the velocity of a fluid at a solid surface is proportional to the shear stress. 31. because some earlier experiments [20. 27] or shear-independent [20. Although it is generally agreed among research findings that surface hydrophobicity aids slip effect [25. with varying degrees of accuracy and uncertainty. These experimental studies produced various magnitudes and dependencies of boundary slip.1) γs ˙ where Uslip is the slip velocity and γs is the fluid shear rate at the solid surface. and an additional slip length of 16 nm is found attributable to surface hydrophobicity. a qualitative conclusion is reached that surface slip length of an aqueous solution would be small and surface chemistry-dependent. [28] to calculate slip velocities.2 were conducted slip effect could be hardly noticeable. hundreds of nanometers [29. [27] and Cottin-Bizonne et al. 30. Experimentally ˙ measured slip lengths were reported to range from micrometers [20. slip velocity was indirectly inferred from measurement of other physical quantities. Although a quantitative conclusion of slip length magnitude could not be drawn due to the uncertainty in determination of the exact imaging range. The accuracy of slip measurements is further improved with the implementation of threedimensional tracking in the TIRV technique. Uslip δ= . again the magnitude of such an effect remains controversial. In the past decade slip flow began to receive more attention again. 30]. [29] and micro-scale particle image velocimetry (µPIV) by Tretheway & Meinhart [30] and Joseph & Tabeling [31]. In Paper 2: Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry (chapter 3). several physical factors that are suspected of inducing slip are also examined. to be introduced later) in performing near-surface measurements makes it a perfect technique for direct observation of slip velocities in a shear flow. That is. 29. . Since then many experimental attempts were made to confirm the existence and the magnitude of slip velocities. we report on a slip length of deionized water being less than 100 nm. 25. [25] to infer slip velocity. The ability of total internal reflection velocimetry (TIRV. In Paper 1: Near-surface velocimetry using evanescent wave illumination (chapter 2). it is shown that velocities of fluid elements within a few hundred nanometers from a solid surface can be accurately measured. δ. Disagreement also exist on whether the slip behavior is shear-dependent [21. In this report. Of particular interest to microfluidics is aqueous solution slip over different surface conditions as water is the most common solvent in biological and chemical applications. 28. while the forces required to move two surfaces separated by a thin film of the test fluid were employed Zhu & Granick [26]. and a slip length above a few hundred nanometers is unlikely. For instance. the relationship between applied pressure drops and measured flow rates was used by Choi et al. 26. Others took a direct approach of studying slip by measuring near-surface flow velocities. 33] to less than 100 nanometers [25. Neto et al. 24] have reported slip boundary conditions in Newtonian liquid. In conducting this experiment. Examples of the direct measurements include fluorescence recovery after photobleaching (FRAP) by Pit et al. 32]. 21] and molecular dynamics simulations [22. The commonly used formulation to characterize boundary slip is the slip length. 28. 27. the obtained slip length values agree with the surfaceroughness dependency previously reported by Granick et al. we report on our first attempt to measure slip velocities of deionized water using the TIRV technique. 26. [34] The issue of apparent slip caused by nano-bubbles at the glass surface is additionally disputed with experiments conducted with sub-nanometer surface roughness and degassed test fluids.

3 Still, questions remained on whether other physical forces, particularly the electrostatic and electrokinetic effects proposed by Lauga et al. [7, 35], would lead to false slip results in TIRV measurements. We address such a concern in Paper 3: Direct measurement of slip length in electrolyte solutions (chapter 4), by repeating the slip measurements with electrolyte solutions whose ionic concentrations have been predicted to reduce the apparent slip effect to sub-molecular level. It is observed that the presence of electrolytes has no effect on the measured slip lengths, suggesting that the observed slip velocities are most likely not due to electrostatic and electrokinetic effects, but are consequences of true boundary slip.

1.2

HINDERED DIFFUSION

Another near-surface phenomenon with great fluid mechanical and mass transport implications is hindered diffusion of colloidal particles. The concept of near-wall hindered diffusion stems from an increased frictional coefficient on a near-wall object. Einstein proposed that the diffusion coefficient of a colloidal object is the ratio of the fluid thermal energy to the object’s frictional coefficient in Stokes flow. Thus if a particle experiences an increase in fluidic friction as it approaches a solid wall, its diffusive translation or Brownian motion will be hindered. Brenner [36] and Goldman et al. [37, 38] were the first to analytically solve the increased frictional coefficients of a near-wall sphere with a lubrication formulation. Their results were later re-confirmed by Chaoui & Feuillebois [12] with an spherical harmonics expansion method. Besides being a fluid mechanics theory, hindered diffusion has additional implications in chemical engineering and drug delivery systems since diffusion is the dominant mass transport mechanism in microfluidics. Because of its significance, advancement of near-surface measurement techniques had lead to several experimental studies of hindered diffusion. Employed experimental techniques include evanescent light scattering off freely suspended particles by Bevan & Prieve [39], evanescent light-scattering spectroscopy by Hosoda et al. [40], catch-and-release video microscopy with optical tweezers by Lin et al. [41], evanescent light scattering coupled with optical radiation pressure by Oetama & Walz [42] and three-dimensional ratiometric total internal reflection fluorescence microscopy with nanoparticles by Banerjee & Kihm [43]. Most of these studies have reported hindered diffusion coefficients agreeing to values derived from the theories of Brenner and Goldman et al. In the theories of Brenner and Goldman et al., the increased friction factor is a function of particle radius and gap size between the particle and the solid wall. Unlike in the fluid bulk, the fluidic friction also becomes anisotropic in the near-wall region. In other words, the hindered diffusion coefficient is different for directions normal and parallel to the solid surface. Few studies have attempted to measure such anisotropicity simultaneously due to experimental limitations. Lin et al. [41] demonstrated such anisotropicity for particles whose gap-size-to-radius ratio is greater than one, and found hindered diffusion coefficient values agreeing with theories. Another study that reported anisotropic hindered diffusion coefficients for gap-size-to-radius ratio greater than one was conducted by Banerjee & Kihm [44]. However, their study reported that the measured diffusion coefficient in the direction normal to the solid wall agreed with the theoretical values only when large particles were tested. By applying the 3D-TIRV technique to freely suspended micron-sized fluorescent particles, we aim to simultaneously observe the three-dimensional anisotropic hindered diffusion for gap-size-toradius ratio much less than one. The theory under this condition is vastly different from that of the gap-size-to-radius ratio greater than one, and based on our literature survey no such experimental confirmation of the theory has been reported. In Paper 4: Direct measurement of anisotropic

4 near-wall hindered diffusion using total internal reflection velocimetry (chapter 5), we demonstrate that the three-dimensional tracking technique of TIRV can be adopted into 3D displacement measurements of freely suspended 1.5-µm radius particles. The displacement measurements reveal that the hindered diffusion coefficients are in close agreement with the theoretical values predicted by the asymptotic solutions of Brenner and Goldman et al. for gap-size-to-radius ratio much less than one, and hindered diffusion anisotropicity is simultaneously demonstrated in all data sets.

1.3

TOTAL INTERNAL REFLECTION VELOCIMETRY

With so much interest in near-surface phenomena, researchers have developed various techniques to study them. Optical microscopy has been widely used to observe interactions in the micrometer scale. However, as fabrication technology advances, the definition of “near-surface” has also evolved from microscopic and to the nano-scale. Traditional optical techniques are no longer sufficient now because the visible wavelength is the physical limit of optical resolution (∼ 0.5 µm). A frequently used technique to overcome this obstacle is evanescent wave imaging [45]. In this technique, a beam of collimated laser light is brought to a solid/liquid interface through the solid phase at a large incident angle, as shown in figure 1.1. Since solids typically have higher indices of refraction than fluids do, total internal reflection occurs at the interface if the incident angle is greater than the critical angle predicted by Snell’s Law. In fact, at the nano-scale this reflection is not “total”, but a small fraction of energy, called evanescent waves, extends into the fluid phase. In the fluid phase the evanescent wave energy or intensity, I, decays exponentially with distance, z, away from the two-medium interface by I(z) = I0 e−z/p , (1.2) where I0 is the intensity at the interface and p is known as the evanescent wave penetration depth. The penetration depth, which characterizes the length scale of the evanescent field, can be calculated from λ0 2 2 −1 p= ns sin θ − n2 2 , (1.3) l 4π where λ0 is the wavelength of the incident light, θ is the beam incident angle and ns and nl are the refractive indices of the solid and liquid, respectively. For visible light, the evanescent wave energy decays to zero within a few hundred nanometers. Thus by using the evanescent wave as an illumination source, one can observe dynamics of fluids and suspended particulates within a few hundred nanometers from the solid surface, while contents outside of this layer would remain invisible. Consequently near-surface phenomena can be observed with clarity and without the interfering noise contributed from objects in the fluid bulk. Since the 1970’s, biologists have combined fluorescence and evanescent wave imaging in studying near-surface biological events [46, 47]. Termed “total internal reflection fluorescence microscopy” (TIRFM), this technique allows observation of fluorescently labelled samples placed in the evanescent field and under a high magnification microscope. Fluorescence is a quantum optical effect of a unique group of molecules called fluorophores, whose quantum states can be excited by a specific range of visible light and release photons of a longer wavelength after initial excitation. Their molecular sizes are excellent for probing near-surface regions at the nano-scale. Examples of TIRFM include the previously mentioned fluorescent recovery after photobleaching (FRAP) and fluorescence resonance energy transfer (FRET). In FRAP, the fluorophores of a chosen spot inside the microscope field of view is purposely photobleached with a high power laser beam. The light-emitting fluorophores in the surrounding area will diffuse into this “dark” spot immediately following the photobleaching. By measuring the diffusivity of these light-emitting fluorophores, one

5

z

liquid ~ penetration depth

θ

solid

Figure 1.1: Schematic of total internal reflection fluorescence microscopy (TIRFM). A fluorescent particle suspended in water was placed in an evanescent field. If the illumination beam incident angle, θ, is greater than the critical angle predicted by Snell’s law, total internal reflection occurs at the solid/liquid interface. The evanescent energy then illuminates the encapsulated fluorophores inside the particle. can study the targeted molecule mobility. For instance, FRAP has been reported in measurement of protein mobility and activity in living cells [48]. FRET, on the other hand, takes advantages of a quantum energy transfer that occurs between matched fluorescent molecules of two different wavelengths [49]. Because such energy transfer can only occur if the two molecules are less than 10 nm apart, they offer an opportunity to examine interactions in the molecular level. Surprisingly, a long time had passed before physical scientists finally caught up with the merits of evanescent wave imaging. At first, evanescent energy scattered off non-fluorescent suspended particles was used in quantitative measurements of near-surface dynamics. Such experimental studies include measurement of surface charge densities on suspended particles by von Grunberg et al. [50], quantifying double-layer repulsion between a colloidal particle and a glass substrate by Flicker et al. [51], effects of particle size on near-wall particle depletion by Kok et al. [14], charged particle adsorption by Kun & Fendler [52], and hindered diffusion by Bevan & Prieve [39]. Only in the past decade were fluorescent tracer particles used in microscopic dynamics measurements. Particle-based velocimetry has long been used in fluid visualization and measurement [53]. It is based on an intuitive and for most part correct assumption that the seeding tracer particles are carried by the fluid surrounding them, and therefore their translational velocities must be that of the local fluid elements. Apparent velocities of the tracer particles are calculated based on displacements of the tracer particles and the time between successive particle imaging. When particle-based velocimetry methods were adopted to study microfluidics, sub-micron fluorescent tracer particles were used to minimize light scattering and imaging noise while attaining spatial resolutions of tens of nanometers [54]. Santiago et al. [55] were the first to demonstrate a correlation-based microscopic particle image velocimetry (µPIV) to map out a velocity field around a Hele-Shaw cell. In their experiment, cross-correlation analysis was performed on successive images of fluorescent tracer particles to identify the most probable ensemble displacements. Since their report, µPIV has been the work horse among diagnostic techniques of microfluidics. When applying to near-surface measurements, µPIV does suffer a major drawback. Its resolution in the direction perpendicular to the imaging focal plane is limited by the focal depth of the

6 microscope objective lens and is at least 0. Both slip and hindered diffusion studies demonstrate the applicability and versatility of the TIRV technique in making near-wall particle velocity and displacement measurements. we present a technique. In Paper 2: Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry (chapter 3).5-µm radius fluorescent particles.5 µm from a solid surface. In Paper 2: Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry (chapter 3) and in Paper 3: Direct measurement of slip length in electrolyte solutions (chapter 4). Zettner & Yoda’s nano-PIV is undoubtedly a novel technique. 100.4 SUBTLETIES OF NEAR-WALL SHEAR FLOW As previous discussion have emphasized. Since the statistical analysis of nano-PIV does not make an attempt to treat the drop-in’s and drop-out’s. owing to the sensitivity of evanescent wave imaging. Zettner & Yoda [56] overcame µPIV’s deficiency by combining it with evanescent wave illumination. Under evanescent wave imaging. further improving its accuracy and applicability. 1. The short effective range of evanescent wave allowed them to measure fluid velocities that were within 280 nm from a solid surface. In Paper 4: Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection velocimetry (chapter 5). and thus the same three-dimensional particle tracking principle applies. Brownian motion of particles in the near-surface region is a non-negligible mechanism and has significant implications in the accuracy of near-wall velocimetry. Brownian motions of the particle ensemble can be monitored and statistically analyzed. it is unsuitable if diffusive dynamics is of interest. Thus numerical approaches offer insightful alternatives to explore the subtleties of near-wall flow. or events where particles diffuse in and out of the evanescent field between consecutive frames. we extend the TIRV technique to three-dimensional. the TIRV technique is extended to observing anisotropic near-wall hindered diffusion of 1. Li et al. where quantitative measurement can be performed on events that occur within 100 nm from the surface.and 200-nm radius fluorescent particles are employed in 3D-TIRV experiments to measure slip lengths over hydrophilic and hydrophobic surfaces. and thus a particle’s peak intensity can be used to infer its distance from the solid surface. called “total internal reflection velocimetry” (TIRV). the peak intensity of these large particles follow an exponential decay relation that is also identical to the exponential decay of evanescent energy. Many of the issues that need to be considered. we use the TIRV technique to measure aqueous solution boundary slip and near-surface hindered diffusion of spherical particles. Secondly. [57] further extended this nano-PIV technique into a multilayer correlation analysis. Consequently. By combining TIRFM with tracking of individual tracer particle. as an alternative to nano-PIV in near-surface measurements. However it still has two significant limitations. could not be easily assessed through experimental data. because diffusion is treated as measurement noise in the cross-correlation analysis of the nano-PIV.5 µm [31]. In Paper 1: Near-surface velocimetry using evanescent wave illumination (chapter 2). µPIV could not distinguish physical features within 0. however. Firstly. The analysis of the image data reveals that a statistical approach is needed in making near-surface velocimetry measurement even for particles whose size variation is as small as a few percent. significant measurement errors could persist. The main concern in near-wall velocimetry is the ensemble behavior of tracer particles in the . As mentioned previously. It is observed that the peak intensity of a fluorescent particle decays exponentially in the same length scale as the evanescent wave energy. leading to a more accurate accounting of all contributing particles. rapid Brownian motions of nanoparticles could lead to a large amount of particle “drop-in” and “drop-out”.

. Sadr et al. the relative strengths of near-wall shear and diffusive forces are explored through Brownian dynamics simulations. how such ensemble behavior affects the accuracy of velocity measurements. hindered diffusion causes an under-estimation of fluid velocity due to biased sampling of tracer particles very close to the wall. this approach has been used in studying the binding rates between attached molecules [60]. a significant portion of the tracer particles would sample planes of higher velocities. ensemble behavior of particles can be obtained through statistical analysis of a large number of repeated simulations. it is found that if the time interval between successive particle image acquisitions is short. and more importantly. however. With a proper time step size. the results found in the Brownian dynamics simulation offer a way where the true fluidic velocities can be inferred from particle apparent velocities through proper scaling. colloidal particle deposition in a microchannel flow [61]. However. initial and boundary conditions. The same Brownian dynamics simulation also provides a tool to assess the effects of hindered diffusion and wall presence on various near-wall velocimetry techniques. The other approach takes advantage of the Langevin equation in which the displacement of a particle consists of a deterministic component (such as displacement of the suspending fluid) and a stochastic component (such as Brownian motion). asymmetric shape and width of apparent velocity distributions are attributed to the shear effect. Still. the time evolution of the particle ensemble phase space configuration function is computed through numerical solution of the momentum-position partial differential equations. Recently. accuracy of potential energy profiles determined by evanescent wave scattering [62] and the accuracy of nano-PIV [63]. respectively. In the Fokker-Planck approach. This scaling method. In addition. In Paper 5: Simulations of hindered diffusion in shear flow and its implications for near-wall velocimetry (chapter 6). Two equally viable numerical approaches have been proposed to study Brownian motion-related problems [58]. is available only for TIRV but not for nano-PIV. The combined effect of the observations described above is that near-wall particle-based velocimetry measurement is not time-invariant. We are most interested in the ensemble behaviors of near-surface particles in a shear flow. Its intuitive resemblance to a physical velocimetry experiment has gained some popularity in colloidal studies. resulting in an over-estimation of overall fluid velocity. Particle drop-in’s and drop-out’s. The ensemble behaviors are then obtained from probability analysis using the phase space configuration function. are found to be a potential source of measurement error for nano-PIV because of the rapid Brownian motion of nanoparticles.7 time interval between consecutive image acquisitions. making TIRV the more superior method in near-surface measurements. [59] had taken the Fokker-Planck approach to study the diffusion-induced bias in near-wall velocimetry. Also called Brownian Dynamics (BD) simulations for its direct emulation of Brownian motions. and the effects of the wall presence and hindered diffusion on the measured velocities because these are issues that we have noticed during our TIRV measurements of boundary slip and anisotropic hindered diffusion. With diffusive time and particle radius as time and length scales. and we believe it can lead to significant inaccuracy in nano-PIV if not treated appropriately. It is observed that the shear force dominates over diffusion at P e > 3. if the time interval between successive particle image acquisitions is long.

1 INTRODUCTION Over the past few years. slip measurement techniques can be classified into indirect and direct methods.Chapter 2 Paper 1: Near-surface velocimetry using evanescent wave illumination Total Internal Reflection Velocimetry (TIRV) is used to measure particle motion in the near-wall region of a microfluidic system. The results suggest that the slip length. the no-slip boundary condition. 21. 2. There are also other groups of researchers who attempted to understand the slip phenomenon from 8 . but uncertainty regarding the exact character of the illumination field prevents a more accurate measurement at this time. if present. fabrication and performance of these devices. is less than 25 nm. the flow rate and pressure within the flow duct [20. The statistical difference between velocities measured over hydrophilic and hydrophobic surfaces was found to be minimal. studies regarding near-wall fluid motion have intensified because of the emergence of micro-electro-mechanical-systems (MEMS). In general. a velocity. Issues associated with the accuracy of the experimental technique and the interpretations of the experimental results are also discussed. In the indirect methods. proportional to the shear rate was observed. Numerical simulations are presented to help understand the results and to provide insight into the mechanisms that result in the experimentally observed distributions. where the fluid-surface interaction is important to the design. a better understanding of fluid-surface interaction becomes necessary in designing applicable devices. In particular. slip velocity is inferred from other measurements. now receives a great amount of attention in microfluidics. 25]. Sub-micron sized fluorescent particles suspended in water are used as seed particles and their images are analyzed with a PTV algorithm to extract information about apparent slip velocity. TIRV images are illuminated with the evanescent field of an incident laser pulse and contain only particles that are very close to channel surface. Thus. many biological applications of microfluidic devices require surface coatings that change the wetting ability of aqueous or biological fluids. Examples of the direct measurement methods include hot film anemometry [32]. At relatively low shear rates (less than 2500 sec−1 ). total internal reflection-fluorescence recovery after photobleaching (FRAP) [29] and micro-PIV [30]. or from forces required to move the surfaces [64]. Moreover. for example. Many experimental techniques have been applied to study the effect of slip on both hydrophilic and hydrophobic surfaces. which has long been experimentally verified in macroscopic flow.

