Carbon Nanotube Synthesis in Supercritical Toluene

Article abstract of DOI:10.1021/ja031522s

Multiwall carbon nanotubes (MWNTs) were synthesized in supercritical toluene at 600 °C and ～12.4 MPa using ferrocene, Fe, or FePt nanocrystals as growth catalysts. Toluene serves as both the carbon source for nanotube formation and the solvent. In contrast to vapor-phase synthetic routes, the supercritical solvent provides high precursor concentration and a homogeneous reaction environment with dispersed growth catalyst particles. Both carbon filaments and MWNTs are produced by this approach, and a growth mechanism is proposed to explain the factors that determine the nanotube versus filament morphology. The plasmon energies of the π and π + σ valence electrons were measured using electron energy-loss spectroscopy (EELS) of individual carbon fibers and MWNTs as a characterization tool to complement the imaging data obtained from electron microscopy.

Full text of DOI:10.1021/ja031522s

Published on Web 03/26/2004
Carbon Nanotube Synthesis in Supercritical Toluene
Doh C. Lee, Frederic V. Mikulec, and Brian A. Korgel*
Contribution from the Department of Chemical Engineering, Texas Materials Institute,
Center for Nano- and Molecular Science and Technology, The UniVersity of Texas at Austin,
Austin, Texas 78712-1062
Received December 4, 2003; E-mail: korgel@mail.che.utexas.edu
Abstract: Multiwall carbon nanotubes (MWNTs) were synthesized in supercritical toluene at 600 °C and
∼12.4 MPa using ferrocene, Fe, or FePt nanocrystals as growth catalysts. Toluene serves as both the
carbon source for nanotube formation and the solvent. In contrast to vapor-phase synthetic routes, the
supercritical solvent provides high precursor concentration and a homogeneous reaction environment with
dispersed growth catalyst particles. Both carbon filaments and MWNTs are produced by this approach,
and a growth mechanism is proposed to explain the factors that determine the nanotube versus filament
morphology. The plasmon energies of the π and π + σ valence electrons were measured using electron
energy-loss spectroscopy (EELS) of individual carbon fibers and MWNTs as a characterization tool to
complement the imaging data obtained from electron microscopy.
nanotubes had been observed as early as 1955;11 however, not
until recently have researchers embarked on a concerted effort
Introduction
Carbon nanotubes exhibit a variety of unique properties,
which make them suitable for many potential uses, including
structural, electronic, and optical applications.1 The commercial
use of carbon nanotubes in real-world applications requires a
cost-effective method to fabricate large amounts of nanotubes
with the desired material properties. With only a couple of
exceptions,2-4 carbon nanotube formation has been explored
exclusively in the gas phase using synthetic methods such as
arc-discharge,5 laser vaporization,6 and chemical vapor deposi-
tion (CVD),7 because they provide access to the high synthetic
temperatures required for nanotube formation. Under the
extremely high temperature conditions of arc-discharge and laser
vaporization (2000-4000 °C), carbon atoms can be vaporized
from a solid carbon target and condensed into high-quality
single-wall nanotubes. The relatively low yield, high operational
costs, and low throughput of these methods, however, limit their
potential use as a manufacturing approach for large-scale
nanotube production.
to produce high-quality nanotubes by metal-catalyzed CVD.12,13
Andrews et al.,14 for example, have produced high-quality
MWNTs by CVD at 625-775 °C and atmospheric pressure
using ferrocene to catalyze the decomposition of aromatic
compounds to nanotubes. In comparison to vapor-phase syn-
thetic methods, supercritical fluid-phase approaches have the
potential for much higher throughput due to orders-of-magnitude
higher precursor concentration and catalyst particle dispersibility,
which would enable a continuous homogeneous synthetic
process. Since carbon nanotube synthesis requires high tem-
peratures (>600 °C), the solvents must be pressurized to
achieve this temperature range. Solvents at high pressures and
temperaturessabove their critical pointsshave been used to
synthesize a variety of materials15-17 and more recently have
played an important role in the synthesis of a variety of
nanostructures.18-24
(8) Dai, H.; Kong, J.; Zhou, C.; Franklin, N.; Tombler, T.; Cassell, A.; Fan,
S.; Chapline, M. J. Phys. Chem. B 1999, 103, 11246-11255.
