applied physics reviews vertically aligned carbon...

APPLIED PHYSICS REVIEWS

Vertically aligned carbon nanofibers and related structures:Controlled synthesis and directed assembly

A. V. MelechkoMolecular-Scale Engineering and Nanoscale Technologies Research Group, Oak Ridge NationalLaboratory, P.O. Box 2008, MS 6006, Oak Ridge, Tennessee 37831-6006 and Materials Scienceand Engineering Department, University of Tennessee, Knoxville, Tennessee 37996-2200

V. I. Merkulov and T. E. McKnightMolecular-Scale Engineering and Nanoscale Technologies Research Group, Oak Ridge NationalLaboratory, P.O. Box 2008, MS 6006, Oak Ridge, Tennessee 37831-6006

M. A. GuillornCornell Nanoscale Science and Technology Facility, Cornell University, Ithaca, New York 14853-2000

K. L. KleinMolecular-Scale Engineering and Nanoscale Technologies Research Group, Oak Ridge NationalLaboratory, P.O. Box 2008, MS 6006, Oak Ridge, Tennessee 37831-6006 and Materials Scienceand Engineering Department, University of Tennessee, Knoxville, Tennessee 37996-2200

D. H. LowndesThin Film and Nanostructured Materials Physics Group, Oak Ridge National Laboratory, Oak Ridge,Tennessee 37831-6056 and Center for Nanophase Materials Science, Oak Ridge National Laboratory,Oak Ridge, Tennessee 37831-6056

M. L. Simpsona!

Molecular-Scale Engineering and Nanoscale Technologies Research Group, Oak Ridge NationalLaboratory, P.O. Box 2008, MS 6006, Oak Ridge, Tennessee 37831-6006, Materials Scienceand Engineering Department, University of Tennessee, Knoxville, Tennessee 37996-2200, and Centerfor Nanophase Materials Science, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6056

sReceived 10 September 2004; accepted 17 December 2004; published online 3 February 2005d

The controlled synthesis of materials by methods that permit their assembly into functionalnanoscale structures lies at the crux of the emerging field of nanotechnology. Although only one ofseveral materials families is of interest, carbon-based nanostructured materials continue to attract adisproportionate share of research effort, in part because of their wide-ranging properties.Additionally, developments of the past decade in the controlled synthesis of carbon nanotubes andnanofibers have opened additional possibilities for their use as functional elements in numerousapplications. Vertically aligned carbon nanofiberssVACNFsd are a subclass of carbonnanostructured materials that can be produced with a high degree of control using catalyticplasma-enhanced chemical-vapor depositionsC-PECVDd. Using C-PECVD the location, diameter,length, shape, chemical composition, and orientation can be controlled during VACNF synthesis.Here we review the CVD and PECVD systems, growth control mechanisms, catalyst preparation,resultant carbon nanostructures, and VACNF properties. This is followed by a review of many of theapplication areas for carbon nanotubes and nanofibers including electron field-emission sources,electrochemical probes, functionalized sensor elements, scanning probe microscopy tips,nanoelectromechanical systemssNEMSd, hydrogen and charge storage, and catalyst support. Weend by noting gaps in the understanding of VACNF growth mechanisms and the challengesremaining in the development of methods for an even more comprehensive control of the carbonnanofiber synthesis process. ©2005 American Institute of Physics. fDOI: 10.1063/1.1857591g

TABLE OF CONTENTS

I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . 2A. Historical background. . . . . . . . . . . . . . . . . . . . 2B. Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2C. Overview of applications of vertically

aligned carbon nanofibers. . . . . . . . . . . . . . . . . 4II. SYNTHESIS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

A. Methods of synthesis of carbon materials. . . . 5B. Catalytic thermal chemical-vapor deposition.. 5

1. Thermal CVD and catalytic thermalCVD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2. C-TCVD growth mechanisms. . . . . . . . . . . 5

JOURNAL OF APPLIED PHYSICS97, 041301s2005d

0021-8979/2005/97~4!/041301/39/$22.50 © 2005 American Institute of Physics97, 041301-1

Downloaded 04 Feb 2005 to 128.219.65.64. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

http://dx.doi.org/10.1063/1.1857591

C. Catalytic plasma-enhanced chemical-vapordeposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71. Introduction to PECVD. . . . . . . . . . . . . . . . 72. C-PECVD experimental systems and

results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83. C-PECVD specific aspects of carbon

nanostructure growth control. . . . . . . . . . . . 12D. Substrates and catalysts. . . . . . . . . . . . . . . . . . 17

1. Substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . 172. Catalysts. . . . . . . . . . . . . . . . . . . . . . . . . . . . 183. Methods of catalyst preparation. . . . . . . . . 184. Catalyst particle evolution. . . . . . . . . . . . . . 21

III. PROPERTIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22A. Structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22B. Electrical properties. . . . . . . . . . . . . . . . . . . . . 24C. Electron field emission. . . . . . . . . . . . . . . . . . . 24D. Mechanical properties. . . . . . . . . . . . . . . . . . . . 26E. Chemical properties. . . . . . . . . . . . . . . . . . . . . 27F. Electrochemical properties. . . . . . . . . . . . . . . . 28G. Postsynthesis processing. . . . . . . . . . . . . . . . . . 29

IV. APPLICATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31A. Microfabricated field-emission electron

sources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31B. Microfabricated x-ray sources. . . . . . . . . . . . . 32C. Electrochemical probes. . . . . . . . . . . . . . . . . . . 32D. Intracellular gene delivery devices. . . . . . . . . 33E. Scanning probe microscopy tips. . . . . . . . . . . 33F. Nanoelectromechanical devicessNEMSd. . . . 34G. Nanoporous membranes. . . . . . . . . . . . . . . . . . 34H. Other applications. . . . . . . . . . . . . . . . . . . . . . . 34

V. FUTURE DEVELOPMENTS ANDCONCLUSIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

I. INTRODUCTION

A. Historical background

It has been known for over a century that filamentouscarbon can be formed by the catalytic decomposition ofcarbon-containing gas on a hot metal surface. In a U.S.Patent published in 1889,1 it is reported that carbon filamentsare grown from carbon-containing gases using an iron cru-cible. It is probable that this early material consisted ofnanofibers arranged in larger filaments; however, the actualobservation of discrete nanofibers had to await the develop-ment of high-resolution electron microscopy. Radushkevichand Lukyanovich obtained transmission electron micro-graphs of nanofibers, or “soot particles,” in 1952.2 Works inthe 1950s established that carbon filaments could be pro-duced by the interaction of a wide range of hydrocarbons andother gases with metals, the most effective of which wereiron, cobalt, and nickel. Early detailed studiess1950sthrough 1970sd of carbon nanofiber growth were promptedby undesirable formation of carbon deposits in industrialsteam cracker tubes used to produce a variety of olefins,3 andin the nuclear industry in which use of gases containing hy-drocarbons or CO for cooling and heat transfer created car-bon deposits on fuel cladding.4 It should be noted that in the

current literature, the term “nanofiber” is preferentially used,featuring distinction in size scale, while in the past simply“filamentous carbon,” “carbon filaments,” and “carbon whis-kers” were applied.4 In 1985 a form of carbon, buckminster-fullerene C60, was observed by a team headed by Krotoetal.,5 which led to the Nobel Prize in chemistry in 1997. Thisdiscovery was followed by IIijima’s demonstration in 1991that carbon nanotubes are formed during arc-discharge syn-thesis of C60 sRef. 6d and other fullerenes, triggering a delugeof interest in carbon nanofibers and nanotubes. In the 1990sthe introduction of catalytic plasma-enhanced chemical-vapor depositionsC-PECVDd provided additional controlmechanisms over the growth of carbon nanostructures. In1997 Chenet al.used PECVD for nanofiber synthesis.7 Theirwork was followed by the better known work of Renet al.8

The use of C-PECVD ultimately has allowed for the highlydeterministic synthesis of carbon nanostructures,9,10 whereby deterministic synthesis we mean control over position,alignment, diameter, length, chemical composition, or othercharacteristics of individual nanostructures.

There are many excellent reviews devoted to various as-pects of synthesis and properties and applications of carbonnanofibers and carbon nanotubes.11–18 Here we focus on theinterplay between controlled synthesis and the resultant ap-plications of vertically aligned carbon nanofibers. That is, weshow that the control of nanofiber properties during the syn-thesis process provides the means for accomplishingbottom-up assembly of microscalesor largerd structures, de-vices, or systems with functional nanoscale elements. Otherclasses of filamentous carbon-based nanomaterials will bereviewed as well since most of the fundamental aspects oftheir synthesis and resulting properties, apart from a few dis-tinctive characteristics, are the same or similar to those ofnanofibers.

B. Definitions

Carbon nanofiberssCNFsd are cylindrical or conicalstructures that have diameters varying from a few to hun-dreds of nanometers and lengths ranging from less than amicron to millimeters. The internal structure of carbonnanofibers varies and is comprised of different arrangementsof modified graphene sheets. A graphene layer can be definedas a hexagonal network of covalently bonded carbon atomsor a single two-dimensionals2Dd layer of a three-dimensionals3Dd graphitefFig. 1sadg. In general, a nanofiberconsists of stacked curved graphite layers that form conesfFig. 1sbdg or “cups.”19,20The stacked cone structure is oftenreferred to as herringbonesor fishboned as their cross-sectional transmission electron micrographssTEMd resemblea fish skeleton, while the stacked cups structure is most oftenreferred to as a bamboo type, resembling the compartmental-ized structure of a bamboo stem. Currently there is no strictclassification of nanofiber structures. The main distinguish-ing characteristic of nanofibers from nanotubes is the stack-ing of graphene sheets of varying shapes.

We can definea as an angle between the fiber axis andthe graphene sheet near the sidewall surfacefFig. 1sbdg. Thenanofiber witha=0 is a special case in which one or moreadElectronic mail: [email protected]

041301-2 Melechko et al. J. Appl. Phys. 97, 041301 ~2005!


graphene layers form cylinders that run the full length of thenanostructurefFig. 1scdg. This arrangement, with its closedand semi-infinite surface,6 results in extraordinary propertiesthat made this type of a nanofiber known to the world as acarbon nanotubesCNTd. However, one cannot expect thesame mechanical, electrical, and chemical properties fornanofibers with a nonzeroa. In this case the nanostructure isassembled from graphene layers that are relatively short andpoorly connected, spanning only a part of the nanofiber. Forexample, similar to graphite, charge transport along thenanofiber witha.0 is determined by both “in-plane” and“interplane” components that are quite different. The same istrue for mechanical properties since the van der Waals bond-ing between the graphene planes differs drastically from thein-plane covalent bonding of true nanotubes. Similarly, thechemical properties of nanofibers and nanotubes are quitedifferent since defect-free nanotube walls do not contain theexposed edges and unsaturated bonds of graphene planes,and as a result nanotubes are far less reactive than nanofi-bers. TEM images displaying nanofibers with various struc-tures are shown in Fig. 2. TEM images of nanotubes arepresented in Fig. 3. The clearly visible fringes correspond tographene layers and run parallel to the axis of nanotubesthroughout their entire length.

Although perhaps not completely warranted by utility,both the scientific community and the popular press havefocused more on the special case of the CNT structure thanthe more general case that includes nanofibers. The reasonfor the attention given to nanotubes is clear: the beauty andperfection of their mathematical description and the resultingextraordinary mechanical andsballisticd electron-transportproperties. However, the distinctive properties of nanotubes

call for a distinction in terminology that is not always madein literature. Instead, there is a tendency to call all thesematerials nanotubes, perhaps due to greater name recogni-tion. For consistency and accuracy in this review, we refer tostructures as nanotubes only if the graphene layers form cyl-inders that extend throughout the length of the material. Ifinstead the graphene sheets form cones, cups, etc., we referto them as nanofibers, even if that differs from the term usedin a particular referenced worksif sufficient data are pro-videdd. Some authors, in order to distinguish this type ofcarbon nanostructure, refer to them as graphitic nanofibers21

sGNFd or multiwalled carbon nanofiberssMWNFsd.22 Theterm nanostructure will be used in cases when some particu-lar aspect is common to both nanotubes and other types ofnanofibers.

Structures that are oriented perpendicular to the substrateon which they are grown are commonly called verticallyaligned. In the case of a nanofiber or nanotube this meansthat the structure is straight and its axis is normal to thesubstratesFig. 4d. In contrast, synthesis methods such as arcdischarge, laser ablation, and certain types of chemical-vapor

FIG. 1. Schematic structure of carbon nanofibers and nanotubes.sadGraphene layer,sbd stacked conesherringboned nanofiber, andscd nanotube.

FIG. 2. sad STEM image of a herringbone carbon nanofiber produced by dcC-PECVD with Ni catalyst;sbd TEM image of a bamboo-type carbonnanofiber grown at the same conditions with Fe catalyst.

FIG. 3. TEM image of bundles of single-walled carbon nanotubes. Inset:TEM image of multiwalled carbon nanotubesscourtesy of D. B. Geoheganand H. Cuid.



deposition with floating catalyst produce nonaligned, en-tangled ropes of nanotubes. The CNT material produced thisway can be used for a variety of applications such as rein-forcement of composite materials, hydrogen storage, catalystsupport, etc. However, for many applications that utilizenanofibers as individual elements, nanoelectronics, forexample,23 tedious manipulations on a microscale are oftenrequired if such entangled material is to be used. Accord-ingly, synthesis of aligned structures at defined positions on asubstrate is preferred. Vertically aligned carbon nanofibersand carbon nanotubes can be prepared as a dense mats“for-est”d or as individual freestanding structuressoften arrangedinto a regular arrayd, as can be seen in Fig. 5. Vertical align-ment in very dense forests is provided by van der Waalsinteraction between nanotubes or nanofibers that keeps themaligned, while isolated elements may be aligned by interac-tion of the growing nanofiber with an electric field.24

Throughout this article we will refer to deterministicsynthesis as a method of growing individual nanostructureswith precisely defined characteristics, such as size, location,chemical composition, internal structure, etc. Recently, it hasbeen demonstrated that carbon nanofibers can be synthesizeddeterministically by PECVD.9,10 In this synthesis process,the location of the VACNF is defined by patterning the cata-lyst material using photo- or electron-beam lithography withthe size of the catalyst nanoparticle controlling the resultantnanofiber diameter, the nanofiber length is controlled by thegrowth rate and duration of the growth process, the align-ment is controlled by the electric field present in the plasmasheath, and the sidewall chemical composition is defined bythe selection of gas composition, substrate materials, andplasma power used during synthesis. In spite of this highdegree of control, command over other VACNF attributes,such as specific internal crystalline structure, remains elu-sive. In particular, the ability to grow either individual free-standing vertically aligned nanofibers or single verticallyaligned nanotubes by selecting appropriate growth condi-tions would represent a major advance in this field. Whilesome fundamental questions remain unanswered, the tech-nology of VACNF synthesis has matured to the point that itcan be used as a standard processing step in complex devicefabrication.25–35

C. Overview of applications of vertically alignedcarbon nanofibers

There are many potential applications of carbon nano-structures which can be subdivided into two categories. Thefirst category includes “bulk material” applications, in whichcarbon nanostructures are utilized as filler in composite ma-terials, as catalyst support, for hydrogen storage, as elec-trodes for fuel cells, as supercapacitors, or for ultrafiltrationmembranes.36 For these applications the properties of indi-vidual nanostructures are controlled to produce a functionalensemble. The second category includes applications inwhich each nanotube or nanofiber is used as an individualfunctional element of a device. For these applications it ishighly important that the nanostructure is a high aspect ratio,mechanically and chemically robust conductor of electronsthat can be deterministically produced on any substrate. Theuse of large-scale growth reactors created the opportunity tosynthesize high-quality VACNFs in precisely defined loca-tions on substrates compatible with microelectronic devicemanufacturing equipmentse.g., 100-mm-diameter Si andquartz wafersd. These applications include field-emission de-vices such as field-emission displays and perhaps nanoscaleelectron sources for electron microscopy and electron-beamlithography, scanning probe tips and functionalized scanningprobes, electrochemical probes, gene delivery arrays, mem-branes for microfluidic devices, and nanoelectronic devices.Some of these applications are discussed in Sec. IV ingreater detail.

FIG. 4. SEM image of a VACNF produced by dc-PECVD in ammonia/acetylene plasma at 700 °C. The image is aquired at a 30° tilt angle, so thatthe apparent height is two times smaller than the actual height of thenanofiber.

FIG. 5. SEM image of regular arrays of freestanding VACNFs and forests ofVACNFs. Catalyst definition was produced by photolithography. VACNFswere grown by dc C-PECVD.



II. SYNTHESIS

A. Methods of synthesis of carbon materials

The methods for synthesizing carbon nanostructures in-clude laser vaporization,5 arc discharge,6,37 catalyticchemical-vapor depositionsC-CVDd, and catalytic plasma-enhanced chemical-vapor depositionsC-PECVDd. While arcdischarge and laser ablation are very efficient methods forproducing high-quality nanotube material in large quantities,they do not offer control over the spatial arrangement of theproduced nanostructures. Complex purification proceduresare also required to remove amorphous carbon particles andentangled catalyst in order to obtain a useful material. At thistime, only C-CVD allows controlled synthesis of carbonnanotubes and nanofibers and only C-PECVD allows deter-ministic synthesis in which the location, alignment, size,shape, and structure of each individual nanofiber are con-trolled during synthesis. The material obtained by thismethod can be considered phase pure, that is, no purificationprocess is necessary. Furthermore, the defined and accessibleposition of the catalyst particle, especially at the tip, makesits optional extraction simplified.38,39 Here we review onlyCVD and PECVD synthesis methods in detail.

B. Catalytic thermal chemical-vapor deposition

1. Thermal CVD and catalytic thermal CVD

Chemical-vapor deposition involves the adsorption, de-sorption, evolution, and incorporation of vapor species at thesurface of a growing film. Since heat is the main energysource for reactions to occur, the CVD process is also oftenreferred to as thermal CVDsTCVDd. Process temperaturesfor TCVD production of carbon nanostructures typically liein the range from 400 to 1000 °C. Catalytic CVDsorC-CVDd differs from CVD by involvement of a catalyst inthe decomposition of vapor species on the catalyst surfaceand requires process temperatures similar to TCVD. Accord-ingly, the catalytically controlled thermal CVD will be ab-breviated as C-TCVD. In addition to thermal excitation,methods include photoexcitation and electrical glow dis-chargesor plasmad.

In a catalytic growth the deposition of carbon usuallyoccurs on one side of the surface of a catalyst particle. Thus,two of the dimensions of the growing “film” are limited bythe size of the particle, while the third dimension is notbound, leading to quasi-one-dimensional growth. In manysynthesis processes, the diameter of the resultant carbon fila-ment is approximately equal to that of the nanoparticle in therange from 5 to 500 nm. However, there is evidence that, forexample, single-walled carbon nanotubes originate on cata-lyst particles that are significantly larger in size than thenanotubes themselvess1 nmd.

The apparatus for C-TCVD usually consists of a quartztube inside a furnace with a controllable source gas flowsFig. 6d. There are two common methods of introducing thecatalyst into the system: supported catalyst and floating cata-lyst. In the supported catalyst method, the catalyst is depos-ited onto a substrate which is then placed inside the tubefurnacese.g., Ref. 40d. In the floating catalyst method, the

catalyst particles form from a source gas and are not attachedto a substrate. For example, the high-pressure carbon mon-oxide sHiPcod process, which allows a high volume produc-tion of carbon nanotubes, is one such floating catalystmethod.41

C-TCVD has been successfully used to synthesize awhole range of carbon nanostructures. Carbon nanofiber syn-thesis using C-TCVD has been observed since the late1950s.2,3 Recently, C-TCVD has been optimized for thegrowth of multiwalled carbon nanotubessMWCNTsd,42 andeven single-walled carbon nanotubessSWCNTsd.40,43 It isbelieved that SWCNTs are favored in CVD if the catalystparticles are small and the carbon supply is low, with suffi-cient energy in the systemsT.900 °Cd.44 The catalyst par-ticle preparation was found to be crucial in control of thestructure of nanotubes.

By combining an efficient trilayer catalyst thin filmssup-ported catalystd, developed by Delzeitet al.,45 with introduc-tion of ferrocenesfloating catalystd in addition to acetylene,Ereset al. produced dense arrays of vertically aligned mul-tiwalled nanotubes by C-TCVD that are 3.5 mm tall.46 Re-cently Ereset al. reported that after optimization, the maxi-mum length of the nanotube achieved by this method was9.25 mm.47

2. C-TCVD growth mechanisms

The catalytic nature of the carbon filament growth pro-cess was established by Tesner and co-workers48,49 whoshowed that carbon filaments had metal particles associatedwith them. The growth mechanism leading to the formationof carbon nanofibers has been studied by many differentgroups. Bakeret al. usedin situ electron microscopy tech-niques to directly observe the manner by which small metalparticles generated carbon nanofibers during the decomposi-tion of acetylene.50 From an analysis of recorded image se-quences they measured the rates of growth of the materialand determined some of the kinetic parameters involved inthe process. On the basis of these experiments, a growthmechanism was proposed that was later refined to include thefollowing steps:sid adsorption and decomposition of the re-actant hydrocarbon molecule on a surface of catalyst,sii ddissolution and diffusion of carbon species through the metalparticle, andsiii d precipitation of carbon on the opposite sur-face of the catalyst particle to form the nanofiber structure.Figure 7 shows a schematic diagram that illustrates the key

FIG. 6. Schematics of a tube-furnace-type TCVD apparatus.



features of this growth model for a tip-type carbon nanofiberstructure, where precipitation occurs on the bottom surfaceof the catalyst particle, thus elevating the particle, which re-mains at the tip throughout the growth. The chemical natureof the metal catalyst, the reaction temperature, and the com-position of the reactant gas dictate the morphology and de-gree of crystalline perfection exhibited by the carbon nanofi-bers.

The kinetics of the three steps listed above determinesthe growth rate. The supply-limited process depends on therates of arrival of different gas species to the catalyst surface,their adsorption rates, and their respective decompositionrates. It has been argued that diffusion of carbon through themetal catalyst particle is the rate-determining step, as sup-ported by the close agreement between the measured activa-tion energy for nanofiber growth and that for carbon diffu-sion through the respective metals used as catalysts. Initially,the driving force for the bulk diffusion of carbon through themetal particle was ascribed to a temperature gradient,50,51

which was believed to develop due to exothermic reactionsof decomposition on the surface of the catalyst. Later, it wasproposed that concentration gradients drove the carbon dif-fusion through the catalyst particle, and there are severalhypotheses about the processes involved in the formation ofsuch concentration gradient. Nielsen and Trimm proposedthat the carbon solubility at the gas/metal interface differsfrom that at the metal/carbon interface, since the activity ofcarbon in the gas phase may be much higher than one.52

Saccoet al. suggested that the mass flux originates from thesolubility difference between carbon at the alpha-iron/Fe3Cinterface and that between alpha-iron and carbon itself.53

Kock et al. proposed that the driving force for bulk carbondiffusion is the gradient of the carbon content of substoichio-metric carbides, whereby the carbon content decreases in thedirection of the metal/carbon interface.54 Central to themodel of Alstrup is the assumption that the carbon atomsentering the selvedge, which consists of subsurface layersthat differ from the ideal structure of bulk crystal, create a“surface carbide” that forms the source of carbon atoms dif-fusing through the bulk of the metal particle.55 In other

words, the driving force for growth is a concentration gradi-ent of carbon atoms from the metal-gas surface to the metal-graphite surface.56 The validity of these models is still hardto assess due to the lack of directin situ observations.

