You are on page 1of 147

A Dissertation

entitled
Percolation Modeling in Polymer Nanocomposites
by
Azita Belashi
Submitted to the Graduate Faculty as partial fulfillment of the
requirements for the Doctor of Philosophy Degree in Engineering




________________________________________
Dr. Lesley M. Berhan, Committee Chair

________________________________________
Dr. Maria R. Coleman, Committee Member

________________________________________
Dr. Yong X. Gan, Committee Member

________________________________________
Dr. Ahalapitiya H. Jayatissa, Committee Member

________________________________________
Dr. Ioan D. Marinescu, Committee Member

________________________________________
Dr. Patricia R. Komuniecki, Dean
College of Graduate Studies


The University of Toledo
May 2011










































Copyright 2011, Azita Belashi
This document is copyrighted material. Under copyright law, no parts of this document
may be reproduced without the expressed permission of the author.
iii

An Abstract of

Percolation Modeling in Polymer Nanocomposites

by

Azita Belashi

Submitted to the Graduate Faculty in partial fulfillment of the
requirements for the Doctor of Philosophy in Engineering

The University of Toledo
May 2011



The electrical conductivity of polymer nanocomposites follows a percolation-like
behavior which is attributed to the formation of a connected network of conductive
inclusions within the insulating polymer matrix. Percolation in polymer nanocomposites
is of high interest due to the potential to create electrically conductive materials with an
optimized conductivity and as low as possible a loading of conductive fillers.
The development of powerful numerical models has enabled researchers to
simulate the microstructure of composites and predict percolation in disordered systems.
Most of the literature studies model fibers as fully interpenetrating objects into each other
which presume that the electrical and geometric percolation thresholds occur
simultaneously which is true only when in a composite the reinforcing particles coalesce
so that a physically connected network is produced.
In this dissertation, an approach has been developed aimed at developing a
qualitative and quantitative picture of the electrical behavior of particulate nanocomposite
materials. The continuum connectedness model was applied to the systems of uniform
and variable objects of a given geometry to simulate the realistic percolation in both
iv

nanosheet and fibrous systems (nanofibers, nanotubes) and predict how shape and size
distribution influence the percolation thresholds. The relationship between percolation
threshold and excluded volume in systems which obey finite size scaling theory was
investigated using Monte Carlo simulations. Voltage-current studies were performed on
the nanocomposite specimens to measure the electrical conductivity. The results were
interpreted using both an analytical percolation theory model and numerical simulations.
The agreement between the predictions and the experimental results establishes the
effectiveness of the model in simulating the electrical percolation phenomenon in
nanofibers or any conductive particles suspended in an electroconductive medium,
demonstrating that the model can be used as a predictive tool for designing
nanocomposite materials.

















To my parents,
for their unconditional love.
Hope they shall be proud of me.
And my sisters and brothers,
who always supported me.

This dissertation is also dedicated to my aunt Mina Mostafavi, passed away at 36
on July 21st, 1984 at north of Iran.


Azita, May 2010, USA











vi



Acknowledgements



I would like to express my gratitude to Dr. Lesley M. Berhan for suggesting the
idea for the project, helping me initiate the simulation studies, for her excellent
recommendations implementing the growth algorithms, for her many insightful
comments during the course of this work, as well as for bringing to my attention many
papers mentioned in the reference.
I am indebted to committee members on final evaluation; Drs. Maria R. Coleman,
Yong X. Gan, Ahalapitiya H. Jayatissa, and Ioan D. Marinescu for a critical reading of
this dissertation and their valuable suggestions.
I am pleased to acknowledge gratefully the assistance of Nima Hakimelahi and
Javed Mapkar, Chemical and Environmental Engineering Department, for their assistance
in preparation of the experimental samples and gathering some of the conductivity data,
and Victor Plotnikov, Department of Physics and Astronomy, for his technical help in
conductivity experiments.
The author would like to thank Leslie Neuendorff and Shannon McKeechen for
editorial assistance.
Last but not least, I wish to thank the staff of University of Toledo for their
hospitality.

vii


Contents

Abstract iii
Acknowledgments vi
Contents vii
List of Tables xi
List of Figures xii
List of Abbreviations xv
List of Symbols xvi
1 Introduction 1
1.1 Motivation and Background 1
1.2 Objectives 3
1.3 Outline of Dissertation 3
2 Literature Review 5
2.1 Introduction 5
2.2 Electrical Conductivity 6
2.3 Electron Hopping and Tunneling 7
2.4 Finite Size Scaling Theory 8
2.5 Percolation and Conductivity 9
2.6 Theory of Percolation Models 11
viii

2.6.1 Lattice Modeling 11
2.6.2 Continuum Modeling 11
2.7 Percolation Threshold Studies Survey 13
2.8 Conclusion 24
3 Modeling and Simulation Method 26
3.1 Introduction 26
3.2 Methods and Approach 27
3.2.1 Monte Carlo Methods 27
3.2.2 Excluded Area and Excluded Volume 28
3.3 Circles and Spheres Simulation 30
3.3.1 Soft-Core Modeling 31
3.3.2 Hard-Core with Soft-Shell Modeling 34
3.4 Spherocylinder Simulation 37
3.4.1 Geometry and Modeling 38
3.4.2 Intersection of Spherocylinders 39
3.4.3 Excluded Volume of a Spherocylinder 46
3.4.4 Soft-Core Modeling 47
3.4.5 Hard-Core with Soft-Shell Modeling 48
3.5 Conclusion 49
4 Results and Discussion 50
4.1 Introduction 50
4.2 Circles Percolation 50
4.3 Spheres Percolation 58
ix

4.3.1 Computational and Experimental Comparison 66
4.4 Spherocylinders Percolation 71
4.4.1 Excluded Volume Modeling 71
4.4.2 Spheres Percolation 73
4.4.3 Soft-core Modeling 74
4.4.4 Hard-Core with Soft-Shell Modeling 79
4.5 Conclusion 87
5 Experiment 89
5.1 Introduction 89
5.2 Electrical Measurement 90
5.3 Theory of Four-Point Probe 91
5.4 CNF Buckypaper Conductivity 93
5.4.1 Materials 93
5.4.2 Chemical Surface Treatment 94
5.4.3 Sample Preparation 96
5.4.4 Experimental Set-up and Test Performance Test 99
5.5 Results and Discussion 101
5.6 Analytical Calculation 106
5.7 Comparison with Other Filled Composites 109
5.8 Experimental and Computational Comparison 110
5.9 Conclusion 112
6 Conclusions 114
6.1 Summary 114
x

6.2 Concluding Remarks 115
6.3 Future Work 117
References 119
Appendix 129















xi


List of Tables
2.1 Key percolation studies in two- and three-dimensional of continuum models 24
4.1 Total excluded area of systems of soft circles with different sizes 54
4.2 Total excluded volume of systems of soft spheres with different sizes 62
4.3 Estimates of shell thickness for composite systems of spherical fillers 69
4.4 Estimates of excluded volume of spherocylinders with different aspect ratios

72
4.5 Total excluded volume in systems of soft spherocylinders 78
4.6 Total excluded volume in the systems of spherocylinders with shell aspect
ratio=10

81
4.7 Total excluded volume in the systems of spherocylinders with shell aspect
ratio=100

85
5.1 Density of materials 99
5.2 Fiber loading levels 99
5.3 Percolation threshold of the collected experimental data 109
5.4 Computational results for critical volume fraction of spherocylinders 111





xii


List of Figures
2.1 The percolation threshold in infinite and finite system 9
3.1 Excluded area (volume) of a circle (sphere) 29
3.2 Schematic presentation of boundary condition 32
3.3 Examples of non-percolating and percolating network of soft circles in 2D
cells

33
3.4 Examples of non-percolating and percolating network of soft spheres in 3D
cells

34
3.5 Hard-core with soft-shell model of a circle or sphere 35
3.6 Examples of non-percolating and percolating network of hard-core-soft-
shell circles in 2D cells

36
3.7 Examples of non-percolating and percolating network of hard-core-soft-
shell spheres in 3D cells

37
3.8 Schematic representation of a spherocylinder model 38
3.9 Nomenclature describing orientation of a single spherocylinder 39
3.10 Schematic representation of three different patterns of contact between
spherocylinders

40
3.11 The side-to-side intersecting of two cylinders 41
3.12 The end-to-side intersecting of two spherocylinders 44
3.13 The end-to-end intersecting of two spherocylinders 45
3.14 Excluded volume of a spherocylinder of length L and radius R 46
3.15 A typical random fiber network generated by the algorithms used in the
present work for modeling soft spherocylinders of uniform aspect ratio


48
xiii

3.16 Schematic representation of the hard-core with soft-shell model of
spherocylinder
48


4.1 Variation of average density of uniform soft circles per square area vs
standard deviation for three different radii

51
4.2 Variation of average density of soft circles vs standard deviation 53
4.3 Variation of average density of hard-core with soft-shell circles vs standard
deviation

56
4.4 Dependence of the total excluded area of the system of hard-core with soft-
shell circles on the ratio of core excluded area to shell excluded area

57


4.5 Variation of density of equal soft spheres vs standard deviation.

59

4.6 Variation of density of soft spheres vs standard deviation.

61

4.7 Variation of density of hard-core with soft-shell spheres vs standard
deviation

64


4.8 Dependence of the total excluded volume of the system of hard-core with
soft-shell spheres on the ratio of core excluded volume to shell excluded
volume
65



4.9 The unit cubic cell and one spherocylinder at the center 72
4.10 Variation of average density of soft spherocylinders with length zero vs
standard deviation

73


4.11 Variation of average density of soft spherocylinders vs standard deviation
for different aspect ratios
75


4.12 Variation of average density of soft spherocylinders with variable aspect
ratios vs standard deviation

77
4.13 Variation of average density of hard-core with soft-shell spherocylinders vs
standard deviation. For all graphs, the shell aspect ratio is 10 = D L

80


4.14 Dependence of the total excluded volume of the system of hard-core with
soft-shell spherocylinders on the ratio of core excluded volume to shell
excluded volume
82



4.15 Variation of average density of hard-core with soft-shell spherocylinders vs
standard deviation. For all graphs, the shell aspect ratio is 100 = D L

84
xiv

4.16 Dependence of the total excluded volume of the system of hard-core with
soft-shell spherocylinders on the ratio of core excluded volume to shell
excluded volume

86



5.1 Enlarged schematic of four-point probe configuration 91
5.2 Carbon nanofiber powder 94
5.3 Functional groups attached to the surface of carbon nanofibers 95
5.4 Steps in casting CNF-composite film 97
5.5 Random CNF-composite samples 98
5.6 The electronic caliper 98
5.7 Four-point probe apparatus 100
5.8 DC surface conductivity as a function of the CNF loading 101
5.9 Conductivity in logarithmic scale as a function of the CNF loading

103
5.10 DC Surface electrical conductivity variations with CNF volume
concentration

105
5.11 Percolation equation fit to the experimental data 108
A.1 Flow chart of network generation 130
















xv



List of Abbreviations
ASTM ...... American Society of Testing and Materials
a.r. .... Aspect ratio
CNF ......... Carbon Nanofiber
DC ....... Direct Current
EMI ................................. Electromagnetic Interference
FSS ...... Finite Size Scaling
MATLAB .... Matrix Laboratory
OCNF ...... Oxidized Carbon Nanofiber
PDMS ...... Polydimethylsiloxane
S ...... Siemens
S.D. ...... Standard Deviation
wt. ........ Weight
vol. ........... Volume








xvi


List of Symbols
A constant of proportionality in classic percolation equation
a . correction factor for resistivity of finite size sample
e
A ...
total excluded area
ex
A ..
excluded area
S
A ...
cross section area
A ...
area

ex
A ..
average excluded area
D .... diameter of circle or sphere
d . diameter of core
ij
d ...
minimum distance between the centers of the two circles
G. electric conductance
I .. electric current
L . length of the cylinder
l .. length
n n
M

..
matrix of objects coordinates
c
P ...
critical density of objects
electric resistivity
0
...
resistivity of semi-infinite sample
R correlation factor
R radius of the shell
r . radius of the core
S
R ...
sheet resistance
electric resistance
s . interprobe spacing in four-point probe instrument
Sin . the average value of Sin
t .. ratio of the radius of the core to the radius of the shell
t .. dimensionless exponent in classic percolation equation
t .
thickness of specimen
U voltage
V. voltage
e
V
total excluded volume
ex
V ..
excluded volume
F
V ..
volume of filler
P
V ...
volume of polymer
xvii

V
Volume of the geometrical object

ex
V ...
average excluded volume
W. size of the system
. angle between two sticks
. azimuthal angle
electrical conductivity
. density or volume fraction
polar angle
C
...
percolation threshold or critical volume fraction
F
...
filler concentration
. ohm












1


Chapter 1
Introduction
1.1 Motivation and Background
Composite materials have attracted considerable attention over the last several
decades due to their economic importance and ease of manufacturing as well as
environmental stability [Gay et al., 2002]. Among different composite systems,
nanocomposites are gaining strong interest for their extraordinary physical properties in
addition to excellent conductive and mechanical performance. Nanocomposites are a new
class of materials containing two or more constituents of which at least one is at the
nanoscale [Ajayan et al., 2003]. In a nanocomposite, the interaction between nanofillers
and polymer media is at the molecular level. The properties of the matrix polymer and the
nanofiller and the amount and distribution of the nanofiller have a strong impact on the
interactions between the nanofillers and polymer.
Of particular interest is the so-called percolation threshold, i.e. the minimum
loading of nanofillers required to form a cluster that spans the whole system [Kyrylyuk
and Schoot, 2008]. Of course, one important and as yet not completely answered question
concerns how to have control over the formation of percolating clusters. Not surprisingly,
2

this would provide opportunities to manipulate the percolation threshold and the level of
conductivity to produce a nanocomposite with an optimized conductivity and as low as
possible a loading of conductive fillers. Different conductive additives are generally very
useful in production of electrically conductive polymer composites. However, loading at
higher content or a larger size of the filler is associated with a degradation of mechanical
properties and manufacturability as well as significant increases in weight and
manufacturing cost [Chisholm et al., 2005; Fu et al., 2008; Manhart1 et al. 2000]. For this
reason, minimizing the experimental percolation threshold without sacrificing the
mechanical characteristics is very important.
At present, it seems that not enough is known about the physics of electrically
conducting networks of fillers suspended in polymeric medium. Various studies of
percolation in composites have been proposed to obtain the minimum amount of fiber
loading which allows maximum electrical conductivity by both numerical and analytical
techniques [Janzen, 1974; Wang and Ogale, 1993; Dani and Ogale, 1994; Nakamura et
al., 1997; Zou et al., 2002]. It is challenging to decouple the effect of nanoparticle size
and their dispersion on the percolating properties of nanocomposites. Hence the particle
size effect on percolation of nanocomposites has not been convincingly established
because the quality of particle dispersion in nanocomposites cannot be assured. A
fundamental understanding of the influence of the material properties and processing
techniques on the percolation threshold of electroconductive composites, however,
remains elusive.
In this dissertation, a detailed study of the electrical behavior of polymer carbon
nanofiber composites was conducted using a combination of computational and
3

experimental approaches which is expandable to any conductive particles with any aspect
ratio.

1.2 Objectives
This study sought to address the percolation issues through a combination of
computational and experimental methods.
The specific objectives of the research are
- Performing numerical simulations using Monte Carlo techniques to model the
percolation for various systems of objects with a uniform shape and different distribution
of size.
- Fabricating nanocomposite samples and performing electric conductivity tests
on specimens over a range of volume fractions.
- Combining simulation and experiment results to infer the critical values of
concentration for electrical conductivity percolation.

1.3 Outline of Dissertation
The material of this dissertation has been arranged in six chapters. Chapter Two
concerns theories of, and models for, percolation, especially those exhibiting a
percolation threshold. The demonstration of a mathematical model and a computational
method proposed to simulate the percolation thresholds for nanocomposites are presented
in Chapter Three. Chapter Four is devoted to the results of the simulation and
discussions. The experimental work is put forward in Chapter Five. The key conclusions
based on the present study and regarding the future research projects are summarized in
4

the final chapter. The description of the algorithm used in numerical simulation and
programming the computer has been provided in the appendix. The organization of this
thesis should facilitate readability and comprehension.








5


Chapter 2
Literature Review
2.1 Introduction
Percolation is a random probabilistic process which exhibits a phase transition.
Different percolation systems may contain clusters of different sizes and shapes. Studying
the statistics of the clusters helps to identify the critical value of density when formation
of infinite or long-range connectivity in random systems first occurs. This is called the
percolation threshold [Grimmett, 1999].
Percolation is a very general phenomenon applicable in almost every area of
science as the simplest model for spatial disorder. It has been shown that percolation has
application to a broad range of topics including but not limited to mathematics, physics,
biology, geography, hydrology, petroleum, ecology, chemistry, and materials science. In
this study we shall focus our attention on implementation of percolation to electrical
conductivity in composite materials.



6

2.2 Electrical Conductivity
Before detailed discussion on percolation, it is helpful to establish some notation
and terminology.
The electrical resistance of a circuit component or device is defined as the ratio of
the voltage applied to the electric current which flows through it:
I
V
= (2-1)
If the resistance is constant over a considerable range of voltage, Ohms law can
be used in the form:

=
V
I (2-2)
Whether or not a material obeys Ohms law, its resistance can be described in
terms of its resistivity :
S
A
l
= (2-3)
where l and
S
A are respectively the length and cross section area of the material.
Resistivity is the inherent property which takes into account the nature of the electrical
resistance material [Northrup, 1912].
Electric conductance is defined as the inverse or reciprocal of electric resistance
and uses the measurement unit mhos, which is also called Siemens after Dr. Ernst Werner
von Siemens [BIPM, 2006].

1
G (2-4)
7

Similarly, the inverse of resistivity is called electrical conductivity which is a
measure of the ability of a material to conduct an electric current.


1
(2-5)

2.3 Electron Hopping and Tunneling
The electrical resistance in a composite system comprising a network of
conducting and insulating phases is the result of a large number of resistors combined in
series and parallel. There are different contributions to the resistance or conductance,
with hopping and tunneling effect on conductance generally dominating the magnitude of
the overall conductance [Ruschau et al., 1992].
When an electric field is applied to two media in contact with each other, charge
polarization occurs at the interface due to the differences between the ratios of the
electrical properties such as resistivity [Tuncer et al., 2002]. It has been shown that, at a
specific low temperature, electrons are able to transfer between localized states through
hopping conduction and electron tunneling.
Hopping refers to the process in which a charge carrier is suddenly displaced from
one occupied state to another equivalent empty donor state of higher energy
[Luk'yanchuk and Mezzane, 2008; Psarras, 2006]. Hopping conduction takes place via
defects or impurities which form potential wells (traps or localized states) that are
favorable for charge carriers (electrons and ions) to hop. The degree of hopping
conductivity is dependent on donor and acceptor atom concentrations in the material. If
8

the applied voltage is assumed to be alternating, the frequency of the field affects the
hopping rate as well [Sahimi, 2003].
The tunneling distance is the average distance between the two closest separated
points of adjacent particles and is only a few nanometers [Compton, 1989]. Tunneling is
a manifested characteristic of the conductivity of nanocomposites. The conductivity of
the percolating network in nanocomposites appears to be the result of electron transport
between the nearest neighbor connections of one fiber to another which denote the
dominant conducting elements across the inter-particle tunneling distance. [Balberg,
2002].

