Synthesis and Electronic Transport in Known Chirality Single Wall Carbon Nanotubes

Bhupesh Chandra

Submitted in partial fulfillment of the Requirements for the degree of Doctor of Philosophy in the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY 2009

 2009 Bhupesh Chandra All Rights Reserved

ABSTRACT Synthesis and Electronic Transport in Known Chirality Single Wall Carbon Nanotubes Bhupesh Chandra
Carbon nanotubes are intriguing new materials with extraordinary electrical properties originating from their quasi 1-dimensional structure and a strong, all carbon lattice. They have great potential for use in electronic and energy related applications, such as low resistance metal interconnects, optically transparent conducting films for display applications, high surface area catalytic support, etc. However, a major roadblock for practical usage of carbon nanotubes is the limited understanding of its synthesis and electrical properties. The first half of the thesis probes into the existing challenges of carbon nanotube synthesis through chemical vapor deposition (CVD) process. Nanotubes grown with various synthesis techniques are characterized with Rayleigh spectroscopy, enabling the development of finely tuned recipes that deliver clean and unbundled single wall carbon nanotubes. Two technological applications which directly utilize nanotubes coming from CVD growth are discussed. The first application uses them as high surface area electrodes in a enzyme mediated bio-fuel cell and the second application finds their use as transparent and conducting films. In the second half of the thesis, electrical transport results on chirality assigned nanotubes are presented. Electronic measurements on a known structure carbon nanotube metalsemiconductor heterojunction suggested the presence of quantum confinement at the point of chirality change; this observation is explained through detailed theoretical calculations. Finally, through an extensive study of electron-phonon scattering on known structure single wall nanotubes, it is concluded that the substrate phonon modes have a dominating effect in nanotube electron scattering. The role of substrate phonon scattering is studied through experimenting with nanotube devices placed on multiple substrates with different phonon energy modes. This result has far reaching implications in the field of nanotube transport.

Contents
1 Introduction to Carbon Nanotubes ................................. 1 1.1 sp2 Hybridization ... 2

1.2 Graphene .... 3 1.3 Carbon Nanotubes .. 6 1.4 Nanotube Crystal Structure .... 7 1.5 Nanotube Electronic Dispersion .... 1.6 Nanotube Phonon Dispersion 9 12

1.7 Summary 14 Bibliography ..... 15 2 Carbon Nanotube Synthesis - Processes and Concepts . 17 2.1 Introduction 18 2.2 Chemical Vapor Deposition ... 20 2.2.1 Carbon Precursor 20 2.2.2 Temperature ... 2.2.3 Catalyst Nanoparticle . 22 23

2.2.4 Gas Flow Rates ... 24 2.3 CNT Growth Experiments . 25

2.3.1 Motivation and Aims .. 25 2.3.2 Growth Setup .. 26 2.3.3 Carbon Monoxide Growth . 2.3.4 Ethanol Growth ...... 27 30

2.3.5 Methane Growth ......... 34 2.3.6 Water-Assisted Ethylene growth ........ 36 2.3.7 Nanotube Growth- Fluid Flow Study ......... 37 2.3.8 Conclusion .. 41 Bibliography . 42

3 Carbon Nanotube Synthesis - Technology Applications ... 3.1 Growth of CNTs on Carbon Toray Paper for Bio-Fuel Cell Applications. 3.1.1 3.1.2 3.1.3 3.1.4 Introduction Experimental Setup and Process Parameters .

46 47 47 48

Growth Results ... 50 Summary 53 54 54

3.2 Growth of Transparent, Conducting CNT Films ... 3.2.1 3.2.2 3.2.3 3.2.4 Introduction

Experiments 55 Measurements Summary 58 61 62

Bibliography .

4 Introduction to Electron Transport in Carbon Nanotubes ..........................

65

4.1 Introduction ............ 66 4.1.1 4.1.2 4.1.3 4.1.4 4.1.5 4.1.6 Metallic and Semiconducting CNTs.. 66

Quantum Transport in a 1-d System .. 66 Schottky Barriers at the Nanotube- Metal Interface .. Electronic Mean Free Path and Phase Coherence Length.. Nanotube Band Diagrams-Field Effect Transistor Action 67 68 69

Coulomb Blockade in CNTs .. 71

4.2 Carbon Nanotube Device Fabrication 77 4.2.1 4.2.2 4.2.3 4.2.4 Introduction 77 Rayleigh Scattering Spectroscopy . Nanotube Transfer Printing ... Metal Electrode Deposition ... 78 81 83

Bibliography . 85

ii

5 Molecular-Scale Quantum Dots from Carbon Nanotube Heterojunctions 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 Introduction ......... Device Fabrication ..

87 88 88

Electronic Transport Measurements 90 Constant Interaction Model . 93

Sequential Tunneling ... 95 Theoretical Modeling .. Explanation for Large Central Diamond and Diode-Like Behavior ... 97 102

Negative Differential Resistance (NDR) . 103 Conclusion ... 104 Acknowledgements . 105

Bibliography . 106 6 Electron Phonon Scattering in Known Chirality Single Wall Carbon Nanotubes... 110 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 Introduction ......... 111 Electron-Phonon Scattering. 112

Electron-Acoustic Phonon Scattering- Theoretical Background 113 Electron-Acoustic Phonon Scattering-Previous Experiments Measurement Method . 115 117

Electrical Transport Measurements T=300K .. 118 Electrical Transport Measurements with Varying T .. Analysis of Results . Substrate Polarized Phonons (SPP) Summary 122 125 126 133 135

Bibliography

Publications . 138 .

iii

List of Figures
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 2.1 2.2 2.3 2.4 2.5 2.6 2.7 sp2 hybridization of carbon and its derived materials 2

Structure of graphene in real and reciprocal space . 3 Graphene energy dispersion ... 5

Single and multiwall carbon nanotubes .. 6 Nanotube crystal structure .. 8 Band structure and Density of State (DoS) Linear dispersion of metallic nanotubes near K-point ... Nanotube phonon dispersion .. Current CNT growth mechanism hypothesis . A schematic of the nanotube growth wafer. CVD growth system ... Carbon monoxide grown nanotubes . Structure of Fe-Mo catalyst nanocluster (Mueller Catalyst) . 10 11 13 19 26 26 28 30

Carbon nanotube CVD growth computer interface 32 Rayleigh scattering spectrum on two different suspended (16, 16) armchair carbon nanotubes 33

2.8 2.9

Diameter and chirality distribution of single wall carbon nanotubes . 34 Methane growth results .. 35 37 38

2.10 Water-assisted ethylene growth results .. 2.11 Flow study surrounding a raised structure on silicon chip .

2.12 Flow movement in side plane of the slits and raised structure ... 39 2.13 Flow chart describing CVD growth parameters studied 3.1 3.2 3.3 3.4 3.5 Schematic of the nanotube growth system Scanning electron images of CNTs grown on carbon paper .. MWNT growth at different time intervals . 41 48 50 51

Effect of carbon nanotube growth time on surface area . 52 Various process parameters studied for carbon nanotube film growth .. 56

iv

3.6 3.7

SEM images of nanotube film growth on quartz substrates ... 57 Light transmission and sheet resistance measurements on as-grown nanotube films 58

3.8

Comparison with results in literature .

60

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9

Nanotube band diagrams .. Capacitive model of nanotube device .. Ground State lines of a nanotube QD ... Stability diagram of the nanotube quantum dot Quantization of nanotube bands ... Quantum dot transport .. Nanotube device fabrication process sequence Single tube Rayleigh spectroscopy setup . Nanotube Rayleigh scattering results ...

69 72 74 74 75 76 78 79 80 82 83

4.10 Nanotube transfer setup .... 4.11 Nanotube device on silicon wafer .

5.1

Device geometry and electrical measurements on semiconducting and metallic sections 89 90 91 93 94 96 96 98 99 100 101 102 104

5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9

Electrical measurements on semiconducting and metallic sections . Temperature dependent measurements across the heterojunction Transport across heterojunction Model of the metal-semiconductor nanotube HJ device .. Rate equations for sequential tunneling through a spin- degenerate level ... Coupling constants of metal nanotube and SC nanotube to QD .. Single heterojunction structures ... Calculated local density of states for different geometries ...

5.10 Defect migration ... 5.11 Interface DoS for double heterojunctions . 5.12 Band diagrams explaining that the central gap . 5.13 Explanation of the negative differential resistance (NDR) ..

6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9

Emission and absorption phonons near K point ... Conductance gate sweep for (21,9) nanotube ... Conductance gate sweep for (26,11) nanotube . Resistance vs. channel length for (26,11) nanotube at T=300K ... Conductance gate sweep for (26,11) nanotube with changing temperatures ... Resistance vs. channel length for (26,11) nanotube at different temperatures . Resistivity vs. temperature for (26,11) nanotube .. Experimental and theoretical resistivity values for (26,11) nanotube .. Experimental and theoretical resistivity values for (21,15) and (24,12) nanotube

113 118 119 120 122 123 124 129

130 131

6.10 Optical image of nanotube device with different substrate stripes ... 6.11 Comparison of experimental and theoretical resistivity values for semiconducting nanotube .

132

vi

List of Tables
6.1 Resistivity values for chirality assigned nanotubes at T=300K121 6.2 Parameters used for curve fitting data with SPP scattering...133

vii

Acknowledgements
Standing on the Shoulders of Giants

It was exactly six years ago when I accepted the offer to join PhD program at Columbia. As a fresh undergraduate I did not know what was in store for me, PhD seemed too big a decision, all I had was just a vague belief that things will turn out to be good. In all these years that belief did not get shattered, but gradually turned into reality, of course, with the help of numerous people that crossed my way during this course. This section is dedicated to them.

I still remember my first day, when Prof. Vijay Modi, then the department chair, introduced me to a new faculty, Prof. James Hone who later became my thesis advisor. Working with Jim has been a tremendously rewarding experience. His insightfulness in plethora of subjects helped me in taking on multiple projects. Apart from research work, Jim showed me how to achieve a good work-life balance. I must admit that I got inspired from him to go into Salsa dancing. Jim, thanks for giving me the absolute freedom and support in pursuing my interests, you have been a wonderful advisor.

How could I forget my lab colleagues with whom I spent most of my time during PhD. Robert Caldwell and Sami Rosenblatt, I will always miss our hour long discussions about current politics and life in general. Changyao, thanks for being my buddy in working late nights in lab and clean room, I am sure you will achieve great results with all the hard work. Mingyuan, you have been a great lab partner and someone I always looked upon for nanotube theory. I know you are the next one to graduate and I wish you all the luck in your job search. Anurag, Yuyao, Changgu, Adam, Bonnie, Josh, its been wonderful working with you guys all these years. Zhengyi and Nick, although I did not have a chance to work with you, I appreciate your help during my last year at Columbia.

viii

My sincere regards go to Mark Hybertsen, Jeffrey Neaton and Joydeep Bhattacharjee, without their help it would not have been possible to answer challenging theoretical problems that crowded my projects. I also want to thank Philip Kim, and his research group, especially Meninder Purewal, Yuri Zuev, Vikram Deshpande, and Mitsuhide Takekoshi. My earnest regards goes to Prof. Tony Heinz and his group for help in Rayleigh spectroscopy, special thanks to Matt Sfeir, Yang Wu, Hugen Yan, Christophe Voisin, and Stephane Berciaud. I also want to acknowledge Limin Huang (Prof. Stephen O Briens group), Gordana Dukovic (Prof. Louis Bruss group) and Inanc Meric (Ken Shepards group) for assisting in nanotube growth, catalyst synthesis and device fabrication.

List of people outside the research circle that helped me go through this long run is endless. I would especially like to acknowledge my long time friends Rohit and Anuj. I wont forget Anujs delicious cooking, and his nook for an evening tea (so British!). Rohit and I shared the liking towards going out and enjoy the city life (and frequent visit to Toms restaurant). Kanishka, Ritvik and Navya, I look forward to more outings and parties with you guys. Kanak, you have been a very supportive roommate all this time. I am sure we will spend more time together. My very special thank goes to my closest friend Shaifali Agarwal. Shaifali, your unwavering support helped me pass through the toughest of times, be it PhD completion or job search. This thesis would have looked awful without your meticulous proof reading. I feel fortunate to have your company.

Lastly, this work could not have been possible without the support of my family members, Mom, you and father are the sole inspiration behind this work. People say you cant do that bad when both your parents are PhDs, now I believe they were right. Brother, I always felt you as the safety net I could lean on if things go wrong, thanks for being there, always. Bua Ji, your letters have always been a source of power and courage during PhD, I hope to keep these values forever. And finally to my beautiful sister-in-law, your regular phone calls always cheered me up for the day. Now, I can also put a Dr. before my name as you do, well of course, I wont still be a real doctor.

ix

Dedicated to my mother and late father

1 ______________________________
CHAPTER

Introduction to Carbon Nanotubes ______________________________

1.1 sp2 Hybridization

a
2pz 2pz

sp2

2pz

2pz

sp2 sp
2

sp2 sp2

sp2

sp2

Figure 1.1: sp2 hybridization of carbon and its derived materials. a) The three sp2 hybridized orbital are in-plane, with 2p orbital orthogonal to the plane, and * denotes the bonding and anti bonding orbital. b) Graphene as the source of three different materials- Fullerene (left), carbon nanotube (Center) and bulk graphite (right). Nature Materials

Carbon is the sixth element of the periodic table. Each carbon atom has six electrons which occupy 1s2, 2s2 and 2p2 atomic orbital. It can hybridize in sp (e.g. C2H2), sp2 (e.g. graphite) or sp3 (e.g. CH4) forms. This property is unique to carbon in its particular group in periodic table. Discoveries of very stable nanometer size sp2 carbon bonded materials such as fullerenes[1], carbon nanotubes[2] and graphene[3] has stimulated the research in this field. In the context of this thesis we will focus mainly on Carbon nanotubes. Since carbon nanotubes derive most of the physical properties from graphene, it is therefore useful to understand this material first.

1.2 Graphene
y x Unit Cell ky kx b1 A B K Brillouin Zone

a1 a2 b2

Figure 1.2: Structure of graphene in real and reciprocal space. a) Crystal structure with positions of two equivalent atoms in the unit cell of graphene. The unit cell is represented by dotted parallelogram. b) 2-D Brillouin zone of graphene with location of high symmetry points K, M and ; b1 and b2 are the reciprocal unit vectors.

Graphene is a one-atom-thick planar sheet of sp2-bonded carbon atoms that are densely packed in a honeycomb crystal lattice[4]. A schematic showing the 1-D graphene lattice is shown in Fig. 1.1. Several graphene layer stacked over each other gives bulk graphite. These layers are loosely bonded to each other and hence can slip freely, making the extraction of graphene possible from graphite crystal through mechanical/chemical exfoliation[5]. Due to the presence of sp2 hybridized carbon atoms in one plane, graphene is atomically two dimensional. Fig. 1.2 shows the unit cell and the Brillouin zone of graphene; a1 and a2 are the unit vectors in real space and b1 and b2 are the reciprocal

4 lattice vectors. In a hexagonal lattice like this, the real space unit vectors can be written as:

r 3 a r 3 a a1 = 2 a, 2 , a 2 = 2 a, 2

(1.1)

Where a is the lattice constant given by

3 aC C where aC C is the carbon-carbon bond

length (0.144 nm). In the Brillouin zone (Fig. 1.2b) there are three high symmetry points at the center, corners and center edge. The approximate energy dispersion relation for piorbitals calculated through tight binding calculation in the Brillouin zone gives [6]:

3k x a k y a k a cos + 4 cos 2 y E (k x , k y ) = t 1 + 4 cos 2 2 2

1/ 2

(1.2)

It is important to understand that in such sp2 hybridize system only electrons contribute to electronic transport; hence the above energy dispersion describes the electronic band structure of graphene. Fig. 1.3 shows the energy dispersion for graphene plotted in the Brillouin zone using equation 1.2. The upper half of energy curve describes anti-bonding (*) orbital, while the lower part describes the bonding () orbital. One of the most striking feature of graphene is that and * bands are degenerate at the K points in the Brillouin zone, through which Fermi energy passes. Due to the symmetry requirements of two equivalent carbon atoms in the unit cell of hexagonal lattice, graphene possesses a zero gap at the K point, making it a zero band gap semiconductor.

Figure 1.3: Graphene energy dispersion. The three dimensional graph has kx, ky and E as axes. The location of high symmetry points K, M and is also shown. The Fermi energy passes through K points (E=0). Its interesting to note that near the K point the graphene dispersion becomes linear (massless).

1.3 Carbon Nanotubes

Figure 1.4: Single and multiwall carbon nanotubes. Hypothetically one can visualize the formation of single wall carbon nanotube through rolling single graphene sheet into a cylinder. For multiwall nanotubes bi-layer graphene sheet will be the starting material

A single wall carbon nanotube is technically defined as a cylinder made up of rolled up sheet of graphene as described artistically in Fig. 1.1. However, rolling up graphene is not the actually way a nanotube forms; its actual synthesis process is explained in chapter 2. The diameter of carbon nanotubes typically vary from 0.7-3 nm. Due to such small diameters, nanotubes become quasi one dimensional. They can posses a single shell or multiple shells, as depicted in Fig 1.4. Tubes with single shell are called single wall carbon nanotubes (SWNT) while once with more than one shell are multiwall carbon nanotubes (MWNT). The length of nanotubes can be up to centimeters, giving them an astonishing length/diameter ratio of 107.

7 Carbon nanotubes (CNT) were first discovered in the black soot product from a CVD process[2, 7]. Since then, their synthesis techniques have evolved considerably. The last 10 years have seen tremendous research in both nanotube synthesis[8] and their potential use in electronic circuits[9], composites, thin films[10], etc. Electronic properties of nanotubes have attracted their use as metallic wires and as semiconducting channels in field effect transistors. Processes have been developed to separate semiconducting from metallic nanotube through solution based techniques. However CVD growth of specific chirality nanotube is still not possible. On the industrial front, carbon nanotubes have found use in making composites and gas sensors. This stems from their extraordinary mechanical properties and high surface area.

1.4 Nanotube Crystal Structure

Unlike graphene, nanotubes can exhibit remarkably different physical properties depending upon their structure. The crystal structure of a nanotube depends upon the axis along which the cylinder is formed from the graphene sheet. Fig 1.5 describes the vectors on graphene plane that are important in understanding the formation of nanotube from a graphene sheet. The vector OB perpendicular to the nanotube axis is called the chiral vector (Ch); the vector OA, which is parallel to the axis is termed as the translational vector (T), this is the unit vector of 1-D carbon nanotube. The chiral vector is denoted by

r r r C h = na1 + ma 2 , where 0 m n

(1.3)

8
5a1

a1 a2

T
O

C h = 5 a1+ 5 a2

5a2

Figure 1.5: Nanotube crystal structure. Formation of a (5,5) CNT is depicted on a graphene plane. The shaded gray rectangle represents the unit cell of the (5,5) CNT. The angle between the unit vector a1 and the chiral vector (Ch) is the chiral angle (0 < < 30o)

The (n, m) indices act as unique tags to every CNT since they define its crystal structure. Fig. 1.5 explains the formation of unit cell for a (5, 5) CNT on the graphene sheet. Red and green colored carbon atoms depict the two atom basis of graphene lattice. Chiral vector, Ch is obtained through the vector addition of 5a1 and 5a2. Once the chiral vector is obtained, the translational vector T is drawn perpendicular to it. The vector T ends at a similar atom of the two atom basis. The unit cell of a CNT is the rectangle generated by chiral vector Ch and the translational vector T as shown in Fig. 1.5. In this example it is easy to verify that there are N=10 hexagons in one unit cell of (5,5) CNT. The resulting CNT diameter (dt) is given by

Ch

1.5 Nanotube Electronic Dispersion

For understanding the electronic band structure of nanotube the reciprocal lattice or kspace is considered. The nanotube reciprocal lattice vectors, K1 (circumferential direction) and K2 (along the nanotube axis) are obtained by the relation Ri K j = 2 ij where Ri is the lattice vector in real space. Note that real space lattice vectors for nanotube are Ch and T. The electronic structure of carbon nanotubes can be obtained from the dispersion of graphene through applying periodic boundary conditions along chiral vector Ch. The main difference being that in CNTs electronic wave functions become quantized along Ch as the circumferential length becomes comparable to Fermi wavelength (F). This concept can be used to obtain the one dimensional dispersion for nanotubes through taking out periodic slices from the graphene dispersion relation. This is summarized in the equation 1.4[6]:

E (k ) = E graphene k

r r K2 r + K1 , where = 0,1, 2........N 1, and < k < T T K2

(1.4)

Due to periodic boundary condition along the vector Ch, the wave functions become discreet in K1 direction wave vector (this happens because C h K 1 = 2 ). In equation 1.4., N is the number of hexagons in a single unit cell of the SWNT. Since each unit cell contains 2N carbon atoms, the final band structure of the nanotube possesses N bonding () and N anti bonding (*) atomic orbitals. As K2 depends on n and m, the (n, m) indices determine the direction at which plane slices will be taken off from the graphene

10

dispersion. Using this approach the dispersion and density of states of (5,5) nanotube and (5,4) nanotube is plotted in Fig. 1.6.