Recently, Helveget al. presented what was described byAjayan in an introductory article as “a long-awaited solutionto the mystery of nanofiber growth.”57,58 They performedtime-resolved, atomic-resolutionin situ transmission electronmicroscope observations of the formation of carbon nanofi-bers from methane decomposition over supported nickelnanocrystals. Carbon nanofibers were observed to developthrough a reaction-inducedreshapingof the nickel nanopar-ticles. The nucleation and growth of graphene layers werefound to be assisted by a dynamic, repetitive formation andrestructuring of monoatomic step edges at the nickel surface.The authors proposed a mechanism, supported by density-functional theory calculations, that involvessurfacediffusionof carbon and nickel atoms. In their picture, the dissociativemethane adsorption is facilitated at the step edges and Catoms adsorb preferentially at the step sites. The graphenelayer forms at the terrace between the steps on the curvedand dynamically changing catalyst nanoparticle surface. Theprocess involves surface diffusion of C and Ni atoms fromthe step edge, and includes the breaking of the C bond to theNi step on the free surface, incorporation under the graphenesheet, and diffusion at the graphene-Ni interface. Moreovertheir observations and calculations suggest that it is not nec-essary to include the bulk diffusion of C through the Niparticle, however, they do not eliminate such possibility.

a. Tip-type and base-type growth modes.Two differentcarbon nanofiber growth modes have been observed:s1dbase-typefFig. 8sadg, in which the catalyst particle remainson the substrate, ands2d tip-type fFig. 8sbdg, in which thecatalyst is detached from the substrate and remains at the tipof the growing nanostructure. Bakeret al. proposed that thegrowth mode depends on the interaction of the catalyst withits support.3 Such interaction is believed to be related to wet-ting, which can be characterized by the contact angle of thecatalyst with the support surface at a given growth tempera-ture. A large angle corresponds to weak interaction, while asmall angle is indicative of strong interaction. The strengthof interaction must depend on the choice of catalyst materials

FIG. 7. Mechanism of carbon nanofiber formation.sad Adsorption and de-composition of the reactant hydrocarbon molecule on the surface of catalyst,sbd dissolution and diffusion of carbon species through or around the metalparticle, andscd precipitation of carbon on the opposite surface of the cata-lyst particle and incorporation into graphene layers.sThe drawing does notrepresent dynamic effects and transient effects on the initial stages.d.

FIG. 8. Tip-type and base-type growth modes. SEM images ofsad tip-typeand sbd base-type carbon nanofibers produced by dc C-PECVD. The brighttriangles are the catalyst particles that are located at the tips insad and on thesubstrate insbd.



and substrates. For example, it it has been reported that Ni onSiO2 has a large contact angle at 700 °C and thus tip growthis favored in this system.59 In experiments where Co or Fewere deposited on Si,60–62 base-type growth was observed.However, the choice of materials does not completely controlthe growth mode. The kinetics of the growth process has avery strong influence as well. For example, using the samematerialssNi on a thin Ti layer deposited on Sid, growth byPECVD splasmad produces tip-type fibers, while growth byCVD sno plasmad produces base-type fibers. The PECVDprocess itself has a very large parameter space, and it ispossible to choose growth parameters in such a way as toproduce base-type or tip-type fibers for the same catalyst-substrate pair, depending on growth conditions.22,63 We dis-cuss this aspect further in a section devoted to the C-PECVDprocess.

b. Vertical alignment.Vertical alignment conventionallymeans that nanostructures are oriented perpendicular to thesubstrate. A variety of methods for production of alignedarrays of carbon nanostructures in C-CVD has been demon-strated and is reviewed, for example, in Ref. 17. Most fre-quently found in the literature are nanotubes or nanofibersgrown in a very dense arrangement with alignment due to thecrowding effect. Fanet al. used porous silicon substrateswith a catalyst patterned by electron-beam evaporationthrough shadow masks to produce nanotube blocks that grewperpendicular to the substrate.60 The high density of nano-tubes within each block confined the nearest neighbors and

attracted the outermost nanotubes to their neighbors via vander Waals forces, thereby producing oriented growthsFig. 9d.In addition, as described in a model by Liet al., the nano-tubes that initially grow in a direction that is not normal tothe substrate eventually encounter their nearest neighborsand either stop growth or change the growth direction andbecome aligned.64

Another method of aligned growth of nanotubes is to usea nanoporous template. In this method, porous alumina pro-duced in an aluminum anodization process is used as a tem-plate for nanotube growth. The anodic alumina has a porousstructure with nanoporess30–50 nmd self-organized in ahexagonal pattern. Catalyst is electrodeposited into the nan-opores and then nanotubes are synthesized by C-CVD. Theporous template may be etched away after nanotube synthe-sis. An array of aligned nanotubes prepared using this fabri-cation process is shown in Fig. 10.65

C. Catalytic plasma-enhanced chemical-vapordeposition

1. Introduction to PECVD

Plasma-enhanced chemical-vapor depositionsPECVDdis similar to chemical-vapor depositionsCVDd which alsouses gaseous sources.sSee Ref. 66 for a review.d The impor-tant difference is that in CVD thermal energy is used toactivate the gas, whereas in PECVD the molecules are acti-vated by electron impact. The gas activation takes place in anonequilibrium plasma, generally referred to as a glow dis-charge. As in the case of C-CVD, the growth of carbonnanofibers occurs through a catalystsnot by direct surfacedepositiond. The main purpose of using plasma enhancementis to reduce the activation energy for a deposition process.

FIG. 9. Vertically aligned towers of dense nanotubes. Reprinted with per-mission from Ref. 60. Copyright 1999 AAAS. Electron micrographs of self-oriented nanotubes synthesized onn1-type porous silicon substrates.sadSEM image of nanotube blocks synthesized on 2503250 mm2 catalyst pat-terns. The nanotubes are 80 mm long and oriented perpendicular to thesubstratefseesfdg. sbd SEM image of nanotube towers synthesized on 38338 mm2 catalyst patterns. The nanotubes are 130 mm long.scd Side viewof the nanotube towers insbd. The nanotubes self-assemble such that theedges of the towers are perfectly perpendicular to the substrate.sdd Nano-tube “twin towers” a zoom-in view ofscd. sed SEM image showing sharpedges and corners at the top of a nanotube tower.sfd SEM image showingthat nanotubes in a block are well aligned to the direction perpendicular tothe substrate surface.sgd TEM image of pure multiwalled nanotubes inseveral nanotube blocks grown on ann1-type porous silicon substrate. Evenafter ultrasonication for 15 min in 1,2-dichloroethane, the aligned andbundled configuration of the nanotubes is still evident. The inset is a high-resolution TEM image that shows two nanotubes bundling together. Thewell-ordered graphitic lattice fringes of both nanotubes are resolved.

FIG. 10. Vertical alignment using a porous template. Adopted with permis-sion from Ref. 65.sad Schematic of fabrication process.sbd SEM image ofthe resulting hexagonally ordered array of carbon nanotubes fabricated usingthe method insad.



Thus, the first step in the proposed growth model describedin Sec. II B 2, namely, decomposition of carbonaceous spe-cies on the surface of a catalyst, should benefit from plasmaassistance. However, it is not clear how the next step, diffu-sion through the catalyst, would be facilitated by plasma ex-citation. In a recent review, Meyyappanet al. questionedwhether plasma-enhanced CVD provides any facilitation ofthe growth process at all compared to thermal CVD in cata-lytically controlled processes.13 They pointed out that inmost of the reported PECVD growth the temperature was ashigh as in the thermal CVD growth, indicating the absence ofapparent benefits. It has been recognized, however, that oneof the most important and unexpected benefits of PECVDgrowth is the alignment of nanofibers due to interaction withthe electric field.24

However, exploration of PECVD’s potential for growthtemperature reduction is far from over. For many applica-tions low substrate temperature growth of carbon nanostruc-tures is highly desirable. For example, integration with stan-dard semiconductor integrated circuit technology requiresreduced fiber-growth temperature and time, since metal con-tacts to shallow diffusion regions of the substrate can onlywithstand temperatures up to 450 °C for a limited time with-out significant degradation of electronic performance. Simi-larly, for field-emission display production the growth tem-perature must be lower than the soda lime glass substratemelting temperature of 500 °C. Recently, there have beenseveral reports of growth at temperatures as low as roomtemperature.67–70 It must be noted that Boscovicet al. pro-posed that a radio frequencysrfd field could selectively heatthe catalyst particle while the substrate temperature remainslow. If true, this speculation could be one of the most impor-tant revelations in the effort to allow a lower substrate tem-perature during nanofiber growth. All the evidences indicatethat the growth of carbon nanotubes and nanofibersrequiresa catalyst particle temperature high enough to provide a highcarbon diffusion rate. However, there seems to be no suchrequirement for the substrate temperature, as long as growthoccurs at the tip. It has been demonstrated that small metalparticles can be heated in high-frequency fields,71 and mag-netic particles have additional mechanisms for high-frequency field heating.72 This brings to attention the unex-ploited potential of PECVD synthesis and the opportunity forfurther exploration. Certainly caution must be taken in mea-surements of the growth temperature in PECVD experi-ments. Power transfer to the substrate from the generatedelectromagnetic field via the plasma, or simply by resistiveheating by induced current, in addition to dedicated tempera-ture control of the substrate holdersor its absenced, can bevery significant. Plasma heating effects in the case of a dcsystem have been investigated by Teoet al.73

In order to understand the mechanisms involved in car-bon nanofiber formation in a PECVD reactor, we briefly re-view some basic processes that occur in plasmas. In the sim-plest case of a dc diode-type reactor, a dc voltage is appliedacross a space filled with a low-pressure gassa few torrsd.The glow discharge that is initiated can be divided into fourvisible regions arranged from cathode to anode:s1d cathodedark space,s2d negative glow,s3d Faraday dark space, and

s4d positive column. The positive column region is not usedin PECVD processes. The dc discharge is maintained by theprocesses at the cathode and in the dark space. The ions areaccelerated by the applied voltage and some of them bom-bard the cathode. This impact generates secondary electronsthat accelerate away from the cathode. The collisions excitemolecules and energetic electrons ionize some of them. Thenegative glow is the result of this excitation process. Thethickness of the dark space is related to the electron meanfree path.74 The current in the dark space is carried primarilyby ions, while in the negative glow it is carried by electrons.Thus, the negative glow is a low impedance region and theapplied voltage drops mostly over the dark space. Since thedark space varies from a few hundred micrometers to a fewmillimeters, application of several hundred volts can createelectric fields on the order of 104 V/cm.

2. C-PECVD experimental systems and results

A PECVD system consists of a vacuum chamber,vacuum pumps, and a pressure control system; a gas flowcontrol system that includes gas manifolds, mass flow con-trollers, and a showerhead for uniform gas mixing and dis-tribution over the substrate; one or two power supplies forplasma excitation with corresponding power coupling sys-tems; and a substrate heater with a temperature control sys-tem. A variety of plasma sources, successfully used for thedeposition of dielectric thin films for semiconductorcircuits75 and carbon-based thin films,76 has been utilized forCNT and CNF growth. These sources include direct-currentsdc PECVDd, hot-filament dcsHF-dc PECVDd, magnetron-type radio frequencysrf PECVDd, inductively coupledplasmasICP PECVDd, microwavesM-PECVDd, electron cy-clotron resonancesECR PECVDd, hollow cathodesHCGDd,and corona discharge plasma. Examples of different systems,growth conditions, and results are summarized in Table I.

a. Direct-current plasma-enhanced chemical-vapordeposition. In the case of a direct-current PECVD system,the substrate is placed on a substrate heater that serves alsoas a cathodesFig. 11d, making it crucial that the substrate iselectrically conductive.9,24,59,77–82 If growth on insulatingsubstrates is desired, a thin metal film may be deposited un-der the catalyst film to establish contact between the cathodeand the metal surface. The gas showerhead, used to producea uniform gas flow distribution over the entire substrate sur-face, also serves as the anode.

Preparation of the growth catalyst is essential to the pro-duction of isolated VACNFs. Many different catalyst prepa-rations can be found in the literature that include physicalvapor depositionsevaporation or sputteringd, electro- andelectroless plating, and preparing nanoparticles from solu-tion, etc.83 For example, to synthesize isolated VACNFs pat-terned nickel sNid catalyst dots 100 nm in diameter and40 nm thick can be prepared on Si substrates using conven-tional electron-beam lithography.9 A 10-nm-thick Ti layer isdeposited between the Ni catalyst and the Si substrate toprevent nickel silicide formation at the moderately highgrowth temperature of 700 °C. Following the preparation ofthe catalyst, the sample is mounted directly onto a heatedcathodefFig. 12sadg. The vacuum chamber is evacuated be-



fore growth to an acceptable base pressurese.g., 1310−5 Torrd. After reaching the base pressure, ammoniasNH3d is introduced into the chamber, and the sample is pre-treated with the NH3 plasma and temperature. As a result ofthis treatment, discrete catalyst nanoparticles are formedfrom the deposited catalyst dotsfFig. 12sbdg. For Ni catalystdots with the diameter and thickness mentioned above, onlya single nanoparticle is formed from each catalyst dot. These

TABLE I. PECVD growth conditions.

Type ofplasmasystem Frequency

Catalyst/substrate

Growth conditions:F-gas flowsSCCMda, P-pressuresTorrd, E-power sWd, T-temperatures°Cd,

U-substrate biassVd Results References

dc HF 8dc 0 Hz VACNF 121Microwave 2.45 GHz Fmethane=10–20,Fhydrogen=80–90

P=1.8–2.2E=500U=250T=650

Nanofibers 112

Microwave 2.45 GHz Nis30–40nmd/Si

Ffs2%dCH4/ s98%dH2g=undefinedP=0.3

E=undefinedT=900–1000

Nanofiberssnonalignedd

99

Microwave 2.45 GHz Co Ftotal=200,Facetylene/Fammonia=10% –30%P=20

E=1000T=825

Denseforests ofVACNF

61 and 102

ICPsdual powersuppliesd

13.56 MHz Fethylene in H2s20%d=100P=3

EICP=100,Ecapacitive=150T=900

MWCNF 22

ICP sdualpowersupplyd

13.56 MHz Ni Facetylene=20–50,Fhyrdogen=30–150P=5310−2

E=1000T=700

VACNF 98

Coronadischarge

ac snotspecifiedd

CoAAO

117

aDenotes standard cubic centimeter per minute.

FIG. 11. sColor onlined dc PECVD reactor for the growth of verticallyaligned carbon nanofibers.sad High current heater wiring;sbd thermocouplewiring; scd mass flow controllers for acetylene, ammonia, and other gasesse.g., H2d; sdd gas inlet;sed glass cylinder vacuum chamber;sfd gas shower-head and anode;sgd cathode glow of acetylene/ammonia plasma above a100-mm diam Si wafer;shd substrate heater and cathode; andsid pressuretransducer.

FIG. 12. Schematic representation of the PECVD process for growing ver-tically aligned carbon nanofibers.sad Catalyst deposition,sbd catalystpretreatment/nanoparticle formation, andscd growth of carbon nanofibers.



nanoparticles act as the necessary seeds for the catalyticgrowth of isolated VACNFs. After the pretreatment step,acetylenesC2H2d is introduced into the chamber while main-taining the NH3 plasma, immediately initiating the growth ofthe VACNF fFig. 12scdg. The VACNF tip diameter is ap-proximately equal to that of the nanoparticle. Theinitial di-ameter of the nanoparticles,d, is roughly determined frommass conservation of the catalyst, that is, 1 /6pd3

=1/4pD2t or d=s3/2tD2d1/3, whereD is the diameter of thecatalyst dot andt is its thickness. However, it has been ob-served thatd decreases as the growth proceeds.30 The NH3

gas flow in Merkulov’s experiments was maintained at a con-stant level of 80 SCCMsstandard cubic centimeter perminuted, and the C2H2 gas flow was varied to produce dif-ferent gas mixtures. The total gas pressure during growthwas maintained at 3 Torr. Typical scanning electron micro-scopesSEMd and TEM images of the resulting nanofibers areshown in Figs. 2, 4, and 5.

dc PECVD growth reactors possess a number of limita-tions when used in the growth of carbon nanostructures. Therequirement that the substrate be electrically conductive inorder to maintain a glow discharge limits the choice of sub-strate materials. One way around this obstacle is to deposit alayer of metal that, if undesired, can be etched away after thegrowth.84 Alternatively, a radio frequencysrfd plasma sys-tem, in which the polarity of the electrodes changes fastenough to avoid surface charging, can be used. Also, in dcPECVD the power delivered into the plasma and the sub-strate bias are inextricably coupled, which limits the processcontrol. One solution to this was offered by Crudenet al.,involving the introduction of graphite spacers into the elec-trode to vary the power input to the reactor while holding thevoltage constant.85 Plasma instability is another drawback ofa dc discharge. Nonuniformity of the substrate surface maycause the formation of so-called “hot spots” and localizedpower dissipation through arcs instead of the glow discharge.The stability of the plasma depends on the substrate surfaceand the output of secondary electrons forms the substrate.The latter is obviously dependent on the substrate materialand the condition of the substrate surface.

b. Hot-filament direct-current PECVD systems.PECVDgrowth of carbon nanofibers was first done in a HF-dcPECVD reactor in 1997 by Chenet al.7,86 Their work wasfollowed by the more well-known work of Renet al.8 whoalso used a hot-filament PECVD reactor. In HF-dc PECVD,which is commonly used for diamondlike carbon production,a filament is needed for the activation and decomposition ofhydrogen gas into atomic hydrogen.87 Recently, Hashet al.modeled processes in a HF-dc PECVD reactor88,89 andshowed that introducing a tungsten filament in the dcplasma produced a negligible influence on the systemcharacteristics.88 Crudenet al.demonstrated that the filamentwire is important in the pretreatment of the substrate, but hasa minor impact on the resulting nanofibers when combinedwith a dc plasma.85

HF-dc PECVD, as in the case of dc PECVD without ahot filament, allows synthesis of freestanding verticallyaligned carbon nanofibers10 if sufficiently small diametercatalyst patterns are prepared on the substrate. This type of

reactor has been used to study different aspects of carbonnanofiber synthesis such as dependence on growth param-etersspressure, plasma power, gas composition, etc.d, elec-tron field-emission properties of the resulting material,90 ori-entation of alignment control,91 and effects of differentbuffer layers between the catalyst and the substrate.92

c. Radio-frequency (capacitively coupled) PECVD.Direct-current reactors have many drawbacks that include plasmainstability and limitations on the choice of substrate material.In fact, for these reasons, the semiconductor industry hasmoved away from dc PECVD reactors and toward higher-frequency plasmas. There are many commercially availablerf reactors suitable for carbon nanofiber growth. Most ofthese reactors have a radio-frequency source coupled to theplasma via a parallel-plate capacitor with the substrate placedon one of the electrodes. Boskovicet al. used a standardindustrial rf s13.56 MHzd PECVD systemsPlasma Technol-ogyd to produce nonaligned carbon nanofibers on powderedNi catalyst using methane and hydrogen mixtures.69 Ho et al.used a modified commercial rf PECVD reactorsPlasmaquestd to grow, reportedly, carbon nanotubes using acetyleneon Ni-coated quartz at 650 °C.93 The fact that a commer-cially available system can be used for this purpose meansthat a custom-made system is not necessary and the growthof carbon nanofibers can be as much as a part of the micro-fabrication process as, for example, deposition of silicon di-oxide or silicon nitride thin films.

Vertically aligned carbon nanofibers similar in structureto the fibers obtained by dc PECVD were produced by Hirataet al. using a magnetron-type rf plasma.94 In their reactor, amagnetic fields0,Bz,340 Gd was externally imposed par-allel to a powered cylindrical rf electrode using solenoidcoils. This allowed them to achieve lower plasma sheathvoltages and higher plasma densities. In this work the au-thors studied the influence of a dc self-bias,Vdc, on the align-ment of nanofibers, which in this case can be changed inde-pendently of other plasma parameters. Poor alignment wasobserved atVdc as low as −180 V, fairly good alignment wasobserved at −235 V, and the nanofibers were damaged due tosputtering atVdc=−570 V. These voltages were estimated tocorrespond to electric-field strengths at the nanofiber of 63,68, and 86 mV/mm, respectively.

Ikuno et al.used a dual power supply rf system in whichone power supply drove one electrode to maintain theplasma, while the other rf power supply was used to controlthe substrate bias.95 This system configuration is commonlyused to control film stress in Si3N4 PECVD processes. Simi-lar to the configuration used by Hirataet al., it offers controlof the plasma power independently of the bias.

d. Radio-frequency inductively coupledPECVD. Relatively high-density plasmas can be achieved inICP reactors. ICP sources are attractive because they arefairly simple in constructionsno external magnetsd and havebeen demonstrated to be applicable to large-area processings.300 mmd,96 which is desirable for large scale manufactur-ing. In such a system a coil is connected to the rf powersource as a part of a highQ network. Resonance currentscirculating through the coil produce an alternating magnetic



field that induces electron currents in the plasma. PracticalICP sources usually utilize a combination of capacitive andinductive couplings.

Delzeit et al. used an ICP system for the growth of car-bon nanofibers and carbon nanotubes.22 They found that it ispossible to produce either MWCNTs or nanofibers depend-ing on the growth conditions. It was determined that theregimes producing the nanofibers coincide with the availabil-ity of large amounts of atomic hydrogen, which is commonin hydrocarbon discharges with H2 as a source gas. Theyshowed that replacing H2 with Ar favored the production ofMWCNTs. Emission spectroscopy was used to explore theintensity of atomic hydrogen emission during the growthruns. These data are plotted in Fig. 13 and provide an inter-esting correlation to the observed transition. The growth ofnanotubes is accompanied by a low intensity peak of atomichydrogen. As the hydrogen emission intensity increases,nanofibers are obtained. The observation is consistent withthe analysis of Nolanet al.97 These observations are alsosupported by a simple zero-dimensionals0Dd model88,89 thatsuggests that the discharge produces large amounts of atomichydrogen along with a variety of CxHy radicals. Stable spe-cies such as CH4, C2H2, C2H4, etc., account for about 5% ina 20:80 CH3/H2 discharge at 3 Torr.

Caughmanet al. synthesized arrays of freestandingVACNFs using an ICP system.98 Their reactor operated at amuch lower total pressure—approximately 50 mTorr—compared to other methods. They note that low-pressure op-eration is potentially advantageous because of the improve-ment in plasma uniformity as the pressure is decreased. Inthis PECVD system, the ICP source was operated at

13.56 MHz. The source coil consisted of a 15-cm-diameterflat spiral coil s6 turnd separated from the chamber vacuumby a 20-cm-diameter, 2.5-cm-thick fused-quartz window.The chamber was made from stainless steel and was 30 cmin diameter. A 4.4-cm-diameter substrate heater, equippedwith rf biasing at 13.56 MHz, was located 20 cm below thewindow. Typical rf-induced self-biases of −50 to −200 Vwere used during growth.