2.4 Finite Size Scaling Theory
Finite size scaling theory is a powerful tool for the study of phase transitions. The
theory describes how long-scale clusters of random media near a critical point manifest
themselves in finite size systems compared to infinite system [Brankov, 2000].
The percolation threshold of an infinite system, in particular, is well defined by a
distinctive phase transition at a finite value of
C
as shown in Figure 2.1 [Stauffer and
Aharony, 1991]. This means the infinite system certainly contains no percolating cluster
for density lower than
C
, and whenever
C
> , the system percolates. Therefore the
probability of a spanning path is either 0 or 1.
In the case of a system which has finite size in at least one space dimension, on
the other hand, such a discrete phase transition technically does not occur. In fact, a
system of finite size might have a percolating cluster for densities lower than
C
and/or
might fail to percolate at densities higher than
C
.
9

The problem of how to deal with percolation in finite size volumes, like those in
numerical computer simulations which are always done on limited-size percolation
models, is discussed through finite size scaling theory. The theory analyzes the formation
of the bulk properties when the size of a finite system is increased. To estimate the
percolation threshold for an infinite-sized system, the approach is to determine the
percolation thresholds in finite sized systems and extrapolate the results using the finite
size scaling theory as the system is growing.


Figure 2.1 The percolation threshold in infinite and finite systems
(W: Size of the system).


2.5 Percolation and Conductivity
An example of the percolation problem is clustering of conductive particles in an
insulating matrix [Sahimi, 1994]. In fact, a major cause of recent interest in percolation is
that percolation effects play an important role in conductor-insulator transition seen in
some composite systems.
10

The electrical conductivity of a composite depends on the conductive filler
loading [Fiske et al., 1999]. At low filler concentration, the composite conductivity is
almost the same as the polymer matrix. By adding more filler, the conductivity of
composite begins to increase very slowly. At the percolation threshold, the amount of
filler is enough to begin the formation of a continuous conductive network throughout the
polymer matrix. In this region, the conductivity of the composite undergoes an increase
of several orders of magnitude. In other words, for a composite, the percolation threshold
is the lowest concentration of filler that forms continuous conductive pathways
throughout the polymer matrix. After that, the conductivity of composite again changes
very little with filler loading until it is almost the same as conductive filler.
This variation of conductivity of the composite materials, as the conductive filler
content is increased, typically shows a classical S-shaped curve (Fig. 2-1) and can be
described by a power-law relationship known as percolation theory [Hunt, 2005]:
t
C F
A ) ( = (2-6)
where is the electrical conductivity of composite,
F
is the filler concentration,
C
is
the percolation threshold, A is the constant of proportionality and t is a dimensionless
exponent. The percolation threshold and critical exponent may be determined by different
methods like series expansion [Haan and Zwanzig, 1977], Monte Carlo simulation [Jain,
1998], the renormalization group approach [Stinchcombe and Watson, 1976], and even
more sophisticated methods. Generally their values are estimated by curve fitting of
experimental results. The value of the constant A theoretically should approach the
conductivity of the filler.

11

2.6 Theory of Percolation Models
Over the last several decades, a tremendous amount of work has gone into the
modeling of the percolation phenomenon [McCullough, 1985; Sevick et al., 1988;
Debondt et. al., 1992]. Two primary formulations are distinguished in finding exact and
approximate values of the percolation thresholds for a variety of systems.

2.6.1 Lattice Modeling
The common lattice models in studying percolation are square, triangular, and
honeycomb structures in two dimensions and cube and diamond in three dimensions
[Vanderzande, 1998]. Lattice models solve percolation problems through the concepts of
bonds and sites [Newman and Ziff, 2001].
Bond Percolation: The bond percolation is the simplest model. In bond
percolation the lattice edges are considered as the relevant entities.
Site Percolation: In site percolation the lattice vertices are considered as the
relevant entities and the system shows behavior qualitatively similar to, though different
in some details from, bond percolation.
Bond percolation can be transformed to site percolation by placing a site on each
bond and then connecting the sites which are closest neighbors [Feng and Jin, 2005].

2.6.2 Continuum Modeling
A variety of percolation phenomena occur in systems that cannot adequately be
represented with lattice models. Continuum or off-lattice models, on the other hand, are
more flexible in modeling the percolation related problems in composites. In this
12

modeling the geometrical objects can be placed randomly and individually in any spot of
the continuum space and the overlap of the objects is defined as connectivity [Meester
and Roy, 1996].
The main classes of continuum modeling are:
Soft-core Modeling: In this approach the fibers are modeled as fully
interpenetrating objects.
Hard-core Modeling: In the soft-core modeling, two particles are considered
bonded when they overlap. This represents a drawback of the soft-core approach since in
a real system solid filler particles may touch each other but do not merge. The hard-core
approach which simulates particles as touching objects has been argued to yield a better
simulating result for behavior of conductivity in real applications.
Hard-core with Soft-shell Modeling: The previous approaches assume the fibers
coalesce and produce a continuous conducting network. This means the electrical and
geometric percolation occur simultaneously. The hard-core with soft-shell approach
presumes the electrical percolation threshold depends not only on the geometry of the
conducting particles but also on the tunneling distance between inclusions.
In this approach, the conductive fibers are considered to be composed of a hard
core and a soft shell. The hard-core is impenetrable and represents the actual conducting
fiber. The penetrable soft shell which surrounds the hard core is aimed to represent the
effective range of electron transferring in the matrix polymer.
While the geometry and distribution of the reinforcing phase are the only required
inputs for previous approaches, here the shell thickness is required for modeling which is
13

complicated and cannot be derived easily by analysis or experiments. In general this
approach is more computationally expensive.

2.7 Percolation Threshold Studies Survey
In order to explain the composites conductivity behavior, considerable research
has been done to determine the percolation threshold for a wide range of systems.
The purpose of this section is to gather in one place an extension of the most
important lattice and continuum models which can be found in literature.
Pike and Seager [1974] were the first to explore percolation problems in two and
three dimensional systems for several types of lattice and continuum models including
circles, squares, sticks, spheres, and hemispheres. Utilizing Monte Carlo simulation, they
proposed approximate solutions to determine the critical value of the cluster parameter
which is the value at which infinite clusters first form. For example, for circles, the
critical value is the critical radius at which an infinite cluster of connected circles is
formed, and the critical size for squares is half of the critical length of the sticks. Their
work was concerned only with the percolation problem of isotropic systems of equal size
of randomly-oriented objects. Therefore, it is not clear whether or not the critical values
are independent of the object size distribution.
In a method similar to Pike and Seagar, Powell [1979] examined lattice
percolation problems in three-dimensional structures of equal-sized randomly packed
hard spheres and calculated the site-percolation threshold of 0.310 (0.005). The spheres
modeled the lattice sites, the contacts between spheres defined the bonds, and from these
contacts or bonds the connecting paths were defined.
14

Gawlinski and Stanley [1981] studied the connectivity properties of a two-
dimensional continuum system of non-interacting discs of a unit radius randomly
distributed within a square. According to their algorithm, an imaginary covering mesh is
overlaid onto the area. Since the diagonal of the mesh cell equals the disc diameter, two
discs can overlap if their centers lie either in the same cell or within the fourth-neighbor
shell of the cell in question. Using Monte Carlo simulations, they estimated the critical
exponents which were in close agreement with accepted values for ordinary lattice
percolation.
Balberg and Binenbaum [1983] and Balberg et al. [1984] investigated the
percolation thresholds in two and three dimensional anisotropic systems of conducting
sticks.
In two dimensions, they found that the critical length of sticks increases with
increasing anisotropy. Because their inclusions consisted of width-less sticks in two
dimensions, the increase in critical length was equivalent to an increase in the percolation
threshold. They also found that increasing the size of the system decreased the difference
between the longitudinal critical length and the transverse critical length. They developed
the work of Pike and Seager [1974] and computed the percolation threshold for a system
in which there is a preferred fiber orientation and in particular a system in which the
orientation is determined by random alignments within a given interval or in which the
alignments are normally distributed around a given direction. Regarding the sticks
lengths, they have studied systems of equal lengths and of distributed lengths. The
samples used for the study were made of a unit-size square. For the completely random
orientation case, they considered an isotropic state where the critical stick length in y
15

direction coincides with the critical stick length in the x direction, both equal to the value
obtained by Pike and Seager [1974] for the completely random orientation case. They
concluded that the broader the stick-length distribution, the lower the mean of the
distribution needed for the onset of percolation. They did not carry out computations for
larger samples than 1000 sticks.
Using the same algorithm used in two dimensions, Balberg et al. [1984] modeled
a three dimensional system of sticks with random orientations in space. They determined
the critical concentration of sticks, which is required for the onset of percolation across
the cube. Similar to the two dimensional case, there was an isotropic percolation
threshold in uni-axially anisotropic three-dimensional systems.
Bug et al. [1985] used the analytic cluster expansion approach for critical density
developed by Coniglio et al. [2003] and applied it to the various continuum systems of
permeable rods and derived the dependence of the percolation threshold on particle
anisotropy. They showed that for the case of long rods with random isotropic orientation,
the proportionality of percolation threshold to the inverse of the excluded volume
obtained by Balberg et al. [1984] becomes equality in the limit of an infinite length-to-
diameter aspect ratio. Later this was confirmed by Neda et al. [1999]. For other cases of
rods in two dimensions with two possible orientations, two dimensions with arbitrary
angular distribution, and three dimensions with three orthogonal orientations it was
confirmed that critical density is proportional to the inverse of excluded volume

in the
long-rod limit as was already shown [Balberg, 1984]. Their results are restricted to the
rod-like particles.
16

Kortschot and Woodhams [1988] used site percolation technique based on those
proposed by Pike and Segar [1974] to determine the critical area fractions for dispersions
of rectangles of various aspect ratios in two dimensions. The simulation results were used
to predict the effect of aspect ratio and orientation on the critical concentration of
conductive flakes filled in a polymer matrix. Kortschot and Woodhams [1988] were
among the first who emphasized the importance of aspect ratio. Their technique is similar
to that used by Balberg and Binenbaum [1983] and has the disadvantages and limitations
of a lattice model.
Munson [1991] developed a statistical formulation for the probability of
intersections in an anisotropic distribution of cylinders and showed the intersection
probability was dependent on the aspect ratio, and increases with increasing randomness
in the plane. The maximum and minimum probabilities of intersection occur respectively
for a distribution of cylinders that is not quite random for the case of aligned cylinders.
Utilizing this analysis, he estimated the critical volume fraction and concluded that both
cylinder aspect ratio and orientation significantly influence the concentrations at which
the formation of the percolation network happens.
Wang and Ogale [1993] established a three dimensional continuum model in the
form of concentric spheres to predict the threshold volume fractions for infinitely large
systems. Each sphere consists of an impenetrable hard core and a permeable soft shell.
The critical volume fractions for different ratios of shell thickness to diameter were
calculated. In all of their work, the size of all of the spheres was held constant.
Ogale and Wang [1993] used a concentric cylindrical model for predicting
thresholds of infinite systems. The model consists of an impenetrable cylindrical core
17

surrounded by a fully penetrable shell. They investigated the effect of shell thickness on
the threshold volume fractions and the effect of aspect ratio on the percolation threshold
of randomly oriented fibers. They showed that the hard core threshold volume fraction
decreases monotically with increasing aspect ratio. Unfortunately, Ogale and Wangs
simulations were limited to the systems with aspect ratios of the hard core between 12
and 50. This condition puts a practical limitation on their model, making it applicable
only to the short-fiber reinforced composites which cannot be applied to the system of
high aspect ratio fibers.
Dani and Ogale [1997] investigated the percolation process in short-fiber
composites in terms of the clusters formed by fibers of different lengths in a three
dimensional continuum system. The shape of the fibers is characterized by the aspect
ratio, defined as the ratio of the length to the diameter. They concluded that a system with
a constant fiber aspect ratio percolates at a smaller fiber content as compared to one
having distributed fiber lengths, due to the fact that in a system of fibers with different
lengths, the shorter fibers can occupy the empty space between the longer fibers and help
in forming clusters. However, since their study is limited to short-fiber systems, this
result cannot be extended to systems with long fibers.
Clezard et al. [1996] studied the percolation threshold of composites containing
fillers with high aspect ratios and based on the excluded volume approach, proposed a
double inequality between the lower limit for three dimensional randomly oriented
infinitely thin disks (or rods) and the upper limit for spheres. In their studies, however,
the fibers were modeled as width-less sticks which their critical length found to be
dependent on the aspect ratio and orientation distribution.
18

Neda et al. [1999] studied the continuum percolation problem in a system of
isotropically oriented sticks. They modeled sticks as permeable three dimensional capped
cylinders and determined the percolation threshold in different directions. Out of their
simulation results, they obtained a relationship between the critical number of sticks at
percolation and the stick aspect ratio. However these results are valid only for the
isotropic case and in the limit of large aspect ratios.
Consiglio et al. [2003] studied the continuum percolation threshold for a three-
dimensional system of two different-sized soft spheres using Monte Carlo methods. Their
results show that for systems composed of two set of spheres, each set have a given
volume, the percolation threshold is a function of the relative concentration of the objects
and the ratio of the volume of the small sphere to the volume of the large sphere. To
determine the location of objects, they generated a random number between 0 and 1, so
that, if the number was less than the ratio of the number of large objects to the number of
total objects, the object was of type A, otherwise it was of type B. According to their
calculations, the critical volume fraction for a mixture of spheres was considerably bigger
than that for the equal-sized spheres. Their work is limited to a system of only two sizes
of spheres and it cannot simulate systems of varied size of objects.
Ounaies et al. [2003] studied electrical properties of single wall carbon nanotube
reinforced composites through analytical modeling and numerical simulation. Nanotubes
are modeled as high aspect ratio spherocylinders which are placed randomly within a unit
cell. Two rods are intersecting if the minimum distance between their axes was less than
the diameter of the rods. Based on the results of Bug et al. [1985] for isotropic capped
cylinders, Ounaies et al.s model yields rather poor predictions for the volume fraction of
19

capped cylinders at percolation which were significantly larger than what was found in
their experimental attempt. The reason is simple: the Ounaies et al.s model uses a soft-
core approach which allows spherocylinders to overlap and results in increasing the total
volume of space occupied by particles. Considering that the cell volume is constant, the
model happens to predict a higher volume fraction at percolation.
Grujicic and Cao [2004] analyzed the percolation thresholds of single wall carbon
nanotube (SWCNT) reinforced composites using an analytical and a numerical approach.
Since the nanotubes tend to be aligned with each other due to the van der Waals
interactions between them, Grujicic and Cao [2004] calculated the percolation threshold
for both the individual single wall carbon nanotubes randomly distributed as well as for
the bundles of three-, seven-, and nineteen-nanotubes. To simplify the calculations, such
bundles were assumed to be spherocylinders with the same length as that of the
individual nanotubes while their radius is increased. Therefore in a bundle the aspect ratio
decreases with the number of nanotubes. The analytical model is an excluded volume
based approach using the results of Bug et al. [1985] which is valid only in the limit of an
infinite length-to-diameter aspect ratio of the dispersed phase. This means as the aspect
ratio decreases, the analytical approach becomes less reliable. Therefore, Grujicic and
Cao [2004] developed a numerical simulation method for calculating the percolation
thresholds. Using the algorithm proposed by Newman and Ziff [2001], they formulated
the clusters of hard-core touching nanotubes, so that the two nanotubes are supposed not
to penetrate each other. According to their results, the percolation threshold increases
with an increase in the number of nanotubes in the bundle and as a result of the van der
Waals interactions between the nanotubes.
20

Natsuki et al. [2005] investigated the percolation behavior in fiber-reinforced
composites using a two-dimensional model of short sticks for both isotropic and
anisotropic distribution. Their results show the percolation threshold has a linear
logarithmic relationship with the aspect ratio above 40 and has larger dependence on the
orientation angle less than 30. According to their results, the percolation threshold is
predicted more sensitive to the orientation angle of fibers with aspect ratios up to 100.
Besides for larger aspect ratios, the body-to-body contacts establish percolation networks
while for lower aspect ratios, the fiber ends strongly affect the connection of crossbar
networks. For hard-core with soft-shell model of sticks, their study predicts a lower
percolation threshold. Their work is concerned with the percolation threshold of short
fibers in two dimensions.
Foygel et al. [2005] used in their research a theoretical and computational
procedure for studying composites of carbon nanotubes which is expandable to any
conductive particles with large aspect ratio. They modeled nanotubes as capped cylinders
which are randomly oriented inside a cube and showed for a system of randomly oriented
sticks, the critical fractional volume is not an invariant but drops as the aspect ratio
increases. Due to the soft-core approach, their study overestimates the values of the
critical fractional volume associated with the onset of the percolation threshold.
Another analytical model was developed recently by Li and Kim [2007] based on
the average interparticle distance (IPD) approach to predict the percolation threshold of
conducting polymer composite. IPD is defined as the shortest distance between the
neighboring platelets. In their study, graphite was modeled as thin and round platelet.
Two- and three-dimensional random orientations of particles were taken into account in
21

the model. Assuming the particles are homogenously distributed within the matrix and
perfectly bonded with the polymer, they divided the composite into cubic elements, each
containing one particle in the center, and the total number of cubic elements equal to the
total number of particles. The model is purposed to predict the percolation behavior of
spheres and disc-shaped nanofillers with high aspect ratio. From their study, Li and Kim
[2007] found that the aspect ratio of conducting filler is a predominant factor of the
nanocomposite percolation threshold. However, the above cubic-element model with
homogenously distributed platelets suffers from the drawback that it can only be applied
to a system containing filler content below the percolation threshold. The assumption of
homogenous distribution and perfect bonding made in the model is different from the
agglomeration of nanoparticles and presence of micro-flaws or debonding between the
matrix and particles, which are normally observed in many polymer nanocomposites.
Kyrylyuk and Schoot [2008] studied the percolation threshold of polymeric
composite of carbon nanotubes by a continuum model of fully penetrable rod-like
particles based on the theory of the structure of fluids. They found that a length
polydispersity (particles of varied sizes) significantly lowers the percolation threshold at
equal average length, so that small quantities of longer nanotubes can dramatically lower
the percolation threshold. They also used slightly flexible hard rods and showed that
finite bending only increases the percolation threshold by a little. Their findings imply
that the nanotubes need not be necessarily perfectly straight or monodisperse (particles of
a uniform size) to act usefully as conductive fillers.
Berhan and Sastry [2007] proposed a more realistic model by investigating the
percolation threshold of systems of three dimensional straight sphereocylinders which
22

randomly oriented within a unit cube using both soft-core and hard-core with soft-shell
models. Parameter t was defined as the ratio of the radius of the core to the outer radius
of the soft shell so that for soft-core model t=0 and for the hard-core model t=1. Based on
the excluded volume concept, Monte Carlo simulations were performed for values of t
equal to 0, 0.2, 0.3, 0.4, 0.5, 0.6 and 1. It was shown that for the hard-core model the
percolation threshold is strongly dependent on the parameter t rather than on the aspect
ratio. Their research emphasizes that the soft core model is more convenient to use due to
its only dependence on the geometry of the fibers; however, the model erroneously
supposes the fibers can merge, which can be significant even for very high aspect ratio
fibers in reinforced composites. In systems where the conducting particles are not in
physical contact and tunneling is the dominant change transport mechanism, the hard
core model is more suitable for modeling electrical percolation provided fiber geometry
and distribution and the tunneling distance are known. If the tunneling distance is five
nanometers or less, both soft-core and hard-core models yield similar results.
Determining the tunneling distance experimentally is difficult. The suggested hard-core
modeling technique can be used in order to determine the likely tunneling distance for the
given fiber geometry, distribution and the experimentally derived percolation threshold.
This can be achieved by finding the value of the parameter t which results in a match
between the values of experimental percolation threshold obtained and the analytical
solution.
Furthermore Berhan and Sastry [2007] investigated the effect of the fibers
waviness on percolation threshold. Five different fiber geometries, all with hemisphere
caps on each end were considered in this study. They concluded that for aspect ratios less
23

than 1000, the effect of fiber waviness on percolation threshold is dependent on both
aspect ratio and parameter t, so that, it increases with a reduction of the aspect ratio and
an increase of parameter t. This result agrees with the previous results for straight fibers
and with the findings of Balberg and Binenbaum [1983]. In summary, Berhan and Sastry
[2007] found that using both soft-core and hard-core approaches the percolation threshold
of systems comprised of high aspect ratio fibers is inversely proportional to the excluded
volume of the fiber, independent of fiber waviness. The excluded volume of an object
with hard-core and soft-shell is equal to the excluded volume of the shell minus that of
the core due to the fact that the interior core is impenetrable. These results are applicable
to fiber reinforced composites. Their results were based on Monte Carlo simulations in
large but finite-sized systems and the finite size scaling concepts were not invoked to
eliminate the finite size effects. Therefore, their results are not exact for infinitely large
systems.
Table 2.1 summarizes key percolation studies.