Similarly. Recently. (2. I. vesicle fusion and single-molecule observation [65. (2. One characteristic of the evanescent wave is that its intensity. such as cell-substrate contacts. The TIRFM technique has long been used in biological studies. 67]. the intensity of an evanescent wave decays exponentially with distance z away from the two-media interface I(z) = I0 e−z/d . images taken under evanescent wave illumination have low background noise. In particular. The maximum incident angle which can be achieved by using an objective lens is given by θmax = sin−1 (N A/nglass ) . and the presence of fluorescent molecules or particles suspended in the sample fluid for imaging and measurement. The first method utilizes a prism to guide the laser beam into the glass cover slip. There have also been several recent experimental studies of fluid mechanics by applying evanescent wave illumination [29. which characterizes the evanescent field. which can be derived from Snell’s Law. can be calculated as: d= λ0 2 2 n sin θ − n2 2 4π 1 1 −2 . 2. It takes advantage of the fact that total internal reflection occurs when light travels from a dense medium (such as a glass cover slip) with refractive index n1 into a less dense medium (such as the sample fluid) with refractive index n2 . allowing total internal reflection to occur. The results from these studies are quite diverse. which then serves as a waveguide. the dependence of the slip on the applied shear rate is also a matter of debate with some results suggesting a constant slip length. while others reporting a slip length dependent on shear rate and/or surface hydrophobicity. The other method. decays exponentially with respect to the distance from the two-media interface. which is prismless. Thus. A particle tracking velocimetry (PTV) technique is applied to track particle displacements and flow velocities near the wall.9 a theoretical perspective through molecular dynamics simulations [22. In this paper. Zettner and Yoda [56] used the PIV technique combined with evanescent wave illumination to measure near-surface velocity field in rotating Couette flow. 24]. in which the TIRFM technique is applied to observe particle motion in the close vicinity of microchannel walls with different surface wetting properties. 2. has the illuminating laser beam directed through a high numerical aperture (NA) microscope objective at an angle that creates total internal reflection at the glass-sample interface. 23. 56. This parallel light wave is known as the “evanescent wave”. 68].3) . When total internal reflection occurs. The penetration depth. (2. we describe the Total Internal Reflection Velocimetry (TIRV) technique. a novel technique termed total internal reflection fluorescence microscopy (TIRFM) has been applied to near-wall fluid measurements. as shown in Fig.2) where I0 is the light intensity at the interface and d is known as the evanescent wave penetration depth. Two different optical configurations are commonly used to create an evanescent field close to the glass-sample interface.2 EXPERIMENTAL PROCEDURES TIRFM requires the creation of total internal reflection of a laser beam at the glass-sample interface. and “apparent slip” lengths ranging from nanometers to microns have been reported. 66.1) As mentioned before. Total internal reflection occurs if the incident angle is larger than the critical angle θc = sin−1 (n2 /n1 ). some fraction of the light energy penetrates into the less dense medium and propagates parallel to the surface. showing clear features of sample structure that is near the interface.1.

respectively. 2. depending on shear rate). The coating process was performed in a nitrogen gas chamber to avoid bulk polymerization.3 ms.33) ~ Penetration depth Glass (n = 1.51) Immersion oil Objective lens Figure 2. 5 nsec laser pulses to illuminate the flow. the experimental setup had a critical angle θc = 61. and baked at 100◦ C on a hotplate for a few hours. patterned PDMS channel to create a hydrophobic microchannel. The laser heads were timed to fire with a pre-determined pulse separation (1. At incident angles equal to or larger than θc . illustrated in Fig. and was made permanent after the glass was rinsed in hexane and ethanol. creating an evanescent wave field. They were bonded to the test surface after exposure to an oxygen plasma which renders both the glass and PDMS surfaces highly hydrophilic for a short time during which they bond on contact.1: Schematic of objective-based TIRFM. Following bonding. one can choose appropriate laser wavelength and microscope objective or prism to achieve desired penetration depth. The coated glass was bonded to a cleaned. Based on these calculations. In the present experiment. The test cell consisted of a precision glass cover slip on which a polydimethylsiloxane (PDMS) microchannel was attached. 250±1 µm wide and 25 mm long. To create hydrophobic microchannels.4◦ . we have adopted the prismless TIRFM method.2. It has been reported that surface roughness can . the laser beam could be totally reflected off the water-glass interface. the channels were immersed in deionized-water to maintain their hydrophilic surface characteristics. The beams were focused off-axis at the back focal plane of a Nikon PL Apo NA 1.1 to 3. With water as the sample fluid.4.10 Fluorescent particle solution U z x θ Water (n = 1. it is necessary to coat the glass surface with a transparent and hydrophobic substance because glass is naturally hydrophilic. A dual head Quantel Brilliant Q-Switched Nd:YAG laser was used to generate a pair of 532 nm.9◦ . λ0 = 532 nm and NA = 1.4 100X oil immersion objective lens mounted on a Nikon Eclipse TE200 inverted epi-fluorescent microscope. Because the laser beam is focused off-axis. dried in nitrogen. illuminating two images for use in a Particle Tracking Velocimetry (PTV) analysis. where λ0 is the wavelength of the incident light and n1 and n2 are the indices of refraction for glass and water. With a water-glass interface. it can be calculated that the penetration depth is approximately 150 nm when the incident angle is θ = 63. We coated glass by immersing glass into a 5-mM octadecyltrichorosilane (OTS) solution (dissolved in hexadecane) at room temperature for several hours. A self-assembled monolayer (SAM) of OTS formed on the glass surface. The PDMS channels were fabricated using standard “soft lithography” techniques [69] and measured 31±1 µm deep. the beam incident angle at the glass-fluid interface can be controlled by changing the radial location of the laser beam.

h2 w (2. the shear rate was estimated from the applied volume flow rate. Thus. at flow rates ranging from 0. Measurements were taken at several shear rates.04% volume fractions respectively. 5 nsec pulse 1.35%).and 300-nm particles. respectively [55].3.8 and 43. Near-wall particle images were projected onto a Q-Imaging Intensified Retiga CCD camera. Particle solutions were flowed through the test channels using a Harvard Apparatus 22 syringe pump.1 µm. Monodisperse Fluorescent Polymer Microspheres (Duke Scientific) with diameters of 200 nm (±5%) and 300 nm (±3%). The recorded image pair was finally analyzed with a PTV algorithm written in our group using the MATLAB programming environment.3 nm in the flow channel and the diffraction-limited spot size is 30. The polished Pyrex glass wafer was measured to have a surface RMS roughness of 2 nm. strongly inhibit boundary slip by transferring momentum if liquid molecules are trapped inside the holes of the monolayer [29]. An air gap between the piston and the fluid in the syringe and long tubes were used to dampen pulsations from the syringe motion. Q.6 to 6. the image diameters are 36. were used as seed particles in this experiment. For 200. an atomic force microscope (AFM) measurement of glass surface roughness was also performed. The choice of Particle Tracking Velocimetry (PTV). Under 100X magnification. NA = 1.3 ms separation Nd-YAG laser 532 nm Dichroic mirror and Barrier filter Convex lens Computer ICCD 1360 x 1036 x 12 bit Pulse generator Figure 2. while the same wafer had a surface roughness of 4 nm after OTS-coating. each image pixel corresponds to 64.1 .4 Objective lens 532 nm.11 PDMS channel Syringe pump 100X.02% and 0. rather than Particle Image Velocimetry (PIV) is an important one and this will be re-addressed later on. suspended in pure water at 0. These microspheres have a peak absorption wavelength of 542 nm and emit a red visible light at 612 nm. Typically 200 image pairs were recorded for each flow condition. For the results presented here. assuming steady laminar flow of an incompressible fluid between two parallel plates without wall slip γwall = ˙ du dy = wall 6Q .9 µm [70].2: Schematic of the experimental setup.4) . which has a capacity of recording 1360-pixelsby-1036-pixels 12-bit images.0 µL/min (±0.

eliminates much of the background noise because the evanescent light penetration depth is restricted to approximately 150 nanometers. Typically about 10-20 particles were identified in each image. This threshold value allowed us to detect almost every particle which can be identified with the naked eye and restricts the analysis to particles that are well within the evanescent field and in focus. and particles in the bulk fluid are not illuminated. where h is the channel depth and w is the channel width.5.3 and 2. The number of particles identified in one image is not very sensitive to the exact threshold value. several hundred image pairs at various low shear rates were taken for subsequent PTV image analysis. In this study. on the other hand.12 Figure 2. flood) illumination and evanescent field illumination. A large amount of background noise is observed in the case of direct illumination. It is important to point out that the determination of an intensity threshold sets an observation range. The lower bound of the observation range is the particle radius. for the current experiments.3: Particle images of different illumination methods: widefield (direct.4 compare particle images of widefield (direct. although approximately 50% of these were not tracked between images due to out-of-plane diffusion which causes particles to enter and leave the observation region. . as illustrated in Fig. the shear rate could be slightly lower than is estimated here [25]. the intensity threshold was set at 50% of the brightest particle intensity (usually about 1500-2000 counts from a 12 bit CCD camera). However.3 RESULTS AND DISCUSSION Figures 2. To measure particle velocities near the channel surface. If there is significant slip at the wall. any possible slip velocity is so small that this uncertainty can be neglected (this presumption will be confirmed by the results in the following section). Evanescent field illumination. Bright particles with intensities above a pre-determined threshold value were identified in each image. 2. 2. This characteristic allows easy detection of only particles that are close to the channel surface. flood) illumination. Such background noise is attributed to the light emitting from the out-of-focus particles in the bulk of the fluid and such images are common in micro-PIV measurements but lead to poor cross-correlations during PIV analysis.

Fluid Velocity Observation Range Glass Cover Slip Figure 2. For this reason we . illumination intensity varies exponentially with distance from the wall (as shown in Eq. Ovryn [71] developed a method to determine the three-dimensional position of a spherical particle.5: Schematic of near-wall particles moving near the surface illustrating the observation range. Thus. As the particle moves farther from the wall. representing a particle in contact with the channel surface. However in the evanescent field. the intensity threshold chosen sets an upper bound on the observation range which is different from the penetration depth of the evanescent field. tracking the particles motion in the z -axis (normal to the surface) is not easy. Although tracking the particle displacements in the x -y plane (parallel to the surface) was straightforward. The exact size of the observation range was not known in these experiments and this uncertainty turned out to be the major source of uncertainty in these experiments and is discussed later on.4: Particle images of different illumination methods: near-wall TIRFM image. 2. based on resolving the spatial variations in the scattering pattern. and hence into a region of lower illumination intensity.13 Figure 2. and consequently it is not easy to directly quantitatively associate the fluorescent scattering of a particle in a non-uniform illumination field with the particle’s distance from the wall. the emitted intensity also falls.2).

There is no reason why this should not be possible in the future. one would expect that the particle motion would be purely driven by thermal fluctuations. A threshold displacement was imposed to detect error vectors. 74]. Figure 2. was used to compute the local diffusion coefficient. The width of the particle displacement distribution. (2. ˙ did not attempt to track the particle’s motion normal to the surface.5) 2 . Indeed. which were defined as displacements larger than deemed reasonable. Particles were matched between images using a custom particle tracking algorithm and displacement vectors were calculated from the displacements of particle centers.6: Distribution of particle velocity vectors of 200-nm particles with γ = 469 sec−1 .1 Cross-stream Velocity Analysis Analysis of the displacements of particles in the cross-stream direction was used as to check the consistency of the experiment results. the total number of validated displacement vectors ranged from 700 to 1500. we found that the distributions of spanwise velocities had a Gaussian distribution with mean and skewness close to zero. These large displacement vectors usually result from matching two different particles mistakenly. 73. The Brownian motion is particularly strong due to the particle’s small size. Using the geometric scale and the applied pulse separation. in the absence of any mean flow. At each testing condition. a velocity vector can be calculated from each displacement vector. σ. In this direction. D: σ 2 ∆t D= . and allowed us to locate the particle center coordinates with sub-pixel resolution. a two-dimensional Gaussian curve fit closely approximates the Lommel function. Because the particles are known to be spherical with diameter standard deviation less than 5%.6 shows an example of a collection of velocity vectors obtained for a single shear rate.3.14 140 120 Streamwise veocity (μm/sec) 100 80 60 40 20 0 20 40 100 50 0 50 100 Cross-stream velocity (μm/sec) Figure 2. 2. The intensities of pixels surrounding the peak intensity were fitted to a two-dimensional Gaussian curve [72.

7.6) 6πη where D0 is the Stokes-Einstein diffusion coefficient. found by projecting the velocity distributions onto the mean flow direction. we find it to be 110 nm for the 200-nm particles and 165 nm for the 300-nm particles. x Radius (cm3/sec) 2. Unfortunately. Firstly.7: Cross-stream Brownian motion of particles in an infinite medium and near wall (experiment). However. This is consistent with the high intensity threshold chosen which defines a small observation range. is independent of particle radius. The diffusion coefficient is scaled by the particle radius. a more precise estimate is not available for this data. and η is the fluid viscosity. The dashed line denotes particle diffusion in an infinite medium calculated on the basis of the Stokes-Einstein relation. ♦. the measured values for D are approximately 30% that of the bulk diffusion expected. Coeff. Several features are worth noting. the location of the peak of the distribution increases with increasing shear rate. the measured spanwise diffusion coefficients are independent of the streamwise shear rate and the diffusion coefficient. 2. and the dotted lines represent scaled particle diffusion in an infinite medium calculated based on the Stokes-Einstein relation: kB T D0 · r = . 38]. and Fig.7 shows the cross-stream diffusion coefficient obtained for 200-nm (solid symbol) and 300-nm (open symbol) particles at a variety of different streamwise shear rates over both hydrophilic and hydrophobic surfaces.5 1 0. 2. r. This is confirmation of the presence of hindered diffusion of particles near a solid surface [37.5 x 10 0 0 500 1000 1500 2000 Shear rate (1/sec) 2500 Figure 2. T is the medium temperature.5 2 1. as indicated by the dotted line in Fig. 2.8 shows an example of the distribution of the streamwise velocities. .15 13 Diff. (2. kB is the Boltzmann constant. If we use the observed value of D with the theory for hindered diffusion [37. As expected. Solid symbol represents 200-nm particles and open symbol represents 300-nm particles ( . defined as the distance between the particle center and the wall. hydrophilic surface. when appropriately scaled.2 Streamwise Velocity Analysis Figure 2. 38] to calculate the ”apparent particle distance” from the wall. hydrophobic surface).3. as one would expect. since we did not carefully monitor the observation range.

leading to the observed broadening of the velocity distribution for the streamwise velocity. including the current TIRV integrates through the observation range and thus samples particle motion from a range of shear planes over which the mean velocity changes. One would expect that the 300-nm particles would sample shear planes at higher local velocities than those sampled by 200-nm particles. for a fixed observation range. and there are no discernible differences between the hydrophilic and hydrophobic surfaces and.15 at the low shear rate (barely noticeable) to a clearly observable positive value of 0.02 234 (1/sec) 1172 (1/sec) 2344 (1/sec) PDF 0. and thus the average velocities . In this case the most-probable velocity is quite different from the average velocity. Thus.16 0. a particle is more likely to move upwards (and thus experience a higher local velocity) than to move downwards (and experience a lower local velocity). The lack of any differences between the mean velocities obtained for 200-nm and 300-nm particles is more troubling. Here. is the observation that the distribution widens with increasing shear rate. and that the skewness clearly shifts from 0. between the 200.85 (a stronger positive tail) at the highest shear rate presented. The velocities fall on a straight line through the origin. the true velocity distribution for tracer particles has a Gaussian shape with a mean equal to the flow velocity at that plane. Less expected.8: Experimentally-measured distribution of streamwise velocities of 200-nm diameter particles over a range of shear rates. due to the presence of the wall. we note that. the error bars depict the standard deviation about the mean.01 0 200 0 200 400 600 Velocity ( μm/sec) 800 Figure 2. The reason for the observed increase in the distribution width and skewness can be explained in terms of the observation range from which particle motions are recorded. However. more interestingly. A similar argument can be used to explain the increasing skewness at higher shear rates. At each shear plane. Thus.and 300-nm diameter particles. raising a fundamental question about the definition of the actual fluid velocity in terms of the observed particle velocity distribution.9 shows the variation of the mean velocities as a function of shear rate. and in contrast to the distributions for the spanwise velocities. higher velocity fluctuations are more probable than lower ones in the vicinity near the wall (a similar argument is used to explain positive skewness in the streamwise velocity distribution in a turbulent wall-bounded flow). any micro-PTV or PIV system. the range of mean velocities sampled increases as the shear rate increases. Figure 2. and standard deviation determined by the temperature and particle size.

if present. we find that the average. this suggests that the slip length. is likely no more than 10 nm. including uncertainty in the measurement of the illumination observation range.150 nm. If we use a Navier representation for the slip velocity. our lack of quantitative knowledge about the evanescent wave penetration depth (which probably varied from experiment to experiment) and the observation range makes a more precise estimate for the slip length impossible at this time. us = l∗ γ. The drift velocities measured can result from two effects: (i) the finite thickness of the observation range and (ii) the possible existence of a slip velocity at the surface. hydrophilic surface. representing this combination of both the penetration depth and the slip length. and other forces on the particles (e.9: Apparent particle velocity vs.g.9 is approximately 100-nm. should be larger for the larger tracer particles. 200-nm particle. At the present. 200-nm particle.8) The slope of the data in Fig. 300-nm particle. However. or ”apparent slip” velocity is u = (l∗ + δ) γ. we note that a slip length of order 1000 nm (as proposed by Tretheway . Given that the observation range is estimated to be approximately 100 . . (where ˙ ∗ is the slip length) and compute the average velocity in the mean depth of the observation range. [25] for these low shear rates. If we assume a constant shear very near the surface. with the appropriate characterization and consistent control of the observation range. particle motions in addition to pure translation (such as rolling motion). hydrophilic surface. ˙ (2. colloidal). We can assess the relative contributions of these effects using a very simple theoretical model. there are several factors that are not yet well understood that could contribute to this counter-intuitive observation. l δ. we can write the local velocity profile as: u (y) = us + γy. . this measurement will be possible and such experiments are currently underway. Finally.7) where us is the slip velocity. hydrophilic surface).17 400 Apparent velocity( μm/sec) 300 200 100 0 0 500 1000 1500 2000 Shear rate (1/sec) 2500 Figure 2. hydrophilic surface. consistent with the results of Choi et al. shear rate (♦. However. The fact that this is not observed remains a somewhat puzzling result and one that needs to be resolved in future measurements. ¯ ˙ (2. ◦. 300-nm particle. 2.

At each time step. as it would predict apparent slip velocities at least an order of magnitude larger than observed. For the Brownian motion parallel to the wall.000 particles were uniformly distributed between r and δ. To emulate the experiment. by the method of reflection [37. After 1000 time steps. the assumed upper bound of the observation range. At the beginning of the simulation. is such that D⊥ = D0 4 sinh α 3 ∞ n=1 n (n + 1) (2n − 1) (2n + 3) −1 (2. + 1. For the Brownian motion normal to the wall [36]. is such that [37] D 9 =1 − D0 16 45 − 256 h r h r −1 + −4 1 8 h r h r −3 1 − 16 −5 +O h r −6 (2.10) . 37. in a no-slip steady-state Couette flow between parallel infinite plates separated by a distance h. r. plate separation distance.18 and Meinhart [30]. 38]. flow shear rate and the total time (1000 time steps) were chosen to be equal to measurement conditions.9) .3 Comparisons with Numerical Simulations The experimental results were compared with a numerical simulation of near-wall Brownian motion in shear flow. 38] for (z − r)/r < 1. D⊥ . 38] for (z − r)/r > 1. the modified diffusion constant. Other effects such as particle rotation. The motivation for this was to generate an intuitive model with which to help understand the experimental results. particles that remained within the observation range were counted and their displacements from their initial positions were calculated to obtain a particle displacement distribution. D . and Zhu and Granick [64]) is completely unsupported by these data.325 ln which is the asymptotic solution [37. and D =− D0 ln 2 ln h−r r 2 h−r r − 0.11) − 4. The Brownian motion was modified to account for the hindered diffusion both parallel and perpendicular to the wall [36. 2 sinh (2n + 1) α + (2n + 1) sinh 2α −1 4 sinh2 n + 1 α − (2n + 1)2 sinh2 α 2 where D0 is the Stokes-Einstein diffusion coefficient. the modified diffusion constant. particle radii. and the assumed observation range was chosen to be 100 . similar to that obtained in the physical experiments. . 2. and would be independent of any reasonable estimate for the observation range.9543 h−r r .591 (2. This simple two-dimensional numerical model was developed to simulate the Brownian motion of particles with radius. each particle was moved by a linear combination of the mean shear flow (evaluated at the particle’s geometric center) and Brownian motion (sampled from a normal distribution with a zero mean and a standard deviation √ of 2D∆t). the experimental channel depth.175 nm from the wall. etc were not considered at this time. 10.3. α = cosh−1 (z/r) and z is the distance between the geometrical center of a particle and the wall.

Nevertheless. 2. and that future experiments should pay more attention to this important issue.19 0. as a function of shear rate. . Only a few particles needed to be treated with these heuristic boundary conditions. and this result adds weight to our suspicion that the observation range in the experiments were not consistent.015 PDF 0. Problems can arise when the simulated particle is predicted to pass through the solid surface during the finite time step. The first condition applied a specular reflection.9).01 0. and it was found that the two models gave almost indistinguishable results. quantitative. Unlike the experimental results in Fig. the similarities between the two figures are striking.005 0 200 0 200 400 600 Velocity (μm/sec) 800 Figure 2. there is a clear difference in the estimated velocity using different particle size in Fig. while Fig. 2. As discussed earlier this difference due to particle size is to be expected for a fixed observation range. The second condition emulated near-wall lubrication condition and simply allowed the particle to remain at the wall until Brownian motion perpendicular to the wall transported it back into the bulk flow. denoted as error bars. but present) skewness of the distributions as the shear rate increases. In addition.11. one or more time steps later. Figure 2. with less than 1% statistical difference in the mean and standard deviation of the resultant particle distributions.10: Streamwise velocity distribution of 200-nm particles determined from Monte Carlo simulation at observation depth = 175 nm. the choice of observation range for the computation has not yet been optimized to match the simulation results with those of the experiments. 2. standard deviation and (less obvious. In the present simulation two different wall boundary conditions were tested to deal with this situation. 2. comparisons with the experimental results (Fig.11 shows the variation of the mean velocities and their standard deviations.10 shows the streamwise velocity distributions of 200-nm particles at various shear rates. The rather simplistic nature of the simulation is such that one should not make direct. in which particle was simply reflected off the wall and back into the flow.9 which did not show any difference in measuring velocity using different particle size. showing similar increases in the mean.02 234 (1/sec) 1172 (1/sec) 2344 (1/sec) 0.8 and 2.

For this reason.11: Streamwise apparent velocity predicted by simulation ( . is of order 10 nm or less. ◦. and is limited by the same constraints . and comparable to ”standard” microPIV techniques. we were not able to measure vertical particle motion and our current resolution of the slip length measurement is limited such that we can only state that a slip length. smaller seed particles can move closer to the channel surface than larger particles. etc [70]. the technique also promises the ability to use intensity information to obtain wall-normal displacements as well as precise estimates for the slip velocity at the surface. Both changes will allow direct measurements of flow velocities that are closer to the channel surface. the penetration depth of evanescent field can be decreased. First. 2. 300-nm particles) at the observation depth of 175 nm. 200-nm particles. The present implementation did not (unfortunately) control with sufficient accuracy or consistency the characteristics of the evanescent wave penetration depth which is very dependent on the incident beam angle. diffraction-limited imaging of sub-micron particles. However. we have demonstrated the ability to measure both velocity distributions and mean velocities in the plane parallel to the surface. The accuracy of the TIRV system can still be improved by several means that are subject to future investigation. The determination of velocities in the streamwise and spanwise directions is extremely accurate.20 500 Apparent velocity (μm/sec) 400 300 200 100 0 0 500 1000 1500 2000 Shear rate (sec 1 ) 2500 Figure 2. where slip might contribute more significantly to the local flow velocity. by using a shorter illumination wavelength. TIRV images have very low background noise that allows accurate PIV and PTV measurements of flow velocities within a few hundred nanometers of the surface. although they also suffer from higher Brownian diffusion and so the particle drop-out will increase. the estimate for the slip length is limited only by the accuracy with which we can measure the observation range and so we expect to be able to make measurements with uncertainties less than 5 nm in the near future. However.4 CONCLUDING REMARKS We have demonstrated that Total Internal Reflection Velocimetry (TIRV) can be a very powerful technique for the study of near-wall flows in microfluidic systems. By using sub-micron seed particles. Secondly. if present. .the accuracy of particle centroid location.

