Large Scale CVD Synthesis of Single-Walled Carbon Nanotubes

Article abstract of DOI:10.1021/jp990957s

The synthesis of bulk amounts of high quality single-walled carbon nanotubes (SWNTs) is accomplished by optimizing the chemical compositions and textural properties of the catalyst material used in the chemical vapor deposition (CVD) of methane. A series of catalysts are derived by systematically varying the catalytic metal compounds and support materials. The optimized catalysts consist of Fe/Mo bimetallic species supported on a novel silica-alumina multicomponent material. The high SWNT yielding catalyst exhibits high surface-area and large mesopore volume at elevated temperatures. Gram quantities of SWNT materials have been synthesized in ～0.5 h using the optimized catalyst material. The nanotube material consists of individual and bundled SWNTs that are free of defects and amorphous carbon coating. This work represents a step forward toward obtaining kilogram scale perfect SWNT materials via simple CVD routes.

Full text of DOI:10.1021/jp990957s

6484
J. Phys. Chem. B 1999, 103, 6484-6492
Large Scale CVD Synthesis of Single-Walled Carbon Nanotubes
Alan M. Cassell,†,‡ Jeffrey A. Raymakers, Jing Kong, and Hongjie Dai*
†
†
,†
Department of Chemistry, Stanford UniVersity, Stanford, California 94305, and National Aeronautics and
Space Administration, Ames Research Center, Moffett Field, California 94035
ReceiVed: March 18, 1999; In Final Form: June 1, 1999
The synthesis of bulk amounts of high quality single-walled carbon nanotubes (SWNTs) is accomplished by
optimizing the chemical compositions and textural properties of the catalyst material used in the chemical
vapor deposition (CVD) of methane. A series of catalysts are derived by systematically varying the catalytic
metal compounds and support materials. The optimized catalysts consist of Fe/Mo bimetallic species supported
on a novel silica-alumina multicomponent material. The high SWNT yielding catalyst exhibits high surface-
area and large mesopore volume at elevated temperatures. Gram quantities of SWNT materials have been
synthesized in ∼0.5 h using the optimized catalyst material. The nanotube material consists of individual and
bundled SWNTs that are free of defects and amorphous carbon coating. This work represents a step forward
toward obtaining kilogram scale perfect SWNT materials via simple CVD routes.
1
. Introduction
including helicity and diameter. Moreover, such controlled
synthesis must be scaled up in bulk or on surfaces.
Since Iijima’s discovery of carbon nanotubes in arc-discharge
Chemical vapor deposition of hydrocarbons over metal
catalysts has been a classical method to produce carbon
materials. Various forms of carbon fibers, filaments, and
multiwalled nanotubes have been synthesized by CVD in the
1
soot materials, nanotubes have generated significant research
activities in chemistry, condensed matter physics, and materials
science. Nanotubes are intriguing one dimensional systems
ideally suited for fundamental studies of atomically well-defined
2
0-24
past.
However, previous CVD approaches mostly obtain
2
-4
materials and have high potential for technological applications.
defective tubular materials. Recently, we developed for the first
time a CVD synthesis approach to high quality SWNTs by using
Nanotubes exhibit the highest Young’s modulus and tensile
strength among all materials. A single-walled nanotube can be
either metallic, semiconducting, or semimetallic depending on
the helicity and diameter. On the practical side, promising
applications of nanotubes have been demonstrated including
2
5,26
methane as the carbon feedstock.
The choice of methane
over other hydrocarbons was essential and based on its high
stability at elevated temperatures (∼1000 °C), which virtually
eliminates hydrocarbon self-pyrolysis that may lead to the
generation of amorphous carbon and cause catalyst poisoning.
We further developed a synthetic strategy using methane CVD
on catalytically patterned substrates to obtain “nanotube chips”
containing isolated single-walled nanotubes at controlled loca-
2
5-7
8
advanced scanning probes, chemical force sensors, electron
field emission sources,9 and nano-electronic components.
-13
14-16
High quality nanotube materials are desired for both funda-
mental and technological applications. High quality refers to
the absence of structural and chemical defects over a significant
length scale (e.g., 1-10 microns) along the tube axes. So far,
2
6
tions. Integrated nanotube circuits were built in combination
with microfabrication techniques, which reliably led to a large
number of ohmically contacted nanotube devices with control-
lable length. Both metallic and semiconducting individual
nanotubes have been observed in recent transport measurements.