(9) Delzeit, L.; Nguyen, C. V.; Chen, B.; Stevens, R.; Cassell, A.; Han, J.;
Meyyappan, M. J. Phys. Chem. B 2002, 106, 5629-5635.
(10) Hofmann, S.; Ducati, C.; Robertson, J.; Kleinsorge, B. Appl. Phys. Lett.
2003, 83, 135-137.
Metal catalysts have enabled high-quality carbon nanotube
synthesis at much lower temperatures (500-1200 °C) using
hydrocarbon species as reactants in conventional CVD re-
actors.8-10 Metal-catalyzed decomposition of CO to carbon
(11) Hofer, L. J. E.; Sterling, E.; McCarthy, J. T. J. Phys. Chem. 1955, 59,
1153-1155.
(12) Nikolaev, P.; Bronikowski, M. J.; Bradley, R. K.; Rohmund, F.; Colbert,
D. T.; Smith, K. A.; Smalley, R. E. Chem. Phys. Lett. 1999, 313, 91-97.
(13) Bronikowski, M. J.; Willis, P. A.; Colbert, D. T.; Smith, K. A.; Smalley,
R. E. J. Vac. Sci. Technol., A 2001, 19, 1800-1805.
(1) Dresselhaus, M. S., Dresselhaus, G., Avouris, P., Eds. Carbon Nanotubes:
Synthesis, Structure, Properties, and Applications; Springer: New York,
2002.
(2) Gogotsi, Y.; Libera, J. A.; Yoshimura, M. J. Mater. Res. 2000, 15, 2591-
(14) Andrews, R.; Jacques, D.; Qian, D. L.; Rantell, T. Acc. Chem. Res. 2002,
35, 1008-1017.
2594.
(3) Libera, J.; Gogotsi, Y. Carbon 2001, 39, 1307-1318.
(4) Shao, M.; Li, Q.; Wu, J.; Xie, B.; Zhang, S.; Qian, Y. Carbon 2000, 40,
2961-2973.
(15) Gogotsi, Y. G.; Yoshimura, M. Nature 1994, 367, 628-630.
(16) Hubert, H.; Garvie, L. A. J.; Devouard, B.; Buseck, P. R.; Petuskey, W.
T.; McMillan, P. F. Chem. Mater. 1998, 10, 1530-1537.
(17) Barns, R. L.; Freeland, P. E.; Kolb, E. D.; Laudise, R. A.; Patel, J. R. J.
Cryst. Growth 1978, 43, 676-686.
(5) Ebbesen, T. W.; Ajayan, P. M. Nature 1992, 358, 220-222.
(6) Thess, A.; Lee, R.; Nikolaev, P.; Dai, H.; Petit, P.; Robert, J.; Xu, C. H.;
Lee, Y. H.; Kim, S. G.; Rinzler, A. G.; Colbert, D. T.; Scuseria, G. E.;
Tomanek, D.; Fischer, J. E.; Smalley, R. E. Science 1996, 273, 483-487.
(7) Ren, Z. F.; Huang, Z. P.; Xu, J. W.; Wang, J. H.; Bush, P.; Siegal, M. P.;
Provencio, P. N. Science 1998, 282, 1105-1107.
(18) Watkins, J. J.; Blackburn, J. M.; McCarthy, T. J. Chem. Mater. 1999, 11,
213-215.
(19) Ziegler, K. J.; Doty, R. C.; Johnston, K. P.; Korgel, B. A. J. Am. Chem.
Soc. 2001, 123, 7797-7803.
9
10.1021/ja031522s CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 4951-4957
4951

A R T I C L E S
Lee et al.
solution was loaded into the 10 mL cell to give ∼12.4 MPa at the
reaction temperature. The sealed reactor was removed from the
glovebox and placed into a heating block preheated to 670 °C. The
reactor temperature was determined using a thermocouple placed inside
the heating block next to the reactor. The reaction cell reached the
reaction temperature within 5 min. Fifteen minutes after placing the
reactor in the heating block, it was removed from the heating block
and cooled rapidly in water bath. The reactor reached room temperature
after 5 min in the water bath.