[51]. the rate at which ions leak from the glass into solution). including non-Gaussian Brownian motion (due to the finite thickness of the observation range) and hindered diffusion due to near-wall lubrication effects. thermal noise and electrostatic effects (both the bead and the surface have surface charges). A full treatment of these issues is beyond the scope of the current paper. tracking each particle independently and then computing statistics from the individual particle motions might yield a more accurate view of the near-wall velocity field. a second is the possibility of particlewall colloidal interactions and the relative roles of gravity. There are also effects other than the translational hydrodynamic drag that can contribute to the particle motion such that the velocity measurements might be compromised. and most importantly. ionic strength (and for pure water such as used in these experiments. This is an area that has been studied by a number of researchers. Crocker and Grier [75] and Behrens and Grier [76] who have identified colloidal forces that are highly dependant on bead diameter. One is the possibility of particle rotation. . the experimental techniques described here to can be used in further studies of such phenomena. Perhaps the most important consequence of this is that techniques that treat ensembles of particles (such as PIV which looks at the collective motion of particles in an interrogation region. For this reason. surface charge. and indeed.21 Finally. the technique does raise fundamental questions regarding the use of ”passive” tracer particles to estimate the velocity field in microfluidic systems. including Flicker et al. and often averages cross-correlations before peak detection) might be suspect close to walls due to particle interference effects which cannot be neglected and where the particle motion is not represented by Gaussian statistics (and hence the average particle velocity is different from the most probably particle velocity). Our results reveal several features of particle motion in the near vicinity of the surface.

27. 64]. 28. 31].Chapter 3 Paper 2: Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry The existence and magnitude of slip velocities between deionized water and a smooth glass surface is studied experimentally. 27. slip velocity is inferred from other flow quantities. 32. alternative techniques such as micro-particle image velocimetry (PIV) have resulted in direct measurements of near-surface fluid velocity using tracer particles [30. For water flowing over a hydrophilic surface. Issues associated with the experimental technique and the interpretation of results are also discussed. 26. 78. Although the direct measurements have a strong appeal. Although there is considerable disagreement regarding the existence of slip over a hydrophilic surface. 30. while the exact mechanism is not yet understood (for a review of this topic. Molecular dynamics simulations.1 INTRODUCTION The century-old assumption of the no-slip boundary condition between a liquid and a solid has been challenged by recent experimental results and molecular dynamic simulations and has been the subject of many recent investigations exploring flows over both wetting and non-wetting surfaces. but increases slightly as the shear rate increases. In indirect measurements. suggest small slip lengths. 33]. The experiments have been conducted under various direct and indirect measurement techniques with varying accuracy and uncertainty. 64]. In contrast. ranging from micrometers [20. on the other hand. Experimental studies have reported a wide range of slip lengths. 31]. hundreds of nanometers [29. 77. to tens of nanometers or smaller (including no-slip) [25. 23. with the slip length reaching a maximum of 96 nm at a shear rate of 1800 s−1 . Surface hydrophobicity is observed to induce a small slip velocity. 24. there are two major drawbacks inher- 22 . 28. 3. the measurements are in agreement with previous experiments and indicate that slip. 79]. such as the relationship between flow rate and pressure drop [25] or the measurement of the forces required to move a pair of crossed cylindrical surfaces separated by a thin film of the test fluid [26. Sub-micron fluorescent particles are suspended in water and imaged using Total Internal Reflection Velocimetry (TIRV). is minimal at low shear rates. if present. mostly less than 100 nm [22. see [7]). it is generally believed that surface hydrophobicity aids slip.

In summary. (3. an evanescent field can be created near a solid-liquid interface where total internal reflection occurs. 63. which is typically of the order of 0. respectively. 47. 3. (3.g. I. as illustrated in figure 4.2) 2 4π where λ0 is the wavelength of the incident light. 56. A second problem is that the size of measurement volume along the optical axis for flood-illuminated microPIV techniques is defined by the focal depth of the optical system.1) where I0 is the intensity at the interface and p is known as the evanescent wave penetration depth. Total internal reflection fluorescence microscopy (TIRFM) has long been used by biologists to study near-surface dynamics [65. The first is that the velocity is determined from the motion of a collection of particles in an interrogation area (IA) and. micro-PIV would still be insufficient to provide an affirmative conclusion if the true slip length is below 100 nm. [29] measured slip velocity of hexadecane over a sapphire substrate using TIRFM and fluorescence recovery after photobleaching. Zettner & Yoda [56] combined micro-PIV and TIRFM to study near-surface Couette flow while Pit et al. but were unable to make a conclusive determination of the size or character of any slip lengths due to difficulties in the determination the location of tracer particles within the narrow evanescent illumination region. in order to obtain a reasonable estimate of the velocity.2. 44. By measuring the intensity of tracer particles’ fluorescence. One attractive approach that solves both of these problems is the use of Total Internal Reflection Velocimetry (TIRV) [80] which uses total internal reflection of an incident laser pulse to generate a highly localized illumination of the surface flow. which characterizes the length scale of the evanescent field. can be calculated from λ0 2 2 −1 p= n1 sin θ − n2 2 .23 ent to conventional micro-PIV. 67]. The field intensity. 40.1. Although Joseph & Tabeling [31] reported a slip length measurement with 100-nm resolution. z. and in conjunction with a statistical model for the optical and hydrodynamic behaviour of small particles near a surface. 31].1 THEORETICAL CONSIDERATIONS Total Internal Reflection Microscopy The evanescent field and its application to fluorescence microscopy has been thoroughly documented in literature [e. The present manuscript reports on experiments in which the TIRV technique has been substantially improved to the point that quantitative statements can be made regarding the nature of slip velocities over hydrophilic and hydrophobic surfaces.5 to 1 micron [30. Hence the determination of phenomena which might measure a few nanometers is very difficult to resolve. The penetration depth.2 3. we are able to track their motion within a narrow region above the solid surface and to compare that motion to the motion predicted for different levels of boundary slip. 80]. . 66. and relies on tracking the motion of single tracer particles rather than the collective particle motion that is used in PIV. We first reported on the TIRV technique in the context of slip velocities for near-wall microflows [80]. the IA must be measuring several hundred nanometers in size [30]. away from the two-medium interface by I(z) = I0 e−z/p . θ is the beam incident angle and n1 and n2 are the indices of refraction of the solid and liquid. decays exponentially with distance.

The intensity decay length can be obtained experimentally by statistically measuring intensities of particles as a function of particles’ distances away from the substrate.24 Figure 3. (3. including statistical variations in illumination intensity. and d is an intensity decay length that need not be the same as the evanescent penetration depth. If we parameterize all of these variations by an “effective emission radius”. of a particle in an evanescent field is I e (r. quantum efficiency of the imaging system and. Kihm et al. r. 3. I0 is the emitted intensity of a single particle with r = a located at h = a (ie. most importantly. [44] and the assumption that emission intensity of a fluorescent particle is proportional to the number of fluorophores within its volume. (3. a distribution in the physical size of the particles.3) and (3.1: Objective-based total internal reflection fluorescence microscopy. [44] calculated the particles’ emission intensities as an exponential function of their distances to a substrate surface. Based on the calculation of Kihm et al. the emission intensity. that the intensity of fluorescence emission from a large number of nominally identical particles can vary due to a variety of factors. Through experience we have found. we can model the particle’s probability density function (PDF) of its effective emission radius with a Gaussian distribution P (r) = P0 exp − r a −1 2 σr 2 .3) where a is the mean effective emission radius of the batch. touching the substrate surface).4) where h is the distance from the particle’s centre to the substrate surface at which total internal e reflection occurs. however. h) = r a 3 e I0 e−(h−a)/d . I e . σr is the characteristic variation (standard deviation) and P0 is a normalization constant. (3.2.4).5) P (I . h) = α exp − 2   σr       . Combining (3. one gets the joint PDF of the emitted intensity and a particle’s distance to the substrate as  2    Ie 3 (h−a)/d − 1    e   I0 e e .2 Emission Intensity of Fluorescent Particles For spherical tracer particles with a uniform volumetric fluorophore distribution in an evanescent field.

For a given imaging range of h1 < h < h2 . the transitions become more blurred. The e normalized intensity in the horizontal axis is defined as I e /I0 .5 1 1.2). When an ensemble of tracer particles are imaged. with the brightest particles located at the surface (I e /I0 = 1) and the faintest particles e /I e = 0. and some larger particles close the wall contribute to the long tail at high observed intensities. however. However. The lower limit of this imaging range is h = a. they would certainly be located at a range of distances from the substrate surface. representing a particle touching the substrate surface.7) e . A sample plot of this distribution (3. such as an intensified CCD camera used in this study.8. I0 α . as the variation in particle radius at the outer edge of the evanescent field (I 0 increases. the PDF of a tracer particle’s intensity is given by P (I e . h1 < h < h2 ) = h2 h1 P (I e .5) could also describe the particle position distribution within a normalized intensity range of e α < I e /I0 < β by β Ie e P (h. α < I e /I0 < β) = c(h) P (I e .6) is shown in figure 3. h)c(h) dh.05 σr = 0. while small particles close to the edge of the imaging e range are represented by the smooth transition close to I e /I0 = 0. h) d (3.25 1. The upper limit of the imaging range is determined by the sensitivity of the recording medium.2 for a range of values of σr .6) where c(h) is the concentration profile of particles in the fluid. we see that the particles in the defined imaging range are sharply defined in e intensity.5 0 0 0. When performing image analysis.2 1 PDF 0. Instead. Equation (3. and the calculation is performed with a normalized observation range of 0 < h/a < 3 and a normalized intensity decay length of d/a = 1.5 Figure 3. one can choose to identify and analyse only the particles within a certain desirable intensity range. For a low variation in effective emission radius (σr small). ignoring particles that appear too bright or too dim.2: The emission intensity distribution of particles in a uniform concentration field.5 σr = 0.1 σr = 0.15 σr = 0.5 Normalized Intensity 2 2. using particle intensity as a guide. one need not analyse all imaged particles. (3.

one can observe that the distribution has a long positive tail that gradually decays to zero at far away into the fluid bulk.5 < I e /I0 < 1 in a uniform concentration field. The most obvious feature of figure 3.3: The predicted position distribution of particles with 0.2. is given by U hS valid for large h/r.3 is that the distribution is nonuniform with the majority of particles contributing to this range of intensities coming from close to the substrate surface (< 250 nm).9) .7) is shown in figure 3. It is well known that shear and near-surface hydrodynamic effects can cause a tracer particle to rotate and translate at a velocity lower than the local velocity of the fluid in the same shear plane [12. These observations reinforce the fact that a particle’s intensity cannot be monotonically related to its distance to the solid surface. In addition. in a shear flow with local shear rate. (3.2 ln h r 1− 5 16 h r −3 . [38] proposed that a particle’s translational velocity.18. we actually measure the velocity of small spheres near the surface. and U hS 0. and it is critical to recognize the differences between these two quantities. S. An example of this position distribution (3. 3.7431 0.8) −1 .5 < I e /I0 < 1.3 for a normalized intensity e range of 0. and attention should be given to any intensity-based quantitative analysis by recognizing the non-uniform position distribution and the contribution of particles far away from wall. a large particle far away (for example.26 8 x 10 −3 6 PDF 4 2 0 100 150 200 250 300 Particle Center Distance to Wall (nm) 350 e Figure 3.6376 − 0. a = 100 nm. d = 181 nm and σr = 0.3 Near-Surface Shear Flow Although we are interested in determining the velocity of a pure fluid near a solid surface. Because of the large particle effective radius variation in this example. > 300 nm) can exhibit in the same the same fluorescence intensity as that of smaller particles close to the substrate surface. (3. In this prediction. Goldman et al. 38]. U .

c(h) is assumed to be uniform. a. calculated for the same imaging range. significantly lower than that of pure fluid velocity when the imaging range is very close to the substrate surface. which is a good approximation near a surface. α < I e /I0 < β) dh.4. of a large ensemble particles chosen from a normalized intensity range of e /I e < β and located in an imaging range of h < h < h is given by the average of the local α<I 0 1 2 velocity integrated over the imaging range: ¯ U= 1 h2 − h1 h2 h1 e U (h. S (h)) P (h.27 1 0.11) is based on a no-slip condition at the fluid/solid interface.6 0.5 0.072332 ln h −1 r 2 (3. [38]: U hS r exp 0.5 4 Figure 3. [60] proposed a cubic approximation to numerical values presented by Goldman et al. U . and failure to recognize this effect can lead to misinterpretation of the measured data and underestimation of slip lengths. (3.10) .54756 ln h +0. In the ¯ computation of U .11) ¯ ¯ This average apparent velocity.5 2 2. This difference must be incorporated into the data analysis. valid for small h/r.4: The ratio of statistical apparent velocity of particles and mean fluid velocity under no ¯ slip in an imaging range of h1 < h < h2 .5 3.8 ¯ U ¯ V 0.4 1 1. at the solid boundary. of the pure fluid. If there exist a slip velocity. V is computed with linear fluid velocity profile. Uslip . which is non-dimensionalized by particle radius.9 0. Pierres et al.68902 + 0. V . Although there exists no analytical approximation for intermediate values of h/r. U .0037644 ln h −1 r h −1 r 3 + 0.7 0. It should be noted that (3. ¯ The apparent velocity. The ratio of the two is shown in figure 3. from which we see that the average apparent velocity of particles is. not surprisingly. can be compared to the mean velocity.5 h2/r 3 h1/r = 1 h1/r = 1. the apparent velocity of the same ensemble .

28 of particles would be ¯ ¯ ¯ Uapp = Uslip + U = δSw + U . a long tubing (> 1 m long) was used to connect .3. capable of recording 1360-pixels-by-1036pixels 12-bit images. 250 ± 1 µm wide and 15 mm long. This volume of air acted as a pressure shock absorber to damp out the oscillatory forcing of the syringe pump screw. These microspheres have a peak absorption wavelength of 542 nm and emit a red light at 612 nm. Fluorescent particles were illuminated with an objective-based TIRFM method. The dimensions of the test channels were 50 ± 1 µm deep.35%). The roughness of the test surface was measured using an atomic force microscope prior to bonding. making relation of particle center distance from the solid wall and their fluorescent intensities less reliable. To maintain steady pumping motions of the Harvard Apparatus stepper motor. 81]. 5nsec laser pulses (Quantel Twins Q-Switched Double Pulse Nd:YAG Laser) were directed through a Nikon PL Apo NA 1.45 100X TIRF oil immersion objective at an angle that created total internal reflection at a glass-water interface. Sw . Under 100X magnification.7 µm (calculated based on [70]). smaller particles do not contain enough fluorophores to provide significant signals with a 5-nsec laser illumination.1 EXPERIMENTAL PROCEDURES Materials and Setup A schematic of the image acquisition system is shown in figure 3.12) where the slip velocity is characterized as the product of the slip length. on the other hand. Test channels were fabricated using a polydimethylsiloxane (PDMS) (Dow Corning Sylgard 184) molding technique [69] and bonded onto polished glass wafers.3 3. In addition. thus illuminating the near-surface region of water with an evanescent field (figure 4.47 nm and 0. 200-nm particles do not diffuse too far during the inter-frame time. Detailed coating procedures can be found in [80].3 nm in the flow channel and the diffraction-limited spot size is 19. Fluorescent images of near-surface particles were projected onto a Q-Imaging Intensified Retiga CCD camera (ICCD). Secondly. (3. A TTL pulse generator (Berkeley Nucleonics Model 555) was used to synchronize laser pulses and image acquisitions. Lastly. To create hydrophilic microchannels. the pre-bonding PDMS surface was exposed to an oxygen plasma and immediately immersed in deionized-water after bonding to a glass wafer.04% volume fraction were used as tracer particles in this experiment. smaller size fluorescent particles have much larger size variation. each image pixel corresponds to 64.5. we purposely introduced a large volume of air between the piston and the fluid in the syringe. Monodisperse Fluorescent Polystyrene Microspheres (Duke Scientific) with diameters of 200 nm (± 5%) suspended in purified water (Fluka) at 0. There are three reasons that 200-nm diameter particles were selected as probes in this experiment. The energy of each illuminating laser pulse was recorded simultaneously with each image acquisition and this was used to adjust the resultant images and thus account for pulse-topulse variations in illumination intensity. 3. A Harvard Apparatus 22 syringe pump was used to maintain particle solution flow rates. as detailed in [80]. Hydrophobic channels. and the wall shear rate. The hydrophilic and hydrophobic surfaces had an RMS surface roughness of 0. First. A pair of 532-nm.1). with diffusion out of imaging range being our major concern. It has been reported that water has a contact angle of 120 degrees with a similarly prepared OTS-coating [30. δ. ranging from 1 to 10 µL/min (±0. were fabricated by bonding untreated PDMS channels onto octadecyltrichorosilane(OTS)-coated glass wafers.35 nm respectively.

the test channel was filled with rhodamine-B solution. Although this is theoretically feasible. the syringe and the channel inlet.45 Objective lens Mirror Dichroic mirror and Barrier filter Convex lens Computer 1360 x 1036 x 12 bit Polarizing Beamsplitter NI_DAQ Pulse generator Nd-YAG laser 532 nm 532 nm. an experimental method was devised to obtain a ratiometric relation between particle emission intensity and its distance to the glass surface. This procedure has been reported to significantly eliminate nanobubble formation [83]. At each distance to the surface. and subsequently fitted to a two-dimensional Gaussian function to find their centre intensities. Once all necessary images of particle motion were captured. It has been suggested that nanobubbles and a gas bubble layer on hydrophobic surfaces could be alternative sources of apparent slip [7. NA = 1. The process was repeated .4-nm precision stage (MadCity Nano-OP25). which were translated perpendicularly to the glass substrate in the evanescent field with a 0.5: Image acquisition system. in practice. 1200 images were captured at each flow rate. Particle flow motion was observed to be very steady under this setup. 34. Thus. successful use of this technique requires precise knowledge of the illumination beam incident angle and a solution of Maxwell’s equation for an evanescent field in a three media system (glass. 5 nsec pulse 2 . [44] used a ratiometric fluorescence intensity to track particle motions three-dimensionally. Kihm et al. 82]. The pixel values of the imaged particles were then averaged. We attached individual fluorescent particles to fine tips of glass micropipettes. Prior to each experiment. which can be difficult to express explicitly. water and polystyrene).29 PDMS channel Syringe pump 100X. Images of the rhodamine-B solution were subsequently captured under the same evanescent illumination.5 ms separation Half-wave plate Convex Lens 10 micron pinhole Convex Lens Pulse Energy Meter ICCD Power Attenuator Figure 3. the fluid was degassed by placing the solution in vacuum for at least 30 minutes. 30 images of the attached particles were captured. It is intuitive to assume that one can infer the location of a particle in the evanescent illumination field based on the fluorescence intensity. as well as in quiescent fluid. and averaged to obtain an image characterizing the spatial distribution of the illuminating laser beam.

several times with different particles. they generally had a very low density and were usually placed at the edge of an image and far away from where the velocimetry measurements were performed. (nm) 150 200 Figure 3. very close to the evanescent penetration depth that we would predict based on the imaging lens. and the displacement was decomposed into streamwise and cross-stream components (for details of the tracking algorithm.6.2 Velocimetry Analysis The analysis of particle image pairs was performed using a custom particle-tracking program written in a MATLAB environment. every image was scaled by the measured laser illumination energy. λ. The particle intensities are shown in figure 3. along with a exponential fit to the measured data: Ipeak = Ae−λ(h−r) + B.6 0.8 0.2 0 0 50 100 h−r. (3.3.30 Normalized Center Intensity 1 0. images of particles stuck on the glass substrate had to be removed from each image pair to prevent undesirable detection by the Particle Tracking Velocimetry (PTV) algorithm. 3. the signals due to stuck particles and background noise were removed by subtracting a reference image formed from the rolling average of 10 previous and 10 subsequent image pairs.13) where A. The fitted curve predicts an intensity decay depth of 181 nm. The particle seeding density was low enough such .6: Fluorescent particle intensity as a function of its distance to the glass surface. is 181 nm.4 0. Prior to velocimetry analysis. B and λ are constants determined by a least-squares fit. wavelength and refractive indices of the system. The particles used here are 200 nm in diameter. Velocimetry analysis started with finding the positions and peak intensities of all particles via threshold-identification and Gaussian fitting of the intensities of the pixels surrounding the peak. see [80]). Subsequently each particle’s displacement was obtained with a limited-range nearest-neighbor matching. The solid line is a least-square exponential fit to the data whose decay length. To account for laser pulse energy fluctuation. Although the presence of these stuck particles could distort the velocity field. For each image.