Honda et al. used an ICP reactor to grow verticallyaligned carbon nanofibers at 500 °C using CH4 as the solesource gas.67 They found that dc bias is essential for growthand possibly plays a critical role in the etching of carbondeposits. This work also explored a variety of growth tem-peratures lower than those typically used to grow VACNF.The nanofibers obtained at 500 °C displayed a herringbone-type structure. When the temperature was reduced to 400 °Cno crystalline structure was observed in the nanofibers andtheir alignment deteriorated.

Lee et al. produced crystalline VACNFs at temperaturesas low as 200 °C using ICP-CVD with a mixture of CH4 andH2 as the source gases.68 The crystallinity of the VACNFswas measured by TEM, electron difraction, and Raman spec-troscopy. In the latter technique, the intensity ratio of theGs1590 cm−1d and D s1350 cm−1d bands, attributed to CNFsand carbonaceous particles, respectively, was observed tochange as a function of the hydrogen-to-methane volumeratios. The intensity of theD band, often referred to as thedefect or disorder band, decreased with increasing hydrogen-to-methane volume ratio. The intensity of theG band dis-played a maximum around the ratio of 30%. These resultsindicate that carbonaceous particles are etched by increasedhydrogen dosing. However, upon excessive hydrogen dosing,not only the carbonaceous particles but also the CNFs weredamaged, resulting in highly defective structures.

e. Microwave PECVD.In the high-frequency fields usedin microwave dischargesstypically 2.45 GHzd the plasmatakes on the character of a free-electron gas. This leads to anincrease in the density of high-energy electrons and increasesthe chemical activation efficiency of gases with high disso-ciation energiesse.g., N2 and H2d. These types of reactors arefrequently used for diamond film synthesis because the effi-cient dissociation of H2 produces atomic H that preferen-cially etches graphitic forms, leaving diamond.

Kuttel et al.grew carbon nanofibers on silicon substratesusing a microwave plasma under conditions that are nor-mally used for the growth of CVD diamond films.99 Thefilms were grown on silicon substrates in a tubular geometrysystem. Except for a slightly increased substrate temperaturessee Table Id, the growth parameters were standard for CVDof diamond films. For catalyst material, they used bothsputter-deposited Ni films on silicon substrates and Fe clus-ters, formed by reduction of FesNO3d3, in a hydrogen atmo-sphere at high temperature. The resulting films containednonaligned 100-mm-long nanofibers 20–60 nm in diameterand onionlike structuressquasispherical particles composedof concentric graphitic shells synthesized and observed byUgarte.100 The catalyst remained at the base of the growingstructures. The best growth conditions were found on thesubstrate in a position where the plasma was well above the

FIG. 13. Atomic hydrogen intensity from emission spectroscopy.sad Varia-tion with substrate power andsbd effect of argon dilution. Adopted withpermission from Ref. 22.



sample. At substrate locations more proximal to the plasma,the high concentration of atomic hydrogen etched the grow-ing film. In another early use of M-PECVD, Qinet al. ob-tained entangled bundles of nanotubes.101 They used an alu-mina substrate and the catalyst was deposited from a ferricnitrate solution. The nanotubes were grown at 850–900 °Cin a mixture of CH4 and H2.

The M-PECVD system used by Boweret al.consisted ofa 2.45-GHz, 5-kW microwave power supply with a rectan-gular waveguide that is coupled to a 6-in. inner diameterstainless-steel cylindrical growth chamber and a molybde-num substrate stage with a rf graphite heater that allows con-trol of the substrate temperature independent of the plasmapower.61,102Dense forests of carbon nanostructures were syn-thesized using this technique. The nanotube orientation wasinfluenced by the self-bias potentialsestimated to be −10 Vdestablished on the immersed substrate surface in the high-frequency plasma, the field of which is invariably terminatedperpendicular to the surface. Boweret al. claimed that elec-trostatic force alone would force these one-dimensional tu-bular structures to align with the field direction, the energeti-cally most favorable orientation. The difference in alignmentduring PECVD and CVD was also observed in this work.102

Many other groups have used this type of growth systemfor the synthesis of carbon nanostructures.101,103–110 Eventhough towers of nanofibers and nanotubes vertically alignedby crowding and van der Waals force have beenproduced,61,102,106,111to date the growth of individual free-standing nanostructuressi.e., single tubes or fibersd has notbeen demonstrated by microwave discharge. Tip-typeherringbone-structured nanofibers were synthesized by Okaiet al.utilizing M-PECVD.112 They used a quartz reactor withflowing methane and hydrogen gases. A voltage of −250 Vwas applied to the substrates during the growth. Unfortu-nately alignment characteristics were not reported in thiswork. The nanofibers M-PECVD-synthesized by Cuiet al.had a bamboo-type structure.106

Other studies of M-PECVD growth of carbon nanostruc-tures include measurements of chemical species byin situoptical emission spectroscopysOESd and quadruple massspectroscopysMSd,104 the effect of substrate bias,105 and theinfluence of catalyst choicesamong Fe, Ni, and Cod.111

f. Electron cyclotron resonance.The main distinguishingcharacteristic of electron cyclotron resonancesECRd sourcesis their ability to produce higher fluxes of low-energy ionsthan other sources. The ratio of ions to neutrals is muchhigher for ECR-PECVD than for other types of PECVD.

Lin et al.synthesized VACNFs using ECR-CVD produc-ing various bamboo and herringbone structures, and con-ducted a comparative study of different catalystssFe, Ni, andCod and of different gas mixturessCH4/H2 and CH4/N2d.113

They also compared M-PECVD and ECR-CVD. They foundthat a nitrogen-containing mixture was more effective in theproduction of bamboolike nanofibers, suggesting that bom-bardment with nitrogen ions provides a more efficient sput-tering agent that keeps the catalyst surface active.

g. Hollow cathode PECVD.A hollow cathode glow dis-chargesHCGDd is another version of a dc source in which acathode is shaped in the form of a cylinder so that the sec-

ondary electrons are self-contained in the negative glow. Theplasma densities are one to two orders of magnitude higherthan in a diode discharge.114 A hollow cathode system wasused by Huczkoet al. for synthesis of a dense forest ofvertically aligned carbon nanofibers115 by decomposition ofFesC5H5d2 in He at 200 °Csdetermined from optical emis-sion spectrumd. In their experiments the substratefanodicaluminum oxide AAOg was not biased and was placed insidethe cathode both perpendicular and parallel to the cathodeaxis. They report that the nanofibers were arranged inbundles with apparent Fe particles at the tips and were not inregistry with AAO pores. However, the alignment mecha-nism was not suggested by the authors. It is worth noting thattheir spectroscopic analysis of the generated plasma revealedthat the strongest emission peaks were those generated by C2

and CH radicals. They went on to speculate that the C2 radi-cals constitute the key intermediate phase in nanofiber for-mation, referring to molecular-dynamics simulations by Xiaet al.,116 in which C2 species were suggested to be respon-sible for the formation of open-ended nanotubes in the ab-sence of catalyst. Consequently, Huczkoet al.concluded thatthe plasma-generated C2 species, together with the catalyticFe, produce nanofibers at a fairly low “effective” tempera-ture of 200 °C.

h. Corona discharge.Li et al. used a corona dischargeplasma combined with template-controlled growth to pro-duce aligned carbon nanotubes at atmospheric pressure andtemperatures below 200 °C.117 However, it has to be notedthat the reported temperature is the temperature of the gasestimated from the measurement of the temperature of theouter walls of the reactor. In order to explain the successfulgrowth at such a low temperature the authors note that non-equilibrium plasmas, such as corona discharges, are charac-terized by high electron energy and low gas temperature, andthe energetic electrons usually can decompose molecules toreactive species without additional heat energy. The coronadischarge plasma reactor used in these experiments consistedof a quartz tube and two axially centered electrodes—a tung-sten wire electrode and lower circular plate stainless-steelelectrode.118

3. C-PECVD specific aspects of carbon nanostructuregrowth control

a. Basic processes at the catalyst particle in a dc plasmaenvironment.So far the majority of the works on the growthand applications of vertically aligned carbon nanofibers weredone using dc and HF-dc systems. As the simplest of thePECVD reactors, the dc systems can be used to understandmost of the fundamental processes involved in C-PECVDgrowth of carbon nanofibers. For clarity of discussion in thissection we focus on dc discharges.

Developing control over catalytically grown nanostruc-tures can only be achieved through a detailed understandingof the behavior of the catalyst nanoparticle. Figure 14 pre-sents a very simplified schematic of the processes that occurin this fascinating nanoscale environment. Many of theseprocesses are present in thermal CVD, namely, arrival ofexcited species to the surfacesAd, catalytic dissociationsBd,departure of undissociated moleculessCd, solution of carbon



into the catalyst particlesGd, formation of carbon film on thesurfacesDd, diffusion of carbon through or around the cata-lyst particle sHd, and incorporation of carbon atoms into agrowing graphene layersId. PECVD introduces a few addi-tional processes due to application of electrical field and par-tial ionization of the gas, such as sputtering due to ion bom-bardmentsFd, chemical etchingsEd, and mechanical forcedue to interaction of a conducting cylinder with high electricfield. In fact, the coupling of the latter to the rates of carbonincorporation into graphene layers around the particle, viastress distribution, is believed to be the mechanism of verti-cal alignment.24

b. Growth-parameter space.Compared to thermal CVD,plasma excitation provides an additional level of controlwhile simultaneously introducing an additional level of pro-cess complexity. Unlike TCVD reactors in which tempera-ture, total gas pressure, and flow govern the nanostructuregrowth process, in the PECVD process, parameters specificto the glow discharge must be considered. The voltage, cur-rent, power, and resultant field distributions within the dis-charge all play a critical role in shaping the outcome of thegrowth process. It is important to note that plasma is usuallyused for both the deposition of thin conformal films and foretching, depending on the choice of conditions. Counter in-tuitively, in order to “deposit” carbon nanostructures aPECVD reactor must be operated in the etching, rather thandeposition, regime to avoid thin-film formation. Special con-sideration must be given to finding the balance between etch-ing and deposition to prevent detrimental formation of car-bon films and at the same time avoid damage to the sidewallsof growing nanofibers.119

The C-PECVD process involves a host of parametersvariable over a multidimensional space, leading to extraordi-nary changes in the structure and morphology of the resultantcarbon deposit, from amorphous carbon films to nanofibersand nanotubes. The parameters that constitute this multidi-mensional parameter space includes1d total pressuresPd, s2dtotal gas flowsFd, s3d carbon source to etchant gas flow ratiose.g., C2H2/NH3d sRd, s4d substrate temperaturesTd, ands5dplasma powerfcurrent sId and voltagesUdg. A slice of thephase diagram along theT-R axis, with all other parameters

FIG. 14. Processes at the nanoparticle in a PECVD environment.

FIG. 15. A slice of growth parameter space. SEM images of VACNFs tiled to illustrate a phase diagram of the VACNF in aT-R slice of the multidimensionalparameter space. The other parameters were the samesI =100 mA,P=2 Torr,FNH3=80 SCCM, and growth time=10 mind. The substrate was Si, the catalystwas 10 nm of NiFe alloy s50%: 50%d a top of 10 nm Ti buffer layer. The areas of the “good” VACNFs aresT,Rd=s750,40/80d ,s700,50/80d ,s650,60/80d ,s600,50/80d. The s750,60/80d are peculiar “fluffy” nanofibers. The part of the diagram to the right of the goodVACNFs represents the sample coated with a graphitic film that prevented the VACNF growth. The VACNFs on the leftfe.g.,s600,40/80dg are damaged andcontained a considerable amount of nitrogen.



kept constant, is presented in Fig. 15. There are specificbounded regions where the resultant material possesses thecharacteristics of the high-quality VACNFs shown, for ex-ample, in Figs. 5 and 6. In general, for a givenT, there is aparticular range ofR values for which favorable growth oc-curs. At higherR, a graphitic carbon film forms on the sub-strate and prevents fiber growth. At small values ofR thefibers are chemically etched during growth, producingheavily damaged fibers. Taken to an extreme, the fibers canbe etched down to the substrate leaving nothing but stublikeremnants. At higher plasma power, the regions for whichhigh-quality VACNF growth occurs are shifted toward higherR values.

The selection of the parameters used to perform VACNFgrowth is dependent upon the combination of catalysts, sub-strates, carbon source and etchant gases, the presence of adilution gas, and the catalyst pattern. Several studies weredevoted to the discovery of trends and interrelationships be-tween parameters.59,63,78,120–122These trends are discussedbelow.

A typical growth rate versus temperature curves forC-TCVD growth of carbon nanostructures is usually similarto the one displayed in Fig. 16 as solid circlesslowercurved.46 On the low-temperature side, the slope and curva-ture of the plot can be attributed to a diffusion-limitedgrowth process. The growth rate reduces abruptly on thehigh-temperature side of the curve, which can be interpretedas a supply-limited growth regime as a consequence of thedrastic reduction in the sticking rate of the species impingingonto the catalyst surface. Chhowallaet al. have observedsimilar behavior in dc PECVD growth, as shown in Fig.17.59 However, the activation energy of 0.56 eV, determinedfrom the slope in the growth rate versus inverse temperatureplot, is significantly lower than the activation energy of Cdiffusion through Ni of 1.5 eV. This suggests either en-hanced diffusion of C through the Ni particle or, as wasproposed by Helveget al.,58 of C along the Ni particle sur-face. Such low activation energy also suggests the possibilityof true room-temperature growth.

The gases used for carbon nanofiber growth from a sup-

ported catalyst include carbon monoxide and hydrocarbons,such as acetylene, methane, benzene, etc. However, as de-scribed above, if carbon source gases are used alone, carbonfilms samorphous or graphiticd are formed and no carbonnanofiber growth occurs. Thus the selection of operating pa-rameters must be made such that etching, rather than depo-sition, is the dominant process. This is achieved in part byselection of an appropriate etchant gas such as hydrogen orammonia.

The optimal value ofR depends on several operatingparameters. Teoet al. studied the dependence of amorphouscarbon film formation during VACNF growth as a functionof acetylene-to-ammonia ratio using Auger electronspectroscopy.79 The chemical composition of species nearthe substrate surface, i.e., at the catalyst particle level, de-pends on the diffusion of active species from the plasmasheath. By changing the operating voltage, the size and shapeof the characteristic regions of the plasma can be altered.This is ultimately reflected in the distance that the activatedspecies must diffuse through to reach the catalyst surface. Adirect relation between plasma power andR has been estab-lished by Merkulovet al.121 In these experiments the gasratios were adjusted at different current values to obtainidentical nanofiberssFig. 18d, showing a linear dependencebetweenR and the discharge currentI. This relationship canbe used interchangeably to control the composition of spe-cies near the substrate surface. It has been reported that dif-ferent species participate in the PECVD VACNF and CVDCNF growth.122 Free radicals are involved in PECVD growthwhile C2H2 is the dominant species in CVD processes.

The VACNF growth rate is proportional to the number ofradicals impinging on the catalyst surface and to the decom-position rate of these radicals at the surface.sHere, by de-composition we mean a complex process involving adsorp-tion and fragmentation of the radicals at the catalyst surfaceas well as desorption of the reaction byproducts from thesurface.d Different types of radicals are expected to havedifferent decomposition rates. One way to increase thegrowth rate is to change the gas mixture and plasma power tocreate more radicals or to shift the radical distribution to-wards those with higher decomposition rates. Simply in-

FIG. 16. Temperature dependence of CNT growth ratessfilm thicknessd.Adopted with permission from Ref. 46. The solid dots correspond to CVDon the Al/Fe/Mo multilayer film using a 100 SCCM/2.9 SCCM hydrogen/acetylene ratio. The solid squares represent the CVD growth with additionof 4 mg/h of ferrocene. The filled triangles represent the CNT growth for anoptimized ratio of 4 mg/h of ferrocene to 12.4 SCCM of acetylene.

FIG. 17. Average nanofiber length vs inverse deposition temperature aftergrowth at all conditions constant in a dc PECVD process. Adopted withpermission from Ref. 59.



creasing the C2H2 content increases the growth rate only upto a certain valuesstill substantially lower than that for ther-mal CVDd, after which the growth rate begins to decrease.59

Above a critical C2H2 content, which depends upon a par-ticular set of experimental conditions such as plasma power,pressure, temperature, etc., more radicals with lower decom-position rates and fewer radicals with higher decompositionrates are produced. Moreover, the radicals with lower de-composition rate may start forming a shell encapsulating thecatalyst particle, thereby limiting the carbon supply availablefor VACNF growth, and may permanently attach to the sub-strate surface. This results in a reduced growth rate and, ifthe C2H2 content is increased even further, leads to the for-mation of a thick carbon film covering the entire substrate.This phenomenon has been experimentally observed in sev-eral studies.79,119

Since some applications may require very longVACNFs, it is highly desirable to develop controllable waysto further increase the growth rate. One way of accomplish-ing this was explored by Merkulovet al. and involved in-creasing the gas flow directed towards the substrate.122 Thiswas achieved by replacing the flat showerhead typically usedin PECVD reactors with a capillary-type gas nozzle featuringa variable orifice. These experiments were carried out usingtwo different orifice diameters: 1 and 5 mm. Figure 19sadshows that the maximum growth rates increase significantly

as smaller nozzle sizes, generating higher local gas flows, areused. The maximum growth rate for the 1-mm nozzle ismore than twice than that for the 5-mm nozzle and about fivetimes higher than that for the showerhead. Another salientpoint of these experiments is that the gas chemistry, i.e.,mutual interactions among the gas species, plays an impor-tant role in determining the growth rate. This is demonstratedin Fig. 19sbd, which shows the growth rate for the 1- and5-mm nozzles as a function of the C2H2/NH3 gas ratio.Clearly, the growth rates in both cases depend strongly uponthe gas mixture and each exhibits a well-defined maximum.Moreover, the maximum growth rates for the two nozzlesizes correspond to different values of the C2H2/NH3 gasratio. This probably occurs because orifices with differentdiameters produce different gas flow velocities, which affectsthe interaction among the gas species within the flow. Thisresults in modification of the chemical composition of theresultant species that reach the sample surface and conse-quently changes the growth rate. In addition to empiricalapproaches to understand how to create the most favorableconditions for the synthesis of a nanofiber, some modelingeffort has been applied to this complicated system. Hashet

FIG. 18. Relationship between plasma power and gas flow ratio in a dcPECVD process, showing the C2H2/NH3 ratio required to produce aVACNF with similar geometry as a function ofsad plasma power andsbdplasma current. Adopted with permission from Ref. 121. Copyright 2002American Chemical Society.

FIG. 19. Influence of gas flow rates on the growth of carbon nanofibers in adc PECVD process.sad Maximum VACNF growth rates for different experi-mental setups and therefore different gas flows to the substrate surface.sbdVACNF growth rate as a function of C2H2/NH3 gas ratio for two nozzlediameters of 1 mmsupper curved and 5 mmslower curved. Adopted fromRef. 122 with permission from Elsevier.



al., in order to understand what species are actually involvedin this processes, simulated the growth conditions using aone-dimensional model88 of hot-filament dc PECVD. Figure20 shows the concentration of the neutral species along theaxis of a dc reactor simulated for an Ar/C2H2/NH3 gas mix-ture without involvement of the filament.

The total pressure is related to the total number of spe-cies near the catalyst surface. For the glow discharge thereare more excited species involved causing an increase in thedeposition rate. This was also demonstrated for dc PECVDby Chhowallaet al., as shown in Fig. 21.59 This plot repre-sents almost the full range in which a dc glow discharge canbe maintained. It is worth noting that at higher pressure,fewer ions reach the surface due to collisions, thus the etch-ing is reduced. This work demonstrates that selecting theproper gas flow ratio is essential for avoiding the depositionof a carbon film which, as discussed above, can deactivatethe catalyst and inhibit growth.

c. Tip-type and base-type growth modes.As in C-TCVD,two growth modes have been observed in PECVDsFig. 22din which the catalyst particle is located either at the tip of thegrowing fiber or at its base. It has been suggested that thegrowth mode depends on the interaction of the catalyst withits supporting substrate.3,12 However, Melechko et al.showed that the kinetics of the growth process also plays a

role in the selection of the growth mode and it is possible togrow base-type or tip-type fibers on the same substrate andusing the same catalyst material by changing the gas flowratio.63

d. Vertical alignment.A direct correlation between thealignment of CNFs and the growth modesbase or tipd hasbeen observed in a PECVD process.24 In the case of tip-typefiber growth, the catalyst particle undergoes a process whereit lifts from the substrate, eventually detaching from it com-pletely. Once this occurs, the particle follows the path of theelectric-field lines present in the plasma sheath. This type ofgrowth leads to field alignment irrespective of the density ofCNF on the substrate, i.e., without the need for crowding forCNF to grow vertically. In contrast, it has been establishedthat base-type CNFs tend to grow in random orientationsunless the density of the fibers is so great that vertical align-ment occurs due to crowding or van der Waals forcessFig.10d.

A model to describe the alignment mechanism of tip-type CNF growth was proposed by Merkulovet al. describ-ing the alignment as the result of a feedback mechanismassociated with a nonuniform stressspart tensile, part com-pressived created across the interface of the catalyst particlewith the CNF by the electrostatic forcessFig. 23d.24 The axisof a CNF growing perpendicular to the substrate coincideswith the direction of the applied electrostatic force, resultingin a uniform tensile stress across the entire nanofiber/catalystparticle interface, as shown in Figs. 23sad and 23sbd. Conse-quently, carbon uniformly precipitates across the interface

FIG. 20. Simulated distribution of the concentration of various neutral spe-cies with distance from the cathode in a dc PECVD system using one-dimensional kinetic model. It is for a 9-Torr, 200-W dc plasma with a677-V negative bias using 54:200 SCCM C2H2:NH3 with a cathode at700 °C heated solely by the plasma. Courtesy of D. B. Hashet al. See alsoRef. 88.

FIG. 21. Dependence of growth rate on total gas pressure in dc PECVDprocess. Adopted with permission from Ref. 59.

FIG. 22. SEM images ofsad vertical alignment of carbon nanofibers in atip-type growth mode, andsbd nonaligned growth of base-type carbonnanofibers by dc PECVD process.



and the fiber continues to grow verticallysperpendicular tothe substrated. However, if there were a spatial fluctuation inthe carbon precipitation at the interface, CNF growth woulddeviate from vertical alignment, as shown in Figs. 23scd and23sdd. In the case of nanofibers growing from the tipscata-lyst particle at the tipd, the electrostatic force produces acompressive stress at the part of the particle/nanofiber inter-face where the greater rate of growth is seenfFig. 23scdg.Likewise, a tensile stress is produced at the part of theparticle/nanofiber interface where the lesser rate of growth isseen. It is proposed that these opposing stresses favor subse-quent carbon precipitation at the interface experiencing ten-sile stress. The net result is a stable, negative feedback thatacts to equalize the growth rate around the entire peripheryof the particle/nanofiber interface, leading to verticallyaligned CNF growth. The difference in the growth rates atthe two boundaries may be caused by stress-induceddiffusion123 due to the stress gradient in the catalyst particleand possibly by the variation in the stress-dependent stickingof diffusing C atoms to the C side of the Ni–C interface. Inany case, the exact mechanism may be quite complex and adetailed study of the stress-induced C diffusion and precipi-tation in the C–Ni system is needed. When the catalyst par-ticle is located at the base of the CNF, however, the situationwith the preferred location of carbon precipitation is differ-ent. Since the nanofiber base is attached to the substrate, thestress created at the particle/nanofiber interface with thegreater growth rate is tensilefFig. 23sddg and acts to continuethe increased growth rate, thus causing the CNF to bend evenfurther.