We have recently developed a supercritical fluid-liquid-
solid (SFLS) approach to synthesize various semiconductor
nanowires, including Si,20,24 Ge,22,23 and GaAs.25 Sterically
stabilized metal nanocrystals are input as seed particles that
direct wire growth at temperatures that exceed the metal/
semiconductor eutectic temperature, which is approximately 360
°C for Au:Ge and Au:Si. The temperature must be sufficiently
high to degrade the molecular precursor and to induce nanowire
formation. In this paper, we present the synthesis of multiwall
carbon nanotubes in supercritical toluene, catalyzed by ferro-
cene, or nanocrystals of Fe or FePt. In this process, toluene
serves as both the carbon source for nanotube growth and the
reaction solvent. Under the synthetic conditions producing the
highest quality nanotubes, toluene degrades catalytically at the
metal particle surfaces, with only minimum homogeneous
toluene degradation. MWNTs ranging from 10 to 50 nm in outer
diameter with wall thicknesses ranging from 5 to 40 nm were
produced along with carbon nanofilaments. High-resolution
transmission electron microscopy (HRTEM) reveals that al-
though the MWNT growth mechanism exhibits similarities to
SFLS growth, the processes are distinct: the morphology of
the nanotubes appears to depend on the growth catalyst, with
larger particles producing solid nanofilaments and smaller
particles yielding MWNTs. Furthermore, nanotube growth
appears to occur on the catalyst particle surface. The nanotubes
were characterized by HRTEM, HRSEM, and electron energy
loss spectroscopy (EELS). EELS of individual MWNTs and
filaments provided a particularly powerful tool for distinguishing
MWNTs from carbon nanofilaments.
The reaction product consisted of a black solution of nanotubes in
toluene. It was collected from the reactor in air. Hexane was used to
extract the remaining product that had adhered to the reactor walls.
The dispersed product removed from the reactor was centrifuged at
8000 rpm for 10 min to isolate a black precipitate containing the
nanotubes. The supernatant was discarded, and the precipitate was
redispersed in hexane and centrifuged again. This washing step was
repeated once more to ensure that all of the organic molecular
byproducts had been separated from the nanotube product.
Characterization. The product was characterized by high-resolution
transmission electron microscopy and scanning electron microscopy
(HRTEM and HRSEM) and electron energy loss spectroscopy (EELS).
For TEM and EELS, the final dry product was redispersed in hexane
and dropped onto a lacey carbon-coated TEM grid (Electron Microscopy
Sciences). The nanotubes are sufficiently long to span to holey regions
of the lacey carbon film, allowing TEM imaging and collection of EELS
without a carbon substrate background. This was of primary importance
for the EELS measurements, where the background carbon signal
obscures the spectroscopic data. A JEOL 2010F operating at 200 kV
accelerating voltage was used for TEM imaging and for EELS. The
JEOL 2010 F was equipped with a Gatan parallel-EELS spectrometer.
Electron energy loss spectra were acquired in STEM mode with the
field emission gun operating at 200 keV, with the EELS aperture size
set at 2 mm and a camera length of 10 cm, which translates into 5
mrad of collection semiangle. The electron beam size was set at 1 nm
for performing the EELS measurements. HRSEM images were obtained
on a field emission LEO 1530 SEM operated at 4 kV. For SEM
observation, the product was dispersed by brief sonication in hexane
and then drop-cast on a piece of silicon wafer (2 cm × 2 cm), which
had been washed ultrasonically in ethanol (5 min)-acetone (5 min)-
ethanol (5 min) beforehand.
Experimental Section
Synthesis. Anhydrous toluene, ferrocene, and hexane were purchased
from Aldrich Chemical Co. (Milwakee, MI), stored under nitrogen,
and used as received. Fe nanocrystals were synthesized by thermal
decomposition of iron pentacarbonyl (Fe(CO)5) (Aldrich) in octyl ether
(Fluka) in the presence of oleic acid (Fluka) as a cappling ligand at
100 °C, following published procedures.26 FePt nanocrystals were
prepared by thermal decomposition of platinum acetylacetonate
(Pt(CH3COCCHCH3)2) (Aldrich) and iron pentacarbonyl (Fe(CO)5) in
the presence of oleic acid and oleylamine (Aldrich).27
Results
Carbon nanotubes were synthesized in a high-pressure 10 mL
stainless steel reactor. Solutions of toluene and catalyst were loaded
into the reactor in a nitrogen-filled glovebox. In the case of ferrocene,
0.1, 2.5, and 5 mM toluene solutions were prepared and tested, and
the nanocrystal concentration was set to 2.5 mM. The volume of the
solutions was adjusted so that the pressure determined from the toluene
phase diagram28 should be ∼12.4 MPa. Note that extreme care must
be taken not to exceed the pressure rating on the reactor, as the
pressure-density isotherm rises sharply just above the critical point
and small deviations in volume can lead to large increases in pressure.
