Springer dresselhaus m dresselhaus g avouris p (eds) carbon nanotubes synthesis structure properties and application (TAP 80 springer 2001)(ISBN 3540410864)(417s)

A review of the close connection between the structure and properties of carbon nanotubes and those of graphite and its related materials is presented in order to gain new insights into

by R E Smalley, Chemistry Nobel Lauveate 1996

Since the discovery of the fullerenes in 1985 my research group and I havehad the privilege of watching from a central location the worldwide scientificcommunity at work in one of its most creative, penetrating, self-correcting,and divergent periods of the last century In his recent book, “The Transpar-ent Society”, David Brin discusses the virtues of an open, information richsociety in which individuals are free to knowledgeably criticize each other,

to compete, to test themselves and their ideas in a free market place, andthereby help evolve a higher level of the social organism He points out thatmodern science has longfunctioned in this mode, and argues that this opencriticism and appeal to experiment has been the keystone of its success Thisnew volume, Carbon Nanotubes, is a wonderful example of this process atwork

Here you will find a summary of the current state of knowledge in thisexplosively growing field You will see a level of creativity, breadth and depth

of understandingthat I feel confident is beyond the capability of any singlehuman brain to achieve in a lifetime of thought and experiment But manyfine brains workingindependently in the open society of science have done itvery well indeed, and in a very short time

While the level of understandingcontained in this volume is immense, it

is clear to most of us workingin this field that we have only just begun Thepotential is vast Here we have what is almost certainly the strongest, stiffest,toughest molecule that can ever be produced, the best possible molecularconductor of both heat and electricity In one sense the carbon (fullerene)nanotube is a new man-made polymer to follow on from nylon, polypropylene,and Kevlar In another, it is a new “graphite” fiber, but now with the ultimatepossible strength In yet another, it is a new species in organic chemistry, andpotentially in molecular biology as well, a carbon molecule with the almostalien property of electrical conductivity, and super-steel strength

VI Foreword

Can it be produced in megatons?

Can it be spun into continuous fibers?

Can it grown in organized arrays or as a perfect single crystal?

Can it be sorted by diameter and chirality?

Can a single tube be cloned?

Can it be grown enzymatically?

Can it be assembled by the molecular machinery of livingcells?

Can it be used to make nanoelectronic devices, nanomechanical memories,nano machines, .?

Can it be used to wire a brain?

There is no way of tellingat this point Certainly for many researchers, thebest, most excitingdays of discovery in this field are still ahead For therest of us, it will be entertainingjust to sit back and watch the worldwideorganism of science at work Hold on to your seats! Watch the future unfold

Houston, Texas

Carbon nanotubes are unique nanostructures with remarkable electronic andmechanical properties, some stemmingfrom the close relation between car-bon nanotubes and graphite, and some from their one-dimensional aspects.Initially, carbon nanotubes aroused great interest in the research communitybecause of their exotic electronic structure As other intriguing propertieshave been discovered, such as their remarkable electronic transport proper-ties, their unique Raman spectra, and their unusual mechanical properties,interest has grown in their potential use in nanometer-sized electronics and

in a variety of other applications, as discussed in this volume

An ideal nanotube can be considered as a hexagonal network of carbonatoms that has been rolled up to make a seamless hollow cylinder These hol-low cylinders can be tens of micrometers long, but with diameters as small as0.7 nm, and with each end of the longcylinder “capped with half a fullerenemolecule, i.e., 6 pentagons” Single-wall nanotubes, having a cylindrical shellwith only one atom in thickness, can be considered as the fundamental struc-tural unit Such structural units form the buildingblocks of both multi-wallnanotubes, containingmultiple coaxial cylinders of ever-increasingdiameterabout a common axis, and nanotube ropes, consistingof ordered arrays ofsingle-wall nanotubes arranged on a triangular lattice

The first reported observation of carbon nanotubes was by Iijima in 1991for multi-wall nanotubes It took, however, less than two years before single-wall carbon nanotubes were discovered experimentally by Iijima at the NECResearch Laboratory in Japan and by Bethune at the IBM Almaden Labora-tory in California These experimental discoveries and the theoretical work,which predicted many remarkable properties for carbon nanotubes, launchedthis field and propelled it forward The field has been advancingat a breath-takingpace ever since with many unexpected discoveries These excitingde-velopments encouraged the editors to solicit articles for this book on thetopic of carbon nanotubes while the field was in a highly active phase ofdevelopment

This book is organized to provide a snapshot of the present status of thisrapidly movingfield After the introduction in Chap 1, which provides somehistorical background and a brief summary of some basic subject matterand definitions, the connection between carbon nanotubes and other carbonmaterials is reviewed in Chap 2 Recent developments in the synthesis and

VIII Preface

purification of single-wall and multi-wall carbon nanotubes are discussed inChap 3 This is followed in Chap 4 by a review of our present limited un-derstanding of the growth mechanism of single-wall and multi-wall carbonnanotubes Chapter 5 demonstrates the generality of tubular nanostructures

by discussing general principles for tubule growth, and providing the readerwith numerous examples of inorganic nanotube formation The unique elec-tronic structure and properties of perfect and defective carbon nanotubes arereviewed from a theoretical standpoint in Chap 6 The electrical properties,transport, and magneto-transport measurements on single-wall nanotubesand ropes, as well as simple device structures based on carbon nanotubes arepresented in Chap 7 Scanningtunnelingmicroscopy is used to study thatnanotube electronic structure and spectra The use of nanotubes as atomicforce microscope tips for ultra-high resolution and chemically sensitive imag-ingis also discussed in Chap 8 The application of optical spectroscopy tonanotubes is presented in Chap 9 In this chapter, the discussion of the opti-cal properties focuses on the electronic structure, the phonon structure, andthe couplingbetween electrons and phonons in observations of resonance Ra-man scatteringand related phenomena The contribution made by electronspectroscopies to the characterization of the electronic structure of the nano-tubes is discussed in Chap 10, in comparison with similar studies devoted tographite and C60 This is followed in Chap 11 by a brief review of the phononand thermal properties, with emphasis given to studies of the specific heatand the thermal conductivity, which are both sensitive to the low-dimensionalaspects of carbon nanotubes Chapter 12 discusses experiments and theory

on the mechanical properties of carbon nanotubes Linear elastic parameters,non-linear instabilities, yield strength, fracture and supra-molecular interac-tions are all reviewed Chapter 13 discusses transport measurements, magne-totransport properties, electron spin resonance, and a variety of other exoticproperties of multiwall nanotubes The volume concludes in Chap 14 with

a brief review of the present early status of potential applications of carbonnanotubes

Because of the relative simplicity of carbon nanotubes, we expect them toplay an important role in the current rapid expansion of fundamental studies

on nanostructures and their potential use in nanotechnology This simplicityallows us to develop detailed theoretical models for predictingnew phenom-ena connected with these tiny, one-dimensional systems, and then look forthese phenomena experimentally Likewise, new experimental effects, whichhave been discovered at an amazingly rapid rate, have provided stimulus forfurther theoretical developments, many of which are expected to be broadlyapplicable to nanostructures and nanotechnology research and development

Relation of Carbon Nanotubes to Other

Carbon Materials

Mildred S Dresselhaus1 and Morinobu Endo2

1 Department of Electrical Engineering and Computer Science

and Department of Physics

MIT, Cambridge, MA 02139, USA

millie@mgm.mit.edu

2 Faculty of Engineering

Department of Electrical and Electronic Engineering

Shinshu University, Nagano-shi, 380 Japan

endo@endomoribu.shinshu-u.ac.jp

Abstract A review of the close connection between the structure and properties

of carbon nanotubes and those of graphite and its related materials is presented

in order to gain new insights into the exceptional properties of carbon nanotubes.The two dominant types of bonding (sp2 and sp3) that occur in carbon materi-

als and carbon nanotubes are reviewed, along with the structure and properties ofcarbon materials closely related to carbon nanotubes, such as graphite, graphitewhiskers, and carbon fibers The analogy is made between the control of the prop-erties of graphite through the intercalation of donor and acceptor species with thecorresponding doping of carbon nanotubes

Carbon nanotubes are strongly related to other forms of carbon, especially tocrystalline 3D graphite, and to its constituent 2D layers (where an individualcarbon layer in the honeycomb graphite lattice is called a graphene layer) Inthis chapter, several forms of carbon materials are reviewed, with particularreference to their relevance to carbon nanotubes Their similarities and dif-ferences relative to carbon nanotubes with regard to structure and propertiesare emphasized

The bonding between carbon atoms in the sp2

and sp3 configurations

is discussed in Sect 1 Connections are made between the nanotube

curva-ture and the introduction of some sp3bonding to the sp2planar bonding ofthe graphene sheet The unusual properties of carbon nanotubes are derived

from the unusual properties of sp2 graphite by imposing additional tum confinement and topological constraints in the circumferential direction

quan-of the nanotube The structure and properties quan-of graphite are discussed inSect.2, because of their close connection to the structure and properties ofcarbon nanotubes, which is followed by reviews of graphite whiskers and car-bon fibers in Sect.3and Sect.4, respectively Particular emphasis is given tothe vapor grown carbon fibers because of their especially close connection tocarbon nanotubes The chapter concludes with brief reviews of liquid carbon

M S Dresselhaus, G Dresselhaus, Ph Avouris (Eds.): Carbon Nanotubes,

Topics Appl Phys.80, 11–28 (2001)

c

Springer-Verlag Berlin Heidelberg 2001

12 Mildred S Dresselhaus and Morinobu Endo

and graphite intercalation compounds in Sect 5 and Sect 6, respectively,relating donor and acceptor nanotubes to intercalated graphite

Carbon-based materials, clusters, and molecules are unique in many ways.One distinction relates to the many possible configurations of the electronicstates of a carbon atom, which is known as the hybridization of atomic or-bitals and relates to the bonding of a carbon atom to its nearest neighbors.Carbon is the sixth element of the periodic table and has the lowest atomicnumber of any element in column IV of the periodic table Each carbon atom

has six electrons which occupy 1s2, 2s2, and 2p2 atomic orbitals The 1s2

orbital contains two strongly bound core electrons Four more weakly bound

electrons occupy the 2s22p2valence orbitals In the crystalline phase, the

va-lence electrons give rise to 2s, 2p x , 2p y , and 2p zorbitals which are important

in forming covalent bonds in carbon materials Since the energy difference

between the upper 2p energy levels and the lower 2s level in carbon is small

compared with the binding energy of the chemical bonds, the electronic wavefunctions for these four electrons can readily mix with each other, thereby

changing the occupation of the 2s and three 2p atomic orbitals so as to

en-hance the binding energy of the C atom with its neighboring atoms The

general mixing of 2s and 2p atomic orbitals is called hybridization, whereas the mixing of a single 2s electron with one, two, or three 2p electrons is called

sp n hybridization with n = 1, 2, 3 [1,2]

