ROLE OF CATALYST SURFACE ENERGY AND SUBSTRATE SUPPORT IN THE GROWTH OF TEMPLATE CONFINED ARRAYS OF CARBON NANOTUBES
By
GREGORY CHESTER
A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE
UNIVERSITY OF FLORIDA
2018
© 2018 Gregory Chester
To my wife who supported and encouraged me throughout this process
4
ACKNOWLEDGMENTS
No project is the work of a single person, especially when involved with the tight
periods and busy schedules of combining work and school. Without the help of people
at Mainstream, at the Navy, and at University of Florida this would not be possible.
Special thanks go to Dr. Ignacio Perez and the Office of Naval Research for their
funding of contract N0014-16-C-2026 which has provided the basis for the work
performed herein. I would also like to thank Mainstream Engineering and specifically my
manager, the principal investigator for this project, Justin Hill, for his support and
allowing me to pursue my degree while working on this project.
As with any project in industry, there have been many hands involved in the
setup and operation of the equipment and experiments. Daniel Ettehadieh and Daniel
Murphy were critical in setting up the test stand and control system to the high
functioning equipment it now is. A very special thanks goes to Kayla O’Neil for her
involvement in the daily experiments on the project. Gibson Scisco at University of
Florida helped with TEM imaging.
Finally, I would also like to thank my advisor, Dr. Juan Nino, for his support of this
work and providing the opportunity to pursue my degree under his tutelage.
5
TABLE OF CONTENTS
page
ACKNOWLEDGMENTS .................................................................................................. 4
LIST OF TABLES ............................................................................................................ 8
LIST OF FIGURES .......................................................................................................... 9
LIST OF ABBREVIATIONS ........................................................................................... 13
ABSTRACT ................................................................................................................... 14
CHAPTER
1 INTRODUCTION .................................................................................................... 17
Statement of Problem and Motivation ..................................................................... 17
Scientific Approach ................................................................................................. 18
Contributions to the Field ........................................................................................ 20
2 CARBON NANOTUBE BACKGROUND ................................................................. 21
Carbon Nanotube Structure .................................................................................... 21
Carbon Nanotube Properties .................................................................................. 23
Carbon Nanotube Applications ............................................................................... 25
Caron Nanotube Composites ........................................................................... 25
Polymer-CNT composites .......................................................................... 25
Metal-CNT composites .............................................................................. 26
6
Composite drawbacks ................................................................................ 27
CNT Ropes ....................................................................................................... 27
Conclusions ............................................................................................................ 28
3 CHEMICAL VAPOR DEPOSITION BASED CARBON NANOTUBE SYNTHESIS . 30
Carbon Nanotube Origins ....................................................................................... 30
Selection and Role of Gaseous Precursors ............................................................ 31
Selection and Role of Catalyst Support Substrate .................................................. 34
Role of Catalyst Selection ....................................................................................... 35
Catalyst Particle Size Effects ............................................................................ 35
Catalyst Composition Effects ............................................................................ 36
Adsorption and Surface Diffusion vs. Absorption and Bulk Diffusion ................ 38
Conclusions ............................................................................................................ 41
4 ULTRA-LONG CARBON NANOTUBE SYNTHESIS .............................................. 43
Failures in Ultra-Long Carbon Nanotube Synthesis ................................................ 43
Segregated-Flow Chemical Vapor Deposition Furnace Design .............................. 44
Investigation of Catalyst Composition ..................................................................... 47
5 TEST STAND DESIGN AND ASSEMBLY .............................................................. 52
System Level Design .............................................................................................. 52
Segregated-Flow Chemical Vapor Deposition Reactor ........................................... 53
High-Temperature Contact Angle Measurement System ....................................... 53
Gas Flow Management .................................................................................... 55
7
Imaging Management ....................................................................................... 56
Controller Design .............................................................................................. 58
High-Temperature Contact Angle Measurement Stand Calibration ........................ 58
Contact Angle Measurements ................................................................................. 60
6 ROLE OF CARBON AND HYDROGEN IN THE GAS STREAM ............................ 66
Contact Angle Measurements ................................................................................. 66
Iron Wetting Properties ..................................................................................... 66
Effects of Carbon and Hydrogen Exposure ...................................................... 71
Effect of Surface Pretreatment ......................................................................... 77
Conclusions ............................................................................................................ 80
7 GROWTH OF ULTRALONG CARBON NANOTUBES ........................................... 81
Ultralong Carbon Nanotube via Segregated-Flow Chemical Vapor Deposition ...... 81
Conclusions ............................................................................................................ 95
8 SUMMARY AND RECOMMENDATIONS ............................................................... 96
Summary ................................................................................................................ 96
Future Work ............................................................................................................ 97
REFERENCE LIST........................................................................................................ 99
BIOGRAPHICAL SKETCH .......................................................................................... 105
8
LIST OF TABLES
Table page
4-1 Physical and catalytic properties of metallic catalysts ........................................ 49
5-1 Calibration data for four measurements of the four calibration images ............... 64
6-1 Contact angles of reference metals on varied substrates ................................... 68
9
LIST OF FIGURES
Figure page
2-1 Arrangement of graphene into CNTs. ................................................................. 22
2-2 SEM image of CNT fiber being wound by pulling process .................................. 28
3-1 Raman spectra of SWNTs grown at low temperature using methanol as a
precursor ............................................................................................................ 32
3-2 Comparison of CNTs grown from different carbon sources ................................ 32
3-3 Comparison of CNTs gron from different carbon concentrations ........................ 34
3-4 Difference in Fe film morphology ........................................................................ 35
3-5 Size of catalyst nanoparticles and the resulting CNT diameters from varied
precursors........................................................................................................... 36
3-6 Images of CNTs grown from varied catalyst metals and the resulting CNT
diameter and growth rate .................................................................................... 37
3-7 Change in melting point of catalytic materials with decreasing nanoparticle
diameter ............................................................................................................. 39
3-8 Crystalline Fe-C catalyst particle ejecting a CNT tip ........................................... 40
3-9 CNT nucleation model ........................................................................................ 41
4-1 Segregated flow CVD furnace diagram .............................................................. 45
10
4-2 Proposed mechanism of ULCNT growth in the SF-CVD reactor ........................ 47
4-3 CNTs grow through a bulk catalyst layer in one of three potential modes. ......... 50
5-1 Process and instrumentation diagram of test-stand ............................................ 52
5-2 Segregating plate cross section.......................................................................... 53
5-3 Diagram of high-temperature contact angle measurement furnace .................... 55
5-4 Mass flow controller containment box ................................................................. 56
5-5 Camera set-up for imaging molten metal contact angle ..................................... 57
5-6 Metal melting process observed in contact angle measurement system ............ 59
5-7 Melting high purity metals in the high temperature contact angle
measurement furnace ......................................................................................... 60
5-8 Focused image of two molten iron droplets on sx-Al2O3 ..................................... 62
5-9 Contact angle of Fe on Al2O3 with respect to droplet mass ................................ 63
5-10 Calibration images used to determine the approximate user error ..................... 64
5-11 Change in contact angle with respect to temperature ......................................... 65
6-1 Contact angle of each molten metal ................................................................... 67
6-2 Zisman plots of non-polar and polar materials .................................................... 69
6-3 Attempted Zisman plot for Al2O3 and pyrolytic graphite ...................................... 70
11
6-4 Effect of ethylene exposure time on Fe sample mass ........................................ 72
6-5 Fe-C phase diagram ........................................................................................... 74
6-6 Effect of ethylene exposure time on melting point of Fe ..................................... 75
6-7 Carbon in Fe weight % vs melting temperature .................................................. 76
6-8 Effect of ethylene exposure time on contact angle of molten Fe on Al2O3 .......... 77
6-9 Effect of pretreating Al2O3 substrate with C ........................................................ 79
6-10 Observed Fe melting on varied substrates ......................................................... 80
7-1 AAO template growth side before catalyst coating ............................................. 83
7-2 AAO template with aluminum annular support after initial test run ..................... 83
7-3 Patchy CNT growth from initial SF-CVD runs showing ....................................... 84
7-4 AAO template after CNT growth in the SF-CVD reactor ..................................... 84
7-5 Raman spectra of CNTs grown in clean stream of SF-CVD reactor ................... 86
7-6 CNT growth from AAO template in clean stream of SF-CVD reactor ................. 87
7-7 TEM image of CNTs harvested from growth side of SF-CVD furnace ................ 88
7-8 Hi-res SEM of CNT nanofiber tip showing individual nanotubes......................... 89
7-9 Base of CNT fiber showing individual CNTs bunching together upon ejection
from the AAO surface ......................................................................................... 90
12
7-10 Tall cluster of ULCNTs lifting away from AAO template ..................................... 91
7-11 EDS spectrum showing no Fe evidence on CNT growth side ............................ 92
7-12 EDS spectrum showing Fe peaks identified on ULCNT cluster .......................... 93
7-13 SF-CVD growth with carbon pretreated template ............................................... 94
13
LIST OF ABBREVIATIONS
AAO Anodized aluminum oxide
AFM Atomic force microscope
CNT Carbon nanotube
CVD Chemical vapor deposition
HT-CAM High-temperature contact angle measurement
MD Molecular dynamics
MFC Mass flow controller
MWNT Multiwall carbon nanotube
SEM Scanning electron microscope
SF-CVD Segregated flow chemical vapor deposition
SWNT Single wall carbon nanotube
TEM Tunneling electron microscope
ULCNT Ultra-long carbon nanotube
XPS X-ray photoelectron spectroscopy
14
Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science
ROLE OF CATALYST SURFACE ENERGY AND SUBSTRATE SUPPORT IN THE GROWTH OF TEMPLATE CONFINED ARRAYS OF CARBON NANOTUBES
By
Gregory Chester
May 2018
ABSTRACT Chair: Juan Nino Major: Material Science and Engineering
Carbon nanotube (CNT) based composite materials often struggle to reach
expected performance due to failures at the CNT/matrix interface. In an effort to reduce
the number of interfaces in a CNT-based composite, a novel method for the growth of
ultra-long CNTs (ULCNTs) has been developed. The segregated-flow chemical vapor
deposition (SF-CVD) reactor provides a method to eliminate many of the current failure
mechanisms in CVD based ULCNT growth and has opened the door to improved
understanding of the bulk diffusion CNT growth mode. Achieving ULCNT growth using
the SF-CVD reactor, though requires understanding and ultimately controlling the
interfacial energies between the gaseous precursor, catalyst material, and the catalyst
nanoporous membrane support which is used as a template for growth.
In order to better understand the interface between a molten catalyst droplet and
the porous anodized aluminum oxide (AAO) support, a custom high-temperature
contact angle measurement system (HT-CAM) was assembled which allows the user to
measure the contact angle of a molten metal droplet on a given substrate and in the
15
desired gaseous environment in order to calculate surface energy. Contact angle
measurements, a representation of surface energy, of Fe on planar Al2O3 agreed with
literature values, showing a high surface energy and an obtuse contact angle. Varying
the carbon content in the catalyst and gaseous environment altered the contact angle
indicating the contact angles most likely to be seen in the nanoporous template. Varying
the surface of the alumina substrate by coating with carbon altered the wetting
properties of the iron, driving a low surface energy or wetting interaction. This allowed
for the comparison of the effects of varied surface energy on ULCNT growth in the SF-
CVD reactor. The results indicate that Fe has high surface energy on bare alumina and
should therefore create an convex meniscus in the nanopores which is favorable for
CNT initiation. It alternately wets alumina coated with carbon, which will induce a
concave meniscus thereby allowing for the comparison of varied meniscus shape on
ULCNT growth.
Using the results of the contact angle measurements and surface energy
calculations, the SF-CVD reactor was successful in demonstrating CNT growth up to 1
mm long. This does not reach the ultimate goal of more than 10 cm; however, CNT
growth was not limited by traditional failure mechanisms and as such this has shown the
potential for this technology to achieve ULCNT growth. The role of catalyst curvature
was verified by demonstrating very little CNT growth when pretreating the alumina
template with carbon which caused iron to wet the anodized alumina (AAO) support
leading to an unfavorable concave meniscus within the nanopore in comparison to an
bare AAO template which showed more, longer CNTs.
16
For the first time, CNT growth has been conclusively shown as possible due
purely to bulk diffusion of C atoms through the catalyst. This discovery is not possible in
traditional catalyzed CVD growth as it has not been possible to dissociate bulk diffusion
from surface diffusion in a catalyst nanoparticle.
17
CHAPTER 1INTRODUCTION
Statement of Problem and Motivation
Carbon nanotubes (CNTs) have been touted as have been touted as having
nearly “limitless potential” since they were initially discovered by IIjima et. al. in 1991.1
While studying fullerenes (C60) produced via an arc-discharge method, Iijima discovered
microtubules of graphitic carbon assembled like a roll of graphite. Initial speculation
postulated that due to their size and structure, CNTs could present with unique
properties. Since that time, the tubular, 1-dimensional arrangement of sp3 hybridized
carbons were discovered to have extremely high electrical conductivity, thermal
conductivity, and modulus in axial direction. The impressive properties led to CNTs
being heavily researched for their potential as structural additives2, conductivity
enhancers3, molecular sieves4, transistor components5, and even for drug delivery6.
Harnessing CNTs impressive properties; however, is difficult for a number of
reasons. Synthesizing perfect, single wall CNTs (SWNTs) has low yield and is
expensive in comparison to synthesizing multiwall CNTs (MWNTs) which tend to have
reduced mechanical, thermal, and electrical properties. Integrating CNTs into composite
materials can improve the composite properties over the bulk, but the benefits are
limited by the CNT/matrix interface as there is an interfacial resistance between the
matrix and CNT. As CNTs tend to be a few nm to µm in length, there are an extremely
high number of interfaces in an average composite solid. By increasing CNT length,
there should be a comparable increase in the composite material properties.
