Citation
Investigation and Characterization of the Dispersion of Nanoparticles in a Polymer Matrix by Scattering Techniques

Material Information

Title:
Investigation and Characterization of the Dispersion of Nanoparticles in a Polymer Matrix by Scattering Techniques
Copyright Date:
2008

Subjects

Subjects / Keywords:
Agglomerates ( jstor )
Composite particles ( jstor )
Light scattering ( jstor )
Nanocomposites ( jstor )
Nanoparticles ( jstor )
Particle density ( jstor )
Particle size classes ( jstor )
Particle size distribution ( jstor )
Polymers ( jstor )
Solvents ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
7/12/2007

Downloads

This item has the following downloads:


Full Text












INVESTIGATION AND CHARACTERIZATION OF THE DISPERSION OF
NANOPARTICLES IN A POLYMER MATRIX BY SCATTERING TECHNIQUES















By

MARIE CHRISTINE KISSINGER-KANE


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2007

































Marie Christine Kissinger-Kane 2007

































To my husband, Kevin















ACKNOWLEDGMENTS

I would first like to acknowledge my husband, Kevin, and my parents for their

constant love and support. Without them, I am sure this would have been much more

difficult to accomplish. Secondly, I would like to thank my advisor, Dr. Brij Moudgil

and my committee co-chair, Dr. Anthony Brennan. Their knowledge and guidance have

been invaluable tools throughout my graduate school journey. Dr. Moudgil, I thank you

for thoroughly preparing me for my future career and making me the confident person

that I am today. Dr. Brennan, I thank you for your unending encouragement and for not

letting me take myself too seriously. I would like to also acknowledge the rest of my

graduate committee for their guidance and support: Dr. Laurie Gower, Dr. Kevin Powers,

Dr. David Tanner, and Dr. Charles Beatty. My special thanks to Dr. Powers; his "out of

the box" ideas and countless discussions with me, academic or not, fueled my passion for

learning and discovery.

I would also like to acknowledge those outside of the University of Florida that

have provided assistance. Dr. Rick Beyer at the Army Research Labs at Aberdeen

Proving Ground provided the small angle x-ray measurements, and Dr. J. David Londono

at DuPont assisted with ultra small angle x-ray experiments. Their collaboration has been

an invaluable tool and I would now consider myself among a small elite group of x-ray

scattering specialists.

Next, I would like to acknowledge my best friend and colleague, Sara Jensen

(Homeijer), for her friendship, encouragement and assistance (patience) with TEM









micrographs. Next, I would like to acknowledge the faculty and staff of the Particle

Engineering Research Center. Especially, I would like to thank Mr. Gary Scheiffele for

laboratory help and Gill Brubaker for helpful discussions on Mie Theory. I would also

like to thank my friends and colleagues at the Particle Engineering Research Center for

their laughter, friendship, and helpful discussions.

The Particle Engineering Research Center (PERC) at the University of Florida, the

National Science Foundation (NSF Grant EEC-94-02989), the United States Air Force,

and the Industrial Partners of the PERC are graciously acknowledged for financial

support of this research.

Finally, I would like honor the memory of my late committee member, Dr. Abbas

Zaman, who took the time to show me the value of rheology measurements and always

had a smile and kind word for me. He will be sorely missed.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................. ............. .............. ... ..... .. ............ .. ix

LIST OF FIGURES ............................... ... ...... ... ................. .x

ABSTRACT ........ .............. ............. ...... ...................... xiv

CHAPTER

1 INTRODUCTION ............... ................. ........... ................. ... ..... 1

2 LITERA TURE REVIEW .......................................................... ..............5

Particle D ispersion Studies ........................................................... .............5
Particle Surface M odification.......................................... ........... ............... 6
M echanical Dispersion ........................ ....... .. ..... .............. ..7
D ispersion Equations .................. .. .... .. ...... ............... .................. ... .9
Correlation of Dispersion with M material Properties................................................. 12
M ech anical P rop erties ........................................... ........................................ 12
O ptical D ensity ................... ............................................................... 13
Molecular Modeling of Nanocomposite Dispersion................................................13
L eight Scattering ....................................................................... 15
B ackgrou n d ...................................... ............................. ................. 15
Particle Size M easurement .......... ......... ....................................... ............ 17
Small Angle and Ultra Small Angle X-ray Scattering..............................................18
X ray D ata A analysis ......................................... ...... ............. ........ 19
T he G uinier regim e .............................................. ................................20
The pow er-law regim e ........................................ ........................... 21
A U unified Fit m odel ... ............................................. ........ .. .......... 22
The Porod regime .............................. ......... .... ........ ... ........22
Scattering Techniques for Dispersion Measurements......................................23
M icro sco p y ......................................... ... ...... ..................... ................ 2 4
ASTM Standard for Dispersion of Carbon Black in Rubber............................. 24
Scope of P proposed R research .......................................... .......... ........ .....................27

3 M ATERIALS AND M ETHODS ........................................ ......................... 29









M materials ....................................................... 2 9
N anoparticles .............................................................................................. 29
P olym er M atrix ............................. ........ ...... ........................ ........... ... 32
Particle Surface Treatment- Silane Coupling Agent .......................................34
M eth o d s .................... ......................................3 5
N anocom posite Form ulation ........................................ .......................... 35
Four levels of particle dispersion .................... .................. ............... 36
Solvent selection for fractionated sample ..............................................44
Particle agglom erate sedim entation .................................. ............... 48
O optical B ench ......................................................................49
O ptical com ponents .......... .............. ........ ......................... .......... 49
Detector Selection .......................................... ................... 50
L aser calib ration ........ ..... .. ...... .................................... .. ......... .. .. .. 5 1
Instrum ent calibration ............................................................................ 52
Snell's L aw C orrection........................................... .......................... 53
Scattering M easurem ents.................................. ....................................... 55
Light scattering (3 [tm < d< 90 nm) ........ ..... ............ ............56
Ultra small angle x-ray scattering (2 [tm < d< 80 nm)............................58
Small angle x-ray scattering (100 nm < d< 1 nm)....................................59
Scattering data analysis with IGOR............... ................ .......... 60
Additional Characterization................... ......... ........................... 61
Optical density................. ................ ........61
Commercial light scattering instruments............................... ............... 62
Transmission electron microscopy.......... ................. ........ .......... 63
Scanning electron m icroscopy ........................................ ............... 64

4 SCATTERING THEORY .......................................................... ............... 65

G general Scattering Theory .......................................................... ............... 65
Boundary Conditions......................................... 67
The A m plitude M atrix ............................................................ ............... 68
T h e M u eller M atrix ..................................................................... ..................70
Scattering of a Sphere............. .................................. ........ ........ .... .......... 7 1
A ngular D dependent Scattering ........................................ ........................ 71
T h e L length S cale ................................................................72
M ie T theory .........................................73
M ie Scattering C oefficients .............................................. .............................. 74

5 RESULTS AND DISCSU SSION ......................................... .......................... 76

D ispersion Force Calculations.................................................. ........... ............... 76
Qualitative Nanoparticle Dispersion Characterization .............................................78
D igital Im ages ................................................... ................. 78
Freeze Fracture SEM M icrographs ........................................ ............... 79
TEM M icrographs ....................... .............. ............. ........... .... 81
Q uantitative M measure of D ispersion ................................... ..................................... 84
Rationale for the Dispersion Index ....................................... ...............85









Surface Chem istry Param eter- D ............................... .. ....................... 85
D1 Calculation for the Nanocomposite Samples ...........................................88
M ixing P aram eter- D 2............................................................................ ... .... 89
D2 Calculation for the Nanocomposite Samples ...........................................92
Scattering T echniques............................................................... ..... ...............94
Small Angle X-Ray Scattering (SAXS) ................................... .................96
Ultra Small Angle X-Ray Scattering (USAXS) ...............................................97
Static Light Scattering (SL S)......................................... .......................... 99
Scattering D ata A analysis ............................................................................100
D ispersion Param eter Calculations...................................................... ................. 105
Relationship Between the Dispersion Parameters and Material Properties..............109

6 CONCLUSIONS AND FUTURE WORK........................................................113

N anocom posite F abrication ...................................................................................... 113
Suggested Future W ork- FIB ........................................................ ............. ..114
Characterization Techniques ............................................................................ 115
Suggested Im provem ents.............................................................. ............... 116
O p tic al b en ch ...................................... ...................... ............... 1 16
D ata an aly sis ................................................................. ............... 1 17
A additional characterization studies ...........................................................117
The D ispersion Index.............................. .............. .... ... ....... .. ........ .... 118
Correlating the Dispersion Index to Material Properties .............. .... ...............119
Suggested Future W ork .................................... .................................................. 119

APPENDIX

A CALCULATION OF THE PARTICLE SIZE DISTRIBUTION BY SOLVING
THE INVERSE M IE PROBLEM ........................................ ........................ 121

B PROCEDURES FOR SCATTERING DATA ANALYSIS USING IRENA ..........123

D esm hearing of U SA X S D ata ......................................................... ..................... 123
Combine U SAX S and SAX S Data ...................................... ......... ........ ....... 123
Calculate Particle Size Distribution and FWHM Values ......................................124
Correlation Length for Interparticle Distance............................. 124
Calculation of Primary Particle Distribution by SAXS ............................... 124

C DERIVATION OF THE CALCULATION OF INTERPARTICLE DISTANCE
FOR THE IDEAL DISPERSION CASE ...................................... ............... 126

LIST OF REFEREN CES .................................................................. ............... 133

BIO GR A PH ICA L SK ETCH .................................... ........... ......................................138
















LIST OF TABLES


Table page

2-1 Several cases of mass-fractal power law scattering .............................................22

3-1 M material properties of the epoxy resin .............................. ....................... .......... 33

3-2 Optical properties of the cured epoxy polymer ...................................................34

3-3 Summary of processing techniques used to fabricate each nanocomposite sample
with varied levels of particle dispersion .... ........... ........ .......................... 36

3-4 Summary of components used to fabricate nanocomposite samples for
undispersed, ultrasonicated, and surface treated samples................................... 39

3-5 Summary of components used to fabricate the fractionated sample........................42

3-7 Stokes Law particle settling calculations based on ten minutes centrifuge time.....49

3-8 Optical components of the light scattering instrument, the manufacturer and part
num ber ............ .. ......... .............. ............. .............. ......... 50

5-1 Surface energy values for the nanocomposite components and the calculated
spreading coefficient for untreated and treated particles..............................77

5-2 Reference for the length scale probed and the range of q values for each
tech n iqu e. ......................................................... ................ 9 5

5-3 Values of T, FWHM, and d and the results of the calculation of D1 and D2.......... 107

5-4 Dispersion Index values compared to the processing methodology where M =
mixing, U = ultrasonication, ST = surface treatment, and S = solvent ................108















LIST OF FIGURES


Figure pge

2-1 Diagram of the hierarchical structures of particles that can be probed with light
and x-ray scattering measurements at different length scales, or q values ..............19

2-2 Transmission electron micrographs of PC-MWNT composites: (a) 5 wt.%
MWNT, (b) 5 wt.% MWNT, (c) 2 wt.% MWNT (d) 1 wt.% MWNT...................27

3-1 Scanning electron micrograph of the Degussa P25 TiO2 powder............................30

3-2 Scanning electron micrograph of dispersed TiO2 primary particles and small
aggregates imaged on carbon tape (background). .................. ....................31

3-3 Small angle x-ray scattering (SAXS) derived particle size distribution of
Degussa P25 titania primary particles .......... ................................ ............... 31

3-4 Log-normal particle size distribution of the Degussa P25 titania particles
dispersed in alcohol media, measured by dynamic light scattering .......................32

3-5 Agglomerated log-normal particle size distribution measured by the dry powder
module of the Coulter LS 13320 laser diffraction instrument .............................32

3-6 Structure of DGEBA and the repeat unit. The reactive epoxide groups are
highlighted in red. .....................................................................33

3-7 Structure of Jeffamine D230 and the repeat unit ....................................................33

3-8 Diagram showing how an amine functionalized hardener attached the reactive
epoxide groups and cross-links the epoxy chains to form a thermoset material......34

3-9 Schematic showing the two silane coupling agents used in this study: APTES
(A ) and A E A P T M S (B )......................................... .............................................35

3-10 Dry titania nanopowder in the isopropanol and acetone solvent mixture, showing
good wetting of the particle surface. ........................... ...... ..................... 43

3-11 Particle size distribution measured after ultrasonication of the dry nanopowder
in a solvent m mixture. ....................... .... ................ ................... .. ..... 43

3-12 Digital image of the dispersed particles in the solvent mixture after
sedimentation and removal of the hard agglomerates ............................................43









3-13 Particle size distribution measured after sedimentation and removal of hard
agglom rates in the solvent mixture ........................... .......... ............. .................. 44

3-14 Particle size distribution measured after aging of the surface treated
nanoparticles in the solvent mixture to verify that the particles remained
dispersed.............................................................................................. 44

3-15 Beer's Law relation for optical density of titania nanoparticles dispersed in
acetone and isopropanol, measured at 600nm wavelength. ....................................47

3-16 Digital pictures of (A) 7g/L TiO2 nanoparticles in pure acetone where
sedimentation has occurred due poor wettability and (B) 7g/L TiO2
nanoparticles in 60% acetone/40% isopropanol with improved particle wetting. ...47

3-17 Schematic representation (A) and digital picture (B) of the optical bench set-up
for light scattering m easurem ents. ........................................ ....................... 51

3-18 Intensity of the incident beam over 8.3 minutes with measurements taken every
5 seconds. The error is assumed to be negligible (-1%) .................... ..................52

3-19 Scattering intensities for 600 nm (A), 800 nm (B), 2 [tm (C)and 5 [tm (D)
diameter polystyrene particles dispersed in water............... .................. ...........54

3-20 Example of a theoretical scattering curve generated by the MiePlot software for
200 nm (radius) TiO2 particles in epoxy with a 10% polydispersity and log-
norm al size distribution .............................................. .. ........ .. ...... ............58

3-21 Custom-made sample holders for measuring the optical density of solid polymer
film s in the UV-Vis spectrom eter. ........................................ ....................... 62

3-22 Polyethylene embedding molds used for TEM polymer sample preparation. .........64

4-1 Forward scattering of a sphere showing the electromagnetic field (wave) vectors
for the incident, particle, and scattered radiation. ............. ..................................... 67

4-2. Schematic depicting the scattering of electromagnetic radiation by a spherical
particle. The incident field is in the z-axis direction. ....... ............... ............... 68

4-3 Diagram of light incident from the left with wave vector ki scattering from an
element at r toward the detector with scattering wave vector ks, at a scattering
an g le 0 ........................................................... ................ 7 3

5-1 Examples of digital images comparing thick films of each dispersion state
against a typewritten background.................................................. ......... ...... 79

5-2 Scanning electron micrographs of the undispersed sample showing the large
particle agglomerates and voids due to poor particle wetting by the epoxy............80









5-3 Scanning electron micrographs of the fractionated sample showing the small
particle agglomerates and individually dispersed particles ....................................81

5-4 Transmission electron microscopy images of 0.5 wt% TiO2 embedded in epoxy:
undispersed (A), ultrasonicated (B) and surface treated (C). ................................83

5-5 High resolution TEM micrographs of the fractionated sample (0.5 wt%)
showing the nano scale dispersion and separated individual particles...................84

5-6 Schematic of the four simulated particle agglomerate morphologies used to
determine the validity of the D1 dispersion index. ........... ....................... 87

5-7 Graphical representation of the four simulated cases of particle or agglomerate
size distribution .................................................... ................. 87

5-8 Surface chemistry parameter D1 values of four simulated size distributions..........88

5-9 Schematic of the four simulated particle agglomerate morphologies used to
determine the validity of the D2 mixing parameter............................ ..............91

5-10 D2 mixing parameter values of four simulated particle morphologies...................92

5-11 Figure representing a typical scattering intensity plot and the various length
scales probed by different scattering techniques.............................................. 96

5-12 Small angle x-ray scattering intensity plot for three differently dispersed
sam ples. ..............................................................................97

5-13 Ultra small angle scattering intensity plot showing the four differently dispersed
samples. This q range covers the length scale of approximately 100 nm to 2000
n m ................... .......................................................... ................ 9 8

5-14 The region 0.001 < q < 0.01, revealing the changes in the USAXS scattering
curve due to changes in agglomerate size. .................................... .................99

5-15 Scattering intensity plot including the static light scattering data for the four
differently dispersed sam ples. ...........................................................................100

5-16 Particle (primary, aggregate, or agglomerate) size distribution as obtained from
data analysis of x-ray scattering data (except dry powder obtained from SLS).....102

5-17 Average particle (primary, aggregate, or agglomerate) size shown for the data
obtained by USAXS measurements (except primary particle size obtained from
S A X S ). ......................................................................... 10 3

5-18 Particle size distribution as obtained from data analysis of static light scattering
data, except primary particles (SAXS) and dry powder (DLS). ..........................104









5-19 A comparison between the average particle (aggregate or agglomerate) sizes
obtained from x-ray measurements (USAXS) and static light scattering
m easurem ents (SL S). ..................... ...... ............ .......................... 105

5-20 Optical density (amount of light transmitted through the sample) at 700 nm
wavelength visible light for each differently dispersed sample (0.5 wt% TiO2 in
epoxy) as compared to a pure epoxy background. .......... ............ .................111

5-21 A strong linear correlation can be made between the D1 dispersion index
component and optical density values (measured at 700 nm).............................112

6-1 Initial attempts at using the FIB technique to three dimensionally image the
nanocom posite sam ples created in this study....................................................... 115

C-l A typical positive-skewed lognormal distribution curve is shown (A). The
experimental particle size distribution of TiO2 particles disbursed in ethanol and
measured with dynamic light scattering is also shown (B)..................................126

C-2 A two-dimensional schematic drawing of the randomly well-dispersed particle
configuration. .......................................................................132











Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

INVESTIGATION AND CHARACTERIZATION OF THE DISPERSION OF
NANOPARTICLES IN A POLYMER MATRIX BY SCATTERING TECHNIQUES

By

Marie Christine Kissinger-Kane

May 2007

Chair: Brij M. Moudgil
Cochair: Anthony B. Brennan
Major: Materials Science and Engineering

There are many material benefits to adding nano-sized filler particles to a

polymeric resin, such as improvements in strength, modulus and optical transparency.

However, the relative amount of material improvement is highly dependent on the degree

of particle dispersion within the nanocomposite. Characterization of the extent of particle

dispersion poses a great challenge because of the length scales of interest, from nano to

macro. In this study, a commercial dry titania nanopowder with a mean primary particle

diameter of approximately 30 nm was mixed into an epoxy resin/hardener system to

produce nanocomposite samples. Processing techniques such as ultrasonication and

particle surface modification were used to produce nanocomposites with varying degrees

of dispersion. The main objective of this study was to characterize the state of dispersion

of the nanoparticles embedded in the epoxy by static light scattering, SAXS, and USAXS

techniques. From these measurements, a quantitative dispersion index was derived based

on the change in particle size distribution and three dimensional relative arrangement of the

nanoparticles as compared to an ideally dispersed system. The state of dispersion was also

verified qualitatively by microscopy.














CHAPTER 1
INTRODUCTION

There are many emerging applications where it is necessary to disperse

nanoparticles into a polymer matrix, such as gas sensing films, adhesives,

pharmaceuticals, aerospace materials, paints, automotive parts, coatings and explosives.

These dispersed particle systems are mainly used to enhance or improve the properties of

the surrounding matrix material, such as conductivity, mechanical and impact strength,

energetic, wear resistance, and optical properties, to name a few. One of the main

reasons for using nanoparticles is their large surface to volume ratio. The large surface

area increases the number of particle-polymer interactions, thus strongly influencing the

overall material properties.1 The growing effort in the area of nanocomposite research

stems from the promising material properties that can be realized by efficient

incorporation of nano-sized filler particles.

Wetzel et al state2 that the unique polymer nanocomposite properties can only be

achieved if the nanoparticles are well dispersed in the matrix. The most optimal material

property benefits can only be realized if an ideal dispersion is achieved. An ideal

dispersion is defined in this study as a three-dimensional array of non-touching primary

particles in which all particles are homogeneously distributed in the matrix and not

confined to any geometrical lattice. Conversely, particle agglomeration decreases

composite material performance by the inclusion of voids that act as preferential sites for

crack initiation and material failure.1 3 Ganesan, et al.4 have concluded that "poor

dispersion would lead to reduced product life or poor performance during service, poor









product appearance, poor processing and product uniformity, waste of raw materials, and

excessive energy usage."

Particle agglomerates typically exist in two different forms. The first form is often

produced during particle synthesis and manufacture, where the primary particles

produced during decomposition and subsequent calcination sinter together to form a hard

agglomerate. In this study, these partially sintered primary particles are termed an

aggwegaite. Aggregates require attrition to break the strong covalent interparticle bonds.

The second agglomerate form is called a soft agglomerate, or simply an aggii,,, tilde.

A particle agglomerate is held together by weaker bonds, such as attractive interparticle

van der Waals forces, capillary forces, electrostatic forces, etc.5 However, van der Waals

forces are much more significant than the long-range electrostatic forces in the nano size

regime.6 The strength of particle adhesion strongly depends on particle size, as well as

the surface roughness, surface energy, and also the effective elastic constant.6 Generally

stated, as particle size decreases, the adhesion forces present between particles in order to

form agglomerates increases.

The most critical endeavor to understanding nanoparticle aggregation and

agglomeration is characterization of not only the change in the size of the agglomerates

(or aggregates)due to adhesion forces, but also the distance between agglomerates and

spatial distribution, reflecting concentration gradients and mixing fluctuations .

Krishnamoorti and Vaia8 state that the final properties of a polymer nanocomposite

system will depend as much on the individual properties of the constituents as on the

relative arrangements and subsequent synergy between them. Thus, characterization of

these changes requires information not only on the nanoscale, but also but also on the









microscale.9 Despite the importance of such information, development of techniques to

quantify the extent of particle dispersion has delayed the investigation of physical and

chemical processing techniques to enhance dispersion further than the results of

conventional procedures.

Unambiguous dispersion characterization is difficult because it also depends on the

complicated interactions between the particles and the matrix, but also the experimental

processing conditions.10 A variety of techniques have been proposed to determine the

degree of particle dispersion of a polymer nanocomposite and can be categorized into

four general areas7: (1) real space observation including optical and electron microscopy;

(2) determination of material property response due to changes in dispersion, such as

rheology, permeability and impact resistance, which provides an indirect verification of

conclusions drawn based on previous models; (3) measurement of a quantitative area of

interaction between the particle and matrix, such as XPS, NMR, or other spectroscopic

techniques; and (4) reciprocal space observation based on scattering of individual

particles and agglomerates (or aggregates). Overall, it is clear that no individual

approach is sufficient to fully characterize nanoparticle dispersion on all relevant length

scales.

The ideal technique for the evaluation of nanoparticle dispersion will probe the

interparticle distance and the relative arrangement of the particles while also accounting

for the primary and agglomerate particle sizes. To date, no one has utilized a range of

length scale measurements, incorporating all aspects of three dimensional particle

homogeneity characterization, to provide a quantitative evaluation of nanoparticle

dispersion. The main objective of this research is to develop a standardized methodology









for the characterization and evaluation of the degree of nanoparticle dispersion in a

polymer matrix. Accordingly, a full range of length scales of the nanocomposite material

will be investigated. Attempts will be made to develop a quantitative measure of

dispersion that includes both a mixing parameter to determine the effectiveness of the

blending process (microscale) and a surface chemistry parameter that reflects the degree

of compatibility of the particles by the surrounding polymer nanoo scale). Additionally,

chemical and physical processing techniques will be investigated to systematically

enhance the dispersion of inorganic nanoparticles in an organic polymer matrix.




















CHAPTER 2
LITERATURE REVIEW

In the following section, a critical review of the characterization techniques

selected for this study: light scattering, small angle x-ray scattering, and microscopy is

presented. Also, previously developed quantitative dispersion indices are reviewed,

along with their calculation and practical limitations. Finally, the ASTM standard for

assessing the state of dispersion of carbon black in rubber is presented as a case study for

what is used in the automobile tire industry.

Particle Dispersion Studies

The dispersion and stabilization of particles suspended in a medium has been

intensely studied for many years. The majority of the work has focused on the dispersion

of particles in aqueous media by electrostatic or steric stabilization mechanisms.1

Electrostatic stabilization can be achieved by creating a high charge density on the

particle surface, which results in a strong electrical double-layer repulsion between

particles. Steric stabilization involves the attachment or adsorption of long chain

molecules (polymers, silanes, surfactants, etc), where the molecules emanating from the

particle surface create a physical barrier to agglomeration. The adsorbed layer opposes

the approach of neighboring particles both entropically and enthalpically11. Widegren

and Bergstrom 2 have found that the control of the dispersion in nano sized powders









becomes even more essential because even small perturbations in the interparticle forces

have drastic effects on the dispersion behavior. Due to the large available surface area of

nanoparticles, the key to effective dispersion lies in control of the chemistry at the

particle surface.

Particle Surface Modification

Shutzberg et a113 suggest that a typical problem leading to particle agglomeration is

incompatibility of the nanoparticles with the organic matrix and suggests organic

modification of the particle surface. One way to decrease the effect of attractive

interparticle forces and increase compatibility of the particles with the surrounding

polymer matrix is by grafting molecules that have an affinity for the surrounding polymer

(eg, similar miscibility parameters) onto the particle surface. Long chain molecules, such

as silane coupling agents, are often used in composite systems to not only form a steric

barrier around the particles to prevent reagglomeration by volume exclusion, but also to

react with the polymer matrix to improve interfacial adhesion.14 However, Fischer notes

that in a polymeric nanocomposite, even when a surface modification with favorable

miscibility with the surrounding matrix is utilized, complete dispersion can be difficult to

achieve due to confinement of the particle agglomerates by the polymer chains.15

Brinker and Scherer11 generalize that an effective barrier around particles to

enhance and stabilize the dispersion must meet certain guidelines. First, the surface of

the particle must be completely covered with the adsorbed molecules. Secondly, the

adsorbed molecule must be firmly anchored to the surface so that it cannot be desorbed

by Brownian motion. Thirdly, the layer must be thick enough so that the minimum

distance between approaching particles still remains well above that necessary for van der

Waals attractive forces to dominate. Finally, the end of the molecule that is not bonded









to the particle surface must be miscible with the surrounding solvent, or covalently

bonded to the surrounding medium11. Effective engineering of the particle surface aids in

the stabilization of particles once interparticle separation has occurred, but cannot break

up existing aggregates. It has been noted in the literature that dispersion of nanoparticles

in a polymer matrix only by surface modification provides limited success without

utilizing other techniques.15 The addition of the appropriate surface chemistry is one

technique used to enhance particle dispersion, while mechanical mixing has also been

used in the literature to overcome the attractive van der Waals forces.

Mechanical Dispersion

Most recent studies have determined that mechanical dispersion methods alone are

not sufficient to disperse nanoparticles to primary size in a viscous medium such as a

polymer. However, many have utilized mechanical energy such as high speed

homogenizers, mixers, and sonication techniques. The objective of using such techniques

is to create eddy currents in the dispersing medium that provide dispersing energy or

shear forces on the appropriate length scale of the particle agglomerate to overcome the

interparticle van der Waals forces.16 Conversely, Yang and Dan17 have stated that

physical mixing alone may lead to increased separation into the discrete phases, rather

than promote particle dispersion. Ultrasonication or bath sonication are common

procedures used to physically break up agglomerates and overcome the van der Waals

attractive forces.

Several groups have used sonication methods to break up agglomerated

nanoparticles for incorporation into polymer matrices.18-19 However, it has been shown

that even high-power ultrasonication does not always break up all agglomerates found in

a suspension of nanoparticles.20 A number of researchers have combined techniques to









provide the optimal level of particle dispersion. For example, Zhao and coworkers21 used

sonication and particle surface treatment combined for more effective dispersion of

carbon nanofibers in water. Lee, et al.22 also employed both physical and chemical

dispersion techniques for Fe203 nanoparticles. Many studies have been devoted to

finding an effective technique for optimal particle dispersion in a polymer matrix.

However, none are reported that have employed a quantitative measurement to define the

level of dispersion that has been attained by using those techniques.

According to the work of Kolmogoroff, the model of isotropic turbulence can be

employed for modeling the mixing process of particles in a fluid. Utilizing this model,

the mixing volume during processing is considered as a distribution of eddy currents of

various sizes, created by the energy input of the mixing technique (e.g.,

ultrasonication).23 24 The larger eddies are slowly transformed into smaller ones at a

constant rate of energy dissipation until they are reduced to the smallest size, 0o, which is

called the inner turbulence scale, or eddy length scale.23'24 The Kolmogoroff equation

can be written as

Ao = (pv3/tot) (1)

where p is the density of the mixture of particles in a liquid, v is the kinematic viscosity

of the mixture, and etot is the total energy input into the mixture with respect to the

mixture volume. A simple calculation reveals that for a low concentration of

nanoparticles (e.g., 0.5 wt%) dispersed in a low viscosity fluid (e.g., isopropanol), placed

in an ultrasonic processor at 250 W, the length scale of the eddy currents is still on the

millimeter scale.









Dispersion Equations

Balek25 first defined an agglomeration parameter, A, in 1970 in order to evaluate

the degree of particle dispersion as follows:

D
A = (2)
d

where D is the median particle aggregate diameter determined by sedimentation and d is

the mean primary particle diameter measured by the BET (Brunauer, Emmet and Teller)

method. Besides the fact that the particles are measured both in the liquid and dry states,

there are several limitations to this approach. First, the sedimentation technique assumes

that the particles or aggregates are spherical in shape, and also that an aggregate has a

constant density, which is not the case for a sometimes porous aggregate. The BET

technique assumes that the primary particle size distribution is monosize and that particle

size can be extracted from a measure of surface area alone. The agglomeration

parameter, A, was the first attempt to evaluate the degree of particle agglomeration, but

was not developed any further.

Liu, et al.26-27 proposed an aggregation equation based on the relative packing of

nanoparticles and the correlation of a dispersed particle system to that of particles

confined to a geometric regular lattice. They calculated the difference in particle spatial

distribution for eight regular lattices and three real particle distributions. They derived a

generalized equation for relating the average matrix ligament thickness (surface to

surface interparticle distance), T, to the average particle size (d), size distribution (o),

volume fraction (p) and spatial distribution factor () for a system containing spherical

particles obeying the log-normal distribution:










T = d exp(1.5 In2 r)- exp(0.5 n2 cr (3)


It has been shown through simulation and experiment that previous models

proposed by Liu and coworkers are appropriate to describe a well-dispersed binary

composite system, but cannot describe a poorly dispersed system.26 The use of previous

equations includes the assumption that the particles form some kind of ordered spatial

lattice, which is seldom the case for a real composite material. Equation (3), however,

includes a spatial distribution parameter, 4, to account for such deviations from geometric

lattice spacing of the particles in the matrix. Premphet and coworkers28 use equation (3)

to characterize the interparticle distance of rubber particles in a polypropylene matrix, but

found that the values obtained by experiment for a non-ideal system did not match well

with the theoretical model.

Other attempts to describe the spatial distribution of randomly dispersed particles

have been made. Gurland29 suggested an approach to quantifying dispersion with a

standard deviation (6,) of the particle spatial distribution in small composite areas,



S= 1 (n n (4)


where Nis the number of samples, n, is the number of observed particles per unit volume

in the ith sample and n is the arithmetic mean of particles in N samples. The major

limitation to this approach is that the equation represents a simple variance in the number

of particles observed in a number of samples. It does not address the particle

agglomerate size or size distribution, or the change in spatial distribution over a large









sample volume. Tanaka, et al.30-32 used the parameter, P, as the spatial distribution factor

to quantify dispersion. The value of P approaches one with increasing particle

dispersion,



= 1= (5)
In


b
P,= (6)
Zb,
1=1

where Mis the total number of grids (divisions) on a transmission electron microscopy

(TEM) micrograph, and b, is the particle area of the ith grid. This method involves

tedious counting of particles on a TEM micrograph and again does not address the

importance of particle agglomerate size or changes in spatial distribution on the macro

scale. The dispersion index developed by Suetsugo33 (1-a), where 0a is the area

fraction of agglomerates, is defined by:

N
( = n D2 (7)
4A ,

where A is the observed area, q is the volume fraction of particles, D, is the agglomerate

diameter and n, is the number of agglomerates in the sample. The dispersion index

decreases to zero for maximum agglomeration and increases to one for maximum particle

dispersion. Although this parameter begins to address the importance of particle size and

arrangement within the sample volume as a measure of dispersion, the measurement

technique still utilizes TEM micrographs that only examine a very small fraction of the









overall bulk sample. The equation also does not address the measurement of particle and

agglomerate spatial distribution.

Although some equations seem more valid than others, all of the proposed

dispersion indices employ measurements made by some microscopy technique, such as

SEM or TEM. This approach can be misleading as to the actual particle arrangement in

space27,34-35. It may be that the characterization of particle dispersion spans many length

scales, which cannot be fully characterized by a small sample area by microscopy

methods alone or just one calculated dispersion parameter.

Correlation of Dispersion with Material Properties

Mechanical Properties

It has been well documented that the degree of particle homogeneity has a great

affect on the resultant mechanical properties of the system. When nanoparticles

agglomerate into large clusters, the reinforcing effect is lost.36 Particle agglomerates

degrade the composite performance through the inclusion of voids that can act as

preferential sites for crack initiation and failure1 For example, the findings of

Thompson and coworkers37suggest that the decrease in tensile properties from the neat

polyimide polymer after addition of metal oxide nanoparticles is indicative of poor

particle dispersion on the nanometer scale. Most studies in literature only address the

change in material properties with varied processing and little progress has been made to

control or quantify the state of particle dispersion. Complications arise when a simple

extrapolation of the design paradigms of conventional composites cannot be used to

predict the behavior of nanocomposites.38









Optical Density

A nanocomposite material in which all of the nanoparticles are dispersed to

primary particle size displays optical properties in the visible range close to that of the

pure matrix material with no filler particles due to low levels of scattering by

nanoparticles. However, when particles agglomerate to sizes that are comparable to the

wavelength of visible light, high levels of light scattering occur and the optical density

decreases.39 Thus, the change in particle agglomerate size as it relates to dispersion has

been related to the degree of particle homogeneity in a polymer nanocomposite. For

example, Shang, et al.40 correlated the high transparency of their polyimide composites to

the compatibility of the SiO2 particles to the polymer matrix, thus providing good

dispersion. Vandervorst and coworkers41 qualitatively related the visual appearance of

several Latex-SWNT (single wall nanotube) composites of varied levels of dispersion.

The worst dispersions (as viewed by microscopy) appeared to have the lowest level of

transparency.

Molecular Modeling of Nanocomposite Dispersion

In order to gain a better understanding of polymer-particle interactions and the

mechanism of nanoparticle agglomeration in polymer matrices, molecular modeling and

simulations have been attempted to further comprehend nanocomposite dispersion.

These molecular dynamics simulations utilize the correlation between polymer matrix

effects with the overall particle dispersion. Smith and coworkers42 have found that

nanoparticle agglomeration increases with polymer molecular weight for composite

samples containing weak polymer-nanoparticle interactions. This finding becomes

significant specifically for fillers that are not covalently bonded to the polymer and have

a high molecular weight matrix material. It has also been observed that when the









polymer-particle interactions are strong, as seen with chemical bonds created by particle

surface modification (eg, silane coupling agents), the molecular weight effect becomes

insignificant.

Starr, et al.43 take a different approach to molecular modeling by investigating the

thermodynamics of particle clustering and the mechanisms involved in agglomeration.

However, they arrived at the same conclusions as Smith and coworkers.42 By observing

changes in potential energy of interparticle interactions with respect to volume fraction,

they determined a boundary between a dispersed and non-dispersed system.

It has been shown through the simulation of simple liquids, which are far from the

case of a viscous polymer melt or resin, that the potential energy is strongly sensitive to

local packing of the particles. The Starr group, as well as others42' 4445, also utilized the

structure factor, S(q) and the pair correlation function, g(r), for an explicit measure of

long and short range ordering of the particles as it relates to particle dispersion. Brechet45

also describes the sensitivity of the nearest neighbor distribution to particle

agglomeration. However, this approach has little practical application outside of

modeled systems due to mathematical complexities. Although these simulations come

closer to an explicit value of dispersion than empirical work, the practical uses of these

results are limiting due to the extremely simplified modeled composite systems. Hooper,

et al.46 reviewed the progress of modeling techniques for polymer nanocomposites and

have found that most work still focuses on dilute polymer systems and the one or two-

particle problem. However, from understanding the models for dispersion measurements

derived previously, it is clear that characterization methods that probe the real composite









material on a wide range of length scales are better suited to determine the degree of

homogeneity of the system.