Transport studies demonstrated ohmic contacts to ∼5 µm long
individual SWNTs, giving two-terminal resistances as low as
arc-discharge1 and laser-ablation are the principal methods
for obtaining high quality nanotube materials. However, there
are several key issues concerning both methods. The first is
that both methods involve evaporating carbon atoms from solid
carbon sources at g3000 °C. This sets a limitation to the
quantity of nanotubes that can be synthesized, and it has been
unclear how to scale-up nanotube production to the kilogram
level using evaporation approaches. The second issue relates
to the fact that evaporation methods grow nanotubes in highly
tangled forms mixed with unwanted forms of carbon or metal
species. The nanotubes are difficult to purify, manipulate, and
assemble for building nanotube device architectures. New
synthetic strategies should open new possibilities in nanotube
science and technology if exquisite control over their growth
can be gained. The ultimate goal in nanotube synthesis should
be gaining control over geometrical aspects of nanotubes, such
as location and orientation, and the atomic structure of nanotubes
7,18
19
1
6
2
0 kΩ, even at 4.2 K.
The current work focuses on the synthesis of bulk quantities
of high quality SWNTs using the methane CVD approach.
Noteworthy is that more recently, SWNTs have also been
2
7
28
synthesized in bulk by CVD of benzene, ethylene, or a
2
9
methane/hydrogen mixture. In our initial bulk synthesis, the
2
5
yield of SWNTs was not high. However, we learned several
important aspects of high quality SWNT growth using CVD.
First, the type of support material for the catalyst was found to
play a key role in the type of SWNTs produced. The Al2O3
supported Fe2O3 catalyst afforded a large fraction of individual
SWNTs when subjected to methane CVD at 900-1000 °C,
whereas silica supported Fe2O3 catalyst led to ropelike bundles
*
To whom correspondence should be addressed.
Department of Chemistry.
NASA Ames Research Center.
2
5
†
of SWNTs. This strongly suggested that a systematic inves-
tigation of various support materials may lead to the optimum
‡
1
0.1021/jp990957s CCC: $18.00 © 1999 American Chemical Society
Published on Web 07/20/1999

Single-Walled Carbon Nanotubes
J. Phys. Chem. B, Vol. 103, No. 31, 1999 6485
catalyst for high yield SWNT production. Second, we found
that base-growth is the dominant growth mechanism for SWNTs
in the CVD process. That is, a significant amount of the ∼1
nm scale metal catalyst particles (responsible for nucleating
SWNTs) remained anchored on the catalyst support surface
during the growth of nanotubes.25 This result hinted that
optimization of metal-support interactions should be pursued
for higher yield and purity SWNTs.
Based on our previous results and understanding of the
methane CVD process, we have carried out a step-by-step
optimization of bulk SWNT growth. We have focused on
investigating a series of catalyst materials prepared by system-
atically varying the metal composition and the support. The
highest yield of SWNTs was achieved with an Fe/Mo bimetallic
catalyst supported on a new type of Al2O3-SiO2 hybrid material.
We found that, in the CVD growth of nanotubes, the active
metal catalytic species, metal-support interactions, resistance
of the catalyst to sintering and the chemical, and physical nature
of the support dictate the yield, purity, and quality of the
nanotubes.
Figure 1. Schematic experimental setup for methane CVD synthesis
of SWNTs. Typical growth temperature is 900 °C. The methane gas
inlet pressure as measured by the gauge is 1.27 atm. Volume flow rate
through the 1 in. quartz tube is ∼6000 cm /min.
3
with stirring, followed by the addition of 0.020 g of (NH4)6-
Mo7O24‚4H2O and 0.167 g of Fe2(SO4)3‚5H2O. The mixture was
stirred for 15 min and heated to 90 °C for 2 h while flowing a
stream of N2 over the mixture to remove the water. The material
was then ground with an agate mortar to obtain a fine orange
powder.
2
. Experimental Section
Materials and Reagents. The main support material used
in this work was the Oxide C alumina obtained from Degussa
Inc. The material consists of crystalline transition alumina (γ-δ
SWNT Synthesis. All the catalyst materials were calcined
in air at 500 °C for 1 h, cooled to room temperature, and ground
prior to CVD growth. Carbon nanotubes were synthesized by
using the experimental setup shown in Figure 1. Typically, 0.1
g of the catalyst was placed into a quartz boat that was inserted
into the center of a 1 in. diameter quartz tube mounted in an
electric tube furnace. Ar was flown over the catalyst as it was
heated from room temperature to 900 °C. The reaction began
2
phase) nanoparticles and has a surface area of ∼100 m /g.