24

Table 2.1 Key percolation studies in two- and three-dimensional of continuum models

Authors Model Contributions Limitations
Balberg and
Binenbaum
[1983]
2D anisotropic
sticks
They found PT increases
with increasing anisotropy.
They did not carry out
computations for samples
larger than 1000 sticks
Balberg et al.
[1984]
Soft-core capped
cylinders
They introduced first
continuum percolation
model in 3D.
Their results restricted only
to fixed-size systems and
were not applicable to
infinitely large systems
Bug et al. [1985] Permeable rods They showed for long
isotropic randomly oriented
rods, the proportionality in
PT turns to equality.
Their results were limited
to rod-like particles.
Wang and Ogale
[1993]
Hard-core with
soft-shell spheres
They showed PT reduced for
non-uniform dispersion of
particles.
Diameter of all fibers was
constant.
Ogale and Wang
[1993]
Hard-core with
soft-shell
cylinders
They showed hard-core
threshold volume fraction
decreases with increasing
aspect ratio.
The model was limited to
short-fibers, and not
applicable to fibers with
aspect ratios higher than
25.
Dani and Ogale
[1997]
Hard-core with
soft-shell
cylinders
They showed system of
constant aspect ratio
percolates at smaller fiber
content in comparison to
variable fiber aspect ratio.
The model was limited to
short fibers.
Ounaies et al.
[2003]
Soft-core long
spherocylinders
They found critical volume
fraction.
Their result was higher
than what their
experiments showed, due
to the soft-core approach.
Also the cell size was kept
constant
Berhan and
Sastry [2007]
3D hard-core
with soft-shell
spherocylinders,
They showed PT for hard-
core model depends on
t=radius of core/ outer radius
of shell rather than aspect
ratio.
The size of the system was
kept constant.

2.8 Conclusion
From the literature review discussed above, it is realized that lattice models are
less flexible than continuum models for modeling the microstructure of composites, and
soft-core models do not adequately reflect the real inclusions.
25

Despite the inconsistencies between the soft-core model and real nanocomposites,
the model employed in the majority of existing studies is soft-core, which assumes the
conducting particles join together so that geometric percolation and electrical percolation
are coincident. However, it has been established in previous studies that nanofibers
embedded in a polymer matrix are not connected physically and the electron charge is
transported over the tunneling distance [Benoit et. al., 2002; Balberg, 2009; Du et. al.,
2004]. Using soft-core in modeling such systems significantly affects calculations,
yielding percolation thresholds far from the experimental values as seen in the literature.
Some studies have applied the hard-core with soft-shell approach for simulating
the percolation phenomenon in nanocomposites. However, in their model, the fiber aspect
ratio was held constant for all the fibers in a given sample. In a real system, fiber
breakage during processing is so common that a fiber size distribution is always
observed. Other researchers did not account for the spatial variation in the fiber
orientation distribution that is also observed in real composite systems.
Therefore the lack of a model that accounts for the fiber size as well as orientation
distribution is felt in the literature. Such a model can closely simulate the microstructural
features of actual composite samples.
The remainder of this thesis shall be confined to an approach for studying the
behavior of nanocomposites and in particular carbon nanofiber reinforced composites
through experimental and analytical work.


26


Chapter 3
Modeling and Simulation Method
3.1 Introduction
During recent decades, a number of approximate theories have been proposed to
describe the phase behavior of disordered systems. The main concern is whether the
answers that such approximate methods give are reliable or not. Development of
powerful computer models empowers researchers to interpret their experimental
observations, and predict the properties of arbitrary systems [Grosberg, 1998].
This chapter focuses on an approach to the percolation problem using a new
method. A number of points were randomly distributed in a unit cell. A geometrical
region was centered on each point. The regions were circles distributed inside a square
cell in two dimensions, and either spheres or spherocylinders distributed throughout a
cube volume in three dimensions. The clusters were formed by the overlapping of the
geometrical regions that are randomly distributed in space.
Simulations were performed for both the soft-core model and the hard-core with
soft-shell model. All the programming was performed using MATLAB 7.6.0 due to its
27

wide use and the clarity of the code. A complete version of the programming
implementation for the percolation problem is provided in the appendix.

3.2 Methods and Approach
Since percolation is a statistical process, percolation problems benefit from Monte
Carlo techniques. Obviously computer programming plays a crucial role. The systems
done through the numerical computer simulations are always of finite sizes, so the
analyses and data extrapolation rely on the finite size scaling theory. The basic idea is to
create a set of randomly distributed particles with a rather regular geometric form by a
pseudo-random number generator inside a unit cell. The program then tests for pairs of
particles that overlap and determines if a continuous path between opposite faces of the
cell exists. If there is a path, the system is said to percolate. The program then alters the
size of the cell and repeats the test for percolation.

3.2.1 Monte Carlo Methods
The term "Monte Carlo methods" describes not a single method but a class of
approaches; all share their reliance on repeated computation of pseudo-random numbers
[Hammersley and Handscomb, 1975]. It was named in the 1940s by physicists working
on nuclear weapon projects in the Los Alamos National Laboratory [Metropolis, 1987].
Opposed to deterministic methods, Monte Carlo algorithms are statistics that are
nondeterministic in manner, because of using random numbers for generating suitable
random solutions. These solutions obey some property or properties. Due to the repetition
28

of computational algorithms, a large number of calculations is involved which makes
utilizing a high-speed computer a necessity for Monte Carlo Methods.
Monte Carlo techniques have been widely used in many areas of physics and
mathematics. Generally, where the problem is too complicated to be solved through
analytical methods, Monte Carlo simulations are used to obtain numerical solutions.
Systems with a large number of coupled degrees of freedom, multidimensional integrals
with complicated boundary conditions, disordered materials, cellular structures, and any
phenomena with significant uncertainty in inputs, are among common areas that broadly
benefit from using Monte Carlo techniques. A classic example is diverse applications of
Monte Carlo methods in percolation problems [Jain, 1992].

3.2.2 Excluded Area and Excluded Volume
In order to estimate the percolation thresholds, several dimensional invariants
have been suggested. The most often referred to among them are the number of bonds per
particle [Balberg and Binenbaum, 1987], and the total excluded area or volume [Pike and
Seager, 1974; Balberg et al., 1984]. These quantities do not offer an accurate value of
threshold for arbitrary continuum systems; however, the analytical approach which
approximates the total excluded area or volume works satisfactorily well for two or three
dimensional random shaped particles. In this approach, the material is assumes to be
statically homogeneous [Weber and Kamal, 1997].
The excluded area or volume of an object is defined as the area or volume around
an object into which the center of another similar object is not allowed to penetrate if
overlapping of the two objects is to be avoided [Onsager and Ann, 1949]. In other words,
29

excluded area or volume around a figure is bounded by a perimeter consisting of all
possible loci of the center of a second figure such that the two figures just make contact.
For a circle (sphere) of diameter D, the excluded area (volume) is simply a circle
(sphere) of diameter 2D (Fig. 3.1).






Figure 3.1 Excluded area (volume) of a circle (sphere)

It is also worthwhile to note that the excluded volume is not always similar in
shape to the actual object. Therefore, while for circles and spheres the excluded area or
volume is just the objects area or volume multiplied by a constant, the excluded area or
volume in general is not proportional to the area or volume of the object [Bug et al.,
1985]. As the complexity of the objects geometry rises, the difference between the true
volume and excluded volume increases.
Of importance here is the fact that the excluded area (volume) rather than the
object area (volume) determines the percolation threshold [Balberg and Binenbaum,
1983]. According to the excluded volume theory, the density of objects at percolation is
inversely proportional to the excluded area (volume) of an object [Sun et al., 2009]. The
approach based on the excluded volume allows evaluating the percolation threshold of
D
.
.

30

continuum systems. If
c
P is the critical density of objects in the system and
ex
V is the
associated excluded volume, then the invariant
c ex
P V is called the total excluded volume
and is linked to the percolation threshold.
For a statistical distribution of objects, an average excluded area
ex
A can be
defined such that the total excluded area is given by [Celzard, 1996]:
c ex e
P A A = (3-1)
and similarly for objects with the average excluded volume
ex
V , the total excluded
volume is given by:
c ex e
P V V = (3-2)
Incidentally, for a system of randomly distributed circular objects, the total
excluded area is:
A P A
c e
4 = (3-3)
where A is the given area of a circle. In a similar way, for spheres of volume V , the
total excluded volume of the system is:
V P V
c e
8 = (3-4)

3.3 Circles and Spheres Simulation
The percolation for continuum models of uniform and variable radius
distributions of circles and spheres were investigated and compared with the literature
data. In the first step, all circles or spheres were of one and the same size R. Two
different models were simulated:

31

3.3.1 Soft-Core Modeling
First a network with penetrable objects corresponding to the traditional soft-core
model was simulated.
The simulations were carried out using the following procedure: Circles were
added one at a time randomly into a unit square. The radius of the circles was measured
in the units of the unit square size. As circles were added, the code determines which, if
any, of the existing circles are intersected by the new circle. The number of contacts is a
sensitive function of the circles density.
The connectivity is determined by the minimum distance between the centers of
the two circles from the equation:
2 2
) ( ) (
j i j i ij
y y x x d + = (3-6)
and the results are recorded in a matrix
n n
M

where n is the number of circles. If such
distance for ith and jth circles is equal to or smaller than the total radii of the circles, the
two circles are considered to be in contact and ) , ( j i m and ) , ( i j m are assigned a value of
1, otherwise ) , ( j i m and ) , ( i j m are assigned a value of 0. The value of ) , ( i i m by
convention is always set to 1. Periodic boundary conditions are applied over the sides of
the unit cell as shown in Figure 3.2. For any circle found to be partially outside of the
cell, the computer program automatically creates similar circle(s) on the opposite side(s)
or corners, so that the parts of the circles inside the box form a complete circle. As the
contacting circles form clusters, an algorithm is activated to check if any of the clusters
extends over the entire square in a given direction of the simulation cell. Once such a
cluster is verified, it means the system percolates. So, the simulation is terminated and the
32


Figure 3.2 Schematic presentation of boundary condition

number of circles is saved. If the percolating cluster is not detected, new circles are added
until percolation is achieved.
One thousand independent runs were performed to calculate values for the
average and standard deviation of the number of circles needed to make the system
percolate. Since the numerical calculations of the percolation threshold are carried out
under the periodic boundary conditions, the size of the unit cell used in the computation
can in general have an effect on the computational results. To assess the potential effect
of the size of the unit cell on the computed percolation threshold, for a given radius of
circles the procedure of numerical simulations of the percolation of objects is repeated for
different values of the unit cell size in a sequence of increasing system size raging from
W=1 to W=2 by 0.1. The increments are equivalent to the diminished size of the radii of
the circles, in order to determine a potential dependence of the computed percolation
threshold on the unit-cell size.
The most time-consuming step is finding the percolating cluster, because a scan
of the whole system is required whenever the two clusters merge. This is especially
W
W
W+2R
W+2R
33

expensive for large cell sizes, as the total number of clusters is proportional to the volume
of the system. The construction of the matrix
n n
M

requires a one-to-one mapping of
allowed connectivities to create integers.
A graphic simulation code, implemented via MATLAB, is used to generate
typical random fiber networks simulation diagrams. Figure 3.3 displays as particle
number increases, clusters of individual particles are formed; forcing percolation to take
place through the system.

(a) Equal size circles


(b) Variable size circles

(1) Individual particles (2) Clusters (3) Percolation
Figure 3.3 Graphical examples of non-percolating and percolating network of soft circles in
2D cells (a)Equal size circles, (b) Variable size circles

Based on the same idea, the simulation was performed for spheres added one at a
time into a cubic box until a percolating cluster could be identified which connected the
34

two opposite faces of the box. At such a time, the box was said to percolate. The
connectivity was determined by the distance d
ij
between the centers of ith and jth spheres
from the equation:
2 2 2
) ( ) ( ) (
j i j i j i ij
z z y y x x d + + = (3-7)
The system size took 10 values in the range of 1 to 2. Figure 3.4 displays the
formation of a sample percolating cluster for soft-core spheres of equal and variable size.

(a) Equal size spheres


(b) Variable size spheres

(1) Individual particles (2) Clusters (3) Percolation
Figure 3.4 Graphical examples of non-percolating and percolating network of soft spheres in
3D cells (a)Equal size spheres, (b)Variable size spheres


35

3.3.2 Hard-Core with Soft-Shell Modeling
Polymer nanocomposites can be considered composed of particles, each of which
is made of a hard core and a soft shell. The inner core represents the actual fiber and is
impenetrable, while the outer soft-shell is the penetrable portion of the particle which
represents the critical tunneling distance of two neighboring particles within which the
two conductive fibers are considered to be in electrical contact. Therefore, the two fibers
are regarded as connected if the soft shells penetrate each other.
Consider a system of randomly dispersed circles (spheres) with a hard core radius
r and a thickness of the soft shell R-r (Fig. 3.5).


Figure 3.5 Hard-core with soft-shell model
of a circle or sphere

Since the hard cores are impenetrable to each other, the location of the circle
(sphere) generated last will be rejected when two hard cores penetrate each other.
Another location is generated for the particle and tested for connectivity. Two objects are
connected if the distance between their centers, which is not allowed to be less than 2r, is
less than 2R. Figure 3.6 illustrates the process of cluster formation for both equal and
variable sizes of hard-core with soft-shell circles. These graphics are produced by graphic
simulation tools of the MATLAB program.
36

(a) Equal size circles


(b) Variable size circles


(1) Individual particles (2) Clusters (3) Percolation
Figure 3.6 Graphical examples of non-percolating and percolating network of hard-core-soft-
shell circles in 2D cells. The large circles contain smaller circles at their centers.
(a)Equal size circles, (b)Variable size circles


Similarly, in Figure 3.7 the procedure of percolation is visualized for a system of
spheres with hard-core and soft-shell in both the cases of equal size and variable size
spheres.




37

(a) Equal size spheres


(b) Variable size spheres

(1) Individual particles (2) Clusters (3) Percolation
Figure 3.7 Graphical examples of non-percolating and percolating network of hard-core-soft-
shell spheres in 3D cells. The large spheres contain smaller spheres at their centers.
(a)Equal size spheres, (b)Variable size spheres


3.4 Spherocylinder Simulation
A similar method is used to simulate the percolation threshold in a system of
spherocylinders, and to provide the role of the percolation threshold in hopping
conduction.
38

3.4.1 Geometry and Modeling
The spherocylinder is taken to mean a right circular cylinder with its ends closed
to form two hemispherical surfaces (Fig. 3.8). In a spherocylinder, the principal
geometric parameter is shown to be the dimensionless ratio of length of the cylindrical
part L to diameter D which is called aspect ratio.

Figure 3.8 Schematic representation of spherocylinder model


The centers of mass of the spherocylinders are given random coordinates inside
the cube.
The spherocylinder orientation is specified in terms of two angles; one describes
the orientation state with respect to an axial orientation and the other with respect to a
plannar orientation. Figure 3.9 shows the azimuthal angle and polar angle which are
defined in the limits of the intervals ) * 2 , 0 ( rand and ) * 2 1 arccos(
2
rand

in
MATLAB programming code, respectively. This geometry is found to be appropriate
for modeling the nanofibers observed in composites.
In all the simulations, the spherocylinders remain in the isotropic phase, meaning
that there is no order in either the positions of centers of mass or the alignments of the
spherocylinders.
L
D
39


Figure 3.9 Nomenclature describing orientation of a single spherocylinder


3.4.2 Intersection of Spherocylinders
The probability of forming a network of intersecting spherocylinders (capped
cylinders or sticks) in various dimensions was estimated. First, taking into account the
orientation distribution and the spherocylinder geometry, the probability of two
spherocylinders intersecting each other were investigated. This probability was used to
estimate the probability of multiple intersections forming clusters.
The intersection of two distinct spherocylinders with finite length is generally
categorized in one of the following three possible patterns (Fig. 3.10):
(a) side-to-side
(b) side-to-end
(c) end-to-end




X
Z
Y
40






(a) (b) (c)
Figure 3.10 Schematic representation of three different patterns of contact between
spherocylinders: (a) side-to-side, (b) side-to-end, (c) end-to-end

In the following, it is shown how we can determine if the two spherocylinders
overlap by any of the above mentioned patterns. The following notations are used:
1
R
and
2
R denote the radii of the spherocylinders,
1
P and
2
P specify the centers of the two
hemispheres of one spherocylinder, while
3
P and
4
P denote the centers of the
hemispheres of the other spherocylinder.
(a) side-to-side intersection: The intersection of two spherocylinders corresponds
to the intersection of the axes of their cylinders (Fig. 3.11). Recall the general method of
finding the shortest distance between two lines [Frost, 1886]. Let line
1
L pass through
points
1
P and
2
P :
1 2
1
1 2
1
1 2
1
P P
P
P P
P
P P
P
z z
z z
y y
y y
x x
x x

(3-8)
or
1 1 1
1 1 1
f
z z
e
y y
d
x x
P P P

=

(3-9)
and line
2
L pass through points
3
P and
4
P :
41

3 4
3
3 4
3
3 4
3
P P
P
P P
P
P P
P
z z
z z
y y
y y
x x
x x

(3-10)
or

2 2 2
3 3 3
f
z z
e
y y
d
x x
P P P

=

(3-11)


Figure 3.11 The side-to-side intersecting of two cylinders

The common perpendicular vector is obtained from the vector product:
k e d e d j f d f d i f e f e
f e d
f e d
k j i
V ) ( ) ( ) (
1 2 2 1 1 2 2 1 1 2 2 1
2 2 2
1 1 1 1
+ = =

(3-12)
which can be written in the form of a unit vector:
2
1 2 2 1
2
1 2 2 1
2
1 2 2 1
1 2 2 1 1 2 2 1 1 2 2 1
1
) ( ) ( ) (
)] ( ), ( ), [(
e d e d f d f d f e f e
e d e d f d f d f e f e
v
+ +

=

(3-13)
The vector connecting the given point
1
P on line
1
L with the given point
3
P on
line
2
L is:
)] ( ), ( ), [(
1 3 1 3 1 3
2 P P P P P P
z z y y x x V =

(3-14)
P
1
P
3
P
4
P
2
d

D
2
D
1
L
1
L
2
42

It is straightforward to observe the scalar product of this vector with the
perpendicular unit vector
1
v

is equal to the shortest distance between the two lines:


2
1 2 2 1
2
1 2 2 1
2
1 2 2 1
1 2 2 1 1 2 2 1 1 2 2 1
1 2
) ( ) ( ) (
) )( ( ) )( ( ) )( (
.
1 3 1 3 1 3
e d e d f d f d f e f e
e d e d z z f d f d y y f e f e x x
v V d
P P P P P P
+ +
+
= =


(3-15)
In order to find out the points of closest approach, we write the vector equation
for lines
1
L and
2
L . For line
1
L :

+ = =

+ = =

+ = =

) (
) (
) (
1 2 1
1 2
1
1 2 1
1 2
1
1 2 1
1 2
1
1 1
1 1
1 1
P P P
P P
P
P P P
P P
P
P P P
P P
P
z z s z z s
z z
z z
y y s y y s
y y
y y
x x s x x s
x x
x x
(3-16)
The same relations hold for line
2
L :

+ = =

+ = =

+ = =

) (
) (
) (
3 4 3
3 4
3
3 4 3
3 4
3
3 4 3
3 4
3
2 2
2 2
2 2
P P P
P P
P
P P P
P P
P
P P P
P P
P
z z s z z s
z z
z z
y y s y y s
y y
y y
x x s x x s
x x
x x
(3-17)
The line joining a general point on line
1
L to a general point on line
2
L has the
vector:

+ +
+ +
+ +
=
)] ( [ )] ( [
)] ( [ )] ( [
)] ( [ )] ( [
1 2 1 3 4 3
1 2 1 3 4 3
1 2 1 3 4 3
1 2
1 2
1 2
3
P P P P P P
P P P P P P
P P P P P P
z z s z z z s z
y y s y y y s y
x x s x x x s x
V

(3-18)
43

Since
1 3
V V

it is easy to see that:
1 2 2 1
1 2
1 2 2 1
1 2
1 2 2 1
1 2
)] ( [ )] ( [
) (
)] ( [ )] ( [
)] ( [ )] ( [
1 2 1 3 4 3
1 2 1 3 4 3
1 2 1 3 4 3
e d e d
z z s z z z s z
f d f d
y y s y y y s y
f e f e
x x s x x x s x
P P P P P P
P P P P P P
P P P P P P

+ +
=

+ +
=

+ +
(3-19)
From these equations,
1
s and
2
s are found. The points of closest approach,
1
D
and
2
D , are therefore:

+ =
+ =
+ =
) (
) (
) (
1 2 1 1
1 2 1 1
1 2 1 1
1
1
1
P P P D
P P P D
P P P D
z z s z z
y y s y y
x x s x x
(3-20)

+ =
+ =
+ =
) (
) (
) (
3 4 3 2
3 4 3 2
3 4 3 2
2
2
2
P P P D
P P P D
P P P D
z z s z z
y y s y y
x x s x x
(3-21)
The key observation is that the two spherocylinders cross over each other by their
sides if the distance between their centric axes is less than
2 1
R R + and both of the points
of closest approach lie on the straight line through on their axes.
(b) side-to-end intersection: The side-to-end intersection of the two
spherocylinders means that the hemispheric part of one object intersects with the
cylindrical part of the other object. Define h as the perpendicular distance between the
center of end hemisphere of one spherocylinder and the axis of the other spherocylinder,
and H as the intersection point of the perpendicular line and the axis (Fig. 3.12).