Light Scattering

Background

Light scattering, also known as laser diffraction, has been used for many years to

characterize the size, shape, and orientation of particles. Generally, light scattering is the

most common method of particle sizing.47 The study of the scattering of light from

matter was most advanced by Gustav Mie in 1908, when he published his ground-

breaking article that provided an exact solution for the scattering of radiation from a

dielectric sphere in a medium.48 It is suggested that this work stems from the study of

colloidal gold suspensions by his student.49 This study lead to the term "Mie scattering",

which is used to describe all scattering of radiation by spherical and non-spherical

particles in a certain size range. It is important to note that the analysis presented in

Mie's paper is restricted to particle sizes up to 0.18 atm, possibly suggesting the

connection between the colloidal gold work and Mie's paper. After 1908, Gustav Mie

published several more papers, but nothing more on light scattering of particles.49

Light scattering occurs due to the gradient of refractive index (electron density)

between the particle and the embedding medium (polymer). The refractive index may be

a function of incident radiation, temperature, or pressure. If the material is nonabsorbent

and nonmagnetic, then the square of the refractive index is equal to the dielectric constant

at that wavelength. For absorbing materials, the complex refractive index, m = n(1 ik),

is related to the absorptive index, k, where the real term describes the refraction and the

imaginary terms describes the absorption.50 If the particles and polymer have the same

refractive index, the particles would appear to vanish because there would theoretically









be no scattering of light, which is called index matching. The greater the refractive index

gradient, the greater the particle scattering. The scattering power (I(q)), or scattering

intensity per unit time and volume, is dependent on the ratio of the refractive indices of

the particle and surrounding medium. For this reason, high refractive index particles

have been chosen for this study (see Chapter three) to achieve high scattering resolution.

Any laser diffraction instrument uses the angular dependent scattering pattern

(scattered intensities) of particles in a dispersing medium to calculate the particle size

distribution. A commercial instrument usually includes a proprietary software package

that solves the inverse Mie problem to mathematically invert the scattering matrix data in

order to calculate particle size distribution (see Appendix B). Generally, large particles

scatter strongly and principally to smaller angles while smaller particles scatter weakly at

much larger angles.47 The relative terms small and large depend solely on the

wavelength of the incident radiation.

Light scattering has been cited by many researchers as a meaningful technique to

characterize the degree of dispersion of particles embedded in a matrix material. McNeil

and coworkers1 state that light scattering can be strongly influenced by inhomogeneities

in the random microstructure, such as particle aggregation and agglomeration.

Specifically, Schnaiter and Wurm52 reported that light scattering is a powerful technique

to characterize dust aggregates. Schaefer and coworkers19 also utilized light scattering to

characterize silica dispersion in a rubber matrix. Wiggins, et al.53 used optical light

scattering in the Rayleigh regime (particles small compared to the wavelength of the light

source) to calculate a bunching parameter, (exp(-i0)) which is an indication of the

density change of particles in the scattering volume over time with respect to the









scattering angle, 0. This type of measurement may be correlated with dispersion

homogeneity, however, there are several limitations of this technique. The Rayleigh

scattering theory assumes a monosized distribution of particles and no multiple

scattering. The measurement does not account for particle settling due to agglomeration

and is not suited for stationary particles embedded in a matrix.

The dispersion state can be monitored by light scattering because a particle

agglomerate made of many primary particles will scatter light as a single, larger particle.

Interparticle interactions also play a role in the scattering pattern when scattered radiation

overlaps due to the small interparticle distance51' 54-55. Although light scattering has been

used for many years to measure changes particle size and size distribution, the efforts to

relate that information to a quantitative measure of dispersion have provided little

practical relevance.

Particle Size Measurement

In the literature, very few groups have used static light scattering to measure

particle size in a solid polymer matrix. Gao and Wu56 used a commercial particle sizing

instrument by modifying the small volume sample holder of the Coulter LS 230 laser

diffraction instrument. Others have chosen build their own static light scattering

instrument to measure particle size52' 57-65. It was shown in all studies that custom-built

optical equipment can be flexible enough to measure the scattering characteristics in any

desired configuration, but also precise enough to deliver meaningful measurements

through careful instrument calibration and measurement.

Although none of the authors mentioned in the light scattering section attempt to

calculate a quantitative dispersion index from light scattering data, almost all mention the

ability to characterize a change in the particle conformation, or the fluctuation in mixture









homogeneity. Many also combine the results from laser light scattering to that of small

angle x-ray scattering to characterize a wider range of length scales66 67

Small Angle and Ultra Small Angle X-ray Scattering

The scattering of x-rays from a radiation source incident on an array of scatterers

(particles) occurs due to the difference in electron densities between the particles and the

surrounding medium, which is defined as the number of electrons per unit volume.68 The

greater the difference in electron density, the greater the scattering of the incident

radiation (scattering power). Small angle and ultra small angle x-ray scattering are

governed by the same fundamental laws of electromagnetic interactions with matter as

with light and neutron scattering.69 The difference between light scattering and x-ray

scattering is the wavelength of the radiation used. A typical x-ray source consists of a

1.54 A wavelength Cu K, target for a conventional system. The scattering pattern ofx-

ray radiation from an array of particles can provide meaningful information on the

relative arrangement, particle size, and microstructure of the particles that are probed. As

with any scattering measurement, the scattered intensity contains contributions from

intra- and interparticle spacing and arrangement.70 The effects of interparticle scattering

can be reduced by using a low concentration or dilute solution of scatterers.

Small angle x-ray scattering is a widely used technique to determine size, shape,

and internal structure of colloidal particles in the 1-100 nm range.70 In order to analyze

data from any scattering technique, the scattering intensity (I(q)) with respect to the wave

vector, q, can be divided into three distinct regions corresponding to specific hierarchical

length scales probed in a sample. These regions are represented by knees in the log-log

plot of I(q) vs. q. At high q values, primary particle size and size distribution can be

obtained (Figure 2-1), regardless of the agglomeration of the particles in the matrix, due








to the small probe size of the technique. In the mid-q range, particle agglomerates or

aggregates in the micron size range can be probed. The lowest q values probe the largest

of the particle agglomerates.71 Koga and coworkers72 used the different scattering

regimes from USAXS and SAXS data to characterize the hierarchical structures formed

by carbon black filler particles in a rubber matrix (Figure 2-1). Others have also used this

technique for morphology characterization of embedded nanoparticles73. The probing of

various length scales in the composite sample is believed to be the best approach to

dispersion characterization by addressing agglomeration from the nano to micron size

range.



I -

primary particle aggregate agglomerate

Figure 2-1. Diagram of the hierarchical structures of particles that can be probed with
light and x-ray scattering measurements at different length scales, or q values.

X ray Data Analysis

Historically, x-ray data analysis has been carried out by using approximations of

particle size based on the shape of the curve in different regions of the I(q) vs. q plot.

These classical techniques are presented with the recognition that a more modern

approach to data analysis is available through the use of modeling software. For the most

ideal case of a monodisperse, homogeneous and isotropic dispersion of spherical particles

in a medium, the general form of the scattering intensity equation can be written as74

I(q) = NP(q)S(q) (8)

where Nis the number of particles in the array, P(q) is the form factor that describes

intraparticle scattering and S(q) is the structure factor describing interparticle scattering.









The Guinier regime

In the Guinier regime of q values, the scattered intensity is used to determine the radius

of gyration (radius of an equivalent sphere), Rg, of the particle agglomerates.73 The

scattered intensities in this region can be generated from either light scattering, USAXS ,

or ultra small neutron scattering (USANS) data. These techniques probe very small q

values that correspond to large sizes in real space, dependent on the wavelength (k) of the

probe radiation, where q = -sin The following Guinier law holds for the region
a 2

of approximately I(q)/I(0) < 4/3 according to Guinier.7


I(q)= I(0) exp Rq2
l3q)=lo)e 3(9)

The radius of gyration, Rg is defined as the radius of a sphere with the same mass and the

same moment of inertia as the object under consideration76. It is related to the correlation

length of the particle agglomerates by77

R, =[D(D, +1)/21 /2 (10)

where Dfis the fractal dimension as determined from the power law regime. Chaker and

coworkers77 assert that the correlation length is the best estimate of the average inter-

agglomerate distance. Glatter and Kratky68 report in their extended review of SAXS

theory and practice that the Guinier formula holds surprisingly well in the majority of

particle size and shape cases. However, this technique does not address the distribution

of particle sizes present in a composite sample, nor the spatial distribution of those

particles.









The power-law regime

The boundary between the Guinier and power law regime was defined by Dobbins

and Megaridis first in 199178, although many describe the boundary as where the log-log

plot of I(q) vs. q displays a power-law decay. This value can be compared to the

crossover point suggested by Sorenson.75 The power law regime describes the mass or

surface fractal scaling at the particular structural level corresponding to the probed q

value.79 The fractal dimension, Df, falls in the range 1 < Df< 3. A value of three

corresponds to a sphere while a value of one corresponds to a rod where the mass is

proportional to the length. Fractal morphologies are often found for particles made by

aerosol techniques between the primary particle size and aggregate size.66

The table below elucidates some of the values of the fractal dimension to be

expected from certain types of aggregation mechanisms, although these values can vary.

Chaker, et al.77 found a fractal dimension of 2.45 for silica in a PPO network and

proposed a particle-cluster growth mechanism, which implies that the primary particles

were in a random walk configuration in the viscous matrix and stick to form a growing

aggregate. Schaefer and coworkers19 also found a fractal dimension of 2.5 for nano-silica

dispersed in a rubber matrix, which they describe as a "bunch of grapes" morphology.

They also suggest that an optimal fractal dimension (high or low) can be tailored based

on the model of reinforcement chosen in a given particle-polymer system. In light of

these suggested results, there is still much debate on the correlation of fractal dimension

and the state of particle dispersion. The change in material properties with respect to

changes in aggregate and agglomerate morphology (shape) has not been established.









Table 2-1. Several cases of mass-fractal power law scattering
Mass fractal structures Power Law slope
Multiparticle diffusion limited aggregate -1.8
Percolation cluster -2.5
Diffusion limited aggregate -2.5

A Unified Fit model

Beacuage and coworkers have developed a unified approach to small angle

scattering by combining both the Guinier and power-law regions into one equation that

only requires the same parameters that are used in the fitting equations for the separate

regimes.67 He terms this combined region as one structural level that more clearly

defines the limits of both regimes to more appropriately calculate Rg and Df. It is also

claimed that the unified approach provides an accurate model for the scattering behavior

for mass fractals, surface fractals, and diffuse interfacial scattering.67 However, it still

does not provide any information about the distribution of particle sizes or the spatial

distribution of the particles.

The Porod regime

The boundary between the power law regime and the Porod regime is characterized

by a change in slope, or "knee" in the scattering data. Several have used the relation q =

27/a at the slope change as an exact location for the calculation of the radius of the

primary particles, a. Hasmy and coworkers80 suggest that the crossover is highly affected

by the polydispersity of the primary particles and must be accounted for, such that


q, = (1 1.6e ) (11)
A

where qc is the crossover q value between the Porod and power law regime and oeff is the

effective polydispersity of the particles (a/a), where a is the radius of the primary

particles.









The "final slope", as termed by Porod in 1965, is expected to reflect the fine

structure of the particle.68 The limiting behavior of scattered intensities at large q values,

known as Porod's law, predicts a decrease in the intensity proportional to the reciprocal

fourth power of s:


I(s)= (12)

where Kp is the power law constant and s = 2sin(O/2)/2 (differs from q by 27C). The

determination of Kp is of interest because it relates to certain structural parameters of the

system.81 It has been generally assumed that a slope of -4 in the Porod regime indicates a

sharp interface (smooth surface) between the particle and surrounding medium, but it has

also been shown that a partially sharp boundary was sufficient to see the Porod-type

behavior.82 The Porod evaluation of the final slope of the scattering intensity plot reveals

the size of primary particles and the morphology of the interface, however, it does not

provide a distribution of primary particle sizes or the distance between those particles.

Scattering Techniques for Dispersion Measurements

Berriot, et al. in 200383 came very close to reporting a quantitative measure of the

state of dispersion of nanosized fillers in an elastomer matrix. They used a ratio of the

scattered intensity of a highly concentrated system to that of a very dilute system, which

is directly related to the structure factor of the particles, which accounts for the

interagglomerate scattering. However, no direct conclusions on a quantitative measure of

particle dispersion could be drawn from their work. In 2005, Causin and coworkers84

also attempted to provide a quantitative dispersion measurement, which consisted of a

comparison of parameters between SAXS and TEM techniques. This particular study

used clay nanocomposite samples, which included the added complexity of high aspect









ratio particles that stack together in tactoids, with very small inter-layer distances. Good

agreement was achieved between SAXS and TEM measurements with respect to size of

the clay nanoparticle cluster size and inter-layer spacing (ratio close to 1), although no

statistical data was reported.

Microscopy

Transmission electron microscopy (TEM) and scanning electron microscopy

(SEM) are both very common methods of obtaining qualitative information about the

dispersion of particles in a polymer matrix. Many of the studies mentioned here have

used microscopy in some way to verify the results from various scattering techniques,

such as primary particle size or average aggregate or agglomerate size. 62, 85, 72, 76

However, some claim that there are many drawbacks to using microscopy. Beaucage and

coworkers85 assert that TEM, although widely applicable, is very tedious and subjective

Later, Hyeon-Lee et al.66 mention that TEM is quite useful in the measurement of

primary particle size, but is very difficult to obtain a statistically meaningful size

distribution. They also suggest that measurement of anything beyond the primary

particle size, such as aggregates or agglomerate structure is very complex due to overlap

in three dimensions of particle structures. Scattering techniques are known to provide

more accurate measurements of size parameters by measuring over a wide range of

scattering vectors, and thus a wider range of length scale.

ASTM Standard for Dispersion of Carbon Black in Rubber

ASTM standard D2663-9586 describes the current standard test method used to

determine the degree of dispersion of nanoparticles in a polymer, specifically carbon

black in a rubber matrix. The test consists of three different methods- visual inspection,

quantitative count of agglomerates, and microroughness measurement. The visual









inspection test consists of looking at several cut rubber samples with a hand lens or low

power microscope and comparing the dispersion on a scale of one to five. Five has 99%

dispersed carbon black with one having 70% dispersed particles. One can only imagine

the amount of subjectivity involved in this sort of test, considering the low resolution of a

hand lens or even an optical microscope. The second method involves the inspection of

microtomed sections of the rubber composite sample under a light microscope. The total

number of agglomerates 5 [tm or larger are counted, and the total percentage of

agglomerates below this critical agglomerate size is calculated and expressed as the

percentage of carbon black dispersed. The final test consists of a microroughness

measurement using an instrument called a Dispersion Analyzer, which is available from

the Federal Products Corporation. This instrument consists of a specimen holder, sample

cutter, and specimen tracking mount. The stylus of the instrument traces across the cut

sample surface, which is mounted on a vibration isolator. The measurements are taken

and a dispersion index is calculated based on three constants that have been previously

determined for the specific material. This instrument can be described as a crude

precursor to the present-day Atomic Force Micrsocope (AFM). The limitations of this

type of technique are that many artifacts can be generated by probing a cut polymer

surface. Deformations and particle pullout can all contribute to artificial changes in the

morphology of the surface, leading to unreliable results.

The measurement standard also makes the claim that important physical properties

are "significantly influenced" by the percentage of carbon black dispersed, such as

abrasion resistance, tensile strength and fatigue life. These properties are said to correlate

with the agglomerate count and microroughness results in the following manner:









Dispersion, % Classification
Above 99 Very High
97 to 99 High
95 to 97 Intermediate
92 to 95 Low
Below 92 Very low

This test method has been applied to other systems in the literature with little

accuracy34 87. In the study by Potschke and coworkers34, all of the samples of multi-

walled carbon nanotubes (MWNT) embedded in a polycarbonate (PC) matrix earned a

dispersion index rating of 100% using the agglomerate counting method described above.

The PC-MWNT samples were observed under an optical microscope to assess the

dispersion of filler particles that are on the order of 12-16nm in diameter. Upon close

inspection of the SEM and TEM micrographs presented in this work (Figure 2-2),

it is obvious that this method of dispersion quantification is not adequate.

In 1995, Ganesan, et al.4 attempted to improve upon the ASTM standard to

describe the dispersion of carbon black in rubber by use of an image processing

technique. This method uses digitally pixilated images with calculated grayscale

tolerances to determine the texture of the micrographs, taken at a magnification of 30X.

Quantitative measures were calculated from these digitized images, including the

statistical weighted mean and variance from the ideal dispersed case. The weighted mean

was plotted against a dispersion rating for a straight-line relationship. There has been no

further use of this particular image analysis technique in the literature, although others

have tried, unsuccessfully, this same approach to quantify dispersion35. Park and

Balatinecz35 proposed a mixing index value that describes the deviation from an ideal

random mixture. This approach of comparing the measured dispersion to an ideal

random mixture provides a good starting point to the techniques developed in this study.









However, using low magnification micrographs to investigate dispersion on the nano

scale is physically impossible due to the resolution issues.






















jU. .






Figure 2-2. Transmission electron micrographs of PC-MWNT composites: (a) 5 wt.%
MWNT, (b) 5 wt.% MWNT, (c) 2 wt.% MWNT (d) 1 wt.% MWNT.

Scope of Proposed Research

The quantification of particle dispersion in a composite material has been intensely

studied in recent years, yet there are still many challenges in providing a meaningful

measure of homogeneity that spans the relevant length scales. Molecular modeling

cannot provide a theoretical solution due to the complexities of a dense polymer melt and

interparticle interactions. Microscopy has proven useful to characterize dispersion on a









small length scale, but has fallen short of long range measurements on a three

dimensional scale. Scattering techniques have been shown to provide information about

particle size distribution and agglomeration of nanoparticles for the various hierarchies of

length scale. It is clear from the literature review that one parameter to describe

nanoparticle homogeneity is not sufficient and one processing technique cannot provide

optimal particle dispersion. The objective of this study is to investigate the effects of

various processing techniques on the homogeneity of nanoparticle dispersion in

polymeric nanocomposites by using scattering techniques and to quantify the degree of

dispersion with parameters reflecting changes at all relevant length scales.














CHAPTER 3
MATERIALS AND METHODS

Materials

Nanoparticles

The nanoparticle system chosen for this study consists of a multi-phase titanium

dioxide commercial dry powder. The particles are made up of 80% anatase and 20%

rutile crystal structures. The particles are manufactured by the Aerosil flame synthesis

process at Degussa Corporation and are used primarily for photocatalytic applications,

but may also be used as a heat stabilizer for silicone rubber (www.degussa.com). The

flame synthesis process by which the particles are manufactured produces particle

aggregates (primary particles held together by covalent bonds) and primary particle

agglomerates, which are primary particles and aggregates that are loosely bound by van

der Waals forces (Figure 3-1). The primary particle size measured by scanning electron

microscopy and small angle x-ray measurements is approximately 25-30 nm in diameter

(Figures 3-2 and 3-3). The size distribution of the P25 titania powder dispersed in

alcohol media was also measured by dynamic light scattering on the UPA 150

(Microtrac) commercial particle sizing instrument and is log-normal (Figure 3-4). The

size distribution of agglomerates in the as-received dry powder was measured by laser

diffraction on the dry powder module of the LS 13320 (Coulter) commercial particle

sizing instrument (Figure 3-5). The average particle agglomerate size in the dry state was

approximately 10 km.









The optical properties of the particles were estimated by using a generic mixing

rule of the weighted anatase and rutile phases of Ti02 as follows

(RI)R (Weight fraction )R + (RI)A (Weight fraction )A = RI parles (1)

where R represents the rutile phase, A represents the anatase phase and RI is the refractive

index. Exact optical properties of materials at the wavelength of interest are difficult to

obtain from the literature, however, values at the wavelength of interest were extrapolated

form the work of McNeil and French.54




















Figure 3-1. Scanning electron micrograph of the Degussa P25 TiO2 powder. The powder
was mounted on carbon tape in the dry state, coated with amorphous carbon to
prevent charging and imaged in as-received form.

































Figure 3-2. Scanning electron micrograph of dispersed TiO2 primary particles and small
aggregates imaged on carbon tape (background). An example of a primary
particle is circled.


3.E-05

3.E-05

2.E-05

2.E-05

1.E-05

5.E-06

O.E+00


1000


Particle Diameter (Angstroms)


Figure 3-3. Small angle x-ray scattering (SAXS) derived particle size distribution of
Degussa P25 titania primary particles. The mean primary particle size is
shown to be approximately 25 nm in diameter.







32



450
40 0
350
D 300
250
>
S200
S 150
100
50
00
000 001 010 100
Particle Diameter (microns)


Figure 3-4. Log-normal particle size distribution of the Degussa P25 titania particles
dispersed in alcohol media, measured by dynamic light scattering.


4.5

4

S3.5



0
S2.5

2

1.5

Q 1

0.5-f


0.1 1 10 10 100 1000
Particle Diameter (microns)


Figure 3-5. Agglomerated log-normal particle size distribution measured by the dry
powder module of the Coulter LS 13320 laser diffraction instrument. The
average particle agglomerate diameter is 10 km.

Polymer Matrix

The matrix polymer for the composite material used in this study is a two-part


thermoset epoxy based on a viscous resin and room temperature hardener. The resin is


manufactured as EPON 828, which is a Bisphenol A diglycidyl ether (DGEBA). This


type of resin makes up approximately 75% of the epoxy market. The industrial









applications of this resin include adhesives, coatings, electronic and structural materials,

and composites (www.miller-stephenson.com/aero_005.htm). Some properties of the

epoxy resin are shown in Table 3-1.

Table 3-1. Material properties of the epoxy resin.
Viscosity at 25C MW per epoxide
EPON 828 110-150 P 185-192

The hardener for this resin is a poly(oxypropylene)diamine, also known as

Jeffamine D230 (Huntsman). This low molecular weight (-230 g/mol) cross linker was

used specifically so that the particles have maximum mobility to disperse within the

curing epoxy matrix. This amine terminated polyether cross links the epoxy resin by

breaking the epoxyide rings and sharing of the amine group to form a three-dimensional

network (Figure 3-8). The hardener was applied in this study without the use of a

catalyst. The optical properties information for this epoxy is also rather difficult to find

in the literature, however Thiele and French have done extensive modeling of rutile

titania particles embedded in a cured epoxy.55




OH _,

Figure 3-6. Structure of DGEBA and the repeat unit. The reactive epoxide groups are
highlighted in red.

H2N -- O x NH2


CH; CH:
x 2.5

Figure 3-7. Structure of Jeffamine D230 and the repeat unit.










S"OH R"

H-NR





CH3 OH

OH CH3
N--

N---

HC OH
'NN 0

HO J/ CH3

Figure 3-8. Diagram showing how an amine functionalized hardener attached the
reactive epoxide groups and cross-links the epoxy chains to form a thermoset
material.

Table 3-2. Optical properties of the cured epoxy polymer.
Wavelength Refractive Index Source
Epoxy 560nm 1.514 Thiele and French (1998)

Particle Surface Treatment- Silane Coupling Agent

Two surface treatments were used to enhance dispersion by making the particles

more compatible with the surrounding polymer. A silane coupling agent functions to

bind both to the functional group of the polymer (in this case, the epoxide rings) and also

the surface hydroxyl groups of the particle. In this way, the coupling agent provides a

covalent bond between the particle and polymer, thus favoring particle dispersion over

the dominant interparticle van der Waals attractive forces that promote particle

agglomeration. The silane coupling agents chosen for this study are aminoethyl

aminopropyltriethoxysilane (Gelest) and aminopropyltriethoxysilane (Gelest), both used

the amine functionality available to bond to the epoxy polymer chains (Figure 3-9). The

triethoxy-type silane coupling agents have been shown in the literature to effectively

increase the compatibility of the metal oxide particle surface with the surrounding









polymer'. 88 Specifically, Zhu and coworkers8 used the amine functional molecules

attached to carbon nanotubes to form a covalent bond between the particles and the

surrounding epoxy polymer. Also, Shutzberg, et al.13 used a series of silane coupling

agents to increase the compatibility of BaTiO3 nanoparticles with an epoxy matrix. Also,

Choi and coworkers utilized a silane coupling agent to increase the adhesion between

ferrite particles and the epoxy matrix.89

The coupling agent bonds to the titania surface first by hydrolysis of the silane.

Next, a condensation reaction proceeds to form hydroxyl groups on the silane, and finally

hydrogen bonding takes place with the hydroxyl groups that are present on the particle

surface to form a physical bond. Finally, water is eliminated and the siloxane bonds are

left between the epoxy and particle surface. The kinetics of these reactions are highly

dependent on the pH of the solution.11 The hydroxyl groups are present on the surface of

the particles due to atmospheric conditions such as humidity.


NH2


0-
IH
.-- --Si-O .-- O-Si- N NH2
I N
__- r

Figure 3-9. Schematic showing the two silane coupling agents used in this study: APTES
(A) and AEAPTMS (B). The APTES is shown bound to the particle surface
through bonding with the surface hydroxyl groups.

Methods

Nanocomposite Formulation

The composite samples were prepared in order to obtain four distinct dispersion

states, using 0.5 wt% TiO2 nanoparticles by weight. It was shown in previous chapters









that 0.5 wt%/ nanoparticles was sufficient to observe improvements in the material

properties of the composite material. The undispersed sample was prepared by the most

basic processing techniques commonly used to produce a nanocomposite material, and is

highly agglomerated. Thefractionated sample, on the other hand, represents the highest

level of nanoparticle dispersion that was achieved in this study. The methodologies used

to create these samples involved the manipulation of surface chemistry and physical

mixing. The development of the methodologies to produce the fractionated

nanocomposite sample involved the selection of the appropriate solvent and surface

modification for optimal particle dispersion. The processing techniques used to fabricate

each sample are summarized in Table 3-3, and a detailed description is provided

subsequently.

Table 3-3. Summary of processing techniques used to fabricate each nanocomposite
sample with varied levels of particle dispersion.
Sample Mixing Ultrasonication Surface Treatment Solvent Dispersion
Undispersed Yes No No No
Ultrasonicated Yes Yes No No
Surface
ra Yes Yes Yes No
Treated
Fractionated Yes Yes Yes Yes

Four levels of particle dispersion

The undispersed sample with the highest degree of agglomeration was prepared by

simply mixing the epoxy resin (EPON 828), curing agent (Jeffamine D230), and the

appropriate amount of titania nanoparticles (Table 3-4) in a polypropylene container.

The three components were mixed at 60 rpm with a homogenizer for approximately five

minutes and degassed for 45 minutes to one hour under reduced pressure to eliminate

bubbles in the polymer-particle mixture that were formed during mixing. The silicone

mold was then placed on a clean glass plate and pressed down to remove all entrapped air









underneath the mold and to ensure planar samples. The epoxy-nanoparticle mixture was

poured slowly into the individual 1 mm deep wells, and a microscope slide was used

horizontally to level off the top of the wells to eliminate any epoxy overflow and ensure

proper filling of each individual well. Any extra epoxy mixture was removed from the

silicone mold before curing. The cast polymer in the mold was placed into an oven at

580 for at least 12 hours to guarantee complete cross-linking of the polymer. After cure,

the nanocomposite samples were easily released from the silicone molds and cut with a

scissors or blade into the desired shape for characterization.

The silicone mold was made from a two-part mixture of T2 Silastic (Dow Coming)

base resin and curing agent. The silicone base resin was mixed in a 10:1 weight ratio

with the curing agent. A typical batch mix consisted of 100g of the base resin and 10g of

the curing agent. The mixture was hand stirred thoroughly (approximately four to five

minutes) and degassed under vacuum until little or no visible bubbles remained. A glass

plate and several 1mm-thick microscope glass slides were cleaned with ethanol and

surface treated with hexamethyldisilazane (Gelest) by vapor deposition or swiping with a

Kimwipe. This surface treatment was used to prevent adhesion of the silicone to the

glass substrate once cured. The treated microscope slides were placed on the treated

glass plate with approximately one inch space in between each slide. Stacks of

microscope slides may be attached to the outer perimeter of the glass plate by double-

sided tape to prevent overflow of the silicone over the plate and to determine the final

thickness of the mold.

The degassed silicone mixture was poured slowly over the glass plate and slides,

starting in one corer of the plate and taking care not to trap air under the poured silicone.









A second glass plate was cleaned and treated with hexamethyldisilazane and placed

carefully on top of the silicone, starting by angling the plate onto one side and slowly

lowering onto the silicone to avoid trapping excess air in the mold. A weight (-5 lbs.)

was placed on top of the upper glass plate to keep the entire set-up in place during cure.

The silicone elastomer may be left to cure overnight at room temperature or more rapidly

cured at 50-600 for 3-4 hours. The silicone was then pulled off of the glass plates after

cure and the glass microscope slides removed to reveal the indentations (wells) left in the

mold. A razor blade was used to cut the overall mold into the desired size.

The silicone mold was placed on a glass plate (with indentations exposed) and the

polymer-nanoparticle mixture was cast into the Imm thick wells as described above.

This procedure yielded 1mm thick nanocomposite films of varied nanoparticle

dispersion. The state of particle dispersion was then characterized by scattering

measurements.

The ultrasonicated sample was systematically formulated to contain particle

agglomerates that are smaller than the undispersed sample by adding a mechanism for

physical dispersion of the loosely bound agglomerates in a reduced viscosity

environment. The sample was fabricated by adding the titania nanoparticles first to the

low viscosity (9 mPa-s, 25 C) curing agent (Jeffamine D230) and ultrasonicating the

mixture at maximum power (250W) for 30 minutes in a cool water (20C) bath. The

instrument used was a cup-horn type ultrasonic processor with a circulating bath. The

particle-curing agent mixture was added to the epoxy resin and mixed by a homogenizer

and cured as in the preparation of the undispersed sample.









The surface treated sample includes the physical dispersion mechanism as

described in the ultrasonicated sample preparation, but also includes a silane coupling

agent covalently bonded to the titania surface, prior to the mixing step. The

aminopropyltriethoxysilane (APTES) coupling agent, was chosen to decrease the surface

energy of the particle surface to make it more compatible with the surrounding polymer,

and also to provide a covalent bond between the particle surface and polymer for

additional cross-link density. The titania nanoparticles were first added to the curing

agent (Jeffamine D230) and ultrasonicated at maximum power for 25 minutes as

described for the ultrasonicated sample. Next, ImL of the APTES silane coupling agent

( or 2% by weight) was added to the dispersed particles and further ultrasonicated at

maximum power for five minutes. The epoxy resin (EPON 828) was then added and

mixed at 60rpm with a homogenizer until the nanoparticles were completely incorporated

into the resin. The polymer was then degassed, cast into the silicone mold and cured as

described previously.

Table 3-4. Summary of components used to fabricate nanocomposite samples for
undispersed, ultrasonicated, and surface treated samples.
0.5 wt% TiO2
EPON 828 29.1 g
Jeffamine D230 8.19 g
APTES 1 mL
TiO2 particles 0.187 g

The fractionated sample preparation yielded the optimal particle dispersion

achieved in this study by employing physical and chemical dispersion techniques used in

the previous three samples, in addition to a solvent mixture (acetone and isopropanol) to

enhance particle wetting and dispersion. First, the dried titania nanoparticles were

weighed and divided equally into 50mL capacity round-bottom Teflon centrifuge tubes.









Each centrifuge tube contained less than 0.5g nanoparticles. A 60/40% vol/vol mixture

of acetone and isopropanol was added to completely fill the centrifuge tubes. The

rationale for the solvent mixture is presented in the following section. The particle

agglomerates were then physically dispersed using a high power cup-horn type ultrasonic

processor for one hour at maximum power (250W) by rotating each individual centrifuge

tube into the ultrasonicator for 4-5 minutes each. The particle/solvent mixture

temperature was regulated by a water bath at 200C. It was determined through

experimentation that one hour ultrasonication time resulted in the smallest

aggregate/agglomerate size as measured by dynamic light scattering with no apparent

fragmentation of the particles or inducing any unwanted chemical changes to the polymer

matrix due to degradation.90

When ultrasonication was applied, the loosely bound agglomerates were broken.

The good wetting characteristics of the solvent molecules with the particle surface aided

the ultrasonication technique and enhanced dispersion. The lower viscosity of the solvent

as compared to the hardener (used as dispersant in other samples) increased particle

wetting simply by reducing viscous forces that oppose wetting. In addition, the cohesive

energy density difference may be lower for the solvent-particle system as compared to

the hardener-particle mixture (Ec information could not be found for Jeffamine D230

hardener).

A bimodal particle size distribution was obtained from the ultrasonicated

nanoparticles by dynamic light scattering (Figure 3-11), which shows the large aggregate

size fraction remaining in the solvent mixture on the right side of the distribution, while

the small size fraction consists of the dispersed particles (broken agglomerates). The









average particle size obtained from this distribution was approximately 280 nm diameter.

The solvent aided in the dispersion of the nanoparticles through ultrasonication

techniques by reducing viscosity and increasing wetting, making the VDW barrier easier

to overcome.

After ultrasonication, the centrifuge tubes were placed immediately into a high

speed centrifuge (Beckman, model J2-21) that was outfitted with a carbon fiber

reinforced rotor (FiberLite, model F13-14x50cy). The samples were centrifuged for ten

minutes at 5000 rpm to settle out any hard agglomerates that remained undispersed after

ultrasonication. The supernatant was removed by pipette after centrifugation and

transferred into a 200 mL round flask. The sedimented particles were recovered, dried

and weighed to determine the weight fraction of particles in the nanocomposite by the

following equation:


particle loading = W tal-, Wdred) x100 (2)
total Wded ) WEPON + Jeffaminme

where Wis weight of the nanoparticles, polymer resin (EPON) or curing agent

(Jeffamine). After sedimentation (Figure 3-12), the particle size distribution was

measured again by dynamic light scattering (Figure 3-13). The average particle size

decreased dramatically to 50 nm, indicating that the majority of the particles remained

dispersed to primary particle size (-30 nm).

The supernatant was agitated by a magnetic stir bar at a speed of 300rpm. The

silane coupling agent, aminoethyl aminopropyltrimethoxy silane (AEAPTMS) was

directly added dropwise to the flask in the amount of 100atL with further stirring for at

least 15 minutes. The silane coupling agent (AEAPTMS) reacted by covalent bonding to

the hydroxyl groups present on the particle surface. The supernatent was left to age for









24 hours to verify that reagglomeration was prevented by surface modification. The

particle size distribution was measured once again and the average size increased slightly

to 55 nm, indicating that the treated nanoparticles remained dispersed to near primary

particle size (Figure 3-14). The rationale for this behavior is that the silane groups

provided a physical steric barrier to approaching nanoparticles and provided a larger

interparticle distance in solution than the long range effects of VDW attractive forces.

The three amine groups present at the end of the silane groups repelled each other mainly

by similar dipole-dipole interactions.

The epoxy resin and curing agent were added to the flask in the ratio 3.55:1 (by

weight) and stirred with a homogenizer at 60 rpm until the resin was completely

solubilized in the particulate suspension. The flask was next affixed to a rotary

evaporator (RotaVAP, Buchi) and rotated at 125 rpm with a fixed water bath temperature

of 600C. The solvent was completely evaporated in a reduced pressure environment in

less than 30 minutes. The final solvent-free polymer particle mixture was cast into 1 mm

deep silicone wells and cured at 580C for at least 12 hours.

Table 3-5. Summary of components used to fabricate the fractionated sample.
0.5 wt% TiO2
EPON 828 9.7 g
Jeffamine D230 2.73 g
AEAPTMS 100tL
TiO2 particles 0.0625 g



























Figure 3-10. Dry titania nanopowder in the isopropanol and acetone solvent mixture,
showing good wetting of the particle surface.

14.00

12.00-

a, 10.00-

> 8.00

6.00-


0.0001 0.0010 0.0100 0.1000 1.0000
Particle Diameter (pm)


Figure 3-11. Particle size distribution measured after ultrasonication of the dry
nanopowder in a solvent mixture.


Figure 3-12. Digital image of the dispersed particles in the solvent mixture after
sedimentation and removal of the hard agglomerates.













20.00-

E
5 15.00-


10.00-


5.00-


0.00
0.0001 0.0010 0.0100 0.1000 1.0000 10.0000
Particle Diameter (pm)

Figure 3-13. Particle size distribution measured after sedimentation and removal of hard
agglomerates in the solvent mixture.

25.00


20.00-


5 15.00-


10.00-


5.00-


0.00-
0.0001 0.0010 0.0100 0.1000 1.0000 10.0000
Particle Diameter (pm)

Figure 3-14. Particle size distribution measured after aging of the surface treated
nanoparticles in the solvent mixture to verify that the particles remained
dispersed.

Solvent selection forfractionated sample

The choice of solvent to disperse the titania nanoparticles was dependent on two

factors; (i) the wettability of the particle surface by the solvent, and (ii) the ability of the

solvent to solubilize the epoxy resin upon mixing. Both acetone and isopropanol were

chosen as candidates due to the solubility parameters being close to that of the epoxy









resin (Table 3-6).91, 92 The solubility parameters are dependent on the cohesive energy

density, which is a measure of how easily the solvent molecules can penetrate the

polymer. Similar solubility parameters between the polymer and solvent indicate a "good

solvent" for a particular polymer.

Table 3-6. Solubility parameters for the chosen solvents to dissolve the epoxy polymer.
The proximity of the values to that of the epoxy represents a good solvent.
Solubility parameter (cal/cm3)12 Source
Epoxy polymer 8.9 Mezzenga, 2000
Acetone 9.9 Sperling, 2001
Isopropanol 11.5 Sperling, 2001

The Beer Lambert Law relation was employed to determine the extent of particle

surface wetting as it relates to the scattering of the particles or agglomerates. For

particles that have been ultrasonicated in a low viscosity solvent, it is assumed that good

particle wetting would produce a scattering response indicative of little or no multiple

scattering in the visible wavelength range due to dispersed particles. If the particles

segregate (phase separate) from the solvent due to poor wetting characteristics, the

agglomeration of the particles will be evident by an increase in scattering in the 400-

700nm wavelength range, which is observed as a deviation from the Beer-Lambert Law

(equation 3).