18
Scientific Approach
As will be described in later sections, the majority of previous investigations into
the growth of ultra-long CNTs (ULCNTs) have been based on controlling catalyst film
deposition to induce very consistent nanoparticle formation7, maintaining extremely
laminar precursor gas flow to prevent turbulence8, and/or utilizing a mobile catalyst
rolling along a substrate9. These methods all face the same difficulties of limited yield,
high cost, and exponential increases in difficulty to increase the length beyond current
limits. Specifically, with carpeted chemical vapor deposition (CVD) there are growth
limitations with merging catalyst particles10, catalyst poisoning11, and blocking of
catalytic sites with amorphous carbon12. To address this, a custom segregated-flow
chemical vapor deposition (SF-CVD) tube furnace was designed which features two
chambers, each totally isolated from the other except at a single point which is
separated by a thin anodized aluminum oxide (AAO) membrane. The membrane
features nanometer scale pores, open on both ends, where one side is coated with a
catalyst to inhibit gas flow between the chambers. The top chamber contains the
catalyst film and a carbon-based precursor gases while the bottom chamber contains
only carrier and reducing gases. Carbon can move between the chambers only by
dissociating into the catalyst, diffusing through the bulk catalyst film, accumulating at the
inner surface, then forming a CNT within the nanopore which grows into the lower
chamber. The SF-CVD reactor is the first of its kind which conclusively demonstrates
the bulk diffusion CNT growth method without the influence of surface adsorption and
diffusion. Many works have argued the role of surface adsorption vs bulk absorption to
19
characterize CNT growth; while this work does not preclude the surface adsorption
mechanism, it does demonstrate the efficacy of bulk absorption and diffusion to create
CNTs. The reactor system also allows for the investigation of the role of individual gas
components on the growth of CNTs. The role of H2 is often debated in the growth of
CNTs; though not explored in this work, the segregation of hydrocarbon precursor from
the actual growth will allow for exploration of the effects on gas composition on CNT
structure and properties.
Optimizing the conditions at which the reactor should operate requires fully
understanding and controlling the catalyst/AAO interface to create a meniscus favorable
for CNT growth. To predict this interface, a second system was set up which is capable
of measuring the contact angle between a molten metal catalyst particle and a substrate
at temperatures up to 1700 °C. Contact angle measurements were taken in order to
determine if standard wetting conditions and if there is a method in which the interaction
between the catalyst and Al2O3 substrate could be varied. The simplest methods to alter
the interfacial energy between the catalyst and substrate with minimal change to the
system was to explore the role of carbon infiltration on the wetting properties, carbon
pretreating the alumina template, and varying the surrounding gas composition. After
determining the effects of varying these conditions on the contact angle, their effect on
the ultimate growth properties of ULCNTs was determined by growing CNTs using the
SF-CVD reactor. Growth in the reactor did not reach beyond 1 mm; however, the CNTs
grown using the SF-CVD reactor did not fail due to amorphous carbon buildup on the
20
CNTs themselves indicating the capability of the SF-CVD reactor to reduce at least one
of the failure mechanisms seen in CNT growth.
Contributions to the Field
This work focused on the development of a novel reactor system to drive growth
of ULCNTs through characterization and optimization of the system by understanding
and tuning the interfacial energy between the substrate and catalyst. The main
contributions to the field of materials science are:
1. Measurement of the effect of carbon absorption, gas composition, and substrate
chemistry on the interaction between an iron catalyst and an alumina substrate
2. A first of its kind demonstration of CNT growth through an SF-CVD reactor
determination of the effects of varying the surface chemistry of the AAO template
on CNT growth and demonstration of CNT growth in which at least two of the
failure mechanisms seen in CVD CNT growth (CNT quantity and amorphous
carbon buildup on the CNTs) are eliminated
3. A confirmation of the bulk diffusion mode of catalyzed CVD CNT growth by
segregating the dissociative surface of the catalyst film from the curved surfaces
within the nanopores where CNT synthesis occurs allowing for complete
elimination of the surface diffusion mode in CNT growth
21
CHAPTER 2CARBON NANOTUBE BACKGROUND
Carbon Nanotube Structure
At their simplest, CNTs are a simple network of sp3 bonded carbon atoms
structured as a single sheet of graphite (graphene) rolled into a tube. Predictive electron
models, such as those developed by Hamada et. al., have indicated that the impressive
properties of CNTs stem from their high electron mobility and strong covalently bonded
atomic network.13 This model, along with others14, indicated that CNTs could have a
variety of properties dependent upon the tube diameter15 and the arrangement of the
carbon atoms with respect to the central axis, known as the chirality. The chiral indices
are presented as a pair of coordinates (n,m) which relate to the unit vectors in wrapping
a hexagonally structured graphene sheet to form a tubule. Changing the chiral index
gives rise to the different chiral structures such as the standard “armchair” and “zig-zag”
conformations of 30º (n = 0) and 0º (n = m) respectively. All others are referred to as
“chiral” with their respective angle or coordinate pair.16 The coordinate pair further has a
direct effect on the diameter of the CNT according to Equation 2-1
d= a�m2+mn+n2
π (2-1)
where a is the lattice constant of a graphite sheet (1.42 x √3 Å).17 The three major CNT
chiralities and their respective graphene roll vectors are shown in Figure 2-1.
22
Figure 2-1: Arrangement of graphene into CNTs. (a) Graphene sheet showing the rolling vector for (b) zig zag, (c) armchair, (d) other n,m chiral CNTs.18
After Iijima’s discovery of the CNT, the electronic models developed assumed
that the tube was formed by a single sheet of graphitic carbon. However, as scanning
electron microscopy (SEM) and tunneling electron microscopy (TEM) technology
improved, researchers discovered that early CNTs, such as those Iijima obtained
through arc discharge synthesis, were assembled from concentric layers of individual
CNTs. Measurements of individual MWNTs showed they could not obtain the predicted
physical properties due to slippage between layers and interactions between the walls
which disturb the sp2 bonding structure and interrupt the conductive pathways. First
23
observed by Iijima in 1993, the SWNT has the ultra-high strength and conductivity
projected by electronic structure models.
Carbon Nanotube Properties
Depending on the chiral index of the CNT, its electronic properties can
demonstrate metal-like conductivity or semiconducting properties with wide or narrow
band gap.19 The changing electronic properties stem from the conductive pathways and
free electrons available in each chiral conformation. Metallic CNTs were initially
predicted to have resistivities on the order of 10-4 Ω-cm.20 These values were later
confirmed along with CNT ampacity up to 107 A/cm2.21 Further research has shown that
doping CNTs with K or Br during the synthesis process can further enhance the
conductivity of a CNT up to 30 times.22
Carbon nanotubes were also predicted to have incredible thermal conductivity
properties. While traditional 1-dimensional systems have reduced conductivities due to
increased phonon scattering, in certain 1-D systems, such as CNTs, theoretical
calculations have revealed nearly infinite intrinsic thermal conductivity. Thermal
conductivity in CNTs has been predicted to be up to 7000 W/m-K, much higher than
diamond which has the highest bulk material thermal conductivity known, 2000 W/m-K.
These ultra-high thermal conductivities are due to the intrinsic properties of the sp2
lattice structure where ballistic transport is achieved.23
Along with surprising electrical and thermal properties, CNTs have were
predicted to have mechanical properties much greater than traditional bulk
carbonaceous materials.24 The bar model developed by Lourie et. al. predicted Young’s
24
modulus up to 3.6 TPa for SWNTs and 2.4 TPa for MWNTs.25 Treacy et. al. was the first
to empirically calculate the Young’s modulus for CNTs by observing the amplitude of the
temperature dependent intrinsic thermal vibrations.26 The MWNTs studied showed a
Young’s modulus between 0.4 and 4.15 TPa, much greater than pure graphite which
has a maximum Young’s modulus of Y = 15.3 GPa. Using direct tensile loads, Lourie et.
al. found the Young’s modulus between 320 and 1470 GPa for SWNTs while MWNTs
measured 270 to 950 GPa.25 The immense strength of CNTs can be attributed to both
the strength of the sp2 C-C bond and the limited sites for defects to form. The role of
defects can be seen in the size dependent strength of graphite where tensile strengths
of thin fibers can reach 20 GPa, much higher than the 1 GPa typically observed in larger
graphite structures. Typical structural materials such as stainless steel and Kevlar have
maximum tensile strengths of 1.5 and 3.8 GPa according to experimental
measurements. Tensile strength testing of CNTs has shown up to 53 GPa while
theoretical analysis has predicted these to reach as high > 100 GPa in zigzag
conformation SNWTs. Unlike bulk materials which are limited by grain slippage, SWNTs
are limited by the strength of bonds and presence of contaminants or imperfections.
Alternately, MWNTs are limited by the strength of the inter-wall forces; it has been
shown that slip occurs between the walls of MWNTs leading to a reduced tensile
strength.27 Due to their hollow structure, CNTs do not have high compressive strengths
in the axial direction. Buckling occurs during axial compression; however, the tubes
return to their original structure upon the release of the compression.28 The ultimate
25
strength properties of a CNT though are ultimately tied to the number and nature of
defects in the structure which is related to the synthesis technique used.
Carbon Nanotube Applications
Though individual CNTs may have impressive properties, they must be
assembled into a bulk material in order to provide these benefits to usable applications.
Composites and CNT ropes have been researched as methods to exploit CNTs in bulk
materials.
Caron Nanotube Composites
Integrating CNTs as the reinforcing phase in composite materials is the most
commonly used method to take advantage of CNT properties in bulk materials. Of the
composites, organic-polymer based CNT composites are commonly researched due to
their ease of fabrication in comparison to metallic-CNT composites. However, metallic-
CNT composites may offer the most significant property benefits.
Polymer-CNT composites
The first polymer-CNT composite was reported by Ajayan et. al. in 1994 and was
fabricated by mixing arc-discharge synthesized CNTs with Epon-812 epoxy resin.29
Extensive research has been paid to variations in the polymer-CNT composite system
through variations in polymer composition, CNT type, CNT loading, and CNT alignment.
Composite materials have been fabricated with heavy- and light-molecular weight
polymers along with low- and high-viscosity polymers. In some cases, the CNTs are
mixed with the precursor while in others CNTs are mixed into a molten polymer. For
26
many thermoplastics and thermosets, the CNTs are dispersed in a solvent with the
thermoset powder then filtered, dried, and pressed.30
Metal-CNT composites
Metal-CNT composites are of great interest to many fields where size and weight
are as critical as the performance of the composite such as in aerospace and space-
bound applications. Integrating CNTs into copper or aluminum composite wire can lead
to reduced density electrical wiring for aircraft while maintaining the same conductivity
and ampacity thereby allowing for a reduction in wire weight and/or volume. Research
by Uddin et. al. showed an increase in conductivity up to 20% over a plain brass alloy
by integrating CNTs into the structure.31 There is also a desire to integrate CNTs into
high-strength, light-weight steel composites for high-performance cabling and structural
materials. Assuming the same benefits seen with polymer-CNT composites, one can
expect an improvement in fatigue life for high-cycling steel components.
Though Cu- and Fe-CNT composites are of significant interest for conductivity
and strength, other metals have been explored for a variety of applications. Composites
of aluminum32, magnesium33 and silver34 have been fabricated to determine the
improvement of hardness, bend strength, tensile strength, and conductivity. Aluminum
and magnesium alloy composites are of interest for use as structural components where
strength-to-weight ratios are of critical importance such as in aerospace applications.
Aluminum composites are also of interest in high conductivity applications where
improved specific conductivity wiring is desired. Adding CNTs into these composites
showed increases in Vickers hardness of silver by 27%, bend strength of silver by 9% at
27
8% CNTs by volume. Tensile strength of magnesium increased by 23% at 1% CNT by
volume. As one of the most researched metal-CNT composites, results for aluminum
composite strength and hardness have varied from increases over 180% to under 80%
likely due to differences in loading, mixing, dispersion, and agglomeration.
Composite drawbacks
In both polymer and metal composites there exist a few issues that hold back the
ultimate composite material properties. First among these is CNTs tendency to
aggregate due to the Van der Wals forces between them. Aggregated CNTs create high
defect densities in the surrounding matrix which can decrease the composites
mechanical strength or fatigue life at this point. Poor distribution also limits the
attainable electrical properties as well due to the lack of a percolated conductive
pathway through the composite.35
CNT Ropes
One alternative to CNT composites materials is pure CNT yarns. These CNT
yarns have extremely high strength-to-weight ratios, flexibilities, and electrical
conductivities that exceed traditional alloys and most polymer- or metal-CNT
composites. To fabricate CNT yarns, one edge of a film of freestanding, aligned CNTs is
electrostatically grabbed and twisted to initialize the yarn. As the CNTs are twisted and
drawn, the surrounding CNTs are pulled from the substrate by the Van der Wals forces
between the CNTs thereby creating a continuous wire as shown in Figure 2-2.36–38
28
Figure 2-2: SEM image of CNT fiber being wound by pulling process.39
Although CNT yarns have many favorable properties, fabrication cost is
extremely high especially with regard to the required CNT quality. The ultimate
achievable strength and conductivity of CNT yarns are lower than what is theoretically
possible due to the large number of CNT interfaces. This same result is seen in CNT
composites as the matrix/CNT interface is the weak point in the structure. As most
CNTs used in composites are on the shorter side of what is achievable in CNTs, there is
a potential method to improve composite and CNT yarn properties by increasing the
length of CNTs.20
Conclusions
Carbon nanotubes present impressive physical, thermal, and electrical properties
when measured individually. However, outside of transistors, single CNTs have very
few legitimate applications. As such materials such as CNT composites and CNT ropes
are expected to allow for the harnessing of the CNT’s properties in an industrially usable
29
structure. When integrating CNTs into composite materials or CNT ropes though, one of
the weakest points of the system is the interface between the neighboring CNTs or
CNTs and the binding matrix. By increasing CNT length, the number of interfaces in the
composite/rope can be reduced which should lead to improved physical, electrical, and
thermal properties. The reactor design presented herein is designed in such a way so
as to prevent the failure mechanisms typically seen in ULCNT growth such to increase
obtainable CNT length.
