Dissertation SwapNil KoHale

MOLECULAR HYDRODYNAMICS IN COMPLEX FLUIDS
by

Swapnil C. Kohale
Bachelor of Chemical Engineering

A Dissertation

In

CHEMICAL ENGINEERING

Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of

DOCTOR OF PHILOSOPHY

IN

CHEMICAL ENGINEERING

December, 2009

Texas Tech University, Swapnil C. Kohale, December 2009
ii
TABLE OF CONTENTS

ABSTRACT .. v
LIST OF TABLES .. vi
LIST OF FIGURES vii

I. INTRODUCTION AND OVERVIEW..................................................................... 1
Hydrodynamics in nanoscale systems .............................................................................. 1
Nanofluidics...................................................................................................................... 2
What is nanofluidics and what is its significance? ....................................................... 2
Applications of nanofluidic devices.............................................................................. 3
Experimental studies using nanofluidic devices ........................................................... 6
Single molecule analysis........................................................................................... 6
Separation ................................................................................................................. 8
Molecule concentration............................................................................................. 9
Particle Micro and Nanorheology................................................................................... 10
Molecular dynamics simulations .................................................................................... 12
Organization.................................................................................................................... 13
II. CROSS-STREAM CHAIN MIGRATION IN NANOFLUIDIC CHANNELS... 15
Introduction..................................................................................................................... 15
Experimental findings................................................................................................. 15
Theoretical findings .................................................................................................... 16
Computational investigations...................................................................................... 16
Chain migration in nanofluidic channels .................................................................... 17
Simulation Method.......................................................................................................... 18
Simulation sets............................................................................................................ 21
Weissenberg number................................................................................................... 22
Results............................................................................................................................. 24
Effect of chain length.................................................................................................. 24
Effect of channel height .............................................................................................. 28
Effect of concentration................................................................................................ 31
iii
Effect of intermolecular interactions .......................................................................... 37
Summary and Discussion................................................................................................ 39
III. MOLECULAR HYDRODYNAMICS IN NANOPARTICLE SUSPENSIONS:
SMOOTH NANOPARTICLES....................................................................................... 42
Introduction..................................................................................................................... 42
Continuum treatment .................................................................................................. 43
Molecular simulation studies: Stokes-Einstein law.................................................... 43
Validity of Stokes law................................................................................................. 44
Cooperative hydrodynamic effects ............................................................................. 46
The smooth nanoparticle............................................................................................. 49
Simulation systems ..................................................................................................... 50
Fluid viscosity............................................................................................................. 52
Results............................................................................................................................. 53
Effect of cooperative hydrodynamic interactions ....................................................... 58
Effect of slip at nanoparticle surface .......................................................................... 62
Effect of confining surfaces ........................................................................................ 66
IV. MOLECULAR HYDRODYNAMICS IN NANOPARTICLE SUSPENSIONS:
FRICTION FORCE AND TORQUE ON A ROUGH NANOPARTICLE................. 72
Introduction..................................................................................................................... 72
Rough nanoparticle generation ................................................................................... 74
Results............................................................................................................................. 79
Friction force and torque on a rough nanoparticle in monomeric solvent .................. 81
Friction force on a rough nanoparticle in a polymer melt .......................................... 88
Concentrated suspension of nanoparticles .................................................................. 91
V. ACTIVE NANORHEOLOGY IN POLYMER MELT......................................... 95
Introduction..................................................................................................................... 95
Microrheology............................................................................................................. 96
iv
Types of microrheological techniques .................................................................... 97
Passive Microrheology........................................................................................ 97
Active Microrheology......................................................................................... 98
Nanorheology............................................................................................................ 100
Simulation Method........................................................................................................ 101
System preparation: Isotropic orientation of polymer chains ................................... 103
Mathematical formulation for active nanorheology ..................................................... 104
Magnitude of inertial effects: Dimensional analysis .................................................... 107
Results........................................................................................................................... 109
Literature studies for comparison ............................................................................. 112
Results: Viscoelastic properties from Active Nanorheology simulations ................ 113
Viscoelastic moduli at different locations................................................................. 115
Slip at the nanoparticle surface................................................................................. 117
Active nanorheology simulations in the absence of slip........................................... 120
Summary and Discussion.............................................................................................. 125
Momentum diffusion time Vs. Period of oscillation ................................................ 126
Dependence on amplitude of oscillation................................................................... 127
VI. CONCLUSIONS AND FUTURE WORK............................................................ 129
VII. APPENDICES......................................................................................................... 133
APPENDIX A............................................................................................................... 133
APPENDIX B............................................................................................................... 134
APPENDIX C............................................................................................................... 139
APPENDIX D............................................................................................................... 144
VIII. BIBLIOGRAPHY ................................................................................................ 148

v
ABSTRACT

Developments in the field of micro- and nanofluidics have renewed interest in the
study of molecular hydrodynamics in confined geometries. The conception and design of
these devices demand accurate knowledge of transport properties of fluids in these small
geometries. The continuum methodologies for the calculation of transport properties of
fluids in the bulk are well established. However, the assumptions made in the continuum
treatment need to be valid for the nanoconfined fluids. In this work molecular dynamics
simulation technique is used to study the hydrodynamic behavior of complex fluids in
nanoconfinement. The goal of the dissertation is to study the hydrodynamic behavior of
fluids in confinement using molecular simulations and to make a quantitative connection
with the continuum predictions.
Shear induced cross stream chain migration in a flowing polymeric solution is studied in
the first part of this dissertation. The role of hydrodynamic interactions and the effect of
chain conformational properties, interactions, channel geometry and chain concentration on
the chain migration phenomenon were studied. Next, molecular hydrodynamics in a
nanoparticle suspension were studied to quantify the effect of confinement, boundary
conditions at the nanoparticle surface and cooperative hydrodynamic interactions between
nanoparticles. Finally, a new technique, termed as active nanorheology, is also presented to
calculate the nanoscale local viscoelastic properties of the complex materials using molecular
dynamics simulations.

vi
LIST OF TABLES

Table 3.1: A matrix of simulation systems used in chapter 3 ................................................ 51
Table 3.2: Comparison of normalized friction force for WCA and LJ interactions. ............. 65
Table 4.1: A table showing the simulation system sizes used in chapter 4............................ 83
Table 5.1: Table of relative magnitudes of the three ratios of coefficients for the frequencies
studied in chapter 5 ............................................................................................................... 109

vii
LIST OF FIGURES

Figure 1.1: A schematic of the cross section of a -hemolysin channel embedded in a lipid
bilayer[49]................................................................................................................................. 7

Figure 2.1: A schematic of the simulation system showing polymer chain, fluid atoms and
the wall atoms. The Couette flow is generated by moving the two walls in opposite direction
with same velocity. ................................................................................................................. 19

Figure 2.2: Chain center of mass density profiles for chains of length 20 (lines with open
symbols) and 4 (filled symbols only) as a function of distance from the wall. Simulation
conditions are: Couette flow (shear rates = 0.05 and 0.075), = 0.8, T
wall
= 0.9 and WCA
interactions. The ratios of channel height (H) to the bulk chain radius of gyration (R
g
) are:
8.1 (N = 20) and 22.4 (N = 4). The Reynolds number (based on channel height) values are:
5.1 for the shear rate of 0.05 and 7.7 for the shear rate of 0.075. ........................................... 24

Figure 2.3: Bead density profiles for chains of length 20 as a function of distance from the
wall. Simulation conditions are same as those for Figure 2.2. .............................................. 25

Figure 2.4: Normalized root mean squared chain end-to-end distance (normalized by the
equilibrium value i.e. at We = 0) as a function of shear rate (filled circles: N = 20 and open
squares: N = 4). Simulation conditions are the same as those for Figure 2.2. The average
error bars on the absolute values of chain end-to-end distance are 0.21 for long chain (N =
20) and 0.01 for short chain (N = 4). ...................................................................................... 26

symbols) and 4 (filled symbols only) as a function of distance from the wall. Simulation
conditions are: Couette flow (shear rates = 0.05 and 0.075), = 0.8, isothermal flow
conditions with T
fluid
= 0.9 and WCA interactions. The channel Reynolds number values are:
4.8 for the shear rate of 0.05 and 7.2 for the shear rate of 0.075. The ratios of channel height
(H) to the bulk chain radius of gyration (R
g
) are: 8.1 (N = 20) and 22.4 (N = 4). .................. 27

Figure 2.6: Chain center of mass density profiles for chains of length 20 as a function of
distance from the wall. Simulation conditions are: Couette flow (shear rates = 0.05 and
0.075), = 0.8, T
wall
= 0.9 and WCA interactions. Density profile is shown for a channel
height of 18. The ratio of channel height (H) to the bulk chain radius of gyration (R
g
) for this
case is 6.9. ............................................................................................................................... 28

distance from the wall. Simulation conditions are identical to those for Figure 2.6. Density
profile is shown for a channel height of 21. The ratio of channel height (H) to the bulk chain
radius of gyration (R
g
) for this case is 8.1............................................................................... 29

distance from the wall. Simulation conditions are identical to those for Figure 2.6. Density
viii
g
) for this case is 10.4............................................................................. 29

distance from the wall. Simulation conditions are identical to those for figure 2.6. Density
g
) for this case is 12.7............................................................................. 30

Figure 2.10: Normalized depletion layer thickness plotted as a function of normalized
channel height for a chain of length 20. Simulation conditions are same as those for Figure
2.6. Depletion layer thickness and channel height are normalized by the radius of gyration of
the chain. ................................................................................................................................. 30

Figure 2.11: Chain center of mass density profiles for chains of length 20. Simulation
conditions are: Couette flow, = 0.8, channel height H = 21, T
wall
= 0.9 and WCA
interactions. Figures show density profiles at equilibrium and at shear rates of 0.05 and
0.075. Density profiles are shown for systems with concentrations of: (a) 0.11 C* and (b)
1.64 C*. ................................................................................................................................... 32

Figure 2.12: Normalized depletion layer thickness (normalized by chain radius of gyration)
plotted as a function of concentration for a chain of length 20 at equilibrium and at shear
rates of 0.05 and 0.075. Simulation conditions: Couette flow, = 0.8, channel height H = 21,
T
wall
= 0.9 and WCA interactions............................................................................................ 33

Figure 2.13: Velocity profiles in the channel for solutions with concentrations 0.11 C* and
1.64 C*. Simulation conditions are the same as those for figure 2.12. The velocities of the
two walls are 0.525. ............................................................................................................ 34

Figure 2.14: Chain longest semi-axis length profile for chain of length 20 at (a) equilibrium
and at shear rates of (b) 0.05 and (c) 0.075. Simulation conditions are same as those for
figure 2.12. Profiles are plotted for lowest (C = 0.11 C*) and the highest (C = 1.64 C*)
values of concentrations studied in this set............................................................................. 36

symbols) and 3 (filled symbols only). Simulation conditions are: Couette flow, = 0.8, T
wall

= 0.9 and LJ interactions. The ratios of channel height (H) to the bulk chain radius of
gyration (R
g
) are: 12.7 (N = 10) and 27.2 (N = 3). The Reynolds number values are: 4.2 for
the shear rate of 0.05 and 6.4 for the shear rate of 0.075........................................................ 38

symbols) and 3 (filled symbols only). Simulation conditions are: Couette flow, = 0.8, T
fluid
= 0.9 and LJ interactions. The ratios of channel height (H) to the bulk chain radius of
gyration (R
g
) are: 12.7 (N = 10) and 27.2 (N = 3). The Reynolds number values are: 4.2 for
the shear rate of 0.05 and 6.4 for the shear rate of 0.075........................................................ 38
ix
Figure 3.1: A schematic of the simulation system showing smooth nanoparticle, fluid and
wall atoms. The nanoparticle is translated in x direction while the confining walls are normal
to z direction............................................................................................................................ 48

Figure 3.2: The pictorial representation of the simulation system showing the central
simulation box and its periodic images along the x and y directions. The spherical solute
particle is pulled along the x direction. ................................................................................... 52

Figure 3.3: Solvent density profile near the channel walls for the system 45.5 x 45.5 x 30.
Five significant layers of solvent were observed near either of the channel walls................. 54

Figure 3.4: Density profile of the solvent atoms as a function of the distance from the
translating sphere. Results are shown for sphere pulling velocity of 0.02 in a system with
box dimensions 45.5 x 45.5 x 30. ........................................................................................... 54

Figure 3.5: A schematic showing the primary quantity of interest in this work the friction
force coefficient. ..................................................................................................................... 55

Figure 3.6: Friction force experienced by the sphere as a function of the sphere velocity for
the system with box dimensions 45.5 x 45.5 x 30. ................................................................. 56

Figure 3.7: Normalized friction force as a function of the normalized channel length for
three different channel widths calculated using bare radius. Channel height is H = 30.
Squares represent data for L
y
= 23.4, triangles for L
y
= 33.8 and circles for L
y
= 45.5. Error
bars on the data points are of the same size as the symbols. .................................................. 58

Figure 3.8: Normalized friction force as a function of the normalized channel length for
three different channel widths calculated using effective hydrodynamic radius. Channel
height is H = 30. Squares represent data for L
y
y
= 33.8 and circles for
L
y
= 45.5. Error bars on the data points are of the same size as the symbols. ....................... 60

Figure 3.9: Normalized friction force as a function of the normalized channel width for three
different channel lengths calculated using bare radius. Channel height is H = 30. Squares
represent data for L
x
x
x
= 45.5. Error bars on
the data points are of the same size as the symbols. ............................................................... 60

Figure 3.10: Normalized friction force as a function of the normalized channel width for
three different channel lengths calculated using effective hydrodynamic radius. Channel
height is H = 30. Squares represent data for L
x
x
= 33.8 and circles for
L
x
= 45.5. Error bars on the data points are of the same size as the symbols. ....................... 61

Figure 3.11: Fluid velocity profile (only the x component of fluid velocity is plotted for the
cubes residing along the axes) in the yz plane for the system with dimensions 33.8 x 33.8 x
30. The sphere resides at the center of the figure and is moving with velocity V = 0.02 along
the x direction. The uncertainties in the numbers range from 0.002 to 0.006, with the average
uncertainty being 0.004........................................................................................................... 63
x
Figure 3.12: Fluid velocity profile in the yz plane for the system with dimensions 33.8 x 33.8
x 30. All conditions are the same as those for Figure 3.11 except that sphere-fluid interaction
is modeled using the full LJ potential. X-components of the fluid velocity are shown for the
cases of (a)
sphere-fluid
= 4.0 and (b)
sphere-fluid
= 7.0. ............................................................... 64

Figure 3.13: Normalized friction force on a translating sphere as a function of the distance d
from the nearest surface. The data points are for the system with dimensions 33.8 x 33.8 x
30............................................................................................................................................. 66

Figure 3.14: Normalized friction force on a translating sphere as a function of normalized
channel length for two different distances from the nearest surface. The data points are for
the system with dimensions L
y
= 33.8 and H = 30.................................................................. 67

Figure 3.15: Normalized friction force on a translating sphere as a function of normalized
channel width for two different distances from the nearest surface. The data points are for
the system with dimensions L
y
= 33.8 and H = 30.................................................................. 68

Figure 4.1: A schematic of the simulation system showing rough nanoparticle, fluid and the
wall atoms. .............................................................................................................................. 74

Figure 4.2: A schematic of the rough nanoparticle made up of beads of the size same as the
solvent atoms and arranged in FCC fashion. The beads of the nanoparticle are connected
together using a simple harmonic potential. ........................................................................... 75

Figure 4.3: Friction force experienced by the translating nanoparticle plotted versus the force
constant for the harmonic potential between the beads of the nanoparticle. The system size
was 33.8 x 23.4 x 30.0. The nanoparticle pulling force constant of 100 was used................ 77

Figure 4.4: Friction force on the translating nanoparticle plotted for three different values of
force constant for the harmonic potential used for pulling the nanoparticle. The force
constant for harmonic potential between beads of the nanoparticle was kept constant at 500.
................................................................................................................................................. 78

Figure 4.5: Density profile of solvent atoms as a function of distance from the center of mass
of the nanoparticle for a translating nanoparticle with a velocity of 0.02 (circles) and for a
rotating nanoparticle with angular velocity of 0.008 (dotted line) in a simulation box of
dimensions 0 . 30 8 . 33 8 . 33 . The average error bars on density values are of size 0.001... 80

Figure 4.6: Normalized friction force (squares) and normalized torque (triangles) as a
function of normalized channel length. The channel height is H = 30.0. Error bars on the data
points are of the same size as the size of symbols used.......................................................... 84

Figure 4.7: Normalized friction force (squares) and normalized torque (triangles) as a
function of normalized channel width. The channel height is H = 30.0. Error bars on the data
points are of the same size as the size of symbols used.......................................................... 85
xi
Figure 4.8: Fluid velocity profile (only the x component of fluid velocity is plotted for the
cubes residing along the axes) in the yz plane for the system with dimensions
0 . 30 8 . 33 8 . 33 . The sphere resides at the center of the figure and is translating with
velocity V = 0.02 along the x direction. The uncertainties in the data range from 0.002 to
0.007, with the average uncertainty being 0.004. ................................................................... 86

Figure 4.9: Fluid velocity profile (x component of fluid velocity is plotted for the cubes
residing along the z axes, while z component of velocity is plotted for cubes residing along x
axes) in the xz plane for the system with dimensions 0 . 30 8 . 33 8 . 33 . The sphere resides at
the center of the figure and is rotating with velocity = 0.008 (corresponds to velocity of
0.02) about the y axis. The uncertainties in the data range from 0.002 to 0.007, with the
average uncertainty being 0.004. ............................................................................................ 87

Figure 4.10: Normalized friction force on a nanoparticle translating in a polymer melt
plotted as a function of normalized channel length. The channel height is H = 30.0. Error bars
on the data points are of the same size as the size of symbols used. ...................................... 89

Figure 4.11: Normalized friction force on a nanoparticle translating in a polymer melt
plotted as a function of normalized channel width. The channel height is H = 30.0. Error bars
on the data points are of the same size as the size of symbols used. ...................................... 90

Figure 4.12: Friction force on a nanoparticle translating in a concentrated suspension of
nanoparticles plotted as a function of normalized channel length. The channel height is H =
30.0. Error bars on the data points are of the same size as the size of symbols used. The
simulation conditions are: =0.844, T=1.0 and WCA interactions. ....................................... 92

Figure 4.13: Friction force on a nanoparticle translating in a concentrated suspension of
nanoparticles plotted as a function of normalized channel width. The channel height is H =
30.0. Error bars on the data points are of the same size as the size of symbols used. The
simulation conditions are: =0.844, T=1.0 and WCA interactions. ....................................... 92

Figure 5.1: A schematic of the simulation system showing a rough nanoparticle in a polymer
melt confined between walls of the nanochannel. ................................................................ 101

Figure 5.2: A schematic showing the rough nanoparticle subjected to an input oscillating
force oscillating in a harmonic trap. ..................................................................................... 104

Figure 5.3: Measured displacement of a nanoparticle oscillating with a frequency of 0.02 in a
polymer melt in a simulation system of size 33.8x33.8x30. All the interactions are WCA.111

Figure 5.4: Fourier transform of the displacement of the oscillating nanoparticle in polymer
melt for the simulation conditions of figure 5.3. .................................................................. 112

Figure 5.5: Comparison of the storage modulus (G) for the polymer melt (N=20) calculated
using active nanorheology simulations with the literature values for a simulation system of
size 33.8x33.8x30. All the interactions in the simulations are of type WCA...................... 114
xii
Figure 5.6: Comparison of the loss modulus (G) for the polymer melt (N=20) calculated
size 33.8x33.8x30. All the interactions in the simulations are of type WCA...................... 114

Figure 5.7: A schematic showing the meaning of the parameter d, i.e. the distance of the
nanoparticle center of mass from the channel surface. ......................................................... 115

Figure 5.8: Comparison of the storage modulus (G) for the polymer melt (N=20) calculated
using active nanorheology simulations at two different d locations with the bulk literature
values for a simulation system of size 33.8x33.8x30. All the interactions in the simulations
are of type WCA. .................................................................................................................. 116

values for a simulation system of size 33.8x33.8x30. All the interactions in the simulations
are of type WCA. .................................................................................................................. 116

Figure 5.10: Velocity field around a nanoparticle translating in a polymer melt with velocity
of 0.039 for the simulation conditions identical to those in figure 5.5. ................................ 118

of 0.169 for the simulation conditions identical to those in figure 5.5. ................................ 118

of 0.149 for the simulation conditions identical to those in figure 5.5, but with LJ interactions
between the nanoparticle and the medium............................................................................ 119

Figure 5.13: Comparison of the storage modulus (G) for the polymer melt (N=20)
calculated using active nanorheology simulations with the literature values for a simulation
system of size 33.8x33.8x30. Nanoparticle-medium interactions are LJ while all other
interactions in the simulations are of type WCA. ................................................................. 121

size 33.8x33.8x30. Nanoparticle-medium interactions are LJ while all other interactions in
the simulations are of type WCA.......................................................................................... 121

Figure 5.15: Comparison of the storage modulus (G) for the polymer melt (N=20)
calculated using active nanorheology simulations at two different d locations with the bulk
literature values for a simulation system of size 33.8x33.8x30. Nanoparticle-medium
interactions are of type LJ while all other interactions in the simulations are of type WCA.
............................................................................................................................................... 122

xiii
values for a simulation system of size 33.8x33.8x30. Nanoparticle-medium interactions are
of type LJ while all other interactions in the simulations are of type WCA......................... 122

Figure 5.17: Comparison of the storage modulus (G) for a different polymer melt (N=10)
calculated using active nanorheology simulations with the bulk literature values for a
simulation system of size 33.8x33.8x30. Nanoparticle-medium interaction are of type LJ
while all other interactions in the simulations are of type WCA. ......................................... 124

Figure 5.18: Comparison of the loss modulus (G) for a different polymer melt (N=10)
calculated using active nanorheology simulations with the bulk literature values for a
simulation system of size 33.8x33.8x30. Nanoparticle-medium interaction are of type LJ
while all other interactions in the simulations are of type WCA. ......................................... 124

1
1. CHAPTER 1
I. INTRODUCTION AND OVERVIEW
Hydrodynamics in nanoscale systems

About half a century ago in his much celebrated talk titled Theres plenty of room
at the bottom, in the annual meeting of American Physical Society, Richard Feynman had
called upon the scientific community to explore the opportunities that exist at the smallest
possible scales and to challenge the definition of smallest possible itself. Science has come
a long way since then and has been in a pursuit of going to increasingly smaller length scales
to achieve bigger and bigger goals. The never-ending quest for miniaturization and the
tremendous potential of small scale devices in scientific development has inspired a
tremendous surge in the conception and development of micro- and nanoscale devices used
for various applications.
The conception and design of these micro- and nanoscale devices demands accurate
knowledge of the transport properties of the fluids involved in these small systems.
Technological advances in techniques of fabricating nanoscale structures,[1-6] have renewed
interest in understanding the transport processes at the nanometer length scales. The
continuum mechanics based methodologies for the calculation of transport properties of
fluids on macroscale are very well established. However, the transport properties of fluids in
micro- and nanofluidic devices could be significantly different from those in the macroscale
devices. The assumptions underlying continuum transport equations, such as of the
homogeneity of the system or the boundary conditions at surfaces that are applicable for
macroscale systems may not hold at the micro- and nanoscales and hence need to be verified.
2
Several studies, that have used both experimental and simulation techniques have
focused in past on studying the hydrodynamics in micro- and nanoscale systems.[7-13] In
these small systems, a large fraction of the fluid in the system is in contact with the surface of
the channel and it was observed that the fluid-surface interactions have a significant impact
on the transport properties of fluid. In simulation studies, the systems usually consisted of
three components solute, solvent and the surface. In most of these studies, while the solute
was treated explicitly, the solvent was assumed to be implicit and bulk continuum
assumptions for the solvent were used to describe the solvent, as will be discussed in
subsequent chapters in detail. One of the primary objectives of this thesis is to treat the
solvent explicitly so as to study the hydrodynamic interactions between the solute particles
that propagate through the solvent and to study their effect on the transport properties of the
fluid. In this approach, diffusion and hydrodynamic interactions are completely determined
by the intermolecular interactions in the system and no prior assumptions (e.g. Oseen
hydrodynamics) need to be made. Furthermore, the molecular nature of the solvent allows
for explicit treatment of the structural heterogeneities in the system which is an essential
feature of the nanoscale systems.
Nanofluidics
What is nanofluidics and what is its significance?

Nanofluidics is a discipline of science that deals with the study of fluid flow in the
geometries where one of the dimensions is less than 100 nm.[14] It is basically the study of
fluid flow in nanoscopic channels or around nanometer sized objects. This form of scientific
study has been around since long time in various branches of science such as biology,
chemistry, physiology etc. although the name Nanofluidics has attracted attention for the
3
past decade due to the growing interest and rapid advances made using the nanoscale
systems. Naturally, nanoscale systems are characterized by nanoscale forces. The theory
developed by Derjaguin, Verwey, Landau and Overbeek (DLVO) in 1940s accounts for
electrostatic and van der Walls forces that act in these geometries.[15, 16] With the advent of
and subsequent improvements in the surface forces apparatus (SFA)[17, 18] and the more
recently developed atomic force microscope (AFM)[19] it was possible to measure these and
other surface forces such as hydration forces, salvation forces and capillary forces. The fluid
behavior in nanoscale systems is governed by these forces. The gravitational and inertial
forces play a significant role in macroscale systems have a negligible effect in nanoscale
systems. In nanoscale systems, the surface area to volume ratio is very high thus resulting in
an increased contact of the fluid molecules with the nanochannel surface. As a result, the
thermodynamic and transport properties of fluids in these small geometries can be
significantly different from their bulk properties, where a relatively smaller proportion of the
fluid is in contact with the channel surface.
Applications of nanofluidic devices

The current and potential applications of nanofluidic devices can primarily be divided
into two broad categories: Single molecule analysis and separations.
A primary application of nanofluidic devices is in single molecule analysis, especially
for biomolecules. As the size of the channel becomes small the volume available for an
individual molecule reduces. The confinement in nanofluidic devices is used to elongate the
biomolecules and high resolution detection techniques are used to analyze the structure,
composition and configuration of the biomolecules.[20, 21] Each molecule in the
nanochannel can be analyzed individually with greater resolution as opposed to the average
4
methods used in bulk measurements. Nanofluidic devices have primarily been used to
perform single molecule analysis of DNA by stretching the DNA molecule and using optical
or electric detection techniques. Several studies performed on DNA molecules such as their
motion in porous media, length of fragments, restriction mapping and their interactions with
proteins have helped in obtaining a deeper insight into the functional properties of the
molecules.[22-24] Excellent reviews dedicated to fabrication of nanochannels can be found
in the literature.[25, 26]. Nanofluidic devices exhibit great potential in biological
applications such as in biomolecule separation, analysis, concentration etc. and have been
looked upon as an effective mean to analyze the Deoxyribonucleic acid (DNA) molecule. In
studying DNA-protein interactions, the critical factors are to determine the number of
proteins attached and the binding sites. To address these issues the DNA molecule has to be
stretched and then the sequence should be read using an analytical technique. Nanochannels,
both naturally and artificially synthesized, with width in nanometers range have been used
for this purpose. The ultimate goal in DNA sequencing research is to reduce the cost of
sequencing the DNA molecule and much of the biological research in the past decade or so
has been dedicated towards developing better sequencing strategies. Nanofluidic devices
with sizes on the order of size of these molecules have been a huge contributor in this
research.
In addition to the development of single molecule analysis techniques, another critical
application is the development of high throughput systems which can perform functions,
such as analysis and separation in a commercially viable way. Lab-on-chip devices or the
Micro Total Analysis Systems offer exciting prospects for creating these high throughput
systems.[27] Several functions are integrated on a small chip in these devices that are a few
5
millimeters in size. These devices hold nanolitres of fluid volumes in nanoscale geometries.
Accurate knowledge of the fluid transport properties on nanoscale is critical in the efficient
design of these systems.
Theoretical models to better understand and mimic the naturally occurring sieving
and separation systems, such as human kidney[28] are difficult to construct due to the
random nature of the structures, such as the pore shape and size, involved in naturally
occurring systems. These systems are mimicked by constructing nanoscale porous devices
and studying the separation efficiency with respect to the pore geometry. Design of such
devices will require a thorough understanding of the nanoscale transport properties.[29]
The comparable sizes of the nanofluidic channel dimensions and the molecules in the
solution under study provide an opportunity to perform separation of solute molecules based
on size considerations. Randomly structured nanoporous materials are used for separation in
techniques such as size exclusion chromatography and high performance liquid
chromatography, while specifically manufactured regular arrays fabricated using
micromachining are also used for constructing various lab-on-chip devices to perform
separation.[30, 31] Flowing solution of biomolecules of different sizes through these
nanostructured arrays causes the size based separation. The shape and structure of these
arrays are tailored based on specific applications and intended separation mechanism such as
sieving, routing, fractionation, shear-flow driven, batch or continuous flow and entropy
driven.[14] These separation techniques are commonly used currently for separation of
various biomolecules such as DNA or proteins and are also popular in genomics and
proteomics.[32]

6
Experimental studies using nanofluidic devices

The developments in photolithographic techniques to construct nanofluidic devices
have enabled researchers to investigate the transport properties in nanoscale geometries. The
majority of this research was targeted towards biomedical systems and can be categorized
into various groups such as single molecule analysis, biomolecule separation, biomolecule
concentration as described in what follows.
Single molecule analysis: Single stranded RNA and DNA were driven through a 2.6nm
channel in a lipid bilayer membrane using an electric field.[33-39] Since the nanochannel
dimensions were just large enough to accommodate only a single strand of DNA the
molecule had to travel through the channel as a stretched molecule. The polymer molecule
translocating through the nanochannel would cause partial blocking of the nanochannel
resulting in reduction in the ionic currents proportional to the length of the molecule. The
technique was used in calculating the polynucleotide length and the characteristics of the
base. Improvements in nanopore detection techniques aided in extension of this study to
identify homopolymers of different polyacids in the solutions based on blockade amplitudes
and kinetics.[35, 36, 40]
Various forms of single molecule detection techniques involved analyte molecules
binding to the channel surface [41-43] with the recognition sites located inside[44, 45] or at
the opening[46] of the channel. Multianalyte detection was made possible by attaching
recognition sites to the polymers which modifies the translocation process of each analyte
differently and hence results in detection.[47]
7
Voltage driven DNA translocation through an -hemolysin pore was carried out to
study the dynamics of the DNA molecule.[35, 48] A schematic of the cross section of a -
hemolysin channel embedded in a lipid bilayer is shown in Figure 1.1.