10 8 4 E(k)/t (eV) 2 0 -2 -4 -6 -8 6

(5,5)

b
Density of States (eV )
-1

(5,5)

-10 -1.5

-1.0

-0.5

0.0 k (5,4)

0.5

1.0

1.5

-8

-6

-4

-2

c 4
2 E(k)/t (eV)

d
Density of State (eV )
-1

0 2 E (eV)

(5,4)

-2

-4 -0.10

-0.05

0.00 k

0.05

0.10

-8

-6

-4

-2

0 E (eV)

Figure 1.6: Band structure and Density of States (DoS). a) (5,5) armchair nanotube,

shows conduction and valence band crossing at E=0 (Fermi Energy). b) DoS plot on the right shows constant states near Fermi energy. At higher energies, van hove singularities are seen as a signature of the one dimensional band structure. c) (5,4) semiconducting nanotube. d) DoS graphs show zero states in a finite energy window near the Fermi energy. This is the semiconducting band gap.

11

Whenever the crossing plane passes through the high symmetry K point, the nanotube will have a finite density of states at the Fermi energy (E=0) and show metallic behavior. In the event the cutting plane does not passes through the K point, the resultant nanotube possesses a finite band gap and it will behave as a semiconductor, as shown for the case of (5,4) nanotube in Fig. 5c,d. The energy gap of a semiconducting nanotube depends only on its diameter, and is given by E g = t aC C , here t denotes the overlap between dt

the wave functions used in the tight binding and aC C is the carbon-carbon atomic distance in the graphene lattice. Semiconducting and metallic CNTs can directly be identified with the chiral indices (n,m). In general, if n-m=3q where q is a whole number, then the tube is metallic, otherwise semiconducting. One important property of metallic nanotubes is its linear (massless) dispersion near the K point. This is shown in Fig 1.7.
E
Cutting Plane passes through K point

E K2

K1

Graphene bands near K point

Metallic CNT bands near K point

Figure 1.7: Linear dispersion of metallic nanotubes near K-point. Green and red

depicts metallic and semiconducting nanotube dispersion respectively.

12

It has some fascinating consequences. For example, due to massless dispersion, metallic nanotubes becomes insensitive to any long range disorders such as localized charges near the tube and hence their mean free paths become order of magnitude higher than semiconducting nanotubes[11].

1.6 Nanotube Phonon Dispersion

Phonons are quantized lattice vibrations. Similar to electronic dispersion, the energy dispersion for phonons (except for radial breathing phonon mode) can also be derived from graphene through the zone folding technique. This is summarized equation 1.5.

m 1D

= w k
m 2D

r r K2 r + K1 Here, m = 1,......6 denotes the modes of vibration K2

(1.5)

A detailed calculation of phonon dispersion of graphene and nanotube can be found in reference 1. Figure 1.8 shows the phonon dispersion of a (10,10) nanotube very near the point. The three lines that intersect the zero energy are the acoustic modes of the nanotube. Overall these lines correspond to four acoustic modes, since one of them represents a doubly degenerate Transverse Acoustic (TA) mode in the x and y direction. The other two modes are longitudinal acoustic (LA) in z direction and the twisting mode (TW). In addition to acoustic modes there are various optical modes for example E2g mode (17.0 cm-1) and radial breathing mode (165 cm-1) near k=0.

13

A
Figure 1.8: Nanotube phonon dispersion. It is obtained through zone folding the

graphene phonon dispersion. Inset shows the phonon density of states of nanotubes with the van hove singularities. The phonon density of states of graphene is shown with a dotted line for comparison [12].

From the phonon dispersion relation it is possible to obtain the phonon density of states (DoS), shown in the Fig. 1.8 (inset). The nanotube phonon DoS is constant below 2.5 meV and then increases as higher sub bands enter. Just like electronic density of states, phonons also possess one dimensional Van Hove singularities near each sub band edge [13].

14

1.7 Summary
Most of the nanotube properties are derived from graphene, as briefly explained in this chapter. However, contrary to graphene, carbon nanotubes exist in various crystal structures depending upon their direction of roll up in the graphene plane. This results in a variety of nanotube types represented by indices (n,m). Electronically, nanotubes exist as metals as well as semiconductors, depending again upon the chiral indices (n,m). The chapter also explained briefly the phonon band structure of nanotubes. This concept will be helpful in later chapters when the origin of resistivity due to scattering of electrons due to various phonon modes in the nanotube is discussed.

15

Bibliography
1. Guo, T., C.M. Jin, and R.E. Smalley, Doping Bucky - Formation and Properties of Boron-Doped Buckminsterfullerene. Journal of Physical Chemistry, 1991.
95(13): p. 4948-4950.

2.

Iijima, S. and T. Ichihashi, Single-Shell Carbon Nanotubes of 1-nm Diameter. Nature, 1993. 363(6430): p. 603-605.

3.

Novoselov, K.S., et al., Electric field effect in atomically thin carbon films. Science, 2004. 306(5696): p. 666-669.

4.

Novoselov, K.S., et al., Two-dimensional gas of massless Dirac fermions in graphene. Nature, 2005. 438(7065): p. 197-200.

5.

Zhang, Y.B., et al., Experimental observation of the quantum Hall effect and Berry's phase in graphene. Nature, 2005. 438(7065): p. 201-204.

6.

Saito, R., G. Dresselhaus, and M.S. Dresselhaus, Physical Properties of Carbon Nanotubes. 1998: Imperial College Press.

7.

Iijima, S., Helical Microtubules of Graphitic Carbon. Nature, 1991. 354(6348): p. 56-58.

8.

Dai, H.J., Nanotube growth and characterization. Topics in Applied Physics, 2001. 80: p. 29-53.

9.

Ouyang, M., J.L. Huang, and C.M. Lieber, Fundamental electronic properties and applications of single-walled carbon nanotubes. Accounts of Chemical Research, 2002. 35(12): p. 1018-1025.

10.

Gruner, G., Carbon nanotube films for transparent and plastic electronics. Journal of Materials Chemistry, 2006. 16(35): p. 3533-3539.

16

11.

McEuen, P.L., et al., Disorder, pseudospins, and backscattering in carbon nanotubes. Physical Review Letters, 1999. 83(24): p. 5098-5101.

12.

Dresselhaus, M.S.E., P.C., Phonons in Carbon Nanotubes. Advances in Physics, 2000. 49(6): p. 705-814.

13.

Hone, J., et al., Quantized phonon spectrum of single-wall carbon nanotubes. Science, 2000. 289(5485): p. 1730-1733.

17

2 ______________________________
CHAPTER

Carbon Nanotube Synthesis - Processes and Concepts ______________________________

18

2.1. Introduction
The first experimental evidence of Carbon nanotubes (CNTs) came in 1991[1] in the form of multi wall nanotubes (MWNT). This prompted a sudden increase in nanotube synthesis research. In 1993, the first experimental evidence of single wall nanotube tubes (SWNT) came [2, 3]. Since then, the synthesis methods for CNTs have been developed tremendously. This chapter explains the basics of CNT growth as well as describes several unique CNT growth processes developed throughout this study.

Production methods for carbon nanotubes (CNTs) can be broadly divided into two categories: chemical and physical depending upon the process used to extract atomic carbon from the carbon carrying precursor. Chemical methods rely upon the extraction of carbon solely through catalytic decomposition of precursors on the transition metal nanoparticles, whereas physical methods also use high energy sources, such as plasma or laser ablation to extract the atomic carbon. Traditionally, physical methods give bulk quantities of CNTs which could then be treated chemically to remove any carbon soot or nanoparticles present in the mixture[4]. These two approaches can further be characterized according to the use of other important aspects of the synthesis process, such as type of precursor and transition metal nanoparticles used. In spite of being thoroughly investigated for the last 10 years, CNT growth process has remained somewhat controversial. Although the exact dynamics of the growth is not yet clear, consensus has been reached on a hypothesis which works pretty well. It is so far been understood that the nanotubes grow from the over saturation of the transition metal nanoparticles with carbon atoms. A sketch of process is shown in Fig. 2.1.

19

Nanotube growth direction Atomic Carbon from precursor gas

Catalyst Nanoparticle

Figure 2.1: Current CNT growth mechanism hypothesis. Transition metal nanoparticle acting as catalyst gets saturated with atomic carbon to grow nanotubes.

The over-saturation of nanoparticles with carbon atoms results in the production of different type of molecular carbon species, like graphitic carbon, carbon filaments, multiwall carbon nanotubes, single wall carbon nanotubes and most recently, graphene. Selecting the right conditions to grow any of these materials (especially, nanotube and graphene) has remained a trial and error approach. This has resulted in a plethora of CNT growth papers, each having a specifically tuned recipe for growth, which is very difficult to reproduce in any other system or setup. The holy grail of controlling the growth has focused the synthesis effort more towards chemical vapor deposition (CVD) process where the yield of tube growth is not that high and it can finely be controlled through the use of catalyst nanoparticles placed on a silicon chip. The next few sections will detail CVD process and advancements.

20

2.2 Chemical Vapor Deposition (CVD)

CVD process consists of placing catalyst nanoparticles (usually supported on a silicon wafer) in a quartz tube inside furnace which is then heated to the desired growth temperature. Carbon carrier gases are then flown through the quartz tube for a predetermined time. After the growth the furnace is cooled down and the wafer is inspected for the resultant nanotube growth. This technique was first used in 2001 [5] to grow single wall carbon nanotubes on flat silicon substrates. The process gained popularity due to its ability to grow well separated, long single wall carbon nanotubes with less defect density and amorphous carbon content, compare to processes that utilize physical route such as plasma. Due to its wide acceptance in research community, several groups have focused on optimizing the CVD process [5-9]. Before going into the details of this technique it will be helpful to look into various parameters that define this type of growth technique.

2.2.1 Carbon Precursor

CVD growth of CNTs start with the gas phase carbon feedstock. Several different carbon containing compounds[10] have been used as precursors, most common ones are: carbon monoxide (CO)[8, 11] , Methane (CH4)[12], Ethylene (C2H4)[6], Acetylene (C2H2)[13], benzene (C6H6)[14], toluene (C7H8), Ethanol (C2H5OH)[15], and Methanol (CH3OH). Each gas has a particular decomposition temperature, thereby resulting in a different nanotube growth temperature. For example CH4 is known to decompose at very high temperatures (>900oC) over catalyst nanoparticles, whereas C2H5OH starts self decomposing at temperatures near 800oC. Although increased temperatures are required

21 for higher rate of nanotube growth, very high temperatures can initiate self dissociation of gases which can cause catalyst poisoning. Apart from the temperature, the dissociation rates of the precursor can also be controlled by the partial pressure of the species. For example, at high pressure CO dissociation rate increases, hence to achieve higher nanotube yield high pressure CO growth processes are used. A well known example is the HiPCO process from Smalleys group[16] at Rice University. Similarly, low pressure growth has also been used to decrease catalyst poisoning to achieve ultra long CNTs [6]. Another important parameter associated with the precursor is its feed rate. At very high temperatures where the precursor is near self-dissociation, the reaction rate gets limited by the precursor feed rate in the system. High feed rates can increase the rate of growth but just like high temperatures can also result in more of carbon soot formation and hence catalyst poisoning. Precursor feed rate is also coupled with the size of catalytic nanoparticles. For example in a study by Cheung et al[17], it is shown that different carbon species can be produced with the same diameter catalyst nanoparticles using different carbon carrier gas flow rates. At higher flow rates, larger diameter nanoparticles will grow more since the smaller ones will quickly get poisoned and vice versa for lower flow rates. In a typical growth scenario, carbon precursors are premixed with other gases such as H2 or other carbon carriers (e.g. C2H5OH+CH3OH)[18]. This is done to have finer control on the reaction rates inside the CVD chamber. For example, if the product after dissociation is H2, then a fine control of premixed hydrogen can be used to check on the dissociation rate of carbon precursor. This is discussed in detail in the succeeding experimental section. Carbon precursor gases can themselves produce unwanted nanotubes due to the

22 presence of premixed nanoparticles coming from the contamination in the gas cylinder[19]. Gradual changes in growth patterns with change in cylinder pressure have been reported before. It is explained in chapter 3 that how this shortcoming can be used for achieving ultra dense nanotube growth on carbon fibers. This suggests that for a wellcontrolled CVD system the use of cleaner gases or gas filters is essential in achieving reproducible CNT growth results.

2.2.2 Temperature
The ideal temperature for CNT growth depends on several factors, mainly carbon feedstock, catalyst, and the type of CNTs desired (single or multi wall). Nanotubes are typically grown in a temperature range from 550oC to 1000oC [10]. In addition to changing the reaction kinetics during growth, the temperature plays an important role in the pre-growth treatment of the catalyst. Small catalytic nanoparticles readily oxidize readily under ambient conditions; therefore, to bring them back to native state a controlled reduction step is required at moderate temperatures (~700oC)[20]. Temperature can also significantly affect the growth depending upon ramping rates. Ramping up the furnace temperature at different rates is shown to have profound effect on the CNT growth kinetics. For example, Lius research group at Duke University has shown that fast heating[8] of the growth tube produces long and aligned CNTs. This was attributed to the effect of temperature on the flow conditions near the surface which cause catalyst particles to leave the substrate and grow long tubes. The role of temperature is becoming more important as nanotube world looks towards growing graphene through CVD process [21-23].

23

2.2.3 Catalyst Nanoparticle

Catalysts are transitional metal nanoparticles obtained through various sources, such as metal salts, evaporated metal films, etc. Two factors that define a catalyst particle are its size and composition. Nanoparticles of Fe, Co, Mo, Ni, Cu, Au, etc. [10] have been tried as catalyst either in pure metallic form or as alloys. These transition metal nanoparticles have common advantages of high carbon solubility, carbon diffusion rates and high melting temperatures. Various approaches are used to obtain the nanoparticles; some of the most common ones use metal salts (nitrates, sulfates and chlorides), where stable nanoparticles complexes are formed in a suitable solvent. Evaporated film of metal can also produce uniform nanoparticles upon controlled annealing treatments. Organic carriers have also been used in the formation of very small size nano clusters; one of such catalyst is used for the current research[24]. Sizes of these particles are reported to be anywhere from 1 nm to 15 nm. There are strong indications of dependence of CNT diameter on the catalyst particles size; hence narrowing the catalyst diameter can help in a controlled diameter nanotube growth. Also of tremendous importance is the method used to apply the nanoparticles on the growth substrate, which affects their resultant morphology significantly. Most nanoparticles are suspended in solvents, which after drying out on the growth substrate give a coffee stain effect; this causes the nanoparticles to clump before growth, making their average size high. Several innovative methods are in use to get around this problem, for example, mixing the nanoparticles with polymers before application on substrates has significantly helped in achieving well-separated particles of uniform size[25]. Another promising approach is to mix the particles with

24 PMMA or photo resist and pattern them using lithography techniques. Every approach has its pros and cons and work best for a particular type of CNT growth.

2.2.4 Gas Flow Rates

Flow rate of carbon precursor affects the rate of carbon dissociation during the growth. In addition to affecting reaction rates, gas flow conditions are also responsible in defining CNT length and orientation during growth. The effect of flow conditions such as turbulence, stable boundary layer, etc. on the tube growth was first demonstrated by Lius[8] research group for CO growth. Longer tubes are regularly seen to grow more from wafer edges than the middle of the wafer. They are also found to be aligned to the gas flow direction. The observation of wavy tubes is attributed to the unsteady flow conditions near the growth wafer (excluding the quartz crystal aligned growth, where the mechanism is faulty crystal steps). A recent publication on suspended growth of nanotubes found local thermal oscillations to play a big role in obtaining aligned growth[26]. In order to stabilize the near surface flow several techniques have been tried, such as bringing down the quartz tube cross section and decreasing the flow rates to very low values. The large variability in nanotube lengths could be attributed to the type of growth steps they go through. It is now generally accepted that nanotube growth processes are of two kinds: 1. tip based, 2.root-based. In the former process the catalyst nanoparticle remains in the end of the CNT, flowing along with the flow while growing CNT in the back. In the latter process the particle stays on the surface while the tube grows up due to stress induced by carbon atoms saturating the catalyst from below (Fig.

25 1.1). Tip-based process achieves very long length tubes since metal particle gets saturated in air while flowing. More of fluid flow studies will be discussed in the coming sections.

2.3 CNT Growth Experiments

2.3.1. Motivation and Aims

Since the start, CNT growth processes have been riddled with problems relating to control over type of CNTs grown and process reproducibility. One major issue with understanding the growth results is a lack of good characterization techniques. Imaging techniques such as Atomic Force Microscopy (AFM) and Scanning Electron Microscopy (SEM) have a high throughput and are used frequently for characterizing grown nanotubes. However, they have big error margins in measuring nanotube diameters and cleanliness. It is also almost impossible to distinguish a double wall nanotube, or a small nanotube bundle from a single tube, using these techniques. Current project aims to characterize grown CNTs much more accurately using new techniques such as Rayleigh spectroscopy on suspended nanotubes (described in section 4.2.2). The high throughput and accuracy of this technique promises to give reliable information about type of nanotubes grown through every growth process. Rayleigh spectroscopy was used through out this project to check if the tubes are single or bundled, whether they have amorphous carbon deposition or not, and if they are single then their diameter and chirality. So much information about every growth done was not available before. Therefore the use of this technique with CNT growth gave the impetus for a detailed CNT synthesis study. Fig.2.2 shows a schematic the substrate used for growing nanotubes in this study.

26

Apart from studying the synthesis process, the use of chirality assigned single tubes in several other projects called for highly tuned recipes to achieve specific goals. These goals can be summarized in following points: Gas Flow Direction 1. Ultra long nanotubes for electron transport measurements such as resistivity and
Carbon Nanotube

mean free path determination. . 2. It was found through Rayleigh spectroscopy that more than 99% of long tubes are

bundles of 2-5 tubes. Finding ways to grow single tubes through changing growth parameters was very important. 3. Clean, amorphous carbon free nanotubes for good electrical measurements.

Figure 2.2: A schematic of the nanotube growth wafer. Suspended nanotubes are grown over a slit, etched through a silicon wafer for Rayleigh characterization.

2.3.2 Growth Setup

Fig. 2.3 shows the schematic of the CVD growth system. This consists of a one inch quartz tube inside an oven with a temperature controller capable of reaching 1200 oC.

vent

feed gases

Flow Controllers

Figure 2.3: CVD growth system. Inside the quartz tube, a quartz sample holder is used to carry the growth substrate. Flow of gases is controlled through electronic mass flow controllers.

27 A variety of parameters were studied for the CVD growth process. This included several carbon precursors, catalyst nanoparticles, various temperature ranges and flow conditions. The effect of each parameter change on the resultant CNT growth is observed and recorded. The next few sections will describe the variety of growth processes classified on the basis of carbon feedstock.