Understanding the e-field mechanism of tip-type nanofi-ber alignment allows control of the orientation. Usually in aplanar geometry the field is perpendicular to a metallic sub-

strate. However, field direction can be changed by placingthe substrate close to the edges of the substrate holder, andthis effect was exploited to grow nanofibers at a variety ofangles other than perpendicular to the surfacesFig. 24d.124 Todistinguish from the alignment by using porous templates orthe crowding effect, carbon nanofibers aligned by the e-fieldpresent in the plasma sheath of a PECVD system should becalled field aligned rather than vertically aligned. However,because the field direction is usually perpendicular to thesubstrate these nanofibers are referred to as vertically alignedin the current literature, and we often retain this terminologyin this review.

D. Substrates and catalysts

1. Substrates

The substrate plays a crucial role in carbon nanostructuresynthesis especially in PECVD processes. The substrate notonly acts as a medium for support but it also interacts with

FIG. 23. Alignment mechanism based on stress-dependent growth rate andstress distribution caused by interaction of nanofiber with electric field.Adopted from Ref. 24.

FIG. 24. Control of electric field and alignment direction of carbon nanofi-bers in a dc PECVD process.sad A schematic representation of the experi-mental setup during the PECVD process, in which the substrate is locatedclose to the sample holder edge, and scanning electron microscopy imagesshowing the resultant CNF forests located atsbd 100 andscd 1000mm awayfrom the edge and aligned at 38° and 12° angles to the substrate normal,respectively.sdd SEM image taken at a 45° tilt angle of an array of indi-vidual CNFs grown in the vicinitys10 mmd of the substrate/the sampleholder edge. Adopted from Ref. 124.



the catalyst and growth environment. Silicon and silicon di-oxide are two of the most common substrates for obviousreasons of application in silicon-based technology, however,the choice is practically unlimited. There are several issuesthat have to be considered when choosing a substrate.

In dc-PECVD systems, with the substrate as a cathode, itis necessary to have an electrically conductive surface, soinsulating substrates such as SiO2 must be covered in a metallayer, which can later be removed, as in Ref. 84. Further-more, substrates also contribute to the radical species in theplasma. During PECVD growth on Si substrates silicon spe-cies can be etched, sputtered, and redeposited onto the side-walls of the fibers, creating an insulating sheath,125 whichmay or may not be desirable. To avoid Si incorporation intothe fiber walls, a metal overlayer can be used to cover thesubstrate.

In addition, there has been some concern about deposit-ing Ni catalyst directly onto Si substrates. Several researchgroups claimed that nickel silicide formation prevents nucle-ation of the catalyst particles from thin films and inhibitsgrowth.59,126 Recently, however, Niheiet al. proved the op-posite by showing that nanotubes can be synthesized directlyfrom nickel silicide.127 Furthermore, there have been relateddebates as to whether CNTs prefer Si or SiO2 surfaces. Somesay that CNTs will selectively grow on Si over SiO2,

128

while others affirm the opposite.126,129Due to incompatibili-ties of substrates with catalyst materials and the growth en-vironment, buffer layers, underlayers, and adhesion layershave come into widespread use. Buffer layers such as Ti, W,and SiO2 have often been used to prevent diffusion or inter-mixing of catalyst and substrate, as in the case of nickelsilicide mentioned above. The significant effect of underlayermaterials on carbon growth is mostly attributed to wettingand particle formation. First of all, it has been shown thatmultilayered catalysts, i.e., metal underneath and sometimeson top of the catalyst material, drastically influence the CVDcarbon nanotube growth. Aluminum, for example, is oftenput underneath the Ni and Fe catalysts because it promotesthe small particle formation needed for CNT growth; in fact,without it no growth is observed.45,46The thicknesses of boththe underlayer and catalyst are crucial to the success of nano-tube growth.45,130 Christenet al. studied the dependence ofnanotube growth rates on thicknesses of both the catalyst andoverlayer in a trilayer system of Mo/Fe/Al.131 They usedpulsed laser deposition to create orthogonally overlappingmetal film gradients of Fe and Mo on top of 10 nm of Al ona Si wafer. In a C-CVD process using C2H2/Ar/H2 mixturethey achieved growth rates exceeding 17mm/min for a spe-cific thickness ratio of Fe/Mo that correspond to an atomicratio of 16:1. In addition, an underlayer can help to “glue”the catalyst to the substrate and act as an adhesion layer forbase-type nanotube growth.45 An extensive list of substrateand underlayer materials, their uses, and references can befound in Table II.239–245

2. Catalysts

The growth of carbon nanostructures is catalytically con-trolled, thus the choice of catalyst plays an imperative role indetermining the outcome.45,132 The catalyst particle is re-

sponsible for breaking bonds and adsorbing carbon at itssurface, then diffusing carbon through or around an interfacewhere the carbon reforms in graphitic planes.133 The proper-ties of the catalyst can therefore determine the rate of each ofthese steps as well as the degree of crystalline perfection andgeometric structure of the resulting carbon fiber.

A variety of metals and their alloys are catalysts for theproduction of carbon nanostructures. A list of all known car-bon catalysts, their uses, and references are compiled inTable II. The most commonly used and studied are Fe, Ni,and Co, whose physical properties and solubility with carbonare shown in Table III.246–248Included in this list are severalbinary or multimetal alloys, which have been shown in somecases to provide certain advantages over single-element cata-lysts for the growth of filamentous carbon.134 Whereas thetrusted transition metals Fe, Ni, and Co are known to be veryactive in their ability to break and reform carbon–carbonbonds, other noncatalytic metals such as Al or Cu, whencombined with a promoter catalyst, can enhance carbon dif-fusion and reaction rates.133 In some cases, alloy catalystshave resulted in higher activity,135 low-temperaturegrowth,136 and branched nanostructures.55,135,137–143

Recently there have been efforts to investigate a batteryof catalyst and substrate combinations by high throughputmethods. Nget al. came up with an efficient methodologyfor evaluating underlayer material compatibility with variouscatalysts for CVD nanotube growth;144 Cassellet al. pub-lished a similar approach for exploring nanofiber PECVDgrowth.145 In their experiments strips of several differentmetal contact layers were deposited onto a Si wafer, then thewafer was turned 90° and the strips of different catalystswere deposited on top. This created an underlayer-catalystgrid, which was then used to grow carbon nanostructureseither by CVD or PECVD. Nget al. concluded that Ti wasthe best underlayer in terms of low contact resistance andhigh growth density, however, Fe–Ni and Ni grew the mostvertically organized on an Al underlayer.144 Cassellet al.found that for their growth process Ni catalyst on a Cr un-derlayer yielded the highest-quality fibers on the basis ofgrowth rate, alignment, and diameter uniformity.145 It is ef-forts like these that will lead to rapid development andimplementation of the best catalyst and substrate for a givenapplication.

3. Methods of catalyst preparation

There are numerous ways to prepare catalyst particlesfor CNF growth including physical vapor depositionsPVDd,electro- and electroless plating, and coprecipitation methods.The most common approach is to deposit a thin film of cata-lyst and then sinter it into discrete nanoparticles. In this ap-proach, a thin film is deposited by electroplating, electrolessplating, or most commonly by PVD. PVD can be accom-plished either by sputtering or evaporating techniques, eachof which has its pros and cons. A sputtered atom typicallyhas tens of eV arriving at the substrate surface, in compari-son to thermal energies of evaporated films, which are on theorder of tens of eV. Thus sputtering leads to better mixing atthe interface and adhesion relative to evaporated films. Inaddition, substrate heating and substrate bias capabilitiy can



TABLE II. Catalysts and substrates.

CatalystUnderlayer

and substrateGrowthmethod

Carbonstructure References

Al–Fe, Cu–Fe,Fe–Ni–Al

Quartz tube CVD CNF, MWCNT 239

Alumina–Ni,Alumina–Ni–Cu

Quartz tube CVD CNF,multidirectional

142

Co Al oxide Corona discharge PECVD CNT 117Co MgO CVD y-junction CNF 140Co Nanochannel

aluminaCVD CNT, y-

junction CNT65 and 240

Co Si M-PECVD CNF 61Co, Co–Cualloy

Si PECVDCVD

VACNFCNF

241

Co, Co–Cualloy powder

Ceramic boat CVD CNF,multidirectional

138

Co, Co–Nialloy, Fe, Fe–Nialloy, Ni

Cr, Ir, Ta, Ti,and W on Si

PECVD VACNF 145

Co, Fe, Fe–Nialloy, Ni

ITO, Ir, Al, Ti,Ta, and W on Si

CVD CNT 144

Co, Fe, Ni Si M-PECVD VACNF 111Co, Fe, Ni Si ECR-PECVD VACNF 113Co, Ni SiO2 on Si PECVD VACNF 59Co, Ni W wires PECVD VACNF 82Cu–Ni alloy Quartz tube CVD CNF 135Cu–Ni alloy SiO2 CVD CNF,

multidirectional141

Cu–Ni alloy, Nipowder

Quartz tube CVD CNF,multidirectional

133

Cu–Ni, Fe–Ni,Pd–Se alloys

Si CVD CNF 136

Fe Al2O3 HCGD CNF 115Fe Alumina M-PECVD CNF 101Fe Fe foil CVD CNF 53Fe Mesoporous Si CVD VA-MWCNT 42Fe Porous Si CVD VA-MWCNT 60Fe Porous Si CVD CNF 62Fe Si CVD CNF 129Fe Si, quartz,

ceramicM-PECVD CNF 108

Fe SiC on Si CVD CNT 201Fe SiO2 CVD CNF 129Fe SiO2 on Si M-PECVD VACNF 106FeNi

SiO2 on SiSi

CVDPECVD

CNTVACNF

77

Fe, Fe–Ni alloy,Fe–Tb alloy, Ni

Silica, alumina CVD CNF, MWCNT 134

Fe, Mo/Fe-layered film

Ir, Al, Nb, andTi on Si

CVD SWCNT 130

Fe, Ni Al on Si, fusedquartz, mica,

HOPG

CVD MWCNT 219

Fe, Ni Si M-PECVD CNF 99Fe–Mo, Mopowder

Alumina CVD SWCNT 242

Fe–Ni alloy Cr, Ti, Ta, andW on Si

HF CVD CNF 92

Fe–Ni alloy Si CVDrf PECVD

CNFVACNF

95

Fe–Ni alloy SiO2 on Si CVD SWCNT 128Fe–Ni–Cr, Ni Si M-PECVD CNF 112Mo/Fe-layeredfilm

Al on Si ICP PECVD MWCNT 22



have a profound effect on the film properties such as crystalstructure, orientation, density, and grain size. Choiet al. re-ported on the controlling of the grain size of sputtered Nifilms by varying the rf power, which in turn affected thediameter, length, and purity of the CNTs grown from thefilm.109 Sputtering also allows for alloy depositions from analloy target, whereas alloys cannot be directly evaporatedaccurately due to differences in the vapor pressure of eachelement. Evaporation does have its advantages, though,mostly attributed to its highly directional deposition, whichenables patterns to be easily transferred to the substrate byresist lift-off methods outlined below.

If catalyst patterning is required, it can be done before orafter the catalyst film is deposited. Commonly, the pattern isdefined beforehand and a lift-off process is employed. First,the resist is applied to the substrate and the desired catalystpattern is exposed and developed. A PVD method is then

used to deposit the catalyst material onto the substrate. Fol-lowing the deposition, the substrate is immersed in a solventcapable of dissolving the resist, causing the metal to be re-moved from the remaining resist-covered areas. Conversely,the film can also be patterned following catalyst depositionby coating it with a resist suitable for either photo or e-beamlithography. The pattern is then exposed and developed suchthat resist remains in the areas where catalyst is desired. Theexposed metal is then removed by wet etching,146 ion-beamsputtering, or reactive ion etching.

In order to grow carbon nanotubes or nanofibers from athin-film catalyst, the film must first be sintered into islandsor discrete nanoparticles.45 While the sintering temperature istypically well below the bulk melting temperature for thecatalyst metal, the reflow of metal indicates that surface dif-fusion is at work.147 The formation of catalyst nanodropletscan also be explained by a stress buildup in the film due to

TABLE II. sContinued.d

CatalystUnderlayer

and substrateGrowthmethod

Carbonstructure References

Mo/Fe-layeredfilm

Al on Si CVD VA-MWCNT 46

Ni Al ICP PECVD VACNF 68

Ni Brass, bronze ICP PECVD VACNF 154Ni Cr on glass M-PECVD CNF 104Ni Cr on Si PECVD VACNF 189Ni Cr on SiO2 on

SiPECVD VACNF 193

Ni Glass PECVD, HFPECVD

VACNF 90

Ni Glass, Si HF PECVD VACNF 8 and 10Ni Graphite, Si,

plasticrf PECVD CNF 69

Ni MgAl 2O4 CVD CNF 58Ni Ni wafer PECVD CNF 86Ni Pt crucible CVD CNT, CNF 97Ni CVD CNF 50Ni Si PECVD CNF 126Ni Si CVD

PECVDCNT

VACNF91

Ni Si M-PECVD VACNF ropes 94Ni Si ICP PECVD VACNF 67Ni Si M-PECVD MWCNT 109 and

110Ni Si M-PECVD VACNF ropes 94Ni Si, PDMS CVD MWCNT 243Ni SiO2 on Si, Cr

on polyimidefoil

PECVD VACNF 70

Ni Ti on Si PECVD VACNF 121Ni W–Ti alloy and

Ti on SiICP PECVD VACNF 98

Ni silicide Si PECVD CNF 127Pd Quartz tube CVD CNF 231Pd powder Quartz tube CVD CNF 244Pt Si M-PECVD Carbon

nanostructures107

Ru Silica, alumina CVD CNF 245titania multidirectional



different expansion coefficients from the substrate.9 In somecases, though, heat alone is not enough to break up the cata-lyst film and additional techniques are required. As was men-tioned in the substrate section, underlayers are often used toincrease surface roughness or wetting capabilities.45 Nano-particle creation can also be encouraged by pretreatment ofthe catalyst film with an etchant gas or ion bombardment ina plasma environment. For example, the research group atCambridge heats the substrate in H2 for 15 min at 750 °CsRef. 59d before initiating PECVD growth. Leeet al. foundthat they could control the diameter of Fe particles for CVDgrowth by adjusting the condition of ammonia pre-treatment.129 Oak Ridge does a combination of heating in anammonia environment and then initiating an ammoniaplasma for up to 2 min before introducing the carbon sourcegas.121 In any case the outcome is the same, the catalyst thinfilm breaks into nanoparticles, which grow fibers when thecarbon source gas is introduced.

The size of the catalyst particles is ultimately determinedby the thickness of the film, in addition to the wetting prop-erties of the catalyst and substrate materials as well as themethod of catalyst preparation and pretreatment. Since nano-tubes are of such small dimensions, the size and uniformityof the nanoparticles become increasingly important. Chhow-alla et al. and Weiet al. reported a direct correlation of Ninanoparticle size increasing with the thickness of the originalfilm.59,77 However, Chhowallaet al. only observed this phe-nomenon on SiO2 substrates and found that the Ni film failedto sinter on clean Si substrates, which they attributed tonickel silicide formation. Cobalt, on the other hand, has beenfound to wet well on both Si and SiO2 substrates, showingthe same correlation of particle size to original filmthickness.59,61 Bower et al. also verified that thinner Cofilms, leading to smaller particles, determined the diameterand length of the resulting nanotubes.61 Wei et al. found thatfor Fe films there was a critical thickness over which nonanotubes would grow. In addition, below this critical filmthickness, there appeared to be no correlation between filmthickness and CNT diameters, which may be a characteristicof CVD systems.77 It can generally be said that for PECVDgrowth, thinner films lead to smaller particles, which in turnlead to denser arrays of smaller diameter fibers.

If periodic arrays of VACNFs are desired, the amount ofmaterial deposited for each catalyst dot is crucial in deter-

mining if single or multiple fibers form. It was shown previ-ously that catalyst film thickness relates to fiber array densityand particle size. Likewise, the diameter and thickness of alithographically defined catalyst dot influence whether onedroplet or thus one fiber forms from each dot. In fact, Merku-lov et al. found that there was a critical dot size resulting insingle fiberssFig. 25d, which was dependent upon severalparameters including the choice of buffer layer, the substrate,and the type and thickness of the catalyst used.9

4. Catalyst particle evolution

Nanoparticles formed by sintering a thin-film metal gen-erally have a disklike hemispherical shape due to a largecontact area with the substrate before the growth is initiated.A succession of “stop-action” images shows that after aninitial few seconds of growth, the particle is pushed upwardby the flux of carbon and becomes elongated. As growthcontinues the bottom surface of the particle begins to slopeupward until it has a conical or teardrop shape, with the tipof the cone directed toward the growing carbon nanostruc-ture and pointing in the direction of carbon diffusion.148 Thishas been found in both tip-type and base-type growths.63

Most of the characterization of catalyst particles by SEMand TEM is performedex situonce the substrate has cooleddown. The particle in this final state can show an elongatedconical end or remain spherical. Some particles have a fac-eted shape on top. This has been repeatedly observed forparticles involved in tip-type growth. An example of Ni par-ticles displaying this feature is shown in Fig. 26. The face-ting might have occurred during the cooling process, how-ever, this cannot be stated conclusively. There are still manymysteries regarding the behavior of the catalyst particle dur-ing growth. One interesting phenomenon is the repeatedstretching or “lurching” behavior of the particle inside thebody of the growing nanostructure. Thanks to recent techno-logical advances, there have been a few accounts that capturethe catalyst’s evolution and the growth phenomenoninsitu.58,149The invoked stretching-retracting mechanism of thecatalyst nanoparticle could explain the periodic formation ofhorizontal graphene planes characteristic of a bamboo struc-ture.

Another interesting phenomenon is the decrease andsubsequent exhaustion of the catalyst material during

TABLE III. Physical properties of Fe, Ni, and Co.sbcc refers to body-centered-cubic structure, hcp refers tohexagonal close-packed structure, and fcc refers to face-centered-cubic structure.d

ElementAtomicnumber

Density ofsolid at20 °C

sg/cm3dMelt temp

s°CdEutectic withcarbons°Cd

Crystalstructure

LatticespacingsAd

Electronicconfiguration

Fe 26 7.87sRef. 169d

1538sRef. 169d

727 sRef. 169d sadbcc

2.8 665 866sRef. 246d

fArg3d64s2

Co 27 8.90sRef. 169d 1495 sRef. 169d 1320 sRef. 247d hcp 2.50714.0 696 514

4.105sRef. 248d

fArg3d74s2

Ni 28 8.91sRef. 169d 1455 sRef. 169d 1326.5sRef. 247d fcc 3.524 540sRef. 246d

fArg3d84s2



growth. It has been reported that the size of the catalystparticle decreases continually during the PECVD synthesisprocess.30 The diminishing of the catalyst material could bedue to ion-beam sputtering or by loss of Ni along the nanofi-ber body. Since particle size correlates to fiber diameter,9

with the particle size reduction the fiber diameter also re-duces. This trend can be used to sharpen the tips of thenanofibers for use in such applications as field emission andintracellular probes. However, this size reduction simulta-neously creates a limitation on the maximum obtainablelength of the freestanding isolated nanofiber due to the lossof catalyst material. In order to achieve a desired final lengththe amount of metal contained in the catalyst particle mustbe sufficient to last the duration of the growth process. Thisis stipulated by the role the catalyst particle plays in protect-ing the core of the growing nanostructure from physical andchemical etchings. In other words, the nanoparticle functionsas an etch mask. Should the nanoparticle disappear before

the nanofiber achieves the targeted length, the fiber corewould no longer be masked and would be etched back.

III. PROPERTIES

A. Structure

Carbon-based materials have unique mechanical, electri-cal, and chemical properties that relate to the various bond-ing arrangements or possible electronic configurations that acarbon atom can form with its nearest-neighbor carbonatoms.150 Each carbon atom has six electrons which occupy1s2, 2s2, and 2p2 atomic orbitals. The 1s2 orbital containstwo strongly bound core electrons. Four more weakly boundelectrons occupy the 2s22p2 valence orbitals for the free-carbon atom. In the crystalline phase, the valence electronsgive rise to 2s, 2px, 2py, and 2pz-orbitals which are importantin forming covalent bonds in carbon materials. Since theenergy difference between the higher-energy 2p levels andthe lower-energy 2s level in carbon is small compared withthe binding energy of the chemical bonds, the electronicwave functions for these four valence electrons can readilymix with each othershybridized, thereby changing the occu-pation of the 2s and three 2p atomic orbitals so as to enhancethe binding energy of the carbon atom with its neighboring

FIG. 25. Formation of multiplesad, sbd, and singlescd Ni droplets on thepatterned evaporated dotss15 nm Ni/10 nm Ti on Sid and subsequentgrowth of multiplesdd, sed, and singlesfd VACNFs. The Ni droplets wereformed during NH3/He plasma preetching and heating up to 600–700 °C.The SEM images were taken at 15 kV and 0° tiltsad–scd and at 50° tiltsdd–sfd. Adopted from Ref. 9.

FIG. 26. An example of one of the possible faceting of Ni catalyst nano-particles on the tips of VACNFs grown by PECVD.sad TEM image of aVACNF at 400 kV,sbd electron diffraction pattern from the Ni nanoparticle,and zone axisf0 0 1g. Adopted with permission from Ref. 155.



atoms. Three possible hybridizations occur in carbon:sp,sp2, andsp3. The various bonding states are associated withcertain structural arrangements, so thatsp bonding gives riseto chain structures,sp2 bonding to planar structures, andsp3

bonding to tetrahedral structures. The major bulk phases ofcarbon aresp2-bonded graphitefFig. 27sbdg andsp3-bondeddiamond fFig. 27sadg. The existence of nanostructuredphases, all based on thesp2-bonded hexagonal network ofatoms sgraphene sheetsd—such as fullerenesfFig. 27scdg,nanotubesfFig. 27sddg, and nanofibersfFig. 27sedg—is attrib-uted to the energetically favored elimination of danglingbonds, even at the expense of increasing strain energy,thereby promoting the formation of closed cage structures.

There is an abundance of literature devoted to the struc-ture of carbon nanotubes and a full description will not berepeated here. In addition to the definitions given in the in-troduction it is worth noting that a nanotube can be formedby rolling a graphene sheet into a cylinder and merging itsopposite edges. The open edge structure—a scroll—has alsobeen observed and has been confirmed by unrolling them inan electric explosion.151 One of the distinctions of nanofibersfrom nanotubes is that for nanofibers the edges of different

graphene sheets end at the sidewalls, for example, as shownby Chenet al. sFig. 28d.152 The annealing experiments con-ducted by Endoet al. showed that when the “stacked cup”nanofibers are subjected to temperatures as high as 3000 °Cthese open edges are removed by formation of closed-loopstructures.153

The structure of VACNFs has been studied by manygroups using TEM. However, in all cases the reported struc-ture was from either the middle of the nanofiber or from itstip. Recently Cuiet al. concentrated on the base of theVACNF.148 They showed that the area in the vicinity of thenanofiber-substrate interface is a telltale of the initial stagesof its growth. Figure 29 shows TEM images of the area nearthe base of the nanofiber. It can be seen that at first thegraphene layers are formed parallel to the substrate and sub-sequent layer edges bend upward as the nanoparticle bottomchanges its shape from a flat surface to a conelike structure.This fact has serious implications for electrical and mechani-

FIG. 27. Carbon structures:sad diamond, sbd graphite, scd buckminsterfullerene,sdd single-walled carbon nanotube, andsed stacked cone carbonnanofibersAdopted with permission from Ref. 20.d

FIG. 28. TEM image of an edge of a nanofiber, displaying the freegraphene-sheet endssindicated by arrowsd.