Figure 1.1: A schematic of the cross section of a -hemolysin channel embedded in a lipid bilayer[49]

The translocation velocity was calculated as a function of polymer length. It was observed
that the molecules longer than the nanopore translocated with the same velocity while, for the
polymers shorter than the nanopore length the velocity increased with decrease in polymer
length. It was proposed that the confinement produces strong drag on the polymer molecules
which can not be approximated from the bulk hydrodynamics.
Statics and dynamics of single DNA molecule studied by electrophoretically
stretching the molecule in a 30 400 nm wide nanochannel revealed a deviation from the
deGennes scaling theory for average extension of the confined self-avoiding polymer
molecule.[50] A crossover in polymer physics was observed with crossover scale to be
roughly twice the persistence length of the molecule.
8
Pressure driven transport of DNA molecule in a nanofluidic channel showed two
distinct transport regimes.[51] The pressure driven mobility increased with the molecular
length for the nanochannel with width greater than few times the radius of gyration of the
molecule whereas the mobility was independent of molecular length for the thinner
nanochannels.
A single molecule barcoding system for DNA molecule analysis was developed using
enzymatic labeling technique that tags specific sequences on the electrokinetically stretched
DNA molecule.[52] Effect of buffer solution strength and confinement was studied on the
molecule stretching. The polymer elongation was observed to increase with decrease in
buffer strength. Polymer elongations studied as a function of confinement yielded deviations
from de Gennes scaling theory as was previously observed.[50]
Separation: Nanofluidic devices prepared using various lithographic techniques, such as
electron, x-beam and ion beam lithographs, have been shown capable of performing DNA
electrophoresis. The polymer length based difference in mobility combined with controlled
electric field was used to perform fractionation.[2, 53] Entropic trapping of DNA molecule
in microfabricated arrays with nanoscale constrictions was demonstrated for separation
purpose.[49] Interesting dynamics were observed with longer polymer molecules escaping
the entropic traps faster than the shorter ones. A model based on the trapping lifetime of
molecules was proposed to explain the observed escape behavior.
Nanofilter array chip for gel free biomolecule separation was proposed by Fu and
coworkers.[54] The proposed technique using Ogston sieving[55] mechanism, in which the
separation is caused by creating energy barriers using deep and shallow regions in the flow
geometry, was described as an alternative to the conventionally used gel electrophoresis
9
technique for which limited knowledge about the sieving mechanism is available. The
technique resulted in effective and efficient separation of sodium dodecyl sulfate-protein
complexes and small DNA molecules. Anisotropic nanofluidic sieving structure for
continuous-flow separation of DNA and proteins was used for a size based and charge based
separation of biomolecules.[56] These devices with their separation efficiency and generality
are expected to be potentially used as a generic molecular sieving structure for an integrated
biomolecule sampler preparation and analysis system.
Molecule concentration: The detection mechanisms used in analysis equipments function
better if the concentration of species to be detected is higher. Nanofluidic sieving structures
could be used as preconcentration devices to increase the concentration of selected species.
Several such techniques have been developed using nanofluidic filters in microfluidic
devices,[57] and using nanolabeling techniques for detection purposes. These systems can be
used in fabrication of microsystems for chemical-biological agent detection. Another such
preconcentration device used a nanofluidic channel created between weak reversibly bonded
glass and polydimethylsiloxane (PDMS).[58-60] The electric field applied across the
channel causes selective movement of ions and results in increased concentration of charged
proteins near the channel opening. These preconcentration devices also help in facilitating
enzymatic reactions due to increased concentrations of reacting species.[61]
Other experimental techniques using nanofluidic devices include diffraction gradient
lithography[62] which uses continuous spatial gradient structures to obtain a smooth
transition of DNA molecules from microchannel region to the nanochannel regions in micro-
nanofluidic devices. Usually the nanofluidic devices are a part of a bigger microfluidic
device, and the biomolecule has to overcome a huge entropic barrier while moving from the
10
micron to nanoscale geometries. The entropic barrier is gradually reduced using this
technique by prestretching the DNA molecules using micropost arrays.
Restriction mapping of individual -DNA molecules were carried out in a 100 200nm
nanochannel by Riehn and coworkers.[63] Two microchannels were connected together
using array of nanochannels and the restriction reactions were carried out in the nanochannel
by electrophoretically driving the DNA in the nanochannel. A study of conformational
response of single DNA molecule to changes in ionic environment showed a tremendous
increase in DNA extension with decrease in ionic strength.[64] It was shown that the
decrease in ionic strength results in a reduced screening of electrostatic interactions leading
to increased self avoidance and hence increased stretching in DNA molecule. The authors
proposed an additional parameter in deGennes theory[65] of average extension of confined
self avoiding molecule an effective DNA width that gives the increase in excluded volume
due to electrostatic repulsions.
Particle Micro and Nanorheology

Calculation of rheological properties of a fluid has traditionally been carried out using
laboratory rheometers. The rheological properties such as the complex viscoelastic moduli
are determined by studying the mechanical response of the fluid to the applied shear. A
typical rheological experiment requires about a milliliter quantity of sample and probes the
sample viscoelastic properties over a limited frequency range. These techniques pose serious
limitations in case of fluids which are precarious in nature, for biological fluids or for fluids
that are expensive to procure. Moreover, the fluid under study is always deformed in the
rheological measurements. In rheological experiments, the fluid under study is assumed to
be homogeneous and the measured response the average, bulk response of the fluid. In many
11
applications that involve biological materials such as cells, the mechanical properties vary
spatially and hence the measurement of local viscoelastic properties of interest.
Developments in the optical techniques of particle manipulation and single particle
tracking have enabled rheologists to devise new methodologies for calculation of local
mechanical properties on micrometer length scales and to resolve the issue of local
heterogeneities. In these techniques, commonly known as Microrheology, micrometer
sized probe particles are embedded in the fluid to be studied. The embedded probe particles
are used to locally deform the sample and optical techniques are used to track the motion of
the particles. The measured response of the fluid on micrometer length scales is used to
determine the local material properties and the technique is called as Microrheology.[66]
Microrheological experiments are typically divided into two broad catagories active and
passive. In active microrheology, the probe particle is actively manipulated by the local
application of force and the material response is studied to the motion of the probe or the
correlated motion of two probes.[67, 68] On the other hand, in passive microrheology, the
passive motion of the probe particle due to thermal or Brownian fluctuations is tracked to
obtain the material response.[69-73] Only microliter quantities of sample are needed in these
experiments thus providing a huge advantage over macrorheological experimental
techniques.
Many systems such as polymer nanocomposites and polymer thin films show
nanoscale structural heterogeneities. The mechanical properties of these systems are thus
expected to show nanoscale variation; these local viscoelastic properties could be
determined by particle nanorheology. Molecular simulation techniques with capabilities to
mimic the systems on the length scales of size of the molecule can be very effective in
12
exploring the physics in nanoscale geometries and to calculate the transport, mechanical as
well as the thermodynamic properties of these systems.

Molecular dynamics simulations

Molecular dynamics (MD) simulation technique is an atomistic simulation method
where each atom is treated as a point mass. MD simulation numerically solves the Newtons
equation of motion for the system to obtain information about its time dependent properties,

dx
dU
F
dt
x d
m = =
2
2
(1.1)
where m is the mass of the atom, x is position of the atom, t is time, F is the force on the atom
and U is the potential between the atoms. Starting with an initial configuration of atoms,
various atoms of the system interact via a chosen potential form which is used to calculate
the forces on each atom center. These forces are then used to advance the particle in time
with chosen time step to obtain the new positions of the atoms using one of the different
finite different methods available such as: predictor-corrector, Verlet, leap-frog, velocity-
Verlet etc. In the simulations described in all of the following chapters we have used
velocity-Verlet algorithm[74] described below.
If r(t), v(t) and a(t) are respectively the position, velocity and acceleration of an atom in the
system at time t, then the position, velocity and acceleration of the atom at time t + t ,
represented by r(t + t), v(t + t) and a(t + t) respectively, is determined using the velocity-
Verlet algorithm as,

13
) ( .
2
1
) ( . ) ( ) (
2
t a t t v t t r t t r + + = + (1.2)
)) ( ) ( .(
2
1
) ( ) ( t t a t a t t v t t v + + + = + (1.3)
The time step t represents advancement in time in each simulation step. The time step must
be small enough to avoid discretization errors in the calculations but at the same time large
enough to capture the effect being modeled without taking an extraordinary period of time.
Hence the positions, velocities, forces etc. on each atom at each time step are obtained and
can be used to calculate various properties of the system using statistical mechanics methods.
The desirable qualities of a molecular simulation algorithm are: it should be fast, it should
take as little memory as possible, it should permit the use of long time step, it should
duplicate the classical trajectories as closely as possible, it should satisfy the known
conservation laws for energy and momentum, it should be time reversible, and it should be
simple in form and easy to program.
Organization

A large number of experiments have captured the features of flow of polymeric
solutions, mainly DNA, in nanoscale channels. The mechanisms underlying these
observations have been hypothesized from the experiments; these assumptions can be
validated by using molecular simulations.
A specific example that of cross stream migration of a polymer chain in a shear
flow in a nanochannel has been selected for detailed study, presented in chapter 2. The
shear induced polymer migration mechanism holds potential for effective separation and pre-
concentration of molecules of interest from a mixture. A detailed study of the cross stream
chain migration phenomenon and the role of hydrodynamic interactions in this process is
14
presented. Furthermore, the effect of parameters such as polymer chain length, nanochannel
dimensions, intermolecular interactions and polymer concentration is studied in detail.
Next, molecular hydrodynamics in confined nanoparticle suspensions are studied
using molecular simulations of smooth (Chapter 3) and rough (Chapter 4) particle translation
and rotation. Simulations are used to quantify three effects: effect of confining channel
surface, effect of slip at the nanoparticle surface and cooperative hydrodynamic interactions
between the particles. It is demonstrated that simulation results can be quantitatively
described by continuum mechanics if these effects are explicitly accounted for in the
continuum treatment. Based on these principles, an Active Nanorheology technique is
presented for the calculation of viscoelastic properties of complex fluids using molecular
dynamics simulations in Chapter 5. The technique provides a very effective way for
calculating the local mechanical properties on nanometer length scales and will serve as a
method to probe structural behavior of heterogeneous materials such as polymer
nanocomposites. The technique is validated by calculating the viscoelastic properties of a
polymer melt and comparing these results with literature values that were obtained using
different simulation techniques.

15
2. CHAPTER 2

II. CROSS-STREAM CHAIN MIGRATION IN NANOFLUIDIC
CHANNELS

Introduction

Flow behavior of dilute polymeric solutions in micro- or nanofluidic channels has
been experimentally studied because of the potential applications of the process for
manipulation of DNA and other biological molecules.[1, 51, 63, 75-80] Cross stream
migration of chains plays an important role in some of these applications; the literature on the
flow behavior and cross-stream chain migration phenomenon in dilute polymer solutions has
been captured in an older as well as a recent review.[81, 82] A major part of this chapter is
taken from our recently published work.[83]

Experimental findings

Fluorescence microscopy experiments on DNA solutions in microfluidic channels
have shown that the chains migrate away from the channel walls when subjected to flow.[84-
86] For the very dilute DNA solutions, the thickness of the chain depletion layer near the
channel walls was found to increase with shear rate in these experiments leading to depletion
layer thicknesses that were several times the size (radius of gyration) of the chains.[84]
These observations were attributed to the hydrodynamic interactions (HI) between the chains
stretched by the flow and the channel walls as was also asserted in Brownian dynamics
simulations.[87] More recent experiments showed that the amount of chain migration due to
wall HI reduced as the solution concentration increased from 0.1C* to 1.0C* (where C* is
16
the chain overlap concentration); the HI effects were almost completely screened out for the
solution of concentration 3.0C* leading to a very small degree of chain migration at this
concentration.[86]

Theoretical findings

The cross-stream migration of chains in flowing polymer solutions has also been
predicted by kinetic theory.[88, 89] The theoretical development presented by Ma and
Graham[89] showed that for polymer solutions undergoing pressure driven flow, cross
stream chain migration can occur both by wall HI and the gradient in chain mobility (which
might arise if chain mobility is a function of chain conformations). Effect of temperature on
cross stream polymer chain migration was also studied in detail using kinetic theory.[90]

Computational investigations

The phenomenon of cross-stream chain migration in confined channels has also been
studied by modeling; these studies have primarily used the mesoscopic modeling techniques
of Brownian dynamics (BD), dissipative particle dynamics (DPD) or Lattice Boltzmann (LB)
simulations.[87, 91-98] BD simulations of Jendrejack et al.[87] showed that for the same
flow rate, the depletion layer thickness increased with an increase in chain length. In another
study, BD simulations carried out using a chain consisting of freely jointed rigid rods showed
that the depletion layer thickness is insensitive to the chain flexibility.[91] The direction of
chain migration (either towards or away from the walls) was found to be governed by the
degree of chain confinement in these simulations. Specifically, for weakly confined chains
17
( 5 >
g
R
H
, where H is the channel height and
g
R is the chain radius of gyration), chain
migration was noted to be away from the walls; on the other hand, for strongly confined
chains ( 3 <
g
R
H
), the chains were found to migrate towards the wall in a flowing
solution.[87, 92, 95, 97, 98] Consistent with the experimental studies noted earlier, BD
simulations carried out by Hernandez-Ortiz et al.[93] indicated that the degree of chain
migration in flowing polymer solutions decreases as chain concentration becomes greater
than 0.1C*.

Chain migration in nanofluidic channels

Most of the previous work has focused on the chain migration phenomenon in
microfluidic channels. Advances in lithography and other experimental techniques have
allowed the development of nanofluidic devices with channel widths in the range of 20-200
nm.[1, 51, 63, 78-80] Experimental work has suggested the usage of these nanofluidic
devices for applications such as measurement of contour length of DNA [78], restriction
mapping of DNA [63] and length based separation of DNA molecules.[51] Fluid flow at the
nanoscale involves rich physics characterized by commensurate length scales of chain size
(both radius of gyration and chain contour length), channel size and most importantly, wall
induced structure formation. Accurate prediction of hydrodynamics of polymer solutions in
these channels necessitates usage of simulation techniques that account for the molecular
nature of the solvent. Previously, three mechanisms of cross-stream chain migration in nano-
confinements were illustrated by Khare and coworkers.[99] That work consisted of
18
molecular dynamics simulations of cross-stream chain migration in nanofluidic channels and
showed that at least three chain migration mechanisms are operational at the nanoscale:
hydrodynamic interactions between the chains stretched by the flow and the channel surface,
thermal diffusion and gradient in chain mobility. The phenomenon of cross-stream chain
migration due to wall hydrodynamic interactions in nanofluidic channels is studied in details
here. In this work, the focus is entirely on the Couette flow; this precludes additional chain
migration mechanisms that could arise in Poiseuille flow due to the parabolic velocity
profile.
The solvent is represented explicitly in our model system. The intermolecular
interactions in the system are modeled using standard functional forms; these intermolecular
interactions completely specify the hydrodynamic interactions in the system. For such a
model system, a detailed analysis of the effects of three factors - chain length, channel height
and chain concentration - on the chain migration process is presented in this chapter. The
degree of chain migration is quantified in terms of the depletion layer thickness at the
channel surface; a discussion of chain conformations in the system is also presented.

Simulation Method

The simulation system consists of a dilute polymer solution that is confined between
two atomistic walls. There are three types of atoms in the system polymer chain atoms,
solvent atoms and the atoms that constitute the channel walls. Figure 2.1 shows a schematic
of the simulation system used in these simulations.

19
Atomistic walls
Sheared
fluid
u
x
z
u
Couette flow
Atomistic walls
Sheared
fluid
u
x
z
u
Couette flow

Figure 2.1: A schematic of the simulation system showing polymer chain, fluid atoms and the wall atoms.
The Couette flow is generated by moving the two walls in opposite direction with same velocity.

The polymer chains are modeled as bead-spring chains. The beads are connected by
finitely extensible non-linearly elastic (FENE) springs represented by the potential:

(
(
|
|
\
|
=
2
2
1 ln
2
) (
Q
r kQ
r
FENE
(2.1)
where k is the spring constant and Q is the maximum extension of the spring. The values of
the parameters are taken from previous work[99], these are: k = 4.0 and Q = 3.0. The solvent
is modeled explicitly with beads that have the same mass and the non-bond interaction
energy parameters as the chain beads. All the atoms in the system interact via a purely
repulsive Lennard-Jones potential (WCA potential)[100] that is obtained by truncating and
shifting the Lennard-Jones (LJ) potential at cutoff distance
6 / 1
2 =
cut
r , where is the
LJ diameter. All quantities are reported in the reduced LJ units in the rest of the paper.
The confining walls are normal to the z direction and consist of atoms that are
attached to the face centered cubic lattice sites by stiff harmonic springs with a spring
constant of 500. Periodic boundary conditions are applied in x and y directions. The number
20
density of the atoms in the simulation box is = 0.8. The simulations were carried out by
maintaining only the wall atoms at a constant temperature T
wall
of 0.9 by coupling these to a
heat bath. At the conditions studied in the simulations, slight temperature increase was noted
at the highest shear rates used. To decouple the effects of the chain migration due to thermal
effects and shear flow, one subset of simulations was carried out by maintaining the entire
mass of the sheared fluid at a constant temperature of 0.9, the y component of the equation of
motion was coupled to a heat bath in this case.[101] These constant fluid temperature
simulations were carried out for illustration purposes; these eliminate any potential effect of
shear induced viscous heating on the chain migration process. The simulation box
dimensions were 42.9 and 14.3 in the x and y directions respectively, while the channel
height i.e. wall to wall distance in z direction was 21 (except in one set of simulations, where
channel height was varied to study the effect of channel height on the chain migration
process). Test runs carried out at a slightly different state point ( = 0.844 and T
wall
= 0.9)
showed that within statistical uncertainties, the same results were obtained for the y
dimension values of 14.3 and 28.6. MD simulations were carried out using the velocity
Verlet algorithm with a time step of 0.004 (0.0035 for the highest shear rate). Couette flow
was simulated by moving the two channel walls with the same speed but in opposite
directions along the x axis.[102, 103] For all systems, simulations were carried out at
equilibrium (i.e. no flow) and at two different shear rates.

21
Simulation sets

Given this basic setup, different sets of simulations were carried out to investigate the
effects of channel height, chain length and the solution concentration on the cross-stream
chain migration process. These are:

(a) Set 1: For studying the effect of chain length, the confined polymer solution consisted of
one short (N = 4, mass fraction = 0.00041) and one long chain (N = 20, mass fraction =
0.002) that were dissolved in the monomeric solvent in the simulation box. Each of the
simulation runs consisted of an equilibration stage followed by a production stage consisting
of ten blocks; the technique of block averaging was used to estimate the uncertainties in the
results. The run lengths were 20 million time steps (equilibration stage) and 200 million time
steps (production stage).

(b) Set 2: For studying the effect of channel height, four different channel heights H of 18.0,
21.0, 27.0 and 33.0 were considered; the mass fractions of the chains (N = 20) in these
systems were 0.0024, 0.002, 0.0016 and 0.0026 respectively. The lengths of the equilibration
stages for the different channel height simulations were (in units of time steps): 20 million (H
= 18 and H = 21), 33 million (H = 27) and 49 million (H = 33). The production stages for
these simulations consisted of 200 million (H = 18 and H = 21), 330 million (H = 27) and
490 million (H = 33) time steps.

(c) Set 3: For investigating the effects of solution concentration on the chain migration
process, the systems studied consisted of polymer solutions (N = 20) with chain
22
concentrations of 0.11 C*, 0.33 C*, 0.66 C*, 0.99 C*, 1.32 C* and 1.64 C*. Here, C* is the
chain overlap concentration (mass fraction = 0.274) and is calculated using the approach
described by Stoltz et al. [104]

3
3
4
g
B
R
N
c
(2.2)
where N
B
is the number of beads in the chain, and R
g
is the chain radius of gyration. For the
systems with the two lowest concentrations (0.11 C* and 0.33 C*), the equilibration stage
consisted of 4 million steps while the production stage consisted of 40 million steps. For the
higher concentration solutions, the equilibration stage consisted of at least 2 million steps and
the production stage had at least 20 million steps.

Weissenberg number

Chain stretching plays an important role in the wall hydrodynamics induced
migration of the chains in shear flow. The Weissenberg number (defined as
= We ,
where is the chain longest relaxation time and
is the shear rate) provides a good

indicator of the tendency of the chains to stretch in the presence of the shear flow. If We >>
1, shear flow stretches the chains on a time scale that is much faster than their natural
relaxation time scale and the chains acquire a stretched configuration. The calculation of We
thus requires determination of the chain longest relaxation time . Here, the chain relaxation
time is determined from equilibrium MD simulations in conjunction with the Zimm model,
as described in previous work.[99] In brief, the chain center of mass diffusion coefficient
23
was obtained by measuring the mean squared displacement of the chain in an equilibrium
MD simulation of a bulk (unconfined) system. In equilibrium simulations a 3-D periodic box
is used and the chain is allowed to move freely, and the mean-squared displacement of the
chain is monitored. The mean squared displacement when plotted against the time gives a
straight line passing through the origin, slope of which yields the diffusion coefficient. The
chain relaxation time was then obtained by using the Zimm model expression that relates the
chain diffusion coefficient with the relaxation time: [105]

D
R
2
0637 . 0 = (2.3)
where R is the chain end-to-end distance and D is the diffusion coefficient calculated from
the equilibrium molecular dynamics simulations.
This procedure is based on the assumptions that the excluded volume interactions
between the beads can be neglected and that the mobility matrix can be represented using the
Oseen tensor. Although based on these approximations, we consider this to be adequate for
our purposes since our only interest is to use the relaxation time value so obtained to
determine the shear rates for which We > 1. It has been pointed out in the literature that the
cross stream chain migration phenomenon due to wall hydrodynamic interactions is only
observed at small values of the Reynolds number.[92] Calculation of the Reynolds number
for the simulated systems necessitates the value of the viscosity. The viscosity of each of the
systems is determined by measuring the shear stress on the channel walls during the Couette
flow simulations and then dividing the wall stress by the shear rate.

24
Results

Effect of chain length

The principal quantity of interest in this work is the distribution of the chain centers
of mass in the channel. Figure 2.2 shows the chain center of mass density profiles for the
longer and the shorter chains (N = 20 and N = 4) obtained from the Couette flow simulation.

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
2.5
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
Equilibrium
We = 48.3
We = 1.0
We = 72.4
We = 1.6

Figure 2.2: Chain center of mass density profiles for chains of length 20 (lines with open symbols) and 4
(filled symbols only) as a function of distance from the wall. Simulation conditions are: Couette flow
(shear rates = 0.05 and 0.075), = 0.8, T
wall
= 0.9 and WCA interactions. The ratios of channel height (H)
to the bulk chain radius of gyration (R
g
) are: 8.1 (N = 20) and 22.4 (N = 4). The Reynolds number (based
on channel height) values are: 5.1 for the shear rate of 0.05 and 7.7 for the shear rate of 0.075.

For the sake of clarity, the uncertainties as calculated from the technique of block
averaging[106] are shown for only one shear rate; these are of similar magnitude for all other
profiles and are not shown in the rest of the figures. The figure compares the distribution of
chain centers of mass at equilibrium (no flow) with those at two different shear rates. Very
different behavior is exhibited by the two chains: the longer chain (N = 20) shows a strong
25
tendency to migrate away from the channel wall (at z = 0) with an increase in the shear rate.
On the other hand, the concentration profiles for the shorter chain (N = 4) at the shear rates
studied are virtually indistinguishable from the profile at equilibrium. A more detailed
picture of the effect of shear flow on the distribution of the longer chain in the channel can be
obtained by focusing on the density of the individual beads in the channel (Figure 2.3). For
the longer chain in this dilute solution at equilibrium, a weak tendency for the beads to layer
against the walls is observed, followed by a bulk-like flat region in the density profile away
from the walls. As the shear rate increases, the beads of the longer chain start migrating
away from the channel walls and the tendency for layering against the wall is further
weakened.

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
2.5
B
e
a
d

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

Figure 2.3: Bead density profiles for chains of length 20 as a function of distance from the wall.
Simulation conditions are same as those for Figure 2.2.

The origin of the different migration behavior shown by the two chains in Figure 2.2
can be deduced by focusing on the stretching behavior of the two chains (Figure 2.4). As
26
determined from the root mean squared end-to-end distance, at the highest shear rate, the
longer chain stretches by about 54% compared to its equilibrium size, whereas the shorter
chain hardly stretches at all. This behavior is to be expected from the values of We for the
two chains: We for the longer chain is much higher than 1 (~ 72) whereas for the shorter
chain, it is barely greater than 1 (~ 2), at the highest shear rate studied.

0 0.02 0.04 0.06 0.08
Shear rate
0.8
1
1.2
1.4
1.6
N
o
r
m
a
l
i
z
e
d

c
h
a
i
n

e
n
d
-
t
o
-
e
n
d

d
i
s
t
a
n
c
e

Figure 2.4: Normalized root mean squared chain end-to-end distance (normalized by the equilibrium
value i.e. at We = 0) as a function of shear rate (filled circles: N = 20 and open squares: N = 4).
Simulation conditions are the same as those for Figure 2.2. The average error bars on the absolute values
of chain end-to-end distance are 0.21 for long chain (N = 20) and 0.01 for short chain (N = 4).

These results indicate that the longer chains that are stretched by the flow exhibit cross-
stream chain migration due to hydrodynamic interactions with the channel walls[87] whereas
the shorter chains do not get stretched by the flow, and hence do not exhibit cross-stream
chain migration.
The simulations reported above were carried out with only the channel walls being
maintained at a constant temperature. At the highest shear rate studied (0.075), there is a
27
slight temperature rise (~ 0.13) in the system due to viscous heating. Previously, it was
shown that a temperature gradient can cause migration of chains in a nanochannel (in
addition to the wall hydrodynamic interaction).[99] To separate out these two effects,
isothermal (constant T
fluid
) shear flow simulations were carried out on the system and the
results for the chain distribution are shown in Figure 2.5 for this case. These results are very
similar to those in Figure 2.2 and indicate that the different migration behavior exhibited by
the two chains in the same solution that is subjected to shear flow is mainly due to the wall
hydrodynamic interactions at these conditions.

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
2.5
3
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
Equilibrium
We = 48.3
We = 1.0
We = 72.4
We = 1.6

(filled symbols only) as a function of distance from the wall. Simulation conditions are: Couette flow
(shear rates = 0.05 and 0.075), = 0.8, isothermal flow conditions with T
fluid
= 0.9 and WCA interactions.
The channel Reynolds number values are: 4.8 for the shear rate of 0.05 and 7.2 for the shear rate of
0.075. The ratios of channel height (H) to the bulk chain radius of gyration (R
g
) are: 8.1 (N = 20) and 22.4
(N = 4).

28

Effect of channel height

We have studied the cross stream chain migration behavior in four different channels
with heights 18.0, 21.0, 27.0 and 33.0. Figure 2.6 to Figure 2.9 show the concentration
profiles of the chain center of mass for these four channel heights. The concentration profiles
for all the channel heights studied are qualitatively similar. In all cases, a steric depletion
layer is observed next to the channel surface in the absence of shear flow. The concentration
profiles in presence of shear flow show that the chains migrate further away from the channel
surface, the amount of migration increasing with an increase in the shear rate. We have
quantified the degree of chain migration by tracking the thickness of the chain depletion layer
which is defined as the distance from the channel surface at which the concentration profile
assumes a value of unity (i.e. uniform concentration).