44










Figure 3.12 The end-to-side intersecting of two spherocylinders


The points
1
P and
3
P are connected by a straight line and form a right triangle.
According to Pythagorean Theorem:
2
1
2
3 1
2
3
H P P P H P = (3-22)
3 1
P P is a vector passing through known points of
1
P and
3
P and its magnitude is
easy to compute:
2 2 2
2
3 1
) ( ) ( ) (
1 3 1 3 1 3
p p p p p p
z z y y x x P P + + = (3-23)
Observe H P
1
is the projection of
3 1
P P on
2 1
P P and is given by the expression
[Corwin and Szczarba, 1994]:

2 1
2
2 1
2 1 3 1
1
P P
P P
P P P P
H P
|
|

\
|

= (3-24)






H

P
4
h
P
3
P
2
R
2
R
1
P
1
45

Knowing
3 1
P P and H P
1
, we can obtain H P
3
. Consequently the distance h is:
2 2 2
3
) ( ) ( ) (
3 3 3
p H p H p H
z z y y x x H P h + + = = (3-25)
The inequality equation
2 1
R R h + together with the condition that H lies
within
2 1
P P are necessary and sufficient for intersecting the end-side intersection of the
two spherocylinders.
(c) end-to-end intersection: The intersection of the two spherocylinders where
they connect together by their hemisphere ends is related to the intersection of two
spheres (Fig. 3.13).









Figure 3.13 The end-to-end intersecting of two spherocylinders


If the distance between the center of each of hemisphere of one spherocylinder
and the center of each of hemisphere of the other spherocylinder is less than
2 1
R R + , it
means the two hemispheres intersect, and therefore the two sphercylinders intersect.
P
3
1
P
46

By the above, the problem of the determination of how the two spherocylinders
connect is completely solved.

3.4.3 Excluded Volume of a Spherocylinder
For an elongated object, the excluded volume depends on the relative orientation
of the object. Balberg et al. [1984] calculated the excluded area and volume for various
sticks. In three dimensions, the stick is a capped cylinder and the excluded volume is
obtained by rotating it around the centric axis (Fig. 3.14) which can be formulated as:
] ) (
8
3
) (
4
3
1 [
3
32
2 3

+ +

= Sin
R
L
R
L
R V
ex
(3-26)
where is the angle between two sticks and Sin is the average value of Sin . For the
random isotropic distribution
4

= , 1 = Sin ; hence, the excluded volume reduces


therefore to:
] ) (
32
3
) (
4
3
1 [
3
32
2 3
R
L
R
L
R V
ex
+ +

= (3-27)


Figure 3.14 Excluded volume of a spherocylinder of length L and radius R



47

3.4.4 Soft-Core Modeling
The exclusion procedure and the determination of the threshold are similar to
those used for spheres. The code works in the following manner. A certain number of
capped cylinders or spherocylinders of a given aspect ratio (a given length L and diameter
D) are randomly planted in a unit cube. The configurations obtained are non-equilibrium
ones, hence, the dispersions were relaxed via Monte Carlo runs, where for each
spherocylinder, a random displacement of its center and a random rotation of its axes
were attempted. As spherocylinders are added the code determines which, if any, of the
existing spherocylinders are intersected by the new spherocylinder, and whether the new
spherocylinder crosses the boundary of the unit cube. Finally it verifies if a percolation
cluster has been established, which connects opposite sides of the cube. If this cluster is
not detected, new spherocylinders are added until percolation is achieved. The simulation
is repeated up to a thousand times in order to find the value of the critical concentration.
The above computer simulations are carried out at successively increased size of the
system in small increments for each given aspect ratio of spherocylinders. Since
orientation effects were being considered, percolation was tested for only in the Z
direction.
Figure 3.15 shows a typical random fiber network for the system of freely
interpenetrable rods.

48


Figure 3.15 A typical random fiber network generated by the algorithms used in the
present work for modeling soft spherocylinders of uniform aspect ratio.


3.4.5 Hard-Core with Soft-Shell Modeling
The electrical conductive fibers can be modeled as having an assumed penetrable
soft-shell as shown in Figure 3.16. The thickness
2
d D
of penetrable shell indicates the
connecting range across which the electron charge transfer between fibers can take place,
which obviously depends on the thickness and material properties of the insulating
polymer film.


Figure 3.16 Schematic representation of the hard-core with soft-shell model of spherocylinder.
The inner portion represents the hard-core whereas the outer portion represents
the penetrable soft shell.

L
D d
49

We generate dispersions of hard-core with penetrable soft-shell spherocylinders
by using an extended version of a previously described algorithm which allows adding
spherocylinders into a cubic cell with periodic boundary conditions through random
sequential addition. The results will be presented in the next chapter.

3.5 Conclusion
We conclude this section by mentioning that the numerical modeling approach
presented here is aimed to simulate the electrical percolation thresholds in carbon
nanofiber-based composites. Monte Carlo methods have been widely used in the
percolation analysis of both soft-core and hard-core modeling. The simulations have been
conducted for systems of similar objects as well as objects of distributed sizes.
In the following chapter, the results of the simulations are presented and
discussed.













50


Chapter 4
Results and Discussion
4.1 Introduction
In the previous chapter we described the numerical modeling procedure used to
generate particle dispersions. This chapter presents the computer simulation results,
directed at determining the percolation thresholds of clusters formed by objects of
different shapes in two- and three- dimensional continuum systems based on the excluded
area and volume approach. Since real composites always have a fiber length distribution,
all results from the present simulations are reported for uniform as well as distributed
particle sizes.

4.2 Circles Percolation
The first system is the model of soft circles. The simple case where all the circles
are of the same size was simulated initially. From a large number of computer
experiments, the data are derived and displayed as individual marks for different sizes of
the system and fitted to a linear function y(x). Figure 4.1 displays the results for systems
of similar permeable circles.
51

Radius = 0.1
y = 0.3121x + 34.672
R
2
= 0.9725
35
36
37
38
39
40
0 5 10 15 20
S.D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

Radius = 0.15
y = 0.9648x + 15.552
R
2
= 0.9893
10
12
14
16
18
20
22
24
26
0 2 4 6 8 10
S.D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

Radius = 0.3
y = 0.3243x + 3.8391
R
2
= 0.983
3
3.2
3.4
3.6
3.8
4
4.2
4.4
0 0.5 1 1.5 2
S.D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

Figure 4.1 Variation of average density of uniform soft circles per square area vs standard
deviation for three different radii R=0.1, R=0.15, R=0.3..
52

The graphs show the average number of circles per square area at percolation
versus the standard deviation while the cell size is increasing. The simulations were
performed for three radius sizes: R=0.1, R=0.15, R=0.3. Each of the blue marks is the
output data of the programming code, associated with the one given size of the square.
The straight lines fitted to the data confirm they obey a linear trend. It is expected from
the finite size scaling theory that as the system size keeps increasing, the standard
deviation becomes smaller. When the size of the system goes to infinite (W ), the
standard deviation goes to zero, and at standard deviation of zero, the critical densities of
circles needed for the occurrence of percolation are obtained from the fit equations.
The results shown in this figure indicate that the unit cell size (within the range
examined) has a very weak effect on the computed percolation threshold.
Next, we expand the computer program for simulating the systems of circles with
variable radius sizes. Here in order to have a better estimation of density of the particles,
instead of the total number of circles, the total area of objects was calculated each time
when percolation occurred. Figure 4.2 shows how the percolation changes with the
different radii ranges.
53

0.13 < Radius < 0.17
y = 1.2949x + 14.855
R
2
= 0.9918
20
21
22
23
24
25
26
27
28
29
0 2 4 6 8 10 12
S. D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

0.12 < Radius < 0.18
y = 1.3232x + 14.736
R
2
= 0.9857
20
21
22
23
24
25
26
27
28
29
0 2 4 6 8 10 12
S. D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

0.11 < Radius < 0.19
y = 1.4663x + 13.472
R
2
= 0.979
20
21
22
23
24
25
26
27
28
29
30
0 2 4 6 8 10 12
S. D.
D
e
n
s
i
t
y

o
f

c
i
r
c
l
e
s

Figure 4.2 Variation of average density of soft circles vs standard deviation. Solid lines are fits.
54

We now need to recall the definition of the excluded area of a circle from Figure
3.1. The excluded area for a circle of radius R is:
2
) 2 ( R A
ex
= (4-1)
and the total excluded area for the system of circles is:
ex c e
A P A = (4-2)
where P
c
is the number of the circles at percolation.
The number of circles at percolation is deduced from the fit equations in Figure
4.2. The percolation threshold is then determined from the total excluded area of all the
circles. The results are summarized in Table 4.1. This helps to predict the percolation
threshold for three different ranges of radius.

Table 4.1
Total excluded area of systems of soft
circles with different sizes

Radius P
c
A
e

0.1 34.672 4.3570
0.15 15.552 4.3972
0.3 3.8391 4.3419
0.13-0.17 14.855 4.2001
0.12-0.18 14.736 4.1665
0.11-0.19 13.472 3.8091

Table 4.1 clearly exhibits that the total excluded volume of system of circles is
invariant with respect to the radius which agrees with the previous studies [Balberg,
1984]. Widening the range of radius size, on the other hand, has reduced the total
excluded area of the system and in consequence the percolation threshold is lowered.
This is comparable to the earlier results by Pike and Seager [1974] and Balberg and
55

coworkers [1984] who also concluded that a distribution in size of the circles results in a
reduction of the critical area fraction. However, the excluded area of the system of soft-
core circles of equal size was found to be 4.3, a slightly lower than the value of 4.48 and
4.5 which has been suggested in the literature for soft circles [Pike and Seager,1974;
Balberg et al., 1984]. Because of the considerably larger systems and the finite size
scaling procedure applied in the present study, the estimate obtained here for the total
excluded area of soft circles, is expected to reach the value with higher accuracy.
The second system is the hard-core with soft-shell circles which consists of
concentric circles. The simulation was done for different values of ratio of shell radius to
core radius, r R , by keeping the radius of the outer circle or shell constant (R=0.15)
while the radius of the inner circle or core r was increased, so that r R decreased from 2
to 1.3 (Figure 4.3). In order to obtain the critical density of objects, each time the
equation of the fitted linear trend line was solved for standard deviation of zero, and the
answer was multiplied by the excluded area of the core. The reason of considering the
core, and not the shell, is that the core represents the actual fiber size.








56

R / r = 2
y = 2.0348x + 9.2253
R
2
= 0.9829
14
15
16
17
18
19
0 1 2 3 4 5
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r =1.9
y = 1.8936x + 9.7555
R
2
= 0.9485
14
15
16
17
18
19
0 1 2 3 4 5
S.D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.8
y = 1.9362x + 9.5601
R
2
= 0.9904
14
15
16
17
18
0 1 2 3 4 5
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.7
y = 2.0203x + 9.4751
R
2
= 0.9875
14
15
16
17
18
0 1 2 3 4
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.6
y = 2.1132x + 9.2643
R
2
= 0.9739
13
14
15
16
17
0 1 2 3 4
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.5
y = 1.7559x + 9.9671
R
2
= 0.9827
13
14
15
16
17
0 1 2 3 4
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.4
y = 1.5941x + 9.8252
R
2
= 0.9628
11
12
13
14
15
16
0 0.5 1 1.5 2 2.5 3
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

R / r = 1.3
y = 1.4618x + 9.2587
R
2
= 0.9628
11
12
13
14
15
16
0 0.5 1 1.5 2 2.5 3
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

c
i
r
c
l
e
s

Figure 4.3 Variation of average density of hard-core with soft-shell circles vs standard deviation
57

Figure 4.4 presents how the total excluded area, associated with the core and shell
of the circles, varies with the ratio of core excluded area to shell excluded area.
y = -9.1556x
4
+ 14.805x
3
- 8.9034x
2
+ 4.4312x - 0.0046
R
2
= 0.9987
0
1
2
3
4
5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Aex Core / Aex Shell
P
c

.

A
e
x
C
o
r
e

y = -9.1764x
3
+ 14.388x
2
- 8.3925x + 4.2867
R
2
= 0.9908
0
1
2
3
4
5
0 0.2 0.4 0.6 0.8 1
Aex Core / Aex Shell
P
c

.

A
e
x
S
h
e
l
l

Figure 4.4 Dependence of the total excluded area of the system of hard-core with soft-
shell circles on the ratio of core excluded area to shell excluded area. The
solid curve represents the apparent polynomial fit.

From the fitting equation of the second graph in Figure 4.4, the total excluded
area of the system can be expressed as a function of the ratio of core excluded area to
shell excluded area:
58

d )
A
A
c( )
A
A
b( )
A
A
( a A . P
shell ex
core ex 2
shell ex
core ex 3
shell ex
core ex
shell ex c
+ + + = (4-3)
By multiplying the two sides of the above equation by
shell ex
core ex
A
A
we have:
d
A
A
)
A
A
c( )
A
A
b( )
A
A
( a A . P
shell ex
core ex 2
shell ex
core ex 3
shell ex
core ex 4
shell ex
core ex
core ex c
+ + + = (4-4)
If the ratio
shell ex core ex
A A is equal to zero, it means the object doesnt include any
core and the model transforms to the soft-core model of circles (soft-core transition). So
that Equation (4-4) becomes zero and Equation (4-3) results in the value similar to that of
the density of soft circles which is equal to 4.29. If the ratio
shell ex core ex
A A is equal to
one, it means the radius of the core has increased so that it occupies all of the shell and
the model transforms to the hard-core model of circles (soft-core-hard-core transition).
Therefore, the two fitting equations should give the similar values for the density of
circles. It is noted that the shape of the two curves in Figure 4.4, however, is not expected
to match. The value obtained for the total excluded area of hard circles is 1.17 associated
with the first graph and 1.11 from the second graph which again is lower than the value
of 2.20.4, reported by Balberg et al. [1984]. This difference can be attributed to using
finite size scaling theory in the present model which was not applied in Balberg et al.s
model.

4.3 Spheres Percolation
The general features of spheres percolation modeling are the same as in the case
of circles. The results for soft-core spheres are presented in Figure 4.5 for three different
radii. The solid lines were fitted to data in the same way as in two dimensions.
59

Radius = 0.1
y = 1.0987x + 77.152
R
2
= 0.9783
35
50
65
80
95
110
125
0 10 20 30 40
S.D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
e
s

Radius = 0.15
y = 2.0326x + 22.615
R
2
= 0.9859
20
25
30
35
40
45
50
55
60
65
0 5 10 15 20
S.D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
e
s

Radius = 0.3
y = 1.035x + 2.8271
R
2
= 0.9836
3
3.2
3.4
3.6
3.8
4
4.2
4.4
0 0.5 1 1.5
S.D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
e
s

Figure 4.5 Variation of density of equal soft spheres vs standard deviation. Solid lines are fits.
60

The second model is characterized by a constant size distribution. For systems of
spheres with different radius sizes, the computer program simulating results are shown in
Figure 4.6. The standard errors of the onset volume fraction for various system sizes also
decrease with increasing the size of the system.














61

0.13 < Radius < 0.17
y = 2.0473x + 21.918
R
2
= 0.9941
0
10
20
30
40
50
60
0 5 10 15 20
S.D.
0.12 < Radius < 0.18
y = 2.0545x + 21.165
R
2
= 0.9732
30
35
40
45
50
55
60
65
0 5 10 15 20
S. D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
e
s

0.11 < Radius < 0.19
y = 2.1209x + 19.217
R
2
= 0.9858
30
35
40
45
50
55
60
0 5 10 15 20
S. D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
e
s

Figure 4.6 Variation of density of soft spheres vs standard deviation. Solid lines are fits.
62

Once more making use of Figure 3.1, the excluded volume for a sphere of radius
R is:
3
) 2 (
3
4
R V
ex
= (4-5)
and the total excluded volume for the system of spheres is:
ex c e
V P V = (4-6)
where P
c
is the critical number of spheres at percolation.
Total volume of spaces occupied by spheres was calculated each time when the
percolation occurred. The density of spheres at percolation obtained from graphs was
used to calculate the total excluded volume for three different ranges of radius. These
results have been quantified in Table 4.2.

Table 4.2
Total excluded volume of systems of soft
spheres with different sizes

Radius P
c
V
e

0.1 77.152 2.5854
0.15 22.615 2.5577
0.3 2.8271 2.5579
0.13-0.17 21.918 2.4789
0.12-0.18 21.165 2.3937
0.11-0.19 19.217 2.1734

Studying the above data, one can see that with widening the range of radius size,
the total excluded volume in a system of spheres is reduced and the percolation threshold
is lowered. This result is in excellent qualitative agreement with previous studies
[Balberg, 1984]. But, the excluded volume of the system of soft-core spheres of equal
63

radius was found in the present study to be 2.5, which is slightly lower than the value of
2.8 reported in the literature for soft spheres by Pike and Seager [1974] and Balberg and
coworkers [1984].
Next we simulated hard-core with penetrable-shell model as concentric spheres
where each sphere is consisting of a hard sphere. Different simulations were performed
by keeping the radius of the shell constant (R=0.15) while the radius of the core r was
increased, so that the ratio of shell radius to core radius, r R , decreased from 2 to 1.3
(Figure 4.7).