Tetraelthyl orthosilicate (TEOS), (NH4)6Mo7O24‚4H2O, Fe2(SO4)3‚
5
H2O, Fe(NO3)3‚9H2O, and Ru(acac)3 were purchased from
Aldrich Chemical. Hydrofluoric acid and deionized water were
supplied by J. T. Baker, Inc. Ethanol was purchased from Gold
Shield Chemical Co. Pyridine and 1,2-dichloroethane were
purchased from E. M. Science. Methane (99.99%) and Ar were
supplied by Praxair, Inc.
3
as Ar was replaced by CH4 (flow rate 6000 cm /min) for the
Preparation of a Al2O3-SiO2 Support Material. An
Al2O3-SiO2 multicomponent (hybrid) support was synthesized
by forming silica sol in the presence of alumina nanoparticles
and gelling the subsequent mixture. First, 1.0 g of alumina
nanoparticles was suspended in 50 mL of ethanol with vigorous
stirring. TEOS (2.2 mL) was introduced into the alumina
suspension via syringe and allowed to stir for 10 min. Water (5
mL) was then added to hydrolyze the TEOS and form a silica
sol mixed with alumina nanoparticles. Concentrated HF (0.47
mL, 28.9 M) solution was added 5 min later to catalyze the
gelation of the silica sol. After 1 h, the suspension was heated
to 90 °C, and a stream of N2 was flown over the samples to
evaporate the water and ethanol. The dried fluffy white powder
was then carefully ground with an agate mortar, which affords
the final Al2O3-SiO2 hybrid support material. The molar ratios
of the compounds involved in making the hybrid support are
TEOS:Al2O3:HF:H2O:EtOH ) 1:1:1.4:28:87.
desired reaction time (2-45 min). The flow was then switched
to Ar and the furnace was cooled to room temperature. When
producing large quantities of SWNTs, we typically used 5.0 g
of catalyst and the CVD reaction time was ∼30 min.
Textural Property Characterization. Surface area and pore
analysis of catalyst materials were performed by standard N2
adsorption and desorption experiments using a Coulter SA-3100
surface area and pore analyzer. All catalysts were heated to 900
°C in Ar for 10 min prior to measurements in order to elucidate
the textural properties of the catalysts at the beginning stage of
the CVD growth. Surface areas were determined using the
30
Brunauer-Emmett-Teller (BET) method. Pore volumes were
obtained at a normalized partial pressure (P /P ) of 0.9814, and
s
o
pore size distributions were calculated using the method of
3
1
Barrett, Joyner, and Halenda (BJH). All results are reported
in accordance with the IUPAC recommendations for the
interpretation of physisorption data. For instance, micropores
are pores with diameters less than 2 nm, mesopores have
diameters between 2 and 50 nm, and macropores have diameters
greater than 50 nm.
Catalyst Preparation. Catalysts were made by impregnating
support materials in salt solutions. For single metallic Fe based
catalysts, we used Fe(NO3)3‚9H2O/methanol solution or Fe2-
(SO4)3‚5H2O aqueous solution. Molar ratio of Fe:Al2O3 ) 1:16
Infrared Spectroscopy. DRFTIRS spectra were recorded
using a Bruker Equinox 55 infrared spectrometer and a Spectra-
Tech diffuse reflectance high-temperature chamber equipped
with KBr windows. Support material or the metal loaded catalyst
was ground to a fine powder with an agate mortar and placed
on top of a powdered KBr layer (which served as the
background) in the sample furnace. Samples for pyridine
adsorption/desorption analysis were dehydrated in-situ by heat-
ing in a vacuum to 500 °C. After the samples were cooled to
room temperature, pyridine vapor was introduced into the
chamber by flowing Ar over the pyridine for 2 min, after which
was used in preparing both catalyst systems. For bimetallic (Fe/
Mo and Fe/Ru) catalysts, we used aqueous Fe2(SO4)3 solutions
and the molar ratios between reagents were kept at Fe: X(Mo
or Ru): Al2O3 )1:0.17:16. When the Al2O3-SiO2 hybrid
support material was used in making the catalyst, the molar
composition of the catalyst was Fe: Mo: SiO2: Al2O3 ) 1:0.17:
1
6:16. The following outlines the detailed procedure for
preparing an Fe/Mo/Al2O3-SiO2 catalyst. Other catalysts were
prepared using the same procedure except for different com-
pounds. Support material (1 g) was suspended in 70 mL of water

6486 J. Phys. Chem. B, Vol. 103, No. 31, 1999
Cassell et al.
