View of Nanotechnology
UNIVERSITY OF CINCINNATI
Date:______
I, ______, hereby submit this work as part of the requirements for the degree of: in:
It is entitled:
This work and its defense approved by:
Chair: ______
Synthesis of Super-Long Carbon Nanotube Arrays by
Chemical Vapor Deposition
by
Andrew J. Gorton
A thesis submitted to the graduate faculty in partial
fulfillment of the requirements for the degree of
MASTER OF SCIENCE
Department: Mechanical Industrial and Nuclear Engineering
Major: Mechanical Engineering
Committee: Dr. Mark J.Schulz, Advisor, Mechanical Engineering
Dr. Vesselin Shanov, Co-Advisor, Materials Engineering
Dr. Jay Kim, Mechanical Engineering
Dr. Randy Allemang, Mechanical Engineering
University of Cincinnati
Cincinnati, Ohio
ABSTRACT
Carbon nanotubes (CNT) that comprise large clusters referred to as carbon nanotube arrays were discovered in 1991 by Sumio Iijima. Carbon nanotubes are the strongest material known to exist and have amazing electrical and thermal properties. Chemical
Vapor Deposition (CVD) is currently the best method for producing relatively pure super-long CNT arrays over large surface areas. The research presented here focuses primarily on new methods for improving the various aspects of the synthesis process with the intent of growing very long CNT arrays over large substrate surface areas.
Experiments were conducted to optimize certain aspects of the substrate preparation and
CVD processes. In addition, amazing advances in CNT growth were obtained by combining gadolinium with iron catalyst. Research was also conducted in two areas of carbon nano-composite materials.
iii
iv ACKNOWLEDGEMENT
As I look back on my research work at the University Cincinnati, I begin to realize just how many people it took to complete this research. In a field as multi-disciplinary as nanotechnology, numerous people with very different skill sets are required to conduct this type of research. At this time I would like to thank all those who contributed to this body of work.
I would first like to thank my advisor Dr. Mark Schulz and co-advisor Dr. Vesselin
Shannov for their constant support and guidance. They were always willing to listen, and were instrumental in honing my ideas. They provided me with the opportunity to explore new research avenues, as well as participate in an array of research topics.
Without this openness I would not have gained the breadth of knowledge that I did.
My other professors that played a significant role in my education in the field of vibration and acoustics are Dr. Randy Allemang and Dr. Jay Kim. While they were not directly involved in the majority of my research, I learned much from them through their teaching as well as their willingness to entertain my endless questions in and out of class. I would also like to thank them, as well as Dr. Teik Lim for providing me with instrumentation and engineering assistance during the preliminary acoustics experiments discussed in the
Chapter 7-Future Work.
Thanks also to all my colleagues in the Smart Structures and Bio-Nanotechnology Lab for their support, inspiration and assistance during my thesis work- special thanks to Dr.
v Yun Yeo-Heung and Dr. Goutham Kirikera for their friendship and willingness to help. I would also like to thank Dr. Robert Jones, for my AFM education and his willingness to answer my questions, Mr. Jeff Simkins for assisting with thin layer
deposition as well as plasma oxidation and Mr. Ron Flenniken for his skill in depositing
numerous thin films of catalyst metals and Mr. Dale Weber, the Instrumentation
Specialist in the Chemical and Materials Engineering Department for his assistance with
TGA analysis.
The following research groups and individuals assisted in numerous parts of this thesis research: Dr. Jandro Abot and Yi Song (UC - Multiscale Material Characterization and
Composite Structures Laboratory) for allowing me to participate in their novel composite material research - Dr. Joseph Caruso, Dr. Douglas Richardson and Mr. Kirk Lokits
(UC) for their assistance with ICP Analysis - Dr. Punit Boolchand and Ping Chen (UC -
Solid State Physics and Electronics Lab) for performing the raman analysis - Dr. Sergie
Yarmolenko and Dr. Sudhere Nurella at North Carolina A&T University for providing combinatorial and layered substrates, not to mention their CNT synthesis expertise
(Center for Advanced Materials and Smart Structures)- and Dr. Wentau Xu at the
University of Kentucky for his expertise in high resolution TEM.
Last but most certainly not least, I would like to thank my parents for their unending love, support and guidance throughout all of my endeavors without them I would never have accomplished all that I have throughout my life including my long road to obtaining a masters degree in Mechanical Engineering.
vi TABLE OF CONTENTS
Chapter 1 - Introduction to Carbon Nanotubes and Their Properties
1.0 Introduction
2.0 Overview of Nanotechnology
3.0 Current Applications and Research Motivation
4.0 Overview of Carbon Nanotubes
4.1 Chirality
4.2 Mechanical Properties
4.3 Vibration
4.4 Electrical Properties
5.0 References
Chapter 2 - Carbon Nanotube Array Synthesis
1.0 Introduction
2.0 CNT Synthesis Methods
2.1 CNT Synthesis Using Arc Discharge and Laser Ablation
2.2 CNT Synthesis Using Chemical Vapor Deposition
3.0 The Physics and Chemistry of CNT Growth
3.1 The Role of Carbon Feedstock Gases
3.2 CNT Growth Termination
3.3 Growth Mode Selection
4.0 The Effects of a Hydrogen Flow Prior to Carbon Precursor Gas Injection
4.1 Motivation for Hydrogen Pre-Deposition Phase
vii 4.2 Hydrogen Pre-Deposition Experiments & Results
4.3 Discussion of Results from Hydrogen Pre-Deposition
5.0 Overview of Substrate Preparation Techniques
5.1 Typical Substrate Preparation Methodology
6.0 The Chemical Vapor Deposition (CVD) System
7.0 Summary of Conclusions regarding the CNT growth process and the Effects of
Hydrogen on CNT growth
8.0 Future Literature Review and Work
9.0 References
Chapter 3 - Overview of Experimental and Material Characterization Techniques
1.0 Material Characterization
1.1 Environmental Scanning Electron Microscope (ESEM)
1.2 Raman Spectroscopy
1.3 High Resolution Transmission Electron Microscope (HR-TEM)
1.4 Atomic Force Microscopy (AFM)
1.5 Thermo-Gravimetric Analysis (TGA)
1.6 Inductively Coupled Plasma Spectroscopy
2.0 Substrate Preparation Methods
2.1 Temescal Electron Beam Evaporation
2.2 March Plasma Oxidation System
2.3 Thermal Annealing
3.0 CVD Furnace Operation
3.1 Easy Tube 1000 User Guide
viii 4.0 Method for Quickly Measuring CNT Array Height
5.0 References
Chapter 4 - Carbon Nanotube Array Synthesis Using Iron Catalyst
1.0 Introduction
2.0 Motivation - How does Iron Function as a CNT Catalyst?
3.0 Substrate Preparation
4.0 Substrate Surface Characterization Using AFM
4.1 Characterization Process Using AFM
4.2 Results of Substrate Surface Characterization Study
4.3 Discussion of Substrate Study Results
5.0 Synthesis Using 100% Iron Catalyst Substrates
5.1 Results of Synthesis Study
5.2 Discussion of Synthesis Study Results
6.0 Fundamental Conclusions Regarding Substrate Preparation and CNT Array
Synthesis
7.0 Future Work in Substrate Preparation and Synthesis
8.0 References
ix Chapter 5 - Carbon Nanotube Array Synthesis Using Gadolinium as an Iron Catalyst Motivator
1.0 Introduction
2.0 Investigation Gadolinium as a Catalyst for CNT Growth
1.1 Motivation for Using 100% Gadolinium
1.2 Synthesis Experiments Using 100% Gadolinium
1.3 Findings of Synthesis Study Using 100% Gd Catalyst Film
2.0 Investigating Gadolinium as an Iron Catalyst Motivator
2.1 Motivation for Combining
2.2 Gd/Fe Substrate Preparation Using E-Beam Evaporation
2.3 Gd/Fe, E-Beam Substrate Surface Characterization Using AFM
2.3.1 Characterization Process Using AFM
2.3.2 Results of Gd/Fe Substrate Characterization
2.3.3 Discussion of Gd/Fe Substrate Study Results
2.4 Gd/Fe Substrates Produced Using Pulsed Laser Deposition (PLD)
2.5 Overview of Synthesis Using Gd/Fe, E-Beam Substrates
2.5.1 Synthesis Results Using Gd/Fe, E-Beam Substrates
2.5.2 Discussion of Synthesis Results Using Gd/Fe, E-Beam Substrates
2.6 Overview of Synthesis Using Gd/Fe, PLD Substrates
2.6.1 Synthesis Results Using PLD Substrates
2.6.2 Discussion of Synthesis Results Using PLD Substrates
2.7 Characterization of Gd/Fe, E-Beam Substrates
2.7.1 Overview of ESEM Study
x 2.7.1.1 Results of ESEM
2.7.1.2 Discussion of ESEM Results
2.7.2 Overview of HR-TEM Study
2.7.2.1 Results of HR-TEM Study
2.7.2.2 Discussion of HR-TEM Results
2.7.3 Overview of Thermal Gravimetric Analysis
2.7.3.1 Results of Thermal Gravimetric Analysis
2.7.3.2 Discussion of Thermal Gravimetric Analysis Results
2.7.4 Overview of Raman Spectroscopy
2.7.4.1 Results of Raman Spectroscopy
2.7.4.2 Discussion of Raman Spectroscopy Results
2.7.5 Overview of ICP Analysis
2.7.5.1 Purpose of ICP Analysis
2.7.5.1.1 ICP Sample Preparation
2.7.5.2 Estimation of Elemental Mass on a Substrate
2.7.5.3 Results and Discussion of ICP Analysis
3.0 Explanation of Why Gadolinium Improves Synthesis with Iron Catalyst
4.0 Fundamental Conclusions Regarding the use of Gadolinium as an Iron Catalyst
Motivator
5.0 Future Work in Gd/Fe Substrate Preparation and Synthesis
6.0 References
xi Chapter 6 - Applications for CNT and CNF Particles
1.0 Overview of Carbon Nano-Composites
2.0 Improving Dispersion in Carbon Nanofiber Composites
2.1 Dispersion Experiments
2.2 Results of Impedance Experiments
2.3 Discussion of Impedance Experiment Results
3.0 Nanotube Reinforced Laminant Composites (NRLC)
3.1 Producing CNT Arrays for Laminant Type Composites
3.2 Producing NRLC Specimens
3.3 Results of First NRLC Composite Sample
3.4 Current State NRLC Research
3.4.1 Technique 1: Sonication Method
3.4.2 Technique 2: Adhesive Tape Method
3.4.3 Technique3: Removal of Substrate Using Mechanical Twist
Method
4.0 Fundamental Conclusions Regarding Impedance Measurements
5.0 Future Work Regarding Impedance Measurements
6.0 References
Chapter 7 - Fundamental Conclusions
1.0 Introduction
2.0 Hydrogen Pre-Deposition Phase
3.0 Effects of Plasma Oxidation
xii 4.0 Iron Catalyst for Growing CNT Arrays
5.0 Gadolinium as a Catalyst Motivator
6.0 Impedance as a Method for Assessing Carbon Nanofiber Dispersion in an Epoxy
Chapter 8 - Future Work
1.0 Introduction
2.0 The Effects and Role of Plasma Oxidation on CNT Array Growth
3.0 Determining Whether Top-growth or Bottom-growth is Dominant
4.0 Iron Catalyst Substrates
5.0 Combination Gadolinium and Iron Substrates
5.1 Optimize Gadolinium to Iron Ratio
5.2 A New Method for Determining if Gadolinium is Encapsulated
6.0 Acoustic Waves for Improving Carbon Flux across Large Surface Areas
6.1 Motivation
6.2 Motivation
6.3 Experiment Design
6.4 Preliminary Experiments
6.5 Results
6.6 Conclusions & Future Work
7.0 Other Uses for Acoustic Excitation
8.0 References
xiii APPENDICES
I. Images of Nanotubes Grown from 100% Iron Substrates
II. ESEM & HR-TEM Images of Nanotubes Grown from 20% Gadolinium, 80% Iron Substrates
III. ICP Data Sheets
IV. UC Clean Room Layout
V. “Temescal” Electron Beam Evaporator Product Data Sheets
VI. “March” Plasma Oxidation System Product Data Sheets
VII. MATLAB Code for Computing Modes of Rectangular Tube (Needs Revision)
VIII. “Nanotechnology Store” - Price List and General Information
IX. C-Gd-Fe Ternary Phase Diagrams
xiv LIST OF FIGURES
Chapter 1 Figures
1.1: The Scale of Micro and Nano-Structures
1.2: Single wall nanotube Structure
1.3: Single and Multiwall Carbon nanotubes
1.4: Different Chiralities of Single Wall Nanotubes
1.5: Formation of Single Wall Nanotube Structure
Chapter 2 Figures
2.1: Side of CNT Array at 25,000 times Magnification
2.2: Growth Mode Selection: (a) Supersaturated catalyst particle, (b) CNT nucleation stage, (c) CNT post nucleation stage showing flux of carbon atoms
2.3: Typical Temperature Profile Including the Hydrogen Pre-Deposition Phase
2.4: Typical Gas Flow Profile Including the Hydrogen Pre-Deposition Phase
2.5: Average Height of CNT Arrays, 3 Hour Growth Time (Sub# 071406 - 1nm Thick Iron Catalyst Film)
2.6: Average Height of CNT Arrays, 3 Hour Growth Time (Sub# B - 2nm Thick Iron Catalyst Film)
2.7: Silicon/Silicon-Oxide – Purchased from wafer supplier
2.8: Aluminum film (15nm) deposited on top of SiO2 layer
2.9: Aluminum film Plasma Oxidized to Create Al2O3 layer
2.10: Iron Catalyst Film (2nm) deposited over Al2O3 layer
2.11: Annealing of Catalyst Film to break-up film into catalyst particles
2.12: CVD system for CNT growth:
2.13. Schematic of a CVD reactor system
xv Chapter 3 Figures
3.1: Environmental Scanning Electron Microscope (ESEM)
3.2: Visual Artifacts due to Electron Charging
3.3a & b: Sample Chamber of Phillips ESEM
3.4: Sample EDX Spectrum of CNT Array
3.5: JEOL JEM-2010F HRTEM
3.6: Nanosurf Easy Scan 2 AFM
3.7: TGA System
3.8: TGA Metal Sample Pan
3.9: Agilent 7500 Series ICP Mass Spectrometer
3.10: Temescal Electron Beam Evaporator
3.11: Schematic Diagram of Electron-Beam Physical Vapor Deposition System
3.12: March CS 1701 Plasma Cleaning System
3.13: UC Cleanroom Area with March Plasma System
3.14: Schematic of Plasma Oxidation System
3.15: Furnace for Annealing Substrates
3.16: Inside of Annealing Furnace
3.17: Inside of First-Nano CVD Reactor
3.18: Substrate Specimens Inside of CVD Reactor
Chapter 4 Figures
4.1: Carbon-Iron Binary Phase Diagram
4.2: Combinatorial Substrates Used to Study Optimum Deposition Thicknesses
xvi
4.3: Patterned CNTs on Silicon Substrate
4.4: AFM Image of 15nm Thick Aluminum Film - “Step 1”
4.5: Representative Topology Sections (15nm Thick Al)
4.6: Fourier Transform of Surface Topology (15nm Thick Al)
4.7: AFM Images of 15nm Thick Aluminum Film Plasma Oxidized - “Step 2” (10, 20 & 30 min Plasma Oxidation: 300W RF, 20% Oxygen & 60mtorr Vacuum)
4.8: Representative Topology Sections (10minute Oxidation of Al)
4.9 Fourier Transform of Surface Topology (10minute Oxidation of Al)
4.10: Representative Topology Sections (20minute Oxidation of Al)
4.11: Fourier Transform of Surface Topology (20minute Oxidation of Al)
4.12 Fourier Transform of Surface Topology (20minute Oxidation of Al)
4.13 Fourier Transform of Surface Topology (30minute Oxidation of Al)
4.14: AFM Images of 2nm Thick Iron Catalyst Film on top of Alumina - “Step 3” (Iron Catalyst Film Deposited using E-Beam Evaporation)
4.15: Representative Topology Sections (2nm Fe, 10minute Oxidation of Al)
4.16: Fourier Transform of Surface Topology (2nm Fe, 10minute Oxidation of Al)
4.17 Representative Topology Sections (2nm Fe, 20minute Oxidation of Al)
4.18: Fourier Transform of Surface Topology (2nm Fe, 20minute Oxidation of Al)
4.19: Representative Topology Sections (2nm Fe, 30minute Oxidation of Al)
4.20: Fourier Transform of Surface Topology (2nm Fe, 30minute Oxidation of Al)
4.21: AFM Images of Thermally Annealed Iron Catalyst Film - “Step 4” (Annealed for 4 hours at 400°C)
xvii 4.22 Representative Topology Sections (Annealed - 2nm Fe, 10minute Oxidation of Al)
4.23 Fourier Transform of Surface Topology (Annealed - 2nm Fe, 10minute Oxidation of Al)
4.24 Representative Topology Sections (Annealed - 2nm Fe, 20minute Oxidation of Al)
4.25: Fourier Transform of Surface Topology (Annealed - 2nm Fe, 20minute Oxidation of Al)
4.26 Representative Topology Sections (Annealed - 2nm Fe, 30minute Oxidation of Al)
4.27 Fourier Transform of Surface Topology (Annealed - 2nm Fe, 30minute Oxidation of Al)
4.28: RMS Surface Roughness
4.29: (a) & (b): CNT Array Produced Using Iron Catalyst
4.30: Array Length versus Plasma Oxidation Time (Iron Catalyst, 1 hour Carbon Deposition)
4.31: Array Length versus Plasma Oxidation Time (Iron Catalyst, 3 hour Carbon Deposition)
4.32: Low Magnification ESEM Images of Iron Catalyst CNT Arrays
4.33: High Magnification ESEM Images of Iron Catalyst CNT Arrays
Chapter 5 Figures
5.1: AFM Image of 15nm Thick Aluminum Film - “Step 1”
5.2: Representative Topology Sections (15nm Thick Al)
5.3: Fourier Transform of Surface Topology (15nm Thick Al)
5.4: AFM Images of 15nm Thick Aluminum Film Plasma Oxidized - “Step 2”
5.5: AFM Images of 2nm Thick 20% Gd / 80% Fe Catalyst on Alumina - “Step 3”
xviii 5.6: AFM Images of Thermally Annealed 20% Gd / 80% Fe Catalyst - “Step 4”
5.7: Representative Topology Sections (2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
5.8: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
5.9: Representative Topology Sections (2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
5.10: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
5.11: Representative Topology Sections (2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
5.12: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
5.13: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
5.14: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
5.15: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
5.16: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
5.17: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
5.18: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
5.19: RMS Surface Roughness at Different Oxidation Times
5.20: Top View of Combinatorial Substrate #149
5.21: PLD Steps Used to Create Combinatorial Substrate #149
5.22: Left Side View of Combinatorial Substrate #149
xix
5.23: Layered Substrate of Iron and Gadolinium Produced Using PLD
5.24: Photographs of CNT Arrays Produced Using 20%Gd, 80%Fe Substrates
5.25: Growth Length Compared to Plasma Oxidation Time (1 hour carbon Deposition)
5.26: Growth Length Compared to Plasma Oxidation Time (3 hour Carbon Deposition)
5.27: Photos of Combinatorial Substrate #149 after CNT Synthesis
5.28: Photos of Layered Substrate #153 after CNT Synthesis (Fe-3 nm, Gd-1nm - Fe 75%, Gd 25%)
5.29: Photos of Layered Substrate #154 after CNT Synthesis (Fe-3 nm, Gd-2nm - Fe 60%, Gd 40%)
5.30: Photos of Layered Substrate #155 after CNT Synthesis (Fe-3 nm, Gd-3nm - Fe 50%, Gd 50%)
5.31: Photos of Layered Substrate #156 after CNT Synthesis (Fe-4 nm, Gd-1nm - Fe 80%, Gd 20%
5.32: Substrate Patterned with Fe/Gd Catalyst
5.33: CNT Array (>4mm) produced from Gd/Fe Substrates
5.34: CNT Array (>4mm) produced from Gd/Fe Substrates
5.35: Close-Up of CNT Array in shown in Figure 5.20
5.36: EDS of Fe/Gd CNT Array Surface
5.37: HRTEM Images of Catalyst Particles Encapsulated by a MWCNT
5.38: Line EDS Analysis across Gd/Fe Catalyst Particles (“Low” Magnification)
5.39: Line EDS Analysis across Gd/Fe Catalyst Particles (“Medium” Mag.)
5.40: Line EDS Analysis across Gd/Fe Catalyst Particles (“High” Mag.)
5.41: Test 1 - TGA plot of 20% Gd / 80% Fe substrate (Oxidation in Air)
5.42: Test 1 – Derivative of Function Plotted in Figure 5.27
xx 5.43: Test 2 - TGA plot of 20% Gd / 80% Fe substrate (Oxidation in Air)
5.44: Test 2 – Derivative of Function Plotted in Figure 5.29
5.45: Raman Spectrum - Side of CNT Array
5.46: Raman Spectrum - Top of CNT Array
5.47 Binary Phase Diagram for Iron and Gadolinium
Chapter 6 Figures
6.1: Illustration of the CNF
6.2: Illustration of Dispersion and Distribution
6.3: Glove Box with Sonicator and Mixer
6.4: Air Bubbles in Clear Epoxy Resin
6.5: Impedance as a Function of Sonication Time (sonicator pow er setting =30)
6.6: Resistance as a Function of Sonication Time (sonicator power setting =30)
6.7: Inductance as a Function of Sonication Time (sonicator power setting =30)
6.8: Capacitance as a Function of Sonication Time (sonicator power setting =30)
6.9: Impedance of Three Different CNF/Epoxy Mixtures
6.10: Typical Carbon Fiber Fabric
6.11: Schematic Cross-Section of NRLC Specimen
6.12: Edge of Cut Through the CNT/Carbon Fiber Sheet
6.13: Exposed CNTs Inside of One Crevasse (indicated by the arrow in Fig 6.12)
6.14: Surface of CNT Array Imbedded in Epoxy Resin
6.15: “Toes” of the Carbon Fiber Fabric Showing Through the Epoxy Impregnated CNT Array
6.16: CNT Array Attached to One Sheet of Carbon Fiber Fabric
xxi
6.17: Cross-Section of Fully Assembled NRLC Specimen
6.18: Fabrication Steps in Sonication Method
6.19: CNT Array Substrates on Fabric Inside of Vacuum Bag
6.20: Schematic of Sonicator Set-up
6.21: CNT Array Bonded to Fabric Ply After Substrate Removal
6.22: Fabrication Steps in “Adhesive Tape Method”
6.23: Wafer Removal Steps in “Adhesive Tape Method”
6.24: Fabrication Steps in “Mechanical Twist Method”
6.25: Wafer Removal Steps in “Mechanical Twist Method”
Chapter 7 Figures
No Figures
Chapter 8 Figures
8.1: 1in2 CNT Array with Decreased Thickness at the Center
8.2: Schematic of Carbon Flux across Substrate Surface
8.3: Concept for Mixing Chambers (Cross Section thru CVD Furnace Tube)
8.4: Stages of Carbon Atoms Inside the CVD Reactor
8.5: Oscillation of Decomposed Carbon Atom
8.6: Schematic Indicating Benefits of Acoustic Excitation
8.7: Test Setup to Measure Acoustic Modes of Replica CVD Reactor
8.8: Schematic of Acoustic Measurement Chain
8.9: Acoustic Pressure FRF of CVD Reactor – Position 1
8.10 Acoustic Pressure FRF of CVD Reactor – Position 2
xxii 8.11 Acoustic Pressure FRF of CVD Reactor – Position 3
8.12 Acoustic Pressure FRF of CVD Reactor – Position 4
8.13 Acoustic Pressure FRF of CVD Reactor – Position 5
8.14 Acoustic Pressure FRF of CVD Reactor – Position 6
8.15 Acoustic Pressure FRF of CVD Reactor – Position 7
8.16 Acoustic Pressure FRF of CVD Reactor – Position 8
xxiii LIST OF TABLES
Chapter 1 Tables
1.1: History of Nanotubes
1.2: Mechanical Properties of Various Carbon Materials
1.3: Natural Frequencies of Individual CNTs
Chapter 2 Tables
No tables
Chapter 3 Tables
No tables
Chapter 4 Tables
4.1: Step 1 - 15nm Thick Aluminum Film
4.2: Step 2 - Oxidized Alumina Film (Al2O3)
4.3: Step 3 - Al2O3 + 2nm Fe Film, No Thermal Annealing
4.4: Step 4 - Al2O3 + 2nm Fe Film, Annealed for 5hrs at 400ºC
Chapter 5 Tables
5.1: Step 1: Surface Metrology of 15nm Thick Aluminum Film
5.2: Step 2: Surface Metrology: Oxidized Alumina Film (Al2O3)
5.3: Step 3: Surface Metrology: Al2O3 + 2nm Fe Film, No Thermal Annealing
5.4: Step 4: Surface Metrology: Al2O3 + 2nm Fe Film, Annealed for 5hrs at 400ºC
Chapters 6, 7 & 8 Tables
No tables
xxiv LIST OF EQUATIONS
Chapter 1
1.1: CNT Modulus of Elasticity
1.2: CNT Bending Stiffness
1.3: CNT Torsional Stiffness
1.4: Equation of Motion for a Simple Beam
1.5: Natural Frequencies of Simple Beam (rad/sec)
Chapter 2
No equations
Chapter 3
No equations
Chapter 4
No equations
Chapter 5
No equations
Chapter 6
6.1: Complex Impedance
Chapter 7
No equations
Chapter 8
8.1: Speed of Sound in an Ideal Gas
xxv NOMENCLATURE
Chapter 1
CNT Carbon Nanotube
nm nanometer
µm micrometer
SWCNT Single-Wall Carbon Nanotube
MWCNT Multi-Wall Carbon Nanotube
TPa Pascal x 1012
E Modulus of Elasticity
V Volume
U Strain Energy
Є Elastic Strain
KB Bending Stiffness
KT Torsional Stiffness
L length of the carbon nanotube
C curvature
θ Torsional Angle
GPa Pascal x 109
ρ Density
A Area
u Deflection
I 2nd moment of area
q(x) Distributed Load
xxvi β boundary condition equation roots
th ωi modal frequency of the i mode
kHz Frequency in Hertz x 103
E(bending) Modulus of Elasticity in Bending
Chapter 2
CVD Chemical Vapor Deposition
τd Bulk diffusion time of catalyst particel
τs Surface diffusion time of catalyst particel
MEMS Micro-Electrical Mechanical Systems
Chapter 3
SEM Scanning Electron Microscope
ESEM Environmental Scanning Electron Microscope
EDS Energy-Dispersive Spectroscopy
Torr Vacuum in millimeters of mercury
kV Voltage x 103
EDX Energy Dispersive X-Ray Spectroscopy
IR Infrared frequency range
HR-TEM High Resolution Transmission Electron Microscope
AFM Atomic Force Microscopy
xxvii TGA Thermal Gravimetric Analysis
ICP Inductively Coupled Plasma mass spectrometry
rf radio frequency
psi pounds per square inch
Chapter 4
PLD Pulsed Laser Deposition
RMS Root-Mean-Square
RF Radio Frequency
Sa Average Roughness
Sm Mean Roughness
Sq Root Mean Square Roughness
Sv Valley Roughness
Sp Peak Height
Sy Peak to Valley Height
MN
Chapter 5
Gd Gadolinium
MRI Magnetic Resonance Imaging
FDA Federal Drug Administration
DTPA diethylenetriaminepentaacetic acid
M Measured mass of CNT array inside TGA system
xxviii T Temperature of TGA system
dM Derivative of CNT array mass versus TGA system dT temperature
D-Peak Raman spectrum peak that corresponds to amorphous carbon or defects in crystalline carbon
G-Peak Raman spectrum peak that corresponds to crystalline carbon that comprises the nanotube shells
ppb parts per billion
Chapter 6
CNF Carbon Nanofiber
R Static resistance
C Capacitance
L Inductance
NRLC Nanotube Reinforced Laminant Composite
Zm Mixture Impedance
Chapter 7
No Nomenclature
Chapter 8
C Speed of sound in an ideal gas
R Ideal gas constant for a mixture
xxix T Temperature of gas mixture
γ Ratio of specific heats for a gas mixture
FFT Fast Fourier Transform
FRF Frequency Response Function
xxx
Chapter 1
Introduction to Carbon Nanotubes and Their Properties
1 1.0 Introduction
Over the past fifteen years substantial research has been done with respect to the growth and application of carbon nanotubes (CNTs). Areas of research include methods for growing bulk, un-oriented nanotubes and nanofibers from powdered catalyst and oriented arrays of nanotubes grown on silicon substrates. Substrate preparation techniques, material characterization and most importantly, the application of carbon nanotubes and nanofibers are also areas of ongoing research. Today CNTs are being used as sensors, actuators, and additives for composite reinforcement and a host of other applications.