The optical density of a range of titania particle concentrations in both solvents was

measured by UV-Vis spectroscopy (see subsequent section on optical density). The

following linear relation holds for single scattering of particles at low concentrations that

are dispersed to a particle size that is small compared to the visible light wavelength:

A = alc (3)

Where A is absorbance, a is the absorption coefficient, I is the pathlength the light

travels through the sample, c is the particle concentration, and I represents the transmitted









and incident light. The wavelength range of interest for this study is visible light (400-

700 nm), as a measure of optical transparency. A sample of dispersed nanoparticles at

primary particle size will scatter a very small amount of light at those wavelengths, thus

producing a high transmitted intensity reading at the detector. However, if the sample is

agglomerated to aggregate sizes comparable to the wavelength of the source, multiple

scattering will occur and the amount of transmitted light will decrease. Conversely, a

highly agglomerated sample will produce a higher absorption measurement than that of a

dispersed sample of nanoparticles.

A Beer's Law plot of titania particles in acetone and isopropanol (Figure 3-15)

reveals that both solvents start to deviate from a linear relation as the particle

concentration was increased. However, if a trendline is fitted to the linear low

concentration portion of the curves (c < 0.2 g/L), the R2 goodness of fit term reveals that

isopropanol (R2 = 0.999 0.002) shows less scattering due to particle agglomerates, and

thus better particle wetting than acetone (R2 = 0.964 0.003). It can also be seen (Figure

3-16) that by adding a small amount of isopropanol to an acetone-nanoparticle mixture,

the particle dispersion is greatly improved due to better wetting of the particle surface by

the isopropanol.







47



2 5 -I--------------------------------


-- -Isopropanol 600nmn
-A-Acetone 600nm


15

-2
0


05-
05


0 005 01 015 02 025 03
Particle Concentration (g/L)


Figure 3-15. Beer's Law relation for optical density of titania nanoparticles dispersed in
acetone and isopropanol, measured at 600nm wavelength.



















Figure 3-16. Digital pictures of (A) 7g/L TiO2 nanoparticles in pure acetone where
sedimentation has occurred due poor wettability and (B) 7g/L TiO2
nanoparticles in 60% acetone/40% isopropanol with improved particle
wetting.

Even though isopropanol most effectively wets the titania nanoparticle surface, it


was found that it does not completely solubilize the epoxy resin used in this study. It was


determined through further optical density measurements that the difference in titania


particle wetting between pure isopropanol and a mixture of equal parts of both solvents


was not experimentally significant (< 5%). Solubility experiments revealed that a 60%









acetone solution was the lowest volume concentration that could completely solubilize

the epoxy resin without additional heating. Therefore, a 60/40% ratio of acetone to

isopropanol was used as an appropriate solvent mixture to both effectively wet the

particle surface and solubilize the epoxy resin.

Particle agglomerate sedimentation

The speed and time of centrifugation were selected such that the particles dispersed

by ultrasonication would stay suspended in the solvent and any agglomerated particles

would sediment out of the solution. The Stokes equation was applied to calculate the

settling velocity of the suspended titania particles

2r g -pf
V 9 (4)
9r7

where r is the radius of the particle (or agglomerate), g is the acceleration due to gravity,

p is the density of the particle and fluid, respectively, and r] is the fluid viscosity. The

optimal speed and time necessary to sediment any large agglomerates was calculated

according to the geometry of the centrifuge (Table 3-7). The particles or agglomerates

would need to travel no more than 3.5 cm to reach the bottom (or side) of the centrifuge

tube, based on the angle during centrifugation. It was determined that a speed of 5000

rpm for ten minutes would be sufficient to settle out the large particle agglomerates and

leave all of the dispersed primary particles suspended in the supernatant. From the

calculated values in Table 3-7, it can be observed that any particle aggregate or

agglomerate larger than 100 nm will settle to the bottom of the centrifuge tube.









Table 3-7. Stokes Law particle settling calculations based on ten minutes centrifuge
time. At 5000 rpm, the primary particles remain suspended, but the large
agglomerates (>100 nm) will settle out.
Size particles (nm) Settle Velocity (cm/s) Distance Settled (cm)
25 3.243E-04 0.195
75 2.919E-03 1.751
100 5.189E-03 3.114
300 4.670E-02 28.022
1000 5.189E-01 311.351

Optical Bench

The state of dispersion of the nanoparticles in the polymer matrix was

characterized, in part, by angular dependent light scattering measurements. The

construction of an optical bench was required due to the absence of commercial

instruments to make such measurements. Commercial laser diffraction instruments used

for particle size analysis commonly measure only a fraction of the full angular scattering

pattern and are not equipped to measure the scattering of solid materials, such as polymer

films. For that reason, an optical bench was constructed to be flexible enough to measure

different sample geometries, such as polymer films or cuvettes with liquid samples, at

any scattering angle. Many instances of custom-built laser light optical benches for

measuring particle scattering patterns are described in the literature and were used as a

guide for the one built for this study.5765' 93

Optical components

The general components needed for the most simplistic optical bench to measure

spherical particles are a radiation source, sample holder, and a detector mounted on a

rotating arm. Note that conversely the sample can also be rotated and measured with a

fixed detector. The instrument built for this study has a 632.8 nm wavelength HeNe

laser, beam expander, static sample holder with several interchangeable components for

different sample geometries, and a photon detector mounted on a goniometer arm, used to









control the measured scattering angle (Figure 3-17). The distance from the sample to the

detector can be changed to allow for maximum angular resolution, depending on the

particle size measured. When measuring small particles (<100 nm), it is important to

maximize the distance from the sample to the detector to identify changes in the

scattering pattern over fractions of a degree.

Optical components such as neutral density filters, a Fourier lens and various

apertures were added to the set-up for additional precision of the laser and scattering

measurements. It was also found that due to high sensitivity of the detector, a light

diffuser or aperture can be placed in front of the detector when measuring high levels of

scattered laser light. All of the major components of the optical bench, the manufacturer,

and part number are listed in Table 3-8. All other components were obtained from

typical optics suppliers (e.g., Edmund Optics, Melles Griot).

Table 3-8. Optical components of the light scattering instrument, the manufacturer and
part number
Optical component Manufacturer Part number
Laser Melles Griot 25LHR121-249
Beam expander Melles Griot 09LBM015
Goniometer Brookhaven Model BI-2030
Photon detector SensL PCSTime

Detector Selection

The measurable dynamic range of the detector must be such that even a small

amount of scattered light from small particles (<200nm) or turbid samples can be easily

measured. However, the range must be wide enough so that a very transparent film or

highly scattering sample will not saturate the detector. A CCD (charge coupled device)

camera detector was not able to provide enough precision to make the scattering

measurements in this study. Therefore, a highly sensitive photon detector (SensL) was










used. Although this type of equipment is more expensive, the limitations of cheaper

equipment are many, especially when measuring low levels of scattered laser light.


BEAM ExP~i44ER
--/l I~ --


GONIrOMErTE


Figure 3-17. Schematic representation (A) and digital picture (B) of the optical bench
set-up for light scattering measurements.

Laser calibration

When making sensitive scattering intensity measurements, it is important that the

incident radiation intensity remains stable within experimental error. If a systematic error

occurs, then it can be accounted for in the light scattering measurements of the

nanocomposite material. For that reason, the beam intensity was measured at 40









scattering angle for 8.3 minutes. Measurements of the intensity were taken every 5

seconds and normalized. The experimental error was approximately 1% for all

measurements and is assumed to be negligible for all measurements taken using this

instrument (Figure 3-18).

2



2 1-------------------------------
E 1.5
z
> 1
(C)
0.5

0
0 20 40 60 80 100
Measurement number

Figure 3-18. Intensity of the incident beam over 8.3 minutes with measurements taken
every 5 seconds. The error is assumed to be negligible (-1%).

Instrument calibration

In addition to checking the incident laser beam for fluctuations, it is also necessary

to verify that the instrument is providing meaningful experimental results. Several

samples were made that contain standard particles that are commonly used for particle

sizing calibration. These samples were made of polystyrene particle standards (Duke

Scientific Corporation) dispersed in water to check the validity of the measurements.

These particles are approved by the NIST organization as acceptable traceable materials

for particle size calibration (www.nist.gov). The polystyrene particles are available in a

range of sizes with very small particle size distributions ( 2%). These particles have

also been used in the literature for scattering instrument calibration.94 The samples were

measured on the light scattering bench and compare to the calculated Mie model. The









results show that the Mie model and the measured scattering intensities were similar

(Figure 3-19) for several particle sizes, indicating that the instrument was properly

calibrated to provide meaningful measurements of scattering intensity.

Snell's Law Correction

It is important to note that the samples used for calibration are particles dispersed in

water contained in a cuvette. Due to the refractive index difference between air and the

cuvette used to hold the sample, a correction had to be made due to refraction. According

to Snell's Law, a simple relation can be used to correct for the error in angular scattering:

n, sin(Oa)= nb sin(Ob) (5)

where the subscript a refers to the refractive index and scattering angle in air and the b

subscript refers to that of the suspension (in this case, water). This correction has been

used by others making similar measurements.59 However, the Snell's Law correction is

not needed for all samples. Since the pathlength of a cuvette is 10 mm, there is a large

difference in the angle of the scattered light due to the refractive index change. If the

pathlength is approximately 1 mm or less, the correction is not needed because the

change in scattering angle would be comparatively small. In the case of measuring the

polymer films used in the present study, the mathematical correction was not required.

















20

A
-Mieplot
2 -I-Experiment



15







05



0
0 5 10 15 20 25 3
Scattering Angle (Degrees)


inf


0 5 10 15 20
Scattering Angle (Degrees)


0 5 10 15 20
Scattering Angle (Degrees)


25 30


25 30


Figure 3-19. Scattering intensities for 600 nm (A), 800 nm (B), 2 [tm (C)and 5 [tm (D)

diameter polystyrene particles dispersed in water.


B
--MiePlot
SExperiment


m
v
1
'(
c
C,
c


- MiePlot
-* Experiment










I UU
D

10 Expenment





0.1


0.01
0 5 10 15 20 25 30
Scattering Angle (Degrees)

Figure 3-19 Continued

Scattering Measurements

Scattering measurements were carried out by placing the nanocomposite sample in

the path of a light or x-ray electromagnetic field and measuring the intensity of scattered

radiation at a given angle perpendicular to the scattering cross sectional area of the

composite film. The theoretical background of scattering of any wavelength radiation by

an array of particles is discussed in Chapter four. The resultant scattering pattern is

directly related to the following factors according to the general scattering equation:

* Refractive index (electron density) difference between particles and medium
* Particle size and size distribution
* Relative three-dimensional spacing of the particles
* Shape of the particles

It is assumed in this study that the refractive indices and particle (and agglomerate)

shape remain the same for all composite samples. Therefore, the objective of the

scattering measurements is to examine the effects of a change in particle (aggregate,

agglomerate) size and size distribution, as well as the relative arrangement of the particles

(aggregates, agglomerates) in space, as the state of dispersion is varied. These changes

can be observed by a change in the scattering pattern over several magnitudes of length









scale: nano to micron size features. Different length scales of a sample can be probed by

either changing the wavelength of the radiation, 2, or the scattering angle measured, 0 as

it relates to the q scattering vector. The q value (see Chapter four) is approximately the

inverse of the length scale probed by a given scattering technique. At very low q values,

the largest micron length scale (3 [tm < d < 90 nm) agglomerates were probed by static

light scattering. Mid-range aggregates were probed by ultra small x-ray scattering

(USAXS) and the smallest length scale (100 nm < d< 1 nm) of the individual primary

particles was measured with small angle x-ray scattering (SAXS) at the highest q values.

Light scattering (3 pm < d < 90 nm)

A custom made light scattering bench set-up was assembled to measure the angular

dependent light scattering patterns of the epoxy nanocomposite films. As described

previously, the instrument was calibrated with standardized particles to validate the

accuracy of such measurements. First, the SensL photodiode software was initialized and

a dark measurement (background correction) was taken for each angle included in the

scattering measurements. This measurement indicated the amount of stray light (or

noise) in each data point that can be subtracted off of the scattering measurements. A

very low dark count is desired (<1000), but it is nearly impossible to prevent any

environmental noise in the data.

The solid epoxy-TiO2 nanocomposite films were mounted onto standard optical

accessories (fixed lens holder) and placed on top of the goniometer. It is important that

the incoming laser impinges on the sample perpendicular to the nanocomposite film

surface so that the scattering is not affected by any angular misalignment. The software

that controls the stepper motor on the goniometer was initialized and the initial scattering

angle set at 5, to avoid sensitivity of the detector at smaller angles. The detector can









essentially be burned out if the laser is directly centered on the detector face. The

electronic stepper motor was calibrated to move the goniometer arm in precise 1

increments.

The SensL software was used to obtain the number of photon counts hitting the

detector and averaged over at least 5 seconds. The measurement interval was set at 2ms

and the amount of signal smoothing was adjusted to 50%. The measurement and

corresponding scattering angle were exported from the software to EXCEL for further

data analysis. The stepper motor was engaged to move the goniometer to 6 and the

measurement process repeated for the desired number of angular measurements. The

data collected was graphed with respect to scattering angle for the raw scattering pattern

of the nanocomposite sample. Each nanocomposite sample was measured three times to

eliminate experimental error. The raw pattern can be further analyzed using the Mie

model for an approximation of particle agglomerate size without complicated data

analysis.

The windows based program, MiePlot v.3.4.18 was chosen (www.philiplaven.com)

to quickly analyze the above light scattering data (Figure 3-20). This particular program,

developed by Philip Laven, calculates the theoretical scattering intensity of an array of

particles based on the refractive indices of the particles and surrounding medium

(polymer), and the size distribution of the particles (or agglomerates) by solving the

scattering equation by considering only Mie-type scattering. This program was also used

to analyze the scattering data used in instrument calibration, as described previously. In

other words, this program was utilized to theoretically model the scattering curve to

obtain a first approximation of the particle aggregate or agglomerate size distribution







58



from static light scattering and to aid in the alignment and calibration of the optical bench


set-up.


File View Advanced Window Help
-lnllensityscal-- Lightourc--- -Dropsize--- Ligh-
Sp gaiti P C Radiu L
j -Anle cale-- I lnlenslyv scaern ange \
Oep r Linear Minmimum angle Wavelength m] n vacuo
IM r rAaroConrced i ..I
Maximum angle fi Dsperse Reilacve de
--''- Real Irra glraly


intensity
100















0 20 40 60 70





log-normal size distribution.

Ultra small angle x-ray scattering (2 pm < d < 80 nm)

The instrument used for USAXS measurements is configured for greater scattering


intensity measurement resolution over a smaller angular range than static light scattering.


Due to the desired resolution in the range of 0-1, the x-ray source is collimated through a


slit instead of a point. The resultant raw data is therefore "slit smeared", and a correction


is typically used to compare USAXS measurements to other scattering data (light


scattering, SAXS). In this study, the Lake method was employed for slit smearing


correction. The x-ray source uses Cu Ka radiation at a wavelength of 1.54 A.


The polymer nanocomposite film sample was mounted in the commercial


instrument and left to run overnight to collect scattering pattern information. Due to the









small angular range, the detector crystal is translated and tilted instead of moved about

the 360 circle as in the SLS measurements. The measured scattering intensity, I(q) was

plotted with respect to the wave vector, q in the range 0.0003 A-' to 0.07 A-1. The

measurements were corrected for any background scattering that may occur due to

environmental conditions when no sample is in the chamber.

Small angle x-ray scattering (100 nm < d < 1 nm)

The SAXS scattering data were collected using a Bruker Hi-Star 2D area detector,

with x-rays produced by a Rigaku Ultraxl8 rotating anode x-ray generator operated at 40

kV and 60 mA. The Cu K, radiation was monochromated with pyrolytic graphite, with a

resultant wavelength of 1.542 A. A series of three pinholes was used to collimate the

beam before impinging on the nancomposite sample. A beamstop was used for better

measurement precision at the lowest angular measurements by absorbing the incident

beam to avoid interference with the scattered signal intensity. The instrument software

corrected the scattering data for any sample absorption and background scattering and

averages the data azimuthally. Corrections were also made using the accompanying

commercial software for differences in sample thickness, as changes in pathlength can

affect the scattered intensity measurement.

After the appropriate corrections, the scattered intensity, I(q), was plotted with

respect to the q value over the range 0.0075 A-' to 0.15 A-'. The scattering of the

polymer matrix background was not subtracted from these measurements, because the

scattering of the polymer was negligible as compared to the scattering of the embedded

particles. The experiment run time was based on the number of scattering centers within

the probed sample volume (beam volume). A higher number of scattering centers leads

to lower running times because a large amount of data can be collected at once. In order









to take a measurement, the nanocomposite sample was cut into the desired size

(approximately 2 in. x 1 in.) and mounted on the sample holder with double sided tape.

Scattering data analysis with IGOR

Scattering raw data required a great deal of analysis before particle (primary,

aggregate, agglomerate) size distribution and interparticle distance information can be

extracted. The software Igor Pro from Wavemetrics, Inc. was used along with the Irena

macros package that was developed by Jan Ilavsky at Argonne National Laboratories.

The Irena macros package is provided as freeware at

http://www.uni.aps.anl.gov/-ilavsky/irena.html. The Irena macros were developed to

analyze scattering data in the format of scattering intensity, I(q), with respect to the q

value for any material. The detailed steps used for data analysis and suggested fitting

parameters are listed in Appendix B, however, a brief description of the data analysis is

provided here.

The first step in analysis of the scattering data was to desmear the USAXS raw data

due to slit collimated rather than point collimated data. The Irena code for desmearing

utilizes the Lake method, which is an iterative curve-fitting procedure developed in 1967

and is described elsewhere.95 The raw data was trimmed to exclude large error

experimental data points, smoothed, and then desmeared according to the specified slit

length. The resultant desmeared data was exported to EXCEL and used for further

analysis.

The second step in data analysis was to combine the SAXS and USAXS data into

one continuous scattering curve. This was achieved by importing both the SAXS and

desmeared USAXS data into the Data Manipulation Tool. The USAXS data was scaled

to the SAXS data, which was measured in absolute units. The third step in the data









analysis process was to calculate the particle (primary, aggregate or agglomerate) size

distribution (PSD) and the interparticle distance for the titania nanoparticles embedded in

epoxy in various states of dispersion. First the combined SAXS and USAXs curve was

imported into the Least Squares Fitting Tool for standard models. The PSD of the

primary particles was also calculated using the SAXS data. The starting parameters for

all PSD fits are located in Appendix B. The interparticle distance, or correlation length,

was then calculated by using the Fractals Model fitting tool with the combined SAXS and

USAXS data. The shape of the particle aggregates and agglomerates were approximated

as spheres with an aspect ratio of one.

It is important to note here that the contrast parameter used for modeling must be

calculated to reflect the electron density (refractive index) difference between the

polymer and nanoparticles. The scattering contrast can be calculated using the scattering

contrast calculator function included in the Irena macros. Other parameters may be

manipulated in the various modeling functions, however, it is imperative that a standard

procedure be used with all samples when analyzing scattering patterns of multiple

samples for reproducible results.

Additional Characterization

Optical density

The optical density measurements of titania nanoparticles dispersed in a solvent

and embedded in the epoxy were made using a Perkin Elmer UV/Vis Spectrometer 900.

The measurements were made over the spectral range 400-700 nm because the visible

wavelength range is of interest to this study for optical measurements. Custom-made

sample holders were designed for these measurements to hold solid polymer films by

cutting a hole in an aluminum L-bracket with a hole drilled large enough for the









transmitted beam to pass through. The bracket was painted black for minimal reflection

of the laser. The sample was affixed to the bracket with magnets and placed into the

sample holder (Figure 3-21).



















Figure 3-21. Custom-made sample holders for measuring the optical density of solid
polymer films in the UV-Vis spectrometer.

Commercial light scattering instruments

Dynamic light scattering measurements were made to determine the particle

(primary or agglomerate) size distribution of the titania nanoparticles in a solvent. A

Microtrac Nanotrac Ultra commercial dynamic light scattering instrument with external

measurement probe was used. The useful range of particle sizes measured by this

instrument is in the range of 0.8 to 6500 nm, which is useful for nanoparticles that have a

30 nm primary particle size. Static light scattering measurements were used to determine

the particle size distribution of the dry TiO2 nanopowder before any dispersion

techniques were used. The static light scattering measurements were made on a Coulter

LS 13 320 particle sizing instrument, with a useful measurement range of 0.4 to 2000 am.









for the dry powder system. All scattering data from commercial instruments was

analyzed using the accompanying commercial software.

Transmission electron microscopy

Transmission Electron Microscopy (TEM) samples were fabricated by slicing the

epoxy-TiO2 composite samples with a Leica ultramicrotome and a Diatome diamond

knife (450) into 100-200 nm thick sections. It was discovered though experimentation

that thinner slices were very prone to beam damage in the microscope due to electron

charge build-up in the sample and subsequent heating. The samples used for sectioning

are made by casting the polymer-particle mixture before cure into polyethylene molds

made specifically for TEM sectioning (Figure 3-22). The molded samples were removed

from the molds after cure and roughly trimmed at the small end with a razor blade into a

trapezoid shape having the dimensions not exceeding 0.5 mm across the base, 0.4 mm

across the top, and 0.3 mm along the sides. The diamond knife was then used for

sectioning. The sections are carefully collected from the water in the boat of the diamond

knife after slicing and mounted onto 300 mesh carbon grids (Electron Microscopy

Sciences). The grids were examined with a JEOL 200CX at 200kV at various

magnifications to determine the qualitative difference between the four nanocomposite

samples with varied particle dispersion and to verify the primary, aggregate and

agglomerate particle size measured by scattering techniques.





















Figure 3-22. Polyethylene embedding molds used for TEM polymer sample preparation.

Scanning electron microscopy

Scanning electron microscopy (SEM) micrographs of the nanocomposite samples were

taken to examine the nanoparticle dispersion in addition to TEM. Samples for SEM were

prepared by immersing a sample prepared for TEM into liquid nitrogen for approximately

two minutes. The sample was removed from the liquid nitrogen and fractured into

several pieces. The smaller "shards" of the nanocomposite sample were mounted onto an

aluminum disk with carbon tape. The entire disk was coated with a thin layer (2-3 nm) of

amorphous carbon to prevent charging of the polymer. The samples were viewed with a

JEOL 6335F Field Emission SEM at 10 kV and a working distance of 15 mm. The SEM

working parameters are indicated on the micrographs shown in this study.














CHAPTER 4
SCATTERING THEORY

In this chapter, the theoretical background of scattering techniques and the rationale

as to why they are used to measure a certain length scale of the nanocomposite samples is

disclosed. The scattering theory for SLS, USAXS, and SAXS measurements is first

presented and the probe length scale is explained. A brief explanation of Mie theory is

outlined as it relates to the modeling of the static light scattering (SLS) measurements.

General Scattering Theory

Scattering measurements entail the quantification of the amount of electromagnetic

energy that is scattered at a specific angle from the composite sample volume of

scattering centers, as compared to the amount of electromagnetic energy that is

transmitted directly through the sample. The scattering of electromagnetic waves

(radiation) is due to the difference in electron density (optical properties, i.e., refractive

index) between the scattering centers (particles) and the surrounding medium. All

scattering techniques are governed by the same physical equations, regardless of the

wavelength of the radiation source.

The scattered intensity can be determined from the Poynting vector, S, which

specifies the magnitude and direction of the rate of transfer of electromagnetic energy

(scattered intensity) at all points of space relative to the scsatterer.96 The Poynting vector,

given by E x H, represents the direction in which energy is being transported (the

scattered direction) and has a magnitude equal to the power density of the wave (scattered

intensity).









The electric and magnetic fields (E and H) of an electromagnetic wave oscillate,

thus the magnitude of the Poynting vector changes with time. Most instruments cannot

measure an instantaneous Poynting vector and as a result, a time averaged value is used.

The average magnitude of the Poynting vector over a long time T (longer than the period

of the wave) is called the irradiance or intensity, .97


(S) = S(t')dt'= I (1)


The time averaged Poynting vector for time-harmonic fields is given by96


(S) = ReEx H* (2)

where E and H cannot be arbitrarily specified, and must satisfy the Maxwell equations

for a plane wave96:

V E= 0
V H= 0
8H
VxE = -/ (3)
at
dE
VxH= e
at

where p is the permeability of radiation through the medium (surrounding material) and e

is the permittivity, which is also a property of the medium. The E and H vectors are

broken down accordingly into the field associated with the particle (El, Hi), the

surrounding medium (E2, H2), the incident field (Ei, Hi) and the scattered field (Es, Hs).









A



I.-




P ParticleE1, H1 I


IncddentEi, ---
Scattered E H,







Figure 4-1. Forward scattering of a sphere showing the electromagnetic field (wave)
vectors for the incident, particle, and scattered radiation.

Boundary Conditions

When crossing the boundary between the surface of the particle and the

surrounding matrix, there is a sudden change in material properties, which is reflected in

a sudden change in optical properties, e.g., e and y. However, some conditions are

imposed on the E and H vectors that preserve the conservation of energy across that

boundary. These rules maintain that the tangential electromagnetic field (from the

surface of the particle) is continuous across a discontinuous boundary96

[E2- E]xf = 0
(4)
[H 2 H1]x fi = 0

where both E and H lie on the surface of the particle.









The Amplitude Matrix

If one considers a particle (or an array of particles) that is illuminated by a plane

wave, the direction of propagation (the z axis) defines the forward direction. The

incident electric field, Ei, can be resolved into the parallel and perpendicular components

respectively, Eli and Er,i. The scattering plane is determined by the azimuthal scattering

angle, p, and the scattering angle, 0. The scattered electric field is defined as the vector,

Es, which can also be resolved in to the parallel and perpendicular components.


Scattering direction
Z




scattering plane
particle











Figure 4-2. Schematic depicting the scattering of electromagnetic radiation by a
spherical particle. The incident field is in the z-axis direction.

The scattering in any direction is described by four amplitude functions, Si, S2, S3,

S4, which are all functions of the angles p and 0 and can be measured by determining the

amplitude and phase of the scattered radiation96. The amplitude function may be

complex, but rarely has it been physically measured for all angular values. The relation

between the incident and scattered electric fields is given by:97









E, S2 S3 e -kr+ kz E 1(

E,s, S4 S, ikr kEr,

where S is the amplitude function matrix. For this case, the scattered electric field is

given by two relations:97

-kr +kz
Er = S(0) e E
ikr (6)
-kr+ikz (6)
El, = S2(0)e
ikr

For spherical particles, S3 = S4 = 0, and Si and S2 are functions only of 0, due to the

scattering symmetries over the full 0-3600 scattering range of 0. In general, the azimuthal

angle can be ignored if the spherical particle assumption can be used. In the present

study, all particles and particle agglomerates are assumed to be spherical. The

calculations involved in calculating the scattered intensity from asymmetrical particles is

beyond the scope of this research. The intensity of the scattered wave for spherical

particles is proportional to the square of the amplitude as follows:

S2() 2r
Iscatt 2 2 k (7)
k 2 r A

For various polarization states of the incident radiation:

s, (o)2
I polarizaton k 2 2

S()2 2
n2ur (s)
// polarization k 2 2 (8)

SIS, (0) 2 + S2 (2
natural light k 2 r 2









For a system of spherical particles, not necessarily similar, each one is characterized by a

unique amplitude function, Si(q). Due to mathematical complexities, the intensities

associated with each amplitude are summed over the entire system97

I(0) = I (0) (9)


The Mueller Matrix

The incident light emitted in a scattering experiment actually consists of many

"simple waves" in very rapid succession. Therefore, measurable intensities always refer

to a superposition of many millions of simple waves with independent phases.97

Following the summation shown above, the Stokes parameters of the beam are

I = YIs

(10)




The factors Q, U, and V are dependent on the polarization of the incident light.97 The

relation between the incident and scattered Stokes parameters make up the full amplitude

scattering matrix, which is often denoted by the matrix F.96



Q 1 S,21 S22,, S,23 S24
U, k2 2 S31 S32 S33 S34 U,
V, S 41 S42 S43 S44 V

The 4 x 4 amplitude matrix is known as the Mueller matrix for scattering of a single

particle. The 16 elements are not all independent and can be determined by knowing

seven individual parameters. The number of parameters may also be reduced because of

symmetry relations. If a collection of randomly distributed particles is present, as in our








model nanocomposite system, the Stokes parameters of the system can be determined by

summing the effects of the individual particles.

Scattering of a Sphere

The following discussion relates only to a single sphere or an array of spheres,

which results in simplifications of the wave equations. Starting with the incident and

scattered electric field amplitudes, the relation becomes96


= Ks .ofe (12)
Er, 0O S1 ikr E(12)

and the Mueller matrix can be simplified as

i / 1S,1 S12 0 0 :I
Q1 S12 0S,, 0 Q,
U k2r2 0 0 S33 S34 U,
V 0 0 S34 3 S3,,V

For the incident radiation in the case of the experimental light scattering set-up used in

this study, the laser is unpolarized and the Stokes parameters can be written as96

Is = S1,I, Qe = S12,,, U, = V, = 0 (14)

Angular Dependent Scattering

In the scattering set-up described previously, the angular dependent forward

scattering (0-90) intensity is measured. The first few scattering coefficients, as

determined in the previous section, will determine the angular dependence of each

scattering event, or particle96. For forward scattering (0 < 90), the intensity plot is highly

peaked whereas the backscattering intensity plot lacks any strong features and is lower in

intensity by orders of magnitude. The pronounced difference between intensities for

forward scattering and backscattering becomes even more dominant as the size of the









particles, or scattering events increase. This factor is very important to the current study

due to the fact that it is assumed that the particle agglomerates are scattering as individual

events and increase in size as the particle agglomeration increases. So, as the size of the

scattering event ranges from primary particle size (nanometers) to large agglomerate

(microns), the forward scattering component increases much faster than the backscatter

component.

The Length Scale

The vector q is called the scattering wave vector. The importance of this vector is

that its approximate inverse represents the length scale, or probe length, of the scattering

experiment19. The q value corresponds to the actual feature size, or length scale, that is

measured by d = 2i / q where dis the feature size. This is the reason for the inverse

relationship between the particle size and the angle at which the features of that size are

observed in a scattering experiment. Consider an incident plane wave, ki on a scattering

element at r. The field scatters toward the detector in the direction ks. If purely elastic

scattering is assumed then75


kj = ks = 2 (15)


The phase of the scattered wave at the detector is a function of the position of the

scattering element and the vector

q = k, k, (16)

which is the scattering wave vector (Figure 4-3). Simple trigonometry and equation (17)

yield the magnitude of this vector as










q = 2k, sin
(17)
q = sin -
q y2

where 0 is the scattering angle.


Phase at detector
em I(q)


I0 tP,










Figure 4-3. Diagram of light incident from the left with wave vector ki scattering from an
element at r toward the detector with scattering wave vector ks, at a scattering
angle 0. The difference in wave vectors, q, is shown. Adapted from Sorensen
(2001).

Mie Theory

Mie Theory was first presented by Gustav Mie in 1908 and is the direct analytical

solution for the scattering, absorption, and transmission of radiation through a dielectric

sphere48. Recently, Mie Theory has come to include all calculations of the interaction of

incident light with particles or material of any shape, size, size distribution, or texture.

However, an exact solution only exists for the single spherical dielectric particle, as

mentioned above. Mie Theory can be used to calculate a particle size distribution from

scattering intensity profiles via a matrix inversion the details of which can be found in

Appendix A. The intensity profiles generated through static light scattering are used to









determine the particle and/or cluster particle size distribution, required for the dispersion

indices calculation. The derivation closely follows that of Bohren and Huffman (1983)

and also van de Hulst (1957) for the scattering of light by small particles.96-97 The

assumptions made are that the particles are spherical, the incident wave satisfies the

Maxwell equation, and the appropriate boundary conditions are valid for the titania

particles embedded in an epoxy matrix.

Mie Scattering Coefficients

The boundary conditions for the electromagnetic field, as explained above, require

the continuity of the transverse components of the electric and magnetic field at the

surface of the sphere. The transverse components of those fields are represented by

Bmn(O,() and Cmn(,(p). The coefficients of those terms can be expressed as a function of

the internal and external fields (of the particle)83


cn a.P' = a ) ca n ( a),
m (18)
d- ( ?) = V n b ( ) V (n
m
CnVn{P a= (a ) Vn' (a),

where a = ka and / = mka are the external and internal size parameters of the particle and

In and n are the spherical Ricatti-Bessel functions. The solutions of these equations

yield the Mie coefficients, of which a, and b, govern the scattering of the radiation and c,

and d, govern the internal field.98








fn f(a)/( -
V (~i, (8)v
an 18n(a?) -
Vfn
in
bn Vf fn(a7)^
M f ^ (On)(a)


M V (On' (a) '
- Vnfvn(a)
- Vfn(On'()a)


(19)


im
d n =(a)- w

The scattering coefficients, an and bn can be determined for an arbitrary spherical particle
and will completely specify the scattered field.














CHAPTER 5
RESULTS AND DISCUSSION

Dispersion Force Calculations

A rough calculation of the van der Waals adhesive forces between two touching

titania particles in an agglomerate yields approximately 93 nJ/m by the following

equation

A RR 2
VD 6H R R, + R (1)


where A is the Hamaker constant for two bare (untreated) titania surfaces as suggested by

French"9, H is suggested as 4 nm for a contact interparticle distance99 and R is the

primary particle radius of 30 nm. That amount of energy, provided at the nano scale at

each particle-particle interface, is required to overcome the van der Waals barrier and

completely break all particle agglomerates for complete dispersion. Additionally, forces

must be provided to prevent subsequent contact of the particles to maintain the particle

dispersion.

In addition, the free energy of the spreading process of the epoxy resin on the bare

titania surface is a large positive value (non-spontaneous), indicating a high probability of

poor wetting of the surface by the polymer and segregation of the nanoparticles (primary

or aggregates) into larger agglomerates. The spreading coefficient, S, can be calculated

using the surface energies of the polymer and particle components as an indication of the

ability of the epoxy resin to flow over and wet the particles. A positive value reflects a

kinetically non-spontaneous wetting condition, whereas a negative value indicates a









spontaneous wetting process, if no other forces are present on the surface affecting

wetting (e.g., surface roughness). The spreading coefficient can be calculated as

follows100

SS12 = CSV LV -- SL (2)

where a is the surface energy of the respective components (titania particle, epoxy,

particle-epoxy). The surface energy of the epoxy was obtained from the work of Page101

and that of the nanoparticles from the work of Navrotsky.102 The surface energy of the

titania particle-epoxy interface was approximated from the Neumann model.103 These

values are listed in Table 5-1. The calculated spreading coefficient for the bare titania

surface, -2527 mJ/m2, indicates a poor compatibility between the particle and epoxy

resin. However, the silane coupling surface treatment (APTES or AEAPTMS) lowers the

surface energy of the particle to 36.5 mJ/m2 (from 1036 mJ/m2) and the spreading

coefficient is increased greatly to 0.19 mJ/m2, which suggests that that epoxy resin would

spontaneously wet the titania nanoparticle surface, given no other forces are present. The

lowering of the surface free energy of the titania and subsequent diffusion of the polymer

in between nanoparticles is a competitive force to the interparticle van der Waals

attractive forces, thus suggesting a decrease in agglomeration.

Table 5-1. Surface energy values for the nanocomposite components and the calculated
spreading coefficient for untreated and treated particles.
Surface Energy After treatment with
(mJ/m2) AEAPTMS (mJ/m2) Reference
TiO2 1036 (ysv) 36.5 (ysv) Navrotsky (2007)
Epoxy 36.3 (yLv) 36.3 (yLv) Page (2001)
TiO2-Epoxy 3527 (sL) 0.01 (TsL)
S12 -2527 0.19









Qualitative Nanoparticle Dispersion Characterization

In order to examine the affects of calculated surface forces on the dispersion of

titania nanoparticles in an epoxy polymer, microscopy was used as the first method of

investigation.

Digital Images

Digital images provided a qualitative indication of the optical density (transmitted

light) of each sample of varied nanoparticles dispersion. The thick films (1 mm) created

in this study were mounted against typewritten pages to illustrate the clarity of the film

for comparison to one another (Figure 5-1). It is obvious from these images that the

fractionated sample has the highest transparency as compared to the other samples.

It can be suggested from these initial results that the new methodology for

fabricating the fractionated sample using a solvent for enhanced particle dispersion must

be employed to produce a nanocomposite with desirable optical properties for use in

applications such as windows and food packaging. According to the Rayleigh regime of

2zr
particle scattering, x = where x is the characteristic Rayleigh size parameter (x


<<1), the particles must be dispersed until a nano size range is reached so that the

agglomerate size is small (-X/10) compared to the wavelength of visible light (400-700

nm) to avoid multiple scattering effects and resulting sample opacity. In other words, the

scatterers must be dispersed to approximately 40-70 nm in diameter to provide an

optically transparent sample. In light of this, it can be surmised from visual inspection of

Figure 5-1, that samples A, B, C and E have agglomerates larger than 70 nm due to the

opacity of the films.



















..ie v ohu
ih Ip h"i m M'Be S llal
h..'.1.