Thus three possible hybridizations occur in carbon: sp, sp2and sp3, while

other group IV elements such as Si and Ge exhibit primarily sp3 tion Carbon differs from Si and Ge insofar as carbon does not have inner

hybridiza-atomic orbitals, except for the spherical 1s orbitals, and the absence of nearby inner orbitals facilitates hybridizations involving only valence s and p orbitals

for carbon The various bonding states are connected with certain structural

arrangements, so that sp bonding gives rise to chain structures, sp2 bonding

to planar structures and sp3bonding to tetrahedral structures

The carbon phase diagram (see Fig.1)guided the historical synthesis ofdiamond in 1960 [4], and has continued to inspire interest in new forms ofcarbon, as they are discovered [3] Although we have learned much aboutcarbon since that time, much ignorance remains about the possible phases of

carbon While sp2bonded graphite is the ground state phase of carbon under

ambient conditions, at higher temperatures and pressures, sp3 bonded cubicdiamond is stable Other regions of the phase diagram show stability rangesfor hexagonal diamond, hexagonal carbynes [5,6,7], and liquid carbon [8] It

is believed that a variety of novel π-electron carbon bulk phases remain to

be discovered and explored

In addition to the bulk phases featured in the carbon phase diagram,much attention has recently focussed on small carbon clusters [9], since the

Relation of Carbon Nanotubes to Other Carbon Materials 13

Fig 1 A recent version of the phase diagram of carbon [3] Solid lines represent

equilibrium phase boundaries A: commercial synthesis of diamond from graphite

by catalysis; B: rapid solid phase graphite to diamond synthesis; C: fast formation of diamond to graphite; D: hexagonal graphite to hexagonal diamondsynthesis; E: shock compression graphite to hexagonal diamond synthesis; F: shockcompression graphite to cubic-type diamond synthesis; B, F, G: graphite or hexag-onal diamond to cubic diamond synthesis; H, I, J: compressed graphite acquiresdiamond-like properties, but reverts to graphite upon release of pressure

trans-discovery of fullerenes in 1985 by Kroto et al [10] and of carbon nanotubes

in 1991 by Iijima [11] The physical reason why these nanostructures form isthat a graphene layer (defined as a single 2D layer of 3D graphite)of finitesize has many edge atoms with dangling bonds,indexdangling bonds and thesedangling bonds correspond to high energy states Therefore the total energy of

a small number of carbon atoms (30–100)is reduced by eliminating danglingbonds, even at the expense of increasing the strain energy, thereby promotingthe formation of closed cage clusters such as fullerenes and carbon nanotubes.The rolling of a single graphene layer, which is a hexagonal network

of carbon atoms, to form a carbon nanotube is reviewed in this volume

in the introductory chapter [12], and in the chapters by Louie [13] and

Saito/Kataura [14], where the two indices (n, m)that fully identify each

carbon nanotube are specified [9,15] Since nanotubes can be rolled from

a graphene sheet in many ways [9,15], there are many possible orientations

of the hexagons on the nanotubes, even though the basic shape of the carbonnanotube wall is a cylinder

A carbon nanotube is a graphene sheet appropriately rolled into a cylinder

of nanometer size diameter [13,14,15] Therefore we can expect the planar sp2bonding that is characteristic of graphite to play a significant role in carbon

nanotubes The curvature of the nanotubes admixes a small amount of sp3

14 Mildred S Dresselhaus and Morinobu Endo

bonding so that the force constants (bonding)in the circumferential tion are slightly weaker than along the nanotube axis Since the single wallcarbon nanotube is only one atom thick and has a small number of atomsaround its circumference, only a few wave vectors are needed to describe theperiodicity of the nanotubes These constraints lead to quantum confinement

direc-of the wavefunctions in the radial and circumferential directions, with planewave motion occurring only along the nanotube axis corresponding to a largenumber or closely spaced allowed wave vectors Thus, although carbon nano-tubes are closely related to a 2D graphene sheet, the tube curvature and thequantum confinement in the circumferential direction lead to a host of prop-erties that are different from those of a graphene sheet Because of the closerelation between carbon nanotubes and graphite, we review briefly the struc-ture and properties of graphite in this chapter As explained in the chapter

by Louie [13], (n, m)carbon nanotubes can be either metallic (n − m = 3q,

q = 0, 1, 2, )or semiconducting (n−m = 3q±1, q = 0, 1, 2, ), the

individ-ual constituents of multi-wall nanotubes or single-wall nanotube bundles can

be metallic or semiconducting [13,15] These remarkable electronic propertiesfollow from the electronic structure of 2D graphite under the constraints ofquantum confinement in the circumferential direction [13]

Actual carbon nanotube samples are usually found in one of two forms:(1)a Multi-Wall Carbon Nanotube (MWNT)consisting of a nested coaxialarray of single-wall nanotube constituents [16], separated from one another

by approximately 0.34 nm, the interlayer distance of graphite (see Sect 2),and (2)a single wall nanotube rope, which is a nanocrystal consisting of∼10–

100 Single-Wall Nanotubes (SWNTs), whose axes are aligned parallel to oneanother, and are arranged in a triangular lattice with a lattice constant that

is approximately equal to dt+ ct, where dt is the nanotube diameter and ct

is approximately equal to the interlayer lattice constant of graphite

The ideal crystal structure of graphite (see Fig.2)consists of layers in whichthe carbon atoms are arranged in an open honeycomb network containingtwo atoms per unit cell in each layer, labeled A and B The stacking of thegraphene layers is arranged, such that the A and A atoms on consecutivelayers are on top of one another, but the B atoms in one plane are over theunoccupied centers of the adjacent layers, and similarly for the B atoms onthe other plane [17] This gives rise to two distinct planes, which are labeled

by A and B These distinct planes are stacked in the ‘ABAB’ Bernal stackingarrangement shown in Fig.2, with a very small in-plane nearest-neighbor

distance aC−Cof 1.421 ˚A, an in-plane lattice constant a0 of 2.462 ˚A, a c-axis lattice constant c0of 6.708 ˚A, and an interplanar distance of c0/2 = 3.354 ˚A

This crystal structure is consistent with the D46h (P 63/mmc)space group and

has four carbon atoms per unit cell, as shown in Fig.2

Relation of Carbon Nanotubes to Other Carbon Materials 15

Fig 2 (a) The crystal structure of hexagonal single crystal graphite, in which

the two distinct planes of carbon hexagons called A and B planes are stacked in

an ABAB sequence with P 63/mmc symmetry The notation for the A and B

planes is not to be confused with the two distinct atoms A and B on a singlegraphene plane (note a rhombohedral phase of graphite with ABCABC stackingalso exists [17]) (b) An STM image showing the trigonal network of highly oriented

pyrolytic graphite (HOPG) in which only one site of the carbon hexagonal network

appears, as for example, the B site, denoted by black balls in (a)

Since the in-plane C-C bond is very strong and the nearest-neighbor ing between carbon atoms in graphite is very small, the in-plane lattice con-stant is quite stable against external perturbations The nearest neighborspacing between carbon nanotubes is essentially the same as the interplanarspacing in graphite (∼3.4 ˚ A) One consequence of the small value of aC−Cingraphite is that impurity species are unlikely to enter the covalently bondedin-plane lattice sites substitutionally (except for boron), but rather occupysome interstitial position between the graphene layer planes which are bonded

spac-by a weak van der Waals force These arguments also apply to carbon tubes and explain why the substitutional doping of individual single wallcarbon nanotubes with species other than boron is difficult The weak inter-planar bonding of graphite allows entire planes of dopant atoms or molecules

nano-to be intercalated between the carbon layers nano-to form intercalation compounds.Also carbon nanotubes can adsorb dopant species on their external and in-ternal surfaces and in interstitial sites between adjacent nanotubes, as isdiscussed in Sect.6

The graphene layers often do not stack perfectly and do not form theperfect graphite crystal structure with perfect Bernal ‘ABAB’ layer stack-ing Instead, stacking faults are often formed (meaning departures from theABAB stacking order) These stacking faults give rise to a small increase inthe interlayer distance from the value 3.354 ˚A in 3D graphite until a value ofabout 3.440 ˚A is reached, at which interplanar distance, the stacking of theindividual carbon layers become uncorrelated with essentially no site bond-

16 Mildred S Dresselhaus and Morinobu Endo

ing between the carbon aatoms in the two layers The resulting structure of

these uncorrelated 2D graphene layers is called turbostratic graphite [1,18].Because of the different diameters of adjacent cylinders of carbon atoms in

a multiwall carbon nanotube [15,16], the structural arrangement of the jacent carbon honeycomb cylinders is essentially uncorrelated with no sitecorrelation between carbon atoms on adjacent nanotubes The stacking ar-rangement of the nanotubes is therefore similar in behavior to the graphenesheets in turbostratic graphite Thus, perfect nanotube cylinders at a largespatial separation from one another should be able to slide past one anothereasily

ad-Of significance to the properties expected for carbon nanotubes is the factthat the electronic structure of turbostratic graphite, a zero gap semiconduc-tor, is qualitatively different from that of ideal graphite, a semimetal with

a small band overlap (0.04 eV) The electronic structure of a 2D graphenesheet [15] is discussed elsewhere in this volume [14], where it is shown thatthe valence and conduction bands of a graphene sheet are degenerate by sym-

metry at the special point K at the 2D Brillouin zone corner where the Fermi

level in reciprocal space is located [19] Metallic carbon nanotubes have an

allowed wavevector at the K-point and therefore are effectively zero gap

semi-conductors like a 2D graphene sheet However, semiconducting nanotubes do

not have an allowed wavevector at the K point (because of quantum

confine-ment conditions in the circumferential direction)[14,15], thus resulting in anelectronic band gap and semiconducting behavior, very different from that of

a graphene sheet

Several sources of crystalline graphite are available, but differ somewhat

in their overall characteristics Some discussion of this topic could be helpful

to readers since experimentalists frequently use these types of graphite ples in making comparisons between the structure and properties of carbon

sam-nanotubes and sp2graphite

Natural single-crystal graphite flakes are usually small in size (typicallymuch less than 0.1 mm in thickness), and contain defects in the form of twin-ning planes and screw dislocations, and also contain chemical impurities such

as Fe and other transition metals, which make these graphite samples lessdesirable for certain scientific studies and applications

A synthetic single-crystal graphite, called “kish” graphite, is commonlyused in scientific investigations Kish graphite crystals form on the surface ofhigh carbon content iron melts and are harvested as crystals from such hightemperature solutions [20] The kish graphite flakes are often larger thanthe natural graphite flakes, which makes kish graphite the material of choicewhen large single-crystal flakes are needed for scientific studies However,these flakes may contain impurities

The most commonly used high-quality graphitic material today is Highly Oriented Pyrolytic Graphite (HOPG), which is prepared by the pyrolysis of

hydrocarbons at temperatures of above 2000◦C and the resulting pyrolytic

Relation of Carbon Nanotubes to Other Carbon Materials 17

carbon is subsequently heat treated to higher temperatures to improve itscrystalline order [21,22] When stress annealed above 3300◦C, the HOPG ex-hibits electronic, transport, thermal, and mechanical properties close to those

of single-crystal graphite, showing a very high degree of c-axis alignment For

the high temperature, stress-annealed HOPG, the crystalline order extends to

about 1 µm within the basal plane and to about 0.1µm along the c-direction.