400 nm.6). 3.2 Hindered Diffusion Near-surface tracer particles are known to exhibit anisotropic hindered diffusion due to hydrodynamic effects. in quiescent fluid).31 2 1. assuming an imaging range of 100 . Intensities of the “trackable” particles were converted to a normalized intensity ratio obtained by dividing the raw intensity counts by the average of all the stuck particles’ initial intensity counts. 3.5 1 1. The accompanying solid line is the theoretical prediction (3. particles within a desired normalized intensity range were selected for each flow rate and their mean streamwise velocity was defined as the apparent velocity of the ensemble.15.4. Finally.7 shows the distribution of particle intensities recorded during a typical data acquisition (in this case.6) for an imaging range of 100 to 400 nm and a 15% intensity statistical variation with a uniform concentration distribution of particles. For the hindered diffusion analysis.5 PDF 1 0.5 0 Intensity Ratio (Ie/Ie) 2 Figure 3. The good agreement between the observed and predicted intensity variation confirms our calibration results and the statistical theory of the observed particle intensities discussed earlier. The solid line represents the theoretical prediction (3. and the hindered diffusion coefficient in the direction parallel to the substrate surface .5 0 0 0. that this tracking was unambiguous.7: Observed fluorescent particle intensity distribution and its probability density function (PDF). the same analysis steps were followed except that diffusion coefficients of particles in quiescent fluid were obtained from the variance of displacements instead of velocities. and an effective emission radius variation of σr = 0.4 3.4.1 RESULTS AND DISCUSSION Validation of Intensity Calibration Curve Figure 3.

+ 1. Physically.4 0. D is [37] D 9 =1− D0 16 h r −1 + 1 8 h r −3 − 45 256 h r −4 − 1 16 h r −5 +O h r −6 . and their asymmetric shapes show an increasing skewness as the shear rate increases. (3. This “Method of Reflection” approximation is generally valid for (h − r)/r > 1. Note that the streamwise velocity distributions widen as shear rate increases. For (h − r)/r < 1.15) − 4. D =− D0 ln 2 ln h−r r 2 h−r r − 0.8: Ratio of hindered diffusion coefficients (Dexp ) for Brownian motion parallel to a surface.8. The results further improve our confidence in the overall experimental method and subsequent analysis.6 0.32 1 0. The increasing distribution width and skewness at high shear rates are both results of sampling particle displacements from an imaging range with finite thickness [80]. Goldman et al.325 ln The combination of our particle-tracking algorithm and the intensity calibration curve was applied to analyse near-surface particle Brownian motion parallel to the glass substrate. 3.8 Dexp/D0 0.591 (3.14) and shown in figure 3.4. including the fall in the effective diffusion coefficient close to the surface.14) where D0 is the Stokes-Einstein diffusivity of a spherical particle in the fluid bulk. The results are compared to the Methods of Reflection calculation (3. The experimental data shows reasonably good agreement with the theory.5 2 Figure 3.2 0 0 Data 1 Data 2 Method of Reflection 0.9543 h−r r . at higher shear rates.9 shows distributions of particle velocities at three different shear rates.5 1 (h−a)/a 1. hence the increasing width of the . [38] proposed an asymptotic solution. the finite-thickness imaging range contains particles that are translated by a wider range of local velocities.3 Velocity Distributions of Particles Figure 3.

s 2D0 ∆t (3. It can be observed that small Peclet numbers are associated with greater distribution widths while the distributions collapse onto a single skewed distribution at high Peclet numbers. Here the Peclet number is defined as the ratio of the shear-induced displacement.9.02 S = 183 sec−1 S = 732 sec−1 0. during the interframe time period. the shear-induced effects in all velocity distributions in figure 3. the effect of shear flow on particle motion increases and when P e > 3. P e. Data include measurements of flow over a hydrophobic surface. V (3. ∆t.17) Because V is shear-dependent only. Thus the distribution exhibits a more pronounced positive skewness at the higher shear rates. particle motions are completely shear-dominant.10 are scaled to equal magnitudes.005 0 −200 0 200 400 600 800 Particle Velocity (µm/sec) 1000 Figure 3. As the Peclet number increases.10. L. It should be noted that the overlapping . In addition. is scaled by V − V ˜ V = . s.9: Distribution of observed particle streamwise velocities under various shear rates. The explanation is the following: at low Peclet numbers the distributions widths scale inversely to shear. V . because the particles are confined by the solid wall. suggesting a diffusion dominance and thus the large widths of the velocity distributions are due to Brownian motion. they are more likely to move away from the wall and into a faster moving shear plane. apparent velocity distributions.015 PDF S = 1465 sec−1 0. the relatively contributions of velocities from shear flow and Brownian motion can be characterized by the Peclet number. Pe = L V ∆t =√ . The scaled velocity distributions are shown in figure 3.16) ˜ where V is the scaled streamwise particle velocity and V is the mean value of each velocity distribution. At each shear rate. causing the distributions to overlap.01 0. The effect of shear with a finite-thickness imaging range can be characterized if the streamwise particle velocity in figure 3.33 0. to the diffusive displacement.

the intensity distribution of all “trackable” particles were compared to the theoretical prediction (3.52 0.10: Distribution of observed particle streamwise velocities after scaling the velocities distributions in figure 3.11)).20. these velocities were compared to the experimental apparent velocities (mean values of the streamwise velocity distribution) of particles from the same normalized intensity range to extract experimental slip velocities and slip lengths.28 Pe = 4. 3.1 and figure 3.12).5 −1 −0.4 Measurements of Apparent Slip Velocities For each flow case.5 Figure 3. Subsequently. A quantitative comparison of the two cases shows a slip length attributed to surface hydrophobicity ranging from -7 nm at low . we calculated the theoretical apparent velocities of a particle ensemble.12 show the experimental and theoretical apparent velocities as a function of wall shear rate over hydrophilic and hydrophobic surfaces.34 1.12)).5 0 0.20 Pe = 1. Figures 3. with a more observable increasing trend of wall shear rate dependence at higher shear rates. 3.93 Pe = 3. on the other hand. The velocities are scaled according to (3.11 and 3.4. The dashed and dotted lines are the expected apparent velocities if one were to assume a 50-nm and a 100-nm slip length.6) to identify the unknown parameters such as h1 .5 1 PDF Pe = 1. With these parameters known.28 and 4. In the case of the hydrophilic surface (figure 3. In both figures.5 Scaled Particle Velocity 1 1. 732 and 1465 sec−1 in figure 3. which are absent in small Peclet number cases. the increasing distribution width and skewness at high shear rates are both effects of shear.17). respectively.43 correspond to the shear rates of 183. h2 and σr through optimization. the solid lines represent the theoretical apparent velocities assuming zero slip (equation (3. suggests a larger slip length ranging from 37 to 96 nm. The data obtained for flow over the hydrophobic surface (figure 3.5 0 −1.9.11).43 Pe = 5. These observations confirm the explanation that in figure 3. chosen from a normalized intensity range. distributions have the same distribution width. respectively (calculated using (3. asymmetric shape and the same positive tail.4. the measured slip length ranges from 26 to 57 nm with a slightly increasing trend as wall shear rate increases. respectively.16 Pe = 4. P e of 1.7.12) and varying the slip length conditions. as demonstrated in section 3. by using equation (3.9.9.

35 700 Apparent Velocity (µm/sec) 600 500 400 300 200 100 0 0 500 1000 1500 Wall Shear Rate (1/sec) 2000 No Slip (Theoretical) Slip length = 50 nm Slip length = 100 nm Exp. [84]). shear rates to 54 nm at the highest tested shear rate.11 and 3.13). which could serve as the upper limit of imaging range uncertainty. There are three possible sources of measurement uncertainty in figure 3. If there is significant slip at the wall. 64].12) do . One should note that the corrections due to shear-induced rotation (figures 3. The additional slip caused by surface hydrophobicity is in agreement with many experimental results [25. In addition. the shear rates could be slightly lowered than the reported values. 28.11: Experimental apparent velocity of particles in a shear flow over a hydrophilic surface e (0. and its comparison to slopes of expected apparent velocities if there exists a 0. In all cases the 95% confidence intervals are within ±3% of the reported mean velocity values. with an uncertainty of 20 nm taken into consideration. 26. The error bar of each experimental data point represents a 95% confidence interval. our PTV algorithm is able to track particle displacements with high accuracy (< 0. Because wall shear rates are calculated based on microchannel dimensions and flow rates which are controlled by the syringe pump. (b) the apparent velocities of particle ensembles and (c) the imaging range. we estimate that the uncertainty in the wall shear rates is less than 3%. 50 or 100-nm slip length.21 < I e /I0 < 0.12: determination of (a) the wall shear rate. 31]. the largest measured slip length still will not exceed 150 nm – much smaller than other experimental studies which reported slip lengths greater than a few hundred nanometers.11 and 3. 33.1 pixel. it should be noted that. with subpixel resolution of the particle centre detection. and thus the slip effect on wall shear rate calculation can be safely neglected. data Figure 3. The determination of apparent velocities of particle ensembles also has extremely low uncertainly because central limit theorem constrains the statistical mean of a large sample (a few thousand trackings) to a very low uncertainty (< ±3% of the reported velocity values for 95% confidence intervals).78). However. but is in sharp disagreement with others [30.6 where intensity measurements were performed in increments of 20 nm. However. The imaging range is determined statistically based on the particle intensity decay length obtained from figure 3. with an average value of 16 nm in between (figure 3. The largest source of uncertainty is the determination of the imaging range. the largest observed slip length is less than 1% of the channel height.

60 Additional Slip Length (nm) 40 20 0 −20 0 500 1000 1500 Wall Shear Rate (1/sec) 2000 Figure 3. and its comparison to slopes of expected apparent velocities if there exists a 0.12: Experimental apparent velocity of particles in a shear flow over a hydrophobic surface e (0. The error bar of each experimental data point represents a 95% confidence interval.36 600 Apparent Velocity (µm/sec) 500 400 300 200 100 0 0 500 1000 1500 Wall Shear Rate (1/sec) 2000 No Slip (Theoretical) Slip length = 50 nm Slip length = 100 nm Exp. 50 or 100-nm slip length.5 < I e /I0 < 1.0). . In all cases the 95% confidence intervals are within ±3% of the reported mean velocity values. data Figure 3.13: The additional slip length due to surface hydrophobicity.

as a first approximation. Because the slip length calculations in figure 3. Secondly. suggesting that our reported slip lengths are most likely the upper bound of the true slip length value. electrostatic repulsion can create a depletion layer near the substrate surface that effectively shifts the imaging range.5 CONCLUDING REMARKS The results presented here represent direct measurements of local velocities in the near wall region of a fluid-solid boundary. Lumma et al. however. based on the overall particle seeding density. calculations suggest that. 3. surface slip is unlikely to be an issue. we can qualitatively estimate its effect.37 not account for the possibility of slip between the fluid and the tracer particle. slip will reduce the shear-induced rotation of the particle. we can ignore this effect. For our experimental setup. Nevertheless. recognizing the possibility of such depletion layer’s existence.9 µm for a pure water and glass combination. Perhaps a more accurate direct measurement of the true surface slip can be obtained if aqueous solutions of moderately high ionic strengths were used. In contrast. [51] experimentally measured the depletion layer thickness to be in the order of 10 to 30 nm. Lauga [35] reported that a streaming potential can induce electrophoretic motion of particles if particle surfaces are charged. because the tracer particles have relatively rough surfaces. Electrostatic interaction between tracer particles and surface could also affect accuracy of velocimetry measurements [7]. Flicker et al. The slip length we measured with sub-nanometer surface roughness. and applying this theory to the current conditions results in an “apparent” slip length of at most 18 nm. [38] to include the effect of surface slip. Granick et al. This is in close agreement with our experimentally-derived images. [33] suggested that the depletion layer thickness can be as large as 0. and we can conclude that the depletion layer thickness must be much smaller than our imaging range and thus effects due to a depletion layer thickness of this magnitude would not be detectable with our current imaging technique. It should also be noted that Joseph & Tabeling [31]. Since these two mechanisms act in opposition to each other. It has also been suggested that electrokinetic effects could significantly affect velocimetry measurements of suspended particles close to a surface. [34] and Lauga et al. Such a thickness seems unreasonably large as it is much bigger than the evanescent field penetration depth and consequently few fluorescent particles would be observed. hydrophobic coating and shear rates. Despite the relative simplicity of the experiment. Physically. reported slip length very close to the present results. agrees well with the surface roughness dependency reported by Zhu & Granick [26]. however.12 are based on the assumption that particle concentration is uniform near surface. and we are unaware of any work that extends the theory of Goldman et al.11 and 3. We found such a shift would result in higher theoretical apparent velocities of particle ensembles and therefore lower slip lengths for all shear rates. we believe that. If the tracer particles and the substrate surface are similarly charged. we re-analysed the theoretical curves in figure 3. However. Firstly. Slip between the fluid and the tracer particle will modify the particle velocity in two ways.11 and 3. if the particles are uniformly distributed in the fluid one should be able to see around 19 particles in a 26 µm-by-26 µm image field. the reduced shear force will result in the particle velocity lagging the local fluid velocity. [7] both pointed out that large surface roughness can significantly decrease slip. The electrostatic and electrokinetic interactions are an interesting pair as their significance decreases with increasing ionic strength of the test fluid. several subtle issues . with flow experiments conducted under similar surface roughness.12 by assuming an imaging range farther away from the substrate surface. a depletion layer could be a source of “apparent slip” if it is not recognized.

and that this might well reveal the reasons behind the range of reported apparent slip phenomena.38 need to be carefully addressed. . However. 31] that there is minimal slip over hydrophilic surfaces. the tracking in the wall-normal direction is currently limited to an accuracy of approximately 10% of the evanescent penetration depth (about 20 nm in the current case). we believe that a consensus is forming from more recent experiments that the slip effect is not as extreme as some studies might have suggested. such as electrokinetics and nanobubbles. Although the present technique does allow for measurement of particle motions very close to the solid surface. measurement and interpretation of slip data in the nanoscale can be very tricky because other physical phenomena. Most notably. and the statistical variations in the effective visibility of the seeding particles. As many authors have pointed out. might lead to observations of an “apparent slip” caused by other complicating factors. and if the imaging system noise can be reduced (both of which are technical improvements on the immediate horizon).12) over the range of shear rates tested. If particles with a higher degree of monodispersity can be reliably found. the effects of the local shear on the dynamics of seeding particles. there is no possibility that the flow exhibits a slip length more than 150 nm (figure 3. it does suffer from some limitations and there is room for improvement. we believe that the technique can be used to track particles with an order of magnitude improvement in accuracy. including the spatial non-uniformity of the illumination region. but small boundary slip. All of these suggest that close attention must be to paid to documenting the experimental conditions. and that hydrophobic surfaces do appear to introduce a discernable. even with the current levels of uncertainty. While the debate of slip versus no-slip will likely continue in the foreseeable future. The current experiment confirms previous measurements in our own group [25] as well as some other recent measurements [26.

By conducting a threedimensional TIRV (3D-TIRV) experiment. and p is the evanescent wave 39 . A total internal reflection velocimetry (TIRV) technique was recently reported to be an effective method in measuring liquid velocities in the region of less than 300 nm from solid surfaces [80]. e I e (z) = I0 e−h/p .Chapter 4 Paper 3: Direct measurement of slip length in electrolyte solutions The emergence of microfluidics over the past decade has brought out increased research interests in liquid boundary slip. 28. 33]. reported that slip length can be obtained within a 20-nm uncertainty [85]. 21. Advancements in flow measurement techniques have resulted in many recent experiments reporting aqueous slip length with a wide range of values and various degrees of dependency on applied shear rate. 35].1). They reported a small but non-negligible slip length for de-ionized aqueous solutions in shear flows over both hydrophilic and hydrophobic surfaces. as an attempt to answer these open questions. Huang et al.1) e where h is the distance between the particle center and the solid phase. In this letter. 31. slip length measurement of ionic aqueous solutions using TIRV is presented and compared to the previous results of Huang et al. and relies on tracking motions of individual tracer particles to determine fluid velocity vectors in the planes parallel to a solid surface (as demonstrated in figure 4. [85]. exhibits the same exponentially decaying characteristic as the evanescent field energy does. 32. Subsequently slip velocities and slip length can be inferred from the measured apparent velocity vectors by applying a statistical model for optical and hydrodynamic behaviors of small particles near a solid/liquid interface. was presented by Huang et al. 25. 26. I e . However. including a proof of concept and a discussion of its subtleties. namely. its emission intensity. The TIRV technique uses total internal reflection of an incident laser beam to generate a highly localized illumination of the near-boundary liquid phase. surface hydrophobicity and surface smoothness [20. A detailed description of the 3D-TIRV theoretical work. When a fluorescent particle of radius a is illuminated by an evanescent field of incident wavelength λ0 (a < λ0 ). A brief summary is presented in this letter to facilitate the readers’ understanding and resultant discussion. which for centuries has been regarded as negligible. The exponentially decaying evanescent field leads to determination of tracer particles’ positions in the direction normal to the solid surface based on their fluorescent intensities. I0 is the emission intensity of the particle when it is in contact with the solid phase (h = a). some questions remained as to whether electrostatic and/or electrokinetic forces acting on the tracer particles could lead to false slip observation [7. (4. 30.

Because tracer particles are used to probe near-surface shear flow. the shear-induced lift under the presented experimental conditions is insignificant. a variety of factors. most importantly. However. and the local fluid shear rate. or effective emission radius. penetration depth. Thus. including statistical variations in illumination intensity. of the particle distance to the solid surface and emitted fluorescent intensity can be obtained.3) h 2 − h1 h 1 Because equation (4. (4. quantum efficiency of the imaging system and. . However. non-uniformity of the particle sizes. U . h. α < Ie < β) dh. ¯ The ensemble-averaged velocity. it is important to recognize that shear and near-surface hydrodynamic effects can cause a tracer particle to rotate and translate at a velocity. which characterizes the length scale of the evanescent field. P (h.2) 2 4π where n1 and n2 are the indices of refraction of the solid boundary and liquid. Uslip . a. I e ). the corrected particle translation velocity does not account for the possibility of slip between the fluid and the tracer particle. assurance is needed that the particle ensemble uniformly samples all shear layers. It should be noted that in the presented analysis. and thus any calculation based on U represents a no-slip case. 60]. The penetration depth. Factors that determine the particle translation velocity include particle radius. and thus alleviating this concern (see Appendix B). based on the theory presented by Cherukat & McLaughlin [87]. can be calculated by h2 1 ¯ U= U (h. a. can be calculated from λ0 2 2 −1 p= n1 sin θ − n2 2 . can lead to variations in the observed fluorescent intensities of an ensemble of nominally identical particles. of a large number of uniformly distributed particles emitting in an intensity range of α < I e < β and located in an imaging range of h1 < h < h2 . 38.1: Objective-based total internal reflection velocimetry (TIRV) system. particle distance to the solid surface. respectively. lower than the velocity of the local fluid element in the same shear plane [12. S (h). with which a joint probability density function (PDF). if there exists a slip velocity. It has been suggested that shear-induced lift can be a source of particle migration away from the solid wall [86].40 Figure 4. Therefore corrections need to be made when one convert fluid velocities into particle velocities. U . S (h)) P (h. If near-wall velocities of a pure fluid are to be measured by tracking motions of tracer particles. a statistical analysis is carried out by characterizing these variations through a single parameter r. (4.3) is based on a no-slip boundary condition.

This procedure has been reported to significantly eliminate nanobubble formation [83]. Prior to each experiment. γw .45 100X TIRF oil immersion objective at an angle that created total internal reflection at a glass-water interface.4) where Uslip is characterized as the product of the slip length. 34. The hydrophilic and hydrophobic glass surfaces had an RMS surface roughness of 0.005-0.35 nm respectively. respectively. the fluid was degassed by placing the solution in vacuum for at least 30 minutes. with special attention paid to removing pulsations and maintaining steady pumping motions. one slip length can be obtained for each wall shear rate. ˙ (4. but is in sharp disagreement with some others [30. These microspheres have a peak absorption wavelength of 542 nm and emit at 612 nm.41 at the liquid/solid substrate interface. the same particle ensemble will exhibit an apparent velocity of ¯ ¯ ¯ Uapp = Uslip + U = δ γw + U . Finally. Hydrophilic and hydrophobic microchannels were created by oxygen plasma treatment of the PDMS mold and by bonding untreated PDMS channels onto octadecyltrichorosilane(OTS)-coated glass wafers. Lauga [35] suggested that for an aqueous solution with 1 . The dimensions of the test channels were 50 ± 1 µm deep. Monodispersed Fluorescent Polystyrene Microspheres (Duke Scientific) with diameters of 100 (±10%) and 200 nm (±5%) suspended in purified water (Fluka) at 0. Detailed description and schematics of the experimental setup have been presented in [85].2 shows the measured slip lengths of aqueous solutions under various experimental parameters. 26. With U ˙ calculated by using equation (4. Slip lengths were then obtained based on equation (4. In summary. thus illuminating the near-surface region of water with an evanescent field (figure 4. Test channels were fabricated using a polydimethylsiloxane (PDMS) (Dow Corning Sylgard 184) molding technique [69] and bonded onto polished glass wafers. The first observation to be made is that all measurements report slip lengths of approximately 100 nm or less.1). The additional slip caused by surface hydrophobicity is in agreement with many experimental results [25. velocimetry analysis started with finding the positions and peak intensities of all particles via threshold-identification and Gaussian fitting of the intensities of the pixels surrounding the peak. The analysis of particle image pairs was performed using a custom particle-tracking algorithm. Slip on a hydrophobic surface appears to be slightly higher than on a hydrophilic surface. 33]. It ˙ ¯app can be experimentally measured by finding an ensemble-averaged velocity should be noted that U ¯ of “trackable” tracer particles while γw is computed from flow rate and channel geometry. The flows were driven by a Harvard Apparatus 22 syringe pump. 250 ± 1 µm wide and 15 mm long. These results further confirm previously reported measurements [25. A quantitative comparison of all measurements shows that the slip length attributed to surface hydrophobicity averages to 22 nm at the tested shear rates. particles within a desired intensity range were selected for each flow rate and their mean streamwise velocity was defined as the apparent velocity of the ensemble and compared to theoretical no-slip values. however. Figure 4. Fluorescent images of near-surface particles were captured by a Q-Imaging Intensified Retiga CCD camera (ICCD). is expected to decrease with increasing ionic strength of test fluids.04% volume fraction were used as tracer particles in this experiment. 85] that there is minimal slip over smooth solid surfaces. Approximately 1000 images were captured at each flow rate. 28. and the wall shear rate. 31]. δ. Electrokinetic effects [35] and electrostatic repulsion between tracer particles and glass surface [7] have both been proposed as sources that might cause an apparent slip and could affect the accuracy of slip velocity measurements.3).47 nm and 0. Subsequently each particle’s displacement was obtained with a limited-range nearestneighbor matching. A continuous-wave 514-nm Argon-ion laser beam (Coherent) was directed through a Nikon PL Apo NA 1. 82]. which could be an alternative source of apparent slip [7. 31. Their significance. 26.4).