Ce' r r* Iy




(' I:
L, r ., .....i hi lv. In l wouldd int ere pt iih s rmaxi mu
ei. Mf '"". f "erametern, wFglhout an\ conl It
,IMIlllf "'' "..t virtuaI sores emhling thi
retd h h eary, he rrr mages a
kie.r.. l Ifyou &an read
*...C a-,.*Crd IY, t,,,yh, pajclc. n iould be Iodte that this kim,-]

d., ar m Irall If you cam r ead ded lt
-i.i iicitni,'W IfYCIMI ,idh arn eailfldeC. lS-'rti)C il1,

Figure 5-1. Examples of digital images comparing thick films of each dispersion state
against a typewritten background. The samples are: (A) undispersed, (B)
ultrasonicated, (C) surface treated, and (D)fractionated. Sample (E) is
undispersed and (F) is thefractionated sample. Sample (G) is the pure epoxy
polymer.

Freeze Fracture SEM Micrographs

Freeze fracture scanning electron microscopy (FF SEM) also provided the

qualitative differences between nanocomposite samples due to variations in particle

dispersion. SEM images also reveal the topography of the fracture surface and a clear

indication of the voids surrounding the particles due to poor wetting. The undispersed

sample micrographs (Figure 5-2) reveal not only large agglomerates and surrounding

voids, but also discontinuities in the epoxy polymer fracture surface at points where the

agglomerates are present (picture C, Figure 5-2). The particle agglomerates provided a

point of crack initiation and failure at the mechanically weakest part of the sample, as

described in the literature review. It can be suggested from these preliminary results that









the mechanical properties of the undispersed sample would be poor, and possibly worse

than that of the pure epoxy polymer.

The fractionated sample micrographs (Figure 5-3) indicate that the nanoparticles

were dispersed to near primary particle size and only a small number of clusters exist at

the fracture surface. No discontinuities or voids in the polymer matrix can be identified.

The large improvement in the particle surface wetting for the fractionated sample is due

to the silane coupling surface treatment applied to the titania surface as described in the

previous sections, which lowers the surface energy of the particle by over an order of

magnitude (Table 5-1).


Figure 5-2. Scanning electron micrographs of the undispersed sample showing the large
particle agglomerates and voids due to poor particle wetting by the epoxy.
The voids are most obvious in picture (B).






















Figure 5-3. Scanning electron micrographs of the fractionated sample showing the small
particle agglomerates and individually dispersed particles. No voids are
visible in the sample due to good particle wetting by the epoxy.

TEM Micrographs

Transmission electron microscopy (TEM) was used to characterize the nano scale

state of dispersion by investigating both the average size of the particle agglomerates

(and aggregates) and also the distance between agglomerates. This method was also used

to obtain approximate values for particle agglomerate size for the modeling program

(Irena), utilized to analyze the scattering data. Figure 5-4 reveals the difference in

agglomerate and aggregate sizes between the differently dispersed nanocomposite

samples.

The undispersed sample has very large agglomerates, on the order of one micron,

which is more than one order of magnitude larger than the primary particles (30 nm).

The simple mixing of the dry nanopowder into the epoxy resin and hardener with no

surface modification provided a highly agglomerated and inhomogeneous composite

material. The as-received dry nanopowder contained many large agglomerates due to the

interparticle van der Waals attractive forces, and mixing provides very little energy to

overcome those forces, thus a highly agglomerated sample is observed.









The ultrasonicated sample agglomerates are approximately 500 nm in diameter,

still much larger than the primary particles. The sample was prepared by ultrasonication

in addition to mechanical mixing of the TiO2 nanoparticles into the epoxy resin. This

procedure results in some reduction in the size of the particle agglomerates. This

reduction is due to the energy input of the ultrasonicator to overcome the van der Waals

(VDW) forces. Ultrasound wavelengths cannot produce dispersion eddies on the nano

length scale, regardless of energy input and duration of ultrasonication, as described by

the Kolmogoroff equation23 24, thus preventing full dispersion of the particles to primary

particle size and bonded aggregates. Additionally, the bare titania surface remains

incompatible with the epoxy resin (no surface treatment) and further prevents particle

dispersion.

The surface treated sample contains agglomerates that are approximately 250 nm

in diameter. The sample preparation involved silane surface treatment in addition to

ultrasonication and mechanical mixing of the particles into the polymer. This procedure

resulted in a further decrease in agglomerate size due to the surface modification of the

titania nanoparticles. This is attributed to the APTES molecules covalently bonding to

the hydroxyl groups present on the metal oxide surface and reduction in the free energy

of mixing to that of a spontaneous process (Table 5-1, similar values as AEAPTMS).

However, as indicated by the above sample preparation technique (ultrasonicated

sample), ultrasonication is not able to fully disperse the nanoparticles before the surface

treatment was applied, thus essentially coating the surface of the agglomerates. The

surface treatment did, however, prevent further agglomeration during mixing and curing









of the epoxy resin and hardener due to steric repulsion. Therefore, a reduction in

agglomerate size from the ultrasonicated sample was observed.

The high resolution TEM images of the fractionated sample (Figure 5-5)

demonstrate that individual particle separation occurs due to good surface wetting by the

epoxy resin. The small clusters seem to be approximately 60-70 nm. Thefractionated

sample involves the use of a solvent to reduce viscosity and increase particle wetting in

addition to surface treatment, ultrasonication, and mechanical mixing. This procedure

generates the smallest particle size and most optimal dispersion of the dry titania

nanopowder in the epoxy matrix achieved in this study. The details and rationale for this

result were shown in Chapter three.

In this study, images provided not only qualitative verification of the state of

nanoparticle dispersion, but also a priori knowledge to formulate an approximate value of

particle (primary, aggregate and agglomerate) size for modeling purposes. The objective

of this study was not to prove that microscopy should not be used for the assessment of

particle dispersion, but rather to show that it can be used with success to complement

more quantitative techniques.













Figure 5-4. Transmission electron microscopy images of 0.5 wt% TiO2 embedded in
epoxy: undispersed (A), ultrasonicated (B) and surface treated (C).











ACAE
-. .~. ~ -


-,% J , .,





Figure 5-5. High resolution TEM micrographs of the fractionated sample (0.5 wt%)
showing the nano scale dispersion and separated individual particles. Note the
difference in scale bars.

Quantitative Measure of Dispersion

It is clear from the qualitative assessment of dispersion that one could perhaps

reliably compare good and bad samples and estimate the approximate average particle

agglomerate size, but it would be extremely difficult to obtain any statistically relevant

information without tedious and time consuming evaluation over a representative sample

area. It is for these reasons that the desire for a more quantitative measure of dispersion

led to the development of a dispersion index concept.









Rationale for the Dispersion Index

The dispersion index concept was developed in this study to obtain a quantitative

measure of the state of dispersion of the nanoparticles embedded in the polymer matrix.

Literature review and preliminary efforts indicated that it may be difficult to quantify

dispersion in a single parameter considering particle-particle interactions are governed by

surface chemistry and spatial distribution are influenced by physical mixing forces.

Accordingly, the nano length scale changes in agglomerate size due to surface chemistry

effects, and also the macro length scale mixing effectiveness of the nanocomposite

fabrication process needed to be taken into account. Thus, the dispersion concept was

developed to include two parameters: one to describe the nano scale wetting of the

individual particles (D1), and the other to reflect the relative particle arrangement in the

matrix (D2) due to mixing. Both of the dispersion indices were expected to range in

value from zero to one: a value of zero representing an infinitely agglomerated system

and a value of one representing a theoretically ideal, random dispersion of primary

particles, homogeneously distributed throughout the polymer matrix.

Surface Chemistry Parameter- D1

The surface chemistry parameter, D1, reflects the compatibility and wetting of the

titania nanoparticle surface with the surrounding polymer on the nano length scale. The

wetting that contributes to particle dispersion is best represented by the particle

agglomerate size and size distribution, as increased wetting has been shown here to

enhance dispersion of the nanoparticles and reduce agglomerate size.104 When an

appropriate solvent was used to increase wetting of the particles and a surface

modification was employed to maintain the nanoparticle dispersion, the epoxy resin,

when added, can permeate the particles and wet the surface of individual particles.









Therefore, increased wetting of the particles is expected to produce smaller agglomerates,

thus creating a narrower size distribution and smaller average particle size that closely

resembles that of a theoretically ideal dispersion (all primary particles).

In this study, the full width half maximum (FWHM) value of the particle

(agglomerate) size distribution is used as a measure of the breadth of the distribution.

The FWHM value combined with the mean particle (or agglomerate) size is used to

compare a theoretically ideal particle dispersion to the experimental nanocomposite

samples as follows,

FWHMWd(
D, = (3)
FWHM, d,

where the subscripts WD and N represent the randomly well dispersed and experimental

nanocomposite samples, respectively. The variable, d, is the average particle or

agglomerate size. The ratio of the average agglomerate size as compared to the primary

particle size of an ideal sample was suggested by Balek25 as a reasonable measure of

dispersion.

In order to demonstrate the relevance of the D, parameter, four different particle

size distributions, including primary particles, aggregates, and agglomerates) were

simulated using a log-normal distribution calculation. Each case reflects a change in

mean particle size and/or distribution breadth from the ideal dispersion of nanoparticles

in Case 1 (Figure 5-6), however, a homogeneous spatial distribution of particles or

agglomerates is assumed. The following four simulated size distribution cases are shown

graphically in Figure 5-7:

* Case 1: the ideal case of randomly well dispersed particles. The average particle
size is 30nm and the standard deviation is 6nm.




Full Text

PAGE 1

INVESTIGATION AND CHARACTERIZ ATION OF THE DISPERSION OF NANOPARTICLES IN A POLYMER MATRIX BY SCATTERING TECHNIQUES By MARIE CHRISTINE KISSINGER-KANE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

Marie Christine Kissinger-Kane 2007

PAGE 3

iii To my husband, Kevin

PAGE 4

iv ACKNOWLEDGMENTS I would first like to acknowledge my husband, Kevin, and my parents for their constant love and support. Without them, I am sure this would ha ve been much more difficult to accomplish. Secondly, I would lik e to thank my advisor, Dr. Brij Moudgil and my committee co-chair, Dr. Anthony Bre nnan. Their knowledge and guidance have been invaluable tools throughout my graduate school journey. Dr Moudgil, I thank you for thoroughly preparing me for my future career and making me the confident person that I am today. Dr. Brennan, I thank you for your unending encouragement and for not letting me take myself too seriously. I woul d like to also acknowledge the rest of my graduate committee for their guidance and supp ort: Dr. Laurie Gower, Dr. Kevin Powers, Dr. David Tanner, and Dr. Charle s Beatty. My special thanks to Dr. Powers; his out of the box ideas and countless discussions with me, academic or not, fueled my passion for learning and discovery. I would also like to acknowledge those outsi de of the University of Florida that have provided assistance. Dr. Rick Beyer at the Army Research Labs at Aberdeen Proving Ground provided the small angle x-ra y measurements, and Dr. J. David Londono at DuPont assisted with ultra small angle xray experiments. Their collaboration has been an invaluable tool and I would now consid er myself among a small elite group of x-ray scattering specialists. Next, I would like to acknowledge my best friend and colleague, Sara Jensen (Homeijer), for her friendship, encouragem ent and assistance (patience) with TEM

PAGE 5

v micrographs. Next, I would like to acknowledge the faculty and staff of the Particle Engineering Research Center. Especially, I would like to thank Mr. Gary Scheiffele for laboratory help and Gill Bruba ker for helpful discussions on Mie Theory. I would also like to thank my friends and colleagues at th e Particle Engineering Research Center for their laughter, friendship, and helpful discussions. The Particle Engineering Research Center (P ERC) at the University of Florida, the National Science Foundation (NSF Grant EEC-9 4-02989), the United States Air Force, and the Industrial Partners of the PERC are graciously acknowledged for financial support of this research. Finally, I would like honor the memory of my late committee member, Dr. Abbas Zaman, who took the time to show me the va lue of rheology measurements and always had a smile and kind word for me. He will be sorely missed.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES.............................................................................................................ix LIST OF FIGURES.............................................................................................................x ABSTRACT.....................................................................................................................xiv CHAPTER 1 INTRODUCTION........................................................................................................1 2 LITERATURE REVIEW.............................................................................................5 Particle Dispersion Studies...........................................................................................5 Particle Surface Modification................................................................................6 Mechanical Dispersion..........................................................................................7 Dispersion Equations....................................................................................................9 Correlation of Dispersion w ith Material Properties....................................................12 Mechanical Properties.........................................................................................12 Optical Density....................................................................................................13 Molecular Modeling of Na nocomposite Dispersion...................................................13 Light Scattering..........................................................................................................15 Background..........................................................................................................15 Particle Size Measurement..................................................................................17 Small Angle and Ultra Small Angle X-ray Scattering................................................18 X ray Data Analysis.............................................................................................19 The Guinier regime......................................................................................20 The power-law regime.................................................................................21 A Unified Fit model.....................................................................................22 The Porod regime.........................................................................................22 Scattering Techniques for Dispersion Measurements.........................................23 Microscopy.................................................................................................................24 ASTM Standard for Dispersion of Carbon Black in Rubber......................................24 Scope of Proposed Research.......................................................................................27 3 MATERIALS AND METHODS...............................................................................29

PAGE 7

viiMaterials.....................................................................................................................29 Nanoparticles.......................................................................................................29 Polymer Matrix....................................................................................................32 Particle Surface TreatmentSilane Coupling Agent...........................................34 Methods......................................................................................................................35 Nanocomposite Formulation...............................................................................35 Four levels of particle dispersion.................................................................36 Solvent selection for fractionated sample....................................................44 Particle agglomerate sedimentation.............................................................48 Optical Bench......................................................................................................49 Optical components......................................................................................49 Detector Selection........................................................................................50 Laser calibration...........................................................................................51 Instrument calibration..................................................................................52 Snells Law Correction.................................................................................53 Scattering Measurements.....................................................................................55 Light scattering (3 m < d < 90 nm)............................................................56 Ultra small angle xray scattering (2 m < d < 80 nm)................................58 Small angle x-ray scattering (100 nm < d < 1 nm).......................................59 Scattering data analysis with IGOR.............................................................60 Additional Characterization.................................................................................61 Optical density..............................................................................................61 Commercial light scattering instruments......................................................62 Transmission electron microscopy...............................................................63 Scanning electron microscopy.....................................................................64 4 SCATTERING THEORY..........................................................................................65 General Scattering Theory..........................................................................................65 Boundary Conditions...........................................................................................67 The Amplitude Matrix.........................................................................................68 The Mueller Matrix.............................................................................................70 Scattering of a Sphere..........................................................................................71 Angular Dependent Scattering............................................................................71 The Length Scale........................................................................................................72 Mie Theory.................................................................................................................73 Mie Scattering Coefficients........................................................................................74 5 RESULTS AND DISCSUSSION...............................................................................76 Dispersion Force Calculations....................................................................................76 Qualitative Nanoparticle Disp ersion Characterization...............................................78 Digital Images.....................................................................................................78 Freeze Fracture SEM Micrographs.....................................................................79 TEM Micrographs...............................................................................................81 Quantitative Measure of Dispersion...........................................................................84 Rationale for the Dispersion Index......................................................................85

PAGE 8

viiiSurface Chemistry ParameterD1.......................................................................85 D1 Calculation for the Nanocomposite Samples.................................................88 Mixing Parameter D2..........................................................................................89 D2 Calculation for the Nanocomposite Samples.................................................92 Scattering Techniques.................................................................................................94 Small Angle X-Ray Scattering (SAXS)..............................................................96 Ultra Small Angle X-Ray Scattering (USAXS)..................................................97 Static Light Scattering (SLS)...............................................................................99 Scattering Data Analysis...................................................................................100 Dispersion Parameter Calculations...........................................................................105 Relationship Between the Dispersion Pa rameters and Material Properties..............109 6 CONCLUSIONS AND FUTURE WORK...............................................................113 Nanocomposite Fabrication......................................................................................113 Suggested Future WorkFIB....................................................................................114 Characterization Techniques....................................................................................115 Suggested Improvements...................................................................................116 Optical bench..............................................................................................116 Data analysis..............................................................................................117 Additional characterization studies............................................................117 The Dispersion Index................................................................................................118 Correlating the Dispersion Inde x to Material Properties..........................................119 Suggested Future Work............................................................................................119 APPENDIX A CALCULATION OF THE PARTICLE SIZE DISTRIBUTI ON BY SOLVING THE INVERSE MIE PROBLEM............................................................................121 B PROCEDURES FOR SCATTERING DATA ANALYSIS USING IRENA..........123 Desmearing of USAXS Data....................................................................................123 Combine USAXS and SAXS Data...........................................................................123 Calculate Particle Size Dist ribution and FWHM Values.........................................124 Correlation Length for In terparticle Distance...........................................................124 Calculation of Primary Part icle Distribution by SAXS............................................124 C DERIVATION OF THE CALCULATI ON OF INTERPARTICLE DISTANCE FOR THE IDEAL DISPERSION CASE.................................................................126 LIST OF REFERENCES.................................................................................................133 BIOGRAPHICAL SKETCH...........................................................................................138

PAGE 9

ix LIST OF TABLES Table page 2-1 Several cases of mass-fract al power law scattering................................................22 3-1 Material properties of the epoxy resin......................................................................33 3-2 Optical properties of the cured epoxy polymer........................................................34 3-3 Summary of processing techniques used to fabricate each nanocomposite sample with varied levels of particle dispersion...................................................................36 3-4 Summary of components used to fabricate nanocomposite samples for undispersed ultrasonicated and surface treated samples.......................................39 3-5 Summary of components used to fabricate the fractionated sample........................42 3-7 Stokes Law particle settling calculations based on ten minutes centrifuge time.....49 3-8 Optical components of the light scatte ring instrument, the manufacturer and part number......................................................................................................................50 5-1 Surface energy values for the nanoco mposite components and the calculated spreading coefficient for untre ated and treated particles..........................................77 5-2 Reference for the length scale probed and the range of q values for each technique..................................................................................................................95 5-3 Values of T FWHM, and d and the results of the calculation of D1 and D2..........107 5-4 Dispersion Index values compared to the processing methodology where M = mixing, U = ultrasonication, ST = surface treatment, and S = solvent..................108

PAGE 10

x LIST OF FIGURES Figure page 2-1 Diagram of the hierarchical structures of particles that can be probed with light and x-ray scattering measurements at different length scales, or q values...............19 2-2 Transmission electron micrographs of PCMWNT composites: (a) 5 wt.% MWNT, (b) 5 wt.% MWNT, (c) 2 wt.% MWNT (d) 1 wt.% MWNT.....................27 3-1 Scanning electron micrograph of the Degussa P25 TiO2 powder............................30 3-2 Scanning electron micrograph of dispersed TiO2 primary particles and small aggregates imaged on carbon tape (background).....................................................31 3-3 Small angle x-ray scattering (SAXS) derived particle size distribution of Degussa P25 titania primary particles......................................................................31 3-4 Log-normal particle size distribution of the Degussa P25 titania particles dispersed in alcohol media, meas ured by dynamic light scattering.........................32 3-5 Agglomerated log-normal particle size distribution measured by the dry powder module of the Coulter LS 13320 laser diffraction instrument.................................32 3-6 Structure of DGEBA and the repeat unit. The reactive epoxide groups are highlighted in red.....................................................................................................33 3-7 Structure of Jeffamine D230 and the repeat unit......................................................33 3-8 Diagram showing how an amine functi onalized hardener a ttached the reactive epoxide groups and cross-links the epoxy ch ains to form a thermoset material......34 3-9 Schematic showing the two silane coup ling agents used in this study: APTES (A) and AEAPTMS (B)............................................................................................35 3-10 Dry titania nanopowder in the isopropanol and acetone solven t mixture, showing good wetting of the particle surface.........................................................................43 3-11 Particle size distributi on measured after ultrasonica tion of the dry nanopowder in a solvent mixture..................................................................................................43 3-12 Digital image of the dispersed pa rticles in the solvent mixture after sedimentation and removal of the hard agglomerates..............................................43

PAGE 11

xi3-13 Particle size distributi on measured after sedimentation and removal of hard agglomerates in the solvent mixture.........................................................................44 3-14 Particle size distribution measured after aging of the surface treated nanoparticles in the solvent mixture to verify that the particles remained dispersed...................................................................................................................44 3-15 Beers Law relation for optical dens ity of titania nanoparticles dispersed in acetone and isopropanol, meas ured at 600nm wavelength......................................47 3-16 Digital pictures of (A) 7g/L Ti O2 nanoparticles in pure acetone where sedimentation has occurred due poor wettability and (B) 7g/L TiO2 nanoparticles in 60% acetone/40% isopropa nol with improved particle wetting....47 3-17 Schematic representation (A) and digita l picture (B) of the optical bench set-up for light scattering measurements............................................................................51 3-18 Intensity of the incident beam over 8.3 minutes with measurements taken every 5 seconds. The error is assume d to be negligible (~1%).........................................52 3-19 Scattering intensities fo r 600 nm (A), 800 nm (B), 2 m (C)and 5 m (D) diameter polystyrene partic les dispersed in water....................................................54 3-20 Example of a theoretical scattering cu rve generated by the MiePlot software for 200 nm (radius) TiO2 particles in epoxy with a 10% polydispersity and lognormal size distribution............................................................................................58 3-21 Custom-made sample holders for measur ing the optical density of solid polymer films in the UV-Vis spectrometer............................................................................62 3-22 Polyethylene embedding molds used fo r TEM polymer sample preparation..........64 4-1 Forward scattering of a sphere showi ng the electromagnetic field (wave) vectors for the incident, particle, and scattered radiation.....................................................67 4-2. Schematic depicting the scattering of electromagnetic radi ation by a spherical particle. The incident field is in the z-axis direction...............................................68 4-3 Diagram of light incident fr om the left with wave vector ki scattering from an element at r toward the detector with scattering wave vector ks, at a scattering angle ......................................................................................................................73 5-1 Examples of digital images comparing thick films of each dispersion state against a typewritten background.............................................................................79 5-2 Scanning electron micrographs of the undispersed sample showing the large particle agglomerates and voids due to poor particle wetting by the epoxy............80

PAGE 12

xii5-3 Scanning electron micrographs of the fractionated sample showing the small particle agglomerates and indi vidually dispersed particles......................................81 5-4 Transmission electron micros copy images of 0.5 wt% TiO2 embedded in epoxy: undispersed (A), ultrasonicated (B) and surface treated (C)..................................83 5-5 High resolution TEM micrographs of the fractionated sample (0.5 wt%) showing the nano scale dispersion an d separated individual particles.....................84 5-6 Schematic of the four simulated part icle agglomerate morp hologies used to determine the validity of the D1 dispersion index....................................................87 5-7 Graphical representation of the four si mulated cases of particle or agglomerate size distribution........................................................................................................87 5-8 Surface chemistry parameter D1 values of four simulated size distributions...........88 5-9 Schematic of the four simulated part icle agglomerate morp hologies used to determine the validity of the D2 mixing parameter..................................................91 5-10 D2 mixing parameter values of four simulated particle morphologies.....................92 5-11 Figure representing a typical scattering intensity plot and the various length scales probed by different scattering techniques......................................................96 5-12 Small angle x-ray scattering intensit y plot for three differently dispersed samples.....................................................................................................................97 5-13 Ultra small angle scattering intensity pl ot showing the four differently dispersed samples. This q range covers the leng th scale of approximately 100 nm to 2000 nm.............................................................................................................................98 5-14 The region 0.001 < q < 0.01, revealing the changes in the USAXS scattering curve due to changes in agglomerate size................................................................99 5-15 Scattering intensity plot including the static light s cattering data for the four differently dispersed samples.................................................................................100 5-16 Particle (primary, aggregate, or agglom erate) size distribu tion as obtained from data analysis of x-ray scattering data (except dry powder obtained from SLS).....102 5-17 Average particle (primary, aggregate, or agglomerate) size shown for the data obtained by USAXS measurements (except primary particle size obtained from SAXS)....................................................................................................................103 5-18 Particle size distribution as obtained from data analysis of static light scattering data, except primary particles (SAXS) and dry powder (DLS).............................104

PAGE 13

xiii5-19 A comparison between the average part icle (aggregate or agglomerate) sizes obtained from x-ray measurements (U SAXS) and static light scattering measurements (SLS)..............................................................................................105 5-20 Optical density (amount of light transmitted through the sample) at 700 nm wavelength visible light for each differe ntly dispersed sample (0.5 wt% TiO2 in epoxy) as compared to a pure epoxy background..................................................111 5-21 A strong linear correlation can be made between the D1 dispersion index component and optical density values (measured at 700 nm)................................112 6-1 Initial attempts at using the FIB t echnique to three dimensionally image the nanocomposite samples created in this study.........................................................115 C-1 A typical positive-skewed lognormal di stribution curve is shown (A). The experimental particle si ze distribution of TiO2 particles disbursed in ethanol and measured with dynamic light sc attering is also shown (B)....................................126 C-2 A two-dimensional schematic drawing of the randomly well-dispersed particle configuration..........................................................................................................132

PAGE 14

xiv Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INVESTIGATION AND CHARACTERIZ ATION OF THE DISPERSION OF NANOPARTICLES IN A POLYMER MATRIX BY SCATTERING TECHNIQUES By Marie Christine Kissinger-Kane May 2007 Chair: Brij M. Moudgil Cochair: Anthony B. Brennan Major: Materials Science and Engineering There are many material benefits to a dding nano-sized fille r particles to a polymeric resin, such as improvements in st rength, modulus and optical transparency. However, the relative amount of material im provement is highly dependent on the degree of particle dispersion within the nanocomposite Characterization of the extent of particle dispersion poses a great challenge because of the length scales of interest, from nano to macro. In this study, a commercial dry titania nanopowder with a mean primary particle di ameter of approximately 30 nm was mixed into an epoxy resin/hardener system to produce nanocomposite samples. Processing techniques such as ultrasonication and partic le surface modification were used to produce nanocomposites with varying degrees of disper sion. The main objective of this study was to characterize the state of dispersion of the nanoparticles embedded in the epoxy by static li ght scattering, SAXS, and USAXS techniques. From these measur ements, a quantitative dispersion index was derived based on the change in particle si ze distribution and thr ee dimensional relative arrangement of the nanoparticles as compared to an ideally dispersed system. The state of dispersion was also verified qualitatively by microscopy.

PAGE 15

1 CHAPTER 1 INTRODUCTION There are many emerging applications where it is necessary to disperse nanoparticles into a polyme r matrix, such as gas sensing films, adhesives, pharmaceuticals, aerospace materials, paints, au tomotive parts, coatings and explosives. These dispersed particle systems are mainly us ed to enhance or improve the properties of the surrounding matrix material, such as c onductivity, mechanical and impact strength, energetics, wear resistance, and optical prope rties, to name a few. One of the main reasons for using nanoparticles is their larg e surface to volume ratio. The large surface area increases the number of particle-polymer interactions, thus strongly influencing the overall material properties.1 The growing effort in the area of nanocomposite research stems from the promising material propert ies that can be realized by efficient incorporation of nano-sized filler particles. Wetzel et al state2 that the unique polymer nanoc omposite properties can only be achieved if the nanoparticles are well dispersed in the matrix. The most optimal material property benefits can only be realized if an ideal dispersion is achieved. An ideal dispersion is defined in this study as a threedimensional array of non-touching primary particles in which all particles are homoge neously distributed in the matrix and not confined to any geometrical lattice. Conversely, partic le agglomeration decreases composite material performance by the inclusion of voids that act as preferential sites for crack initiation and material failure.1, 3 Ganesan, et al.4 have concluded that poor dispersion would lead to redu ced product life or poor perf ormance during service, poor

PAGE 16

2 product appearance, poor proces sing and product uniformity, wast e of raw materials, and excessive energy usage. Particle agglomerates typically exist in two different forms. The first form is often produced during particle synthesis and ma nufacture, where the primary particles produced during decomposition and subsequent calcination sinter to gether to form a hard agglomerate. In this study, these par tially sintered primary particles are termed an aggregate. Aggregates require attrition to break the strong covalent in terparticle bonds. The second agglomerate form is called a soft agglomerate, or simply an agglomerate A particle agglomerate is held together by weaker bonds such as attractive interparticle van der Waals forces, capillary forces, electrostatic forces, etc.5 However, van der Waals forces are much more significant than the long-ra nge electrostatic forces in the nano size regime.6 The strength of particle adhesion strong ly depends on particle size, as well as the surface roughness, surface energy, and al so the effective elastic constant.6 Generally stated, as particle size decreases, the adhesion forces present between particles in order to form agglomerates increases. The most critical endeavor to unders tanding nanoparticle aggregation and agglomeration is characterizati on of not only the change in the size of the agglomerates (or aggregates)due to adhesion forces, but also the distance between agglomerates and spatial distribution, reflecting concentration gradients and mixing fluctuations .7 Krishnamoorti and Vaia8 state that the final propert ies of a polymer nanocomposite system will depend as much on the individual properties of the constituents as on the relative arrangements and subsequent synergy between them. Thus, characterization of these changes requires information not only on the nanoscale, but also but also on the

PAGE 17

3 microscale.9 Despite the importance of such info rmation, development of techniques to quantify the extent of partic le dispersion has delayed the investigation of physical and chemical processing techniques to enhance dispersion further th an the results of conventional procedures. Unambiguous dispersion characterization is difficult because it also depends on the complicated interactions between the particle s and the matrix, but also the experimental processing conditions.10 A variety of techniques have been proposed to determine the degree of particle dispersion of a polymer nanocomposite and can be categorized into four general areas7: (1) real space observa tion including optical and electron microscopy; (2) determination of material property res ponse due to changes in dispersion, such as rheology, permeability and impact resistance, wh ich provides an indirect verification of conclusions drawn based on previous models; (3) measurement of a quantitative area of interaction between the particle and matrix, such as XPS, NMR, or other spectroscopic techniques; and (4) reciprocal space obser vation based on scattering of individual particles and agglomerates (or aggregates). Overall, it is clear that no individual approach is sufficient to fully characterize nanoparticle disp ersion on all relevant length scales. The ideal technique for the evaluation of nanoparticle dispersion will probe the interparticle distance and the relative arrangement of the particles while also accounting for the primary and agglomerate particle size s. To date, no one has utilized a range of length scale measurements, incorporating a ll aspects of three di mensional particle homogeneity characterization, to provide a quantitative evalua tion of nanoparticle dispersion. The main objective of this rese arch is to develop a standardized methodology

PAGE 18

4 for the characterization and evaluation of the degree of nanoparticle dispersion in a polymer matrix. Accordingly, a full range of length scales of the nanocomposite material will be investigated. Attempts will be made to develop a quantitative measure of dispersion that includes both a mixing paramete r to determine the effectiveness of the blending process (microscale) and a surface chem istry parameter that reflects the degree of compatibility of the particles by the su rrounding polymer (nano scale). Additionally, chemical and physical processing techniques w ill be investigated to systematically enhance the dispersion of inorganic nanopa rticles in an organic polymer matrix.

PAGE 19

5 CHAPTER 2 LITERATURE REVIEW In the following section, a critical revi ew of the characterization techniques selected for this study: light scattering, small angle x-ra y scattering, and microscopy is presented. Also, previously developed quant itative dispersion indices are reviewed, along with their calculation a nd practical limitations. Finally, the ASTM standard for assessing the state of dispersi on of carbon black in rubber is presented as a case study for what is used in the au tomobile tire industry. Particle Dispersion Studies The dispersion and stabilization of part icles suspended in a medium has been intensely studied for many years. The major ity of the work has focused on the dispersion of particles in aqueous media by electrost atic or steric stabilization mechanisms.11 Electrostatic stabilization can be achieve d by creating a high charge density on the particle surface, which results in a str ong electrical double-layer repulsion between particles. Steric stabilization involves the attachment or adsorption of long chain molecules (polymers, silanes, surfactants, etc), where the molecules emanating from the particle surface create a physical barrier to agglomeration. The adsorbed layer opposes the approach of neighboring particle s both entropically and enthalpically11. Widegren and Bergstrom12 have found that the c ontrol of the dispersion in nano sized powders

PAGE 20

6 becomes even more essential because even sma ll perturbations in the interparticle forces have drastic effects on the disp ersion behavior. Due to the la rge available surface area of nanoparticles, the key to effective dispersi on lies in control of the chemistry at the particle surface. Particle Surface Modification Shutzberg et al13 suggest that a typical problem lead ing to particle agglomeration is incompatibility of the nanopa rticles with the organic ma trix and suggests organic modification of the particle surface. One way to decrease the effect of attractive interparticle forces and in crease compatibility of the particles with the surrounding polymer matrix is by grafting molecules that have an affinity for the surrounding polymer (eg, similar miscibility parameters) onto the par ticle surface. Long chain molecules, such as silane coupling agents, are often used in composite systems to not only form a steric barrier around the particles to prevent reagglomeration by vo lume exclusion, but also to react with the polymer matrix to improve interfacial adhesion.14 However, Fischer notes that in a polymeric nanocomposite, even when a surface modification with favorable miscibility with the surrounding matrix is utili zed, complete dispersion can be difficult to achieve due to confinement of the particle agglomerates by the polymer chains.15 Brinker and Scherer11 generalize that an effectiv e barrier around particles to enhance and stabilize the dispersion must meet certain guidelines. Fi rst, the surface of the particle must be completely covered w ith the adsorbed molecules. Secondly, the adsorbed molecule must be firmly anchored to the surface so that it cannot be desorbed by Brownian motion. Thirdly, the layer must be thick enough so that the minimum distance between approaching pa rticles still remains well above that necessary for van der Waals attractive forces to dominate. Finall y, the end of the molecule that is not bonded

PAGE 21

7 to the particle surface must be miscible with the surrounding solvent, or covalently bonded to the surrounding medium11. Effective engineering of the particle surface aids in the stabilization of particle s once interparticle separation has occurred, but cannot break up existing aggregates. It has been noted in the literature that disp ersion of nanoparticles in a polymer matrix only by surface modi fication provides limited success without utilizing other techniques.15 The addition of the appropr iate surface chemistry is one technique used to enhance particle disper sion, while mechanical mixing has also been used in the literature to overcome the attractive van der Waals forces. Mechanical Dispersion Most recent studies have determined that mechanical dispersion methods alone are not sufficient to disperse na noparticles to primary size in a viscous medium such as a polymer. However, many have utilized mechanical energy such as high speed homogenizers, mixers, and sonication techniques The objective of using such techniques is to create eddy currents in the dispersing medium that provide dispersing energy or shear forces on the appropriate length scale of the particle agglomerate to overcome the interparticle van der Waals forces.16 Conversely, Yang and Dan17 have stated that physical mixing alone may lead to increased se paration into the disc rete phases, rather than promote particle dispersion. Ultras onication or bath sonication are common procedures used to physically break up aggl omerates and overcome the van der Waals attractive forces. Several groups have used sonicati on methods to break up agglomerated nanoparticles for incorporation into polymer matrices.18-19 However, it has been shown that even high-power ultrason ication does not always break up all agglomerates found in a suspension of nanoparticles.20 A number of researchers ha ve combined techniques to

PAGE 22

8 provide the optimal level of particle disp ersion. For example, Zhao and coworkers21 used sonication and particle surf ace treatment combined for more effective dispersion of carbon nanofibers in water. Lee, et al.22 also employed both physical and chemical dispersion techniques for Fe2O3 nanoparticles. Many studies have been devoted to finding an effective technique for optimal pa rticle dispersion in a polymer matrix. However, none are reported that have employe d a quantitative measurement to define the level of dispersion that has been attained by using those techniques. According to the work of Kolmogoroff, th e model of isotropic turbulence can be employed for modeling the mixing process of part icles in a fluid. Utilizing this model, the mixing volume during processing is consider ed as a distribution of eddy currents of various sizes, created by the ener gy input of the mixing technique (e.g., ultrasonication).23, 24 The larger eddies are slowly transformed into smaller ones at a constant rate of energy dissipation until th ey are reduced to the smallest size, 0 which is called the inner turbulence sc ale, or eddy length scale.23, 24 The Kolmogoroff equation can be written as 0 = ( 3/ tot) (1) where is the density of the mixt ure of particles in a liquid, is the kinematic viscosity of the mixture, and tot is the total energy input into the mixture with respect to the mixture volume. A simple calculation re veals that for a low concentration of nanoparticles (e.g., 0.5 wt%) disper sed in a low viscosity flui d (e.g., isopropanol), placed in an ultrasonic processor at 250 W, the le ngth scale of the eddy cu rrents is still on the millimeter scale.