This material is commonly used because of its good physical properties, highchemical purity and relatively large sample sizes Thin-film graphite materi-als, especially those based on Kapton and Novax (polyimide)precursors, arealso prepared by a pyrolysis/heat treatment method, and are often used

A graphite whisker is a graphitic material formed by rolling a graphene sheet

up into a scroll [23] Except for the early work by Bacon [23], there is tle literature about graphite whiskers Graphite whiskers are formed in a dcdischarge between carbon electrodes using 75–80 V and 70–76 A In the arcapparatus, the diameter of the positive electrode is smaller than that of thenegative electrode, and the discharge is carried out in an inert gas using ahigh gas pressure (92 atmospheres) As a result of this discharge, cylindri-cal boules with a hard shell were formed on the negative electrode Whenthese hard cylindrical boules were cracked open, scroll-like carbon whiskers

lit-up to ∼3 cm long and 1–5µm in diameter were found protruding from the

fracture surfaces The whiskers exhibited great crystalline perfection, highelectrical conductivity, and high elastic modulus along the fiber axis Sincetheir discovery [23], graphite whiskers have provided the benchmark againstwhich the performance of carbon fibers is measured The growth of graphitewhiskers by the arc method has many similarities to the growth of carbonnanotubes [24], especially MWNTs which do not require the use of a catalyst,except that graphite whiskers were grown at a higher gas pressure than iscommonly used for nanotube growth While MWNTs are generally found to

be concentric cylinders of much smaller outer diameter, some reports havebeen given of scroll-like structures with outer diameters less than 100 nm [23]

Carbon fibers represent an important class of graphite-related materialswhich are closely connected to carbon nanotubes, with regard to structureand properties Despite the many precursors that can be used to synthe-size carbon fibers, each having different cross-sectional morphologies (Fig.3),the preferred orientation of the graphene planes is parallel to the fiber axisfor all carbon fibers, thereby accounting for the high mechanical strength

of carbon fibers [1] Referring to the various morphologies in Fig.3, the prepared vapor-grown fibers have an “onion skin” or “tree ring” morphology

as-18 Mildred S Dresselhaus and Morinobu Endo

Fig 3 Sketch illustrating the morphology of Vapor-Grown Carbon Fibers (VGCF):

(a) as-deposited at 1100◦C [1], (b) after heat treatment to 3000◦C [1] The

mor-phologies for commercial mesophase-pitch fibers are shown in (c) for a “PAC-man”

cross section with a radial arrangement of the straight graphene ribbons and a

miss-ing wedge and (d) for a PAN-AM cross-sectional arrangement of graphene planes.

In (e) a PAN fiber is shown, with a circumferential arrangement of ribbons in the

sheath region and a random structure in the core

and after heat treatment to about 2500◦C bear a close resemblance to carbonnanotubes (Fig.3a) After further heat treatment to about 3000◦C, the outerregions of the vapor grown carbon fibers form facets (Fig.3b), and becomemore like graphite because of the strong interplanar correlations resultingfrom the facets [1,25] At the hollow core of a vapor grown carbon fiber is

a multiwall (and also a single wall)carbon nanotube (MWNT), as shown inFig.4, where the MWNT is observed upon fracturing a vapor grown carbonfiber [1] Of all carbon fibers, the faceted vapor grown carbon fibers (Fig.3b)are closest to crystalline graphite in both crystal structure and properties.The commercially available mesophase pitch-based fibers, are exploitedfor their extremely high bulk modulus and high thermal conductivity, whilethe commercial PAN (polyacrylonitrile)fibers are widely used for their hightensile strength [1] The high modulus of the mesophase pitch fibers is related

to the high degree of c-axis orientation of adjacent graphene layers, while the

high strength of the PAN fibers is related to defects in the structure Thesestructural defects inhibit the slippage of adjacent graphene planes relative toeach other, and inhibit the sword-in-sheath failure mode (Fig.5)that dom-inates the rupture of a vapor grown carbon fiber [1] Typical diameters forindividual commercial carbon fibers are∼ 7µm, and they can be very long.

These fibers are woven into bundles called tows and are then wound up as

a continuous yarn on a spool The remarkable high strength and modulus

of carbon fibers (Fig.6)are responsible for most of the commercial interest

Relation of Carbon Nanotubes to Other Carbon Materials 19

(a)

(b)

Fig 4 (a) Carbon nanotube exposed on

the breakage edge of a vapor grown

car-bon fiber as grown (a) and heat-treated

at 3000◦C (b) The sample is fractured

by pulverization and the core diameter is

∼5 nm (b) These photos suggest a

struc-tural discontinuity between the nanotubecore of the fiber and the outer carbon lay-ers deposited by chemical vapor depositiontechniques The photos show the strongmechanical properties of the nanotube corewhich maintain its form after breakage ofthe periphery [26]

Fig 5 The sword-in-sheath

fail-ure mode of heat treated vaporgrown carbon fibers Such failuremodes are also observed in multi-wall carbon nanotubes [1]

in these fibers and these superior mechanical properties (modulus and sile strength)should be compared to steel, for which typical strengths andbulk modulus values are 1.4 and 207 GPa, respectively [1] The excellent me-chanical properties of carbon nanotubes are closely related to the excellentmechanical properties of carbon fibers, though notable differences in behaviorare also found, such as the flexibility of single wall carbon nanotubes and thegood mechanical properties of multiwall nanotubes (with only a few walls)under compression, in contrast with carbon fibers which fracture easily undercompressive stress along the fiber axis

ten-Vapor-grown carbon fibers can be prepared over a wide range of ters (from∼10 nm to more than 100µm)and these fibers have central hollow

diame-cores A distinction is made between vapor grown carbon fibers and fibers

20 Mildred S Dresselhaus and Morinobu Endo

Fig 6 The breaking strength

of various types of carbon fibersplotted as a function of Young’smodulus Lines of constantstrain can be used to estimatethe breaking strains [1,27,28]

with diameters in the range 10–100 nm, which are called nanofibers, and

ex-hibit properties intermediate between those of typical vapor grown carbonfibers, on the one hand, and MWNTs, on the other [24] The preparation ofvapor grown carbon fibers is based on the growth of a thin hollow tube ofabout 100 nm diameter (a nanofiber)by a catalytic process based on ultra-fineparticles (∼10 nm diameter)which have been super-saturated with carbon

from the pyrolysis of a hydrocarbon gas at ∼1050 ◦C [1,29] The thickening

of the vapor-grown carbon fiber occurs through an epitaxial growth process,whereby the hydrocarbon gas is dehydrogenated at the∼1050 ◦C growth tem-

perature, and the carbon deposit is adsorbed on the surface of the growingfiber Subsequent heat treatment to ∼2500 ◦C anneals the disordered car-

bon deposit and results in vapor grown carbon fibers with a tree ring coaxialcylinder morphology [29] Further heat treatment to 2900◦C results in facetedfibers (Fig.3b)which exhibit structural and electronic properties very close

to those of single crystal graphite [1,29] If the growth process is stoppedbefore the thickening step starts, MWNTs are obtained [30]

These vapor grown carbon fibers show (h, k, l)X-ray diffraction lines

in-dicative of the 3D graphite structure and a semimetallic band overlap andcarrier density similar to 3D graphite The infrared and Raman spectra areessentially the same as that of 3D graphite Vapor-grown carbon fibers andnanofibers with micrometer and several tens of nanometer diameters, respec-tively, provide intermediate materials between conventional mesophase pitch-derived carbon fibers (see Fig.3)and single wall carbon nanotubes Since theirsmallest diameters (∼10 nm)are too large to observe significant quantum con-

finement effects, it would be difficult to observe band gaps in their electronicstructure, or the radial breathing mode in their phonon spectra Yet subtledifferences are expected between nanofiber transport properties, electronic

Relation of Carbon Nanotubes to Other Carbon Materials 21

structure and Raman spectra relative to the corresponding phenomena ineither MWNTs of graphitic vapor grown carbon fibers At present, little isknown in detail about the structure and properties of nanofibers, except thattheir properties are intermediate between those of vapor grown carbon fibersand MWNTs

4.1 History of Carbon Fibers in Relation to Carbon Nanotubes

We provide here a brief review of the history of carbon fibers, the macroscopicanalog of carbon nanotubes, since carbon nanotubes have become the focus

of recent developments in carbon fibers

The early history of carbon fibers was stimulated by needs for materialswith special properties, both in the 19th century and more recently afterWorld War II The first carbon fiber was prepared by Thomas A Edison

to provide a filament for an early model of an electric light bulb Speciallyselected Japanese Kyoto bamboo filaments were used to wind a spiral coilthat was then pyrolyzed to produce a coiled carbon resistor, which could beheated ohmically to provide a satisfactory filament for use in an early model

of an incandescent light bulb [31] Following this initial pioneering work byEdison, further research on carbon filaments proceeded more slowly, sincecarbon filaments were soon replaced by a more sturdy tungsten filament inthe electric light bulb Nevertheless research on carbon fibers and filaments

proceeded steadily over a long time frame, through the work of Sch¨ utzenberger and Sch¨ utzenberger (1890)[32], Pelabon [33], and others Their efforts weremostly directed toward the study of vapor grown carbon filaments, showingfilament growth from the thermal decomposition of hydrocarbons