0 3 6 9
z
0
0.5
1
1.5
2
2.5
3
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

Figure 2.6: Chain center of mass density profiles for chains of length 20 as a function of distance from the
wall. Simulation conditions are: Couette flow (shear rates = 0.05 and 0.075), = 0.8, T
wall
= 0.9 and WCA
interactions. Density profile is shown for a channel height of 18. The ratio of channel height (H) to the
bulk chain radius of gyration (R
g
) for this case is 6.9.
29

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
2.5
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

wall. Simulation conditions are identical to those for Figure 2.6. Density profile is shown for a channel
g

0 1 2 3 4 5 6 7 8 9 10 11 12 13
z
0
0.5
1
1.5
2
2.5
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

wall. Simulation conditions are identical to those for Figure 2.6. Density profile is shown for a channel
g
30

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
z
0
0.5
1
1.5
2
2.5
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

wall. Simulation conditions are identical to those for figure 2.6. Density profile is shown for a channel
g

0.75
1.25
1.75
2.25
2.75
3.25
3.75
6 8 10 12 14
Equilibrium
We = 48.3
We = 72.4

Figure 2.10: Normalized depletion layer thickness plotted as a function of normalized channel height for
a chain of length 20. Simulation conditions are same as those for Figure 2.6. Depletion layer thickness
and channel height are normalized by the radius of gyration of the chain.

d
/
R
g

H/R
g
31
A plot of the normalized depletion layer thickness as a function of channel height is
shown in Figure 2.10 for conditions of equilibrium and shear flow. We note that the
depletion layer thickness at equilibrium shows a very weak variation (if at all) with channel
height as is expected for a steric depletion layer. On the other hand, for chains subjected to
shear flow, the thickness of the depletion layer increases monotonically with an increase in
the channel height.

Effect of concentration

The hydrodynamic interactions are expected to be increasingly shielded as the
solution concentration increases and hence the cross-stream chain migration process is
expected to be strongly affected by the chain concentration in the solution. We investigated
this effect by studying the chain migration phenomenon in solutions in the concentration
range of 0.11 C* - 1.64 C*. Our results show that the solutions with a lower concentration of
chains show a qualitatively different chain migration behavior than the higher concentration
solutions. Specifically, at the lower concentration (0.11 C*, see Figure 2.11a), chains
migrate away from the surface when subjected to the shear flow.
The tendency for the migration of chains away from the surface in presence of shear
flow decreases with an increase in the solution concentration. At the higher concentrations
(results for solution concentration of 1.64 C* are shown in Figure 2.11b), there is a weak
tendency for chain migration towards the surface in the presence of shear flow.

32

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
We = 48.3
We = 72.4

Figure 2.11: Chain center of mass density profiles for chains of length 20. Simulation conditions are:
Couette flow, = 0.8, channel height H = 21, T
wall
= 0.9 and WCA interactions. Figures show density
profiles at equilibrium and at shear rates of 0.05 and 0.075. Density profiles are shown for systems with
concentrations of: (a) 0.11 C* and (b) 1.64 C*.

(a)
(b)
33
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
0 0.5 1 1.5 2
Equilibrium
We = 48.3
We = 72.4

Figure 2.12: Normalized depletion layer thickness (normalized by chain radius of gyration) plotted as a
function of concentration for a chain of length 20 at equilibrium and at shear rates of 0.05 and 0.075.
Simulation conditions: Couette flow, = 0.8, channel height H = 21, T
wall
= 0.9 and WCA interactions.

At this point, we note that the relaxation time of the chain is a function of solution
concentration. We have chosen to compare the chain migration behavior of different
solutions at the same shear rate rather than at the same Weissenberg number. The relaxation
time of chain increases with the solution concentration and thus, the systems with the higher
concentration will, in fact, have a larger Weissenberg number than those with the lower
concentration at the same shear rate. The degree of chain migration is quantified by plotting
the normalized depletion layer thickness as a function of the solution concentration (Figure
2.12). As seen, the depletion layer thickness in shear flow is larger than that at equilibrium
for the lower concentration solutions, whereas opposite behavior is observed in solutions
with concentrations comparable to or greater than the overlap concentration.
d
/
R
g

C/C*
34
In addition to the chain density across the channel, the solution concentration is also
expected to affect the velocity profile in the system. Previous molecular simulations of shear
flow of polymer melts in nanochannels have shown the occurrence of velocity slip at the
channel surface.[103, 107] The effect of polymer solution concentration on the velocity
boundary condition at the channel surface is shown in Figure 2.13. The velocity profiles
show that the amount of velocity slip increases significantly with an increase in the solution
concentration from 0.11 C* to 1.64 C*.

0 3 6 9 12 15 18 21
z
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
V
e
l
o
c
i
t
y
C = 0.11 C*
C = 1.64 C*
Wall velocity

Figure 2.13: Velocity profiles in the channel for solutions with concentrations 0.11 C* and 1.64 C*.
Simulation conditions are the same as those for figure 2.12. The velocities of the two walls are 0.525.

Molecular simulations provide the ability for developing a detailed molecular level
understanding of the chain conformational behavior in these flowing solutions. Firstly, the
g
) of the polymer chain was seen to show a very weak dependence on
the concentration: the chain R
g
in the solution with concentration C = 1.64 C* was seen to be
35
lower than the value at C = 0.11 C* by about 5% at equilibrium and by about 2% at the two
shear rates studied. In addition to the overall chain size, the differences in the chain
stretching behavior in the solutions of different concentrations were also determined by
monitoring the chain dimensions. This was achieved by representing the polymer chain in
terms of an equivalent ellipsoid with the same moment of inertia and then calculating the
semi-axis lengths of this ellipsoid.[103] As can be seen from Figure 2.14a, at equilibrium
(no-flow), the chain longest dimension shows little variation across the channel (with the
exception of a small peak near the two walls) for a solution of concentration 0.11 C*.
On the other hand, at equilibrium, the chains near the channel surface in the higher
concentration solution (1.64 C*) are much longer than the ones away from the surface. In the
bulk like region in the middle of the channel, the chains in this higher concentration
solution are shorter than the ones in the dilute solution. As shear rate is increased (Figure
2.14b and 2.14c), the chain size difference between the solutions of different concentrations
gradually diminishes. At a shear rate of 0.075, the chain longest dimension profile for
solution with concentration 1.64 C* still shows a well defined peak near the channel wall,
however, away from the surface, the length of the chains is similar to that for the dilute
solution (0.11 C*). In summary, the overall chain size as determined by R
g
showed a small
variation with concentration over the concentration range studied in this work. On the other
hand, significant differences in the shape of the chains were observed as the solution
concentration was varied.

36
0 3 6 9 12 15 18 21
z
4
5
6
C
h
a
i
n

l
o
n
g
e
s
t

s
e
m
i
-
a
x
i
s

l
e
n
g
t
h
C = 0.11 C*
C = 1.64 C*

0 3 6 9 12 15 18 21
z
5
6
7
C
h
a
i
n

l
o
n
g
e
s
t

s
e
m
i
-
a
x
i
s

l
e
n
g
t
h
C = 0.11 C*
C = 1.64 C*

0 3 6 9 12 15 18 21
z
5
6
7
C
h
a
i
n

l
o
n
g
e
s
t

s
e
m
i
-
a
x
i
s

l
e
n
g
t
h
C = 0.11 C*
C = 1.64 C*

Figure 2.14: Chain longest semi-axis length profile for chain of length 20 at (a) equilibrium and at shear
rates of (b) 0.05 and (c) 0.075. Simulation conditions are same as those for figure 2.12. Profiles are
plotted for lowest (C = 0.11 C*) and the highest (C = 1.64 C*) values of concentrations studied in this set.
(a)
(b)
(c)
37
Effect of intermolecular interactions

The results reported above were obtained with WCA non-bonded interactions (which
are purely repulsive) in the system. In order to assess the importance of attractive
interactions on the process of cross-stream chain migration, one set of simulations was
carried out with the full LJ interactions between various species with a cut-off distance of
2.1. The box dimensions were 42.9 x 11.7 x 21.0 in this set of simulation runs. The
simulated system consisted of a one long (N = 10, mass fraction = 0.00125) and one short
chain (N = 3, mass fraction = 0.00037) that is dissolved in a monomeric solvent. The
equilibration stage of the simulation consisted of 11 million time steps and the production
stage consisted of 110 million steps. The simulations took a much longer time in this case
due to the longer range of the interactions, and hence a smaller system, shorter chains and
shorter run times were used.
The concentration profiles of the longer and the shorter chain in this LJ interaction
case are shown in Figure 2.15. As seen from the figure, the effect of shear rate on the
distribution of longer and shorter chains in the channel is qualitatively very similar to that
obtained for the WCA system i.e. the longer chain migrates away from the channel walls
with an increase in the shear rate, a behavior not shown by the shorter chains. The difference
in the migration behavior of the two chains is smaller for the LJ system due to the smaller
ratio of the lengths (relaxation times) of the two chains in this case compared to the WCA
system. At the highest shear rate studied, the longer chain in the LJ system stretches by ~
20% whereas the shorter chain is hardly stretched; this observation supports the assertion that
the observed cross-stream chain migration of the long chain is a result of the hydrodynamic
interactions between the stretched chains and the channel walls. Finally, isothermal (constant
38
T
fluid
) shear flow simulations were also carried out for this LJ system and the results for chain
migration (Figure 2.16) were seen to be very similar to those in Figure 2.15.

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
Equilibrium
We = 12.6
We = 0.7
We = 18.9
We = 1.1

(filled symbols only). Simulation conditions are: Couette flow, = 0.8, T
wall
= 0.9 and LJ interactions.
The ratios of channel height (H) to the bulk chain radius of gyration (R
g
) are: 12.7 (N = 10) and 27.2 (N =
3). The Reynolds number values are: 4.2 for the shear rate of 0.05 and 6.4 for the shear rate of 0.075.

0 1 2 3 4 5 6 7 8 9 10
z
0
0.5
1
1.5
2
C
h
a
i
n

c
e
n
t
e
r

o
f

m
a
s
s

d
e
n
s
i
t
y
Equilibrium
Equilibrium
We = 12.6
We = 0.7
Re = 18.9
Re = 1.1

(filled symbols only). Simulation conditions are: Couette flow, = 0.8, T
fluid
= 0.9 and LJ interactions.
The ratios of channel height (H) to the bulk chain radius of gyration (R
g
) are: 12.7 (N = 10) and 27.2 (N =
3). The Reynolds number values are: 4.2 for the shear rate of 0.05 and 6.4 for the shear rate of 0.075.
39
Summary and Discussion

We have used MD simulations to study the cross-stream chain migration
phenomenon in polymer solutions that are flowing in nanochannels. The diffusion and
hydrodynamic interactions in our model systems are determined only by the intermolecular
interaction potentials. Simulations of Couette flow of dilute polymeric solutions containing
chains of different lengths show a strong tendency of longer chains for cross stream
migration toward the channel center with an increase in the shear rate; on the other hand,
shorter chains do not exhibit such a cross-stream migration behavior. Similar results for two
different kinds of interaction potentials used (WCA and LJ) and for non-isothermal and
isothermal simulations, indicate that the cross stream migration process is primarily governed
by the hydrodynamic interactions between the stretched chains and the channel walls at the
conditions studied. The differences in the chain stretching as well as chain migration
behavior of long and short chains in the same solution also support the relationship between
chain stretching and cross-stream migration in polymer solutions flowing in nanochannels.
The chain migration process is shown to be strongly affected by the channel size at the
nanoscale. While the thickness of the chain depletion layer near the channel surface is shown
to be insensitive to channel height in the absence of flow, the depletion layer thickness is
shown to grow monotonically with channel height in the presence of Couette flow. For the
channel with the largest height studied, the depletion layer thickness was seen to be more
than 3 times the chain radius of gyration.
Our results show that the solution concentration strongly influences the cross-stream
chain migration process. We show that while the direction of chain migration is away from
the walls at lower concentrations, at concentrations comparable to or higher than the overlap
40
concentration, there is weak migration towards the channel wall in presence of shear flow.
We attribute this behavior mainly to the shielding of hydrodynamic interactions in higher
concentration solutions. Our simulations show the occurrence of wall slip in higher
concentration solutions at the nanoscale; this phenomenon could also play an important role
in determining the system behavior. Both experimental work and previous BD simulations
have also shown a reduction in the degree of chain migration away from the walls with an
increase in solution concentration.[86, 93] However, these BD simulations did not report
chain migration towards the wall as compared with the equilibrium profile as we have shown
for the higher concentration solutions (we note that these BD simulations did not consider
concentrations higher than the overlap concentration).
For the chain model employed in this work, at the highest shear rate studied, the
largest value of the spring lengths connecting the neighboring beads is found to be around
2.7. Thus under the right set of circumstances (chain configuration and direction of bead
motion), it is possible that two chains or different parts of the same chain might cross each
other. This is not a particular concern for dilute solutions, however, the results for the semi-
dilute solutions must be considered keeping in mind that our model chains might cross each
other during the course of the simulation. It is expected that the excluded volume
interactions between the beads will discourage this process. Furthermore, at the higher
values of We where chain migration is observed, the chains are predominantly aligned in the
flow direction,[103, 104] thus further reducing the possibility of chain crossing. In spite of
this, the possibility of chain crossing can not be completely ruled out in the chain model that
is employed in this work. Future improvements to this work will consist of employing
spring-spring repulsion potentials that will discourage such chain crossings.[108]
41
Only preliminary investigation of the effect of the specific interactions on the chain
migration phenomenon was presented in this work. An extension of this work will consist of
varying the solvent quality from poor to good and characterizing its effects on polymer chain
migration.

42
3. CHAPTER 3

III. MOLECULAR HYDRODYNAMICS IN NANOPARTICLE
SUSPENSIONS: SMOOTH NANOPARTICLES

Introduction

Molecular hydrodynamics phenomena near a solid surface has been the widely
explored area of research in past couple of decades due to its significance in micro and
nanofluidics. A number of recent studies have focused on the hydrodynamic interactions
experienced by colloidal particles in the vicinity of a solid surface.[7-13] Both experimental
and simulation studies of the effects of wall hydrodynamic interactions on the cross-stream
migration of chain molecules have been reported as well.[84, 87, 99, 109] The interparticle
hydrodynamic interactions are an important consideration in the analysis of the coupled
motions of colloidal particles observed in microrheology experiments carried out using
optical tweezers.[110-112] Cooperative hydrodynamic interactions between particles have
also been shown to play an important role in the ordering of particles in a vibrated fluid, self
organization of biological cells and collective dynamics of swimming particles.[113-115]
These cooperative hydrodynamic interactions between neighboring particles in either a
periodic grid or a train moving between parallel walls have recently been analyzed using a
continuum treatment.[116, 117] In this chapter (a major part of this chapter is taken from our
previously published work)[118], molecular dynamics simulations are used to study the
molecular scale hydrodynamic interactions between spherical solute particles arranged in a
periodic grid that is translating between parallel walls. The hydrodynamics is governed by
the intermolecular interactions in such a model system.
43
Continuum treatment

At the macroscopic scale, the friction force experienced by a single sphere moving
through an infinite fluid medium is given by the familiar Stokes law. For a spherical particle
moving near a solid surface, the friction force shows deviations from the Stokes law value.
Several continuum treatments have presented expressions for the force experienced by a
sphere that is moving in a fluid medium either towards a solid surface[119] or translating
parallel to a solid surface[120, 121] Of particular interest to the current work, continuum
mechanics results for the drag force on a spherical particle translating parallel to a solid
surface have shown approximately 10% increase in the drag force over the Stokes law value
for the case when the particle is about five diameters away from the surface.[120] Most of
these early continuum treatments assumed no-slip boundary condition at the sphere surface,
an assumption that may not be valid at the nanoscale. A theoretical and numerical study of
the hydrodynamic interactions in suspension of spherical particles confined between planar
walls was performed by Bhattacharya et al.[11] They had observed an increase in frictional
force on spherical particles in creeping flow conditions with an increase in the distance
between the particles before attaining a steady value at long lengths. The effect was more
pronounced in direction of flow than in the transverse direction.

Molecular simulation studies: Stokes-Einstein law

At the nanoscale, a number of molecular simulation studies have investigated the
validity of the Stokes-Einstein law:

r c
kT
D
= (3.1)
44
where D is the diffusion coefficient of the particle diffusing through a solvent of viscosity ,
T is the temperature, c is the coefficient that assumes the value of 4 or 6 for slip and stick
boundary conditions respectively, and r is the radius of the particle. Walser et al. carried out
MD simulation of a system of water molecules where different simulation systems were
generated by changing the masses of oxygen and hydrogen atoms.[122] The product of
diffusion coefficient and viscosity was found to depend on the mass of the molecule and
hence it was concluded that Stokes-Einstein law is not applicable at the molecular level.
Equilibrium MD simulation studies for simple Lennard-Jones fluids were also carried out to
investigate the applicability of Stokes-Einstein law at the nanoscale.[123-125] A main focus
of these studies was to determine the boundary conditions (slip or no-slip) at the solute
surface. The results indicated the existence of different boundary conditions (slip or stick)
depending on both the nature of the solute surface and the solute-solvent interaction
potential. Specifically, it was shown that a sufficiently large and smooth spherical solute
followed Stokes-Einstein law with slip boundary conditions; on the other hand, a spherical
solute with a rough surface followed this relationship with stick boundary conditions.[123,
124]

Validity of Stokes law

In an early study the dependence of the Stokes friction coefficient on the size of a
Brownian particle was studied by Alley and Alder for a hard sphere system.[126]
Hydrodynamic forces experienced by a spherical solute that is moving through an atomistic
fluid interacting via Lennard-Jones potential have also been studied by MD simulation.
Specifically, Vergeles et al.[127, 128] studied the force and the torque on a spherical solute
45
that is either moving towards a solid surface or that is rotating near a solid surface. These
authors showed that Brenners continuum result for dependence of friction force on distance
from the wall[119] was applicable until the solute-wall distance became of the order of a few
molecular radii. The breakdown of continuum predictions at shorter distances was attributed
to the depletion of solvent molecules in the near-wall region and to the breakdown of the no-
slip boundary conditions used in the continuum treatment. Recently, the problem of drag
force on a sphere that is moving towards a smooth planar surface was revisited by Challa and
van Swol using MD simulations.[129] These authors showed that as the sphere approaches
the surface, the drag force on the sphere can be represented by a superposition of two
contributions: the static solvation force (present even when the sphere is not in motion) and
the dynamic force.
The main focus of the studies described above was on investigating the
hydrodynamics for the motion of a single particle through a fluid. Two recent studies have
presented a continuum level analysis of the cooperative hydrodynamics between the
neighboring particles in a grid of spheres that is moving through a confined fluid.[116, 117]
The continuum Stokes equation was solved using a Stokesian dynamics technique and it was
shown that for a periodic grid moving through a confined fluid medium, there are significant
long range hydrodynamic cooperative effects between neighboring spheres.[116] In another
study, the synchronous motion of beating cilia was explained in terms of the cooperative
hydrodynamic interactions between the beating cilia which were modeled as moving
spheres.[130]

46
Cooperative hydrodynamic effects

It is known that for a train of spheres moving through channels with one dimensional
confinement such as a cylindrical tube, the hydrodynamic interactions between successive
particles become negligible if the distance between them is greater than the width of the
channel or the diameter of the tube.[131, 132] The range of hydrodynamic cooperativity
between spherical particles arranged in an array has been studied using two different
mesoscopic modeling techniques of Lattice-Boltzmann (LB) and Dissipative Particle
Dynamics (DPD) simulations. LB simulations of a system consisting of a sphere that is held
stationary in Poiseuille flow between parallel plates showed that the sphere is
hydrodynamically decoupled from its periodic images (that form a two dimensional grid)
when the grid spacing was greater than 30 times the sphere radius.[133] The DPD
simulations focused on a similar system consisting of a sphere that is held stationary in either
planar or uniform shear flow between two parallel plates.[134] It was shown that for the
hydrodynamic interactions between the sphere and its images along the grid to be negligible,
the box size in the flow direction needed to be greater than 30 times the sphere diameter (for
Reynolds numbers of 0.5 and 50), while in the other periodic (span-wise) direction, the box
size needed to be 3.3 or 6.6 times the sphere diameter for Reynolds numbers of 0.5 and 50
respectively.
The focus of this study is on developing a detailed understanding of these cooperative
hydrodynamic effects and to quantify these effects in confined nanoparticle suspensions.
Molecular dynamics simulations are used to study the effect of slip, the solid surface and
cooperative hydrodynamic interactions on the friction force experienced by a translating
nanoparticle. Another goal of this work is to assess the applicability of continuum mechanics
47
to quantify simulation results. To this end, a two dimensional grid of spherical particles that
is translating parallel to the wall in an atomistically explicit solvent which is confined
between two atomistic, planar surfaces is studied taking advantage of the periodic boundary
conditions used in simulations to represent the grid by a single spherical particle that is
moving through the solvent. In this scenario, the central sphere and its periodic images
constitute the periodic grid. By construction then, each sphere in the grid shows identical
hydrodynamic behavior, at the same time, this approach allows studying large grid spacing
values. MD simulations at low enough sphere velocities are carried out to ensure that the
particle Reynolds number is 0.1 or lower and the state point chosen for simulations is typical
of that used in MD simulation of fluid flow phenomena. Quantitative comparison of our
MD simulation results for an atomistic solvent model with the continuum analysis recently
presented by Bhattacharya for the same problem is also made.[116]

Simulation Method

The simulated system consists of a dilute nanoparticle suspension that is confined
between two parallel atomistic walls. There are three kinds of particles in the simulation
box: the solvent atoms, the spherical solute and the wall atoms. Figure 3.1 shows a
schematic of the system used in these simulations.

48
Fluid
u
Fluid
u

Figure 3.1: A schematic of the simulation system showing smooth nanoparticle, fluid and wall atoms.
The nanoparticle is translated in x direction while the confining walls are normal to z direction.

The solvent atoms interact with each other via purely repulsive Lennard-Jones (WCA
potential):[100]

(
(
\
|
\
|
=
6 12
4 ) (
r r
r
WCA
solvent solvent

(3.2)
where and are the Lennard-Jones parameters and the potential is truncated and shifted
at a cut-off distance of
6 / 1
2 =
c
r . All the quantities are reported in reduced Lennard-Jones
units. The interaction between the solute sphere and the solvent atoms is given by the
following functional form (at the conditions simulated in this work, the spherical solute never
gets close to the channel surface in order to interact with it).[123]

(
(
\
|
\
|
6 12
4 ) (
a r a r
r
WCA
solvent solute

(3.3)
where a r > and the cut-off distance is
6
1
2 + = a r
c
. This potential is of the same form
as the solvent-solvent interaction but is shifted in the radial direction by a parameter a
which accounts for the size of the solute particle. The radius of the spherical solute can be
varied by changing the value of the parameter a.
49
The smooth nanoparticle

The initial configuration for the simulation is prepared by starting with a system in
which the solute particle is of the same size as the solvent particles. The solute particle size
is then gradually increased in a stage-wise manner while allowing the system to equilibrate
using short MD simulations. In each stage, as the solute size and mass are increased, the
proportionate amount of solvent molecules are removed to maintain the number density of
the system using the algorithm described in the work by Schmidt and Skinner.[123] If m is
the mass of the solvent and M is the mass of the solute then the mass of solute is changed
with parameter a such that the mass density of solute and solvent remained the same. The
radius of the solvent atoms are defined to be
2
whereas the radius of solute is equal to

2
+ a . Hence, to keep the mass density of solute and solvent to be same,

(
(
\
|
|
\
|
= 1
2
3
1
m
M
a

. (3.4)
For 1 = m and 1 = rearranging above equation gives mass of solute as a function of the
size of the solute (parameter a) as,

3
1
2
|
\
|
+ =
a
m M (3.5)
For this work, we have used the value of a = 2.0 and the mass of the solute is taken to be 125
times the mass of the solvent molecule. In order to generate the initial configuration for
these simulations, the size of the solute (parameter a) and, hence the mass, is gradually
increased in stages starting from a = 0 to a = 2.0. In each stage the system is equilibrated
such that the energy of system attains a minimum.
50
The confining surfaces are normal to the z axis and consist of beads that are
connected to FCC lattice sites (lattice spacing = 1.3) by stiff harmonic springs with a spring
constant value of 500. Periodic boundary conditions are applied in the x and y directions.
The number density of the system is kept at = 0.844 and the temperature of the system is
maintained at T = 1.0 by coupling the wall atoms in the system to a heat bath using a method
described in earlier work.[101] The solute sphere was constrained to be at the center of the
parallel channel by using a harmonic potential with a spring constant of 100 (constraint along
the y and z directions). The translation of the spherical solute was simulated by pulling it
along the x direction using a harmonic spring with the same spring constant value of 100.
Trial runs with spring constant values of 50 or 200 were shown to give identical results for
the friction force on the sphere within statistical uncertainties.
Molecular dynamics (MD) simulations were carried out using the velocity Verlet
algorithm with a time step of 0.003. All of the simulations (except for the largest systems)
had an equilibration stage of 3 million time steps followed by a production stage of 30
million steps. For the four largest systems studied in the work, the equilibration stage
consisted of 2.5 million steps with the production stage being of 25 million steps duration.
The production stage was divided into ten blocks; the technique of block averaging was used
to determine the average values of quantities and the associated uncertainties.

Simulation systems

The moving sphere and its periodic images form a two dimensional periodic grid that
translates through the fluid. Each moving sphere generates a velocity field in the
surrounding fluid, this effect, in turn, gives rise to hydrodynamic interactions between the
51
spheres in the grid. To quantify the dependence of these hydrodynamic effects on the grid
spacing, simulations were carried out for a number of different box dimensions in the x and y
directions resulting in a total of 27 different simulation systems as highlighted (in yellow) in
Table 3-1.
Table 3.1: A matrix of simulation systems used in chapter 3
L
y
\ L
x
23.4 33.8 45.5 55.9 67.6 79.3 91
23.4
33.8
45.5
55.9
67.6
79.3
91

The simulation system sizes (channel lengths L
x
and channel widths L
y
) ranged from 23.4 to
91.0 in both the x and y directions. The wall to wall distance (along the z direction) was kept
to be H = 30.0, for all of the systems. The volume fraction of the spherical solute in these
simulation systems varied in the range 0.004 to 0.007. Figure 3.2 shows the pictorial
representation of the two dimensional grid of the nanoparticles. For each system, simulations
at five different sphere pulling velocities ranging in values from 0.002 to 0.05 were carried
out. The friction force on the sphere was measured by averaging the x component of the
force experienced by the spherical solute. It was verified that the value so calculated agreed
(within statistical uncertainties) with the average value of the spring pulling force on the
sphere, as required for a sphere that does not have a net acceleration.

52
L
y
L
x
x
y
z
L
y
L
x
x
y
z
x
y
z

Figure 3.2: The pictorial representation of the simulation system showing the central simulation box and
its periodic images along the x and y directions. The spherical solute particle is pulled along the x
direction.

Fluid viscosity

Quantitative comparison of the friction drag on the translating sphere with the Stokes
law friction force requires the value of the fluid viscosity. The fluid viscosity was
determined by carrying out a planar Couette flow simulation for one of the systems (box size
= 23.4 x 23.4 x 30.0) at the given state point. This was achieved by moving the two walls
with the same speed but in opposite directions and then measuring the shear stress on the
channel wall. The fluid viscosity value of 2.16 was determined at a shear rate of 0.05 from
this procedure; this value of viscosity is used for all of the calculations presented in the rest
of the paper. The value so used must be considered as an approximation to the true zero
shear viscosity of the fluid for the following reasons: (1) We have not considered the shear
rate dependence of viscosity that has been demonstrated in the literature[135, 136] (2) There
is a small temperature rise in the sheared fluid at this shear rate (3) The imposed shear rate
(0.05) is used for viscosity calculation whereas there is small velocity slip (~ 1%) at the
53
channel wall (factors 2 and 3 are expected to change the value of viscosity in opposite
directions and are thus expected to mitigate each other to some extent). However,
considering that the slightly approximate value that is used will not have a major impact on
any of the findings, this value is considered to be adequate for the purpose of our
calculations.
Results

It is well known that presence of a solid surface such as the channel wall in our
system causes the fluid atoms to form several layers next to these surfaces.[101, 137-139]
Such local structural heterogeneities could potentially lead to the deviation of the system
behavior from that predicted by a continuum model of the fluid. We have verified the
existence of similar structure formation in our system with at least 5 significant fluid layers
next to either of the confining surfaces in our system as shown in Figure 3.3. In addition, the
presence of the spherical nanoparticle is expected to cause structural heterogeneities in the
fluid surrounding the particle. Figure 3.4 shows the density profile of the solvent atoms as a
function of their distance from the center of the moving sphere (normalized such that uniform
distribution of the solvent atoms will yield a value of unity). As can be seen from the figure,
the fluid surrounding the moving sphere also forms at least three well defined layers around
it, thus clearly illustrating the nanoscale density heterogeneities in the system due to the
molecular nature of the solvent.
54
0 5 10 15 20 25 30
z
0
0.5
1
1.5
2
2.5
3
S
o
l
v
e
n
t

D
e
n
s
i
t
y

Figure 3.3: Solvent density profile near the channel walls for the system 45.5 x 45.5 x 30. Five significant
layers of solvent were observed near either of the channel walls.

0 1 2 3 4 5 6 7
r
0
1
2
3
4
S
o
l
v
e
n
t

d
e
n
s
i
t
y

Figure 3.4: Density profile of the solvent atoms as a function of the distance from the translating sphere.
Results are shown for sphere pulling velocity of 0.02 in a system with box dimensions 45.5 x 45.5 x 30.