the sample chamber was evacuated. Pyridine desorption studies
were performed by collecting spectra as the samples were heated
-
3
in steps to 500 °C under vacuum (2.0 × 10 Torr).
X-ray Photoemission Spectroscopy. X-ray photoelectron
spectroscopy was performed using a Surface Science Instru-
ments S-Probe, employing monochromated Al KR (1.486 keV)
as the source.
Characterization of Nanotube Materials. For all CVD
grown samples, we first employed transmission electron mi-
croscopy (TEM) and scanning electron microscopy (SEM) to
qualitatively estimate the yield of nanotubes. TEM was per-
formed on a Phillips CM20 TEM operating at 200 kV. SEM
was performed at 25 kV using a Hitachi H-800 field-emission
instrument. Sample preparation for TEM involves sonicating
the synthesized material in 1,2-dichloroethane for ∼10 min and
placing a few drops of the resulting suspension onto holey
carbon grids (Ted Pella, Inc.). To prepare samples for SEM,
the synthesized materials were sonicated in 1,2-dichloroethane
for 10 min, and a few drops of the suspension were placed onto
a silicon substrate and allowed to evaporate. Samples were also
prepared for SEM examination by placing the catalyst material
onto a silicon substrate and performing nanotube growth using
CH4 CVD. For quantitative measurements, we carried out weight
gain studies of the materials after CVD. Typically 0.1 g of
catalyst material was first heated to 900 °C in flowing Ar, cooled
to room temperature and weighed. The percent weight loss was
calculated using the weight before and after heating in Ar. After
CVD growth, the weight gain was calculated by using the known
weight loss of the catalyst heated to 900 °C in Ar. We note
that weight gain is a valid estimate of the yield of SWNTs only
when the SWNTs are free of amorphous carbon coating and
graphitic particles are absent in the CVD material, which is the
case for the optimized SWNT materials in the current work.
3
. Results and Disussion
Systematic Catalyst Optimization for SWNT Growth. In
the initial methane CVD method, we used an Fe(NO3)3/Al2O3
catalyst prepared by impregnation in methanol and produced
high quality individual SWNTs and a small fraction of bundled
SWNTs.25 The yield of SWNTs, however, is not sufficient for
a bulk growth process where large quantities of SWNTs are
desired. Also, a noticeable amount of graphitic particles exist
in the material as byproducts. A number of steps were taken to
systematically improve the yield of SWNTs.
Figure 2. TEM images recorded with SWNT materials (a) made from
the Fe(NO
3
)
3
/Al
2 3
O
-based catalyst after 10 min of CH
4
CVD and (b)
CVD.
made from Fe
2
(SO
4
)
3
/Al -based catalyst after 10 min of CH
2
O
3
4
Step 1: Varying the Catalytic Metal-Bearing Compounds.
Different Fe-based salts were screened for improving the yield
of SWNTs. We found that the Fe2(SO4)3/Al2O3 prepared in an
aqueous solution was superior to the Fe(NO3)3/Al2O3-based
catalyst. A comparison of the TEM images of the SWNTs
generated from the Fe2(SO4)3/Al2O3 catalyst indicates the
increased efficiency for SWNT growth compared with the
Fe(NO3)3/Al2O3 system (Figure 2). Since SEM and TEM images
are only qualitative in determining nanotube yield, weight gain
results were used in correlation with microscopy data to quantify
yield. Figure 3 shows the weight gain versus reaction time for
a series of catalysts. The Fe(NO3)3/Al2O3 system initially has a
steady growth rate, but levels off near 20 min, giving a 16 wt
%
gain after 45 min. The Fe2(SO4)3/Al2O3 system has a higher
initial growth rate but behaves similarly to the Fe(NO3)3/Al2O3
system giving a 21 wt. % yield after 45 min of CH4 CVD. Upon
closer inspection of the CVD material synthesized by the
Fe(NO3)3/Al2O3 based catalyst using high magnification TEM,
we detect the presence of graphitic particles with encapsulated
metallic crystals inside (∼10 nm, data not shown). Their
Figure 3. Weight gain vs time for a series of catalyst materials.