For example, for a reaction carried out at 600 °C, 2 mL of toluene
Ferrocene-Catalyzed Nanotube Synthesis. Parts a and b of
Figure 1 show HRSEM images of the crude carbonaceous
product obtained by degrading toluene at 600 °C and ∼12.4
MPa in the presence of 2.5 mM ferrocene. The product appears
as an entangled mesh of fibrous material. TEM images reveal
that the wires are a mixture of MWNTs and solid carbon
nanofilaments. In the absence of ferrocene, there is no ap-
preciable degradation of toluene at 600 °C at ∼12.4 MPa.
HRTEM images of MWNTs and nanofilamentsssuch as those
in Figure 1c-fsshow the filament core to consist of randomly
stacked graphene sheets (Figure 1e), whereas the nanotubes
exhibit coaxially stacked graphene sheets (Figure 1c) with a
layer spacing of 0.344 nm, in good agreement with previously
reported intertubule spacing.29 The yield of carbon nanotubes
relative to the total carbonaceous material produced in the
reaction is approximately 2% based on TEM observations.
Catalyst particles were always observed at the tips of the
MWNTs and nanofilaments. The fiber morphologyswhether
the carbon structures end up as tubes or solid filamentssappears
(20) Holmes, J. D.; Johnston, K. P.; Doty, R. C.; Korgel, B. A. Science 2000,
287, 1471-1473.
(21) Holmes, J. D.; Ziegler, K. J.; Doty, R. C.; Pell, L. E.; Johnston, K. P.;
Korgel, B. A. J. Am. Chem. Soc. 2001, 123, 3743-3748.
(22) Hanrath, T.; Korgel, B. A. J. Am. Chem. Soc. 2002, 124, 1424-1429.
(23) Hanrath, T.; Korgel, B. A. AdV. Mater. 2003, 15, 437-440.
(24) Lu, X.; Hanrath, T.; Johnston, K. P.; Korgel, B. A. Nano Lett. 2003, 3,
93-99.
(25) Davidson, F. M.; Schricker, A. D.; Wiacek, R.; Korgel, B. A. AdV. Mater.
2004, 16, in press.
(26) Hyeon, T.; Lee, S. S.; Park, J.; Chung, Y.; Na, B. H. J. Am. Chem. Soc.
2001, 123, 12798-12801.
(27) Sun, S.; Murray, C. B.; Weller, D.; Folks, L.; Moser, A. Science 2000,
287, 1989-1992.
(28) Yaws, C. L. Handbook of Thermodynamic Diagrams; Gulf Publishing
Co.: Houston, TX, 1996.
(29) Burian, A.; Dore, J. C.; Fischer, H. E.; Sloan, J. Phys. ReV. B: Condens.
Matter Mater. Phys. 1999, 59, 1665-1668.
9
4952 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbon Nanotube Synthesis in Supercritical Toluene
A R T I C L E S
Figure 1. High-resolution scanning electron microscopy (HRSEM) images
(a, b) of the crude reaction product consisting of carbon nanofilaments and
MWNTs obtained by heating 2.5 mM ferrocene in toluene solutions for 15
min at 600 °C and ∼12.4 MPa. HRTEM images of a MWNT (c, d), and a
carbon filament (e, f), isolated from the reaction at 600 °C, ∼12.4 MPa,
using ferrocene as the growth catalyst. The MWNTs exhibit concentrically
stacked graphite sheets (d), whereas the nanofilaments exhibit disordered
stacking along the length of the filament (f).
to depend on the size and shape of the catalyst particle, with
nanotubes being generated from smaller catalytic particles.
After examining the effects of catalyst concentration, hydro-
carbon species, reaction pressure, and temperature on the
reaction product, the temperature appears to be the most
significant parameter in determining the quality of the nanotube
product (see the Supporting Information for a complete listing
of the experimental results). Figure 2 shows TEM images of
product obtained after heating solutions of toluene and ferrocene
(2.5 mM) to temperatures ranging from 400 to 650 °C for 15
min, with a reaction pressure of ∼12.4 MPa. At 400 °C, the
product isolated from the reactor consists only of Fe particles.