PAGE 23

9 Dispersion Equations Balek25 first defined an agglomeration parameter, A in 1970 in order to evaluate the degree of particle dispersion as follows: d D A (2) where D is the median particle aggregate di ameter determined by sedimentation and d is the mean primary particle diameter measured by the BET (Brunauer, Emmet and Teller) method. Besides the fact that the particles ar e measured both in the liquid and dry states, there are several limitations to this approach. First, the sedimentation technique assumes that the particles or aggregates are spherical in shape, and also that an aggregate has a constant density, which is not the case fo r a sometimes porous aggregate. The BET technique assumes that the primary particle si ze distribution is monosi ze and that particle size can be extracted from a measure of surface area alone. The agglomeration parameter, A, was the first attempt to evaluate the degree of particle agglomeration, but was not developed any further. Liu, et al.26-27 proposed an aggregation equation based on the relative packing of nanoparticles and the correlati on of a dispersed particle sy stem to that of particles confined to a geometric regular lattice. They calculated the difference in particle spatial distribution for eight regular lat tices and three real particle distributions. They derived a generalized equation for relating the averag e matrix ligament thickness (surface to surface interparticle distance), T, to the average particle size (d), size distribution (), volume fraction () and spatial distribution factor () for a system containing spherical particles obeying the log-normal distribution:

PAGE 24

10 2 2 3 1ln 5 0 exp ln 5 1 exp 6 d T (3) It has been shown through simulation and experiment that previous models proposed by Liu and coworkers are appropria te to describe a we ll-dispersed binary composite system, but cannot desc ribe a poorly dispersed system.26 The use of previous equations includes the assumption that the pa rticles form some kind of ordered spatial lattice, which is seldom the case for a real composite material. Equation (3), however, includes a spatial distribution parameter, to account for such de viations from geometric lattice spacing of the particles in the matrix. Premphet and coworkers28 use equation (3) to characterize the inte rparticle distance of rubber partic les in a polypropylene matrix, but found that the values obtained by experiment for a non-ideal system did not match well with the theoretical model. Other attempts to describe the spatial di stribution of randomly dispersed particles have been made. Gurland29 suggested an approach to quantifying dispersion with a standard deviation ( n) of the particle spatial distri bution in small composite areas, N i i nn n N1 21 (4) where N is the number of samples, ni is the number of observed particles per unit volume in the ith sample and n is the arithmetic mean of particles in N samples. The major limitation to this approach is that the equati on represents a simple variance in the number of particles observed in a number of samp les. It does not address the particle agglomerate size or size distribution, or the change in spatial distribution over a large

PAGE 25

11 sample volume. Tanaka, et al.30-32 used the parameter, P as the spatial distribution factor to quantify dispersion. The value of P approaches one with increasing particle dispersion, M p p PM i i i1 ln ln1 (5) M i i i ib b p1 (6) where M is the total number of grids (divisions ) on a transmission electron microscopy (TEM) micrograph, and bi is the particle area of the ith grid. This method involves tedious counting of particles on a TEM mi crograph and again does not address the importance of particle agglomerate size or changes in spatial di stribution on the macro scale. The dispersion index developed by Suetsugo33 (1a), where a is the area fraction of agglomerates, is defined by: N i i i aD n A1 24 (7) where A is the observed area, is the volume fraction of particles, Di is the agglomerate diameter and ni is the number of agglomerates in the sample. The dispersion index decreases to zero for maximum agglomeration and increases to one for maximum particle dispersion. Although this parameter begins to address the importance of particle size and arrangement within the sample volume as a measure of dispersion, the measurement technique still utilizes TEM micrographs that only examine a very small fraction of the

PAGE 26

12 overall bulk sample. The equation also does not address the measurement of particle and agglomerate spatial distribution. Although some equations seem more valid than others, all of the proposed dispersion indices employ measurements made by some microscopy technique, such as SEM or TEM. This approach can be misleadi ng as to the actual particle arrangement in space27,34-35. It may be that the characterization of particle disper sion spans many length scales, which cannot be fully characteri zed by a small sample area by microscopy methods alone or just one cal culated dispersion parameter. Correlation of Dispersion with Material Properties Mechanical Properties It has been well documented that the degree of particle homogeneity has a great affect on the resultant mechanical propert ies of the system. When nanoparticles agglomerate into large clusters, the reinforcing effect is lost.36 Particle agglomerates degrade the composite performance through th e inclusion of voids that can act as preferential sites for crack initiation and failure1 For example, the findings of Thompson and coworkers37suggest that the decrease in te nsile properties from the neat polyimide polymer after addition of metal oxide nanoparticles is indicative of poor particle dispersion on the nanometer scale. Most studies in literat ure only address the change in material properties with varied pr ocessing and little progress has been made to control or quantify the state of particle dispersion. Complications arise when a simple extrapolation of the design paradigms of conventional composites cannot be used to predict the behavior of nanocomposites.38

PAGE 27

13 Optical Density A nanocomposite material in which all of the nanoparticles are dispersed to primary particle size displays optical propertie s in the visible range close to that of the pure matrix material with no filler particles due to low levels of scattering by nanoparticles. However, when particles agglom erate to sizes that are comparable to the wavelength of visible light, hi gh levels of light scattering occur and the optical density decreases.39 Thus, the change in particle agglomer ate size as it relates to dispersion has been related to the degree of particle homogeneity in a polymer nanocomposite. For example, Shang, et al.40 correlated the high transparency of their polyimide composites to the compatibility of the SiO2 particles to the polymer matrix, thus providing good dispersion. Vandervorst and coworkers41 qualitatively related the visual appearance of several Latex-SWNT (single wall nanotube) com posites of varied levels of dispersion. The worst dispersions (as view ed by microscopy) appeared to have the lowest level of transparency. Molecular Modeling of Nanocomposite Dispersion In order to gain a better understanding of polymer-parti cle interactions and the mechanism of nanoparticle agglomeration in polymer matrices, molecular modeling and simulations have been attempted to furt her comprehend nanocomposite dispersion. These molecular dynamics simulations utili ze the correlation between polymer matrix effects with the overall particle dispersion. Smith and coworkers42 have found that nanoparticle agglomeration increases with polymer molecular weight for composite samples containing weak polymer-nanoparticle interactions. This finding becomes significant specifically for fillers that are no t covalently bonded to the polymer and have a high molecular weight matrix material. It has also been observed that when the

PAGE 28

14 polymer-particle interactions ar e strong, as seen with chem ical bonds created by particle surface modification (eg, silane coupling agen ts), the molecular weight effect becomes insignificant. Starr, et al.43 take a different approach to mo lecular modeling by investigating the thermodynamics of particle clustering and the mechanisms involved in agglomeration. However, they arrived at the same conclusions as Smith and coworkers.42 By observing changes in potential energy of interparticle in teractions with respect to volume fraction, they determined a boundary between a dispersed and non-dispersed system. It has been shown through the simulation of simple liquids, which are far from the case of a viscous polymer melt or resin, that the potential energy is strongly sensitive to local packing of the pa rticles. The Starr group, as well as others42, 44-45, also utilized the structure factor, S ( q ) and the pair correlation function, g ( r ), for an explicit measure of long and short range ordering of the particles as it relates to particle dispersion. Brechet45 also describes the sensitivity of the n earest neighbor distri bution to particle agglomeration. However, this approach ha s little practical application outside of modeled systems due to mathematical complexities. Although these simulations come closer to an explicit value of dispersion than empirical work the practical uses of these results are limiting due to the extremely simp lified modeled composite systems. Hooper, et al.46 reviewed the progress of modeling t echniques for polymer nanocomposites and have found that most work still focuses on d ilute polymer systems and the one or twoparticle problem. However, from understand ing the models for dispersion measurements derived previously, it is clear that characterization methods that probe the real composite

PAGE 29

15 material on a wide range of length scales ar e better suited to determine the degree of homogeneity of the system. Light Scattering Background Light scattering, also known as laser diffr action, has been used for many years to characterize the size, shape, and orientation of particles. Gene rally, light scattering is the most common method of particle sizing.47 The study of the scattering of light from matter was most advanced by Gustav Mie in 1908, when he published his groundbreaking article that provided an exact solu tion for the scattering of radiation from a dielectric sphere in a medium.48 It is suggested that this work stems from the study of colloidal gold suspensions by his student.49 This study lead to th e term Mie scattering, which is used to describe all scattering of radiation by spherical and non-spherical particles in a certain size range. It is impor tant to note that the analysis presented in Mies paper is restricted to particle sizes up to 0.18 m, possibly suggesting the connection between the colloidal gold work and Mies paper. After 1908, Gustav Mie published several more papers, but nothi ng more on light scattering of particles.49 Light scattering occurs due to the gradie nt of refractive inde x (electron density) between the particle and the embedding medium (polymer). The refractive index may be a function of incident radiation, temperature, or pressure. If the ma terial is nonabsorbent and nonmagnetic, then the square of the refractiv e index is equal to th e dielectric constant at that wavelength. For absorbing ma terials, the complex refractive index, m = n (1 ik ), is related to the absorptive index, k where the real term desc ribes the refraction and the imaginary terms describes the absorption.50 If the particles and polymer have the same refractive index, the particles would appear to vanish becau se there would theoretically

PAGE 30

16 be no scattering of light, which is called inde x matching. The greater the refractive index gradient, the greater the particle scattering. The scattering power ( I (q)), or scattering intensity per unit time and volume, is dependent on the ratio of the refractive indices of the particle and surrounding me dium. For this reason, high refractive index particles have been chosen for this study (see Chapter three) to achieve high scattering resolution. Any laser diffraction instrument uses th e angular dependent scattering pattern (scattered intensities) of particles in a disp ersing medium to calculate the particle size distribution. A commer cial instrument usually includes a proprietary software package that solves the inverse Mie problem to mathema tically invert the scattering matrix data in order to calculate particle size distribution (see Appendix B). Generally, large particles scatter strongly and principally to smaller angl es while smaller particles scatter weakly at much larger angles.47 The relative terms small and large depend solely on the wavelength of the incident radiation. Light scattering has been cited by many rese archers as a meaningful technique to characterize the degree of disp ersion of particles embedded in a matrix material. McNeil and coworkers51 state that light scattering can be strongly influenced by inhomogeneities in the random microstructure, such as pa rticle aggregation and agglomeration. Specifically, Schnaiter and Wurm52 reported that light scatte ring is a powerful technique to characterize dust aggregates. Schaefer and coworkers19 also utilized light scattering to characterize silica dispersion in a rubber matrix. Wiggins, et al.53 used optical light scattering in the Rayleigh regime (particles sm all compared to the wavelength of the light source) to calculate a bunching parameter, i exp, which is an indication of the density change of particles in the scatte ring volume over time with respect to the

PAGE 31

17 scattering angle, This type of measurement may be correlated with dispersion homogeneity, however, there are several lim itations of this technique. The Rayleigh scattering theory assumes a monosized di stribution of particles and no multiple scattering. The measurement does not account for particle settling due to agglomeration and is not suited for stationary particles embedded in a matrix. The dispersion state can be monitored by light scattering because a particle agglomerate made of many primary particles will scatter light as a single, larger particle. Interparticle interactions also play a role in the scattering pattern when scattered radiation overlaps due to the small interparticle distance51, 54-55. Although light scattering has been used for many years to measure changes partic le size and size distribution, the efforts to relate that information to a quantitative measure of dispersion have provided little practical relevance. Particle Size Measurement In the literature, very few groups have used static light scattering to measure particle size in a solid polymer matrix. Gao and Wu56 used a commercial particle sizing instrument by modifying the small volume sa mple holder of the Coulter LS 230 laser diffraction instrument. Others have chosen build their own static light scattering instrument to measure particle size52, 57-65. It was shown in all studies that custom-built optical equipment can be flexible enough to measure the scattering characteristics in any desired configuration, but al so precise enough to deliver meaningful measurements through careful instrument calibration and measurement. Although none of the authors mentioned in the light scattering section attempt to calculate a quantitative dispersion index from light scattering data, almost all mention the ability to characterize a change in the particle conformation, or the fluctuation in mixture

PAGE 32

18 homogeneity. Many also combine the results from laser light scattering to that of small angle x-ray scattering to characteri ze a wider range of length scales6667. Small Angle and Ultra Small Angle X-ray Scattering The scattering of x-rays from a radiation source incident on an array of scatterers (particles) occurs due to the difference in el ectron densities between the particles and the surrounding medium, which is defined as the number of electrons per unit volume.68 The greater the difference in electron density, th e greater the scattering of the incident radiation (scattering power). Small angle and ultra small angle x-ray scattering are governed by the same fundamental laws of electromagnetic interactions with matter as with light and neutron scattering.69 The difference between light scattering and x-ray scattering is the wavelength of the radiati on used. A typical x-ray source consists of a 1.54 wavelength Cu K target for a conventional system. The scattering pattern of xray radiation from an array of particles can provide meaningful information on the relative arrangement, particle size, and microstr ucture of the particles that are probed. As with any scattering measurement, the scatte red intensity contains contributions from intraand interparticle spacing and arrangement.70 The effects of interparticle scattering can be reduced by using a low concentration or dilute solution of scatterers. Small angle x-ray scattering is a widely us ed technique to determine size, shape, and internal structure of colloidal particles in the 1-100 nm range.70 In order to analyze data from any scattering techniqu e, the scattering intensity ( I ( q )) with respect to the wave vector, q can be divided into three distinct regions corresponding to specific hierarchical length scales probed in a sample. These re gions are represented by knees in the log-log plot of I ( q ) vs q At high q values, primary particle size and size distribution can be obtained (Figure 2-1), regardless of the agglomeration of the particles in the matrix, due

PAGE 33

19 to the small probe size of the technique. In the midq range, particle agglomerates or aggregates in the micron size range can be probed. The lowest q values probe the largest of the particle agglomerates.71 Koga and coworkers72 used the different scattering regimes from USAXS and SAXS data to charac terize the hierarchical structures formed by carbon black filler particles in a rubber matrix (Figure 2-1). Others have also used this technique for morphology characteri zation of embedded nanoparticles73. The probing of various length scales in the composite sample is believed to be the best approach to dispersion characterization by addressing agglomeration from the nano to micron size range. primary particleaggregateagglomerate primary particleaggregateagglomerate Figure 2-1. Diagram of the hierarchical stru ctures of particles th at can be probed with light and x-ray scattering measuremen ts at different length scales, or q values. X ray Data Analysis Historically, x-ray data analysis has been carried out by using approximations of particle size based on the shape of th e curve in different regions of the I ( q ) vs q plot. These classical techniques are presented with the recognition that a more modern approach to data analysis is available throug h the use of modeling so ftware. For the most ideal case of a monodisperse, homogeneous and isotropic dispersion of spherical particles in a medium, the general form of the scat tering intensity equation can be written as74 ) ( ) ( ) ( q S q NP q I (8) where N is the number of particles in the array, P ( q ) is the form factor that describes intraparticle scattering and S ( q ) is the structure factor describing interparticle scattering.

PAGE 34

20 The Guinier regime In the Guinier regime of q values, the scattered intensity is used to determine the radius of gyration (radius of an equivalent sphere), Rg, of the particle agglomerates.73 The scattered intensities in this region can be ge nerated from either light scattering, USAXS or ultra small neutron scattering (USANS) data. These techniques probe very small q values that correspond to large sizes in real space, dependent on the wavelength ( ) of the probe radiation, where 2 sin 4 q The following Guinier law holds for the region of approximately I(q)/I(0) < 4/ 3 according to Guinier.75 3 exp 02 2q R I q Ig (9) The radius of gyration, Rg is defined as the radius of a sphere with the same mass and the same moment of inertia as the object under consideration76. It is related to the correlation length of the particle agglomerates by77: 2 / 12 / 1 f f gD D R (10) where Df is the fractal dimension as determined from the power law regime. Chaker and coworkers77 assert that the correlation length is the best estimate of the average interagglomerate distance. Glatter and Kratky68 report in their exte nded review of SAXS theory and practice that the Guinier formula holds surprisingly well in the majority of particle size and shape cases. However, th is technique does not a ddress the di stribution of particle sizes present in a composite sa mple, nor the spatial distribution of those particles.

PAGE 35

21 The power-law regime The boundary between the Guinier and pow er law regime was defined by Dobbins and Megaridis first in 199178, although many describe the bo undary as where the log-log plot of I ( q ) vs q displays a power-law decay. This value can be compared to the crossover point suggested by Sorenson.75 The power law regime describes the mass or surface fractal scaling at the particular st ructural level corresponding to the probed q value.79 The fractal dimension, Df, falls in the range 1 < Df < 3. A value of three corresponds to a sphere while a value of one corresponds to a rod where the mass is proportional to the length. Fractal morphol ogies are often found for particles made by aerosol techniques between the primar y particle size and aggregate size.66 The table below elucidates some of the values of the fractal dimension to be expected from certain types of aggregation mechanisms, although these values can vary. Chaker, et al.77 found a fractal dimension of 2. 45 for silica in a PPO network and proposed a particle-cluster gr owth mechanism, which implies that the primary particles were in a random walk configuration in the viscous matrix and stick to form a growing aggregate. Schaefer and coworkers19 also found a fractal dimens ion of 2.5 for nano-silica dispersed in a rubber matrix, which they de scribe as a bunch of grapes morphology. They also suggest that an optimal fractal di mension (high or low) can be tailored based on the model of reinforcement chosen in a given particle-polymer system. In light of these suggested results, there is still much de bate on the correlation of fractal dimension and the state of particle dispersion. The cha nge in material properties with respect to changes in aggregate and agglomerate morphology (shape) has not been established.

PAGE 36

22 Table 2-1. Several cases of ma ss-fractal power law scattering Mass fractal structures Power Law slope Multiparticle diffusion limited aggregate -1.8 Percolation cluster -2.5 Diffusion limited aggregate -2.5 A Unified Fit model Beacuage and coworkers have developed a unified approach to small angle scattering by combining both the Guinier and power-law regions into one equation that only requires the same parameters that are used in the fitting equations for the separate regimes.67 He terms this combined region as one structural level that more clearly defines the limits of both regimes to more appropriately calculate Rg and Df. It is also claimed that the unified approach provides an accurate model for the scattering behavior for mass fractals, surface fractals, and diffuse interfacial scattering.67 However, it still does not provide any information about the distribution of particle sizes or the spatial distribution of the particles. The Porod regime The boundary between the power law regime and the Porod regime is characterized by a change in slope, or kn ee in the scattering data. Se veral have used the relation q = 2 /a at the slope change as an exact location for the calculation of the radius of the primary particles, a Hasmy and coworkers80 suggest that the crossover is highly affected by the polydispersity of the primary particle s and must be accounted for, such that eff cq 6 1 1 2 (11) where qc is the crossover q value between the Porod and power law regime and eff is the effective polydispersity of the particles ( /a ), where a is the radius of the primary particles.

PAGE 37

23 The final slope, as termed by Porod in 1965, is expected to reflect the fine structure of the particle.68 The limiting behavior of scattered intensities at large q values, known as Porods law, predicts a decrease in the intensity proportional to the reciprocal fourth power of s : 4s K s Ip (12) where Kp is the power law constant and s = 2sin ( / 2)/ (differs from q by 2 ). The determination of Kp is of interest because it relates to certain structural parameters of the system.81 It has been generally assumed that a sl ope of -4 in the Porod regime indicates a sharp interface (smooth surf ace) between the particle and surrounding medium, but it has also been shown that a partially sharp bounda ry was sufficient to see the Porod-type behavior.82 The Porod evaluation of the final slope of the scattering intensity plot reveals the size of primary particles and the mor phology of the interface, however, it does not provide a distribution of primary particle sizes or the distance betw een those particles. Scattering Techniques for Dispersion Measurements Berriot, et al. in 200383 came very close to reporti ng a quantitative measure of the state of dispersion of nanosized fillers in an elastomer matrix. They used a ratio of the scattered intensity of a highly concentrated syst em to that of a very dilute system, which is directly related to the structure factor of the par ticles, which accounts for the interagglomerate scattering. However, no dire ct conclusions on a quantitative measure of particle dispersion could be drawn from th eir work. In 2005, Causin and coworkers84 also attempted to provide a quantitative dispersion measur ement, which consisted of a comparison of parameters between SAXS and TEM techniques. This particular study used clay nanocomposite samples, which incl uded the added complexity of high aspect

PAGE 38

24 ratio particles that stack together in tactoids with very small inter-layer distances. Good agreement was achieved between SAXS and TEM measurements with respect to size of the clay nanoparticle cluster size and inte r-layer spacing (ratio close to 1), although no statistical data was reported. Microscopy Transmission electron microscopy (TEM ) and scanning electron microscopy (SEM) are both very common methods of obt aining qualitative information about the dispersion of particles in a polymer matrix. Many of the studies mentioned here have used microscopy in some way to verify the results from various s cattering techniques, such as primary particle size or average aggreg ate or agglomerate size. 62, 85, 72, 76 However, some claim that there are many dr awbacks to using microscopy. Beaucage and coworkers85 assert that TEM, although widely app licable, is very tedious and subjective. Later, Hyeon-Lee et al.66 mention that TEM is quite useful in the measurement of primary particle size, but is very difficult to obtain a statistically meaningful size distribution. They also suggest that measurement of anything beyond the primary particle size, such as aggregates or agglomer ate structure is very complex due to overlap in three dimensions of partic le structures. Scattering tec hniques are known to provide more accurate measurements of size parame ters by measuring over a wide range of scattering vectors, and thus a wider range of length scale. ASTM Standard for Dispersion of Carbon Black in Rubber ASTM standard D2663-9586 describes the current standard test method used to determine the degree of dispersion of nanopa rticles in a polymer, specifically carbon black in a rubber matrix. The test consists of th ree different methodsvisual inspection, quantitative count of agglomerates, and microroughness measurement. The visual

PAGE 39

25 inspection test consists of l ooking at several cut rubber samp les with a hand lens or low power microscope and comparing the dispersion on a scale of one to five. Five has 99% dispersed carbon black with one having 70% di spersed particles. One can only imagine the amount of subjectivity involved in this sort of test, considering th e low resolution of a hand lens or even an optical microscope. The second method invol ves the inspection of microtomed sections of the rubber composite sample under a light microscope. The total number of agglomerates 5 m or larger are counted, a nd the total percentage of agglomerates below this critical agglomerat e size is calculated and expressed as the percentage of carbon black dispersed. The final test consists of a microroughness measurement using an instrument called a Di spersion Analyzer, which is available from the Federal Products Corporation. This instru ment consists of a specimen holder, sample cutter, and specimen tracking mount. The styl us of the instrument traces across the cut sample surface, which is mounted on a vibra tion isolator. The measurements are taken and a dispersion index is calculated based on th ree constants that have been previously determined for the specific material. This instrument can be described as a crude precursor to the present-day Atomic Force Mi crsocope (AFM). The limitations of this type of technique are that many artifacts can be generated by probing a cut polymer surface. Deformations and particle pullout can all contribute to artificial changes in the morphology of the surface, leading to unreliable results. The measurement standard also makes the claim that important physical properties are significantly influenced by the percentage of carbon black dispersed, such as abrasion resistance, tensile stre ngth and fatigue life. These pr operties are said to correlate with the agglomerate count and micror oughness results in the following manner:

PAGE 40

26 Dispersion, % Classification Above 99 Very High 97 to 99 High 95 to 97 Intermediate 92 to 95 Low Below 92 Very low This test method has been applied to othe r systems in the literature with little accuracy34, 87. In the study by Ptschke and coworkers34, all of the samples of multiwalled carbon nanotubes (MWNT) embedded in a polycarbonate (PC) matrix earned a dispersion index rating of 100% using the a gglomerate counting method described above. The PC-MWNT samples were observed under an optical microscope to assess the dispersion of filler particles that are on the order of 12-16nm in diameter. Upon close inspection of the SEM and TEM micrographs pres ented in this work (Figure 2-2), it is obvious that this method of di spersion quantification is not adequate. In 1995, Ganesan, et al.4 attempted to improve upon the ASTM standard to describe the dispersion of carbon black in rubber by use of an image processing technique. This method uses digitally pi xilated images with calculated grayscale tolerances to determine the texture of the mi crographs, taken at a magnification of 30X. Quantitative measures were calculated fr om these digitized images, including the statistical weighted mean and variance from th e ideal dispersed case. The weighted mean was plotted against a dispersion rating for a straight-line relati onship. There has been no further use of this particular image analysis technique in the lite rature, although others have tried, unsuccessfully, this sa me approach to quantify dispersion35. Park and Balatinecz35 proposed a mixing index value that de scribes the deviation from an ideal random mixture. This approach of compar ing the measured dispersion to an ideal random mixture provides a good star ting point to the techniques developed in this study.

PAGE 41

27 However, using low magnification microgra phs to investigate dispersion on the nano scale is physically impossible du e to the resolution issues. Figure 2-2. Transmission elec tron micrographs of PCMWN T composites: (a) 5 wt.% MWNT, (b) 5 wt.% MWNT, (c) 2 wt.% MWNT (d) 1 wt.% MWNT. Scope of Proposed Research The quantification of particle dispersion in a composite ma terial has been intensely studied in recent years, yet there are stil l many challenges in providing a meaningful measure of homogeneity that spans the re levant length scales. Molecular modeling cannot provide a theoretical solu tion due to the complexities of a dense polymer melt and interparticle interactions. Microscopy has proven useful to char acterize dispersion on a

PAGE 42

28 small length scale, but has fallen short of long range measurements on a three dimensional scale. Scattering techniques have been shown to provi de information about particle size distribution and a gglomeration of nanoparticles fo r the various hierarchies of length scale. It is clear from the literature review that one parameter to describe nanoparticle homogeneity is not sufficient a nd one processing tec hnique cannot provide optimal particle dispersion. The objective of this study is to investigate the effects of various processing technique s on the homogeneity of na noparticle dispersion in polymeric nanocomposites by using scattering techniques and to qua ntify the degree of dispersion with parameters reflecting ch anges at all relevant length scales.

PAGE 43

29 CHAPTER 3 MATERIALS AND METHODS Materials Nanoparticles The nanoparticle system chosen for this study consists of a multi-phase titanium dioxide commercial dry powder. The particles are made up of 80% anatase and 20% rutile crystal structures. The particles ar e manufactured by the Aerosil flame synthesis process at Degussa Corporation and are used primarily for photocatalytic applications, but may also be used as a heat stabilizer for silicone rubber (www.degussa.com ). The flame synthesis process by which the partic les are manufactured produces particle aggregates (primary particles held togeth er by covalent bonds) and primary particle agglomerates, which are primary particles and aggregates that are loosely bound by van der Waals forces (Figure 3-1). The primary particle size measured by scanning electron microscopy and small angle x-ray measurements is approximately 25 -30 nm in diameter (Figures 3-2 and 3-3). The size distribution of the P25 titania powder dispersed in alcohol media was also measured by dynamic light scattering on the UPA 150 (Microtrac) commercial particle sizing instru ment and is log-normal (Figure 3-4). The size distribution of agglomerates in the as -received dry powder was measured by laser diffraction on the dry powder module of the LS 13320 (Coulter) commercial particle sizing instrument (Figure 3-5). The average pa rticle agglomerate size in the dry state was approximately 10 m.

PAGE 44

30 The optical properties of the particles were estimated by using a generic mixing rule of the weighted anatase and rutile phases of TiO2 as follows particlesRI fraction Weight RI fraction Weight RI A A R R (1) where R represents the rutile phase, A represents the anatase phase and RI is the refractive index. Exact optical properties of materials at the wavelength of inte rest are difficult to obtain from the literatu re, however, values at the wavelengt h of interest were extrapolated form the work of McNeil and French.54 Figure 3-1. Scanning electron mi crograph of the Degussa P25 TiO2 powder. The powder was mounted on carbon tape in the dry st ate, coated with amorphous carbon to prevent charging and imaged in as-received form.

PAGE 45

31 Figure 3-2. Scanning electron micrograph of dispersed TiO2 primary particles and small aggregates imaged on carbon tape (b ackground). An example of a primary particle is circled. 0.E+00 5.E-06 1.E-05 2.E-05 2.E-05 3.E-05 3.E-05 1001000 Particle Diameter (Angstroms)Volume Distribution Figure 3-3. Small angle x-ray scattering (S AXS) derived particle size distribution of Degussa P25 titania primary particles. The mean primary particle size is shown to be approximately 25 nm in diameter.

PAGE 46

32 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 0.000.010.101.00Particle Diameter ( microns ) Differential Volume (%) Figure 3-4. Log-normal partic le size distribution of the De gussa P25 titania particles dispersed in alcohol media, meas ured by dynamic light scattering. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0.11 10 101001000Particle Diameter (microns) Differential Volume (%) Figure 3-5. Agglomerated log-normal par ticle size distribution measured by the dry powder module of the Coulter LS 13320 laser diffraction instrument. The average particle agglomerate diameter is 10 m. Polymer Matrix The matrix polymer for the composite mate rial used in this study is a two-part thermoset epoxy based on a viscous resin and ro om temperature hardener. The resin is manufactured as EPON 828, whic h is a Bisphenol A diglycid yl ether (DGEBA). This type of resin makes up approximately 75% of the epoxy market. The industrial

PAGE 47

33 applications of this resin include adhesives, coatings, electronic and structural materials, and composites (www.miller-stephenson.com/aero_005.htm ). Some properties of the epoxy resin are shown in Table 3-1. Table 3-1. Materi al properties of the epoxy resin. Viscosity at 25 C MW per epoxide EPON 828 110-150 P 185-192 The hardener for this resin is a po ly(oxypropylene)diamine, also known as Jeffamine D230 (Huntsman). This low mol ecular weight (~230 g/mol) cross linker was used specifically so that the particles have maximum mobility to disperse within the curing epoxy matrix. This amine terminated polyether cross links the epoxy resin by breaking the epoxyide rings and sharing of th e amine group to form a three-dimensional network (Figure 3-8). The hardener was a pplied in this study without the use of a catalyst. The optical properties information fo r this epoxy is also rather difficult to find in the literature, however Thiele and Fren ch have done extensive modeling of rutile titania particles embedded in a cured epoxy.55 Figure 3-6. Structure of DGE BA and the repeat unit. Th e reactive epoxide groups are highlighted in red. Figure 3-7. Structure of Jeffa mine D230 and the repeat unit.

PAGE 48

34 Figure 3-8. Diagram showing how an amin e functionalized hardener attached the reactive epoxide groups and cross-links th e epoxy chains to form a thermoset material. Table 3-2. Optical properties of the cured epoxy polymer. Wavelength Refractive Index Source Epoxy 560nm 1.514 Thiele and French (1998) Particle Surface TreatmentSilane Coupling Agent Two surface treatments were used to e nhance dispersion by making the particles more compatible with the su rrounding polymer. A silane coupling agent functions to bind both to the functional group of the polymer (in this case, the epoxide rings) and also the surface hydroxyl groups of the particle. In this way, the coupling agent provides a covalent bond between the particle and polym er, thus favoring particle dispersion over the dominant interparticle van der Waals at tractive forces that promote particle agglomeration. The silane coupling agents chosen for this study are aminoethyl aminopropyltriethoxysilane (Geles t) and aminopropyltriethoxysilane (Gelest), both used the amine functionality available to bond to the epoxy polymer chains (Figure 3-9) The triethoxy-type silane coupling agents have b een shown in the liter ature to effectively increase the compatibility of the metal oxide particle surf ace with the surrounding

PAGE 49

35 polymer,. 88 Specifically, Zhu and coworkers18 used the amine functional molecules attached to carbon nanotubes to form a c ovalent bond between th e particles and the surrounding epoxy polymer. Al so, Shutzberg, et al.13 used a series of silane coupling agents to increase the compatibility of BaTiO3 nanoparticles with an epoxy matrix. Also, Choi and coworkers utilized a silane coup ling agent to increase the adhesion between ferrite particles and the epoxy matrix.89 The coupling agent bonds to the titania su rface first by hydrolysis of the silane. Next, a condensation reaction proceeds to form hydroxyl groups on the silane, and finally hydrogen bonding takes place with the hydroxyl groups that are present on the particle surface to form a physical bond. Finally, wate r is eliminated and the siloxane bonds are left between the epoxy and particle surface. The kinetics of these reactions are highly dependent on the pH of the solution.11 The hydroxyl groups are pr esent on the surface of the particles due to atmospheri c conditions such as humidity. Figure 3-9. Schematic showing the two silane coupling agents used in this study: APTES (A) and AEAPTMS (B). The APTES is shown bound to the particle surface through bonding with the surface hydroxyl groups. Methods Nanocomposite Formulation The composite samples were prepared in order to obtain four distinct dispersion states, using 0.5 wt% TiO2 nanoparticles by weight. It wa s shown in previous chapters

PAGE 50

36 that 0.5 wt% nanoparticles was sufficient to observe improvements in the material properties of the composite material. The undispersed sample was prepared by the most basic processing techniques commonly used to produce a nanocomposite material, and is highly agglomerated. The fractionated sample, on the other hand, represents the highest level of nanoparticle dispersi on that was achieved in this st udy. The methodologies used to create these samples involved the mani pulation of surface chemistry and physical mixing. The development of the methodologies to produce the fractionated nanocomposite sample involved the selecti on of the appropriate solvent and surface modification for optimal particle dispersion. The processing techniques used to fabricate each sample are summarized in Table 3-3, and a detailed description is provided subsequently. Table 3-3. Summary of processing techniques used to fabricate each nanocomposite sample with varied levels of particle dispersion. Sample Mixing Ultrasonication Su rface Treatment Solvent Dispersion Undispersed Yes No No No Ultrasonicated Yes Yes No No Surface Treated Yes Yes Yes No Fractionated Yes Yes Yes Yes Four levels of particle dispersion The undispersed sample with the highest degree of agglomeration was prepared by simply mixing the epoxy resin (EPON 828), curing agent (Jeffamine D230), and the appropriate amount of titania nanoparticles (T able 3-4) in a pol ypropylene container. The three components were mixed at 60 rpm with a homogenizer for approximately five minutes and degassed for 45 minutes to one hour under reduced pressure to eliminate bubbles in the polymer-particle mixture that were formed during mixing. The silicone mold was then placed on a clean glass plate and pressed down to remove all entrapped air

PAGE 51

37 underneath the mold and to ensure planar samples. The epoxy-nanoparticle mixture was poured slowly into the individual 1 mm deep wells, and a microscope slide was used horizontally to level off the top of the wells to eliminate any epoxy overflow and ensure proper filling of each individual well. A ny extra epoxy mixture was removed from the silicone mold before curing. The cast polym er in the mold was placed into an oven at 58 for at least 12 hours to guarantee complete cross-linking of the polymer. After cure, the nanocomposite samples were easily released from the silicone molds and cut with a scissors or blade into the desired shape for characterization. The silicone mold was made from a two-part mixture of T2 Silastic (Dow Corning) base resin and curing agent. The silicone base resin was mixed in a 10:1 weight ratio with the curing agent. A typi cal batch mix consisted of 100g of the base resin and 10g of the curing agent. The mixture was hand sti rred thoroughly (approximately four to five minutes) and degassed under vacuum until little or no visible bubbles remained. A glass plate and several 1mm-thick microscope gla ss slides were cleaned with ethanol and surface treated with hexamethyl disilazane (Gelest) by vapor de position or swiping with a Kimwipe. This surface treatment was used to prevent adhesion of the silicone to the glass substrate once cured. The treated microscope slides were placed on the treated glass plate with approximately one inch sp ace in between each slide. Stacks of microscope slides may be attached to th e outer perimeter of the glass plate by doublesided tape to prevent overflow of the silicone over the plate and to determine the final thickness of the mold. The degassed silicone mixture was poured slowly over the glass plate and slides, starting in one corner of the plate and taking ca re not to trap air under the poured silicone.

PAGE 52

38 A second glass plate was cleaned and treate d with hexamethyldi silazane and placed carefully on top of the silicone, starting by a ngling the plate onto one side and slowly lowering onto the silicone to avoid trapping excess air in th e mold. A weight (~5 lbs.) was placed on top of the upper glas s plate to keep the entire set-up in place during cure. The silicone elastomer may be left to cure ove rnight at room temper ature or more rapidly cured at 50-60 for 3-4 hours. The silicone wa s then pulled off of the glass plates after cure and the glass microscope slides removed to reveal the indentations (wells) left in the mold. A razor blade was used to cut the overall mold into the desired size. The silicone mold was placed on a glass pl ate (with indentations exposed) and the polymer-nanoparticle mixture was cast into the 1mm thick wells as described above. This procedure yielded 1mm thick nanoc omposite films of va ried nanoparticle dispersion. The state of particle disp ersion was then characterized by scattering measurements. The ultrasonicated sample was systematically formulated to contain particle agglomerates that are smaller than the undispersed sample by adding a mechanism for physical dispersion of the loosely bound agglomerates in a reduced viscosity environment. The sample was fabricated by adding the titania nanopa rticles firs t to the low viscosity (9 mPa s, 25 C) curing agent (Jeffamine D230) and ultrasonicating the mixture at maximum power (250W) fo r 30 minutes in a cool water (20 C) bath. The instrument used was a cup-horn type ultrasoni c processor with a circulating bath. The particle-curing agent mixture was added to the epoxy resin and mixed by a homogenizer and cured as in the preparation of the undispersed sample.

PAGE 53

39 The surface treated sample includes the physical dispersion mechanism as described in the ultrasonicated sample preparation, but al so includes a silane coupling agent covalently bonded to the titania surface, prior to the mixing step. The aminopropyltriethoxysilane (APTES ) coupling agent, was chosen to decrease the surface energy of the particle surface to make it mo re compatible with the surrounding polymer, and also to provide a cova lent bond between the particle surface and polymer for additional cross-link density. The titania na noparticles were first added to the curing agent (Jeffamine D230) and ultrasonicat ed at maximum power for 25 minutes as described for the ultrasonicated sample. Next, 1mL of the APTES silane coupling agent ( or 2% by weight) was added to the disper sed particles and further ultrasonicated at maximum power for five minutes. The epoxy resin (EPON 828) was then added and mixed at 60rpm with a homogenizer until the na noparticles were completely incorporated into the resin. The polymer was then degasse d, cast into the silicone mold and cured as described previously. Table 3-4. Summary of co mponents used to fabricate nanocomposite samples for undispersed ultrasonicated and surface treated samples. 0.5 wt% TiO2 EPON 828 29.1 g Jeffamine D230 8.19 g APTES 1 mL TiO2 particles 0.187 g The fractionated sample preparation yielded th e optimal particle dispersion achieved in this study by employing physical a nd chemical dispersion techniques used in the previous three samples, in addition to a solvent mixture (acetone and isopropanol) to enhance particle wetting and di spersion. First, the drie d titania nanoparticles were weighed and divided equally into 50mL cap acity round-bottom Teflon centrifuge tubes.