The “Holy Grail” of the nanotube community is to develop a method in which the growth process can be carried out indefinitely, such that the length of the nanotube is not limited by the synthesis process. If this can one day be accomplished, it will be an incredible breakthrough -- one that brings nanotubes to the forefront of engineering where applications range from the smallest structures used in medicine to the largest buildings, aircrafts, bridges and dams. The majority of the research discussed in this thesis is targeted at further developing methods to improve CNT growth characteristics and develop new applications for these materials.
2.0 Overview of Nanotechnology
Webster’s dictionary defines nanotechnology as “the art of manipulating materials on an atomic or molecular scale especially to build microscopic devices” [1]. This definition is appropriate; even going so far as to call it “art”. During my work with nanotubes, other nano-materials and processes, it became increasingly clear that there are a vast number of factors and nuances that are very difficult to control. To call it a science in the purest
2 sense, especially since exact descriptions of the behavior of nanoscale systems is extremely difficult, may be misleading. Just as with art, the nuances are the key. Science can help us to “sketch-out” the interaction of these nanoscale systems; however, the ability to effectively manipulate the science to achieve the desired outcome is on the fringe of art. When dealing with molecular and atomic systems, one must be able to manipulate, sense and visualize these system on the nanoscale -- not an easy task considering that a nanometer is one billionth of a meter.
Figure 1.1: The Scale of Micro and Nano-Structures (Figure from http://www.sc.doe.gov/bes/Scale_of_Things_07OCT03.pdf)
3 3.0 Current Applications and Research Motivation
There are numerous ways that carbon nanotubes are being utilized. Currently nanotubes are being used commercially in numerous structural systems. You may find them in the highest-end tennis rackets, bike frames, baseball bats, golf clubs, etc. Other applications include rip proof clothing and flack jackets that track information on the condition of the person wearing the jacket.
Carbon nanotubes are used in micro and nanoscale fluidic devices where fluid velocities can reach speeds up to five times faster than predicted by traditional fluid mechanics.
This is particularly useful in the design of “lab-on-a-chip” devices, which allow for chemical and biological experiments to take place on a much smaller scale, allowing for decreased analysis time and greater assay through-put. Substantial research is also being
conducted on the use of carbon nanotubes for drug delivery. CNTs provide a very useful
platform for encapsulating drugs or genes for targeted drug and gene therapy. The
primary reasons there is so much excitement over this application, is that the size of the
CNTs allows them to pass through cell membranes and even into the nucleus of a cell.
4.0 Overview of Carbon Nanotubes
Carbon nanotubes are amazing structures; especially considering they are merely a tube-
like lattice of carbon atoms; the basic element of life itself. Carbon nanotubes are the
strongest substance known to exist. They are highly electrically and thermally
conductive and have one of the highest aspect ratios of any material. Carbon nanotubes
are essentially graphite sheets rolled up to form a tube like structure [2]. Considering that
4 graphite itself is considered to be a relatively soft material, it is quite amazing that the reorganization of a single sheet of graphite can result in the world’s strongest structure.
The discovery of CNTs was a direct consequence of the discovery of fullerenes by Kroto,
Smalley and others at Rice University [3]. Smalley and other researchers speculated that single wall carbon nanotube (SWCNT) tubes having diameters on the nanoscale might also be possible. This was an astute observation given that direct observation of CNTs later confirmed that CNTs have diameter as the carbon C60 molecule, otherwise known as a “Buckyball”-the smallest fullerene which follows the isolated pentagon rule [4]. At the same time Smalley and his group were discovering carbon nanotubes were simultaneously finally discovered by Sumio Ijima and Donald Bethune [8 & 9].
Figure 1.2: Single wall nanotube Structure ( http://www.physik.uni-tuebingen.de/kern/sxm_files/res_sxm_ger.html)
Figure 1.3: Single and Multiwall Carbon nanotubes (Figure from www.images.google.com).
5 When Who Events 1970 Harry Kroto & Try to synthesize long carbon chains. Dave Walton Late Scientists around the world Buckyball was synthesized and confirmed as 1980’s C60. 1991 Sumio Iijima Discovery of multi wall carbon nanotubes.
1993 S. Iijima and T. Ichihashi Synthesis of single wall carbon nanotubes. 1996 Robert Curl, Harry Kroto Nobel prize in chemistry for the & Richard E. Smalley discovery of the “Buckyball”, C60. 1999 Samsung Flat Panel display prototype
2001 IBM The first computer circuit comprised of one carbon nanotube. Table 1.1: History of Nanotubes [12]
4.1 Chirality
Not all nanotubes have identical properties. Depending on the exact formation of the carbon lattice structure that forms the shells of the nanotube, a nanotube will exhibit different properties. There are three primary classifications of nanotubes: “arm-chair”,
“zig-zag” and “chiral”.
mixture of “armchair” (c) Chiral & zig-zag
Figure 1.4: Different Chiralities of Single Wall Nanotubes (Figure from reference 2)
6 The concept of chirality is crucial in understanding how the graphitic lattice structures of nanotubes vary. The arrangement of carbon atoms (i.e. the tube’s chirality) relates directly to the optical, magnetic, electronic and other properties of the nanotube; however the Young’s Modulus is independent of the chirality [5]. The ability to grow nanotubes of a particular chirality is especially important for electronic applications [4]. The chirality does not affect the Young’s Modulus of the CNT.
The vectors OA and OB vectors in Figure 3 define the chiral vector (Ch) and the translational vector (T) respectively. The direction of OB lies along the axial direction of the nanotube, while OA defines the circumferential direction. The unit cell of the nanotube is defined by OA and OB. The vector R defines the direction about which, the nanotube would be symmetric, and the a1 and a2 are the lattice unit vectors [2]. These vectors are important in describing the lattice structure, and in essence define the properties of the nanotube.
Figure 1.5: Formation of Single Wall Nanotube Structure (a) Unrolled lattice structure of a SWNT (b) Diagram illustrating how graphite is “rolled” to form a SWNT (Figure from reference 2)
7 4.2 Mechanical Properties
The desire to utilize the high strength of carbon nanotubes is a great motivator for a substantial amount of research. The modulus of elasticity is a good measure of the strength of the CNT due to small axial tension, axial compression, non-axial bending and torsion. The exceptional strength of the CNT comes from the in-plane covalent C-C bonds [2, 5]. For axial strains, the Modulus of Elasticity (E) is defined by Equation 1 below, where V is the volume, U is the strain energy and є is the elastic strain [5].
1 ∂ 2U E = V ∂ε 2
Equation 1.1: CNT Modulus of Elasticity
The mechanical properties of CNTs must be described differently depending on their diameter. For small diameter CNTs, mechanical properties may be calculated based on a
“thin rod” approximation; however for multi-wall carbon nanotubes (MWCNTs) which have larger diameters, shell theory must be employed. Using “tight-binding” molecular dynamic methods, total energy calculations in conjunction with measurements resulted in accurate estimation of the Modulus of Elasticity, which is in the range of 1TPa and 5 to
10% tensile strain before failure [5].
For single wall carbon nanotubes, the following mathematical relationships may be used to approximate the bending stiffness (KB) and torsional stiffness (KT), where L is the length of the CNT, C is the curvature and θ is the torsional angle.
8 d 2U K = B dC 2
Equation 1.2: CNT Bending Stiffness
1 d 2U K = T L dθ 2 Equation 1.3: CNT Torsional Stiffness
When under axial strain of more than a few percent, CNTs tend to deform in two primary modes:
(1) Compressive Strain - CNTs are prone to sideways buckling and collapse
(2) Tensile Strain – Plastic failure occurs when Stone-Wales bond rotation defects occur in the lattice structure.
The following table provides a comparison of approximate mechanical properties for different CNT arrangements, graphite and steel.
Material Type Young’s Modulus Tensile Strength Density (GPa) (GPa) (g/cm2) MWCNT 1200 150 2.6 SWCNT 1054 75 1.3 SWCNT Bundle 563 150 1.3 Graphite (in-plane) 350 2.5 2.6 Steel 208 0.4 7.8 Table 1.2: Mechanical Properties of Various Carbon Materials
9 4.3 Vibration
The vibration modes of nanotubes can be used to indirectly measure the Young’s
Modulus of the CNT [4, 6 & 7]. This method assumes that the nanotube behaves as a
“beam”. Taking advantage of the beam approximation allows one to develop the classic
equation of motion for small deflections, where ρ is the density, A is the area, u is the
deflection, I is the 2nd moment of area and q(x) is a distributed load [6].
Equation 1.4: Equation of Motion for a Simple Beam
Based on this equation and the fundamental assumptions, the natural frequencies for the
CNT can be calculated per Equation 5 below; where β is the root of the boundary
th condition equation, and ωi is the modal frequency of the i mode [6].
Equation 1.5: Natural Frequencies of Simple Beam (rad/sec)
As indicated by the natural frequencies report in Table 2, CNTs have very high natural frequencies in their first bending mode. The natural frequency tends to decrease with increasing length and increase bending rigidity. Understanding what factors influence
CNT natural frequencies is valuable when trying to design materials and micro-electro mechanical systems that utilize these vibration modes. These modes of vibration may potentially be harnessed to improve carbon flux to the catalyst nanoparticles. This is discussed further in Chapter 7.
10
Nanotube Natural Frequency CNT Length E(bending) (kHz) (µm, +/-0.05) (GPa) 1 658 5.5 32.0 +3.6 2 644 5.7 26.5 +3.1 3 791 5.0 26.3 +3.6 4 908 5.3 31.5 +3.6 5 1,420 4.6 32.1 +3.6 6 968 5.7 23.0 +3.6 Table 1.3: Natural Frequencies of Individual CNTs
4.4 Electrical Properties
The electrical properties of carbon nanotubes are some of the most exciting and widely studied since their very small size and their symmetrical structure produces interesting quantum effects, electronic and magnetic properties.
Depending on their chirality, nanotubes are either semi-conducting or metallic. Typically semi-conductors are used to produce diodes, pn junctions, and field-effect transistors; while metallic tubes are used for electron tunneling transistors. In general, heterogeneous, bent or branched nanotubes are of the greatest interest for these type of applications [2, 4
& 5].
11
5.0 References
[1] Merriam-Webster 1974, "Definition of Nanotechnology," 2009 (March, 23, 2008)
[2] Saito, R.; Dresselhaus, G.; Dresselhaus, M. S., 1999, “Physical Properties of Carbon Nanotubes”, Imperial College Press, London,
[3] Kroto, H., Heath, J., O'Brien, S., Curl, R., Smalley, R., 1985, "C60: Buckminsterfullerene". Nature 318: 162 - 163
[4] Dresselhaus, M.S., Dresselhaus, G., and Phaedon, A., 2000, "Carbon Nanotubes: Synthesis, Structure, Properties and Applications," Springer, Germany
[5] Meyyappan, M., 2005, “Carbon Nanotubes: Science and Applications”, CRC Press, Washington D.C.
[6] Qian, D., Wagner, G., Liu, W., 2002, “Mechanics of Carbon Nanotubes”, Applied Mechanics Review 55, 495-533.
[7] Gao, R., Wang, Z., Bai, Z., 2000, “Nanomechanics of Individual Carbon Nanotubes from Pyrolytically Grown Arrays”, Physical Review Letters, 65, 622-625.
[8] Iijima, Sumio; Toshinari Ichihashi,1993. "Single-Shell Carbon Nanotubes of 1-nm Diameter". Nature 363: 603–605.
[9] Bethune, D., Klang, C., Vries, M., 1993, Cobalt-Catalyst Growth of Carbon Nanotubes with Single Atomic-Layer Walls, Nature, 363, 605-607
12
Chapter 2
Carbon Nanotube Array Synthesis
13 1.0 Introduction
Clusters of millions, to billions of vertically oriented carbon nanotube “forests” often referred to carbon nanotube arrays, are of particular interest for a variety of reasons.
Each CNT is not perfectly straight, and there is substantial entanglement between the individual CNTs. When viewed at 25,000 time magnification (Figure 2.1), a CNT array resembles the entwined fibers of a thick shag carpet.
Figure 2.1: Side of CNT Array at 25,000 time Magnification
The primary parameters that control nanotube growth are the choice of: hydrocarbon feedstock gas, catalyst (typically a transition metal) and the chemical vapor deposition
(CVD) furnace settings. The transition metal catalyst particles are typically supported on a base of aluminum oxide (Al3O2), also known as alumina. In general the CNT growth process, using CVD techniques, occurs when the atomic bonds of the hydrocarbon feedstock are broken at high temperature. The decomposed hydrocarbon gas results in a
“vapor” consisting of carbon and hydrogen atoms inside the CVD furnace. Through gas
14 diffusion, the carbon atoms are deposited in the molten catalyst particles. Once the catalyst particles become saturated, CNT nucleation occurs and the nanotube begins to grow in a way similar to that of a plant growing from a seed [10]. In most CVD systems the gas diffusion occurs naturally without any mechanical mixing. This is one area where additional work should be conducted, and will be of greater interest as the CVD process is scaled-up for large scale manufacturing.
2.0 CNT Synthesis Methods
There are a number of methods for producing carbon nanotubes. These methods include
arc-discharge [1, 9], laser ablation [2, 9], and Chemical Vapor Deposition (CVD) [3].
2.1 CNT Synthesis Using Arc Discharge and Laser Ablation
Arc discharge CNTs are produced when carbon atoms are evaporated in helium plasma,
which is produced when high current is passed through a carbon anode and cathode that oppose one another. The arc-discharge method is effective at producing CNTs with few structural defects. The reason for this is the high temperature (>3000°C) plasma tends to provide the energy required to fully bond the atoms to one another. A few grams of these nanotubes can be made at one time; however the length of the tubes is typically limited to tens of microns and 5 to 30nm diameters. MWCNTs are produced by controlling the pressure of the helium and the current through the anode and cathode; production of
SWCNTs requires a metal catalyst. These CNTs are very straight and are bound together by strong van der Waals forces. The first SWCNTs were produced by using a carbon anode containing a small amount of cobalt [9, 10, 11].
15 A method invented by Richard Smalley’s research group employed an intense laser to ablate a carbon target, which also contained 0.5 atomic percentages of nickel and cobalt.
The carbon target is placed in a high temperature furnace at 1200°C. A flow of inert gas, typically argon, is used to carry the CNTs to a low-temperature collector. The SWCNTs produced are typically ropes consisting of tens of individual nanotubes [9].
Arc-discharge and laser ablation methods typically produced carbon based bi-products such as fullerenes, graphitic polyhedrons and amorphous carbon often found on the walls of the CNTs. Purification is typically accomplished through use of a nitric acid reflux technique [9].
2.2 CNT Synthesis Using Chemical Vapor Deposition
Chemical Vapor Deposition (CVD) techniques are more suitable for manufacturing scale- up than the arc-discharge or laser ablation methods. Arc-discharge does have the advantage of producing large quantities of short CNTs; which are essentially in
“powdered” form and are desirable for composite materials and other applications where the CNTs must be mixed with another material or liquid.
However, CVD methods are more suitable for scale-up for a number of reasons:
(1) Nanotube orientation can be controlled – typically vertically grown.
(2) CNT arrays can be patterned through substrate preparation.
(3) Individual CNTs can potentially be grown across large surface areas.
16 Because of these advantages, CVD methods have dominated much of the research in large-scale production and applications. Typically CVD methods use a silicon substrate, similar to those used to make micro-electronics chips. Proper substrate preparation methods can produce a substrate surface covered with millions of transition-metal catalyst nanoparticles; typically iron (Fe), cobalt (Co) and nickel (Ni). These metal nanoparticles make up the chemical reaction catalysts that produce millions of individual, vertically oriented, multi-wall, and some single-wall, CNTs.
3.0 The Physics and Chemistry of CNT Growth
The physical and chemical processes that culminate in the creation of these amazing
nano-structures are highly complex. Much research is still being done to fully understand
the factors that influence CNT growth; many questions are yet un-answered.
3.1 The Role of Carbon Feedstock Gases
The CVD process for growth of a CNT involves the introduction of a carbon precursor
gas such as methane, ethylene, acetylene, benzene, carbon monoxide or ethanol into a
high temperature “reactor” chamber. These chambers are often long tubes of quartz,
which has negligible expansion at high temperature. During the course of this research,
ethylene was used exclusively for producing Multi-Wall Carbon Nanotubes (MWCNT).
Other gases can be used to obtain varying morphology, length and purity.
The gaseous environment inside the reactor tube is typically held at temperatures between
7000 C and 10000 C, high enough to break the bonds of the carbon precursor gas
17 molecules. Higher temperatures are typically used to produce single wall CNTs, while lower temperatures typically result in multi-wall CNTs [9, 10, 11].
Decomposition of ethylene (C2H4) in a high temperature environment results in two carbon atoms and two H2 molecules. Compared with methane, which only yields one carbon atom from each methane molecule, ethylene has the advantage of providing two times more feedstock carbon atoms per molecule of the feedstock gas.
C2H4 → 2C + 2H2
As the carbon atoms are diffused through the gaseous environment, they are deposited on the surface of the molten metal catalyst nanoparticles located on the substrate surface.
Nanoparticles often exhibit drastically different properties than their macroscopic counterparts; this is true of the catalyst metals nanoparticles such as iron, which will melt well below their macroscopic melting point (1535.0°C).
After a short time, the molten metal nanoparticles become super-saturated with carbon atoms. Once this occurs, the carbon atoms begin to precipitate and form rings of carbon atoms which eventually comprise the concentric shell(s) of the nanotube [5, 6]. The thickness of the molten layer of catalyst surrounding the catalyst particle is likely controlled by the number of carbon feedstock atoms decomposed on the surface of the nano-particle. This highly disordered molten layer, made up of carbon mono-layers, helps to channel carbon atoms at a constant rate [6].
18 In addition to the carbon feedstock ethylene gas, water (H2O), hydrogen gas (H2) and argon (Ag) are also introduced into the CVD reactor during the synthesis process, primarily to counteract chemical reactions that contaminate the active catalyst particle.
Contamination resulting in deactivation of the active catalyst nanoparticle is considered to be the primary reason for CNT growth to terminate.
3.2 CNT Growth Termination
The primary reason why CNTs stop growing has to do with the deactivation of the catalyst nanoparticle. Depending on the temperature at which the CNTs are grown in the
CVD process, different mechanisms work to slowly “poison” the catalyst particle, eventually halting CNT growth. The three explanations for catalyst particle deactivation or “poisoning” are:
(1) Amorphous carbon layer which impedes carbon flux to the catalyst particle
(2) Oxidation of the catalyst particle
(3) Decreased temperature at the CNT tip during top-growth mode.
(4) Oswald Ripening
The commonly held belief is that catalyst deactivation occurs primarily as a result of a carbonaceous or amorphous carbon layer that covers the catalyst particle and impedes the flux of carbon feedstock atoms into the catalyst nanoparticle. This layer forms when excess carbon feedstock atoms combines at various locations across the catalyst particle surface. As the surface area taken up by these platelets increases, the flux of carbon into the particle decreases; finally halting CNT growth. This simple assumption tends to correlate well with the predicted deactivation rate of CNT growth using CVD
19 temperatures between 535 and 700˚C. At CVD temperatures between 700 and 900˚C factors such as [6]: (1) poisoning due to gas-phase pyrolysis and (2) diffusion of silicon
(Si) atoms from the substrate into the catalyst particle.
To remove the excess carbon forming the carbonaceous layer, which impedes the flux of carbon atoms from the feedstock and eventually results in CNT growth termination, water is continually introduced into the high temperature CVD reactor. The water is continually carried into the CVD reactor by “bubbling” argon through highly purified, low capacitance (HPLC) water. The water molecules are carried by the argon gas into the CVD reactor [6]. In the reactor, amorphous carbon (C) atoms are removed as they bond with water molecules to produce carbon dioxide (CO2) and hydrogen [20, 21].
C + 2H2O → CO2 + 2H2
A potential side effect of the water molecules decomposing into hydrogen and oxygen is that the oxygen continually oxidizes the metal catalyst nanoparticle during the growth process. To counteract this effect, additional hydrogen is introduced into the system to
assist in the removal any oxide layer that may be forming on the surface of the metal
catalyst-in the case of iron (Fe) catalyst, the layer could potentially be iron oxide (FeO2).
Another condition which may also contribute to the CNT growth termination has to do with the decreased temperature at the tip of the CNT. This would be true for CNTs undergoing top-growth mode. Growth mode selection is discussed in the following section. Louchev et al proposed a set of coupled equations aimed at studying the thermal kinetics of the single-wall CNTs growing in a high temperature, inert gas environment.
20 This analytical study indicates that various stages of CNT growth may have different temperatures. It was found that the highest temperatures occur near the substrate.
Temperatures along the length of the CNT decrease with increasing height. For CNTs undergoing top growth, this decrease in tip temperature could also play a role in catalyst
deactivation [19].
Small catalyst nanoparticles may also be deactivated as their carbon and catalyst atoms
are attracted to larger, neighboring particles. This process is known as Oswalt Ripening
and may be a new possible for emplaning the deactivation of some catalyst sites. In
addition, very large nanoparticles may also be more prone to deactivation for a couple of
reasons: (1) decreased temperature due to the particle size and (2) more carbon required
for before the particle can be saturated [22].
3.3 Growth Mode Selection
There are two primary methods for orienting the growth of a CNT; “bottom-growth” and
“top-growth”. Bottom-growth occurs when conditions are such that the metal catalyst
nano-particle remains on the surface of the silicon wafer and the carbon rings which form
the nanotube are produced on top of the particle. Conversely, top-growth occurs when
the metal catalyst nano-particle are lifted off of the silicon wafer, as the rings of carbon
atoms are formed below the nanoparticle.
The specific conditions that contribute to the growth mode selection have been
investigated by a number of researchers. It has been suggested that the primary
contributor in growth mode selection is the adhesion force between the catalyst
21 nanoparticles and the substrate surface [14]. The mechanical properties that may play a role in controlling the adhesion forces are the size of the nanoparticle, how deeply imbedded they are in the substrate surface, and if there is any connection between multiple particles. It has been verified experimentally that if the catalyst particles are interconnected, that the adhesion force is stronger and the catalyst will be more likely to remain on the surface of the substrate. If however the catalyst particles are separate from one another, then the overall adhesive force is reduced and the catalyst particle can be lifted off of the substrate surface [7].
Recent research on how the substrate affects growth mode selection has been conducted at the University of Texas, at Austin [16, 17]. These experiments compared the growth mode and growth rate differences for iron catalyst on silicon (SiO2) against iron catalyst on tantalum (Ta). The size, density and morphologies of the catalyst particles were substantially different between the two substrate types. In general the SiO2 exhibited top growth mode, while the tantalum exhibited bottom growth. The surface property which seems to control the surface adhesion forces is the contact angle between the substrate and the catalyst particle. These experiments indicate that the growth rate for bottom growth is 10 times faster than the top growth mode [17]. I contend that for growing long nanotube arrays, fast initial growth is not as important as ensuring there is sufficient carbon flux to catalyst particles buried underneath the carbon nanotubes. If the catalyst particles stay on the substrate, it is likely that carbon flux will be significantly reduced as the nanotubes grow longer; however bottom growth is desirable because it should allow
22 the nanotubes to grow straighter and will not diminish the desirable interaction between the catalyst particles and the alumina layer.