64

R / r = 2
y = 1.9005x + 27.27
R
2
= 0.9937
30
35
40
45
50
55
0 5 10 15
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.9
y = 1.9985x + 26.362
R
2
= 0.99
30
35
40
45
50
55
0 5 10 15
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.8
y = 1.9831x + 26.602
R
2
= 0.9937
30
35
40
45
50
55
0 2 4 6 8 10 12 14
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.7
y = 2.0857x + 25.55
R
2
= 0.9955
30
35
40
45
50
55
0 2 4 6 8 10 12 14
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.6
y = 2.1667x + 25.179
R
2
= 0.992
30
35
40
45
50
55
0 2 4 6 8 10 12 14
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.5
y = 2.3416x + 24.609
R
2
= 0.9977
30
35
40
45
50
55
0 2 4 6 8 10 12 14
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.4
y = 2.3879x + 24.725
R
2
= 0.9888
30
35
40
45
50
55
0 2 4 6 8 10 12
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

R / r = 1.3
y = 2.2811x + 25.294
R
2
= 0.9871
30
35
40
45
50
55
0 2 4 6 8 10 12
S. D.
D
e
n
s
i
t
y

o
f

c
o
r
e

s
p
h
e
r
e
s

Figure 4.7 Variation of density of hard-core with soft-shell spheres vs standard deviation.
65

Figure 4.8 represents how the total excluded volume of the system, associated
with the core and shell of the spheres, varies with the ratio of the objects shell excluded
volume to core excluded volume.

y = 1.8174x
3
- 0.655x
2
+ 2.6107x - 0.0007
R
2
= 0.9998
0
0.5
1
1.5
2
2.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Vex Core / Vex Shell
P
c
.

V
e
x

C
o
r
e

y = 1.829x
2
- 0.6451x + 2.5999
R
2
= 0.9726
0
1
2
3
4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Vex Core / Vex Shell
P
c
.

V
e
x

S
h
e
l
l

Figure 4.8 Dependence of the total excluded volume of the system of hard-core with soft-
shell spheres on the ratio of core excluded volume to shell excluded volume. The
solid curve represents the apparent polynomial fit.

Based on the graphs, the total excluded volume of the system can be written as a
function of the ratio of core excluded volume to shell excluded volume. According to the
first graph in Figure 4.8:
66

c )
V
V
b( )
V
V
( a V . P
shell ex
core ex 2
shell ex
core ex
shell ex c
+ + = (4-7)
By multiplying the two sides of the above equation by
shell ex
core ex
V
V
we have:
)
V
V
c( )
V
V
b( )
V
V
( a V . P
shell ex
core ex 2
shell ex
core ex 3
shell ex
core ex
core ex c
+ + = (4-8)
which is what the second graph in Figure 4.8 expresses.
When
shell ex core ex
V V is zero, the particle doesnt have a core and the modeling is
soft-core. Thus, the first graph in Figure 4.8 equates to zero and the second graph yields
2.6, which is equal to the density of soft-core spheres. When
shell ex core ex
V V is one, it
means the radius of core has increased so that it has occupied all of the shell, which
induces the hard model of spheres. In general, the introduction of a hard structure
decreases a considerably large fraction of percolating inclusions, leading to a raise in the
percolation threshold. In this case, the density of objects estimated from both equations
(4-7) and (4-8) is 3.8; significantly higher than the value of 1.4 obtained by Balberg,
[1984] for excluded volume of continuum hard-core spheres. This larger value is
probably due to the smaller system simulated by Balberg.

4.3.1 Computational and Experimental Comparison
Now these general observations are applied to particular composite systems.
The shell thickness represents tunneling distance which is a characteristic
electrical property of the matrix polymer. Since electron transport in electrically
conductive composites consists of several different mechanisms, theoretical
67

determination of the shell thickness is a complicated procedure. Instead, it is rather
preferred to estimate the shell thickness based on the experimental thresholds.
In their research, Ogale and Wang [1993] estimated the shell thickness by
matching the experimental threshold for a composite filled with inclusions of a known
size with the hard-core volume fractions computed from simulation results. A concentric
spherical model was used to predict the percolation behavior of the composite specimens.
Similar to their work, in this study, the shell thickness of the system of spheres is
computed by utilizing the values of experimental thresholds extracted from the literature
for composites filled with spherical inclusions.
The percolation threshold is explained based on
exShell
exCore
V
V
for the system of spheres
with hard-core and soft-shell according to the Equation (4-8):
d cx bx ax V . P
exCore C
+ + + =
2 3
(4-9)
where x is
3
3
3
2
3
4
2
3
4
)
R
r
(
) R (
) r (
V
V
x
exShell
exCore
=

= = (4-10)
If V
Core
and V
exCore
respectively represent the volume and excluded volume of one
spherical particle, then
8
1
2
3
4
3
4
3
3
=

=
) r (
) r (
V
V
exCore
Core
(4-11)
By multiplying Equation (4-9) by
exCore
Core
V
V
one obtains:
68

8
1
2 3
) d cx bx ax (
V
V
V . P
exCore
Core
exCore C
+ + + = (4-12)
Or:
Core C
V . P d cx bx ax 8
2 3
= + + + (4-13)
where
Core C
V . P is the critical volume fraction or experimental percolation threshold in
vol%.
From Equation (4-13), x can be easily calculated and then the obtained value may
be substituted in Equation (4-10) to calculate the ratio of
R
r
:
3
x
R
r
= (4-14)
R and r are correspondingly the radius of shell and radius of the core. From the
above equation, R is calculated and knowing the radius of the particle r, the shell
thickness R-r is finally obtained.
Table 4.3 shows the values calculated for x and shell thickness based on this
Monte Carlo modeling for each of the spherical conductive fillers in previous studies.
The values for particle size and percolation threshold have been extracted from literature.
The last column in the Table 4.3 demonstrates the estimated shell thickness. In gathering
the published results of experiments on composites loaded with spherical particles, we
have considered only those works where the percolation threshold (critical volume
fraction) and particle size were explicitly reported.
It should be noticed that not all kinds of carbon black are spherical. Only carbon
blacks of small and geometrically simple aggregates, which are called low-structure, are
69

morphologically sphere-shaped [Balberg, 1987]. Therefore only composites of low
structure carbon black were considered here.

Table 4.3 Estimates of shell thickness for composite systems of spherical fillers
Author Matrix Fiber
Particle
Diameter
(m)
Percolation
Threshold
Shell
thickness
(m)
Aharoni [1972] polymide-amide iron 10 7 3.2888
Ruschau [1992] silicon rubber Ag:TiO
2
1 0.15 0.1536
Wang & Ogale
[1993]
polyethylene nickel grade A 2.9 0.1 0.6922
Wang & Ogale
[1993]
polyethylene nickel grade B 8 0.23 0.6650
Wang & Ogale
[1993]
polyethylene nickel grade C 68 0.43 0.5985
Karasek et al. [1996]
styrene butadiene
rubber
carbon black 0.15 0.25 0.0107
Nakamura et al.
[1997]
high density
polyethelene
carbon black 0.09 0.2 0.0094
Foulger [1998] polyethylene carbon black 0.03 3.3 0.0170
Nagata et al. [1999]
low-density
polyethylene
graphite 5.1 0.292 0.2619
Rubin et al. [1999] polymer carbon black 0.32 0.4 0.5276
Flandin et al., [2000]
ethylene-octene
elastomer
carbon black 3 0.35 0.0906
Brechet et al. [2001] polybutylacrylate silica 0.120 0.24 0.0093
Coso [2004] Al
2
O
3
Cu 0.002 0.07 0.0065

As can be seen in Table 4.3, the shell thickness was found within the range 0.6 to
0.68 m for composites filled with nickel powder with different grades, in agreement
with 0.7 m reported in an earlier study by Wang and Ogale [1993].
70

Shell thickness depends primarily on the matrix polymer and the reinforced
conductive phase. If the conductivity mechanism is tunneling, the shell thickness is a
representation of tunneling distance, but it does not always equate to it. If the tunneling
distance is indeed about 5 or less nanometer as suggested by Du et al. [2004], the values
of 0.009 m for the silica/poly butyle acrylate composite and 0.006 m for Cu:Al
2
O
3

composites appear to be reasonable estimates for the average shell thickness. However
the estimated shell thickness of 0.6 m for polyethylene composites filled with nickel
powder is too large. The same can be said for the shell thickness of 0.26 microns obtained
for the low-density polyethylene composite of spherical graphite particles or shell
thickness of 0.36 microns estimated for carbon black polymer composite.
The quantum mechanics requires a tunneling distance of few nanometers if the
electrical transport between the conductive inclusions occurs only by tunneling
mechanism. In composites in which the primary electric charge transferring mechanism is
given by electron hopping and tunneling, the conducting objects are electrically
connected, and not necessarily physically connected. The interparticle distance decreases
proportionally with decreasing particle diameter [Jing et al., 2000]. That is why the
particle size effect plays an important role on the percolation threshold. In the case of
large particles, the conductance between two particles is non-zero only when they
essentially touch each other. The approach outlined in this dissertation does not test the
average shell thickness or tunneling distance in composites in which the main
interparticle transport mechanism is not electron tunneling. The large shell thickness
estimated for metal reinforced composites can be explained by the fact that for metal
71

particles with micro-scale size, the electron current flows through circular pathways with
very low resistances which are created between conductive fibers [Bridge et al. 1988].
The same procedure can be applied to the system of hard-core circles to evaluate
the shell thickness given that the experimental data is available.

4.4 Spherocylinders Percolation
In this section, the results of different cases of simulation for a system of
spherocylinders are presented. Each evaluation of the percolation probability was
obtained from a thousand randomly generated configurations, based on the connectivity
criterion in spherocylinders expressed in section 3.4.2.
First, the simulating result for excluded volume of spherocylinder is presented as
a verification of the validity of the algorithm. Also, continuum results for spherical
particles are well known and can be utilized as a check for simulation veracity. Finally,
the simulation results for soft-core and hard-core modeling will be presented.

4.4.1 Excluded Volume Modeling
The excluded volume of a spherocylinder can in principle be obtained by
Equation (3-27). An alternative would be a simulating scheme in which a single
speherocylinder of a given length and diameter is planted in the middle of a unit cube
(Fig. 4.9). Then one million similar objects are randomly placed inside the simulation
cell. If n is the number of those which intersect the central sphereocylinder, the excluded
volume of a spherocylinder is calculated through the equation:
3
1000000
W
n
V
ex
= (4-15)
72

where W is the width of the cube. Because of its random nature, the entire process is
repeated a number of times, so that a mean of the excluded volume could be calculated.






Figure 4.9 The unit cubic cell and one spherocylinder at the center. All
dimensions are measured in the units of the simulation cell.

Table 4.4 presents the results for the excluded volume deduced from the
simulation program. It also presents values obtained from the Equation (3-27). Both L
and D are measured in the units of the cube size.

Table 4.4 Estimates of excluded volume of spherocylinders
with different aspect ratios

D
L

V
ex

Equation
Results
Simulation
Results
0 0.1130973 0.1131592
0.1 0.1304861 0.1306368
10 0.0060507 0.0059403
100 0.0004412 0.0004414
600 0.0000712 0.0000714
1000 0.0000426 0.0000419

1
1
73

It can be seen that the model closely matches the equation. The comparison
between the results of two methods verifies the correctness of the computer model in
calculation of excluded volume of the spherocylinders with different aspect ratios.

4.4.2 Spheres Percolation
It is important to check the present spherocylinder percolation modeling against
the well studied case of spheres, in order to establish the reliability of the procedure used
for spherocylinders. Hence, the above procedure was used in the special case (L=0) for a
fixed radius (R=0.15). In this case the spherocylinders practically degenerate into
spheres. Figure 4.10 shows the results.

L=0, R=0.15
y = 2.2104x +23.045
R
2
= 0.9985
20
30
40
50
60
70
80
90
100
0 5 10 15 20 25 30 35
S.D.

Figure 4.10 Variation of average density of soft spherocylinders with length
zero vs standard deviation. The straight line through the points is
a straight line fit obtained from the linear least squares method.

As it was expected the total excluded volume was the same as seen in the
simulations of spherical particles:
74

6063 2 15 0 2
3
4
045 23
3
. ) . ( . =

(4-16)
Moreover, the result would also support the value obtained for the total excluded
volume of the system of soft spheres in other studies [Pike and Seger, 1974; Balberg et
al., 1984]. This observation then lends further support to the computer programming
model.

4.4.3 Soft-Core Modeling
In this section, the simulation results for percolation of a system of soft
spherocylinders are presented. Figure 4.11 distinguishes the simulation results for three
different aspect ratios where again L denotes the length and D the width of the rods. The
spherocylinders are of equal size. The length is kept constant and equal to 0.3 while the
radius is changed to produce aspect ratios range from 10 to 1000. In all our computations
we have assumed particles dispersed randomly in a cube of a unit volume, hence all the
lengths mentioned are in units of the cubes edge.









75

a.r.=10
y = 2.3834x +219.13
R
2
=0.9807
0
50
100
150
200
250
300
350
400
450
0 20 40 60 80
S.D.

a.r.= 50
y = 7.4958x +1370.8
R
2
= 0.9796
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
0 20 40 60 80 100
S.D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
o
c
y
l
i
n
d
e
r
s

a.r.=100
y = 0.2772x +2049.8
R
2
= 0.9702
1000
1200
1400
1600
1800
2000
2200
2400
0 200 400 600 800 1000
S.D.
a.r.=1000
y = 8.9164x +15396
R
2
= 0.9722
0
4000
8000
12000
16000
20000
24000
28000
0 200 400 600 800 1000
S.D.

Figure 4.11 Variation of average density of soft spherocylinders vs standard
deviation for different aspect ratios. The straight lines through each
data set are best fits base on least squares method.


Next, we expand the computer program for simulating system of spherocylinders
with variable aspect ratios. We have considered the isotropic case for the spherocylinders
with a fixed length L and a variable radius R, so as to scan the wide range of aspect ratios
(L/2r) from L/2R <<1 to L/2R >>1 with an average fiber aspect ratio of 10.
Distribution in aspect ratio can also be acquired by keeping the radius of spherocylinders
76

constant and changing their length which is expected to give the same result. This is
because as mentioned before in a spherocylinder the important geometrical feature in
percolation is aspect ratio.
Here in order to have a better estimate of density of the particles, instead of the
total number of spherocylinders, the total volume of objects was calculated each time
when percolation occurred. Figure 4.12 shows the average density of soft spherocylinders
at percolation in systems with different aspect ratios.














77

9 < a.r. < 11
y = 0.894x +198.91
R
2
= 0.9781
0
60
120
180
240
300
0 20 40 60 80
S.D.
D
e
n
s
i
t
y

o
f

s
p
h
e
r
o
c
y
l
i
n
d
e
r
s

8 < a.r. < 12
y = 0.7746x + 189.92
R
2
= 0.9793
0
40
80
120
160
200
240
280
320
0 20 40 60 80 100 120
S.D.

7 < a.r. < 13
y = 1.0711x +177.92
R
2
= 0.9885
100
130
160
190
220
250
280
310
0 20 40 60 80 100
S.D.

Figure 4.12 Variation of average density of soft spherocylinders with variable
aspect ratios vs standard deviation. Solid lines are linear fits.
78

The density of spherocylinders at percolation obtained from graphs was used to
calculate the total excluded volume for different aspect ratios. The numerical results of
these calculations are presented in Table 4.5.

Table 4.5
Total excluded volume in systems of
soft spherocylinders

Aspect Ratio P
c
V
e

10 219.13 1.3259
50 1370.8 1.2337
100 2049.8 0.9044
1000 15398 0.6557
9-11 198.91 1.2035
8-12 189.92 1.1492
7-13 177.92 1.0765


In their study, Balberg and his coworkers [1984] showed that for three
dimensional sticks with random orientations, the maximum total excluded volume is
significantly lower than the total excluded volume of spheres or parallel-aligned objects.
They reported this maximum value is 1.8. Table 4.5 confirms this statement; as all of the
numerical values obtained for the total excluded volume of spherocylinders are lower
than 1.8 and they are less than the values obtained for total excluded volume of spheres
and total excluded area of circles.
The fractional decrease in the total excluded volume and thus percolation
threshold due to the increase in aspect ratio is obvious, rising gradually over 1.0 due to a
distribution in the aspect ratio.
79

The average number of contacts per spherocylinder increases with the aspect
ratio. The shortest distance between pairs of spherocylinders axes is crucial to cluster
formation. As expected, the percolation threshold of conductive filler particles decreases
with increasing aspect ratio. For spherocylinders of large aspect ratio, L/D where L is the
cylinders length and D is its diameter, it was found that in the soft-core limit as the
aspect ratio increases to 1000, total excluded volume of system decrease. A reduction in
total excluded volume is also observed for a system of sticks with variety of aspect ratios.
From the results of the simulation, the advantages of high aspect ratio particles are
apparent, because long spherocylinders contribute more to the percolation threshold than
short ones do; therefore, a lower aspect ratio means the system percolates in a higher
percolation threshold. Also a system which contents spherocylinders with variable aspect
ratios percolates in a lower threshold. These results are in agreement with the literature
data [Balberg, 1983; Kyrylyurk and Schoot, 2008].

4.4.4 Hard-Core with Soft-Shell Modeling
To test this case, a Monte Carlo computation was carried out on a system of
randomly aligned capped cylinders of length L , hard-core radius r, and soft shell
thickness R-r. The core aspect ratio decreased from 90 to 17, while the shell aspect ratio
is kept constantly equal to 10. It is convenient to use the volume fraction of the core
growing phase as the density parameter in order to calculate the excluded volume for the
system. The simulation results for percolation of a system of spherocylinders with hard-
core and soft-shell for selected aspect ratios are shown in Figure 4.13.

80

core a.r.= 90
y = 1.0446x + 144.42
R
2
= 0.9459
30
80
130
180
230
280
0 20 40 60 80 100 120
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 80
y = 1.8383x + 132.21
R
2
= 0.9808
30
80
130
180
230
280
330
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 70
y = 1.6194x + 105.24
R
2
= 0.9902
30
80
130
180
230
280
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s
core a.r.= 60
y = 1.4028x + 106.54
R
2
= 0.9701
30
70
110
150
190
230
0 20 40 60 80
S. D.
D
e
n
s
it
y

f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 50
y = 1.6303x + 102.38
R
2
= 0.9132
30
80
130
180
230
280
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 40
y = 1.3416x + 98.432
R
2
= 0.9695
30
70
110
150
190
230
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 30
y = 1.3804x + 77.82
R
2
= 0.9634
30
60
90
120
150
180
210
0 20 40 60 80
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.= 20
y = 2.5794x + 69.52
R
2
= 0.9819
30
60
90
120
150
180
210
0 10 20 30 40 50
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

Figure 4.13 Variation of average density of hard-core with soft-shell spherocylinders vs
standard deviation. For all graphs, the shell aspect ratio is 10 = D L
81

The numerical results of these calculations are presented in Table 4.6. As can be
seen a very low excluded volume is observed for system of hard-core spherocylinders.

Table 4.6
Total excluded volume in the systems of
spherocylinders with shell aspect ratio=10

Core Aspect Ratio P
c
V
e

90 144.42 0.0711
80 132.21 0.0736
70 105.24 0.0674
60 106.54 0.0804
50 102.38 0.0939
40 98.432 0.1150
30 77.82 0.1250
20 69.52 0.1779
19 78.93 0.2146
18 95.942 0.2782
17 110.453 0.3429

As the core aspect ratio decreases, the shell thickness also decreases, so that a
significantly larger threshold is required for percolation. The reason is that the reduction
of shell thickness represents a decrease in the effective range of electron transfer. This
was found to be in agreement with the result that Ogale and Wang [1993] proposed for
their model consisting of hard-core with soft-shell cylinders with maximum aspect-ratio
of 25.
Figure 4.14 represents how the total excluded volume of the system, associated
with the core and shell of spherocylinders, varies with the ratio of objects shell excluded
volume to core excluded volume.

82

y = 8.3109x
3
- 5.5337x
2
+ 1.3212x- 0.0062
R
2
= 0.9902
0
0.2
0.4
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6
VexCore / VexShell
P
c
.