2.3.3 Carbon Monoxide Growth

Carbon monoxide is a highly efficient feedstock for CNT growth, both for bulk synthesis[27] and for direct CVD growth on substrates. For direct growth onto substrates, CO is mixed with H2 to optimize nanotube growth and minimize the deposition of amorphous carbon. A significant advantage of CO as a feedstock is that the process window for CNT growth is very wide: good growth is generally achieved for hydrogen concentrations of ~20% to ~70% at 900 C[28]. Another advancement in CO CVD has been the efficient production of long SWNTs by a fast heating technique[8]. In this method, a substrate with catalyst is inserted directly into the hot zone of a furnace under flow of feedstock gas. Under these conditions, a significant proportion of CNTs evidently grow by tip growth mechanism described previously. This allows for very fast nanotube growth and alignment of the tubes with the gas flow in the CVD reactor. Iron nanoparticle catalysts (diameter ~ 2 nm) are deposited on Si/SiO2 substrates, either randomly or in patterned areas. The substrates are then inserted into a 1 diameter quartz tube that fits into a tube furnace (18 hot zone). The quartz tube is long enough (48) so that the sample can be moved in and out of the hot zone by simply sliding the tube. The substrates are first placed in the hot zone of the furnace and kept in pure oxygen at 650

28 C for 30 minutes to burn off the organic capping layers on the nanoparticles and any residual photo resist on the substrate. The substrate is then removed from the hot zone. The oxygen flow is turned off and replaced (after Argon purge) with CO/H2 at flow rates of 1000 SCCM for both gases. The oven is heated to 900 C, allowed to stabilize, and the sample is quickly inserted into the hot zone. The substrate is maintained at the growth temperature for 20-30 minutes, followed by slow cooling in pure H2 to room temperature. The effect of hydrogen on the nanotube growth can be attributed to two factors. First, the direct hydrogenation of CO produces carbon as well as water (H2O), which helps in cleaning the catalyst particles off any amorphous carbon deposition. The second reported effect of H2 is that its adsorption on the catalyst nanoparticles increases the catalytic dipropotionation of CO [16] which increases the nanotube growth rates. In addition to this hydrogen also provided a reducing atmosphere during the processs. Figure 2.4a shows arrays of long SWNTs that can be achieved by fast-heating CO CVD.

Figure 2.4: Carbon monoxide grown nanotubes. a) Scanning Electron Microscope (SEM) images tubes grown through fast heating method. b) Tubes grown through patterning local catalyst islands.

29 It is also possible to grow long nanotubes from individually patterned catalyst pads, as shown in Fig. 2.4b. The catalyst pads were defined by photolithography employing a two-layer resist consisting of a layer of SU-8 photo resist on top of PMMA. Because PMMA is soluble in SU-8 developer, patterning the top SU-8 layer also selectively removes the underlying PMMA. After deposition of catalyst, the bilayer is lifted off in acetone. The PMMA acts as a release layer, while the SU-8 prevents stray redeposition of the catalyst onto the substrate. Although carbon monoxide showed a very consistent and dense growth it suffered from some major drawbacks: 1. The self decomposition of CO resulted in considerable amorphous carbon deposition in the quartz tube. nanotubes. 2. Due to the hazardous nature of the gas, the CVD setup has to be enclosed and continuously vented. 3. The presence of iron-carbonyl as impurity inside the gas frequently gave spontaneous gas phase growth of nanotubes. In order to reduce this, the gas had to be passed through a preheating oven to deactivate the impurity metal contamination through carbon crackdown, making the process more complex. 4. The metal impurities in the gas were destroying the mass flow controllers, since these controllers used an active micro heater inside them to calibrate flow rates. This threatened the cleanliness of the grown

These disadvantages shifted the focus towards ethanol based process, which was already proven to be clean and non-hazardous [29].

30

2.3.4 Ethanol Growth

Ethanol process is a low temperature method for growing single wall CNTs[30]. The method uses Ar as carrier gas to introduce ethanol in the growth chamber as a carbon feedstock. The amount of ethanol taken away by per unit volume of gas is controlled by keeping the bubbler in an ice bath which controls the ethanol vapor pressure. Traditionally, ethanol growth process was used with Co catalyst either in solution phase (Co salt) or in evaporated thin film form[15]. Both the approaches worked well for growing ultra long nanotube bundles; though single tubes were rare. This may be due to the clumping of catalyst nanoparticles in solution phase or during the growth process. In order to solve this problem a new type of catalyst was prepared from a published nanoparticle cluster synthesis process[31]. The catalyst nanoparticle consisted of Fe and Mo atoms arranged in a cage like structure. The exact structure is shown in Fig. 2.5.

Figure 2.5: Structure of Fe-Mo catalyst nanocluster (Mueller Catalyst), green spheres on the outside wire frame are 30 FeIII atoms, encapsulated inside are 12 {(Mo)(Mo)5} pentagons. Image taken from ref [31].

31 The circumferential atoms are bonded with acetate ligands making the cluster readily soluble in hydrophilic solvents like water, ethanol. The goal was to apply this catalyst on silicon oxide substrate without letting it clump once solvent is evaporated. This was achieved through mixing it with a surfactant polymer (Pluronic F-127) before its application on growth wafers.

The high viscosity of surfactant polymer prevented the catalyst from clumping after the solvent is evaporated. The polymer is then burned in air to get mono dispersed cluster of nano-particles. An important step in this process is diluting the catalyst solution sufficiently so that resultant nanoparticle clusters are too well separated to clump even at high growth temperatures. In addition to addressing the catalyst problem various small changes to the CVD system helped in achieving the desired growth.

1. Replacement of the Teflon tubing with the No-Ox tubing helped in diffusion of oxygen inside the growth system. Oxygen diffusion through sintered polymer tubes such as Teflon was found to be a major source of irreproducibility in the growth results. 2. Use of leak proof quartz tube fittings helped in checking for any potential leakage at the quartz tube-gas inlet joint. 3. With the use of industry standard mass flow controllers the growth system was made completely computer controlled. This resulted in the execution of precise values of flow rate and temperature that eventually made recipes repeatable. An image of the computer interface of the CVD growth system is shown in Fig. 2.6.

32

Figure 2.6: Carbon nanotube CVD growth computer interface. The program controls the flow rates of gases, oven temperature as well as process timings.

Encouraging results were obtained using the new catalyst nanoparticles with ethanol CVD process. Furthermore, optimized flow and temperature conditions were achieved by testing CNT growths at different conditions. One major issue with ethanol grown CNTs was their cleanliness. The CNTs easily get coated with amorphous carbon if the growth conditions are not ideal. It was found that above 860oC, self-dissociation of ethanol caused significant amorphous carbon deposition over the CNTs. This effect could be observed through Rayleigh scattering as shown in Fig. 2.7. Typically, single nanotubes with higher amorphous carbon deposition show a higher light scattering intensity in the

33 background than the cleaner tube. Using this feedback from the Rayleigh technique it was possible to get a finely tuned recipe that grows cleaner CNTs.

Intensity (a.u.)

2.0

2.2 2.4 E(eV)

2.6

Intensity (a.u.) 1.6

1.8

2.0

2.2 E(eV)

2.4

2.6

Figure 2.7: Rayleigh scattering spectrum on two different suspended (16, 16) armchair carbon nanotubes, spectrum (a) has less signal to background ratio than spectrum (b). This suggests that the (16,16) nanotube in (b) is cleaner than in (a).

The growth of clean, single walled CNTs has made possible to gather sufficient number of suspended tube samples on TEM compatible chips for preparing the database which that relates the tube chirality with Rayleigh scattering peaks. This database is updated continuously and is currently used to directly assign the tube indices by obtaining the Rayleigh scattering peak position. Chirality and diameter values assigned using such a database for nanotubes grown through ethanol CVD process are shown in Fig. 2.8. Nanotube diameter distribution ranges from 1.2 to 2.8 nm with a clear peak at ~ 2nm. This is two to three times higher than what is expected from the size of the catalyst nanoclusters, this indicates toward possible clumping of two to three catalyst nanoparticles during growth. The chirality distribution is more towards higher chiral

34 angles, something which is still not understood, although there have been theoretical predictions of near-armchair tubes being more stable in structure than tubes of other chiralities.

20 16 15 No. of tubes 10 5 0 12 16 20 24 Diameter (A) 28 No. of tubes 12 8 4 0 0 5 10 15 20 25 30 Chiral Angle (degree)

Figure 2.8: Diameter and chirality distribution of single wall carbon nanotubes grown through ethanol CVD process using Mueller catalyst.

The ethanol growth process could also be tuned to the length requirement of CNTs. Ultra long nanotubes (~ 10 cm) were produced through increasing the growth temperature (~885 oC) and using higher gas flow rates.

2.3.5 Methane Growth

The very first CVD growth of nanotube achieved on silicon wafers used CH4 as the feedstock[12]. The interest in CH4 based growth increased with the reporting of ultra long and aligned nanotube growth through ultra low flow of methane[32]. This approach was counterintuitive to the high flow rate growth and the fast heating growth mechanism suggested previously. Since CH4 doesnt dissociate that easily, it requires high

35 temperature during growth. It is hypothesized that due to the use of high temperature, CH4 process grows smaller diameter nanotubes than the ethanol based growth. A separate growth system with ultra low manual flow controllers was designed to make methane growth possible. Mueller catalyst was used for this process. The growth temperature was tested in the range of 940-980 oC. Methane flow rate was at ~2 sccm while hydrogen at ~4 sccm. Further details of the process are provided in reference[32]. Fig.2.9 shows the resultant nanotube growth on silicon wafers.
a b

Intensity (a.u.)

2.2

2.4 Energy (eV)

2.6

Figure 2.9: Methane growth results. a) SEM image of ultra long nanotubes (> 2mm) are grown over silicon wafers, the tubes are crossing a slit on the wafer as shown in the right SEM image. b) Rayleigh spectrum of one of the single, suspended tube section grown through methane process.

The nanotubes were very straight and ultra long. These tubes were grown crossing the slits on the silicon wafers for Rayleigh characterization. Through Rayleigh scattering technique, it is found that nearly all of these long tubes were big bundles of single wall nanotubes. Also they did not seem to be very clean. The Rayleigh spectrum of one of the

36 single tubes obtained from this growth process is shown in Fig. 2.8b. Due to the virtual absence of any single tube, the process was not continued.

2.3.6 Water-Assisted Ethylene Growth

Ethylene growth combined with water vapors has been proven to grow ultra tall vertical nanotube forests, which can be used for applications such as nanotube yarns and composites. One of the salient features of this process was the use of water during growth, which was carried by argon gas into the growth chamber. It is hypothesized that water helped in cleaning the catalyst hence increasing its lifetime[6]. Thereby tubes can be grown for a longer time. The growth process to achieve these tall nanotube forests is as follows:

1. Substrate preparation: 10 nm of aluminum oxide is deposited through Atomic Layer Deposition (ALD) system on a Si/SiO2 wafer. Thereafter, 1 nm thin film of iron as a catalyst is evaporated using Electron Beam Evaporation (EBE). 2. Prepared substrate is placed in the quartz growth tube on a quartz sample holder and the oven is heated to a temperature of 750oC in a gas flow of Ar (600 sccm) and hydrogen (400 sccm). 3. At 750oC, ethylene (150 sccm) mixed with argon (40 sccm) coming separately through the water bubbler is introduced in the growth tube. The growth is done for 15-25 minutes. Once the growth is over, the system is cooled down to room temperature in Argon (600 sccm) and hydrogen (400 sccm).

37

Figure 2.10: Water-assisted ethylene growth results. Left SEM image shows the ultra tall vertical CNTs forest. A high magnification SEM image of these tubes is shown in the right.

Millimeter tall and straight nanotube forest can be seen in Fig.2.10. A closer look through scanning electron microscope suggests that the CNTs are not as aligned as they looked at lower magnifications. The forest contains a directional growth CNTs which meshed up with each other. Such growth has been utilized for making nanotube composites and probing the photo mechanical response of CNTs. These CNTs are also finding research applications in fabricating flow channels for understanding transport of water inside carbon nanotubes.

2.3.7 Nanotube Growth: Fluid Flow Study

Gas flow conditions play an enormous role in deciding the length and alignment of CNTs during CVD growth. Ultra high flow rates, fast heating approach[8], as well as ultra low

38 flow rates[32] of gases have been tried to achieve directionally uniform, ultra long growth of CNTs. Although there are several dissimilarities among these approaches, they all agree on the fact that CNTs grow in the direction of the gas flow. A major concern was that these ultra long nanotubes were rarely single. Through Rayleigh spectroscopy it is found that more that 99% of long nanotube were bundle of single wall tubes. One approach to reduced bundling is to separate the catalyst particles far away so that the CNT grown from them dont bundle up due to close proximity. This was achieved by the use of Mueller catalyst and polymer combination as explained in section 2.2.4. However, it was found that the tubes still get bundled because of flow conditions near the substrate, influenced heavily by surface geometry.

Figure 2.11: Flow study surrounding a raised structure on silicon chip. a) SEM image of the nanotube growth obtained through a raised pillar. Crossing of tubes suggests that nanotubes follow the flow lines while growing. b) Simulated flow lines around the structure.

39 Fig. 2.11a shows the SEM image of a pillar, etched on the surface of the silicon wafer. Catalyst is stamped on the surface of the pillar and CNTs are grown from it. As seen clearly in the image, tubes do tend to bundle up following the flow lines around the raised structure. The flow lines around the pillars are shown in Fig. 2.11b. The fluid flow simulations are done using COMSOL FemLabTM finite element analysis software.

Figure 2.12: Flow movement in side plane of the slits and raised structure. a) SEM image of the tubes following a parabolic trajectory while crossing the slit. b) Color shows the average flow velocity with red being the higher. The flow stream lines are represented by red lines. The white line represents a particles trajectory with the mass of a catalyst nanoparticle. The particles origin was set very close to the surface. This suggests that

40 with the right flow lines, particles from the surface can lift up to give ultra-long nanotubes. Fig. 2.12a shows the SEM image of grown tubes in the side plane. The trajectory that tubes follow when grown over the slits from a catalyst particles right underneath the pillar is highlighted through the arrows. This matches well with flow lines simulated in the Fig. 2.12b. The silicon chip in the model of Fig. 2.12b is tilted at an angle to show the leading edge effects, which can make a catalyst particle pulled up from a leading edge and follow the flow streamline while simultaneously growing the CNT from back (tipbased growth). This also explains why most of the long tubes grow from the wafer edges. It is also evident that CNTs follow the flow streamlines during growth. Since the flow velocity near to the surface is very low, there is no fluid turbulence. Therefore, the tube directionality is determined mainly by local flow perturbations created by the raised structures near the surface. With the knowledge of flow simulation results, raised platforms are designed across the slits to help grow unbundled tube. These platforms served multiple functions: 1. With platform so close to the slit, its edge can be used to grow long, flow aligned tubes. This is opposed to using the chip edge which is far away from the slit. This decreases the chances of tube bundling on way to the slit. 2. Less catalyst is needed, because most of the tubes that grow from platform edge cross the slits. The high growth efficiency through edge allows the use ultra low catalyst density (~ 10000 nanoparticles/ mL) hence directly reducing bundling. 3. Once characterized through Rayleigh, it was much easier to transfer suspended tubes onto the target chip with raised platforms across the slit.

41 Developing an understanding of the nanotube growth dynamics and its relation with local flow conditions helped in achieving unbundled, yet straight and clean single wall carbon nanotubes. More work towards studying local flow conditions along with temperature gradients occurring near the growth substrate can deliver better results.

2.3.8 Conclusion
This study contributes significantly to the understanding of nanotube growth through CVD process. A wide growth parameter range is covered, as shown in Fig. 2.13 below.

Growth Parameters Studied

Carbon Feedstock CO CH4 C2H4 Ethanol

Catalyst Fe, Co, Mo Salts Fe Nanoparticles FeCl3 Fe-Mo Complex

Temperature Low (<750oC) Medium (750oC<T<850oC) High (>850oC)

Fe, Co evaporated films Fe Carbonyl

Figure 2.13: Flow chart describing CVD growth parameters studied

The effect of local flow conditions in the resultant CNT growth is also studied through experiments as well as theoretical models. Overall, the research furthers the understanding of various ways through which multiple CVD growth parameters can be chosen and controlled to attain the type of growth desired.

42

Bibliography
1. Iijima, S., Helical Microtubules of Graphitic Carbon. Nature, 1991. 354(6348): p. 56-58. 2. Iijima, S. and T. Ichihashi, Single-Shell Carbon Nanotubes of 1-nm Diameter. Nature, 1993. 363(6430): p. 603-605. 3. Bethune, D.S., et al., Cobalt-Catalyzed Growth of Carbon Nanotubes with SingleAtomic-Layer walls. Nature, 1993. 363(6430): p. 605-607. 4. Thess, A., et al., Crystalline ropes of metallic carbon nanotubes. Science, 1996. 273(5274): p. 483-487. 5. Dai, H.J., Nanotube growth and characterization. Topics in Applied Physics, 2001. 80: p. 29-53. 6. Hata, K., et al., Water-Assisted Highly Efficient Synthesis of Impurity-Free SingleWalled Carbon Nanotubes. Science, 2004. 306(5700): p. 1362-1364. 7. James, H., et al. Growth of nanotubes and chemical sensor applications. 2004: SPIE. 8. Liu, J., Growth of ultralong and aligned single walled carbon nanotubes using a "fast heating" chemical vapor deposition method. Abstracts of Papers of the American Chemical Society, 2004. 227: p. U273-U273. 9. Lu, C.G. and J. Liu, Controlling the diameter of carbon nanotubes in chemical vapor deposition method by carbon feeding. Journal of Physical Chemistry B, 2006. 110(41): p. 20254-20257.

43 10. Moisala, A., A.G. Nasibulin, and E.I. Kauppinen, The role of metal nanoparticles in the catalytic production of single-walled carbon nanotubes - a review. Journal of Physics-Condensed Matter, 2003. 15(42): p. S3011-S3035. 11. Zheng, B., et al., Efficient CVD Growth of Single-Walled Carbon Nanotubes on Surfaces Using Carbon Monoxide Precursor. Nano Lett., 2002. 2(8): p. 895-898. 12. Cassell, A.M., et al., Large scale CVD synthesis of single-walled carbon nanotubes. Journal of Physical Chemistry B, 1999. 103(31): p. 6484-6492. 13. Patole, S.P., et al., Water-assisted synthesis of carbon nanotubes: Acetylene partial pressure and height control. Epl, 2008. 81(3): p. 6. 14. Yang, Q.H., et al., Direct growth of macroscopic fibers composed of large diameter SWNTs by CVD. Chemical Physics Letters, 2003. 370(1-2): p. 274-279. 15. Huang, L.M., et al., Cobalt ultrathin film catalyzed ethanol chemical vapor deposition of single-walled carbon nanotubes. Journal of Physical Chemistry B, 2006. 110(23): p. 11103-11109. 16. Hafner, J.H., et al., Catalytic growth of single-wall carbon nanotubes from metal particles. Chemical Physics Letters, 1998. 296(1-2): p. 195-202. 17. Cheung, C.L., et al., Diameter-controlled synthesis of carbon nanotubes. Journal of Physical Chemistry B, 2002. 106(10): p. 2429-2433. 18. Cassell, A.M., et al., Combinatorial optimization of heterogeneous catalysts used in the growth of carbon nanotubes. Langmuir, 2001. 17(2): p. 260-264. 19. Chandra, B., Joshua T. Case, Yuhao Sun, S.C. Barton, James Hone, Growth of Carbon Nanotubes on Carbon Toray Paper for Bio-Fuel Cell Applications.