30
CHAPTER 3CHEMICAL VAPOR DEPOSITION BASED CARBON NANOTUBE SYNTHESIS
Carbon Nanotube Origins
Iijima originally observed the CNT while studying fullerenes produced via arc-
discharge methods. In the decades since that discovery, CNTs have been synthesized
in a variety of methods. Methods such as arc discharge and laser ablation present with
their own advantages and disadvantages which may make them better suited for
synthesizing CNTs for specific applications. However, CVD has been shown to grow
large quantities of highly controllable, aligned CNTs.
Chemical vapor deposition (CVD) was first used to grow CNTs by Yacaman et.
al. in 1993. By passing a hydrocarbon gas over a catalytic material at elevated
temperature, the gas decomposes and restructures on the catalyst to form a CNT.40
This method is often used to synthesize large arrays of high-purity CNTs and is used to
produce the longest CNTs currently observed (0.5 m in length).41 The first instance of
CNT growth via CVD processing utilized the decomposition of ethylene over iron
particles supported on a graphite substrate. Reactor temperatures up to 700 ºC were
used in this set up leading to a tangle of CNTs which were difficult or isolate or order in
a controllable way. The desire for isolated or aligned high-purity CNTs led to
researching the roles of each component of the CVD process including: gas
composition, catalyst chemistry, pressure, substrate chemistry, and temperature. The
addition of microwave plasma enhancements to the CVD process further affected the
growth rate and properties.42
31
Selection and Role of Gaseous Precursors
Many hydrocarbon precursors have been explored for their effect on the growth
of CNTs. These gases include ethylene (as originally used by Yacaman40), methane43,
acetylene44, benzene45, and alcohols like ethanol46. The selection of gaseous precursor
is a major driving force for the operating temperature range and in turn the purity and
structure of the as-synthesized CNT. Using a traditional hydrocarbon precursors and
transition metal catalyst, lower temperature reactions (< 900 °C) are more likely to drive
MWNT growth while higher temperatures (> 900 °C) tend to drive SWNT growth which
indicates that SWNTs have a higher heat of formation.47 However, some research has
shown that utilizing methanol as a precursor can reduce the temperature required for
SWNT formation down to as low as 550 °C.48 The CNTs grown at low temperature had
a lower Raman D/G ratio than those grown at 800 °C as shown in Figure 3-1 where the
D-peak at 1380 cm-1, indicative of disordered bonding, decreases as synthesis
temperature increases.
Work by Campbell’s group has indicated that use of simple hydrocarbons such
as ethylene and acetylene are more likely to induce growth of straight, hollow CNTs
from an Fe catalyst while complex precursors such as benzene or C60 are more likely to
form tangls of curved CNTs, as shown in Figure 3-2, due to incomplete catalytic
breakdown of the molecule.49,50
32
Figure 3-1: Raman spectra of SWNTs grown at low temperature using methanol as a precursor.48
Figure 3-2: Comparison of CNTs grown from different carbon sources. (a) Straight CNTs grown from acetylene in comparison to (b) tangled CNTs grown for C60 (right)50
The ratio of partial pressure of the precursor gas can affect the resulting CNTs
growth CVD. As shown in Figure 3-3, increasing the partial pressure of carbon
33
precursor in the gas stream increases the amount of carbon that can be absorbed by a
catalyst particle according to Equation 3-1
𝑄 = 𝑃𝑐 ∗ (2𝜋𝜋𝑘𝐵𝑇)−12 (3-1)
Where 𝑄 is the flux of carbon into the catalyst, 𝑃𝑐 is the partial pressure of carbon in the
precursor gas stream, 𝜋 is the particle mass, 𝑘𝐵 Boltzmann’s constant, and 𝑇 the
reaction temperature. This calculation is further adjusted according to the particle form
factor for carbon absorption according to Equation 3-2
QT = QψπRp2 (3-2)
Where 𝑄𝑇 is the total flux, 𝑄 is the flux from equation 2, 𝜓 is the unitless form factor
adjustment ≈ 2, and 𝑅𝑝 is the particle radius.
Oxygen has also been shown to potentially play a critical role in the synthesis of
CNTs. Work by Zhang et. al. has shown that vertically aligned CNTs grown using an
iron catalyst via methane with 0.8% O2 were dense and well aligned while those grown
from methane with 7.4% H2 were sparse and flat on the silica substrate.51 Others has
shown that O2 plays a critical role in promoting the growth of CNTs from non-metal
surfaces such as SiO2 or Al2O3 by enhancing the capture of hydrocarbon species.52 This
may play an important role in the SF-CVD reactor where the oxygen from the silica tube
or alumina template may provide the necessary oxygen to capture hydrocarbons, then
crack them to release free carbon for CNT formation.
34
Figure 3-3: Comparison of CNTs gron from different carbon concentrations. (a, c) CNTs grown from 2.5% C2H2 in N2. (b, d) CNTs grown from 10% C2H2 in N2.53
Selection and Role of Catalyst Support Substrate
The substrate material and precursor gas composition also have significant
effects on the growth properties of CNTs. Growth has been demonstrated on Ti44, Si54,
SiO255, amorphous carbon56, and alumina43 among others. Pairing the catalyst selected
with an appropriate substrate is critical to forming the desired catalyst architecture as
shown in Figure 3-4 where the substrate chemistry and thickness of the applied catalyst
film have significant effects on the catalyst particle size.55
35
Figure 3-4: Difference in Fe film morphology (a,d) 9, (b,e) 3, and (c,f) 1.5 nm were deposited on (a-c) Ta or (d-f) SiO2. Numbers in lower right hand corners represent applied film thickness prior to melting. Scale bars are (a, c, d) 100 nm, (b, e, f) 40 nm.55
Role of Catalyst Selection
Catalyst Particle Size Effects
As described previously, careful control of the substrate composition, catalyst
composition, and catalyst film thickness allows for the precise tuning of the catalytic
nanostructure to induce growth and tune the CNT diameter. Work by Li et. al. has
shown that CNT diameter is critically dependent on the diameter of the catalyst
nanoparticle. Their work showed that CNTs grown from different forms of ferritin (s- and
m-ferritin) had resulting diameters approximately 30% lower than the diameter of the
catalyst nanoparticle.43
36
Figure 3-5: Size of catalyst nanoparticles and the resulting CNT diameters from varied precursors. (a) Fe particle diameter from s-ferritin and (c) m-ferritin with the resulting CNT diameters from (b) s-ferritin and (d) m-ferritin.43
According to Sinnott et. al. this correlation between CNT diameter and catalyst
particle diameter is due to the precipitation of graphite sheets with their basal plane
around the circumference of the nanoparticle at its lowest energy state i.e. its widest
point. This observation is congruent with carbon film precipitation on catalyst sheets in
which the basal plane is parallel to the surface.57
Catalyst Composition Effects
Nearly all transition metals and many of their alloy compositions have been
explored for their ability to synthesize carbon nanotubes in a CVD system. The most
commonly researched are iron group elements Fe, Ni, and Co. Huang et. al. performed
an analysis of the difference between CNTs grown from a 25 nm thick film of each metal
sputtered on a titanium substrate, with the resultant CNTs for Ni, Fe, and Co shown in
37
Figure 3-6 A-C respectively. This research showed that CNTs grown from Ni catalyst
were the most aligned, consistently tubular, had the largest diameter (Figure 3-6D), and
had the fastest growth rate (Figure 3-6E). This was followed by Fe then Co in all
categories.
Figure 3-6: Images of CNTs grown from varied catalyst metals and the resulting CNT diameter and growth rate (A) Ni, (B) Fe, (C) Co. At varied catalyst thickness, resulting CNT (D) diameters and (E) growth rate.
The catalyst plays a variety of roles in the formation and growth of CNTs beyond
providing a nucleation site. Catalyst particle size determines the overall size of the
CNT58 and also determines the number of walls based on the atomic step structure of
38
the catalyst at the CNT ejection point.59 Also due to the higher binding energy of C in its
hexagonal structure over the pentagonal structure, the catalyst can easily dissolve a
pentagonal bonded carbon while the hexagonal bonded carbon structures stay stable
on the catalyst surface.60
Adsorption and Surface Diffusion vs. Absorption and Bulk Diffusion
According to the previously described energy model, carbon will initially adsorb
on the catalyst particle surface until the concentration at the surface is high enough for
absorption and diffusion into the particle bulk and/or diffusion across the surface along
the concentration gradient. Early work, such as that Tibbets, assumed that carbon
filaments grew via a vapor-liquid-solid (VLS) growth method, originally proposed by
Wagner for the growth of Si filaments,61 where an initially solid catalyst particle becomes
a supersaturated liquid through the absorption of solute as the precursor gas
decomposes at the particle surface. Upon supersaturation, the solute precipitates as a
solid cylinder from the low concentration side of the particle in order to maintain a lower
energy state. The VLS growth mechanism though relies on the formation of a liquid
catalyst droplet. Based on the previously described observation of CNT growth as low
as 550 °C, there would need to be a significant effect on the melting temperature of the
catalytic particles to induce melting. There is an observable drop in the melting
temperature in response to decreasing particle size, especially below 20 nm according
to Equation 3-3
𝑇𝑐 = 𝑇0 − 2𝑇0∆𝐻𝑓𝜌𝑠𝑟
∗ �𝜎𝑠𝑠 + �1 − 𝜌𝑠𝜌𝑙� 𝜎𝑠� (3-3)
39
where the melting temperature of the nanoparticle (𝑇𝑐) is a function of the bulk catalyst
melting temperature (𝑇0), heat of fusion (∆𝐻𝑓), densities of the liquid (𝜌𝑠) and solid (𝜌𝑠)
phases, solid-liquid interfacial energy (𝜎𝑠𝑠), and surface energy of the liquid phase (𝜎𝑠).
The resulting change is melting point for Fe and Ni nanoparticles (along with Au and Ag)
is shown in Figure 3-7.62
Figure 3-7: Change in melting point of catalytic materials with decreasing nanoparticle diameter.62
There is also a decrease in melting temperature of transition metal catalysts as
carbon diffuses into the system; an example of which is provided in the Fe-C phase
diagram in Figure 6-5. However, combining this with the nanoscale depression does not
provide enough of a melting point depression to cause melting of a 20 nm Fe particle at
< 700 °C. In turn, most modern work indicates that the VLS mechanism is unlikely for
low temperature CVD of CNTs. High-resolution, in-situ TEM has been used to confirm
40
this surface adsorption and transport method by imaging the crystal structure of the
catalyst particle during CNT growth as shown Figure 3-8.
Figure 3-8: Crystalline Fe-C catalyst particle ejecting a CNT tip.63
In light of this, researchers have sought to determine the driving mechanism for
CNT growth. Work by Hofmann investigated the activation energy of thermal CVD and
found that the lowest energy pathway for CNT nucleation on a Ni catalyst is adsorption
and surface diffusion to the CNT nucleation site as shown in Figure 3-9.64 This implies
that CNT growth from a catalyst particle should be driven almost entirely by surface
adsorption and transport as it is the most energetically favorable process. Early work by
Baker assumed the pure metal particles were responsible for the growth of CNTs in this
mode.65 However, as explained by Louchev, there remains debate regarding whether
CNT growth is driven by the surface diffusion or a bulk diffusion mode.60 As shown in
the TEM image in Figure 3-8, shows the iron particle in as iron carbide, indicating that
carbon had completely dissolved in the particle, not just at the surface.
41
Figure 3-9: CNT nucleation model due to adsorption of a gaseous carbon species, surface diffusion along the concentration gradient, and CNT nucleation and ejection.64
While the chemical state of the catalyst is often explored upon cooling after CNT
growth, this may not accurately represent the actual chemical state the catalyst is in
during CNT growth. In-situ TEM and x-ray photoelectron spectroscopy (XPS) has
brought to light more information regarding the catalyst chemical state. However, while
these analysis methods only allow scientists to determine what state the catalyst is in,
they do not allow for determination of the contributions of the individual components of
surface or bulk diffusion as in all forested CNT growth methods carbon has the ability to
both surface adsorb and bulk absorb in the catalyst.
Conclusions
The reactor design presented in the following chapter allows us to explore the
purely the role of bulk diffusion in CNT growth but separating the dissociative surface
42
from the CNT forming surface. Until this point, it has been impossible to eliminate the
influence of surface diffusion on the growth of CNTs via CVD. Based on the literature
studied, we expect to observe CNT growth in this manner, though it may be slowed in
comparison to traditional CVD growth as bulk diffusion through a solid catalyst particle
has high activation energy in comparison to surface diffusion or VLS. Proper selection
and pairing of the substrate and catalyst pairing will be critical to inducing CNT growth in
the desired manner in order to grow CNT with the desired properties.
43
CHAPTER 4ULTRA-LONG CARBON NANOTUBE SYNTHESIS
Failures in Ultra-Long Carbon Nanotube Synthesis
As described in the previous sections CNT based composites may have limited
properties due to poor interfacial adhesion between the matrix and CNT which is
exacerbated by the very short length of high-quality CNTs leading to a high number of
interfaces. By increasing CNT length it is possible to improve the overall composite
properties through decreasing the number of CNT interfaces while increasing the
volume fraction of CNTs in the composite. In traditional CVD CNT growth, catalyst
poisoning and amorphous carbon buildup limit the ultimate achievable length.