existence suggests that relatively large metal particles (∼10 nm)
have formed in the CVD process due to weakly anchored metal
species “lifting off” from the support material and sintering

Single-Walled Carbon Nanotubes
J. Phys. Chem. B, Vol. 103, No. 31, 1999 6487
TABLE 1: Textural Properties of Catalysts after Heating to 900 °C in Ar
catalysts heated to 900 °C in Ar
support
Fe(NO
3
)/Al
O
2 3
Fe
2
(SO
105
0.89
4
)
3
/Al
O
2 3
Fe/Mo Al
O
2 3
Fe/Ru Al
O
2 3
Fe/Mo hybrid (1:1)
Fe/Mo hybrid (2.5:1)
2
surface area (m /g)
pore volume (mL/g)
85
0.19
79
0.61
108
0.59
151
0.71
196
0.79
open pore structures allowing efficient diffusion of reactant and
product molecules in and out of the catalyst.
Step 2: Bimetallic Catalysts. Further increase in the yield of
SWNTs was obtained by employing Fe/Mo and Fe/Ru catalysts
after investigating various combinations of bimetallic catalyst
materials. Figures 5 and 6 show SEM/TEM images of CVD
materials synthesized by CH4 CVD for 10 min at 900 °C using
the Fe/Mo/Al2O3 and Fe/Ru/Al2O3 catalyst, respectively. In these
materials, we observe more abundant SWNTs compared to the
material produced by the Fe2(SO4)3/Al2O3 catalyst. Results from
weight gain experiments also show that the addition of Mo or
Ru to the catalyst leads to a substantial increase in weight gain
over the Fe2(SO4)3/Al2O3 single-metal catalyst under identical
reaction conditions (Figure 3), consistent with the qualitative
TEM and SEM results. The Fe/Mo/Al2O3 catalyst led to a weight
gain of 30% after 45 min of CH4 CVD, and the Fe/Ru/Al2O3
catalyst showed a weight gain of 31% after 45 min of CVD.
The Fe/Mo/Al2O3, Fe/Ru/Al2O3, and Fe2(SO4)3/Al2O3 catalysts
have similar textural structures as shown in Table 1, although
we do observe the tendency of reduction in surface area and
pore volume with the addition of Mo or Ru in the catalyst.
Clearly, the increase in SWNT yield by the Mo or Ru addition
cannot be accounted for by the textural properties of the three
catalysts.
Figure 4. N
after heating to 900 °C in Ar. The isotherm curves show a dramatic
increase in mesopore volume (large absorption at P /P
∼ 0.9 in the
upper curve) for the Fe (SO /Al -based catalyst over the Fe(NO
Al -based catalyst.
2
adsorption and desorption isotherms of catalyst materials
s
0
2
)
4 3
O
2 3
3 3
) /
2 3
O
together. On the other hand, the CVD synthesized material using
the Fe2(SO4)3/Al2O3-based catalyst was nearly free of these
unwanted graphitic structures. These results suggest that strong
interactions exist between the metal species and the Al2O3
support for the Fe2(SO4)3 derived catalyst, which contributes
to the increased SWNT yield.
We tentatively attribute the increase in yield of SWNTs by
the Fe/Mo catalyst to the roles played by Mo species in
promoting the aromatization of methane at elevated tempera-
tures. It has been established that supported Mo compounds on
zeolite as well as transition alumina convert methane into
Different textural properties between the Fe2(SO4)3- and
Fe(NO3)3-derived catalysts provide further explanation to the
observed difference in the yield of SWNTs for the two catalyst
materials. Surface analysis has proven to be an excellent tool
to probe the structures of catalyst materials. The pore structures
of a catalyst could strongly influence the diffusion rates of
molecular species and thus play important roles in determining
the reaction kinetics. In Table 1, we show the textural properties
of a series of catalysts after heating to 900 °C in Ar. We observe
a dramatic increase in pore volume (0.89 vs 0.19 mL/g) and a
34
aromatic species. For the Fe/Mo/Al2O3 catalyst in our methane
CVD process, it is plausible that intermediate aromatic hydro-
carbons are generated on Mo sites. Owing to the close proximity
between Mo and Fe catalytic sites where SWNT growth takes
place, the intermediate aromatic species can feed into the SWNT
growth sites with high efficiency, without being limited by
diffusion.