44 Proceedings of the 2nd Energy Nanotechnology International Conference, 2007. ENIC 2007, 45038. 20. O'Brien, S.P., L.M. Huang, and S. Wind. Controlled growth of single walled carbon nanotube assemblies from thin film mesoporous silica templates. 2003: Amer Chemical Soc. 21. Kim, K.S., et al., Large-scale pattern growth of graphene films for stretchable transparent electrodes. Nature, 2009. 457(7230): p. 706-710. 22. Campos-Delgado, J., et al., Bulk production of a new form of sp(2) carbon: Crystalline graphene nanoribbons. Nano Letters, 2008. 8(9): p. 2773-2778. 23. Reina, A., et al., Large Area, Few-Layer Graphene Films on Arbitrary Substrates by Chemical Vapor Deposition. Nano Letters, 2009. 9(1): p. 30-35. 24. Fan, Y.W., B.R. Goldsmith, and P.G. Collins, Identifying and counting point defects in carbon nanotubes. Nature Materials, 2005. 4(12): p. 906-911. 25. Huang, L.M. Oriented growth of single-walled carbon nanotube arrays. 2003: Amer Chemical Soc. 26. Marcus, M.S., et al., Predicting the Results of Chemical Vapor Deposition Growth of Suspended Carbon Nanotubes. Nano Letters. 0(0). 27. Nikolaev, P. Gas-phase production of single-walled carbon nanotubes from carbon monoxide: A review of the HiPco process. 2004: Amer Scientific Publishers. 28. Zheng, B., et al., Efficient CVD growth of single-walled carbon nanotubes on surfaces using carbon monoxide precursor. Nano Letters, 2002. 2(8): p. 895-898.

45 29. Huang, L.M., et al., Long and oriented single-walled carbon nanotubes grown by ethanol chemical vapor deposition. J. Phys. Chem. B, 2004. 108: p. 16451. 30. Maruyama, S., et al., Low-temperature synthesis of high-purity single-walled carbon nanotubes from alcohol. Chemical Physics Letters, 2002. 360(3-4): p. 229-234. 31. Muller, A., et al., A new type of supramolecular compound: Molybdenum-oxidebased composites consisting of magnetic nanocapsules with encapsulated kegginion electron reservoirs cross-linked to a two-dimensional network. Angewandte Chemie-International Edition, 2000. 39(19): p. 3414-+. 32. Jin, Z., et al., Ultralow feeding gas flow guiding growth of large-scale horizontally aligned single-walled carbon nanotube arrays. Nano Letters, 2007. 7(7): p. 2073-2079.

46

3 ______________________________
CHAPTER

Carbon Nanotube Synthesis - Technology Applications

______________________________

47

3.1 Growth of CNTs on Carbon Toray Paper for BioFuel Cell Applications

3.1.1 Introduction
Since their discovery in 1991[1], CNTs have been extensively researched for their potential applications in semiconductor devices, solar cells, composites, displays, etc[2]. Their unique quasi-one dimensional structure gives them extraordinary electronic (high current carrying capability) as well as mechanical properties[3]. In this work, we examine CNTs as high surface area support for one such application: a bio-fuel cell. Bio-fuel cells utilize enzymes as catalysts; however, because of the relatively low enzymatic activity, large amounts of them are required. In a previous work, research has employed paper made up of carbon fibers mixed in a binder material as a high surface area catalyst support[4] to enhance the activity. The aim here is to utilize the extraordinary high surface area per unit volume of CNTs to further enhance the fuel cell reaction rates Presently, carbon nanotubes are grown by various chemical vapor deposition (CVD) techniques, viz. arc discharge process, plasma enhanced chemical vapor deposition (PECVD), high pressure carbon monoxide process (HiPCO), etc. All these processes use some kind of catalyst and carbon carrier gas combination for the nanotube growth [5]. The main disadvantage with all the above listed processes is inefficient use of resources (gases, high power consumption) and long time of operation (typically 4-5 hours per growth). The next section describes a newly developed approach to grow CNTs on carbon TorayTM paper through utilizing joule heating and gas phase catalyst.

48

3.1.2 Experimental Setup and Process Parameters

CVD growth of CNTs on carbon fibers process involves the local heating of growth substrate, which is in contrast to the regular CVD processes in which the whole growth chamber is heated [6, 7]. The method is fast, and excellent for the application involving growing multiwall carbon nanotubes (MWNT) through a porous, conducting carbon paper. A schematic of such a growth setup is shown in Fig 3.1.

Pressure Gauge

Glass Flask

Electrode

Carbon paper

To Pump/Exhaust

Gas Inlet Power supply electrode

Figure 3.1: Schematic of the nanotube growth system. The carbon paper (Toray inc., 90m thickness, 40mmx20mm) is clamped between two copper electrodes over a 20mm x 13 mm rectangular slot made in a ceramic plate. Carbon carrier gases containing a gas phase catalyst (Fe (CO)5) is passed through the porous paper via a passage below it. Carbon paper is sandwiched between two identical ceramic plates placed over the gas inlet. This arrangement makes sure that the gases pass

49 through the paper, to help achieve CNT growth throughout the paper thickness. The growth assembly is enclosed in a leak proof glass enclosure as shown in Fig 3.1. The chamber is first evacuated to the base pressure of 50 m-torr. It is then purged with argon gas at a steady flow rate of 2500 sccm, until the pressure reaches close to atmosphere. Finally, the chamber is pumped back to the base pressure and the throttle valve connected to the pump is closed. This process makes the chamber environment free of any traces of oxygen. Thereafter, a mixture of approximately 75% CO, 25% H2 (both Ultra high purity grade) is fed into the chamber at a flow rate of 300 sccm and 100 sccm respectively. It takes 8-10 minutes for the pressure in the chamber to reach 350 torr; here, the throttle valve for pump is opened slowly to make the pressure constant inside the chamber. Current is then passed through the carbon paper by an external DC power supply (Agilent E3633A) till the carbon paper reaches the growth temperature. Most of the heating takes place in the paper only since it is the most resistive part in the circuit (sheet resistance 1.4 / ). The selection of CO as the carbon feedstock is due to its longer CNT growth temperature window than other gases.

The effect of hydrogen on the nanotube growth can be attributed to two factors: first, the direct hydrogenation of CO produces carbon as well as water (H2O). This increases the reaction of carbon production as the byproduct H2O helps in cleaning the catalyst particles off any amorphous carbon deposition[8]. The second reported effect of H2 is that its adsorption on the catalyst nanoparticles increases the catalytic dipropotionation of CO through Boudouard reaction [9, 10] which increases the nanotube growth rates.The

50 catalyst used in these experiments is gas-based iron penta-carbonyl (Fe(CO)5), trace amounts of which is present as impurities in the carbon monoxide.

3.1.3 Growth Results

Growth of CNTs is characterized by Scanning electron microscopy. The overall surface area of nanotube grown substrate is measured with electrochemical capacitive measurements as well as with nitrogen physiosorption measurements (BET).

Figure 3.2: Scanning electron images of CNTs grown on carbon paper Fig. 3.2 shows a high magnification SEM image of the CNTs grown on an individual carbon fiber. The average length of the CNTs is less than 500 nm. Scanning electron microscope images of the CNTs grown over the carbon fibers for different time intervals is shown in Fig. 3.3. Density of nanotubes increases continuously with growth time; this is mainly due to the use of fresh gas phase catalyst coming in continuously with the incoming gas. This would not have been possible with the substrate-based catalyst due to steady deactivation of catalyst nano-particles with growth time. After 24 minutes of

51 growth nanotubes significantly start filling up the voids between these carbon fibers as seen in Fig. 3.3.

15 sec

30 sec

24 min 45 sec

30 min 28 min

Figure 3.3: MWNT growth at different time intervals. Very thin layer of nanotube is seen at just the beginning of growth (t=15 sec). However as the growth time increases the tube density increases too. After 30 minutes the tube density becomes high enough to start filling the void space between the carbon fibers.

Fig. 3.4 shows capacitive and BET surface area measurements of nanotubes grown over Toray paper. Capacitive measurements give lesser surface area estimates than nitrogen

52 physiosorption (BET) measurements because of inaccessibility of large electrolyte ions to pores smaller than their size. The rate of surface area increase is highest for the first few minutes of the growth. After 30 minutes of growth the overall surface area is increased by a factor of 180 as compared to bare Toray paper. Increasing the growth time more than 30 minutes does not result in any appreciable increase in surface area due to increase in clumping of CNTs, which decreases the effective area available for nitrogen physiosorption in BET measurements. This also decreases the pore size for ions to go through in an electrochemical capacitive surface area measurement setup.

Figure 3.4: Effect of carbon nanotube growth time on surface area

53

3.1.4 Summary
Use of gas phase catalyst with local resistive heating of the substrate grows ultra-dense CNTs on a carbon fiber substrate. The CNT mats increase the surface area of the carbon fiber paper by more than two orders of magnitude. This promises an improvement in the catalytic efficiency of bio-fuel cells where carbon substrates are used as enzyme support. Study of the nanotube growth time on increasing surface area suggests an ideal time after which the surface area saturates. Study of various other growth conditions like temperature, flow rates, catalyst concentration can help in growing longer tubes which can fill the voids between the carbon fibers with less clumping; this has the potential of further improving the surface area of the carbon paper.

54

3.2 Growth of Transparent, Conducting CNT Films

3.2.1 Introduction
Thin, transparent and electrically conducting films are frequently used in consumer appliances, such as LCD touch screen displays, organic LEDs, etc. Their future use in thin film solar cells, e-paper and foldable displays is imminent. However, current thin film materials used for this purpose (such as Indium Tin Oxide or ITO) are very expensive and not flexible enough to allow for such advanced applications. CNTs can act as a viable transparent, conducting thin film material because of its extraordinary electronic properties such as high conductivity and mobility, as well as excellent mechanical and electronic stability towards deformation. A random network of CNTs deposited on a substrate is a quasi 2-D network of metallic and semiconducting wires. Due to 1-D nature of each wire, the network is transparent for infrared and visible wavelength regions. The reflectivity of these films is also very low. One interesting aspect of the nanotube network is that its DC conductivity is decoupled from its optical transparency. This means that the conductivity of a network can be independently increased by growing high quality nanotubes of same density so that the optical transparency remains unchanged[11].

Most of the previous work on using carbon nanotube films is focused on solution based tubes, which are filtered through a membrane and deposited on the transparent quartz substrate [12-15]. This approach has the advantages of being easy and scalable. However, there are several shortcomings too, such as:

55 1. Carbon nanotubes become defective after going through strong chemical treatments and sonication in solvents. This has already been proved through transport experiments on solution deposited nanotubes on a silicon wafer. 2. It is difficult to clean carbon nanotubes completely out of any surfactant remains. This causes low conductivity between individual nanotubes in the film. 3. Strong acid chemical treatments dope the tubes. Although doping the tubes makes semiconducting tubes in the film act as metal thereby increasing the film conductivity, yet, this effect is not permanent, and slowly goes away with time. Extrinsic doping also makes it difficult to extract the true performance of the films

Acknowledging these facts about solution deposited nanotubes, this project started off to find the ultimate performance of as-grown nanotube films obtained through CVD process. It is well known through electronic measurements that pristine CVD grown nanotubes are less defective and have a longer average length in a nanotube film. This should make CVD grown films better conductors and more optically transparent than solution deposited CNTs.

3.2.2 Experiments
Nanotube films are grown over flat, transparent quartz substrate in a CVD oven. The growth chart presented in Fig. 3.5 explains the process parameters that are tested for the film growth. Carbon monoxide and ethanol processes were chosen for growing films. The former is known to give high yields of nanotubes while the latter) gives much cleaner and longer nanotubes. Catalyst densities were increased as compared to single CNT growths

56 to achieve dense films. For CO process, salts of Fe, Co and Mo worked best[16] as catalyst.

Growth Parameters Studied

Carbon Feedstock CO CH4 C2H4 Ethanol

Catalyst Fe, Co, Mo Salts Fe Nanoparticles FeCl3 Fe-Mo Complex

Temperature Low (<750oC) Medium (750oC<T<850oC) High (>850oC)

Fe, Co evaporated films Fe Carbonyl

Figure 3.5: Various process parameters studied for carbon nanotube film growth. The connecting red line shows the combination of carbon feedstock, catalyst and temperature conditions that provided best quality films.

In order to achieve uniform coating of catalyst over quartz, it was deposited using a slow substrate pull out method. Here the quartz slide is first dipped in the catalyst solution and then slowly pulled out at speeds of 1 cm/min. Organic ligands such as acetate that binds the metal in the salt solutions are oxidized through heating the film in air 450oC. Finally the usual CO process explained in chapter 2 is used to grown dense CNT films on quartz.

57 The second process used for growing CNT films was ethanol CVD. It worked best with physically evaporated metal films[17]. The process required the reduction of films to achieve uniformly sized metal nanoparticles. Process conditions (temperature, gas flow rates) were varied to achieve the maximum dense growth with least amorphous carbon deposition (explained in chapter 2).

200 m

30 m

Figure 3.6: SEM images of nanotube film growth on quartz substrates. a) Growth through CO process with Iron Nitrate as the deposited catalyst. b) Ethanol process with evaporated Cobalt film as catalyst.

Fig. 3.6 shows the growth results from carbon monoxide and ethanol processes. Ethanol process worked best with evaporated Cobalt films. As evident from these two images, less catalyst density in case of Fe salts used for CO process, led to longer and less dense tubes. With the use of high density Co films CNT density increased (Fig. 3.6b), however it adversely affected the average tube length. Low nanotube lengths give rise to more tube-tube contacts for the same tube density, resulting in higher sheet resistance. On the other hand, very sparse and long tubes make poor films since tube resistance increases for very long lengths CNTs. As a result of these studies, an optimum evaporated film

58 thickness, with ethanol growth process seemed to work best in providing dense CNT films with longer average length and less amorphous carbon deposition.

3.2.3 Measurements
Electrical sheet resistance measurements were performed using a four-probe measurement setup. Light absorption measurements were done in a UV-Vis light spectrometer with wavelength range 150 nm- 900 nm. Figure 3.7a shows the absorption measurements on quartz samples grown with different densities of CNT films.

100 98

a
98.6 98.4 %T (550 nm) 98.2 98.0 97.8 97.6 97.4

96 %T 94 92 90 400 600 (nm) 800 1000

2.0 k/sq 5.5 k/sq 6.75 k/sq 7.2 k/sq 6.0 k/sq 4.6 k/sq

4 5 6 Rs(k/sq.)

Figure 3.7: Light transmission and sheet resistance measurements on as-grown nanotube films. a) Light transmission measurements on substrate with varying CNT densities. The high absorption at low wavelengths probably comes from metal nanoparticles, whose density is quite high in the films. Note that as the film density increases, the absorption of light increases too. Also, a higher density film gives lower sheet resistance. b) Transmission at technology relevant (550 nm) wavelength. The results suggest that increasing the density of tubes does not severely affect the light

59 transmission (98.6%-97.5%). However, the sheet resistance can be decreased to as much as four times. The lowest value of sheet resistance that is obtained through this process is close to 2.0 k/sq .This value is staggering low when considering that the grown CVD CNTs have a thickness close to a mono layer. Due to low thickness these films are highly transparent in the visible and near IR region. As an example, a 2.0 k/sq sheet resistance film has a transparency of 97.5% at 550 nm wavelength. It is worth noting that at such high transparencies, the reflective loss of light passing through the films on quartz substrates becomes important, which the instrument does not register. Therefore the actual light transparency could be much higher than current values. Figure 3.8 shows the transparency vs. sheet resistance for CVD grown CNT films on quartz substrates (red circles). A comparison is done with the solution deposited films[11] and other conducting thin materials such as graphene oxide, ITO and EL-P ink. At higher values of transmittance, the non-doped CVD films perform similar to the solution deposited (gray curve). Doping the CVD films with Thionyl Chloride decreases the sheet resistance to 1/3rd of the value, with no loss of transparency. Doping makes the semiconducting CNTs which, statistically constitutes 2/3rd of the film, conducting. This data point suggests that CVD grown CNTs can perform better than solution deposited in medium to high sheet resistance range. However, it is difficult to bring the sheet resistance of CVD grown CNT films to less than a 2 k/sq from current processing technique. This may be due to ultra thin nature of as-grown CNT films.

60

Reduced GO

Rs (O/sq.)

Thionyl Chloride doped films

CVD grown

%T, 550 nm

Figure 3.8: Comparison with results in literature. The graph shows the sheet resistance and transmittance of solution deposited nanotube films (gray circles and lines)[11], best results from graphene oxide (GO) films[18] (brown ellipse), EL-P ink thin film material (black squares), as grown films through floating catalyst CVD process[19] (brown ellipse) and ITOPET as the benchmark material. The red circles are the data points from Fig 3.7b, with the green circle representing the drop in sheet resistance after doping the nanotube films with Thionyl Chloride solution.

Attempts to grow thicker CNT films (through depositing more catalyst, increasing growth time, increasing carbon feedstock) resulted in an increase in sheet resistance. A probable cause for this is either increased catalyst clumping in case of thicker catalyst layer or more amorphous carbon deposition as the growth time and/or carbon feedstock is increased. Recently, lower sheet resistances have been reported through a different CVD

61 method[19], the data point for this result is shown in Fig. 3.8 (brown ellipse with an arrow), however, the light transmission from these films was not good enough.

3.2.4 Summary
It is shown that CVD grown carbon nanotube films can perform much better as transparent conductors than the solution deposited films. However, as grown CNT films falls short of the low sheet resistances desired for practical applications. Doping them, post growth makes their conductivity far better than solution deposited CNT films, with no decrease in transparency. The results of this study suggest the potential of ultra clean CNT films as an alternative to currently used transparent conductors. Although, direct growth of CNTs on substrates not scalable as well as practical for low T materials, it successfully sheds light on the performance limits of ultra clean CNT films. The results demonstrate that there is still a large room for improvement in the performance of solution deposited CNT films, if they could be made as clean as directly grown films, they can certainly beat ITO in electrical and optical performance.

62

Bibliography
1. Iijima, S., Helical Microtubules of Graphitic Carbon. Nature, 1991. 354(6348): p. 56-58. 2. Endo, M., et al., Synthesis and application of carbon nanotubes. Chimica OggiChemistry Today, 2005. 23(4): p. 16-+. 3. Saito, R., G. Dresselhaus, and M.S. Dresselhaus, Physical Properties of Carbon Nanotubes. 2001, Singapore: Imperial College Press. 4. Barton, S.C., et al., The "Wired" Laccase Cathode: High Current Density Electroreduction of O2 to Water at +0.7 V (NHE) at pH 5. J. Am. Chem. Soc., 2001. 123(24): p. 5802-5803. 5. Moisala, A., A.G. Nasibulin, and E.I. Kauppinen, The role of metal nanoparticles in the catalytic production of single-walled carbon nanotubes - a review. Journal of Physics-Condensed Matter, 2003. 15(42): p. S3011-S3035. 6. Smiljanic, P., et al., Growth of carbon nanotubes on Ohmically heated carbon paper. Chemical Physics Letters, 2001. 342(5-6): p. 503-509. 7. Sun, X., et al., Growth of carbon nanotubes on carbon paper by Ohmically heating silane-dispersed catalytic sites. Chemical Physics Letters, 2002. 363(5-6): p. 415-421. 8. Hata, K., et al., Water-Assisted Highly Efficient Synthesis of Impurity-Free SingleWalled Carbon Nanotubes. Science, 2004. 306(5700): p. 1362-1364. 9. Bladh, K., L.K.L. Falk, and F. Rohmund, On the iron-catalysed growth of singlewalled carbon nanotubes and encapsulated metal particles in the gas phase. Applied Physics A: Materials Science & Processing, 2000. 70(3): p. 317-322.