55
The primary quantity of interest in our simulations is the friction force experienced by
the spherical particle that is moving through the fluid with a constant velocity. For a
spherical object moving through an unbounded fluid, the friction force F on the sphere is
given by the Stokes law as:
V r c F = (3.6)
where r is the radius of the sphere which is moving with a velocity V through a fluid of
viscosity . The coefficient c is calculated from slope of the friction force versus the
velocity plot as shown in the following schematic.

F
v
V r c F = ) (
r
slope
c
=
F
v
V r c F = ) (
r
slope
c
=

Figure 3.5: A schematic showing the primary quantity of interest in this work the friction force
coefficient.

The coefficient c assumes the values of 4 and 6 for slip and no-slip (stick) boundary
conditions respectively. The existence of Stokes law like behavior in our model system was
checked by investigating the relationship between the friction force and the velocity of the
sphere.
56
As mentioned above, for each system studied, MD simulations were performed at 5
different velocities, and the friction force on the sphere was measured. A plot of friction
force vs. velocity for one of the systems considered is shown in Figure 3.6. The figure
clearly shows a linear relationship between the friction force and the velocity of the sphere.
Similar linear relationship between the friction force and the velocity was observed for all
other systems that were studied thus indicating that our model systems show Stokes law like
behavior.

0.0
1.0
2.0
3.0
4.0
5.0
0 0.01 0.02 0.03 0.04 0.05
Sphere velocity
F
r
i
c
t
i
o
n

f
o
r
c
e

o
n

s
p
h
e
r
e

Figure 3.6: Friction force experienced by the sphere as a function of the sphere velocity for the system
with box dimensions 45.5 x 45.5 x 30.

The value of the coefficient c in equation (3.6) was then evaluated from the slope of
the linear fit to the friction force-velocity data. Two possible definitions for the radius of the
spherical solute have been used in the literature, namely, the bare radius
57
( 5 . 2 5 . 0 = + = a ) or the effective hydrodynamic radius ( 0 . 3 = + = a ).[123] The
former considers the sphere radius to be defined by the region occupied by the spherical
solute while the latter considers the sphere radius to be defined by the radius of the region
from which solvent atom centers are excluded. We have primarily used the bare radius for
reporting the results of our calculations, this gives us a value of
12
1
for the ratio of sphere
radius (r) to channel height (H), the same value for the ratio as that was used in the
continuum treatment[116] with which we will be comparing our results (note that the particle
Reynolds number values mentioned earlier were calculated using particle diameter value of
5.0 ). All of the reported numbers will be scaled by a proportionate amount if the effective
hydrodynamic radius value were to be used for the sphere radius; these numbers are also
given in relevant places. For each box size (grid spacing), the average value of c and the
associated uncertainty in it was estimated by Monte Carlo sampling of data sets.[140] For
the 27 systems considered, the values of the coefficient c were found to span the range from
4.29 to 5.81 (the range will be 3.58 to 4.85 based on the effective hydrodynamic radius).
Here we note that several factors affect the value of this coefficient. In addition to the
boundary conditions on the surface of the moving sphere, the presence of the confining solid
walls also affects the value of c (the solid surfaces will give rise to higher values of c). More
importantly, the cooperative hydrodynamic interactions between the spheres in the moving
grid have a strong influence on the value of c; this physical effect tends to lower the value of
c. The wide range of values of c (4.29 to 5.81) that is observed can be mainly attributed to
these cooperative hydrodynamic effects which show a strong dependence on the grid spacing
(i.e. simulation box size in the periodic directions).

58
Effect of cooperative hydrodynamic interactions

To quantify this dependence of the hydrodynamic cooperation on the grid spacing, we
plot the normalized values of c as a function of channel length in Figure 3.7 for three
different channel widths. The coefficient c was normalized by dividing it by the Stokes law
value of 6 (this is akin to normalizing the friction force by the Stokes drag in the absence of
slip); this normalization also allows us to directly compare our MD simulation results with
the recent continuum solution presented by Bhattacharya for the same problem.[116] For the
purpose of this comparison, both the channel length (L
x
) and channel width (L
y
) are also
normalized by the channel height (H).

0.7
0.75
0.8
0.85
0.9
0.95
1
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 3.7: Normalized friction force as a function of the normalized channel length for three different
channel widths calculated using bare radius. Channel height is H = 30. Squares represent data for L
y
=
23.4, triangles for L
y
y
= 45.5. Error bars on the data points are of the same size as
the symbols.

59
These normalized friction force values show a strong dependence on the distance
between the neighboring spheres along the direction of motion of the sphere, with the values
increasing from 0.72 to 0.97 over the range of grid spacing studied. As was observed in the
continuum solution, the curves for different channel widths obtained from MD simulation
also cross each other albeit at a slightly smaller value of L
x
/H than the continuum solution.
The uncertainties in the normalized force values are of the same size as the symbols used in
the graph; this suggests that the observed crossing behavior is a systematic effect. Our MD
simulations thus qualitatively reproduce the main features of the continuum predictions for
the hydrodynamic interactions between moving spheres. We note that the hydrodynamics is
determined only by the intermolecular interactions in our model systems. A quantitative
comparison of MD results with the continuum solution shows that the continuum predictions
of friction force are consistently larger than the simulated values by approximately 21%. The
disagreement is expected to be in the range of 53 to 60% if the effective hydrodynamic radius
is used in the calculations. In that case, the normalized friction force values derived from
MD simulations will be even lower while the continuum prediction values will be higher due
to a change in the ratio of
H
r
from
12
1
to
10
1
(this increase in the continuum values due to
the change in the
H
r
ratio is expected to be around 5 to 10% [11]). Figure 3.8 shows the
plot normalized value of c plotted against normalized channel length calculated using
effective hydrodynamic radius. Figure 3.9 shows the channel width (L
y
) dependence of the
normalized friction force for three different channel lengths (L
x
).

60
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 3.8: Normalized friction force as a function of the normalized channel length for three different
channel widths calculated using effective hydrodynamic radius. Channel height is H = 30. Squares
y
y
y
= 45.5. Error bars on the data
points are of the same size as the symbols.

0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 3.9: Normalized friction force as a function of the normalized channel width for three different
channel lengths calculated using bare radius. Channel height is H = 30. Squares represent data for L
x
=
23.4, triangles for L
x
x
= 45.5. Error bars on the data points are of the same size as
the symbols.
61

It can be seen that for a given channel length, the normalized friction force values
increase with an increase in L
y
but attain a stable value at a shorter distance (L
y
/H = 2 to 3)
than that suggested by figure 4 for L
x
dependence. The three curves in figure 3.9
qualitatively show very similar behavior to that predicted by the continuum treatment. Once
again, a quantitative comparison of these friction force values with the Stokesian dynamics
results[116] shows that the values obtained from Stokesian dynamics are consistently larger
than the MD simulations by about 22%. The weak dependence of the friction force on L
y
as
opposed to that on L
x
was also observed in the continuum treatment and was attributed to the
interplay of the far-field and near-field hydrodynamic interactions in that work (these effects
augment each other along the x direction while they oppose each other along the y
direction).[116] Figure 3.10 shows the normalized value of c plotted against the normalized
channel width calculated using the effective hydrodynamic radius.
0.55
0.6
0.65
0.7
0.75
0.8
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 3.10: Normalized friction force as a function of the normalized channel width for three different
channel lengths calculated using effective hydrodynamic radius. Channel height is H = 30. Squares
x
x
x
= 45.5. Error bars on the data
points are of the same size as the symbols.
62
Effect of slip at nanoparticle surface

From Figures 3.7 and 3.9, it is thus seen that the simulation results are in qualitative
agreement with the continuum predictions, albeit consistently lower than the continuum
values. The continuum results were obtained for the case of no-slip boundary condition on
the sphere surface whereas this boundary condition is not satisfied in our simulations. A set
of simulations was carried out to quantify the effect of this velocity slip on the friction force.
First, to determine the exact boundary condition at the surface of the sphere, we
monitored the fluid velocity field around an individual translating sphere in our system. This
was achieved by generating a cubic grid around the sphere; this grid moved with the sphere
during the course of the simulation. The fluid velocity values in this cubic grid were
averaged to obtain the fluid velocity field around the sphere. Figure 3.11 shows the fluid
velocity profile (only along the axes) in the yz plane for one of the systems. The figure
shows the velocity profile around a sphere that is translating with a velocity of 0.02 in a
simulation box of dimensions 33.8 x 33.8 x 30.0. It can be seen from the figure that the fluid
next to the sphere surface along the yz plane moves with a significantly lower velocity than
the sphere thus indicating a strong tendency for velocity slip at the sphere surface. Note that
given the magnitude of the statistical uncertainties in the velocity profiles, it is not possible to
determine the exact functional dependence of the velocity on the distance (e.g. decay as
r
1
)
without a very significant computational effort.
63

Figure 3.11: Fluid velocity profile (only the x component of fluid velocity is plotted for the cubes residing
along the axes) in the yz plane for the system with dimensions 33.8 x 33.8 x 30. The sphere resides at the
center of the figure and is moving with velocity V = 0.02 along the x direction. The uncertainties in the
numbers range from 0.002 to 0.006, with the average uncertainty being 0.004.

The existence of a slip boundary condition at the surface of a smooth spherical
particle such as the one used here was also noted previously in the equilibrium MD
simulation by Schmidt and Skinner.[123] In order to estimate the effect of this slip boundary
condition on the friction force, we performed a set of simulations on a system where the
sphere-fluid interaction was modified to eliminate the slip at the sphere surface. Specifically,
only the sphere-fluid interaction was changed to the full LJ potential with a cutoff distance of
2.3 (all other interactions in the system were of the purely repulsive, WCA functional form as
before). Two sets of simulations were carried out using this modified interaction potential
and the LJ interaction parameter values of
sphere-fluid
= 4.0 and
sphere-fluid
= 7.0. The fluid
velocity profiles around the translating sphere for these two systems are shown in Figures
3.12a and 3.12b. As can be seen, due to the strong attractive interactions between the sphere
64
and the fluid atoms, the slip at the sphere surface is diminished for the
sphere-fluid
= 4.0
system, while the slip is almost completely eliminated for the
sphere-fluid
= 7.0 system.

Figure 3.12: Fluid velocity profile in the yz plane for the system with dimensions 33.8 x 33.8 x 30. All
conditions are the same as those for Figure 3.11 except that sphere-fluid interaction is modeled using the
full LJ potential. X-components of the fluid velocity are shown for the cases of (a)
sphere-fluid
= 4.0 and (b)

sphere-fluid
= 7.0.

(a)
(b)
65
The values of the normalized friction force obtained for these systems are presented
in Table 3.2 and are compared with the original system (all WCA interactions) and the
continuum mechanics solution that was obtained using the no-slip boundary condition. From
these values, one can conclude that the quantitative differences in the friction force values
between the simulation results and the continuum solution can be attributed to the velocity
slip at the sphere surface (assuming that the change in the sphere-fluid interaction primarily
affects the boundary condition and its effect on the fluid viscosity is small). In fact, when
slip is nearly eliminated (for
sphere-fluid
= 7.0), the value of the normalized friction force
obtained using the effective hydrodynamic radius is in good agreement with the continuum
prediction of Bhattacharya for the same problem[116] (note that when the calculations are
based on the hydrodynamic radius, the ratio
H
r
is
10
1
for the simulations, while the
continuum results are for
H
r
=
12
1
; changing
H
r
to
10
1
should further increase the
continuum values by 5 to 10% [11]).

Table 3.2: Comparison of normalized friction force for WCA and LJ interactions.

sphere-fluid
= 1.0

WCA interaction
sphere-fluid
= 4.0

LJ interaction
sphere-fluid
= 7.0

LJ interaction
Normalized friction force
(using bare radius value = 2.5)
0.82 1.12 1.27
Normalized friction force
(using effective hydrodynamic radius
value = 3.0)
0.69 0.93 1.06

66
Effect of confining surfaces

In addition to slip, the other aspect of interest is the effect of the confining surface on
the friction force. To obtain an estimate of this effect, a set of simulations was performed by
varying the distance of the translating spheres from the surface (the results presented in
Figures 3.7 to 3.10 were obtained for the case where the sphere was translating at the channel
center i.e. at a distance of 15 from both of the confining surfaces). A plot of the normalized
friction force as a function of the sphere distance from the nearest surface is shown in Figure
3.13.
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
0 2.5 5 7.5 10 12.5 15 17.5
d
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 3.13: Normalized friction force on a translating sphere as a function of the distance d from the
nearest surface. The data points are for the system with dimensions 33.8 x 33.8 x 30.

As can be seen from the figure, the friction force is higher for smaller separations of
the particle from the solid surface. The simulated values of the normalized friction force
qualitatively exhibit the same dependence on the distance from the surface as predicted by
continuum mechanics for the case of a single sphere translating parallel to a surface.[120]
67
However, the simulated values are smaller than the continuum predictions due to the
presence of slip and the cooperative hydrodynamic interactions in the simulated system. In
terms of quantitative comparison of the dependence of the friction force on the distance from
the surface (d), the friction force in simulations shows a decrease of 8.6% as d is increased
from 8 to 15, for a corresponding increase in the ratio of d to particle radius, the continuum
results predict an approximate decrease of 12% in the friction force.[120] The cooperative
hydrodynamic effects for a translating smooth nanoparticle at different locations from the
channel surface were also studied by performing simulations at a different d location for a
range of simulation box sizes. A comparison of normalized friction force as a function of
normalized channel length for two different d locations is presented in Figure 3.14.

0.7
0.75
0.8
0.85
0.9
0.95
1
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e
d = 15
d = 8

Figure 3.14: Normalized friction force on a translating sphere as a function of normalized channel length
for two different distances from the nearest surface. The data points are for the system with dimensions
L
y
= 33.8 and H = 30.

68
It can be observed that the friction force is higher on the nanoparticle closer to the surface
and that the cooperative hydrodynamic effects are present similar to the case with
nanoparticle at the center of nanochannel. A similar plot for channel width dependence can
be seen in Figure 3.15.

0.7
0.75
0.8
0.85
0.9
0.95
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e
d = 15
d = 8

Figure 3.15: Normalized friction force on a translating sphere as a function of normalized channel width
for two different distances from the nearest surface. The data points are for the system with dimensions
L
y
= 33.8 and H = 30.

Encouraged by the qualitative and quantitative agreement (see earlier text for the case
with no-slip at the sphere surface) between the simulation results and the continuum
predictions, we can estimate the relative contribution of the cooperative hydrodynamic
interactions in our system. For this purpose, we will focus on the particular grid spacing of
L
x
= 33.8 and L
y
= 33.8. For this case, the normalized friction force value (based on effective
hydrodynamic radius) obtained from simulations in the absence of slip is 1.06. The deviation
69
from the Stokes law value occurs in this case because of the presence of the two confining
surfaces and the cooperative hydrodynamic interactions between the spheres on the grid.
Bhattacharya et al.[11] have calculated the normalized friction force for a spherical particle
in a suspension that is confined between two parallel walls. For our geometric configuration
(H = 30, hydrodynamic radius = 3 and sphere is moving at the channel center), they predicted
a value of 1.25 for the normalized friction force. If we assume that the effect of confinement
is quantitatively similar for both the continuum and simulation results, then the difference
between the simulation and continuum values (1.06 and 1.25) can be attributed to the
cooperative hydrodynamic interactions. In other words, for these grid spacing values,
cooperative hydrodynamic interactions reduced the friction force by about 15% in our model
systems. We note that this simple analysis assumes that the effects of slip, solid surfaces
(confinement) and the cooperative hydrodynamic interactions are not strongly coupled.


MD simulations were used to study the hydrodynamic interactions experienced by
spherical particles in a periodic grid that is moving parallel to two confining surfaces. The
friction force experienced by these spheres showed a linear dependence on the velocity thus
indicating Stokes law like behavior of our model system. Our results show that significant
hydrodynamic cooperative effects exist between spheres in the grid; these effects reduce the
value of the friction force on an individual sphere compared with the Stokes law value. The
dependence of these cooperative hydrodynamic effects on the grid spacing was quantified in
this work. The simulation results for these cooperative hydrodynamics were found to
70
qualitatively agree with a recent continuum treatment of hydrodynamic interactions for a
periodic grid that is moving parallel to two confining surfaces[116], however, the simulated
values were consistently lower than the continuum predictions. This difference in the friction
force values was attributed to the fluid slip at the surface of the sphere as was seen by
monitoring the fluid velocity field around the moving sphere. When this slip at the surface of
the moving sphere was effectively eliminated by using a strongly attractive interaction
between the sphere and the fluid, the normalized friction force values from simulations
(obtained using the effective hydrodynamic radius) were in good agreement with the
continuum predictions. This observation also suggests that the effective hydrodynamic
radius is better suited for the analysis of hydrodynamic effects in molecular modeling than
the bare radius.
The removal of slip at the sphere surface was shown to lead to an increase in the
friction force value by approximately 55% (this analysis neglects the coupling between the
three effects: slip, presence of solid surface and cooperative hydrodynamics). The presence
of the solid surface (i.e. confinement) was found to lead to an increase in the friction force on
the moving sphere, with the friction force being higher at shorter distances from the surface
as expected from the continuum analysis for a similar situation albeit for the case of a single
surface.[120] Even though the simulated values of friction force were obtained in the
presence of two surfaces, these were consistently lower than the continuum predictions
which are for the case of one surface; this observation was attributed to slip and cooperative
hydrodynamics effects. A simple analysis based on a comparison with continuum mechanics
was presented to estimate the magnitude of the cooperative hydrodynamic interactions for a
specific grid spacing. Our results show a clear evidence for very long range hydrodynamic
71
effects between spherical particles that are moving in a confined fluid which is represented
using an atomistic model. For the given geometrical configuration and the state point chosen
(which is a typical state point used in flow simulations), the range of this hydrodynamic
interactions is shown to be longer than 90 i.e. much longer than the length scale of a typical
intermolecular interaction potential such as the Lennard-Jones potential. This observation
suggests that molecular simulations of processes dominated by interparticle hydrodynamics
will necessitate usage of very large box sizes to avoid artifacts due to hydrodynamic
interactions between periodic images of the moving particles. The observed qualitative and
quantitative (for the one case of no-slip at the sphere surface) agreement between the
molecular simulation and continuum results is encouraging. Based on these observations,
one can conclude that the continuum results can be utilized for determining the magnitude of
hydrodynamic interactions between periodic images in molecular simulations of particulate
systems and thus help determine the box sizes that need to be used to avoid artifacts due to
hydrodynamic interactions between periodic images. At higher particle volume fractions, the
magnitude and range of the hydrodynamic interactions between the particles will show a
strong dependence on the concentration of the suspension; these effects will be studied in
future work.

72
4. CHAPTER 4

IV. MOLECULAR HYDRODYNAMICS IN NANOPARTICLE
SUSPENSIONS: FRICTION FORCE AND TORQUE ON A
ROUGH NANOPARTICLE

Introduction

The effect of the solid surface on both the friction force and the torque experienced
by a macroscopic spherical particle near a surface has been studied using the continuum
theory.[119-121] Quantitative description of the hydrodynamic interactions experienced by a
nanoscopic particle necessitates appropriate accounting of the effects such as the boundary
conditions on the nanoparticle surface. The importance of slip for interpreting the
experimental results on the rotational relaxation times of molecules has been recognized[141]
and subsequently expressions were developed for the rotational friction coefficients of non-
spherical objects and molecules in the presence of slip boundary conditions[142, 143] At the
molecular scales, both collisional and hydrodynamic or collective modes of momentum
transport near the surface of a moving particle were shown to play an important role.[144-
148] As opposed to translation, for the case of rotation, a very small displacement of solvent
atoms takes place and thus collisional dynamics was considered to play a more important role
than the collective effects for suspended particles of size comparable to the solvent particles.
It was shown that this breakdown of the collective or hydrodynamic effects near the particle
surface could be captured by a partial slip boundary condition at the particle surface.[144,
145]
Indeed, the hydrodynamic boundary condition at the nanoparticle surface was shown
to be a strong function of the particle-solvent interactions and the particle surface structure,
73
and the presence of slip at the particle surface was demonstrated in equilibrium molecular
dynamics (MD) simulations under certain conditions.[124] MD simulations also showed the
presence of slip at the nanoparticle surface for the case of a nanoparticle moving towards a
solid surface and the case of a nanoparticle rotating in a molecular solvent.[128] In recent
work,[118] and as discussed in the previous chapter, MD simulations were used to study the
hydrodynamic interactions experienced by a nanoparticle that was translating parallel to the
solid surface in a confined suspension. A molecular solvent was considered explicitly in that
study; this allows for the direct determination of the velocity field in the solvent and hence
the determination of the boundary condition at the nanoparticle-fluid interface. In addition to
both the effect of the confining surfaces and the slip at the nanoparticle surface, the
hydrodynamic interactions between the periodic images of the moving particle in the
simulation were also found to play an important role in determining the value of the friction
force experienced by the particle. The nanoparticle was represented as a smooth sphere in
that work and the solvent could not exert any torque on the particle.
The nanoparticle used in this work is formed by connecting a number of beads of the
same size as the solvent beads; such a nanoparticle exhibits surface roughness. The
roughness of the particle is expected to alter the boundary conditions at the particle surface.
More importantly, such a model allows for characterization of the torque experienced by the
nanoparticle while rotating in the solvent. The translation and rotation of this model
nanoparticle in a molecular solvent near a solid surface is studied using MD simulations.
The simulation results for the friction force and the torque experienced by this nanoparticle
are compared with the values predicted by the continuum theory. The effect of the
cooperative hydrodynamic interactions between the periodic images of the moving particle
74
on the friction force and the torque values is quantified and this behavior is compared with
the previous results for the translation of a smooth nanoparticle near a solid surface.

Simulation Method

Simulation system consists of three types of atoms: the wall atoms, the solvent atoms
and the atoms comprising the suspended nanoparticle. Figure 4.1 shows a schematic of the
system used in these simulations.

Fluid Fluid

Figure 4.1: A schematic of the simulation system showing rough nanoparticle, fluid and the wall atoms.

All the atoms in the system interact via a purely repulsive Lennard-Jones (i.e.
WCA)[100] potential with the same length and energy parameters; reduced Lennard Jones
(LJ) units are used to report all of the quantities in this work.

Rough nanoparticle generation

The spherical particle used in the previous chapter was with a smooth surface. To
generate a rough nanoparticle, a face centered cubic (FCC) lattice was generated and then a
75
spherical region of required radius was carved out from the lattice. The FCC lattice spacing
is chosen such that the lattice points can be of the size of a solvent atom without overlapping
with other atoms. The lattice spacing should also be small enough such that no solvent atom
can penetrate through the nanoparticle. The nanoparticle of radius 2.5 thus generated
consists of 87 beads that are arranged in the face centered cubic fashion with a lattice spacing
of 1.42 as shown Figure 4.2. A similar procedure can be applied to generate nanoparticles of
different shapes such as ellipsoidal, cubical etc. A spherical shape was chosen in this study
to make direct comparison with previous studies in the literature with spherical particles.

Figure 4.2: A schematic of the rough nanoparticle made up of beads of the size same as the solvent atoms
and arranged in FCC fashion. The beads of the nanoparticle are connected together using a simple
harmonic potential.

These beads of the nanoparticle are bonded to each other using a harmonic potential:

2
0
) (
2
) ( b b
k
b U
b
= (4.1)
where k
b
is the force constant, b is the distance between beads and b
0
is the equilibrium bond
distance. The distances between the beads of the nanoparticle in the initial structure are
stored to be used as the equilibrium distances in the harmonic potential. A force constant of
500 was used in this work. The integrity of the nanoparticle was verified at all times;
furthermore, the nanoparticle was observed to undergo a very small shape change, if at all,
76
for this value of the spring constant used. The mass of each bead was calculated using the
method described in the previous chapter used by Schmidt and Skinner.[123] If M is the
mass of the whole nanoparticle then,

3
1
2
|
\
|
+ =
a
m M (4.2)
where m is the mass of solvent atom and a is the size of the nanoparticle. Using radius of
gyration of the nanoparticle as the size of the nanoparticle (
g
R a = ), 1 = m and 1 = ,
mass of the nanoparticle is obtained. The mass of the nanoparticle is then divided by the
number of beads in the nanoparticle to obtain the mass of each bead.

The periodic boundary conditions were applied in x and y directions while the
confining channel walls were normal to the z direction. These channel walls consisted of
atoms placed in a FCC lattice, the wall atoms were attached to these lattice sites by harmonic
springs with a force constant of 500.
Two types of simulations were performed: translation and rotation of the
nanoparticle. For translation, the nanoparticle was moved along the x direction by applying a
harmonic potential to the center of mass of the nanoparticle. The calculated force on the
center of mass of the nanoparticle was uniformly distributed over all of the constituent beads.
The nanoparticle was constrained in y and z directions at the channel center using a harmonic
potential with force constant of 100. The translation was performed by pulling the
nanoparticle in positive x direction using a harmonic potential with a force constant of 100.
The use of force constants in various potential forms necessitates a study of effect of these
force constants on the quantity of interest. To this end a comparative study was performed to
77
study the effect of force constant for harmonic force between the beads on the friction force
keeping the pulling force constant to be the same. Figure 4.3 shows the friction force on a
nanoparticle measured for a system size of 33.8 x 23.4 x 30.0, for four different values of
bead force constants keeping the pulling force constant to be 100. It was observed that the
friction force values are independent of the bead force constants considering the uncertainties
on the individual data points.
1.75
1.85
1.95
2.05
2.15
2.25
2.35
0 200 400 600 800 1000 1200
Force constant for beads
F
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.3: Friction force experienced by the translating nanoparticle plotted versus the force constant
for the harmonic potential between the beads of the nanoparticle. The system size was 33.8 x 23.4 x 30.0.
The nanoparticle pulling force constant of 100 was used.

A similar study was performed to study the effect of puling force constant on the
friction force observed by the rough nanoparticle. Figure 4.4 shows a plot for friction force
on the nanoparticle for three different pulling force constants (half and twice the value used
in these simulation) for the same system (33.8 x 23.4 x 30.0) and for a bead force constant of
500. The friction force values for the three pulling force constants were similar considering
78
the uncertainties on each data point. Hence, it was observed that the quantity of interest (i.e.
the friction force) was independent of the force constants used for harmonic potentials for
keeping the beads of the rough nanoparticle together or for the harmonic potential used for
pulling the nanoparticle in a particular direction of translation or rotation.

1.85
1.9
1.95
2
2.05
2.1
2.15
2.2
2.25
2.3
0 50 100 150 200 250
Pulling force constant
F
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.4: Friction force on the translating nanoparticle plotted for three different values of force
constant for the harmonic potential used for pulling the nanoparticle. The force constant for harmonic
potential between beads of the nanoparticle was kept constant at 500.

For the purposes of rotation, again harmonic spring force was applied on each bead of
the nanoparticle based on the target positions of these beads at the given time for the given
angular velocity. The nanoparticle was rotated about the y axis. The number density of the
simulation system was kept at = 0.844 which is representative of a condensed phase, while
the temperature of the walls was maintained at T = 1.0 by coupling the wall atoms to a heat
bath.[102] MD simulations were carried out using the velocity Verlet algorithm with a time
79
step of 0.003. The simulation runs consisted of an equilibration stage of at least
6
10 5 . 2
steps and a production stage of at least
6
10 25 steps. The production stage was divided into
10 equal blocks and the method of block averaging was used to calculate the average
properties and associated uncertainties. The periodic boundary conditions in the y and z
directions create a grid consisting of the moving nanoparticle and its images. The
hydrodynamic interactions between these image particles in the grid are a function of the
distance between these. As described in the previous chapter for a smooth nanoparticle, we
had quantified the effect of this grid spacing on the hydrodynamic interactions between the
images for a translating nanoparticle. The effect of grid spacing on hydrodynamic
interactions is captured in this work by studying a range of simulation box sizes for
simulations consisting of both nanoparticle translation and rotation. For these simulations,
the channel height (i.e. wall to wall separation) is maintained at a constant value of H = 30.0
while the lateral i.e. the x and y dimensions of the simulation box were changed from 23.4 to
91.0 resulting in 13 different simulation box sizes. Quantitative comparison of the
simulation results with the continuum theory necessitates the value of fluid viscosity; at the
simulation conditions studied in this work, the fluid viscosity was determined to be 2.16 in
the previous work.[118]

Results

The presence of a solid nanoparticle in the simulation system induces structure in the
fluid surrounding the nanoparticle. Figure 4.5 shows the solvent density profile for both
translating nanoparticle and a rotating nanoparticle plotted as a function of the distance from
center of mass of the nanoparticle.
80

0 1 2 3 4 5 6 7
r
0
0.5
1
1.5
2
S
o
l
v
e
n
t

d
e
n
s
i
t
y

p
r
o
f
i
l
e

Figure 4.5: Density profile of solvent atoms as a function of distance from the center of mass of the
nanoparticle for a translating nanoparticle with a velocity of 0.02 (circles) and for a rotating nanoparticle
with angular velocity of 0.008 (dotted line) in a simulation box of dimensions 0 . 30 8 . 33 8 . 33 . The
average error bars on density values are of size 0.001.