Another explanation for growth mode selection is based on an analytical model of particle kinetics involved in the nucleation of a CNT. It is proposed that the growth mode selection is dependant on the time required for carbon atoms to diffuse to the bottom of the metal nanoparticle (τd), versus the time required for the upper surface of the metal particle to become saturated with carbon atoms (τs). If the surface becomes saturated before the carbon atoms have an opportunity to diffuse and saturate the catalyst particle, the carbon atoms will begin forming carbon rings on the top surface, whereas if the carbon atoms diffuse to the bottom before the surface of the particle becomes saturated, the carbon rings will begin to form below the metal particle, thus raising it up off the substrate surface [8]. Figure 2.2 illustrates this aspect of growth mode selection.
Figure 2.2: Growth Mode Selection: (a) Supersaturated catalyst particle, (b) CNT nucleation stage, (c) CNT post nucleation stage showing flux of carbon atoms [8]
23
As carbon atoms are continually supplied to the catalytic site, the CNT continues to grow in the same fashion as it was initially nucleated; be it top or bottom growth method- armchair, zig-zag or chiral. While nucleation is not fully understood, it is known that initial CNT nucleation occurs by means of carbon flux through the molten metal catalyst particle; however this may not be the only means for carbon to reach the active catalyst particle.
It has been proposed by Louchev et al [8] that at some distance away from the catalyst particle, the surface of the CNT is highly effective at attracting carbon atoms toward the nucleation site where they combine to produce the carbon shells of the CNT. Louchev contends that the rate of carbon atom diffusion through the catalyst particle is slow with respect to the rate at which carbon atoms can be supplied to the nucleation site by surface diffusion effects [8]. Recent experiments have indicated that catalyst particles over
100nm in size are primarily fed by bulk diffusion [14], while smaller catalyst particles are equally supplied with carbon through bulk and surface diffusion means [15, 16]. Surface diffusion should become the primary source of carbon flux when thick (e.g. millimeters to centimeter range) bottom growth CNT arrays are grown over large substrate areas.
Recent research has indicated that the CNT arrays grown by the Smart Structures and
Nano-biotechnology Laboratory at the University of Cincinnati are primarily bottom growth. Additional studies are currently underway to further verify this observation.
Under normal bulk gas flow conditions, decomposed carbon precursor gases flow relatively slowly. In this case carbon atoms may not be able to penetrate into the dense
CNT forest. Instead, the only area that may benefit from surface diffuse carbon atoms is
24 along the sides of the CNT forest. To counteract this effect and provide carbon atoms more efficiently to the catalyst particles, I propose that oscillation of the carbon/gas mixture at high particle velocity centered on the nucleation site may assist the carbon atoms in being distributed across the CNT surface, helping them to penetrate deeper into the CNT forest thus promoting surface and bulk diffusion. This would potentially result in the growth of longer nanotubes with greater macroscopic structural uniformity over large surface areas. This method and preliminary experiments are summarized in
Chapter 7.
4.0 The Effects of a Hydrogen Flow Prior to Carbon Precursor Gas Injection
As previously discussed, hydrogen has the potential to remove oxide layers that form on
the iron catalyst particles. It is also possible that introducing hydrogen prior to injection
of the carbon precursor gas ethylene, may help to reduce any oxide layer that forms
before the substrate is used to produce CNT arrays.
4.1 Motivation for Hydrogen Pre-Deposition Phase
Assuming that the iron oxide layer impedes carbon atom flux to the active catalyst
particle, introducing carbon precursor gases at the same time as the hydrogen may not be
the most effective method for reducing the oxide layer and improving growth. If the two
are simultaneously introduced into the CVD furnace, it stands to reason that it would take
time before the hydrogen could reduce the oxide layer sufficiently to allow carbon atoms
to reach the catalyst particle. Introducing hydrogen before the carbon precursor gas may
25 increase the initial growth rate and increase CNT height. It was hoped that a “Hydrogen
Pre-Deposition Phase” would help to increase the overall height of the CNT arrays.
4.2 Hydrogen Pre-Deposition Experiments & Results
A preliminary study was conducted to determine if the Hydrogen Pre-Deposition Phase has any effect on CNT height. Two substrates were used for this study: one had a 1nm thick iron catalyst film and the other 2nm. Four pieces of each substrate type (total of 8 samples) were used. A baseline experiment without the Hydrogen Pre-Deposition Phase was conducted. The same CVD settings were used again, but this time hydrogen was introduced prior to ethylene being introduced into the furnace. Figures 2.3 and 2.4 illustrate the different temperature and flow rate profiles.
Figure 2.3: Typical Temperature Profile Including the Hydrogen Pre-Deposition Phase
26
Figure 2.4: Typical Gas Flow Profile Including the Hydrogen Pre-Deposition Phase
Since this was a preliminary study, the optimum hydrogen gas flow rate, furnace temperature and length of time of the Hydrogen Pre-Depositions Phase were chosen arbitrarily. A hydrogen flow rate of 200 standard cubic centimeters (sccm) was started
15 minutes prior to ethylene gas injection. The effects of different temperatures during the Pre-Deposition Phase were investigated, using constant temperatures of 500, 550, 600,
650, 700 and 750ºC.
After the synthesis experiments the height of each CNT array was measured using the procedure described in Chapter 3. Due to unknown environmental conditions probably during the substrate preparation, a couple of substrate samples exhibited uncharacteristically poor results and were excluded from the study. The remaining samples were measured and their heights averaged. The results for the two substrates are provided in Figures 2.5 & 2.6 below. The results are plotted against the baseline condition of no Hydrogen Pre-Deposition Phase.
27 1.8
1.6
1.4
1.2
1.0
0.8 Height (mm) Height 0.6
0.4 Height with no H2 reduction phase 0.2
0.0 500 550 600 650 700 750 Temperature During Hydorgen Pre-Deposition Phase (C)
Figure 2.5: Average Height of CNT Arrays, 3 Hour Growth Time (Sub# 071406 - 1nm Thick Iron Catalyst Film)
1.8
1.6
1.4
1.2
1.0
0.8 Height (mm)
0.6
0.4
0.2 Height with no H2 reduction phase
0.0 500 550 600 650 700 750 Temperature During Hydorgen Pre-Deposition Phase (C)
Figure 2.6: Average Height of CNT Arrays, 3 Hour Growth Time (Sub# B - 2nm Thick Iron Catalyst Film)
28 4.3 Discussion of Results from Hydrogen Pre-Deposition
The results clearly indicate that the Hydrogen Pre-Deposition Phase is generally effective
in increasing the height of CNT arrays above the baseline condition for all six constant
temperatures tested. The substrate having a 1nm thick catalyst film showed very clear
improvement. The optimum temperature for this substrate appears to be between 550 and
600ºC. In this temperature range it is possible to double, if not triple the length of the
CNT array. Above 600ºC, the 1nm thick catalyst film showed a steady decline in height
up to 750ºC, where improvement over the baseline condition is negligible. The 2nm
thick catalyst layer showed consistent improvement of 0.2mm above the baseline, but no
optimum temperature is clearly identifiable.
Further study of the Hydrogen Pre-Deposition Phase is required to optimize the process
with regards to the length of time used for this phase, temperature and hydrogen flow rate.
It is possible, however, that each substrate will respond differently to this process. Since
the results of this preliminary study indicated that CNT array height increased, all
subsequent synthesis attempts incorporated the Hydrogen Pre-Deposition Phase with a
constant temperature of 600ºC. In general, growth results after adopting this phase were
significantly improved.
5.0 Overview of Substrate Preparation Techniques
Substrate preparation is a critical step in ensuring that “long” carbon nanotube arrays
with proper orientation can be produced with repeatable results. There are many choices
that influence the end result of the substrate. By controlling the substrate preparation
methods, and thus the surface morphology, one can attempt to control the variables that
29 influence growth mode type, the uniformity and size of CNTs, the chemical make-up of the Alumina layer as well as patterning for use in micro-scale systems such as micro- electro mechanical systems (MEMS) devices.
Typically, 3 to 4-inch diameter polished silicon wafers are purchased with a layer of silicon dioxide (SiO2) already deposited on one or both sides. This is the base configuration upon which various deposition, oxidation and lithographic techniques can be used to control the growth mode, morphology, surface adhesion properties and patterning of the nanotube arrays. Many studies are currently under way to determine the optimum substrate design for various types of CNT growth. The research presented here is limited to only a few variations on a standard substrate, and experiments were conducted to study the effect of plasma oxidation and different catalyst types. Substrate preparation was primarily done inside the clean-room facility at the University of
Cincinnati. A typical substrate preparation scheme is illustrated below. Details about particular substrates will be provided as needed in the body of this report.
5.1 Typical Substrate Preparation Methodology
The following outlines the primary steps for preparing silicon substrates for producing super long CNT arrays over large surface areas.
Step 1:
Purchase silicon wafers of 3inch or 4inch diameter with a layer of silicon oxide (SiO2) already present on one or both sides. The silicon layer should be polished on the side
30 where the deposition will occur. The purpose of the silicon oxide layer is to prevent the carbon atoms from “leaching” to the silicon substrate. The oxidized layer provides a barrier to reduce this effect. A typical silicon wafer used in these studies is shown in
Figure 2.7. In addition, the silicon is P-type, boron doped and has a resistivity of 0.001-
0.005 Ohm-cm.
0.5µm (SiO2)
500-550µm (Si)
Figure 2.7: Silicon/Silicon-Oxide – Purchased from wafer supplier
Step 2:
A 15nm thick film of aluminum (Al) is deposited on top of the SiO2 layer using electron beam evaporation. At a thickness of 15nm, the surface profiler is able to monitor the thickness of the deposited layers accurately. The surface of the coated substrate is very shiny and looks like a mirror. The back side remains rough and is not shiny; this allow for easy identification of the aluminum coated side.
15nm (Al)
0.5µm (SiO2)
500-550µm (Si)
Figure 2.8: Aluminum film (15nm) deposited on top of SiO2 layer
31 Step 3:
Once the substrate is coated with aluminum, plasma oxidation or “etching” is used to oxidize the aluminum layer thus converting it to aluminum oxide (Al2O3) also known as
“alumina”. An overview of plasma-oxidation is provided in Chapter 3.
Originally it was believed that oxidation would increase the surface roughness and thus help separate the catalyst particles from one another by providing indentations for the catalyst particles to settle in; however the RMS surface roughness data presented in
Chapter 3, along with similar studies conducted recently, indicate that the opposite may be true. Based on the data provided in Chapter 3, the plasma oxidation actually smoothes the substrate surface.
There is another possibility, which is currently being investigated as part of this ongoing research at UC. Recent experiments with catalyst support materials other than aluminum have not produced equally good results as substrates made with alumina. This suggests that the Alumina is not passive, and may actually interact with the catalyst in ways that promote growth. While much study is necessary to make any definitive conclusions, the following are possible reasons why the alumina functions more effectively as a catalyst support layer.
The plasma oxidation process has a number of parameters that may be varied to influence the alumina layer. The primary parameters are as follows:
1. Time of exposure to the plasma environment
32 2. Power used to generate the plasma
3. Percent of oxygen used with respect to the diluting gas (typically argon)
4. Amount of vacuum produced in the chamber
The process tends to oxidize more aggressively as the exposure time, power and percentage of oxygen are increased. Decreasing the amount of vacuum also increases the resulting oxidation because the oxygen is not evacuated from the chamber as rapidly.
The role of Argon is purely for diluting the oxygen. Current substrate preparation studies with 100% oxygen are currently underway at UC.
Plasma Oxidation to Increase surface roughness for better isolation of Metal catalyst particles
<15nm (Al2O3)
0.5µm (SiO2)
500-550µm (Si)
Figure 2.9: Aluminum film Plasma Oxidized to Create Al2O3 layer
Step 4:
Once the aluminum film has been oxidized, the catalyst film may then be deposited using electron-beam evaporation. Catalyst films vary in thickness and element type; however transmission metals such as iron, nickel and cobalt are typically used. This research is limited to iron and iron combined with gadolinium, a rare-earth metal. Current research is underway to study the effect of adding other rare-earth metals to known transmission metal catalyst.
33
2nm Catalyst Film
<15nm (Al2O3)
0.5µm (SiO2)
500-550µm (Si)
Figure 2.10: Iron Catalyst Film (2nm) deposited over Al2O3 layer
Step 5:
The final step before the substrates are ready to be used for CNT growth is to thermally anneal the substrates. To accomplish this, the substrates must be placed in clean Pyrex
Petri dishes and then placed inside a high-temperature furnace dedicated to the thermal annealing process. The substrates are then heated to 400˚C for a minimum of 4 hours, but longer exposure times will not damage the substrates.
Metal Catalyst Particles for CNT Nucleation
<15nm (Al2O3)
0.5µm (SiO2)
500-550µm (Si)
Figure 2.11: Annealing of Catalyst Film to break-up film into catalyst particles
34
6.0 The Chemical Vapor Deposition (CVD) System
An EasyTube 1000 chemical vapor deposition system produced by First Nano
(Ronkonkoma, NY) is the primary means of producing oriented carbon nanotube arrays
at the University of Cincinnati’s Smart Structures and Bio-Nanotechnology lab (Figure
2.12).
The system is comprised of a quartz reactor tube, approximately 2-inches in diameter and
2-feet long (Figure 2.12b). Quartz is ideal because of its negligible thermal expansion
coefficient. The quartz reactor is encased in ceramic insulation to maintain a constant
temperature environment as well as to protect the surrounding environment from the high
temperature environment inside the reactor. Temperatures inside the CVD reactor can
range from 700-10000 C. The left end of the quartz tube is connected to a series of MTS mass-flow controllers manufactured by ElectroChem, Inc. (Woburn, MA.). There is one flow controller for each gas being injected into the system. The flow controllers are pneumatically controlled from a nitrogen (N2) supply source. Electrical controls are avoided due to the potential for some of the gases to ignite. Nitrogen flow is controlled via a remote electric valve. Since nitrogen is inert, there is no danger of the nitrogen gas igniting.
Depending on the type of deposition process, the temperature and gas flow profiles can be set using the National Instruments, Labview interface developed by First Nano (Figure
2.12c). Excess gases are discharged through an exhaust port at the right end of the
35 reactor tube. A quartz tray is used to load substrate in and out of the CVD reactor tube.
Figure 2.12d shows two substrate pieces inside of the reactor tube.
Figure 2.12: CVD system for CNT growth: (a) CVD controller and furnace, (b) Quartz tube in furnace (c) Display of process temperature and flow rates (d) Two samples on the loader in the quartz tube (e) Exhaust tube leading to primary exhaust (f) Stainless steel cap that seals the CVD furnace
36 Bubbler Heating Coil Gas Out Quartz Tube
H2 Mass Spectrometer Ar Quartz Tray Substrate: Heating Coil Silicon Wafer Mass Flow Substrate Controllers Catalyst: Fe Carbon T Ethylene (C2H4) Precursor C2H4(g) 2C(s) + 2H2(g) H2 Figure 2.13. Schematic of a CVD reactor system (Figure by Y. Yeo-Heung with permission)
7.0 Summary of Conclusions regarding the CNT growth process and the Effects of Hydrogen on CNT growth
(1) Hydrogen pre-deposition phase is generally effective in increasing the height
of CNT arrays.
(2) The use of a 600ºC hydrogen pre-deposition phase temperature seems to be
generally acceptable for most substrates.
(3) Different substrate types will respond differently to the hydrogen pre-
deposition phase. There may be one particular optimum setting for each
substrate, but in an average sense, there are probably a few key factors that
generally produce the best results.
37 8.0 Future Literature Review and Work
(1) Catalyst deactivation should be studied through additional literature and
experiments studies. Of particular is the impact of Oswalt ripening,
amorphous carbon growth and oxidation of the catalyst particle.
(2) Studies should be conducted to optimize the length of the hydrogen pre-
depositions phase and hydrogen flow rate.
(3) Studies currently underway to confirm that the substrates currently being
produced at UC are prone to bottom growth mode.
38 9.0 References
[1] Ebbesen T W and Ajayan P M 1992 Nature 358 220
[2] Thess A et al 1996 Science 273 483
[3] Tsai S H, Chao C W, Lee C L and Shih H C 1999 Appl. Phys. Lett. 74 3462
[4] Fan S, ChaplineM G, Franklin N R, Tombler T W, Cassell A M and Dai H 1999 Science 283 512
[5] Hafner, J. H., Bronikowski, M. J., and Azamian, B. R., 1998, "Catalytic Growth of Single-Wall Carbon Nanotubes from Metal Particles," Chemical Physics Letters, 296(1- 2) pp. 195-202.
[6] Puretzky, A. A., Geohegan, D. B., Jesse, S., 2005, "In Situ Measurements and Modeling of Carbon Nanotube Array Growth Kinetics during Chemical Vapor Deposition," Applied Physics A, 81(2) pp. 223.
[7] Song, I. K., Yu, W. J., and Cho, Y. S., 2004, "The Determining Factors for the Growth Mode of Carbon Nanotubes in the Chemical Vapour Deposition Process," Nanotechnology, 15(10) pp. S590-S595.
[8] Louchev, O. A., Laude, T., and Sato, Y., 2003, "Diffusion-Controlled Kinetics of Carbon Nanotube Forest Growth by Chemical Vapor Deposition," The Journal of Chemical Physics, 118(16) pp. 7622-7634.
[9] Meyyappan, M., 2005, “Carbon Nanotubes: Science and Applications”, CRC Press, Washington D.C.
[10] Dresselhaus, M.S., Dresselhaus, G., and Phaedon, A., 2000, "Carbon Nanotubes: Synthesis, Structure, Properties and Applications," Springer, Germany
[11] Saito, R.; Dresselhaus, G.; Dresselhaus, M. S., 1999, “Physical Properties of Carbon Nanotubes”, Imperial College Press, London,
[13] Wang, Y., Yao, Z., Shi, L., 2007, “Comparison Study of Catalyst Nanoparticle Formation and Carbon Nanotube Growth: Support Effect”, Journal of Applied Physics, 101, pp 124310-1 thru 8
[14] Baker, R. T. K., 1989, "Catalytic Growth of Carbon Filaments", Carbon, 27, pp. 315-323
39
[15] Lee, C.J and Park, J., 2000, “Growth Model of Bamboo-Shaped Carbon Nanotubes by Thermal Chemical Vapor Deposition”, Appl. Phys. Lett. 77, 3397. [16] Wang, Y.Y., Gupta, S., Nemanich, R. J., Liu, Z., and Qin, L.-C., 2005, “Hollow to Bamboolike Internal Structure Transition Observed in Carbon Nanotube Films”, J. Appl. Phys. 98, 014312
[17] Wang, Y.Y., Li, B., Z. Yao, Shi, L. and Ho, P.S., 2006, “Effect of Supporting Layer on Growth of Carbon Nanotubes ny Thermal Chemical Vapor Deposition”, Appl. Phys. Lett. 89, 183113.
[18] Louchev, O.A., Sato, Y., and Kanda, H., 2002, “Growth Mechanisms of Carbon Nanotube Forests by Chemical Vapor Deposition”, Appl. Phys. Lett. 80, 2752.
[19] Louchev, O. A., Kanda, H., Rosén, A., and Bolton, K., 2004, “Thermal Physics in Carbon Nanotube Growth Kinetics”, J. Chem. Phys. 121, 446.
[20] Hata, K., Futaba, D. N., Mizuno, K., and Namai, T., 2004, “Water-Assisted Highly Efficient Synthesis of Impurity-Free Single-Walled Carbon Nanotubes”, Science 306, 1362-1364.
[21] Futaba, D.N., Hata, K., Yamada, T., Mizuno, K., Yumura, M., and Iijima, S., 2005, “Kinetics of Water-Assisted Single-Walled Carbon Nanotube Synthesis Revealed by a Time-Evolution Analysis”, Phys. Rev. Lett. 95, 056104
[22] Redmond, P.L., Walter, E.C., and Brus, L.E., 2006, “Photoinduced Thermal Copper Reduction onto Gold Nanocrystals under Potentiostatic Control”, J. Phys. Chem. B 110, 25158-25162
40
Chapter 3
Overview of Experimental and Material Characterization Techniques
41 1.0 Material Characterization
Characterization is an extremely important aspect of nanotechnology research. Obtaining
high quality images of macro, micro and nanoscale materials and structures is critical to
advancing research. Information regarding the chemical species, chemical nature and
physical properties is also highly valuable information.
There are many techniques that can be employed to characterize a nanoscale material or
system. The techniques reviewed in this chapter were all used to characterize materials
produced for this research with the Smart Structures and Bio-Nano Technology group.
1.1 Environmental Scanning Electron Microscope (ESEM)
Figure 3.1: Environmental Scanning Electron Microscope (ESEM)
The Environmental Scanning Electron Microscopes (ESEM) at the University of
Cincinnati’s Characterization center (Figure 3.1) is a favored method for quickly imaging carbon nanotube arrays. It is a preferred method because little sample preparation is
42 required, and images can be obtained in about 15 minutes after closing the specimen chamber; the 15 minute delay is primarily due to the time required to evacuate the chamber.
The major benefit of the ESEM system is that the evacuated specimen chamber can be filled with ionized gas molecules which help to amplify the image signal. The gaseous environment also allows for un-coated, non-conducting samples to be imaged with minimal electron charging. This is possible because the ionized gas molecules compensate for the charging effects [1].
While charging is reduced by the ionized gases, it is still an issue that needs to be addressed in order to obtain very high quality images. The lightly colored area that appears to be raised at the edges of the CNT surfaces is the result of electron charging
(Figure 3.2). To reduce this effect, a small amount of silver paste is placed at the edge of the CNT array to allow the electrons to drain off of the sample surface. This is particularly important for imaging CNT arrays, since the silicon wafers that the CNT arrays are grown on are semi-conductors. CNT arrays are typically mounted to the rotating aluminum specimen stand with carbon tape that has adhesive on both sides.
43
Figures 3.2: Visual Artifacts due to Electron Charging
The ESEM at the University of Cincinnati’s Characterization center is equipped with a chamber housing a motorized stage with an internal camera for monitoring specimens.
The size of the chamber allows for large specimens to be inserted as indicated in Figures
3.3a & b below. Chamber pressures range from 1 to 20Torr and can be maintained by means of variable pressure limiting apertures. Image resolution of 1.2 to 1.5nm can be obtain using a Shottky hot-field emission tip as the electron source. An accelerating voltage of 10-30 KV, chamber pressure of 0.9-1.3Torr and a working distance of 8-10 mm are typically suitable for obtaining high resolution images of the CNT arrays.
Figures 3.3a & b: Sample Chamber of Phillips ESEM
44
Energy Dispersive X-Ray Spectroscopy (EDX) is available in most SEMs, and is widely used to “probe” a material surface for information regarding the chemical compositions of the specimen. The EDX method takes advantage of the X-rays emitted when a sample is irradiated by an electron beam. The energies and wavelengths of the emitted x-ray photons are associated with the chemical species present in the irradiated specimen [1].
By measuring the energies or wavelengths, the elemental spectrum can be obtained.
Figure 3.4 below is a typical EDX spectrum indicating that silicon (Si) and carbon (C) are the two primary elements detected in the sample area shown in the boxed-in area.
Element Wt% At%
C K 87.11 93.98
O K 00.59 00.48
AlK 00.58 00.28
SiK 11.35 05.24
GdL 00.37 00.03
Figure 3.4: Sample EDX Spectrum of CNT Array
1.2 Raman Spectroscopy
Raman spectroscopy measures molecular vibration energy amplitudes in the infrared (IR) frequency range of the electro-magnetic spectrum. The wavelength of the Raman spectrum typically ranged between 400-5000 cm-1 or 2-16µm. The spectrometer measures the vibration energy of the molecules in the test specimen by splitting a beam of IR light. One portion of the split beam irradiates the sample; the other “reference” beam is diverted; and its original phase and amplitude remain unchanged [1].
45 Molecular vibrations excited in the irradiated sample absorb energy of various resonant frequencies. Vibrational modes tend to be characterized as “stretching”, “bending”,
“rocking” and twisting. Raman spectroscopy is only able to excite molecular modes of vibration that are symmetrical about the center of symmetry. These modes are known as
IR-inactive.
Excited vibration modes tend to absorb IR energy at the frequencies of vibration, and as a result the spectrum of the IR beam is altered. The altered (due to absorption) IR beam is then re-combined with the diverted reference beam. Combination of the two IR beams results in positive and negative wave interference. These interferences are a direct result of molecular absorption that occurred as the irradiating beam came in contact with the test sample. By correcting for the path length difference between the two beams, and taking the Fourier Transform of the recombined IR beams, the resulting Raman spectrum is obtained. The peak amplitudes in the Raman spectrum are proportional to the number of molecules of a particular species in the path of the IR beam. Raman spectroscopy can be used for qualitative and quantitative analysis of CNT arrays [1, 2, 3 & 4].
46 1.3 High Resolution Transmission Electron Microscope (HR-TEM)
Figure 3.5: JEOL JEM-2010F HRTEM
Specifications Ancillary Equipment Imaging Modes: TEM/STEM CCD Camera Cs: 0.5mm EDAX Acquisition System PTP Resolution: 0.19nm X-Ray Detector Focused Probe: 0.2nm Double-Tilt Holder Source: Schottky Field Emission Tilt Range: ± 15°
Transmission electron microscopy has the advantage of higher magnification than SEM techniques. It is possible to effectively image down to the atomic scale. The increased resolution is primarily due to the high accelerating voltage of the electron gun; which decreases the electron wavelength and helps the electron beam penetrate through material layers. Accelerating voltages are typically in the range of 100 to 400kV. The Jeol JEM-
47 2010 has an accelerating voltage of 200kV and a probe width of 0.2nm. A Cs correction of 0.5mm helps to increase image clarity [1].