Hydrophilic 200−nm. Still. 31]. Hydrophilic” and “200nm. To assess this. variations in aqueous ionic concentration do not change measured slip lengths significantly. a TIRV system equipped with a very sensitive high-speed camera would be needed for measurements at high shear rates. as the conducted measurements span only one order of magnitude of shear rates. DI−water. Zhu & Granick [26]. but are consequences of true boundary slip. As shown in figure 4. DI−water. 0. Hydrophobic 200−nm. Hydrophobic” have been previously reported by Huang et al. 1mM NaCl.1mM NaCl. DI-water.2. The uncertainty of each slip length measurement (not drawn for plot clarity) is less than 20 nm. mM NaCl concentration. 29. The error bar on each data point represents a 95% confidence interval of each ensemble calculation. It is physically plausible that the shear dependence of slip length is nonlinear and the shear effect is significant only at higher shear rates. Neto et al. estimated based on penetration depth calibration and particle radius variation [85]. Hydrophobic 50 0 0 500 1000 1500 Wall Shear Rate (1/sec) 2000 Figure 4.2: Measured slip lengths of aqueous solutions. [25] all tested over larger ranges of shear rates (> 2 orders of magnitude) and reported shear-dependent slip lengths. [27] and Choi et al. Hydrophobic 100−nm. the apparent slip length would become sub-molecular. Hydrophobic 200−nm. Data of “200-nm. DI−water. DI-water. This observation suggests that the measured slip lengths are most likely not due to ionic effects. this result does not suggest that slip length is independent of shear rates. This observation agrees with experimental studies conducted under similar ranges of moderate shear rates [28. Another important observation is that the slip lengths do not vary significantly over the tested range of shear rates.42 Apparent Slip Length (nm) 150 100 200−nm. .

Chapter 5 Paper 4: Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection velocimetry By applying the three-dimensional total internal reflection velocimetry (3D-TIRV) technique to freely suspended micron-sized fluorescent particles. Inside a fluid bulk the diffusivity of an isolated particle follows the Stokes-Einstein relation. In the presence of a nearby solid boundary. a particle would experience an increased. [37] for gap-size-to-radius ratio much less than one. 42]. total internal reflection fluorescence microscopy (TIRFM) [43. many experimental studies have been conducted to observe hindered diffusion and to verify the correction factors. which is the ratio of fluid thermal energy to the particle’s hydrodynamic mobility. and hindered diffusion anisotropicity is simultaneously observed in all data sets. The displacement measurements reveal that the hindered diffusion coefficients are in close agreement with the theoretical values predicted by the asymptotic solutions of Brenner [36] and Goldman et al.and nano-particles in engineering research (eg. particle imaging velocimetry) and biotechnology (eg. 40. Demonstrated experimental techniques include evanescent light-scattering [39. simultaneous three-dimensional measurements of the anisotropic diffusivity could not be 43 . however. 5. [37] were the first to developed drag force correction factors for near-wall spheres under no inertial effects and a no-slip boundary condition. Brenner [36] and Goldman et al. These correction factors have since evolved into theories of hindered diffusion [88].1 INTRODUCTION Advancement in colloidal sciences have lead to many applications of micro. We demonstrate that the three-dimensional tracking technique of TIRV used for nanoparticles can be adopted into 3D displacement measurements of freely suspended 1. At these length scales Brownian motions are significant and understanding these thermal agitations are essential in making practical use of these tiny particles. Following these works. In most of the above mentioned studies. anisotropic drag which hinders its mobility. drug delivery).5-µm radius particles. 44] and combined optical tweezers and digital video microscopy [41]. we aim to simultaneously observe the threedimensional anisotropic hindered diffusion for gap-size-to-radius ratio much less than one.

its Brownian motion is hindered due to an increase in hydrodynamic drag. [37] offered an asymptotic solution. [41].2 THEORY OF HINDERED DIFFUSION When an isolated particle is at the vicinity of a solid boundary. and the gap size between the particle and the wall. using a regression method.9543 h a . In addition. defined by h/a 5.2).3 EXPERIMENTAL PROCEDURES To experimentally verify equations (5. diffusion of micron-sized fluorescent particles in an evanescent field was tracked with digital imaging microscopy. in measuring motions of near-wall nanoparticles [89]. no study has demonstrated the anisotropic hindered diffusion for h/a 1 with a simultaneous three-dimensional measurement as studies of Banerjee & Kihm and Lin et al. Later Bevan & Prieve [39]. who reported anisotropic diffusivities that only partially agree with the theories.1) is a close approximation to Brenner’s infinite series solution. who demonstrated that the anisotropic hindered diffusion could extend far away from the wall. Based on our literature survey. + 1. fluorescent intensity of the particles can be used to infer the instantaneous separation distance between the particle and the two-media interface. Another exception is a study reported by Lin et al. The presence of the solid wall also breaks the symmetry of particle dynamics. a. have been demonstrated by Huang et al. In the theories of Brenner and Goldman et al. resulting in anisotropic Brownian motion. reported that β⊥ ≡ D⊥ 6h2 + 2ah ≈ 2 D0 6h + 9ah + 2a2 (5. β . an exact analytical solution was never found. h. They proposed analytical solutions of the hindered mobility that are vastly different for h/a 1 and for h/a 1. a. β ≡ D =− D0 ln 2 ln h a 2 h a − 0. 5. both attempted to measure hindered diffusion at h/a 1. β⊥ . probably due an inaccurate accounting of particle positions. TIRV allows particle centers to be identified on the captured fluorescent images. 90]. the hindered mobility of a near-wall particle is a function of the particle’s radius. as commonly done in traditional TIRFM experiments [44. and consequently particle motions can be tracked via the displacements of particle centers. Unlike evanescent light-scattering experiment of non-fluorescent particles.44 conducted due to experimental limitations. . termed 3D-TIRV. Brenner [36] successfully solved the lubrication equation of particle motion normal to the solid wall and proposed an infinite series solution of the drag correction factor. One exception is a ratiometric-TIRFM measurement conducted by Banerjee & Kihm [43].325 ln 1.2) − 4. In this paper we present our experimental study aimed to validate the anisotropic mobility correction coefficients for h/a 1 through a direct measurement of hindered diffusion using three-dimensional total internal reflection velocimetry (3D-TIRV). (5. Instead. For the diffusivity correction factor in the direction parallel to the solid wall. and the gap size between the particle and the solid wall. He reported that the correction factor is a function of the particle radius. The technique.591 for particles very close to the wall. h.1) and (5. Goldman et al.

The water phase was an aqueous solution consisted of 10mM NaCl and individual 1.5-µm radius fluorescent polystyrene particles were attached to polished conical tips of thin graphite rods. Spherotec Inc. Our 3D-TIRV setup was created on an inverted epi-fluorescent microscope (Nikon TE-2000).51) Figure 5.4-nm precision translation stage.9◦ . Still.33) ~ penetration depth θ glass (n = 1. especially in determining the relationship between the particle/wall gap size and the particle’s peak fluorescent intensity. To do so. collimated 514-nm Argon-ion laser beam was directed through a Nikon PL Apo NA 1. a calibration experiment was necessary to verify its validity for 1. 200 µL of the solution was injected into a closed reservoir formed by a piece of carved polydimethylsiloxane (PDMS) sandwiched between glass coverslips. This method has been previously demonstrated in TIRV with nanoparticles. and recorded at a 10-Hz frequency. ensured the isolated-particle assumption of the theories was met. total internal reflection occurs at the glass/water interface. or 32 nm.02% volume fraction. By mounting the assembly to a 0.5µm-radius (±5%) fluorescent polystyrene particles (refractive index 1. greater than 61. A red fluorescent microparticle suspended in water is placed in a green evanescent field created within the focal depth of a high numerical aperture microscopy objective. At the illumination beam incident angle. Subsequently TIRV imaging of hindered particle Brownian motions was performed at one of the glass/solution interface. the fluorescent particles were traversed through the evanescent field and their fluorescent images were digitally captured at various distances to the glass surface. A particle’s 3D position was determined through threshold-identification and Gaussian fitting of the intensities of the pixels surrounding the peak.1: Schematic of TIRV. 2000 to 2500 sample images were captured for each experiment for reliable statistics. Before each measurement the solution was sonicated in intervals of 30 seconds for 10 minutes to disperse coagulated particles. individual 1.45 100X TIRF oil immersion objective at an angle that created total internal reflection at a glass/water interface.2.5-µm particles. and thus low particle number density. Because the particle radius is much larger than the evanescent penetration depth. Overall.5 pixels. We also found that the precision of center identification was approximately 0.) suspended at 0. A continuous-wave. A collimated laser beam through the objective is used as the illumination light source. The result of this calibration is shown in figure 5.59. θ. The low volume fraction.45 z x water (n = 1. the evanescent energy illuminates only the lower portion of the encapsulated fluorophores. . The close agreement between particle intensity’s exponential decay length and the evanescent field penetration depth suggests that particle-glass gap size can be reliably inferred from the particle peak intensity. For image acquisition. The fluorescent images of the particles were projected onto a Q-Imaging Intensified Retiga CCD camera (ICCD).

The hindered diffusion coefficient in the direction parallel to glass surface.4 RESULTS AND DISCUSSION Shown in figure 5.2: Gaussian-fitted peak intensities. diffusion coefficient of the same 1. The laser incident angle is 64◦ and the resulting evanescent field penetration depth is 148 nm.3 are hindered diffusion correction factors measured by 3D-TIRV. of 1. and represents ensemble average.5-µm radius fluorescent particles in evanescent field. I0 is the fitted intensity at h = 0. 5. Finally. hindered diffusion coefficients from a Brownian dynamics simulation [89] are also plotted for comparison. These results are compared to the asymptotic solution of Goldman et . With an exponential decay length of 142 nm.5-µm particles were measured in the fluid bulk.6 I/I0 0. ∆t is the time between consecutive image acquisitions (= 100 millisecond). In addition to experimental results. D⊥ . Similarly. Each data point represents the particle peak intensity averaged over 10 images.2 0 0 Particle 1 Particle 2 Particle 3 Exp. the hindered diffusion coefficient in the direction normal to glass surface.4) where ∆z is the particle displacement normal to the glass surface. All data series exhibit an exponential decay tendency versus gap size. to obtain the hindered diffusion correction factors. taken at 0 to 275 nm away from the glass surface. fit (d = 142 nm) 50 100 150 200 Gap size.8 0.4 0. using regular fluorescent digital microscopy. while the length of the error bar spans two standard deviations. β and β⊥ .46 1 0. the best fit of the intensity data shows close agreement to the penetration depth. D . particle displacements were calculated between consecutive frames.3) where ∆R is the averaged radial displacement of particles. Once positions of all imaged particles were determined. was calculated from the equation ∆R 2 = πD ∆t (5. h (nm) 250 300 Figure 5. I. was calculated from (∆z)2 − ∆z 2 = 2D⊥ ∆t (5.

1 in the same h/a range.) DX (Exp. β ≡ Dx /D0 . Theoretically equation (5. (equation (5. “Exp.3 0. Each error bar represents the 95% confidence interval of measurement.025.4 at h/a ≈ 1.) DZ (Sim. who conducted one-dimensional hindered diffusion measurement in the range h/a < 0.2) is more appropriate for the measured gap sizes of h/a 1 while the method of reflection solution is more accurate at h/a > 1.2) but significantly deviate from the MOR solution.1 0 0 0.3: Hindered diffusion correction factor (Dx /D0 . demonstrating the anisotropicity of near-wall hindered diffusion.#1) DX (Exp.2 at h/a ≈ 0 to 0. Each data point is obtained from an ensemble of particles found within ±0.2)). the Bevan approximation and “Method of Reflection” solution. Bevan’s approximation (equation (5. The transition from the asymptotic solution to the method of reflection solution is demonstrated at h/a > 0. The correction factor in the direction parallel to glass surface.4 0.1 h/a GCB MOR Bevan DX (Sim. while the correction factor in the direction normal to glass surface.3.#2) 0.7 0.05 are very close to that predicted by equation (5.05 0. One of these forces is gravitational . A close agreement between the measured and the predicted β⊥ values at extremely small h/a has also been reported by Oetama & Walz [42]. Finally. respectively. we realize the importance of recognizing other acting physical forces that could lead to measurement inaccuracy.01 of the targeted gap size values.” represents experimental data while “Sim. As in most near-wall colloidal measurements.. increases from 0 to 0.” means data obtained from Brownian dynamics simulation.2 0. confirming the validity of our measurement technique. It is observed that the measured β values at h/a < 0.5 D/D0 0.15 0.2 Figure 5. Dz /D0 ) vs. “MOR” and “Bevan” represent asymptotic solution of Goldman et al. β⊥ = Dz /D0 .1 where measured β⊥ is slight larger than the predicted values possibly due to insufficient or slightly biased particle sampling.1 as the measured β values fall between the two solutions. increases from approximately 0. D0 is the diffusion coefficient measured in the fluid bulk while Dx = D and Dz = D⊥ .#1) DZ (Exp. the experimental data also agree well with the simulation results. The measured β⊥ agrees well with Bevan’s approximation.47 0. “GCB”. non-dimensional gap size between particles and glass surface (h/a). but nevertheless they are both plotted for a quantitative comparison.#2) DZ (Exp.6 0.05 < h/a < 0. confirming the validity of the asymptotic solution at h/a 1. except at 0.1)) and a “Method of Reflection” (MOR) solution [37]. al.005 or ±0.

when the particle approaches the solid boundary. In the fluid bulk of density ρf and dynamic viscosity µ. Zhu & Granick [26] concluded that apparent slip lengths are only a few nanometers if the rms surface roughness is 3.5 nm or larger. Besides hydrodynamic interaction. based on the theory of particle mobility presented by Vinogradova [93]. By adding 10 mM of NaCl to the solution. we found few particles came close to glass surface. electrostatic repulsion can also exist between the polystyrene particles and the glass surface as they both carry negative surface charges when immersed in aqueous solution [51]. this terminal settling velocity would be 245 nm/sec. Thus under the current experimental conditions. it is expected that the theories of Brenner and Goldman et al. Because the polystyrene microspheres used in the current hindered diffusion experiment are hydrophilic and have surface roughness of at least 10 nm [92]. First. slip of such magnitude would have an insignificant effect on the particle mobility for the particle diameter and particle-wall gap size under consideration. Indeed when the same fluorescent particles were suspended in de-ionized water and observed under evanescent wave imaging. However. if not subnanometer. However the length scale of electrostatic repulsion can be easily decreased by increasing the electrolyte concentration of the suspending aqueous solution.1). we decreased the electric double layer thickness surrounding the particles and the glass surface to a Debye length ∼ 3 nm. would under-estimate near-surface particle mobility and the hindered diffusion coefficients. offer insights into whether slip effect should be taken into consideration. or equivalently a non-dimensional gap size of 0. for a particle/glass gap size of 10 nm and greater.5 nm during the imaging inter-frame time of 0. equivalent to a settling displacement of 24.3. At a first glance. within the evanescent penetration depth the particles would settle for a distance of less 2.0016 for our measurement gap size of h/a < 0. van der Waals force scales with h−3 [91]. The hydrodynamic mobility correction factors proposed by Brenner and Goldman et al. ws . For a 1.1 seconds. h/δ 10.1 seconds.5-µm radius polystyrene particle of density 1050 kg/m3 . and is additionally verified by the agreement between experimental data and theoretical values shown in figure 5. Based on equation (5. Consequently. and thus significantly reduced the effect of electrostatic repulsion. If fluid slip exists at the particle surfaces. Another colloidal force that could potentially affect measurement accuracy is van der Waals force between the particles and the glass. its terminal velocity is expected to decrease with a correction factor identical to β⊥ for the same hydrodynamic reason of decreased mobility.002. the effective slip length at particle surfaces. Therefore the no-slip assumption is considered valid in the current experiment. . assume a no-slip boundary condition at particle surfaces. δ. the terminal settling velocity. however. Several recent theoretical and experimental studies.5) ws = 9µ where g is the gravitational acceleration. making its effective range even shorter than that of electrostatic repulsion. is expected to be of only a few nanometers in magnitude. or h/a ∼ 0. of a particle with radius a and density ρp is 2 (ρp − ρf ) ga2 .14. by conducting slip measurements of water flowing over hydrophobic surfaces.45 nm in the inter-frame time of 0. it is perceivable that sedimentation could lead to bias in diffusion measurement normal to the glass boundary. (5. Therefore in the time scale of our Brownian motion measurements. However.48 pull or sedimentation of the particles. sedimentation can be considered as a negligible factor. Thus it is safe to assume that van der Waals force is also insignificant for the range of h/a under consideration.

5 SUMMARY A direct measurement of hindered diffusion was conducted with total internal reflection velocimetry. . and such correction shall be applied to not only diffusion but also other translational motion of particles where the drag force is of concern.49 5. We believe that this is the first simultaneous three-dimensional measurement of hindered diffusion coefficients for h/a 1. The results confirm the increase of hydrodynamic drag when a particle approaches a solid boundary. The anisotropic characteristic of hindered diffusion is experimentally confirmed and the correction factors are found to agree with the theories of Brenner and Goldman et al.

if the time interval between successive particle image acquisitions is long. 6. are found to be a potential source of measurement error for nano-PIV because of the rapid Brownian motion of nanoparticles.[55] and nano-scale [56] where size of tracer particles has also decreased from microns to nanometers. The combined effect of the observations described above is that near-wall particle-based velocimetry measurement is not time-invariant. and the effects of the wall presence and hindered diffusion on the measured apparent velocities are explored through Brownian dynamics simulations with diffusive time and particle radius as time and length scales. 85] or a method of statistical tracking [94]. In the past decade. an under-estimation of fluid velocity can result from biased sampling of tracer particles very close to the wall. Its accuracy heavily relies on the assumption that tracer particles will accurately trace local fluid velocities. Brownian motion of tracer particles can be quite significant. This scaling method. a statistical analysis of the particle ensemble [80. resulting in an over-estimation of overall fluid velocity. At these physical length scales. thus violating the assumption of particles travelling at velocities equal to that of the surrounding fluids. Particle drop-in’s and drop-out’s. It is observed that the shear force dominates over diffusion at P e > 3. It is also found that if the time interval between successive particle image acquisitions is short. respectively. Still. however. However. Consequently. an ensemble correlation averaging technique [54]. conducting an accurate velocimetry experiment requires a good understanding and a careful treatment of Brownian motion via either a thorough error analysis [55]. making TIRV the more superior method in near-surface measurements. particle image velocimetry (PIV) has been extended to the micro. and asymmetric shape and width of apparent velocity distributions are attributed to the shear effect.1 INTRODUCTION Colloidal particle-based image velocimetry has long been used as an experimental method in measuring fluid velocity profile and other physical quantities. a significant portion of the tracer particles would sample planes of higher velocities. is available only for TIRV but not for nano-PIV.Chapter 6 Paper 5: Simulations of hindered diffusion in shear flow and its implications for near-wall velocimetry The ensemble behaviors of near-surface particles in a shear flow. 50 . the results found in the Brownian dynamics simulation offer a way where the true fluidic velocities can be inferred from particle apparent velocities through proper scaling.

both Dx and Dz would be equal to the Stoke-Einstein . its displacement between simulation time step i and step i + 1 with a step size δt is. In most cases. We then discuss the simulation results and their implications. (6. 42. The other approach is a Brownian dynamics (BD) simulation. 38] and experimentally [39. The usefulness of BD simulations has been demonstrated in studying the impact of hindered Brownian diffusion on the accuracy of nano-PIV [63]. Only a few researchers managed to turn the disadvantages into merits by proposing usage of the extent of hindering to reconstruct near-wall shear flow velocity profile [96] and to probe the no-slip boundary condition [97]. dz kB Θ 2Dz δt . which is to simulate motion of each particle in a large ensemble through a stochastic differential equation. kB is the Boltzmann constant and Θ is the fluid temperature. 37. This is known as hindered diffusion and has been reported both theoretically [12. in their total internal reflection velocimetry (TIRV) experiment [80]. where (x . To study the ensemble behavior of hindered and non-hindered Brownian motions. [59] had taken the Fokker-Planck approach to study the hindered diffusion of particles in the direction normal to the wall and its effect on the bias of velocimetry.1b) dDz Dz δt + Fz δt + N 0. on the colloidal particle deposition in a microchannel flow [61]. which solves the momentum-space partial differential equations to obtain a time evolution of spatial configuration function of the particle ensemble. both deterministic (such as fluid flow and sedimentation) and stochastic (such as Brownian motion) processes contribute to particle displacements. such a spatially dependent Brownian motion would lead to greater difficulties in accessing measurement uncertainties. Dx and Dz are the diffusion coefficients in the directions parallel and normal to the solid surface.51 In the vicinity of a solid wall. Sadr et al. Details provided by this method are not only intuitive in understanding ensemble behavior but also bear a close resemblance to physical velocimetry experiments. respectively. In this paper the theories of near-wall particle dynamics will first be presented and followed by a detailed description of the BD simulation algorithm. it would be of interest to examine how the relative strength of hindered diffusion and shear flow leads to skewed near-wall apparent velocity distributions first reported by Jin et al. xi+1 − xi = U δt + zi+1 − zi = dDx Dx δt + Fx δt + N 0. dx kB Θ 2Dx δt . 43.1 THEORIES AND COMPUTATIONS The Langevin Equation In Brownian dynamics simulation. One is the Fokker-Planck approach. Recently.2 6. For a particle in a shear flow as shown in figure 6. time between image acquisitions and the accuracy of velocity measurements based on ensemble averages. In addition. particle displacements are computed based on a stochastic Langevin equation [58]. 95]. and on the accuracy of potential energy profiles determined via total internal reflection microscopy [62]. there are two equally appropriate approaches [58].1.2. a colloidal particle will undergo a “hindered” Brownian motion due to an lubricational increase of its hydrodynamic drag. 36. and the relationship between particle diffusion. 6. In this approach. What has yet to be examined is the effect of hindered diffusion and the presence of the wall on the accuracy of velocimetry. z ) is the particle’s center position.1a) (6. In the fluid bulk. and conclude with a summary and our thoughts on the accuracies of various velocimetry methods. The grand simulation then yields a spatial distribution of the particle ensemble after each time step. 41.