FIG. 29. Transverse TEM images of a VACNF after 5-min growth:sadcomplete TEM image of the VACNF on the silicon substrate;sbd high-resolution transmission electron microscopysHRTEMd image of the inter-face between the VACNF and the silicon substrate;scd interface images ofthe left edge of the VACNF, silicon substrate, and amorphous resin;sdd ofthe right edge;sed a void in the sidewalls; andsfd a void between the cupbottoms. Note thatsed and sfd are taken from a different VACNF fromsad.Due to sputtering during the VACNF growth, the silicon substrate aroundthe VACNF is etched, producing the protruding appearance of the siliconsubstrate immediately under the VACNF insad. The dotted lines indicatehow the wall angle is measured with the assistance of the voids in thesidewalls. To guide the reader’s eyes, circular dots are used to mark thetransition between the cup bottoms and cup walls.sbd shows a clear bound-ary between the VACNF and the silicon substrate with an interface layer ofonly two graphitic planes. The interface layer that appears insad is thuscaused by these graphitic planes. Adopted with permission from Ref. 148.



cal properties of the VACNFs. For example, this explainswhy when lateral force is exerted on nanofibers they typi-cally break at their bases by a horizontal shear along graphiteplanessFig. 30d.

Catalyst particle crystallographic orientation and shapehave been identified as two of the factors that govern thestructure of carbon nanofibers. Recently, Kiselevet al. havereported on their studies of the relation of the structure ofVACNFs to the orientation and faceting of Ni catalystparticles.154 VACNFs were grown in acetylene/ammonia at-mosphere by ICP PECVD from Ni 10-mm-thick film electro-plated onto bronze plates. TEM and diffraction observationsshowed that Ni catalyst particles are faceted single crystals.They found thath100j facets preferentially decompose car-bon while h031j facets deposit carbon. Kuanget al. showedthat the axial direction of the VACNFs grown by dc PECVDon Ni wafers is mainly parallel to thek110l andk042l direc-tions of Ni.155 The graphene cone angle varied around 30°which they concluded to be not matchingh110j planes andthus corresponding to high-index crystal planes of Ni. Fromthese experiments it appears that the graphene cone angle isnot completely governed by the nanoparticle shape and ori-entation. At the same time, thein situ video recordings byHelveget al. clearly show the formation of graphene layerson the step edge of af111g surface of Ni. It is interesting tonote that in their work two different types of nanofibers arepresented that are herringbone and bamboo in our definitionsare possibly the result of different orientations of Ni nano-particles. Another set of observations that might connect thenanofiber structure with the crystallography of catalyst nano-particles is the difference in the nanofiber structure producedby different catalysts. Fe, Ni, and Co have different crystal

structuressTable IIId. Donget al., for example, showed thatat the same CVD conditions Fe and Co produced nanotubes,while Ni resulted in herringbone nanofibers.156 It is not clearat the moment how to control the catalyst nanoparticle ori-entation and shape but it might be a key factor to be able tocontrol the structure of the nanofibers. However, catalystcontrol alone might be not sufficient for that because thegrowth conditions influence structure formation as well. No-lan et al., based on their observations, asserted that the coneangle is controlled by the amount of atomic hydrogen presentduring CVD process.97 Specifically, the cone angle increaseswith the increase of atomic hydrogen concentration.

B. Electrical properties

Lee et al. have investigated the electrical characteristicsof individual nanofibers by measuring theI –V characteristicsof suspended nanofiber bridges.157 The nanofibers exhibitedlinear I –V characteristics at low positive and negative ap-plied voltages. Note that this two-contact measurement pro-vides a resistance that is the sum of the contact and thenanofiber resistances. For this particular section of thenanofiber the resistance was 622V, which corresponds to1 kV /mm of nanofiber length. Using the assumption thatconduction was through the entire cylindrical cross-sectionalarea of the nanofiber, the range of resistivity was estimated tobe between 10−6 and 10−5 V m. These values compare verywell to arc discharge multiwall nanotubes,158 for which theresistivity was calculated to be 9310−6 V m for a350-nm-long section of a 20-nm-diameter nanotube whoseresistance was 10 kV. In general, the resisitivities of multi-wall nanotubes are comparable to those of arc-grown graph-ite fibers and ropes of single-wall nanotubes whose resistiv-ities are,10−6 V m.159 Schonenbergeret al. also demon-strated that electrical breakdown occured at current densitiesof 107–108 A/cm2, which is much higher than the typicalcurrent density of,106 A/cm2, sCRC Handbookd at whichmetal wires undergo electromigration, demonstrating the ad-vantage of covalently bonded conductors.

While two-probe measurements provide an estimate ofthe resistivity of a CNF, precise measurements should bedone using a four-point probe technique to exclude the con-tact resistance contribution. Such measurements were re-cently performed by Zhanget al. using multiple Ti/Auohmic contacts patterned on top of the nanofibers usingelectron-beam lithographysFig. 31d. These fibers were har-vested from a VACNF forest and deposited onto an insulat-ing substrate, and the contacts on individual fibers exhibitedresistances of a few kilohms.160 These four-point probe mea-surements demonstrate that VACNFs exhibit linearI –V be-havior at room temperature, with a resistivity of approxi-mately 4.2310−5 V m. Zhanget al. propose that this valueis consistent with a dominant transport mechanism of elec-trons traveling through intergraphitic planes in the VACNFs.

C. Electron field emission

Field emission involves the extraction of electrons froma solid by tunneling through the surface-potential barrier,typically equal to the work function of the material,f. An

FIG. 30. SEM images of VACNFs mechanically sheared from their bases.The arrows indicate the stumps of VACNFs demonstrating that nanofibersare weakest to lateral shearing near the substrate.



applied electric field,E, lowers the barrier height by anamountDf described by the following relation:

Df = S qE

4p«0D1/2

. s1d

The emitted current depends on the local electric field at theemitting surface,E, and onf. A model offered by Folwerand Nordheim based on early experimental evidence, showsthat the dependence of emitted current onE andf should beexponential,161

I =q3E2a

8phft2sydexpF− 8ps2md1/2f3/2

3hqEnsydG , s2d

where a is the emission site area in units of cm2, h isPlanck’s constant,m is the electron mass,y=Dw /w and tsydandnsyd are the Nordheim elliptic functions.

The local electric field strongly depends on the shape ofa surface and can be significantly enhanced at the apex ofsharp features. Such influence can be expressed as a geomet-ric field-enhancement factor,b. This leads to the followingsimple relation for the field at the tip,Etip, of a field-emittingsurface:

Etip = Eb. s3d

For a thin cylinder,b is roughly proportional to the height-to-diameter ratio, or aspect ratio.162 For electron emission tooccur, the electric field at the emitting surface must be on theorder of 1–3 V/nm, which at reasonable applied voltagesrequires large values ofb.

Carbon field-emission cathodes have some distinct ad-vantages compared to other materials. It has been shown byBakeret al. and Lea that graphitic carbon microfibers can beoperated continuously in ambient gas pressures two decadeshigher than conventional tungsten field emitters without ob-servation of degraded performance.163,164 Graphitic carbonalso possesses the lowest sputter yield of any material. Thisstems from the covalent bonding present in graphite. Conse-quently, carbon is far more resistant to sputtering from ion-ized residual gas molecules than conventional field-emissioncathodes composed of refractory metals.

Field emission from nanostructured graphitic materialsincluding single- and multiwalled CNTs and CNFs has beenan area of intense investigation in recent years. This body ofresearch indicates that these materials possess the environ-mental robustness inherent to microscopic carbon fiber fieldemitters while offering highb values as a consequence oftheir nanoscale dimensions and intrinsically high aspectratios.165,166 Almost every publication on the synthesis ofcarbon nanostructures contains data on electron field emis-sion from the material that is presented. The main impetusfor this focus is the potential use of such materials in appli-cations that involve electron field-emission sources. Thistopic is discussed in detail in theApplicationssection of thisarticle.

There are many reports with measurements of field emis-sion from thin films or mats of carbon nanostructures.167,168

In such arrangement it is difficult to determine the contribu-tion of individual nanofibers or nanotubes to the observedexperimental data. Moreover, a dense arrangement of nano-structures marginalizes the geometric field enhancement dueto the screening of the electric field by the neighboring nano-structures. In order to determine the actual field-emissionproperties of individual nanotubes or nanofibers they shouldbe isolated. For this purpose Merkulovet al. prepared arraysof conical VACNFs with average base diameters of 200 nm,average tip diameters of 25 nm, and average heights of2 mm, spaced at 50-mm intervals onn-type Si substrates.168

Field-emission properties of individual VACNFs were mea-sured with a movable current probe capable of positioning a2-mm-diameter probe tip above an individual VACNF withsubmicron accuracy. Field-emmissionsFEd measurements re-vealed that isolated nanofibers are good field emitters withemission threshold fields between 15 and 50 V/mm. IsolatedVACNFs displayed stable emission for 175 hsthe longesttest periodd of continuous 10-nA operation at vacuum levelsof 10−6 Torr. FE current versus voltagesI –Vd analysis hasshown a maximum measured FE current exceeding 5mAwithout any degradation to the VACNF tip. This correspondsto a current density of approximately 500 kA/cm2. The FEI –V characteristics of isolated VACNFs display Fowler–Nordheim-like tunneling behaviorsFig. 32d.

FIG. 31. Four-probe measuremnts of VACNF resistivity.sad A SEM imageof a completed VACNF structure with five metal electrodess200 nm each ofTi/Aud contacting the nanofiber. The inset is a SEM image of the as-grownVACNF forest.sbd I −V curves for the four-probefelectrodes 1, 2, 3, and 4of the device insadg and two-probeselectrodes 2 and 3d measurements.Adopted from Ref. 160.



D. Mechanical properties

It is well known that nanoscale structures exhibit differ-ent mechanical properties than their bulk counterparts. Fur-thermore, an important characteristic of most materials isthat a small diameter fiber is much stronger than the bulkmaterial169 due to the diminished probability of critical sur-face flaws with decreasing specimen volume. The vast ma-jority of engineering forms of carbon have more or less dis-ordered graphite microstructures. The microstructure of thecrystalline layers within the fiber influences properties suchas strength, stiffness, deformation modes, fracture behavior,and toughness.170 The literature shows great variance in themechanical properties of carbon forms such as carbon nano-tubes and carbon fibers. Table IV249–259gives some generalranges for the strength and modulus of engineering carbons.

The high volumetric density of short, strongsp3 bondsgives diamond the highest stiffness of any known materialssYoung’s modulus,1 TPad. Similarly, the high areal densityof short, strongsp2 bonds in the basal plane of graphite

results in a Young’s modulus that is comparable to that ofdiamond.170 The sp2 graphite bond is the strongest of allchemical bonds, but the weakness of the interplanar bondingmeans that an ordinary graphite is of little value as a struc-tural material.171 One way the great in-plane strength ofgraphite can be exploited is in the development of high-modulus carbon fibers that have basal planes preferentiallyoriented along the fiber axis. Even so, we cannot neglect thegraphite’s highly anisotropic mechanical properties. Single-crystal graphite has a Young’s modulus over 28 times higherin the direction parallel to the basal planes than in the per-pendicular direction.170 Thus, while there is a preferred ori-entation of basal planes parallel to the fiber axis, this createspoor transverse properties. Carbon fibers can also have ashigh as 100 times more stiffness along the fiber axis thanperpendicular to it.170

Single-crystal carbon nanotubes have extremely largelength-to-diametersaspectd ratios, with diameters as small asonly a few nanometers. They are the strongest known mate-rial, with a specific tensile strength as high as 100 times thatof steel.172 CNTs also have high stiffness and ductility. ForSWCNTs, tensile strengths range between 50 and 200 GPa,Young’s modulus is on the order of 1 TPa and fracturestrains are between 5% and 20%.169 As a consequence oftheir size and high degree of crystalline perfection, CNTs arevirtually flaw-free, which contributes to their exceptionalstrength.169 However, in spite of their incredible mechanicalproperties, SWCNTs are not utilized extensively as a rein-forcement medium because they are expensive to produceand purify.

On the other hand, MWCNTs are easier to produce sincethey do not require as stringent catalyst particle preparation.MWCNTs are composed of concentric graphene sheets andhave diameters on the order of tens of nanometers dependingon the number of graphite layers. One drawback is that theselayers can slide past each other easily, often failing by the

TABLE IV. Mechanical properties of carbon-based materials.

MaterialDensitysg/cm3d

Tensile strengthsGPad

E Young’s modulussGPad

Diamond 3.52sRef. 249d3.20–3.52sRef. 169d

0.800–1.4sRef. 169d 1054 sRef. 249d700–1200sRef. 169d

Graphite 2.26sRef. 250d1.71–1.78sRef. 169d

0.0138–0.069sRef. 169d0.01–0.08sRef. 251d

1060 in planesRef. 252d1020 parallel to basal,36.3 perp.sRef. 250d

11 sRef. 169dC60 1.72sRef. 253d 16 sRef. 253dSWCNTs 50–200sRef. 169d 1000 sRef. 169d

900–1700sRef. 254dMWCNTs 6.2–2.2sRef. 252d

11–63sRef. 173d1800 averagesRef. 255d

690–1870sRef. 252d270–1280sRef. 171d270–950sRef. 173d

Graphitewhiskers

2.2 sRef. 169d 20 sRef. 169d20 sRef. 256d

700 sRef. 169d

Carbonfibers

1.78–2.15sRef. 169d 2–5 sRef. 257d0.3–8sRef. 170d

1.5–4.8sRef. 169d0.6–3.7sRef. 258d

,80–700sRef. 170d228–724sRef. 169d

680 averagesRef. 259d

FIG. 32. Electron field emission from a VACNF.sad SEM image taken at45° of an individual VACNF grown from an electron-beam lithographysEBLd-patterned catalyst site.sbd Fowler–Nordheim plots of scanned-probeFE measurements from three individual fibers. Maximum field-emissioncurrents exceeding 5mA were observed corresponding to a current densityof approximately 500 kA/cm2 for a nominal VACNF tip diameter of 30 nm.Adopted from Ref. 195.



“sword-in-sheath” mechanism.173 This failure and the in-creased probability of defects with greater mass makeMWCNTs less desirable than SWCNTs, but their strengthstill surpasses that of steel and they have a high modulus.

Carbon whiskers are very thin single crystals that havehigh aspect ratios. As a result of their small size and crystal-line perfection they, too, have very desirable mechanicalproperties, including a Young’s modulus of 700 GPa and ten-sile strength of 20 GPa.169 In spite of these appealing prop-erties, again whiskers are extremely expensive to produceand can be impractical to incorporate into a matrix.169

Carbon fibers, composed of polycrystalline and amor-phous carbon, have diameters ranging from hundreds tothousands of nanometers and can be grown thousands ofmicrometers long. For carbon fibers, tensile strengths areabout an order of magnitude lower than for CNTs, rangingfrom 1.5 to 4.8 GPa and the Young’s modulus is between228 and 724 GPa.169 Carbon fibers retain their high tensilemodulus and strength at elevated temperatures and are notaffected by water, solvents, acids, or bases at room tempera-ture. Even though their mechanical properties are not as as-tounding as those of carbon nanotubes, carbon fibers are rela-tively easy to produce as well as economical.

In order to measure the mechanical properties of indi-vidual carbon nanofibers produced by C-CVD, Gaoet al.observed electromechanical resonance in TEM.174 A nanofi-ber was excited by applicaton of a periodic voltage withrespect to a counterelectrode. The bending modulus ofbamboo-type nanofibers with point defects was,30 GPaand that of nanofibers with volume defect was 2–3 GPa. Thetime-decay constant of nanotube resonance in a vacuum of10−4 Torr was,85 ms. Suchin situTEM method provides apowerful approach to study the mechanics of high aspectratio nanostructures.

One of the applications for which the mechanical prop-erties of carbon fibers are of high importance is in compositematerials.175 Fibers have relatively low densities, makingthem attractive for lightweight applications. Most impor-tantly though, carbon fiber and composite manufacturingprocesses have been developed that are relatively cost effec-tive. In fact, carbon fibers are the most commonly used re-inforcement in advanced polymer-matrix composites.169

Recently, Weisenbergeret al. showed that the mechani-cal properties of polyacrylonitrile fibers could be enhancedby creating a composite fiber reinforced with MWCNTs.They reported a 36% initial modulus increase and a 46%yield strength increase.176

E. Chemical properties

McCreery provides an excellent overview of both thechemical and electrochemical properties of carbon-based ma-terials, including “carbon fibers” that correspond to the ver-tically aligned, catalytically synthesized fibers of thisreview.177 In general, nanofibers are chemically and me-chanically resistant to many standard microfabrication etchand deposition processes including nitric acidsused for Nicatalyst dissolutiond, HF and buffered HF, photoresist spincasting, exposure and development, fluorine-based dry etch

processing, and a variety of PVD, CVD, and PECVD coatingprocesses, including metal, oxide, and nitride depositions.Calcination, electrochemical oxidation, and oxygen-containing plasma processing can be used to etch a nanofibermaterial or to provide oxygen-rich moieties on the nanofibersurface.

Carbon electrode functionalization strategies are a vasttopic in the literature, ranging, for example, from direct im-mobilization of materials to the carbon surface to entrapmentof materials in polymer films or nucleated islands on thesurface of the electrode. Noncovalent modification has beenwidely used, specifically for nanotubes for which the aro-matic nature of the sidewalls provides for strong adsorptionof aromatic compounds.178 These noncovalent strategies areoften preferred when the nanotube’ssp2 structure is to bepreserved. Bahr and Tour provide an overview of covalentfunctionalization strategies specifically for carbon nanotubes,but all are likely applicable to nanofiber modification aswell.179 Due to lack of defect sites, covalent functionalizationof nanotubes is often preceeded by mechanical and oxidativetreatments, such as sonication in sulfuric and nitric acid,pirhanna etchssulfuric acid and hydrogen peroxided, electro-chemical oxidation, or exposure to air or oxygen plasmas.Resultant defect sites are predominantly formed on the endsof nanotubes. The sidewalls of nanotubes, due to their aro-matic nature, are far less reactive, although several sidewallfunctionalization strategies have been developed.180 In con-trast, the edge planes of the stacked structure of nanofibersprovide convenient covalent functionalization along their en-tire length, particularly if these surfaces are oxidatively pre-treated.

Carbodiimide chemistries are some of the most widelyused functionalization strategies for both carbon nanotubesand nanofibers with surfaces activated by oxidative treat-ment. Peptide bonding, i.e., condensation of primary aminesto carboxylic acids, is perhaps one of the more commonlyemployed techniques. Amine-containing materials, includingproteins and DNA, can be covalently coupled to generatedcarboxylic acid sites by use of, for example, 1-ethyl-3-s3-dimethylaminopropyld carbodiimidesEDCd which forms areactive intermediate at the carboxylic acid site that cancouple with available amines to form amides, or to sulf-hydryls forming thiol ester linkages.181

N-hydroxysuccinimidesNHSd is often used to stabilize thereactive intermediate,o-acylisourea in these reactions.

Lin et al.have demonstrated the direct immobilization offunctional enzymessglucose oxidase, GOxd on the tips of thevertical nanostructured carbon elements for the purpose ofmediatorless electrochemical sensing of glucose.182 Here car-boxylic acid sites were generated by electrochemical pre-treatment in 1N NaOHs1.5 V for 90 sd. NHS-stabilized ex-posure to EDCs2 hd was then followed by a 2-h incubationin 2 mg/mL glucose oxidase in 100-mM phosphate- buff-ered salinesPBSd, resulting in condensation of primaryamines of the GOx with activated sites on the carbon ele-ments. Also, using EDC/NHS carbodiimide chemistry,Nguyen et al. have demonstrated the immobilization ofamine- and dye-terminated single-stranded oligonucleotideson the opened ends of the aligned forests of carbon nano-



structures, ultimately for hybridization and detection ofDNA.183 McKnight et al. have employed EDC condensationof whole plasmidss5031 bp pGreenlantern-1d to carboxylicacid groups on nanofiber scaffolds and demonstrated tran-scriptional activity of the bound DNA in nanofiber-penetrated live cells.184 As opposed to using amine-terminated DNA constructs, double-stranded plasmid DNAwas bound putatively at base amines on cytosine, guanine,and adenine.

F. Electrochemical properties

Carbon materials have been used extensively as elec-trodes in chemical production and in measurement equip-ment. In general, carbon-based electrodes feature relativelywide potential windows in aqueous media, are low cost, havelow chemical and electrochemical reactivity, and provide re-producible responses following appropriate surface treat-ments. Various conventional carbon-based electrode systemsinclude glassy carbon, carbon paste, highly ordered pyrolyticgraphite sHOPGd, and wax-impregnated graphites, as re-viewed in Refs. 177 and 185. In line with the intent of thisreview, we focus on the literature aimed directly at verticallyaligned carbon-based nanomaterials.

Several research groups have investigated forests ofclosely packed vertical nanofibers and nanotubes as elect-rodic material. Liet al. characterized densely packed multi-walled nanotube towerssMWNTTd grown by thermal CVDand aligned by the crowding effect.186 Individual nanostruc-tured elements within the electrode matrix featured lengthsfrom 30 to 100mm and diameters ranging from 15 to 80 nmat a densely packed spacing of 100–200 nm. Compared to“paper-film” electrodes prepared from laser ablation synthe-sized and acid-treated SWCNT in the same study, the three-dimensional MWNTT matrix featured lower volumetric ca-pacitances2.0 F/cm3 vs 6.25 F/cm3d but a redox response tothe FesCNd6

3−/4− couple that scaled with the thickness of theMWNTT film. Heat treatment of 450 °C in air for 10 h wasfound to increase the volumetric capacitance to 4.7 F/cm3,perhaps due to removal of amorphous carbon within thethree-dimensional matrix. This scaling of redox with thethicknessslengthsd of the vertical electrode material demon-strated that electron transfer was occurring within the matrixalong the lengths of the nanostructured elements. Peak sepa-ration of the FesCNd6

3−/4− couple was larger than for the pre-pared SWCNT materials170 mV vs 100 mVd and increasedwith heat treatment to 230 mV.

Murphy et al. studied the electrochemical behavior ofless aligned, more crowded nanostructured carbon, grownalso by thermal CVD using Fe2O3 as catalyst.187 For theRusNH3d2+/3+ couple, slightly wider peak separation was ob-served than that obtained with a conventional glassy carbonelectrode, with some peak separation being attributed to theresistive drop in the electrode material. It was recognizedthat volumetric redox of analyte within the nanotube matrixwas occurring. In the same study, oxidation of hydroquinoneresulted in significantly higher charge transfer than that ex-pected from the material within the electrode matrix and thiswas attributed to strong adsorption. Thorough rinsing of the

electrode in 1-M phosphate buffer did not remove the ad-sorbed hydroquinone-benzoquinone analyte, but 30-minsoaking in ethanol could remove these materials.