Previous methods of synthesizing ULCNTs have focused on one of two methods:
precise control of system parameters to grow CNT forests normal to the substrate or
use of mobile catalysts to grow CNTs in plane with the substrate. Hong et. al.
demonstrated CNT growth up to 10 cm by placing iron catalyst nanoparticles on a
quartz substrate in a tube furnace then using the flow of gas to roll the catalyst along the
substrate. Though this method formed very long CNTs, the synthesis quantity is so low
as to be unusable in realistic systems. Li et. al. synthesized aligned CNT forests up to
4.7 mm by precisely controlling the thickness of the deposited Fe catalyst layer down to
precisely 1 nm thickness on a dense 10 nm Al2O3 buffer layer in order to create very
consistent catalyst nanoparticles and limit Ostwald ripening to maintain catalyst
nanoparticle spacing. Utilizing this precise catalyst control with very low, laminar gas
flow rates allows precise control of CNT growth. Though this process was more scalable
than the mobile catalyst method it still faces the same draw backs of other carpeted
44
CNT growth methods along with limited scalability of the catalyst deposition reducing
the quantity and ultimate length of CNTs which can be synthesized.66 Alternate CVD
growth methods have been developed where catalyst precursors such as ferrocene are
regularly added to the system to replenish the catalyst. This process demonstrated CNT
forests up to 15 mm; however, the CNTs are low quality and have a high concentration
of catalyst particles in and on the CNT.67
Segregated-Flow Chemical Vapor Deposition Furnace Design
Knowing these limitations, we sought to develop a method which could
synthesize ultra-long or continuous length CNTs over a broad area. The first step to
increasing achievable CNT length was to design a method to inhibit the buildup of
amorphous carbon and catalyst poisoning. We designed a CVD chamber which was
segregated into two chambers, separated at a single point by a porous AAO template
which has open pores on both ends. The AAO template is the only region in the furnace
in which matter can transfer between the two chambers. A thin catalyst layer is applied
to the top surface of the AAO template which completely covers and seals the pore
openings. In the upper chamber, the standard CNT precursor gases are supplied,
namely Ar, H2, and a carbon precursor such as ethylene or acetylene. The lower
chamber contains only Ar and H2. The catalyst layer performs two functions: first it
prevents the carbon precursor from crossing into the lower chamber in its stable form
and second it acts as the catalyst for CNT growth in the AAO template. The carbon
precursor is broken down at high temperature and absorbs into the catalyst layer. The
carbon ions then naturally diffuse through the catalyst layer based on the concentration
45
gradient until a high enough concentration of C is obtained at the AAO surface. The
nanopore diameter induces the catalyst curvature required for CNT initiation. The basic
design of the system and a close up of the catalyst/AAO interface is provided in Figure
4-1.
Figure 4-1: Segregated flow CVD furnace diagram
Utilizing this system, we anticipate the capability of growing continuous length
CNTs limited only by the dimensions of the furnace. This would involve tuning the
optimized growth conditions and inhibiting catalyst blockage. Growing arrays of ultra-
long CNTs would then lead to composites with higher CNT loadings and reduced
CNT/matrix interfaces. In order to improve the chances of synthesizing ultra-long CNTs,
we required a better understanding of catalyst/substrate interface within the AAO
nanopore. This would allow us to design a catalyst and substrate pairing which will best
induce ULCNT growth.
46
The proposed mechanism for CNT growth in the SF-CVD reactor is shown in
Figure 4-2. This consists of (A) an initial AAO template closed with a catalyst layer on
one side where (B) at the start of the run, carbon begins to diffuse into the catalyst and
forms nucleation sites for CNTs on the outer surface. As the run progresses, (C) carbon
saturates the catalyst, increasing the number of nucleation sites and forming nucleation
sites within the pores. Finally, (D) CNTs began to grow from the nucleation sites; the
increased number of nucleation sites on the top surface of the catalyst lead to a higher
density of CNTs on the catalyst side. As is evidenced through Figure 4-2, the layout of
the catalyst and pore structure inhibits the influence of the surface diffusion and
transport mode by eliminating a surface pathway for carbon to reach the CNT ejection
site. Instead, the process relies on the bulk diffusion of carbon completely through the
catalyst to drive CNT growth.
47
Figure 4-2: Proposed mechanism of ULCNT growth in the SF-CVD reactor where (a) an AAO template is coated with a catalyst, (b) carbon dissolves into the catalyst at elevated temperature, (c) carbon saturates the catalyst leading to (d) ejection of CNTs from nucleated regions.
Investigation of Catalyst Composition
A thorough literature study was performed to determine which elements could
effectively catalyze CNT growth. Catalyst choice in this system depends not only on the
ability of the catalyst material to produce the desired type and quality of CNT. It also
depends on the melting point and whether the catalyst wets the alumina pores to some
degree. As shown in Table 1, only pure indium wets pure alumina, and all other
catalysts are phobic to the substrate—the contact angle is greater than 90˚. Indium will
readily wet an alumina pore; however, a pressure gradient exceeding the capillary
48
pressure would need to be imposed to cause the other molten catalyst materials to
penetrate a 10 nm pore, which is not possible. However, the contact angle
measurements in Table 4-1are of a pure metal on alumina and do not consider the
effects that a solute, such as carbon, present in the melt will have. Without detailed
experimental data, it is difficult to exactly calculate the change in surface tension (γlg)
and contact angle (θc) as a function of solvated carbon within the molten catalyst.
Traditional carpeted CNT growth requires highly controlled catalyst/substrate
interaction to form nanoparticles and the formation and growth of the CNT structure.
The high curvature and surface energy of the catalyst particle is the driving force behind
CNT formation and growth.68 On traditional substrates, the catalyst nanoparticle
diameter is controlled by the thickness of the initial catalyst film and the
catalyst/substrate interaction or texturing the substrate to prevent Ostwald ripening of
the nanoparticles. Conversely, templated CNT growth does not require the creation of
ultra-thin films because the catalyst is constrained to nanoparticle scale by the diameter
of the AAO nanopores. In the SF-CVD reactor, the catalyst does not initially penetrate
the pores but covers over the pore openings. Upon heating, the surface energies
between the catalyst and substrate determine how the catalyst interacts within the pore.
49
Table 4-1. Physical and catalytic properties of metallic catalysts
Catalyst metal
Melting point (°C)
Surface tension (N/m)
Contact tngle with Al2O3 (°)
Capillary pressure (x107 ATM)
SW/MW
Fe 1538 165025 13426 –5 SW27 Co 1495 177925 11328 –3 SW27 Ni 1455 181025 14728 –6 SW27 Au 1064 121125 13928 –4 SW/MW27,29 Ag 961 92525 14028 –3 SW27 Cu 1083 135225 13728 –4 SW27 Pt 1768 189625 15428 –7 SW27 In (ITO) 156 56030 < 90 (wets) > +4 SW/MW31,32, 33 Ga 30 72534 118.435 –10 SW/MW33, 36 Ge 938 60037 111.135 –9 SW38
Depending on the surface energies between the liquid catalyst, solid substrate,
and gaseous environment, the catalyst-nanopore interface can act in one of three
manners as shown in Figure 4-3. If the catalyst is repelled by the substrate it will form a
convex meniscus at the pore as shown in Figure 4-3 (A) and Figure 4-3 (C). In this
mode, the catalyst adhesive force between the catalytic particle and bulk catalyst or
CNT will drive tip growth (C) or the preferred root growth (A). Tip growth has no means
to form a continuous ULCNT as the catalyst particle dislodges from the bulk catalyst
and no longer has a means to absorb extra carbon. If the catalyst wets the substrate it
is likely to form a concave meniscus, as shown in Figure 4-3 (B). Based on the
curvature, it is unlikely this will grow CNTs as CNT cap and wall formation is driven by
the curvature of the catalytic particle.
50
Figure 4-3: CNTs grow through a bulk catalyst layer in one of three potential modes. (a) the catalyst does not wet the substrate and drives root growth from the convex, (b) the catalyst wets the substrate driving root growth from the concave meniscus, or (c) the catalyst does not wet the matrix but dislodges a particle and drives tip growth.
In macroscopic systems, one would utilize the Young-Laplace equation to
determine the determine the rise in the capillary and the wetting properties within.
Young-Laplace relates the height of a fluid drawn into a pore (h) to the surface tension
(γ), contact angle (θ), density (ρ), gravitational force (g) and pore radius (a) as shown in
Equation 4-1.
ℎ = 2𝛾 cos𝜃𝜌𝜌𝜌
(4-1)
where surface tension and density are material constants. Young-Laplace is
often used to describe droplet behavior at the nanoscale interfaces.69,70 However,
conflicting molecular dynamics (MD) studies have shown that surface tension can either
be influenced by pore diameter in either direction or be independent of curvature.70–72 A
51
comprehensive MD study by Liu and Cao has showed the efficacy of Young-Laplace at
nanoscale dimensions down to 0.5 nm pore diameter using water and a nanopore on a
carbon plate. Work by Li and Akkutlu has indicated that Young-Laplace is accurate
down to 10 nm, smaller than the pores used for this research.73 The results of these
studies support the use of the Young-Laplace relationship to relate the curvature of a
macroscopic droplet to confinement at the nanoscale.
Determining the contact angle and surface tension of various metals in view of
the efficacy of Young-Laplace in a nanopore, should allow us to predict the wetting
environment in the nanopore based on the interfacial energy. To do this, we will need to
develop a Zisman plot, which correlates the surface tension of the liquid metal to the
contact angle on a desired substrate. In this graph the cosine of the contact angle of
varied liquids is compared to their surface tension and the intersection of this trend at
180° is considered the solid/gas interfacial energy (γsg).
52
CHAPTER 5TEST STAND DESIGN AND ASSEMBLY
System Level Design
The experiments performed for this project are intended to provide an
understanding of how catalyst and substrate selection affect the growth of CNTs in
dense packed porous membrane. In order to predict the optimal substrate/catalyst
system, the interfacial energy between the catalyst and substrate must be determined in
order to estimate the curvature of a catalyst within a nanopore. To do this a custom
high-temperature contact angle measurement (HT-CAM) system was assembled in
conjunction with the SF-CVD furnace. An overview of the two furnace system is
provided in Figure 5-1.
Figure 5-1: Process and instrumentation diagram of test-stand
53
Segregated-Flow Chemical Vapor Deposition Reactor
The SF-CVD furnace is assembled starting with a Carbolite-Gero EHA 12/600
tube furnace which features a 1200 °C maximum temperature in a 600 mm heated
section. The quartz tube is 2 in outer diameter with a 1.8 in inner diameter. The tube is
sealed on the inlet end with a custom tube furnace flange designed to be similar to a
standard MTI Corp but with two vertically aligned inlet tubes to allow for gases to be
properly segregated. A single outlet flange seals the opposite end of the 1200 mm long
tube. The segregating plate is custom designed and fabricated by Technical Glass
Products (Painesville TWP, OH). It consists of a 5 mm thick quartz plate with a 20 mm
radius on the side to allow the plate to seal neatly against the curved interior surface of
the quartz tube. In the center of the segregating plate is a 23 mm hole completely
through the plate with a 27 mm diameter x 1 mm deep annular region for the AAO
template to rest on. A cross sectional drawing of the segregating plate is provided in
Figure 5-2.
Figure 5-2: Segregating plate cross section
High-Temperature Contact Angle Measurement System
As described previously, contact angle measurements are the critical
measurement used to calculate the interfacial energy between a liquid droplet, solid
substrate, and the gaseous environment. In low temperature liquid systems this
experimental is simple to perform with a high-fidelity camera, macro lens, and a well
54
leveled substrate. However, performing this measurement on liquid metals is more
complicated. First, the measurement must be performed while the metal is in its liquid
state. At these high temperatures the metal will rapidly oxidize if not held in an oxygen-
free or reducing environment. Holding this oxygen-free atmosphere and high-
temperature requires a closed system which makes performing high-fidelity imaging
difficult. To address these issues in performing this analysis we assembled a custom
high-temperature contact angle measurement system based on a tube furnace featuring
a flange with a quartz window and a fixed focal length camera. A basic diagram of the
system is presented in Figure 5-3. The system consists of a (1) high temperature tube
furnace (MTI GSL-1700x) capable of short temperature holds up to 1700 °C and
sustained 1650 °C. The furnace uses an (2) alumina tube to prevent deformation at high
temperature and is sealed with (3) water cooled tube flanges (MTI EQ-FG-60WC) with a
single compression fitting for flowing gas into and out of the tube which are connected
to a recirculating chiller operating at 5 °C. The flanges compress two O-rings around a
centering ring to completely seal the tube on each end and are water cooled to slow the
rate at which the high-temperature silicone O-rings degrade. One flange features (4) a
quartz viewport sealed with a graphite gasket to prevent leakage (MTI EQ-FG-60W). An
(5) alumina D-tube is used to lift the (6) substrate and (7) catalyst up to the center of
tube in order to allow for easy imaging with the (8) camera and fixed focal length lens.
The design of the system is loosely based on that of Zhao et. al.74 and similar systems
have been used by Ogino75 and Kapilashrami76 using visual analysis and x-ray analysis
respectively.