As it has been mentioned earlier. as the time scale. 2Dδt denotes the stochastic displacement of Brownian motion randomly sampled √ from a normal distribution with a zero mean and a standard deviation 2Dδt. respectively.1b) are non-dimensionalized by choosing the particle radius.1a) and (6. √ and N 0. diffusivity. dDx /dx = 0 in equation (6. a2 /D0 . The z -dependency of Dx and Dz is discussed in section 6.1: A schematic of the simulation geometry.3b) . A colloidal particle of radius a is freely suspended in a fluid but near a solid boundary. The gravitation sinking force acting on the particles. Fg . Even though existence of slip has been reported [85].3. is thus the only external force in the z-direction and Fg = 4π 3 a ∆ρg. near-wall diffusion is anisotropic and hindered.1a) and (6.2. and g is the gravitational acceleration. Therefore equations (6. 2βz (Zi ) δT (6. and thus Dx < D0 and Dz < D0 . as the length scale and time required for the particle to diffuse a distance of one radius.3a) dβz dZ δT + G · βz Zi δT + N Zi 0. on the other hand. its presence would not alter the simulation results and is thus neglected. 2βx (Zi ) δT (6. 3 (6. D0 .and z-directions.52 a U S z’ z x x’ Fluid Solid Figure 6.1a). The fluid is undergoing a shear flow of shear rate S. It should be noted that because Dx is only a function of z . and thus the particle translates due to both fluid flow and its Brownian motion. In the current BD simulation scheme there exists no force in the x-direction and Fx = 0. Density mismatch between tracer particles and the suspending fluid. A no-slip boundary condition is assumed at the fluid/solid interface. The equations of motion (6.1b) become Xi+1 = Xi + F Zi · P e · Zi · δT + N and Zi+1 = Zi + 0.2) where ∆ρ is the density difference between the tracer particles and the suspending fluid. Fx and Fz represent external forces acting on the particle in the x. leads to either sedimentation or floatation of the particles. a.

042 × 10−5 -0.2 and 6. ∆T . another important control parameter is the time between consecutive image acquisitions. where X ≡ x /a. P e. G.630 4630 G= −5.1: Sample values of the Peclet number. Pierres et al. D0 G≡ 4πa4 g∆ρ .54756 ln Z − 1 (6.2. Z ≡ z /a.5) 1− 5 Z 16 −3 . ∆t. (6. U .2 ln (Z − 1) (6. respectively and F Z ≡ U .00022 ∆t = 10 msec 2.2: Representative values of the non-dimensional time between consecutive image acquisition. and T ≡ D0 t/a2 are the non-dimensional x .1 msec 0.072332 ln Z − 1 + 0.0463 46. Some sample values of non-dimensional time between consecutive image acquisition.68902 + 0.6) 2 +0.22 0.022 0. and U ≡F Z zS 0.1. [60] proposed a cubic approximation to numerical values presented by Goldman et al.4630 463.4) for small Z . 38]. [38] proposed that the translational velocity. for 1-µm and 100-nm particles are shown in table 6. G.5042 a = 100 nm a = 1 µm Table 6.3 while Pe ≡ Sa2 .000022 ∆t = 1 msec 0. 6.0 103 sec−1 4. z and t.0046 4. of a particle with radius a in a shear flow of a local shear rate S is given by U ≡F Z zS which is valid for large Z .0022 ∆t = 100 msec 22 0.022 Table 6.2. Although there exists no analytical approximation for intermediate values of Z . a = 100 nm a = 1 µm ∆t = 0.0037644 ln Z − 1 3 .630 101 sec−1 0. and sedimentation coefficient.7431 0. characterizes the gravitational pull on the particles. In image-based velocimetry measurements. D0 βz Z ≡ Dz (z ) D0 as described in sections 6. and the sedimentation coefficient.2. Goldman et al.2 Effects of Shear on Particle Velocities It is well known that shear and near-surface hydrodynamic effects can cause a tracer particle to rotate and translate at a velocity lower than the local fluid velocity in the same shear plane [12.2 0. . in [38]: U ≡F Z zS 1 Z exp 0. zS βx Z ≡ Dx (z ) .53 S = 100 sec−1 P e = 0.2. P e. 3kB Θ The Peclet number. represents the relative strength of shear force and diffusion. Representative values of P e and G are shown in table 6. ∆T .30 102 sec−1 0.6376 − 0.

Like the particle translational velocity. Dx ≡ βx Z D0 =− 2 [ln (Z − 1) − 0.2.4 Implementation of Simulation Since the experiments are conducted using dilute suspensions. based on the theory presented by Cherukat & McLaughlin [87]. 2 sinh (2n + 1) α + (2n + 1) sinh 2α −1 4 sinh2 n + 1 α − (2n + 1)2 sinh2 α 2 where α = cosh−1 (Z ). and thus will be neglected (for details. where the particle would be in contact with the wall.3 Hindered Diffusion Near-wall tracer particles are known to exhibit anisotropic hindered Brownian motion due to hydrodynamic effects. A boundary condition was needed in the event that a particle attempted to enter the solid wall during a simulation step. (6. Goldman et al.3a) and (6. Dx . the modified diffusion constant. .9543] .54 Because a particle’s translational velocity depends on the shear rate and the distance to the wall.and nano-particles in a low shear flow regime.10) well approximates equation (6. For Z < 2. the smallest Z value a particle could have was Z = 1. it needs to be updated in equation (6. the dependence of these hindered diffusion coefficients on Z requires them to be updated at every time step of the simulation. Because the solid wall was located at Z = 0. is such that [36] Dz ≡ βz Z D0 = 4 sinh α 3 ∞ n=1 n (n + 1) (2n − 1) (2n + 3) −1 (6. The simulation then progressed for a total of ∆T /δT steps as prescribed by equations (6. and their hindered diffusion coefficient in the direction parallel to the solid surface. the BD simulations were conducted under an assumption of no particle-particle hydrodynamic interactions. [38] proposed an asymptotic solution.8) In the direction normal to the wall. [ln (Z − 1)]2 − 4. Z ) of the particle recorded after each time step. see Appendix B). It has also been suggested that shear-induced lift can be a source of particle migration away from the wall [86]. the shear-induced lift is insignificant for micro. Bevan & Prieve [39] reported that the equation βz Z = 6 (Z − 1)2 + 2 (Z − 1) 6 (Z − 1)2 + 9 (Z − 1) + 2 (6.1a) at every time step of the simulation. 6.3b).2. with F and β’s updated and the position (X . This “Method of Reflection” approximation is more accurate for Z > 2. is [37] Dx ≡ βx Z D0 =1− 9 Z 16 −1 + 1 Z 8 −3 − 45 Z 256 −4 − 1 Z 16 −5 +O Z −6 . Dz .591 (6.325 ln (Z − 1) + 1. a particle was situated at X = 0 and randomly placed in a pre-determined range of 1 < Z < 10. However. At the beginning of each simulation.9) .7) where D0 is the Stokes-Einstein diffusivity of an isolated spherical particle in the fluid bulk.9) and is much less computationally intensive than the infinite series. 6.

5-µm particles. Therefore δT was chosen to be 10−4 . on the other hand. the time step must be much greater than the particle relaxation time. 2D0 ∆t . Particles that start off farther away from the wall can diffuse a longer distance within a given amount of time. δT 1). as demonstrated by the wider displacement distribution widths. nano-PIV and TIRV as tracers. However. Finally. or the z-direction.1 RESULTS AND DISCUSSIONS Sedimentation Accurate particle velocimetry measurements require unbiased sampling of fluid velocities by particle ensembles. these displacement PDF’s are more symmetric and collapse onto each other. mD0 /kB Θ. it is important to mention the selection of the size of the computational time step. Shown in figures 6. In both case the final spatial distribution agrees with the Boltzmann distribution prediction [91]. 6. With the time step taken small enough. non-uniform spatial distribution after a transient period. with the presence of the wall hindered diffusion is expected to break the symmetry of the displacement PDF.001% of the simulated displacements required applications of the boundary condition.3 6. Secondly. 99] have studied the efficiencies of different boundary conditions for BD simulations. It was also repeated for various values of parameters P e and G to study their effects. it can be observed that the distribution started out uniformly.3. but do serve to validate our simulation. which has two physical constraints [58].4. which satisfies both constraints and was numerically efficient. respectively. sampled after various times ∆T . With sedimentation being negligible for small particles.55 Peters & Barenbrug [98. The single particle simulation was then repeated 105 times to obtain a large ensemble. In the case of 1. as these particles are commonly used in micro-PIV. the use of the boundary condition was seldom triggered and less than 0.3. The displacement PDF’s of near-wall . These results are not surprising.2 Particle Displacement Due to Hindered Diffusion The first observation made is the displacement probability density function (PDF) of hindered Brownian motion in the direction perpendicular to the wall. In addition. numerical accuracy requires that the time step must be short enough such that the diffusion coefficients and their gradients are essentially constant during the time step (that is. δT .3 are BD simulations of particle distribution for 1.5-µm and 100-nm radius particles. where m is mass of one particle. It is well known that the one-dimensional displacement PDF of an isolated particle in the fluid bulk is a Normal √ distribution N 0.2 and 6. First. 6. The initial positions of particles and the seeding of our random number generator were made identical for all runs of simulations to ensure that the differences in results found between runs were the consequences of simulation parameters only. A common bias of this sort is sedimentation of particles between image acquisitions.11) kB Θa2 for a > 100 nm. but evolves with time and eventually settle into a steady. and is demonstrated in figure 6. we concentrate on presenting results of sub-micron particle simulations. Here we chose a simple and yet effective specular reflection to prevent a particle from entering the wall. The 100-nm particles. remain dispersed through the volume at all times. In nondimensional terms it is equivalent to mD0 2 δT ∼ O 10−6 (6.

56 (a) 1 0.2 0 1 2 3 4 5 Z (b) 2. . The particles are uniformly distributed in the volume (1 < Z < 10) at T = 0 and allowed to diffuse only within this volume. which is equivalent to a 1.2: Sedimentation of particles at G = −2. PDF stands for probability density function.5.8 PDF 0.5 ΔT=1 ΔT=2 ΔT=3 ΔT=5 Figure 6.5 1 0.5 0 1 2 3 4 5 Z 6 7 8 9 10 ΔT=5 ΔT=6 ΔT=7 ΔT=8 ΔT=9 Δ T = 10 Boltzmann Distribution 6 7 8 9 10 ΔT=0 Δ T = 0.5 2 PDF 1.4 0.6 0.5µm-radius polystyrene particle (density 1050 kg/m3 ) suspended in pure water.

12 0.5 T=1 T=2 T=3 T=5 9 10 Figure 6.06 0.06 0.04 1 2 3 4 5 Z 6 7 T=5 T=6 T=7 T=8 T=9 T = 10 Boltzmann Distribution 8 9 10 6 7 8 T=0 T = 0. which is equivalent to a 100nm-radius polystyrene particle (density 1050 kg/m3 ) suspended in pure water. .57 (a) 0. The particles are uniformly distributed in the volume (1 < Z < 10) and allowed to diffuse only within this volume.1 PDF 0.08 0.04 1 2 3 4 5 Z (b) 0.3: Sedimentation of particles at G = −5 × 10−5 .08 0.1 PDF 0.12 0.

After long periods of time when the particles are allowed to reach the wall through diffusion (∆T > 0.5. initially uniformly distributed within the imaging range. with the spatial distribution skewed toward the wall (Z = 0).3 Particle Drop-in’s and Drop-out’s The accuracy of velocimetry measurements relies on the ability to detect the same particles in consecutive image acquisitions. The particles. more particles would drop out with longer inter-acquisition time. The portion of the particles that are not present in the imaging range at the second image acquisition are said to have “dropped out”.1.5.5 0 −2 −1 0 Z − Z0 1 2 Figure 6. which is a good estimate for evanescent wave imaging using 100-nm radius particles.3. Further proof of the hindered diffusion effect is shown in figure 6. on the other hand. which is a direct consequence of hindered diffusion. Suppose the imaging range of near-wall microscopy is 1 < Z < 5.58 2 Z0 = 1. One such situation is demonstrated in figure 6. Clearly. images of the dropping in particles at the second image acquisition would lead to additional errors . In PIV. many “outside” particles would have to “drop in” to take over the space.3). This is especially difficult in near-wall microscopy measurements because near-wall sub-micron particles can diffuse away from the near-wall region quickly.4: Displacement distribution of small particles from various initial positions. 6. the skewness of the spatial PDF is even more pronounced. are significantly skewed with a more pronounced tails in the positive values and are narrower. Such time-dependence of the particle spatial distribution could lead to time-dependent apparent velocity measurements and is further discussed in section 6.5 PDF Z0 = 2 Z0 = 6 Z0 = 11 1 Z0 = 51 Z0 = 101 0. leading to a high concentration of particles at the wall.6. For the overall concentration of the particles to remain uniform. This observation suggests that hindered diffusion could lead to a bias of measurements toward values obtained in the near-wall region. during which time no particle is able to reach the wall. Particles start to diffuse away from the initial position of Z = 2 as time evolves. All particles are allowed to diffuse for an identical time ∆T = 0. will start to leave the imaging range as time evolves.3. particles.5 1. or Z0 .

25 0. .05 0 0 5 Z’ 10 ΔT=0 Δ T = 0.5 0 1 2 3 Z 4 5 Figure 6. All particles are located at Z = 2 at T = 0.1 0.0001 Δ T = 0. The distributions are taken after various inter-acquisition times. ∆T .15 0.59 1.01 Δ T = 0.2 PDF 0.5: Spatial distribution of small particles at various times due to hindered diffusion.5 1 PDF ΔT=0 Δ T = 0.5 ΔT=1 ΔT=5 0.6: Spatial distribution of particles located within the range of 1 < Z < 5 at T = 0.001 Δ T = 0.1 ΔT=1 Δ T = 10 15 Figure 6. 0.1 Δ T = 0.3 Δ T = 0.

V = ∆X /∆T .7: Percentage of particles in the imaging range for different inter-acquisition time. however. the focus is on the effect of shear on apparent velocity measurements. which is in agreement with a separate numerical study conducted by Sadr et al.8 is the distribution of non-dimensional apparent velocity. or equivalently the shear rates. [80] and Huang et al. or ∆t = 0.0 < Z0 < 3. this percentage drops to 65% if the inter-acquisition time is 10 msec. The imaging range is the same as the initial particle distribution range.2.0 1. present less of a problem for Particle Tracking Velocimetry (PTV) as the technique makes an effort to distinguish the particle drop-in’s and drop-out’s. which are particles that do not drop out. the acquisitions should be taken with ∆T < 0. Shown in figure 6. the major concern for PIV is the percentage of particle drop-in’s and drop-out’s because all imageable particles at the time of image acquisition are analyzed without any effort to distinguish the entering and the leaving. emulating physical TIRV measurements. In dimensional terms. The percentage of particles that do not drop out for various inter-acquisition time is shown in 6. more than 90% of 100-nm particles remain in the imaging range of 100 < z < 300 nm (10% drop-out) if image acquisitions are taken at 1 msec apart.5 msec for 100-nm radius tracer particles. a longer times between consecutive acquisition results in a lower percentage of “good” particles. near-wall PIV measurement with sub-micron particles should be conducted with a inter-acquisition time as short as possible to ensure accuracy. Based on this result.60 100 % Remaining in Imaging Range 90 80 70 60 50 40 30 0 2 4 6 8 10 1. [85]. suggested that the increasing skewness and widening of the apparent velocity distribution . in measurements. P e.3.0 < Z0 < 5. In particular. Z0 .7. The dropout’s. If an acceptable percentage of drop-out’s is 95%.0 < Z0 < 5.0 3.4 Horizontal Apparent Velocity Distributions Subsequent analysis concentrates on horizontal motion of particles that do not drop out. As mentioned previously. [59] Not surprisingly. for particles under various Peclet number. 6. This agrees well with experimentally measured PDF’s reported by Jin et al. It is clearly observable that the apparent velocity PDF’s widen and skew as shear rate increases. Jin et al. However.0 ΔT Figure 6.

Vavg is the mean of each apparent velocity PDF.6 PDF Pe = 0.8 0. larger P e has a narrower but more skewed normalized PDF.61 0. In general.5 Pe = 1 Pe = 3 Pe = 5 Pe = 10 Pe = 30 0.9: Normalized apparent velocity distribution of particles remaining in the imaging range 1 < Z < 3 after ∆T = 10.8: Apparent velocity distribution of particles remaining in the imaging range 1 < Z < 3 after ∆T = 10. 1. width.5 Pe = 1 Pe = 3 Pe = 5 Pe = 10 0.5 1 PDF Pe = 0. . and skewness.2 0 0 10 20 V 30 40 50 Figure 6.3 Pe = 0.5 0 −3 −2 −1 (V−V 0 avg 1 )/V avg 2 3 Figure 6. The normalized PDF’s collapse onto each other at P e > 3. Larger P e leads to an apparent velocity distribution with larger mean.4 0.

[85].02 0. It is noted that the collapsed PDF’s for P e > 3 have a smaller distribution width but a larger skewness. and the agreement of BD simulations further validates the presented conclusion.10 that the apparent velocity PDF’s of imaging ranges farther away from the wall are wider and less skewed than that of imaging ranges closer to the wall.12. it is observed that diffusion is quite dominant at small ∆T . the horizontal displacement of the particle ensemble is dominated by diffusion. One can observe in figure 6. the time evolution of apparent velocity distributions was examined and shown in figure 6. leading to apparent . while the distributions of P e < 3 are wider but more symmetric.9. as shown in figure 6. The explanation is that at low shear rates.5 Time Evolution of Apparent Velocity Distributions Finally.03 PDF 0. the finite thickness of the imaging range and shear effect give rise to the skewness and distribution widening.05 1 < Z0 < 2 0. as plotted in figure 6.01 0 0 2 < Z0 < 3 3 < Z0 < 4 4 < Z0 < 5 5 < Z0 < 6 50 V 100 150 Figure 6. All apparent velocity distributions are measured at ∆T = 10 and P e = 10. This physical explanation is supported by the fact that both skewness and distribution width are scalable by the shear rate only at P e > 3. Another source of apparent velocity distribution skewness is the hindered diffusion and the presence of the wall. At high shear rates.10: Apparent velocity distribution of particles at various imaging ranges.3. violates the symmetric PDF assumption of PIV analysis and could lead to measurement errors. on the other hand. it is observed that the apparent velocity PDF’s can collapse onto a single distribution for P e > 3. The combined effect of the hindered diffusion and the shear-induced rotational slow-down in the near-wall region causes the apparent velocity PDF’s to skew. Upon further analysis. and their distributions are more symmetric due to less influence of the wall and hindered diffusion. while at far away the wall presence is not felt and the symmetric distributions again show scaling similarity. The same shear-dominance effect at P e > 3 has been reported experimentally by Huang et al. For a fixed P e. 6. Particles that start off farther away from the surface move faster because they are carried by fluids at higher velocity planes.11. The difference is even more obvious if each PDF is re-scaled by its distribution mean.04 0. Vavg . leading to a wider normalized PDF.62 0.

01 Δ T = 0.12: Time evolution of apparent velocity distribution. Note that the apparent velocity distribution narrows and skews with increasing inter-acquisition time.03 PDF 0.01 0.05 1 < Z0 < 2 0.0001 Δ T = 0. All apparent velocity PDF’s are obtained at P e = 30 and with particles within the imaging range of 1 < Z < 3 at T = 0. .01 0 −100 2 < Z0 < 3 3 < Z0 < 4 4 < Z0 < 5 5 < Z0 < 6 6 < Z0 < 7 7 < Z0 < 8 −50 0 V − Vavg 50 100 Figure 6.01 Δ T = 0.11: Collapsed apparent velocity distribution shown in figure 6.1 ΔT=1 Δ T = 10 0 V 100 200 300 400 Figure 6. 0.025 0.015 PDF 0. Vavg .04 0.02 0.005 0 −300 −200 −100 Δ T = 0. ∆T .02 0.63 0. The PDF’s are scaled by subtracting their corresponding distribution mean.10.

3 Pe = 0.95 0.1 1.64 1.15 Vavg/Vshear 1.12. . Vshear is the calculated velocity of the particles in a near-wall shear flow if there is no Brownian motion. and is the target quantity that velocimetry is trying to measure. All mean apparent velocities are obtained from particles that remained at imaging range of 1 < Z < 3.05 1 0.5 Pe = 1 Pe = 3 Pe = 5 Pe = 10 Pe = 30 8 10 Figure 6.13: Time evolution of mean particle apparent velocity.9 0 2 4 ΔT 6 Pe = 0. Thus Vavg /Vshear can be interpreted as the accuracy of velocimetry measurement. Each data point represents the mean value of one apparent velocity distribution shown in figure 6. Note that data of all P e show a minimum velocity ratio at approximately ∆T = 1.