The increase in SWNT yield due to the bimetallic Fe/Ru
catalyst appears to have a different mechanism than that due to
the Fe/Mo catalyst. We have observed an interesting and unique
aspect of the CVD material synthesized by using the Fe/Ru/
Al2O3 catalyst. That is, in the TEM, we observe abundant
2
slight increase in surface area (105 vs 85 m /g) for the Fe2-
(SO4)3/Al2O3-based catalyst over the Fe(NO3)3/Al2O3-based
catalyst (Figure 4). The large increase in pore volumes for the
former could be attributed to two aspects. First, in an aqueous
solution, strong interactions exist between sulfate ions and
surface sites of alumina,3 which allows the metal species to
be uniformly dispersed and anchored on the alumina surface.
This is advantageous because metal species can form active
catalytic sites with high efficiency and density without being
aggregated to form large particles. Second, the decomposition
of Fe2(SO4)3 to Fe2O3 occurs at ∼ 750 °C whereas the
decomposition temperature is ∼ 140 °C for Fe(NO3)3. While a
detailed understanding is required, we speculate that the
decomposition of the highly dispersed and strongly anchored
sulfate species at high temperature leads to the generation of
the observed textural pores. In the CVD material synthesized
by Fe2(SO4)3-based catalyst, the absence of “lifted-off” metal
particles and associated graphitic structures provides direct
evidence for strong metal-support interactions. Thus, the
increase in SWNT yield can be rationalized by better metal
dispersion leading to higher density of catalytic sites and more
3
5
circular tubes (known as crop circles as in laser ablation
materials) and short tubes (0.1-2 µm in length) that are
disconnected from the catalyst materials. These disconnected
tubes appear to be grown in the gas phase, instead of grown
off the catalyst particles or sonicated off the catalyst particles
during TEM sample preparation. This is because for the CVD
materials synthesized by Fe/Mo/Al2O3 catalysts, no appreciable
disconnected tubes and circular tubes have been seen by TEM.
Furthermore, the formation of disconnected tubes in the gas
phase for the Fe/Ru/Al2O3 CVD system was confirmed by
patterning silicon substrates with Fe/Ru/Al2O3 catalyst islands
and characterizing the CVD grown nanotubes on the substrate
by atomic force microscopy (AFM). Indeed, AFM imaging
reveals nanotubes with both ends disconnected from the catalyst
islands as well as circular tubes formed between the catalyst
2,33
3
6
islands. These results suggest that during the methane CVD
process, metallic species are volatilized from the Fe/Ru/Al2O3

6488 J. Phys. Chem. B, Vol. 103, No. 31, 1999
Cassell et al.
Figure 5. TEM micrographs of SWNT materials made from a bimetallic Fe/Mo/Al
O
2 3
catalyst after 10 min CH
4
CVD. (a) Low magnification
SEM, (b) low magnification TEM, and (c) high magnification TEM.
catalyst and catalyze the formation of nanotubes in the gas phase.
Further experimental data and discussion of nanotube growth
mechanism involving the Fe/Ru/Al2O3 catalyst are presented
in a later section of this paper.
presence of double walled nanotubes, but they are essentially
negligible relative to the abundant SWNTs. Importantly, we
observe nearly no graphitic particles due to lifted-off metal
particles or amorphous carbon particles due to pyrolysis and
buildup. Weight gain measurements with the Fe/Mo/hybrid
catalyst for 45 min approaches 45 wt %, which is the highest
among all catalysts studied in the current work. These results
show that the methane CVD SWNT materials are of comparable
quality and purity to those obtained by laser ablation and arc
discharge, which is very significant because the CVD approach
has the potential to be scaled up for the production of large
quantities of perfect nanotube materials. For instance, starting
with 5 g of the Fe/Mo/hybrid catalyst, we synthesized ∼1.5 g
of SWNTs by methane CVD in 30 min.