63 10. P. L. Walker, J.F.R., and G. R. Imperial, Carbon Mixture over Iron Catalysts. II. Rates of Carbon Formation. Journal of Physical Chemistry, 1959. 63(2): p. 140149. 11. Gruner, G., Carbon nanotube films for transparent and plastic electronics. Journal of Materials Chemistry, 2006. 16(35): p. 3533-3539. 12. Chhowalla, M., Transparent and conducting SWNT thin films for flexible electronics. Journal of the Society for Information Display, 2007. 15(12): p. 10851088. 13. Shin, J.H., et al., Smooth, transparent, conducting and flexible SWCNT films by filtration-wet transfer processes. Journal of Physics D-Applied Physics, 2009. 42(4): p. 4. 14. Takenobu, T., et al., High-performance transparent flexible transistors using carbon nanotube films. Applied Physics Letters, 2006. 88(3): p. 3. 15. Venancio, E.C., et al., Flexible Thin Films of Single-Walled Carbon Nanotubes Deposited on Plastic Substrates. Journal of Nanoscience and Nanotechnology, 2009. 9(1): p. 567-571. 16. J. Hone, P.K., X.M.H. Huang, B Chandra, R. Caldwell, J. Small, B. H. Hong, T. Someya, L. Huang, S. OBrien, C. P. Nuckolls, Growth of Nanotubes and Chemical Sensor Applications. Proceedings of the SPIE (2005), 2005. 17. Huang, L.M., et al., Cobalt ultrathin film catalyzed ethanol chemical vapor deposition of single-walled carbon nanotubes. Journal of Physical Chemistry B, 2006. 110(23): p. 11103-11109.

64 18. Wang, X., L. Zhi, and K. Mullen, Transparent, Conductive Graphene Electrodes for Dye-Sensitized Solar Cells. Nano Letters, 2008. 8(1): p. 323-327. 19. Ma, W., et al., Directly Synthesized Strong, Highly Conducting, Transparent Single-Walled Carbon Nanotube Films. Nano Letters, 2007. 7(8): p. 2307-2311.

65

4 ______________________________
CHAPTER

Introduction to Electron Transport in Carbon Nanotubes ______________________________

66

4.1 Introduction
Single wall carbon nanotubes (CNTs) provide a perfect model system for 1-D conductors due to its nano size cross- section and atomically uniform surface. Depending on the structure defined by (n,m) indices, nanotubes can be metallic or semiconducting which is quite a unique property among other nano structures. The most striking aspect of transport in carbon nanotubes is that it can culminate itself into various classical and quantum regimes. In the first part of this chapter we will discuss briefly some of these transport phenomenons. In the second half, fabrication of nanotube devices to study these processes will be discussed along with several new tools that enable to study the transport properties of known structure CNTs.

4.1.1 Metallic and Semiconducting CNTs

Metallic CNTs have finite density of states at the Fermi energy. As explained in section 1.5 of this thesis, for a (n,m) type CNT, if n-m is a multiple of integer 3, then it shows metallic behavior[1]. Extensive tight binding calculations which take tube curvature into account predicts a small band gap opening (~ 10 meV) for metallic nanotubes with nm. Semiconducting nanotubes have a finite gap where density of states is zero. This makes them an insulator unless an external field is applied to move the tube Fermi level to conduction or valence band.

4.1.2 Quantum Transport in a 1-D System

An interesting aspect of low dimensional materials is the quantization of electron wave functions. In CNTs, the quantization is in circumferential direction due to its ultra small

67 diameter. This causes large energy spacing between 1-D sub-bands. Therefore when an electron jumps from a 3-D metal contact to a 1-D system like CNT it sees only two sub bands near its Fermi energy. These two sub bands are present due to the orbital degeneracy. As the conductance quantum for each sub band is e2/h, therefore an electron will see combined conductance of 2e2/h. In addition to this each sub band can accommodate two electrons of different spins. Therefore the total conductance will be 4e2/h or 155 S. In resistance terms this corresponds to 6.5 k. Hence, the minimum resistance that an electron will see when it enters a nanotube is 6.5 k, this is independent of nanotube length or chirality. For a more elaborate complete proof for the quantum of conductance please see ref[1, 2].

4.1.3 Schottky Barriers at the Nanotube- Metal Interface

Whenever materials of different work functions are connected together, there is a charge transfer and an equilibrium electric field. This field can bend the electronic bands so that a barrier for the flow of electrons is created. Such a barrier is called a Schottky barrier. This is a well understood phenomenon in traditional semiconductors. Since nanotubes also have different work function than metals contacting them, there always exists a Schottky barrier at the nanotube-metal interface. This barrier is more pronounced in semiconducting CNTs than metals since SC possess a band gap. Henceforth in addition to the minimum resistance that comes from only two sub bands present in the transport, there is an additional resistance at the contacts due to schottky barriers[3].

68 Another source of contact resistance is bad contact between metal and nanotubes, which may occur due to unclean nanotubes or polymer scum layer left after lithography processes. Usually this resistance is also included in contact resistance.

4.1.4 Electronic Mean Free Path and Phase Coherence Length

At different channel lengths (L) CNTs can exhibit completely different electronic phenomenon. Phase Coherence Length (L) is described as the length at which electrons can retain their phase information. Mean Free Path (Lm) is the average length an electron traverses before it encounters a scattering event. Sources of electron scattering in an ideal crystal are mainly thermal vibrations of the lattice or phonons. However, in real crystalline material systems lattice defects, charge impurities and environmental effects such as structure deformation can also cause an electron to scatter. Various electron transport regimes can be divided on the basis of the above mentioned length values.

1. L > Lm, L : This is the diffusive transport regime, since the length of nanotube channel is greater than electron mean free path as well as phase coherence length. Here nanotube will act as a normal metallic wire following ohms law, where resistance of a nanotube varies linearly with length. 2. L < Lm, L: Fabry-Perot like resonances are observed, when the contacts are electron transparent. These are the standing electron waves formed due to the backscatter of the electron waves from contacts.

69 3. Lm< L: Anderson localization occurs due to localized wave functions around the elastic scatters. At very low temperatures this results in a complete blockade of current through the channel.

4.1.5 Nanotube Band Diagrams-Field Effect Transistor Action

Transport through conduction, valence bands as well as the schottky barriers[3] at the metal nanotube contacts can be explained through simple band diagram. Fig. 4.1 shows the standard three-terminal nanotube device geometry used throughout this work.
Drain

280 nm Gate

B
Vg=0 c a a Vg=-(b + c)/

C
Vg=(a- c)/

a b

EF
b b

25.0 Conductance (S) 20.0 15.0 10.0 5.0 0.0 -20

Vg=-(b + a)/

B C A
-10 0 10 Gate Voltage (V) 20

70 Figure 4.1: Nanotube band diagrams. The diagrams explain the transport mechanism for a semiconducting nanotube with a band gap. Image at the top of the figure depicts the nanotube device geometry, showing source, drain electrodes and a back gate. In these diagrams it is assumed that nanotube is p-doped, is the gate efficiency. A) Shows alignment in the absence of a gate field. B) Nanotube bands move up as negative back gate voltage is applied. A negative voltage on gate injects positive charge on nanotube, this brings the Fermi level closer to the valence band, or in other words the bands move up with respect to the Fermi level. C) Opposite to what happens in B, this time with positive gate voltage move the bands down with respect to the Fermi level.

A gate voltage applies field on nanotube which can change its electrostatic potential. A positive gate voltage injects negative charge onto the tube and therefore decreases its potential with respect to a zero vacuum potential level (bands move down). Whereas a negative gate voltage moves the potential of the tube higher. Here the assumption is that source and drain voltages are small compare to gate. Therefore channel potential is governed mainly through gate. In figure, the semiconducting nanotube device is assumed to be p-doped (which is true, experimentally due to oxygen doping), this causes a schottky barrier with low work function metal. Due to the p type doping of the tube schottky barriers height for n-channel and p-channel is different. As shown in Fig. 4.1A barrier for hole transport is b while for electron transport is a. Changing the gate voltage moves the bands up or down, which also changes the tunnel thickness of the schottky barrier. Transport occurs through conduction or valence band when there are available states in the bias window of source and drain

71 electrodes and barriers at the contacts are thin enough to cause the electron to tunnel into the available state. For this reason, nanotube transistors action is said to be schottky barrier mediated. a>b results in a lesser conductance through the n-channel compare to the p-channel as shown in the conductance gate sweep of a semiconducting nanotube device. Resistance of nanotube is usually measured at high gate voltages (regions B and C) since in these regions the contacts do not dominate the transport through the tube. Similar band diagrams can be made for metallic nanotubes. If they possess a small band gap[4] (for e.g. chiral metals) then the result will look exactly like a semiconductor with a much smaller conductance gap and lower schottky barriers.

4.1.6 Coulomb Blockade in CNTs

As mentioned in previous sections, a single wall carbon nanotube can exhibit different regimes of transport depending upon the transparency of the nanotube-metal contacts and the nanotube channel length. At very short channel length the capacitance of tube becomes small too (C~L) resulting in an increased charging energy (U~1/C) of the nanotube channel. If the electrostatic potential of the nanotube becomes less than the charging energy required (in quantum regime a separate orbital energy should also be included) the transport through the channel becomes blocked. This phenomenon is termed as Coulomb blockade[5, 6]. Figure 4.2 shows a schematic of a nanotube circuit. Here a tube of length L is connected though metal electrodes which can apply potential Vs. In addition to metal source and drain capacitances, the nanotube is also capacitive

72 coupled to the back gate. Electrons can tunnel in and out through the metal nanotube barriers but no electrons can tunnel through the gate.

e-

e-

Cs

Cg

Cd

Vs

Vg

Figure 4.2: Capacitive model of nanotube device. The source, drain and gate capacitances are in parallel, hence the total capacitance of the system is the sum of all three capacitances.

For an electron to tunnel into the tube, it needs an available state in the tube. Whether this state is present in the bias window of source and drain electrode or not depends upon the electrochemical potential of the nanotube channel. Assuming there are N electrons residing in the nanotube channel, the electrochemical potential of the tube is given by

U ( N ) = E n ,l +
B

2 Qdot

2C

(4.1.1)

Here En,l denotes the energy level spacing in the nanotube channel due to quantum confinement in the nanotube channel at short lengths. C is the total capacitance of the system.

C = Cs + C g + Cd

(4.1.2)

73 The net charge on the nanotube channel (Qdot) is given by:

Qdot ( N ) = e( N N 0 ) (C gVg + CsVs )

(4.1.3)

Here, Cg and Cs are the gate and source capacitances respectively (Fig. 4.2). For an electron to enter inside the nanotube channel, the change of electrochemical potential from N to N-1 electron should match the applied source potential. This condition is indicated by the following relation:

t = U ( N ) U ( N 1)

(4.1.4)

Putting in the relation for U (N) from equation 4.1.1, this condition becomes: 1 e2 e t = E N + ( N N 0 ) (C gVg + CsVs ) C 2 C (4.1.5)

Since the bias window is between 0 and Vs, the above relation is solved for two scenarios,
t = 0 and t = eVs . Fig.4.3 shows the resultant straight lines for N=0 and En=0.

N=0

Vsd

t = eVs

Slope=

Cg (C C s )

Vg

t = 0

Slope=

Cg Cs

74 Figure 4.3: Ground state lines of a nanotube QD. Obtained from equation 4.1.5, with

N=0 and EN=0. It is possible to find the individual capacitances of the system by using slope of these lines along with the equation 4.1.2. These slopes can also give valuable information regarding the coupling strength of contacts to the nanotube channel. Such an approach has been utilized in understanding the transport through a nanotube quantum dot (QD) in succeeding chapter.

t = eVs

Slope=

Cg (C C s )
e C
E e + e C

Addition energy

Addition energy

C e ( ) Cg C

C E e ( + ) Cg e C

N=1

N=2

t = 0

Slope=

Cg Cs

Figure 4.4: Stability diagram of the nanotube quantum dot. The thicker lines define

the diamond like regions inside which the electron transport is blocked. The addition energy, which is the energy required to add one electron in the QD, alternates between just being the charging energy and a sum of charging energy and single particle level spacing.

75

The slope of source and drain lines is directly related to the system capacitances. As the gate voltage and source bias changes the number of electrons in the channel will change, giving rise to the repeated line patterns as shown in Figure 4.4. This is termed as a stability diagram of a Coulomb blockaded system. The thicker lines represent the ground state of the nanotube energy levels. Inside the diamonds surrounded by the lines the number of electrons in the nanotube channel becomes a constant and no current passes through it.

E
k

DOS

Figure 4.5: Quantization of nanotube bands. With the decrease in nanotube length (L), the wave vector in the translational direction (T) also becomes quantized. This

gives rise to discrete energy levels, similar to a zero dimensional molecular quantum dot. In addition to L, the energy level spacing depends on the Fermi velocity and the level degeneracy of the system. If the channel is only spin degenerate (2 fold degeneracy) then denominator will be 4L. For ultra clean nanotube QD, it is possible to obtain 4- fold degeneracy (2 spin, 2 orbital). In that case the expression will have 2L in the denominator.

76

Outside the diamonds, at higher source bias voltages, lines running parallel to the diamond edges represent excited states of the system[5].

Extra lines parallel to ground state lines shown in Fig.4.4 are the excited state of the QD. These states become available at higher source bias. They occur due to reduction in nanotube length, which leads to a quantum confinement, making the nanotube channel a zero dimensional quantum dot. The further quantization of nanotube bands is illustrated through Fig. 4.5. Here E is the orbital level spacing for a two-fold shell filling. Due to the presence of charging energy and level spacing, the energy required to fill an electron alternates between U and U+ E. This is illustrated through Figure 4.6.

VGate= 0

VGate= U

VGate= U+E

U U+E

Vs

Figure 4.6: Quantum dot transport. The discreet energy level spacing of the QD

system alternates between U and U+E. In reality nanotubes are a spin degenerate system (except in case of magnetic impurities), but in a nanotube QD the degeneracy gets lifted due to single electron charging energy.

77

4.2 Carbon Nanotube Device Fabrication

4.2.1 Introduction
In the beginning years of carbon nanotube science, the transport measurements were dominated by the process of depositing carbon nanotubes on pre patterned electrodes[7]. Through this approach the contact resistances were found to be very high; this was because nanotubes were not contacting the metal well. With the advent of on chip nanotube growth through CVD process, top contacts on nanotubes become a preference. As these nanotubes were long, oriented and could be grown from specific catalyst sites, the process of electron beam lithography for putting leads on each of them became much easier. In our study same process is used to deposit metal contacts on carbon nanotubes. One major difference between previous electron transport studies and present is the use of known chirality carbon nanotubes for transport measurements. For a long time, transport measurements on carbon nanotubes were done without the knowledge of their structure. Due to the limitations of microscopy techniques such as AFM and SEM, it was also not clear if the nanotube in study is a single tube or a two tube bundle or a double wall nanotube. Our process of device fabrication contains several steps to obtain known chirality nanotube devices for electron transport studies[8]. These steps are shown in Figure 4.7.

78

Figure 4.7: Nanotube device fabrication process sequence.

Step1, which is growth of suspended carbon nanotubes on Si/SiO2 wafer, has already been presented in chapter 2. Here we use meticulously designed growth recipes and special catalyst preparation to achieve good quality single wall carbon nanotubes. The other there process steps are characterization of tubes through Rayleigh scattering, Nanotube Transfer Printing, and Metal Electrodes Deposition. We will go through these steps one by one in coming sections.

4.2.2 Rayleigh Scattering Spectroscopy

Optical spectroscopy techniques, such as fluorescence or Raman scattering, have been shown to be powerful tools for characterization of nanotubes. The suspended nanotube structures are particularly well suited for optical characterization due to lack of background signal from the substrate. In this study, the nanotubes are characterized primarily by Rayleigh scattering spectroscopy[9]. In a Rayleigh scattering spectrum excitonic transitions in the nanotube appear as peaks in the scattering spectrum. Here, super-continuum white light is focused using a microscope objective to 2 m2 area and is directed at normal incidence onto the suspended SWNT as shown in Figure 4.8. The

79

scattered light is collected at an oblique angle in a confocal arrangement and is analyzed by a two-dimensional charge coupled device (CCD) array cooled by liquid nitrogen. The Rayleigh scattering spectra are then obtained after correction for the profile of the supercontinuum light and the detection sensitivity of the system. Power integrated over the entire spectrum is of a few mW, comparable to what is typically used in Raman scattering experiments.

CCD Sensor

Catalyst

Carbon Nanotube

Super Continuum White light

Non-linear optical fiber

Figure 4.8: Single tube Rayleigh spectroscopy setup.

Under these conditions, possible contributions from nonlinear and thermal effects are ruled out experimentally because the Rayleigh scattering spectra are independent of excitation intensity. Rayleigh scattering data in the spectral range of 450 to 850 nm were

80

recorded with integration times of 10 to 30 s at different locations on the nanotube with a step size of 5 m. Rayleigh spectrum taken from single nanotubes is shown in Figure 4.9.

a
(15,11) Intensity (a.u.) S33 S44
Intensity (a.u.)

b
(21,15) M22(+) M22(-)

1.6

1.8

2.0 2.2 Energy (eV)

2.4

2.6

1.6

1.8

2.0 2.2 2.4 Energy (eV)

2.6

c
(18,18) Intensity (a.u.) M22

1.6

1.8

2.0 2.2 2.4 Energy (eV)

2.6

Figure 4.9: Nanotube Rayleigh scattering results: a) S33 and S44 excitonic transition

of a (15,11) nanotube. b) Chiral metal nanotube, (21,15). The split peaks result from the trigonal warping effect. c) Armchair nanotube, a single peak in scattering intensity is observed. Sometimes a confirmation through Raman signal becomes necessary to assign a definite chirality.

81

For semiconducting nanotubes usually two peaks in intensity are observed, one corresponding to S33 and other to S44 transition (Fig. 4.9a). For chiral metal nanotubes M22 (+) and M22 (-) split peak are observed (Fig 4.9b). , this comes from the trigonal warping effect in chiral metal nanotubes[9]. For armchair only one peak is observed usually corresponding to M22 and M33 transition (Fig. 4.9c). One advantage with Rayleigh scattering is that it works for every type of nanotube, whereas fluorescence only works for semiconducting tubes. Also, the technique is fast and does not require a tunable laser source, as is required for resonance Raman scattering. Rayleigh scattering can also distinguish single tubes from bundles, and can examine the structure at multiple points along the tube. After the spectrum is taken, the intensity peaks are correlated with the trends of an already existing database consisting of TEM structural assignments of Rayleigh peaks for various nanotubes[10]. This helps in assigning the appropriate chirality to the nanotubes.

4.2.3 Nanotube Transfer Printing

It is important for nanotubes to be on a solid substrate with a back gate for performing electron transport measurements. For measurements to be done on suspended Rayleigh characterized nanotubes, a process is developed in our lab to transfer these nanotubes to the desired location on a Si/SiO2 transfer wafer. To increase the repeatability and quality of tube transfer, this process has undergone several changes since its first publication[11], although the basic idea of transferring suspended nanotubes remained unchanged. The latest version of nanotube transfer process is shown in Figure 4.10.

82

Optical Microscope

xyz control

Aluminum Plate Nanotube Growth Wafer

PMMA Transfer Wafer Heater

Figure 4.10: Nanotube transfer setup. The growth wafer contains Rayleigh

characterized (known chirality) suspended nanotube, shown in red in the figure. Stage XY movements have an accuracy of 5 m.

First, the substrate on which nanotubes need to be transferred is patterned with metal alignment marks. These marks define the location and alignment of the transferred nanotube. In the second step, a thin layer of PMMA resist (1.5%, A3) is spun over the substrate. For tube transfer, the wafers containing suspended, characterized nanotubes are flipped and their top surface (side containing nanotube) is carefully brought into contact with PMMA spun transfer substrate, this is shown in Fig 4.10. This step is done under a microscope and position of transfer is controlled using a micrometer with xyz controls on which the suspended nanotube substrate is fixed. In the last step, after both the chips are in contact, substrate temperature is raised to 180o C, well above the PMMA glass transition temperature. After 1 minute, the temperature is brought down and substrates are pulled away from each other. At this point the nanotube is transferred over the

83

substrate with alignment marks, although it is sitting on a thin layer of PMMA. To make it anchor to the bottom substrate, an array of windows are opened using electron beam lithography. Finally, the rest of resist is slowly dissolved in acetone. This completes the nanotube transfer. Nanotube transfer can also be possible with contacting the suspended nanotube with patterned substrate directly without the aid of resist. However, use of PMMA layer makes it much more repeatable and clean. The PMMA layer solves two purposes: 1. it acts as glue for transferred nanotube. 2. It protects the transfer chip from getting dirt particles through the growth chip. Removal of PMMA after transfer lifts off all the unwanted materials coming from the process, hence, provide a much cleaner chip to fabricate devices on.