Phase contrast allows for lattice imaging, a powerful technique that allows one to view individual atoms. Phase contrast occurs when electrons of different phases interact with one another and form an image. By passing multiple electron beams through a sample, multiple lattice images are formed and can be combined to obtain a complete picture of the atomic structure [1]. Samples should be no thicker than a few hundred nanometers and are typically placed on a 3mm diameter disc. For CNT imaging, short CNTs were produced and dispersed in ethanol. The dispersed tubes were then placed on an amorphous carbon grid similar to the one shown in Figure 3.6 below.
1.4 Atomic Force Microscopy (AFM)
Figure 3.6 Nanosurf Easy Scan 2 AFM
Atomic Force Microscopy (AFM) was conducted using the Nanosurf Easy Scan 2. This particular AFM is relatively easy to operate; however getting clear images can be very difficult. AFM can be used to image the surface of CNT arrays; however the tip can get
48 caught in the surface of the array. To minimize the risk of this occurring, it is critical that the AFM be run in non-contact mode.
When imaging a non-reflective surface such as a CNT array, great care must be taken when “Advancing” toward the surface. With reflective surfaces, one can see the image of the AFM tip approach the actual tip. When the two tips get close together, you can then have the system approach, as described in the Standard Operating Procedure below.
When first becoming familiar with this piece of equipment it is recommended to use silicon substrate as the test specimen, as it is easier to see where the AFM tip is in relation to the surface.
AFM are sensitive to vibrations and acoustic energy. Most AFMs come with acoustical enclosures to reduce the chance for excessive noises that may produce image errors. A limitation of the Nanosurf system is that there is no acoustic enclosure and when located in a clean room environment which has substantial acoustic noise (typically NC50-60).
This will always be a limitation of this system when trying to obtain clear, high magnification images- This particular AFM cannot be used to obtain clear images below
1micron.
The following outlines some general notes, consideration and a typical operating procedure for the Nanosurf Easy Scan 2 AFM.
General Notes:
(1) Using 0.5-1.0 second scan rate/line yields about the same quality image.
49 (2) Use a resolution 128 for the initial scan. You can then adjust to higher scan rates once you have a reasonably clear picture of the surface.
(3) To reduce noise in the signal, you can increase the Z-axis set-point.
Notes on Imaging Silicon Substrates:
(1) Since the image of the tip can be seen clearly, you can set the tip relatively close to the surface before pushing the “Approach” button.
(2) Typical surface morphologies that would be of interest can be seen clearly from 20 X 20 micron windows down to 0.5 x 0.5 micron window
(3) Longer scan times per line does not necessarily translate into improved images. If the surface is relatively smooth, a scan time of 0.5-1.0 seconds may be acceptable for an initial pass.
(4) Points per line can be 256 or above. Perform an initial scan to identify the structures of interest and to ensure that there are no defects or contaminants within the scan area; 128 points per line may be used for the initial scan.
(5) In general start with settings that allow fast imaging of the surface. Once you have an idea of what you are going to scan, refine the settings as well as zoom in to particular areas of interest.
Standard Operating Procedure (SOP):
This procedure should be used when operating the Nanosurf Easy Scan 2 located inside the clean room area at the University of Cincinnati.
(1) Sign in using log sheet
(2) Log on to computer & turn on AFM controller; the switch is located on the back side of the long controller)
(3) Place level on camera/tip head and adjust the 3 leveling screws until level
(4) Start Easy Scan software
(5) Place sample on micrometer stage, near the edge of the orange colored head. DO NOT try to slide the sample under the head!
50 (6) Position the sample under the tip using the micrometer stage adjustments. Use the camera image on the software, as well as the naked eye to ensure that the sample can be safely positioned underneath the tip.
a. Use the “side-view” camera angle to monitor the positioning of the sample, and during tip approach
b. The camera focus is fixed, so, depending on the sample size, it may be difficult to see clearly
(7) Set the AFM mode to “Non-Contact”. This is a tapping type scan. Also set to “Dynamic”.
(8) Use the “approach” button to move the tip close to the sample surface. If the surface is reflective, you should be able to see the image of the tip. Once the image of the tip is 5-10 tip widths away from the tip itself, you may proceed with the final approach.
(9) Once the automatic approach is complete, you should see the tip scanning back and forth.
1.5 Thermo-Gravimetric Analysis (TGA)
Figure 3.7: TGA System
51 Thermal Gravimetric analysis (TGA) is used in materials engineering to measure material weight loss or gain as a function of temperature. This analysis provides information related to phase changes, oxidation or chemical reactions that result in weight loss or gain. TGA is one method used to characterize the purity of carbon nanotubes. It is well
documented that amorphous carbon burns at a lower temperature than graphitic carbon
(400°C). Therefore if high amounts of amorphous carbon or other non-crystalline carbon
are present, appreciable weight loss should occur around 700˚C. After this initial weight
loss, a second weight change should occur when the graphitic carbon comprising the
CNT walls oxidizes and converts to CO2. Identifying if there are multiple temperatures where appreciable weight loss occurs is difficult through visual inspection of the weight
loss as a function of temperature. Taking the derivative of the weight as a function of
temperature allows for easy identification of the temperature(s) where weight loss occurs.
In order for the instrument to remain relatively stable before oxidation occurs, over 1mg
of material should be contained inside the hanging sample pan (Figure 3.8); however 3-
5mg is desirable. If too little initial mass is in the system significant fluctuations in the
measured data, are observed.
Figure 3.8: TGA Metal Sample Pan
52 After the CNT array has been entirely burned away, the remaining mass should theoretically be the catalyst particles that were encapsulated by the graphitic shells; or catalyst particles that were “stuck” to the outside of the nanotubes. In either case, the total mass of all of the residual catalyst particles will be very small, most likely on the order of a few micrograms. For this reason, typical TGA equipment may not be sensitive enough to measure the catalyst weight accurately. Even if accurate measurement is possible, it will take a long time for the instrument to become stable. The sensitivity of the instrument should always be verified and reported if necessary [8, 9].
1.6 Inductively Coupled Plasma Spectroscopy
Inductively Coupled Plasma (ICP) mass spectrometry has been used as an analytical
method for identifying virtually all elements (transition and other metals, alkali and
alkaline earth elements, rare earth elements, metalloids, most halogens and some non-
metals) and in very small amounts [8].
Some of the advantages of ICP analysis include:
(a) Wide element range
(b) High sensitivity and low background levels provide extremely low detection limits; below ng/L or parts-per-million
(c) Fast analysis using a high speed quadrapole mass spectrometer
53
Figure 3.9: Agilent 7500 Series ICP Mass Spectrometer (Figure taken from Agilent ICP-MS Primer [8])
Current ICP systems in use, such as the Agilent 7500, include the following primary process stages:
(a) An aerosol spray containing the sample is produced by a component known as the Nebulizer
a. Nebulizer types include: Babington, Micro Flow and Concentric
(b) The Spray Chamber cools the aerosol spray and helps to filter out the larger particles while allowing the smaller droplets to enter the plasma torch
a. Low spray chamber temperature (~2ºC) improves sample interaction with the plasma flame
(c) The Plasma Torch is used to ionize the sample spray by stripping away one electron from each atom as well as atomizing larger sample pieces. A good plasma torch design is critical in achieving high sensitivity across the range of elements.
(d) The Plasma/Vacuum Interface is critical in extracting a representative sample of the plasma ion population and efficiently transferring it to the ion focusing, mass spectrometer and detector systems.
(e) Electrostatic plates that function as the Ion Focusing Lenses are used to focus the ionized and decomposed sample while directing it towards the
54 detector. It also helps to remove neutral species and photons that must be prohibited from reaching the detector.
(f) A Quadrapole mass spectrometer is typically used for the Mass Analyzer because of its mass range, fast scanning speeds, low cost and ease of use.
a. Quadrapole acts as a mass filter, separating ions based on their mass to charge ratio.
b. By oscillating an AC voltage at high frequency, in combination with a DC voltage, a range of mass can be allowed to reach the detector. Particles outside this mass range lose their trajectory and are not allowed to reach the detector.
c. The range of masses allowed to pass can be scanned very rapidly to obtain a mass spectrum for all elements in the sample.
(g) The Detector used in ICP analysis is largely responsible for the very high sensitivity and low background noise that makes ICP analysis so powerful.
a. Known as an “electron multiplier”, meaning that it can generate a measurable signal from the impact of a single ion.
b. A positive ion reaches the detector opening and is deflected into the first dynode, held at a negative voltage. The ion impact results in electron release from the dynode surface.
c. Each release electron strikes several successive dynode surfaces, each resulting in multiple electron release from each electron impact. Hence the multiplier effect.
2.0 Substrate Preparation Methods
The following equipment was used to produce silicon substrates for growing carbon
nanotube arrays discussed in this body of work. The electron-beam evaporator and the
March plasma oxidation system are both located in the University of Cincinnati
Engineering Research Center Cleanroom.
55 2.1 Temescal Electron Beam Evaporation
The electron-beam evaporation system is used to “sputter” or deposit thin layers of aluminum and catalyst metals onto the silicon substrates. Electron beam evaporation
(also know as e-beam evaporation) falls into the larger category of Physical Vapor
Deposition (PVD). This process is used to coat surfaces (typically silicon substrates) with other materials with other materials.
The evaporation process consists heating a block or “ingot” of source material with an electron beam. Once the source material is heated to its melting point and begins to boil, it evaporates and condenses on the substrate surface creating a thin coating
So that the molecules can evaporate freely and be dispersed to all surfaces where they condense, the chamber is evacuated [12, 13, 14].
The electron beam evaporation process typically involves the following components:
• Electron Beam Evaporation gun
• A System Controller
• Power Supply
• Crucibles for the evaporation material
• Materials for Evaporation – Aluminum and catalyst metals
• Material to be coated – Silicon Substrates
56
Figure 3.10: Temescal Electron Beam Evaporator
Figure 3.11: Schematic Diagram of Electron-Beam Physical Vapor Deposition System (http://en.wikipedia.org/wiki/Electron_beam_evaporation)
57
2.2 March Plasma Oxidation System
The March plasma oxidation system is located inside the UC ERC Cleanroom area. The system is controlled via a windows based computer interface. This system was used for plasma oxidation of all substrate used to produce nanotube arrays for this research. Other plasma systems are now available for oxidation as well as thermal treatment.
Figure 3.12: March CS 1701 Plasma Cleaning System
Figure 3.13: UC Cleanroom Area with March Plasma System
58 Plasma oxidation of substances has been used by the micro-electronics industry for some time. The same technology is also used to remove or “etch” materials from a surface, as well as clean surfaces of impurities. There are two types of etching, dry and wet. Wet etching uses chemicals to etch the surface whereas dry etching, which can also be divided into plasma and vapor phase etching, will be discussed here.
Plasma oxidation/etching is accomplished when a reactive gas is mixed with a diluting gas. The gas mixture is then subjected to high-energy radio frequency (rf), electric and
magnetic fields. The combined field ionizes the gas mixture which results in further
ionization of the gas. The result is a stable phase of ions and electrons. During plasma
oxidation/etching, the ions move towards the substrate surface. Once in contact with the
surface, the reactive ions diffuse over the surface and react with the molecules and then
diffuse back into the plasma. If the ions are highly energized, physical etching of the
surface can occur by impact momentum transfer between the ions and the molecules on
the substrate surface [13, 14 ]. This effect was observed for oxidation times longer than
10 minutes, and power setting of 300W and 400W.
TOP ELECTRODE
RF VOLTAGE PLASMA
VOLTAGE SILICON WAFER BIAS
HEATED ELECTRODE
Figure 3.14: Schematic of Plasma Oxidation System
59 2.3 Thermal Annealing
Figure 3.15: Furnace for Annealing Substrates
Annealing of the prepared substrates for CNT growth is a critical step. The annealing helps to form the Iron catalyst nanoparticles from which each CNT “sprouts”. The formation of the catalyst particles can be observed thru the increase in surface roughness of the substrate.
The furnace used to anneal the substrates can be operated as high as 1000˚C, but it has been found than an annealing temperature of 450˚C for 4 hours is optimal for growing
CNT arrays. To anneal, each piece of substrate is placed in a clean Pyrex Petri dish and covered. No markings should be used on the Petri dish, because they could contaminate the substrate surface. Even Sharpie marker marking will not remain, so if different types of substrates are being annealed at the same time, care must be taken not to confuse one substrate with another.
60
Figure 3.16: Inside of Annealing Furnace
After the substrates have cooled, they should be labeled and placed in an evacuated desiccator, which will help to minimize oxidation of the substrate surface. Oxidation of the substrate has negative effects which are discussed in greater detail in Chapters 2, 4 and 5.
3.0 CVD Furnace Operation
Figure 3.17: Inside of FirstNano CVD Reactor
61
Figure 3.18: Substrate Specimens Inside of CVD Reactor
3.1 Easy Tube 1000 User Guide
Step 1: Turn on computer by opening the right door on the front face, and pressing the black switch.
Step 2: Turn on all four gases:
Gases used in the synthesis process are located in the blue cabinets outside of the furnace room. Nitrogen is located next to the furnace. Make sure to open both the main valve on each tank and the secondary valves located on the side of the fume hood where the CVD reactor is located.
Step 3: Check to make sure that all gases have adequate pressure. The total pressure of each gas should be recorded at the time the gas cylinder is installed. Supply pressures to the system are generally between 40 & 60 psi.
Step 4: Open the correct shut-off valves on the left end of the fume hood holding the furnace.
Step 5: Double click on the “First Nano” icon on the desktop to start the Labview program
Step 6: Load the substrates making sure that they are inside of the furnace tube, and not outside the opening. Placing them too far in or out will result in substantial change in the temperature at the substrate surface because there
62 is only a very small area where the temperature is at the desired temperature.
Step 7: Close the loader
Step 8: Ensure that the gasket makes contact with the inside of the quartz tube and that the seal is maintained all the way around.
Step 9: Click on the recipe button. Select the desired recipe, and ensure that it loads correctly.
Step 10: Make any notes you desire if you are going to screen print the results screen.
Step 11: Double check that all gases have been turned on
Step 12: Close fume hood and CLICK RUN!
Step 13: If system loses power and needs to be restarted, you may have to set the “parity” on the MKS control module to “none”. This can be done by scrolling through the menu on the face-plate of the controller located in the rack spaces below the computer.
4.0 Method for Quickly Measuring CNT Array Height
This provides a simple method for quickly measuring the height of a CNT array:
Step 1: Adhere a small piece of double sided tape to the metal tip of the dial micrometer
Step 2: Using forceps, adhere the back side of the silicon substrate to the micrometer tip.
Step 3: Under low magnification, dial the micrometer in until the surface of the CNT array comes in contact with the other side of the micrometer surface.
Step 4: Record the height indicated by the micrometer. This measurement includes the thickness of the silicon substrate.
Step 5: Subtract the silicon substrate thickness and the double sided tape thickness (~ 53microns) from the total thickness. This should be an approximate measure of the CNT array height.
63 Step 6: Make multiple measurements of the same array to ensure repeatability of the measurement.
64 5.0 References
[1] Kelsall, R., Hamley, I., Geogehan, M., Nanoscale Science and Technology, Wiley, New Jersey, 2005.
[2] Lewis, I., and Edwards, H., “Handbook of Raman Spectroscopy: From the Research Laboratory to the Process Line”, Marcel Dekker, New York, 2001.
[3] Szymanski, H., “Raman Spectroscopy: Theory and Practice”, Plenum Press, New York, 1967-70.
[4] McCreery, R., “Raman Spectroscopy for Chemical Analysis”, John Wiley & Sons, New York, 2000.
[5] Meyyappan, M., 2005, “Carbon Nanotubes: Science and Applications”, CRC Press, Washington D.C.
[8] Agilent Technologies, Inc. 2005, “Inductively Coupled Plasma Mass Spectrometry: A Primer”, Publication Number 5989-3526EN, USA
[9] Thomas, R., 2004, “Practical Guide to ICP-MS”, Marcel Dekker Inc, New York
[10] Brown, M., 2001, “Introduction to Thermal Analysis“, Lluwer Academic Publishers, Massechusetts.
[11] Sorai, M., 2004, “Calorimetry & Thermal Analysis”, Wiley, England.
[12] http://www.ferrotec.com/technology/electronbeam.php
[13] Movchan, B. A. (2006). "Surface Engineering" 22 (1): 35-46.
[14] Wolfe, D.; J. Singh (2000). "Surface and Coatings Technology" 124: 142-153.
65
Chapter 4
Carbon Nanotube Array Synthesis Using Iron Catalyst
66 1.0 Introduction
A significant amount of research has been dedicated to CNT arrays grown from iron catalyst particles. In general, CNTs produced from iron catalyst have larger diameters and grow more slowly than CNTs grown from cobalt (Co) or nickel (Ni) [1]. To grow
MWCNT arrays, the CVD reactor temperature can be less than 850˚C. SWCNT synthesis requires that CVD temperatures are between 850 and 1000˚C due to the higher energy required to produce defect free SWCNTs which have very high strain energies [2].
2.0 Motivation - How does Iron Function as a CNT Catalyst?
There is much debate on the chemical state of the carbon-infused iron catalyst particles during CNT growth. In order to explain the observed thermodynamics, some scientists propose that intermediate carbide species are formed, but these have rarely been observed
[3, 4]. It is believed that catalyst particles must be in a liquid state before that may accept carbon atoms and nucleate CNT growth-because the catalyst is isolated into nanoparticles, and are not in bulk quantities, it is very likely that the catalyst particle melts at a much lower temperature.
Perez-Cabero et al have used Mossbauer and Auger spectroscopy to analyze the chemical state of iron catalyst films at different temperatures when acetylene or ethane are used as the carbon pre-cursor gas. They were able to confirm experimentally that Fe or FeO is the active center for CNT nucleation for large surface area applications [5]. In 1984
Sacco verified through X-Ray diffraction that cementite (Fe3C) is present before CNT nucleation; however this conclusion was made under a very specific set of growth
67 conditions, where alpha-iron particles were heated in a mixture of carbon pre-cursor gases [6]. These results should be verified when ethylene is used as the precursor gas, assuming that carbon concentrations are equivalent.
It was also verified experimentally that Fe/SiO2 catalyst particles were reduced to metallic iron when hydrogen treatments were conducted at 800°C. In all cases cementite, which is a ceramic material, was present and increased its concentration with reaction temperature, up to 800°C [5]. While these experiments tend to indicate that cementite is present, cementite is a ceramic material which is hard and not in liquid form-this indicates that CNT nucleation and growth cannot occur when 100% cementite is present.
There may be some intermediate chemical phase when cementite is present, but a significant amount of research on many fronts is necessary to determine conclusively what chemical species are present during CNT nucleation and growth.
In order to gain a better understanding of how iron and carbon interact in the range of
CVD temperatures, a binary phase diagram in Figure 4.1 can be used to study the state of iron and carbon at different temperatures and carbon concentrations. This binary phase diagram may be of limited value since nanoparticles typically behave differently bulk quantities of a material. In general, we expect that the melting point of iron nanoparticles would be less than what is indicated in the phase diagram below.
68
Figure 4.1: Carbon-Iron Binary Phase Diagram (http://en.wikipedia.org/wiki/Image:Phase_diag_iron_carbon.PNG)
Initially the iron particles inside the CVD reactor are at room temperature and contain no carbon. The CVD reactor is then heated to 750°C, at which time, the carbon pre-cursor gases are injected into the furnace. As the carbon precursor gases are decomposed into carbon and hydrogen, the carbon atoms are deposited on the surface of the iron particles.
It is likely that the catalyst particles will pass through the eutectic point (the lowest
melting point and percentage of carbon atoms). At bulk quantities, the eutectic point
occurs at 1,100°C and 4% carbon by mass-however where nanoparticles are concerned
the temperature range where the liquid phase occurs is expected to be lower.
As previously discussed, spectrometry studies have revealed that cementite (Fe3O) may be the state in which CNTs nucleation and growth begins; however this tends to indicate that the percentage of carbon in an active catalyst particle is greater than 6% by mass.
69 3.0 Substrate Preparation
Substrate preparation is one of the most crucial steps in producing long, uniform CNT arrays for a number of reasons:
(1) Production of catalyst chemical compounds capable of CNT nucleation.
(2) Control of catalyst particle morphology.
(3) Control of factors that tend to promote one type of growth mode.
(4) Allow CNTs to be produced over large surface areas.
For these and other reasons, it is desirable to have sufficient control over the substrate chemical composition and mechanical properties. Typical substrate preparation for iron catalyst substrates involve four primary phases:
(1) Deposition of aluminum film.
(2) Plasma oxidation of aluminum film.
(3) Deposition of iron catalyst film.
(4) Thermal annealing of substrate.
By altering some key parameters in the four primary substrate preparation steps, the resulting surface morphology can be influenced. Previous experiments have demonstrated that some of the key factors in producing long CNT arrays are:
(1) Thickness of the aluminum oxide layer.
(2) Thickness of the catalyst layer.
(3) Length of time and temperature of the thermal annealing process.
70 Combinatorial substrates produced using a Pulsed Laser Deposition (PLD) technique are useful in determining under what substrate preparation conditions oriented growth is optimal. Based on earlier experiments where CNT arrays were grown on combinatorial substrates (Figure 4.2), it was determined from that a 15nm thick film of Al203 and 2nm thick iron catalyst film produced the longest CNT arrays with minimum structural defects.
Subsequently, 15nm alumina and 2nm catalyst films were generally used to produce the
CNT arrays studied during this research.
Figure 4.2: Combinatorial Substrates Used to Study Optimum Deposition Thicknesses (Figure provided by: North Carolina A&T University)
It has been suggested that plasma oxidation used to convert the aluminum film to aluminum-oxide (Al2O3) may play a critical role in producing long CNT arrays with
71 consistent and repeatable structural features. Aluminum-oxide is a ceramic material, with a very rough, porous surface [7]. The roughness provides “wells” where isolated catalyst particles can form. The alumina layer also helps to prevent the iron from coming in contact with the silicon, resulting in reduced CNT growth due to carbon atoms “leaching” into the silicon layer [8]. It is also plausible that the alumina surface chemically interacts with the iron catalyst, improving CNT nucleation and growth.
Plasma oxidation is used in many aspects of clean-room processing. It is typically used to clean wafer surfaces; however it can also be used to oxidize materials-refer to Chapter
3 for a more information regarding plasma oxidation. For our purposes plasma oxidation was used to impart oxygen atoms to the aluminum, converting it to aluminum oxide. To dilute the reactive oxygen gas, argon is also fed into the evacuated chamber. The gas mixture is then subjected to high-energy radio-frequency, electric and magnetic fields; this produces plasma and ion transfer between the ionized gas and the substrate surface.
In addition to oxidation of the aluminum collisions between the energized plasma ions and the [something is missing here], the plasma also removes material from the substrate thus “smoothing” the surface. This change in the substrate surface has been studied using an atomic force microscope (AFM) and is presented herein. By altering the length of the time the substrate is oxidized, one can affect the relative roughness of the surface in an attempt to improve surface characteristics, resulting in better catalyst particle formation.
Characterization at each of the primary substrate preparation steps are presented below.
For a more detailed discussion refer to Chapter 2.
72 4.0 Substrate Surface Characterization Using AFM
A preliminary study of the substrate surface characteristics was conducted at each substrate preparation step used to produce the iron catalyst substrates for oriented CNT array growth. The Nanosurf Easy Scan 2 atomic force microscope (AFM) was used to scan the substrate surface after each process step.
4.1 Characterization Process Using AFM
The surface topology data obtained from the AFM scans was used to measure the surface properties of the substrate. Post processing of the topology data can provide information about the catalyst particle grain size and Root-Mean-Square (RMS) surface roughness; a commonly used metric for comparison of different substrates [1]. The measured surface topology (height) across each scan area was also exported to Excel in matrix format.
Each row of the topology matrix was Fourier transformed and the maximum and linear average for all spectra obtained. A Hanning window was used to reduce leakage effects.
In general the effects of the substrate surface are not well understood-however the plasma does changes the chemical structure of the aluminum surface. Some preliminary studies comparing synthesis with aluminum and alumina have been conducted. In general the alumina layer produces more favorable CNT growth conditions. This tends to indicate that the alumina layer is not passive. The images for Steps 1 through Step 4 are provided below and include images of substrates plasma oxidized for 10, 20 and 30 minutes.
4.2 Results of Substrate Surface Characterization Study
The following provides the results of the substrate surface topology study conducted at the four primary substrate preparation steps-these steps are covered in detail in Chapter 2.