55
Figure 5-3: Diagram of high-temperature contact angle measurement furnace
Gas Flow Management
The high temperature measurement system utilizes a mixture of Ar, H2, and
C2H4. Gas flow is controlled using five Bronkhorst EL-FLOW mass flow controllers. The
system needs 5 controllers in order to provide Ar, H2, and C2H4 to the HTCAM through
the first three while allowing the SF-CVD furnace to use Ar, H2, and C2H4 in the top
chamber and Ar and H2 in the bottom. Each controller has a solenoid valve to shut off
the flow to the furnace(s) at any time. The mass flow controllers are housed inside an
aluminum enclosure, shown in Figure 5-4, and is connected to the ventilation lines to
vent any gas leaks. The enclosure has been designed with extra space to allow for
expanded functionality via an increased number of gas lines or the addition of a bubbler
to inject water vapor. The effluent lines of the SF-CVD and HT-CAM connect to the
same ventilation line which provides venting for the MFC box. The vent fan is simple
56
explosion proof, in-line muffin fan (Shield Air MSFX) which helps to prevent effluent gas
from building up in the lab.
Figure 5-4: Mass flow controller containment box featuring (1) Bronkhorst MFCs with (2) integrated solenoid valves controlled by (3) an electronic relay integrated with the custom labview program
Imaging Management
Obtaining high quality contact angle measurements requires consistent camera
positioning and precision focusing. The camera is a Mako G234B GigE camera with a
Sony IMX249 CMOS sensor and a fixed focal length lens. The lens has a magnification
of 0.5, f # of 13.9, field depth of 0.12” and working distance of 15.7”. The camera is
mounted on top of a XYZ linear translation stage with a maximum travel distance of 1”,
1” and .55” in the X, Y, and Z directions, respectively, allowing us to ensure proper
focus on the droplet. A hot mirror is positioned between the camera and the viewing
flange in order to reduce the thermal stress on the camera lens. A hot mirror is used to
57
reflect infrared radiation caused away from the camera and prevent degradation of the
lens. The hot mirror has a multi-layer dielectric coating and can reflect more than 90%
of infrared radiation (700 nm – 1 mm) and transmit more than 80% of visible light (380 –
700 nm) for a 0° incident angle. Additionally, a small fan will be used to provide forced
air cooling over the hot mirror, camera lens, and tube furnace flange. The four main
components of the camera system are shown in Figure 5-5 where the high resolution
camera body (1) is attached to a 400 mm fixed focal length lens (2) then mounted on an
X-Y-Z translational stage for focusing on the sample of interest (3). The hot mirror is in
place to prevent the lens from overheating without blocking visible light (4).
Figure 5-5: Camera set-up for imaging molten metal contact angle
58
Controller Design
A custom LabView program was assembled to control, record, and visualize the
components of the HTCAMS and SF-CVD systems. This process required replacing the
stock temperature controllers with Watlow EZ-ZONE PM4 digital controllers capable of
receiving temperature profiles and relaying the furnace temperature to the computer.
The MFCs have custom LabView drivers directly from Bronkhorst and the camera
system uses a simple interface to take still images or record video.
High-Temperature Contact Angle Measurement Stand Calibration
Determining the role of carbon infiltration on the catalyst requires knowing the
exact melting point of the catalyst droplet. The actual temperature within the alumina
tube is offset from the measured temperature due the thermocouple is location in an
alumina sheath touching the outside of the alumina tube. The simplest method for
calibrating the furnace involves inserting a ceramic sheathed thermocouple into the
center of the tube through a flange at the end. This method is costly though due to the
need for an extended length high temperature thermocouple (Type-B), an alumina
sheath to enclose thermocouple and a new flange to allow the thermocouple to pass
through. Instead of purchasing these components for a single furnace calibration we
calibrated the furnace by measuring the melting points of high purity metals. This is
similar to processes used by Domagala77 and Allen78 which look for the deformation of a
metal piece to obtain its melting point.
To observe the melting point, metal foils were cut into thin strips approximately 2-
3 mm across and folded into a pyramid shape so that a sharp point existed at the top.
59
Tracking the peak allowed easy observation for when deformation and melting began.
The melting process consists of 4 steps, as shown in Figure 5-6, where the foil is initially
upright (Figure 5-6A) then begins to deform as the melting point approaches (Figure
5-6B). The sample then completely melts (Figure 5-6C) and aggregates into a single
droplet (Figure 5-6D). The temperatures at which each step occurs can be observed in
the upper left corner of the video screen shots.
Figure 5-6: Metal melting process observed in contact angle measurement system where (a) there is a pyramidal shaped metal foil, (b) the foil begins to collapse as it approaches its melting point, (c) the metal completely melts, and (d) forms into a molten droplet
60
For this melting point calibration, samples with 99.999% purity of each copper,
nickel, titanium and iron were individually melted in the furnace under argon flow with <
4% H2 to prevent oxidation. We were unable to melt titanium in the furnace for this test
and thus it is not used as part of the calibration. The measured melting points of the four
metals showed a 2.6% linear offset and are plotted in comparison to the measured
melting points inside the furnace in Figure 5-7.
Figure 5-7: Melting high purity metals in the high temperature contact angle measurement furnace
Contact Angle Measurements
The interface between the catalyst and substrate can be explained
mathematically through quantifying the interfacial energy. This interfacial energy is
calculated by measuring the contact angle of a liquid droplet on a level substrate. The
61
contact angle (θc) is dependent on the interfacial energy between the liquid and solid
(γSL), liquid and gas (γLG), and solid and gas (γsg) according to the Young equation,
Equation 5-1.79
𝛾𝑆𝑆 = 𝛾𝑆𝑆 + 𝛾𝑆𝐿 cos 𝜃𝑐 (5-1)
Combining this with Young-Laplace will give an estimation of the contact angle
inside the nanopore and the wetting properties.
After verifying the temperature reading on the furnace is equal to the melting
temperature we needed to verify consistent measurement using the camera system. To
start, two iron foils were cut with similar masses (114.5 mg and 115.7 mg), then folded
into quarters and placed on new single crystal alumina substrates and inserted into the
center of the furnace. The furnace was then ramped up in temperature according a
standard program. This program consists of a 10 min purge cycle of 1000 sccm Ar at
room temperature. The temperature is then ramped at the recommended limit for an
alumina tube (5 °C/min) under 300 sccm Ar and 10 sccm H2 to maintain a reducing
atmosphere and prevent oxidation of the metal. For melting measurements, the
temperature is taken to 10% over the estimated melting point or up to the limit of the
furnace, whichever is less. During the heating process, the hot mirror is intermittently
removed to check the camera focus and the camera position is adjusted until the
material is in focus. An example of a focused camera image is provided in Figure 5-8
62
Figure 5-8: Focused image of two molten iron droplets on sx-Al2O3
We next needed to determine what size catalyst droplet we would need to obtain
an accurate and consistent measurement. This required a droplet small enough to
minimize the gravitational effects on contact angle while having a large enough droplet
to observe the contact angle and melting point. Samples weighing 23 – 1500 mg were
melted on alumina substrates and the contact angle was measured to determine if there
was effect. As shown in Figure 5-9, there was an observable increase in the contact
angle at higher droplet masses indicating that gravitational forces were affecting the
particle shape. The 23 mg sample had the truest contact angle then according this,
however it was difficult to measure which can lead to increased error. For this reason,
we focused on using a 100 – 150 mg metal sample as that was the easiest to observe
and consistently measure while being within a standard deviation of the “true” contact
angle.
63
Figure 5-9: Contact angle of Fe on Al2O3 with respect to droplet mass
It would be fairly straightforward to calibrate the error in a dimensional
measurement by imaging an item with a known dimension. However; calibrating the
error in a contact angle measurement is more complicated. To provide some
quantification to the error, three molten metal droplet images with different contact
angles were provided to 4 engineers who were asked to use their preferred image
editing software to measure the contact angle. The calibration images, shown in Figure
5-10, and the resulting measurements, shown in Table 5-1, showed an approximate
standard deviation of 4.5°. Further calibration was performed by repeating the iron on
sx-Al2O3 measurement six times under the same melting conditions. This set of
64
experiments showed a standard deviation of 4.1°. For data presented herein, a standard
deviation error bar of the larger error, 5.5°, will be assumed.
Figure 5-10: Calibration images used to determine the approximate user error in the measurement of molten metal contact angles used for the calibration in Table 5-1
Table 5-1. Calibration data for four measurements of the four calibration images
Calibration image Person 1 Person 2 Person 3 Person 4 Average Standard
deviation (a) Top Left 138.8 137.9 145 132.8 138.63 5.00 (b) Top Right 108.4 105.9 101 99.8 103.78 4.07 (c) Lower Left 117.2 130.4 114 102.1 115.93 11.63 (d) Lower Right 104.7 103.7 107 103.4 104.70 1.63
Further calibration was performed in order to determine the effect of
measurement temperature and time at temperature. The same Fe-Al2O3 system was
used with melting performed under the same conditions. In this test contact angle was
measured every 50 °C between 1500 and 1650 °C on both the heating and cooling
cycles. This test showed not consistent change in the contact angle with respect to time
65
at temperature or measurement temperature as shown in Figure 5-11. This is further
correlated in work by Zhao et. al. which demonstrated similarly no change in Fe on
Al2O3 contact angle with changing temperature.74 Utilizing this data implies that
measurements are not required to be performed at the exact moment in time, time after
melting, or temperature. Ultimately this means that the varied melting points and
exposure times for each experiment should not affect the contact angle, only the
experimental conditions should.
Figure 5-11: Change in contact angle with respect to temperature
66
CHAPTER 6ROLE OF CARBON AND HYDROGEN IN THE GAS STREAM
Contact Angle Measurements
The major work of this project, outside of the assembly of the contact angle
measurement system and SF-CVD reactor, was (1) the measurement of molten metal
contact angles in order to predict the behavior of a catalyst droplet within a nanopore
and (2) attempted growth of ULCNTs using the SF-CVD reactor. The first step of
growing ULCNTs is to characterize the interfacial energy between the catalyst and
nanoporous support.
Iron Wetting Properties
Understanding the growth properties of ULCNTs in the SF-CVD reactor requires
first understanding the interfacial energy between the catalyst and substrate then using
this information to predict the curvature of the catalyst within the nanopores. As it’s not
possible to measure the catalyst curvature in situ, CNT growth rate and CNT quality are
used to determine the effect of catalyst structure on CNT growth. Iron was initially
melted on single crystal alumina substrates leading to the measurements shown in
Table 6-1. The experiment was then repeated on aluminum nitride (AlN), pyrolytic
graphite, and beryllium oxide (BeO) using both Cu and Ni as well as Fe as shown in
Figure 6-1. These substrates were chosen for the low reactivity with molten metals and
high melting temperatures far above that of iron.
67
Figure 6-1: Contact angle of each molten metal (Cu, Ni, and Fe) on the defined substrates
68
Table 6-1. Contact angles of reference metals on varied substrates
Metal MP (°C) ƔSG AlN BeO Pyro G Al2O3 Cu 1084 1.29 129 139 122 107 Ni 1454 1.77 105 115 60 110 Fe 1538 1.92 107 112 30 111
As can be seen in the table, the only wetting conditions found were Ni and Fe on
pyrolytic graphite. For this work the focus remained on iron due to the well-defined
phase diagram, catalytic activity, and ability to transition between wetting and non-
wetting systems using only the desired chemical components.
As explained previously, the solid gas interfacial energy by developing a Zisman
plot. Recent work by Brillo in 2005 analyzed the effect of temperature on the surface
tension of iron, nickel, and copper droplets and their alloys. This work showed
decreases in surface tension from 1.92 N/m at the melting point (1538 °C) to 1.82 N/m
as temperature increased to 1800 °C. Nickel surface tension decreased from 1.90 N/m
at 1150 °C to 1.70 N/m at 1700 °C; specifically, 1.77 N/m at the melting point (1454 °C).
Finally copper showed a surface tension of 1.29 at the melting point of 1084 °C.80 Using
these values, a Zisman plot can be generated for each substrate to verify the validity of
the measurements, then the value for γsg can be plugged into the Young equation to
solve for γsl. However, as explained by Kruss GmbH, the Zisman plot is ineffective on its
own when used for Al2O3 due to its polarity. The graphs in Figure 6-2, produced by
Kruss GmbH, show how a polar substrate affects the accuracy of the Zisman
estimation.81
69
Figure 6-2: Zisman plots of non-polar and polar materials (a) LDPE is non-polar and (b) PMMA is polar.81
This is further shown experimentally where analyzing the contact angles of Cu,
Ni, and Fe on Al2O3 and plotting them with their surface tensions leads to a negative
70
surface energy as shown in Figure 6-3. Obviously, a negative surface energy is not
possible as it would indicate that the material is actively dissolving. In an attempt to
verify the efficacy of the measurement method the process was repeated on pyrolytic
graphite, a non-polar material, and found to give a surface energy of 1.95 J-m-2 as
further shown in Figure 6-3. However, work by Morcos82 and Ooi83 showed the surface
energy of graphite to be between 0.035 and 0.2 J-m-2, a full order of magnitude lower
than found with the Zisman plot. This difference in surface energy could be due to the
temperature and gas environment dependence of surface energy at 1600 °C
Figure 6-3: Attempted Zisman plot for Al2O3 and pyrolytic graphite
In this situation where the Zisman plot is infeasible. We could use Owens-Wendt
or Fowkes theory to account for the polar and dispersive components of the liquid and
71
solid surface energies. The polar and dispersive components of the liquid metals are
simple to calculate as they are non-polar and as such the surface tension (surface
energy) is entirely due to the dispersive component. Estimating the polar and dispersive
components of the solid surface energy is more difficult. Experimentally, three liquids
would be chosen which feature different polar and dispersive surface tension
requirements then the contact angle measurements of these liquids is used to
determine the polar and dispersive components of the solid.