4.2.4 Metal Electrode Deposition

Figure 4.11 shows a known chirality nanotube transferred between pre patterned alignment marks. The metal electrodes are made up of palladium (Pd).
Carbon Nanotube Gold electrodes Palladium electrodes

Figure 4.11: Nanotube device on silicon wafer. a) SEM image shows single wall

carbon nanotube (24, 12) with palladium leads. b) Optical microscope image of the nanotube device. Outside bond pads are made of gold with a Chromium adhesion layer.

84

Standard electron beam lithography (EBL) along with electron beam evaporation is used to deposit metal electrodes on the nanotube[8]. For contacting the nanotube palladium metal is chosen due to its high work function which makes less resistance contacts with p-doped nanotube[12]. The bonding pads were made out of gold with titanium as the sticking layer. In the final step one more EBL is performed followed by oxygen plasma treatment to burn off any other nanotube which might have gotten transferred and could short the metal electrodes. This finishes the fabrication of a known chirality nanotube device.

85

Bibliography
1. Saito, R., G. Dresselhaus, and M.S. Dresselhaus, Physical Properties of Carbon

Nanotubes. 1998: Imperial College Press.

2. Datta, S., Electronic transport in mesoscopic systems. 1997: Cambridge University Press. 3. Heinze, S., et al., Carbon nanotubes as Schottky barrier transistors. Physical Review Letters, 2002. 89(10). 4. Ouyang, M., et al., Energy gaps in "metallic" single-walled carbon nanotubes. Science, 2001. 292(5517): p. 702-705. 5. Beenakker, C.W.J., Theory of Coulomb-Blockade Oscillations in the Conductance

of a Quantum Dot. Physical Review B, 1991. 44(4): p. 1646-1656.

6. Kouwenhoven, L.P., D.G. Austing, and S. Tarucha, Few-electron quantum dots. Reports on Progress in Physics, 2001. 64: p. 701-736. 7. Yao, Z., et al., Carbon nanotube intramolecular junctions. Nature, 1999.
402(6759): p. 273-276.

8.

Chandra, B., et al., Electrical transport measurements of nanotubes with known (n,

m) indices. Physica Status Solidi B-Basic Solid State Physics, 2006. 243(13): p.
3359-3364. 9. Sfeir, M.Y., et al., Probing electronic transitions in individual carbon nanotubes

by Rayleigh scattering. Science, 2004. 306(5701): p. 1540-1543.

10. Sfeir, M.Y., et al., Optical spectroscopy of individual single-walled carbon

nanotubes of defined chiral structure. Science, 2006. 312(5773): p. 554-556.

86

11.

Huang, X.M.H., et al., Controlled placement of individual carbon nanotubes. Nano Letters, 2005. 5(7): p. 1515-1518.

12.

Javey, A., et al., Ballistic carbon nanotube field-effect transistors. Nature, 2003.
424(6949): p. 654-657.

87

5 ______________________________
CHAPTER

Molecular-Scale Quantum Dots from Carbon Nanotube Heterojunctions ______________________________

88

5.1 Introduction
Soon after the discovery of carbon nanotubes, it was realized that nanotubes of different chiralities can be covalently joined through the introduction of a small number of pentagon-heptagon defects, forming a nanoscale metal/semiconductor or metal/metal carbon nanotube heterojunction (HJ)[1]. In spite of the interest in such devices, there have been only a few direct observations of carbon nanotube HJs through scanning tunneling microscopy[2, 3] and Raman spectroscopy[4, 5] techniques. The only

electrical transport study reported diode-like behavior across a possible metalsemiconductor HJ device[6], identified by the presence of a sharp kink in the nanotube. However, no prior work has combined structural characterization and transport measurements on such HJs. Here we report detailed electrical transport measurements of a HJ connecting a (21,9) metallic nanotube and an (18,13) semiconducting nanotube, as identified by optical spectroscopy.

5.2 Device Fabrication

For this study, single-wall carbon nanotubes (SWCNTs) were grown by chemical vapor deposition across open slits (~100 m wide) on Si/SiO2 wafers[7], then characterized using Rayleigh scattering spectroscopy[8]. Typically the Rayleigh spectrum remains unchanged along a suspended nanotube. The insets of figures 1b and 1c show the Rayleigh spectra for the nanotube used in this study, a rare case in which the spectrum changed from one typical of a semiconducting (S) SWCNT to that of a metal (M) SWCNT near the center of the suspended structure. Previous structural studies[9] allow us to assign chiral indices of (18,13) and (21,9) for the S and M sections, respectively. Because of the nanotubes mechanical integrity, and because the structure change is

89 abrupt within the size of the laser spot (~1 m), we conclude that the two sections are covalently joined by a linear HJ. For electrical measurements, the suspended nanotube containing the HJ was transfer printed[10] onto a Si / SiO2 wafer, and multiple Pd electrodes were patterned using electron-beam lithography (Fig 5.1a).

Heterojunction Metal Semiconductor

10 m
b
800 nm

Figure 5.1: Device geometry and electrical measurements on semiconducting and metallic sections (a) SEM image of the nanotube device, taken after electronic measurements. Multiple electrodes were patterned on each side of the junction to characterize the semiconducting (SC) and metallic (M) sections. A larger spacing (15.5 m) was left for the HJ segment to allow for a possible misalignment during the tube transfer process. The arrow points toward the approximate location of the HJ. (b) and (c) Room-temperature Rayleigh scattering spectra (insets) and low-bias conductance vs. gate voltage for the (21,9) M and (18,13) SC sections.

90 Imaging of the central section containing the HJ by atomic force microscopy (Fig.5.1b) revealed no visible kink, and a uniform diameter of ~2 nm, consistent with the assigned chiral indices.

5.3 Electronic Transport Measurements

Transport measurements on the S and M sections (Figures 5.2a and 5.2b) show behavior uniformly consistent with the structural assignments.
a b
Intensity (a.u.)

100 Conductance (S) 10 1 100n

100 Conductance (S)

800 nm

18.5 m
Intensity (a.u.)

10 1 100n

400 nm

(18,13)

(21,9)

1.6

6.0 m

1.8 2.0 2.2 Energy (eV)

2.4

10n -10

1.8 2.0 2.2 2.4 2.6 Energy (eV)

-5 0 5 Gate Bias (V)

10

10n -10

-5 0 5 Gate Bias (V)

10

Figure 5.2: Electrical measurements on semiconducting and metallic sections. Room-temperature Rayleigh scattering spectra (insets) and low-bias conductance vs. gate voltage for a) (21,9) M and b) (18,13) SC sections.

All of the (21,9) devices (length 18.5 m and 800 nm) show gate response characteristic of chiral metal nanotubes, which possess a small curvature-induced bandgap[11]. Analysis of the length-dependent resistance shows a low contact resistance (~15 k) and a resistivity of 9 k/m, typical for clean metallic nanotubes at room temperature[12]. The (18,13) devices (length 6.0 m and 400 nm) likewise display behavior typical of large-

91 diameter S nanotubes with Pd contacts[13, 14]. Both chiralities show symmetric currentbias behavior. Low-bias gate sweeps (Fig. 5.3a) of the HJ device show ambipolar response, similar to the S sections. However, the HJ device uniquely displays strong conductance oscillations that appear below ~100 K, and intensify at lower temperatures. Furthermore, the currentvoltage behavior of the HJ (Fig. 5.3b) is highly asymmetric, on first glimpse resembling the rectification typically seen for a M-S Schottky diode. However, the off region both shifts laterally and changes width as the gate voltage is varied, resulting in a diamondshaped pattern as seen in inset of Fig. 5.3b.
a
10 Conductance (S) 1 100n 10n 1n -5 0 5 Gate Bias (V)
10 K 20 K 50 K 100 K

b
Vg=3.5 V

450 Current (nA) 300 150 0

Current (nA)

0.8
Vg = 0 V

0.4 0.0 -0.4

Vg = 3.5 V

-0.8 -500 -250 0 250 500 Source Bias (mV)

Vg= 1 V

-150 -600 -300 0 300 Source Bias (mV) 600

Figure 5.3: Temperature dependent measurements across the heterojunction (a) Conductance gate sweep at four different temperatures, source bias is 27 mV. (b) Current-Voltage measurements at 10 K for gate voltages from 1.0 V to 3.5 V. The off region shifts and changes width with gate voltage, displaying a clear diamond pattern (inset, with curves offset by 0.1 nA).

92 In addition, the I-V curves show considerable structure at low temperature, in particular the presence of sharp current steps that are often succeeded by pronounced negative differential resistance. The observed conductance oscillations and current steps are

inconsistent with a model of a nanotube M-S heterojunction as a simple Schottky diode, and strongly suggest QD behavior. The presence of a QD is confirmed by measurement of the low-temperature (10 K) differential conductance of the HJ as a function of source and gate voltage. The resulting stability diagram (Fig. 5.3a and 5.4b) shows a series of triangles of zero conductance, characteristic of the Coulomb charging spectrum of a single QD[15, 16].

a
Source Bias (mV)

5 4 3 2 1 0

Gate Bias (V)

b
200

Source Bias (mV)

100 0 -100 -200 -2.0 -1.5 -1.0 -0.5 0.0

Gate Bias (V)

Conductance (S)

93 Figure 5.4: Transport across heterojunction (a) Differential conductance of the HJ at T=10K, measured as a function of source-drain and gate bias for positive biases at the onset of the p-type conduction. The edge of the large diamond appears near Vg = 0.25 V. Below this value, smaller half-diamonds with sharp edges are visible, inside which conductance is zero. Bright lines parallel to the diamond edges indicates excited states. The solid black lines next to some of the excited-state lines show regions of pronounced negative differential resistance. This is likely due to the entry of loosely coupled excited states into the bias window which then block the current through the ground state of the QD[17]. (b) Lower-resolution plot showing both positive and negative biases. Qualitatively, looking at the stability diagram, the absence of excited state lines with a negative slope indicates a much stronger coupled semiconducting nanotube than metallic one. The coupling of metal and semiconducting tube to the QD is discussed more thoroughly in the next sections.

5.4 Constant Interaction Model

In a constant interaction model[15] (Fig. 5.5a), the stability diagram can be analyzed to yield energy scales and coupling parameters for the QD. The height of the triangles gives the electron addition energy Eadd 135 meV , slightly higher than the largest previously reported for a nanotube QD device[18]. The distance between the excited state lines running parallel to the triangles gives the single-particle level spacing E 75 meV, and the difference between E add and E is the Coulomb charging energy, U 60 meV. A second indication of these two energy scales appears at larger negative gate voltages

94 (Fig. 5.5b), where the addition energy alternates between U and U + E due to the spin degeneracy of the electronic states. The capacitances to the QD can be extracted from the stability diagram of Fig 5.4a by measuring the ground state lines forming the charging diamonds, with the positive and negative slopes given by
Cgate Csource
C gate C Csource

and

a
(18,13) Source
Csource

HJ

(21,9) Drain

Conductance (S)

16 14

b
U U U

Cdrain

U + E

Cgate

12
U + E

U + E

10

Gate

-9

-8 -7 -6 -5 -4 Gate Bias (V)

-3

Figure 5.5: Model of the metal-semiconductor nanotube HJ device. (a) Source, drain, and gate capacitances to the QD are denoted by Csource , Cdrain and C gate respectively,
m and s

are the coupling constants of metal and semiconducting nanotube to the

quantum dot. (b) Low-bias conductance vs. gate voltage measurements away from the gap at T=10K. The peak separation shows a two-fold shell filling pattern, alternating between a value corresponding to addition energy of U and one corresponding to U+E.

Here C is the sum of all the three capacitances ( C = C source charging energy by U = e 2 C . This

+ C drain + C gate )

and is related to
C drain = 0.46aF

yields C source = 1.86aF ,

95 and C gate = 0.37aF . Again the capacitances show that Source (Semiconducting nanotube) is much highly coupled to the QD than the drain (metal nanotube). This is illustrated more with the sequential tunneling mode in the following section.

5.5 Sequential Tunneling

In a sequential tunneling model, the coupling constants of the nanotube leads to the QD can be derived from the magnitude of the current steps across an excited state line[19, 20]. At Vg = 4 V the coupling of the M-SWCNT to the QD is m ~ 1.5meV 2meV , and the coupling between the SC-SWCNT and the QD is s ~ 5meV 12meV . These values remain in the 1-10 meV range at all gate voltages, while the ratio ( 6.

m ) varies from ~3-

I+ = 2

| e | r l h 2 r + l l r
R F

I = 2

| e | r l h r + 2 l l r

L F R F

L F

Figure 5.6: Rate equations for sequential tunneling transport through a single spindegenerate level. In (a) electrons can tunnel in QD from right lead only, this brings and extra factor of 2 with in the denominator of rate equation, opposite situation happens in (b).

96

20 Vg = 4 V Current (nA) 10

s ~3 m
Vg = -1 V

s ~6 m
Current (nA)

12.5 nA 0 8.85 nA -10 -20 -150 -75 0 75 Source Bias (mV) 150

10.5 nA 0 6.5 nA

-75 0 75 Source Bias (mV)

Figure 5.7: Coupling constants of metal nanotube and SC nanotube to QD ( m and s , respectively) solved using current steps originating from sequential tunneling through a single, spin-degenerate sublevel. The coupling ratios for a state on each side of the band gap (Vg=4V and Vg=-1V) is shown.

The energy scales derived above can be used to estimate the size of the QD. In a simple picture, the level spacing is given by E = hvF 4 L where vF is the Fermi velocity (8x105 m/s), h is Plancks constant, and L is the quantum confinement length. Using
E =75 meV, a QD length of 10 nm is obtained. The measured gate capacitance of 0.37

aF yields a similar QD length of 11 nm through Cgate =

2 L , where d and L are the ln ( 4h / d )

wire diameter and length, h (285 nm) is the gate oxide thickness and = 3.9 0 is the dielectric permittivity for thermally-grown SiO2.

97 We propose that the QD is located at the HJ, based on several considerations. First, the size scale of the QD (10 nm)is much smaller than the electrode spacing (15 m), and does not match any other physical length scale in the experiment. Second, we observe QD behavior only in the section containing the HJ, and moreover have never observed this type of behavior in any single-chirality sample prepared in a similar manner. Defects such as vacancies, bond rotations, and sidewall functionalization can act as tunnel barriers[21], and pairs of defects can produce similar QD behavior[22, 23]. However, the high conductance and uniform behavior of the M and S sections, as well as the clean charging spectrum of the QD, indicate a very low density of such defects. Third, such defects typically act as resonant scatterers over a finite energy window, whereas the barriers and multiple weakly-coupled electronic levels in the HJ device are evident over a wide range of gate voltages. For these reasons, it is highly unlikely that defects unrelated to the HJ are responsible for the QD behavior.

5.6 Theoretical Modeling

To understand what atomic-scale features might give rise to the observed behavior, we developed several detailed atomistic models of the HJ, and compared their calculated electronic structure to the level spacing and coupling parameters extracted from the experiment. In principle, using just two defects, any two SWCNTs of differing

chiralities can be joined[24] to form a HJ while maintaining the topological integrity of the carbon network.

98 a

Figure 5.8: Single HJ structures a) Interface geometries and b) corresponding energy minimized structures of the HJ.

However, (21,9)-(18,13) HJs formed in this way introduce a bend of about 20 degrees, whereas the experiments observed a linear HJ. To construct linear interfaces, we developed a geometrical method to connect any two SWCNTs using a minimum, welldefined number of pentagon-heptagon (5-7) pair defects. We find that (18,13) and (21,9) CNTs can form topologically linear HJs using just four or nine 5-7 defects and that the HJ is most linear and also lowest in energy if the 5-7 defects are distributed uniformly along the circumference of the interface (Fig. 5.8). The calculated local density of states for these HJ structures reveals localized states at the interface, much more so for ninedefect HJs (Fig. 5.9), with the degree of spatial confinement being sensitive to their arrangement. However, these new interface states are strongly resonant to the electronic states in the (21,9) nanotube and broadened considerably compared with their energy

99 spacing, in contrast to the experimental observation that m , s << E . Evidently QDs cannot be formed from HJs consisting of a single interface.

Figure 5.9: Calculated local density of states for different geometries. Spatiallyresolved local density of states (LDOS), averaged over rings 0.5 A wide, plotted as a function of energy along the nanotube axis for HJs with (a) a ring of nine 5-7 defects uniformly distributed over the entire circumference; (b) the nine 5-7defects nonuniformly distributed over a HJ interfacial section 1 nm wide, and (c) a single pentagon and heptagon at the interface. White suggests a high LDOS. The Fermi level EF is at zero. The gap of the (18,13) CNT is evident from the dark regions on the left that bracket EF; the (21,9) tube is metallic with a finite LDOS everywhere on the right, its small curvature-induced gap neglected within the tight-binding approximation. As can be seen in panel (a) A junction with uniform circumferential defect distribution results in three localized states with spacing of about 200 meV and level broadening on the order of 10 meV. Absence of LDOS peaks in (c) reflects our finding that several 5-7 defects are required to localize electronic states at the interface.

100 A tube segment of a third chirality introduced between the (18,13) and (21,9) nanotubes could naturally confine electronic states at the HJ. In fact, we find that the atomistic model that best explains the experiment is a composite HJ consisting of a roughly 10 nm long metallic tube between (18,13) and (21,9) SWCNTs. This composite interface can arise from a nine-defect HJ if some of the nine 5-7 defects migrate (Fig. 5.10) towards (21,9) and the rest towards (18,13). Previous simulations at high temperatures indicate that Stone-Wales defects can dissociate into two 5-7 defects migrating in opposite directions[25], resulting in a new chirality between them; calculations have also predicted[26] spatial electronic confinement and sharp resonances by separating two 5-7 defects in a carbon nanotube of a single chirality. Composite HJ structures have also previously been observed with scanning tunneling microscopy[27].

Figure 5.10: Evolution of the (15,15) SWCNT at the interface due to migration of defects.

101 The only metallic chiralities that can link (18,13) and (21,9) without introducing more than nine defects are (15,15), (18,12) and (17,14).

Figure 5.11: (a) Interface projected DOS with a) (15,15) SWCNT and (b) (27,0) in the middle. (d) and (e) show similar LDOS contour plots for the (18,13)-(15,15)-(21,9) and (18,13)-(27,0)-(21,9) double heterojunctions (DHJs), respectively. States confined

within the 10 nm (15,15) and (27,0) interior SWCNT segments exhibit level broadenings of about 1 meV, in agreement with experiment and significantly less than the level spacing, and the LDOS between two successive states is also zero. The ladder of states confined within the 10 nm (27,0) SWCNT is due to the two interfaces on either side of the dot which are almost entirely covered by 5-7 defects. Introduction of a (15,15) segment results in several new confined states with relatively uniform spacing and weak electronic coupling ( m + s 1 10 meV) to itinerant states in the (21,9) tube, in good agreement with experiment (Fig. 5.11a and b). However, allowing more 5-7 defects (at the expense of increasing the interface formation energy) can lead to additional possibilities for intermediate metallic chiralities. Among such

102 possibilities, a (27,0) segment allows the maximum possible number of 5-7 defects at the two interfaces covering the entire circumference. As is shown in Fig. 5.11c and d, the (27,0) intermediate chirality greatly enhances the degree of spatial confinement and level spacing uniformity over the (15,15) case, and provides the best explanation for the experimental data, with predicted level spacing of ~100 meV, and electronic coupling ~1 meV.