73
Figure 4.3: Patterned CNTs on Silicon Substrate
Figure 4.4: AFM Image of 15nm Thick Aluminum Film - “Step 1”
74 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.5: Representative Topology Sections (15nm Thick Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.6: Fourier Transform of Surface Topology (15nm Thick Al)
75 10 Minute Oxidation
20 Minute Oxidation
30 Minute Oxidation
Figure 4.7: AFM Images of 15nm Thick Aluminum Film Plasma Oxidized - “Step 2” (10, 20 & 30 minute Plasma Oxidation: 300W RF, 20% Oxygen & 60mtorr Vacuum
76 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.8: Representative Topology Sections (10minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.9: Fourier Transform of Surface Topology (10minute Oxidation of Al)
77 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.10: Representative Topology Sections (20minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.11: Fourier Transform of Surface Topology (20minute Oxidation of Al)
78 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.12: Representative Topology Sections (30minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.13: Fourier Transform of Surface Topology (30minute Oxidation of Al)
79
10 Minute Oxidation
20 Minute Oxidation
30 Minute Oxidation
Figure 4.14: AFM Images of 2nm Thick Iron Catalyst Film on top of Alumina - “Step 3” (Iron Catalyst Film Deposited using E-Beam Evaporation)
80 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.15: Representative Topology Sections (2nm Fe, 10minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.16: Fourier Transform of Surface Topology (2nm Fe, 10minute Oxidation of Al)
81
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.17: Representative Topology Sections (2nm Fe, 20minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.18: Fourier Transform of Surface Topology (2nm Fe, 20minute Oxidation of Al)
82 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.19: Representative Topology Sections (2nm Fe, 30minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.20: Fourier Transform of Surface Topology (2nm Fe, 30minute Oxidation of Al)
83 10 Minute Oxidation
20 Minute Oxidation
30 Minute Oxidation
Figure 4.21: AFM Images of Thermally Annealed Iron Catalyst Film - “Step 4” (Annealed for 4 hours at 400°C)
84 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.22: Representative Topology Sections (Annealed - 2nm Fe, 10minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.23: Fourier Transform of Surface Topology (Annealed - 2nm Fe, 10minute Oxidation of Al)
85 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.24: Representative Topology Sections (Annealed - 2nm Fe, 20minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.25: Fourier Transform of Surface Topology (Annealed - 2nm Fe, 20minute Oxidation of Al)
86 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 4.26: Representative Topology Sections (Annealed - 2nm Fe, 30minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 4.27: Fourier Transform of Surface Topology (Annealed - 2nm Fe, 30minute Oxidation of Al)
87 The following surface roughness data was obtained for each of the AFM scan areas
provided above. The values obtained are based on the entire scan area; however multiple
scan areas were not analyzed. To obtain statistical information, additional studies should
be conducted and standard statistical processes applied to the overall data. Surface
roughness descriptors provided in tables 4.1, 4.2, 4.3 & 4.4 are as follows:
(a) Sa – Average Roughness
(b) Sm – Mean Roughness
(c) Sq – Root Mean Square Roughness
(d) Sv – Valley Roughness
(e) Sp – Peak Height
(f) Sy – Peak to Valley Height
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 0 minute 1.2 0.22 1.5 -5.1 9.7 15 Table 4.1: Step 1 - 15nm Thick Aluminum Film
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.38 0.21 0.46 -1.8 1.8 3.6 20 minute 0.36 0.21 0.48 -1.8 8 9.8 30 minute 0.26 0.000046 0.33 -1.4 3.9 5.2 Table 4.2: Step 2 - Oxidized Alumina Film (Al2O3)
88 Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.39 0.00014 0.52 -3.4 9.9 13 20 minute 0.39 0.21 0.49 -1.6 2.4 4 30 minute 0.49 -0.00003 0.61 -2.2 2.4 4.6 Table 4.3: Step 3 - Al2O3 + 2nm Fe Film, No Thermal Annealing
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.65 0.0000016 0.82 -3.9 4.3 8.2 20 minute 0.38 0.00012 0.5 -3.4 3.7 7.1 30 minute 0.5 -0.00019 0.63 -2.8 2.8 5.6 Table 4.4: Step 4 - Al2O3 + 2nm Fe Film, Annealed for 5hrs at 400ºC
1.75
10minute oxidation 20minute oxidation 30minute oxidation
1.5
1.25
1
0.75 RMS Roughness (nm) Roughness RMS 0.5
0.25
0 Step 1Step 2Step 3Step 4 Substrate Preperation Step
Figure 4.28: RMS Surface Roughness
89 4.3 Discussion of Substrate Study Results
Studying spatial FFT spectra in the figures above, some interesting information can be extracted.
(1) The 15nm thick aluminum layer has the highest surface-to-valley amplitude
out of all the process steps (1nm amplitude) -- a grain size of 55nm is most
prevalent.
(2) The oxidation decreases the surface-to-valley height to a maximum of
0.45nm. The grain size is generally larger than 33nm.
(3) The addition of the catalyst layer did increase the maximum amplitude, and
also shifted the grain size towards 100nm in width.
(4) The annealing process greatly increased the height of the larger catalyst
particles --- nearly 1µm maximum height for particles having a width of
approximately 200nm.
Figure 4.8 illustrates the RMS surface roughness at each of the primary substrate preparation steps. The RMS roughness of the 15nm thick aluminum film, prior to plasma oxidation is approximately 1.6nm. Plasma oxidation, even for a 10minute period, substantially decreased the surface roughness as the material is removed by high velocity ions bombarding the substrate surface. Longer exposure times to the plasma further decreases the surface roughness. The final substrate surface roughness varied between
0.5nm and 0.82nm with the 10minute oxidation resulting in the roughest surface. Recent experiments conducted at UC have confirmed these results.
5.0 Synthesis Using 100% Iron Catalyst Substrates
90
Using the substrates characterized in the previous section, CNT synthesis was conducted using the CVD techniques discussed in Chapter 2. These experiments were conducted to ascertain if the plasma oxidation time of the aluminum film has an impact on the resulting CNT height. Six small substrate pieces from each of substrates plasma oxidized for 10, 20 and 30 minutes (18 total substrate samples) were placed in the furnace and
CNT synthesis was conducted simultaneously. Synthesis experiments were conducted for one and three hour ethylene flow durations. The height of each CNT arrays from the two experiments was measured using the technique described in Chapter 3.
5.1 Results of Synthesis Study
Figure 4.29: (a) & (b): CNT Array Produced Using Iron Catalyst
The six substrates for each oxidation length were used to obtain the statistics provided in the Figure 4.10 & 4.11 box-plots. The primary features of the box-plots are: (1) the lower and upper blue lines of the box are the 25th and 75th percentiles respectively; (2) the red center line is the median or 50th percentile; (3) the upper and lower “whiskers” indicate the range of data used in the statistical analysis; (4) red plus signs above or below whiskers are considered to be outliers and are not included in the analyzed data set.
91 0.7
0.65
0.6
0.55
0.5
0.45
Height of Array (mm) 0.4
0.35
0.3
10 20 30 Oxidation Time (min)
Figure 4.30: Array Length versus Plasma Oxidation Time (Iron Catalyst, 1 hour Carbon Deposition)
2.6
2.4
2.2
2
1.8
1.6
1.4
Height of Array (mm) 1.2
1
0.8
0.6 10 20 30 Oxidation Time (min)
Figure 4.31: Array Length versus Plasma Oxidation Time (Iron Catalyst, 3 hour Carbon Deposition)
92
Figure 4.32: Low Magnification ESEM Images of Iron Catalyst CNT Arrays
Figure 4.33: High Magnification ESEM Images of Iron Catalyst CNT Arrays
93 5.2 Discussion of Synthesis Study Results
Some observations are apparent from the data presented in Figures 4.30 and 4.31
(1) Longer synthesis time results in longer CNT arrays, the minimum length for 3
hours of carbon deposition was near the maximum length achieved for a 1
hour growth period.
(2) For 1hour synthesis, the longest CNT arrays can be achieved using 30 minute
oxidation.
(3) For 3 hour synthesis, the longest CNT arrays are achieved using 10minute
oxidation time.
One possible explanation for conclusions (2) and (3) are as follows. Knowing that plasma oxidation removes, or etches material away from the surface as it oxidizes the aluminum, one can surmise that there is an optimum length of time where oxidation is sufficient to produce the desired chemical change without reducing the thickness of the aluminum to where iron “leaches” into the silicon layer below [1].
6.0 Fundamental Conclusions Regarding Substrate Preparation and CNT Array
Synthesis
(1) Plasma etching removes substrate surface material resulting in a smoother
surface.
(2) Iron catalyst particles may not be predisposed to top growth when surface
roughness is decreased
(3) Iron carbide may be present prior to CNT nucleation, but is unlikely that it is
the chemical state of the functional catalyst.
94 (4) Iron catalyst substrates do not form arrays over 2mm in height with
repeatability or with good structural uniformity - cracking and substantial
variation in length occur.
(5) Decreased roughness, leading to dense clusters of nanoparticles that remain
on the substrate surface, may promote fast initial growth, but would not result
in optimum conditions for growing long CNT arrays.
(6) Annealing process substantially increases particle grain size (width) and the
depth of the catalyst particles.
(7) Prolonged CNT growth requires that sufficient carbon flux be maintained as
the CNTs grow longer. Oxidation of the alumina layer may help to promote
carbon diffusion along the substrate surface.
7.0 Future Work in Substrate Preparation and Synthesis
(1) Develop method for determining whether top-growth mode is occurring.
Conduct experiments to assess if top-growth is the preferred option for
growing long CNT arrays.
(2) Use in-situ spectroscopy to determine the effect of the hydrogen pre-
deposition phase.
(3) Correlate experimentally surface roughness with adhesion force and how this
effects growth mode selection.
(4) Conduct addition experiments in plasma oxidation to obtain the optimal
amount of oxygen, vacuum and argon gas.
95 8.0 References
[1] Lee, C. J., Park, J., and Yu, J. A., 2002, "Catalyst Effect on Carbon Nanotubes Synthesized by Thermal Chemical Vapor Deposition," Chemical Physics Letters, 360(3-4) pp. 250-255.
[2] Dresselhaus, M.S., Dresselhaus, G., and Phaedon, A., 2000, "Carbon Nanotubes: Synthesis, Structure, Properties and Applications," Springer, Germany
[3] Lee, S.H. ; Ruckenstein, E., 1987, “Simulation of the behavior of supported metal catalysts in real reaction atmospheres by means of model catalysts”, Journal of Catalysis, 107 pp. 23.
[4] Kock, A.; Fortuin, H.; Geus, W.; 1985, “The reduction behavior of supported iron catalysts in hydrogen or carbon monoxide atmospheres,” Journal of Catalysis, 96 pp. 261.
[5] Pérez-Caber, M., Taboad, J. B., Guerrero-Rui, A., 2006, "The Role of Alpha-Iron and Cementite Phases in the Growing Mechanism of Carbon Nanotubes: A 57Fe Mössbauer Spectroscopy Study," Physical Chemistry Chemical Physics, 8(10) pp. 1230-1235.
[6] Sacco, A.; Thacker, P.; Tzyh, N.; Chiang, A.; 1984, “The initiation and growth of filamentous carbon from α-iron in H2, CH4, H2O, CO2, and CO gas mixtures,” Journal of Catalysis, 85 pp. 224.
[7] J. Geng, C. Singh, D.S. Shepard, M.S.P. Shaffer, B.F.G. Johnson, A.H. Windle, Chem. Commun. (2002) 2666.
[8] Puretzky, A. A., Geohegan, D. B., Jesse, S., 2005, "In Situ Measurements and Modeling of Carbon Nanotube Array Growth Kinetics during Chemical Vapor Deposition," Applied Physics A, 81(2) pp. 223.
96
Chapter 5
Carbon Nanotube Array Synthesis Using Gadolinium as an Iron Catalyst Motivator
97 1.0 Introduction
The rare earth metal, gadolinium (Gd) is a commonly used as a tracer element in
Magnetic Resonance Imaging (MRI). The gadolinium tracer is excited by the strong
magnetic field and tends improves MR imaging response. [1].
To improve MR images, gadolinium is typically injected into the subject near the region of the body where the images will be taken. This method of administering the gadolinium as a contrast agent has some ramifications. The Federal Drug Administration
(FDA) recently updated the required warning labels for commercially available gadolinium based contrast agents. In general, the FDA is concerned that patients with renal problems will have an increased chance of developing Nephrogenic Systemic
Fibrosis when exposed to gadolinium [2]. In addition to this concern, gadolinium as a free ion is toxic. To counter act its toxicity, gadolinium is typically combined with an organic ligand, such as diethylenetriaminepentaacetic acid (DTPA). In-spite of this, there is still some concern that the could become detached from the ligand, resulting in increased toxicity [3]. To further reduce the chances that the gadolinium will become a free ion that can come into direct contact with living tissues, encapsulating the gadolinium in an inert, non-toxic substance may help to shield living tissues from the gadolinium. carbon nanotubes (CNTs) have shown promise as drug delivery agents [4, 5]. A natural extension of this would be to use CNTs to encapsulate gadolinium for use as an MRI contrast agent, thus “shielding” living tissues from direct contact with gadolinium.
Since CNTs are considered to be non-toxic and inert, they are a good candidate for
98 delivery of toxic agents into the body. There is also potential for intracellular imaging since CNTs can be produced at short lengths (<200nm in length) that allow them to penetrate the cellular membrane with minimal toxicity to the cell [6, 7].
Previous research has indicated that liquid gadolinium Chloride (GdCl3) can be partially encapsulated by fluorinated nanotubes. These “gadonanotube” have shown 40 to 90 times better relaxation times [8]. While not fully understood at this point, it is believed that synergy between the particles of gadolinium and the carbon shells of the nanotubes improve relaxation times and thus MR imaging response.
While these results are very encouraging for use of nanotubes to encapsulate gadolinium, there are a several issues with this particular method:
(1) Fluorination requires the use of Hydrofluoric acid, which is extremely dangerous
(2) Fluorination opens the nanotubes, so that the gadolinium still has the potential to
come in contact with living tissues
(3) The gadolinium could potentially be removed from the surface of the nanotube
Complete encapsulation of gadolinium particles inside of nanotubes without the use of dangerous chemicals would be a step forward in developing an improved method for producing these particles while further reducing the exposure of living tissues to gadolinium.
It has been shown that iron catalyst particles are completely encapsulated by the carbon shells of the CNT-this has also been verified through TEM images of CNTs produced during this research [9, 10, 11]. If Gd alone, or in conjunction with a functional catalyst
99 could produce CNTs that encapsulate gadolinium particles, this may provide a convenient method for encapsulating gadolinium without the use of dangerous, costly or time consuming processes. The initial work presented in this chapter outlines a preliminary method for producing large areas comprising billions of CNTs, each with a small amount of gadolinium encapsulated inside. This work is intended to lay a foundation for producing CNT particles safely and quickly, that could be easily prepared for delivery into the body and enhanced imaging.
1.0 Investigation Gadolinium as a Catalyst for CNT Growth
1.1 Motivation for Using 100% Gadolinium
Encapsulating gadolinium inside of a carbon nanotube has the potential to move medical imaging to a new level by: (1) reducing Gd toxicity, (2) improving imaging response and
(3) allowing for imaging at the cellular level.
Since combing other known catalyst materials with gadolinium may have unforeseen ramifications, a method where CNTs can be “grown” from 100% gadolinium nanoparticles is preferred. The following experiments were conducted to determine if gadolinium showed any promise as a CNT catalyst. A literature survey should be conducted in an effort to understand how iron or another known catalyst might impact the functioning of gadolinium as well as potential medical ramafications.
1.2 Synthesis Experiments Using 100% Gadolinium
Experiments were conducted to ascertain if gadolinium alone could function as a catalyst
for CNT growth. To ensure that the results obtained were not influenced by any
100 environmental conditions arising during the synthesis process, iron catalyst, and later on, combination gadolinium and iron catalyst substrates were used as control samples.
Control substrates were selected based on their repeatability. Substrates having a catalyst film of 100% gadolinium were produced by both PLD and E-Beam Evaporation.
Three substrates (062706_A, B & C) were prepared using PLD at North Carolina A&T
University. These substrates had 2, 4 and 6nm thick gadolinium films. 100% Gd substrates (substrates: A & 100406_A) were also produced using E-Beam Evaporation at
UC. Substrate preparation was identical to other iron catalyst substrates with the only process change being the deposition of Gd in lieu of iron. For a detailed description of the process steps, refer to Chapter 2-the four primary substrate preparation steps are outline below.
Step 1: Deposit 15nm thick Al film on 500micron thick Si/SiO2 wafers
Step 2: Plasma Oxidation of Al film – 10, 20 or 30 min., 400W, 20% O2 80% Ar
Step 3: Deposition of 2nm thick Gd film using E-Beam evaporative
Step 4: Thermal Annealing of substrate - 400˚C for 5 hours
CNT synthesis using substrates produced by E-Beam evaporation was attempted first under a number of different CVD conditions. After no CNT nucleation occurred, synthesis using the PLD substrates was conducted. The synthesis reference numbers for these experiments are provided below.
(a) 060706_1 - 5 minute carbon deposition
(b) 060706_2 - 10 minute carbon deposition
101 (c) 060906_1 – 2.5 hour carbon deposition
(d) 070206_1 – 3 hour carbon deposition
1.3 Findings of Synthesis Study Using 100% Gd Catalyst Film
All attempts to grow CNT arrays from 100% Gd substrate were unsuccessful. No trace of CNT could be observed with either optical or Scanning Electron Microscopes; thus it was determined that gadolinium alone cannot function as a catalyst for the nucleation of
CNTs produced using traditional CVD methods. Further study was suspended due to lack of positive results.
Since it was determined that gadolinium alone does not function as a catalyst, the next logical step was to investigate if gadolinium in combination with a known catalyst such as iron would be effective in producing arrays of carbon nanotubes.
2.0 Investigating Gadolinium as an Iron Catalyst Motivator
2.1 Motivation for Combining
It was proposed that gadolinium, which cannot function as a CNT catalyst, in
combination with a known CNT catalyst such as iron, might result in arrays of vertically
aligned nanotubes. If catalyst nanoparticles that incorporate Gd and iron can be formed
through typical substrate preparation techniques, gadolinium, in addition to iron will
likely be encapsulated by the carbon shells of the CNT. These CNT particles produced
could then be harvested and used to conduct MR imaging studies to see if the imaging
response is improved by the encapsulated iron and gadolinium nanoparticles.
102 2.2 Gd/Fe Substrate Preparation Using E-Beam Evaporation
The same substrate preparation techniques that had worked so well to produce the iron catalyst substrates were also used to produce combination gadolinium and iron catalyst substrates.
To deposit the combined iron and gadolinium onto silicon substrates using the same E-
Beam evaporation techniques used to produce the iron catalyst substrate discussed in
Chapter 4, the two metals were combined together into one “ingot” that could be used in the E-Beam Evaporator. Previously 100% Fe and 100% Gd ingots had been used to produce the substrates-the combined 20% Gd, 80% Fe ingot was produced by Kurt J.
Lesker Company.
A higher percentage of iron was chosen so that there would be a greater chance for CNT nucleation. The optimum ratio for medical imaging applications would be the highest percentage of gadolinium to iron that can be used and still produce CNT arrays. For MRI imaging purposes, CNTs do not need to be very long so CNT growth would only need to be sustained for very short periods of time (<<1 minute), and substrate preparation should be focused on increasing the amount of encapsulated gadolinium while optimizing the length of the CNT for delivery into the body. Very short CNTs would be best suited for liquid suspension and for injection into the body, and may also assist in imaging of very small organs or cells. The growth conditions that result in optimum CNT length and diameter have yet to be studied. The length and diameter of the CNT would be crucial to ensure that the carbon structure can pass through the body and cellular membranes.
103 Several different types of substrates were produced to investigate optimum preparation methods and the ratio of gadolinium to iron. Combination gadolinium and iron substrates produced using E-Beam Evaporation and PLD include the following substrate types:
(a) 2nm Thick Catalyst film deposited by E-Beam evaporation from combination
metal ingot (20% Gd and 80% Fe). Substrates numbers include:
a. 072106_1, 2 & 3
b. 080306_1, 2 & 3
c. 083001_A & B
(b) 50%Gd and 50%Fe produced by depositing 1nm of 100% Gd on top of a 1nm
thick film of 100% Gd. E-Beam evaporation was used to deposit both 1nm thick
films. The substrate produced is designated as 100406_B.
(c) Combinatorial substrates produced using PLD with different layers of gadolinium
and iron. PLD substrates were produced by North Carolina A&T University.
2.3 Gd/Fe, E-Beam Substrate Surface Characterization Using AFM
A preliminary study of the substrate surface characteristics was conducted at each substrate preparation step used to produce the iron catalyst substrates for oriented CNT array growth. The Nanosurf Easy Scan 2 atomic force microscope (AFM) was used to scan the substrate surface after each process step. The surface topology data obtained from the AFM scans was then used to measure the surface properties of the substrate.
Post processing of the topology data provides information about the catalyst particle grain
104 size and Root-Mean-Square (RMS) surface roughness; a commonly used metric for comparison of different substrate types [1].
2.3.1 Characterization Process Using AFM
The surface topology data obtained from the AFM scans was then used to measure the surface properties of the substrate. Post processing of the topology data can provide information about the catalyst particle grain size and Root-Mean-Square (RMS) surface roughness; a commonly used metric for comparison of different substrates. In general the effects of the substrate surface are not well understood-however the plasma does changes the chemical structure of the aluminum surface. Some preliminary studies comparing synthesis with aluminum and alumina have been conducted. In general the alumina layer produces more favorable CNT growth conditions. This tends to indicate that the alumina layer is not passive. The images for Steps 1 through Step 4 are provided below and include images of substrates plasma oxidized for 10, 20 and 30 minutes.
2.3.2 Results of Gd/Fe Substrate Characterization
The following provides the graphic and surface topology data obtained through the AFM study. The images for Steps 1 through Step 4 are provided below, and include images of substrates plasma oxidized for 10, 20 & 30 minutes with a 60mtorr Vacuum, 300W RF power, 20% oxygen and 80% argon environment. The AFM scans were used to obtain the surface roughness data for each scan area. The measured surface topology (height) across each scan area was exported to Excel in a matrix format. Each row of the topology matrix was Fourier Transformed and the maximum and linear average for all
105 spectra was obtained. The Fourier Transformed spectra have a wave-number resolution of 0.5. A Hanning window was used to reduce leakage effects.
Figure 5.1: AFM Image of 15nm Thick Aluminum Film - “Step 1”
5
0
Surface Topography(nm) -5 0 0.5 1 1.5 2 Sample Width (micro-meter)
Figure 5.2: Representative Topology Sections (15nm Thick Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.3: Fourier Transform of Surface Topology (15nm Thick Al)
106 10 Minute Oxidation 20 Minute Oxidation 30 Minute Oxidation
Figure 5.4: AFM Images of 15nm Thick Aluminum Film Plasma Oxidized - “Step 2”
10 Minute Oxidation 20 Minute Oxidation 30 Minute Oxidation
Figure 5.5: AFM Images of 2nm Thick 20% Gd / 80% Fe Catalyst on Alumina - “Step 3”
10 Minute Oxidation 20 Minute Oxidation 30 Minute Oxidation
Figure 5.6: AFM Images of Thermally Annealed 20% Gd / 80% Fe Catalyst - “Step 4”
107 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.7: Representative Topology Sections (2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.8: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
108 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.9: Representative Topology Sections (2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.10: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
109 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.11: Representative Topology Sections (2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.12: Fourier Transform of Surface Topology (2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
110 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.13: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.14: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 10minute Oxidation of Al)
111 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.15: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.16: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 20minute Oxidation of Al)
112 5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
5 5
0 0
Surface Topography(nm) -5 Surface Topography(nm) -5 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Sample Width (micro-meter) Sample Width (micro-meter)
Figure 5.17: Representative Topology Sections (Annealed - 2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
1 Average 0.9 Maximum
0.8
0.7
0.6
0.5 nm
0.4
0.3
0.2
0.1
0 0 10 20 30 40 50 60 70 1/micro-meter
Figure 5.18: Fourier Transform of Surface Topology (Annealed - 2nm 20% Gd, 80% Fe, 30minute Oxidation of Al)
113 The following surface roughness data was obtained for each of the AFM scan areas provided above. The values obtained are based on the entire scan area; however multiple scan areas were not analyzed. To obtain statistical information, additional studies should be conducted and standard statistical processes applied to the overall data. The surface roughness descriptors provided in Tables 5.1, 5.2, 5.3 & 5.4 are as follows:
Sa – Average Roughness
Sm – Mean Roughness
Sq – Root Mean Square Roughness
Sv – Valley Roughness
Sp – Peak Height
Sy – Peak to Valley Height
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 0 minute 1.2 0.22 1.5 -5.1 9.7 15 Table5.1: Step 1: Surface Metrology of 15nm Thick Aluminum Film
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.38 0.21 0.46 -1.8 1.8 3.6 20 minute 0.36 0.21 0.48 -1.8 8 9.8 30 minute 0.26 0.000046 0.33 -1.4 3.9 5.2 Table 5.2: Step 2: Surface Metrology: Oxidized Alumina Film (Al2O3)
114 Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.43 0.21 0.55 -3 4.7 7.7 20 minute 0.43 0.21 0.57 -3.2 4.5 7.7 30 minute 0.27 -0.00013 0.34 -1.6 1.8 3.3 Table 5.3: Step 3: Surface Metrology: Al2O3 + 2nm Fe Film, No Thermal Annealing
Plasma Oxidation Sa Sm Sq Sv Sp Sy Time (nm) (nm) (nm) (nm) (nm) (nm) 10 minute 0.65 0.0000016 0.82 -3.9 4.3 8.2 20 minute 0.38 0.00012 0.5 -3.4 3.7 7.1 30 minute 0.5 -0.00019 0.63 -2.8 2.8 5.6 Table 5.4: Step 4: Surface Metrology: Al2O3 + 2nm Fe Film, Annealed for 5hrs at 400ºC
1.75
10minute oxidation 20minute oxidation 30minute oxidation
1.5
1.25
1
0.75
0.5 RMS Surface Roughness (nm) Surface Roughness RMS
0.25
0 Step 1 Step 2 Step 3 Step 4 Substrate Preperation Step
Figure 5.19: RMS Surface Roughness at Different Oxidation Times
2.3.3 Discussion of Gd/Fe Substrate Study Results
Figure 5.5 above indicates that the initial plasma oxidation of the 15nm thick Aluminum layer results in a substantial decrease in surface roughness (Steps 1 to Step 2). The peak
115 to valley height of the aluminum layer before oxidation is 15nm-the thickness of the aluminum layer. After oxidation, the height is reduced to 3.6, 9.8 and 5.2nm for the 10,
20 and 30 minute oxidation times respectively. This indicates that material is removed as the plasma ions bombard the substrate surface at high velocity; refer to Chapter 3 for a general discussion of plasma oxidation.
There is very little change in surface roughness when the catalyst film is applied. This indicates that the catalyst layer coating is relatively smooth, and the surface features are predominately controlled by the Alumina layer below.
Thermal annealing of the catalyst layer changes the surface roughness, and in general increases the surface roughness. For the 10 and 30 minute plasma oxidation times, while the 20 minute oxidation time was not as impacted by the thermal annealing. It is anticipated that further study would indicate that thermal annealing generally increases the surface roughness as it forms the catalyst nanoparticle layer.
The surface topology frequency spectra data compared to 100% iron catalyst substrates tends to indicate that there is an increase amount of smaller catalyst particles. The may be one more explanation why the gadolinium helps to synthesis process, it essentially regulates the particle grain size; which may help to speed up the nucleation process due to the small particle size.