Toluene does not degrade at this temperature. At 500 °C, larger
Fe particles were produced that were coated with a thin layer
of amorphous carbonaceous material. The amorphous coating
indicates that 500 °C is below the temperature required for
graphitization. Fibrous structures, which form at 550 °C, are
filamentous but not graphitic. All of these fibers embed Fe
particles ranging from 30 to 50 nm in diameter at their tips,
indicating that the particles promote fiber formation in the
reactor. Interestingly, smaller particles isolated from the reaction
mixture did not appear to be associated with the carbonaceous
material. Reactions at 600 °C produced MWNTs and graphitic
carbon nanofilaments. The optimum reaction temperature for
MWNTs was found to be 600 °C. Ferrocene thermally
decomposes to form Fe particles, which catalyze toluene
degradation and promote nanotube and nanofilament formation.
Pure toluene is stable at 550 °C and ∼12.4 MPa and begins to
partially degrade at 600 °C. At 650 °C, toluene degrades rapidly
and reactions carried out at 650 °C did not produce filaments
or nanotubes. At 650 °C, primarily amorphous carbonaceous
material is obtained: the homogeneous solvent pyrolysis rate
overwhelms nanotube and nanofilament growth, leading to
amorphous particulate formation and catalyst poisoning.
EELS Characterization of Nanotubes and Filaments.
Figure 3 shows electron energy loss spectra of an individual
multiwall nanotube compared to spectra obtained from a
nanofilament. In the experiments, a 1 nm diameter electron beam
is scanned radially across the nanotube, while simultaneously
Figure 2. HRTEM images of reaction product obtained at (a) 400, (b)
500, (c) 550, (d) 600, and (e) 650 °C. All reactions were conducted at ∼12.4
MPa for 15 min with 2.5 mM ferrocene in toluene loaded into the cell. At
temperatures lower than 500 °C, only Fe particles were produced. At 550
°C, amorphous carbon fibers with Fe particles embedded at their tips were
produced. At 600 °C, multiwall carbon nanotubes were produced. At 650
°C, severe toluene degradation results in the production of mostly
carbonaceous byproduct without MWNTs. Panel f shows enlarged images
of the nanotubes in d.
collecting EELS data as a function of probe position. These
EELS line scans confirm that the structure in Figure 3a is a
multiwall nanotube and the structure in Figure 3b is a nano-
filament. EELS spectra taken in the low-loss region exhibit two
plasmon peaks corresponding to inelastic electron scattering
from collective oscillations (plasmons) of π (∼6 eV) and π +
σ (∼26 eV) valence electrons. The key differences in the
nanotube and filament spectra appear when the electron beam
is positioned at the outer edge of the structures. For the MWNT,
the π + σ plasmon shifts to significantly lower energys
approximately 18 eVswhen the probe is positioned at the
nanotube surface. An additional lower lying sideband at
approximately 14 eV also appears in the spectra. In the
nanofilament, the π + σ plasmon energy does not depend on
the probe position. Energy-dispersive X-ray spectroscopy (EDS)
did not reveal the presence of any impurities, such as iron debris
from catalyst, and furthermore the peak shift is not due to a
change in carbon-carbon bonding. The peak shifting that occurs
in the nanotube EELS spectra is the result of the anisotropic
symmetry of the π electrons delocalized parallel to the graphene
sheets in the nanotube wall. Due to the plasmon dispersion
relation, when the beam is positioned orthogonally incident to
a (002) graphite plane, as is the case in the center of the
nanotube, the π + σ plasmon energy is 26.5 eV. When the beam
is positioned at the nanotube edge, the beam is directed parallel
to the graphitic sheets, resulting in a lower energy plasmon peak
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4953

A R T I C L E S
Lee et al.
Figure 3. Dark-field TEM images and EELS line scans across (a) a multiwall nanotube and (b) a carbon nanofilament. The numbered labels on the spectra
correspond to the labeled axial positions in the corresponding TEM image.
Figure 4. MWNTs formed from toluene catalyzed using (a) ferrocene, (b) Fe nanocrystals (9.2 nm diameter), and (c) FePt nanocrystals (4 nm diameter).