5.7 Explanation for Large Central Diamond and Diode-Like Behavior

Vg=-(b + c)/ a

Vg=(a- c+ U)/ a

a b

a b

(a+b+U)/a
10 Conductance(S) 1 100m 10m 1m -8 -6 -4 -2 0 2 4 6 Gate Voltage (Volts) 8

Figure 5.12: Band diagrams explaining that the central gap (rectification region) is equal to the band gap of the semiconducting nanotube plus the charging energy of the first electron in the quantum dot.

103 The existence of the QD explains the ambipolar rectification behavior of the HJ device, which leads to the large asymmetric central diamond in the stability diagram (Fig. 5.3b inset). The height (in bias voltage) of the diamond (~600 meV) is consistent with the sum of the S band gap (predicted to be 410 meV for an (18,13) SWCNT[28]) and the QD addition energy E add for first hole addition (~190 meV - note that this value is higher than E add =135 meV for electron addition at high negative voltages, this is due to high SC tube resistance near the gap), both of which must be overcome to permit conduction through the device, this is explained through band diagrams in figure 5.12. The rectification is a natural result of asymmetric contacts to the semiconducting tube[29, 30]. At the Pd contact, pinning of s-NT bands results in a voltage-tunable barrier through which electrons (or holes) must tunnel[31]. At the other S-NT contact, the existence of the HJ-QD implies that the bands of the S-NT and the QD are decoupled, so that there is no (or at least much smaller) barrier due to band bending at the S-NT-HJ interface. Finally, at gate voltages outside the central diamond, the Fermi level is outside the semiconducting gap, so the S-NT acts as a simple metallic contact to the QD. In this region, the large density of states near the semiconducting band edge explains the observed stronger coupling of the S-NT to the QD.

5.8 Negative Differential Resistance (NDR)

Figure 5.13 explains negative differential resistance through the help of simple band diagrams. The presence of Negative Differential Resistance (NDR) is probably due to the blocking of current through ground state when an electron tunnels into a loosely coupled excited state. NDR can also exist due to voltage tunable barriers which cause the barrier height to change with applied voltage, however in this case since NDR peaks sometimes

104 occur one after other at different bias voltages, the former reason seems to hold more ground.

V1

V2

V=-2.3 Volts 20.0n Current (nA)

V1<V<V2

10.0n

V>V2

V2

50

100 Bias (mV)

150

200

Figure 5.13: Explanation of the negative differential resistance (NDR)

5.9 Conclusion
We can conclude that in order to form good molecular scale quantum dots from carbon nanotube HJs, a short segment of intermediate metallic chirality is necessary, and that both interfaces should be maximally linear to ensure enough 5-7 defects at each interface to cover the entire circumference. Our result demonstrates the potential of atomically defined interfaces between carbon nanotubes as model systems for molecular scale electronic devices[32, 33]. It also opens up the possibility of creating robust, room

105 temperature single electron transistors through controlled CVD growth of nanotube HJs[34].

5.10 Acknowledgements
We thank Mark Hybertsen and Marc Bockrath for valuable discussions. We also thank Sami Rosenblatt, Matthew Sfeir and Christophe Voisin for assistance in AFM and Rayleigh spectroscopy. We acknowledge support from the National Science Foundation under award CHE-0117752, and from Intel Corporation. Work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. Computational resources required for this work were partially provided by NERSC. We also acknowledge the Network for Computational Nanotechnology NCN) for support.

106

Bibliography
1. Chico, L., et al., Pure carbon nanoscale devices: Nanotube heterojunctions. Physical Review Letters, 1996. 76(6): p. 971-974. 2. Kim, H., et al., Direct observation of localized defect states in semiconductor nanotube junctions. Physical Review Letters, 2003. 90(21). 3. Ouyang, M., et al., Atomically resolved single-walled carbon nanotube intramolecular junctions. Science, 2001. 291(5501): p. 97-100. 4. Anderson, N., A. Hartschuh, and L. Novotny, Chirality Changes in Carbon Nanotubes Studied with Near-Field Raman Spectroscopy. Nano Lett., 2007. 7(3): p. 577-582. 5. Doorn, S.K., et al., Raman spectral imaging of a carbon nanotube intramolecular junction. Physical Review Letters, 2005. 94(1). 6. Yao, Z., et al., Carbon nanotube intramolecular junctions. Nature, 1999. 402(6759): p. 273-276. 7. Huang, L.M., et al., Long and oriented single-walled carbon nanotubes grown by ethanol chemical vapor deposition. J. Phys. Chem. B, 2004. 108: p. 16451. 8. Sfeir, M.Y., et al., Probing electronic transitions in individual carbon nanotubes by Rayleigh scattering. Science, 2004. 306(5701): p. 1540-1543. 9. Sfeir, M.Y., et al., Optical spectroscopy of individual single-walled carbon nanotubes of defined chiral structure. Science, 2006. 312(5773): p. 554-556. 10. Huang, X.M.H., et al., Controlled placement of individual carbon nanotubes. Nano Letters, 2005. 5(7): p. 1515-1518. 11. Ouyang, M., et al., Energy gaps in "metallic" single-walled carbon nanotubes. Science, 2001. 292(5517): p. 702-705.

107 12. Purewal, M.S., et al., Scaling of resistance and electron mean free path of singlewalled carbon nanotubes. Physical Review Letters, 2007. 98(18). 13. Chen, Z.H., et al., The role of metal-nanotube contact in the performance of carbon nanotube field-effect transistors. Nano Letters, 2005. 5(7): p. 1497-1502. 14. Javey, A., et al., Ballistic carbon nanotube field-effect transistors. Nature, 2003. 424(6949): p. 654-657. 15. Kouwenhoven, L.P., D.G. Austing, and S. Tarucha, Few-electron quantum dots. Reports on Progress in Physics, 2001. 64: p. 701-736. 16. Beenakker, C.W.J., Theory of Coulomb-Blockade Oscillations in the Conductance of a Quantum Dot. Physical Review B, 1991. 44(4): p. 1646-1656. 17. Weis, J., et al., Competing channels in single-electron tunneling through a quantum dot. Physical Review Letters, 1993. 71(24): p. 4019. 18. Postma, H.W.C., et al., Electrical transport through carbon nanotube junctions created by mechanical manipulation. Physical Review B, 2000. 62(16): p. 1065310656. 19. Deshmukh, M.M., et al., Equilibrium and nonequilibrium electron tunneling via discrete quantum states. Physical Review B, 2002. 65(7). 20. Bonet, E., M.M. Deshmukh, and D.C. Ralph, Solving rate equations for electron tunneling via discrete quantum states. Physical Review B, 2002. 65(4). 21. Fan, Y.W., B.R. Goldsmith, and P.G. Collins, Identifying and counting point defects in carbon nanotubes. Nature Materials, 2005. 4(12): p. 906-911. 22. Bockrath, M., et al., Resonant electron scattering by defects in single-walled carbon nanotubes. Science, 2001. 291(5502): p. 283-285.

108 23. Maltezopoulos, T., et al., Direct observation of confined states in metallic singlewalled carbon nanotubes. Applied Physics Letters, 2003. 83(5): p. 1011-1013. 24. Melchor, S. and J.A. Dobado, CoNTub: An algorithm for connecting two arbitrary carbon nanotubes. Journal of Chemical Information and Computer Sciences, 2004. 44(5): p. 1639-1646. 25. Nardelli, M.B., B.I. Yakobson, and J. Bernholc, Brittle and Ductile Behavior in Carbon Nanotubes. Physical Review Letters, 1998. 81(21): p. 4656. 26. Chico, L., M.P. Lpez Sancho, and M.C. Muoz, Carbon-Nanotube-Based Quantum Dot. Physical Review Letters, 1998. 81(6): p. 1278. 27. Ouyang, M., J.L. Huang, and C.M. Lieber, Fundamental electronic properties and applications of single-walled carbon nanotubes. Accounts of Chemical Research, 2002. 35(12): p. 1018-1025. 28. Saito, R., G. Dresselhaus, and M.S. Dresselhaus, Physical Properties of Carbon Nanotubes. 1998: Imperial College Press. 29. Yang, M.H., et al., Carbon nanotube Schottky diode and directionally dependent field-effect transistor using asymmetrical contacts. Applied Physics Letters, 2005. 87(25). 30. Freitag, M., et al., Controlled creation of a carbon nanotube diode by a scanned gate. Applied Physics Letters, 2001. 79(20): p. 3326-3328. 31. Heinze, S., et al., Carbon nanotubes as Schottky barrier transistors. Physical Review Letters, 2002. 89(10). 32. Guo, X.F., et al., Covalently bridging gaps in single-walled carbon nanotubes with conducting molecules. Science, 2006. 311(5759): p. 356-359.

109 33. Guo, X.F., et al., Conductivity of a single DNA duplex bridging a carbon nanotube gap. Nature Nanotechnology, 2008. 3(3): p. 163-167. 34. Yao, Y., et al., Temperature-mediated growth of single-walled carbon-nanotube intramolecular junctions. Nat Mater, 2007. 6(4): p. 293-296.

110

6 ______________________________
CHAPTER

Electron-Phonon Scattering in Known Chirality Single Wall Carbon Nanotubes ______________________________

111

6.1 Introduction
The intrinsic resistivity of carbon nanotube comes from many electron scattering processes, such as electron-phonon (lattice vibrations) interactions[1], lattice defects (such as broken bonds, ad atoms, structural change, etc.)[2, 3], electron-substrate interactions (substrate optical phonons, potential wells, charge scatters)[4], as well as electron-electron interactions[5]. Each of these scattering processes displays different temperature and bias dependence. At moderate to high temperatures, the resistivity in nanotubes mainly comes from electronphonon scattering processes. Understanding of electron-phonon scattering length in metallic carbon nanotubes is important for their potential use in electronic circuits for e.g. as lower level interconnects in microprocessors[6]. This chapter addresses significant similarities and dissonances between the existing theoretical calculations and latest experimental results on electron-phonon scattering. The first ever resistivity measurements on known chirality nanotubes provide a reliable data to compare with the theoretical estimates of electron-phonon scattering. Finally, it is found that the surface optical phonon modes of the substrate play an important role in electron scattering. Their contribution to scattering significantly affects nanotubes resistivity at moderate to high temperatures. This finding takes nanotube transport physics into a completely new direction where studying the effect of surrounding substrate becomes vital in understanding nanotube electronic response.

112

6.2 Electron-Phonon Scattering

Phonons are lattice vibrations whose occupation depends on the tube temperature and the energy of the phonon mode[7]. Phonons create a deformation potential through the movement of the lattice atoms, which scatters the electrons and causes resistance for current flow[8]. Phonon modes relevant to this study are divided in two parts:

1. Acoustic Phonons: At moderate temperatures and low bias voltages, nanotubes mainly possess acoustic phonon modes. These modes have velocities ( v s ) very less than the Fermi velocity ( v F ) of electrons, and therefore scatter electrons elastically. This process is shown Fig. 3a. The scattering rate increases linearly with temperature as the phonon occupation increases. Since these scattering events happen near the K-point, mainly low momentum point phonons participate in the process.

2. Optical and Zone boundary Phonons: These modes do not exist at room temperature (except for radial breathing mode for nanotube).However they can be excited through a high energy electron using high source drain bias. These modes can decrease the current flow considerably, due inelastic scattering (Fig. 3b). The saturation current at high bias in a nanotube is supposed to be mainly from scattering due to these modes[9]. Zone boundary phonons are also optical phonons; however they can scatter electrons across the zone boundary (Fig. 3c). These are mainly high momentum K-point phonons.

113

Figure 6.1: Emission and absorption phonons near K point. a) Elastic scattering through acoustic phonon, b) Emission of an optical phonon, c) Zone-boundary phonon

Electron-phonon interaction in carbon nanotubes has been a busy area of research in past few years. With the advent of improved carbon nanotube growth processes, ultra-long nanotubes are now available for mean free path calculations[10].

6.3 Electron-Acoustic Phonon Scattering- Theoretical Background

Scattering due to acoustic phonons are very weak in carbon nanotubes (especially metallic ones). This is due to the presence of orthogonal bands which do not allow scattering due to a slowly varying potential created by acoustic phonons. Several groups have studied the electron-acoustic phonon scattering rates in nanotubes through theoretical calculations. This approach broadly contains the determination of electronphonon coupling parameters (or in other words, the deformation potential that electrons find due to phonons) and the use of Fermis golden rule in calculating the scattering rates. The seminal work in this area was done by Ando et al.[8] They calculated the deformation potential through studying atomic displacements for various phonon modes. It was found that the scattering rate for acoustic phonons increases linearly with

114 temperature. Also it is dependent mainly on nanotube diameter above the breathing mode temperature which is close to 150 K for nanotube diameter ~2 nm. The chirality dependence of scattering gets noticed at lower temperature values. Theoretical calculations suggest that for armchair nanotubes shear and twist waves contribute to electron scattering, whereas for zigzag tubes main contribution is from stretching and breathing modes. At temperature higher than the breathing mode temperature, resistivity of zigzag nanotubes becomes same as armchair nanotubes hence resistivity becomes independent of tube chirality. McEuen et al. [11] proposed the scattering rates through a simple formula

ac

=2

2 2 k BT 1 2 h 2 vs hvF

h q << kT

(6.2)

Here vs = q / q is the acoustic phonon velocity and is the deformation potential[11]. This is produced by a long wavelength acoustic phonon and is proportional to local contraction and dilation of tube. Here v F is the Fermi velocity which is constant for metallic tubes near the K point and is the nanotube tube mass density. Using this expression they calculated a scattering rate of 1ac = 3x1011 s-1 for a 1.8 nm diameter nanotube at room temperature (300K). Perebeinos et al.[12], have calculated the

scattering rate through tight binding calculations for semiconducting carbon nanotubes. Their value for scattering rate near the conduction band edge is four times of the previous work ( 1ac = 12x1011 s-1). However, they have included scattering from all the phonon modes at room temperature including Radial Breathing Mode (RBM). Pennington et

115

al.[13] used the deformation potential approximation along with the first order perturbation theory to obtain a scattering rate of 1011s-1 at room temperature. This is much lower than both of the previous estimates. A recent publication by Perebeinos et al.[14] has also speculated on the possibility of electron scattering through substrate polar phonon modes. This work suggested that scattering in nanotubes, even at low temperatures, can be dominated by optical phonon modes that arise from the surface of the silicon dioxide substrate.

6.4 Electron-Acoustic Phonon Scattering-Previous Experiments

Two probe resistance of a nanotube connected with metal electrodes can be written as:

R Total = R C + R Q + L

(6.1)

Here RC is the contact resistance between nanotube and carbon metal electrode and RQ is the quantum mechanically fixed resistance value of 6.5 k [15]. Resistivity ( ) of a nanotube is a constant in diffusive limit, therefore resistance offered by a nanotube section is resistivity multiplied by the length of the channel. In order to separate the contact effect (bad contacts, schottky barriers, etc.) from the transport properties of bulk of nanotube it is important to study the tube at various lengths. This is because the contact effects will be same in all the sections and the only change will come from the bulk. Devices of varying lengths are fabricated on a known chirality nanotube through the process[16] discussed in chapter 4 of this thesis. The substrate on which tube sits is degenerately doped silicon with an epi-layer of silicon oxide (285 nm). This arrangement enables to perform a three terminal measurement of the nanotube. Room temperature measurements are performed in a probe station with a semiconductor

116

parameter analyzer. The low temperature measurements are done in a liquid helium cryostat. The nanotube resistivity measurement through depositing multiple electrodes with varying distances has been used in a previous collaborative work with Meninder et al.[17] on resistivity measurements of CVD-grown single wall carbon nanotubes. In this work resistivity measurements on a number of as-grown CVD tubes were performed. A room temperature resistivity of 8.2 k/m was obtained for the best samples. Majority of metal tubes were between 10 to 12 k/m. Detailed temperature dependent study showed the decrease of tube resistivity with temperature, however the trend was not very linear at higher temperature values (T>200K). Some of the tubes were highly resistive; it was suggested that static scatters are responsible for this increase in resistance. Another approach for resistivity measurements on nanotubes is through using a conducting AFM tip as a traversing electrode. A first comprehensive resistivity measurements using this approach were done by Park et al.[11]. Room temperature resistivity of 12 k/m was reported at room temperature for a 1.8 nm diameter nanotube. This value is among the lowest value reported in open literature. A recent work by Sundqvist et al.[18] using the conducting AFM tip electrodes found that nearly all of the nanotubes showed a low bias resistivity of 102 k/m at room temperature. The resistivity values were found to be independent of the nanotube diameter. It is evident from the literature that there is still a lack on consensus on the exact resistivities of nanotubes and no dependence has been seen on its diameter and chirality, which is contrary to what theory suggests. One reason for this is the lack of resistivity studies on known structure nanotubes. Since AFM and other techniques have a good

117

margin of error in measuring diameters, it was not possible to relate results to any known diameter nanotube. Other major drawback was the inability to know whether a nanotube is single and not multiwall or a few tube bundles. It is shown in chapter 2 on nanotube synthesis that most of the tubes obtained through CVD are bundles of several single wall nanotubes. This suggests a greater need for measurements on known structure nanotubes. The coming sections will show results from such a study.

6.5 Measurement Method

Three terminal measurements are done using metal electrodes as voltage source/drain and silicon as the back gate. Gate voltage changes the tube potential that increases or decreases the tube Fermi level with respect to the vacuum. A semiconducting nanotube starts conducting when the tube Fermi level is pulled to conduction or valence band with the help of back gate[19]. The back gate voltage is applied in a range of -10 to +10V. For low bias measurements, source voltage Vsd is kept below 10 mV for metallic tubes. As an example, Fig. 2 shows the low bias conductance variation of a (21,9) chiral metal nanotube for two different lengths at room temperature. The ON conductance of the tube is defined as the maximum value of conductance in such a plot. Conductance for (21,9) metal nanotube is 190 k for a 18.5 m channel length. In figure 6.2, it can be seen that conductance changes with the length of the nanotube channel. For multiple nanotube channels of same length on the tube, the conductance remains constant. Similar conductance sweeps are done at varying temperatures to study temperature dependent properties of nanotubes such as the change in tube conductance with increase or decrease of phonon population.

118

100 800 nm G (S) 10 18.5 m

1 -10 -5 0 VG (V) 5 10

Figure 6.2: Conductance gate sweep for (21,9) nanotube. The red and blue curves

represent variation of conductance with gate voltage for 800 nm and 18.5 m nanotube channel length.

6.6 Electrical Transport Measurements T=300K

As described in the previous chapter, the transport measurements on the known chirality nanotubes are achieved through a series of processing steps - suspended tube growth over slits on silicon wafer, tube characterization through Rayleigh spectroscopy, tube transfer over silicon oxide substrate and finally the device fabrication. This section describes the final part consisting of measurements. Figure 6.3 shows conductance measurement results on a (26, 11) chiral metal nanotube. Conductance of a 7m tube section is

119

measured at varying gate voltage. The small conductance dip at the center comes from the band gap in chiral metal nanotubes. The conductance at the extreme positive side of the gate voltage is a bit lower than negative side due to higher schottky barrier heights from the n-channel transport (please refer to section for more nanotube band diagrams).

25.0 20.0 15.0 G (S) 10.0 5.0 0.0 -10 -5 0 VG (V) 5 10

Figure 6.3: Conductance gate sweep for (26,11) nanotube. Taken at T=220K, the

sweep shows characteristic transport through a p-doped small band gap metallic nanotube.

Similar measurements are done at different lengths of the nanotube and the ON conductance (or resistance) is extracted. A comprehensive result of resistance change for with length (26,11) nanotube is shown in Fig. 6.4. It can be seen that at very low channel

120

lengths the resistance approaches a constant. This suggests that the mean free path of electrons has become less than the channel length and hence the transport becomes ballistic. Here, no scattering is offered from the tube and the only resistance left is RQ+RC or the contact resistance. For contacts with good tunnel probability, this value is independent of temperature. In the case of (26,11) tube this is 10 k, very close the quantum resistance limit of 6.5 k. This suggests that the contacts are highly transparent. Such low contact resistance devices are necessary for studying the resistivity variations in the nanotubes.

low bias = 9.4 K/m

100k R() 10k 1 L(m) 10

Figure 6.4: Resistance vs. channel length for (26,11) nanotube at T=300K. At long

nanotube length segments resistance varies linearly with length, suggesting diffusive transport. Signs of resistance flattening out can be seen below 600 nm length.