2.4 Gd/Fe Substrates Produced Using Pulsed Laser Deposition (PLD)
A number of substrates prepared using Pulsed Laser Deposition (PLD) were provided by
North Carolina A&T University. As was done previously with 100% iron & 100%
116 gadolinium catalyst substrates, these combinatorial substrates were used to gain insight on how to tailor the binary-catalyst layers such that growth is optimized in terms of length and structural consistency.
Back
Region I
Region II Left Right
Region III
Region IV
Front
Figure 5.20: Top View of Combinatorial Substrate #149 (figured prepared by Dr. Sudhere Nurella at NC A&T, edited by Andrew Gorton)
I II III Fe Gd Gd Al2O3 Al O 2 3 Al2O3
Fe IVGd V Gd Fe Fe Gd Gd Al O Al2O3 2 3
Figure 5.21: PLD Steps Used to Create Combinatorial Substrate #149 (figured prepared by Dr. Sudhere Nurella at NC A&T, edited by Andrew Gorton)
117 Region IV Region III Region II 1nm Fe Region I
1nm Gd
20nm Al2O3
500µm Si/SiO 2
Figure 5.22: Left Side View of Combinatorial Substrate #149 (figured prepared by Dr. Sudhere Nurella at NC A&T, edited by Andrew Gorton)
Combinatorial substrates (substrate #’s: 153, 154, 155 & 156) were produced to gain further insight into the optimum ratio of iron to gadolinium. The laminate layers were deposited at a temperature of 200°C. Alternate layers of Fe and Gd were deposited on a
20nm thick layer of Al2O3 on Si/SiO2 substrate. The given thickness of Fe is the total thickness of the 10 deposited layers, and the thickness of Gd is the total thickness of 9 layers.
PLD Sample #153 Fe-3 nm; Gd-1nm (Fe 75%, Gd 25%)
PLD Sample #154 Fe-3 nm; Gd-2nm (Fe 60%, Gd 40%)
PLD Sample #155 Fe-3 nm; Gd-3nm (Fe 50%, Gd 50%)
PLD Sample #156 Fe-4 nm; Gd-1nm (Fe 80%, Gd 20%)
118 Gd
Al2O3 Fe
Figure 5.23: Layered Substrate of Iron and Gadolinium Produced Using PLD (figured prepared by Dr. Sudhere Nurella at NC A&T)
2.5 Overview of Synthesis Using Gd/Fe, E-Beam Substrates
The original purpose of these experiments was to determine if Gd could be encapsulated inside the CNT for imaging purposes. An unexpected and very positive result was that that the addition of gadolinium to iron catalyst substrates resulted in greatly increased array height, improved repeatability and structural uniformity across the array surface.
This remained true as the substrate surface area was increased to approximately one square inch. In general, gadolinium combined with iron produced super-long, very robust substrates-meaning that they were less prone to structural variations and produced very predictable, repeatable results.
119 2.5.1 Synthesis Results Using Gd/Fe, E-Beam Substrates
The CNT arrays shown below were produced using substrates having a 2nm catalyst film of 20% gadolinium and 80% iron. The height of CNT arrays produced using these substrates generally ranged between 5 and 10mm. Over relatively large surface areas,
CNT arrays can be grown 4-5mm in length consistently. CNT arrays produced using these substrates are highly repeatable. Multiple substrates were produced independently using the same preparation techniques with equivalent results.
Figure 5.24: Photographs of CNT Arrays Produced Using 20%Gd, 80%Fe Substrates
As discussed in this chapter and in Chapter 4, oxidation of the Aluminum substrate layer is a critical step in producing substrates that grow long, vertically oriented CNT arrays.
To study the effect of plasma oxidation time on combination iron/gadolinium CNT array height, substrates were oxidized in a 20% oxygen, 80% argon gas environment at 400W
120 and in a vacuum of 30mTorr for 10, 20 and 30 minutes. Multiple 20%Gd, 80%Fe substrates of each oxidation time were grown for a period of one hour and three hours.
Each substrate piece was nominally 30mm2.
The height of each CNT arrays from the two experiments was measured using the technique described in Chapter 3. The data obtained was used to generate the statistical
“box plots” provided in Figures 5.11 and 5.12 below. The maximum length achieved during the 1 & 3 hour growth periods was approximately 2mm & 5mm respectively.
The primary features of the box-plots are as follows: (1) the lower and upper blue lines of the box are the 25th and 75th percentiles respectively; (2) the red center line is the median or 50th percentile; (3) the upper and lower “Whickers” indicate the range of data used in the statistical analysis; (4) red plus signs outside beyond the ends of the whiskers are considered to be “outliers” and are not included in the analyzed data set.
2
1.9
1.8
1.7
1.6
1.5
Height of Array (mm) 1.4
1.3
1.2
1.1 10 20 30 Oxidation Time (min)
Figure 5.25: Growth Length Compared to Plasma Oxidation Time (1 hour carbon Deposition)
121 4
3.5
3
Height of Array (mm) 2.5
2
10 20 30 Oxidation Time (min)
Figure 5.26: Growth Length Compared to Plasma Oxidation Time (3 hour Carbon Deposition)
2.5.2 Discussion of Synthesis Results Using Gd/Fe, E-Beam Substrates
It was observed that structural uniformity decreases as the array height and surface area increases. Films of carbon nanotubes (<3mm) have excellent structural uniformity and can be produced with highly repeatable results. This result is very important when considering how these CNT arrays could be used for many different applications, including intermediate layers in laminant carbon composites. For a discussion of this particular application, please refer to Chapter 6.
For many applications, other than MR contrast agent improvement, structural uniformity will become increasingly important as the process is scaled-up to larger and larger surface areas. As illustrated by Figure 5.10 b above, one major structural defect common to CNT arrays covering a large surface area is the difference in array height between the edges and the center of the array. This difference is most pronounced when the height at the center is compared with the height of the edge closest to the gas inlet inside the CVD
122 furnace. This is likely a result of decreased carbon flux to the center of the array. What is likely happening is that the edges of the CNT are depleting the amount of decomposed carbon before it has can reach the center of the array; resulting in shorter CNTs at the center of the array. A molecular dynamics model could be run to assess the rate of carbon depletion in the gaseous environment.
To counteract this effect, one possible method may be to oscillate the gas particles at maximum velocity at the location of the CNT array. This potential method is discussed in Chapter 8.
Studying the result data for the plasma oxidation study provided in figures 5.11 and 5.12 above, some reasonable conclusions can be drawn.
(a) Longer synthesis time results in longer CNT arrays
(b) For 1hour synthesis, the longest CNT arrays can be achieved using 20 minute oxidation
(c) For 3 hour synthesis, the longest CNT arrays are achieved using 10minute oxidation time.
Knowing that plasma oxidation removes, or etches material away from the surface as it oxidizes the aluminum, one can surmise that there is an optimum length of time where oxidation is sufficient, and the aluminum layer has not been reduced in thickness to an extent that results in the iron catalyst leaching into the silicon-limiting CNT growth [1].
The shorter oxidation time may be optimum for longer synthesis times because iron atoms tend to be more protected from the silicon below, and larger catalyst particle may be better for longer synthesis times.
123 2.6 Overview of Synthesis Using Gd/Fe, PLD Substrates
Pulsed Laser Deposition (PLD) substrates produced by North Carolina A&T University were used to investigate if there were any benefits to layering the iron and gadolinium.
2.6.1 Synthesis Results Using PLD Substrates
The Figures 5.13, 5.14, 5.15, 5.16 & 5.17 below are the results of synthesis using a 3 hour carbon deposition time and hydrogen pre-deposition phase.
Front Left Back Right
Figure 5.27: Photos of Combinatorial Substrate #149 after CNT Synthesis
Front Left Back Right Figure 5.28: Photos of Layered Substrate #153 after CNT Synthesis (Fe-3 nm, Gd-1nm - Fe 75%, Gd 25%)
Front Left Back Right
Figure 5.29: Photos of Layered Substrate #154 after CNT Synthesis (Fe-3 nm, Gd-2nm - Fe 60%, Gd 40%)
124
Front Left Back Right
Figure 5.30: Photos of Layered Substrate #155 after CNT Synthesis (Fe-3 nm, Gd-3nm - Fe 50%, Gd 50%)
Front Left Back Right
Figure 5.31: Photos of Layered Substrate #156 after CNT Synthesis (Fe-4 nm, Gd-1nm - Fe 80%, Gd 20%)
2.6.2 Discussion of Synthesis Results Using PLD Substrates
In general, combinatorial and layered PLD substrates comprised of gadolinium and iron did not produce any particularly interesting results. The combinatorial substrate does indicate that the 1nm iron strip produced the best results, while the layered areas of Gd and Fe had poor uniformity and tended to grow in a very unpredictable fashion-there also appears to be some sort of surface impurity that limited growth along one edge of the substrate. The layered substrates tended to produce thin layers of growth that seems to occur as CNTs grow from the edges. The probably occurs because the layers restrain the
CNTs from growing vertically, and they are forced to find other directions to grow in.
Further study may be conducted with a clear objective, but these results do not provide any particularly interesting information.
125
2.7 Characterization of Gd/Fe, E-Beam Substrates
The following sections provide results and observations for various CNT characterization techniques. The techniques covered here include:
(a) Environmental Scanning Electron Microscopy (ESEM)
(b) High-Resolution Transmission Electron Microscopy (HRTEM)
(c) Thermal Gravimetric Analysis (TGA)
(d) Raman Spectroscopy
(e) Inductively Coupled Plasma Spectroscopy (ICP)
2.7.1 Overview of ESEM Study
2.7.1.1 Results of ESEM
The following images illustrate the potential of gadolinium as a catalyst motivator to produce CNT arrays with a lot of potential for development.
Figure 5.32: Substrate Patterned with Fe/Gd Catalyst
126
Figure 5.33: CNT Array (>4mm) produced from Gd/Fe Substrates
Figure 5.34: CNT Array (>4mm) produced from Gd/Fe Substrates
127 Figure 5.35: Close-Up of CNT Array in shown in Figure 5.20 - Nanoropes, shown here are an interesting Byproduct of the Gd/Fe CNT Arrays.
Element Wt% At%
C K 87.11 93.98
O K 00.59 00.48
AlK 00.58 00.28
SiK 11.35 05.24
GdL 00.37 00.03
Figure 5.36: EDS of Fe/Gd CNT Array Surface
128 2.7.1.2 Discussion of ESEM Results
Visual inspection of the CNT arrays at high magnification did not indicate any visible layer of amorphous carbon. In general the CNT arrays produced using gadolinium in combination with iron, were dense with few structural defects. For shorter array lengths, the edges of the CNT arrays were clean and relatively free from cracks or clusters of
CNTs pealing away.
EDS using the Philips ESEM indicated only trace amounts of gadolinium, while counts for carbon and silicon were substantially higher. Amount of gadolinium in this range may be within the noise floor of the system and may not be accurate. Further studies should be conducted to determine if this method is viable for measuring the amount of gadolinium present in the analysis region.
2.7.2 Overview of HR-TEM Study
High resolution transmission electron microscopy was performed at the University of
Kentucky by Dr. Wentao Xu. The primary objective of the HRTEM imaging was to try and detect trace amounts of gadolinium encapsulated inside of a CNT. Prior attempts to detect gadolinium across the surface of several CNT arrays using the EDS system of the
Phillips ESEM were unsuccessful (refer to Chapter 3 for a more thorough description of this system). It was hoped that a more sensitive EDS system, under very high magnification, would be able to determine if gadolinium is present in the CNT sample.
The JEOL JEM-2010F HRTEM EDS system has vastly better sensitivity which may help
129 to determine if gadolinium from the substrate surface was in fact encapsulated inside of the CNT.
In order to more easily locate the catalyst nanoparticles, which should be located at one of the two ends of a CNT, a 1-minute deposition time was used to limit the height of the
CNTs. Like searching for the proverbial “needle in the hay-stack”, by reducing the size of the hay-stack, it should make finding the “needle” that mush easier.
Samples were prepared by carefully removing the CNT layer from a substrate having a catalyst film consisting of 20% gadolinium & 80% iron. Care was taken not to scrape the substrate surface and remove any gadolinium that may have remained on the substrate surface. The layer of nanotubes was then placed in a small glass vial containing Ethyl alcohol. The mixture was shaken by hand and briefly sonicated in a water bath until the
CNT clumps began to disperse.
\The prepared samples were taken to the Electron Microscopy Center at the University of
Kentucky, deposited on a carbon grid and placed in their HRTEM. The HRTEM system enables one to obtain very detailed images down to the atomic scale. Images of nanotubes grown from 80% Fe, 20% Gd substrates are provided in Figures 5.23 A thru F.
The figures A thru D show three different catalyst particles totally encapsulated by the shells of the nanotubes.
2.7.2.1 Results of HR-TEM Study
The crisscrossing pattern shown in black the catalyst particle in Figures C & D is the atomic lattice structure of the metal catalyst particle. CNTs shown in Figures A thru D
130 are multiwall particles with between 17 and 20 concentric carbon shells; while the CNTs shown in Figures E & F are double wall nanotubes. The black lines running parallel with one another are the individual carbon shells of the nanotube.
131 Figure 5.37: HRTEM Images of Catalyst Particles Encapsulated by a MWCNT
(A) Particle over 100nm long, and 40nm wide across at the largest point (B) Particle approximately 15nm long and <10nm wide (C) Particle <15nm long and <10nm wide (D) Close-up of (C), Crystalline lattice structure of metal particle is visible (E) Double Wall Carbon Nanotubes (F) Same as E
Initially, point Energy Dispersive Spectroscopy (EDS) was attempted at 1 million times magnification. Accurate analysis was not possible due to the CNTs being excited by the electron beam. The excitation of these particle results in drifting of the particles making
EDS at a fixed point over the required length of time nearly impossible. In addition to particle motion, there is also a risk of the highly focused electron beam damaging the graphitic structure of the CNT.
As an alternative, line EDS analysis was used in the scanning mode. This allows for EDS to be done along a predefined line. By selecting the line accordingly, we were able to compensate for the “drifting” motion of the particles being analyzed; in hopes that a more accurate representation of the elements contained in the sample.
Gadolinium (Gd) 1000 Iron (Fe) Carbon (C)
Counts 500
0
100 nm 0 50 100 150 200 250 300 Position (nm)
Figure 5.38: Line EDS Analysis across Gd/Fe Catalyst Particles (“Low” Magnification)
132 200 Gadolinium (Gd)
150 Iron (Fe) Carbon (C)
100 Counts
50
0
50 nm 0 50 100 150 200 250 300 Position (nm)
Figure 5.39: Line EDS Analysis across Gd/Fe Catalyst Particles (“Medium” Mag.)
1000 Gadolinium (Gd) Iron (Fe) Carbon (C)
Counts 500
0
10 nm 0 10 20 30 40 Position (nm)
Figure 5.40: Line EDS Analysis across Gd/Fe Catalyst Particles (“High” Mag.)
2.7.2.2 Discussion of HR-TEM Results
Figure 5.24 provides the best indications that trace amount of gadolinium are present in the catalyst particle. The red line across the two catalyst particles (represented by the white areas in the picture) is the path along which element counts are taken. The graph on the right provides the element counts for gadolinium, iron and carbon. Careful observation of the data indicates that when element counts are done across one of the iron particles, the counts in gadolinium also increase. Vise-versa, when the element counts are done away from the catalyst particle, the amount of carbon peaks, while both the iron and gadolinium counts approach zero. This analysis tends to indicate that there is some
133 amount of gadolinium contained in both of the two particles analyzed. Figures 5.25 &
5.26 tend to confirm these results, but do not illustrate this as conclusively as Figure 5.24.
2.7.3 Overview of Thermal Gravimetric Analysis
TGA was initially used to study the purity of CNT arrays grown using the 20%Gd and
80% Fe. The experiment was repeated twice to verify the results. Typically amorphous carbon oxidizes around 400 ˚C; while crystalline carbon typically oxides around 700˚C.
Figures 5.27 and 5.29 provide the % of total mass as a function of increasing temperature.
To more precisely locate the temperature(s) where the change in mass occurs most
⎛ dM ⎞ rapidly, the derivative ⎜ ⎟ of the measured mass (M) as a function of temperature (T) ⎝ dT ⎠ was computed and is plotted in Figures 5.28 and 5.30.
2.7.3.1 Results of Thermal Gravimetric Analysis
120
100
80
60
40
20 Percent of Starting Mass (%) 0 0 100 200 300 400 500 600 700 800 900 1000 Temperature (˚C)
Figure 5.41: Test 1 - TGA plot of 20% Gd / 80% Fe substrate (Oxidation in Air) Starting Mass =3.332g
134 10
1 C)
˚ 0.1
0.01
0.001 dM/dT (mg/ 0.0001
0.00001 600 610 620 630 640 650 660 670 680 690 700 710 720 730 740 750
Temperature (˚C)
Figure 5.42: Test 1 – Derivative of Function Plotted in Figure 5.27
120
100
80
60
40
20 Percent of Starting Mass (%)
0 0 100 200 300 400 500 600 700 800 900 1000 Temperature (˚C)
Figure 5.43: Test 2 - TGA plot of 20% Gd / 80% Fe substrate (Oxidation in Air) Starting Mass (3.371g)
135 1
0.1
C) 0.01 ˚
0.001
` 0.0001 dM/dT (mg/
0.00001
0.000001 600 610 620 630 640 650 660 670 680 690 700 710 720 730 740 750 Temperature (˚C)
Figure 5.44: Test 2 – Derivative of Function Plotted in Figure 5.29
2.7.3.2 Discussion of Thermal Gravimetric Analysis Results
Figures 5.28 and 5.30 indicate that there is virtually no mass change until approximately
575˚C. Between 600˚C and 700˚C, the change in mass due to oxidation of the crystalline carbon is very pronounces. Above 750˚C, 0% of the initial mass remains. Below 680˚C, the rate of mass change indicated is most likely a result inaccuracies (“noise”) in the
measurement system. Only one peak in the derivative was observed. This peak occurred
at 690˚C for both experiments. No peaks could be identified at the temperature where
amorphous carbon would tend to oxidize-this indicates that nearly 100% of the mass
change is due to oxidation of the crystalline carbon which makes up the walls of the
CNTs.
The data obtained tends to indicates that amorphous carbon is not detectable by the TGA
method. Other data has indicated (HRTEM) images and Raman Spectroscopy that
amorphous carbon is produced as part of the synthesis process is 20%Gd / 80%Fe
136 substrates. We anticipate that some amount of amorphous carbon oxidation could be observed using TGA if the mass of the sample could be substantially increased, or a more sensitive system used.
2.7.4 Overview of Raman Spectroscopy
Raman spectroscopy was performed on the top and side of two CNT arrays produced using 20%Gd / 80%Fe. One CNT array was thermally treated in argon at 1200°C.
Thermal treatment should improve the purity of the CNTs by burning off the amorphous carbon, which will oxidize at a lower temperature than the graphitic carbon. Thermal treatment will also help to “heal” flaws in the CNT walls by providing enough energy to form bonds at the discontinuities.
2.7.4.1 Results of Raman Spectroscopy
MWCNT Arrays Catalyst: 20%Gd & 80% Fe Side of Array Scanned 60 No Thermal Treatment "G" peak -1 Thermally Treated (1200˚F) 1575cm 50 "D" peak 1340cm-1 40
30 Counts
20
10
0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Wave Number (cm-1)
Figure 5.45: Raman Spectrum - Side of CNT Array
137 MWCNT Arrays 20%Gd & 80% Fe Catalyst Top of Array Scanned 60 No Thermal Treatment Thermally Treated (1200˚F) 50 "D" peak 1350cm-1 40 "G" peak 1590cm-1 30 Counts
20
10
0 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Wave Number (cm-1)
Figure 5.46: Raman Spectrum - Top of CNT Array
2.7.4.2 Discussion of Raman Spectroscopy Results
The peak in the Raman spectrum (Figure 5.31 & 5.32) at 1340 & 1350cm-1 for the treated and untreated samples is known as the “D” peak. This peak occurs as a result of amorphous or non-graphitic carbon present-the peaks at 1575 & 1590cm-1 correspond to the crystalline carbon that comprises the carbon forming the CNT walls. Amorphous carbon does not have any particularly useful properties, and reducing the amount of amorphous carbon is necessary to produce CNT arrays that have high purity. The purity of a CNT array can be evaluated by comparing the magnitude of the D and G spectrum peaks. Ideally the D peak would be reduced to a level that is undetectable; however results from numerous research groups indicate that there generally a pronounced D peak
[11, 12, 13].
138 The Raman spectra provided in Figure 5.31 and 5.32 indicate that the sides of the CNT array have a stronger G peak with respect to the D peak. This is likely due to amorphous carbon build-up on top of the CNT array grows. It is anticipated that build-up of amorphous carbon on top is most pronounced for bottom growth CNT arrays. This layer probably forms as excess carbon atoms collect on top of the array. It is clear from the spectra plots that thermal treatment of the arrays does decrease the D peak in relation to the G peak-thus improving CNT array purity.
2.7.5 Overview of ICP Analysis
Inductively Coupled Plasma (ICP) mass Spectrometry allows for quantitative and qualitative analysis of elements. ICP has the capability of sensing trace amounts of elements. This makes it an attractive method for further investigating if gadolinium is encapsulated by the CNTs.
ICP analysis was conducted at the University of Cincinnati Mass Spectrometry Facility.
Computer print-out results from these tests are provided in appendix. The data sheets provided indicate that Gd was not present in the solution analyzed. It is very possible that through the preparation and analysis process, the trace amounts of gadolinium were substantially reduced, below detectable limits.
2.7.5.1 Purpose of ICP Analysis
For applications in medical imaging, it is crucial to know the ratio of the substrate surface area to amount of gadolinium contained within the CNTs. This way, the amount of gadolinium being delivered to a patient can be estimated. The ratio between the weights of the Gd and the CNTs is not useful because the weight of the catalyst particle
139 containing gadolinium would generally remain constant while the CNTs can have vastly different weights because due their ability to grow to long lengths (<1cm).
2.7.5.1.1 ICP Sample Preparation
One major issue that arises when trying to use ICP techniques to detect elements encapsulated inside the CNT is that the Nebulizer can become clogged by the CNTs, which tend to stick together in large clumps. Forming an aerosol spray from clumps of carbon nanotubes cannot be accomplished by current ICP sample preparation methods.
Exploring this area further may be interest if ICP is to be used more extensively to investigate particles encapsulated inside the CNT.
In order to overcome the problem of clumping, the CNT samples prepared for ICP analysis were oxidized in air and the crystalline carbon converted to carbon Dioxide. In theory this should leave all of the catalyst metal particles behind, along with any gadolinium that was incorporated; however there are a number of issues with the sample preparation process.
Once the crystalline carbon has been oxidized, there is no visible material left behind, making it impossible to determine at a glance if any of the catalyst material remains.
Giving the very small size of the gadolinium/iron particles contained inside each CNT, it is plausible that catalyst particles could simply be released into the air during transport or during the thermal oxidation process. For this reason, other methods for preparing ICP samples should be investigated.
140 Perhaps a better method would be to use Laser Ablation ICP, which uses a laser, often of
Ultra-Violet light. This method of ICP analysis allows for direct analysis of solids and powders. This method is favorable because it eliminates the need for sample digestion, minimizes contaminants, and reduces polyatomic interferences [14].
2.7.5.2 Estimation of Elemental Mass on a Substrate
As a crude method for estimating the weight of gadolinium deposited on a substrate surface through E-Beam Evaporation is described below. For a substrate approximately
28mm x 18mm, having a 2nm thick film of 20%Gd and 80%Fe, the approximate amount of metal contained on the surface is 1.6µg and 6.5µg respectively.
(1) Measured the area of the substrates being used to grow the CNT array. Try to
obtain as accurate a measurement as possible. Estimating to surface area by
weight may be less prone to error, as assuming one can determine the correct
eight per unit area.
(2) Assume that the layer deposited is truly 2nm thick. This is a very large
assumption, as the layer thickness can very greatly when dealing with such
thin layers. Calculating a range of values may be a better approach.
(3) The density of the metal catalyst at the deposition temperature can be found in
most texts covering Electron Beam Evaporation. The densities of iron and
gadolinium at 293K are 7.86g/cm3 and 7.895g/cm3 respectively.
141 (4) Once the volume of the catalyst film is known, and the percentage of metals
used to form the catalyst, one can determine the approximate amount of each
metal contained on the surface of the substrate.
2.7.5.3 Results and Discussion of ICP Analysis
The data sheets for the ICP analyses conducted by Dr. Caruso’s lab are provided in the appendix. The data sheets indicate that for substrates with 100% iron, concentration of iron is substantially greater than when gadolinium and iron are combined together; however gadolinium is not clearly detectable and 0.000 ppb was reported.
While ICP was not affective in measuring the gadolinium, further research should be conducted to determine if any errors occurred in the experimental process: sample preparation, analysis and interpretation of data.
3.0 Explanation of Why Gadolinium Improves Synthesis with Iron Catalyst
In order to gain a better understanding of how iron, gadolinium and carbon might chemically combine during CNT nucleation and growth, the binary Fe-Gd phase diagram can provide some insight into how these two elements react with one another. Since the binary phase diagram in Figure 5.33 was developed based on how bulk material reacts with different percentages of gadolinium and at different temperatures-the same phase diagram may not be entirely accurate for nanoparticles; we expect that the melting temperature at different mixture percentages, along with the eutectic point, would be reduced.
142 The binary phase diagram of Fe-Gd indicates that the melting point is drastically lowered even with small weight percentages of gadolinium. Gadolinium is approximately three times heavier than iron and has a standard atomic weight of 157.23g/mol-whereas iron has an atomic weight of 55.845g/mol. This indicates that only 28%of gadolinium by weight is needed to reap the potential benefits of lowering the catalyst particle melting point.
Recent studies have indicated that 20% Gd and 80% Fe yields the best results, and that
80% gadolinium is the maximum percentage that can be used to produce CNT growth.
The weight percentage for 20%Gd, 80%Fe is approximately 41%, while the 80%Gd,
20%Fe condition has a gadolinium weight percentage of approximately 92%.