All reactions are carried out at 600 °C, ∼12.4 MPa, for 15 min. Scale bars are 10 nm.
position. Results from angle-dependent EELS measurements
from graphite are well-known, and our measurements match
the expected values for the plasmon energies at both the center
(orthogonal to the (002) plane) and surface (parallel to the (002)
plane) of the nanotubes.30 The plasmon energy does not shift
as a function of probe position in the nanofilament since
the graphitic carbon is randomly oriented. The EELS spectra
clearly demonstrate that MWNTs were obtained from these
reactions.
Fe and FePt Nanocrystal-Seeded MWNT Formation.
MWNTs were also synthesized by direct injection of sterically
stabilized Fe and FePt nanocrystals. The Fe and FePt particles
were injected with diameters less than 10 nm, which is
significantly less than the 20-50 nm diameter Fe particles
formed by in situ ferrocene degradation. The MWNTs generated
using the preformed catalyst particles were generally of smaller
diameter, reflecting the more stable smaller diameter seed
particles; however, the nanotube yield was lower. Figure 4
shows high-resolution TEM bright-field images of carbon
nanotubes grown using the different catalysts. The smaller
MWNTs were typically straighter than the larger tubes, most
likely as a result of greater surface stress due to the increased
curvature of the graphite sheets in the nanotube shell.31,32 The
lower yield could be due to the adsorbed organic passivation
layer on the nanocrystals, which initially could potentially inhibit
surface-directed nanotube growth, or smaller diameter nanotube
formation may simply be less efficient under these low
temperature growth conditions. Although smaller nanotubes
(31) Kim, N. S.; Lee, Y. T.; Park, J.; Han, J. B.; Choi, Y. S.; Choi, S. Y.;
Choo, J. B.; Lee, G. H. J. Phys. Chem. B 2003, 107, 9249-9255.
(32) Dai, H.; Rinzler, A. G.; Nikolaev, P.; Thess, A.; Colbert, D. T.; Smalley,
R. E. Chem. Phys. Lett. 1996, 260, 471-475.
(30) Zeppenfeld, K. Z. Phys. 1968, 211, 391-399.
9
4954 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbon Nanotube Synthesis in Supercritical Toluene
A R T I C L E S
should form more efficiently at higher reaction temperatures,
we could not explore this possibility since the homogeneous
solvent degradation at 650 °C and above eliminated nanotube
formation, thus preventing the exploration of a wider range of
synthetic temperatures in this system.
Discussion
Role of Catalyst in MWNT Formation. Fe catalyst particles
were observed at the ends of the nanotubes and the filaments
grown using ferrocene injection. Transition metal particle
catalyzed carbon filament and nanotube growth has been
extensively studied since the 1950s, and a few critical steps in
the carbon filament formation process are well-known:33,34 (1)
catalytic hydrocarbon decomposition occurs at the particle
surface, which results in (2) carbidization of the catalyst,
subsequently saturating the particle with carbon and giving rise
to (3) surface-directed graphitization and nanotube or nano-
filament formation. Figure 5 shows a schematic representation
of the growth process. Although the mechanistic steps have been
identified, many important details about nanotube formation are
still not well understood. For example, carbon will alloy with
the Fe particle but may be associated with either the Fe
nanoparticle surface or the interior of the particle. And nanotube
growth itself has been proposed to occur by either “root
growth”,32 in which the nanotube base interfaces directly with
the nanoparticle, or a “folded growth mode”,35 in which the
carbon shell that forms the nanotube wraps around the nano-
particle leading to the curved graphitic layers that extrude from
the particle surface. In fact, it appears from extensive data in
the literature that both of these mechanisms can occur depending
on the synthetic conditions.32,35 Information about the growth
mechanism in the supercritical solvent can be deduced by
examining the particle tips of the nanotubes and filaments.
Figure 6 shows two representative TEM images of the catalyst
particles at the end of a MWNT and a carbon filament, along
with the fast Fourier transform (FFT) of the image. We observed
two primary differences between particles associated with
nanotubes and those associated with filaments. First, the particles
associated with the nanotubes were smaller: nanotubes had
particles less than 30 nm in diameter, whereas larger particles
were attached to filaments. Second, the seed particle composi-
tions of nanotubes and filaments were different: the particles
at the tips of the nanotubes were composed of pure fcc Fe, while
the particles at the end of the nanofilaments were composed of
an Fe-C alloy. Both seed particles are coated with a carbon
shell. However, the filament seed particles exhibit a thin Fe-C
shell that surrounds the particle core, which appears to be
composed of a different Fe-C composition or structure. The
Fe seed particle embedded at the tip of the MWNT interfaces
directly with the graphitic carbon shell.