The velocity ratio would keep increasing as ∆T increases. With further analysis. If Vshear is defined as the mean velocity of all fluidic planes inside the imaging range and Vavg is the mean value of an apparent velocity distribution such as the ones shown in figure 6. except near the minimum. For each data series. Figures 6. [59] suggested an alternative time scale. the imaging range is set at 1 < Z < W and the non-dimensional time is rescaled with 1/W 2 .2 1.9 −3 10 10 −2 10 −1 10 0 10 1 ΔT/W2 Figure 6.12 as ∆T increases. The time dependence of the velocity ratio Vavg /Vshear is shown in figure 6. Vavg /Vshear . This alternative time scale. Thus diffusion accounts for a large fraction of the measured apparent velocity at short inter-acquisition time. This explains the narrowing of the apparent velocity distributions in figure 6. collapse onto each other under this rescaled time scheme. Note that the velocity ratios. Due to increased shear dominance. it can be observed that the values of ∆T at which the minima of Vavg /Vshear ratio occur depend on the thickness of the imaging range. All data series have P e = 30.65 1.3 Vavg/Vshear 1. then Vavg /Vshear can be interpreted as velocimetry measurement accuracy with Vavg /Vshear = 1 being the ideal measurement. it would be of interest to see if the mean value of the velocity distributions evolve as a function of time as well.5 W=2 W=4 W=6 W=8 W = 10 1.12 and 6.13 demonstrate that both the shape and the mean of an apparent velocity distribution are not time-invariant.12) while the shear-induced flow velocity distribution width scales linearly with ∆T . while the majority of the measure velocity is due to the shear flow at large ∆T .14: Rescaled time evolution of mean particle apparent velocity. ∆X ∼ √ ∆T ⇒ V = 1 ∆X ∼√ . It can be observed that all Vavg /Vshear ratios exhibit the form of a concave-up function of ∆T . can be interpreted as the time it takes for a particle in contact . Minimal velocity ratios for all W ’s occur near an alternative non-dimensional time scale of ∆T /W 2 ≈ 10−1 . ∆T ∆T (6. with a minimum occurring at approximately ∆T = 1. Clearly. the apparent velocity distribution narrows as time increases. if the imaging range is defined at 1 < Z < W . therefore.1 1 0.12.4 1. In their paper Sadr et al. velocity distributions that are significantly wider than that of at large ∆T . The goal of vecimetry measurements is to accurately estimate fluid velocities using the tracer particle apparent velocity as an estimator. ∆T /W 2 . Such change can be explained by the fact diffusive velocity distribution wideth scales with ∆T −1/2 .13. if the apparent velocity distributions evolve with time.

For small ∆T .14. used the Fokker-Planck approach to study the PDF of fluidic planes that an ensemble of particles would have sampled between image acquisitions in nano-PIV. The time dependence of this ratio is in striking resemblance to that shown in figure 6. the turning point in the relative contribution of the two groups of particles (the ones that move closer to the surface and the ones that sample the higher shear plane and return) occurs at ∆T /W 2 ≈ 10−1 .13 and 6.13 that the velocity ratio is P e- . with the wall to diffuse out of imaging range due to Brownian motion. Two potential paths of a particle are shown in this figure. That is. it samples the higher velocities of planes that are out of the imaging range. The larger the ∆T . the physical reasoning described in the previous paragraph suggests that the thickness of the imaging range would be an important parameter.15.15. and thus the mean apparent velocity decreases. Clearly. it should be noted that in both figures 6. the higher probability that particles can sample farther away from the surface and still manage to return to the imaging range at the time of second image acquisition. and increases monotonically afterward? The answer to this question lies in figure 6.14 demonstrates that the velocity ratio Vavg /Vshear also has a scaling similarity. such path 2 illustrated in figure 6.5 and is illustrated in figure 6. Figures 6. Vavg /Vshear → 0 at ∆T → 0 as the instantaneous apparent velocities of particles would be the closest to the local fluidic velocities. Using this non-dimensional time scale. the consequence is that the tracer particles sample toward the lower velocity planes. This suggests that the percentage of particle drop-out and the positions whose fluidic velocities are sampled by the particles play a crucial role in the accuracy of velocimetry measurements. one can observe in figure 6.15 is more likely to be taken than path 2. On the other hand for large ∆T . these contributions of fluidic velocities outside of the imaging range are reflected in the higher values of mean particle apparent velocities. In both cases. Between image acquisitions. In addition.66 path 2 path 1 Imaging range Figure 6. it samples the velocities of the lower planes and thus translates a short horizontal distance. If the particle takes path 1. particles that have taken different paths could have travelled different distances while sampling different velocities along the way. Because they have sampled the fluidic velocities of the higher shear planes. On the other hand if the particle takes path 2.14 can also serve as guides for choosing the appropriate ∆T for near-wall particle tracking velocimetry. and thus translates a larger horizontal distance. which justifies the scaling of non-dimensional time ∆T with 1/W 2 . First. path 1 in figure 6. Indeed Sadr et al. They obtained a ratio of averaged Z value sampled by the particles to the mid-distance of the imaging range. So why is the velocity ratio minimal at ∆T /W 2 ≈ 10−1 . (W + 1) /2.15: Schematic of 2 potential paths for particle translation.14. some of the particles that have left the imaging range are given enough time to return at the second acquisition. figure 6. Since drop-out’s are less significant for small ∆T .9 and 6. the skewed distribution of particles toward the wall suggests that the imaged particles are more likely to move closer to the wall.

4 CONCLUDING REMARKS Brownian dynamics simulation offers an attractive approach to the study of near-wall particle dynamics. velocity distributions evolve as a function of time. with hindered diffusion also contributing to changes in the velocity distribution width and its skewness. PIV will remain accurate as long as the time between acquisition is much smaller than the diffusion time of tracer particles. Furthermore. with results confirmed by experimental data. and the problem can be mathematically closed by specifying a Peclet number and a sedimentation constant. The potential error of both methods would stem from biased sampling of the velocities at various shear planes. By using it as a scaling guide. colloidal particle motion is dominated by shear when Peclet number is greater than 3. These observations suggest that choosing an inter-acquisition time of ∆T /W 2 ∼ O 10−1 would most likely yield measurement inaccuracy. In near-wall colloidal dynamics. In the direction parallel to the wall. PTV and SPTV (Statistical Particle Tracking Velocimetry) [94] are much more reliable measurements because PTV attempts to exclude particle drop-in’s and drop-out’s while SPTV treats the drop-in’s and drop-out’s as statistical noise. It can be clearly observed that the “dip” in velocity ratio depends on the values of W . the incapability of imaging equipments to make consecutive image acquisitions with an extremely short inter-acquisition time would limit the applicability of PIV in near-wall measurements. reporting mean velocities that are time-dependent with a minimum at ∆T /W 2 ≈ 10−1 . An alternative method is to calculate which inter-acquisition time would result in the measured mean velocity being equal to the shear-induced velocity before experiments are conducted. Neither correction method can be applied to PIV measurements. Thus the selection of ∆T such that the velocity ratio would be minimum should be avoided.14 is not W -invariant in regions near the minimum. physical length and time scales are particle radius and diffusive time. where large W values lead to a lower minimal velocity ratios. the correct shear-induced velocity of the particles can be retrieved from the measured mean velocity values. while the velocity ratio shows no dependence on W at both the higher and lower ends of the time scale ∆T /W 2 . As tracer particles get smaller and their diffusion become greater. . hindered diffusion skews the displacement distributions toward to wall. We demonstrate that wall-induced hindered diffusion and translational slow-down of a colloidal particle in a simple shear flow can be accurately captured by Brownian dynamics simulations. This study also provides an assessment tool to the relative accuracy of near-wall PIV and PTV. In the direction perpendicular to the wall.14 offers a way to correct for this bias. 6. On the other hand. Because of a large amount of drop-in’s and drop-outs when nanoparticles are used. the inaccuracy of PIV in near-wall measurement could be significant. Such avoidance is further reinforced by the observation that the velocity ratio in figure 6. figure 6. However.67 invariant except near the minimum where fluctuations exist.

quantum dots have been viewed as the next generation of tracer probes. TIRV can also be applied to experimentally investigate flow characteristics within the gas boundary layer. These quantum dots are small (a few nanometer in diameter). Thus a new breed of tracer particles or molecules in the order of 1 nanometer would be needed. advancement of technology would soon deem particles of 100 nanometer insufficient in probing the near-wall region as scientists would like to look even closer to the solid surface.Chapter 7 Concluding Remarks and Suggested Studies In this dissertation we have presented our experimental studies of aqueous solution boundary slip and anisotropic hindered diffusion. and their usage under evanescent wave imaging settings have been demonstrated by Pouya et al. In fact. and if the solid surface charges would interact with the weak charges carried by the solve molecules. Besides binding. It would certainly be of scientific interest to see if the solvent molecular polarity would have an effect on boundary slip. and described the total internal reflection velocimetry (TIRV) method used in these experiments. thus making them an ideal candidate to probe regions that are even closer to the wall with a variety 68 . Of course. The area of interest include both specific and non-specific binding of suspensions and the solid boundary. Understandably. more uniform in size than polystyrene fluorescent particles. In addition. [100] and Guasto et al. First of all in pursuing a better understanding of boundary slip. [94]. and their chemical properties are vastly different from the non-polar group of solvents. such as ligands and receptors to study blood cell adhesion. Water and similar solvents such as alcohols have a polar molecular configuration. and the TIRV technique is ideal for studying these phenomena. Returning to the liquid phase. such as hexane and toluene. Furthermore. The TIRV technique can certainly be extended to near-surface measurements in a gas phase. Although the existence of boundary slip at a gas/solid interface has been well documented. kinetics of surface-based chemical reactions can also be studied with the same mechanism. these studies open up opportunities for future experimental works in the areas of fluid mechanics and colloidal science. A Brownian dynamics simulation study is also presented to compare the effectiveness and accuracy of TIRV and nano-PIV. particle or cell adhesion and desorption are both areas of great interests in colloidal science and biology. fluorescent aerosols would be required as tracer particles. To investigate specific chemical binding. water is by no means the only test fluid of interest. the TIRV technique offers a method to make direct measurement of the gas slip velocities. the particles and the solid boundary surface can be coated with molecules that bind correspondingly. and available in water or organic solvent soluble forms. and we believe such technology is possible in the near future.

provided that experimental parameters such as camera noise. The Brownian motion of these quantum dots is undoubtedly much more significant and the particle tracking method in the TIRV will be incapable of dealing with the large proportion of particle drop-in’s and drop-out’s between image acquisitions. however. The SPTV thus eliminates the needs to keep track of particle drop-in’s and drop-out’s and to unambiguously matching detected particles. especially in near-surface measurements. . a high seeding density might be desirable to preserve the particle-particle interactions and/or to increase measurement efficiency. An additional limitation of the particletracking based TIRV is its requirement of low particle seeding density to avoid tracking ambiguity. Perhaps a more accurate measurement of the slip velocities can be achieved by using quantum dots in TIRV. They proposed that the true statistics of the desired physical displacement can be extracted by matching each detected particle to all others. and seeding density are known. In some experimental situations.69 of liquids. [94] is a perfect solution to these problems. It is certainly more advantageous than other particle-based velocimetries. The statistical particle tracking velocimetry (SPTV) developed by Guasto et al. particle size and diffusivity. imaging depth.

such measurement is extremely difficult. The best fit line was then extended to φ greater than critical angle to determine incident angle. respectively.2). and this relation should be true for all incident angles despite of total internal reflection.1) where n1 and n2 are refractive indices of glass and air. The refracted beam in air is then projected to a far-away wall or ceiling where the refractive angle can be measured accurately (shown in figure A. x. as shown in figure 2. The illumination laser beam is directed into the objective lens by passing through the a converging lens. The empirical equation was found to accurately predict the lens position of critical angle. it would be ideal if one can measure the beam incident angle directly when determining the penetration depth. the total internal reflection was created at the glass/water interface by directing a laser beam through a microscope objective lens. and sin θ. Knowing the beam incident angle is critical in determining evanescent field penetration depth. The beam is then allowed to refract at a glass/air interface. To determine this relation. In practice. an empirical linear relation between x and sin φ was found (shown in figure A. Using equation (A.1). Thus a geometric relation should exist between the the lens translation distance and the beam incident angle. because the total internal reflection bends the reflected beam back into the objective lens. (A.3).Appendix A Calibration of beam incident angle In our TIRV setup.1. In this appendix. the laser beam is directed through the objective as used in a non-TIRF setup. The bending of the beam is achieved by translating the converging lens [47]. the measurement method of the illumination beam incident angle is presented.3). as shown in figure A. Snell’s law determines the refraction angle by n1 sin φ = n2 sin θ. By examining the data. As demonstrated in equation (2. 70 . we found a linear relationship between the lens position.1.

71 θ Air φ Glass Figure A.1: Schematic of laser beam refraction at the glass/air interface.1. . θ Figure A.2: Photo and schematic of beam incident angle measurement. The refraction angle θ is the same as that of figure A.

8 Data Linear fit 0. .5 10 x (mm) 10.4 0.2 8.72 0.5 sin(φ) 9 9.5 11 Figure A.6 0.33 x + 3. converging lens position.5 0 −0.2 sin(φ) = − 0.3: Plot of laser beam angle vs.

Appendix B

Shear-induced lift force (or more precisely, lack of ) on near-wall submicron-particles
It has been suggested that a shear-induced lift force can act on suspended particles in a linear shear flow, making them more likely to move away from a solid boundary. This would result in a non-uniform distribution of particles near our glass solid surface, where our total internal reflection velocimetry (TIRV) measurements take place. Because our statistical analysis is based on the assumption of uniform particle distribution, the shear-induced lift force would lead to significant inaccuracies. However, we believe that such a lift force is insignificant in our flow regime, and in this appendix, an established theory is presented to support our argument. The subject of lift forces acting on a small sphere in a wall-bounded linear shear flow has been thoroughly studied by Cherukat & McLaughlin [87]. Here we will present only the theory that applies to the flow conditions under consideration. Suppose that a free-rotating rigid sphere of radius a is in a Newtonian incompressible fluid of kinematic viscosity ν and is in the vicinity of a solid wall. In the presence of a linear shear flow, the free-rotating sphere can can travel at a velocity Usph that is different from the fluid velocity, UG , of the shear plane located at its center [38]. We can define a characteristic Reynolds number based on the velocity difference, Us = Usph − UG , with Res = Us a . ν (B.1)

Another characteristic Reynolds number based on shear rate can be defined as ReG = Ga2 , ν (B.2)

where G is the wall shear rate. In this geometry, the wall can be considered as located in the ”inner region” of flow around the particle if Res κ and ReG κ2 , where κ ≡ a/h and h is the distance between the particle’s center and the wall. For near-wall particle velocimetry described in chapters 3 and 4, Res ∼ ReG ∼ 10−7 while κ ∼ O (1), and thus the inner region theory of lift force applies. For a flat wall located in the inner region of flow around a free-rotating particle, the lift force, FL , which is perpendicular to the wall, is scaled by [87] FL ∼ Res · I, (B.3)

73

74 where I is a coefficient that can be numerically estimated by I = 1.7631 + 0.3561κ − 1.1837κ2 + 0.845163κ3 − 3.21439 + 2.6760 + 0.8248κ − 0.4616κ2 κ ReG Res +
2

(B.4) .

1.8081 + 0.879585κ − 1.9009κ2 + 0.98149κ3

ReG Res

Again for the velocimetry conditions described in chapters 3 and 4, I ∼ O 102 . Therefore FL ∼ Res · I 10−7 102 1, (B.5)

and the lift force acting on near-wall particles is insignificant and can be neglected for all practical purposes.

Appendix C

Evanescent wave image of micron-sized fluorescent particles
Imaging of fluorescent nanoparticles under TIRFM have been well reported by many groups, particularly in the exponential decay relation between particle intensity and its distance to the substrate surface [44, 85, 90]. But for micron-sized fluorescent particles in evanescent field, there has been no report on their imaging characteristics. In this appendix, numerical simulations were performed to investigate some of the imaging characteristics associated with fluorescent particles under evanescent wave imaging. The numerical simulation was conducted in the commercially packaged COMSOL Multiphysics software environment. The simulation geometry is simplified from three-dimensional to twodimensional but is still a representative of near-surface imaging, and shown in figure C.1. Polystyrene particles (refractive index 1.59) of three different diameters, d/λ = 0.39, 6 and 12 where λ is the incident beam wavelength, are immersed in the water phase (refractive index 1.33) and placed in the vicinity of the glass phase (refractive index 1.51). A Gaussian beam enters from the lower-left boundary and reflects off the water/glass interface at a 64.54◦ angle, while scattering boundary conditions are applied to all other external boundaries. Simulations were repeated for various gap sizes, h, between the particle and the glass, with h/λ ranging from 0 to 2. The effectiveness of the COMSOL Multiphysics package in modelling total internal reflection and evanescent field was first demonstrated in the absence of a polystyrene particle, and is shown in figure C.2. The simulation model perfectly captures total internal reflection and an exponential decay of evanescent energy density in the water phase is found to be in agreement with theories (shown in figure C.3). Time-averaged energy densities of particles in evanescent field are shown in figures C.4, C.5 and C.6, respectively. In figure C.4, it can be observed that the presence of a particle with diameter d/λ = 0.39 does not distort the energy field. In contrast, larger particles such as d/λ = 6 and 12 can distort the energy field quite significantly. Their presence can be seen to create a shadow in the reflected beam. These lost energy leaks into the interior of the large particles through a “frustrated-total internal reflection mechanism”, illuminating regions beyond the penetration depth of evanescent wave. These energy continue to reflect and refract at the polystyrene/water interface, resulting in a high energy density in the lower-right side of the particle, while some additional energy escapes into the right half of the water phase. Using the subdomain integration function of COMSOL, one can find out the overall energies inside the polystyrene particles. This quantity is important because in TIRFM experiments, only the light emitted by the fluorophores inside the polystyrene particles is imageable, and the fluorophores are excited only by the light energy that enters the particles. The overall energies that 75

except the presence of a suspended particle. a total internal reflection occurs at the glass/water interface. Scattering (or floating) boundary condition is applied to all external boundaries at the water and glass phases.2: COMSOL simulation of total internal reflection. The geometry of this simulation is identical to that of figure C.1: Schematic of COMSOL simulation geometry. Figure C. .76 Polystyrene Particle Water Phase Glass Phase Incident Gaussian Beam Figure C. Because the incident angle of the Gaussian beam is 64. Plotted in the figure is time-averaged total energy density of total internal reflection at a glass/water interface. A polystyrene particle is suspended in the water phase and located in the vicinity of glass’interface. The particle shown in the figure has a diameter of d/λ = 6. except at the boundary inside the glass phase where the incident Gaussian beam enters.54◦ .1.

except that the particle diameter is with d/λ = 0.77 1 Simulation Theory Normalized Intensity 0.39 in evanescent field.4: COMSOL simulation of a particle with d/λ = 0. Plotted in the figure is time-averaged total energy density.4 0.39 and the gap size between the particle and the glass surface is h/λ = 0. and the geometry of this simulation is identical to that of figure C. Close agreement between simulation results and theoretical calculations is obtained.2 0 0 0.8 0.1.5 2 Figure C. .6 0. Particle Figure C.5 1 h/λ 1.3: The exponential intensity decay of evanescent field in figure C.1.2.

1. Plotted in the figure is time-averaged total energy density. except that the gap size between the particle and the glass surface is h/λ = 0. except that the particle diameter is d/λ = 12 and the gap size between the particle and the glass surface is h/λ = 0.5: COMSOL simulation of a particle with d/λ = 6 in evanescent field. Figure C. .1.1. Plotted in the figure is time-averaged total energy density. and the geometry of this simulation is identical to that of figure C. and the geometry of this simulation is identical to that of figure C.6: COMSOL simulation of a particle with d/λ = 12 in evanescent field.78 Figure C.1.

39) follows an exponential decay function that is identical to the evanescent energy decay. This forms the basis of using the overall intensity of nanoparticles for 3D displacement tracking. does not follow the same exponential decay function of the evanescent field. However. The intensities are normalized by the intensity observed when the particle is in contact with the glass surface. however. the overall intensity of a small particle (d/λ = 0. such a convolution is computationally complicated and expensive.5 2 Figure C. a is the radius of the objective lens and f is the objective focal length. high magnitude objective whose focal depth.7: Overall evanescent energy in the suspended particles at various gap sizes.2) Here λ is the wavelength. δ.7. The evanescent field intensity is obtained from the cross-sectional time-averaged total energy density profile of figure C. (C. enter the particles at gap sizes h/λ ranging from 0 to 2 is shown in figure C. As it has been demonstrated by Huang et al. For a point light source that is a distance z off the focal plane. [85] and Sarkar et al. The overall intensity of larger particles. it’s intensity is [101] sin (u/4) 2 I (z) = I0 (C.79 1 0.7.2 0 0 d/λ = 0. This is probably due to the fact that a larger radius of curvature provides more surface area for light to leak into the particles through the frustrated-total internal reflection mechanism. A simpler way is to look at the intensity profile when the fluorescent energy is projected onto the focal plane.6 0. A more relative question is what will a fluorescent particle with diameter greater than the illumination wavelength look like under a high numerical aperture (NA).1) u/4 where I0 is the point source emission intensity and u= 2π λ a f 2 z. [90]. One of the ways to find out is to convolute the energy density inside the particles with diffraction equation and point spread function.39 d/λ = 6 d/λ = 12 Evanescent Field Normalized Intensity 0. In . with which one could still gain some qualitative understanding of the image. particularly at small gap sizes.4 0. a near-exponential dependency of large particle intensity on the gap size is observed in figure C.8 0. Still.5 1 h/λ 1.2. is roughly δ/λ = 2.

The intensity projection of a particle with d/λ = 6 is shown in figure C. The direction of the evanescent wave propagation is toward positive x/λ. z) · zdz (C. a/f is commonly referred to as the f-number of lens and for a microscope objective. which is important in microscopy.8 as blue squares. I0 is the peak intensity and x/λ = 0 corresponds to the particle geometric center. the intensity profile from actual imaging of a particle appears to be wider than that of the simulated result. z) is calculated using equation (C.4 0. This leads to physical inaccuracy as we will demonstrate subsequently. This projection is obviously 0 a simplified calculation as it assumes negligible diffraction. the . The red dotted line. represents the projected intensity profile based on results of COMSOL simulation of a particle of the same size and equation (C.9. it is assumed that both a and f are much larger than λ. A peak intensity near the center can be clearly observed in the image. This is due to the fact that the computation leading to the projection (red dotted line) in figure C.8 0.3). For comparison. However. A TIRFM image of a 3-µm particle (equivalently d/λ = 6) is shown in figure C. its projected intensity at point x = x0 on the z = z0 plane is [101] 2a−b Ip (xo ) = b I (x0 .80 1 0.8 as a red dotted line. Therefore. however. for a particle in contact of the wall and its center is located at x = 0 and z = a.2 0 −4 −2 0 2 4 x/λ Figure C.1). In the projection.9 is also shown in figure C. while a longer tail is observed in the direction of the laser beam propagation. This intensity profile is similar in shape when compared to that of red dotted line from simulation. the intensity exhibits a peak value near the particle center. getting equation (C.3) where b = a− a2 − x2 and I (x0 .9.6 I/I0 0. it is equal to the numerical aperture [73]. A peak intensity is found near the particle center. while a slightly longer tail is found in the direction of evanescent wave propagation. Nevertheless. where the focal plane is set at the glass surface.8 neglects diffraction of light.8: Cross-sectional intensity profile of a 3-µm particle (d/λ = 6). The blue squares represent the cross-section intensity profile indicated as a red line in figure C. a cross-sectional intensity profile of figure C.1).