As can be seen from the figure, the solvent density profile is virtually identical for
these two cases. The profile shows that there are two strong layers and a weak third layer of
the solvent around the nanoparticle indicating the presence of nanoscale density
heterogeneities in these systems due to the molecular nature of the solvent. The rough nature
of the nanoparticle surface leads to the occurrence of a small shoulder in the density profile
right after the first peak, such a shoulder in the density profile was not observed for fluid
structuring around a smooth nanoparticle.[118]

81
Friction force and torque on a rough nanoparticle in monomeric solvent

The principal quantities of interest in this work are the friction force acting on the
translating nanoparticle and the torque experienced by the rotating nanoparticle. For a
spherical solute in an unbounded fluid, the friction force (F) experienced by the translating
solute and the torque (T) experienced by the rotating solute are given by the continuum
theory as:
V r c F
p F
= (4.3)
=
3

p T
r c T (4.4)
where is viscosity of the fluid, r
p
is radius of the solute, V is the translation velocity and
is the angular velocity of rotation. The coefficient
F
c assumes the values of 6 and 4 for no-
slip and slip boundary conditions respectively, whereas the coefficient
T
c has values of 8
and 0 for these conditions respectively. The value of r
p
that is to be used when applying
these expressions requires some consideration. It has been suggested that rather than using
the actual radius of the nanoparticle, the hydrodynamic radius should be used in the
application of the continuum theory expressions.[123] In our previous work as well, the
simulation data were found to be well represented quantitatively by the continuum
expressions if the hydrodynamic radius is used.[118] Given these observations, in this work,
we have employed the effective hydrodynamic radius represented by,
+ =
g H
R R (4.5)
as the radius of the nanoparticle in our analysis, where R
g
is radius of gyration of the
nanoparticle and is the LJ diameter of the solvent atoms. The value of the hydrodynamic
82
radius thus determined was 2.93 . We note that the first peak in the radial distribution
function of the solvent (Figure 4.5) around the nanoparticle occurs at a distance of 2.85
which is consistent with the above value of the hydrodynamic radius.
We begin with an investigation of the hydrodynamic interactions between the moving
nanoparticle and its periodic images that form a grid. As described in the previous chapter
for a smooth nanoparticle, these interactions were shown to be a function of the grid spacing.
In this work, for quantifying the dependence of the hydrodynamic interactions on the grid
spacing in the x direction (box dimension in the x direction), the grid spacing in the y
direction was held constant at 33.8 and the x direction grid spacing was varied between 23.4
and 91. Two sets of simulations were carried out for each grid configuration: the
nanoparticle was translated in the x direction and the friction force experienced by it was
calculated in the first set, while in the second set, the nanoparticle was rotated about the y
axis and the torque experienced by it was calculated. For this purpose, the friction force on
the nanoparticle was determined as the sum of the LJ forces exerted by the solvent beads on
the nanoparticle beads. Similarly the torque on the nanoparticle was obtained by a vector
sum of the torques on all of the individual beads that comprise the nanoparticle. These
friction force and torque values were normalized by the continuum theory values given by
equations (4.3) and (4.4) respectively for the no-slip conditions.
For each simulation box size (i.e. periodic grid spacing), particle translation
simulations were carried out at six different velocities spanning the range 0.002 to 0.05. The
friction force was verified to vary linearly with the velocity and the value of the coefficient
F
c was then obtained from the slope of the linear fit to the friction force-velocity results as
described in chapter 3. Similarly, the particle rotation simulations were also carried out at
83
five different values of the angular velocity spanning the range 0.002 to 0.016 and it was
verified that the torque showed a linear dependence on the angular velocity. The value of the
coefficient
T
c was then obtained analogously from the slope of the linear fit to the torque-
angular velocity data. These friction force and torque values were normalized by dividing
these
F
c and
T
c values by the no-slip boundary condition continuum values of 6 and 8
respectively. Figure 4.6 shows the normalized friction force and normalized torque on a
rough nanoparticle plotted against the grid spacing in x direction (L
x
) normalized by the
channel height (H) for a constant grid spacing in the y direction (L
y
). Table 4.1 shows a grid
of various simulation systems studied in these work.
Table 4.1: A table showing the simulation system sizes used in chapter 4
L
Y
\ L
X
23.4 33.8 45.5 55.9 67.6 79.3 91
23.4
33.8
45.5
55.9
67.6
79.3
91

The friction force experienced by the translating nanoparticle increases with an
increase in the grid spacing along the direction of motion of the nanoparticle and does not
attain a steady value even at a grid spacing of 91. These results are consistent with the
continuum predictions[116] and the previous MD simulation results for a translating smooth
nanoparticle[118] and can be attributed to the cooperative hydrodynamic interactions
between the image particles, the magnitude of which decreases with an increase in the grid
spacing. The torque experienced by a rotating nanoparticle, on the other hand, seems to be
insensitive to the x grid spacing. This result can be explained by noting that unlike the
84
translating grid of nanoparticles, a grid of rotating nanoparticles does not set up a long range
flow field in the system which could lead to hydrodynamic interactions on length scales
much longer than the range of the intermolecular interactions.

0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e
,

N
o
r
m
a
l
i
z
e
d

t
o
r
q
u
e

Figure 4.6: Normalized friction force (squares) and normalized torque (triangles) as a function of
normalized channel length. The channel height is H = 30.0. Error bars on the data points are of the same
size as the size of symbols used.

Similarly, in order to quantify the effect of y grid spacing on the hydrodynamic
interactions, two sets of simulations were carried out by keeping the x grid spacing constant
at 33.8 and varying the y direction grid spacing. The results shown in Figure 4.7 indicate that
in this case, the friction force on the translating nanoparticle increases with an increase in the
y grid spacing, but ultimately approaches a constant value in the range of y grid spacings
studied. This observation is consistent with that from the MD simulation of a translating
smooth nanoparticle[118] and the continuum results[116]. On the other hand, the torque
85
experienced by the rotating nanoparticle again is not affected by the change in the y direction
grid spacing.
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e
,

N
o
r
m
a
l
i
z
e
d

t
o
r
q
u
e

Figure 4.7: Normalized friction force (squares) and normalized torque (triangles) as a function of
normalized channel width. The channel height is H = 30.0. Error bars on the data points are of the same
size as the size of symbols used.

In Chapter 3, normalized values of friction force (calculated using the effective
hydrodynamic radius) for a translating smooth nanoparticle were shown to be lower than the
continuum predictions[116] and this observation was attributed to the slip at the nanoparticle
surface. The surface roughness of the nanoparticle is expected to reduce the degree of slip.
A comparison of the friction force values obtained in this work with those in the previous
chapter shows that, the friction force experienced by the rough nanoparticle is higher than the
simulation results for a smooth nanoparticle at identical conditions[118] but the values are
still lower than the continuum predictions.[116] Figure 4.8 shows the velocity field along the
axes in the yz plane around the nanoparticle translating in x direction with velocity 0.02. The
velocity of the fluid adjacent to the nanoparticle surface is lower than the nanoparticle
86
velocity (by ~ 20%) thus confirming the presence of slip at the surface, the degree of slip is
observed to be lower than that for a smooth nanoparticle (comparing with Figure 3.11 in the
previous chapter).

Figure 4.8: Fluid velocity profile (only the x component of fluid velocity is plotted for the cubes residing
along the axes) in the yz plane for the system with dimensions 0 . 30 8 . 33 8 . 33 . The sphere resides at
the center of the figure and is translating with velocity V = 0.02 along the x direction. The uncertainties in
the data range from 0.002 to 0.007, with the average uncertainty being 0.004.

The reduction in the degree of slip for the rough nanoparticle could be a function of
just the surface topology or a combination of the surface topology and solvent packing
around the nanoparticle. A comparison of the radial distribution function (RDF) of the
solvent around the nanoparticle (Figure 4.5 in this chapter and Figure 3.4 from the previous
chapter) indicates that although the height of the first peak in the solvent RDF is much higher
for the smooth nanoparticle, there are approximately 12% more solvent beads in the first
87
shell (defined as the region covered by the first peak up to the minimum between the first and
the second peaks) around the rough nanoparticle than the smooth nanoparticle. The presence
of slip at the rough nanoparticle surface is also observed for the rotating nanoparticle.
Finally, changing the nanoparticle-solvent interactions from WCA potential to LJ potential,
and increasing the strength of the attraction, is found to lead to a small increase in the friction
force but a substantial increase in the torque values due to a reduction in the slip as can be
seen in figure 4.9.

Figure 4.9: Fluid velocity profile (x component of fluid velocity is plotted for the cubes residing along the
z axes, while z component of velocity is plotted for cubes residing along x axes) in the xz plane for the
system with dimensions 0 . 30 8 . 33 8 . 33 . The sphere resides at the center of the figure and is rotating
with velocity = 0.008 (corresponds to velocity of 0.02) about the y axis. The uncertainties in the data
range from 0.002 to 0.007, with the average uncertainty being 0.004.

Specifically, simulations carried out with LJ interaction potential between the
nanoparticle and solvent beads (all other interactions being WCA) with the parameter value
88
of 5 . 1 =
solvent le nanopartic
, resulted in about 4% increase in the friction force and
more than 80% increase in the torque over the values obtained using the purely repulsive,
WCA potential for this interaction.
Friction force on a rough nanoparticle in a polymer melt

Friction force experienced by a rough nanoparticle translating in a polymer melt was
also studied using a similar simulation setup. The simulation system consisted of a rough
nanoparticle, wall atoms and a polymeric melt as a medium. All the atoms in the system
interacted via WCA potential. The characteristics of the nanoparticle and the wall atoms and
the force constants used were same as described earlier in this chapter. The polymeric melt
was made of polymer chains each of length N = 20 and the preparation procedure of polymer
melt will be discussed in details in the next chapter. The state point used in these simulations
was of = 0.84 and T = 1.2 to ensure that the system is at a temperature far away from its
glass transition temperature and is in a condensed phase. Nanochannel walls were made up
of beads arranged in FCC fashion. Periodic boundary conditions were applied in x and y
directions whereas the system was confined in z direction.
The nanoparticle was pulled in x direction through the polymer melt and friction force
experienced by the nanoparticle was measured. The friction force plotted against the pulling
velocity demonstrated a linear dependence of friction force on the velocity. The friction
coefficient was obtained from the slope of the linear plot of the friction force and the
velocity. This calculation of friction coefficient requires the value of viscosity for the
polymer melt. The viscosity of the polymer melt was calculated by performing planar
Couette flow simulations (as described in the previous chapter) on the polymer melt and was
89
obtained to be 7.58. The friction coefficient thus obtained was normalized by the Stokes
law value of the friction coefficient for no-slip boundary condition.
0.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.10: Normalized friction force on a nanoparticle translating in a polymer melt plotted as a
function of normalized channel length. The channel height is H = 30.0. Error bars on the data points are
of the same size as the size of symbols used.

To study the effect of cooperative hydrodynamic interactions simulations were
performed for a range of system sizes mentioned earlier in the chapter. Figure 4.10 shows a
plot of normalized friction coefficient plotted against normalized channel length (normalized
using the channel height). It can be observed that the cooperative hydrodynamic interactions
are operative over long length scales and the friction force experienced by the nanoparticle
doesnt attain a steady value even at a grid spacing of 91.0, similar to the behavior observed
for monomeric solvent described earlier. A similar set of calculations were performed for
dependence of simulation box size in the y direction. Figure 4.11 shows a plot of normalized
90
friction coefficient plotted as a function of normalized channel width (normalized using
channel height).

0.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
N
o
r
m
a
l
i
z
e
d

f
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.11: Normalized friction force on a nanoparticle translating in a polymer melt plotted as a
function of normalized channel width. The channel height is H = 30.0. Error bars on the data points are
of the same size as the size of symbols used.

Consistent with the behavior observed for the monomeric solvent, the friction force
experienced by translating nanoparticle increases with increase in box dimension in y
direction but attains a stable value at a smaller distance as compared to earlier case of effect
in x direction.

91
Concentrated suspension of nanoparticles

A study of cooperative hydrodynamic interactions was performed for a suspension
with increased concentration of nanoparticles in monomeric solvent. The simulation
conditions were identical to those described earlier in the chapter for monomeric solvent: =
0.844, T = 1.0 and WCA interactions. The nanoparticle volume fraction of = 0.1 was
maintained for all the systems studied. Simulation system sizes identical to those mentioned
earlier in Table 4.1 were used to study cooperative hydrodynamic effects in x and y direction.
The friction force on the nanoparticle, translating through this concentrated suspension of
identical nanoparticles and the monomeric solvent, was calculated. However, the systems
simulated here had issues related to the viscosity and pressure and the results obtained should
be considered as preliminary results. Figure 4.12 shows the channel length, L
x
, dependence
of friction force experienced by the translating nanoparticle. Similar to the behavior obtained
earlier for the case, of a dilute suspension of nanoparticles, the friction force increases with
the increase in the channel length and eventually attains a steady value. The results show the
presence of long range cooperative effects for these systems. Similar plot for the dependence
on channel width, L
y
, is presented in Figure 4.13. The friction force increases with the
channel width and attains a steady value at much shorter distances as compared to the earlier
case of channel length dependence. The friction force values for the biggest system
simulated (last point in figures 4.12 and 4.13) show deviation from the general behavior and
could be because of the bad quality of data or because of the increase effect of the issues
mentioned earlier with the increase in number of nanoparticles in the system. The possibility
of a shift in the friction force curve could not be checked due to the unavailability of the data
for the bigger simulation systems due to computational limitations.
92
5
5.5
6
6.5
7
7.5
8
0 0.5 1 1.5 2 2.5 3 3.5
L
x
/H
F
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.12: Friction force on a nanoparticle translating in a concentrated suspension of nanoparticles
plotted as a function of normalized channel length. The channel height is H = 30.0. Error bars on the data
points are of the same size as the size of symbols used. The simulation conditions are: =0.844, T=1.0 and
WCA interactions.

5
5.5
6
6.5
7
7.5
0 0.5 1 1.5 2 2.5 3 3.5
L
y
/H
F
r
i
c
t
i
o
n

f
o
r
c
e

Figure 4.13: Friction force on a nanoparticle translating in a concentrated suspension of nanoparticles
plotted as a function of normalized channel width. The channel height is H = 30.0. Error bars on the data
points are of the same size as the size of symbols used. The simulation conditions are: =0.844, T=1.0 and
WCA interactions.

93

In this work, we have studied the translation and rotation of a rough nanoparticle in
molecular solvent using MD simulations. The measured values of the friction force and the
torque experienced by the nanoparticle are analyzed in terms of the hydrodynamic radius of
the nanoparticle. Simulations carried out using different periodic box sizes indicate that
while the friction force values show a strong dependence on the box size, the torque values
are insensitive to the box size. This result can be explained in terms of the cooperative
hydrodynamic interactions between the periodic images of a translating particle that result
from the collective, long range flow field that is set up in the system. On the other hand,
such a long range flow field is not set up by a periodic grid of rotating particles, thus leading
to a lack of dependence of the torque values on the simulation box size. An analysis of the
velocity field of the solvent beads around the moving nanoparticle indicates that although
partial slip boundary conditions hold for the rough nanoparticle, the degree of slip is reduced
as compared with a smooth nanoparticle. The magnitude of the friction force on the
translating nanoparticle and especially the torque experienced by a rotating nanoparticle are
reduced as compared to the continuum expectations due to the occurrence of slip at the
particle surface. The friction force experienced by the rough nanoparticle translating in a
polymer melt was also studied. The cooperative hydrodynamic interactions were found to
exist in the polymer melt as well with systems exhibiting the effect of these interactions over
a long range.
The systems simulated and the state points used in this chapter are the commonly
studied systems in literature for the prediction of the transport properties. However, the
simulation box sizes used are not always chosen considering the effect of the long range
94
hydrodynamic interactions. Incorrect choice of simulation box sizes can introduce artifacts
in the calculations of transport properties if the effect of hydrodynamic interactions is not
taken into consideration. The results of this study highlight the importance of careful
consideration of boundary conditions and the long range cooperative hydrodynamic
interactions. The qualitative agreement of the simulation results with the continuum
mechanics predictions show that continuum methods, which are computationally lesser
expensive, can be used to capture the hydrodynamic behavior if the effects of slip at the
nanoparticle surface are accounted for.

95
5. CHAPTER 5

V. ACTIVE NANORHEOLOGY IN POLYMER MELT

Introduction

Rheology is the study of flow and deformation. It comes from a Greek verb
(meaning flow) and was coined by a professor of chemistry at Lehigh University, Eugene
Bingham, in 1920.[149] It is primarily used to study the mechanical properties of
viscoelastic materials by studying the mechanical response of the material to the applied
shear using a Rheometer. Several types of geometries have conventionally been used in a
Rheometer such as cone-and-plate, concentric cylinders (Couette), parallel plates etc. A
typical rheological measurement requires a sample volume in milliliters range. This puts a
severe limitation on the investigation of biological and expensive samples. Limitations also
come in the form of the range of frequency over which the sample can be probed since the
conventional rheological techniques can probe material behavior over a frequency range of
tens of Hz. Several biological materials such as living cells operate on the range of
micrometer length scales. To investigate the properties of the cellular fluids, the techniques
need to consider the length and time scales relevant to these systems. Moreover, the
materials under consideration are not always homogeneous and can contain artificial or
inherent inhomogeneities. The inhomogeneous nature of the material can be captured via
local rheological measurements, which is not possible in a conventional rheometer. These
limitations, along with the advent of particle tracking techniques, were the major factors
behind the development of a new experimental technique that uses microliter quantity of
sample volume and can probe the material response over a wide range of frequencies without
96
the use of time-temperature superposition. The technique is based on the idea of using
micron sized spheres as probes in microliter sample volumes to study the local material
response on micrometer scales and is termed as Microrheology.

Microrheology

Advances in the particle tracking videomicroscopy, optical/magnetic tweezers and
atomic force microscope(AFM) have been instrumental in the development of newer
experimental methodologies to use microrheology for over more than the past decade to
explore the local viscoelastic properties.[69, 70, 150-154] Detailed reviews on particle
tracking techniques[155, 156], laser tweezers[157-159], magnetic tweezers[160], force
microscopy[161] and diffusing wave spectroscopy[162] can be found in literature. The
potential of microrheological techniques for probing viscoelastic behavior of living cells and
cellular components has attracted a great deal of attention towards these techniques. Several
studies were targeted towards determination of microstructure and calculation of viscoelastic
properties for cellular components such as F-actin and other cytoskeletal networks using
microrheology.[163-171]
Suspension of particles dispersed in a complex fluid are also of primary importance
to various industrial applications such as in coatings, paints, pharmaceuticals,
nanocomposites etc. and microrheological techniques have been used to study the
viscoelastic behavior of these suspensions.[158, 172, 173] A comprehensive review of
several other systems that have been explored using microrheology is given by Waigh.[66]
Stokesian dynamics simulations of microrheology of colloidal dispersions have also been
97
published by Brady and coworkers.[174-177] Furthermore Brownian dynamics simulations
of particle-tracking rheology in colloidal suspensions have also been reported.[178]

Types of microrheological techniques

The microrheological techniques can be divided into two broad categories: In the
first category, termed Passive Microrheology, the probe particle is suspended in the sample
and the thermal fluctuations of the probe particle due to Brownian motion are tracked, while
the second category involves active manipulation of the suspended probe particle and the
surrounding fluid interaction with the probe is studied; and the technique is termed as Active
Microrheology.

Passive Microrheology

The probe particle that is embedded in the viscoelastic medium undergoes thermal
fluctuations due to the Brownian motion. The passive microrheology techniques have been
carried out by tracking the single probe particle embedded in complex fluid directly [70-72,
179] or by more indirect observations of the intensity fluctuations by a group of embedded
probe particles.[69, 73, 151, 180] The underlying physics behind passive microrheology
techniques is that the macroscopic nature of the polymer solution governs the response of the
probe particle to the thermal fluctuations. Mason and Weitz in their pioneering work in 1995
led foundations of most of the passive nanorheology techniques developed thereafter by
using Dynamic Light Scattering (DLS) to track the probe particle.[69, 151] They proposed
98
the usage of generalized Stokes-Einstein relation (GSER) to obtain the viscoelastic modulus
of the medium from the measured mean squared displacement of the probe particle:

) (
~

) (
~
) (
~
2
s r as
T k
s G s s G
B
r

= (5.1)
where ) (
~
s G is the Laplace transform of G(t), the viscoelastic modulus, s is the Laplace
transform variable, ) (
~
s G
r
is the Laplace transform of stress relaxation modulus ) (t G
r
, T is
the temperature, a is the bead radius, and ) (
~2
s r is the Laplace transform of mean squared
displacement. Equation 5.1 is the frequency dependent form of the GSER. Using this
equation the macroscopic viscoelasticity of the complex fluid can be determined using the
microscopic response to the bead motion. The MSD in time domain is transformed into
MSD in Laplace frequency domain and then substituted in equation 5.1 to calculate ) (
~
s G . A
number of different studies building on this framework have later been used to explore the
microrheological behavior.[70-72, 153, 181, 182]

Active Microrheology

In active microrheological techniques the probe particle is manipulated by application
of a systematic force on the probe particle. The response of the material to the motion of the
probe particle is measured and is used to calculate the viscoelastic properties. A high-
bandwidth one- and two-particle microrheology technique was developed by Mizuno and
coworkers using an optical trap combined with high resolution tracking methods.[153] The
Fourier transformed response to the probe motion combined with generalized Stokes
99
relation[69] was used to calculate the complex viscoelastic moduli. The frequency
dependent response of the probe particle was defined as:
) ( ) ( ) ( F A x = (5.2)
where x is the displacement of the probe particle, F is the force on the particle and A is the
complex response function. The true response function based on the simulation setup was
calculated as:

) ( ) ( 1
) (
) (
2 1

A k k
A
+
= (5.3)
where k
1
and k
2
are the trap constants for the laser drive and the probe respectively. The
response function combined with GSER then was used to calculate the complex shear
modulus:

) ( 6
1
) (

a
G = (5.4)
where,
) ( ) ( ) ( G i G G + = . (5.5)
Sinusoidally oscillating optical tweezers were used to manipulate the probe particle to
develop a technique to calculate forces on a colloidal particle.[173] An extension of this
study resulted in a development of a probe for mechanical testing of nanostructures in soft
materials.[183] Optical tweezers were also used to study the correlated motion of two
hydrodynamically coupled oscillating particles trapped in two different potential wells and to
deduce the response as a function of frequency of oscillations.[110]
A theoretical framework for active and nonlinear microrheology was developed by
Squires and Brady to study microstructural deformations.[184] The authors discussed the
various modes of active microrheology experiments such as fixed-velocity, fixed-force and a
100
mix of two modes. They calculated the microviscosities for these different modes and
presented a new theory to probe microrheological properties of concentrated suspensions.
Nonlinear microrheology with special attention to bulk stresses was studied in detail by
Squires recently.[185] In his study of driven motion of a probe particle through the bath
particles, he studied the different sources of stresses in the probe-bath collisions and
discussed their corresponding analogs in the macrorheological experiments.

Nanorheology

Molecular Dynamics simulation technique with its capabilities to perform atomistic
simulations provides an opportunity to investigate the viscoelastic behavior in the nanoscale
systems and to perform Nanorheology. Although Microrheology and Nanorheology have
different scales of operation, they work on the same theoretical framework of tracking the
motion of very small probe particle suspended in a medium and then inferring the
viscoelastic properties of the medium from these measurements. Nanorheological techniques
can be specifically useful for the investigation of systems where the properties can change on
nanoscopic length scales such as in polymer nanocomposites or in polymer thin films.[186]
With regards to the classification for microrheological techniques described above we
have developed an active nanorheology technique using molecular dynamics simulations to
calculate the viscoelastic properties of a polymer melt. The technique is based on the idea of
applying an oscillating force to the nanoparticle suspended in a polymer melt and then
measuring the response of the polymeric melt by tracking the displacement of the
nanoparticle. The nanoparticle is made to oscillate in a harmonic trap and the complex
viscoelastic moduli are determined using the calculation procedure similar to the one
101
described by Hough and Ou-Yang [187] and is described in detail later in the chapter. The
simulation results are compared with two different literature simulation studies where
viscoelastic properties of an identical polymer melt were calculated using different
simulation techniques.[188, 189]

Simulation Method

The simulation setup used in these active nanorheology simulations is essentially an
extension of the setup described in chapter 4. The simulation system consists of three kinds
of atoms atoms comprising a rough nanoparticle, polymer melt and wall atoms. Figure 5.1
shows a schematic of the system used in these simulation.
Nanoparticle
wall
wall
Polymeric melt
z
x
Nanoparticle
wall
wall
Polymeric melt
z
x
z
x

Figure 5.1: A schematic of the simulation system showing a rough nanoparticle in a polymer melt
confined between walls of the nanochannel.

All the atoms in the system interact via a Lennard-Jones (LJ) potential or the purely repulsive
Lennard-Jones (WCA) potential with same length and energy parameters. All the quantities
expressed in this chapter are in reduced Lennard-Jones units.
The nanoparticle used in this work is the same as described in Chapter 4. The
nanoparticle is a rough nanoparticle made up of 87 beads arranged in an FCC lattice fashion
102
with a lattice spacing of 1.42. The beads of the nanoparticle are connected together using
harmonic potential:

2
0
) (
2
) ( b b
k
b U
b
= (5.6)
where k
b
is the force constant, b is the distance between beads and b
0
is the equilibrium bond
distance. A force constant of 500 was used in this work. The integrity of the nanoparticle
was verified at all times; furthermore, the nanoparticle was observed to undergo a very small
shape change, if at all, for this value of the spring constant used. The shape of the
nanoparticle is verified by keeping track of the x, y and z component of the radius of gyration
of the rough nanoparticle. The periodic boundary conditions were applied in x and y
directions while the confining channel walls were normal to the z direction. These channel
walls consisted of atoms placed in a FCC lattice and the wall atoms were attached to these
lattice sites by harmonic springs with a force constant of 500.
The polymer chains are modeled as bead-spring chains. The beads are connected by
finitely extensible non-linearly elastic (FENE) springs represented by the potential:

(
(
|
|
\
|
=
2
2
1 ln
2
) (
Q
r kQ
r
FENE
(5.7)
where k is the spring constant and Q is the maximum extension of the spring. The values of
the parameters chosen are k = 30 and Q = 1.5. The number density of the system is kept at
= 0.85 while the temperature of the system is maintained at T = 1.0 by coupling the wall
atoms in the system to a heat bath using a method described in an earlier work.[102] The
state point and the parameters in the FENE potential are the same as those used in the
literature studies with which our results will be compared.[188, 189] To perform the
calculations at any given location in the nanochannel the rough nanoparticle is oscillated in
103
the x direction keeping y and z direction coordinates constrained. The constraint is employed
by using a harmonic potential with a force constant of 1000. The constraint in y and z
direction is required to avoid the motion of the rough nanoparticle in the direction lateral to
the direction of oscillations.

System preparation: Isotropic orientation of polymer chains

To prepare a polymer melt the coordinates are first generated on a lattice, then the
adjacent coordinates are connected together to form the polymer chains. As a result of this
process of connecting the adjacent atoms, the polymer chains are all aligned parallel to the
walls in this configuration. Before using this polymer melt configuration as an input
structure, the polymer melt needs to be equilibrated so that the polymer chains attain
isotropic orientation to avoid the introduction of any artifacts in the system. The isotropic
orientation of the chains can be ensured by calculating the molecular axis orientational
correlation function as described by Khare et al. [103] The molecular axis orientational
correlation function is defined as:

2
1 cos 3
) (
2
2
z S
M
(5.8)
where z is the position of the center of mass of the chain and is the angle between the
molecular axis vector and the z axis which is normal to the channel walls. If the polymer
chain molecules are on average aligned parallel to the channel walls then S
2M
= -0.5, if the
chains are aligned normal to the channel walls then S
2M
= 1.0 whereas if the chains have
isotropic orientation then S
2M
= 0. Hence the system is equilibrated in step-by-step manner
and at the end of each step the molecular axis orientational correlation function is calculated,
104
and the process is repeated till it becomes zero. Once the polymer melt is equilibrated the
nanoparticle is introduced by removing a few chains from the system and then the system is
equilibrated some more to accommodate the nanoparticle.

Mathematical formulation for active nanorheology

Input Force
(sinusoidal)
Trap
Input Force
(sinusoidal)
Trap

Figure 5.2: A schematic showing the rough nanoparticle subjected to an input oscillating force oscillating
in a harmonic trap.