McKnight et al. characterized the electrochemical prop-erties of as-grown as well as treated VACNF forest elec-trodes of different dimensions, which were grown using dc-PECVD and nickel catalyst.188 Various electrode configur-ations were evaluated using a variety of outer sphere, qua-sireversible analytes including FesCNd6

3−/4− andRusNH3d6

2+/3+ at 0.5-, 1-, and 5-mM concentrations in 100-mM KCl. Figures 33sad and 33sbd provide representativecyclic voltammogramssCV: 100 mV/sd for a large surfaceareas5-mm diameterd VACNF forest corrected for resistivedrop of the electrode, interconnects, and solution. For theseas-grown electrodes, which can be described as stackedfunnel-type carbon nanofibers, the quasireversible couplesexhibited rapid electron-transfer kinetics, with cathodic andanodic peak separations at sweep rates of 100 mV/s verynear the ideal 59 mV for a single electron-transfer reaction,and the anodic/cathodic peak ratio near unity. As with the Liand Murphy study, the CVs in Fig. 33 indicate the diffusion-

FIG. 33. Cyclic voltammogramms of VACNF forests. Responses of5-mm-diameter nanofiber forest electrodes to varying concentrations ofsadruthenium hexamine trichloride andsbd potassium ferrocyanide. Thesevoltammograms are IR compensated to account for resistive losses in theinterconnect/fiber-substrate junction/solution, which are pronounced due tothe relatively large currents from the high-concentration analytes. All sub-sequent voltammograms are not IR compensated. Reprinted with permissionfrom Ref. 191, Copyright 2003 American Chemical Society.



limited electrode responses with peak currents of 10%–15%greater than those expected from a 5-mm-diameter planarelectrode where analyte access is limited to the electrodesurface by semi-infinite planar diffusion. The additionalcharge transferred can be explained by redox of the analyteadsorbed to and within the free volume of the porous elec-trode matrix.186

Tu et al. reported on the use of sparse arrays of verticalnanostructured carbon as electrode material.189 Low site-density arrays were grown by PECVD from randomly elec-trodeposited Ni on a Cr-coated silicon substrate. Magnetron-sputtered SiO2, followed by thick layers of a two-componentsolvent-thinned epoxysM-Bond 610d, were used to insulateand fill the substrate and interfiber spaces. The potted struc-ture was then polished to expose sparse nanofiber tips in aplanar format on the electrode surface. This sparse array ofelectrodes acted effectively as the sum of manys106/cm2dindividual ultramicroelectrodes, providing a redox responseto the FesCNd6

3−/4− couple that indicated radial diffusivetransport to the active nanofiber sites. Liet al. generatedpolished planarized electrodes of Ni-catalyzed PECVD-grown nanofibers insulated and tip exposed from thick layersof tetraethoxysilane CVD SiO2.

190 In this case, individuallyaddressed pads of closely packed forests of fibers presentedplanar diffusion-limited redox responses to FesCNd6

3−/4−,while lower-density nanofiber spacingsachieved viaelectron-beam lithographyd produced the sigmoidal responseof radial diffusion. Using ac and differential pulse voltamme-try, low-density electrodes provided higher sensitivitysver-sus high densityd of response to surface-immobilized fer-rocene.

Guillorn et al. demonstrated the ability to electricallyconnect with an individually addressed nanofiber by using amultistep microfabrication process.84 Here dc-PECVD-synthesized nanofibers were grown from electron-beam pat-terned nickel dots on tungsten interconnects. This work wascontinued by McKnightet al. as the electrochemical proper-ties of photolithographically defined, individually addressednanofiber elementsstypical lengths of 4–10mmd and nanofi-ber tips s,250-nm-diameter semihemisphered were charac-terized against FesCNd6

3−/4−, RusNH3d62+/3+, and IrCl6

2−/3−.191

The redox response of electrodes was dependent on the sur-face area of the exposed nanofiber material, thus indicatingthat electron transfer occurs along the entire length of thesevertically aligned structures. Electrode interfacial capaci-tance was evaluated at,140 mF/cm2 by performing volta-mmetry of various length fiberss2–12mmd at multiple scanrates in 100-mM KCl. The potential window of these elec-trodes was evaluated in several aqueous solvents, including100-mM KCl and phosphate-buffered salinef1.2 to −1.3 Vvs Ag/AgCls3-M KCldg, methanols1.2 to −0.6 Vd, 1-N sul-furic acid s1.5 to −0.5 Vd, and sodium hydroxides0.8 to−1.5 Vd. Tip exposedssemihemisphericald and electrodeswith an overall surface area,10 mm2 adhered well to clas-sic ultramicroelectrode steady-state response. Longer elec-trodes sfibers with lengths .10 mm and surface areas.10 mm2d began to indicate a quasisteady-state response ofcylindrical ultramicroelectrodes, with slight diffusion-limited

peaking of the oxidation and reduction curves at scan rates of200 mV/s for ruthenium hexamine trichloride in 100-mMKCl.

Recent work with vertically aligned structures exploitssome of the classic electrochemical modifications performedwith other carbon-based electrode systems to provide spe-cific modification of the electrode response. Chenet al. havedemonstrated the electrodeposition of polypyrrolespPydfilms over well-aligned vertical nanostructured carbon.192

Conformal polypyrrole films were electropolymerized ontoeach individual element of vertical forests of carbon fibersfrom 17.3-mM pyrrole monomer in 0.1-M LiClO4 at anodicpotentials f.0.4 V versus standard calomel electrodesSCEdg. As an electronically conductive polymer film, pPycan be used to immobilize redox mediators, including en-zymes, to provide specificity to nanostructured carbon-basedelectrodes. Direct immobilization of enzymes to the nanofi-ber material has also been demonstrated by Linet al. sseeprevious sectiond to provide probing specificity to glucose.Similarly, direct immobilization of oligonucleotides has beendemonstrated by Liet al. and Koehneet al. for detection ofhybridization of subattomole DNA targets by usingRusbpyd3

2+-mediated guanine oxidation.190,193

G. Postsynthesis processing

There are many applications of carbon nanofibers thatrequire the integration of their synthesis process into theoverall device microfabrication process. Very often synthesisis not the first or the last step, which generates an issue ofcompatibility of the processes and materials. First of all, theintegration of the growth process requires use of large-scalereactors for substrates compatible with microelectronic de-vice manufacturing equipmentse.g., 100 mm or larger diam-eter round Si and quartz wafersd. Second, multilevel process-ing requires the ability to synthesize high-quality VACNF inprecisely defined locations, as subsequent fabrication pro-cesses need to be performed in registry with defined patterns.This can be achieved by deposition of catalyst in areas de-fined by lithography. Third, the nanofibers must be able towithstand the conditions to which they are exposed duringstandard microfabrication processing used in the productionof integrated circuits and microelectromechanical systemssMEMSd. Fortunately, in many cases this compatibility hasbeen demonstrated.26,28,30,31,33,39,190,194As in any process en-gineering, there are still many issues that need to be resolvedfor each specific application and set of materials. We brieflydescribe some of the more useful processing steps below.

Perhaps the most fundamental operation is the deposi-tion of thin films on VACNF arrays. The effect of SiO2 andamorphous Si deposition onto VACNF using rf PECVD hasbeen investigated.195 The layers could be uniformly depos-ited onto the fibers resulting in a conformal coating. ThermalCVD using tetraethylorthosilicatesTEOSd at 700 °C in aquartz tube furnace was also employed for conformal SiO2

coating of VACNFs.190,193 Physical vapor deposition tech-niques including sputtering and electron-beam evaporationhave also been used successfully for this purpose, providinga mechanism to modify the surface of the VACNF. Of par-



ticular interest was the coating onto the VACNF of dielectriclayers that could be selectively removed from the regions ofthe fiber with subsequent microfabrication processes. Figure34 shows the VACNFs conformally coated with and subse-quently directionally emancipated from SiO2. This providesthe ability to control the amount of the fiber body participat-ing in electron transport independent of the aspect ratio orgeometry of the fiber. The detailed studies of controlled re-moval of insulator coating from the tips of the nanofiberswere performed by Ikunoet al.196

Once the material has been deposited onto a substrate,typically some sort of patterning is performed. Photolithog-raphy has long been established as the standard workhorsetechnique used in the microelectronics industry for this pur-pose. This process involves the patterning of ultravioletsUVdsensitive polymer layerssphotoresistsd. Photoresists are typi-cally spin-cast onto substrates at speeds ranging from1000 to 6000 rpm. Simple experiments involving the depo-sition of photoresist layers between 200 nm and 2-mm thick-ness demonstrated that even high aspect ratio VACNFs sur-vive this processing. Not only can photoresist be applied tosubstrates with VACNF on them, but it can also be exposedand developed using well-established techniques. Moreover,photoresist can be stripped from substrates containing fibersby using a combination of organic solvents and ultrasonicagitation with no damage to the structural integrity of theVACNF. Furthermore, brief exposure to an oxygen rf plasma,usually used to remove remains of photoresist, does not sig-nificantly affect nanofibers. Once a pattern has been exposedin a layer of photoresist and developed, it is transferred ontothe substrate by either the addition or removal of material,referred to as additive or subtractive pattern transfer, respec-tively. In the case of subtractive pattern transfer, some formof etching is used to remove the desired layer or layers from

the patterned area. These processes include various forms ofplasma-based etching along with wet chemical etching. Ma-terial deposited onto VACNF can be removed using severalcombinations of these techniques without significantly dam-aging the fiber. The fact that the VACNF is composed ofgraphitic carbon provides it with a relatively robust body thatcan withstand bombardment of ions during plasma-basedprocesses and any sort of chemical reaction with the excep-tion of those designed to attack carbonaceous materials. Thetolerance of VACNFs and CNTs to HF provides an isotropicmethod of SiO2 removal. This property was also used for theremoval of CNT from a SiO2 substrate for successive use ortransfer to another substrate, e.g., one that cannot withstandCNT growth temperatures.197

The processing techniques described above have resultedin the fabrication of several microscale structures thatexploit the unique nanoscale properties of theVACNF.26–28,30,32,33,35,39,190,194A process for passivating thebody of the fibers with an insulating thin film while leavingthe tips electrically and electrochemically active was devel-oped for the fabrication of electrochemical probes with highspatial resolution. This technology was then combined with aprocess for creating individually electrically addressableVACNF on an insulating substrate to produce arrays of highaspect ratio electrochemical probes.84 The changes in theelectrochemical activity of the carbon nanofiber-based elec-trodes, due to exposure to different microfabrication pro-cesses, are reviewed in Ref. 188.

Chemical mechanical polishingsCMPd is frequentlyused in microelectronic circuit manufacturing to planarizesubstrate morphology, and can be applied to films depositedonto VACNF. Deep layers of SiO2 have been successfullyplanarized without damaging the buried VACNF. Continua-tion of this process has been shown to remove sectionsof the VACNF at the same rate as the SiO2, leaving theexposed fiber core coplanar with the surrounding oxidetopography.190,193This strategy has enabled the synthesis ofcoplanar electrode arrays for electrochemical applicationswhere many fibers perform transduction in a parallel fashion.

All of the above show the ability of carbon nanostruc-tures to withstand numerous processing conditions but meth-ods of CNF modification or removal are also important fordevice fabrication. For example, a VACNF can be used as asacrificial template for the creation of vertically orientednanofluidic devices.33,63 In this process, arrays of VACNFsare grown on either Si or Si3N4 membranes and are coatedwith a thin conformal film using PECVD or low-pressurechemical-vapor depositionsLPCVDd. The tips of the fibersare liberated using the process described above for the indi-vidual electrochemical probes. Following the removal of theremaining photoresist, the substrates are subjected to a briefetch in concentrated nitric acid to remove the Ni catalystparticle at the tip, and an O2 reactive ion etch is then used toremove the body of the fiber from within the thin-film SiO2

tube encasing it. Once the body of the fiber has been entirelyremoved, the bottom of the tube can be opened to create ananometer-sized pipe structure, or nanopipesFig. 35d.

The accessibility of the catalyst particle allows for itseasy removal from the VACNF. When Ni is used as a cata-

FIG. 34. Deposition and removal of conformal PECVD SiO2 on VACNFarrays.sad VACNF array on Si as grownsbd following the deposition of a1-mm-thick PECVD SiO2 layer scd, sdd after 40 min of reactive ion etchingsRIEd using a CHF3/O2 plasma. All micrographs were taken in a SEM at a45° angle. Adopted from Ref. 195.



lyst, the catalyst particle can be removed by nitric acid etchthat is preceded by a brief oxygen plasma, as describedabove. A simple purification process for the removal of ironcatalyst particles at the tip of VACNFs synthesized usingC-PECVD is reported by Nguyenet al.38 In this process, thefirst step involves thermal oxidation in air, at temperatures of200–400 °C, resulting in the physical swelling of the ironparticles due to formation of iron oxide. Subsequently, thecomplete removal of the iron oxide particles is achieved withdiluted acids12% HCld. The purification process appears tobe very efficient at removing all of the iron catalyst particles.Electron microscopy images and Raman spectroscopy dataindicate that the purification process does not damage thegraphitic structure of the nanofibers.

In addition to thermal oxidation in oxygen atmosphereand using an oxygen plasma etch, carbon nanostructures canbe removed electrochemically. MWNTs were etched at po-tentials more positive than 1.7 V vs Ag/AgCls3-M NaCld inan aqueous 0.1-M KCl electrolyte solution.198 Hinds et al.used this same method to tailor the length of CNTs.199

IV. APPLICATIONS

A. Microfabricated field-emission electron sources

The design and fabrication of vacuum microelectronicdevices based on field-emission cathodes of various typeshave proven to be a fruitful area of research for many groupsover the last three decades.200 Recently, the construction ofdevices using various forms of nanostructured graphitic car-bon has received a significant amount of attention.195,201,202

Xu and Brandes presented data on the operating CNT-basedgated cathode device in 1998 by employing disordered matsof multiwalled CNTs sMWNTd grown within electrostaticgating structures by thermal CVD.203 Several other groupshave fabricated similar devices using a variety of differenttechniques, all achieving similar results. Due to the multi-plicity of CNTs in the cathodes of these devices, numerousemission sites are present, yielding devices capable of deliv-ering high emission currents at relatively low operating volt-

ages. However, it has been demonstrated that this featuremay be the source of significant device-to-device nonuni-formity.27

Devices using mats of CNTs for the field-emission cath-ode possess no way to precisely control the location or den-sity of the emission sites. This significantly complicates theconstruction of field-emission devices that produce well-focused emitted electron beams. While focused beams arenot required for every application of microscale field-emission guns, beam focusing is essential for more demand-ing uses such as electron sources for electron-beam lithogra-phy or electron microscopy. In these applications, VACNFsoffer some additional advantages over other nanostructuredgraphitic carbon-based cathodes. In particular, the ability toconstruct devices with single VACNF cathodes greatly sim-plifies the design of integrated focusing and deflection opticsand has a significant impact on the minimum achievablebeam diameter. This stems from the fact that the location ofthe VACNF can be precisely controlled within the devicestructure, at least within the dimensional tolerances of theVACNF themselves. It follows that the emission site, pre-sumed to be at or near the tip of the VACNF, can also becontrolled with accuracy of probably a few nanometers. Thiscannot be said of devices employing ensembles of nanotubesor nanofibers, for which multiple active emission sites arespread out over an entire micron-scale cathode.

In situ growth of VACNF into prefabricated electrostaticgating structures was demonstrated by Guillornet al.32 Whilethis approach is capable of realizing device structures, thereare some disadvantages to this process. Primarily, the dcplasma discharge used to grow the VACNFs was found tocause charge-induced damage to the gating structure. Never-theless, Pirioet al. have used a variant of this process tofabricate gated cathode structures with multiple VACNFemitters on each cathode.202 This work demonstrated thatarrays of such devices displayed field-emission characteris-tics that agreed with the Fowler–Nordheim model with field-emission threshold voltages as low as 10 V. However, thepresence of multiple VACNF withinin the cathode negatesthe point of using VACNF for this type of device, and pro-vides no advantage over CNT-based devices. In a later work,the ability to fabricate devices with single, low-qualityVACNF cathodes was demonstrated using this technique.204

The operation of these devices has not been published at thistime.

A robust process for fabricating integrated field-emissionelectron sources using single VACNF cathodes was demon-strated by Guillornet al.29,195This process utilized the abilityalready noted of the VACNF to withstand microfabricationprocessing to synthesize functional devices in a wafer-scalefabrication technology. Tested devices displayed behaviorconsistent with the Fowler–Nordheim model161 of field emis-sion, and operated with very high efficiency with less than2% of the emitted current being collected by the gate elec-trode. This process was later improved upon by forming thegate electrode using a self-aligned process.31 This eliminatedthe need for definition of the gate electrode by an aligned

FIG. 35. Nanopipes made using VACNFs as nanotemplates. Adopted fromRef. 33.



lithographic exposure step, thus significantly simplifying de-vice construction without compromising the operational per-formance.

A microfabricated electron gun structure with an inte-grated electrostatic focusing electrode using a single VACNFas the field-emission cathode was reported by Guillornet al.in 2004sFig. 36d.205 This device was fabricated using a vari-ant of the device processing techniques reported in their ear-lier works, with the addition of an out-of-plane focusingelectrode that is aligned coaxially with the VACNF cathodeand gate electrode apertures. The performance of the inte-grated focusing electrode was investigated by modulating thepotential on this electrode while projecting the image of theemitted electron beam onto a phosphor screen. This wasfound to have a significant impact on the beam diameter. Theobserved experimental results were also in good agreementwith numerical simulations of the device performance,thereby revealing that true refocusing of the electron beamhad occurred. Devices of this design were used to image

submicron lines in an electron-beam resist to demonstratetheir potential use in a massively parallel electron-beam li-thography system.206

B. Microfabricated x-ray sources

X-ray sources are widely used in industrial, medical, andscientific instruments. X rays are usually generated by bom-bardment of a metal target with energetic electrons. Theproperties of the x-ray source, such as beam size, cruciallydepend on the electron source. Cold electric-field enhancedsEFEd sources with carbon nanostructure cathodes show im-proved performance and have lead to portable and miniaturex-ray sources for industrial and medical applications.207

Sugieet al.used carbon nanotubes grown by C-CVD on a Wwire as a primary electron source.208 They showed at a mod-erately low pressure of 2310−7 Torr, the CNT had a stableoperation for more than 1 h and enabled the acquisition ofclear x-ray images. The intensity of the x rays passingthrough a beryllium window was measured as a function ofthe target potential from 10 to 60 kV at a constant current of10−7 A. The x-ray intensity increased approximately expo-nentially with the linear increase in the target potential:0.005 mSv/h at 20 kV and 4 mSv/h at 50 kV.

Yue et al. fabricated a device that can readily produceboth continuous and pulsed x rays with a programmablewave form and repetition rate.209 A total emission current of28 mA was obtained from a 0.2-cm2-area CNT cathode. Thex-ray intensity is shown to be sufficient to image a humanextremity at 14 kVp and 180 mA s. Pulsed x-ray generationwith a repetition rate greater than 100 kHz was readilyachieved by programming the gate voltage.

The advantages of carbon nanofibers for a point x-raysource have been demonstrated by Matsumoto andMimura.210 A point x-ray emission was obtained from a di-ode configuration composed of a graphite-nanofiber coldcathode and a conical-shaped copper metal anode. Whencombined with a highly sensitive charge coupled devicesCCDd camera with a scintillator, this x-ray source was thenused to obtain x-ray transmission images of a tungsten meshand an x-ray test chart. The spatial resolution of this x-rayradiography system was on the order of 10mm and the ac-quisition time for these images was less than 10 s. This com-bination of the point x-ray source and the sensitive detectionsystem offers a high-resolution x-ray radiography system.

C. Electrochemical probes

The incorporation of micro- and nanoscale carbon mate-rials into electrodes has long served to augment electrodeperformance for specific applications. Modification of carbonpaste electrodes with nanotubes is frequently employed, forwhich the nanoscale structure of the incorporated nanotubesenables improved coupling of the bulk electrode with theredox center of, for example, immobilized redox mediatingenzymes. Bundles of nanotubes have also been incorporatedinto individual microscale probing elements, and in this con-figuration serve as a material of choice for a variety of mi-croelectrophysiological measurements and electroanalyticaldetection techniques and applications, including electro-

FIG. 36. Microfabricated structure with a single VACNF as a cathode andintegrated focusing electrode. Adopted from Ref. 205.



chemical detection of microfluidic separations,211 scanningelectrochemical microscopy,212 and dynamic electrophysi-ological measurements in and around excitable cells.213 Ad-vances in the ability to incorporate vertically aligned nanos-cale carbon electrodes into multielement devices, such asarrays of individually addressable probing elements, shouldprovide unique platforms that will introduce a high level ofparallelism into these well-established electrophysiologicaland electroanalytical techniquessFig. 37d.

D. Intracellular gene delivery devices

Methods to introduce genetic material into cells haveremained a central theme in molecular biology, biotechnol-ogy, and medicine. While many methods exist, includingtarget-specific viral vectors, electroporation,214 liposomefusion,215 and microprojectile bombardment,216 it is perhapsmicroinjection that provides the most effective mechanismacross a broad range of cell types. In this application, indi-vidual microscale needles are used for the direct injectionand delivery of genetic material to the intracellular domainand to the nuclear domain of eukaryotes. Recently, a parallelmicroinjection-based method termedimpalefectionhas beendemonstrated using vertical aligned arrays of carbon nanofi-bers sFig. 38d.184,217 In this technique, arrays of nanofibersare either coated with DNA or covalently modified withDNA and are directly inserted into many cells simulta-neously using a variety of techniques, including centrifuga-tion of cells onto the nanofiber array or direct pressing of theDNA-modified array into tissue matrices. The size scale andmorphology of the vertically aligned nanofibers provide ef-fective penetration into cells, including into the nucleus ofmammalian cells, where the DNA is either released from thenanofiber and expressed by the cell, or transcribed evenwhile tethered to the nanofiber scaffold. In either case, pen-etrated cells can recover from the procedure, with theirplasma membranes effectively resealing around the penetrantnanoscale structure. Proliferation of cells occurs upon thenanofiber substrate and, often, these proliferating cells tendto maintain intracellular residence of the penetrant nanofibersfor extended periods of timesweeksd.

In the course of impalefection experiments, the require-ments for successful use of VACNF arrays as material deliv-ery vectors have been identified: nanofibers are required tobe sharp, sparse, considerably tall, rigid, and able to retainDNA during the insertion. For typical mammalian cells thathave linear dimensions on the order of 10mm it has beendetermined that conical 6–10-mm-long nanofibers with tipdiameters of 100 nm or less, positioned in an array with5-mm spacing between fibers, provide effective impalefec-tion.

E. Scanning probe microscopy tips

The atomic force microscopesAFMd has emerged as avaluable tool for probing various surfaces in semiconductormetrology and profilometry, imaging of biological samples,

FIG. 37. Microarrays of electrochemical nanoprobes. Scanning electron mi-crographs of an individual chip of 40 nanofiber probing elements. At far left,image taken following the oxide deposition step and prior to SU-8 passiva-tion to facilitate imaging the underlying interconnect structuresfirst insetd.At right, a four-element array is shown following device completion. Indi-vidual fibers feature oxide and some SU-8 passivation, with only the ex-treme tip being exposed for electrochemical activity. SU-8 on the substratehides the underlying interconnect structure in these images. Reprinted withpermission from Ref. 191, Copyright 2004 American Chemical Society.