PAGE 54

40 Each centrifuge tube contained less than 0. 5g nanoparticles. A 60/40% vol/vol mixture of acetone and isopropanol was added to comp letely fill the centrifuge tubes. The rationale for the solvent mixture is presente d in the following sect ion. The particle agglomerates were then physic ally dispersed using a high po wer cup-horn type ultrasonic processor for one hour at maximum power (250W) by rotating each individual centrifuge tube into the ultrasonicator for 4-5 minut es each. The particle/solvent mixture temperature was regulated by a water bath at 20C. It was determined through experimentation that one hour ultrasoni cation time resulted in the smallest aggregate/agglomerate size as measured by dynamic light scattering with no apparent fragmentation of the particles or inducing any unwanted chem ical changes to the polymer matrix due to degradation.90 When ultrasonication was applied, the l oosely bound agglomerates were broken. The good wetting characteristics of the solven t molecules with the particle surface aided the ultrasonication technique and enhanced disp ersion. The lower visc osity of the solvent as compared to the hardener (used as disper sant in other samples) increased particle wetting simply by reducing viscous forces that oppose wetting. In addition, the cohesive energy density difference may be lower for the solvent-particle system as compared to the hardener-particle mixture (Ec information could not be found for Jeffamine D230 hardener). A bimodal particle size distribution was obtained from the ultrasonicated nanoparticles by dynamic light scattering (Figur e 3-11), which shows the large aggregate size fraction remaining in the solvent mixture on the right side of th e distribution, while the small size fraction consists of the disp ersed particles (broken agglomerates). The

PAGE 55

41 average particle size obtained from this di stribution was approximately 280 nm diameter. The solvent aided in the dispersion of the nanoparticles thro ugh ultrasonication techniques by reducing viscos ity and increasing wetting, ma king the VDW barrier easier to overcome. After ultrasonication, the centrifuge t ubes were placed immediately into a high speed centrifuge (Beckman, model J2-21) that was outfitted with a carbon fiber reinforced rotor (FiberLite, model F13-14x50c y). The samples were centrifuged for ten minutes at 5000 rpm to settle out any hard a gglomerates that remained undispersed after ultrasonication. The supernatant was re moved by pipette after centrifugation and transferred into a 200 mL round flask. The se dimented particles were recovered, dried and weighed to determine the weight fracti on of particles in the nanocomposite by the following equation: 100 loading particlemin e Jeffa EPON dried total dried totalW W W W W W (2) where W is weight of the nanoparticles, pol ymer resin (EPON) or curing agent (Jeffamine). After sedimentation (Figur e 3-12), the particle size distribution was measured again by dynamic light scattering (F igure 3-13). The average particle size decreased dramatically to 50 nm, indicating th at the majority of the particles remained dispersed to primary particle size (~30 nm). The supernatant was agitated by a magnetic stir bar at a speed of 300rpm. The silane coupling agent, aminoethyl am inopropyltrimethoxy silane (AEAPTMS) was directly added dropwise to th e flask in the amount of 100 L with further stirring for at least 15 minutes. The silane coupling agen t (AEAPTMS) reacted by covalent bonding to the hydroxyl groups present on the particle surfa ce. The supernatent was left to age for

PAGE 56

42 24 hours to verify that reagglomeration wa s prevented by surface modification. The particle size distribution was measured once ag ain and the average size increased slightly to 55 nm, indicating that the treated nanopart icles remained dispersed to near primary particle size (Figure 3-14). The rationale for this behavior is that the silane groups provided a physical steric barrier to appr oaching nanoparticles and provided a larger interparticle dist ance in solution than the long range effects of VDW attractive forces. The three amine groups present at the end of th e silane groups repelled each other mainly by similar dipole-dipole interactions. The epoxy resin and curing agent were added to the flask in the ratio 3.55:1 (by weight) and stirred with a homogenizer at 60 rpm until the resin was completely solubilized in the particulate suspension. The flask was next affixed to a rotary evaporator (RotaVAP, Buchi) and rotated at 125 rpm with a fixed water bath temperature of 60C. The solvent was completely evaporated in a reduced pressure environment in less than 30 minutes. The final solvent-free polymer particle mixture was cast into 1 mm deep silicone wells and cured at 58C for at least 12 hours. Table 3-5. Summary of com ponents used to fabricate the fractionated sample. 0.5 wt% TiO2 EPON 828 9.7 g Jeffamine D230 2.73 g AEAPTMS 100 L TiO2 particles 0.0625 g

PAGE 57

43 Figure 3-10. Dry titania nanopowder in the isopropanol and acetone solvent mixture, showing good wetting of the particle surface. 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 0.00010.00100.01000.10001.0000 Particle Diameter ( m)Relative Volume 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 0.00010.00100.01000.10001.0000 Particle Diameter ( m)Relative Volume Figure 3-11. Particle size distribution measured after ultrasonication of the dry nanopowder in a solvent mixture. Figure 3-12. Digital image of the disperse d particles in the solvent mixture after sedimentation and removal of the hard agglomerates.

PAGE 58

44 0.00 5.00 10.00 15.00 20.00 25.00 0.00010.00100.01000.10001.000010.0000 Particle Diameter ( m)Relative Volume 0.00 5.00 10.00 15.00 20.00 25.00 0.00010.00100.01000.10001.000010.0000 Particle Diameter ( m)Relative Volume Figure 3-13. Particle size di stribution measured after sedime ntation and removal of hard agglomerates in the solvent mixture. 0.00 5.00 10.00 15.00 20.00 25.00 0.00010.00100.01000.10001.000010.0000 Particle Diameter ( m)Relative Volume 0.00 5.00 10.00 15.00 20.00 25.00 0.00010.00100.01000.10001.000010.0000 Particle Diameter ( m)Relative Volume Figure 3-14. Particle size distribution measured after ag ing of the surface treated nanoparticles in the solvent mixture to verify that the particles remained dispersed. Solvent selection for fractionated sample The choice of solvent to disperse the tit ania nanoparticles was dependent on two factors; (i) the wettability of the particle surface by the solvent, and (ii) the ability of the solvent to solubilize the epoxy resin upon mi xing. Both acetone and isopropanol were chosen as candidates due to th e solubility parameters bei ng close to that of the epoxy

PAGE 59

45 resin (Table 3-6).91, 92 The solubility parameters are dependent on the cohesive energy density, which is a measure of how easily the solvent molecules can penetrate the polymer. Similar solubility parameters be tween the polymer and solvent indicate a good solvent for a particular polymer. Table 3-6. Solubility parameters for the c hosen solvents to disso lve the epoxy polymer. The proximity of the values to that of the epoxy represents a good solvent. Solubility parameter (cal/cm3)1/2 Source Epoxy polymer 8.9 Mezzenga, 2000 Acetone 9.9 Sperling, 2001 Isopropanol 11.5 Sperling, 2001 The Beer Lambert Law relation was employed to determine the extent of particle surface wetting as it relates to the scattering of the part icles or agglomerates. For particles that have been ultras onicated in a low viscosity solv ent, it is assumed that good particle wetting would produce a scattering re sponse indicative of little or no multiple scattering in the visible wavele ngth range due to dispersed pa rticles. If the particles segregate (phase separate) from the solven t due to poor wetting characteristics, the agglomeration of the particles will be evid ent by an increase in scattering in the 400700nm wavelength range, which is observed as a deviation from th e Beer-Lambert Law (equation 3). The optical density of a range of titania particle concentrations in both solvents was measured by UV-Vis spectroscopy (see subseque nt section on optical density). The following linear relation holds for single scatteri ng of particles at low concentrations that are dispersed to a particle size that is sma ll compared to the visible light wavelength: c l A (3) Where A is absorbance, is the absorption coefficient, l is the pathlength the light travels through the sample, c is the particle concentration, and I represents the transmitted

PAGE 60

46 and incident light. The wavele ngth range of interest for this study is visible light (400700 nm), as a measure of optical transparency A sample of dispersed nanoparticles at primary particle size will scatter a very sma ll amount of light at those wavelengths, thus producing a high transmitted intensity reading at the detector. However, if the sample is agglomerated to aggregate sizes comparable to the wavelength of the source, multiple scattering will occur and the amount of tran smitted light will decrease. Conversely, a highly agglomerated sample wi ll produce a higher absorption measurement than that of a dispersed sample of nanoparticles. A Beers Law plot of titania particles in acetone and isoprop anol (Figure 3-15) reveals that both solvents start to deviate from a linear relation as the particle concentration was increased. However, if a trendline is fitted to the linear low concentration portion of the cu rves (c < 0.2 g/L), the R2 goodness of fit term reveals that isopropanol (R2 = 0.999 0.002) shows less scattering due to particle agglomerates, and thus better particle wetting than acetone (R2 = 0.964 0.003). It can also be seen (Figure 3-16) that by adding a small amount of isopr opanol to an acetone-nanoparticle mixture, the particle dispersion is greatly improved due to better wetting of the particle surface by the isopropanol.

PAGE 61

47 0 0.5 1 1.5 2 2.5 00.050.10.150.20.250.3 Particle Concentration (g/L)Absorbance (a. u.) Isopropanol 600nm Acetone 600nm Figure 3-15. Beers Law relation for optical density of titania nanopa rticles dispersed in acetone and isopropanol, meas ured at 600nm wavelength. Figure 3-16. Digital pictur es of (A) 7g/L TiO2 nanopart icles in pure acetone where sedimentation has occurred due poor wettability and (B) 7g/L TiO2 nanoparticles in 60% acetone/40% isopropanol with improved particle wetting. Even though isopropanol most effectively we ts the titania nanoparticle surface, it was found that it does not completely solubilize the epoxy resin used in this study. It was determined through further optical density m easurements that the difference in titania particle wetting between pure isopropanol and a mixture of equal parts of both solvents was not experimentally significant (< 5%). Solubility experiments revealed that a 60% A B

PAGE 62

48 acetone solution was the lowest volume concentr ation that could completely solubilize the epoxy resin without additional heating. Therefore, a 60/40% ratio of acetone to isopropanol was used as an appropriate solvent mixture to both effectively wet the particle surface and solubilize the epoxy resin. Particle agglomerate sedimentation The speed and time of centrifugation were se lected such that the particles dispersed by ultrasonication would stay suspended in th e solvent and any aggl omerated particles would sediment out of the so lution. The Stokes equation wa s applied to calculate the settling velocity of the su spended titania particles 9 22 f p sg r V (4) where r is the radius of the particle (or agglomerate), g is the acceleration due to gravity, is the density of the particle and fluid, respectively, and is the fluid viscosity. The optimal speed and time necessary to sediment any large agglomerates was calculated according to the geometry of the centrifuge (T able 3-7). The particles or agglomerates would need to travel no more than 3.5 cm to reach the bottom (or side) of the centrifuge tube, based on the angle during centrifugation. It was determined that a speed of 5000 rpm for ten minutes would be sufficient to set tle out the large particle agglomerates and leave all of the dispersed pr imary particles suspended in the supernatant. From the calculated values in Table 3-7, it can be observed that any particle aggregate or agglomerate larger than 100 nm will settle to the bottom of the centrifuge tube.

PAGE 63

49 Table 3-7. Stokes Law partic le settling calculations base d on ten minutes centrifuge time. At 5000 rpm, the primary partic les remain suspended, but the large agglomerates (>100 nm) will settle out. Size particles (nm) Settle Velocity (cm/s) Distance Settled (cm) 25 3.243E-04 0.195 75 2.919E-03 1.751 100 5.189E-03 3.114 300 4.670E-02 28.022 1000 5.189E-01 311.351 Optical Bench The state of dispersion of the nanopa rticles in the polymer matrix was characterized, in part, by angular depe ndent light scattering measurements. The construction of an optical bench was requi red due to the absence of commercial instruments to make such measurements. Co mmercial laser diffraction instruments used for particle size analysis commonly measure only a fraction of the full angular scattering pattern and are not equipped to measure the scat tering of solid materials, such as polymer films. For that reason, an optical bench wa s constructed to be flexible enough to measure different sample geometries, such as polymer films or cuvettes with liquid samples, at any scattering angle. Many instances of custom-built laser light optical benches for measuring particle scattering pa tterns are described in the lit erature and were used as a guide for the one built for this study.57-65, 93 Optical components The general components needed for the mo st simplistic optical bench to measure spherical particles are a radi ation source, sample holder, and a detector mounted on a rotating arm. Note that conversely the sample can also be rotated and measured with a fixed detector. The instrument built for this study has a 632.8 nm wavelength HeNe laser, beam expander, static sample holder with several interchangeable components for different sample geometries, and a photon dete ctor mounted on a goniometer arm, used to

PAGE 64

50 control the measured scattering angle (Figure 3-17). The distance from the sample to the detector can be changed to allow for ma ximum angular resolution, depending on the particle size measured. When measuring sma ll particles (<100 nm), it is important to maximize the distance from the sample to th e detector to identify changes in the scattering pattern over fractions of a degree. Optical components such as neutral dens ity filters, a Fourier lens and various apertures were added to the set-up for addi tional precision of the laser and scattering measurements. It was also found that due to high sensitivity of the detector, a light diffuser or aperture can be pl aced in front of the detector when measuring high levels of scattered laser light. All of the major compone nts of the optical benc h, the manufacturer, and part number are listed in Table 3-8. All other components were obtained from typical optics suppliers (e.g., Edmund Optics, Melles Griot). Table 3-8. Optical component s of the light scattering inst rument, the manufacturer and part number Optical component Manufacturer Part number Laser Melles Griot 25LHR121-249 Beam expander Melles Griot 09LBM015 Goniometer Brookhaven Model BI-2030 Photon detector SensL PCSTime Detector Selection The measurable dynamic range of the detect or must be such that even a small amount of scattered light from small particle s (<200nm) or turbid samples can be easily measured. However, the range must be wide enough so that a very transparent film or highly scattering sample will not saturate th e detector. A CCD (charge coupled device) camera detector was not able to provid e enough precision to make the scattering measurements in this study. Therefore, a highly sensitive photon detector (SensL) was

PAGE 65

51 used. Although this type of equipment is more expensive, the limitations of cheaper equipment are many, especially when measur ing low levels of scattered laser light. Figure 3-17. Schematic representation (A) a nd digital picture (B) of the optical bench set-up for light scattering measurements. Laser calibration When making sensitive scattering intensity measurements, it is important that the incident radiation intensity remains stable within experimental error. If a systematic error occurs, then it can be accounted for in th e light scattering measurements of the nanocomposite material. For that reason, the beam intensity was measured at 4

PAGE 66

52 scattering angle for 8.3 minutes. Measurements of the intensity were taken every 5 seconds and normalized. The experimental error was approximately 1% for all measurements and is assumed to be negligible for all measurements taken using this instrument (Figure 3-18). 0 0.5 1 1.5 2 0 20 40 60 80 100Measurement numberIntensity (Norm.) Figure 3-18. Intensity of the incident beam over 8.3 minutes with measurements taken every 5 seconds. The error is as sumed to be negligible (~1%). Instrument calibration In addition to checking the incident laser be am for fluctuations, it is also necessary to verify that the instrument is providing meaningful experimental results. Several samples were made that contain standard pa rticles that are commonl y used for particle sizing calibration. These samples were made of polystyrene part icle standards (Duke Scientific Corporation) disper sed in water to check the validity of the measurements. These particles are approved by the NIST or ganization as acceptabl e traceable materials for particle size calibration (www.nist.gov). Th e polystyrene particles are available in a range of sizes with very small particle size distributions ( 2%). These particles have also been used in the literature for scattering instrument calibration.94 The samples were measured on the light scattering bench and co mpare to the calculat ed Mie model. The

PAGE 67

53 results show that the Mie model and the m easured scattering intensities were similar (Figure 3-19) for several particle sizes, indicating that the instrument was properly calibrated to provide meaningful meas urements of scattering intensity. Snells Law Correction It is important to note that the samples us ed for calibration are particles dispersed in water contained in a cuvette. Due to the re fractive index differen ce between air and the cuvette used to hold the sample a correction had to be made due to refraction. According to Snells Law, a simple relation can be used to correct for the error in angular scattering: b b a an n sin sin (5) where the subscript a refers to the refractive index a nd scattering angle in air and the b subscript refers to that of the suspension (in this case, water). This correction has been used by others making similar measurements.59 However, the Snells Law correction is not needed for all samples. Since the pathle ngth of a cuvette is 10 mm, there is a large difference in the angle of the scattered light due to the refr active index change. If the pathlength is approximately 1 mm or less, the correction is not needed because the change in scattering angle would be compara tively small. In the case of measuring the polymer films used in the present study, th e mathematical correc tion was not required.

PAGE 68

54 0 0.5 1 1.5 2 2.5 051015202530 Scattering Angle (Degrees)Intensity (a.u.) Mieplot Experiment 0.1 1 10 0 5 10 15 20 25 30Scattering Angle (Degrees) Intensity (a.u.) MiePlot Experiment 0.0 0.1 1 10 0 5 10 15 20 25 30Scattering Angle (Degrees) Intensity (a.u.) MiePlot Experiment Figure 3-19. Scattering intensitie s for 600 nm (A), 800 nm (B), 2 m (C)and 5 m (D) diameter polystyrene partic les dispersed in water. A B C

PAGE 69

55 0.01 0.1 1 10 100 0 5 10 15 20 25 30Scattering Angle (Degrees) Intensity (a.u.) MiePlot Experiment Figure 3-19 Continued Scattering Measurements Scattering measurements were carried out by placing the nanocomposite sample in the path of a light or x-ray electromagnetic field and measuring the intensity of scattered radiation at a given angle perpendicular to the scattering cross sectional area of the composite film. The theoretical background of scattering of any wa velength radiation by an array of particles is discussed in Chapter four. The resultant scattering pattern is directly related to the following factors acco rding to the general scattering equation: Refractive index (electron density) difference between particles and medium Particle size and size distribution Relative three-dimensional spacing of the particles Shape of the particles It is assumed in this study that the refr active indices and particle (and agglomerate) shape remain the same for all composite sa mples. Therefore, the objective of the scattering measurements is to examine the e ffects of a change in particle (aggregate, agglomerate) size and size distri bution, as well as the relative arrangement of the particles (aggregates, agglomerates) in space, as the st ate of dispersion is varied. These changes can be observed by a change in the scattering pattern over several magnitudes of length D

PAGE 70

56 scale: nano to micron size feat ures. Different lengt h scales of a sample can be probed by either changing the wavelength of the radiation, or the scattering angle measured, as it relates to the q scattering vector. The q value (see Chapter four) is approximately the inverse of the length scale probed by a gi ven scattering technique. At very low q values, the largest micron length scale (3 m < d < 90 nm) agglomerates were probed by static light scattering. Mid-range aggregates were probed by ultra small x-ray scattering (USAXS) and the smallest length scale (100 nm < d < 1 nm) of the individual primary particles was measured with small angl e x-ray scattering (SAXS) at the highest q values. Light scattering (3 m < d < 90 nm) A custom made light scattering bench setup was assembled to measure the angular dependent light scattering patterns of the epoxy nanocomposite films. As described previously, the instrument was calibrated w ith standardized particles to validate the accuracy of such measurements. First, the SensL photodiode software was initialized and a dark measurement (background correction) wa s taken for each a ngle included in the scattering measurements. This measuremen t indicated the amount of stray light (or noise) in each data point that can be subtracted off of the scattering measurements. A very low dark count is desired (<1000), but it is nearly impossible to prevent any environmental noise in the data. The solid epoxy-TiO2 nanocomposite films were mounted onto standard optical accessories (fixed lens holder) and placed on top of the goniometer. It is important that the incoming laser impinges on the sample perpendicular to the nanocomposite film surface so that the scattering is not affected by any angular misalignment. The software that controls the stepper motor on the goniomet er was initialized and the initial scattering angle set at 5 to avoid sensitivity of the detector at smaller angles. The detector can

PAGE 71

57 essentially be burned out if the laser is di rectly centered on the detector face. The electronic stepper motor was calibrated to move the goniometer arm in precise 1 increments. The SensL software was used to obtai n the number of photon counts hitting the detector and averaged over at least 5 seconds. The measurem ent interval was set at 2ms and the amount of signal smoothing was adjusted to 50%. The measurement and corresponding scattering angle were exporte d from the software to EXCEL for further data analysis. The stepper motor was engaged to move the goniometer to 6 and the measurement process repeated for the desire d number of angular measurements. The data collected was graphed with respect to scattering angle for the raw scattering pattern of the nanocomposite sample. Each nanocompos ite sample was measured three times to eliminate experimental error. The raw patte rn can be further an alyzed using the Mie model for an approximation of particle a gglomerate size without complicated data analysis. The windows based program, MiePlot v.3.4.18 was chosen (www.philiplaven.com) to quickly analyze the above lig ht scattering data (Figure 3-20 ). This particular program, developed by Philip Laven, calculates the theoretical scattering intensity of an array of particles based on the refractive indices of the particles and surrounding medium (polymer), and the size distribution of the particles (or agglomer ates) by solving the scattering equation by considering only Mie-type scattering. This program was also used to analyze the scattering data used in instrume nt calibration, as desc ribed previously. In other words, this program was utilized to theoretically model the scattering curve to obtain a first approximation of the particle aggregate or agglomer ate size distribution

PAGE 72

58 from static light scattering and to aid in th e alignment and calibrati on of the optical bench set-up. Figure 3-20. Example of a theoretical scatte ring curve generated by the MiePlot software for 200 nm (radius) TiO2 particles in epoxy with a 10% polydispersity and log-normal size distribution. Ultra small angle x-ray scattering (2 m < d < 80 nm) The instrument used for USAXS measuremen ts is configured for greater scattering intensity measurement resolution over a smaller angular range than stat ic light scattering. Due to the desired resolution in the range of 0-1 the x-ray source is collimated through a slit instead of a point. The resultant raw data is therefore slit smeared, and a correction is typically used to compare USAXS measur ements to other scattering data (light scattering, SAXS). In this study, the La ke method was employed for slit smearing correction. The x-ray source uses Cu K radiation at a wavelength of 1.54 The polymer nanocomposite film sample was mounted in the commercial instrument and left to run overnight to collect scattering pattern information. Due to the

PAGE 73

59 small angular range, the detector crystal is translated and tilted instead of moved about the 360 circle as in the SLS measurements The measured scattering intensity, I ( q ) was plotted with respect to the wave vector, q in the range 0.0003 -1 to 0.07 -1. The measurements were corrected for any background scattering that may occur due to environmental conditions when no sample is in the chamber. Small angle x-ray scattering (100 nm < d < 1 nm) The SAXS scattering data were collected using a Bruker Hi-Star 2D area detector, with x-rays produced by a Rigaku Ultrax18 ro tating anode x-ray generator operated at 40 kV and 60 mA. The Cu K radiation was monochromated w ith pyrolytic graphite, with a resultant wavelength of 1.542 A series of three pinholes was used to collimate the beam before impinging on the nancomposite sample. A beamstop was used for better measurement precision at the lowest angular measurements by absorbing the incident beam to avoid interference with the scattered signal intensity. Th e instrument software corrected the scattering data for any sample absorption and background scattering and averages the data azimuthally. Correcti ons were also made using the accompanying commercial software for differences in sample thickness, as changes in pathlength can affect the scattered intensity measurement. After the appropriate corrections, the scattered intensity, I(q) was plotted with respect to the q value over the range 0.0075 -1 to 0.15 -1. The scattering of the polymer matrix background was not subtract ed from these measurements, because the scattering of the polymer was negligible as compared to the scattering of the embedded particles. The experiment run time was base d on the number of scattering centers within the probed sample volume (beam volume). A higher number of scattering centers leads to lower running times because a large amount of data can be collected at once. In order

PAGE 74

60 to take a measurement, the nanocomposite sample was cut into the desired size (approximately 2 in. x 1 in.) and mounted on the sample holder with double sided tape. Scattering data analysis with IGOR Scattering raw data required a great deal of analysis before particle (primary, aggregate, agglomerate) size distribution a nd interparticle distance information can be extracted. The software Igor Pro from Wavemetrics, Inc. was used along with the Irena macros package that was developed by Jan Ilavsky at Argonne Nati onal Laboratories. The Irena macros package is provided as freeware at http://www.uni.aps.anl.gov/~ilavsky/irena.html The Irena macros were developed to analyze scattering data in the fo rmat of scattering intensity, I ( q ), with respect to the q value for any material. The detailed steps us ed for data analysis and suggested fitting parameters are listed in Appe ndix B, however, a brief descrip tion of the data analysis is provided here. The first step in analysis of the scattering data was to desmear the USAXS raw data due to slit collimated rather than point col limated data. The Irena code for desmearing utilizes the Lake method, which is an itera tive curve-fitting pro cedure developed in 1967 and is described elsewhere.95 The raw data was trimmed to exclude large error experimental data points, smoothed, and th en desmeared according to the specified slit length. The resultant desmeared data wa s exported to EXCEL and used for further analysis. The second step in data analysis was to combine the SAXS and USAXS data into one continuous scattering curve. This was achieved by importing both the SAXS and desmeared USAXS data into the Data Manipul ation Tool. The USAXS data was scaled to the SAXS data, which was measured in absolute units. The third step in the data

PAGE 75

61 analysis process was to calculate the particle (primary, aggregate or agglomerate) size distribution (PSD) and the inte rparticle distance for the ti tania nanoparticles embedded in epoxy in various states of dispersion. Fi rst the combined SAXS and USAXs curve was imported into the Least Squares Fitting Tool for standard models. The PSD of the primary particles was also calculated using th e SAXS data. The starting parameters for all PSD fits are located in Appendix B. The interparticle distance, or correlation length, was then calculated by using the Fractals M odel fitting tool with the combined SAXS and USAXS data. The shape of the particle aggregates and agglomerates were approximated as spheres with an aspect ratio of one. It is important to note here that the contrast parameter used for modeling must be calculated to reflect the electron density (refractive index) difference between the polymer and nanoparticles. The scattering cont rast can be calculate d using the scattering contrast calculator function included in th e Irena macros. Other parameters may be manipulated in the various modeling functions, however, it is imperative that a standard procedure be used with all samples when analyzing scattering patterns of multiple samples for reproducible results. Additional Characterization Optical density The optical density measurements of titani a nanoparticles dispersed in a solvent and embedded in the epoxy were made using a Perkin Elmer UV/Vis Spectrometer 900. The measurements were made over the spect ral range 400-700 nm because the visible wavelength range is of interest to this study for optical measurements. Custom-made sample holders were designed for these meas urements to hold solid polymer films by cutting a hole in an aluminum L-bracket with a hole drilled large enough for the

PAGE 76

62 transmitted beam to pass through. The bracket was painted black for minimal reflection of the laser. The sample was affixed to the bracket with magnets and placed into the sample holder (Figure 3-21). Figure 3-21. Custom-made sample holders fo r measuring the optical density of solid polymer films in the UV-Vis spectrometer. Commercial light scattering instruments Dynamic light scattering measurements were made to determine the particle (primary or agglomerate) size distribution of the titania nanoparticles in a solvent. A Microtrac Nanotrac Ultra commercial dynamic light scattering instrument with external measurement probe was used. The useful ra nge of particle sizes measured by this instrument is in the range of 0.8 to 6500 nm, which is useful for nanoparticles that have a 30 nm primary particle size. Static light sca ttering measurements were used to determine the particle size distribution of the dry TiO2 nanopowder before any dispersion techniques were used. The static light sca ttering measurements we re made on a Coulter LS 13 320 particle sizing instrument, with a useful measurement range of 0.4 to 2000 m.

PAGE 77

63 for the dry powder system. All scatteri ng data from commerci al instruments was analyzed using the accompanying commercial software. Transmission electron microscopy Transmission Electron Microscopy (TEM) sa mples were fabricated by slicing the epoxy-TiO2 composite samples with a Leica ultramicrotome and a Diatome diamond knife (45) into 100-200 nm thick sections. It was discovered though experimentation that thinner slices were very prone to beam damage in the microscope due to electron charge build-up in the sample and subsequent heating. The samples used for sectioning are made by casting the polymer-particle mixt ure before cure into polyethylene molds made specifically for TEM sectioning (Figure 3-22). The molded samples were removed from the molds after cure and roughly trimmed at the small end with a razor blade into a trapezoid shape having the dimensions not exceeding 0.5 mm across the base, 0.4 mm across the top, and 0.3 mm along the sides. The diamond knife was then used for sectioning. The sections are carefully collected from the wa ter in the boat of the diamond knife after slicing and mounted onto 300 mesh carbon grids (Electron Microscopy Sciences). The grids were examined with a JEOL 200CX at 200kV at various magnifications to determine the qualitativ e difference between the four nanocomposite samples with varied particle dispersion a nd to verify the primary, aggregate and agglomerate particle size meas ured by scattering techniques.

PAGE 78

64 Figure 3-22. Polyethylene embedding molds us ed for TEM polymer sample preparation. Scanning electron microscopy Scanning electron microscopy (SEM) micrographs of the nanocomposite samples were taken to examine the nanoparticle dispersion in addition to TEM. Samples for SEM were prepared by immersing a sample prepared for TEM into liquid nitrogen for approximately two minutes. The sample was removed from the liquid nitrogen and fractured into several pieces. The smaller shards of the nanocomposite sample were mounted onto an aluminum disk with carbon tape. The entire di sk was coated with a thin layer (2-3 nm) of amorphous carbon to prevent charging of the pol ymer. The samples were viewed with a JEOL 6335F Field Emission SEM at 10 kV a nd a working distance of 15 mm. The SEM working parameters are indicated on th e micrographs shown in this study.

PAGE 79

65 CHAPTER 4 SCATTERING THEORY In this chapter, the theoretical background of scattering techniques and the rationale as to why they are used to measure a certain length scale of the nanocomposite samples is disclosed. The scattering theory for SLS, USAXS, and SAXS measurements is first presented and the probe length sc ale is explained. A brief e xplanation of Mie theory is outlined as it relates to the modeling of the static light scattering (SLS) measurements. General Scattering Theory Scattering measurements entail the quantif ication of the amount of electromagnetic energy that is scattered at a specific a ngle from the composite sample volume of scattering centers, as compared to the amount of electromagnetic energy that is transmitted directly through the sample. The scattering of electromagnetic waves (radiation) is due to the diffe rence in electron density (optic al properties, i.e., refractive index) between the scattering centers (p articles) and the surrounding medium. All scattering techniques are governed by the sa me physical equations, regardless of the wavelength of the radiation source. The scattered intensity can be determined from the Poynting vector, S, which specifies the magnitude and dire ction of the rate of transf er of electromagnetic energy (scattered intensity) at all points of space relative to the scsatterer.96 The Poynting vector, given by E x H, represents the direction in whic h energy is being transported (the scattered direction) and has a magnitude equal to the power density of the wave (scattered intensity).

PAGE 80

66 The electric and magnetic fields (E and H) of an electromagnetic wave oscillate, thus the magnitude of the Poynting vector ch anges with time. Most instruments cannot measure an instantaneous Poynting vector and as a result, a time averaged value is used. The average magnitude of the P oynting vector over a long time (longer than the period of the wave) is called the irradiance or intensity, I.97 I dt tt t ' S 1 S (1) The time averaged Poynting vector for time-harmonic fields is given by96 *Re 2 1H E S (2) where E and H cannot be arbitrarily sp ecified, and must satisf y the Maxwell equations for a plane wave96: t t E H H E H E 0 0 (3) where is the permeability of radiation through the medium (surrounding material) and is the permittivity, which is also a property of the medium. The E and H vectors are broken down accordingly into the field associated with the particle (E1, H1), the surrounding medium (E2, H2), the incident field (Ei, Hi) and the scattered field (Es, Hs).

PAGE 81

67 Figure 4-1. Forward scatteri ng of a sphere showing the electromagnetic field (wave) vectors for the incident, partic le, and scattered radiation. Boundary Conditions When crossing the boundary between th e surface of the particle and the surrounding matrix, there is a sudden change in material properties, wh ich is reflected in a sudden change in optical properties, e.g., and However, some conditions are imposed on the E and H vectors that preserve the c onservation of energy across that boundary. These rules maintain that the ta ngential electromagnetic field (from the surface of the particle) is con tinuous across a discontinuous boundary96 0 0 1 2 1 2 n H H n E E (4) where both E and H lie on the surface of the particle.

PAGE 82

68 The Amplitude Matrix If one considers a particle (o r an array of particles) that is illuminated by a plane wave, the direction of propagation (the z ax is) defines the forward direction. The incident electric field, Ei, can be resolved into the para llel and perpendicular components respectively, El,i and Er,i. The scattering plane is determ ined by the azimuthal scattering angle, and the sca ttering angle, The scattered electric fiel d is defined as the vector, Es, which can also be resolved in to th e parallel and perpe ndicular components. Figure 4-2. Schematic depicting the scat tering of electromagnetic radiation by a spherical particle. The incident field is in the z-axis direction. The scattering in any direction is de scribed by four amplitude functions, S1, S2, S3, S4, which are all functions of the angles and and can be measured by determining the amplitude and phase of the scattered radiation96. The amplitude function may be complex, but rarely has it been physically m easured for all angular values. The relation between the incident and scattere d electric fields is given by:97

PAGE 83

69 i r i l ikz ikr s r s likr e S S S S, 1 4 ,E E E E3 2 (5) where S is the amplitude function matrix. For this case, the scattered electric field is given by two relations:97 i l ikz ikr s l i r ikz ikr s rik r e S ikr e S, 2 , 1 ,E E E E (6) For spherical particles, S3 = S4 = 0, and S1 and S2 are functions only of due to the scattering symmetries over the fu ll 0-360 scattering range of In general, the azimuthal angle can be ignored if the s pherical particle assumption can be used. In the present study, all particles and part icle agglomerates are assu med to be spherical. The calculations involved in calculati ng the scattered intensity from asymmetrical particles is beyond the scope of this research. The inte nsity of the scattered wave for spherical particles is proportional to the squa re of the amplitude as follows: 2 2 2 r k S Iscatt, 2 k (7) For various polarization states of the incident radiation: 2 2 2 2 2 1 2 2 2 2 // 2 2 2 1r k S S I r k S I r k S Ilight natural on polarizati on polarizati (8)

PAGE 84

70 For a system of spherical particles, not necessa rily similar, each one is characterized by a unique amplitude function, Si(q). Due to mathematical complexities, the intensities associated with each amplitude are summed over the entire system97 j jI I (9) The Mueller Matrix The incident light emitted in a scattering experiment actually consists of many simple waves in very rapid succession. Th erefore, measurable intensities always refer to a superposition of many millions of si mple waves with independent phases.97 Following the summation shown above, the Stokes parameters of the beam are j j j jV V U U Q Q I I (10) The factors Q, U, and V are dependent on the polariza tion of the incident light.97 The relation between the incident and scattered Stokes parameters make up the full amplitude scattering matrix, which is often denoted by the matrix F.96 i i i i s s s sV U Q I r k V U Q I44 43 42 41 34 33 32 31 24 23 22 21 14 13 12 11 2 2S S S S S S S S S S S S S S S S 1 (11) The 4 x 4 amplitude matrix is known as the Mueller matrix for scattering of a single particle. The 16 elements are not all independent and can be determined by knowing seven individual parameters. The number of parameters may also be reduced because of symmetry relations. If a collection of randomly distributed particles is present, as in our

PAGE 85

71 model nanocomposite system, the Stokes parameters of the system can be determined by summing the effects of the individual particles. Scattering of a Sphere The following discussion relates only to a single sphere or an array of spheres, which results in simplifications of the wave equations. Starting w ith the incident and scattered electric field amp litudes, the relation becomes96 i r i l ikz ikr s r s likr e S S, 1 ,0 0E E E E2 (12) and the Mueller matrix can be simplified as i i i i s s s sV U Q I S S S S S S S S r k V U Q I33 34 34 33 11 12 12 11 2 20 0 0 0 0 0 0 0 1 (13) For the incident radiation in the case of th e experimental light s cattering set-up used in this study, the laser is unpolarized and the Stokes parameters can be written as96 0 ,12 11 s s i s i sV U I S Q I S I (14) Angular Dependent Scattering In the scattering set-up described prev iously, the angular dependent forward scattering (0-90) intensity is measured. The first few scattering coefficients, as determined in the previous section, w ill determine the angular dependence of each scattering event, or particle96. For forward scattering ( < 90 ), the intensity plot is highly peaked whereas the backscatteri ng intensity plot lacks any str ong features and is lower in intensity by orders of magnitude. The pr onounced difference between intensities for forward scattering and backscattering become s even more dominant as the size of the

PAGE 86

72 particles, or scattering events increase. This factor is very important to the current study due to the fact that it is assumed that the particle agglomerates are scattering as individual events and increase in size as the particle aggl omeration increases. So, as the size of the scattering event ranges from primary particle size (nanometers) to large agglomerate (microns), the forward scattering component in creases much faster than the backscatter component. The Length Scale The vector q is called the scattering wave vector. The importance of this vector is that its approximate inverse represents the le ngth scale, or probe le ngth, of the scattering experiment19. The q value corresponds to the actual feat ure size, or length scale, that is measured by q d / 2 where d is the feature size. This is the reason for the inverse relationship between the particle size and the an gle at which the features of that size are observed in a scattering experiment. Consider an incident plane wave, ki on a scattering element at r. The field scatters toward the detector in the direction ks. If purely elastic scattering is assumed then75 2 s ik k (15) The phase of the scattered wave at the dete ctor is a function of the position of the scattering element and the vector s ik k q (16) which is the scattering wave vector (Figure 4-3). Simple trigonome try and equation (17) yield the magnitude of this vector as

PAGE 87

73 2 sin 4 2 sin 2 q qik (17) where is the scattering angle. Figure 4-3. Diagram of light incident from the left with wave vector ki scattering from an element at r toward the detector with scattering wave vector ks, at a scattering angle The difference in wave vectors, q, is shown. Adapted from Sorensen (2001). Mie Theory Mie Theory was first presented by Gustav Mie in 1908 and is the direct analytical solution for the scattering, ab sorption, and transmission of radiation through a dielectric sphere48. Recently, Mie Theory has come to include all calculations of the interaction of incident light with particles or material of a ny shape, size, size distribution, or texture. However, an exact solution only exists for th e single spherical dielectric particle, as mentioned above. Mie Theory can be used to calculate a particle size distribution from scattering intensity profiles via a matrix inve rsion the details of which can be found in Appendix A. The intensity profiles generate d through static light sc attering are used to I0 I(q)

PAGE 88

74 determine the particle and/or cluster particle size distributi on, required for the dispersion indices calculation. The derivation closely fo llows that of Bohren and Huffman (1983) and also van de Hulst (1957) for the scattering of light by small particles.96-97 The assumptions made are that the particles are spherical, the incident wave satisfies the Maxwell equation, and the a ppropriate boundary conditions are valid for the titania particles embedded in an epoxy matrix. Mie Scattering Coefficients The boundary conditions for the electromagne tic field, as explained above, require the continuity of the transverse components of the electric and magnetic field at the surface of the sphere. The transverse components of those fields are represented by Bmn( ) and Cmn( ). The coefficients of those term s can be expressed as a function of the internal and external fields (of the particle)83 , ,' ' ' n n n n n n n n n n n n n n n n n n n na c b m d a m c a c (18) where = ka and = mka are the external and internal size parameters of the particle and n and n are the spherical Ricatti-Bessel function s. The solutions of these equations yield the Mie coefficients, of which an and bn govern the scattering of the radiation and cn and dn govern the internal field.98

PAGE 89

75 , ,' ' ' ' ' ' n n n n n n n n n n n n n n n n n n n n n n n n n n n nm im d m im c m m b m m a (19) The scattering coefficients, an and bn can be determined for an arbitrary spherical particle and will completely specify the scattered field. .