V
e
x

C
o
r
e


y = 8.4519x
2
- 5.4794x+ 1.2537
R
2
= 0.9371
0
0.4
0.8
1.2
1.6
0 0.1 0.2 0.3 0.4 0.5 0.6
VexCore / VexShell
P
c
.

V
e
x

S
h
e
l
l

Figure 4.14 Dependence of the total excluded volume of the system of hard-core with soft-
shell spherocylinders on the ratio of core excluded volume to shell excluded
volume. The solid curve represents the apparent polynomial fit.

From the above graphs, the total excluded volume of the system can be expressed
as a function of the ratio of core excluded volume to shell excluded volume:
d )
V
V
c( )
V
V
b( )
V
V
( a V . P
shell ex
core ex 2
shell ex
core ex 3
shell ex
core ex
shell ex c
+ + + = (4-17)
By multiplying the two sides of the above equation by
shell ex
core ex
V
V
:
d
V
V
)
V
V
c( )
V
V
b( )
V
V
( a V . P
shell ex
core ex 2
shell ex
core ex 3
shell ex
core ex 4
shell ex
core ex
core ex c
+ + + = (4-18)
83

If the ratio
shell ex core ex
V V is equal to zero, it means the object doesnt include any
core and the model transforms to the soft core model of spherocylinders (soft-core
transition). So that the first graph gives the zero value and the second graph gives the
value similar to that of the density of soft spherocylinders which is equal to 1.3.
If the ratio
shell ex core ex
V V is equal to one, it means the radius of core has increased
so that it occupies all of the shell and the model transforms to the hard core model of
spherocylinders (soft-core-hard-core transition). Therefore, the two equations give the
similar values for the density of objects (4.09 from the first graph and 4.23 from second
graph in Figure 4.14 ).
The lower threshold of 1.3 for the system of soft-core spherocylinders in
comparison to 4 for system of hard-core spherocylinders is due to the fact that the soft-
core approach allows the objects to overlap which results in a lower excluded volume and
percolation threshold, as mentioned before.
The same procedure was repeated for a system of hard-core spherocylinders
where the range of core aspect ratios increased from 160 to 900 while the shell aspect
ratio was kept constantly equal to 100. The simulation results for percolation of a system
of spherocylinders with hard-core and soft-shell for selected aspect ratios are shown in
Figure 4.15.
84

core a.r.=900
y = 1.3934x + 1307.21
R
2
= 0.9943
400
800
1200
1600
2000
2400
2800
0 200 400 600 800 1000
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
l
in
d
e
r
s

corel a.r.=800
y = 8.5295x + 1217.34
R
2
= 0.9746
400
800
1200
1600
2000
2400
2800
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
l
in
d
e
r
s

core a.r.=700
y = 9.0858x + 1109.32
R
2
= 0.981
400
800
1200
1600
2000
2400
2800
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

corel a.r.=600
y = 8.089x + 1028.43
R
2
= 0.9769
400
800
1200
1600
2000
2400
2800
0 20 40 60 80 100
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
l
in
d
e
r
s

core a.r.=500
y = 10.587x + 890.81
R
2
= 0.9649
400
800
1200
1600
2000
2400
2800
0 10 20 30 40 50 60 70
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.=400
y = 12.558x +645.32
R
2
= 0.9721
400
800
1200
1600
2000
2400
2800
0 10 20 30 40 50 60 70
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.=300
y = 13.273x + 645.32
R
2
= 0.9739
400
800
1200
1600
2000
2400
2800
0 10 20 30 40 50 60
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

core a.r.=200
y = 13.48x + 907.45
R
2
= 0.9485
400
800
1200
1600
2000
2400
2800
0 10 20 30 40 50 60
S. D.
D
e
n
s
it
y

o
f

s
p
h
e
r
o
c
y
lin
d
e
r
s

Figure 4.15 Variation of average density of hard-core with soft-shell spherocylinders vs
standard deviation. For all graphs, the shell aspect ratio is 100 = D L
85

The numerical results of these calculations are presented in Table 4.7.

Table 4.7
Total excluded volume in the systems of
spherocylinders with shell aspect ratio=100

Core Aspect Ratio P
c
V
e

900 1377.2 0.0652
800 1359.3 0.0724
700 1297.3 0.0791
600 948.44 0.0675
500 860.81 0.0736
400 802.34 0.0859
300 645.32 0.0924
200 907.45 0.1963
190 1217.3 0.2775
180 1585.16 0.3818
170 1660.423 0.4240
160 1790.423 0.4865


Figure 4.16 represents how the total excluded volume of the system, associated
with the core and shell of spherocylinders, varies with the ratio of objects shell excluded
volume to core excluded volume.

86

y = 6.0209x
3
- 3.8122x
2
+ 0.8898x+ 0.0043
R
2
= 0.9844
0
0.2
0.4
0.6
0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Vex Core / Vex Shell
P
c
.

V
e
x

C
o
r
e


y = 6.0897x
2
- 3.9291x+ 0.9401
R
2
= 0.9363
0
0.2
0.4
0.6
0.8
1
1.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Vex Core / Vex Shell
P
c
.

V
e
x

S
h
e
l
l

Figure 4.16 Dependence of the total excluded volume of the system of hard-core with soft-
shell spherocylinders on the ratio of core excluded volume to shell excluded
volume. The solid curve represents the apparent polynomial fit.

If the ratio
shell ex core ex
V V is equal to zero, it means the object doesnt include any
core and the model transforms to the soft core model of spherocylinders (soft-core
transition). So that the first graph gives the zero value and the second graph gives the
value similar to that of the density of soft spherocylinders which is about 0.9.
If the ratio
shell ex core ex
V V is equal to one, it means the radius of core has increased
so that it occupies all of the shell and the model transforms to the hard core model of
87

spheres (soft-core-hard-core transition). Therefore, the two equations should give the
similar values for the density of objects (3.1028 from the top graph and 3.1007 from the
bottom graph).
Again, the lower threshold of 0.9 for the system of soft-core spherocylinders in
comparison to 3.1 for system of hard-core spherocylinders is because the soft-core
approach allows the objects to overlap which causes a less excluded volume and
percolation onset.

4.5 Conclusion
From the above discussion, it can be conclude that for circular and spherical
particles, a smaller particle size as well as widening the radius range will lower the
percolation threshold. Soft-core approach allows particles to overlap and increases the
total area or volume of space occupied by the particles.
One of the early percolation models often referenced was originally proposed by
Pike and Seager [1974]. Their work was concerned only with the percolation threshold
of the isotropic case of randomly distributed sticks. Later, Balberg and Binenbaum [1983]
have performed extended work of Pike and Seager [1974]. They have presented results
from a series of Monte Carlo simulations for both hard-core and soft-core spheres as well
as soft-core circles, using both continuum and lattice percolation models. They
considered the dependence of the percolation threshold on the anisotropic system with a
fiber orientation distribution. Furthermore, in terms of total excluded area or volume
approaches, a two- and three-dimensional analysis of percolation threshold has been
presented [Balberg et al., 1984].
88

In this study, a spherocylindrical model was used for predicting the percolation
threshold of electrical conductivity for fiber-reinforced composites. The proposed
procedure computes average onset volume fractions of percolation for various finite
system sizes, and applies finite size scaling (FSS) arguments to predict the thresholds for
infinitely large systems. Also, for random fiber orientations, the finite size scaling
hypothesis was found to be applicable.
The maximum possible numerical value for the total excluded volume associated
with randomly oriented capped cylinders is found to be considerably less than the total
excluded volume of spheres and total excluded area of circles. Also a length polydipersity
lowers the percolation threshold of the spherocylinders system. This is in agreement with
the results of Balberg and his coworkers [1984] and Kyrylyuk and Schoot [2008].
The validity of our formulation is assessed by reinterpreting existing experimental
results on nanosphere composites and by extracting the characteristic shell thickness,
which is found to be within the expected range of its values. The numerous simulations
confirmed the validity of the simulation method proposed in this study, so that it can be
applied with sufficient accuracy to determine the percolation thresholds in continuum
models of circles and spheres with or without a radius size distribution.
Our model can be used to simulate the electrical properties of real composites,
and can be generalized to include different filler shapes, filler size, and aspect-ratio
polydispersity within the insulating matrix.
The shortcoming of the model is that it simulates the fibers as individual straight
sticks distributed inside the matrix. In reality, the nanofibers are not monodisperse, nor
are they infinitely rigid or even straight.
89


Chapter 5
Experiment
5.1 Introduction
The electrical conductivity of a polymeric material can be increased by the
addition of conductive fillers such as metallic powder [Clingerman et al., 2003;
Mamunya et al., 2002; Araujot and Rosenberg, 1976]. As such, polymer nanocomposites
using these materials have outstanding properties and can be used in many potential and
functional applications like electrostatic charge dissipation and shielding/absorbing
devices for electromagnetic and radio-frequency interference, where metals have
conventionally been used for years [Ma et al., 2005].
Polymeric composites of different types of carbon fillers including carbon black,
graphite, carbon fibers, and carbon nanotubes are an extensively used group of
composites and have recently been used for many technological applications. Due to its
excellent conductivity and low cost, carbon nanofiber has been widely used as the filler
of choice to enhance the electrical properties and mechanical strength of polymeric
nanocomposites [Taipalus et al., 2005; Agari et al., 1994].
90

The objective of this chapter is to set an example process of fabricating a polymer
carbon fiber nanocomposite that has electrochemical properties. The chapter is concerned
with the preparation of flexible high loading nanofiber-elastomer of composites and
investigating the usage of conductive fibers such as carbon nanofiber (CNF) to increase
the electrical conductivity of polymeric nanocomposites. The electrical conductivity
properties of these composites and the effect of chemical surface treatment are
investigated and electrical and percolative results are presented. Based on the
experimental results of electrical conductivity measurements, the critical exponent and
the constant of proportionality in the classic percolation equation were determined.

5.2 Electrical Measurement
The electric conductivity of a material is determined by measuring the resistivity.
The test standard often used for measuring resistivity of insulators is ASTM D-257
[2005].
Depending upon the application, there are two main techniques for measuring the
resistance: two-point probe method and four-point probe (Kelvin) method [Northrup,
1912]. A problem with the two-point probe technique is that the voltage is measured not
only across the resistance, but includes the resistance of the leads and contacts as well. If
the resistance is great enough (>10) to ignore the unwanted resistance of the leads, this
error is not a problem. But in measuring low resistances like nanocomposites the error is
not negligible.
To solve this problem a second set of wires is used in the four-wire or four-point
method. Here, negligible current flows in the second set of wires; therefore only the
91

voltage across the unknown resistance is measured. By using Ohms law resistance can
be calculated [Keithley et al., 1998].

5.3 Theory of Four-Point Probe
This technique is one of the most widely used methods for the measurement of
resistivity of semiconductors. The four-point probe system has four collinear tungsten
tips. There is a very small yet equal distance between each two tips (Fig. 5.1). At one end
the tips contact the sample and at the other end they are connected to springs to minimize
the damage introduced to the sample. A voltmeter measures the voltage across the inner
two probes and a current source supplies current to flow though the material by two outer
probes [Chan, 1994].


Figure 5.1 Enlarged schematic of the four-point probe
configuration


92

In the ideal circumstance where it can be considered that the sample is semi-
infinite (infinite extent and very small thickness), it can be shown that the resistivity of it
is [Smits, 1985]:
I
V
s = 2
0
(5-1)
where s is the interprobe spacing. However, in principle, samples are of finite size.
Therefore, the above equation should be corrected to be valid for typical samples. Valdes
[1954] showed that if the thickness of the sample is at least five times that of the probe
spacing, no correction is needed. Otherwise the resistivity of the sample can be computed
by using a correction factor in equation (6-1):
I
V
s a = 2 (5-2)
where a is the correction factor and is obtained in terms of s t [Valdes, 1954].
Generally in four-point probe apparatus, the probe spacing is 25-60 mils and
typical samples used in tests are of 0.5-20 mils thickness. This means that usually 5 . 0
s
t

and hence a is determined to be:
s
t
a 72 . 0 = (5-3)
and when substituted into the Equation (5-2), the resistivity of the sample for the cases of
5 . 0
s
t
is :
I
V
t 53 . 4 = (5-4)
93

This equation can be manipulated into a more useful form by dividing both sides
by t and replacing
t

by
S
R :
I
V
R
S
53 . 4 = (5-5)
S
R is the sheet resistance, since it can be interpreted as the resistance of a square
sample. This can be clear if one recalls the Equation (5-3) for a rectangular sample where
t w A = .

t w
l
R = (5-6)
For square l w = , so
t
R

= (5-7)
which is equal to
S
R .

5.4 CNF Buckypaper Conductivity
5.4.1 Materials
The polymer used in this work was polydimethylsiloxane (PDMS), purchased
from Gelest Inc.
1
(Morrisville, PA, USA). PDMS is one of the most widely used silicon-
based organic polymers [Utracki, 2002]. The chemical formula of PDMS is
(H
3
C)
3
SiO[Si(CH
3
)
2
O]
n
Si(CH
3
)
3
, where n is the number of repeating monomer
[SiO(CH
3
)
2
] units, and it has an average molecular weight of 1100 gm/mol [Mark, 2006].

1
http://www.gelest.com
94

Like most polymers, PDMS is naturally insulating and its room temperature conductivity
is no higher than 10
-14
S/cm [Barton et al., 2007].
The conducting fillers were carbon nanofibers, constituted of natural pure carbon
or pristine fibers which were supplied by Applied Sciences Inc
2
(Cedarville, OH, USA)
and were about 100-200 nm in diameter and 30,000-100,000 nm in length (Fig. 5.2). The
range of electrical conductivity of carbon fillers is from 10
2
to 10
5
S/cm [Barton et al.,
2007], compared to metals which are typically around 10
6
S/cm [Clingerman et al.,
2002].

Figure 5.2 Carbon nanofiber powder

Adhesion between fibers and polymer matrix is considered necessary and
governed by using Poly-diethoxy siloxane (PSI-021) as the cross-linker to induce the
chemical and physical interactions at the interface and increase the solubility of the CNF
in conducting form in polymer. Di-n-Butyldilauryl (SND 3260) was used as Tin based
catalyst to speed up the process.

5.4.2 Chemical Surface Treatment
Proper dispersion and good interfacial bonding between the CNFs and the
polymer matrix have to be guaranteed to form a consistent network of fiber nanoscale

2
http://www.apsci.com
95

[Baraton 2003]. Chemical surface treatment can be used to alter the surface composition
of particles to promote the incorporation of the nanoparticles into the matrices and
enhance the adhesion and interfacial region. It is considered an area of importance for
high added value of nanoparticles in the emerging field of nanofabrication.
Oxidation is the creation of defect sites on the structure of the CNF walls for
breaking up the CNF strong bundles, though it is impossible to break up all the
entanglements of the CNF material. This can result in reduction of their aspect ratio.
Functionalization is the modification of the CNFs walls by grafting specific chains onto
the surface of the CNF walls to improve the wetting of the filler as well as the dispersion
in the medium. It has been shown that oxidized and functionalized nanofillers have a
higher toughness than pristine nanofillers which increases the toughness of the whole
nanocomposite formed using them [Jeong and Kessler, 2008]. Meanwhile oxidation and
functionalization of CNF helps in dispersion of the fibers in the polymer matrix [Baraton
2003].
In this research, in addition to the direct mixing of the CNFs and the polymer,
oxidized fiber (OCNF) and functionalized fibers with Amine-terminated PDMS group
(CNF-PDMS-NH
2
) were also employed as conductive material. Figure 5.3 represents the
structure of the above functional groups on CNF surface.

(a) Oxidized fiber (b) PDMS functionalized fiber
Figure 5.3 Functional groups attached to the surface of carbon nanofibers
96

A detailed description of sample fabrication and the process of oxidation and
functionalization to form the elastomer DMS-S14, known as two-step condensation cure
chemistry, can be found in the research by Mapkar et al. [2011].

5.4.3 Sample Preparation
Polymer matrix nanofiber composite films were designed with a high degree of
CNF dispersion using a novel synthesis method developed by Dr. M. R. Colemans
research group in the University of Toledos Chemical Engineering Department
laboratory. The method is described below.
The carbon nanofibers were first ground using a pestle to break up agglomerated
nanofibers. Then carbon nanofibers and matrix polymer were well mixed by slowly
adding 10 wt.% cross linker and using 0.04-0.08 grams catalyst until a homogeneous
distribution of fillers within the polymer matrix was achieved. The conductive materials
concentration range was from 0.5 to 50 wt.% so that the total concentration of fiber and
polymer was kept always 90 wt.%. The remaining 10 wt.% of the sample constitutes of
cross-linker. The total weight of each sample was considered constant and equal to 0.4
grams. Particle dispersion is affected adversely as the loading weight of nanoparticles
increases. At higher loadings, due to the strong interaction between the carbon
nanofibers, obtaining a uniform dispersion of the fiber in the polymer matrix was more
difficult and an electronic SpeedMixer DAC 150 SP was used. The samples with
functionalized CNFs showed much better dispersion in polymer matrix than those
without functionalization.
97

After the CNF was dispersed in the PMDS matrix, at low loading of CNF, the
nanosheet was obtained by pouring the above mixture in a desired plate and letting it dry
for 24 hours at room temperature for the catalyst to evaporate prior to testing. The
nanosheet made of higher percentage of fibers was obtained by putting them under a
weight. Figure 5.4 displays the processing steps of the CNF-composite material.
Appropriate dispersion of the nanofiber sheets is essential for the formation of continuous
conducting network.


Figure 5.4 Steps in casting CNF-composite film


Typically for each formulation and each kind of carbon filler weight percentage,
two specimens were prepared to obtain a total of 106 test specimens. Only single fillers
were added to the polymer in this experiment and there were no combination of fillers
studied.
98

Because of the significant increase in brittleness, the maximum amount of fiber
that could be blended and solved into test specimens was 40% for the pristine type of
fibers. However, the samples with a concentration lower than 0.5wt% for the pristine
fibers could be prepared. Each specimen was approximately a disk of average diameter of
3-4 cm which varied in thickness from 0.3 to 1.2 mm, thus with an aspect ratio of
thickness about 100. Figure 5.5 shows the final CNF-composite nanosheets.


Figure 5.5 Random CNF-composite samples


All dimensions were measured by the electronic caliper (Fig. 5.6).


Figure 5.6 The electronic caliper


99

The volume fraction of the CNF was determined from the weight and density of
the material used. So the final samples are characterized by their vol.% in CNF, :

P F
F
V V
V
+
= (5-8)
where
P
V is the volume of the dry polymer and
F
V is the volume of CNF. The
calculation is obviously based on supposing total evaporation of the solvent.
Table 5.1 shows the densities of CNF [Xing et al., 2008], PDMS [Mark, 1996],
and the cross linker [Gelest Inc., 2007].

Table 5.1 Densities of materials

Material CNF PDMS Cross linker
Density (g/cm
3
) 1.95 0.97 1.05-1.07

The concentrations in wt.% and the corresponding vol.% for all of the filler
composites tested in this research are shown in Table 5.2.

Table 5.2 Fiber loading levels

wt.% 0.5 1 2 4 5 7 10 12 15 20 30 40 50
vol.% 0.3 0.5 1.0 2.0 2.6 3.6 5.2 6.4 8.1 11.1 17.6 25.0 33.4


5.4.4 Experimental Set-up and Test Performance
The electrical input was provided by means of a DC power source. Current and
voltage measurements were performed using a Keithley 617 digital electrometer and
100

Fluke 8840 digital multimeter. The resistivity test fixture is model A of A&M Fell Ltd, a
UK-based company (Fig. 5.7).



Figure 5.7 Four-point probe apparatus


A 000 , 10 resistor is connected in series with the specimen. By applying a
constant current (typically mA 10 5 ) the voltage drop over the sample during the test is
measured using Ohms law,

=
U
I , where U is the voltage across the resistance and
is the resistance of the resistor. So the sheet resistance of the specimen is:
U
V
U
V
R
S
= = 10000 53 . 4
10000
53 . 4 (5-9)
Finally the resistivity of the specimen is obtained by
t R
S
= (5-10)
where t is the thickness of the specimen. The electrical conductivity of all samples is
determined by taking the inverse of the electrical resistivity.