The second applications-driven stimulus to carbon fiber research came

in the 1950’s from the needs of the space and aircraft industry for strong,stiff light-weight fibers that could be used for building lightweight compositematerials with superior mechanical properties This stimulation led to greatadvances in the preparation of continuous carbon fibers based on polymer pre-cursors, including rayon, polyacrylonitrile (PAN)and later mesophase pitch.The late 1950’s and 1960’s was a period of intense activity at the Union Car-bide Corporation, the Aerospace Corporation and many other laboratoriesworldwide This stimulation also led to the growth of a carbon whisker [23](see Sect.3), which has become a benchmark for the discussion of the mechan-ical and elastic properties of carbon fibers The growth of carbon whiskerswas also inspired by the successful growth of single crystal whisker filaments

at that time for many metals such as iron, non-metals such as Si, and oxidessuch as Al2O3, and by theoretical studies [34], showing superior mechanicalproperties for whisker structures [35] Parallel efforts to develop new bulk syn-thetic carbon materials with properties approaching single crystal graphiteled to the development of highly oriented pyrolytic graphite (HOPG)in 1962

by Ubbelohde and co-workers [36,37], and HOPG has since been used as one

of the benchmarks for the characterization of carbon fibers

22 Mildred S Dresselhaus and Morinobu Endo

While intense effort continued toward perfecting synthetic filamentarycarbon materials, and great progress was indeed made in the early 1960’s, itwas soon realized that long term effort would be needed to reduce fiber de-fects and to enhance structures resistive to crack propagation New researchdirections were introduced because of the difficulty in improving the structureand microstructure of polymer-based carbon fibers for high strength and highmodulus applications, and in developing graphitizable carbons for ultra-highmodulus fibers Because of the desire to synthesize more crystalline filamen-tous carbons under more controlled conditions, synthesis of carbon fibers by

a catalytic Chemical Vapor Deposition (CVD)process proceeded, laying thescientific basis for the mechanism and thermodynamics for the vapor phasegrowth of carbon fibers in the 1960’s and early 1970’s [1] In parallel to thesescientific studies, other research studies focused on control of the processfor the synthesis of vapor grown carbon fiber [38,39,40,41], leading to themore recent commercialization of vapor grown carbon fibers in the 1990’s forvarious applications Concurrently, polymer-based carbon fiber research hascontinued worldwide, mostly in industry, with emphasis on greater control ofprocessing steps to achieve carbon fibers with ever-increasing modulus andstrength, and on fibers with special characteristics, such as very high thermalconductivity, while decreasing costs of the commercial products

As research on vapor grown carbon fibers on the micrometer scale ceeded, the growth of very small diameter filaments less than 10 nm (Fig.7),was occasionally observed and reported [42,43], but no detailed systematicstudies of such thin filaments were carried out An example of a very thin va-por grown nanofiber along with a multiwall nanotube is shown in the brightfield TEM image of Fig.7[42,43,44,45]

pro-Reports of thin filaments below 10 nm inspired Kubo [46] to ask whetherthere was a minimum dimension for such filaments Early work [42,43] onvapor grown carbon fibers, obtained by thickening filaments such as the fiberdenoted by VGCF (Vapor Grown Carbon Fiber)in Fig.7, showed very sharplattice fringe images for the inner-most cylinders Whereas the outermostlayers of the fiber have properties associated with vapor grown carbon fibers,

Fig 7 High-resolution TEM micrograph

showing carbon a vapor grown carbonnanofiber (VGCF) with an diameter lessthan 10 nm and a nanotube [42,43,44,45]

Relation of Carbon Nanotubes to Other Carbon Materials 23

there may be a continuum of behavior of the tree rings as a function of eter, with the innermost tree rings perhaps behaving like carbon nanotubes.Direct stimulus to study carbon filaments of very small diameters moresystematically [11] came from the discovery of fullerenes by Kroto and Smal- ley [10] In December 1990 at a carbon-carbon composites workshop, papers

diam-were given on the status of fullerene research by Smalley [47], the discovery

of a new synthesis method for the efficient production of fullerenes by man [48], and a review of carbon fiber research by Dresselhaus [49] Discus-sions at the workshop stimulated Smalley to speculate about the existence

Huff-of carbon nanotubes Huff-of dimensions comparable to C60 These conjectureswere later followed up in August 1991 by discussions at a fullerene workshop

in Philadelphia [50] on the symmetry proposed for a hypothetical single-wallcarbon nanotubes capped at either end by fullerene hemispheres, with sugges-tions on how zone folding could be used to examine the electron and phonondispersion relations of such structures However, the real breakthrough oncarbon nanotube research came with Iijima’s report of experimental observa-tion of carbon nanotubes using transmission electron microscopy [11] It wasthis work which bridged the gap between experimental observation and thetheoretical framework of carbon nanotubes in relation to fullerenes and as

theoretical examples of 1D systems Since the pioneering work of Iijima [11],the study of carbon nanotubes has progressed rapidly

Liquid carbon refers to the liquid phase of carbon resulting from the melting

of pure carbon in a solid phase (graphite, diamond, carbon fibers or a variety

of other carbons) The phase diagram for carbon shows that liquid carbon isstable at atmospheric pressure only at very high temperatures (the melting

point of graphite Tm ∼ 4450 K)[4] Since carbon has the highest meltingpoint of any elemental solid, to avoid contamination of the melt, the crucible

in which the carbon is melted must itself be made of carbon, and sufficientheat must be focused on the sample volume to produce the necessary temper-ature rise to achieve melting [8,51] Liquid carbon has been produced in thelaboratory by the laser melting of graphite, exploiting the poor interplanarthermal conductivity of the graphite [8], and by resistive heating [52], thetechnique used to establish the metallic nature of liquid carbon

Although diamond and graphite may have different melting temperatures,

it is believed that the same liquid carbon is obtained upon melting either solidphase It is likely that the melting of carbon nanotubes also forms liquidcarbon Since the vaporization temperature for carbon (∼4700 K)is only

slightly higher than the melting point (∼4450 K), the vapor pressure over

liquid carbon is high The high vapor pressure and the large carbon–carbonbonding energy make it energetically favorable for carbon clusters rather thanindependent atoms to be emitted from a molten carbon surface [53] Energetic

24 Mildred S Dresselhaus and Morinobu Endo

considerations suggest that some of the species emitted from a molten carbonsurface have masses comparable to those of fullerenes [51] The emission offullerenes from liquid carbon is consistent with the graphite laser ablation

studies of Kroto, Smalley, and co-workers [10]

Resistivity measurements on a variety of vapor grown carbon fibers vided important information about liquid carbon In these experiments fiberswere heated resistively by applying a single 28µs current pulse with currents

pro-up to 20 A [52] The temperature of the fiber as a function of time was mined from the energy supplied in the pulse and the measured heat capacityfor bulk graphite [54] over the temperature range up to the melting point,assuming that all the power dissipated in the current pulse was convertedinto thermal energy in the fiber The results in Fig.8 for well-graphitized

deter-fibers (heat treatment temperatures THT = 2300◦C, 2800◦C)show an

ap-proximately linear temperature dependence for the resistivity ρ(T )up to the melting temperature, where ρ(T )drops by nearly one order of magnitude,

consistent with metallic conduction in liquid carbon (Fig.8) This figure

fur-ther shows that both pregraphitic vapor grown carbon fibers (THT= 1700◦C,

2100◦C), which are turbostratic and have small structural coherence lengths,and well graphitized fibers, all show the same behavior in the liquid phase,

although their measured ρ(T )functional forms in the solid phase are very

different Measurements of the resistance of a carbon nanotube through themelting transition have not yet been carried out

Fig 8 The electrical resistivity vs temperature for vapor grown carbon fibers with

various heat treatment temperatures (THT= 1700, 2100, 2300, 2800◦C) The sharpdecrease inρ(T ) above ∼4000 K is identified with the melting of the carbon fibers.

The measured electrical resistivity for liquid carbon is shown [52]

Relation of Carbon Nanotubes to Other Carbon Materials 25

Because of the weak van der Waals interlayer forces associated with the sp2bonding in graphite, graphite intercalation compounds (GICs)may be formed

by the insertion of layers of guest species between the layers of the graphitehost material [56,57], as shown schematically in Fig.9 The guest speciesmay be either atomic or molecular In the so-called donor GICs, electrons aretransferred from the donor intercalate species (such as a layer of the alkali

metal potassium)into the graphite layers, thereby raising the Fermi level EF

in the graphitic electronic states, and increasing the mobile electron tration by two or three orders of magnitude, while leaving the intercalatelayer positively charged with low mobility carriers Conversely, for acceptorGICs, electrons are transferred to the intercalate species (which is usually

concen-molecular)from the graphite layers, thereby lowering the Fermi level EF inthe graphitic electronic states and creating an equal number of positively

charged hole states in the graphitic π-band Thus, electrical conduction in

GICs (whether they are donors or acceptors)occurs predominantly in thegraphene layers and as a result of the charge transfer between the intercalateand host layers The electrical conductivity between adjacent graphene layers

is very poor Among the GICs, Li-based GICs are widely commercialized in

Fig 9 Schematic model for a graphite

intercala-tion compound showing the stacking of graphitelayers (networks of hexagons on a sheet) and ofintercalate (e.g., potassium) layers (networks oflarge hollow balls) For this stage 1compound,each carbon layer is separated by an intercalatelayer [55]

26 Mildred S Dresselhaus and Morinobu Endo

Li-ion secondary batteries for cell phones, personal computers, and electricvehicle batteries [58]

Carbon nanotubes also provide a host material for intercalation of donors(e.g., alkali metals)or acceptors (e.g., halogens such as bromine and iodine)

In the case of nanotubes, the guest species is believed to decorate the rior of the single wall nanotubes, and also to enter the hollow cores of thenanotubes A similar charge transfer process is observed for alkali metalsand halogens in SWNTs, based on measuring the downshifts and upshifts in

exte-the characteristic Raman G-band mode frequencies, similar to exte-the standard

characterization techniques previously developed for GICs [9,56] Also, thedistances of the dopants to the nearest-neighbor carbon atoms are similarfor doped carbon nanotubes and for GICs [9,56] Practical applications ofintercalated carbon nanotubes are also expected in analogy with Li-GICs,especially for super high capacity batteries