Investigating these two points on the binary phase diagram, one observes that for bulk materials, the melting points are approximately 1307°C and 950°C respectively. The
80% gadolinium condition is just right of the eutectic point; above which the melting temperature increases quickly. This may be the reason why above 80%, CNT nucleation is poor or nonexistent. Near 41% of gadolinium by weight, the molecule is Fe23Gd6. The
C-Fe-Gd ternary phase diagrams provided in the appendix indicate a small region where
Fe23Gd6 may exist-studying this region more closely may help to gain an understanding
of the anticipated range of carbon percentages.
143
Figure 5.47 Binary Phase Diagram for Iron and Gadolinium [15]
Three ternary phase diagrams for C-Fe-Gd are provided in the appendix. These ternary phase diagram were developed at 900°C, and may also be of limited value since nanoparticles typically behave differently than macroscopic clusters of atoms or molecules. However, these phase diagrams may be worth studying. Ideally, new binary and ternary phase diagrams would be developed for nanoparticle clusters.
4.0 Fundamental Conclusions Regarding the use of Gadolinium as an Iron Catalyst Motivator
(1) Gadolinium alone cannot produce CNT growth
(2) Gadolinium in combination with iron “motivates” catalyst and significantly
improves the length of carbon nanotube arrays, structural uniformity and
repeatability over synthesis using 100% iron-potential reasons for this are:
144
a. Gadolinium lowers the melting point of iron, allowing carbon to be
deposited at lower catalyst particle temperatures.
b. Gadolinium may help to regulate carbon flux. This might minimize the
build-up of an amorphous carbon layer on the catalyst particle surface.
(3) Line EDS results indicate that trace amount of gadolinium are encapsulated, along
with iron inside of the nanotube. This result should be replicated, but does
suggest that this method may be valid for complete encapsulation of gadolinium
particles; however iron or some other known catalyst will always be required and
present due to the fact that 100% gadolinium
(4) RMS surface roughness for annealed substrate of 100%Fe and 20%Gd, 80%Fe
are nearly the same and approximately range between 0.5nm and 0.8nm for the
three difference plasma oxidation times (10, 20 & 30 minutes).
(5) For 20%Gd, 80%Fe substrates, a 10 to 20 minute plasma oxidation time is fine,
however if very log growth time (over 1 hour), 10 minutes of oxidation is
preferred.
(6) Gadolinium levels detected using the Philips ESEM EDX system may be within
the noise floor of the instrument, and should be verified through other means.
145 (7) HR-TEM images of CNTs produced using 20%Gd, 80%Fe substrates indicate
that double and multi-wall CNTs were produced.
(8) Visual inspection using ESEM and HR-TEM indicated that amorphous carbon
was not abundant. TGA analysis did not show any mass change at the oxidation
temperature of non-crystalline (amorphous) carbon-therefore CNT arrays
produced using these substrates are relatively free of amorphous carbon.
(9) Raman spectrum data indicates that thermal treatment increases the magnitude of
the G peak relative to the D peak. Raman spectrum from the side is more pure
than the top, were amorphous carbon is more prone to form on top of the array as
it grows from the bottom.
(10) Combinatorial and layered substrates produced using PLD did not provide much
insight into how these substrate function. No real benefits were seen from these
studies.
(11) ICP analysis indicated that 0.000ppb of gadolinium was present in the CNT
sample produced using 20%Gd, 80%Fe substrates-the same was true for CNTs
produced using 100% Fe substrates.
146 (12) Considering the binary phase diagram for iron and gadolinium, it is clear that
only a small amount of gadolinium is necessary to substantially lower the
melting point of iron.
(13) Then binary phase diagram also indicates that for 20%,Gd, 80% Fe, the
molecule formed may be Fe23Gd6.
5.0 Future Work in Gd/Fe Substrate Preparation and Synthesis
(1) Develop a phase diagram of Fe-Gd in nanoparticle form.
(2) Conduct synthesis experiments with varying percentages of gadolinium-the
results should be correlated with the liquid phase transition.
(3) Grow very short (<100micron long) CNTs from substrates that have the
maximum allowable gadolinium to iron ratio. Line EDS using the HR-TEM at
the University of Kentucky should be conducted to determine if gadolinium is
clearly detectable in the catalyst particles; and if so, what is the resulting
percentage of gadolinium to iron.
(4) Conduct imaging response experiments using CNTs produced using a high ratio
of Gd to Fe. Accurate baseline experiments should be conducted first to form a
comparative imaging response study.
147
(5) Conduct additional ICP analyses, and investigate the use of laser ablation ICP to
minimize the amount of sample preparation required. Chapter 7 discussion the
proposed laser ablation process is greater detail.
148 6.0 References
[1] Weishaupt, D., Köchli, V., Marincek, B., “How Does MRI Work?”, Springer, 2003.
[2] Federal Drug Admistration, 2007 “Gadolinium-Based Contrast Agents for Magnetic Resonance Imaging (marketed as Magnevist, MultiHance, Omniscan, OptiMARK, ProHance)
[3] Moran, G., Pekar, J., Bartolini, D., Chettle, F., McNeil, F., Scott, A., Gibbons, J., Prato, F., 2002, "An Investigation of the Toxicity of Gadolinium Based MRI Contrast Agents", International Society for Magnetic Resonance in Medicine.
[4] Cai, D., Mataraza, J., Qin, Z., 2005, “Highly efficient molecular delivery into mammalian cells using carbon nanotube spearing,” Nature Methods, 2 (6) pp. 449- 454
[5] Bianco, A., Kostarelos, K., and Prato, M., 2005, "Applications of Carbon Nanotubes in Drug Delivery," Current Opinion in Chemical Biology, 9(6) pp. 674-679.
[6] N.W. S. Kam, T. C. Jessop, P. A.Wender and H. Dai, J. Am. Chem. Soc., 2004, 126, 6850.
[7] Q.Lu, J.M.Moore, G.Huang, A. S.Mount, A.M. Rao,L. L. Larcom and P. C. Ke, Nano Lett., 2004, 4, 2473.
[8] Sitharaman, B., Kissell, R., Hartman, K., 2005, “Superparamagnetic gadonanotubes are high-performance MRI contrast agents,” The Royal Society of Chemistry- Chemical Communications, pp3915-3917
[9] Hart, A., Slocum, A., 2006, “Force Output, Control of Film Structure and Microscale Shape Transfer by Carbon Nanotube Growth under Mechanical Pressure,” Nano Letters, 6(6), pp 1254-1260.
[10] Deck, C., Vecchio, K., 2005, “Growth mechanism of vapor phase CVD-grown multi-walled carbon nanotubes,” Carbon, 43, pp2608-2617
[11] Saito, R.; Dresselhaus, G.; Dresselhaus, M. S., “Physical Properties of Carbon Nanotubes”, Imperial College Press, London, 1999
[12] Dresselhaus, M.S., Dresselhaus, G., and Phaedon, A., 2000, "Carbon Nanotubes: Synthesis, Structure, Properties and Applications," Springer, Germany
[13] Kelsall, R., Hamley, I., Geogegan, M., 2005, “Nanoscale Science and Technology”, Wiley, New Jersey.
149 [14] Agilent Technologies, Inc. 2005, “Inductively Coupled Plasma Mass Spectrometry: A Primer”, Publication Number 5989-3526EN, USA
[15] Zhang, W., Li, C., Su, X., Han, K., 1998, „An Updated Evaluation of the Fe-Gd (Iron-Gadolinium) System“, Journal of Phase Equilibria, 19, pp. 56-62.
150
Chapter 6
Applications for CNT and CNF Particles
151 1.0 Overview of Carbon Nano-Composites
There are many possible applications for carbon nanotubes (CNT) and carbon nanofibers
(CNF). One area where CNT and CNF have begun to have a significant impact is in the engineering of exceptionally strong and light-weight structural materials. Currently carbon nanotube composites are being used to make bicycle frames, tennis rackets, etc.
The major advantage of composite materials is the improved strength to weight ratio.
Typical composite materials are comprised of a matrix material that surrounds and binds together fibers which have higher strength properties than that of the matrix. The matrix is typically a liquid resin, which becomes irreversibly solid after a curing agent is added and the mixture is heated and baked at the prescribed curing temperature. In carbon nano-composites, the fiber material that is added to the matrix material is typically CNT or CNF, which due to their extremely high aspect ratio, provide a very large bonding surface for attachment of the matrix material. Assuming that the matrix fiber material adheres well to the fiber additive, stresses can be transferred effectively from the “weak” matrix material to the stronger fiber additive. In addition to their very high strength to weight ratio, these nanoscale structures can also have the following advantages:
(a) Electrical conductivity
(b) Improved thermal resistance
(c) Unidirectional thermal conductivity
(d) Self-healing and “smart material” applications
(e) Decreased delamination
152 Traditionally, CNT and CNF fibers are added to the matrix in powdered form. Fibers in powdered form are easy to produce, but are randomly oriented. While their length, purity
(i.e. amount of amorphous carbon), molecular structure (single, double or multi-wall for
CNT) can be controlled during the process used to make the fibers, some degree of variability should always be anticipated. Powdered CNT and CNF of a specific type can be purchased from a number of different companies. While powdered fibers are typically used to produce nano-composite materials, research in the use of vertically oriented CNT arrays in multi-layered laminate composites is also presented in this chapter.
2.0 Improving Dispersion in Carbon Nanofiber Composites
This section provides some preliminary results pertaining to the use of impedance measurements as a method to estimate the degree of dispersion of carbon nanofibers
(CNF) in an epoxy resin. CNF are typically cheaper than carbon nanotubes, which makes them a good candidate for large scale applications. CNFs however have thermal, electrical and strength properties below that of carbon nanotubes. Carbon nanofibers have many carbon walls that are angled with respect to the axis of the fiber.
Figure 6.1: Illustration of the CNF
153 For these experiments, Pyrograf III PR24-XT-LHT fibers, produced by Applied Sciences,
Inc. were used. The outer fiber diameters are typically in the range of 60 to 150 nanometers and the fiber lengths range from of 30 to 100 microns [1]. These particular fibers, which are not yet commercially available, were selected for their improved dispersion properties. Adequate dispersion of fibers in the matrix material is fraught with many difficulties. In bulk powdered form, the carbon fibers tend to stick to each other in large clumps. In order for the matrix and the fiber additive to have “good” surface interaction, it is crucial that the fibers be well dispersed and distributed in the matrix.
Figure 6.2 illustrates the difference between dispersion and distribution of fibers in a matrix.
(a) Ideal distribution and “poor” dispersion
(b) “good” distribution and “good” dispersion
(c) “good” distribution and “poor” dispersion
(d) “poor” distribution and “good” dispersion
(e) “poor” distribution and “poor” dispersion
Figure 6.2: Illustration of Dispersion and Distribution
154 It is anticipated that the overall impedance of the CNF/epoxy mixture will decrease as the carbon fibers become evenly dispersed and with “good” distribution as is illustrated in
Figure 6.2b. Impedance is a complex quantity that describes the overall electrical system characteristics as a function of frequency. Impedance takes into account the static resistance (R) as well as the frequency dependant parameters, capacitance (C) and inductance (L) [2].
1 Z()ω = R + jωL + jωC
Equation 6.1: Complex Impedance [2]
The epoxy resin has very high impedance (can be assumed to be infinite); while the carbon nanofibers are highly conductive. Thus, while it is plausible that optimum distribution and dispersion is obtained (as illustrated by Figure 6.2b), the impedance of the CNF/epoxy mixture will be at a minimum. Essentially the CNF will provide conductive pathways through the epoxy resin. Because the fibers are better “linked” together, the flow of electrons from one side of the system to the other will be less impeded. The experiments discussed in this paper were conducted to ascertain if impedance measurement is a potential candidate for monitoring CNF dispersion. A method for monitoring the degree of dispersion by measuring the impedance of the epoxy and CNF mixture is investigated here. Another objective was to gain a better understanding of how resistance, capacitance and inductance change independently of one another as dispersion and distribution approaches an optimum condition.
155 2.1 Dispersion Experiments
To determine if impedance as a measure of dispersion is a viable option, resistance (R),
inductance (L) and capacitance (C) were measured as a function of time during
sonication.
Two electrodes were submerged in EPON 862 Resin and connected to a BK Precision
Model 878A RLC meter. The RLC meter was set to measure the system parameters at a
frequency of 1000Hz. An optical data cable was used to stream the data from the RLC
meter to a computer using software provided by BK Precision. Unfortunately only one
parameter (i.e. R, L or C) could be monitored at a time.
Carbon nanofibers were added to the epoxy in amounts of 1, 2 and 3% by weight of the
epoxy resin. Tip ultra-sonication, using a Fisher Sonic Dismembrator model 300 and
high speed mixing, was used to disperse the CNF in the epoxy. The use of the tip
sonicator has been shown to provide better results than sonication in a water bath since
higher wave intensities can be achieved. The tip sonication method helps to break down
the larger CNF clumps. To maximize the amount of CNF that comes in contact with the
tip sonicator, a high speed mixer is used to assist in the movement of CNF clumps past
the sonicator tip [3]. Figure 6.3 illustrates the mixing configuration and the glove box
within which the experiments were conducted.
156
Figure 6.3: Glove Box with Sonicator and Mixer
An unwanted side effect of high speed mixing is the production of air bubbles in the epoxy resin. Initial experiments exhibited a high degree of variability in the time domain.
One explanation for this phenomenon was that the air bubbles moving through the epoxy/CNF mixture were causing the resistance, capacitance and inductance to vary greatly with time. In an attempt to reduce the amount of air bubbles in the resin during the mixing process, a vacuum equal to 5 inches of water was produced in the glove box shown in Figure 6.3. To verify that the vacuum truly aided in the reduction of air bubbles produced in the epoxy, mixing was performed under vacuous conditions with only epoxy.
As illustrated by Figure 6.4, even in a vacuum environment, air bubbles were still produced in the clear epoxy sample.
157
Figure 6.4: Air Bubbles in Clear Epoxy Resin
2.2 Results of Impedance Experiments
Experiments indicate that measurement of the electrical properties of the mixture during mixing is not feasible due to large fluctuations in the R, L and C values. This could possibly be attributed to the CNF particles moving during mixing. The movement of the conductive particles in the epoxy is likely making the electrical properties of the mixture time varying; this would certainly make accurate impedance measurements difficult.
Figure 6.5 provides a typical curve for the impedance as a function of sonication time.
For this particular experiment a 3% mixture of carbon nanofibers were premixed with a hand-held mixer. After the fibers were mixed reasonably well, the mixture was sonicated at a power setting of 30 while the R, L and C were monitored. The results indicate that impedance decreases as a function of sonication time. This suggests that as the clumps of nanofibers are dispersed, the mixture becomes increasingly conductive, thus lowering the impedance of the mixture.
158 140 135 )
Ώ 130 125 120 115 i 110 Impedance (k Impedance 105 100 02468101214 time (min.)
Figure 6.5: Impedance as a Function of Sonication Time (sonicator power setting =30)
Figure 6.6: Resistance as a Function of Sonication Time (sonicator power setting =30)
159
Figure 6.7: Inductance as a Function of Sonication Time (sonicator power setting =30)
Figure 6.8: Capacitance as a Function of Sonication Time (sonicator power setting =30)
It is believed that proper control of the sonication level and duration is a critical step in better control of dispersion. Other results have indicated that sonicating at high power levels for long periods of time break apart carbon nanofibers.
160
Three mixtures containing varying amounts of carbon nanofibers were mixed under similar conditions and impedance was measured after the solutions sat overnight. Figure
6.9 illustrates the decrease in impedance as the amount of CNF is increased. This is consistent with what was expected given that the nanofibers are highly conductive.
1.00E+02
1.00E+01 ) Ώ
1.00E+00 1% 2% 3%
Impedance (M 1.00E-01
1.00E-02 Carbon Nanofibers (Percent by Weight)
Figure 6.9: Impedance of Three Different CNF/Epoxy Mixtures
2.3 Discussion of Impedance Experiment Results
There is much that is unknown about the issues associated with the dispersion of carbon nanofibers. The results obtained in these experiments indicate the impedance may be a viable technique for monitoring the degree of dispersion and interconnectivity of CNF.
Figure 6.9 verifies the assumption that with larger amounts of CNF, the impedance will decrease. This should also hold true as the fibers become more evenly dispersed.
Plotting the resistance, inductance and capacitance values (Figures 6.6, 6.7 & 6.8) used to produce the impedance plotted in Figure 6.9 indicate that resistance and inductance are
161 the two most important parameters. Both tend to follow the same general trend during sonication, while capacitance remains relatively constant.
3.0 Nanotube Reinforced Laminant Composites (NRLC)
Development of a new composite material, which uses carbon nanotubes as intermediary layers in traditional woven carbon fiber composite materials, is currently underway at UC with Dr. Abot and the Multiscale Material Characterization and Composite Structures
Laboratory. Traditional carbon fiber composites are made of multiple layers of woven carbon fiber layers imbedded in hardened epoxy resin. Multiple sheets of woven carbon fibers are stacked on top of one another and the epoxy matrix material is used to bind the woven carbon fiber sheets together.
It has been proposed to use thin layers of vertically oriented carbon nanotubes between each woven carbon fiber sheet; similar to the carbon “fabric” shown in Figure 6.10.
Throughout this chapter the new composite material will be referred to as a Nanotube
Reinforced Laminant Composite (NRLC). Figure 6.11 provides a schematic diagram of how a typical NRLC would be constructed.
Figure 6.10: Typical Carbon Fiber Fabric
162
Epoxy Resin
Woven Carbon Fiber Fabric CNT Array
Figure 6.11: Schematic Cross-Section of NRLC Specimen
To move forward with this technology, the methods used to produce such a composite material must be carefully developed. If the bonds between the CNT layers and the epoxy are not sufficiently strong, the material will fail below its true failure limit. In- plane compression tests would probably not be affected, but in-plane or out-of-plane tensioning, or shear tests would probably be greatly affected by poor bonding between layers. Since nanotubes are inert, it is difficult to get substances to bond strongly with them. In addition there are other challenges that will need to be overcome before this can become a viable composite material. The following two sections outline some of the major challenges that were encountered while trying to produce the first composite samples.
3.1 Producing CNT Arrays for Laminant Type Composites
The thickness of the CNT array is crucial in the process of producing laminant composites. It was determined early on that the CNT array thickness should be somewhere between 300-400 microns. It may be advantageous to have arrays shorter
163 than 300microns, but this poses some challenges, as it was generally observed that shorter arrays would often have insufficient CNT nucleation in some areas. Providing some method for mixing the gases inside the chamber may help to diffuse the decomposed carbon to all areas of the substrate, and minimize any difference in the CNT nucleation rate across the substrate surface; using substrates that have not been left sitting for a long period of time will also help to reduce the chance of surface contamination that results in poor growth conditions.
Once a target length of 300-400 microns was established, it was simply a matter of finding a synthesis time that would produce these results. While the optimum time was not systematically determined, a 15 minute carbon deposition time resulted in substrate lengths slightly over 400 microns when substrates pieces, approximately 10mm x 10mm are used. Further study should be conducted to determine the optimum synthesis time, but shorter lengths are more difficult to produce repeatably and are also more difficult to handle when producing the NRLC specimen.
3.2 Producing NRLC Specimens
It was quickly discovered that handling the CNT arrays was very challenging. The primary issues associated with using these CNT arrays to produce a composite material were: (1) how to remove the CNT from the substrate without it tearing or breaking apart and (2) how to transfer the CNT arrays to the carbon fabric ply.
164 The first method tried consisted of placing the CNT array on the carbon fiber fabric, while still attached to the silicon substrate. The array was placed so it was upside down and the nanotubes were in contact with the carbon fiber sheet below. The sample was then put under pressure; epoxy was injected and hardened. The next challenge was how to remove the silicon substrate after the epoxy had bonded with it around the edges.
Physically pulling the substrate off was attempted unsuccessfully.. We then tried placing the whole specimen in hydro-fluoric acid, but this took several days to remove the substrate and damaged the CNT and carbon fiber layers in the process; not to mention that hydro-fluoric acid is also very dangerous to work with.
It was found that submersing the array in acetone would typically remove the CNT from the silicon substrate, leaving the array floating in the acetone. If the CNT array does not come off on its own, briefly sonicate the submersed substrate; this should cause it to detach from the silicon. Wax paper may then be used to pick up the CNT and move it to the carbon fabric for placement. The key is that once the array gets moist, it must stay moist otherwise it will shrivel; therefore it is important to transfer the array and produce the composite material specimen immediately.
3.3 Results of First NRLC Composite Sample
Figures 6.12 through 6.15 are ESEM pictures of first NRLC specimen prepared at the
University of Cincinnati. These images indicate that the CNT fibers became imbedded in
the epoxy resin and bonded with the carbon fiber fabric. These preliminary findings
provided a starting point for preparation and characterization of other NRLC specimens.
165
Figure 6.12: Edge of Cut Through the CNT/Carbon Fiber Sheet
Figure 6.13: Exposed CNTs Inside of One Crevasse (Indicated by the arrow in Figure 6.12)
166
Figure 6.14: Surface of CNT Array Imbedded in Epoxy Resin
Figure 6.15: “Toes” of the Carbon Fiber Fabric Showing Through the Epoxy Impregnated CNT Array
167
3.4 Current State NRLC Research
Over the past year and a half since these first ESEM images of the NRLC composite material were taken, Yi Song, a Ph.D candidate in Dr. Abot’s lab, has made substantial progress in developing methods to remove the CNT array and has used these processes to make several material specimens. The following figures and general descriptions provide the three removal techniques developed after my involvement with this research [4].
Figures 6.16 through 6.25 were generously provide by Yi Song for use in this thesis.
Figure 6.16: CNT Array Attached to One Sheet of Carbon Fiber Fabric (photo by Yi Song)
Figure 6.17: Cross-Section of Fully Assembled NRLC Specimen (photo by Yi Song)
168 3.4.1 Technique 1: Sonication Method
In this method, the CNT array with the substrate still intact is attached to the carbon fiber fabric ply by first bonding epoxy to the assembled layer in a vacuum bag. The completed layers are then placed in a beaker of acetone and sonicated in a water-bath. After a couple of hours, the substrates can be removed from the carbon fiber fabric and used to produce the sandwich construction illustrated in Figure 6.17. The disadvantages of this method are:
(1) Ultrasonic waves can damage CNT array
(2) Epoxy of one layers is only able to bond with the adjacent layer due to the curing
the epoxy twice.
Figure 6.18: Fabrication Steps in Sonication Method (figure by Yi Song)
Figure 6.19: CNT Array Substrates on Fabric Inside of Vacuum Bag (photo by Yi Song)
169
Figure 6.20: Schematic of Sonicator Set-up (figure by Yi Song)
Figure 6.21: CNT Array Bonded to Fabric Ply after Substrate Removal (photo by Yi Song)
3.4.2 Technique 2: Adhesive Tape Method
This method uses tape to remove the CNT array from the substrate surface. The carbon fiber fabric then has a layer of epoxy spread or sprayed onto it, and the CNT array is placed upside down on the fabric ply. Once the epoxy has cured the tape is peeled off of the back of the array. This method has worked well, but was modified into Technique 3 which is generally easier and reduces the number of steps.
170
Figure 6.22: Fabrication Steps in “Adhesive Tape Method” (figure by Yi Song)
Figure 6.23: Wafer Removal Steps in “Adhesive Tape Method” (photo by Yi Song)
3.4.3 Technique 3: Removal of Substrate Using Mechanical Twist Method
The mechanical twist method has proven to be the easiest and most effective method for removing the substrate from the CNT array. The technique consists of spreading or spraying a thin layer of epoxy onto the carbon fiber fabric. The CNT array is then placed upside down and allowed to bond with the thin epoxy layer. Once the epoxy has cured, the substrate is gently rotated and simply pulls away leaving the array intact upon the surface of the carbon fiber fabric.
171
Figure 6.24: Fabrication Steps in “Mechanical Twist Method” (figure by Yi Song)
Figure 6.25: Wafer Removal Steps in “Mechanical Twist Method” (photo by Yi Song)
4.0 Fundamental Conclusions Regarding Impedance Measurements
(1) Preliminary experiments indicate that impedance decreases as nanofibers are
dispersed in an epoxy resin.
(2) Bubbles formed during mixing tend to increase the impedance and would need to
be minimized in order for this method to be useable.
(3) The resistance tends to be impacted most by the dispersion of nanofibers and was
reduced by approximately 10ohms during the 12.5 minute sonication time.
172 5.0 Future Work Regarding Impedance Measurements
The following items should be considered as potential areas where further investigation is warranted:
(1) Optimization of electrode design to improve impedance measurement.
(2) Methods for controlling the velocity of the mixture as it makes contact with the
electrode surface may help to improve measurement accuracy.
(3) Develop an automated method for continually sampling the mixture. Removing a
small specimen from the larger volume may make it easier to obtain accurate
impedance measurements. Consider intermittent testing of small samples
extracted from the total mixture volume.
173 6.0 References
[1] Anonymous 2001, "Http://www.Apsci.com/ppi-pyro3.Html," 2006(May/28) pp. 2.
[2] Alciatore, D.; Histand, M., “Introduction to Mechatronics and Measurement System” McGraw Hill, Boston, 2007.
[3] Narasimhadevara, S. 2005, “Processing of Carbon Nanotube Epoxy Composites”, ICCE Conference Proceedings
[4] Jandro L. Abot, J., Song, Y., Schulz, M., Shanov, V., “Novel Carbon Nanotube Array- Reinforced Laminated Composite Material”
174
Chapter 7
Fundamental Conclusions
175 1.0 Introduction
The following outlines the fundamental conclusions obtained through my extensive study of carbon nanotube synthesis.
2.0 Hydrogen Pre-Deposition Phase
Hydrogen introduced into the CVD reactor prior to injection of the carbon pre-cursor gas
significantly increases CNT array height. This result may occur because the hydrogen
removes oxide formed on the CNT substrate surface before carbon atoms begin to deposit
on the substrate. It is anticipated that the oxide layer would be more prevalent the longer
the substrates are allowed to sit in air before being used to produce CNT arrays. The
hydrogen may be most effective on highly oxidized substrates.