Figure 5. Schematic representation of the nanotube growth process. (a)
Hydrocarbon decomposes at the catalytic surface. (b) Decomposed carbon
diffuses into the nanoparticle and eventually saturates the catalyst. Surface-
directed graphitization can take place mainly in two ways: (c-d) “root
growth mechanism”,32 where the nanotube edge is bound to the surface of
the catalyst and the nanotube wall is thrust up with a continuous supply of
carbon source; (e-g) “folded growth mode”,35 where graphitic nanosheets
evolve as carbon precipitates on the catalyst surface. Nanotubes grown in
the present study exhibit particles embedded at the tip, indicating that the
folded growth mode dominates in our case.
nanocrystals to seed growth, a process termed supercritical
fluid-liquid-solid (SFLS) growth. The carbon filaments pro-
duced in supercritical toluene form by this mechanism, with
the exception that the seed particles may be a solid Fe-C alloy,
instead of a liquid, since the growth temperatures are well below
the Fe-C eutectic temperature. However, nanometer-size
particles exhibit a melting point depression, which means that
a liquid-phase seed particle cannot be entirely ruled out.
Regardless, the only requirement from VLS-type growth from
nanometer particles is limited solubility of the wire material in
the seed. The fact that the graphitic layers in the carbon
nanofilament cores form disordered stacks oriented orthogonal
to the growth direction further supports the idea that the
filaments form at the surface of a metal seed.
On the basis of TEM observations, it appears that the
nanotubes produced in supercritical toluene form by the folded-
growth mechanism, whereas the nanofilaments grow by a VLS-
type (vapor-liquid-solid) mechanism. In our laboratory, we
have exploited the VLS mechanism to produce Si,20,24 Ge,22,23
and GaAs25 nanowires in supercritical solvents using gold
(33) Evans, E. L.; Thomas, J. M.; Thrower, P. A.; Walker, P. L. Carbon 1973,
11, 441-445.
(34) Bower, C.; Zhou, O.; Zhu, W.; Werder, D. J.; Jin, S. Appl. Phys. Lett.
2000, 77, 2767-2769.
(35) Louchev, O. A.; Hester, J. R. J. Appl. Phys. 2003, 94, 2002-2010.
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4955

A R T I C L E S
Lee et al.
Figure 6. HRTEM images of catalyst particles at the tip of a MWNT (a,
b), and a carbon filament (c, d), produced at 600 °C, ∼12.4 MPa, using
ferrocene as the catalyst. The particle at the nanotube tip is Fe, while the
particle in the nanofilament has a core-shell structure with a crystal structure
different from pure Fe. Insets in b and d show FFTs of the HRTEM images.
FFTs of the Fe-C alloy seeds, as in d, do not match any Fe or Fe-C alloy
crystal structure available from the literature; however, the presence of both
Fe and C are confirmed by EDS.
Figure 7. HRTEM images of (a) an embryonic MWNT, (b) a fully grown
MWNT, and (c) a MWNT with two Fe particles trapped inside the nanotube.
Scale bars are 10 nm.
In order for nanotubes to form, the graphitic layers must be
curved by the seed particle. The higher surface curvature of
the smaller seed particles presumably adds the needed driving
force to induce nanotube formation. Below some critical particle
diameter, the relatively strong van der Waals attraction between
the condensing graphite sheets and the nanoparticle surface sta-
bilizes the formation of hollow nanotubes.35 Surface-templated
graphite sheet formation is clearly evident in the TEM image
of the Fe particle at the MWNT tip in Figure 6. Figure 7 shows
additional TEM images of MWNTs that reveal the tube
morphology near the seed particle. In the folded-growth
mechanism, a graphene cap forms at the particle surface, and
the carbon layers extend as additional carbon diffuses through
the graphitic shell to reach the Fe-nanotube interface. We did
not find evidence of the root-growth mechanism for nanotube
formation.