FIG. 38. sColor onlined SEM images of VACNF arrays as gene deliverydevices. Electron micrographs of mammalian cellssSP2/0-AG14; mousemyeloma13d following centrifugation at 600 G onto an indexed nanofibergene delivery chip and culture for 3 dayssviewing angle is 30°d. The cellswere fixed with 2% gluteraldehyde in PBS and dehydrated with methanolprior to electron microscopy. SP2 cells were used for this image as they area suspension cell line and therefore do not attach and spread on the nanofi-ber platform, thus facilitating clearer images. Nonetheless, some materialhas clearly agglomerated onto the fiber surfaces, as seen in panel Csfalsecolor addedd. Reprinted with permission from Ref. 217, Copyright 2004American Chemical Society.



and numerous other surfaces since its inception.218 Conven-tional AFM cantilevers have a pyramidal-shaped siliconprobe, and advances in micromachining and fabrication tech-nology allow this probe tip to be made as small as 10 nm. Asmall tip diameter is important for achieving high resolution,but a critical problem with the conventional silicon tips istheir brittleness. As the tip wears out quickly, degradation ofimage resolution is unavoidable, resulting in the need tochange the probe frequently. CNT probes, due to their smalltip diameter s1–10 nmd, offer very high resolution.219,220

Their extraordinary strength and the ability to retain struc-tural integrity during operation in contact or tapping modemake CNT tips very robust. Figure 39 shows an example ofa CNT tip grown by thermal CVD at the end of a siliconcantilever. The ability to functionalize carbon nanostructuresbrings extended range of applications to scanning probetechniques.221 Deterministic synthesis of vertically alignedcarbon nanofibers by C-PECVD allows large-scale produc-tion of AFM tips on a wafer scale.35

F. Nanoelectromechanical devices „NEMS…

Nanoelectrochemical devicessNEMSd utilize geometri-cal, electrical, and mechanical properties of carbon nano-structures. Atomic force microscopy tips, discussed in the

previous section, can also be classified as NEMS devices. Inthis case, CNTs and CNFs are parts of the AFM cantilevermicroelectromechanical device. Nanotubes themselves canbe used as cantilevers as was demonstrated by Poncharaletal. who constructed a nanobalance that can be used forweighing nanoparticles.222 Even though manipulation ofsingle atoms has been demonstrated,223 manipulation ofnanoscale objects still remains a challenge. Kim and Lieberfabricated nanotweezers constructed using carbon nanotubesthat may provide a universal means of nanomanipulation.224

The integration of nanotweezers with AFM cantileversbrings an additional level of control over positioning andimaging.225

G. Nanoporous membranes

Membranes are used in a multitude of applications suchas chemical separations, drug delivery, and waste water re-mediation, and membranes with nanometer-size pores are ofspecial importance. Carbon nanotubes and carbon nanofiberscan be used for this purpose in geometries parallel and per-pendicular to the transport directions. Vertically aligned car-bon nanotubes can be incorporated into a membrane by em-bedding into a polymer film.199 In this case transport occursinside the nanotubes perpendicular to the substrate. In an-other geometry the nanotubes or nanofibers can be placedperpendicular to the transport direction.194 In this case theeffective “pore size” is controlled by their density and size,which depends in turn on the catalyst particle preparation.Such membranes can be incorporated into microfluidic chan-nels, for example, to mimic cell membrane functionalitysFig. 40d.226

H. Other applications

There are many potential applications for carbon nano-tubes and nanofibers that the research community continuesto explore.227 Some, such as composite materials, do not relyon deterministic synthesis.175 In this case the bulk material isused as a filler of a polymer matrix, though directional align-ment or alternating alignment may be needed to develop en-hanced thermal or electrical conduction or great strength.Composite fibers produced by spinning CNTs with polymerscan be made tougher than any natural or synthetic organicfiber. They can be used to make superstrong and superlighttextiles.228 They can also be used to make fiber supercapaci-tors that are suitable for weaving into textiles.

Hydrogen and energy storage are other areas in whichcarbon nanofibers can be of high importance, but there ismuch debate over the actual potential.229–233There is a con-tinuous search for cleaner renewable energy sources, andhydrogen-based alternatives are being considered as attrac-tive candidates. One of the main challenges for use of hydro-gen is its storage, as currently there are no efficient and safemethods. Approaches now in use or under development in-clude compressed hydrogen gas, liquid hydrogen, chemicalstorage in hydridesse.g., metal hydridesd, and gas-on-solidphysisorption. Recently there were several reports on theperhaps extraordinary hydrogen storage ability of carbonnanostructured materials. For example, Chamberset al. re-

FIG. 39. VACNF as scanning probe tips synthesized on AFM cantilevers onthe wafer scale. Adopted with permission from Ref. 35, Copyright 2004American Chemical Society.



ported that a material consisting of carbon nanofibers is ca-pable of sorbing and retaining in an excess of 20 LsSTPd ofhydrogen per gram of carbon when the nanofibers are ex-posed to the gas at pressures of 120 atm at 25°C,21 which isan order of magnitude better than with the other means ofhydrogen storage. They argued that this result, based on themeasurements of pressure drop, is due to the unique crystal-line arrangement existing within the graphite nanofiber struc-ture, in which the graphene planes make a system of subna-nometer pores with only the edge sites exposed. Since theinterplanar distance within the material is 3.37 Å, sorption ofmolecular hydrogen, which possesses a kinetic diameter of2.89 Å, is a facile process owing to the short diffusion path.Other reports with lower but still impressive values for thespecific adsorption have also appeared. However, latelymany of these results were disputed and much more moder-ate adsorption capacities in the range 0–2.5 wt %sat lowtemperatures and ambient pressuresd were quoted.sSee Ref.234 and references within.d

VACNFs have been proposed for use in charge storagedevices.192 Ren et al. proposed to use VACNF electrodescoated with an electrically conducting polymerssuch aspolypyrroled for electrical energy storage devices such as re-chargeable batteries, fuel cells, and capacitors. For manysuch applications, a high charge capacity is required. In orderto increase the capacity of a polymer battery without dete-riorating the electrodic performancescharging/dischargingratesd due to long ion diffusion time and migration length inthe thick film, high surface area is required. The VACNFforest electrodes offer a high surface area conductive sub-strate.

The high surface area, size, hollow geometry, and chemi-cal inertness of carbon nanotubessCNTsd make them attrac-tive for demanding applications in the field of gas sensing.235

To date, studies on possible applications of CNTs have beenfocused either on individual single-walled carbon nanotubessSWNTsd as sensitive materials for O2, NO2 and NH3 sens-ing, or on multiwalled carbon nanotubesMWNTd mats asNH3, CO, CO2, humidity, and O2 gas sensors.

Carbon materials have important advantages for use incatalysis including236 sid resistance to acid/base media,sii dpossibility to control, up to certain limits, the porosity andsurface chemistry, andsiii d easy recovery of precious metalsby support burning, resulting in low environment impact. Inrecent years, carbon nanofibers and carbon nanotubes havebeen shown to be promising candidates as catalyst supports.The major limitation for their extensive use so far remainshigh-ost and low-roduction volumes. Although a relativelyhigh-yield process, CVD is a batch-type process and cannotbe easily scaled up. There are many reports demonstratingthat the characteristics of CVD-grown carbon nanofibers arepromising for catalyst support. For example, Pesantet al.recently demonstrated that CNF-supported palladium cata-lyst showed higher conversion performance in liquid-phasehydrogenation of nitrobenzene into aniline, compared to pal-ladium catalyst supported on high surface area activatedcharcoal.237 They explain such increase by the highersurface-to-volume ratio of CNF material.

Another use of enhanced electron field-emission charac-teristics of carbon nanotubes is in electric lighting devices.238

It is interesting to note that the proposed application for “car-bon filaments” in 1889 was for a lighting device as well asfilament of an electric light bulb.1

V. FUTURE DEVELOPMENTS AND CONCLUSIONS

Significant progress has been made to achieve controlledsynthesis and directed assembly of carbon nanostructures.However, there is still a need for a process that operates attemperatures that can be tolerated by a larger variety of sub-strate materials, yet provides greater control of nanostructureproperties. A better understanding of the fundamental mecha-nisms of nanostructure nucleation and growth points the wayfor future work. The key to the catalytic growth of carbonnanofibers is in the processes that occur at the surface of orwithin the catalytic particle. Plasma processing provides away to control these processes via interaction with the elec-tromagnetic field. The electric field plays a crucial role in the

FIG. 40. VACNF forests as cell mimic membranes. Electron micrographs ofthe cell mimic structure. Panel A illustrates the overall design of the cellmimic device and the integrated microfluidic structure. The microfluidicinlet and outlet channels that allow fluid flow to the array of the cell struc-tures are 50mm wide and 10mm deep. Panels B and Cs30° viewing angledshow close-ups of a single cell and the component nanofibers, respectively.Reprinted with permission from Ref. 226, Copyright 2004 American Chemi-cal Society.



alignment control of carbon nanofibers by the coupling ofmechanical force to the synthesis process. High-frequencyelectromagnetic fields, augmented by field enhancement atthe tip of a nanofiber, may be the way to control the tem-perature of the catalyst separately from the substrate. An-other major challenge is the precise control of the nanofiberstructure. The growth of isolated single vertically alignedcarbon nanotubes, mulitwalled or singlewalled, is yet to bedemonstrated. We are just beginning to gain an understand-ing of the processes that control the arrangement of graphenelayers of a growing nanostructure. The relation of a carbonnanofiber structure and catalyst particle shape, crystallo-graphic orientation, and their dynamics will eventually beunderstood and an additional level of control of synthesiswill be achieved, bringing future possibilities for practicalnanoscale devices due to improved characteristics.

ACKNOWLEDGMENTS

We would like to thank I. A. Merkulov and G. Eres forfruitful discussions. This work was sponsored by MaterialSciences and Engineering Division Program of the DOE Of-fice of Science under Contract No. DE-AC05-00OR22725with UT-Battelle LLC, the Defense Advanced ResearchProjects AgencysDARPAd under Contract No. DE-AC05-00OR22725, and the Center for Nanophase Materials Sci-encessCNMSd Research Scholar Program.

1T. V. Hughes and C. R. Chambers,Manufacture of Carbon Filaments, USPatent No. 405, 480,s1889d.

2L. V. Radushkevich and V. M. Lukyanovich, Zh. Fiz. Khim.26, 88s1952d.

3R. T. K. Baker, Carbon27, 315 s1989d.4R. T. K. Baker and P. S. Harris, inChemistry and Physics of Carbon,edited by P. L. WalkersMarcel Dekker, New York, 1978d, pp. 83–165.

5H. W. Kroto, J. R. Heath, S. C. Obrien, R. F. Curl, and R. E. Smalley,NaturesLondond 318, 162 s1985d.

6S. Iijima, NaturesLondond 354, 56 s1991d.7Y. Chen, Z. L. Wang, J. S. Yin, D. J. Johnson, and R. H. Prince, Chem.Phys. Lett. 272, 178 s1997d.

8Z. F. Ren, Z. P. Huang, J. W. Xu, J. H. Wang, P. Bush, M. P. Siegal, and P.N. Provencio, Science282, 1105s1998d.

9V. I. Merkulov, D. H. Lowndes, Y. Y. Wei, G. Eres, and E. Voelkl, Appl.Phys. Lett. 76, 3555s2000d.

10Z. F. Renet al., Appl. Phys. Lett.75, 1086s1999d.11K. P. De Jong and J. W. Geus, Catal. Rev. - Sci. Eng.42, 481 s2000d.12N. M. Rodriguez, J. Mater. Res.8, 3233s1993d.13M. Meyyappan, L. Delzeit, A. Cassell, and D. Hash, Plasma Sources Sci.

Technol. 12, 205 s2003d.14H. J. Dai, Surf. Sci.500, 218 s2002d.15M. S. Dresselhaus and M. Endo, inCarbon Nanotubess2001d, pp. 11–28.16K. B. K. Teo, C. Singh, M. Chhowalla, and W. I. Milne, inEncyclopedia

of Nanoscience and Nanotechnology, edited by H. S. NalwasAmericanScientific Publishers, 2003d, pp. 1–22.

17A. Huczko, Appl. Phys. A: Mater. Sci. Process.74, 617 s2002d.18L. M. Dai, A. Patil, X. Y. Gong, Z. X. Guo, L. Q. Liu, Y. Liu, and D. B.

Zhu, ChemPhysChem4, 1150s2003d.19A. Krishnan, E. Dujardin, M. M. J. Treacy, J. Hugdahl, S. Lynum, and T.

W. Ebbesen, NaturesLondond 388, 451 s1997d.20M. Endoet al., Appl. Phys. Lett.80, 1267s2002d.21A. Chambers, C. Park, R. T. K. Baker, and N. M. Rodriguez, J. Phys.

Chem. B 102, 4253s1998d.22L. Delzeit, I. McAninch, B. A. Cruden, D. Hash, B. Chen, J. Han, and M.

Meyyappan, J. Appl. Phys.91, 6027s2002d.23Y. Huang, X. F. Duan, Y. Cui, L. J. Lauhon, K. H. Kim, and C. M. Lieber,

Science294, 1313s2001d.24V. I. Merkulov, A. V. Melechko, M. A. Guillorn, D. H. Lowndes, and M.

L. Simpson, Appl. Phys. Lett.79, 2970s2001d.

25J. Koehne, H. Chen, J. Li, A. M. Cassell, Q. Ye, H. T. Ng, J. Han, and M.Meyyappan, Nanotechnology14, 1239s2003d.

26X. J. Yanget al., Nano Lett. 3, 1751s2003d.27M. A. Guillorn, M. D. Hale, V. I. Merkulov, M. L. Simpson, G. Y. Eres, H.

Cui, A. A. Puretzky, and D. B. Geohegan, Appl. Phys. Lett.81, 2860s2002d.

28L. R. Baylor et al., J. Vac. Sci. Technol. B20, 2646s2002d.29M. A. Guillorn, A. V. Melechko, V. I. Merkulov, E. D. Ellis, C. L. Britton,

M. L. Simpson, D. H. Lowndes, and L. R. Baylor, Appl. Phys. Lett.79,3506 s2001d.

30V. I. Merkulov, A. V. Melechko, M. A. Guillorn, D. H. Lowndes, and M.L. Simpson, Chem. Phys. Lett.350, 381 s2001d.

31M. A. Guillorn, A. V. Melechko, V. I. Merkulov, D. K. Hensley, M. L.Simpson, and D. H. Lowndes, Appl. Phys. Lett.81, 3660s2002d.

32M. A. Guillorn, M. L. Simpson, G. J. Bordonaro, V. I. Merkulov, L. R.Baylor, and D. H. Lowndes, J. Vac. Sci. Technol. B19, 573 s2001d.

33A. V. Melechko, T. E. McKnight, M. A. Guillorn, V. I. Merkulov, B. Ilic,M. J. Doktycz, D. H. Lowndes, and M. L. Simpson, Appl. Phys. Lett.82,976 s2003d.

34J. Moser, R. Panepucci, Z. P. Huang, W. Z. Li, Z. F. Ren, A. Usheva, andM. J. Naughton, J. Vac. Sci. Technol. B21, 1004s2003d.

35Q. Ye, A. Cassell, H. Liu, K.-J. Chao, J. Han, and M. Meyyappan, NanoLett. 4, 1301s2004d.

36Z. G. Mao, A. Garg, and S. B. Sinnott, Nanotechnology10, 273 s1999d.37W. Kratschmer, L. D. Lamb, K. Fostiropoulos, and D. R. Huffman, Nature

sLondond 347, 354 s1990d.38C. V. Nguyen, L. Delziet, K. Matthews, B. Chen, and M. Meyyappan, J.

Nanosci. Nanotechnol.3, 121 s2003d.39A. V. Melechkoet al., J. Vac. Sci. Technol. B20, 2730s2002d.40A. M. Cassell, J. A. Raymakers, J. Kong, and H. J. Dai, J. Phys. Chem. B

103, 6484s1999d.41M. J. Bronikowski, P. A. Willis, D. T. Colbert, K. A. Smith, and R. E.

Smalley, J. Vac. Sci. Technol. A19, 1800s2001d.42W. Z. Li, S. S. Xie, L. X. Qian, B. H. Chang, B. S. Zou, W. Y. Zhou, R. A.

Zhao, and G. Wang, Science274, 1701s1996d.43A. Moisala, A. G. Nasibulin, and E. I. Kauppinen, J. Phys.: Condens.

Matter 15, S3011s2003d.44H. Kanzow and A. Ding, Phys. Rev. B60, 11180s1999d.45L. Delzeit, C. V. Nguyen, B. Chen, R. Stevens, A. Cassell, J. Han, and M.

Meyyappan, J. Phys. Chem. B106, 5629s2002d.46G. Eres, A. A. Puretzky, D. B. Geohegan, and H. Cui, Appl. Phys. Lett.

84, 1759s2004d.47G. Eres, A. A. Puretzky, D. B. Geohegan, and H. Cui, In situ Control of

Catalyst Efficiency in Chemical Vapor Deposition of Long VerticallyAligned Carbon Nanotubes, in March Meeting of American Physical So-ciety, Montreal, Canada, 22–26 March 2004sunpublishedd.

48P. A. Tesner and I. S. Rafalkes, Dokl. Akad. Nauk SSSR87, 821 s1952d.49P. A. Tesner, R. Ey, I. S. Rafalkes, and E. F. Arefieva, Carbon8, 453

s1970d.50R. T. K. Baker, M. A. Barber, R. J. Waite, P. S. Harris, and F. S. Feates, J.

Catal. 26, 51 s1972d.51R. T. Yang and J. P. Chen, J. Catal.115, 52 s1989d.52J. R. Nielsen and D. L. Trimm, J. Catal.48, 155 s1977d.53A. Sacco, P. Thacker, T. N. Chang, and A. T. S. Chiang, J. Catal.85, 224

s1984d.54A. Kock, P. K. Debokx, E. Boellaard, W. Klop, and J. W. Geus, J. Catal.

96, 468 s1985d.55I. Alstrup, J. Catal.109, 241 s1988d.56J. W. Snoeck, G. F. Froment, and M. Fowles, J. Catal.169, 240 s1997d.57P. M. Ajayan, NaturesLondond 427, 402 s2004d.58S. Helveg, C. Lopez-Cartes, J. Sehested, P. L. Hansen, B. S. Clausen, J. R.

Rostrup-Nielsen, F. Abild-Pedersen, and J. K. Norskov, NaturesLondond427, 426 s2004d.

59M. Chhowallaet al., J. Appl. Phys.90, 5308s2001d.60S. S. Fan, M. G. Chapline, N. R. Franklin, T. W. Tombler, A. M. Cassell,

and H. J. Dai, Science283, 512 s1999d.61C. Bower, O. Zhou, W. Zhu, D. J. Werder, and S. H. Jin, Appl. Phys. Lett.

77, 2767s2000d.62J. I. Sohn, S. Lee, Y. H. Song, S. Y. Choi, K. I. Cho, and K. S. Nam, Appl.

Phys. Lett. 78, 901 s2001d.63A. V. Melechko, V. I. Merkulov, D. H. Lowndes, M. A. Guillorn, and M.

L. Simpson, Chem. Phys. Lett.356, 527 s2002d.64D. C. Li, L. M. Dai, S. M. Huang, A. W. H. Mau, and Z. L. Wang, Chem.

Phys. Lett. 316, 349 s2000d.



65J. Li, C. Papadopoulos, J. M. Xu, and M. Moskovits, Appl. Phys. Lett.75,367 s1999d.

66F. Jansen,Plasma Enhanced Chemical Vapor Depositon, in AVS Mono-graph Series, edited by H. G. TompkinssAmerican Vacuum Society, NewYork, 1998d.

67S. Honda, M. Katayama, K. Y. Lee, T. Ikuno, S. Ohkura, K. Oura, H.Furuta, and T. Hirao, Jpn. J. Appl. Phys., Part 242, L441 s2003d.

68K. Y. Lee et al., Jpn. J. Appl. Phys., Part 242, L804 s2003d.69B. O. Boskovic, V. Stolojan, R. U. A. Khan, S. Haq, and S. R. P. Silva,

Nat. Mater. 1, 165 s2002d.70S. Hofmann, C. Ducati, J. Robertson, and B. Kleinsorge, Appl. Phys. Lett.

83, 135 s2003d.71K. Hamad-Schifferli, J. J. Schwartz, A. T. Santos, S. G. Zhang, and J. M.

Jacobson, NaturesLondond 415, 152 s2002d.72R. Hergt, W. Andra, C. G. d’Ambly, I. Hilger, W. A. Kaiser, U. Richter,

and H. G. Schmidt, IEEE Trans. Magn.34, 3745s1998d.73K. B. K. Teo et al., Nano Lett. 4, 921 s2004d.74N. Mutsukura, K. Kobayashi, and Y. Machi, J. Appl. Phys.68, 2657

s1990d.75D. R. Cote, S. V. Nguyen, A. K. Stamper, D. S. Armbrust, D. Tobben, R.

A. Conti, and G. Y. Lee, IBM J. Res. Dev.43, 5 s1999d.76M. Funer, C. Wild, and P. Koidl, Appl. Phys. Lett.72, 1149s1998d.77Y. Y. Wei, G. Eres, V. I. Merkulov, and D. H. Lowndes, Appl. Phys. Lett.

78, 1394s2001d.78V. I. Merkulov, M. A. Guillorn, D. H. Lowndes, M. L. Simpson, and E.

Voelkl, Appl. Phys. Lett.79, 1178s2001d.79K. B. K. Teo et al., Appl. Phys. Lett.79, 1534s2001d.80K. B. K. Teo et al., J. Vac. Sci. Technol. B20, 116 s2002d.81K. B. K. Teo et al., Nanotechnology14, 204 s2003d.82M. Tanemura, K. Iwata, K. Takahashi, Y. Fujimoto, F. Okuyama, H. Sugie,

and V. Filip, J. Appl. Phys.90, 1529s2001d.83J. H. Yim, S. Choi, S. Lee, and K. H. Koh, J. Vac. Sci. Technol. B22,

1308 s2004d.84M. A. Guillorn, T. E. McKnight, A. Melechko, V. I. Merkulov, P. F. Britt,

D. W. Austin, D. H. Lowndes, and M. L. Simpson, J. Appl. Phys.91,3824 s2002d.

85B. A. Cruden, A. M. Cassell, Q. Ye, and M. Meyyappan, J. Appl. Phys.94, 4070s2003d.

86Y. Chen, L. P. Guo, D. J. Johnson, and R. H. Prince, J. Cryst. Growth193,342 s1998d.

87R. Haubner and B. Lux, Diamond Relat. Mater.2, 1277s1993d.88D. Hash, D. Bose, T. R. Govindan, and M. Meyyappan, J. Appl. Phys.93,

6284 s2003d.89D. B. Hash and M. Meyyappan, J. Appl. Phys.93, 750 s2003d.90J. Han, W. S. Yang, J. B. Yoo, and C. Y. Park, J. Appl. Phys.88, 7363

s2000d.91Q. Yang, C. Xiao, W. Chen, A. K. Singh, T. Asai, and A. Hirose, Diamond

Relat. Mater.12, 1482s2003d.92K. M. Lee, H. J. Han, S. H. Choi, K. H. Park, S. Oh, S. Lee, and K. H.