121

At longer lengths, the resistance becomes linear with temperature- a clear signature of the diffusive transport in a metal. Here, the slope of the line is measured to extract the resistivity of the tube. For (26,11) tube this value is 9.4 K/m at T=300K. Similar measurements were done at T=300K for different tube chiralities. A summary of their room temperature resistivities is shown in table 6.1.

Nanotube (n,m)

(18,18) 7.9 k/m

(26,11) 9.4 k/m

(21,9) 10 k/m

(24,12) 17 k/m

(16,16) 18 k/m

Table 6.1: Resistivity values for chirality assigned nanotubes at T=300K

The resistivity value ranges from 8-18 k/m with no visible dependence on the tube diameter or chirality. Moreover, our result of a narrow resistivity distribution at room temperature matches well with the experimental results of Sundqvist et al.[18] where they found nearly the same resistivities for different diameter nanotubes at room temperature. To study the origin of resistivity in detail, measurements are performed at different temperatures. This is discussed in the next section.

122

6.7 Electrical Transport Measurements with Varying T

1E-4
77K

1E-5 G (S)

220K

1E-6

-9

-6

-3

0 3 VG(V)

Figure 6.5: Conductance gate sweep for (26,11) nanotube with changing temperatures. As the temperature is decreased from 220K to 77K the conductance of the

tube goes up.

Figure 6.5 shows the variation of conductance in the 7 m (26,11) nanotube device with changing temperature. It can be seen that the ON conductance of the nanotube increases with a decrease in temperature. This is due to the decrease in the population of phonon modes. The conductance in the gap increases with temperature increase. Although metallic, these nanotubes do possess a small gap due to the tube curvature. The value of this gap can be extracted from the temperature dependence of the gap conductance

123
E as G ~ e g K T . This gives an energy gap of around 7 meV which is very close to the B

curvature induced gap prediction for a (26,11) nanotube from tight binding (Eg~12 meV).

180.0k 150.0k R() 120.0k 90.0k 60.0k 30.0k 5

100 K 120 K 140 K 160 K 180 K 200 K 220 K 240 K 260 K 280 K 300 K

10

15 L(m)

20

Figure 6.6: Resistance vs. channel length for (26,11) nanotube at different temperatures. Since the channel lengths are greater than 5 m the transport is still in the

diffusive regime at T=100 K.

Figure 6.6 shows the distribution of ON resistance with length at various temperatures. The perfectly linear dependence of resistance on tube length suggests that the transport is diffusive throughout the temperature range. Therefore a direct slope measurement of these points conveys the resistivity of the nanotube at that temperature. Figure 6.7 shows the variation of resistivity with temperature. At lowest temperatures the resistivity reaches to ~2 k/m.

124

Resistivity is related to mean free path through the relation.

h 1 4e 2 l mfp

(6.3)

Therefore this corresponds to a mean free path of 3.25 m.

25.0k 20.0k
(/m)

15.0k 10.0k 5.0k 0 100 200 300 400 500

T (K)
Figure 6.7: Resistivity vs. temperature for (26,11) nanotube. It can be seen that the

resistivity deviates from linearity near T=100K.

Mean free path at low temperatures is slightly less than the best values found by Meninder et.al. (~5 m) [17]. However, it is very close to the average value obtained by them for different tubes. Measurements on other known chirality nanotubes show

similar resistivity dependence on temperature.

125

6.8 Analysis of Results

Since most of the theoretical estimates available relates to room temperature scattering rates, a comparison is done to check on these theories with the known chirality tube data presented above. 1. All the nanotubes studied possess low defect density, as observed through Rayleigh spectrum and verified through transport measurements at low temperatures. Contact resistances are also very low, suggesting electron transparent contacts 2. The mean free path at room temperature is close to 0.8 to 0.65 m for most of the nanotubes which matches well with previous results of Meninder et al.[17] and Sundqvist et al.[18]. It corresponds to a scattering rate of ~1012 sec-1 at T=300K. This value of scattering rate is much higher than what suggested by McEuen et al.[11] Although it comes very close to the theoretical estimates of Perebeinos et al.[12] for semiconducting nanotubes. Since SC tubes have a very high density of states at the conduction band edge and more scattering than a metallic nanotube, the metallic tubes should show even lower resistivity values. 3. Temperature dependence of resistivity is not completely linear (Fig. 6.7), especially at temperatures above 100K. This has been reported by Meninder et al. Since the scattering at low bias is mainly dominated by well thermalized acoustic phonon modes, this should not be the case. Clearly, the resistivity results do not show signatures of a dominating electron-acoustic phonon scattering mechanism. The non-linear temperature dependence of resistivity points towards a possibility of an additional phonon scattering mechanism. The scattering

126

through nanotube RBM phonons can also cause such dependence. However, RBM modes are not known to couple very strongly to the electrons[20]. Furthermore the total resistivity in their presence should still be linear above T~2TB. i.e. ~2 nm diameter tubes
B

should all show linear R vs. T above room temperature, which is not the case here. One possible reason for a super linear resistivity could be scattering due to substrate polarized phonons as predicted by Perebeinos et al.[14] in a recent work. Experimental work in this direction is explained in the next sections.

6.9 Substrate Polarized Phonons (SPP)

Because nanotubes are typically incorporated into devices on substrates, it is important to take into account scattering by substrate phonons. In fact, a recent work has revealed that substrate optical phonons provide the dominant scattering mechanism in graphene on SiO2 at room temperature [21]. It has further been demonstrated that the removal of the substrate in graphene substantially increases the mean free path and results in linear (T) above ~100K [22]. In addition, it has been demonstrated that emission of substrate optical phonons of the SiO2 substrate constitutes the dominant mechanism for velocity saturation in graphene field effect transistors [23]. The experimental observations from graphene strongly hint towards the substrate phonon effects in carbon nanotubes. Recent theoretical work by Perebeinos [14] has also predicted the possibility of substrate phonons interaction with nanotubes.

127

The measurements on known chirality tubes in previous sections are done over silicon dioxide substrate, which has five surface optical phonon modes hwSO ,v =1, 2....., 5 = 50, 62, 100, 149, 149 meV[24]. The coupling strength of these modes is defined by a parameter called Frohlich coupling given by the relation[14]:

Fv =

hwSO ,v 2

1 1 + 1 + 1 0

(6.4)

Here 0 and are the low and high frequency dielectric constants of silicon oxide. Using this relation, couplings for various SO modes for oxide can be obtained as

F 2 v=1, 2....,5 =0.042, 0.38, 0.069, 1.08, 1.08 meV.

These coupling parameters are included in the usual relation for acoustic phonon resistivity in carbon nanotubes. This is summarized in the following relation:

Total = static + acoustic T + SPP (n B (hw2 ) + n B (hw4 ,5 ))

(6.5)

In equation 6.5, static describes the static scattering due to defects. The linear temperature dependence of acoustic phonons of carbon nanotube is included in the second term, here acoustic is the electron-acoustic phonon coupling parameter.The third term describes the scattering due to SO phonons from silicon oxide. Here, SPP

128

represents the electron- SPP coupling and nB is the phonon occupation number for Silicon oxide, given by:

nB (hwv ) =

1 (e
hwv k BT

(6.6)

1)

And is the ratio of the electron-phonon coupling strengths fixed by the ratio of the Frohlich couplings; for silicon dioxide it is:

( F4 + F5 ) =5.6 2 F2

(6.7)

Using the above values, equation 6.5 is fitted to the data for the (26, 11) chiral metal carbon nanotube. This is shown in Fig. 6.8. For this fitting, acoustic = 0.022 and SPP = 27 are used as the free parameters. The red line shows the curve fit with SPP scattering included, while the green line shows the acoustic phonon scattering dependence only with acoustic = 0.022 . This value falls in between theoretical prediction done by reference [11, 12] at 300K.

129

25.0k 20.0k
(/m)

15.0k 10.0k 5.0k 0 100 200 300 T(K) 400 500

Figure 6.8: Comparison of experimental and theoretical resistivity values for (26,11) nanotube. Red curve represents the theoretical fit taking into account the SPP scattering.

Green line represents the acoustic phonon scattering.

Similar measurements and curve fittings done on (21,15) chiral metal nanotube (Fig. 6.9a) show very close value for SPP =23 as the (26,11) nanotube. However acoustic = 0.0025 is far lower than (26,11). This discrepancy might be due to less number of length section studied for (21,15) nanotube which makes resistivity measurements at lower temperature

130

bit erroneous. Measurements on (24,12) showed (Fig. 6.9b) the superlinear temperature dependence with acoustic = 0.018 and SPP = 40.

12.50k 10.00k
(/m)

a (21,15)
18.0k 16.0k
(/m)

b (24,12)

14.0k 12.0k 10.0k 8.0k

7.50k 5.00k 2.50k 100 200 300 T(K) 400 500

100

200 T(K)

300

Figure 6.9: Comparison of experimental and theoretical resistivity values for a) (21,15) and b) (24,12) nanotube. Red curves depict the theoretical fittings with SPP

scattering taken into consideration.

In order to ascertain the effects of the substrate, experiments were planned to measure the same nanotube on different substrate stripes pattern on a wafer. Sample device geometry is shown in Fig. 6.10.

Here, Aluminum Nitride and Hafnium oxide stripes are made over silicon oxide wafer. Since aluminum nitride has a high energy SPP mode (83.60 and 104.96 meV)[25] it is expected to give less SPP scattering than Hafnium oxide (19.42, 52.87).

131

HfO2 AlN

Figure 6.10: Optical image of nanotube device with different substrate stripes.

Hafnium stripes are patterned through atomic layer deposition (ALD) over aluminum nitride (AlN) substrate. AlN is deposited through plasma ALD process.

Figure 6.11 shows the results of transport measurement on an unknown chirality semiconducting nanotube (probably a bundle) on the AlN-HfO2 striped substrate. It is clear that the resistivity values diverge at T~125K, which is very close to the temperature value at which any SPP scattering will take effect. Below T=100K, SPP scattering is not present due to Pauli blocking[14].

132

22.0 20.0 18.0 16.0 14.0 12.0 10.0 8.0 6.0 4.0 100 200

HfO2

(/m)

AlN

300 T(K)

400

Figure 6.11: Comparison of experimental and theoretical resistivity values for semiconducting nanotube. Red curve (dots) represents theoretical (experimental)

resistivity values nanotube resistivity on Hafnium oxide. Blue curve (dots) represents theoretical (experimental) resistivity values nanotube resistivity on Aluminum nitride.

The different variation of resistivity on AlN and HfO2 further strengthens the SPP scattering hypothesis. The data fits perfectly to the AlN substrate with its two lower most SPP modes (83.6 meV and 104.96 meV). Interestingly, the value of SPP = 45 is nearly twice for the single metallic nanotube. Curve fits also show the acoustic scattering parameter acoustic =0.005. The reason for such a low value may lie in the bundled nature

133

of the nanotube. For HfO2, the fit is not good with the same acoustic and SPP parameters as used for AlN. This may be due to different roughness of the HfO2 film as compared to ALN which changes the average tube distance from substrate. This has been theoretically shown to have significant effect on the SPP scattering rates [14]. In addition to surface roughness, film quality may also affect the values of the SPP couplings due to a change in the film dielectric constant.

6.10 Summary
This work confirms the existence of nanotube electron scattering through substrate polarized phonon modes. Transport measurements on known chirality nanotubes as well as tube bundles suggest a super linear change in tube resistivity with increase in temperature. Resistivity equations, when combined with SPP scattering parameters provided good fit for the data observed. The fitting parameters and tube resistivities are summarized in table 6.1.
(n,m) Substrate
(k/m) at 300K

(26,11) SiO2

(24,12) SiO2

(21,15) SiO2

SC Bundle HfO2 AlN

9.4 1.4 0.022 27

17.0 6.0 0.018 40

7.0 3.5 0.0025 23

8.0 3.0 0.005 45

9.0 3.5 0.005 45

static (k/m) acoustic (k/m-K) SPP (k/m)

Table 6.2: Parameters for curve fitting data with SPP scattering

134

It is possible to estimate the true electron-acoustic phonon scattering rate ( 1acoustic ) using the values of acoustic .For (26,11) nanotube, 1acoustic = 8x1011 sec-1 while for (24,12) nanotube 1acoustic = 6.5x1011 sec-1 is obtained. These values match closely to the existing theoretical results. Also, they correspond to a room temperature mean free path of ~1 m; 50% more than the values obtained experimentally. More work may be needed to ascertain the effect of SPP scattering on other substrates such as SiC where the scattering is predicted to be less[14]. Using these results, in future, it will be possible to study the true nanotube acoustic phonon scattering and its dependence on tube structure. This work also opens up possibilities for engineering the right substrate for obtaining minimum tube resistivity.

135

Bibliography
1. Dresselhaus, M.S.E., P.C., Phonons in Carbon Nanotubes. Advances in Physics, 2000. 49(6): p. 705-814. 2. Biel, B., et al., Anderson Localization in Carbon Nanotubes: Defect Density and Temperature Effects. Physical Review Letters, 2005. 95(26): p. 266801. 3. Park, J.-Y., Electrically tunable defects in metallic single-walled carbon nanotubes. Applied Physics Letters, 2007. 90(2): p. 023112-3. 4. Petrov, A.G. and S.V. Rotkin, Transport in nanotubes: Effect of remote impurity scattering. Physical Review B, 2004. 70(3): p. 10. 5. Deshpande, V.V., et al., Mott Insulating State in Ultraclean Carbon Nanotubes. Science, 2009. 323(5910): p. 106-110. 6. List, S., et al., A global view of interconnects. Microelectronic Engineering, 2006.
83(11-12): p. 2200-2207.

7.

Hone, J., et al., Quantized phonon spectrum of single-wall carbon nanotubes. Science, 2000. 289(5485): p. 1730-1733.

8.

Suzuura, H. and T. Ando, Phonons and electron-phonon scattering in carbon nanotubes. Physical Review B, 2002. 65(23).

9.

Yao, Z., C.L. Kane, and C. Dekker, High-Field Electrical Transport in SingleWall Carbon Nanotubes. Physical Review Letters, 2000. 84(13): p. 2941.

10.

Huang, L.M., et al., Long and oriented single-walled carbon nanotubes grown by ethanol chemical vapor deposition. J. Phys. Chem. B, 2004. 108: p. 16451.

11.

Park, J.Y., et al., Electron-Phonon Scattering in Metallic Single-Walled Carbon Nanotubes. Nano Lett., 2004. 4(3): p. 517-520.

136

12.

Perebeinos, V., J. Tersoff, and P. Avouris, Electron-phonon interaction and transport in semiconducting carbon nanotubes. Physical Review Letters, 2005.
94(8): p. 4.

13.

Pennington, G. and N. Goldsman, Semiclassical transport and phonon scattering of electrons in semiconducting carbon nanotubes. Physical Review B, 2003. 68(4): p. 11.

14.

Perebeinos, V., et al., The Effects of Substrate Phonon Mode Scattering on Transport in Carbon Nanotubes. Nano Letters, 2009. 9(1): p. 312-316.

15.

Saito, R., G. Dresselhaus, and M.S. Dresselhaus, Physical Properties of Carbon Nanotubes. 1998: Imperial College Press.

16.

Chandra, B., et al., Electrical transport measurements of nanotubes with known (n, m) indices. Physica Status Solidi B-Basic Solid State Physics, 2006. 243(13): p. 3359-3364.

17.

Meninder, S.P., et al., Scaling of Resistance and Electron Mean Free Path of Single-Walled Carbon Nanotubes. Physical Review Letters, 2007. 98(18): p. 186808.

18.

Sundqvist, P., et al., Voltage and Length-Dependent Phase Diagram of the Electronic Transport in Carbon Nanotubes. Nano Lett., 2007. 7(9): p. 2568-2573.

19.

Heinze, S., et al., Carbon nanotubes as Schottky barrier transistors. Physical Review Letters, 2002. 89(10).

20.

Verma, A., M.Z. Kauser, and P.P. Ruden, Effects of radial breathing mode phonons on charge transport in semiconducting zigzag carbon nanotubes. Applied Physics Letters, 2005. 87(12): p. 3.

137

21.

Chen, J.H., et al., Intrinsic and extrinsic performance limits of graphene devices on SiO2. Nature Nanotechnology, 2008. 3(4): p. 206-209.

22.

Bolotin, K.I., et al., Temperature-Dependent Transport in Suspended Graphene. Physical Review Letters, 2008. 101(9): p. 096802.

23.

Meric, I., et al., Current saturation in zero-bandgap, top-gated graphene fieldeffect transistors. Nat Nano, 2008. advanced online publication.

24.

Petrov, A.G. and S.V. Rotkin, Energy relaxation of hot carriers in single-wall carbon nanotubes by surface optical phonons of the substrate. Jetp Letters, 2006.
84(3): p. 156-160.

25.

Aniruddha Konar, T.F., Debdeep Jena, Effect of high-K dielectrics on charge transport in graphene. arXiv:0902.0819v1, 2009.

138 PUBLICATIONS

B .Chandra, Y. Wu, M. Purewal, M. Huang, H. Yan, L. Huang, S. Brien, T.F. Heinz, P. Kim, J. Hone, Electronic Transport across a Carbon Nanotube Heterojunction, Nano Letters, Publication Date (Web): March 11, 2009

B. Chandra, R. Caldwell, M. Huang, L. Huang, M.Y. Sfeir, S. Brien, T. F. Heinz, J. Hone, Electrical transport measurements of nanotubes with known (n, m) indices, Phys. stat. sol.(b) 243,No.13, 3359-3364 (2006)

V. Deshpande, B. Chandra, R. Caldwell, D. Novikov, J. Hone, M. Bockrath, Observation of a Mott insulating state in ultra-clean carbon nanotubes, Science, 323, 5910, 106-110 (2009)

M. Huang, Y. Wu, B. Chandra, H. Yan, Y. Shan, et al., Direct measurement of straininduced changes in the band structure of carbon nanotubes, Physical Review Letters, 100 , 136803 (2008)

S. Calabrese Barton, Y. Sun, B. Chandra, S. White, and J. Hone, Utilization and Transport in Mediated Enzyme Electrodes with Multiscale Supports, Electrochemical and Solid- State Letters, Volume 10, Issue 5, pp. B96-B100 (2007)

Purewal MS, Hong BH, Ravi A, Chandra B, Hone J, Kim P, Scaling of Resistance and Electron Mean Free Path of Single-Walled Carbon Nanotubes, Physical Review Letters 98 (18): Art. No. 186808 May 4 (2007)

Y. Wu, J. Maultzsch, E. Knoesel, B. Chandra, M. Huang, M.Y. Sfeir, L. E. Brus, J. Hone, T. F. Heinz, Variable Electron-Phonon Coupling in Isolated Metallic Carbon Nanotubes Observed by Raman Scattering, Physical Review Letters 99, 027402 (2007)

139 S. Calabrese Barton, Y. Sun, B. Chandra, S. White, and J. Hone, Utilization and Transport in Mediated Enzyme Electrodes with Multiscale Supports, ECS Transactions 3, 1341 (2006)

B. Chandra, J. Case, J. Hone, Growth of Carbon nanotubes on Carbon Paper for fuel cell applications, Proceedings of ASME energy nanotechnology conference, September 2007, Santa Clara

J. Hone, P. Kim, X.M.H. Huang, B. Chandra, R. Caldwell, J. Small, B. H. Hong, T. Someya, L. Huang, S. OBrien, C. P. Nuckolls, "Growth of Nanotubes and Chemical Sensor Applications," Proceedings of the SPIE (2005).