The simulation system consists of a suspension of nanoparticles in a polymer melt
confined between two atomistic walls. The nanoparticle is oscillated at the center of the
channel by applying an oscillatory force on the nanoparticle center of mass in the x direction.
Furthermore a harmonic trap is placed around the nanoparticle to maintain the nanoparticle at
the center of the nanochannel as shown in Figure 5.2. The force balance on the oscillating
nanoparticle in x direction can be written as:
( ) x m kx Ae x a
t i
& & & = +

6 (5.9)
105
where, a is the radius of nanoparticle, m is the mass, ( ) is the complex viscosity, x
is displacement of the nanoparticle center of mass, A is amplitude of oscillation, is
frequency of oscillation, t is time and k is force constant for the harmonic trap. The first term
on left hand side represents the force experienced by the nanoparticle, the second term is the
external oscillatory force and the third term is the harmonic trap force. This expression
assumes that the friction force on the nanoparticle follows the Stokes law with no-slip
boundary conditions. Neglecting the inertia of the particle and the fluid (the assumption of
negligible inertial effects is validated later in this chapter),
( ) 0 6 = + kx Ae x a
t i
& (5.10)
The complex viscosity is defined in terms of complex shear modulus as,
( )
( )

i
G
= (5.11)
Substituting in equation 5.10,

( )
0 6 = + kx Ae x
i
G
a
t i
& (5.12)
Rearranging,
( ) ( )
x a
kx i
e
x a
iA
G
t i
& &

6 6
= (5.13)
The displacement, x, of the nanoparticle is also expected to be sinusoidal (an assumption that
we verify) and can be represented as:
( )
( ) [ ]

=
t i
e D x (5.14)
where ( ) D is amplitude and ( ) is the phase lag between the applied sinusoidal force and
the observed sinusoidal displacement. The velocity is then obtained as follows:
( )
( ) [ ]

=
t i
e D i x & (5.15)
106
Substituting the displacement and velocity in equation 5.13 and rearranging,
( )
( )
( ) [ ]
( )
( ) [ ]
( )
( ) [ ] |
|
\
|
|
|
\
|
=
t i
t i
t i
t i
e D
e D
a i
k i
e
e
aD i
iA
G
6 6
(5.16)
( )
( )
( )
(
= k e
D
A
a
G
i

6
1
(5.17)
The exponential term can also be expressed as,

( )
( ) ( )

sin cos i e
i
+ = (5.18)
Substituting in equation 5.17 and rearranging,
( )
( )
( )
( )
( )
(
+ = k
D
A
i
D
A
a
G
sin cos
6
1
(5.19)
The complex viscoelastic shear modulus is defined in terms of real ) (G and imaginary
) (G parts as,
( ) ( ) ( ) G i G G + = (5.20)
Combining equations 5.19 and 5.20 the storage and elastic modulus can be expressed as,
( )
( )
( )
(
= k
D
A
a
G
cos
6
1
(5.21)
( )
( )
( )
(
D
A
a
G
sin
6
1
(5.22)
Thus the complex viscoelastic moduli are expressed as a function of the amplitude of
displacement ( ) D and the phase lag ( ) between the applied oscillating force and the
displacement curve. Thus for the calculation of the complex viscoelastic moduli, the
simulation data are first collected, then the displacement data averaged over many oscillation
periods are plotted against the applied oscillating input force to obtain the amplitude of
107
displacement ( ) D and the phase lag ( ) . Plugging the values of these quantities and the
constants in equations 5.21 and 5.22 gives the complex viscoelastic moduli.

Magnitude of inertial effects: Dimensional analysis

Here the dimensional analysis of the force balance equation is performed to get an
insight about the ranges of the parameters that can be used while keeping the inertial term
negligible.
Consider the force balance equation:
( ) x m kx Ae x a
t i
& & & = +

6 (5.9)
Rearranging and simplifying

|
|
\
|
= |
\
|
2
2
6
dt
x d
m kx
dt
dx
a Ae
t i

(5.23)
Nondimensionalizing this equation using a length scale of H (height of the channel) and then
by the time scale of oscillation, ,

|
|
\
|
=
|
|
\
|
2
2
6
dt
x d
m kx
dt
dx
a e
H
A
t i

(5.24)

|
|
\
|
=
|
|
\
|
2
2
2
2
6
dt
x d
m x k
dt
dx
a e
H
A
t i

(5.25)
where,

2
= (5.26)
substituting in equation 5.25,

|
|
\
|
=
|
|
\
|
2
2
2
2 2
2
2
4 12 4
dt
x d
m x
k
dt
dx a
e
H
A
t i

(5.27)
108

|
|
\
|
=
|
|
\
|
|
|
\
|
|
|
\
|
|
|
\
|
2
2
2
2 2
2
2
4 12 4
dt
x d
m x
k
dt
dx a
e
H
A
t i

(5.28)

|
|
\
|
=
|
|
\
|

2
2
3 2 1
dt
x d
m x C
dt
dx
C e C
t i
(5.29)
Now from equation 5.29 we can compare the three ratios of the coefficients-

2
2
1
1
4
A
H m
C
m
R = = (5.30)

a
m
C
m
R
2
2
2
12
= = (5.31)

2
2
3
3
4
k
m
C
m
R = = (5.32)
For our simulation setup, the values of the various parameters are:
m = mass of nanoparticle = 114.7
H = height of channel = 30.0
A = amplitude of the oscillatory force on nanoparticle = 100.0
k = force constant for the harmonic trap = 10.0
a = radius of the nanoparticle = 2.5
approximate viscosity of the melt 9.0
= frequency of oscillation of nanoparticle = 0.004, 0.006, 0.01 and 0.02

For all the values of the frequency of oscillation used in these simulations the values
of the three ratios calculated using these parameters are tabulated in the following table:

109
Table 5.1: Table of relative magnitudes of the three ratios of coefficients for the frequencies studied in
chapter 5
R
1
R
2
R
3
0.004 0.0000139 0.0001721 0.0000015
0.006 0.0000314 0.0002582 0.0000035
0.01 0.0000871 0.0004303 0.0000097
0.02 0.0003486 0.0008606 0.0000387

From the above table it can be seen that the ratios of coefficients are much smaller than 1
indicating that the terms on the left hand side of the equation are much greater than the term
on the right hand side of the equation (inertial term), and hence it validates the assumption of
neglecting the inertia in the calculation procedure. It can be observed that by increasing the
frequency of oscillation the magnitude of R
2
increases rapidly and this consideration can be
used for determining the highest frequency value that can be employed in our calculations.

Results

The primary quantity of interest in these simulations is the displacement of the
oscillating rough nanoparticle. The sinusoidally oscillating input force on the nanoparticle
causes displacement of the particle, the magnitude and phase of which could be used to
calculate the complex viscoelastic moduli for the system. As described earlier the complex
viscoelastic moduli are given by:

110
( )
( )
( )
(
= k
D
A
a
G
cos
6
1
(5.21)
( )
( )
( )
(
D
A
a
G
sin
6
1
(5.22)

The only unknowns in these expressions are the magnitude of the displacement of the
nanoparticle, ( ) D , and phase lag ( ) . The calculations were performed for four different
frequencies 0.004, 0.006, 0.01 and 0.02 to determine ( ) D and ( ) .

In the mathematical formulation of complex moduli we had assumed that the nature
of the displacement of the nanoparticle in response to the sinusoidally oscillating force would
also be sinusoidal. This assumption was verified by inspecting the displacement of the
nanoparticle. Figure 5.3 shows the displacement of the nanoparticle for one of the systems
used in our simulations. It can be seen that the nature of the displacement is perfectly
sinusoidal with slight phase lag in displacement (shifted curve) with respect to the input
oscillating force which would have the maximum at origin.

111
-10
-8
-6
-4
-2
0
2
4
6
8
10
0 0.25 0.5 0.75 1
D
i
s
p
l
a
c
e
m
e
n
t
Time
-10
-8
-6
-4
-2
0
2
4
6
8
10
0 0.25 0.5 0.75 1
D
i
s
p
l
a
c
e
m
e
n
t
Time

Figure 5.3: Measured displacement of a nanoparticle oscillating with a frequency of 0.02 in a polymer
melt in a simulation system of size 33.8x33.8x30. All the interactions are WCA.

.
Furthermore the Fourier transform of the displacement data as shown in Figure 5.3
was used to study the excitation in the system (Figure 5.4). The single peak in the Fourier
transform corresponds to the frequency of oscillation of the nanoparticle. The occurrence of
a single peak confirms that the nanoparticle is oscillating with the desired frequency and
there are no higher order harmonics in the system.
112
0
50000
100000
150000
200000
250000
300000
0 50 100 150 200
( ) x f
)
f
0
50000
100000
150000
200000
250000
300000
0 50 100 150 200
( ) x f
)
f

Figure 5.4: Fourier transform of the displacement of the oscillating nanoparticle in polymer melt for the
simulation conditions of figure 5.3.

Literature studies for comparison

The simulation results were compared with the results of two different studies in
literature for calculation of viscoelastic properties of a polymer melt for identical simulation
conditions but using different simulation techniques. Sen, Kumar and Keblinski used
equilibrium molecular dynamics simulations to calculate the viscoelastic properties of a
polymer melt.[188] The complex moduli were calculated using Green-Kubo expressions and
stress autocorrelation functions. The simulation conditions were identical to our simulations
with density of 85 . 0 = and temperature of 0 . 1 = T . Polymer chains in the melt were
connected together using the FENE potential described earlier with identical parameters.
Polymer chain lengths of N = 20 and higher were used in that work with all the atoms in the
system interacting via WCA potential. In an another study by Vladkov and Barrat,
113
nonequilibrium molecular dynamics (NEMD), Green-Kubo, and Rouse modes analysis were
used to calculate the complex moduli of the polymer melt.[189] The simulation conditions
used in their study were identical to those used by Sen et. al. with respect to density of the
polymer melt, system temperature and FENE potential parameters. The polymer chain
length of N=10 and N=20 were studied and all the atoms in the system interacted via LJ
potential.
The studies mentioned above provide the average bulk values for the complex moduli
and the values obtained from the simulation results using the technique described in this
chapter (active nanorheology) will be compared with these studies. The active nanorheology
technique holds an advantage over the techniques mentioned above as it can calculate the
local viscoelastic properties and the comparison will be presented subsequently.

Results: Viscoelastic properties from Active Nanorheology simulations

Figures 5.5 and 5.6 show the complex viscoelastic moduli obtained for WCA
interactions between all the atoms in the system in the simulation box size of dimensions
0 . 30 8 . 33 8 . 33 . This simulation conditions are as described earlier and the polymer chain
length is of N = 20. The nanoparticle is oscillating at the center of the channel and the
simulation results are compared with the literature data with identical simulation conditions.
The active nanorheology data presented here are averaged over 50 periods of oscillations for
the lowest frequency and about 200 periods of oscillations for the highest frequency of
oscillations.

114
0.001 0.01 0.1
Frequency
0.001
0.01
0.1
G
Active nanorheology
Vladkov and Barrat (2006)
Sen, Kumar and Keblinski (2005)

Figure 5.5: Comparison of the storage modulus (G) for the polymer melt (N=20) calculated using active
nanorheology simulations with the literature values for a simulation system of size 33.8x33.8x30. All the
interactions in the simulations are of type WCA.

0.001 0.01 0.1
Frequency
0.01
0.1
G
"
Active nanorheology

Figure 5.6: Comparison of the loss modulus (G) for the polymer melt (N=20) calculated using active
nanorheology simulations with the literature values for a simulation system of size 33.8x33.8x30. All the
interactions in the simulations are of type WCA.
115
Viscoelastic moduli at different locations

The active nanorheology simulations can be used to calculate the local values of
complex viscoelastic moduli. To this end we used a bigger simulation box of dimensions
0 . 45 8 . 33 8 . 33 and calculated the viscoelastic moduli at two different locations for
identical simulation conditions of density, temperature and polymer chain length (N = 20). If
d is the distance between the nanochannel wall and the nanoparticle center of mass as shown
in the following schematic then the viscoelastic moduli were calculated at d = 7.5 and d =
22.5 from the channel walls.

dd

Figure 5.7: A schematic showing the meaning of the parameter d, i.e. the distance of the nanoparticle
center of mass from the channel surface.

In these simulations the nanoparticle is constrained at a distance d from the wall and
the oscillating force is applied on the nanoparticle in x direction. Figures 5.8 and 5.9 show
the complex viscoelastic moduli calculated using active nanorheology simulations for two
different d locations. The data are compared with the bulk values from the literature.
116
0.01
Frequency
0.001
0.01
0.1
G
d = 22.5 (Active nanorheology)


nanorheology simulations at two different d locations with the bulk literature values for a simulation
system of size 33.8x33.8x30. All the interactions in the simulations are of type WCA.

0.01
Frequency
0.01
0.1
G
"

system of size 33.8x33.8x30. All the interactions in the simulations are of type WCA.
117
Slip at the nanoparticle surface

It can be observed that the values of complex moduli obtained using active
nanorheology simulations as shown in figures 5.5, 5.6, 5.8 and 5.9 are lower than the
literature values. We attribute this to the presence of slip at the nanoparticle surface. In the
previous chapter, it was shown that the no-slip boundary condition at the nanoparticle surface
was not satisfied when WCA interactions were used. The solvent was monomeric in those
studies while it is a polymer melt in these simulations.
To investigate the effect of slip at the nanoparticle surface translation simulations as
described in the previous chapter were performed with nanoparticle translating through the
polymer melt. A periodic grid was placed around the nanoparticle to track the velocities of
the beads of the polymers adjacent to the nanoparticle surface to quantify the velocity slip at
the nanoparticle surface. Figures 5.10 and 5.11 show the velocity field in the yz plane for a
nanoparticle translating in x direction (into the plane of the figure) for two different
translation velocities (V
NP
= 0.039 and V
NP
= 0.169). For no-slip boundary condition to
apply, the velocity of polymer beads in cubes adjacent to the nanoparticle surface should be
identical to the nanoparticle velocity. As can be seen from the figures the velocity of fluid is
much smaller than the nanoparticle velocity in both the cases resulting in about 35-40%
velocity slip at the nanoparticle surface. The large amount of velocity slip at the nanoparticle
surface leads to the lower values of viscoelastic moduli observed in figures 5.5, 5.6, 5.8 and
5.9.

118

Figure 5.10: Velocity field around a nanoparticle translating in a polymer melt with velocity of 0.039 for
the simulation conditions identical to those in figure 5.5.

the simulation conditions identical to those in figure 5.5.

119
The velocity slip at the nanoparticle surface can be reduced by using LJ interactions
between the beads of the nanoparticle and the medium, as was seen in the previous chapter.
To verify this for the system under consideration, a translation simulation as described earlier
was performed with nanoparticle translating in the polymer melt and the interactions between
nanoparticle and the medium defined using LJ potential. The LJ parameters used for the
nanoparticle-medium interactions were, 3 . 2 =
c
r and 0 . 2 =
Medium NP
. All other interactions
in the system were defined using WCA potential. Figure 5.12 shows the velocity field in yz
plane around the nanoparticle translating in x direction with velocity of V
NP
= 0.149. It was
observed that the slip is almost completely eliminated (~1%) as the fluid velocity at the cubes
adjacent to the nanoparticle surface is virtually the same as the nanoparticle velocity.

the simulation conditions identical to those in figure 5.5, but with LJ interactions between the
nanoparticle and the medium.

120
Active nanorheology simulations in the absence of slip

The use of LJ interactions between nanoparticle and solvent has been shown to
eliminate the slip at the nanoparticle surface. Hence, to study the effect of slip on the
complex moduli active nanorheology simulations were performed with nanoparticle-medium
interactions defined using LJ potential with same parameters, 3 . 2 =
c
r and 0 . 2 =
Medium NP
,
all other interactions (medium-medium, medium-wall, wall-wall, NP-NP and NP-wall) being
WCA. Figures 5.13 and 5.14 show the complex viscoelastic moduli obtained using these
active nanorheology simulations compared with the data obtained from literature mentioned
earlier. It can be seen that the values of complex viscoelastic moduli are higher than those
obtained using WCA potential between nanoparticle and medium. Also the agreement with
the literature values is much better than that was in earlier case. The simulations with LJ
interactions are computationally more expensive than those with WCA interactions and
hence the data shown here for LJ interactions is for less number of output data sets. For the
lowest frequency of oscillations the data are averaged over 30 periods of oscillations whereas
for the highest frequency of oscillation the data are averaged over 100 periods of oscillations.
These results demonstrate the effect of slip at the nanoparticle surface on the magnitude of
the complex viscoelastic moduli. However the agreement with literature data in these
partial-LJ simulations proves that the active nanorheology simulation technique is very
effective in calculating complex viscoelastic moduli. Moreover, the complex viscoelastic
moduli calculated at two different d locations (d = 7.5 and d = 22.5) in a simulation box of
size 0 . 45 8 . 33 8 . 33 showed increased value of moduli as shown in figure 5.15 and 5.16.
121
0.001 0.01 0.1
Frequency
0.001
0.01
0.1
G
Active nanorheology

nanorheology simulations with the literature values for a simulation system of size 33.8x33.8x30.
Nanoparticle-medium interactions are LJ while all other interactions in the simulations are of type WCA.

0.001 0.01 0.1
Frequency
0.01
0.1
G
"
Active nanorheology

nanorheology simulations with the literature values for a simulation system of size 33.8x33.8x30.
Nanoparticle-medium interactions are LJ while all other interactions in the simulations are of type WCA.
122
0.01
Frequency
0.01
0.1
G


system of size 33.8x33.8x30. Nanoparticle-medium interactions are of type LJ while all other interactions
in the simulations are of type WCA.

0.01
Frequency
0.01
0.1
G
"

system of size 33.8x33.8x30. Nanoparticle-medium interactions are of type LJ while all other interactions
in the simulations are of type WCA.
123
To verify the applicability of the active nanorheology method for another system we
used a polymer melt with polymer chain length of N = 10. The choice of this system was
again influenced by the availability of the data in literature for the identical system and ease
of computation. For this system active nanorheology simulations were performed in the
same simulation box size 0 . 30 8 . 33 8 . 33 and with identical simulation conditions:
85 . 0 = , 0 . 1 = T , partial LJ interactions with identical parameters ( 3 . 2 =
c
r and
0 . 2 =
Medium NP
) , and same FENE parameters for the polymer chain ( k = 30.0 and Q = 1.5).
The oscillation frequencies used were identical to the previous system ( = 0.004, 0.006,
0.01 and 0.02). The polymer melt was well equilibrated in the starting configurations to
ensure isotropic orientation of the polymer chains in the melt using method described earlier.
Figures 5.17 and 5.18 show the complex viscoelastic moduli obtained using active
nanorheology simulations for polymer melt with N = 10. The results are compared with the
data obtained from the literature. As can be seen from the figure the simulation results are in
very good agreement with the literature data. The data in these simulations were averaged
over about 40 periods of oscillation for the lowest frequency of oscillation while for the
highest frequency of oscillation the data was averaged over 80 periods of oscillations.
The good agreement of the results of active nanorheology simulations with the data obtained
from literature for different systems underlines the effectiveness of this method to calculate
the complex viscoelastic moduli locally.

124

0.01
Frequency
0.001
0.01
0.1
G
Active nanorheology

Figure 5.17: Comparison of the storage modulus (G) for a different polymer melt (N=10) calculated
using active nanorheology simulations with the bulk literature values for a simulation system of size
33.8x33.8x30. Nanoparticle-medium interaction are of type LJ while all other interactions in the
simulations are of type WCA.

0.01
Frequency
0.01
0.1
G
"
Active nanorheology

Figure 5.18: Comparison of the loss modulus (G) for a different polymer melt (N=10) calculated using
active nanorheology simulations with the bulk literature values for a simulation system of size
33.8x33.8x30. Nanoparticle-medium interaction are of type LJ while all other interactions in the
simulations are of type WCA.
125

In this chapter a new approach for calculation of local viscoelastic properties using
active nanorheology simulations was presented. These active nanorheology simulations use
molecular dynamics simulation technique to probe the material viscoelastic response at the
nanometer length scales. The calculation procedure and the formulation of expressions for
complex moduli are presented in detail. The complex viscoelastic moduli obtained using this
approach seem to agree well with the available literature values for an identical polymer melt
calculated using different simulation techniques. The simulations performed using WCA
interactions yielded lower values of complex moduli as compared to the literature values. It
was observed that the no-slip boundary condition was not satisfied in these simulations and
had an effect on the calculated values. To avoid the effect of slip at the nanoparticle surface
the nanoparticle-medium interactions were described using LJ interactions which removed
the slip effects and resulted in better agreement with the literature values for complex
viscoelastic moduli. The method was further validated by performing calculations on a
different polymer melt system (N = 10). Active nanorheology simulation technique provides
an advantage over conventional methods of calculating viscoelastic properties as it can
investigate the local behavior of the system. The results obtained using the proposed method
are encouraging, however, following are a few factors that need to be considered to refine the
method and to obtain a more accurate representation of the system using these active
nanorheology simulations.

126
Momentum diffusion time Vs. Period of oscillation

The comparison of time taken for momentum to diffuse and the time taken for the
oscillating nanoparticle to complete a period can provide an insight into the size of the
simulation box and the frequencies of oscillation to be used.
The time for momentum diffusion can be written as:

2
x
diff mom
L
=
(5.33)
where
x
L is the box length in direction of oscillation and is the kinematic viscosity
defined as
= .
For our simulation setup the values are:
x
L = simulation box length = 33.8
= dynamic viscosity 9.0
= density = 0.85
Substituting these values in equation 5.33:
9 . 107 =
diff mom

The time taken for the nanoparticle to complete an oscillation is given by:

2
=
osc
(5.34)
For the lowest value of frequency ( = 0.004) used in these simulations,
1571 =
highest osc

whereas for the highest value of frequency ( = 0.02) used,
314 =
lowest osc
.
127
Now for avoiding the effect of cooperative hydrodynamic interactions between the
oscillating nanoparticle and its image, the nanoparticle should complete oscillation before the
momentum diffuses over the length of the box, i.e.

osc diff mom
>
(5.35)
For the values of frequency of oscillation and simulation box size used in these simulations
the time for momentum diffusion is less than the period of oscillation, even for the highest
frequency used, suggesting the effect of cooperative interactions on the calculated values of
complex moduli. We note that for translation of nanoparticle, these cooperative
hydrodynamic interactions were shown to have ~15% contribution to the friction force. The
effect of cooperativity can be avoided by increasing the simulation box size and also by
increasing the frequency of oscillation, so that the time for momentum diffusion becomes
greater than the period of oscillation. The increase of simulation box size is governed by the
cost of simulation whereas the increase in frequency of oscillation is governed by the inertial
effects as explained in earlier in this chapter in dimensional analysis section.

Dependence on amplitude of oscillation

The active nanorheology simulation technique needs to be verified for the
dependence on the parameters involved in the mathematical formulation, such as on the
amplitude (A) of oscillatory force. This can be verified by performing active nanorheology
simulations for the values of amplitude of oscillatory force which are twice and half of the
amplitude of oscillation used in this chapter. Increase in amplitude of oscillatory force
should result in increase of amplitude of displacement of the nanoparticle. If the simulation
box size, in the direction of oscillation, is not big enough then these high displacement
128
amplitudes might result in increased cooperative effects and hence can introduce artifacts in
the calculation. Hence a big enough simulation box should be used for these amplitude
dependence studies. Similarly, dependence on the harmonic trap constant should also be
studied on the complex moduli. The dependence on the size of the simulation box size can
be studied by using larger simulation boxes and calculating the viscoelastic properties at the
center of the box for each. The calculated values should not show a variation with box sizes
used.

129
6. CHAPTER 6
VI. CONCLUSIONS AND FUTURE WORK

Molecular dynamics simulation technique was used in this thesis to study the
hydrodynamic interactions in complex fluids in nanoconfinement. The goal was to study the
role of hydrodynamic interactions that exist on these small scales, by explicitly defining the
solvent in the simulations. Three different systems were studied: sheared dilute polymer
solution, translating or rotating grid of nanoparticles in monomeric medium, and an
oscillating nanoparticle in polymer melt.

A study of the cross stream chain migration phenomenon in a sheared dilute polymer
solution (Chapter 2) indicated a preferential size based cross-stream migration of longer
polymer chains away from the channel walls. It was observed that the longer polymer chains
in a shear flow stretch more than the short chains and migrate away from the channel walls
with an increase in shear rate due to the hydrodynamic interactions with the channel walls.
This behavior can potentially be used as a scheme for separating polymer chains of different
lengths in a shear flow by using a Poiseuille flow. The practical implementation of the
technique would need a detailed study of the achievable flow strengths and pressure
conditions in these nanochannels and can be a part of a future research in this area. The
effect of polymer concentration on migration phenomenon was studied in detail and a
reversal of direction of migration behavior with increase in concentration was observed. The
parameters used in the interaction potential between the beads of the polymers (FENE
potential) in this study, for the shear flow conditions used, may allow unrealistic chain
130
crossing in these systems. A modified form of this study would be using spring-spring
repulsion potentials which will avoid the crossing of the polymer chains.

Cooperative hydrodynamic effects studied using the translating suspension of
nanoparticles (Chapter 3 and 4) showed the long range nature of these effects in simulations.
This study highlighted the importance of giving careful consideration to the size of
simulation box sizes used to avoid any artifacts due to cooperative hydrodynamic interactions
between the molecules under study. The study of the effect of nanoparticle surface topology
and the nanochannel confinement on the friction force and torque on the nanoparticles
showed that these factors should be taken into account before applying the continuum bulk
expressions, and the assumptions involved such as of the slip or no-slip boundary conditions
at nanoparticle surface, to calculate the transport properties in these small scale geometries.
The study also established a link between the continuum predictions and the molecular
simulations so that, the lesser expensive continuum methods can be used to predict the
transport properties taking into account the effect of slip at the nanoparticle surface and
cooperative hydrodynamic interactions presented using molecular simulations. The
methodology developed in this work can be used to study various other systems in detail,
such as concentrated suspension of nanoparticles to explore the cooperative effects in the
presence of many particles in the system.

An oscillating nanoparticle was used to calculate the local viscoelastic properties of a
polymer melt (Chapter 5) confined between nanochannel walls by measuring the
displacement of the nanoparticle in response to an imposed oscillating force. This technique,
131
termed as Active Nanorheology, can be used to probe the local viscoelastic behavior of
materials with structural inhomogeneities such as polymer nanocomposites or polymer thin
films. The results of the simulations performed using this method agree very well with the
available literature values, for an identical polymer melt, obtained using other equilibrium
and nonequilibrium simulation methods. Though encouraging, the proposed method needs to
be rigorously validated with regards to the dependence on the parameters used in the
mathematical formulation to calculate the complex viscoelastic moduli. The dependence of
results obtained on various other factors such as the size of the simulation system, size of the
nanoparticle, cooperative hydrodynamic effects etc. should also be investigated in the future
work. Moreover, as was observed from the radial distribution profile of the solvent around
the nanoparticle, nanoscale structural heterogeneities are introduced in the medium due to the
introduction of the nanoparticle in the medium. The presence of these structural
heterogeneities needs to be taken into account while probing the local behavior of the
material, and hence needs to be studied in detail.

The current molecular simulation set up can be applied to several hydrodynamic
problems for complex fluids. The molecular dynamics code in its current form is a serial
code, i.e. it can run only on a single processor and hence poses a limitation with respect to
the size of the system that can be simulated. The various future extensions mentioned earlier
in this chapter, to different projects studied, will require bigger simulation system sizes
which mean bigger memory requirements. Moreover, the single processor codes take more
computation time which increases with the number of atoms in the system. These problems
can be addressed to a large extent by making the current code capable of running on multiple
132
processors, i.e. by parallelizing the molecular dynamics code. The molecular dynamics
code parallelization involves distributing the work to multiple processors so that each
processor performs calculations on less number of particles and thus reducing the total
computation time. The two common methods of performing the molecular dynamics code
parallelization are: atom decomposition and space decomposition. In atom decomposition
method, the atoms in the simulation system are divided equally among the processors and
each processor performs the calculations involving the atoms assigned to it, while in the
space decomposition method the simulation system is divided into equal parts and each
processor performs calculation on the atoms in the simulation space assigned. A detailed
algorithm for parallelizing our molecular dynamics code using the space decomposition
technique is designed and presented in Appendix D.

133
VII. APPENDICES
APPENDIX A

Reduced Lennard-Jones units
(Reference: M. P. Allen and D. J. Tildesley, Computer simulation of liquids, Oxford University
Press Inc.: New York, Reprint 2005)

Density
3
=

Temperature
T k
T
B
=

Energy
E
E =

Pressure
3
P
P =

Time
2
m
t t =

Force
f
f =

Torque

where, is number density, T is temperature, E is energy, P is pressure, t is time, m is mass
and f is force and is the torque. and are the Lennard-Jones parameters for energy and
distance respectively. In usual practice the dimensionless quantities are written normally
without the asterisk (*) i.e. T* is expressed as T, as it is done in this thesis.