Power supply
Test fixture
101

5.5 Results and Discussions
The electrical conductivity of all the samples was measured perpendicular to the
plane of the film across the four ends of both sides of the nanosheet using the four-point
resistivity method, so that for each sample formulation eight measurements were
performed and then the average value of electrical conductivity was obtained for the
different loading of nanofiber.
It is of significant interest to compare the results of the same polymer with the
fillers of differing types. Figure 5.8 compares the mean value of DC surface conductivity
of nanocomposites as a function of the CNF content.

0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
C
o
n
d
u
c
t
i
v
i
t
y

(
S
/
c
m
)
0.3 0.5 1.0 2.0 2.6 3.6 5.2 6.4 8.1 11.1 17.6 25.0 33.4
CNF content (vol.%)
Pristine
OCNF
CNF-PDMS-NH2

Figure 5.8 DC surface conductivity as a function of the CNF loading


102

It can be seen that the different functionalization has an effect on the conductivity
of the composite. Fiber functionalization is often necessary to improve mechanical
properties and dispersion in the structures of nanofiber-reinforced composite. On the
other hand, surface modification of carbon nanofibers changes the graphitization extent
of the fibers and increases their surface area. In general, conductivity decreases with
increasing filler hardness and/or elastic modulus [Ruschau et al., 1992].
As anticipated, the composites containing the pristine fibers achieve a much
higher conductivity level. It exhibited high electrical conductivity of 1.6 S/cm when the
carbon content was about 11 vol.%.
At 17.6 vol.%, the conductivity of oxidized composite reached 0.35 S/cm, about
five orders of magnitude higher than the value of composites of functionalized fibers
(0.076 S/cm), and about four orders of magnitude lower than the value of composites of
pristine nanofillers (1.46 S/cm) at the same fiber loading. This phenomenon could be
attributed to the destruction of the conducting network during oxidation.
The above values of conductivity might seem low especially for composites
loaded with oxidized and functionalized fibers; still these values are high enough for
some electrical applications including electrostatic dissipation, electrostatic painting and
electromagnetic interference shielding as demonstrated in Figure 5.9 [Ramasubramaniam
et al, 2003].




103

1E-11
1E-10
1E-09
1E-08
1E-07
1E-06
1E-05
0.0001
0.001
0.01
0.1
1
10
0.3 0.5 1.0 2.0 2.6 3.6 5.2 6.4 8.1 11.1 17.6 25.0
CNF loading (vol%)
E
l
e
c
t
r
i
c
a
l

C
o
n
d
u
c
t
i
v
i
t
y

(
S
/
c
m
)
Pristine
OCNF
CNF-PDMS-NH2

Figure 5.9 Conductivity in logarithmic scale as a function of the CNF loading. The
inset shows the required conductivity levels for electrical applications:
Electrostatic dissipation, electrostatic painting and EMI shielding.

In Figure 5.10 the measured electrical conductivity of the CNF-based materials as
a function of the CNF volume concentration are plotted on the logarithmic scale,
allowing observing the percolation. The experimental result points are connected by
straight line segments. The behavior of a classical percolating system for the composite
system can be clearly described where a low conducting region with conductivity of the
order of the insulating matrix is well separated from a high-conducting region with
conductivity close to the additive conductivity. The composites of pristine, oxidized, and
functionalized fibers have rather different features, as shown in Figure 5.10. The figure
clearly shows that the conductivity of the composites is strongly dependent on the CNF
loading. A feature of interest in this figure is the critical filler content that yields a
104

conductivity transition, which can be designated as the percolation threshold. In general,
the percolation threshold is characterized by a sharp increase in the electrical
conductance. Upon further increase of the carbon nanofiber content, the electrical
conductivity only showed a slight increase until a plateau was observed.
The high increase of the conductivity value was observed between 1.0 and 2.03
vol.% for composites of pristine , where the conductivity changed from 0.00059 to about
0.0016 S/cm. The oxidized composite conductivity increased highly between 2.5 and 3.6
vol.% when the conductivity increased from 1.3810
-6
S/cm to a magnitude of four
orders. Functionalized composites conductivity improved from 1.610
-6
to 1.110
-5
S/cm
between 3.6 and 5.2 vol.%. At loading levels in excess of 11.0 vol.%, the conductivity of
all three formulations increased only moderately.
In general, Figure 5.10 indicates that the percolation threshold for this material
resides between 1.0 and 5.0 vol.%. The percolation threshold of pristine composite is
observed at about 1.0 vol.%, which was much lower than that for CNF modified by
functionalization (5 vol.%). As the filler concentration increases beyond the percolation
threshold, a continuous conducting network has been formed, so the conductivity of the
three kinds of composites is seen to increase gradually.



105

Pristine
0.00
0.00
0.00
0.00
0.00
0.10
10.00
0 5 10 15 20 25 30 35
Vol %
C
o
n
d
u
c
t
i
v
i
t
y

(
S
/
c
m
)

OCNF
0.00
0.00
0.00
0.00
0.00
0.10
10.00
0 5 10 15 20 25 30 35
Vol %

C
o
n
d
u
c
t
i
v
i
t
y

(
S
/
c
m
)

CNF-PDMS-NH2
0.00
0.00
0.00
0.00
0.00
0.10
10.00
0 5 10 15 20 25 30 35
Vol %

C
o
n
d
u
c
t
i
v
i
t
y

(
S
/
c
m
)

Figure 5.10 DC Surface electrical conductivity variations with CNF volume concentration
106

5.6 Analytical Calculation
In this section, the conductivity relationship arising from percolation theory is
applied to the composites. As discussed in section 2.5 of this dissertation, the
conductivity is linear with ) (
C
in a logarithmic scale and the relationship is
described by the equation [Stauffer, 1991]:
t
C
A ) ( = (5-11)
where is the conductivity of the composite, is the volume fraction of the CNF in the
composite,
C
is the critical volume fraction (volume fraction at percolation), A and t
are fitted constants. The determination of
C
is based on the value of at which the
slope of ) ( is the greatest.
In order to determine t , the surface conductivity data was fitted to a power law in
terms of volume fraction of nanofibers. Figure 5.11 represents the logarithm of
conductivity as a function of ) ln(
C
corresponding to Figure 5.10.
A best fit to the data in the log-log plot of Figure 5.11 resulted in a value of
A=552.4 S/cm and t=2.67 for pristine made composite when
C
was assumed to be 1.15
vol.%, A=76.1 S/cm and t=3.18 for oxidized fibers composite assuming
C
equal to 3.0
vol.%, and A=7.51 S/cm and t=2.27 for functionalized fibers composite assuming
C
to
be 5.0 vol.%, with a correlation factor R=96.2% for pristine, R=96.4% for oxidized, and
R=98.8% for functionalized fibers. This demonstrates that a good fit was achieved
between the experimental data (squares) and fit function (straight line) and compares
experimental results to the values predicted. Theoretically, the value of the constant A
should approach the conductivity of CNF. The intrinsic electrical conductivity of pristine
107

carbon nanofiber was reported to be 500 S/cm [Tibbetts, 2007] which is close enough to
the above mentioned value of A. The universal value of the conductivity exponent of t ,
theoretically, is predicted to lie from 1.6 to 2.0 [Stauffer, 1991]. Yet, experimental values
have been found in the range from 2 to 4.5 [Balberg, 1987; Carmona et al., 1984; Quivy
et al., 1989; Kanapitsas et al., 2008]. Some significantly higher values of 5.1 [Ezquerra et
al., 1988], 6.27 [Ezquerra et al., 1990], and even 8 [Mamunya et al., 2002], also have
been reported in literature.









108


Pristine
y = 2.6721x + 6.3143
R
2
= 0.9622
-25.00
-20.00
-15.00
-10.00
-5.00
0.00
5.00
-7 -5 -3 -1 1
Ln (-c)
L
n




OCNF
y = 3.1859x + 4.3321
R
2
= 0.9636
-25.00
-20.00
-15.00
-10.00
-5.00
0.00
5.00
-7 -5 -3 -1 1
Ln (-c)
L
n



CNF-PDMS-OH
y = 2.2696x + 2.0159
R
2
= 0.9811
-25.00
-20.00
-15.00
-10.00
-5.00
0.00
5.00
-7 -5 -3 -1 1
Ln (-c)
L
n



Figure 5.11 Percolation equation fit to experimental data (DC conductivity data from Fig. 5.10).
109

5.7 Comparison with Other Filled Composites
We have gathered together a certain number of literature results of percolation
threshold values for different reinforced polymer composites. Table 5.3 presents results
obtained by other investigators.

Table 5.3 Percolation threshold of the collected experimental data
Reference Reinforcement Matrix
Percolation
threshold
(vol%)
Critical
exponent
Balberg [1987] Carbon black Polyethylene 39 6.4
Ezquerra et al.
[1988]
Polypyrole
Polyethylene
Oxide
13.4 1.8
Barton et al. [2007]
Carbon fiber
Fortafil
Vectra 4 n/a
Celzard et al. [1996]
Single crystal
graphite flakes
Epoxy resin 1.3 1.7-2
Kanapitsas et. al.
[2008]
MWCNT Polypropylene 0.6 4.5
Stankovich et al.
[2006]
Graphene Polystyrene 0.1 n/a
Ounaies et al. [2003] SWNT Polyimide 0.05 n/a
Foygel et al. [2005] SWNT CP2 0.05 n/a
Kim et al.[2005] MWNT Epoxy 0.1 n/a

The host matrix has an effect on the threshold through the interactions between
the dispersed reinforcements (e.g. fibers or particles). As can be seen in the table, the
reported values for percolation threshold of carbon naofiber composites are bigger than
those of single-wall or multi-wall carbon naotubes. This is primarily because percolation
threshold is low for the filler material with high aspect ratios. For carbon black
110

composites, the value of percolation threshold is considerably higher than for composites
made with other forms of carbon. Most reports in the literature indicate that at least 8
wt% carbon black filler is needed in order to achieve percolation [Talapatra and Gerhardt,
2006].
In a comparison to the previous works done by the research group at Michigan
Technological University by Rodwick et al [2007], the percolation threshold of carbon
fiber Fortafil/Vecta (~4 vol %) is comparable to that of the carbon nanofiber/PDMS
composites (1.15~5 vol %).
Table 5.3 shows that there are some experimental values of critical exponent t
which are not within the theoretical range.

5.8 Experimental and Computational Comparison
From the values obtained for the critical density of spherocylinders in chapter
four, the critical volume fraction of the system can be calculated to report a direct
comparison between experimental and computational results.
The relation between the critical volume fraction
C
, total excluded volume of the
system
e
V , excluded volume of one object
ex
V , and the volume available for one object
V , can be expressed as:
V
V
V
ex
e
C
= (5-12)
Table 5.4 shows the Monte Carlo results for estimated critical volume fraction of
different systems of spherocylinders. It is noted that the aspect ratio of CNFs can exceed
1000, which is beyond the reach of our simulations. Simulations for the aspect ratios
111

higher than 1000 for soft-core modeling could not be conducted due to the computational
limitations. Also, the simulations for hard-core modeling limited to aspect ratios as high
as 100 for the same reason.

Table 5.4
Computational results for critical
volume fraction of spherocylinders

Aspect
Ratio

C
(vol. %)
Soft core Hard core
10 4.86 15.5
100 0.44 1.5
1000 0.03 -

It is well known that critical volume fraction decreases rapidly with increasing
fiber aspect ratio [Yi and Sastry, 2002] which is confirmed by the results presented here.
Table 5.4 shows that as the aspect ratio of fibers increases, the threshold value for a 3-D
random system decreases monotonically. A critical fractional volume of 0.05% for CP2
composites containing SWNT with the average aspect ratio as high as 1200 has been
reported by Ounaies and colleagues [2003] which is still higher than the threshold value
of 0.03 vol.% obtained for the soft-core spherocylinders with aspect ratio equal to 1000.
Therefore, the developed Monte Carlo simulations successfully show the extremely low
thresholds which a model based on the soft-core approach estimates for fibrous systems.
Simulations of the hard-core model yield significantly higher results. It is expected that
the values for percolation threshold using the hard-core with soft-shell modeling lie
between the data obtained from the previous approaches.
112

Computational studies represent an ideal case. Moreover, real composites are not
always in the form of random distribution of independent fibers. A good comparison
between the simulation results and experiment prediction could, in principle, be tested if
composites of shorter nanofibers can be synthesized and characterized, and the nanofiber
orientations and distribution in the composite experimentally measured. Then
experimental parameters could be incorporated in the computational model. It should be
noted, however, that exact thresholds are only known for certain two dimensional lattices.
Unfortunately, there is not much published data about a whole range of thresholds for
continuum systems.

5.9 Conclusion
In this research, composite samples of electrically conductive CNF-PDMS were
produced utilizing three types of fillers, measured the electrical conductivity, and applied
the analytical model for our carbon fiber composites. The percolation threshold depends
on type, size, shape, surface area, and distribution of the filler particles as well as
processing conditions such as temperature and mixing time. In future experiments, it
seems natural to explore how functioanlizing of the CNF surfaces would change the
nanofiber-nanofiber contact conductivity that dominates the electrical properties of the
composites.
Based on the results obtained in the present chapter, the following main
conclusions can be drawn:
1- The addition of carbon greatly improved the polymer conductivity with a sharp
transition from an electrical insulator to an electrical semiconductor.
113

2- The resulting conductive polymer nanocomposites can be used in commercial
elastomer polymer processes with composite bulk conductivities of 10
-5
-10 S/cm
at various volume percentages of the filler.
3- The CNF pristine composites had excellent electrical performances, with a
conductivities as high as 10 S/cm.
4- The electrical properties of the composites depended very much on the dispersion
of CNF.
5- Improvement the CNF dispersion into the polymer matrix via functionalization,
decreased the electrical properties of the polymer.
6- Loading of filler needed for composite to reach a satisfactory conductivity is
usually as high as 20-30 wt%.
7- The variation in volume electrical conductivity with respect to filler content
reveals that the electrical conducting property of the composites is greatly
affected by the processes of oxidation and functionalization.
8- The percolation threshold of the nanocomposites depends strongly on the different
types of fibers.



114


Chapter 6
Conclusions
6.1 Summary
As noted in chapter one, there is a growing interest in polymer nanocomposites
and development of new applications. A comprehensive picture of the influence of the
material parameters on the percolation threshold of fiber dispersions in composites
remains an ongoing progress. Compared to analytical methods, computer simulations
have been used more successfully to model the percolation phenomenon in composite
materials.
In this study, Monte Carlo studies were performed, aimed at finding a critical
excluded volume associated with the onset of percolation. Alternate models were
analyzed for the purpose of understanding the role of shape and geometry of the
conducting filler in determining the resultant percolation threshold. For Monte Carlo
simulations, tests for percolation were conducted on randomly generated samples under a
set of constraints (particle size, boundary condition, etc.) utilizing a continuum (off-
lattice) approach. Calculations are three dimensions except for two-dimensional circles.
Specifically included in this development is the distribution in size of the particles, since
115

real fiber composites always have a fiber size distribution. The percolation study was also
extended from soft-core models to the hard-core models with soft-shell. The influence of
particle size on percolation behavior, as well as the effect of shell thickness on the
percolation threshold was investigated. The validity of the algorithm developed for the
programming code is confirmed by the agreement of its results for shell thickness in three
dimensions and the predictions for literature data.
The production and application of the carbon-based polymer nanocomposites
were discussed, conductivity of the samples was obtained by a standard four-probe
technique and their physicochemical property changes were evaluated.
The theoretical study of percolation phenomena agreed well with experimental
data obtained for conductive composites and provided some practical suggestions for
maximizing the efficiency of conductive fillers.

6.2 Concluding Remarks
There appears to be no previous study or reference on the effect of particle size
distribution on percolation onset using the hard-core with soft-shell approach in a system
which obeys finite size scaling theory. The weight-average cluster size observed in
composite systems dictates the growth of the clusters, and is affected by the distribution
in fiber lengths which can be seen in the actual composite specimens. Therefore, the fiber
length distribution plays an important role in determining the cluster statistics and
percolation mechanism in these composites. In contrast to most of the cluster statistics
reported in the literature that are for uniform fiber lengths, the one presented here is for
distributed fiber lengths. The originality of this research lies in its having created a new
116

analytical model to evaluate the electrical conductivity properties in nanocomposites
which can be used for different applications in the field of nanotechnology.
From the work accomplished in this study, it was possible to draw several
conclusions concerning the modeling of conductive composites. The main conclusion of
this study is that the finite size scaling hypothesis was established to be applicable to
infinitely large systems of randomly-distributed particles with different shapes and sizes.
The total excluded volume needed for the onset of percolation always decreases
when the degree of randomness increases, as confirmed by the Monte Carlo results.
Increasing the degree of randomness by using objects of different sizes but of the same
shape, lowers the average excluded volume (and the corresponding percolation threshold)
to a value which is system invariant. Another increase in the degree of randomness, by
allowing variable orientation of the objects, causes a variation in the total excluded
volume as well.
A stick-like filler with a higher aspect ratio has an advantage in forming the
conducting network in the polymer matrix over that of a round-shaped filler, either a
circle or sphere, which has a lower aspect ratio. For spherical particles, a smaller particle
size will lower the percolation threshold, while for particles with an aspect ratio greater
than one, larger aspect ratios and a broader range of aspect ratios will lower the
percolation threshold. In a random distribution of positions and orientations, a single rod
can provide connectivity over a longer distance than several spheres with the same total
volume. This means that the polymer matrix can be conductively loaded with a smaller
amount of conducting filler, resulting in a lighter composite. For maximum efficiency,
117

very high aspect ratio fibers are recommended, and these particles will be most effective
when they have random orientations in the composite.
The overall agreement of the experimental and computational results may reflect
the fact that the present approach can be used to study the percolating behavior of
nanocomposites reinforced with any conductive fillers to facilitate a better understanding
of the electrical conductivity of nanocomposites.

6.3 Future Work
The present study further highlighted the importance of theoretical and numerical
analyses in the exploration of percolation in conductive nanocomposites. In addition, it
allowed us to extract some important information concerning the nature of the percolating
networks responsible for the transport phenomena in nanocomposites.
Future work may include the use of this approach to investigate the effect of
functionalization on the onset of percolation in polymer nanocomposites. More
experimental work is needed for making samples of nanocomposites loaded with
spherical conductive particles as well as short fibers in order to achieve a complete
comparison between the simulation results with experiment data.
Significant advancement in modeling gradual microstructrual variations in layers
from the skin to the core observed in composites could be achieved by placing a large
number of boxes, one inside the other, which will provide a more accurate representation
of the real composites over the single box geometry; however, the limitations in the
current computer resources available would have to be overcome.
118

Mechanical experiments on the various filler-matrix combinations are required to
obtain new data about percolating CNF nanocomposite samples, which maintain high
electrical conductivity while improving on the mechanical behavior, an interesting and
important topic in its own right.


































119


References


A
Agari Y., Ueda A., and Nagai S., Electrical and thermal conductivities of polyethylene
composites filled with biaxial oriented short-cut carbon fibers, Journal of Applied
Polymer, Vol. 52, pp. 1223-31, 1994.

Ajayan P. M., Schadler L. S., and Braun P. V., Nanocomposite Science and
Technology, Preface, Weinheim: Wiley-VCH, 2003.

Araujot F. F. T., and Rosenberg H. M., Switching behaviour and DC electrical
conductivity of epoxy-resin/metal-powder composites, Journal of Physics D: Applied
Physics, Vol. 9, pp. 1025-1030,1976.

ASTM D257-99, Standard Test Methods for DC Resistance or Conductance of
Insulating Materials, American Society of Testing Materials, Vol.10.01, 2005.
B
Balberg I., Tunneling and nonuniversal conductivity in composite materials, Physical
Review Letters, Vol. 59, page 1305, 1987.