References

1 M S Dresselhaus, G Dresselhaus, K Sugihara, I L Spain, H A Goldberg,

Graphite Fibers and Filaments, Springer Ser Mater Sci., Vol 5 (Springer,

Berlin, Heidelberg 1988) 12,16,17,18,19,20,22

2 B T Kelly, Physics of Graphite (Applied Science, London 1981) 12

3 F P Bundy, W A Bassett, M S Weathers, R J Hemley, H K Mao,

6 A G Whittaker, P L Kintner, Carbon 23, 255 (1985) 12

7 V I Kasatochkin, V V Korshak, Y P Kudryavtsev, A M Sladkov,

I E Sterenberg, Carbon 11, 70 (1973) 12

8 M S Dresselhaus, J Steinbeck, Tanso 132, 44–56 (1988) (Journal of the

Japanese Carbon Society) 12,23

9 M S Dresselhaus, G Dresselhaus, P C Eklund, Science of Fullerenes and Carbon Nanotubes (Academic, New York 1996) 12,13,26

10 H W Kroto, J R Heath, S C O’Brien, R F Curl, R E Smalley, Nature

(London) 318, 162–163 (1985) 13,23,24

11 S Iijima, Nature (London) 354, 56 (1991) 13,23

12 M S Dresselhaus, P Avouris, chapter 1 in this volume 13

13 S G Louie, chapter 6 in this volume 13,14

14 R Saito, H Kataura, chapter 9 in this volume 13,16

15 R Saito, G Dresselhaus, M S Dresselhaus, Physical Properties of Carbon Nanotubes (Imperial College Press, London 1998) 13,14,16

16 L Forr´o, C Sch¨onenberger, chapter 13 in this volume 14,16

17 R W G Wyckoff, Crystal Structures, (Interscience) New York 1964, Vol 1

14,15

18 J Maire, J M´ering, Proceedings of the First Conference of the Society of Chemical and Industrial Conference on Carbon and Graphite (London, 1958)

p 204 16

Relation of Carbon Nanotubes to Other Carbon Materials 27

19 P R Wallace, Phys Rev 71, 622 (1947) 16

20 S B Austerman, Chemistry and Physics of Carbon, Vol 7,

P L Walker, Jr (Ed.) (Dekker, New York 1968) p 137 16

21 A W Moore, Chemistry and Physics of Carbon, Vol 1 1 , P L Walker, Jr.,

P A Thrower (Eds.), (Dekker, New York 1973) p.69 17

22 A W Moore, Chemistry and Physics of Carbon, Vol 17, P L Walker, Jr.,

P A Thrower (Eds.), (Dekker, New York 1981) p 233 17

23 R Bacon, J Appl Phys 31, 283–290 (1960) 17,21

24 J C Charlier , Carbon Nanotubes, M S Dresselhaus, G Dresselhaus,

P Avouris (Eds.), (Springer, Berlin, 2000 Springer Series in Solid-State ences) 17,20

Sci-25 T C Chieu, G Timp, M S Dresselhaus, M Endo, A W Moore, Phys Rev

B 27, 3686 (1983) 18

26 M Endo and M S Dresselhaus, Science Spectra (2000) (in press) 19

27 M Endo, A Katoh, T Sugiura, M Shiraishi, Extended Abstracts of the 18th Biennial Conference on Carbon, (Worcester Polytechnic Institute, 1987) p 151

20

28 M Endo, T Momose, H Touhara, N Watanabe, J Power Sources 20, 99

(1987) 20

29 M Endo, CHEMTECH 18 568 (1988) (Sept.) 20

30 M Endo, K Takeuchi, K Kobori, K Takahashi, H Kroto, A Sarkar, Carbon

33, 873 (1995) 20

31 T A Edison, US Patent 470,925 (1892) (issued March 15, 1892) 21

32 P Sch¨utzenberger, L Sch¨utzenberger, Compt Rendue 111, 774 (1890) 21

33 C H Pelabon, Compt Rendue 137, 706 (1905) 21

34 C Herring, J K Galt, Phys Rev 85, 1060 (1952) 21

35 A P Levitt, Whisker Technology (Wiley-Interscience, New York 1970) 21

36 A W Moore, A R Ubbelohde, D A Young, Brit J Appl Phys 13, 393

40 G G Tibbetts, Appl Phys Lett 42, 666 (1983) 22

41 G G Tibbetts, J Cryst Growth 66, 632 (1984) 22

42 M Endo, Mecanisme de croissance en phase vapeur de fibres de carbone (The growth mechanism of vapor-grown carbon fibers), PhD thesis, University of

Orleans, Orleans, France, (1975) (in French) 22

43 M Endo, PhD thesis, Nagoya University, Japan, (1978) (in Japanese) 22

44 A Oberlin, M Endo, T Koyama, Carbon 14, 133 (1976) 22

45 A Oberlin, M Endo, T Koyama, J Cryst Growth 32, 335–349 (1976) 22

46 R Kubo, (private communication to M Endo at the Kaya Conference) (1977)22

47 R E Smalley, DoD Workshop in Washington, DC (Dec 1990) 23

48 D R Huffman, DoD Workshop in Washington, DC (Dec 1990) 23

49 M S Dresselhaus, DoD Workshop in Washington, DC (Dec 1990) 23

50 M S Dresselhaus, G Dresselhaus, P C Eklund, University of PennsylvaniaWorkshop (August 1991) 23

28 Mildred S Dresselhaus and Morinobu Endo

51 J Steinbeck, G Dresselhaus, M S Dresselhaus, Int J Thermophys 11, 789

(1990) 23,24

52 J Heremans, C H Olk, G L Eesley, J Steinbeck, G Dresselhaus, Phys Rev

Lett 60, 452 (1988) 23,24

53 J S Speck, J Steinbeck, G Braunstein, M S Dresselhaus, T Venkatesan, in

Beam-Solid Interactions and Phase Transformations, MRS Symp Proc., Vol.

51, 263, H Kurz, G L Olson, J M Poate (Eds.) (Materials Research Society

Press, Pittsburgh PA, 1986) 23

54 F P Bundy, H M Strong, R H Wentdorf, Jr Chemistry and Physics of Carbon, Vol 1 0, P L Walker, Jr., P A Thrower (Eds.), (Dekker, New York

1973) p 213 24

55 W R¨udorff, E Shultze, Z Anorg allg Chem 277, 156 (1954) 25

56 M S Dresselhaus, G Dresselhaus, Adv Phys 30, 139–326 (1981) 25,26

57 H Zabel, S A Solin (Eds.) Graphite Intercalation Compounds I: Structure and Dynamics, (Springer, Berlin, Heidelberg 1990) 25

58 M Endo, C Kim, T Karaki, Y Nishimura, M J Matthews, S D M Brown,

M S Dresselhaus, Carbon 37, 561–568 (1999) 26

Nanotube Growth and Characterization

Hongjie Dai

Department of Chemistry, Stanford University

Stanford, CA 94305-5080, USA

hdai@chem.stanford.edu

Abstract This chapter presents a review of various growth methods for carbon

nanotubes Recent advances in nanotube growth by chemical vapor deposition(CVD) approaches are summarized CVD methods are promising for producinghigh quality nanotube materials at large scales Moreover, controlled CVD growthstrategies on catalytically patterned substrates can yield ordered nanotube archi-tectures and integrated devices that are useful for fundamental characterizationsand potential applications of nanotube molecular wires

In 1991, Iijima of the NEC Laboratory in Japan reported the first tion of multi-walled carbon nanotubes (MWNT) in carbon-soot made by arc-discharge [1] About two years later, he made the observation of Single-WalledNanoTubes (SWNTs) [2] The past decade witnessed significant research ef-forts in efficient and high-yield nanotube growth methods The success innanotube growth has led to the wide availability of nanotube materials, and

observa-is a main catalyst behind the recent progress in basobserva-is physics studies andapplications of nanotubes

The electrical and mechanical properties of carbon nanotubes have tured the attention of researchers worldwide Understanding these propertiesand exploring their potential applications have been a main driving force forthis area Besides the unique and useful structural properties, a nanotube hashigh Young’s modulus and tensile strength A SWNT can behave as a well-defined metallic, semiconducting or semi-metallic wire depending on two keystructural parameters, chirality and diameter [3] Nanotubes are ideal sys-tems for studying the physics in one-dimensional solids Theoretical and ex-perimental work have focused on the relationship between nanotube atomicstructures and electronic structures, electron-electron and electron-phononinteraction effects [4] Extensive effort has been taken to investigate the me-chanical properties of nanotubes including their Young’s modulus, tensilestrength, failure processes and mechanisms Also, an intriguing fundamen-tal question has been how mechanical deformation in a nanotube affects itselectrical properties In recent years, progress in addressing these basic prob-lems has generated significant excitement in the area of nanoscale science andtechnology

cap-Nanotubes can be utilized individually or as an ensemble to build tional device prototypes, as has been demonstrated by many research groups.Ensembles of nanotubes have been used for field emission based flat-panel

func-M S Dresselhaus, G Dresselhaus, Ph Avouris (Eds.): Carbon Nanotubes,

Topics Appl Phys.80, 29–53 (2001)

c

Springer-Verlag Berlin Heidelberg 2001

30 Hongjie Dai

display, composite materials with improved mechanical properties and tromechanical actuators Bulk quantities of nanotubes have also been sug-gested to be useful as high-capacity hydrogen storage media Individual nano-tubes have been used for field emission sources, tips for scanning probe mi-croscopy and nano-tweezers Nanotubes also have significant potential as thecentral elements of nano-electronic devices including field effect transistors,single-electron transistors and rectifying diodes

elec-The full potential of nanotubes for applications will be realized until thegrowth of nanotubes can be optimized and well controlled Real-world ap-plications of nanotubes require either large quantities of bulk materials ordevice integration in scaled-up fashions For applications such as compositesand hydrogen storage, it is desired to obtain high quality nanotubes at thekilogram or ton level using growth methods that are simple, efficient andinexpensive For devices such as nanotube based electronics, scale-up willunavoidably rely on self-assembly or controlled growth strategies on surfacescombined with microfabrication techniques Significant work has been car-ried out in recent years to tackle these issues Nevertheless, many challengesremain in the nanotube growth area First, an efficient growth approach tostructurally perfect nanotubes at large scales is still lacking Second, growingdefect-free nanotubes continuously to macroscopic lengths has been difficult.Third, control over nanotube growth on surfaces should be gained in order toobtain large-scale ordered nanowire structures Finally, there is a seeminglyformidable task of controlling the chirality of SWNTs by any existing growthmethod