134
APPENDIX B

c This program constructs a spherical nanoparticle of
c specified radius and an elipsoidal nanoparticle of
c specified semi-axis lengths.
c The nanoparticles are made up of of atoms arranged in a
c Face Centered Cubic (FCC) lattice.
c The code also writes out the atoms forming bonds
c and bond length of each bond pair in the nanoparticles.
c The input to the code is the box size, FCC lattice spacing,
c and the radius of the nanoparticle and the semi-axis lengths

program nanoparticles

implicit double precision (a-h,o-z)

parameter (ndim = 30000)

parameter (tiny = 0.0001)
parameter (bx = 14.2, by = 14.2, bz = 14.2)
parameter (fccspc = 1.42)
parameter (rsphere = 2.5)
parameter (a = 6.0, b = 4.0, c = 3.0)

dimension qix(ndim), qiy(ndim), qiz(ndim)
dimension xsph(ndim), ysph(ndim), zsph(ndim),dist(ndim)
dimension nb1(ndim),nb2(ndim),bondlength(ndim)
dimension nbep1(ndim),nbep2(ndim),bondlengthep(ndim)
dimension xelip(ndim),yelip(ndim),zelip(ndim)

open(1,file='InputDatafile',status='replace')
open(2,file='FCCcoordinates',status='replace')
open(3,file='sphere.crd',status='replace')
open(4,file='ellipsoid.crd',status='replace')
open(5,file='spherebonds',status='replace')
open(6,file='ellipsoidbonds',status='replace')

c Number of FCC lattices that will fit
numbx = int((bx/fccspc)+tiny)
numby = int((by/fccspc)+tiny)
numbz = int((bz/fccspc)+tiny)

bx2 = bx/2.0
by2 = by/2.0
bz2 = bz/2.0
fcc2 = fccspc/2.0
n = 0

c **********************************************
c FCC LATTICE COORDINATES GENERATION
C **********************************************

c Atoms on lattice sides:
135

do 10 k=1,numbz
do 11 i=1,numby
do 21 j=1,numbx
n = n+1
qix(n) = (j-1)*fccspc-bx2
qiy(n) = (i-1)*fccspc-by2
qiz(n) = (k-1)*fccspc-bz2
21 continue
11 continue
10 continue

c Face centered atoms in XY plane:

do 20 k=1,numbz
do 31 i=1,numby
do 41 j=1,numbx
n=n+1
qix(n) = (j-1)*fccspc+fcc2-bx2
qiy(n) = (i-1)*fccspc+fcc2-by2
qiz(n) = (k-1)*fccspc-bz2
41 continue
31 continue
20 continue

c Face centered atoms in YZ plane:

do 180 k=1,numbx
do 190 i=1,numby
do 200 j=1,numbz
n=n+1
qix(n) = (k-1)*fccspc-bx2
qiy(n) = (i-1)*fccspc+fcc2-by2
qiz(n) = (j-1)*fccspc+fcc2-bz2
200 continue
190 continue
180 continue

c Face centered atoms in XZ plane:

do 210 k=1,numby
do 220 i=1,numbx
do 230 j=1,numbz
n=n+1
qix(n) = (i-1)*fccspc+fcc2-bx2
qiy(n) = (k-1)*fccspc-by2
qiz(n) = (j-1)*fccspc+fcc2-bz2
230 continue
220 continue
210 continue

136

do 160 nt = 1,n
write(2,170)qix(nt),qiy(nt),qiz(nt)
170 format(3(2x,f10.5))
160 continue

c ****************************************************
c SPHERE GENERATION
C ****************************************************

xref = 0.0
yref = 0.0
zref = 0.0
nsph = 0

do 50 i=1,n
xs = qix(i) - xref
ys = qiy(i) - yref
zs = qiz(i) - zref

r = ((xs)**(2.0)+(ys)**(2.0)+(zs)**(2.0))**(0.5)

if(r.le.rsphere)then
nsph = nsph + 1
xsph(nsph) = qix(i)
ysph(nsph) = qiy(i)
zsph(nsph) = qiz(i)
dist(nsph) = r
end if

50 continue

do 60 i=1,nsph
write(3,70)xsph(i),ysph(i),zsph(i)
70 format(10x,3f10.5)
60 continue

c **********************************************************
c BONDS FOR SPHERE
C **********************************************************

nbond = 0

do 110 i=1,nsph-1

do 120 j=i+1,nsph

bdistx = xsph(i) - xsph(j)
bdisty = ysph(i) - ysph(j)
bdistz = zsph(i) - zsph(j)

bdtotal=((bdistx)**(2.0)+(bdisty)**(2.0)+(bdistz)**(2.0))**0.5
137

if(bdtotal.le.(1.5))then
nbond = nbond + 1
nb1(nbond) = i
nb2(nbond) = j
bondlength(nbond) = bdtotal
end if

120 continue

110 continue

write(5,150)nbond
150 format(i6)

do 140 k = 1,nbond
write(5,130)nb1(k),nb2(k),bondlength(k)
130 format(i8,2x,i8,2x,f10.5)
140 continue

c ***************************************************************
c ELLIPSOID GENERATION
c ***************************************************************

nelip = 0

a2 = a**(2.0)
b2 = b**(2.0)
c2 = c**(2.0)

do 240 ii = 1,n

xp = qix(ii) - xref
yp = qiy(ii) - yref
zp = qiz(ii) - zref

distep=((xp)**2.0)/a2+((yp)**2.0)/b2+((zp)**2.0)/c2

if(distep.lt.(1.0))then
nelip = nelip + 1
xelip(nelip) = qix(ii)
yelip(nelip) = qiy(ii)
zelip(nelip) = qiz(ii)
end if

240 continue

do 250 jj=1,nelip
write(4,260)xelip(jj),yelip(jj),zelip(jj)
260 format(10x,3f10.5)
250 continue

c ****************************************************************
c BONDS FOR ELLIPSOID
c ****************************************************************
138

nbondep = 0

do 270 kk=1,nelip-1
do 280 mm = kk+1,nelip

edistx = xelip(kk)-xelip(mm)
edisty = yelip(kk)-yelip(mm)
edistz = zelip(kk)-zelip(mm)

edist=((edistx)**(2.0)+(edisty)**(2.0)+(edistz)**(2.0))**(0.5)

if(edist.le.(1.5))then
nbondep = nbondep + 1
nbep1(nbondep) = kk
nbep2(nbondep) = mm
bondlengthep(nbondep) = edist
end if

280 continue
270 continue

write(6,290)nbondep
290 format(i6)

do 300 nn = 1,nbondep
write(6,310)nbep1(nn),nbep2(nn),bondlengthep(nn)
310 format(i8,2x,i8,2x,f10.5)
300 continue

c ****************************************************************

write(1,*)'FCC lattice spacing = ',fccspc
write(1,*)'Box dimensions:'
write(1,*)'bx = ',bx
write(1,*)'by = ',by
write(1,*)'bz = ',bz
write(1,*)'Sphere radius = ',rsphere
write(1,*)'Semi-axis lengths for ellipsoid:'
write(1,*)'a = ',a
write(1,*)'b = ',b
write(1,*)'c = ',c

c ****************************************************************

stop
end

139
APPENDIX C

c This code calculates the velocity field around
c the nanoparticle

subroutine vxfield(ndatvx,idstep)

implicit double precision (a-h,o-z)

parameter (ndim=84000)
parameter (npart = 100)

parameter (nbinx = 8, nbiny = 8, nbinz = 8)
c Total number of bins = (2*nbinx)+1
parameter (binsizex = 1.0, binsizey = 1.0, binsizez = 1.0)
parameter (ntotbinx = (2*nbinx)+1)
parameter (ntotbiny = (2*nbiny)+1)
parameter (ntotbinz = (2*nbinz)+1)

common/box/bx,by,bz
common/syssiz/na,nb,nw
common/coord/qix(ndim),qiy(ndim),qiz(ndim)
common/velocs/vx(ndim),vy(ndim),vz(ndim)
common/trajct/ieqtim,iwofrq
common/vtraj/icvfrq,ibvfrq
common/vxtraj/avgvx(ntotbinx,ntotbiny,ntotbinz)
common/vxtraj2/avg2vx(ntotbinx,ntotbiny,ntotbinz)
common/bincount/ndat2vx(ntotbinx,ntotbiny,ntotbinz)
common/COMforNP/xcom(npart),ycom(npart),zcom(npart)
common/numbden/avgden(ntotbinx,ntotbiny,ntotbinz)

dimension vxsum(ntotbinx,ntotbiny,ntotbinz)
dimension natmvx(ntotbinx,ntotbiny,ntotbinz)

ndatvx = ndatvx + 1

c For the main simulation box

bx2 = bx/2.0
bx2m= -bx2
by2 = by/2.0
by2m= -by2
bz2 = bz/2.0
bz2m= -bz2

c Length of region over which velocity will be tracked

lengthx = dble(ntotbinx) * binsizex
140
lengthy = dble(ntotbiny) * binsizey
lengthz = dble(ntotbinz) * binsizez

x2 = lengthx / 2.0
y2 = lengthy / 2.0
z2 = lengthz / 2.0

delrx = binsizex
delrix = 1.0/delrx
delry = binsizey
delriy = 1.0/delry
delrz = binsizez
delriz = 1.0/delrz

volcube = binsizex * binsizey * binsizez
avgrhocube = dble(na+nb)/dble(bx*by*(bz-1))

c For writing out data for the cubes on axes

grdx = x2
grdy = y2
grdz = z2

if (mod(idstep,ibvfrq).eq.icvfrq) then
ndatvx = 1
do 111 k=1,ntotbinx
do 11 i=1,ntotbiny
do 12 j=1,ntotbinz
avgvx(k,i,j) = 0.0
avg2vx(k,i,j)= 0.0
ndat2vx(k,i,j)= 1
avgden(k,i,j)= 0
12 continue
11 continue
111 continue
endif

do 133 k=1,ntotbinx
do 13 i=1,ntotbiny
do 14 j=1,ntotbinz
vxsum(k,i,j) = 0.0
natmvx(k,i,j) = 0
14 continue
13 continue
133 continue

c Only calculate the solvent velocity (na-solvent, nb-solute/particle)

do 21 i= 1,na

c Initialize flags

141
ixflag = 0
iyflag = 0
izflag = 0

c Distance are calculated from COM of the rotating nanoparticle

RXIJ=(qix(i) - xcom(1))-bx*ANINT((qix(i) - xcom(1))/bx)
if(abs(RXIJ).le.x2)then
ixflag = 1
end if

RYIJ=(qiy(i) - ycom(1))-by*ANINT((qiy(i) - ycom(1))/by)
if(abs(RYIJ).le.y2)then
iyflag = 1
end if

RZIJ=(qiz(i) - zcom(1))-bz*ANINT((qiz(i) - zcom(1))/bz)
if(abs(RZIJ).le.z2)then
izflag = 1
end if

if((ixflag.eq.1).and.(iyflag.eq.1).and.(izflag.eq.1))then
ibinx = 1 + int((RXIJ+x2) * delrix)
ibiny = 1 + int((RYIJ+y2) * delriy)
ibinz = 1 + int((RZIJ+z2) * delriz)

vxsum(ibinx,ibiny,ibinz) = vxsum(ibinx,ibiny,ibinz)+vx(i)
natmvx(ibinx,ibiny,ibinz) = natmvx(ibinx,ibiny,ibinz)+1
end if

21 continue

c Calculate the instantaneous value

do 411 k=1,ntotbinx
do 41 i=1,ntotbiny
do 42 j=1,ntotbinz
if (natmvx(k,i,j).ne.0) then
vxinst = vxsum(k,i,j)/natmvx(k,i,j)
ndat2vx(k,i,j) = ndat2vx(k,i,j) + 1
temp = (vxinst-avg2vx(k,i,j))/ndat2vx(k,i,j)
avg2vx(k,i,j) = avg2vx(k,i,j)+ temp
else
vxinst = 0.0
endif
temp2 = (vxinst-avgvx(k,i,j))/ndatvx
avgvx(k,i,j) = avgvx(k,i,j)+ temp2
c Instantaneous normalized number density
deninst = dble(natmvx(k,i,j))/(volcube*avgrhocube)
c Running average number density in the cube
avgden(k,i,j)=avgden(k,i,j)+(deninst-avgden(k,i,j))/ndatvx

42 continue
41 continue
142
411 continue

if (mod(idstep,ibvfrq).eq.0) then

c First open the file in the append mode

open(62,file='VX_BrutForce',status='old',access='append')
open(65,file='VX_Conditional',status='old',access='append')
open(67,file='VX_axialBF',status='old',access='append')
open(68,file='VX_axialCOND',status='old',access='append')

write (62,60) idstep
60 format ('Begin Vx Profile at step # =',1x,i9)

64 format(4(f8.3))
74 format(4(f8.3),1x,f8.3)

do 611 k=1,ntotbinx
do 61 i = 1,ntotbiny
do 63 j = 1,ntotbinz
xcrd = (k-1)*delrx + delrx/2.0
ycrd = (i-1)*delry + delry/2.0
zcrd = (j-1)*delrz + delrz/2.0
write (62,74)xcrd,ycrd,zcrd,avgvx(k,i,j),avgden(k,i,j)
write (65,74)xcrd,ycrd,zcrd,avg2vx(k,i,j),avgden(k,i,j)
63 continue
61 continue
611 continue

c Following loops are to write out data only
c for the cubes along the axes

do 511 k=1,ntotbinx
do 51 i = 1,ntotbiny
do 53 j = 1,ntotbinz
xcrd = (k-1)*delrx + delrx/2.0
ycrd = (i-1)*delry + delry/2.0
zcrd = (j-1)*delrz + delrz/2.0

if((xcrd.eq.grdx).and.(ycrd.eq.grdy))then
elseif((xcrd.eq.grdx).and.(zcrd.eq.grdz))then
elseif((ycrd.eq.grdy).and.(zcrd.eq.grdz))then
else
avgvx(k,i,j) = 0.0
avg2vx(k,i,j) = 0.0
143
end if

53 continue
51 continue
511 continue

78 format ('End Vx Profile at step # =',1x,i9)

c Close the velocity profile file
close (62)
close (65)
close (67)
close (68)

endif

return
end

c
**********************************************************************

144
APPENDIX D

An algorithm to parallelize the molecular dynamics simulation code using space
decomposition method is as follows. In this method the one of the available processors is
acts as a master processor which controls the synchronization of the computations and
controls the flow of information to and from other processors called as the slave
processors.

**************************************************************************

1. Input is fed to the MASTER (Input parameters file, coordinates, bonds information for
nanoparticle, number of processors etc.)

2. Initial set up by master
a) Distribution of atoms to processors
b) COM* calculation for nanoparticle
c) Atom assignments (giving identity to atoms, solute/solvent/wall etc; bond lists)
c) Linked cells construction
d) Opening/creating all the output files to be written
e) Velocity generation for each atom
f) Quantities for brownian dynamics
g) Initial neighbor list construction

3. Required initial information is distributed to respective processors
a) Coordinates assigned
b) Neighbor lists for each atom
c) Velocities of each atom
d) Linked cells (Linked cell map, list of atoms in each cell)
e) Brownian dynamics parameters

4. Main dynamics loop starts

-----------------------------------------------------------------------------------------

5a. At t=0 initial force calculations by each processor for its atoms

5b. Each processor performs move1 updates coordinates and velocities on its atoms

6. Communication 1:
New coordinates are sent to Master

Master does calculations
a) New COM of the nanoparticle is calculated by master
b) Spring force on nanoparticle calculated by master
Master sends back the force information to respective processors

145
7. Processors perform calculations
a) Buffer on each processor updated using updated coordinates
b) Neighbor lists are formed using shared buffer information (done every 5 steps)
c) LJ force calculations
d) Spring force calculations (Each nanoparticle bead needs to know
its bonded atoms a bond list for each bead)
e) Forces on wall atoms (only by the processors who have wall atoms)
f) Each processor performs move2 (velocities, energy, temperature) for
its atoms

8. Communication 2:

Updated velocities and forces sent to master.

Master does calculations and writes out output files for
a) Energies
b) Different profiles, RDF**
c) Torque on nanoparticle
d) Velocity field around nanoparticle
e) Friction force calculations
f) Chain/solvent properties
g) Rg of nanoparticle
h) Wall friction force
i) Stress
j) Restart data

9. Repeat steps 5 to 8 for each simulation step

(Step 5 doesnt need to wait while master is doing step 8 work, although synchronization
needs to be ensured)

--------------------------------------------------------------------------------------

10. Main dynamics loop ends

11. Final output (Coordinates, forces, velocities) written out by master

**************************************************************************

*COM Center of mass

**RDF Radial distribution function

146
In the current molecular dynamics code the primary subroutine where most of the
calculations are performed is the dynamics subroutine called dynmbd.f. Following are
some of the important suggestions with regards to the distribution of various calculations
performed in this subroutine and the kind of processor involved with those.

***************************************************************************
Linked cell maps should be generated on each processor for each domain.
Initial velocities generation should be done on master processor as this subroutine is
called only once.
Neighbor list should be built on each processor for each of its atoms. This will need
the buffer information being passed to the neighboring processors, as the boundary
atoms on each processor will need the atoms in their neighboring cells on other
processors to build the neighbor list.
Force evaluation (LJ force, spring force, pulling force should be calculated by master)
should be done on each processor. This can also be done by the master for the first
time, before going into the main loop, and then inside the loop by each processor.
Force information can be distributed to each processor for its atoms to start with.
Each processor should perform move1 and update the coordinates and velocities of
its atoms.
After the positions are updated the buffer needs to be formed of the atoms that need to
be exchanged.
Processors need to be identified who will need mutual exchange of information.
Exchange of buffer information should be carried out now.
Force calculation is done by each processor for each of its atoms using the primary
and the buffer information.
Spring forces for bonds should be calculated by each processor for its bonding atoms.
Each atom should have its own bond list.
Forces on wall atoms should be calculated by processors containing wall atoms.
Pulling force is based on the COM location, and can be calculated by master and the
force contribution to each nanoparticle bead can be sent to respective processors.
Each processor should perform move2 and send the output to master.
Master should add up the total potential and total energy values to be printed out at
each step.
All the profiles generation should be done by master. Information should be provided
to the master by the slaves.
Torque calculation should be done by master.
Velocity field around nanoparticle should be calculated by master.
Friction forces should be calculated by master.
Chain/Solvent properties should be calculated my master
Radius of gyration of nanoparticle should be calculated by master (master has all the
updated coordinate information)
Density profiles should be calculated by master.
RDF should be calculated by master
Wall friction force should be calculated by master
147
Stress should be calculated by master
Temperature profiles should be calculated by master
Master should print out the total energy, potential energy, kinetic energy and
temperature at each step.
Intermediate restart files should be written down by master.
Output files (those written within the various subroutines intermediately) should be
written by master including the final output.