To provide a best estimate of how the contact angle will change with respect to
the nanopore interaction, the surface energy of Al2O3 is estimated to be approximately
1.25 J-m-2 according to best fitting experimental estimates.84 Using this value for the
surface energy of Al2O3 along with the estimate of 1.92 J-m-2 for the surface tension of
Fe allows us to determine the value of γLG as a function of the 110.5° contact angle
which leads to an interfacial energy of 1.82 J-m-2. Plugging this value into Young-
Laplace gives us an estimated wetting height of -0.5 mm indicating that the iron catalyst
will not wet the pores. Performing the same analysis using the values for pyrolytic
graphite and the Fe on graphite wetting angle gives a wetting height of 20 um.
Effects of Carbon and Hydrogen Exposure
One of the first steps to determining the role of carbon infiltration into the catalyst
was to determine the rate at which carbon absorption occurs. This was accomplished by
exposing Fe samples with nearly equal surface areas, approximately equal to as those
used in the initial contact angle experiments, to a mixture of 57 sccm ethylene, 28 sccm
H2 and 200 sccm Ar at 650 °C for 1 – 6 hours. After the exposure time, the sample was
72
heated to its melting point and its contact angle was measured. Sample mass was
measured before and after melting for experiments both with and without H2 in the gas
stream during the melting process. This test showed linear mass gain for both melting
processes however the amount of mass gained by the Fe sample was greater in tests
without H2 in the gas stream, as shown in Figure 6-4.
Figure 6-4: Effect of ethylene exposure time on Fe sample mass with and without H2 in the stream
Based on the data from this experiment we can expect a mass gain of 0.1 – 0.2
mg/h when H2 is in the gas stream and 0.6 – 0.7 without H2. At two hours exposure time
the mass gained by the sample melted without hydrogen was lower than that melted
with hydrogen. The difference in weights is relatively low though and could be due to
incomplete reformation of the molten droplet during melting or within the error of the
73
measurement. When melting without hydrogen at times the molten iron would form a
series of extremely small droplets that did not coalesce into the main droplet, leading to
measurement error. The difference in the observed mass gain could be due to hydrogen
reducing free carbon atoms from the surface of the catalyst before they could be
absorbed or a reaction occurring between the catalyst and substrate at high
temperature. At these temperatures (>1500 °C), the H2 in the gas stream could reduce
the Al2O3 into O2 and free Al which could be alloyed with the Fe catalyst.
One way of checking whether the mass gained by the sample is completely due
to carbon absorption is to verify the melting point according to the phase diagram
provided in Figure 6-5, which shows that the Fe melting point should drop from 1538 °C
to approximately 1450 °C as the carbon content increases up to 1% by mass.
74
Figure 6-5: Fe-C phase diagram85
However, upon examining the melting point of samples when melting with
hydrogen in the gas stream we found there was no noticeable reduction in the melting
temperature, as shown in Figure 6-6. This indicates that the carbon was not absorbing
into the catalyst but instead adsorbing on the surface then reducing off the surface as a
hydrocarbon when heated to melting. This is further emphasized in view of the Fe-Al
phase diagram which shows a limited drop in melting temperature up to 15% Al by
weight. Samples melted without H2 in the gas stream showed a pronounced linear
75
decrease in melting temperature, as shown in Figure 6-6, with respect to ethylene
exposure time more closely resembling the Fe-C phase diagram.
Figure 6-6: Effect of ethylene exposure time on melting point of Fe with and without H2 in the stream
Combining the results from Figure 6-4 and Figure 6-6 using data for samples
melted without H2 in the gas stream showed that the measured melting point was less
than the anticipated melting point based on calculated carbon content as shown in
Figure 6-7.
76
Figure 6-7: Carbon in Fe weight % vs melting temperature
The offset in melting anticipated vs measured melting point could be due to
interactions with aluminum absorbing into the catalyst or carbon mass loss occurring
after melting. If the carbon is vaporized from the catalyst particle during/after melting, it’s
possible that we observed the melting point at a carbon content that is different from the
measured carbon content after cooling.
Having determined the role of carbon on the melting temperature we measured
the contact angle of the molten iron droplets with and without hydrogen in the gas
stream. An initial offset due to the change in γlg and γsg is expected with the changing
gas composition; however, this difference of approximately 1° (110° vs 111°) at zero
minutes was within the error of the measurement. This measurement correlates well
with literature values which are shown between 100° and 140°.74,76 There was an
77
increase in contact angle (less wetting) with respect to increasing ethylene exposure
when hydrogen was in the stream as shown in Figure 6-8. However, without hydrogen
in the gas stream the contact angle was nearly constant around 105° after an initial drop
in contact angle from 110° at 60 min ethylene exposure. The constant contact angle
with increasing mass and decreasing melting point, indicative of increasing carbon
content, indicates that the contact angle of the catalyst should not change as carbon
diffuses through the catalyst and reaches saturation.
Figure 6-8: Effect of ethylene exposure time on contact angle of molten Fe on Al2O3 with and without H2 in the stream
Effect of Surface Pretreatment
Having shown that adding carbon to iron did not have a significant effect on the
interaction between the catalyst and substrate we explored pretreating the Al2O3
78
substrate with ethylene to deposit a thin carbon film on the substrate surface. To do this,
the substrates were placed in the tube furnace, and heated to 650 °C under H2 and Ar
flow then ethylene was mixed in and flowed for 2 h before cooling under H2 and Ar. An
iron piece was placed on the treated substrate then and subject to the same melting
process used in previous experiments. Upon melting, the iron wetted the C-coated
Al2O3 substrate with a contact angle of 68°. This aligns with the initial contact angle
seen by Zhao for iron on pure graphite. A second test was done to expose the Fe
catalyst to additional ethylene for 2 h before melting which reduced the contact angle to
91° upon melting as shown in Figure 6-9. This is contraindicative to Zhao’s work which
showed that for extended Fe-C reaction times the carbon content in the Fe droplet
increased and the sample more thoroughly wetted the graphite. This decrease in
wetting could be due to C absorption causing the catalyst/substrate interface to be more
determined by the Fe/Al2O3 interaction as the presence of alumina can affect the
contact angle.74
79
Figure 6-9: Effect of pretreating Al2O3 substrate with C for 60 minutes on contact angle of molten Fe droplet
The results of the carbon pretreating the substrate correlate with previous
experiments looking at the contact angle of Fe on pyrolytic graphite. A comparison of Fe
contact angles on Al2O3 and pyrolytic graphite are shown in Figure 6-10. These results
indicate that if we desire a wetting interaction between the catalyst and nanoporous
support it may be beneficial to pretreat the substrate with carbon.
80
Figure 6-10: Observed Fe melting on varied substrates (a) Al2O3 and (b) pyrolytic graphite
Conclusions
This data indicates that the change in contact angle observed when H2 is in the
gas stream is actually due to changing substrate chemistry. This would correlate with
the melting point and mass change data that is not easily explained by carbon
absorption alone. This data also shows that carbon absorption does not have a
significant effect on the contact angle of iron on alumina. This is encouraging that during
ULCNT growth we can expect the system to remain in a steady state and not alter
between wetting and non-wetting properties during the growth process. It also indicates
that there is not a simple way to get iron to wet alumina through carbon absorption or
hydrogen gas manipulation. Based on the analysis of alumina substrates pretreated
with carbon we expect that we can vary the wetting properties of the catalyst within the
nanopores by pretreating the template with carbon to induce wetting. The wetting
property expectations are further confirmed by calculating the Young-Laplace
relationship to determine wetting height where Fe would not wick into an alumina pore
but will when its coated with carbon.
81
CHAPTER 7GROWTH OF ULTRALONG CARBON NANOTUBES
Ultralong Carbon Nanotube via Segregated-Flow Chemical Vapor Deposition
Using the information gathered from the contact angle measurements, we moved
into ULCNT growth using the SF-CVD reactor. A polished, high-purity aluminum disk
(27 mm diameter, 24 mm exposed diameter) was immersed in 0.05M oxalic acid
maintained at 0 °C. An initial anodization step was performed at 160 V to prepattern the
aluminum surface with a thin AAO layer in order to prevent runaway heating in the
second anodization step. After 2 hours in the initial anodization, the sample was
removed and immersed in a 0.3 M oxalic acid and anodized at 140 V for 24 h to create
an approximately 200 - 300 um AAO template. After 24 h at 140 V, the voltage is
reduced to 40 V to transition the pores from 140 nm to 40 nm. This performs two roles,
first the 40 nm pores should produce CNTs with diameters less than 40 nm which
should produce higher quality CNTs than 140 nm. Secondly the 140 nm region acts as
both a thick support layer to improve the structural stability of the template but also the
pore diameter allows the maximum 40 nm diameter CNT to grow freely with less
interaction with the pore walls.
After anodization is complete, the aluminum disk is then inverted in its holder and
the aluminum is removed by immersing a mixture of 0.5 M CuCl2 with 10% w/w HCl.
This reaction takes approximately 4 hours to complete and leaves the alumina layer in
the center supported by a aluminum annular region. After thoroughly rinsing, the sample
is soaked in 25% H3PO4 to dissolve the remaining barrier layer leaving an AAO
nanoporous membrane open on both sides with approximately 140 nm pores on the
82
growth side and 40 nm pores on the catalyst coated side. After pore opening is
complete, 500 nm of Fe is applied to the smooth, freshly exposed nanopore surface
using a thermal evaporator. The Fe completely covers the open pores creating the
desired sealed membrane structure shown previously in Figure 4-2. After Fe deposition,
the sample is placed in the SF-CVD furnace in a recessed hole in the divider plate. A lip
around the hole aligns with the aluminum annular region and ensures that proper
sealing between the clean and dirty sides of the chamber.
The furnace is heated to 650 °C under 200 sccm Ar and 28 sccm H2 before
ethylene is switched on to the furnace at 57 sccm. The furnace runs under these
conditions then for the desired period of time up to 72 hours. After the growth period is
complete, the sample is cooled under Ar and H2 before being removed. The initial
samples, which appeared similar to Figure 7-2, had amorphous carbon growth in the
dirty stream and small regions of black discoloration in the clean side. Imaging the clean
side with SEM, Figure 7-3, showed patchy spots of heavily coiled tubes with long,
smooth tubes reaching out from the central cluster. Of the initial samples run under
these production conditions we discovered some incidences where Fe had traveled to
the clean side of the template and others where they had not. There was no tell-tale
reason for samples which had iron and those that did not. Raman analysis did not show
the tell-tale peaks at 1350 cm-1 and 1580 cm-1 Raman shifts which we use as the first
indication of CNT growth. This lack of Raman signal could be due to the limited, spotty
CNT growth and the reliance on random chance to position the Raman laser perfectly
over a CNT cluster.
83
Figure 7-1: AAO template growth side before catalyst coating
Figure 7-2: AAO template with aluminum annular support after initial test run showing the (a) catalyst coated side and (b) growth side
84
Figure 7-3: Patchy CNT growth from initial SF-CVD runs showing (a) the full cluster and(b) a zoomed in region of the same growth
Decreasing the carbon flow and increasing the time proportionally to deliver the
same amount of carbon led to increased CNT growth on the growth side and
significantly increased amorphous carbon buildup on the dirty side of the template as
shown in Figure 7-4 .
Figure 7-4: AAO template after CNT growth in the SF-CVD reactor showing (a) amorphous carbon overgrowth on the catalyst coated side and (b) patchy dark spots on the ULCNT growth side
After removing the sample, it was weighed and excess carbon removed then
analyzed using various methods including Raman spectroscopy, SEM, and TEM. The
85
Raman spectra from this sample are shown in Figure 7-5 with the growth side in red
and the catalyst coated side in blue. According to work by Dresselhaus in 2005, the
large peak at 1350 cm-1 represents the “dispersive” band of carbon bonding, indicative
of non-graphitic bonding, while the smaller peak at 1600 cm-1 represents the amount of
graphitic bonding.86 The D/G ratio is the ratio between the peaks heights, which
representation of the amount of disordered carbon bonding to graphitic carbon bonding.
For this sample, the D/G ratio is approximately 1.6 on the growth side and 1.4 on the
catalyst coated. Typically “high quality” MWNTs are considered those with D/G ratios
less than 1, thus a lower D/G ratio is indicative of more carbon bonding. The higher D
peak on the growth side could be due to increased inter-wall bonding in these thick
CNTs or poor graphitization due to lower catalyst curvature angle leading to more CNT
walls and interwall bonding. The D/G ratio on the catalyst-coated side of the template is
lower which indicates more graphitic bonding; however, this measurement was
performed after the removal of a large mass of amorphous carbon, which showed no
discernible D/G signal as shown in green in Figure 7-5. The catalyst-coated side D/G
ratio could also be explained by planar graphite forming on the surface of the catalyst,
which would elicit the same G-peaks. Finally, spot to spot variation could be responsible
for some change in the D/G ratio which could align the ratios on the catalyst coated and
growth sides.
86
Figure 7-5: Raman spectra of CNTs grown in clean stream of SF-CVD reactor showing the difference in Raman signal between the growth side (red), catalyst coated side (blue) and the amorphous build-up from the catalyst coated side (green)
Upon close examination of the dark spots on the surface, large clumps of what
appear to be carbon nanotubes are seen. The CNT clumps consist of short, individual
tubes and what appear to be long nanofibers as shown in Figure 7-6. These fibers have
a larger diameter than nanopores, indicating they are not individual CNTs.
87
Figure 7-6: CNT growth from AAO template in clean stream of SF-CVD reactor showing (a) a wide angle view showing full CNT cluster and (b) a zoomed in portion of same cluster.