Balberg I., A comprehensive picture of the electrical phenomena in carbon black-
polymer composites, Carbon 40, pp. 139-143, 2002.

Balberg I., Tunnelling and percolation in lattices and the continuum, Journal of
Physics D: Applied Physics, Vol. 42, No. 6, pp. 1-16, 2009.

Balberg I., and Binenbaum N., Computer study of the percolation threshold in a two-
dimensional anisotropic system of conducting sticks, Physical Review B, Vol. 28, No. 7,
page 3, 1983.

120

Balberg I., and Binenbaum N., Invariant properties of the percolation thresholds in the
soft-core hard-core transition , Physical Review A, Vol. 35, No. 12, pp. 5174-5177,
1987.

Balberg I., Binenbaum N., and Wagner N., Percolation thresholds in the three-
dimensional sticks system, Physical Review Letters, Vol. 52, No. 17, pp. 1465-1468,
1984.

Balberg I., Anderson C. H., Alexander S., and Wagner N., Excluded volume and its
relation to the onset of percolation, Physical Review B, Vol. 30, No. 7, pp. 3933-3943,
1984.

Baraton M. I., Synthesis, functionalization and surface treatment of nanoparticles,
American Scientific Publishers, 2003.

Barton R. L., Keith J. M., and King J. A., Electrical conductivity model evaluation of
carbon fiber filled liquid crystal polymer composites, Journal of Applied Polymer
Sceince, Vol. 106, pp. 2456-2462, 2007.

Benoit J. M., Corraze B., and Chauvet O., Localization, Coulomb interactions, and
electrical heating in single-wall carbon nanotubespolymer composites, Physical
Review B, Vol. 65, No. 24, 241405(R), 2002.

Berhan L., and Sastry A. M., Modeling percolation in high-aspect-ratio fiber systems.
I. Soft-core versus hard-core models Physical Review E, Vol. 75, 041120, 2007.

Berhan L., and Sastry A. M., Modeling percolation in high-aspect-ratio fiber systems.
II. The effect of waviness on the percolation onset, Physical Review E, Vol. 75, 041121,
2007.

Brankov J. G., Danchev D. M., and Tonchev N. S., Theory of critical phenomena in
finite-size systems: Scaling and quantum effects, Series in Modern Condensed Matter
Physics, Vol. 9, World Scientific, 2000.

BIPM, Brochure The International System of Units, 8
th
edition, 2006.

Bug A. L. R., Safran S. A., and Webman I., Continuum percolation of rods, Physical
Review Letters, Vol. 54, No. 13, 1985.
C
Carmona F., Prudhon P., Barreau G., Solid State Commun. Vol. 51, page 255, 1984.

Celzard A., McRae E., Mareche J. F., Furdin G., Dufort M., and Deleuze C.,
Composites based on micron-sized exfoliated graphite particles: Electrical conduction,
121

critical exponents and anisotropy, Journal of Physics and Chemistry of Solids, Vol. 57,
Nos. 6-8, pp. 715-718, 1996.

Chan J., EECS 143 Microfabrication Technology Lab Manual, 1994.

Chisholm N., Mahfuz H., Rangari V. K., Ashfaq A., and Jeelani S., Fabrication and
mechanical characterization of carbon/SiC-epoxy nanocomposites, Composite
Structures, Vol. 67, pp. 115-124, 2005.

Clingerman M. L., Weber E. H., King J. A., and Schulz K. H., Development of an
additive equation for predicting the electrical conductivity of carbon-filled composites,
Journal of Applied Polymer Science, Vol. 88, No. 9, pp. 2280-2299, 2003.

Clingerman M. L., King J. A., Schulz K. H., and Meyyers J. D., Evaluation of
electrical conductivity models for conductive polymer composites, Journal of Applied
Polymer Science, Vol. 83, pp. 1341-1356, 2002.

Compton R. G., Electron tunneling in chemistry: chemical reactions over large
distances, Amsterdam: Elsevier, page 123, 1989.

Consiglio R., Baker D. R., Paul G., and Stanley H. E., Continuum percolation
thresholds for mixtures of spheres of different sizes, Physica A 319, pp. 49-55, 2003.

Corwin L. J. and Szczarba, R. H., Calculus in vector spaces, CRC Press, 1994.

Coso R. D., Requejo-Isidro, J., Solis J., Gonzalo J., and Afonso C. N., Third order
nonlinear optical susceptibility of Cu:Al2O3 nanocomposites: From spherical
nanoparticles to the percolation threshold, Journal of Applied Physics, Vol. 95, No. 5,
pp. 2755-2762, 2004.
D
Dani A., and Ogale A. A., Percolation in short-fiber composites: cluster statistics and
critical exponents, Composite Science and Technology, Vol. 57, pp. 1355-1361, 1997.

Debondt S., Feroyen L., and Deruyttere A., Electrical conductivity of composites: A
percolation approach, Journal of Material Science, Vol. 27, pp. 1983-1988, 1992.

Du F. M., Scogna R. C., Zhou W., Brand S., Fischer J. E., and Winey K. I.,
Macromolecules, Vol. 37, 9048, 2004.
E
Ezquerra T. A., Moammadi M., Kremer F., Viligis T., and Wegner G., On the
percolative behaviour of polymeric insulatorconductor composites: polyethylene-oxide-
polypyrrole, Journal of Physics C, Solid State Physics, No. 21, pp. 927-941, 1988.
122


Ezquerra T. A., Kulescza M., Cruz C. and Calleja F. J., Advanced Materials, Vol. 2,
page 597, 1990.
F
Feng D. and Jin G., Introduction to condensed matter physics, Vol. 1, World Scientific,
page 107, 2005.

Fiske T., Gokturk H., and Kalyon D. M., "Percolation in Magnetic Composites",
Journal of Materials Science, Vol. 32, pp. 5551-5560, 1999.

Flandin L., Chang A., Nazarenko S., Hiltner A., and Baer E., Effect of strain on the
properties of an ethylene-octene elastomer with conductive carbon fibers, Journal of
Applied Polymer Science, Vol. 76, pp. 894-905, 2000.

Foygel M., Morris R. D., Anez D., French S., and Sobolev V. L., Theoretical and
computational studies of carbon nanotube composites and suspensions: Electrical and
thermal conductivity, Physical Review B, Vol. 71, pp. 104201-8, 2005.

Foulger S. H., Electrical properties of composites in the vicinity of the percolation
threshold, Journal of Applied Polymer Science, Vol. 72, pp. 1573-15-82, 1999

Frost P., Solid geometry, London: Macmillan, 1886.

Fu S. Y., Feng X. O., Lauke B., and Mai Y. W., Effects of particle size,
particle/matrix interface adhesion and particle loading on mechanical properties of
particulate-polymer composites, Composites B, Vol. 39, pp. 933961, 2008.
G
Gawlinski E. T., and Stanley H. E., Continuum percolation in two dimensions: Monte
Carlo tests of scaling and universality for non-interacting discs, Journal of Physics A:
Math. Gen. Vol. 14, pp. L291-9, 1981.

Gay D., Hoa S. V., and Tsai S. W., Composite materials: design and applications, CRC
Press, 2002.

Gelest Inc. Catalogue, Reactive silicones, page 44, 2007.

Grimmett G., Percolation, second edition, Berlin; New York: Springer, 1999.

Grosberg A., Theoretical and Mathematical Models in Polymer Research, Academic
Press, 1st edition, 1998.

123

Grujicic M., and Cao G., A computational analysis of the percolation threshold and the
electrical conductivity of carbon nanotubes filled polymeric materials, Journal of
Materials Science Vol. 39, pp. 4441-4449, 2004.
H
Haan S. W., and Zwanzig R., Series expansions in a continuum percolation problem,
Journal Phys. A: Math. Gen., Vol. 10, No. 9, pp.1547-1555, 1977.

Hammersley J. M., Handscomb D. C., Monte Carlo methods, London: Methuen, 1975.

Hunt A., Percolation theory for flow in porous media, New York: Springer, 2005.
J
Jain S., Monte Carlo simulations of disordered systems, World Scientific, 1992.

Janzen J., On the critical conductive filler loading in antistatic composites, Journal of
Applied Physics, Vol. 46, No. 2, pp. 966-969, 1974.

James E. M., Physical properties of polymers handbook, page 451, Table 25.4, N.Y.:
AIP Press, 1996.

Jeong W., and Kessler M. R., Toughness enhancement in ROMP functionalized
carbon nanotube/polydicyclopentadiene composites, Chem. Mater., Vol. 20, No. 22, pp
7060-7068, 2008.

Jing X., Zhao W., and Lan L.., The effect of particle size on electric conducting
percolation threshold in polymer/conducting particle composites, Journal of Materials
Science Letters, Vol. 19, pp 377-379, 2000.
K
Kanapitsas A., Logakis E., Pandis C., Zuburtikudis I., Pissis P., Delides C. G., and
Vatalis A. S., Dielectric and Thermomechanical Properties of Polypropylene/Multi-
Walled Carbon Nanotubes Nanocomposites, Materials Research Society Symposium
Proceeding, Vol. 1056, HH11-39, 2008.

Keithley J. F., Yeager J. and Hrusch-Tupta M. A., Keithleys Low Level
Measurements handbook, Fifth Edition, Keithley Instruments, Inc., pp.3-18,19, 1998.

Kim Y. J., Shin T. S., Choi H. D., Kwon J. H., Chung Y., and Yoon H. G.,

Electrical
conductivity of chemically modified multiwalled carbon nanotube/epoxy composites,
Carbon, Vol. 43, Issue 1, pp. 23-30, 2005.

124

Kortschot M. T., and Woodhams R. T., Computer simulation of the electrical
conductivity of polymer composites containing metallic fillers, Polymer composites,
Vol. 9, No. 1, pp 60-71, 1988.

Kyrylyuk A. V. and Schoot P., Continuum percolation of carbon nanotubes in
polymeric and colloidal media, PNAS, Vol. 105, No. 24, pp. 8221-8227, 2008.
L
Li J., and Kim J., Percolation threshold of conducting polymer composites containing
3D randomly distributed graphite nanoplatelets, Composite Science and Technology,
Vol. 67, pp. 2114-2120, 2007.

Luk'yanchuk L.A., and Mezzane, D., Smart Materials for Energy, Communications
and Security, page 172, Dordrecht: Springer, 2008.
M
Ma C. M., Huang Y., Kuan H. C., Chiu Y. S., Preparation and electromagnetic
interference shielding characteristics of novel carbon-nanotube/siloxane/poly(urethane)
nanocomposites, Journal of Polymer Science B: Polymer Physics, Vol. 43, No. 4, pp.
345-358, 2005.

Mamunya Y. P., Davydenko V. V., Pissis P.,

and Lebedev E. V., Electrical and
thermal conductivity of polymers filled with metal powders, European Polymer Journal
Vol. 38, pp. 1887-1897, 2002.

Manhart1 J., Kunzelmann K. H. , Chen H. Y., and Hickel R., Mechanical properties
and wear behavior of light-cured packable composite resins, Dental Materials, Vol. 16,
pp. 3340, 2000.

Mapkar J. A., Belashi A., Berhan L. M., and Coleman M. R., Mechanical and
electrical properties of high loading functional carbon nanofiber sheets, submitted to
Composite Science and Technology, 2011.

Mark J. E., Physical properties of polymers handbook, New York: Springer, 2006.

McCullough, R. L., Generalized combining rules for predicting transport properties of
composite materials, Composite Science and Technology, Vol. 22, pp. 3-21, 1985.

Meester R., and Roy R., Continuum percolation, Cambridge Univ. Press 1996.

Metropolis N., "The beginning of the Monte Carlo method", Los Alamos Science, pp.
125-130, 1987.

125

Munson-McGee S. H., Estimation of the critical concentration in an anisotropic
percolation network, Physical Review B , Vol. 43, No 4, pp. 3331-3336, 1991.
N
Nagata K., Iwabuki H., and Nigo H., Effect of particle size graphites on electrical
conductivity of graphite/polymer composite, Composites Interfaces, Vol. 6, No. 5, pp.
483-495, 1999.

Nakamura S. Saito K., Sawa G., and Kitagawa K., Percolation threshold of carbon
black-polyethylene composites, Jpn. Journal of Applied Physics, Vol. 36, No. 8, pp.
5163-5168, 1997.

Natsuki T., Endo M., and Takahashi T., Percolation study of orientated short-fiber
composites by a continuum model Physica A 352, pp. 498-508, 2005.

Neda Z., Florian R., and Brechet Y., Reconsideration of continuum percolation of
isotropically oriented sticks in three dimensions Physical Review E, Vol. 59, No. 3, pp.
3717-3719, 1999.

Newman M. E. J., and Ziff R. M., A fast Monte Carlo algorithm for site or bond
percolation, Physical Review E, Vol. 64, pp. 1-16, 2001.

Northrup E. F., Methods of Measuring Electrical Resistance, New York, McGraw-Hill,
1912.
O
Ogale A. A., and Wang S. F., Simulation of the percolation behavior of quasi- and
transversely isotropic short-fiber composites with a continuum model, Composites
Science and Technology, 46, pp. 379-388, 1993.

Onsager L., and Ann. N.Y, Acad. Sci. Vol. 51, page 627, 1949.

Ounaies Z., Park C., and Wise K. E., Electrical properties of single wall carbon
nanotube reinforced polyimide composites, Composite Science and Technology, Vol.
63, pp.1637-1646, 2003.
P
Pike G. E., and Seager C. H., Percolation and conductivity: A computer study. I,
Physical Review B, Vol. 10, No. 4, pp. 1421-1434, 1974.

Powell M. J., Site percolation in randomly packed spheres, Physical Review B, Vol.
20, No. 10, pp. 4194-4198, 1979.

126

Psarras G. C.,Hopping conductivity in polymer matrixmetal particles composites,
Composites A, Vol. 37, pp. 1545-1553, 2006.
Q
Quivy A., Deltour R., Jansen A. G. M., and Wyder P., Transport phenomena in
polymer-graphite composite materials, Physical Review B, Vol. 39, No. 2, page 1026,
1989.
R
Ramasubramaniam R., Chen J., and Liu H.Y., Homogeneous carbon
nanotube/polymer composites for electrical applications, Applied Physics Letters; Vol.
83, No.14, pp. 2928-30, 2003.

Rodwick L. B., Jason M. K., and Julia A. K., Electrical Conductivity Model
Evaluation of Carbon Fiber Filled Liquid Crystal Polymer Composites, Journal of
Applied Polymer Science, Vol. 106, pp. 2456- 2462, 2007.

Rubin Z., Sunshine S. A., Heaney M. B., Bloom I., and Balberg I., Critical behavior
of the electrical transport properties in a tunneling-percolation system, Physical Review
B, Vol. 59, No. 19, pp. 12196-12199, 1999.

Ruschau G. R., and Newnham R. E., Critical volume fractions in conductive
composites, Journal of Composite Materials, Vol. 26, No. 18, pp. 2727-2735, 1992.

Ruschau G. R., Yoshikawa S., and Newnham R. E., "Resistivities of conductive
composites", Journal of Applied Physics, Vol. 72, No. 3, pp. 953-959, 1992.
S
Sahimi M., Applications of percolation theory, CRC Press, 1994.

Sahimi M., Heterogeneous Materials: Linear transport and optical properties, New
York; London : Springer, 2003.

Sevick E. M., Manson P. A., and Ottino J. M., Monte carlo calculations of cluster
statistics in continuum models of composite morphology, Journal of Chemical Physics,
Vol. 88, pp. 1198-1206, 1988.

Smits F. M., Measurements of sheet resistivity with the four-point probe, BSTJ, Vol.
37, page 711, 1958.

Stankovich S., Dikin D. A., Dommett G. H. B., Kohlhaas K. M., Zimney E. J., Stach
E. A., Piner R. D., Nguyen S. T., and Ruoff R. S., Graphene-based composite
materials, Nature Vol. 442, pp. 282-286, 2006.
127


Stauffer D., and Aharony A., Introduction to percolation theory, Taylor & Francis,
London, second edition, 1991.

Stinchcombe R. B., and Watson B. P., Renormalization group approach for
percolation conductivity, Journal of Physics, C: Solid State Physics, Vol. 9, pp. 3221-
3247, 1976.

Sun Y., Bao H. D., and Guo, Z. X., and Yu J., Modeling of the Electrical Percolation
of Mixed Carbon Fillers in Polymer-Based Composites, Macromolecules, Vol. 42, pp.
459-463, 2009.
T
Taipalus R., Harmia T., Zhang M. Q., and Friedrich K., The electrical conductivity
of carbon-fibre-reinforced polypropylene/polyaniline complex-blends: experimental
characterisation and modeling, Composites Science and Technology, Vol. 61, pp. 801-
914, 2001.

Talapatra S., and Gerhardt R. A., Optimization of the Electrical Conductivity of ABS
Nanocomposites filled with Carbon Black and Carbon Nanotubes, Materials Research
Society Symposium Proceedings Vol. 977, pp. 126-131, 2006.

Tibbetts G. G., Lake M. L., Strong K. L., Rice B. P., A review of the fabrication and
properties of vapor-grown carbon nanofiner/polymer composites, Composites Science
and Technology, 2007, 67, 1709.

Tuncer E., Serdyuk Y. V., and Gubanski S. M., Dielectric mixtures- electrical
properties and Modeling, IEEE Transactions on Dielectrics and Electrical Insulation,
Vol. 9, pp. 809-28, 2002.
U
Utracki L. A., Polymer blends handbook, Boston: Kluwer Academic Publishers, page
80, 2002.
V
Valdes L. G., Resistivity Measurements on Germanium for Transistors, Proc. I.R.E.,
Vol. 42, pp. 420-427, 1954.

Vanderzande C., Lattice models of polymers, Cambridge: Cambridge Univ. Press, 1998.


128

W
Wang S. F., and Ogale, A. A., Continuum space simulation and experimental
characterization of electrical percolation behavior of particulate composites, Composites
Science and Technology, Vol. 46, pp. 93-103, 1993.

Weber M. and Kamal M. R., Estimation of the volume resistivity of electronically
conductive composites, Polymer Composites, Vol. 18, No. 6, pp. 711-725, 1997.
X
Xing H., Sun, L., Song G., Gou J., and Hao Y. M.,

Surface coating of carbon
nanofibers / nanotubes by electrode position for multifunctionalization, Nanotechnology,
Vol. 19, 2008.
Y
Yi Y. B., and Sastry A. M., Analytical approximation of the percolation threshold for
overlapping ellipsoids of revolution, Proc. R. Soc., London, A 460, pp. 2353-2380,
2004.

Yi Y. B., and Sastry A. M., Analytical approximation of the two-dimensional
percolation threshold for fields of overlapping ellipses, Physical Review E, Vol. 66,
066130, 2002.
Z
Zou J., Yu Z., Pan Y., Fang X., and Ou Y., Conductive mechanism of
polymer/graphite conducting composites with low percolation threshold, Journal of
Polymer Science B: Polymer Physics, Vol. 40, pp. 954-963, 2002.



129


Appendix


The complete program codes are provided in the MATLAB 7.6.0 programming
language that performs the algorithm implementation of the continuum percolation on a
linear cell with periodic boundary conditions. The algorithm to apply the actual code is
quite brief. It performs exactly as described in Figure A.1. The function of boundary
condition is used to find out if any of the objects crosses each of the adjacent boundaries
of the cell. The detailed formulation of the algorithm is sufficiently obvious and is
omitted here.











130





Figure A.1 Flow chart of network generation



Read Network Parameters:
Fiber and particle sizes such as
diameter and length, and center
Generate Network
Inside Unit Cell
Enforce Boundary Condition on Cell:
Add fibers if they cross over boundaries

Network Addition:
Add fibers for cluster formation
Circuit Analysis

You might also like