**************************************************************************

148
VIII. BIBLIOGRAPHY

1. Abgrall, P. and Nguyen, N.T., "Nanofluidic devices and their applications",
Analytical Chemistry, 2008. 80(7): p. 2326-2341.
2. Volkmuth, W.D. and Austin, R.H., "DNA electrophoresis in microlithographic
arrays", Nature, 1992. 358(6387): p. 600-602.
3. Gajar, S.A. and Geis, M.W., "An ionic liquid-channel field-effect transistor", Journal
of the Electrochemical Society, 1992. 139(10): p. 2833-2840.
4. Kramer, S., Fuierer, R.R., and Gorman, C.B., "Scanning probe lithography using self-
assembled monolayers", Chemical Reviews, 2003. 103(11): p. 4367-4418.
5. Eigler, D.M. and Schweizer, E.K., "Positioning single atoms with a scanning
tunneling microscope", Nature, 1990. 344(6266): p. 524-526.
6. Tas, N.R., Berenschot, J.W., Mela, P., Jansen, H.V., Elwenspoek, M., and van den
Berg, A., "2D-confined nanochannels fabricated by conventional micromachining",
Nano Letters, 2002. 2(9): p. 1031-1032.
7. Brenner, M.P., "Screening mechanisms in sedimentation", Physics of Fluids, 1999.
11(4): p. 754-772.
8. Squires, T.M. and Brenner, M.P., "Like-charge attraction and hydrodynamic
interaction", Physical Review Letters, 2000. 85(23): p. 4976-4979.
9. Lin, B.H., Yu, J., and Rice, S.A., "Direct measurements of constrained Brownian
motion of an isolated sphere between two walls", Physical Review E, 2000. 62(3): p.
3909-3919.
149
10. Dufresne, E.R., Squires, T.M., Brenner, M.P., and Grier, D.G., "Hydrodynamic
coupling of two Brownian spheres to a planar surface", Physical Review Letters,
2000. 85(15): p. 3317-3320.
11. Bhattacharya, S., Blawzdziewicz, J., and Wajnryb, E., "Hydrodynamic interactions of
spherical particles in suspensions confined between two planar walls", Journal of
Fluid Mechanics, 2005. 541: p. 263-292.
12. Oetama, R.J. and Walz, J.Y., "Simultaneous investigation of sedimentation and
diffusion of a single colloidal particle near an interface", Journal of Chemical
Physics, 2006. 124(16): p. 164713.
13. Swan, J.W. and Brady, J.F., "Simulation of hydrodynamically interacting particles
near a no-slip boundary", Physics of Fluids, 2007. 19(11): p. 113306.
14. Eijkel, J.C.T. and van den Berg, A., "Nanofluidics: what is it and what can we expect
from it?" Microfluidics and Nanofluidics, 2005. 1(3): p. 249-267.
15. Derjaguin, B.V. and Landau, L., "Theory of the stability of strongly charged
lyophobic sols and the adhesion of strongly charged particles in solutions of
electrolytes", Acta Physico-chimica (URSS), 1941. 14: p. 633-662.
16. Verwey, E.J.W. and Overbeek, J.T.G., Theory of the Stability of Lyophobic Colloids.
1948, Amsterdam, The Netherlands: Elsevier.
17. Derjaguin, B.V., Rabinovich, Y.I., and Churaev, N.V., "Direct measurement of
molecular forces", Nature, 1978. 272: p. 313-318.
18. Israelachvili, J., Intermolecular and Surface Forces. 1991, Academic Press: New
York.
150
19. Heinz, W.F. and Hoh, J.H., "Spatially resolved force spectroscopy of biological
surfaces using the atomic force microscope", Trends in Biotechnology, 1999. 17(4): p.
143-150.
20. Mannion, J.T. and Craighead, H.G., "Nanofluidic structures for single biomolecule
fluorescent detection", Biopolymers, 2007. 85(2): p. 131-143.
21. Foquet, M., Korlach, J., Zipfel, W.R., Webb, W.W., and Craighead, H.G., "Focal
volume confinement by submicrometer-sized fluidic channels", Analytical Chemistry,
2004. 76(6): p. 1618-1626.
22. Foquet, M., Korlach, J., Zipfel, W., Webb, W.W., and Craighead, H.G., "DNA
fragment sizing by single molecule detection in submicrometer-sized closed fluidic
channels", Analytical Chemistry, 2002. 74(6): p. 1415-1422.
23. Riehn, R. and Austin, R.H., "Wetting micro- and nanofluidic devices using
supercritical water", Analytical Chemistry, 2006. 78(16): p. 5933-5934.
24. Wang, Y.M., Tegenfeldt, J.O., Reisner, W., Riehn, R., Guan, X.J., Guo, L., Golding,
I., Cox, E.C., Sturm, J., and Austin, R.H., "Single-molecule studies of repressor-DNA
interactions show long-range interactions", Proceedings of the National Academy of
Sciences of the United States of America, 2005. 102(28): p. 9796-9801.
25. Mijatovic, D., Eijkel, J.C.T., and van den Berg, A., "Technologies for nanofluidic
systems: top-down vs. bottom-up - a review", Lab on a Chip, 2005. 5(5): p. 492-500.
26. Perry, J.L. and Kandlikar, S.G., "Review of fabrication of nanochannels for single
phase liquid flow", Microfluidics and Nanofluidics, 2006. 2(3): p. 185-193.
27. King, T.L., Jin, X., Aluru, N., and Bohn, P.W., Molecular transport and fluidic
manipulation in three dimensional integrated nanofluidic networks, in Nanofluidics:
151
Nanoscience and Nanotechnology, J.B. Edel and A.J. deMello, Editors. 2009, RSC
Publishing: Cambridge.
28. Benthansen, L., Feldtrasmussen, B., Kverneland, A., and Deckert, T., "Plasma
disappearance of glycated and non-glycated albumin in type-1 (insulin-dependent)
diabetes-mellitus - evidence for charge dependent alterations of the plasma to lymph
pathway", Diabetologia, 1993. 36(4): p. 361-363.
29. Huang, L.R., Tegenfeldt, J.O., Kraeft, J.J., Sturm, J.C., Austin, R.H., and Cox, E.C.,
"A DNA prism for high-speed continuous fractionation of large DNA molecules",
Nature Biotechnology, 2002. 20(10): p. 1048-1051.
30. Giddings, J.C., Unified Separation Science. 1991, New York: Wiley.
31. Blom, M.T., Chmela, E., Oosterbroek, R.E., Tijssen, R., and van den Berg, A., "On-
chip hydrodynamic chromatography separation and detection of nanoparticles and
biomolecules", Analytical Chemistry, 2003. 75(24): p. 6761-6768.
32. Tegenfeldt, J.O., Prinz, C., Cao, H., Huang, R.L., Austin, R.H., Chou, S.Y., Cox,
E.C., and Sturm, J.C., "Micro- and nanofluidics for DNA analysis", Analytical and
Bioanalytical Chemistry, 2004. 378(7): p. 1678-1692.
33. Kasianowicz, J.J., Brandin, E., Branton, D., and Deamer, D.W., "Characterization of
individual polynucleotide molecules using a membrane channel", Proceedings of the
National Academy of Sciences of the United States of America, 1996. 93(24): p.
13770-13773.
34. Wanunu, M., Chakrabarti, B., Mathe, J., Nelson, D.R., and Meller, A., "Orientation-
dependent interactions of DNA with an alpha-hemolysin channel", Physical Review
E, 2008. 77(3): p. 5.
152
35. Meller, A., Nivon, L., and Branton, D., "Voltage-driven DNA translocations through
a nanopore", Physical Review Letters, 2001. 86(15): p. 3435-3438.
36. Mathe, J., Aksimentiev, A., Nelson, D.R., Schulten, K., and Meller, A., "Orientation
discrimination of single-stranded DNA inside the alpha-hemolysin membrane
channel", Proceedings of the National Academy of Sciences of the United States of
America, 2005. 102(35): p. 12377-12382.
37. Deamer, D.W. and Akeson, M., "Nanopores and nucleic acids: Prospects for
ultrarapid sequencing", Trends in Biotechnology, 2000. 18(4): p. 147-151.
38. Dekker, C., "Solid-state nanopores", Nature Nanotechnology, 2007. 2(4): p. 209-215.
39. Austin, R., "Nanopores - The art of sucking spaghetti", Nature Materials, 2003. 2(9):
p. 567-568.
40. Akeson, M., Branton, D., Kasianowicz, J.J., Brandin, E., and Deamer, D.W.,
"Microsecond time-scale discrimination among polycytidylic acid, polyadenylic acid,
and polyuridylic acid as homopolymers or as segments within single RNA
molecules", Biophysical Journal, 1999. 77(6): p. 3227-3233.
41. Prodhom, B., Pietrobon, D., and Hess, P., "Direct measurement of proton-transfer
rates to a group controlling the dihydropyridine-sensitive Ca-2+ channel", Nature,
1987. 329(6136): p. 243-246.
42. Barann, M., Wenningmann, I., and Dilger, J.P. Interactions of general anesthetics
within the pore of an ion channel. in 5th International Conference on Molecular and
Cellular Mechanisms of Anaesthesia. 1997. Calgary, Canada: Elsevier Sci Ireland
Ltd.
153
43. Bezrukov, S.M. and Kasianowicz, J.J., "Current noise reveals protonation kinetics
and number of ionizable sites in an open protein ion channel", Physical Review
Letters, 1993. 70(15): p. 2352-2355.
44. Walker, B., Kasianowicz, J., Krishnasastry, M., and Bayley, H., "A pore-forming
protein with a metal-actuated switch", Protein Engineering, 1994. 7(5): p. 655-662.
45. Braha, O., Walker, B., Cheley, S., Kasianowicz, J.J., Song, L.Z., Gouaux, J.E., and
Bayley, H., "Designed protein pores as components for biosensors", Chemistry &
Biology, 1997. 4(7): p. 497-505.
46. Kasianowicz, J.J., Burden, D.L., Han, L.C., Cheley, S., and Bayley, H., "Genetically
engineered metal ion binding sites on the outside of a channel's transmembrane beta-
barrel", Biophysical Journal, 1999. 76(2): p. 837-845.
47. Kasianowicz, J.J., Henrickson, S.E., Weetall, H.H., and Robertson, B., "Simultaneous
multianalyte detection with a nanometer-scale pore", Analytical Chemistry, 2001.
73(10): p. 2268-2272.
48. Marziali, A. and Akeson, M., "New DNA sequencing methods", Annual Review of
Biomedical Engineering, 2001. 3: p. 195-223.
49. Han, J., Turner, S.W., and Craighead, H.G., "Entropic trapping and escape of long
DNA molecules at submicron size constriction", Physical Review Letters, 1999.
83(8): p. 1688-1691.
50. Reisner, W., Morton, K.J., Riehn, R., Wang, Y.M., Yu, Z.N., Rosen, M., Sturm, J.C.,
Chou, S.Y., Frey, E., and Austin, R.H., "Statics and dynamics of single DNA
molecules confined in nanochannels", Physical Review Letters, 2005. 94(19): p.
196101.
154
51. Stein, D., van der Heyden, F.H.J., Koopmans, W.J.A., and Dekker, C., "Pressure-
driven transport of confined DNA polymers in fluidic channels", Proceedings of the
National Academy of Sciences of the United States of America, 2006. 103(43): p.
15853-15858.
52. Jo, K., Dhingra, D.M., Odijk, T., de Pablo, J.J., Graham, M.D., Runnheim, R.,
Forrest, D., and Schwartz, D.C., "A single-molecule barcoding system using nanoslits
for DNA analysis", Proceedings of the National Academy of Sciences of the United
States of America, 2007. 104(8): p. 2673-2678.
53. Turner, S.W., Perez, A.M., Lopez, A., and Craighead, H.G. Monolithic nanofluid
sieving structures for DNA manipulation. in 42nd International Conference on
Electron, Ion, and Photon Beam Technology and Nanofabrication (EIPBN). 1998.
Chicago, Illinois: Amer Inst Physics.
54. Fu, J.P., Mao, P., and Han, J.Y., "Nanofilter array chip for fast gel-free biomolecule
separation", Applied Physics Letters, 2005. 87(26): p. 263902.
55. Ogston, A.G., "The spaces in a uniform random suspension of fibres", Transactions
of the Faraday Society, 1958. 54: p. 1754-1757.
56. Fu, J.P., Schoch, R.B., Stevens, A.L., Tannenbaum, S.R., and Han, J.Y., "A patterned
anisotropic nanofluidic sieving structure for continuous-flow separation of DNA and
proteins", Nature Nanotechnology, 2007. 2(2): p. 121-128.
57. Wang, Y.C., Stevens, A.L., and Han, J.Y., "Million-fold preconcentration of proteins
and peptides by nanofluidic filter", Analytical Chemistry, 2005. 77(14): p. 4293-4299.
155
58. Kim, S.M., Burns, M.A., and Hasselbrink, E.F., "Electrokinetic protein
preconcentration using a simple glass/poly(dimethylsiloxane) microfluidic chip",
Analytical Chemistry, 2006. 78(14): p. 4779-4785.
59. Lee, J.H., Chung, S., Kim, S.J., and Han, J.Y., "Poly(dimethylsiloxane)-based protein
preconcentration using a nanogap generated by junction gap breakdown", Analytical
Chemistry, 2007. 79(17): p. 6868-6873.
60. Kim, S.J. and Han, J.Y., "Self-sealed vertical polymeric nanoporous-junctions for
high-throughput nanofluidic applications", Analytical Chemistry, 2008. 80(9): p.
3507-3511.
61. Lee, J.H., Song, Y.A., Tannenbaum, S.R., and Han, J., "Increase of reaction rate and
sensitivity of low-abundance enzyme assay using micro/nanofluidic preconcentration
chip", Analytical Chemistry, 2008. 80(9): p. 3198-3204.
62. Cao, H., Tegenfeldt, J.O., Austin, R.H., and Chou, S.Y., "Gradient nanostructures for
interfacing microfluidics and nanofluidics", Applied Physics Letters, 2002. 81(16): p.
3058-3060.
63. Riehn, R., Lu, M.C., Wang, Y.M., Lim, S.F., Cox, E.C., and Austin, R.H.,
"Restriction mapping in nanofluidic devices", Proceedings of the National Academy
of Sciences of the United States of America, 2005. 102(29): p. 10012-10016.
64. Reisner, W., Beech, J.P., Larsen, N.B., Flyvbjerg, H., Kristensen, A., and Tegenfeldt,
J.O., "Nanoconfinement-enhanced conformational response of single DNA molecules
to changes in ionic environment", Physical Review Letters, 2007. 99(5): p. 058302.
65. deGennes, P.G., Scaling Concepts in Polymer Physics. 1979, Ithaca, NY: Cornell
University Press.
156
66. Waigh, T.A., "Microrheology of complex fluids", Reports on Progress in Physics,
2005. 68(3): p. 685-742.
67. Zaner, K.S. and Valberg, P.A., "Viscoelasticity of f-actin measured with magnetic
microparticles", Journal of Cell Biology, 1989. 109(5): p. 2233-2243.
68. Ziemann, F., Radler, J., and Sackmann, E., "Local measurements of viscoelastic
moduli of entangled actin networks using an oscillating magnetic bead micro-
rheometer", Biophysical Journal, 1994. 66(6): p. 2210-2216.
69. Mason, T.G. and Weitz, D.A., "Optical measurements of frequency-dependent linear
viscoelastic moduli of complex fluids", Physical Review Letters, 1995. 74(7): p.
1250-1253.
70. Mason, T.G., Ganesan, K., vanZanten, J.H., Wirtz, D., and Kuo, S.C., "Particle
tracking microrheology of complex fluids", Physical Review Letters, 1997. 79(17): p.
3282-3285.
71. Gittes, F., Schnurr, B., Olmsted, P.D., MacKintosh, F.C., and Schmidt, C.F.,
"Microscopic viscoelasticity: Shear moduli of soft materials determined from thermal
fluctuations", Physical Review Letters, 1997. 79(17): p. 3286-3289.
72. Schnurr, B., Gittes, F., MacKintosh, F.C., and Schmidt, C.F., "Determining
microscopic viscoelasticity in flexible and semiflexible polymer networks from
thermal fluctuations", Macromolecules, 1997. 30(25): p. 7781-7792.
73. Gisler, T. and Weitz, D.A., "Scaling of the microrheology of semidilute F-actin
solutions", Physical Review Letters, 1999. 82(7): p. 1606-1609.
74. Allen, M.P. and Tildesley, D.J., Computer Simulation of Liquids. 1987, New York:
Oxford University Press.
157
75. Han, J. and Craighead, H.G., "Separation of long DNA molecules in a
microfabricated entropic trap array", Science, 2000. 288(5468): p. 1026-1029.
76. Doyle, P.S., Ladoux, B., and Viovy, J.L., "Dynamics of a tethered polymer in shear
flow", Physical Review Letters, 2000. 84(20): p. 4769-4772.
77. Dimalanta, E.T., Lim, A., Runnheim, R., Lamers, C., Churas, C., Forrest, D.K., de
Pablo, J.J., Graham, M.D., Coppersmith, S.N., Goldstein, S., and Schwartz, D.C., "A
microfluidic system for large DNA molecule arrays", Analytical Chemistry, 2004.
76(18): p. 5293-5301.
78. Tegenfeldt, J.O., Prinz, C., Cao, H., Chou, S., Reisner, W.W., Riehn, R., Wang,
Y.M., Cox, E.C., Sturm, J.C., Silberzan, P., and Austin, R.H., "The dynamics of
genomic-length DNA molecules in 100-nm channels", Proceedings of the National
Academy of Sciences of the United States of America, 2004. 101(30): p. 10979-10983.
79. Mannion, J.T., Reccius, C.H., Cross, J.D., and Craighead, H.G., "Conformational
analysis of single DNA molecules undergoing entropically induced motion in
nanochannels", Biophysical Journal, 2006. 90(12): p. 4538-4545.
80. Liang, X.G., Morton, K.J., Austin, R.H., and Chou, S.Y., "Single sub-20 nm wide,
centimeter-long nanofluidic channel fabricated by novel nanoimprint Mold
fabrication and direct imprinting", Nano Letters, 2007. 7(12): p. 3774-3780.
81. Agarwal, U.S., Dutta, A., and Mashelkar, R.A., "Migration of macromolecules under
flow - the physical origin and engineering implications", Chemical Engineering
Science, 1994. 49(11): p. 1693-1717.
82. Shaqfeh, E.S.G., "The dynamics of single-molecule DNA in flow", Journal of Non-
Newtonian Fluid Mechanics, 2005. 130(1): p. 1-28.
158
83. Kohale, S.C. and Khare, R., "Cross stream chain migration in nanofluidic channels:
Effects of chain length, channel height, and chain concentration", Journal of
Chemical Physics, 2009. 130(10): p. 014904.
84. Fang, L., Hu, H., and Larson, R.G., "DNA configurations and concentration in
shearing flow near a glass surface in a microchannel", Journal of Rheology, 2005.
49(1): p. 127-138.
85. Chen, Y.L., Graham, M.D., de Pablo, J.J., Jo, K., and Schwartz, D.C., "DNA
molecules in microfluidic oscillatory flow", Macromolecules, 2005. 38(15): p. 6680-
6687.
86. Fang, L. and Larson, R.G., "Concentration dependence of shear-induced polymer
migration in DNA solutions near a surface", Macromolecules, 2007. 40(24): p. 8784-
8787.
87. Jendrejack, R.M., Schwartz, D.C., de Pablo, J.J., and Graham, M.D., "Shear-induced
migration in flowing polymer solutions: Simulation of long-chain deoxyribose
nucleic acid in microchannels", Journal of Chemical Physics, 2004. 120(5): p. 2513-
2529.
88. Jhon, M.S. and Freed, K.F., "Polymer migration in newtonian fluids", Journal of
Polymer Science: Polymer Physics edition, 1985. 23(5): p. 955-971.
89. Ma, H.B. and Graham, M.D., "Theory of shear-induced migration in dilute polymer
solutions near solid boundaries", Physics of Fluids, 2005. 17(8): p. 083103.
90. Choi, H.J., Inn, Y.W., and Jhon, M.S., "Effect of temperature on polymer migration
.2. Concentration equation", Korean Journal of Chemical Engineering, 1994. 11(3):
p. 145-152.
159
91. Saintillan, D., Shaqfeh, E.S.G., and Darve, E., "Effect of flexibility on the shear-
induced migration of short-chain polymers in parabolic channel flow", Journal of
Fluid Mechanics, 2006. 557: p. 297-306.
92. Hernandez-Ortiz, J.P., Ma, H.B., Pablo, J.J., and Graham, M.D., "Cross-stream-line
migration in confined flowing polymer solutions: Theory and simulation", Physics of
Fluids, 2006. 18(12): p. 123101.
93. Hernandez-Ortiz, J.P., de Pablo, J.J., and Graham, M.D., "Fast computation of many-
particle hydrodynamic and electrostatic interactions in a confined geometry",
Physical Review Letters, 2007. 98(14): p. 140602.
94. Fan, X.J., Phan-Thien, N., Yong, N.T., Wu, X.H., and Xu, D., "Microchannel flow of
a macromolecular suspension", Physics of Fluids, 2003. 15(1): p. 11-21.
95. Millan, J.A., Jiang, W.H., Laradji, M., and Wang, Y.M., "Pressure driven flow of
polymer solutions in nanoscale slit pores", Journal of Chemical Physics, 2007.
126(12): p. 124905.
96. Fedosov, D.A., Karniadakis, G.E., and Caswell, B., "Dissipative particle dynamics
simulation of depletion layer and polymer migration in micro- and nanochannels for
dilute polymer solutions", Journal of Chemical Physics, 2008. 128(14): p. 144903.
97. Usta, O.B., Ladd, A.J.C., and Butler, J.E., "Lattice-Boltzmann simulations of the
dynamics of polymer solutions in periodic and confined geometries", Journal of
98. Usta, O.B., Butler, J.E., and Ladd, A.J.C., "Flow-induced migration of polymers in
dilute solution", Physics of Fluids, 2006. 18(3): p. 031703.
160
99. Khare, R., Graham, M.D., and de Pablo, J.J., "Cross-stream migration of flexible
molecules in a nanochannel", Physical Review Letters, 2006. 96(22): p. 224505.
100. Weeks, J.D., Chandler, D., and Andersen, H.C., "Role of Repulsive Forces in
Determining the Equilibrium Structure of Simple Liquids", The Journal of Chemical
Physics, 1971. 54(12): p. 5237-5247.
101. Khare, R., dePablo, J., and Yethiraj, A., "Rheological, thermodynamic, and structural
studies of linear and branched alkanes under shear", Journal of Chemical Physics,
1997. 107(17): p. 6956-6964.
102. Khare, R., dePablo, J., and Yethiraj, A., "Molecular simulation and continuum
mechanics study of simple fluids in non-isothermal planar couette flows", Journal of
Chemical Physics, 1997. 107(7): p. 2589-2596.
103. Khare, R., dePablo, J.J., and Yethiraj, A., "Rheology of confined polymer melts",
Macromolecules, 1996. 29(24): p. 7910-7918.
104. Stoltz, C., de Pablo, J.J., and Graham, M.D., "Concentration dependence of shear and
extensional rheology of polymer solutions: Brownian dynamics simulations", Journal
of Rheology, 2006. 50(2): p. 137-167.
105. Doi, M. and Edwards, S.F., The Theory of Polymer Dynamics. 1995, New York:
Oxford Science Publications.
106. Frenkel, D. and Smit, B., Understanding Molecular Simulations: From Algorithm to
Applications. 2002, San Diego, CA: Academic press.
107. Khare, R., Keblinski, P., and Yethiraj, A., "Molecular dynamics simulations of heat
and momentum transfer at a solid-fluid interface: Relationship between thermal and
161
velocity slip", International Journal of Heat and Mass Transfer, 2006. 49(19-20): p.
3401-3407.
108. Kumar, S. and Larson, R.G., "Brownian dynamics simulations of flexible polymers
with spring-spring repulsions", Journal of Chemical Physics, 2001. 114(15): p. 6937-
6941.
109. Usta, O.B., Butler, J.E., and Ladd, A.J.C., "Flow-induced migration of polymers in
dilute solution", Physics of Fluids, 2006. 18(3): p. 4.
110. Hough, L.A. and Ou-Yang, H.D., "Correlated motions of two hydrodynamically
coupled particles confined in separate quadratic potential wells", Physical Review E,
2002. 65(2): p. 021906.
111. Mellor, C.D., Sharp, M.A., Bain, C.D., and Ward, A.D., "Probing interactions
between colloidal particles with oscillating optical tweezers", Journal of Applied
Physics, 2005. 97(10): p. 103114.
112. Martin, S., Reichert, M., Stark, H., and Gisler, T., "Direct observation of
hydrodynamic rotation-translation coupling between two colloidal spheres", Physical
Review Letters, 2006. 97(24): p. 248301.
113. Voth, G.A., Bigger, B., Buckley, M.R., Losert, W., Brenner, M.P., Stone, H.A., and
Gollub, J.P., "Ordered clusters and dynamical states of particles in a vibrated fluid",
Physical Review Letters, 2002. 88(23): p. 234301.
114. Riedel, I.H., Kruse, K., and Howard, J., "A self-organized vortex array of
hydrodynamically entrained sperm cells", Science, 2005. 309(5732): p. 300-303.
162
115. Hernandez-Ortiz, J.P., Stoltz, C.G., and Graham, M.D., "Transport and collective
dynamics in suspensions of confined swimming particles", Physical Review Letters,
2005. 95(20): p. 204501.
116. Bhattacharya, S., "Cooperative motion of spheres arranged in periodic grids between
two parallel walls", Journal of Chemical Physics, 2008. 128(7): p. 074709.
117. Baron, M., Blawzdziewicz, J., and Wajnryb, E., "Hydrodynamic crystals: Collective
dynamics of regular arrays of spherical particles in a parallel-wall channel", Physical
Review Letters, 2008. 100(17): p. 174502.
118. Kohale, S.C. and Khare, R., "Molecular simulation of cooperative hydrodynamic
effects in motion of a periodic array of spheres between parallel walls", Journal of
119. Brenner, H., "The slow motion of a spherical particle through a viscous fluid towards
a plane surface", Chemical Engineering Science, 1961. 16: p. 242.
120. Goldman, A.J., Cox, R.G., and Brenner, H., "Slow viscous motion of a sphere parallel
to a plane wall - I Motion through a quiescent fluid", Chemical Engineering Science,
1967. 22: p. 637-651.
121. O'Neill, M.E., "A slow motion of viscous liquid caused by a slowly moving solid
sphere", Mathematika, 1964. 11: p. 67-74.
122. Walser, R., Mark, A.E., and van Gunsteren, W.F., "On the validity of Stokes' law at
the molecular level", Chemical Physics Letters, 1999. 303(5-6): p. 583-586.
123. Schmidt, J.R. and Skinner, J.L., "Hydrodynamic boundary conditions, the Stokes-
Einstein law, and long-time tails in the Brownian limit", Journal of Chemical Physics,
2003. 119(15): p. 8062-8068.
163
124. Schmidt, J.R. and Skinner, J.L., "Brownian motion of a rough sphere and the Stokes-
Einstein Law", Journal of Physical Chemistry B, 2004. 108(21): p. 6767-6771.
125. Cappelezzo, M., Capellari, C.A., Pezzin, S.H., and Coelho, L.A.F., "Stokes-Einstein
relation for pure simple fluids", Journal of Chemical Physics, 2007. 126(22): p.
224516.
126. Alley, W.E. and Alder, B.J., "Generalized transport-coefficients for hard-spheres",
Physical Review A, 1983. 27(6): p. 3158-3173.
127. Vergeles, M., Keblinski, P., Koplik, J., and Banavar, J.R., "Stokes drag at the
molecular-level", Physical Review Letters, 1995. 75(2): p. 232-235.
128. Vergeles, M., Keblinski, P., Koplik, J., and Banavar, J.R., "Stokes drag and
lubrication flows: A molecular dynamics study", Physical Review E, 1996. 53(5): p.
4852-4864.
129. Challa, S.R. and van Swol, F., "Molecular simulations of lubrication and solvation
forces", Physical Review E, 2006. 73(1): p. 016306.
130. Vilfan, A. and Julicher, F., "Hydrodynamic flow patterns and synchronization of
beating cilia", Physical Review Letters, 2006. 96(5): p. 058102.
131. Cui, B.X., Diamant, H., and Lin, B.H., "Screened hydrodynamic interaction in a
narrow channel", Physical Review Letters, 2002. 89(18): p. 188302.
132. Drazer, G., Khusid, B., Koplik, J., and Acrivos, A., "Wetting and particle adsorption
in nanoflows", Physics of Fluids, 2005. 17(1): p. 017102.
133. Clague, D.S. and Cornelius, P.J., "The hydrodynamic force and torque on a bounded
sphere in Poiseuille flow", International Journal for Numerical Methods in Fluids,
2001. 35(1): p. 55-70.
164
134. Chen, S., Phan-Thien, N., Khoo, B.C., and Fan, X.J., "Flow around spheres by
dissipative particle dynamics", Physics of Fluids, 2006. 18(10): p. 103605.
135. Travis, K.P., Searles, D.J., and Evans, D.J., "Strain rate dependent properties of a
simple fluid", Molecular Physics, 1998. 95(2): p. 195-202.
136. Borzsak, I., Cummings, P.T., and Evans, D.J., "Shear viscosity of a simple fluid over
a wide range of strain rates", Molecular Physics, 2002. 100(16): p. 2735-2738.
137. Toxvaerd, S., "The structure and thermodynamics of a solid-fluid interface", Journal
of Chemical Physics, 1981. 74(3): p. 1998-2005.
138. Israelachvili, J.N. and Pashley, R.M., "Molecular layering of water at surfaces and
origin of repulsive hydration forces", Nature, 1983. 306(5940): p. 249-250.
139. Thompson, P.A. and Robbins, M.O., "Shear-flow near solids - epitaxial order and
flow boundary-conditions", Physical Review A, 1990. 41(12): p. 6830-6837.
140. Press, W.H., Teukolsky, S.A., Vellerling, W.T., and Flannery, B.P., Numerical
recipes in FORTRAN. 1992, Cambridge, UK: Cambridge University Press.
141. Alms, G.R., Bauer, D.R., Brauman, J.I., and Pecora, R., "Depolarized Rayleigh
scattering and orientational relaxation of molecules in solution. I. Benzene, toluene,
and para-xylene", The Journal of Chemical Physics, 1973. 58(12): p. 5570-5578.
142. Hu, C.-M. and Zwanzig, R., "Rotational friction coefficients for spheroids with the
slipping boundary condition", The Journal of Chemical Physics, 1974. 60(11): p.
4354-4357.
143. Youngren, G.K. and Acrivos, A., "Rotational friction coefficients for ellipsoids and
chemical molecules with the slip boundary condition", The Journal of Chemical
Physics, 1975. 63(9): p. 3846-3848.
165
144. Hynes, J.T., Kapral, R., and Weinberg, M., "Slip boundary condition for rough sphere
rotation", Chemical Physics Letters, 1977. 46(3): p. 463-466.
145. Hynes, J.T., Kapral, R., and Weinberg, M., "Slip boundary condition and rough
sphere angular velocity correlations", Chemical Physics Letters, 1977. 47(3): p. 575-
577.
146. Hynes, J.T., Kapral, R., and Weinberg, M., "Molecular rotation and reorientation:
Microscopic and hydrodynamic contributions", The Journal of Chemical Physics,
1978. 69(6): p. 2725-2733.
147. Hynes, J.T., Kapral, R., and Weinberg, M., "Molecular theory of translational
diffusion: Microscopic generalization of the normal velocity boundary condition",
The Journal of Chemical Physics, 1979. 70(3): p. 1456-1466.
148. Padding, J.T., Wysocki, A., Lowen, H., and Louis, A.A. Stick boundary conditions
and rotational velocity auto-correlation functions for colloidal particles in a coarse-
grained representation of the solvent. in 6th Liquid Matter Conference. 2005.
Utrecht, NETHERLANDS: Iop Publishing Ltd.
149. Macosko, C.W., Rheology: Principles, Measurements and Applications. 1993, Ney
York: Wiley-VCH.
150. MacKintosh, F.C. and Schmidt, C.F., "Microrheology", Current Opinion in Colloid &
Interface Science, 1999. 4(4): p. 300-307.
151. Mason, T.G., Gang, H., and Weitz, D.A., "Rheology of complex fluids measured by
dynamic light scattering", Journal of Molecular Structure, 1996. 383(1-3): p. 81-90.
166
152. Dasgupta, B.R., Tee, S.Y., Crocker, J.C., Frisken, B.J., and Weitz, D.A.,
"Microrheology of polyethylene oxide using diffusing wave spectroscopy and single
scattering", Physical Review E, 2002. 65(5): p. 051505.
153. Mizuno, D., Head, D.A., MacKintosh, F.C., and Schmidt, C.F., "Active and Passive
Microrheology in Equilibrium and Nonequilibrium Systems", Macromolecules, 2008.
41(19): p. 7194-7202.
154. Tseng, Y. and Wirtz, D., "Mechanics and multiple-particle tracking
microheterogeneity of alpha-actinin-cross-linked actin filament networks",
Biophysical Journal, 2001. 81(3): p. 1643-1656.
155. Tseng, Y., Kole, T.P., Lee, S.H.J., and Wirtz, D., "Local dynamics and viscoelastic
properties of cell biological systems", Current Opinion in Colloid & Interface
Science, 2002. 7(3-4): p. 210-217.
156. Tseng, Y., An, K.M., and Wirtz, D., "Microheterogeneity controls the rate of gelation
of actin filament networks", Journal of Biological Chemistry, 2002. 277(20): p.
18143-18150.
157. Furst, E.M., "Applications of laser tweezers in complex fluid rheology", Current
Opinion in Colloid & Interface Science, 2005. 10(1-2): p. 79-86.
158. Meyer, A., Marshall, A., Bush, B.G., and Furst, E.M., "Laser tweezer microrheology
of a colloidal suspension", Journal of Rheology, 2006. 50(1): p. 77-92.
159. Furst, E.M., "Interactions, structure, and microscopic response: Complex fluid
rheology using laser tweezers", Soft Materials, 2003. 1(2): p. 167-185.
167
160. Strick, T.R., Dessinges, M.N., Charvin, G., Dekker, N.H., Allemand, J.F., Bensimon,
D., and Croquette, V., "Stretching of macromolecules and proteins", Reports on
Progress in Physics, 2003. 66(1): p. 1-45.
161. Mukhopadhyay, A. and Granick, S., "Micro- and nanorheology", Current Opinion in
Colloid & Interface Science, 2001. 6(5-6): p. 423-429.
162. Harden, J.L. and Viasnoff, V., "Recent advances in DWS-based micro-rheology",
Current Opinion in Colloid & Interface Science, 2001. 6(5-6): p. 438-445.
163. Mason, T.G., Gisler, T., Kroy, K., Frey, E., and Weitz, D.A. Rheology of F-actin
solutions determined from thermally driven tracer motion. in 71st Annual Meeting of
the Society-of-Rheology. 1999. Madison, Wisconsin.
164. Claessens, M., Tharmann, R., Kroy, K., and Bausch, A.R., "Microstructure and
viscoelasticity of confined serniflexible polymer networks", Nature Physics, 2006.
2(3): p. 186-189.
165. Liu, J., Gardel, M.L., Kroy, K., Frey, E., Hoffman, B.D., Crocker, J.C., Bausch, A.R.,
and Weitz, D.A., "Microrheology probes length scale dependent rheology", Physical
Review Letters, 2006. 96(11): p. 118104.
166. Mizuno, D., Tardin, C., Schmidt, C.F., and MacKintosh, F.C., "Nonequilibrium
mechanics of active cytoskeletal networks", Science, 2007. 315(5810): p. 370-373.
167. Wilhelm, C., "Out-of-equilibrium microrheology inside living cells", Physical Review
Letters, 2008. 101(2): p. 028101.
168. Bausch, A.R., Moller, W., and Sackmann, E., "Measurement of local viscoelasticity
and forces in living cells by magnetic tweezers", Biophysical Journal, 1999. 76(1): p.
573-579.
168
169. Bausch, A.R. and Kroy, K., "A bottom-up approach to cell mechanics", Nature
Physics, 2006. 2(4): p. 231-238.
170. Valentine, M.T., Bausch, A., Kaplan, P.D., Crocker, J.C., and Weitz, D.A.,
"Measuring the local mechanical response of cells", Biophysical Journal, 2001.
80(1): p. 2117.
171. Li, Y.X., Vanapalli, S.A., and Duits, M.H.G., "Dynamics of ballistically injected
latex particles in living human endothelial cells", Biorheology, 2009. 46(4): p. 309-
321.
172. Solomon, M.J. and Lu, Q., "Rheology and dynamics of particles in viscoelastic
media", Current Opinion in Colloid & Interface Science, 2001. 6(5-6): p. 430-437.
173. Valentine, M.T., Dewalt, L.E., and OuYang, H.D. Forces on a colloidal particle in a
polymer solution: A study using optical tweezers. in 3rd Liquid Matter Conference.
1996. Norwich, England.
174. Sierou, A. and Brady, J.F., "Rheology and microstructure in concentrated
noncolloidal suspensions", Journal of Rheology, 2002. 46(5): p. 1031-1056.
175. Khair, A.S. and Brady, J.F., ""Microviscoelasticity" of colloidal dispersions", Journal
of Rheology, 2005. 49(6): p. 1449-1481.
176. Khair, A.S. and Brady, J.F., "On the motion of two particles translating with equal
velocities through a colloidal dispersion", Proceedings of the Royal Society a-
Mathematical Physical and Engineering Sciences, 2007. 463(2077): p. 223-240.
177. Khair, A.S. and Brady, J.F., "Microrheology of colloidal dispersions: Shape matters",
Journal of Rheology, 2008. 52(1): p. 165-196.
169
178. Carpen, I.C. and Brady, J.F., "Microrheology of colloidal dispersions by Brownian
dynamics simulations", Journal of Rheology, 2005. 49(6): p. 1483-1502.
179. Xu, J.Y., Viasnoff, V., and Wirtz, D., "Compliance of actin filament networks
measured by particle-tracking microrheology and diffusing wave spectroscopy",
Rheologica Acta, 1998. 37(4): p. 387-398.
180. Palmer, A., Mason, T.G., Xu, J.Y., Kuo, S.C., and Wirtz, D., "Diffusing wave
spectroscopy microrheology of actin filament networks", Biophysical Journal, 1999.
76(2): p. 1063-1071.
181. Mason, T.G., "Estimating the viscoelastic moduli of complex fluids using the
generalized Stokes-Einstein equation", Rheologica Acta, 2000. 39(4): p. 371-378.
182. Chen, D.T., Weeks, E.R., Crocker, J.C., Islam, M.F., Verma, R., Gruber, J., Levine,
A.J., Lubensky, T.C., and Yodh, A.G., "Rheological microscopy: Local mechanical
properties from microrheology", Physical Review Letters, 2003. 90(10): p. 108301.
183. Hough, L.A. and Ou-Yang, H.D., "A new probe for mechanical testing of
nanostructures in soft materials", Journal of Nanoparticle Research, 1999. 1: p. 495-
499.
184. Squires, T.M. and Brady, J.F., "A simple paradigm for active and nonlinear
microrheology", Physics of Fluids, 2005. 17(7): p. 073101.
185. Squires, T.M., "Nonlinear microrheology: Bulk stresses versus direct interactions",
Langmuir, 2008. 24(4): p. 1147-1159.
186. Khare, R., de Pablo, J., and Yethiraj, A., "Molecular simulation and continuum
mechanics investigation of viscoelastic properties of fluids confined to molecularly
thin films", Journal of Chemical Physics, 2001. 114(17): p. 7593-7601.
170
187. Hough, L.A. and Ou-Yang, H.D., "Viscoelasticity of aqueous telechelic poly(ethylene
oxide) solutions: Relaxation and structure", Physical Review E, 2006. 73(3): p.
031802.
188. Sen, S., Kumar, S.K., and Keblinski, P., "Viscoelastic properties of polymer melts
from equilibrium molecular dynamics simulations", Macromolecules, 2005. 38(3): p.
650-653.
189. Vladkov, M. and Barrat, J.L., "Linear and nonlinear viscoelasticity of a model
unentangled polymer melt: Molecular dynamics and rouse modes analysis",
Macromolecular Theory and Simulations, 2006. 15(3): p. 252-262.

Dissertation SwapNil KoHale

Uploaded by

Copyright:

Available Formats

Dissertation SwapNil KoHale

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dissertation SwapNil KoHale

Uploaded by

Copyright:

Available Formats

MOLECULAR HYDRODYNAMICS IN COMPLEX FLUIDS

is the shear rate) provides a good

whereas the radius of solute is equal to

+ a . Hence, to keep the mass density of solute and solvent to be same,

d = 22.5 (Active nanorheology)

d = 22.5 (Active nanorheology)

You might also like