This chapter summarizes the progress made in recent years in carbonnanotube growth by various methods including arc-discharge, laser ablationand chemical vapor deposition The growth of nanotube materials by Chem-ical Vapor Deposition (CVD) in bulk and on substrates will be focused on

We will show that CVD growth methods are highly promising for scale-up

of defect-free nanotube materials, and enable controlled nanotube growth onsurfaces Catalytic patterning combined with CVD growth represents a novelapproach to ordered nanowire structures that can be addressed and utilized

Arc-discharge and laser ablation methods for the growth of nanotubes havebeen actively pursued in the past ten years Both methods involve the conden-sation of carbon atoms generated from evaporation of solid carbon sources.The temperatures involved in these methods are close to the melting temper-ature of graphite, 3000–4000◦C

1.1 Arc-Discharge and Laser Ablation

In arc-discharge, carbon atoms are evaporated by plasma of helium gas nited by high currents passed through opposing carbon anode and cathode

ig-Nanotube Growth and Characterization 31

(Fig.1a) Arc-discharge has been developed into an excellent method forproducing both high quality multi-walled nanotubes and single-walled nano-tubes MWNTs can be obtained by controlling the growth conditions such asthe pressure of inert gas in the discharge chamber and the arcing current In

1992, a breakthrough in MWNT growth by arc-discharge was first made byEbbesen and Ajayan who achieved growth and purification of high qualityMWNTs at the gram level [5] The synthesized MWNTs have lengths on theorder of ten microns and diameters in the range of 5-30 nm The nanotubes aretypically bound together by strong van der Waals interactions and form tightbundles MWNTs produced by arc-discharge are very straight, indicative oftheir high crystallinity For as grown materials, there are few defects such

as pentagons or heptagons existing on the sidewalls of the nanotubes Theby-product of the arc-discharge growth process are multi-layered graphiticparticles in polyhedron shapes Purification of MWNTs can be achieved byheating the as grown material in an oxygen environment to oxidize away thegraphitic particles [5] The polyhedron graphitic particles exhibit higher ox-idation rate than MWNTs; nevertheless, the oxidation purification processalso removes an appreciable amount of nanotubes

For the growth of single-walled tubes, a metal catalyst is needed in thearc-discharge system The first success in producing substantial amounts ofSWNTs by arc-discharge was achieved by Bethune and coworkers in 1993 [6].They used a carbon anode containing a small percentage of cobalt catalyst

in the discharge experiment, and found abundant SWNTs generated in thesoot material The growth of high quality SWNTs at the 1–10 g scale was

achieved by Smalley and coworkers using a laser ablation (laser oven) method

(Fig.1b) [7] The method utilized intense laser pulses to ablate a carbontarget containing 0.5 atomic percent of nickel and cobalt The target wasplaced in a tube-furnace heated to 1200◦C During laser ablation, a flow

of inert gas was passed through the growth chamber to carry the grownnanotubes downstream to be collected on a cold finger The produced SWNTsare mostly in the form of ropes consisting of tens of individual nanotubesclose-packed into hexagonal crystals via van der Waals interactions (Fig.2)

The optimization of SWNT growth by arc-discharge was achieved by Journet

and coworkers using a carbon anode containing 1.0 atomic percentage ofyttrium and 4.2 at % of nickel as catalyst [8]

In SWNT growth by arc-discharge and laser ablation, typical by-productsinclude fullerenes, graphitic polyhedrons with enclosed metal particles, andamorphous carbon in the form of particles or overcoating on the sidewalls ofnanotubes A purification process for SWNT materials has been developed

by Smalley and coworkers [9] and is now widely u sed by many researchers.The method involves refluxing the as-grown SWNTs in a nitric acid solutionfor an extended period of time, oxidizing away amorphous carbon species andremoving some of the metal catalyst species

32 Hongjie Dai

Fig 1 a–c Schematic

experi-mental setups for nanotubegrowth methods

The success in producing high quality SWNT materials by laser-ablationand arc-discharge has led to wide availability of samples useful for studyingfundamental physics in low dimensional materials and exploring their appli-cations

1.2 Chemical Vapor Deposition

Chemical vapor deposition (CVD) methods have been successful in makingcarbon fiber, filament and nanotube materials since more than 10–20 yearsago [10,11,12,13,14,15,16,17,18,19]

1.2.1 General Approach and Mechanism

A schematic experimental setup for CVD growth is depicted in Fig.1c Thegrowth process involves heating a catalyst material to high temperatures

in a tube furnace and flowing a hydrocarbon gas through the tube reactorfor a period of time Materials grown over the catalyst are collected upon

Nanotube Growth and Characterization 33

Fig 2 a,b. Single-wallednanotubes grown by laserablation (courtesy of

R Smalley)

cooling the system to room temperature The key parameters in nanotubeCVD growth are the hydrocarbons, catalysts and growth temperature Theactive catalytic species are typically transition-metal nanoparticles formed

on a support material such as alumina The general nanotube growth anism (Fig.3) in a CVD process involves the dissociation of hydrocarbonmolecules catalyzed by the transition metal, and dissolution and saturation

mech-of carbon atoms in the metal nanoparticle The precipitation mech-of carbon fromthe saturated metal particle leads to the formation of tubular carbon solids in

sp2structure Tubule formation is favored over other forms of carbon such asgraphitic sheets with open edges This is because a tube contains no danglingbonds and therefore is in a low energy form For MWNT growth, most of theCVD methods employ ethylene or acetylene as the carbon feedstock and thegrowth temperature is typically in the range of 550–750◦C Iron, nickel orcobalt nanoparticles are often used as catalyst The rationale for choosingthese metals as catalyst for CVD growth of nanotubes lies in the phase di-agrams for the metals and carbon At high temperatures, carbon has finitesolubility in these metals, which leads to the formation of metal-carbon solu-tions and therefore the aforementioned growth mechanism Noticeably, iron,

34 Hongjie Dai

Fig 3 Two general growth

modes of nanotube in

chemi-cal vapor deposition Left agram: base growth mode

di-Right diagram: tip growthmode

cobalt and nickel are also the favored catalytic metals used in laser ablationand arc-discharge This simple fact may hint that the laser, discharge andCVD growth methods may share a common nanotube growth mechanism,although very different approaches are used to provide carbon feedstock

A major pitfall for CVD grown MWNTs has been the high defect densities

in their structures The defective nature of CVD grown MWNTs remains

to be thoroughly understood, but is most likely be due to the relativelylow growth temperature, which does not provide sufficient thermal energy

to anneal nanotubes into perfectly crystalline structures Growing perfectMWNTs by CVD remains a challenge to this day

1.2.2 Single-Walled Nanotube Growth and Optimization

For a long time, arc-discharge and laser-ablation have been the principalmethods for obtaining nearly perfect single-walled nanotube materials Thereare several issues concerning these approaches First, both methods rely onevaporating carbon atoms from solid carbon sources at ≥ 3000 ◦C, which

is not efficient and limits the scale-up of SWNTs Secondly, the nanotubessynthesized by the evaporation methods are in tangled forms that are diffi-cult to purify, manipulate and assemble for building addressable nanotubestructures

Recently, growth of single-walled carbon nanotubes with structural tion was enabled by CVD methods For an example, we found that by usingmethane as carbon feedstock, reaction temperatures in the range of 850–

perfec-1000◦C, suitable catalyst materials and flow conditions one can grow highquality SWNT materials by a simple CVD process [20,21,22,23] High tem-

Nanotube Growth and Characterization 35

perature is necessary to form SWNTs that have small diameters and thus highstrain energies, and allow for nearly-defect free crystalline nanotube struc-tures Among all hydrocarbon molecules, methane is the most stable at hightemperatures against self-decomposition Therefore, catalytic decomposition

of methane by the transition-metal catalyst particles can be the dominantprocess in SWNT growth The choice of carbon feedstock is thus one of thekey elements to the growth of high quality SWNTs containing no defectsand amorphous carbon over-coating Another CVD approach to SWNTs wasreported by Smalley and coworkers who used ethylene as carbon feedstockand growth temperature around 800◦C [24] In this case, low partial-pressureethylene was employed in order to reduce amorphous carbon formation due

to self-pyrolysis/dissociation of ethylene at the high growth temperature.Gaining an understanding of the chemistry involved in the catalyst andnanotube growth process is critical to enable materials scale-up by CVD [22].The choice of many of the parameters in CVD requires to be rationalized inorder to optimize the materials growth Within the methane CVD approachfor SWNT growth, we have found that the chemical and textural proper-ties of the catalyst materials dictate the yield and quality of SWNTs Thisunderstanding has allowed optimization of the catalyst material and thusthe synthesis of bulk quantities of high yield and quality SWNTs [22] Wehave developed a catalyst consisting of Fe/Mo bimetallic species supported

on a sol-gel derived alumina-silica multicomponent material The catalyst hibits a surface are of approximately 200 m2/g and mesopore volume of 0.8mL/g Shown in Fig 4 are Transmission Electron Microscopy (TEM) andScanning Electron Microscopy (SEM) images of SWNTs synthesized withbulk amounts of this catalyst under a typical methane CVD growth condi-tions for 15 min (methane flow rate = 1000 mL/min through a 1 inch quartztube reactor heated to 900◦C) The data illustrates remarkable abundance

ex-of individual and bundled SWNTs Evident from the TEM image is that thenanotubes are free of amorphous carbon coating throughout their lengths.The diameters of the SWNTs are dispersed in the range of 0.7–3 nm with apeak at 1.7 nm Weight gain studies found that the yield of nanotubes is up

to 45 wt.% (1 gram of catalyst yields 0.45 gram of SWNT)

Catalyst optimization is based on the finding that a good catalyst terial for SWNT synthesis should exhibit strong metal-support interactions,possess a high surface area and large pore volume Moreover, these texturalcharacteristics should remain intact at high temperatures without being sin-tered [22] Also, it is found that alumina materials are generally far superiorcatalyst supports than silica The strong metal-support interactions allowhigh metal dispersion and thus a high density of catalytic sites The inter-actions prevent metal-species from aggregating and forming unwanted largeparticles that could yield to graphitic particles or defective multi-walled tubestructures High surface area and large pore volume of the catalyst supportfacilitate high-yield SWNT growth, owing to high densities of catalytic sites

ma-36 Hongjie Dai

Fig 4 Bulk SWNT

ma-terials grown by ical vapor deposition of

chem-methane (a) A low

mag-nification TEM image

(b) A high

magnifica-tion TEM image (c)