Though SEM images and Raman spectroscopy showed encouraging results
toward the nanomaterial growth being CNTs, TEM images would further improve this
standing by finding crystallographic planes and measuring the lattice spacing between
the planes. A representative TEM image is shown in Figure 7-7 where the red circles
88
show clear crystallographic planes. The lattice spacing was measured by measuring the
pixel spacing in reference to the pixel length of the 10 nm scale bar. The provided TEM
image shows few regions with clear planes making it not possible to obtain a wall count
for the CNTs. Measuring the lattice spacing indicated an average spacing of 3.30 Å,
which aligns well with the lattice spacing of 3.38 – 3.40 Å, depending on chirality,
measured by Jindal.87
Figure 7-7: TEM image of CNTs harvested from growth side of SF-CVD furnace
Upon closer examination of the tip of the one of the large fibers we discovered
that the large fibers were bundles of individual CNTs, as shown in Figure 7-8, The
89
individual tubes which compose the larger fibers are the expected diameter for CNTs
grown from a 140 nm pore diameter. Following a similar CNT bundle to its source
showed where the individual CNTs coil together and form into a single fiber, as shown
Figure 7-9. This bundling is not unusual due to the high Van der Wals forces drawing
tubes together. Some of these long fibers are as much as 1 mm in length, though they
are heavily tangled. There is also evidence that many of the shorter, thick tubes are
actually tightly coiled individual CNTs indicating there is the potential to increase their
length by preventing the coiling.
Figure 7-8: Hi-res SEM of CNT nanofiber tip showing individual nanotubes
90
Figure 7-9: Base of CNT fiber showing individual CNTs bunching together upon ejection from the AAO surface
Measuring a tall cluster of CNTs, Figure 7-10, shows approximately 820 um CNT
length. This measurement is made using short, linear, one-dimensional length
measurements. This does not fully account for the length of the CNT in
three-dimensional space, nor does it fully account for the tangled growth near the
surface.
91
Figure 7-10: Tall cluster of ULCNTs lifting away from AAO template. Cluster is ~ 200 um tall with individual CNTs over 800 um
A thorough SEM analysis of the Fe coated side of the template showed that
where we expected 40 nm pores from the anodization step down to 40 V, we were
actually seeing 140 nm pores indicative of the 140 V anodization. This was shown to be
due to the 40 nm pore layer delaminating from the 140 nm pore layer after etching
open. Various methods for switching between 140 V and 40 V were explored such as a
direct voltage ramp from 140 – 40 V, ramping the voltage from 140 – 0 V then restarting
the anodization with a ramp from 0 – 40 V, and finally a direct change from 140 – 40 V
without ramping. None of these methods were able to consistently produce a stable 40
nm pore layer that could withstand the phosphoric acid pore opening step. At this point,
we determined that we could obtain a thick enough anodization at 40 V to provide
structural stability then reduce the anodizing voltage to 20 V to provide a lower catalyst
particle diameter while still allowing the CNT to grow through the majority of the
92
template with minimal interference with the pore walls. Samples grown with the 20 nm
pores showed similar growth patterns and Raman signals.
Based on our initial analysis, we expect to need root-type CNT growth which will
allow the CNT and its corresponding catalyst particle to remain in constant
“communication” with the precursor gases. To determine if root- or tip-growth was the
dominant growth mode, energy dispersive x-ray spectroscopy (EDS) was performed on
a selection of CNT clusters. The first sample presented showed the EDS spectrum in
Figure 7-11 which shows no Fe K-α peaks indicating the catalyst remained adherent to
the substrate.
Figure 7-11: EDS spectrum showing no Fe evidence on CNT growth side
93
The TEM image, presented in Figure 7-7, provides further confirmation of the
root growth methodology, as there are no metal particles at the CNT tips. Throughout
the ULCNT growth tests, EDS scans were performed on the “clean” and “dirty” sides of
each sample before and after growth. The initial scan is to ensure that no iron is present
on the clean side of the template before CNT growth. No specific pattern was observed
in the EDS spectra with respect to pore diameter or growth time. A representative EDS
spectra of a sample with Fe present is shown in Figure 7-12.
Figure 7-12: EDS spectrum showing Fe peaks identified on ULCNT cluster
As a further comparison on the role of contact angle in the growth of ULCNTs,
we wanted to investigate how changing the wetting properties of the catalyst on the
porous template would affect the CNT growth. To do this, AAO templates with 40 nm
94
pores stepped down to 20 nm pores were coated with carbon by placing them in the
HTCAM furnace at 650 °C under C2H2 flow for 30 minutes. The pretreated templates
were then Fe coated and subject to CNT growth in the SF-CVD furnace under the same
conditions. These templates showed less CNT growth than seen in templates without
carbon pretreating, as shown in Figure 7-13, which we can attribute to the wetting of Fe
on carbon. This causes two issues where the iron to wicks into the pores creating a
much longer diffusion length required for CNT growth and creates a concave meniscus
which is unfavorable for CNT growth.
Figure 7-13: SF-CVD growth with carbon pretreated templated showed less CNT growth than without carbon pretreating
Another sample was tested using the same gas flow rates, no carbon
pretreatment, and an iron catalyst coating. This sample however was tested at room
temperature to prove the critical nature of heat to the formation of carbon breakdown
95
over the catalyst and diffusion through the catalyst. After 24 h carbon exposure, no
carbon deposition could be seen on either side of the template.
Conclusions
The SF-CVD reactor has been demonstrated for its ability to grow long CNTs
through growth in a stream of H2 and Ar, eliminating the failure mechanisms typically
seen in CVD based CNT growth. Growth occurred more favorably in samples with
smaller pores as expected due to the small pores confining smaller catalyst particles
leading to better CNTs. The growth also followed in line with the contact angle
measurements where non-wetting Fe on Al2O3 demonstrated large, regular CNT
clusters while wetting Fe on carbon-coated Al2O3 showed little CNT growth with shorter
tubes indicative of the expected concave meniscus which is unfavorable for CNT
growth. There was no correlation found between pore diameter or pretreatment on
driving tip vs. root growth in the reactor as similarly run samples would show EDS both
with and without Fe on the clean side. Nanotubes up to 800 um beyond the template
surface, therefore at least 1 mm in total length, have been synthesized. This is far below
the target length of greater than 5 mm; however, this is the first time CNT have been
synthesized in this method, thus there is a learning curve in optimizing the growth rate
and time. These CNTs do not show growth failure due to the traditional mechanism of
catalyst tip poisoning indicating the potential to utilize this method for the growth of
ULCNTs. The SF-CVD reactor is also the first reactor of its kind which can completely
eliminate surface adsorption and diffusion from the CNT growth process which shows
that CNT growth can occur from purely bulk diffusion and ejection.
96
CHAPTER 8SUMMARY AND RECOMMENDATIONS
Summary
Reducing CNT/matrix interfaces in composite materials is critical to improving the
achievable composite properties. This can be achieved by increasing the length of
individual CNTs. The SF-CVD reactor design seeks to increase the length of CNTs by
inhibiting the failure mechanisms which limit the obtainable length of CNTs synthesized
through traditional catalyzed CNT growth. Achieving ULCNT growth requires tuning the
catalyst/substrate interaction to drive high-quality growth within the nanopores. To
quantify parts of this interfacial interaction, a high-temperature contact angle
measurement system was assembled and used to measure the contact angle of a
molten metal catalyst material on a ceramic substrate. In agreement with literature
values, iron was showed to be non-wetting on alumina and wetting on graphite or
carbon-coated alumina. The role of H2 in the gas stream was investigated and showed
to inhibit the absorption of carbon in the iron catalyst particle as shown by a lack of
change in melting temperature with increased carbon exposure. Considering the
constant melting temperature, it is likely that the H2 reduces the surface of the Al2O3
causing Al and O to absorb into the iron. This would explain the mass gain after melting
and the change in contact angle without a change in melting point. Removing H2 from
the gas stream caused the Fe/C interaction to behave as expected, with an initial
increasing in wetting as C is absorbed into the Fe followed by a constant contact angle.
The melting point of the Fe in a H2 free stream was in line with literature values based
on estimated carbon content from weight gain measurements in view of the Fe-C phase
97
diagram. Pretreating the alumina substrate with carbon led to a wetting interaction
indicating the ability to tune the wetting properties in the SF-CVD reactor by pretreating
the AAO template.
This work presents the first demonstrated growth of CNTs up to 1 mm using the
SF-CVD reactor. Growth in the SF-CVD was shown to be affected by the pore diameter
of the nanoporous membrane where smaller pores showed more consistent and longer
CNT growth. Confirmation of CNT growth was achieved using both Raman
spectroscopy and TEM analysis by verifying indicating peaks and lattice spacing
respectively. Decreasing the flow rate of the carbon precursor while increasing the flow
time to maintain the same total carbon flow led to longer CNTs and larger clusters
indicating that restrictions in growth can be due to diffusion times or carbon overgrowth
in the dirty side of the furnace. There was no identifiable trend in iron appearance on the
clean side of the stream, indicative of tip growth, where some runs showed Fe on the
growth side and others did not, even with all others factors equal. Pretreating the AAO
template with carbon decreased the CNT cluster size and showed less overall CNT
growth. This coincides with the contact angle data indicating that Fe wets C-coated
Al2O3 leading to a concave contact angle which is unfavorable for CNT growth. Utilizing
this data indicates the ideal conditions for CNT growth in the SF-CVD reactor.
Future Work
There are many factors involved in the SF-CVD system that have not yet been
addressed on the path to ULCNT growth. Presently, there is no reliable method to
measure ULCNT length. This is in part due to needs SEM images which cannot
98
effectively measure in 3-D space and struggle to measure individual CNTs within the
tangled clusters. Upon determining a reliable method to measure CNT length, a
relationship between growth time and CNT length needs to be developed and the
mechanisms which limit the growth need to be explored. One potential method to
improve the measurability of the CNTs is to straighten them. As described earlier, Ni
catalyst has been shown to synthesize straighter tubes. Introducing water vapor into the
reactant gas stream has previously been shown to reduce amorphous carbon buildup.
Adding water vapor could fairly easily be explored as a method for increasing ULCNT
length. Further TEM work could be performed to determine if there is a correlation
between pore diameter, catalyst curvature, and CNT wall structure. If it were possible to
tune this system to synthesize SWNTs the application potential would be even greater.
As the Zisman plot has been deemed infeasible due to the polarity of Al2O3, we
could use Owens-Wendt or Fowkes theory to account for the polar and dispersive
components of the liquid and solid surface energies. Utilizing these mathematical
processes would require careful selection of the liquid phase and temperature, along
with some changes to the measurement to ensure accurate calculation of the interfacial
energy.
Harnessing the benefits of ULCNTs will require developing a process to collect
the synthesized CNTs. This may include needing to separate the ULCNTs to disperse
them in a composite material or could be based on harnessing the bundled fibers found
during this experimental process.
99
REFERENCE LIST
1. Iijima, S. Helical microtubules of graphitic carbon. Nature 354, 56–58 (1991). 2. George, R., Kashyap, K. T., Rahul, R. & Yamdagni, S. Strengthening in carbon
nanotube/aluminium (CNT/Al) composites. Scr. Mater. 53, 1159–1163 (2005). 3. Allaoui, A., Bai, S., Cheng, H. . & Bai, J. . Mechanical and electrical properties of
a MWNT/epoxy composite. Compos. Sci. Technol. 62, 1993–1998 (2002). 4. Ren, Z. F. et al. Synthesis of Large Arrays of Well-Aligned Carbon Nanotubes on
Glass. Science (80-. ). 282, 1105–1107 (1998). 5. Franklin, A. D. et al. Sub-10 nm Carbon Nanotube Transistor. Nano Lett. 12, 758–
762 (2012). 6. Bianco, A., Kostarelos, K. & Prato, M. Applications of carbon nanotubes in drug
delivery. Curr. Opin. Chem. Biol. 9, 674–679 (2005). 7. Patole, S. P., Alegaonkar, P. S., Shin, H.-C. & Yoo, J.-B. Alignment and wall
control of ultra long carbon nanotubes in water assisted chemical vapour deposition. J. Phys. D. Appl. Phys. 41, 155311 (2008).
8. Wang, X. et al. Fabrication of Ultralong and Electrically Uniform Single-Walled Carbon Nanotubes on Clean Substrates. Nano Lett. 9, 3137–3141 (2009).
9. Byung Hee Hong, † et al. Quasi-Continuous Growth of Ultralong Carbon Nanotube Arrays. (2005). doi:10.1021/JA054454D
10. Luo, C. et al. Growth mechanism of Y-junctions and related carbon nanotube junctions synthesized by Au-catalyzed chemical vapor deposition. Carbon N. Y. 46, 440–444 (2008).
11. Amama, P. B. et al. Role of Water in Super Growth of Single-Walled Carbon Nanotube Carpets. Nano Lett. 9, 44–49 (2009).
12. Stadermann, M. et al. Mechanism and Kinetics of Growth Termination in Controlled Chemical Vapor Deposition Growth of Multiwall Carbon Nanotube Arrays. Nano Lett. 9, 738–744 (2009).
13. Hamada, N. & Sawada, S.-I. New One-Dimensional Conductors: Graphitic Microtubules. 68,
14. Mintmire, J. W., Dunlap, B. I. & White, C. T. Are fullerene tubules metallic? Phys. Rev. Lett. 68, 631–634 (1992).
15. Forró, L. & Schönenberger, C. in Carbon Nanotubes 329–391 (Springer Berlin Heidelberg, 2001). doi:10.1007/3-540-39947-X_13
100
16. Wilder, J. W. G., Venema, L. C., Rinzler, A. G., Smalley, R. E. & Dekker, C. Electronic structure of atomically resolved carbon nanotubes. Nature 391, 59–62 (1998).
17. Eatemadi, A. et al. Carbon nanotubes: properties, synthesis, purification, and medical applications. Nanoscale Res. Lett. 9, 393 (2014).
18. Jasti, R. & Bertozzi, C. R. Progress and Challenges for the Bottom-Up Synthesis of Carbon Nanotubes with Discrete Chirality. Chem. Phys. Lett. 494, 1–7 (2010).