3) is narrower than that of an actual image because of the neglected diffraction in numerical integration.9: Digital image of a 3-µm particle (d/λ = 6) in contact with glass surface. toward the right tip. .10. we can again see that there is a bright peak near its center. In addition. blue squares). while the asymmetric tail on the right side is even more pronounced. However.81 1 µm Figure C.1) and (C. Again. The intensity profile along the red cross-section shown is plotted in figure C. With a particle of this size. which can be also be observed in the right tail of the red simulation intensity profile in figure C. the intensity profile obtained from the simulation results and equation (C. The same imaging projection is also performed for a 6-µm particle (d/λ = 12) and shown in figure C. as observed in figure C.8.3). there are small.10.11) as well as on a cross-sectional intensity profile (figure C. the modelling results and the actual image show qualitative agreement despite the simplicity of the integration model in equations (C. resemblance of of the two profiles is striking and the profiles are in qualitative agreement. alternating regions of bright spots. A similar observation can be made on an actual image of a 6-µm particle (figure C.8.10. A large bright center is clearly visible while the right half of the particle is significantly brighter.

8 I/I 0 0. The direction of the evanescent wave propagation is toward positive x/λ.11. The blue squares represent the cross-section intensity profile indicated as a red line in figure C.11: Digital image of a 6-µm particle (d/λ = 12) in contact with glass surface.3). I0 is the peak intensity and x/λ = 0 corresponds to the particle geometric center. however. 2 µm Figure C. The intensity profile along the red cross-section shown is plotted in figure C.2 0 −6 −4 −2 0 2 4 6 x/λ Figure C. The red line.10: Cross-sectional intensity profile of a 6-µm particle (d/λ = 12). .10.6 0.4 0.82 1 0. represents the projected intensity profile based on results of COMSOL simulation of a particle of the same size and equation (C.

Electrophoresis. [7] Eric Lauga. Olbricht. 2002. Science. Katarzyna Jaszczolt. and William R. and Peter O’Brien. [8] Bin Zhao. [2] Samuel K. Sia and George M. 291:1023–1026. Principles of surface-directed liquid flow in microfluidic channels. Analytical Chemistry. Tropea and A. Handbook of Experimental Fluid Dynamics. Mark Saltzman and William L. Microfluidics: The no-slip boundary condition. Foss anc C. [9] Bin Zhao. [10] Zbigniew Adamczyk. 2002. Michael P. Physics and applications of microfluidics in biology. 120:11155– 11162. Kirat Singh. Stone. Surface-directed liquid flow inside microchannels. Mensing. 12:486–494.Bibliography [1] Adam Bange. Glennys A. 2002. Alan Morrison. and David J. 1996. 24:3563–3576. Nature Reviews. Jeffrey S. Microfluidic immunosensor systems. Beebe. Brian Halsall. Heineman. Building drug delivery into tissue engineering. Development of a microfluidic device for fluorescence activated cell sorting. [3] Jan Kruger. Moore. 2001. 56:381–410. [4] W. Quarterly Journal of Mechanics and Applied Mathematics. 2002. Langmuir. 2005. [6] George E. Hammer. Journal of Chemical Physics. Karniadakis and Ali Beskok. Microfluidic devices fabricated in poly(dimethylsiloxane) for biological studies. 74:4259–4268. Journal of Micromechanics and Microengineering. 2002. 2004. Influence of direction and type of applied force on the detachment of macromolecularly-bound particles from surfaces. Chaoui and F. Feuillebois. New York. [5] David J. Springer. In J. Yarin. 4:261–286. [11] Kai-Chien Chang and Daniel A. 2003. Irreversible adsorption of particles at random-site surfaces. Beebe. Barbara Siwek. Creeping flow around a sphere in a shear flow close to a wall. 20:2488–2503. 83 . Springer. chapter 15. editors. 1:177–186. Moore. [12] M. Micro flows: fundamentals and simulation. 2003. Beebe. Annual Review of Biomedical Engineering. Carl Jackson. Jeffrey S. Biosensors and Bioelectronics. and David J. and Howard A. H. 12:2271–2282. Whitesides. Walker. and Glenn M. and Pawel Weronski. 2006. Brenner. Alan O’neill.

B. . Lawrence. 88:106102. 2003. J. Evidence of shear-dependent boundary slip in newtonian liquids. [19] K. Kenneth S. 56:221–251. Westin. 1995. Troian. 2003. Nature. Journal of Physics: Condensed Matter. R. Steinberger. V. Limites of the hydrodynamic no-slip boundary condition. Slippage of liquids over lyophobic solid surfaces. V. [22] Peter A. Apparent slip flows in hydrophilic and hydrophobic microchannels.84 [13] Poppo J. D. Boundary slip on smooth hydrophobic surfaces: intrinsic effects and possible artifacts. 1969. Johan A. Polymer depletion-induced slip near an interface. Boundary conditions at a fluid-solid surface. Briscoe. Effects of particle size on near-wall depletion in mono-dispersed colloidal suspensions. 82:4671–4674. Neto. Kazarian. J. Journal of Applied Physics. and Henk J. [28] C. [14] P. Physical Review Letters. The European Physical Journal E. Journal of Colloid and Interface Science. Slippage of water over nonwettable surfaces. S. and D. 1984. and E. Large slip effect at a nonwetting fluid-solid interface. Jaap Noordmans. [27] C. [17] Howard A. J. 97:574–581. Breuer. Johan A. 15:2620–2626. van der Mei. Craig. [24] Marek Cieplak. 2005. [26] YingXi Zhu and Steve Granick. Bruce Caswell. N. 280:511–517. Journal of Colloid and Interface Science. Charlaiz. 27:1149–1152. Albert Poortinga. A review of the slip (wall depletion) of polymer solutions. 389:360–362. 1:1–29. Physics of Fluids. Physical Review Letters. [21] N. A general boundary condition for liquid flow at solid surfaces. Liquid transport properties in submicron channel flows. A. and Jayanth R. A. Williams. Busscher. Sobolev. 12:S71–S74. Zhiqiang Cao. Richardson. S. Annual Review of Fluid Mechanics. 94:056102. V. 17:L9–L14. Physical Review Letters. In Proceedings of 2001 ASME International Mechanical Engineering Congress and Exposition. 1999. and Merwin Sibulkin. [16] Yoram Cohen and A. M. Barnes. Cross. Henry C. 2005. Banavar. Deposition of polystyrene particles in a parallel plate flow chamber under attractive and repulsive electrostatic conditions. 15:2897–2902. [15] Remco Tuinier and Takashi Taniguchi. Peter D. Churaev. 29:67–102. [18] Sydney Goldstein. and C. 86:803–806. and A. Somov. Journal of Non-Newtonian Fluid Mechanics. Peter Huang. Physical Review Letters. [23] Jean-Louis Barrat and Lyderic Bocquet. [20] Erhard Schnell. G. [25] Chang-Hwan Choi. 2002. 1997. B. Breuer. emulsions and particle suspensions in viscometers: its cause character and cure. Cottin-Bizonne. 1956. Joel Koplik. 1999. Langmuir. Metzner. Apparent slip flow of polymer solutions. Fluid mechanics in the first half of this century. Chang-Hwan Choi. 1985. Journal of Rheology. 2001. Hartman Kok. J. Wit. 2001. Westin. B. Thompson and Sandra M. and Kenneth S. 2004.

113:1228–1236. Banerjee. [40] M. J. 2:221–227. 85:980–983. and H. The slow motion of a sphere through a viscous fluid towards a plane wall. J. Kihm. J. 2004. Tretheway and Carl D. Prieve. Physical Review E. Flow profile near a wall measured by double-focus fluorescence cross-correlation. 2002. Nature Materials. 2003. Near-wall hindered brownian diffusion of nanoparticles examined by three-dimensional ratiometric total internal reflection fluorescence microscopy (3-d r-tirfm). [35] Eric Lauga. 62:3909–3919. Physical Review E. Simultaneous investigation of sedimentation and diffusion of a single colloidal particle near an interfce. [36] Howard Brenner. 2004. G. 22:653–660. Hosoda. Cox. and Hiroshi Mizunuma. I. 2000. 2003. Feuillebois. [32] Keizo Watanabe. 1961. R. Apparent slip due to the motion of suspended particles in flows of electrolyte solutions. Takagi. Chemical Engineering Science. Leger. 1967. Hindered diffusion of colloidal particles very near to a wall: revisited. K. Jonathan Yu. [41] Binhua Lin. [30] Derek C. 20:8924–8930. Yanuar. Brenner. Chemical Engineering Science. and K. 71:035303(R). Experiments in Fluids. . [31] Pierre Joseph and Patrick Tabeling. Direct experimental evidence of slip in hexadecane: solid interface. K. O.85 [29] R. and H. Physical Review Letters. [43] Arindam Banerjee and Kenneth D. Pit. Brenner. 72:042101. 41:525–529. Gansen. 16:242–251. and Stuart A. Kihm. F. [44] K. Vinogradova. Walz. Oetama and John Y. R. 67:056313. Slow viscous motion of a sphere parallel to a plane wall . Experimental verification of near-wall hindered diffusion for the brownian motion of nanoparticles using evanescent wave microscopy. D. Measurement of anisotropic brownian motion near an interface by evanescent light-scattering spectroscopy.ii: Couette flow. Sakai. and T. Journal of Chemical Physics. 124:164713. A. Physical Review E. Langmuir. Goldman. 2006. [37] A. Bevan and Dennis C. Meinhart. Cox. Radler. [42] Ratna J. [34] Steve Granick. 37:811–824. G. H. C. Takagi. Direct measurements of constrained brownian motion of an isolated sphere between two walls. Physics of Fluids. 1998. A. A. 2005. 58:6275–6280. Physical Review E. JSME International Journal. Yingxi Zhu. [39] Michael A. Physical Review E. Best. Slow viscous motion of a sphere parallel to a plane wall . 1998. 2005. 22:637–651. [33] D. and O. Slippery questions about complex fluids flowing past solids. Apparent fluid slip at hydrophobic microchannel walls. 2000. and L. Chemical Engineering Science. 14:L9–L12.i: motion through a quiescent fluid. Hervet. Direct measurement of the apparent slip length. 1967. Choi. Lumma. 2000. and Hyunjung Lee. Slip of newtonian fluids at solid boundary. The Journal of Chemical Physics. Goldman. Rice. [38] A.

G. Mucha. 34:115–121. 1994. Flicker. [50] H. Diffusioninduced bias in near-wall velocimetry. and Stacy G. 2005. and Minami Yoda. Experiments in Fluids. Wereley. 81:25–42. 1993. Graham Wigley. 1978. P. J. Biosensors and Bioelectronics. . Micron-resolution particle image velocimetry. Jennifer L. M. 25:78–86. Use of attenuated total internal reflection-fourier transform infrared spectroscopy to investigate the adsorption of and interactions between charged latex particles. Ermak and J. George P. [51] Scott G. L. 2:764–774. Neefjes. Journal of Colloid and Interface Science. editor. Zettner and M. [52] Robert Kun and Janos H. Experiments in Fluids. Buchhave. McCammon. Journal of Physical Chemistry. and R. 41:185–194. 2001. and C. Journal of Chemical Physics. From fixed to frap: measuring protein mobility and activity in living cells. Particle velocity field measurements in a near-wall flow using evanescent wave illumination. [60] Anne Pierres. Adrian.A. 3:E145–E147. Methods. 2003. Optical diagnostics for flow processes. [56] C. 25:316–319. D. Anderson. Measurement of surface charge densities on brownian particles using total internal reflection microscopy. Christel Hohenegger. pages 247–270. Brownian dynamics with hydrodynamic interactions. 2006. J. Submitted to Journal of Fluid Mechanics. 114:10094–10104. Nature Cell Biology. D. 2006. [53] P. Cheng Zhu. Wereley and C. D. [57] Haifeng Li. Peter J. [46] Chris Rowe Tiatt. 2005. and Preben Buchhave. S. [55] J. Meinhart. editors. Inc. In Methods in Cell Biology. and Minami Yoda. and Frances S. 158:317–325. Beebe. Biophysical Journal. Bechinger. [48] Eric A. volume 30. [49] Taekjip Ha. Tipa. Academic Press. Total internal reflection fluorescence microscopy in cell biology. [54] S. Santiago. Springer. In K. Plenum Press. Total internal reflection fluorescence microscopy. Reits and Jacques J. 2001. 108:3462–3468. New York. Experiments in Fluids. Journal of Chemical Physics. 2001. J. pages 245–270. Quantifying double-layer repulsion between a colloidal sphere and a glass plate using total internal reflection microscopy. 20:2470–2487. 1998. 2001. 2004. Haifeng Li.86 [45] Daniel Axelrod. Breuer. Fendler. [59] Reza Sadr. In Lars Lading. Traffic. Microscale Diagnostic Techniques. H. Single-molecule fluorescence resonance energy transfer. C. Meinhart. Reza Sadr. A particle image velocimetry system for microfluidics. Diffusion of microspheres in shear flow near a wall: use to measure binding rates between attached molecules. Yoda. and Pierre Bongrand. Leiderer. 69:1352–1360. [58] Donald L. Helden. T. T. Particle image velocimetry. [47] Daniel Axelrod. pages 51–112. chapter 9. Multilayer nano-particle image velocimetry. 1989.. Ligler. Evanescent wave fluorescence biosensors. Bike. 2001. Anne-Marie Benoliel. von Grunberg.

[68] J. Cooper McDonald. 1996. Haifeng Li. Ovryn. Brownian dynamics simulation of the motion of a rigid sphere in a viscous fluid very near a wall. 2003. Experiments in Fluids. [66] Jeffrey S.87 [61] H. 2001. 19:307–325. [65] N. Lighting up the cell surface with evanescent wave microscopy. [67] Derek Toomre and Dietmar J. Olivier. 2005. Olivier J. Langerholm. [63] Reza Sadr. and Minami Yoda. Three-dimensional forward scattering particle image velocimetry applied to a microscopic field-of-view. [75] J. Validation of an analytic solution for depth of correlation in microscopic particle image velocimetry. Physical Review Letters. 70:4974–4984. 77:1897–1900. 2001. [76] S. Manstein. 29:S175–S184. Sholl. Analytical Chemistry. Thompson and B. Physical Review E. 2000. Biomaterials. [74] C. [64] YingXi Zhu and Steve Granick. Truskey. and George M. Reichert. 113:9268–9278. [71] B. and George A. The theory of diffraction-limited resolution in microparticles image velocimetry. Bourdon. [72] R. Measurement Science and Technology. C. Behrens and D. . 1998. Physical Review Letters. When like charges attract: The effects of geometrical confinement on long-range colloidal interactions. Adrian and C. Application of total internal reflection fluorescence microscopy to study cell adhesion to biomaterials. Crocker and D. 24:44–52. 291:28– 36. G. Michael K. Current Opinion in Biotechnology. D. Applied Physics Letters. A. [62] David S. J. Yao. Out-of-focus effects on particle image visibility and correlation in microscopic particle image velocimetry. Olsen and R. Evanescent wave doppler velocimetry for a wall’s near field. C. 8:58–64. G. Wereley. Rate-dependent slip of newtonian liquid at smooth surfaces. 1998. Schueller. 38:90–98. The pair interaction of charged colloidal spheres near a charged wall. N. 2000. Yamada. Grier. Lauri A. J. Edward Atman. Prieve. Measurement Science and Technology. Unni and C. Gorby. 1997. 15:318–327. Fenwick. 64:050401. M. Pulsed laser technique application to liquid and gaseous flows and the scattering power of seed materials. M. Applied Optics. J. Duffy. J. 1999. L. Whitesides. 11:298–303. [69] David C. Grier. Rapid prototyping of microfluidic systems in poly(dimethylsiloxane). G. S. Adrian. 2000. 14:1047–1053. G. Journal of Chemical Physics. Brownian dynamics simulation and experimental study of colloidal particle deposition in a microchannel flow. Burmeister. Olsen. W. 87:096105. Meinhart and Steve T. Experiments in Fluids. Impact of hindered brownian diffusion on the accuracy of particle-image velocimetry using evanescent-wave illumination. 1985. Experiments in Fluids. 75:1805–1806. 29:S166–S174. Journal of Colloid and Interface Science. Total internal reflection fluorescence: applications in cellular biophysics. [73] M. 2005. and A. H. Yang. 2001. Trends in Cell Biology. 2004. and Dennis C. [70] Carl D.

In preparation. . 263:1–18. 2001. 2004. To slip or not to slip . The inertial lift on a rigid sphere in a linear shear flow field near a flat wall. L. 81:2378–2388. 2005. Proceedings of the National Academy of Sciences. S. Zhang. and K. 2006. Ragan B. Quantitative comparison of algorithms for tracking single fluorescent particles. 1994. Walker. Prieve. McLaughlin. Zhang. [82] Eric Lauga and Michael P. Journal of Fluid Mechanics. William F. Oxford University Press. [78] Gyoko Nagayama and Ping Cheng. and William H. Fernandez. 37:825–833. and Elisabeth Charlaiz. 2:237–240. Oxford. Measurement of the inertial lift on a moving sphere in contact with a plane wall in a shear flow. Journal of Fluid Mechanics. 1993. Lou. Breuer. [91] Richard A. Yoo. Breuer. Hindered diffusion of a single sphere very near a wall in a nonuniform force field. Langmuir. J. Breuer. 20:3813–3815. [84] Michael K. [80] S. Y. Physical Review E. 2004. and Julio M. 566:447–464. Xiao D. Physics of Fluids. Simulations of hindered diffusion in shear flow and its implications for near-wall velocimetry. Breuer. [87] Pradeep Cherukat and John B.88 [77] T. International Journal of Heat and Mass Transfer. England. 101:12882–12886. and Kenneth S. [88] Nasser A. Dynamic mechanisms for apparent slip on hydrophobic surfaces. 2004. Robertson. Jean-Louis Barrat. pages IMECE2002–33707. Langmuir. King and David T. J. Jeffrey S. Shi T. 1997. M. [90] Atom Sarkar. Cheezum. Zhi X. Molecular dynamics study of the effect of atomic roughness on the slip length at the fluid-solid boundary during shear flow. Westin. 98:7552–7564. [86] Michael R. Effects of interface wettability on microscale flow by molecular dynamics simulation. Experiments in Fluids. 47:501– 513. [81] Chang-Hwan Choi. 2004. Park. Jie. Near-surface velocimetry using evanescent wave illumination. Guilford. Frej and Dennis C. 2004. 2002. and Jun Hu. Jin. Zhang. Lyderic Bocquet.water flows in hydrophilic and hydrophobic microchannels. Leighton Jr. and Kenneth S. Jones. Nature Materials. Direct measurement of slip velocities using three-dimensional total internal reflection velocimetry. Soft Condensed Matter. P. Biophysical Journal. Journal of Chemical Physics. [83] Xue H. Simultaneous atomic force microscope and fluorescence measurements of protein unfolding using a calibrated evanescent wave. 2004. [85] Peter Huang. K. In Proceedings of IMECE 2002. Degassing and temperature effects on the formation of nanobubbles at the mica/water interface. 2006. 70:026311. Galea and Phil Attard. [89] Peter Huang and Kenneth S. Guasto. 2002. Sun. L. Brenner. 20:3477–3482. Huang. Lowfriction flows of liquid at nanopattened interfaces. 9:1248–1255. [79] Cecile Cottin-Bizonne. Johan A.

Squires. [99] E. reflecting and absorbing walls. F. 2005. 2002. [97] Eric Lauga and Todd M. O. M. Direct measurement of the effect of surface roughness on the colloidal forces between a particle and flat plate. J. Efficient brownian dynamics simulation of particles near wall. 2006. A. 39:784–786. 2002. . M. Statistical particle tracking velocimetry using molecular and quantum dot tracer particles. 66:056702. i. Manoochehr Koochesfahani. Direct measurement of anisotropic near-wall hindered diffusion using total internal reflection velocimetry. Moungi Bawendi. Journal of Colloid and Interface Science. Breuer. Preston Snee. Peters and Th. Cambridge University Press. M. sticky walls. Peters and Th. [95] Peter Huang and Kenneth S. 17:103102. Physical Review E. 56:31–60. [94] Jeffrey S.89 [92] Lakkapragada Suresh and John Y. and Daniel Nocera. A. 2005. i. 2006. Principles of Optics: Electromagnetic theory of propagation interference and diffraction of light. Physical Review E. Barenbrug. Statistical reconstruction of velocity profiles for nano particle image velocimetry. A. Barenbrug. Mucha. [98] E. in press. Peter Huang. Experiments in Fluids. Single quantum dot (qd) imaging of fluid flow near surfaces. Physics of Fluids. 1999. and Kenneth S. Guasto. 1997. 66:056701. 6th edition. O. [101] Max Born and Emil Wolf. Vinogradova. Breuer. [100] Shahram Pouya. 2006. A. Walz. 1980. Slippage of water over hydrophobic surfaces. [96] Christel Hohenegger and Peter J. To be published. M. 196:177–190. [93] Olga L. Brownian motion near a partial-slip boundary: A local probe of the no-slip condition. F. J. Internal Journal of Mineral Processing. In preparation. Efficient brownian dynamics simulation of particles near wall.