An SEM image of theas-grown material

made possible by the former and rapid diffusion and efficient supply of carbonfeedstock to the catalytic sites by the latter

Mass-spectral study of the effluent of the methane CVD system has beencarried out in order to investigate the molecular species involved in the nano-tube growth process [25] Under the typical high temperature CVD growthcondition, mass-spectral data (Fig.1.2.2) reveals that the effluent consists

of mostly methane, with small concentrations of H2, C2 and C3 bon species also detected However, measurements made with the methanesource at room temperature also reveals similar concentrations of H2and C2–

hydrocar-C3 species as in the effluent of the 900◦C CVD system This suggests thatthe H2 and C2–C3 species detected in the CVD effluent are due to impuri-ties in the methane source being used Methane in fact undergoes negligibleself-pyrolysis under typical SWNT growth conditions Otherwise, one would

Nanotube Growth and Characterization 37

Fig 5 Mass spectrum

recorded with the fluent of the methaneCVD system at 900◦C

ef-observe appreciable amounts of H2and higher hydrocarbons due to methanedecomposition and reactions between the decomposed species This result isconsistent with the observation that the SWNTs produced by methane CVDunder suitable conditions are free of amorphous carbon over-coating.The methane CVD approach is promising for enabling scale-up of defect-free nanotube materials to the kilogram or even ton level A challenge istwheter it is possible to enable 1 g of catalyst producing 10, 100 g or evenmore SWNTs To address this question, one needs to rationally design andcreate new types of catalyst materials with exceptional catalytic activities,large number of obtain active catalytic sites for nanotube nucleation with agiven amount of catalyst, and learn how to grow nanotubes continuously intomacroscopic lengths

A significant progress was made recently by Liu and coworkers in

obtain-ing a highly active catalyst for methane CVD growth of SWNTs [26] Liuused sol-gel synthesis and supercritical drying to produce a Fe/Mo catalystsupported on alumina aerogel The catalyst exhibits an ultra-high surfacearea (∼ 540 m2

/g) and large mesopore volume (∼ 1.4 mL/g), as a result of

supercritical drying in preparing the catalyst Under supercritical conditions,capillary forces that tend to collapse pore structures are absent as liquid andgas phases are indistinguishable under high pressure Using the aerogel cat-alyst, Liuand coworkers were able to obtain∼ 200% yield (1 g of catalyst

yielding 2 g of SWNTs) of high quality nanotubes by methane CVD idently, this is a substantial improvement over previous results, and is anexcellent demonstration that understanding and optimization of the catalystcan lead to scale-up of perfect SWNT materials by CVD

Ev-The growth of bulk amounts of SWNT materials by methane CVD has

been pursued by several groups Rao and coworkers used a catalyst based

on mixed oxide spinels to growth SWNTs [27] The authors found that goodquality and yield of nanotubes were obtainable with FeCo alloy nanoparticles.Colomer and coworkers recently reported the growth of bulk quantities ofSWNTs by CVD of methane using a cobalt catalyst supported on magnesiumoxide [28] They also found that the produced SWNTs can be separated fromthe support material by acidic treatment to yield a product with about 70–80% of SWNTs

38 Hongjie Dai

1.2.3 Growth Mode of Single-Walled Nanotubes in CVD

The states of nanotube ends often contain rich information about nanotubegrowth mechanisms [22] High resolution TEM imaging of the SWNTs syn-thesized by the methane CVD method frequently observed closed tube endsthat are free of encapsulated metal particles as shown in Fig.6 The oppo-site ends were typically found embedded in the catalyst support particleswhen imaged along the lengths of the nanotubes These observations sug-gest that SWNTs grow in the methane CVD process predominantly via thebase-growth process as depicted in figure 3 [10,14,16,22,29] The first step

of the CVD reaction involves the absorption and decomposition of methanemolecules on the surface of transition-metal catalytic nanoparticles on thesupport surface Subsequently, carbon atoms dissolve and diffuse into thenanoparticle interior to form a metal-carbon solid state solution Nanotubegrowth occurs when supersaturation leads to carbon precipitation into a crys-talline tubular form The size of the metal catalyst nanoparticle generally dic-tates the diameter of the synthesized nanotube In the base-growth mode, thenanotube lengthens with a closed-end, while the catalyst particle remains onthe support surface Carbon feedstock is thus supplied from the ‘base’ wherethe nanotube interfaces with the anchored metal catalyst Base-growth oper-ates when strong metal-support interactions exist so that the metal speciesremain pinned on the support surface In contrast, in the tip-growth mecha-nism, the nanotube lengthening involves the catalyst particle lifted off fromthe support and carried along at the tube end The carried-along particle isresponsible for supplying carbon feedstock needed for the tube growth Thismode operates when the metal-support interaction is weak [22]

In the methane CVD method, we have found that enhancing support interactions leads to significant improvement in the performance ofthe catalyst material in producing high yield SWNTs [22] This is due tothe increased catalytic sites that favor base-mode nanotube growth On the

metal-Fig 6 a,b TEM images of the ends of SWNTs grown by CVD

Nanotube Growth and Characterization 39

other hand, catalysts with weak metal-support interactions lead to tion of metal species and reduced nanotube yield and purity Large metalparticles due to the aggregation often lead to the growth of multi-layeredgraphitic particles or defective multi-walled tube structures Metal-supportinteractions are highly dependent on the type of support materials and thetype of metal precursor being used in preparing the catalyst [22]

aggrega-1.3 Gas Phase Catalytic Growth

It has also been demonstrated that catalytic growth of SWNTs can be grown

by reacting hydrocarbons or carbon monoxide with catalyst particles

gener-ated in-situ Cheng and coworkers reported a method that employs benzene

as the carbon feedstock, hydrogen as the carrier gas, and ferrocene as thecatalyst precursor for SWNT growth [30] In this method, ferrocene is va-porized and carried into a reaction tube by benzene and hydrogen gases.The reaction tube is heated at 1100–1200◦C The vaporized ferrocene de-composes in the reactor, which leads to the formation of iron particles thatcan catalyze the growth of SWNTs With this approach however, amorphouscarbon generation could be a problem, as benzene pyrolysis is expected to

be significant at 1200◦ C Smalley and coworkers has developed a gas phase

catalytic process to grow bulk quantities of SWNTs [31] The carbon stock is carbon monoxide (CO) and the growth temperature is in the range

feed-of 800–1200◦C Catalytic particles for SWNT growth are generated in-situ

by thermal decomposition of iron pentacarbonyl in a reactor heated to thehigh growth temperatures Carbon monoxide provides the carbon feedstockfor the growth of nanotubes off the iron catalyst particles CO is a very stablemolecule and does not produce unwanted amorphous carbonaceous material

at high temperatures However, this also indicates that CO is not an efficientcarbon source for nanotube growth To enhance the CO carbon feedstock,Smalley and coworkers have used high pressures of CO (up to 10 atm) tosignificantly speed up the disproportionation of CO molecules into carbon,and thus enhance the growth of SWNTs The SWNTs produced this wayare as small as 0.7 nm in diameter, the same as that of a C60 molecule Theauthors have also found that the yield of SWNTs can be increased by intro-ducing a small concentration of methane into their CO high pressure reactor

at 1000–1100◦C growth temperatures Methane provides a more efficient bon source than CO and does not undergo appreciable pyrolysis under theseconditions The high pressure CO catalytic growth approach is promising forbulk production of single-walled carbon nanotubes

by Chemical Vapor Deposition

Recent interest in CVD nanotube growth is also due to the idea that alignedand ordered nanotube structures can be grown on surfaces with control that

is not possible with arc-discharge or laser ablation techniques [23,32]

40 Hongjie Dai

2.1 Aligned Multi-Walled Nanotube Structures

Methods that have been developed to obtain aligned multi-walled nanotubesinclude CVD growth of nanotubes in the pores of mesoporous silica, an ap-

proach developed by Xie’s group at the Chinese Academy of Science [33,34].The catalyst used in this case is iron oxide particles created in the pores

of silica, the carbon feedstock is 9% acetylene in nitrogen at an overall 180torr pressure, and the growth temperature is 600◦C Remarkably, nanotubeswith lengths up to millimeters are made (Fig.7a) [34] Ren has grown rel-

atively large-diameter MWNTs forming oriented ‘forests’ (Fig.7b) on glasssubstrates using a plasma assisted CVD method with nickel as the catalystand acetylene as the carbon feedstock around 660◦C [35]

Our group has been devising growth strategies for ordered multi-walledand single-walled nanotube architectures by CVD on catalytically patternedsubstrates [23,32] We have found that multi-walled nanotubes can self-assem-ble into aligned structures as they grow, and the driving force for self-align-ment is the Van der Waals interactions between nanotubes [36] The growthapproach involves catalyst patterning and rational design of the substrate

to enhance catalyst-substrate interactions and control the catalyst particle

Fig 7 Aligned multi-walled

nano-tubes grown by CVD methods

(a) An ultra-long aligned nanotube bundle (courtesy of S Xie) (b) An

oriented MWNT forest grown onglass substrate (courtesy of Z Ren)

Nanotube Growth and Characterization 41

size Porous silicon is found to be an ideal substrate for this approach andcan be obtained by electrochemical etching of n-type silicon wafers in hy-drofluoric acid/methanol solutions The resulting substrate consisted of athin nanoporous layer (pore size∼ 3 nm) on top of a macroporous layer (with

submicron pores) Squared catalyst patterns on the porous silicon substrateare obtained by evaporating a 5 nm thick iron film through a shadow maskcontaining square openings CVD growth with the substrate is carried out

at 700◦C under an ethylene flow of 1000 mL/min for 15 to 60 min Figure8shows SEM images of regularly positioned arrays of nanotube towers grownfrom patterned iron squares on a porous silicon substrates The nanotubetowers exhibit very sharp edges and corners with no nanotubes branchingaway from the blocks The high resolution SEM image (Fig.8c) reveals thatthe MWNTs (Fig.8c inset) within each block are well aligned along the direc-tion perpendicular to the substrate surface The length of the nanotubes andthus the height of the towers can be controlled in the range of 10–240µm byvarying the CVD reaction time The width of the towers is controlled by thesize of the openings in the shallow mask The smallest self-oriented nanotubetowers synthesized by this method are 2µm × 2µm.