PAGE 90

76 CHAPTER 5 RESULTS AND DISCSUSSION Dispersion Force Calculations A rough calculation of the van der Waals adhesive forces between two touching titania particles in an agglomerate yiel ds approximately 93 nJ/m by the following equation 2 1 2 1 26 R R R R H A FVDW (1) where A is the Hamaker constant for two bare (u ntreated) titania surfaces as suggested by French99, H is suggested as 4 nm for a contact interpar ticle distance99 and R is the primary particle radius of 30 nm. That am ount of energy, provided at the nano scale at each particle-particle interface, is required to overcome the van der Waals barrier and completely break all particle agglomerates for complete di spersion. Additionally, forces must be provided to prevent subsequent contact of the particles to maintain the particle dispersion. In addition, the free energy of the spreadi ng process of the epoxy resin on the bare titania surface is a large positive value (non-sp ontaneous), indicating a high probability of poor wetting of the surface by the polymer and segregation of the nanoparticles (primary or aggregates) into larger agglomer ates. The spreading coefficient, S can be calculated using the surface energies of the polymer and pa rticle components as an indication of the ability of the epoxy resin to flow over and we t the particles. A positive value reflects a kinetically non-spontaneous wetting conditi on, whereas a negative value indicates a

PAGE 91

77 spontaneous wetting process, if no other fo rces are present on the surface affecting wetting (e.g., surface roughness). The spread ing coefficient can be calculated as follows100 SL LV SV SS 12 (2) where is the surface energy of the respective components (titania particle, epoxy, particle-epoxy). The surface energy of the epoxy was obtained from the work of Page101 and that of the nanoparticles from the work of Navrotsky.102 The surface energy of the titania particle-epoxy inte rface was approximated from the Neumann model.103 These values are listed in Table 5-1. The calculated spreading coefficient for the bare titania surface, -2527 mJ/m2, indicates a poor compatibility between the particle and epoxy resin. However, the silane coupling surface treatment (APTES or AEAPTMS) lowers the surface energy of the particle to 36.5 mJ/m2 (from 1036 mJ/m2) and the spreading coefficient is increased greatly to 0.19 mJ/m2, which suggests that that epoxy resin would spontaneously wet the titania na noparticle surface, given no othe r forces are present. The lowering of the surface free energy of the titania and subsequent diffusion of the polymer in between nanoparticles is a competitive force to the interpar ticle van der Waals attractive forces, thus suggesting a decrease in agglomeration. Table 5-1. Surface energy values for the na nocomposite components and the calculated spreading coefficient for untre ated and treated particles. Surface Energy (mJ/m2) After treatment with AEAPTMS (mJ/m2) Reference TiO2 1036 ( SV) 36.5 ( SV) Navrotsky (2007) Epoxy 36.3 ( LV) 36.3 ( LV) Page (2001) TiO2-Epoxy 3527 ( SL) 0.01 ( SL) S12 -2527 0.19

PAGE 92

78 Qualitative Nanoparticle Dispersion Characterization In order to examine the affects of calcu lated surface forces on the dispersion of titania nanoparticles in an epoxy polymer, microscopy was used as the first method of investigation. Digital Images Digital images provided a qualitative indi cation of the optical density (transmitted light) of each sample of varied nanoparticles dispersion. Th e thick films (1 mm) created in this study were mounted agai nst typewritten pages to illust rate the clarity of the film for comparison to one another (Figure 5-1). It is obvious from these images that the fractionated sample has the highest transparency as compared to the other samples. It can be suggested from these init ial results that the new methodology for fabricating the fractionated sample using a solvent for enha nced particle dispersion must be employed to produce a nanocomposite with desirable optical properties for use in applications such as windows and food packag ing. According to th e Rayleigh regime of particle scattering, r x 2 where x is the characteristic Rayleigh size parameter (x <<1), the particles must be dispersed until a nano size range is reached so that the agglomerate size is small (~ /10) compared to the wavelengt h of visible light (400-700 nm) to avoid multiple scattering effects and resu lting sample opacity. In other words, the scatterers must be dispersed to approxim ately 40-70 nm in diameter to provide an optically transparent sample. In light of this, it can be surmised from visual inspection of Figure 5-1, that samples A, B, C and E have agglomerates larger than 70 nm due to the opacity of the films.

PAGE 93

79 Figure 5-1. Examples of digital images co mparing thick films of each dispersion state against a typewritten background. The samples are: (A) undispersed (B) ultrasonicated (C) surface treated and (D) fractionated Sample (E) is undispersed and (F) is the fractionated sample. Sample (G) is the pure epoxy polymer. Freeze Fracture SEM Micrographs Freeze fracture scanning electron micr oscopy (FF SEM) also provided the qualitative differences between nanocomposite samples due to variations in particle dispersion. SEM images also reveal the t opography of the fracture surface and a clear indication of the voids surrounding the particles due to poor wetting. The undispersed sample micrographs (Figure 5-2) reveal not only large agglomerates and surrounding voids, but also discontinuities in the epoxy polymer fracture surface at points where the agglomerates are present (picture C, Figure 5-2). The particle a gglomerates provided a point of crack initiation and failure at the m echanically weakest part of the sample, as described in the literature revi ew. It can be suggested from these preliminary results that A B C D E F G

PAGE 94

80 the mechanical properties of the undispersed sample would be poor, and possibly worse than that of the pure epoxy polymer. The fractionated sample micrographs (Figure 5-3) indicate that the nanoparticles were dispersed to near primary particle size a nd only a small number of clusters exist at the fracture surface. No discontinuities or voids in the polymer matrix can be identified. The large improvement in the particle surface wetting for the fractionated sample is due to the silane coupling surface treatment applie d to the titania surface as described in the previous sections, which lowers the surface en ergy of the particle by over an order of magnitude (Table 5-1). Figure 5-2. Scanning electron micrographs of the undispersed sample showing the large particle agglomerates and voids due to poor particle wetting by the epoxy. The voids are most obvious in picture (B). A D C B

PAGE 95

81 Figure 5-3. Scanning electron micrographs of the fractionated sample showing the small particle agglomerates and individually dispersed pa rticles. No voids are visible in the sample due to good particle wetting by the epoxy. TEM Micrographs Transmission electron microscopy (TEM) was used to characterize the nano scale state of dispersion by investigating both the average size of the particle agglomerates (and aggregates) and also the distance between agglomerates. This method was also used to obtain approximate values for particle agglomerate size for the modeling program (Irena), utilized to analyze the scattering data. Figure 5-4 reveals the difference in agglomerate and aggregate sizes between the differently dispersed nanocomposite samples. The undispersed sample has very large agglomer ates, on the order of one micron, which is more than one order of magnitude la rger than the primary particles (30 nm). The simple mixing of the dry nanopowder in to the epoxy resin and hardener with no surface modification provided a highly agglomerated and inhomogeneous composite material. The as-received dr y nanopowder contained many large agglomerates due to the interparticle van der Waals attractive forces and mixing provides very little energy to overcome those forces, thus a highly agglomerated sample is observed.

PAGE 96

82 The ultrasonicated sample agglomerates are appr oximately 500 nm in diameter, still much larger than the primary particles. The sample was prepared by ultrasonication in addition to mechanical mixing of the TiO2 nanoparticles into th e epoxy resin. This procedure results in some reduction in the size of the particle agglomerates. This reduction is due to the energy input of the u ltrasonicator to overcome the van der Waals (VDW) forces. Ultrasound wavelengths ca nnot produce dispersion eddies on the nano length scale, regardless of energy input and dur ation of ultrasonicat ion, as described by the Kolmogoroff equation23, 24, thus preventing full dispersion of the particles to primary particle size and bonded aggregates. Additi onally, the bare titania surface remains incompatible with the epoxy re sin (no surface treatment) and further prevents particle dispersion. The surface treated sample contains agglomerates that are approximately 250 nm in diameter. The sample preparation invol ved silane surface treatment in addition to ultrasonication and mechanical mixing of the pa rticles into the polymer. This procedure resulted in a further decrease in agglomerat e size due to the surface modification of the titania nanoparticles. This is attributed to the APTES molecule s covalently bonding to the hydroxyl groups present on the metal oxid e surface and reduction in the free energy of mixing to that of a spontaneous process (Table 5-1, similar values as AEAPTMS). However, as indicated by the above sample preparation technique ( ultrasonicated sample), ultrasonication is not able to fully disperse the nanoparticles before the surface treatment was applied, thus essentially coat ing the surface of the agglomerates. The surface treatment did, however, prevent furt her agglomeration during mixing and curing

PAGE 97

83 of the epoxy resin and hardener due to ster ic repulsion. Therefore, a reduction in agglomerate size from the ultrasonicated sample was observed. The high resolution TEM images of the fractionated sample (Figure 5-5) demonstrate that individual particle sepa ration occurs due to good surface wetting by the epoxy resin. The small clusters seem to be approximately 60-70 nm. The fractionated sample involves the use of a solvent to reduc e viscosity and increase particle wetting in addition to surface treatment, ultrasonication, and mechanical mixing. This procedure generates the smallest particle size and mo st optimal dispersion of the dry titania nanopowder in the epoxy matrix achieved in this study. The de tails and rationale for this result were shown in Chapter three. In this study, images provided not only qualitative verificatio n of the state of nanoparticle dispersion, but also a priori knowledge to formulate an approximate value of particle (primary, aggregate and agglomerat e) size for modeling purposes. The objective of this study was not to prove that micros copy should not be used for the assessment of particle dispersion, but rather to show that it can be used with success to complement more quantitative techniques. Figure 5-4. Transmission electron mi croscopy images of 0.5 wt% TiO2 embedded in epoxy: undispersed (A), ultrasonicated (B) and surface treated (C). A C B

PAGE 98

84 Figure 5-5. High resolution TEM micrographs of the fractionated sample (0.5 wt%) showing the nano scale dispersion and sepa rated individual particles. Note the difference in scale bars. Quantitative Measure of Dispersion It is clear from the qualitative assessme nt of dispersion that one could perhaps reliably compare good and bad samples and estimate the approximate average particle agglomerate size, but it would be extremely di fficult to obtain any statistically relevant information without tedious and time consumi ng evaluation over a representative sample area. It is for these reasons that the desire for a more quantitative measure of dispersion led to the development of a dispersion index concept.

PAGE 99

85 Rationale for the Dispersion Index The dispersion index concept was developed in this study to obtain a quantitative measure of the state of dispersion of the na noparticles embedded in the polymer matrix. Literature review and preliminary efforts i ndicated that it may be difficult to quantify dispersion in a single parameter considering pa rticle-particle inter actions are governed by surface chemistry and spatial distribution are influenced by physical mixing forces. Accordingly, the nano length scale changes in agglomerate size due to surface chemistry effects, and also the macro length scal e mixing effectiveness of the nanocomposite fabrication process needed to be taken into account. Thus, the dispersion concept was developed to include two parameters: one to describe the nano scale wetting of the individual particles ( D1), and the other to reflect the rela tive particle arrangement in the matrix ( D2) due to mixing. Both of the dispersion indices were expected to range in value from zero to one: a value of zero repr esenting an infinitely agglomerated system and a value of one representing a theoreti cally ideal, random dispersion of primary particles, homogeneously distribut ed throughout the polymer matrix. Surface Chemistry ParameterD1 The surface chemistry parameter, D1, reflects the compatibility and wetting of the titania nanoparticle surface w ith the surrounding polymer on the nano length scale. The wetting that contributes to pa rticle dispersion is best represented by the particle agglomerate size and size distribution, as increased wetting has been shown here to enhance dispersion of the nanopar ticles and reduce agglomerate size.104 When an appropriate solvent was used to increase wetting of the particles and a surface modification was employed to maintain the nanoparticle dispersion, the epoxy resin, when added, can permeate the particles and wet the surface of individual particles.

PAGE 100

86 Therefore, increased wetting of the particles is expected to produce smaller agglomerates, thus creating a narrower size distribution and smaller average particle size that closely resembles that of a theoretically ideal dispersion (all primary particles). In this study, the full width half ma ximum (FWHM) value of the particle (agglomerate) size distribution is used as a measure of the breadth of the distribution. The FWHM value combined with the mean pa rticle (or agglomerate) size is used to compare a theoretically ideal particle di spersion to the experimental nanocomposite samples as follows, N N WD WDd FWHM d FWHM D 1 (3) where the subscripts WD and N represent the randomly well dispersed and experimental nanocomposite samples, respectively. The variable, d is the average particle or agglomerate size. The ratio of the average agglomerate size as compared to the primary particle size of an ideal sa mple was suggested by Balek25 as a reasonable measure of dispersion. In order to demonstrate the relevance of the D1 parameter, four different particle size distributions, including primary particles, aggregates, and agglomerates) were simulated using a log-normal distribution ca lculation. Each case reflects a change in mean particle size and/or di stribution breadth from the id eal dispersion of nanoparticles in Case 1 (Figure 5-6), however, a homoge neous spatial distribution of particles or agglomerates is assumed. The following four simulated size distri bution cases are shown graphically in Figure 5-7: Case 1 : the ideal case of randomly well disper sed particles. The average particle size is 30nm and the standard deviation is 6nm.

PAGE 101

87 Case 2 : has the same average particle size as the ideal case, 30nm, but has a larger standard deviation of 12nm, which provide s a wider distribution breadth. This condition may occur if some of the particle s from Case 1 were to agglomerate and some particles were fractured to smaller sizes. Case 3 : the average particle size is increased to 75nm and the standard deviation is 6nm. The distribution shifts to a larger size, but th e width of the distribution remains the same as the ideal case. This condition may occur if the particles from Case 1 were to agglomerate into clus ters of similar size for a narrow size distribution. Case 4 : the average particle size is increased to 75nm, and standard deviation is increased to 15nm. This di stribution has a large averag e size and large breadth of distribution. This condition may occur if some particles from Case 1 agglomerate into differently sized clusters, a nd some particles remain dispersed. Figure 5-6. Schematic of the four simulated particle agglomerate mo rphologies used to determine the validity of the D1 dispersion index. 0 0.02 0.04 0.06 0.08 0.1 0.12 0.010.11 Particle Diameter (um)Volume Frequenc y Case 1 Case 2 Case 3 Case 4 Figure 5-7. Graphical repres entation of the four simulated cases of particle or agglomerate size distribution. Ca se1 Case3 Case2 Case4

PAGE 102

88 The results of the D1 calculation of four simulated partic le size distributio ns (Figure 5-8) reveal a trend of decreasing value as the average particle agglomerate size and/or distribution breadth is changed from the ideal case (Case 1). It can also be observed that simply agglomerating the particles in Case 1 into the schematic shown in Case2 provided the largest change in the D1 parameter. The difference between Case 2 and Case 3, where the agglomerate size was increased but the distribution was narrowed, the smallest change in D1 was observed. Using a similar comp arison between Case 1 and Case 4, both agglomeration of the particles and a re spective change in the breadth of the distribution provided the most significant change in the D1 parameter. It is expected that the nanocomposite samples fabricated for this study would most closely resemble that of Case 4. 0 0.2 0.4 0.6 0.8 1Case 1Case 2Case 3Case 4D1 value Figure 5-8. Surface chemistry parameter D1 values of four simula ted size distributions. D1 Calculation for the Nanocomposite Samples The D1 component utilizes the FWHM value and average particle agglomerate size to determine the change in particle surface wett ing and compatibility of the particles with the polymer matrix as the particle dispersi on state varies. The FWHM value for the well

PAGE 103

89 dispersed ideal system (Case 1) was obtained experimentally from a measurement technique that probes the nano length scale of the individual nanoparticles. Therefore, small angle x-ray scattering (SAXS) was empl oyed to measure the scattering of the primary particles and obtain the average si ze and FWHM value of the particle size distribution. The FWHM values for the e xperimental nanocomposite samples of varied particle dispersion were obtai ned from USAXS data, due to the larger probe length scale on the order of the size of the particle agglomerates in those samples. The details and results of the scattering measurements are shown in subsequent sections. Mixing ParameterD2 The mixture homogeneity component, D2, reflects changes in the spatial distribution of the particles or clusters within the composite material. Practically, it can be considered an indication of the eff ectiveness of the mixing process for the nanocomposite fabrication. These changes in mixing are best characterized by measuring the interparticle chord length, or surface to su rface distance between particles or particle agglomerates. In a theoretically ideal disper sion of particles, the average interparticle distance is dependent on the par ticle size distribution and the volume fraction of particles in the sample. Assuming optimal surface chemistry, the average interparticle distance in an agglomerated composite sample will depend on the clustering effect of the particles as a result of mixing (Figure 5-9). If the agglomerates are dist ributed homogeneously within the sample volume, then the interparticle dist ance will be larger than that of the ideal dispersion. If the particles were to segreg ate to one fraction of the sample volume, the interparticle distance would be smaller than th at of the ideal dispersion. A preliminary assessment by techniques such as microsc opy is helpful in assessing the qualitative

PAGE 104

90 homogeneity of the particle dispersion. The D2 mixing parameter is the average interparticle distance ca lculated for the ideal particle dispersion as compared to the experimental dispersion as follows, N WDT T D 2 (5) where T is the average interparticle (inter-aggl omerate or inter-aggregate) distance and the subscripts WD and N represent the randomly well disp ersed (ideal) and experimental nanocomposite particle dispersions, respectively. The following four theoretical cases of partic le spatial distribution are considered to explain how the D2 mixing parameter changes with particle (agglomerate) spatial distribution. The results of the interparticle distance calcula tions for the four simulated cases were weighted by the volume fraction in which the particle agglomerates were located. Case 1: represents the ideal randomly well dispersed condition. The average particle size is 30nm. Case 2: the average particle size is increased to 75nm due to agglomeration. The agglomerates are dispersed homogeneously within the entire sample volume. Case 3: the average agglomerate size is 75nm. The agglomerates are not dispersed homogeneously within the sample volume. All of the particle agglomerates segregated to 50% of the sample volume due to poor mixing. The D2 value was weighted by half the sample volume to account for other agglomerates located elsewhere in the sample that would pr oduce a very large c hord length between agglomerates due to segregation. Case 4: the average agglomerate size is 75nm The clusters are not dispersed homogeneously within the sample volume. All of the particle agglomerates segregated to 25% of the sample volume due to poor mixing. The D2 value was weighted by one quarter of the sample volume to account for other agglomerates located elsewhere in the sample that would produce a very large chord length between agglomerates due to segregation.

PAGE 105

91 Figure 5-9. Schematic of the four simulated particle agglomerate mo rphologies used to determine the validity of the D2 mixing parameter. The results of the D2 calculations for the four simulated particle distributions presented in Figure 5-10 indi cate that the value of D2 decreases when the particles agglomerate and/or segregate into a fraction of the overall sample volume. It can be observed that just agglomerati on of the particles, even wh en mixing was optimal, reduced the D2 value by more than 50% due to the change in interparticle distance. It is clear by comparing Case 1 and Case 2 that the D2 value is greatly affected by the parameters that govern the D1 value, such as particle wetti ng and dispersion. However, the D1 value is not affected by segregation issues due to poor mixing, as long as the particle agglomerate size and distribution remain the same. In other words, the D1 value can stand alone as a measure of how well the surface of the pa rticle was made compatible with the surrounding polymer to enhance dispersion, but the D2 mixing parameter is affected not only by the mixing of the particles within the sample volume, but also by the particle agglomeration dominated by the surface chemistry. Case1 Case2 Case3 Case4

PAGE 106

92 0 0.2 0.4 0.6 0.8 1 Case 1Case 2Case 3Case 4D2 values Figure 5-10. D2 mixing parameter values of four simulated particle morphologies. D2 Calculation for the Nanocomposite Samples The D2 mixing parameter employs the calculation or measurement of the interparticle distance, according to equation (5). The theoretical distan ce between particles for an ideal, random dispersion can be calculated according to the derivation of Irani and Callis105, and also Liu, et. al26-27 and can be found in Appendix C. The equation used to calculate the theoretical dist ance between particles for an ideally dispersed system of primary particles, TWD, is as follows, 2 2 3 1ln 2 1 ln 2 3 exp 6 d TWD (6) where d is the average diameter of the primary particles, is the volume fraction of primary particles in the nanocomposite sample, and is the polydispersity of the size distribution where = 1 represents a monodisperse distribution. The parameter is the spatial distribution factor26, which takes into account the ra ndom dispersion of particles

PAGE 107

93 in the ideal case. If th e particles were fixed to a cubic lattice, then = 1. However, it was empirically determined that = 1.1 for a random dist ribution of particles20. Due to the nano length scale of the primary particles in the ideal dispersion, the parameters necessary for the calculation of TWD in equation (6) were measured using the SAXS technique. The inter-agglomerate distance required for TN in equation (6) is characterized by the correlation length, which is measured by scattering experiments that probe the length scale between the agglomerates. When ex amined by microscopy, the length scale was observed to be in the micron size range, whic h is best measured by USAXS or static light scattering (SLS). TN is the number average of chord le ngths that are drawn through every particle in every direction68. It can be calculated theoretically from the correlation function, which is the averag e of the product of two fluctu ations in electron density (refractive index) between two particles or particle agglomerates at a distance r r r0 2 (7) where 0(r) is only related to the geometry of the particle. In this st udy, all particles and particle agglomerates are a ssumed to be spherical, thus 0(r) can be easily calculated for each set of two particle s or agglomerates by 3 02 1 2 3 1 d r d r r (8) where r is the chord length between particles and d is the average par ticle diameter. The correlation function can then be mathematically related to a distributi on of chord lengths, G ( T ), by

PAGE 108

94 D rdl T G r T T r 10 (9) where the average value of T is simply the average of all chord lengths measured over the entire particle or particle agglomerate, which in this case would be constant due to particle sphericity. The correlation length, TN, is then the mean width of the correlation function. After integration of equation (9) (see reference 68 for complete derivation), the correlation length, TN can be calculated as D NT T dr r T0 2 02 (10) Where T is the average value of T The scattering experime nts used in this study measure the total scattered intensity of the nanocomposite sample, I ( h ), where h is related to q by a difference of 2 The scattered intensity is related to the interparticle or interagglomerate distance, TN, by68 N DT V dr r V hdh h I 2 42 0 0 0 2 1 (11) where V is the volume of the scatterers, or particles and is the difference in electron density. The results of the scattering experi ments to calculate the interparticle (interagglomerate) distance, TN, of the samples with varied particle dispersion are shown in the following section. Scattering Techniques In order to observe the wide range of length scales enc ountered in the differently dispersed composite samples, ra nging from nano to micron, th ree different approaches to characterization were employed. The SAXS technique is sensitive to intraparticle scattering to probe the shape of the primary pa rticles and the interparticle scattering of the

PAGE 109

95 individual primary particles and primary particle size distribution in the composite to measure an ideal particle dispersion and calculate FWHMWD, dWD and TWD. The USAXS technique is sensitive to the inter-agglomera te scattering to measure the size of the particle agglomerates and inte r-agglomerate distance due to the larger length scale than the primary particles for a measure of FWHMN, dN and TN. The static light scattering technique was also used to pr obe the size of the particle a gglomerates and was compared to the USAXS data as an alternative measur ement technique. However, the static light scattering q regime extends into length scales larger than that of the USAXS technique. Table 5-2 and Figure 5-11 provide a reference for the length sc ale probed by the different scattering techniques. The results of the sca ttering experiments contain data points that were averaged over at least three measurements. Table 5-2. Reference for the le ngth scale probed and the range of q values for each technique. q value () Length scale probed Sca ttering technique Scattering type 0.0001 6280 nm USAXS/SLS inter-agglomerate 0.001 628 nm USAXS/SLS inter-agglomerate 0.01 62.8 nm USAXS/SA XS interparticle 0.1 6.28 nm SAXS intraparticle

PAGE 110

96 Log q Log I(q) Light Scattering 3 m 90nm USAXS 2 m 75nm SAXS 1-100nm Log q Log I(q) Light Scattering 3 m 90nm USAXS 2 m 75nm SAXS 1-100nm Figure 5-11. Figure representing a typical scattering intensity plot and the various length scales probed by different scattering techniques. Small Angle X-Ray Scattering (SAXS) The SAXS technique was used to measur e the primary partic le distribution and primary particle average size. Any features that were larger than the 1-100 nm resolution limit were not observed and only the primary partic les were resolved. It is for this reason that any of the scattering intensity plots from each of the four nanocomposite samples, regardless of the degree of pa rticle dispersion, could have been used to measure the primary particle size distribution. Therefore, the SAXS plots were the same for all of the samples (Figure 5-12) because all of the composites contained the same primary particles. Further analysis of these plots using the Irena modeling program, as discussed in Appendix B, revealed that the averag e primary particle size was 27.5nm with a standard deviation of 7.3 nm for all sample s. The particle si ze distribution for the

PAGE 111

97 primary particles, used as the reference idea l dispersion, is shown to be narrow and has the smallest average size of all the nanocomposite samples due to full dispersion of the particles (Figure 5-16). 1.00E-01 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 0.01 0.1 q (1/)Intensity (a.u.) 0.5 wt% Undispersed 0.5 wt% Ultrasonciated 0.5 wt% Surface Treated Figure 5-12. Small angle x-ray scattering inte nsity plot for three differently dispersed samples. Due to the fact that all samples have the same primary particle size, all of the plots overlapped. Ultra Small Angle X-Ray Scattering (USAXS) The USAXS technique was used to probe f eatures on the micron length scale (~0.1-2 m resolution). The raw scattering data does not seem to present any significant changes between differently dispersed samples, but sm all changes in the features of the curve represent large changes in the average partic le agglomerate size (Figure 5-13). However, a closer observation of the region 0.001 < q < 0.01 in Figure 5-14 more clearly reveals the changes in the shape of th e curve due to changes in agglomerate size. This region

PAGE 112

98 (0.001 < q < 0.01) corresponds to a length scale of approximately 700 nm to 100 nm, which correlates well with the range of aggl omerate sizes shown in the four differently dispersed samples. The convergence of all but the fractionated sample USAXS curve at the higher q values ( q = 0.01) reveals that the undispersed ultrasonicated and surface treated samples do not contain small particles in the nano size regime, whereas the fractionated sample contains small aggregates at th e 70 nm length scale, which is probed at that q value (0.01). This is confirmed by the calculated agglomerat e size distribution in Figure 5-16, showing that the smallest aggl omerate size observed for those three samples is approximately 90 nm. 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09 0.00010.0010.01 q (1/)Intensity (a.u.) 0.5 wt% Undispersed 0.5 wt% Ultrasonciated 0.5 wt% Surface Treated 0.5 wt% Fractionated Figure 5-13. Ultra small angle x-ray scattering intensity plot showing the four differently dispersed samples. This q range covers the length scale of approximately 100 nm to 2000 nm.

PAGE 113

99 1.00E+03 1.00E+04 1.00E+05 1.00E+06 0.001 0.01 q (1/)Intensity (a.u.) 0.5 wt% Undispersed 0.5 wt% Ultrasonciated 0.5 wt% Surface Treated 0.5 wt% Fractionated Figure 5-14. The region 0.001 < q < 0.01, revealing the changes in the USAXS scattering curve due to changes in agglomerate size. Static Light Scattering (SLS) The raw scattering intensity plot including the static light scattering data (Figure 515) indicates that the fractionated sample scattered at the lowest intensity at the smaller q values ( q ~ 0.0001) of all the samples because ag glomerates were not present in the sample at that length scale. The undispersed ultrasonicated and surface treated scattered more intensely in that region, indi cating the presence of large agglomerates in the micron size range. The characteristic upturn in the data (as compared to the fractionated sample) indicates some agglomerates that may be on the order of the resolution of this technique (~3 m). Particle agglomerates of this size were observed through TEM micrographs in the undispersed sample, as shown previously.

PAGE 114

100 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07 0.0001 0.001 q (1/)Intensity (a. u.) 0.5 wt% Undispersed 0.5 wt% Ultrasonicated 0.5 wt% Surface Treated 0.5 wt% Fractionated Figure 5-15. Scattering intensity plot including the static light scattering data for the four differently dispersed samples. Scattering Data Analysis The scattering data from the SAXS, USAX S, and SLS techniques were combined, scaled to the SAXS absolute intensity measurements, a nd analyzed using the Irena macros for average particle (primary, aggregat e, or agglomerate) si ze, size distribution and interparticle distance. The particle si ze distribution data revealed that as the nanoparticles underwent dispersion from a hi ghly agglomerated state to an ideal randomly well dispersed state, the averag e particle size decreased and the size distribution narrowed (Figure 5-16 ). The particle size distri bution of the as-received dry powder was measured by commercial static light scattering and the remaining distributions were calculated from x-ray scatte ring data. The decreasing average particle

PAGE 115

101 size (Figure 5-17) with changing particle di spersion clearly indicates the quantitative difference between the four nanocomposite samp les with varying degree of dispersion. A similar particle size distribution graph is also shown for the static light scattering data analysis (Figure 5-18), which indicates si milar trends as Figure 5-16. However, the light scattering results produced a systematically larger average particle size as compared to the size derived from x-ray scattering (Figur e 5-18). This may be due to the fact that static light scattering (SLS) probes a smaller q value range than USAXS, thus resolving larger agglomerates that were in the na nocomposite sample. The USAXS measurements can probe agglomerates up to approximately 2 m, whereas the SLS measurements can observe particle agglomerates that are approximately 3 m. Regardless of the difference in results from USAXS and SLS measurem ents both techniques have shown a quantitative distinction between the samples of varied nanoparticle dispersion and the SAXS technique provided informa tion about the primary particles.

PAGE 116

102 0.00E+00 1.00E-03 2.00E-03 3.00E-03 4.00E-03 5.00E-03 6.00E-03 0.010.1110100 Particle Diameter ( m)Volume Distribution (cm^3/cm^3) Primary Particle Distribution 0.5 wt% Fractionated 0.5 wt% Surface Treated 0.5 wt% Ultrasonicated 0.5 wt% Undispersed Dry Powder Figure 5-16. Particle (primary aggregate, or agglomerate) size distribution as obtained from data analysis of x-ray scattering data (except dry powder obtained from SLS). As the dry powder agglomerates disperse in the polymer due to an increase in processing techniques, the size distribution shifts to the left and narrows.

PAGE 117

103 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 UndispersedUltrasonicatedSurface TreatedFractionatedIdeal CaseParticle diameter ( m) Figure 5-17. Average particle (primary, aggr egate, or agglomerate) size shown for the data obtained by USAXS measurements (e xcept primary particle size obtained from SAXS).

PAGE 118

104 0.00E+00 1.00E-03 2.00E-03 3.00E-03 4.00E-03 5.00E-03 6.00E-03 0.010.1110100 Particle Diameter ( m)Volume Distribution (cm^3/cm^3) Primary Particle Distribution 0.5 wt% Fractionated 0.5 wt% Surface Treated 0.5 wt% Ultrasonicated 0.5 wt% Undispersed Dry Powder Figure 5-18. Particle size distribution as obt ained from data analysis of static light scattering data, except primary par ticles (SAXS) and dry powder (DLS).

PAGE 119

105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 UndispersedUltrasonicatedSurface TreatedFractionatedIdeal CaseParticle diameter ( m) USAXS SLS Figure 5-19. A comparison between the averag e particle (aggregate or agglomerate) sizes obtained from x-ray measurements (USAXS) and static light scattering measurements (SLS). Dispersion Parameter Calculations Table 5-3 summarizes the results of the s cattering data analysis, including the calculation of scatteri ng parameters, and D1 and D2 calculations for the epoxy-TiO2 nanocomposite samples. The results s are ex pected to yield a quantitative comparison between the differently dispersed nanocomposite samples. The mixing parameter ( D2) and surface chemistry parameter ( D1) both indicate similar tre nds in dispersion because both parameters are greatly affected by th e size of the particle agglomerates, as mentioned previously. According to Table 5-3, the ch ange in processing from the undispersed to ultrasonicated sample is the addition of ultrason ic physical mixing, however, both the D1 and D2 values approximately doubled. This indicat es that ultrasonicati on had as much of an affect on the three dimensional spacing (ove rall mixing) of the particles in the matrix

PAGE 120

106 as it had on the size of the agglomerates (na no scale dispersion). The addition of the surface modification step to the surface treated sample resulted in a larger change in the D1 parameter, indicating that the particles dispersed by ultrasonication were stabilized sterically by the su rface modification and made compatible with the polymer matrix, thereby promoting greater size reduc tion than mixing alone. However, the D2 parameter did not change dramatically because the amount of physi cal mixing (ultrasonication) remained unchanged. Therefore, the only change in D2 for the surface treated sample was due to the reduction in agglomerate size only, and not a significant improvement in overall mixing. The difference between the surface treated and fractionated samples is quite large for both the D1 and D2 values. The D1 value more than doubled due to the addition of a solvent in the fractionated sample to reduce viscosity duri ng ultrasonication, but also the addition of a surface modification (ST2) that lowered the surface energy 10mJ/m2 more than that of the surface treated sample (ST1), according to given values by the manufacturer of the silane surface treatments. The lowered viscosity due to the solvent can be attributed to smaller dispersion e ddy currents as calculated by the Kolmogoroff equation, mentioned previously, thus prom oting even further reduction of particle agglomerate size, and overcoming the van der Waals forces to produce nearly primary particles. The change in the D2 value was affected by the cha nges in particle agglomerate size reduction and also by overall mixi ng improvement, again due to smaller eddy currents, because the same physical mixing by ultrasonication used in the other samples was enhanced by the addition of a solvent.

PAGE 121

107 Table 5-3. Values of T FWHM, and d and the results of the calculation of D1 and D2. Sample TN ( m) FWHMN or FWHMWD ( m) dN ( m) D1 D2 Undispersed 1.50 0.04 0.704 0.02 0.861 0.005 0.0243 0.0001 0.113 0.003 Ultrasonicated 0.775 0.12 0.404 0.005 0.507 0.003 0.0423 0.0002 0.218 0.035 Surface Treated 0.446 0.05 0.159 0.002 0.218 0.001 0.107 0.0006 0.380 0.046 Fractionated 0.233 0.05 0.0693 0.0004 0.0816 0.001 0.247 0.004 0.727 0.142 Ideal Case 0.169 ( TWD) 0.0171 0.0001 0.0275 ( dWD) 1 1 An attempt was made to separate the e ffects of surface modification and particle mixing by subtracting D1 from D2, to eliminate effects in D2 from particle agglomerate size changes (Table 5-4). It is ob served that the mi xing portion of the D2 parameter ( D2 D1) accounted for more than 50% of the D2 value for each sample, which indicates that investigating the interparticle (inter-aggl omerate) separation is important in the quantification of dispersion, but has often been neglected in the literature. The value of the mixing contribution to the D2 dispersion parameter increase d as particle/agglomerate size decreased, indicating that the additio nal processing methodologies enhanced the mixing homogeneity of the sample in addition to compatibilizing the particles with the surrounding polymer, as indicated in D1. Additionally, the separation of particles, aggregates, or agglomerates from one another as it relate s to dispersion becomes more important as the system approaches the id eal, homogeneously dispersed state. In conclusion, if both D1 and D2 dispersion parameters are used in tandem, the mechanism for particle dispersi on can be realized.