3.0 Effects of Plasma Oxidation
Plasma oxidation tends to remove material from the substrate surface, thus reducing the
surface roughness and thickness of the 15nm thick aluminum film. While not clearly
identified, there is likely a range of plasma oxidation settings that are suitable to many
different substrate types. It was found that shorter oxidation times are required for
prolonged CNT growth. Longer oxidation times may deplete the aluminum film to an
extent that exposes the iron or iron alloy catalyst particles to the silicon below.
4.0 Iron Catalyst for Growing CNT Arrays
CNT synthesis using 100% iron catalyst particles does not generally produce CNT arrays
over 2nm in height. The arrays are prone to breaking apart and often have very poor
176 structural uniformity and numerous visible defects. With the discovery of the benefits of using gadolinium in combination with iron, there is now little need to spend significant time studying the iron catalyst substrates; however gaining a better understanding of the thermal physics of how the CNT nucleates and what chemical compounds make up the catalyst particles throughout the CNT growth process would be beneficial. The information obtained may elucidate how gadolinium or other lanthanide metals promote super-long CNT growth over large surface areas.
5.0 Gadolinium as a Catalyst Motivator
It was determined conclusively that 100% gadolinium cannot be used as a catalyst for
CNT nucleation in a CVD furnace; however, gadolinium in combination with iron
“motivates” the catalyst and significantly improves the length, structural uniformity, and repeatability of carbon nanotube arrays over synthesis using 100% iron. Arrays consisting of double-wall and multi-wall CNTs were produced using 20% Gd, 80% Fe substrates. Some of these arrays were up to 4.5mm long and grown over large surface areas (up to 1in2).
One potential reason gadolinium facilitates better CNT growth is that adding gadolinium to iron lowers the melting point drastically. This may allow carbon atoms to be deposited sooner and at lower catalyst particle temperatures. Gadolinium may also help to regulate carbon flux in some way. These two possibilities might help to minimize the build-up of an amorphous carbon layer on the catalyst particle surface.
177 It was observed through line EDS using an HR-TEM that combinations of gadolinium and iron nanoparticles are encapsulated inside the CNTs. This could prove very useful for imaging applications where gadolinium is required, but cannot be used for medical imaging due to toxicity of the patient. This is particularly true for patients with renal failure.
The CNT arrays produced using 20% Gd, 80% Fe substrates seem to be relatively free of amorphous carbon and structural defects. Visual inspection using ESEM and HR-TEM indicated that amorphous carbon was not abundant; TGA analysis did not show any mass change at the oxidation temperature of non-crystalline (amorphous) carbon and Raman spectroscopy indicated that the D peak was only slightly reduced by thermal treatment to remove amorphous carbon and “heal” defects in the CNT walls,.
Combinatorial and layered substrates produced using PLD did not provide any useful information on how these substrates function. ICP analysis indicated that no gadolinium was present in the CNT sample produced using 20% Gd, 80% Fe substrates. The same was true for CNTs produced using 100% Fe substrates. The fact that gadolinium was undetectable with the ICP instrument may be the result of an involved sample preparation process. Laser ablation ICP may have some desirable features; the use of this tool should be considered and investigated.
6.0 Impedance as a Method for Assessing Carbon Nanofiber Dispersion in an Epoxy
Preliminary experiments indicate that impedance decreases as nanofibers are dispersed in
an epoxy resin. Based on these experiments, it was determined that the resistance of the
178 epoxy nanofibers mixture decreases by approximately 10ohms during a 12.5 minute sonication time. The total system impedance decreases linearly with the sonication time.
Mixing is required to disperse the clustered nanofibers; however numerous air bubbles, which tend to affect impedance measurements, are produced in the epoxy. A method for periodically extracting small samples of epoxy, placing them in a vacuum to remove all of the bubbles and measuring the impedance of these samples may help to significantly improve the accuracy of impedance measurements during the mixing and sonication process.
179
Chapter 8
Future Work
1.0 Introduction
The intent of this chapter is to summarize potential avenues of future work and experimentation that have been touched upon throughout this thesis. Much of the potential work that could be conducted is an extension of experiments that are summarized within these chapters. There is great value in conducting further experiments in these areas to confirm the repeatability of experiments,, to develop new
180 methods for improving the experimental process and, in general, to obtain a greater understanding of these systems and methods.
2.0 The Effects and Role of Plasma Oxidation on CNT Array Growth
Originally it was believed that plasma oxidation was merely a way to control the structural characteristics of the catalyst particles. New experimental evidence has indicated that the alumina layer produced as a result of aluminum oxidation may interact chemically with the catalyst particles, improving CNT growth.
Further efforts should be made to develop the optimum bounds for the plasma oxidation parameters: power, oxygen and argon percentages, vacuum pressure and time. It is likely that there is not one clearly identifiable optimum setting; however some degree of control over the catalyst particle density, grain size and surface adhesion force can be maintained.
In addition, studies should be conducted to ascertain how the oxidation changes the alumina layer chemically, and how this chemical change promotes catalysis.
One experiment that should be conducted is to deposit the catalyst on the 15nm layer of aluminum without any oxidation. This would provide a baseline condition for additional studies. Additional growth studies at different oxidation settings should also be conducted. Some form of surface spectroscopy may also be useful in determining the chemical make up the supporting alumina surface and oxidized catalyst particles.
Currently XPS is being used to study the chemical species on the substrate surface, however other methods such as Mossbauer spectroscopy should also be considered.
181 Additional AFM studies should be conducted to gain a better understanding of how plasma oxidation affects the catalyst nanoparticle size, density and adhesion force. It would be a great benefit to have a method for measuring the adhesion force of the particles and, through measurement of the other pertinent particle properties, determine in an average sense whether the catalyst particles will stay on the substrate surface or be lifted off as the CNTs grow.
3.0 Determining Whether Top-growth or Bottom-growth is Dominant
Recent in situ visual observation of the CNT growths tends to indicate that the catalyst particles remain on the surface of the substrate. Additional studies should also be conducted to confirm this new observation. One possible method is to remove the CNT array from the substrate after synthesis and determine the density of the CNTs making up the array.. The substrate without the CNT array can then be put back into the furnace and synthesis conducted again using the same substrate. If another CNT array is produced, this may indicate that the CNTs are growing from catalyst particles that remain on the surfaces of the substrate. However; it is possible that catalyst particles not activated during the initial growth were used. To rule this out, the CNT density of the new array and the previous array should be compared. If they are similar, this would tend to indicate that the catalyst is remaining on the surface of the substrate. Similar studies were conducted by Iijima et al [1].
182 4.0 Iron Catalyst Substrates
Given that iron combined with gadolinium produces substantially better growth results than iron alone, the study of 100% iron catalyst should shift to a new regime. I propose that the thermal physics and chemistry of why iron functions as growth catalyst should be thoroughly studied with the intent of understanding how other metals combine with the iron to promote CNT growth. Numerical simulations of the active catalyst particles may be useful in studying these chemical interactions. COMSOL is a widely used multi- physics package that may be used to study these interactions for comparison with experimental results.
5.0 Combination Gadolinium and Iron Substrates
A significant amount of research is currently being done to obtain a better understanding
of why gadolinium assists the catalyst and how it might be optimized for improved
growth. The majority of research avenues that I would recommend are currently
underway at UC. However one area which requires further work is with regard to using
these particles to improve MR imaging.
5.1 Optimize Gadolinium to Iron Ratio
As discussed in Chapter 6, maximizing the percentage of gadolinium with respect to iron
is critical in producing the optimum MR imaging response. CNT synthesis has been
achieved with 95% gadolinium and 5% iron substrates; while long CNT growth was not
observed, the results are still very valuable since long CNTs are not necessary for
imaging In fact very short CNTs would be preferred for injection into a living organism.
183 The two experimental paths listed below should be fully investigated to determine the best way to verify conclusively if gadolinium encapsulated inside nanotubes improves imaging response over the same amount of gadolinium.
(1) Grow very short (<100 micron long) CNTs from substrates that have the
maximum allowable gadolinium to iron ratio. Line EDS using the HR-TEM at
the University of Kentucky should be conducted to determine if gadolinium is
clearly detectable in the catalyst particles; and if so, determine the resulting
percentage of gadolinium to iron.
(2) Conduct imaging response experiments using CNTs produced using a high ratio
of Gd to Fe. Accurate baseline experiments should be conducted first to form a
comparative imaging response study.
5.2 A New Method for Determining if Gadolinium is Encapsulated
Initial efforts to use ICP mass spectrometry that uses a nebulizer to “vaporize” the sample being analyzed have been unsuccessful. It is my belief that this method is not the optimum methodology for determining the percentage of gadolinium that is encapsulated inside of a CNT array. As discussed in Chapter 3, during the process of converting the
CNT array into a form that can be converted into a spray without clogging the nebulizer, the gadolinium particles are being reduced to levels that are below detectable limits. It is also possible that errors were made in the sample preparation or the ICP analysis.
184 Another form of ICP analysis that may provide better results is laser ablation ICP. This method is highly desirable for the following reasons [2]:
(1) Direct analysis of powders and solids
(2) No need to digest samples in acids
(3) Eliminates contaminants during the sample preparation
(4) Can perform surface mapping
6.0 Acoustic Waves for Improving Carbon Flux across Large Surface Areas
The following outlines the motivation for using acoustic waves to improve carbon flux over large substrate surfaces with the intent of reducing CNT array height variation and increasing growth length. Preliminary experiments were conducted to assess the feasibility of introducing acoustic energy into the CVD reactor.
6.1 Motivation
As illustrated in Figure 1 below, as the surface area across which the CNTs are grown
increases, the thickness at the center of the CNT arrays tends to be less than the edges
(Figure 1). This is most likely due to decreased carbon flux at the center of the array
(Figure 2). As the CVD process is scaled up for large-scale manufacturing purposes, this effect will likely become more of an issue, particularly when trying to use these CNT surfaces for structural applications.
185 Figure 8.1: 1in2 CNT Array with Decreased Thickness at the Center
Figure 8.2: Schematic of Carbon Flux across Substrate Surface
(1) Decomposed carbon atoms are depleted first as they are absorbed by the nanoparticles they come in contact with (2) Catalyst nanoparticles further upstream and at the center of the CNT arrays are deprived of carbon
To address this issue, mixing of the gases to more rapidly and evenly distribute the decomposed carbon atoms across the CNT array surface should be investigated. Mixing could be achieved through mechanical means, however due to the high temperature environment this may be difficult. One option would be to provide multiple mixing chambers through which the gases would flow before reaching the substrate. The mixing could be accomplished by inducing turbulence at multiple stages; this should help to mix the gases and distribute the carbon more evenly inside the CVD furnace tube. Figure 3 below provides a preliminary concept for mixing. Since these mixing chambers would introduce substantial pressure-drops, greater pressure (i.e. gas flow rates) may be required. A computation fluid dynamics model could be used to evaluate design alternatives for improving mixing and thus carbon atom diffusion.
186
Figure 8.3: Concept for Mixing Chambers (Cross Section thru CVD Furnace Tube)
(1) Perforated barriers with short tubes to produce turbulence (2) Finely perforated barrier to smooth-out fluid flow (3) CNT array substrate
6.2 Motivation
It has been briefly discussed that acoustic energy may potentially be harnessed to improve carbon diffusion across the CNT array surface. Assuming the CNT array would be located at the center of the CVD furnace tube, a number of acoustic modes should be present where minimum acoustic pressure, and therefore maximum particle velocity (90° out of phase with the pressure), occurs at the location of the substrate [3].
By introducing acoustic energy into the CVD reactor at the appropriate frequency or frequencies, gas particles could be oscillated across the substrate surface (Figures 8.4 &
8.5). The oscillatory displacements may help to distribute carbon atoms across the substrate surface, while high particle velocity may help to physically push the carbon atoms into the dense carbon nanotube forest (Figure 8.5). Substrate preparation methods used to produce CNT arrays at UC should generally result in top-growth CNTs, however it is probable that bottom-growth CNTs are also present. Top-growth CNTs could
187 potentially benefit from large particle displacements that move the decomposed carbon atoms across the substrate surface; while bottom-growth CNTs may benefit more from high particle velocities that help the carbon atoms to penetrate deep into the dense nanotubes array where they can come in contact with the sides of the growing nanotube and be attracted to the catalyst nucleation site through surface diffusion; (Refer to
Chapter 2 for a brief discussion of surface diffusion [4, 5]).
Figure 8.4: Stages of Carbon Atom Oscillation inside the CVD Reactor
(1) Hydrocarbon gas decomposes in carbon atoms (2) Decomposed carbon atoms are oscillated at the substrate location (3) Unabsorbed carbon atoms are exhausted
Figure 8.5: Oscillation Diagram of Decomposed Carbon Atom
188
Figure 8.6: Schematic Indicating Benefits of Acoustic Excitation
(1) Distribution of carbon atoms to top-growth catalyst particles imbedded in the CNT “forest” (2) Increase CNT surface diffusion to bottom growth catalyst particles through greater penetration in the CNT “forest”
6.3 Experiment Design
In order to properly design a system for introducing acoustic energy into the CVD reactor at high amplitude, it is critical to analyze the system to ensure that it is not filtering out the frequencies of interest. Ideally in situ measurements would be used to analyze the system; however this is not possible due to the high temperature environment.
To assist in the experiment design process, an analytical model describing the mechanical system supplying the acoustic energy, and the acoustic environment inside the CVD reactor, is desirable. In order to calibrate, a replica CVD reactor tube at ambient temperature in air was used to conduct some preliminary measurements of the tube resonances. The results of these experiments are provided below.
To determine the acoustic modes of the CVD under operating temperature and gas mixture percentages, the speed of sound inside the CVD reactor must be determined. By
189 knowing the approximate ratio of gases and the temperature inside the CVD furnace, the speed of sound can be calculated using the following relationship where T is the temperature in degrees Kelvin, the gas constant R and “gamma”, the ratio of specific heats, are determined for the appropriate gas mixture percentages.
c = γRT
Equation 8.1: Speed of Sound in an Ideal Gas
Once the speed of sound is determined through the ideal gas relationship above, the acoustic modes can be calculated by assuming that the CVD furnace behaves as a circular
“duct”, approximately 3’ long that is closed at both ends. The “closed” boundary conditions are not exact, and a sensitivity study should be conducted to see how changes
to this assumption change the modal frequencies. The measured results from the replica
furnace should also be used to calibrate the analytical model. Matlab code to calculate
the acoustic mode of a rectangular tube is provided in the appendix.
6.4 Preliminary Experiments
Many thanks to Dr. Allemang, Dr. Kim, and Dr. Lim for providing me with
instrumentation to conduct these experiments, as well as their guidance in the design of
the preliminary experiments and interpretation of the data.
Some preliminary experiments were conducted to assess the feasibility of introducing
acoustic energy into the CVD reactor. In general the concept is to introduce acoustic
energy at one or many frequencies that produce maximum particle velocity at the
190 substrate location. This may be accomplished by connecting a horn driver (speaker), via a tube, to the CVD reactor. The tube connection to the CVD reactor should be designed to minimize filtering at the frequencies of interest.
The experimental apparatus consisted of the equipment listed below and indicated in
Figure 8.7.
(1) Hewlett Packard FFT Analyzer
(2) Signal Generator
(3) Amplifier
(4) Horn Driver
(5) Microphone at connection to replica CVD reactor tube
(6) Bruel & Kjaer Microphone Preamp
(7) Bruel & Kjaer Condenser Microphone
(8) Replica CVD reactor tube
A glass replica of the CVD reactor was provided at no cost. This was done because quartz is considerably more expensive than glass, and breaking the tube mechanically or through acoustic excitation was a concern.
191
Figure 8.7: Test Setup to Measure Acoustic Modes of Replica CVD Reactor
192
Figure 8.8: Schematic of Acoustic Measurement Chain
6.5 Results
The following frequency response function (FRF) plots were measured at 8 positions along the length of the glass tube to develop an understanding of the modal characteristics of the replica CVD reactor for use in calibration of an analytical model to design the system for introducing acoustic energy into the CVD reactor. The plots were generated by taking 100 linear averages of random pink noise.
Frequency Response - Position 1 Microphone at left end of the tube 10
1
0.1 Output (Volts) / Input Voltage Input / (Volts) Output
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz)
Figure 8.9: Acoustic Pressure FRF of CVD Reactor – Position 1
193 Frequency Response - Position 2 Microphone 1/8 of Tube Length 10
1
0.1 Output (Volts) / Input Voltage / Input (Volts) Output
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz) Figure 8.10: Acoustic Pressure FRF of CVD Reactor – Position 2
Frequency Response - Position 3 Microphone 1/4 of Tube Length
10
1
0.1 Output (Volts) / Input Voltage
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz)
Figure 8.11: Acoustic Pressure FRF of CVD Reactor – Position 3
194 Frequency Response - Position 4 Microphone 3/8 of Tube Length 1.00E+01
1.00E+00
1.00E-01 Output (Volts) / Input Voltage
1.00E-02 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz))
Figure 8.12: Acoustic Pressure FRF of CVD Reactor – Position 4
Frequency Response - Position 5 Microphone 1/2 of Tube Length
10
1
0.1 Output (Volts) / Input Voltage
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz)
Figure 8.13: Acoustic Pressure FRF of CVD Reactor – Position 5
195 Frequency Response - Position 6 Microphone 5/8 of Tube Length 10
1
0.1 Output Voltage / Input Voltage
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz) Figure 8.14: Acoustic Pressure FRF of CVD Reactor – Position 6
Frequency Response - Position 7 Microphone 3/4 of Tube Length 10
1
0.1 Input Voltage Voltage / Output
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz)
Figure 8.15: Acoustic Pressure FRF of CVD Reactor – Position 7
196 Frequency Response - Position 8 Microphone 7/8 of Tube Length
10
1
0.1 Input Voltage / Output Voltage
0.01 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 Frequency (Hz)
Figure 8.16: Acoustic Pressure FRF of CVD Reactor – Position 8
6.6 Conclusions & Future Work
The most prominent mode appears to be at approximately 320Hz. The frequency and mode shape of the 320Hz mode should be confirmed through a calibrated analytical study. Once the results are confirmed, work should continue on determining the modal frequencies and mode shapes inside the CVD reactor during CNT growth. A paper study should be conducted to determine what velocity and displacement range would be the most desirable. Once the desired range of frequencies and the electro-acoustic power required to produce the desired displacement and/or velocity are determined, the CVD reactor should be outfitted with a properly designed system for introducing the acoustic energy. A preliminary comparative study using well developed CNT growth conditions
197 should be conducted to determine if the addition of acoustic energy provides any substantial benefits.
7.0 Other Uses for Acoustic Excitation
Ultrasonic frequency waves may potentially be used to assist in a number of different
CNT synthesis issues:
(1) Use ultrasonic waves to “shake” the CNT array. This may allow more carbon
atoms to penetrate into the dense CNT “forest” as spaces between CNTs or
groups of CNTs are formed. A metal CVD tube should be used due to
potential fatigue produced during ultrasonic excitation.
(2) While cavitation is not likely to occur in the all gas CVD reactor environment
(two-phase mixture required), ultrasonic waves may still help to remove
amorphous carbon by breaking it down, or even shaking it loose from the
array.
Ultrasonic waves may also help to provide additional energy to the chemical system in hopes that it may prolong the catalyst particle “life”.
198 8.0 References
[1] Futaba, D., Hata, K., Yamada, T., 2005, “Kinetics of Water-Assisted Single-Walled Carbon Nanotube Synthesis Revealed by a Time-Evolution Analysis”, Physical Review Letters, vol 95.
[2] Agilent Technologies, Inc. 2005, “Inductively Coupled Plasma Mass Spectrometry: A Primer”, Publication Number 5989-3526EN, USA
[3] Kinsler, L. Frey, A., Coppens, A., “Fundamentals of Acoustics” Wiley, New York, 2000.
[4] Louchev, O., Thomas, L., Sato, Y., 2003, “Diffusion-controlled kinetics of carbon nanotube forest growth by chemical vapor deposition”, Journal of Chemical Physics, vol. 118.
[5] Louchev, O., Thomas, L., Kanda, H., 2002, “Growth mechanical of carbon nanotube forest by chemical vapor deposition”, Applied Physics Letters, vol. 80.
199
APPENDICES
200
I. Images of Nanotubes Grown from 100% Iron Substrates
201
202 II. ESEM & HR-TEM Images of Nanotubes Grown from 20% Gadolinium, 80% Iron Substrates
203
204
205
206
207
208
209
210
211
III. ICP Data Sheets
212
213
214
215
216 IV. UC Clean Room Layout
217 V. “Temescal” Electron Beam Evaporator Product Data Sheets
218
219
220
221
222
223
224
225 VI. “March” Plasma Oxidation System Product Data Sheets
226
227
VII. MATLAB Code for Computing Modes of Rectangular Tube (Needs Revision) %cut off frequency for this problem is when w=688*pi. At that frequency %the pt matrix is all infinite values. clear all, close all; rho=1.168; U0=.001; np=50; % We use 50 x 50 points for plotting nterm=10; % no. of non-plane wave mode terms to use in the expansion Ly=1; % height of the duct w=700*pi; c=344; k=w/c; zmax=3; % z range; therefore we plot 1 m by 3 m duct space y=linspace(0,1,np); % sample y by np points z=linspace(0,zmax,np); % sample z by 50 points pt = zeros(np,np); % initialize for m=1:nterm % Calculate km kdet=k*k-(m*pi/Ly)^2; if kdet >= 0.; kzm(m)=sqrt(k*k-(m*pi/Ly)^2); % Propagating modes else kzm(m)=-sqrt(kdet); %Evanescent modes; note the - sign in front. This is to make sure the mode is e^(-kmz): end end for m1=1:np % m1 = iteration for y points for m2=1:np % m2 = iteration for z points pt(m1,m2)=-1/3*rho*(U0/k)*exp(-j*k*z(m2)); %Plane mode term for m3=1:nterm % m3 is mode terms, this loop is non-plane mode terms. this is to sum up all modes that are being considered. pt(m1,m2)=pt(m1,m2)+2*rho*U0/(m3*pi*kzm(m3))*(sin(2*m3*pi/3)- sin(m3*pi/3))* ... cos(m3*pi*y(m1)/Ly)*exp(-j*kzm(m3)*z(m2)); %pt(m1,m2)=pt(m1,m2)+2/(m3*pi)*(sin(2*m3*pi/3)-sin(m3*pi/3))* ... %cos(m3*pi*y(m1)/Ly)*exp(-j*kzm(m3)*z(m2)); end end end pmag=abs(pt); clim=max(max(pmag)); clim1=min(min(pmag)); V=[-clim clim]; % sets color scheme with caxis(V) % red: maximum positive % blue: minimum negative pressure
pcolor(z,y,real(pt)); axis equal shading interp; caxis(V);
228 axis([0 zmax 0 Ly])
VIII. “Nanotechnology Store” - Price List and General Information The Nanotechnology Store
University of Cincinnati Smart Materials Nanobiotechnology Laboratory
Product Catalog
The University of Cincinnati (UC) Nanotechnology Store will sell products based on nanotechnology that are developed at UC. The products will include nanotube and nanowire raw materials, biosensors, nanotube laced polymers, and NanoSpray Film for broad applications. The store is nanoprofit. All revenues go to support research. This document is a catalog of products. The catalog will be regularly updated with new products.
Ordering Information. Contacts:
Mark J. Schulz Vesselin Shanov University Of Cincinnati University of Cincinnati Department of Mechanical Engineering Department of Chemical and Materials Engineering Smart Materials Nanotechnology Lab Cincinnati, OH 45221-0012 598 Rhodes Hall Phone: (513) 556-2461 Cincinnati, OH 45221-0072 Fax: (513) 556-3773 Phone: (513) 556-4132, Fax: (513) 556-3390 E-mail: [email protected] Email:
229 Product Description: Multi-Wall Carbon Nanotube Arrays
Research grade multi-wall carbon nanotube (MWCNT) arrays may be purchased from the University of Cincinnati, Smart Materials Nanotechnology Laboratory.
Carbon Nanotube Array Specifications The carbon nanotube arrays you will receive are grown on 500-530µm thick Silicon/Silicon-Oxide wafers. Catalyst film deposition and other processes are proprietary. The carbon nanotubes are multi-wall structures, typically comprised of 15 to 20 shells. Outside diameters typically range from 20 to 40nm. Raman spectra and other characterization information for thermally treated arrays and untreated arrays can be provided. The arrays may be used for reinforcing polymers, biosensors, and for making sensors and actuators. The arrays have high electrical and thermal conductivity in the vertical direction. The nanotubes in the arrays may also be dispersed in a solvent and dried and used in the powder form. The mm long nanotubes provide improved properties in many applications compared to micron long nanotubes produced by other processes. The prices are given below.
Prices for MWCNT Arrays. CNT Film Thickness Area <200µm 200-500µm 500µm-1mm 2mm 3mm 4mm
1 mm2 $20 $40 $80 $120 $160 $200
5 mm2 $40 $80 $120 $160 $200 $240
1 cm2 $80 $120 $160 $200 $240 $280
2 cm2 $160 $200 $240 $280 $320 $360
4 cm2 $240 $280 $320 $360 $400 $440
5 cm2 $320 $360 $400 $440 $480 $520 6 cm2 $400 $480 $560 $600 $700 $800
Additional Products and Services related to nanotube arrays. 1. Thermal Treatment of Arrays for higher purity - $20.00 per run. 2. Larger area arrays are available by quote. 3. Gram quantity of 4 mm long MWCNT thermally treated - $4,000. About a 120 cm2 area of 4 mm long MWCNT will produce one gram of nanotubes. Larger quantities are available by quote. 4. Functionalized arrays are available by quote. Acid functionalization, plasma functionalization, and biofunctionalization are available. 5. Patterned arrays are available by quote. Examples of the arrays are shown below.
230
1cm2 Sample 6cm2 Sample
Patterned arrays of different shapes and sizes can be grown on substrates. The UC logo shown contains millions of individual MWCNT.
1 mm
231
IX. C-Gd-Fe Ternary Phase Diagrams
232
233
234