Koh, J. Vac. Sci. Technol. B21, 623 s2003d.93G. W. Ho, A. T. S. Wee, J. Lin, and W. C. Tjiu, Thin Solid Films388, 73

s2001d.94T. Hirata, N. Satake, G. H. Jeong, T. Kato, R. Hatakeyama, K. Motomiya,

and K. Tohji, Appl. Phys. Lett.83, 1119s2003d.95T. Ikuno et al., Surf. Interface Anal.35, 15 s2003d.96S. S. Kim, H. Y. Chang, C. S. Chang, and N. S. Yoon, Appl. Phys. Lett.

77, 492 s2000d.97P. E. Nolan, D. C. Lynch, and A. H. Cutler, J. Phys. Chem. B102, 4165

s1998d.98J. B. O. Caughman, L. R. Baylor, M. A. Guillorn, V. I. Merkulov, D. H.

Lowndes, and L. F. Allard, Appl. Phys. Lett.83, 1207s2003d.99O. M. Kuttel, O. Groening, C. Emmenegger, and L. Schlapbach, Appl.

Phys. Lett. 73, 2113s1998d.100D. Ugarte, NaturesLondond 359, 707 s1992d.101L. C. Qin, D. Zhou, A. R. Krauss, and D. M. Gruen, Appl. Phys. Lett.72,

3437 s1998d.102C. Bower, W. Zhu, S. H. Jin, and O. Zhou, Appl. Phys. Lett.77, 830

s2000d.103M. Yoshimura, S. Jo, and K. Ueda, Jpn. J. Appl. Phys., Part 142, 4841

s2003d.104Y. S. Woo, D. Y. Jeon, I. T. Han, N. S. Lee, J. E. Jung, and J. M. Kim,

Diamond Relat. Mater.11, 59 s2002d.105C. L. Tsai, C. F. Chen, and L. K. Wu, Appl. Phys. Lett.81, 721 s2002d.106H. Cui, O. Zhou, and B. R. Stoner, J. Appl. Phys.88, 6072s2000d.

107S. H. Tsai, C. W. Chao, C. L. Lee, and H. C. Shih, Appl. Phys. Lett.74,3462 s1999d.

108Q. Zhang, S. F. Yoon, J. Ahn, B. Gan, Rusli, and M. B. Yu, J. Phys.Chem. Solids61, 1179s2000d.

109Y. C. Choi, Y. M. Shin, S. C. Lim, D. J. Bae, Y. H. Lee, B. S. Lee, and D.C. Chung, J. Appl. Phys.88, 4898s2000d.

110Y. C. Choi, D. J. Bae, Y. H. Lee, B. S. Lee, I. T. Han, W. B. Choi, N. S.Lee, and J. M. Kim, Synth. Met.108, 159 s2000d.

111H. Sato, H. Tagegawa, and Y. Saito, J. Vac. Sci. Technol. B21, 2564s2003d.

112M. Okai, T. Muneyoshi, T. Yaguchi, and S. Sasaki, Appl. Phys. Lett.77,3468 s2000d.

113C. H. Lin, H. L. Chang, C. M. Hsu, A. Y. Lo, and C. T. Kuo, DiamondRelat. Mater.12, 1851s2003d.

114D. J. Willins and R. L. F. Boyd, J. Phys. D6, 1447s1973d.115A. Huczko, H. Lange, M. Sioda, Y. Q. Zhu, W. K. Hsu, and D. R. M.

Walton, J. Phys. Chem. B106, 1534s2002d.116Y. Y. Xia, Y. G. Mu, Y. C. Ma, S. Y. Li, H. D. Zhang, C. Y. Tan, and L.

M. Mei, Nucl. Instrum. Methods Phys. Res. B155, 395 s1999d.117M. W. Li, Z. Hu, X. Z. Wang, Q. Wu, and Y. Chen, J. Mater. Sci. Lett.

22, 1223s2003d.118A. Marafee, C. J. Liu, G. H. Xu, R. Mallinson, and L. Lobban, Ind. Eng.

Chem. Res.36, 632 s1997d.119V. I. Merkulov, A. V. Melechko, M. A. Guillorn, D. H. Lowndes, and M.

L. Simpson, Appl. Phys. Lett.80, 476 s2002d.120J. Ma, C. Park, N. M. Rodriguez, and R. T. K. Baker, J. Phys. Chem. B

105, 11 994s2001d.121V. I. Merkulov, D. K. Hensley, A. V. Melechko, M. A. Guillorn, D. H.

Lowndes, and M. L. Simpson, J. Phys. Chem. B106, 10 570s2002d.122V. Merkulov, A. V. Melechko, M. A. Guillorn, D. H. Lowndes, and M. L.

Simpson, Chem. Phys. Lett.361, 492 s2002d.123M. Laudon, N. N. Carlson, and M. P. Masqueller, Appl. Phys. Lett.78,

201 s2001d.124V. I. Merkulov, A. V. Melechko, M. A. Guillorn, M. L. Simpson, D. H.

Lowndes, J. H. Whealton, and R. J. Raridon, Appl. Phys. Lett.80, 4816s2002d.

125A. V. Melechko, T. E. McKnight, D. K. Hensley, M. A. Guillorn, A. Y.Borisevich, V. I. Merkulov, D. H. Lowndes, and M. L. Simpson,Nanotechnology14, 1029s2003d.

126K. M. Ryu, M. Y. Kang, Y. D. Kim, and H. T. Jeon, Jpn. J. Appl. Phys.,Part 1 42, 3578s2003d.

127M. Nihei, A. Kawabata, and Y. Awano, Jpn. J. Appl. Phys., Part 242,L721 s2003d.

128P. M. Campbell, E. S. Snow, and J. P. Novak, Appl. Phys. Lett.81, 4586s2002d.

129C. J. Lee, S. C. Lyu, Y. R. Cho, J. H. Lee, and K. I. Cho, Chem. Phys.Lett. 341, 245 s2001d.

130L. Delzeit, B. Chen, A. Cassell, R. Stevens, C. Nguyen, and M. Meyyap-pan, Chem. Phys. Lett.348, 368 s2001d.

131H. M. Christen, A. A. Puretzky, H. Cui, P. H. Belay, P. H. Fleming, D. B.Geohegan, and D. H. Lowndes, Nano Lett.4, 1939s2004d.

132S. Y. Zaginaichenko, D. V. Schur, and Z. A. Matysina, Carbon41, 1349s2003d.

133M. S. Kim, N. M. Rodriguez, and R. T. K. Baker, J. Catal.131, 60s1991d.

134P. D. Kichambare, D. Qian, E. C. Dickey, and C. A. Grimes, Carbon40,1903 s2002d.

135Y. Nishiyama and Y. Tamai, J. Catal.33, 98 s1974d.136K. Kamadaet al., Appl. Surf. Sci. 212, 383 s2003d.137I. Alstrup, M. T. Tavares, C. A. Bernardo, O. Sorensen, and J. R. Rostrup-

Nielsen, Mater. Corros.49, 367 s1998d.138A. Chambers, N. M. Rodriguez, and R. T. K. Baker, J. Mater. Res.11,

430 s1996d.139D. Pradhan, M. Sharon, M. Kumar, and Y. Ando, J. Nanosci.

Nanotechnol.3, 215 s2003d.140W. Z. Li, J. G. Wen, and Z. F. Ren, Appl. Phys. Lett.79, 1879s2001d.141C. A. Bernardo, I. Alstrup, and J. R. Rostrup-Nielsen, J. Catal.96, 517

s1985d.142L. B. Avdeeva, O. V. Goncharova, D. I. Kochubey, V. I. Zaikovskii, L. M.

Plyasova, B. N. Novgorodov, and S. K. Shaikhutdinov, Appl. Catal., A141, 117 s1996d.

143B. Ganet al., Diamond Relat. Mater.9, 897 s2000d.144H. T. Ng, B. Chen, J. E. Koehne, A. M. Cassell, J. Li, J. Han, and M.

Meyyappan, J. Phys. Chem. B107, 8484s2003d.



145A. M. Cassell, Q. Ye, B. A. Cruden, J. Li, P. C. Sarrazin, H. T. Ng, J. Han,and M. Meyyappan, Nanotechnology15, 9 s2004d.

146D. H. Kim and H. R. Lee, J. Korean Phys. Soc.44, L208 s2004d.147A. G. Naumovets and Z. Y. Zhang, Surf. Sci.500, 414 s2002d.148H. Cui, X. Yang, M. L. Simpson, D. H. Lowndes, and M. Varela, Appl.

Phys. Lett. 84, 4077s2004d.149T. Ichihashi, J. Fujita, M. Ishida, and Y. Ochiai, Phys. Rev. Lett.92,

215702s2004d.150M. S. Dresselhaus,Encyclopedia of Materials: Science and Technology

sElsevier Science, New York, 2001d.151R. Bacon, J. Appl. Phys.31, 283 s1960d.152Y. Chen, D. T. Shaw, and L. P. Guo, Appl. Phys. Lett.76, 2469s2000d.153M. Endoet al., Carbon 41, 1941s2003d.154N. A. Kiselev et al., Carbon 42, 149 s2004d.155M. H. Kuang, Z. L. Wang, X. D. Bai, J. D. Guo, and E. G. Wang, Appl.

Phys. Lett. 76, 1255s2000d.156L. Dong, J. Jiao, C. Pan, and D. W. Tuggle, Appl. Phys. A: Mater. Sci.

Process.78, 9 s2004d.157S. B. Lee, K. B. K. Teo, M. Chhowalla, D. G. Hasko, G. A. J. Amara-

tunga, W. I. Milne, and H. Ahmed, Microelectron. Eng.61–62, 475s2002d.

158C. Schonenberger, A. Bachtold, C. Strunk, J. P. Salvetat, and L. Forro,Appl. Phys. A: Mater. Sci. Process.A69, 283 s1999d.

159A. Thesset al., Science273, 483 s1996d.160L. Zhang, D. Austin, V. I. Merkulov, A. V. Melechko, K. L. Klein, M. A.

Guillorn, D. H. Lowndes, and M. L. Simpson, Appl. Phys. Lett.84, 3972s2004d.

161R. H. Fowler and L. Nordheim, Proc. R. Soc. London, Ser. A119, 173s1928d.

162T. Utsumi, IEEE Trans. Electron Devices38, 2276s1991d.163F. S. Baker, J. Williams, and A. R. Osborn, NaturesLondond 239, 96

s1972d.164C. Lea, J. Phys. D6, 1105s1973d.165K. A. Dean and B. R. Chalamala, Appl. Phys. Lett.75, 3017s1999d.166L. R. Baylor sprivate communicationd.167W. A. Deheer, A. Chatelain, and D. Ugarte, Science270, 1179s1995d.168V. I. Merkulov, D. H. Lowndes, and L. R. Baylor, J. Appl. Phys.89,

1933 s2001d.169W. D. J. Callister,Materials Science and Engineering an Introduction,

6th ed.sWiley, New York, 2003d.170B. McEnaney,Encyclopedia of Materials: Science and TechnologysAm-

sterdam, New York, 2001d, pp. 967–975.171B. G. Demczyk, Y. M. Wang, J. Cumings, M. Hetman, W. Han, A. Zettl,

and R. O. Ritchie, Mater. Sci. Eng., A334, 173 s2002d.172R. S. Ruoff, D. Qian, and W. K. Liu, C. R. Phys.4, 993 s2003d.173M. F. Yu, O. Lourie, M. J. Dyer, K. Moloni, T. F. Kelly, and R. S. Ruoff,

Science287, 637 s2000d.174R. P. Gao, Z. L. Wang, Z. G. Bai, W. A. de Heer, L. M. Dai, and M. Gao,

Phys. Rev. Lett.85, 622 s2000d.175E. T. Thostenson, Z. F. Ren, and T. W. Chou, Compos. Sci. Technol.61,

1899 s2001d.176M. C. Weisenberger, E. A. Grulke, D. Jacques, T. Rantell, and R. An-

drews, J. Nanosci. Nanotechnol.3, 535 s2003d.177R. L. McCreery, inElectroanalytical Chemistry, edited by A. J. Bard

sMarcel Dekker, New York, 1991d.178R. J. Chen, Y. G. Zhan, D. W. Wang, and H. J. Dai, J. Am. Chem. Soc.

123, 3838s2001d.179J. L. Bahr and J. M. Tour, J. Mater. Chem.12, 1952s2002d.180E. T. Mickelson, C. B. Huffman, A. G. Rinzler, R. E. Smalley, R. H.

Hauge, and J. L. Margrave, Chem. Phys. Lett.296, 188 s1998d.181G. T. Hermanson,Boiconjugate TechniquressAcademic, San Diego,

1996d.182Y. H. Lin, F. Lu, Y. Tu, and Z. F. Ren, Nano Lett.4, 191 s2004d.183C. V. Nguyen, L. Delzeit, A. M. Cassell, J. Li, J. Han, and M. Meyyap-

pan, Nano Lett.2, 1079s2002d.184T. E. McKnight et al., Nanotechnology14, 551 s2003d.185K. Kinoshita, Carbon: Electrochemical and Physiological Properties

sWiley, New York 1988d.186J. Li, A. Cassell, L. Delzeit, J. Han, and M. Meyyappan, J. Phys. Chem.

B 106, 9299s2002d.187M. A. Murphy, G. D. Wilcox, R. H. Dahm, and F. Marken, Electrochem.

Commun. 5, 51 s2003d.188T. E. McKnight et al., J. Phys. Chem. B107, 10722s2003d.189Y. Tu, Y. H. Lin, and Z. F. Ren, Nano Lett.3, 107 s2003d.

190J. Li, H. T. Ng, A. Cassell, W. Fan, H. Chen, Q. Ye, J. Koehne, J. Han,and M. Meyyappan, Nano Lett.3, 597 s2003d.

191T. E. McKnight, A. V. Melechko, D. W. Austin, T. Sims, M. A. Guillorn,and M. L. Simpson, J. Phys. Chem. B108, 7115s2004d.

192J. H. Chen, Z. P. Huang, D. Z. Wang, S. X. Yang, W. Z. Li, J. G. Wen,and Z. F. Ren, Synth. Met.125, 289 s2001d.

193J. Koehne, J. Li, A. M. Cassell, H. Chen, Q. Ye, H. T. Ng, J. Han, and M.Meyyappan, J. Mater. Chem.14, 676 s2004d.

194L. Zhang, A. V. Melechko, V. I. Merkulov, M. A. Guillorn, M. L. Sim-pson, D. H. Lowndes, and M. J. Doktycz, Appl. Phys. Lett.81, 135s2002d.

195M. A. Guillorn, A. V. Melechko, V. I. Merkulov, E. D. Ellis, M. L.Simpson, L. R. Baylor, and G. J. Bordonaro, J. Vac. Sci. Technol. B19,2598 s2001d.

196T. Ikuno, M. Katayama, K. Y. Lee, T. Kuzuoka, J. G. Lee, S. Honda, H.Mori, and K. Oura, Jpn. J. Appl. Phys., Part 243, L987 s2004d.

197S. M. Huang, L. M. Dai, and A. W. H. Mau, J. Phys. Chem. B103, 4223s1999d.

198T. Ito, L. Sun, and R. M. Crooks, Electrochem. Solid-State Lett.6, C4s2003d.

199B. J. Hinds, N. Chopra, T. Rantell, R. Andrews, V. Gavalas, and L. G.Bachas, Science303, 62 s2004d.

200D. Temple, Mater. Sci. Eng., R.24, 185 s1999d.201X. P. Xu and G. R. Brandes, Appl. Phys. Lett.74, 2549s1999d.202G. Pirio, P. Legagneux, D. Pribat, K. B. K. Teo, M. Chhowalla, G. A. J.

Amaratunga, and W. I. Milne, Nanotechnology13, 1 s2002d.203X. Xu and G. R. Brandes,Carbon nanotube-based vacuum microelec-

tronic gated cathode, in Materials Issues in Vacuum MicroelectronicsSymposium, 13–16 April 1998sMaterials Research Society, San Fran-cisco, CA, 1998d.

204K. B. K. Teo et al., J. Vac. Sci. Technol. B21, 693 s2003d.205M. A. Guillorn et al., J. Vac. Sci. Technol. B22, 35 s2004d.206L. R. Baylor et al., J. Vac. Sci. Technol. B22, 3021s2004d.207OxfordInstruments, Oxford instruments.208H. Sugie, M. Tanemura, V. Filip, K. Iwata, K. Takahashi, and F.

Okuyama, Appl. Phys. Lett.78, 2578s2001d.209G. Z. Yueet al., Appl. Phys. Lett.81, 355 s2002d.210T. Matsumoto and H. Mimura, Appl. Phys. Lett.82, 1637s2003d.211R. A. Wallingford and A. G. Ewing, Anal. Chem.59, 1762s1987d.212A. J. Bard, F. R. F. Fan, J. Kwak, and O. Lev, Anal. Chem.61, 132

s1989d.213R. M. Wightmanet al., Proc. Natl. Acad. Sci. U.S.A.88, 10754s1991d.214G. L. Prasanna and T. Panda, Bioprocess Eng.16, 261 s1997d.215A. Radford, S. Pope, A. Sazci, M. J. Fraser, and J. H. Parish, Mol. Gen.

Genet. 184, 567 s1981d.216N. J. Taylor and C. M. Fauquet, DNA Cell Biol.21, 963 s2002d.217T. E. McKnight, A. V. Melechko, D. K. Hensley, D. G. J. Mann, G. D.

Griffin, and M. L. Simpson, Nano Lett.4, 1213s2004d.218G. Binnig, C. F. Quate, and C. Gerber, Phys. Rev. Lett.56, 930 s1986d.219L. Delzeit, C. Nguyen, R. Stevens, J. Han, and M. Meyyappan,

Nanotechnology13, 280 s2002d.220C. V. Nguyen, K. J. Chao, R. M. D. Stevens, L. Delzeit, A. Cassell, J.

Han, and M. Meyyappan, Nanotechnology12, 363 s2001d.221J. H. Hafner, C. L. Cheung, and C. M. Lieber, J. Am. Chem. Soc.121,

9750 s1999d.222P. Poncharal, Z. L. Wang, D. Ugarte, and W. A. de Heer, Science283,

1513 s1999d.223D. M. Eigler and E. K. Schweizer, NaturesLondond 344, 524 s1990d.224P. Kim and C. M. Lieber, Science286, 2148s1999d.225S. Akita et al., Appl. Phys. Lett.79, 1691s2001d.226B. L. Fletcheret al., Nano Lett. 4, 1809s2004d.227P. M. Ajayan and O. Z. Zhou, Top. Appl. Phys.800, 391 s2001d.228A. B. Dalton et al., NaturesLondond 423, 703 s2003d.229A. Zuttel, P. Sudan, P. Mauron, T. Kiyobayashi, C. Emmenegger, and L.

Schlapbach, Int. J. Hydrogen Energy27, 203 s2002d.230Y. S. Nechaev and O. K. Alexeeva, Int. J. Hydrogen Energy28, 1433

s2003d.231D. Lupu, A. R. Biris, I. Misan, A. Jianu, G. Holzhuter, and E. Burkel, Int.

J. Hydrogen Energy29, 97 s2004d.232D. Lupu, A. R. Biris, I. Misan, N. Lupsa, A. S. Biris, D. A. Buzatu, and

M. Kleeve, Part. Sci. Technol.20, 225 s2002d.233S. Orimo, A. Zuttel, L. Schlapbach, G. Majer, T. Fukunaga, and H. Fujii,

J. Alloys Compd.356, 716 s2003d.234H. G. Schimmel, G. J. Kearley, M. G. Nijkamp, C. T. Visserl, K. P. de



Jong, and F. M. Mulder, Chem.-Eur. J.9, 4764s2003d.235C. Cantalini, L. Valentini, I. Armentano, L. Lozzi, J. M. Kenny, and S.

Santucci, Sens. Actuators B95, 195 s2003d.236P. Serp, M. Corrias, and P. Kalck, Appl. Catal., A253, 337 s2003d.237L. Pesant, G. Wine, J. Matta, R. Vieira, N. Keller, C. Pham-Huu, and G.

Ledoux, Large Scale Synthesis of Carbon Nanofibers for Use as CatalystSupport in Liquid-Phase Reactions, in 12th Brazillian Congress on Ca-talysis, Augra dos Reis, Rio de Janeiro, 2003sunpublishedd.

238M. Croci, I. Arfaoui, T. Stockli, A. Chatelain, and J.-M. Bonard, Micro-electron. J.35, 329 s2004d.

239G. H. Luo, Z. F. Li, F. Wei, L. Xiang, X. Y. Deng, and Y. Jin, Physica B323, 314 s2002d.

240J. Li, C. Papadopoulos, and J. Xu, NaturesLondond 402, 253 s1999d.241W. Hofmeister, W. P. Kang, Y. M. Wong, and J. L. Davidson, J. Vac. Sci.

Technol. B 22, 1286s2004d.242J. H. Hafner, M. J. Bronikowski, B. R. Azamian, P. Nikolaev, A. G.

Rinzler, D. T. Colbert, K. A. Smith, and R. E. Smalley, Chem. Phys. Lett.296, 195 s1998d.

243H. T. Ng, M. L. Foo, A. P. Fang, J. Li, G. Q. Xu, S. Jaenicke, L. Chan,and S. F. Y. Li, Langmuir18, 1 s2002d.

244H. Terroneset al., Chem. Phys. Lett.343, 241 s2001d.245M. L. Mabudafhasi, R. Bodkin, C. P. Nicolaides, X. Y. Liu, M. J. Wit-

comb, and N. J. Coville, Carbon40, 2737s2002d.

246A. W. Hull, Phys. Rev.10, 661 s1917d.247Binary Alloy Phase Diagrams, 2nd ed., edited by T. B. MassalskisASM

International, Metals Park, OH, 1990d, Vol. 1 and 2.248A. W. Hull and W. P. Davey, Phys. Rev.14, 54 s1919d.249The Properties of Natural and Synthetic Diamond, edited by J. E. Field

sAcademic, London, 1992d.250B. T. Kelly, The Physics of GraphitesApplied Science Publishers, Lon-

don, 1981d.251D. P. Kim, Y. R. Suhng, and M. M. Labes, Carbon30, 729 s1992d.252E. W. Wong, P. E. Sheehan, and C. M. Lieber, Science277, 1971s1997d.253M. S. Dresselhaus, G. Dresselhaus, and P. C. Eklund,Science of

Fullerenes and Carbon NanotubessAcademic, San Diego, 1996d.254A. Krishnan, E. Dujardin, T. W. Ebbesen, P. N. Yianilos, and M. M. J.

Treacy, Phys. Rev. B58, 14013s1998d.255M. M. J. Treacy, T. W. Ebbesen, and J. M. Gibson, NaturesLondond 381,

678 s1996d.256R. Bacon, J. Appl. Phys.31, 323 s1960d.257J. D. H. Hughes, J. Phys. D20, 276 s1987d.258T. Hashishin, H. Kohara, H. Iwanaga, and S. Takeuchi, J. Ceram. Soc.

Jpn. 110, 772 s2002d.259R. L. Jacobsen, T. M. Tritt, J. R. Guth, A. C. Ehrlich, and D. J. Gillespie,

Carbon 33, 1217s1995d.



applied physics reviews vertically aligned carbon...

Documents