The carbon content in the seed particles of different size most
likely plays a significant role in determining the nanotube and
nanofilament morphology as well. The larger seed particles
exhibit higher carbon content, which could favor filament
formation due to enhanced wetting between the graphite layers
and the seed particle that makes nanotube formation less
energetically favorable over filament formation.36
pletely quenches nanotube and nanofilament formation. In the
available temperature window <∼625 °C, the thermal energy
cannot overcome the higher energetic barrier (relative to
MWNTs) to surface curvature needed to produce SWNTs. This
curvature energy relates to the critical radius of the nanotube
nucleus that forms on the particle surface. At low temperatures,
the critical radius of the nucleus is simply too large to permit
SWNT formation. In metal-particle-catalyzed CVD, higher
reaction temperatures are reached by using methane and CO as
precursors, which are relatively stable to homogeneous decom-
position, yet degrade catalytically at the particle surface.12,13,37,38
Other potential supercritical solvents available for nanotube
formation that could help limit homogeneous precursor degrada-
tion are water and CO2. These two solvents provide different
engineering and safety challenges that must be addressed;
however, there is no fundamental reason to believe that the
supercritical approach to nanotube formation could not be
successful for SWNTs as well as MWNTs.
Summary
Carbon MWNTs and nanofilaments can be produced in
supercritical toluene using either the molecular additive, ferro-
cene, or prefabricated sterically stabilized nanocrystals of Fe
or FePt, to catalyze their growth. The carbon nanostructure
morphology depends on the reaction temperature and the seed
particle size. The temperature must be sufficiently high for (1)
significant hydrocarbon decomposition, which occurs hetero-
Although we certainly tried, we did not produce SWNT under
any of the conditions we explored. This does not appear to be
a problem related to the seed particle diameter in the reactor
but rather a limitation imposed by the relatively low tempera-
tures available in supercritical toluene. Homogeneous solvent
degradation at temperatures much higher than ∼625 °C com-
(36) Kuznetsov, V. L.; Usoltseva, A. N.; Chuvilin, A. L.; Obraztsova, E. D.;
Bonard, J. M. Phys. ReV. B: Condens. Matter Mater. Phys. 2001, 64,
235401.
(37) Kong, J.; Cassell, A. M.; Dai, H. Chem. Phys. Lett. 1998, 292, 567-574.
(38) Kong, J.; Soh, H. T.; Cassell, A. M.; Quate, C. F.; Dai, H. Nature 1998,
395, 878-881.
9
4956 J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004

Carbon Nanotube Synthesis in Supercritical Toluene
A R T I C L E S
geneously at the particle surfaces at temperatures exceeding at
least 500 °C for toluene, and (2) carbon graphitization at the
metal seed particle surface, which requires temperatures greater
than ∼550 °C. At temperatures much exceeding ∼625 °C,
homogeneous solvent decomposition spoils nanotube growth.
Smaller seed particles (<∼30 nm diameter) promote MWNT
formation, while larger particles produce nanofilaments with
disorderd graphitic cores.
The supercritical fluid approach to carbon nanotube synthesis
has the potential for much higher throughput relative to the
heterogeneous gas-phase approaches. However, the quality of
the nanotubes currently does not match that of nanotubes
produced in the vapor phase. There are many scientific and
engineering considerations in process optimization, including
reactor design, as well as optimization of catalyst and precursor
reactivity. In comparison to the vapor-phase process, which has
been studied and optimized by many researchers over the past
10 years, little effort has been spent developing wet nanotube
synthesis. One currently open question in the supercritical fluid
approach relates to single-wall nanotube growth. Considering
that SWNTs are known to require more energy to form than
MWNTs, the use of thermally stable alternative supercritical
solvents, e.g. supercritical CO2, could improve the quality of
nanotubes, eventually leading to SWNT formation. The use of
these kinds of supercritical solvents could also improve MWNT
yield by limiting homogeneous solvent decomposition and
catalyst poisoning at higher temperature.
Acknowledgment. We thank the National Science Foundation
(NSF), the Welch Foundation, and the Texas Higher Education
Coordinating Board (ATP program) for financial support of this
work. We are grateful to M. Jose-Yacaman, T. Hanrath, and A.
E. Saunders for fruitful discussions, and we thank J. P. Zhou
for assistance with the EELS and TEM measurements.
Supporting Information Available: Description of the reac-
tion products from different reaction conditions (Table S1). This
material is available free of charge via the Internet at
http://pubs.acs.org.
JA031522S
9
J. AM. CHEM. SOC. VOL. 126, NO. 15, 2004 4957