PAGE 122

108 Table 5-4. Dispersion Inde x values compared to the processing methodology where M = mixing, U = ultrasonication, ST = surface treatment, and S = solvent. Sample D1 D2 D2 D1 Processing methods Undispersed 0.0243 0.0001 0.113 0.003 0.0887 M Ultrasonicated 0.0423 0.0002 0.218 0.035 0.176 M, U Surface Treated 0.107 0.0006 0.380 0.046 0.273 M, U, ST1 Fractionated 0.247 0.004 0.727 0.142 0.480 M, U, S, ST2 Ideal Case 1 1 0 It appears from the above disc ussion that particle-particle interactions and particle compatibility with the surrounding polymer ma trix, both governed by surface chemistry, play a more dominant role in the dispersion of nanoparticles in a polymer because of the effect at all length scales of dispersion. Appropriate surfac e modification is necessary to enhance or stabilize dispersion by tailoring the correct chemistry at the particle-polymer interface. Conventional mixing techniques, such as high shear homogenizers, ultrasonication, and extrusion, have been shown to be su ccessful for incorporation of micron-scale particles into a polymer matrix by observi ng improvements in material properties. Chuang, et al.23, suggest that dispersion of glass mi croparticles was achieved with simple mixing into a liquid polyurethane resin for onl y two minutes, and that a surface treatment was needed solely to promote adhesion with the polymer matrix. However, conventional mixing techniques such as ultrasonica tion cannot produce eddy currents on the nano length scale for this system to completely overcome the van der Waals interparticle attractive forces purely by physical means.

PAGE 123

109 Relationship Between the Dispersion Parameters and Material Properties An attempt was made to correlate the quantitative dispersion parameters, D1 and D2, to a material property, such as optical de nsity. Optical density is a measure of the amount of light that is allowed to transmit through the polymer film sample, as compared to a background material. If a composite samp le has many particle agglomerates that are on the order of the size of visible light (400-700 nm), then high intensity multiple scattering will take place and the amount of transmitted light will decrease rapidly. On the other hand, if the particles (or agglomerates) in the sample are in the nano size range and small compared to visible light, then very minimal scattering will occur and the sample will remain relatively transparent. It is important to note that scattering can never be completely eliminated unless the refract ive indices of the na noparticles and polymer matrix material were exactly the same, t hus 100% transmission through a point in a composite material can never be theoretically achieved. Therefore, optical density is a relative measurement that compares a sample containing particles to that of the pure polymer without nanoparticles. The optical density measurements for the four differently dispersed nanocomposite samples (Figure 5-20) revealed a similar tre nd as the dispersion i ndices, as expected. Furthermore, the optical density meas urements correlate better with the D1 dispersion index component with a systematic linear difference between the two values. Although all of the dispersion index values correspond to the optical density the good correlation with the D1 value in particular may be due to the strong dependency of visible light scattering on the size of the partic le agglomerates. However, the D2 parameter also includes these affects, as discussed previously.

PAGE 124

110 The undispersed sample, with a mean agglomerate size of 860 nm and an interagglomerate distance of 1.5 m, has particle clusters that ar e much larger than the visible light wavelength but are also close enough that scattering be tween the agglomerates may take place. It is expected that this sample will have very low light transmission due to high intensity scattering away from the detect or due to large agglomerates and multiple scattering. The reduction in agglomerate size to 507 nm and interagglomerate distance to 775 nm for the ultrasonicated sample also indicate a low optical density, as the agglomerates are similar to the light wavele ngth. The scattering intensity will be lower due to smaller agglomerates than the undi spersed sample, thus the optical density increased between the two samples. Th e near 50% reduction in both the average agglomerate size and also inter-a gglomerate distance results in a two-fold increase in the optical density between samples. An increase in the optical density of 2.5x correlates well with the average agglomerate size reduction (~2.5x) for the surface treated sample, thus correlating with the D1 value. The change in interparticle distan ce is not as significant for this sample (less than two-fold reduction), which is indicated by less of a correlation of the D2 value with optical density. Finally, when cons idering the average agglomerate size and separation of the fractionated sample, again, the optical density change from the surface treated sample to the fractionated sample is best estimated by direct correlation between agglomerate size reduction and optical density. In other words, it can be suggested from these results that the following relation may be used to predict optical density when the change in dispersion index is known between samples 1~ D Itrans (12)

PAGE 125

111 where Itrans is the light transmitted through the sample, or optical density. This relationship may be utilized for ex trapolation to other material properties, such as tensile strength. A two-fold reduc tion in particle agglomerate size may be correlated with a two-fold increase in yield strength or ultimate tensile strength, if the same particle volume fraction is used. It was observed through SEM microscopy that the highly agglomerated samples co ntained voids around the part icle clusters, which would have a drastic affect on mechanical properties. Therefore, an optimal dispersion obtained by careful selection of surface modifica tion and mixing techniques would minimize agglomerates in the nanocomposite samp le, thereby leading to improved optical, mechanical, and other material properties. 0 10 20 30 40 50 60 70 80 90 100 Pure epoxyFractionatedSurface TreatedUltrasonicatedUndispersed% Transmission Figure 5-20. Optical density (amount of light transmitted through the sample) at 700 nm wavelength visible light fo r each differently dispersed sample (0.5 wt% TiO2 in epoxy) as compared to a pure epoxy background. The sample pathlength is 1 mm.

PAGE 126

112 0.01 0.1 1 UndispersedUltrasonicatedSurface TreatedFractionatedOptical Density or D Value Optical Density D1 Figure 5-21. A strong linear corre lation can be made between the D1 dispersion index component and optical density values (measured at 700 nm).

PAGE 127

113 CHAPTER 6 CONCLUSIONS AND FUTURE WORK Nanocomposite Fabrication The objective set forth in th is study was to quantitativel y characterize the state of dispersion of nanoparticles embedded in a polymer matrix. Several challenges became apparent when starting this i nvestigation. The first barr ier was to fabricate polymer composite samples with varied levels of nanoparticle disp ersion. An epoxy resin was chosen as the polymer matrix and titani a nanoparticles as the filler material. Nanocomposite samples with four discrete dispersion states were prepared and new methodologies were developed to create an optimal dispersion of a dry titania nanopowder in an epoxy matrix The particle agglomerat e size was reduced by over two orders of magnitude from the dry agglomerated powder to the fractionated nanocomposite sample. It was determined that the pr imary factors to improving (enhancing) the dispersion of a dry nanopowder in a polymer resi n is the choice of an appropriate solvent and also particle surface modi fication. The solvent was used reduce viscosity, thus reducing dispersion eddy currents, while th e surface modification lowered the surface energy to provide spontaneous spreading of the polymer resin on the particles. The particle agglomerates in the nanoco mposite samples spanned the length scale of large agglomerates in the micron size range down to near primary particle size. The qualitative differences between polymer na nocomposite samples were observed through optical micrographs and TEM/SEM microgr aphs. Optical micrographs provided information about the larger length scale of the sample, while SEM/TEM revealed the

PAGE 128

114 nano length scale of nanoparticle dispersion. The micrographs also provided preliminary data of the particle size distribution for use in the Irena modeling program. Suggested Future WorkFIB Microscopy has been used in the literature as the most common technique for qualitatively investigating particle dispersion, and advances have been made in the field of focused ion beam (microscopy) technology to image a bulk composite sample three dimensionally. The focused ion beam (FIB ) technique requires SEM microscopy and gallium ion beam etching to be used together First, the SEM beam scans the sample to create an image. Next, the ion beam etches away a designated thic kness of the sample to reveal a new surface. The SEM beam then images the new surface and the process continues. The series of SEM images are then stacked together to form a 3-D image. Although this process can be automated, the image processing involved to create the image can be tedious. This promising tec hnique can be employed to validate or develop more robust techniques for qualitative dispersion measur ements in nanocomposite materials. In this study, the FIB techni que was briefly explored to support the results of the SEM and TEM micrographs of the nanocomposite materials (Figure 7-1). However, the resolution was not adequate to i nvestigate individual features of the nanoparticles. It is recommended that this technique be used with either larger particles than those used in this study (>30nm diameter), or that a smaller volume of the sample be scanned with the SEM beam at higher magnification.

PAGE 129

115 Figure 6-1. Initial attempts at using the FIB technique to three dimensionally image the nanocomposite samples created in this study. Characterization Techniques The second barrier to charac terizing the state of partic le dispersion was to choose an appropriate technique or seri es of techniques to probe the relevant length scales in the nanocomposite system that spanned more than two orders of magnitude. Light and x-ray scattering techniques were employed to measure the scattering intensity behavior of each of the four differently dispersed nanocomposite samples. Small angle x-ray (SAXS) measurements revealed that the average primar y particle size of the titania particles was 27.5nm in diameter. The probe length of the SAXS instrument was small enough (1100nm) that only primary particle s were detected and measured.

PAGE 130

116 Ultra small angle x-ray measurements (U SAXS) provided the ne xt larger probe length in order to measure the average size a nd size distribution of the agglomerates for the composite samples. The calculation results of the USAXS measurements revealed the agglomerate average size ra nged from nearly 900nm for the undispersed state, down to 75 nm for the fractionated state. Light scattering measurements also provi ded a method to calc ulate the particle agglomerate average size and size distribution. However, the SLS provided a larger wavelength probe length ( q value) range than USAXS meas urements. Considering this fact, the average agglomerate sizes calculat ed from the light scattering measurements were approximately 20% larger than the aggl omerate size calculated from USAXS. This result is due to the difference in probe length ( q range) between the two measurement techniques. Static light scattering was able to resolve some large agglomerates in the samples that were not visible using th e USAXS technique, thus shifting the size distribution to higher values. Suggested Improvements Optical bench There are several improvements that can be made to the light scattering data measurement and data analysis techniques em ployed in this study. The optical bench built specifically for this study needs to be further calibrated to ensure more accurate measurements. In this study, calibration wa s carried out by using monodisperse particle standards of various sizes and measuring the light scattering pattern while dispersed in water in a cuvette. While this method provi ded a basis for selecting the appropriate optical components and aligning the apparatus by comparing the scattering behavior to the calculated Mie model, a more rigorous calibration procedure is required. Known

PAGE 131

117 bimodal and multimodel mixtures of particle standards should be used to test the accuracy of the scattering measurements. The theoretical scattering curve can be calculated using the same MiePlot program us ed in this study by calculating the fraction of particles of one size that is used in the distribution as it relates to the overall scattering pattern. Resultant multiple modeled scattering graphs can be combined into a single curve, representing all size fractions of the distribution. The mode of the distribution should be increased until a Gaussian or lognormal distribution can be created experimentally to match with the theoreti cal scattering model. At this time, only monodisperse standard particle s are used for calibration. Data analysis The raw scattering data analysis was acco mplished in this study by utilizing the Irena macros in the Igor Pro software. However, no standa rd operating procedures have been published describing the use of the Ir ena macros for acquiring information about particle size and interparticle distance. In fact, there are several procedures within the Irena macros that can be used to obtain diffe rent solutions for the same size distribution of particles. From the limited discussions between the author a nd others using these analysis techniques, no standa rd protocol for their use could be found. The somewhat subjective use of data trimming, smoothi ng and manipulation should be used with extreme caution. It is further recommended th at standard operating procedures for this type of data analysis be publis hed either in the literature or by creating a standard in the ISO or ASTM configuration. Additional characterizati on studieschemistry of the nanoparticle surface The silane coupling agent effectively preven ted reagglomeration of the particles in the solvent and lowered the titania surface energy to enhance wetting of the surface by

PAGE 132

118 the epoxy resin. However, verification is required by further ch aracterization. XPS spectroscopy and Bio-Raman spectroscopy may be used for characterization purposes. Some groups have characte rized the Ti-O-Si bond by Ra man Spectroscopy at 950cm-1 in order to identify the presen ce of the bonded silane groups to the particle surface.105 The advantage to using XPS is the small resolu tion and the ability to quantify specific bonds at the surface. Bio-Raman could provide a faster method to characterize the surface, however, the concentration of the coupling agent must be high enough for the resolution of the instrument (~100mmol). Another chal lenge to Bio-Raman measurements is the identification of the appropriate peaks that correspond to the bonds formed at the particle surface. Information on the peak location of the Ti-O-Si bond is not readily found in the literature. The Dispersion Index The last challenge to meeting the obj ective of this study was to develop a quantitative dispersion index th at indicated changes in part icle dispersion on both the macro and nano length scales The dispersion indices, D1 and D2, were proposed to quantify changes in particle su rface wetting and also mixture homogeneity, respectively. The values for each component can vary from zero, for a completely agglomerated system, to one for a completely random well-dispersed system. The results of the D1 and D2 calculations for the four differently disp ersed samples illustrated the quantitative differences between nanocomposite samples. The results also indicated a strong correlation of the processing techniques to th e dispersion indices and the affect of each. It was concluded that the D1 parameter is governed solely by the change in particle agglomerate size, however, the D2 parameter is affected by bo th the interparticle distance and also the agglomerate size due to the interrelation of these parameters. The

PAGE 133

119 quantitative results of the di spersion index verified the conclusions drawn from the qualitative dispersion data. Furthermore, if both D1 and D2 dispersion indices are used in tandem, the mechanism for particle dispersion can be realized by careful analysis of the D values for each sample. Correlating the Dispersion Index to Material Properties The development of the dispersion index development was enhanced by correlating the results to a measurable material property. The optical density of the nanocomposite was chosen as an important mate rial property due to applications in many industries, such as food and beverage packaging and window materials. The optical density of nanocomposite samples fabricated in this st udy revealed the same dispersion trends as shown from the dispersion i ndex calculations. A relations hip between the change in optical density and the change in the D1 value was proposed based on the strong correlation of optical density with changes in particle agglomerate size due to scattering of visible light. Suggested Future Work The results of this study have been pr esented at various industry and academic settings, and many have expressed an intere st in also correlating the mechanical properties (performance) of a composite material to the state of dispersion of a composite material. However, when this prospect was explored in this study, several problems were encountered. Most importantly, th e mixing rule equations used in the literature to predict the mechanical behavior of composite material s cannot be used when the filler particles have nano scale dimensions. An empirical relationship can be formed to relate the dispersion to material properties such as st rength and modulus, but the performance of an ideally dispersed material cannot be properly predicted with the equations in use

PAGE 134

120 currently. It is suggested th at this avenue be further pur sued as a separate study on the prediction of mechanical properties of na nocomposites and later correlated to the quantitative dispersion index sugge sted in this study. The re lationship between dielectric properties and the degree of par ticle dispersion was also explor ed briefly in this study and may also correlate well to the calculated value of D1. Choi, et al.89 have found that a relationship exists between particle size and the composite permeability.

PAGE 135

121 APPENDIX A CALCULATION OF THE PARTICLE SI ZE DISTRIBUTION BY SOLVING THE INVERSE MIE PROBLEM The inverse Mie problem can be solved to determine the particle size distribution of a sample from scattering data. However, in this study, modeling is used to analyze all scattering data (see Chapter three). The scattering inverse problem is generally written in the form93 N K S (1) where S is the scattering signature over the detectors, N is the unknown size distribution of bins vector so that Nj represents the quantity of particles in the j th size class, and K is the n x n scattering matrix so that element Kij indicates the amount of energy that a unit quantity of would scatter into the i th detector, assuming single scattering. The elements of K can be determined by Mie theory following the equations presented in chapter four or any approximation that would be appropria te to the system of particles and medium used for measurements (e.g., Rayleigh-Debye -Gans). Often the matrix is calculated using weighting functions for both the si ze classes and the detector geometries.93 The calculation of the particle size bins is given in a similar manner as other commercial laser diffraction instruments The distribution of the size ranges is logarithmic, which correlates well with the log-normal distribution of the primary particles used in this study, as well as the hypothesized log-no rmal distribution of aggregate sizes. The equation for the logarithmic size classes is given as

PAGE 136

122 1 min max min N nd d d d (2) where d is the width of the size class, dmin is the minimum particle size expected in the distribution, dmax is the largest expected particle (agglomerate) size, n is the bin number, and N is the number of elements in the intensity matrix given by S. The inversion of the matrix is given by S K N 1 (3) The values of the resultant N vector can be plotted with re spect to the size classes for a particle size distribution. Th ere are a number of ways in which this deceptively difficult calculation can be accomplished, including the Philips-Twomey inversion, which is cited most often in literature. However, this me thod is beyond the scope of this work and is presented as an alternative to using the Irena modeling programs, as described in Chapter three and Appendix B.

PAGE 137

123 APPENDIX B PROCEDURES FOR SCATTERING DATA ANALYSIS USING IRENA The standard procedure used for all data anal ysis in this study is presented here as a starting point for publication a nd standardization. These pr ocedures were obtained in part from the Irena user manual as provide d by Jan Ilavsky and partially from user experience. The steps to an alyzing the scattering data ar e presented in chronological order and with useful suggestions for starting parameters. The same data analysis can be completed with either SLS or USAXS data used interchangeably, keeping in mind that they will not produce the same solution. Desmearing of USAXS Data 1. Import raw USAXS scattering data into the Irena SAS macros in proper format (Q, R, S wave names) 2. Open the desmearing macro, import and gra ph the USAXS data, and input the slit length (0.1 will suffice for most measurements). 3. Trim high q value data at q = 0.0229 (-1) 4. Smooth fully before desmearing. 5. Use the power law with flat model for extrapolation. 6. Run the desmearing macro ten times, or unt il the error is small and the modeled data points do not change. 7. No further data smoothing was need ed. Export data to spreadsheet. Combine USAXS and SAXS Data 8. Import SAXS and USAXS data (in Q, R, S waves) to the data manipulation tool. 9. Combine the data sets by placing the cu rsors on the overlapping part of the two data curves.

PAGE 138

124 10. Scale the USAXS data as necessary and export new data set. Calculate Particle Size Distribution and FWHM Values 11. Import the combined data set into the sta ndard models Least Squares Fitting tool. 12. Fit a small part of the distribution for a PSD curve first, then move cursors out with each fitting until the region of approximately 0.000237 < q < 0.0027 is covered. 13. Fit only one size distribution, unle ss a bimodal distribution is known. 14. Use the suggested starting parameters and adjust as necessary Number of bins = 300 Neglect tail dist fractions = 0.05 Background = 0 (no fitting) Use initial guesses of mean and minimum size from microscopy data Correlation Length for Interparticle Distance 15. Use the fractals model macro and impor t the combined SAXS/USAXS data. 16. Use one mass fractal model only and set cursors over the same q range used for obtaining the PSD. 17. Use the suggested starting parameters and adjust as necessary Radius = 135 Dv = 2.5 Background = 0 (no fitting) Aspect ratio = 1 (spherical particles) 18. Set the number of integration points to 100 and gradually increase to 500 as the solution becomes more stable. 19. The solution is stable when the correlati on length does not change (< 1%) when the number of integration points is increased. 20. If the solution does not stabilize, change the volume filling from 0.4 to 0.6. Calculation of Primary Particle Distribution by SAXS 21. Use the Least Squares Fitting tool and import the SAXS data of any composite sample. 22. Fit the entire SAXS curve, using small q range at first and gradually expand to entire data range.

PAGE 139

125 23. Use the starting parameters similar to the USAXS data, adjusting as necessary. The values should be well known and solution should diverge quickl y to an expected solution. 24. It is not recommended to fit the standard deviation until a very stable solution can be found.

PAGE 140

126 APPENDIX C DERIVATION OF THE CALCULATION OF INTERPARTICLE DISTANCE FOR THE IDEAL DISPERSION CASE The log-normal distribution has been shown to accurately describe the particle size distribution of the particle system used in this study of nano-sized titanium dioxide (see Chapter three). Clearly, the agglomerate size distribution would also follow this type of distribution. A 0.00 10.00 20.00 30.00 40.00 50.00 0.0010.0100.1001.000 Size (microns)Differntial Number B Figure C-1. A typical positive-skewed lognor mal distribution curve is shown (A). The experimental particle si ze distribution of TiO2 particles disbursed in ethanol and measured with dynamic light scattering is also shown (B).

PAGE 141

127 The forming and dispersing of particle clusters can be treated kinetically by the following equation: M dt dM (1) where M is the size of the cluster and ( M ) can be expressed as: 0 0M M M M M M K M (2) where M0 and M are the minimum and maximum cluste r sizes formed respectively, and K is a velocity constant that can take a positive or negative value dependent on cluster growth or dispersion. The smallest value for M0 would be the primary particle size and the largest value of M constitutes an agglomerate made up of all the particles in the system. Since clustering/dispersing time has no defi ned starting point, the units of growth or separation of the particle clusters can be lengthened or shortened arbitrarily. This point justifies the assumption that the relative times for clustering or dispersing are normally distributed, which would be expressed by the following: 2 exp 2 12t t f (3) where f(t) is the time distribution function for clustering or disp ersing of particles. If equations (2) and (3) are combined, then M M M M M M b a t 0 0ln (4) where a and b are constants. Also, by combining equations (3) and (4)

PAGE 142

128 ln 2 ln exp ln 2 10 0M M M M M M M M f (5) where and M can be assumed to be the measure of polydispersity and mean particle size respectively, as proven elsewhere105. Under specific conditions, where the minimu m particle size is very small and the maximum particle size approaches infinity, equation (5) becomes the simple log-normal distribution as shown in many textbooks: 2ln 2 / ln exp ln 2 1 M M M f (6) which is equivalent to the probabi lity density of a particle size, di, with a geometric mean particle size, d and a geometric standard deviation (polydispersity), 2 2ln 2 ln ln exp ln 2 1 d d d fi (7) The mean size and polydispersity are thus defi ned over the entire composite system as the following summations: N i i N i i in d n d1 1ln ln (8) N i i N i i in d d n1 1 2ln ln ln (9) where = 1 for a monodisperse system.

PAGE 143

129 However, the D2 component of the dispersion index is dependent on the relative arrangement of particle agglomerates in the polymer matrix. Accordingly, the center-tocenter average interpar ticle distance between two particles, di and dj, can be calculated using the following equation26-27: j i k kd d T L 2 1 (10) where Tk is the average surface-to-surface interpar ticle distance, or chord length, of two adjacent hard spheres and Lk is the average center-to-center distance between particles. The hard-sphere assumption is used to prev ent the overlapping of particles and the complicated calculations that would ensue. E quation (10) can be adde d up over the entire system as N i N j j j i i N i k k N i k kd n d n T n L n11 1 12 1 (11) The average values for the center-to-cen ter and surface-to-surface interparticle distances, L and T are the arithmetic mean of all the values in the system. From equation (11) it follows that N k N k k k kn L L n11 (12) and N k N k k k kn T T n11 (13) Through an integration of the average of the log-normal particle size distribution by a simple transformation105, it is shown that the average pa rticle size, represented in this size distribution as th e summation of all d is shown to be

PAGE 144

130 N i i N i N i j j i in d d n d n1 11 2ln 2 1 ln exp (14) Also, it can be seen that N i i N i kn n1 1 (15) By substituting equations (12)-(15) into equation (11) we find that 2ln 2 1 expd T L (16) which gives the exact solution for the averag e interparticle spacing with relation to d the average (mean) particle size, and also the polydispersity for particles in any configuration that follows the log-normal distribution. If the particles are affixed to a given spatial distribution lattice, for example the simple c ubic lattice, then the interparticle distance may be written as follows: 2ln 2 1 exp d T LSC SC (17) where the subscript SC denotes parameters for particles occupying a simple cubic lattice. In the case of the same mean particle size, polydispersity, and volume fraction of particles in samples with the simple c ubic lattice and a randomly well dispersed configuration, the spatia l distribution parameter, can be defined as26 2 2ln 2 1 exp ln 2 1 exp d T d T L LSC SC (18) An increase in the value of would indicate that the partic le morphology is further away from that of the simple cubic lattice. The spatial distribution parameter has been calculated by Liu, et al26 to be approximately 1.1 for a randomly well dispersed

PAGE 145

131 morphology of particles. The th eoretically ideal, random disp ersion is defined as a threedimensional array of non-touching nanopa rticles in which all particles are homogeneously distributed in the sample vol ume and not confined to any geometrical lattice (Figure C-2). All of the polymer na nocomposite samples fabri cated in this study were compared to this ideal disp ersion state as a reference point. The average interparticle dist ance for particles that are confined to a simple cubic lattice can be exactly calculated by26 2 2 3 / 1ln 2 1 exp ln 2 3 exp 6 d TSC (19) If equation (19) is rearranged then 2 3 / 1 2ln 2 3 exp 6 ln 2 1 exp d d TSC (20) The combination of equation (18) and (20) yields the calculation of the average interparticle distance of the ideal dispersion st ate using the spatial di stribution factor for the randomly well dispersed morphology of particles 2 2 3 1ln 2 1 ln 2 3 exp 6 d T (21) where is the volume fraction of partic les in the nanocomposite material.

PAGE 146

132 Figure C-2. A two-dimensi onal schematic drawing of the randomly well-dispersed particle configuration.

PAGE 147

133 LIST OF REFERENCES 1. Ash, B.; Rogers, D.; Wiegand, C.; Schadler L.; Siegel, R.; Benicewicz, B.; Apple, T. Polymer Composites 23 (2002) 1014-1025. 2. Wetzel, B., Haupert, F., Zhang, M. Q. Composites Sci. and Tech. 63 (2003) 20552067 3. Lozano, T., Lafleur, P., Grmela, M. Canadian J. of Chem. Eng. 80 (2002) 11351142 4. Ganesan, L.; Bhattacharyya, P.; Bhowmich, A.; J. Appl. Polymer Sci 56 (1995) 1739-1747. 5. Shin, H.; Lee, B. J. Matls. Sci. 32 (1997) 4803-4806. 6. Chow, T. S. J. Phys.: Condens. Matter 15 (2003) L83-L87 7. Vaia, R.; Liu, W.; Koerner, H. J. Polym. Sci. B: Polym. Phys. 41 (2003) 32143236. 8. Krishnamoorti, R., Vaia, R. Polymer NanocompositeSynt hesis, Characterization and Modeling ACS Symposium Se ries 804. (2002). 9. Starr, F., Douglas, J., Glotzer, S. J. Chem. Phys 119 (2003) 1777-1788 10. Liang, J.; Li, R. J. Reinforced Plastics and Composites 20 (2001) 630-638. 11. Brinker, C. J., Scherer, G. W. Sol-Gel Science Academic Press, Inc., San Diego, CA, 1990. 12. Widegren, J., Bergstrom, L. J. Euro. Ceramic Soc. 20 (2000) 659-665. 13. Shutzberg, B.; Huang, C.; Ramesh, S.; Giannelis, E. Electrical Components and Technology Conference (2000) 1564-1567. 14. Vassileva, E.; Friedrich, K. J. Appl. Poly. Sci. 2003, 89, 3774. 15. Fischer, H. Matls. Sci. and Eng. C 23 (2003) 763-772. 16. Lozano, T.; Lafleur, P.; Grmela, M. Canadian J. Chem. Eng. 80 (2002) 1135-1142. 17. Yang, Y.; Dan, Y. Colloid and Poly. Sci. 284 (2005) 243-250.

PAGE 148

134 18. Zhu, J.; Peng, H. Adv. Func. Mater. 14 (2004) 643-648. 19. Schaefer, D.; Suryawanshi, C.; et al. Physica A 2004, 314, 686. 20. Martyanov, I. N.; Savinov, E. N.; Klabunde, K. J. J Colloid Int. Sci. 267 (2003) 111-117. 21. Zhao, J.; Schaefer, D.; et al J. Phys. Chem. B 109 (2005) 23351-23357. 22. Lee, C.; Lee, J.; Oh, S. Matls. Letters 57 (2003) 2643-2646. 23. Chuang, T.; Yang, T.; Chang, A. Int. J. Polym. Matls 52 (2004) 465-474. 24. Landau, L.; Lifshitz, E. Fluid Mechanics Pergamon, London, 1959. 25. Balek, V. J. Matls. Sci. Lett. 5 (1970) 714-716. 26. Liu, Z.; Li, R.; Tjong, S.; Choy, C.; Zhu, X.; Qi, Z. Wang, F.; Choy, C. Polymer 39 (1997) 4433-4436. 27. Liu, Z.; Li, R.; Tjong, S.; Choy, C.; Zhu, X.; Qi, Z. Wang, F.; Choy, C. Polymer 40 (1999) 2903-2915. 28. Premphet, K.; Paecharoenchai, W. J. Appl. Poly. Sci. 82 (2001) 2140-2149. 29. Gurland, J. In: Dehoff, R.T., Rhines, F. N., editors. Quantitative microscopy New York: McGraw-Hill, 1968. 30. Tanaka, H.; Hayashi, T.; Nishi, T. J. Appl. Phys. 59 (1986) 653. 31. Tanaka, H.; Hayashi, T.; Nishi, T. J. Appl. Phys. 59 (1986) 3627. 32. Tanaka, H.; Hayashi, T.; Nishi, T. J. Appl. Phys. 59 (1986) 4480. 33. Suetsugo, Y. Intern. Poly. Proc. 3 (1990) 184. 34. Ptschke, P.; Bhattacharyya, A.; Janke, A.; Euro. Poly. J. 40 (2004) 137-148. 35. Park, B.; Balatinecz, J. Intern. J. Polymeric Mater. 37 (1997) 133-150. 36. Becu, L.; Taha, M.; Maazouz, A.; Merle, G. J. Matls. Sci. 37 (2002) 41-49. 37. Thompson, C.; Herring, H.; Gates, T.; Connell, J. Composites Sci. and Technol. 63 (2001) 1591-1598. 38. Gersappe, D. Phys. Rev. Letters 89 (2002) 1-4. 39. Tishkovets, V.; Petrova, E.; Jockers, K. J. Quant. Spec. and Rad. Trans. 86 (2004) 241-265.

PAGE 149

135 40. Shang, Z.; Lu, C.; Gao, L. Polymer International 55 (2006) 1277-1282. 41. Vandervorst, P.; Lei, C.; Lin, T. et al. Progress in Organic Coatings 57 (2006) 9197. 42. Smith, J.; Bedrov, D.; Smith, G. Comp. Sci. and Tech. 63 (2003) 1599-1605. 43. Starr, F.; Douglas, J.; Glotzer, S. J. Chem. Phys. 119 (2003) 1777-1788. 44. Vad, T.; Haubold, H.; Waldfner, N.; Bnnemann. Appl. Cryst. 35 (2002) 459-470. 45. Brechet, Y. J. M. Matls. Sci. and Eng. A175 (1994) 63-69. 46. Hooper, J.; Schweizer, K.; et al. J. Chem. Phys. 121 (2004) 6986-6997. 47. Wedd, M. Education Resources For Particle Technology 032Q-Wedd (2003). 48. Gustav Mie, Beitrage zur Optik tr uber Medien, speziell Kolloidaler Metallosungen, A nnalen der Physik, 25 (1908) 377. 49. Lilienfeld, P. Applied Optics 30 (1991) 4696-4698. 50. Xu, R. Particle Characterization: Light Scattering Methods Springer, New York, 2002. 51. McNeil, L., Hanuska, A., French, R. Applied Optics 40 (2001) 3726-3736. 52. Schnaiter, M., Wurm, G. Applied Optics 41 (2002) 1175-1180. 53. Wiggins, S.; Robb, G., et al. Meas. Sci. Technol. 13 (2002) 263-269. 54. McNeil, L., French, R. Acta Materiala 48 (2000) 4571-4576. 55. Thiele, E., French, R. Journal of the American Ceramic Society 81 (1998) 469-479. 56. Gao, J.; Wu, C. J Appl. Poly. Sci. 77 (2000) 1165-1168. 57. Stone, S., Bushell, et al. Meas. Sci. Technol. 13 (2002) 357-364. 58. Hirelman, E. Applied Optics 30 (1991) 4832-4838. 59. Weiner, I., Rust, M., Donnelly, T. American Journal of Physics 69 (2001) 129-136. 60. Ofoli, R.; Prieve, D.; Langmuir 13 (1997) 4837-4842. 61. Bonczyk, P.; Hall, R. Langmuir 7 (1991) 1274-1280. 62. Xing, Y., Koylu, U., Rosner, D. Applied Optics 38 (1999) 2686-2697.

PAGE 150

136 63. Thill, A., Desert, S., Delsanti, M. Eur. Phys. J. Appl. Phys. 17 (2002) 201-208. 64. Romer, S.; Urban, C.; et al J. Appl. Cryst. 36 (2003) 1-6. 65. Buehler, C., Caruthers, J., Franses, E. J. Chem. Phys. 98 (1993) 3600-3611. 66. Hyeon-Lee, J., Beaucage, G., Pratsinis, S., Vemury, S. Langmuir 14 (1998) 57515756. 67. Beaucage, G. Journal of Applied Crystallography28 (1995) 717-728. 68. Glatter, O., Kratky, O. Small Angle X-Ray Scattering Academic, New York, 1982. 69. Chu, B., Hsiao, B. Chemical Review 101 (2001) 1727-1761 70. Brunner-Popela, J.; Glatter, O. A. Appl. Cryst. 30 (1997) 431-442. 71. Suryawanshi, C., Pakdel, P., Schaefer, D. J. Appl. Cryst. 36 (2003) 573-577. 72. Koga, T., Takenaka, M., et al. Langmuir 21 (2005) 11409-11413. 73. Chu, B., Liu, T. Journal of Nanopar ticle Research 2 (2000) 29-41. 74. Weyerich, B.; Bruner-Popella, J.; Glatter, O. J. Appl. Cryst. 32 (1999) 197-209. 75. Sorensen, C. Aerosol Science and Technology 35 (2001) 648-687. 76. Skillas, G., Agashe, N., et al. Journal of Applied Physics 91 (2002) 6120-6124. 77. Chaker, J., Dahmouche, K., Santilli, C., et al. J. Appl. Cryst. 36 (2003) 689-693. 78. Dobbins, R., Megaridis, C. Applied Optics 30 (1991) 4747-4754. 79. Beaucage, G. J. Appl. Cryst. 29 (1996) 134-146. 80. Hasmy, A., Vacher, R., Jullien, R. Physical Review B 50 (1994) 1305-1308. 81. Koberstein, J., Morra, B., Stein, R. J. Appl. Cryst. 13 (1980) 34-45. 82. Schelten, J., Hendricks, R. J. Appl. Cryst. 11 (1978) 297-324. 83. Berriot, J., Montes, H., et al. Polymer 44 (2003) 4909-4919. 84. Causin, V., Marega, C., et al. Polymer 46 (2005) 9533-9537. 85. Beaucage, G., Kammler, H., Pratsinis, S.; J. Appl. Cryst. 7 (1994) 523-535. 86. ASTM. D2663-95a, Standard Test Met hods for Carbon BlackDispersion in Rubber. Annual Book of ASTM Standards, v.9.01.

PAGE 151

137 87. Premphet, K.; Paecharoenchai, W. J. Appl. Poly. Sci. 82 (2001) 2140-2149. 88. Schaefer, D.; Vu, B.; Mark, J. Rubber Chem. and Technol. 75 (2002) 795-810. 89. Choi, H.; Moon, K., et al. J. Appl. Polym. Sci. 66 (1997) 477-482. 90. Miyagawa, H.; Rich, M. J.; Drzal, L.T. J. Poly. Sci. B: Poly. Phys. 2004, 42, 4384. 91. Sperling, L. Introduction to Physical Polymer Science John Wiley & Sons, New York, 2001. 92. Mezzenga, R.; Boogh, L.; Manson, J. J. Polym. Sci. B: Polym. Phys. 38 (2000) 1883-1892. 93. Hirleman, E. Applied Optics 30 (1991) 4832-4838. 94. Chen, K., Kromin, A., et al. Optics Communicatio ns 228 (2003) 1-7. 95. Lake, J. Acta Cryst. 23 (1967) 191-194. 96. Bohren, C; Huffman, D. Adsorption and Scattering of Light by Small Particles John Wiley & Sons, New York, 1983. 97. van de Hulst, H. Light Scattering by Small Particles Wiley &Sons, New York, 1957. 98. Ludlow, I.; Everitt, J. Phys. Rev. E 53 (1996) 2909-2924. 99. French, R. J. Am. Ceram. Soc. 83 (2000) 2117-2146. 100. Myers, D. Surfaces, Interfaces, and Colloid s: Principles and Applications John Wiley & Sons, New York, 1999. 101. Page, S.; Berg, J.; Manson, J. J. Adhesion Sci. Technol. 15 (2001) 153-170. 102. Navrotsky, A. J. Chem. Thermodynamics 39 (2007) 2-9. 103. Zenkiewicsz, M. Polymer Testing 26 (2007) 14-19. 104. Jia, N.; Martin, R.; Qi, Z. ; Lefebvre, M.; Pickup, P. Electrochimica Acta 46 (2001) 2863-2869. 105. Yakolev, V.; Scarel, G.; Aita, C.; Mochizuki, S. Appl. Phys. Letters 76 (2000) 1107-1109. 106. Irani, R.; Callis, C. Particle Size: Measurement, Interpretation and Application, John Wiley, New York, 1963.

PAGE 152

138 BIOGRAPHICAL SKETCH Marie was born in Michigan on January, 23, 1980 and moved to Orlando, Florida in 1990, where she attended Dr. Phillips High School. After receivi ng her diploma with a concentration in Internationa l Studies, she decided to pursue chemical engineering and music at the University of Florida. After receiving her Bachelors degree in Chemical Engineering and a Minor in Clarinet Perf ormance in December of 2002, she attended graduate school at the Univer sity of Florida in the Materials Science and Engineering department. She received her Doctor of Philosophy degree in May of 2007 and plans to pursue a career in polymer composites. Marie currently resides in Gainesville with her loving husband, Kevin, and her dogs, Bailey and Killian.