19. Dresselhaus, M. S., Dresselhaus, G. & Saito, R. Physics of Carbon Nanotubes. Carbon N. Y. 33, 883–891 (1995).
20. Thess, A. et al. Crystalline Ropes of Metallic Carbon Nanotubes. Science (80-. ). 273, (1996).
21. Frank, Poncharal, Wang & Heer. Carbon nanotube quantum resistors. Science 280, 1744–6 (1998).
22. Fischer, J. E., Lee, R. S., Kim, H. J., Thess, A. & Smalley, R. E. Conductivity enhancement in single-walled carbon nanotube bundles doped with K and Br. Nature 388, 255–257 (1997).
23. Li, J. et al. Bottom-up approach for carbon nanotube interconnects. doi:10.1063/1.1566791 ͔
24. Dresselhaus, M. S., Dresselhaus, G., Sugihara, K., Spain, I. L. & Goldberg, H. A. in 12–34 (Springer, Berlin, Heidelberg, 1988). doi:10.1007/978-3-642-83379-3_2
25. Lourie, O. & Wagner, H. D. Evaluation of Young’s modulus of carbon nanotubes by micro-Raman spectroscopy. J. Mater. Res. 13, 2418–24 (1998).
26. Treacy, M. M. J., Ebbesen, T. W. & Gibson, J. M. Exceptionally high Young’s modulus observed for individual carbon nanotubes. Nature 381, 678–680 (1996).
27. Yu, M.-F. et al. Strength and Breaking Mechanism of Multiwalled Carbon Nanotubes Under Tensile Load. Science (80-. ). 287, (2000).
28. Dekker, C. Carbon Nanotubes as Molecular Quantum Wires. Phys. Today 52, 22–28 (1999).
29. Ajayan, P. Aligned Carbon Nanotube Arrays Formed by Cutting a Polymer R - ProQuest. Science (80-. ). 265, 1212 (1994).
30. Seyhan, A. T., Gojny, F. H., Tanoğlu, M. & Schulte, K. Critical aspects related to processing of carbon nanotube/unsaturated thermoset polyester nanocomposites. Eur. Polym. J. 43, 374–379 (2007).
31. Uddin, S. M. et al. Effect of size and shape of metal particles to improve hardness and electrical properties of carbon nanotube reinforced copper and copper alloy composites. Compos. Sci. Technol. 70, 2253–2257 (2010).
101
32. He, C. et al. An Approach to Obtaining Homogeneously Dispersed Carbon Nanotubes in Al Powders for Preparing Reinforced Al-Matrix Composites. Adv. Mater. 19, 1128–1132 (2007).
33. Shimizu, Y. et al. Multi-walled carbon nanotube-reinforced magnesium alloy
composites. Scr. Mater. 58, 267–270 (2008). 34. Feng, Y., Yuan, H. L. & Zhang, M. Fabrication and properties of silver-matrix
composites reinforced by carbon nanotubes. Mater. Charact. 55, 211–218 (2005). 35. Spitalsky, Z., Tasis, D., Papagelis, K. & Galiotis, C. Carbon nanotube–polymer
composites: Chemistry, processing, mechanical and electrical properties. Prog. Polym. Sci. 35, 357–401 (2010).
36. Zhang, X. et al. Self-organized arrays of carbon nanotube ropes. Chem. Phys. Lett. 351, 183–188 (2002).
37. Termeh Yousefi, A., Bagheri, S., Shinji, K., Rusop Mahmood, M. & Ikeda, S. Highly oriented vertically aligned carbon nanotubes via chemical vapour deposition for key potential application in CNT ropes. Mater. Res. Innov. 19, 212–216 (2015).
38. Ren, Y., Li, F., Cheng, H.-M. & Liao, K. T ension–tension fatigue behavior of unidirectional single-walled carbon nanotube reinforced epoxy composite. Carbon N. Y. 41, 2159–2179 (2003).
39. Zhao, H. et al. Carbon nanotube yarn strain sensors. Nanotechnology 21, 305502 (2010).
40. José‐Yacamán, M., Miki‐Yoshida, M., Rendón, L. & Santiesteban, J. G. Catalytic growth of carbon microtubules with fullerene structure. Appl. Phys. Lett. 62, 657–659 (1993).
41. Zhang, R. et al. Growth of Half-Meter Long Carbon Nanotubes Based on Schulz–Flory Distribution. ACS Nano 7, 6156–6161 (2013).
42. Choi, Y. C. et al. Controlling the diameter, growth rate, and density of vertically aligned carbon nanotubes synthesized by microwave plasma-enhanced chemical vapor deposition. Appl. Phys. Lett. Appl. Phys. Lett. 76, (2000).
43. Yiming Li et al. Growth of Single-Walled Carbon Nanotubes from Discrete Catalytic Nanoparticles of Various Sizes. (2001). doi:10.1021/JP012085B
44. Huang, Z. P. et al. Effect of nickel, iron and cobalt on growth of aligned carbon nanotubes. Appl. Phys. A Mater. Sci. Process. 74, 387–391 (2002).
45. Yajun Tian et al. In Situ TA-MS Study of the Six-Membered-Ring-Based Growth of Carbon Nanotubes with Benzene Precursor. (2003). doi:10.1021/JA037561L
102
46. Azam, M. A., Fujiwara, A. & Shimoda, T. Thermally oxidized aluminum as catalyst-support layer for vertically aligned single-walled carbon nanotube growth using ethanol. Appl. Surf. Sci. 258, 873–882 (2011).
47. Kumar, M. & Ando, Y. Chemical Vapor Deposition of Carbon Nanotubes: A Review on Growth Mechanism and Mass Production. J. Nanosci. Nanotechnol. 10, 3739–3758 (2010).
48. Maruyama, S., Kojima, R., Miyauchi, Y., Chiashi, S. & Kohno, M. Low-temperature synthesis of high-purity single-walled carbon nanotubes from alcohol. Chem. Phys. Lett. 360, 229–234 (2002).
49. Nerushev, O. A. et al. Particle size dependence and model for iron-catalyzed growth of carbon nanotubes by thermal chemical vapor deposition. J. Appl. Phys. J. Chem. Phys. Appl. Phys. Lett. 93, 2775–3282 (2003).
50. Morjan, R. E. et al. Growth of carbon nanotubes from C60. Appl. Phys. A 78, 253–261 (2004).
51. Zhang, G. et al. Ultra-high-yield growth of vertical single-walled carbon nanotubes: Hidden roles of hydrogen and oxygen.
52. Liu, B. et al. Importance of Oxygen in the Metal-Free Catalytic Growth of Single-Walled Carbon Nanotubes from SiO x by a Vapor-Solid-Solid Mechanism. doi:10.1021/ja107855q
53. Escobar, M. et al. Synthesis of carbon nanotubes by CVD: Effect of acetylene pressure on nanotubes characteristics. (2007). doi:10.1016/j.apsusc.2007.07.044
54. Nishimura, K., Okazaki, N., Pan, L. & Nakayama, Y. In Situ Study of Iron Catalysts for Carbon Nanotube Growth Using X-Ray Diffraction Analysis. Jpn. J. Appl. Phys. 43, L471–L474 (2004).
55. Wang, Y. et al. Comparison study of catalyst nanoparticle formation and carbon nanotube growth: Support effect. J. Appl. Phys. 101, 124310 (2007).
56. Kukovitsky, E. F., L ’vov, S. G., Sainov, N. A., Shustov, V. A. & Chernozatonskii, L. A. Correlation between metal catalyst particle size and carbon nanotube growth.
57. Sinnott, S. B. et al. Model of carbon nanotube growth through chemical vapor deposition. Chem. Phys. Lett. 315, 25–30 (1999).
58. Kukovitsky, E. F., L’vov, S. G., Sainov, N. A., Shustov, V. A. & Chernozatonskii, L. A. Correlation between metal catalyst particle size and carbon nanotube growth. Chem. Phys. Lett. 355, 497–503 (2002).
59. Helveg, S., Lopez-Cartes, C., Sehested, J. & Hansen, P. Atomic-scale imaging of carbon nanofibre growth. Nature (2004).
103
60. Louchev, O. A., Laude, T., Sato, Y. & Kanda, H. Diffusion-controlled kinetics of carbon nanotube forest growth by chemical vapor deposition. J. Chem. Phys. 118, (2003).
61. Wagner, R. & Ellis, W. Vapor‐liquid‐solid mechanism of single crystal growth. Appl. Phys. Lett. (1964).
62. Moisala, A., Nasibulin, A. G. & Kauppinen, E. I. The role of metal nanoparticles in the catalytic production of single-walled carbon nanotubes—a review. J. Phys. Condens. Matter J. Phys. Condens. Matter 15, 3011–3035 (2003).
63. Yoshida, H., Takeda, S., Uchiyama, T., Kohno, H. & Homma, Y. Atomic-Scale In-situ Observation of Carbon Nanotube Growth from Solid State Iron Carbide Nanoparticles. Nano Lett. 8, 2082–2086 (2008).
64. Hofmann, S., Csanyi, G., Ferrari, A. & Payne, M. Surface diffusion: the low activation energy path for nanotube growth. Phys. Rev. (2005).
65. Baker, R., Barber, M., Harris, P. & Feates, F. Nucleation and growth of carbon deposits from the nickel catalyzed decomposition of acetylene. J. Catal. (1972).
66. Li, Q. et al. Sustained Growth of Ultralong Carbon Nanotube Arrays for Fiber Spinning**. doi:10.1002/adma.200601344
67. Li, X. et al. Air-assisted growth of ultra-long carbon nanotube bundles. Nanotechnology 19, 455609 (2008).
68. Shibuta, Y. & Maruyama, S. Molecular dynamics simulation of formation process of single-walled carbon nanotubes by CCVD method. Chem. Phys. Lett. 382, 381–386 (2003).
69. Cottin-Bizonne, C., Barentin, C., Charlaix, É., Bocquet, L. & Barrat, J.-L. Dynamics of simple liquids at heterogeneous surfaces: Molecular-dynamics simulations and hydrodynamic description. Eur. Phys. J. E 15, 427–438 (2004).
70. Rezaei Nejad, H., Ghassemi, M., Mirnouri Langroudi, S. M. & Shahabi, A. A molecular dynamics study of nano-bubble surface tension. Mol. Simul. 37, 23–30 (2011).
71. Langroudi, S. M. M., Ghassemi, M., Shahabi, A. & Nejad, H. R. A molecular dynamics study of effective parameters on nano-droplet surface tension. J. Mol. Liq. 161, 85–90 (2011).
72. Homman, A.-A. et al. Surface tension of spherical drops from surface of tension. J. Chem. Phys. 140, 34110 (2014).
73. Li, B., Bui, K. & Akkutlu, I. Y. Capillary Pressure in Nanopores: Deviation from Young-Laplace Equation. in SPE Europec featured at 79th EAGE Conference and Exhibition (Society of Petroleum Engineers, 2017). doi:10.2118/185801-MS
74. Zhao, L. & Sahajwalla, V. Interfacial Phenomena during Wetting of Graphite/Alumina Mixtures by Liquid Iron. ISIJ Int. 43, 1–6 (2003).
104
75. Ogino, K., Nogi, K. & Yamase, O. Effects of Selenium and Tellurium on the Surface Tension of Molten Iron and the Wettability of Alumina by Molten Iron*.
76. Kapilashrami, E., Jakobsson, A., Lahiri, A. K. & Seetharaman, S. Studies of the Wetting Characteristics of Liquid Iron on Dense Alumina by the X-Ray Sessile Drop Technique.
77. Domagala, R. F. & Heckenbach, E. High Temperature Furnace for Melting Point Determination. Cit. Rev. Sci. Instruments 35, (1964).
78. Allen, R. D. Techniques for Melting-Point Determination on an Electrically Heated Refractory Metal. Nature 193, 769–770 (1962).
79. Adamson, A. W. & Gast, A. P. Physical Chemistry of Surfaces. 80. Brillo, J. & Egry, I. Surface tension of Nickel, Copper, Iron and their Binary Alloys.
J. Mater. Sci. 40, 2213–2216 (2005). 81. GmbH, K. So You Want to Measure Surface Energy? A tutorial designed to
provide basic understanding of the concept of solid surface energy, and its many complications. (1999).
82. Morcos, I. Surface Tension of Stress-Annealed Pyrolytic Graphite. J. Chem. Phys. 57, 1801–1089 (1972).
83. Ooi, N., Rairkar, A. & Adams, J. B. Density functional study of graphite bulk and surface properties. (2005). doi:10.1016/j.carbon.2005.07.036
84. Cheng, T., Fang, D. & Yang, Y. The temperature-dependent surface energy of ceramic single crystals. J. Am. Ceram. Soc. 100, 1598–1605 (2017).
85. Chipman, J. Thermodynamics and Phase Diagram of the Fe-C System. 86. Dresselhaus, M. S., Dresselhaus, G., Saito, R. & Jorio, A. Raman spectroscopy of
carbon nanotubes. Phys. Rep. 409, 47–99 (2005). 87. Jindal, V. K., Gupta, S. & Dharamvir, K. Bulk and Lattice Properties for Rigid
Carbon Nanotubes Materials.
105
BIOGRAPHICAL SKETCH
Gregory Chester grew up in Wadsworth, OH before moving to Boston, MA to
attend Northeastern University following his graduation. At Northeastern, Greg obtained
a Bachelor of Science in Chemical Engineering and participated in three cooperative
education programs where he was exposed to research from the start-up level working
on nanostructured electroplated thin films to highly insulating aerogel for oil and gas
applications. After college, Greg moved to Orlando, FL and began working at
Mainstream Engineering where his research focus has been on utilizing nanomaterials,
namely CNTs, to improve the mechanical, thermal, and electrical properties of
advanced materials.
Top Related