The mechanism of nanotube self-orientation involves the nanotube growth mode [36] Since the nanoporous layer on the porous silicon substrateserves as an excellent catalyst support, the iron catalyst nanoparticles formed

base-on the nanoporous layer interact strbase-ongly with the substrate and remainpinned on the surface During CVD growth, the outermost walls of nano-tubes interact with their neighbors via van der Waals forces to form a rigidbundle, which allows the nanotubes to grow perpendicular to the substrate(Fig.8d) The porous silicon substrates exhibit important advantages overplain silicon substrates in the synthesis of self-aligned nanotubes Growth onsubstrates containing both porous silicon and plain silicon portions finds thatnanotubes grow at a higher rate (in terms of length/min) on porous siliconthan on plain silicon This suggests that ethylene molecules can permeatethrough the macroporous silicon layer (Fig.8d) and thus efficiently feed thegrowth of nanotubes within the towers The nanotubes grown on poroussilicon substrates have diameters in a relatively narrow range since cata-lyst nanoparticles with a narrow size distribution are formed on the poroussupporting surface, and the metal-support interactions prevent the catalyticmetal particles from sintering at elevated temperatures during CVD

2.2 Directed Growth of Single-Walled Nanotubes

Ordered, single-walled nanotube structures can be directly grown by methaneCVD on catalytically patterned substrates A method has been devised togrow suspended SWNT networks with directionality on substrates contain-ing lithographically patterned silicon pillars [25,37] Contact printing is used

to transfer catalyst materials onto the tops of pillars selectively CVD of

42 Hongjie Dai

Fig 8 Self-oriented MWNT arrays grown by CVD on a catalytically patterned

porous silicon substrate (a) SEM image of tower structures consisted of aligned nanotubes (b) SEM image of the side view of the towers (c) A high magnification

SEM image showing aligned nanotubes in a tower Inset: TEM image showing two

MWNTs bundling together (d) Schematic diagram of the growth process

methane using the substrates leads to suspended SWNTs forming nearly dered networks with the nanotube orientations directed by the pattern of thepillars (Fig.9)

or-The growth approach starts with developing a liquid-phase catalyst cursor material that has the advantage over solid-state supported catalysts inallowing the formation of uniform catalyst layers and for large-scale catalyticpatterning on surfaces [25,37] The precursor material consists of a triblockcopolymer, aluminum, iron and molybdenum chlorides in mixed ethanol andbutanol solvents The aluminum chloride provides an oxide framework whenoxidized by hydrolysis and calcination in air The triblock copolymer directsthe structure of the oxide framework and leads to a porous catalyst structureupon calcination The iron chloride can lead to catalytic particles needed forthe growth of SWNTs in methane CVD The catalyst precursor material isfirst spun into a thin film on a poly-dimethyl siloxane (PDMS) stamp, fol-lowed by contact-printing to transfer the catalyst precursor selectively onto

pre-Nanotube Growth and Characterization 43

Fig 9 Directed growth of suspended SWNT (a) A nanotube power-line structure.

(b) A square of nanotubes (c) An extensive network of suspended SWNTs

the tops of pillars pre-fabricated on a silicon substrate The stamped strate is calcined and then used in CVD growth

sub-Remarkably, the SWNTs grown from the pillar tops tend to direct frompillar to pillar Well-directed SWNT bridges are obtained in an area of thesubstrate containing isolated rows of pillars as shown in Fig.9a, where sus-pended tubes forming a power-line like structure can be seen In an areacontaining towers in a square configuration, a square of suspended nanotubebridges is obtained (Fig.9b) Suspended SWNTs networks extending a largearea of the substrate are also formed (Fig.9c) Clearly, the directions of theSWNTs are determined by the pattern of the elevated structures on the sub-strate Very few nanotubes are seen to extend from pillars and rest on thebottom surface of the substrate

The self-directed growth can be understood by considering the SWNTgrowth process on the designed substrates [25] Nanotubes are nucleated only

on the tower-tops since the catalytic stamping method does not place any alyst materials on the substrate below As the SWNTs lengthen, the methaneflow keeps the nanotubes floating and waving in the ‘wind’ since the flow ve-locity near the bottom surface is substantially lower than that at the level ofthe tower-tops This prevents the SWNTs from being caught by the bottomsurface The nearby towers on the other hand provide fixation points for thegrowing tubes If the waving SWNTs contact adjacent towers, the tube-towervan der Waals interactions will catch the nanotubes and hold them aloft

cat-44 Hongjie Dai

Growing defect-free nanotubes continuously to macroscopic lengths hasbeen a challenging task In general, continuous nanotube growth requiresthe catalytic sites for SWNTs remaining active indefinitely Also, the carbonfeedstock must be able to reach and feed into the catalytic sites in a continu-ous fashion For CVD growth approaches, this means that catalyst materialswith high surface areas and open pore structures can facilitate the growth oflong nanotubes, as discussed earlier

We have found that within the methane CVD approach, small trations (∼0.03%) of benzene generated in-situin the CVD system can lead

concen-to an appreciable increase in the yield of long nanotubes Individual SWNTswith length up to 0.15 mm (150µm) can be obtained [25] An SEM image of

an approximately 100µm long tube is shown in Fig.10 The SWNTs appearcontinuous and strung between many pillars In-situ generation of benzene

is accomplished by catalytic conversion of methane in the CVD system ing bulk amounts of alumina supported Fe/Mo catalyst [25] This leads toenhanced SWNT growth, and a likely explanation is that at high tempera-tures, benzene molecules are highly reactive compared to methane, thereforeenhancing the efficiency of carbon-feedstock in nanotube growth However,

us-we find that when high concentrations of benzene are introduced into theCVD growth system, exceedingly low yields of SWNTs result with virtually

no suspended tubes grown on the sample This is because high tions of benzene undergo extensive pyrolysis under the high temperatureCVD condition, which causes severe catalyst poisoning as amorphous carbon

concentra-is deposited on the catalytic sites therefore preventing SWNT growth

Fig 10 SEM image of a CVD grown approximately 100m long SWNT strungbetween silicon pillars

2.3 Growth of Isolated Nanotubes on Specific Surface Sites

A CVD growth strategy has been developed collaboratively by our group

and Quate to grow individual single-walled nanotubes at specific sites on flat

silicon oxide substrates [20] The approach involves methane CVD on strates containing catalyst islands patterned by electron beam lithography

sub-‘Nanotube chips’ with isolated SWNTs grown from the islands are obtained.Atomic Force Microscopy (AFM) images of SWNTs on such nanotube-chipsare shown in Fig.10, where the synthesized nanotubes extending from thecatalyst islands are clearly observed This growth approach readily leads

Nanotube Growth and Characterization 45

to SWNTs originating from well controlled surface sites, and have enabled

us to develop a controlled method to integrate nanotubes into addressablestructures for the purpose of elucidating their fundamental properties andbuilding devices with interesting electrical, electromechanical and chemicalfunctions [23,38,39,40,41,42,43,44]

2.4 From Growth to Molecular-Wire Devices

Integrating individual nanotubes into addressable structures is important

to the characterization of nanotubes It is necessary to investigate ual tubes because the properties of nanotubes are highly sensitive to theirstructural parameters, including chirality and diameter Currently, all of thegrowth methods yield inhomogeneous materials containing nanotubes withvarious chiralities Therefore, measurements of ensembles of nanotubes canonly reveal their bulk averaged properties

individ-2.4.1 Electrical Properties of Individual Nanotubes

Previous approaches to individual SWNT electrical devices include randomlydepositing SWNTs from liquid suspensions onto pre-defined electrodes [4,45],

or onto a flat substrate followed by locating nanotubes and patterning trodes [46,47,48,49] We have demonstrated that controlled nanotube growthstrategies open up new routes to individually addressable nanotubes TheSWNTs grown from specific sites on substrates can be reliably contacted byelectrodes (Figs.11c,12a) and characterized [38] Metal electrodes are placedonto the two ends of a nanotube via lithography patterning and electronbeam evaporation Detailed procedure for contacting a SWNT can be found

elec-in [38]

The formation of low ohmic contacts with SWNTs is critical to elucidatingtheir intrinsic electrical properties [40] and building devices with advancedcharacteristics This is accomplished by our controlled approach of growingand contacting nanotubes We found that titanium metal contacts give rise

to the lowest contact resistance compared to other metals Metallic SWNTsthat are several microns long typically exhibit two-terminal resistance onthe order of tens to hundreds of kilo-ohms The lowest single-tube resistancemeasured with our individual metallic SWNT is∼ 12 k Ω (Fig.12b) The lowcontact resistance achieved in our system can be attributed to several fac-tors The first is that our method allows the two metal electrodes to contactthe two ends of a nanotube Broken translational symmetry at the nanotubeends could be responsible for the strong electrical coupling between the tubeand metal [50] Secondly, titanium–carbon (carbide) bond formation at themetal–tube interface may have occurred during the electron-beam evapora-tion process This is based on our result that titanium yields lower contactresistance than other metals, including aluminum and gold, and the fact that

46 Hongjie Dai

Fig 11 Growth of single-walled nanotubes on controlled surface sites and device

integration (a) An atomic force microscopy image of SWNTs grown from patterned catalyst dots (while spots) (b) AFM image of a SWNT grown between two catalyst sites (white corners) (c) AFM image of a SWNT grown from a catalyst island and

contacted by metal electrodes

aluminum and gold do not form strong bonding with carbon and stable bide compounds

car-For individual semiconducting SWNTs grown on surfaces, relatively lowresistance devices on the order of hundreds of kilo-ohms can be made byour approach [40] These nanotubes exhibit p-type transistor behavior at

room temperature as their conductance can be dramatically changed bygate voltages (Fig.12c) This property is consistent with the initial observa-

tion made with high resistance samples (several mega-ohms) by Dekker [51]

and by Avouris’ group [52] The transconductance (ratio of current changeover gate-voltage change) of our semiconducting tube samples can be up to

∼200 nA/V [40] which is two orders of magnitude higher than that measuredwith earlier samples The high transconductance is a direct result of the rela-tively low resistance of our semiconducting SWNT samples, as high currentscan be transported through the system at relatively low bias voltages Thisresult should not be underestimated, given the importance of high transcon-ductance and voltage gain to transistors Nevertheless, future work is needed

in order to create semiconducting SWNT devices with transconductance andvoltage gains that match existing silicon devices