The increase in SWNT yield with the Fe/Mo/hybrid catalyst
can be at least partly attributed to its more open textural structure
compared to the Fe/Mo/Al2O3 catalyst. As shown in Table 1
and Figure 9, the Fe/Mo/hybrid catalyst exhibits the highest
surface area and largest pore volume among all the catalysts
investigated in the current work. The large pore volume consists
of a large fraction of mesopores (Figure 9) that allow efficient
diffusion of reactant and intermediate hydrocarbon species,
which could be responsible for efficient SWNT growth and high
yield. To understand the textural properties of the Al2O3-SiO2
hybrid support and the catalyst at a microscopic level, we have
carried out IR spectroscopy (see the next subsection) and TEM
studies. Both studies found that the silica phase in the catalyst
does not conform as uniform layers on the alumina surface.
We found by TEM that the silica phase partially covers and
interconnects the alumina particles. Therefore, the silica particles
effectively act as “spacers” that prevent the catalyst material
from sintering and collapsing, which provides a rationale for a
Step 3: Varying the Support Material. We have systematically
investigated various support materials to further optimize the
performance of the catalyst for the high yield synthesis of
SWNTs. So far, we find that the best catalysts for producing
high quality SWNTs using CH4 CVD were made from an Fe/
Mo catalyst supported on the Al2O3-SiO2 hybrid support
described in the Experimental Section. Figure 7 shows two TEM
images of the synthesized nanotube material using the Fe/Mo/
hybrid catalyst. These low magnification images illustrate the
remarkable abundance of nanotubes emanating from the catalyst
particles, extending and forming a massive web. In Figure 8,
we show a representative SEM image and high magnification
TEM images. The SEM image (Figure 8a) clearly reveals the
abundant fiberlike nanotube networks. The material is very
uniform and has a consistent appearance throughout the sample
under the SEM. High magnification TEM images (Figures 8b-
d) reveal that the synthesized material consists of high quality
individual and bundled single-walled nanotubes. At high
magnification, the walls of the nanotubes are clearly resolved
and show no amorphous carbon covering the nanotubes. The
SWNTs are nearly perfect because they contain no sharp
structural variations due to topological defects along their axes,
only buckling and smooth bending of the nanotubes are seen
due to nanotube entanglement. Quite often, ends (Figure 8c) as
well as end-on views of nanotube bundles (Figure 8d) are
observed. By measuring over 100 SWNTs, we find that SWNTs
have diameters dispersed in the range of 0.7-5 nm with a peak
diameter close to 1.7 nm. Occasionally, we also detect the

Single-Walled Carbon Nanotubes
J. Phys. Chem. B, Vol. 103, No. 31, 1999 6489
2 3 4
Figure 6. TEM micrographs of SWNT materials made from a bimetallic Fe/Ru/Al O catalyst after 10 min of CH CVD. (a) Low magnification
SEM, (b) low magnification TEM, (c) intermediate, and (d) high magnification TEM.
highly open pore structure of the support and the catalyst. To
reduce sintering, the silica phase must be thermally stable at
elevated temperatures and is achieved by our sol-gel process
using HF as the gel-forming catalytic agent. Hybrid materials
obtained using other types of acid or base all exhibit poor
textural properties at high temperatures and are inferior for CVD
SWNT growth. Si-F is known as one of the strongest chemical
bonds and remains intact up to ∼2000 °C. Furthermore, sol-
gel chemistry has established the displacement of -OH groups
lower yield and purity SWNTs than the materials obtained by
the Fe/Mo/Al2O3 and Fe/Mo/hybrid catalysts. Graphitic particles
with enclosed metal nanocrystals were frequently observed in
the synthesized material, indicating weaker metal-support
interactions on silica supports than on alumina and the hybrid
supports. This led us to investigate the surface chemistry aspects
of the support materials.
In Figure 10a, we show DRFTIRS spectra of hydroxyl groups
on the surfaces of alumina, silica and the hybrid material. The
spectra were recorded at room temperature after removing
physisorbed water by heating the samples to 500 °C in a vacuum
-
on the silica surface by F to give Si-F groups, which leads to
less water content and enhances the thermal stability of the silica
structure.3
7-39
We indeed confirmed the presence of F in the
-3
(1 × 10 Torr) for 0.5 h. The pure alumina material was found
hybrid based catalyst after heating to 900 °C by XPS. A peak
positioned at 686 eV was observed corresponding to F (1s) in
the XPS spectrum. The idea of using silica phase to improve
textural properties and reduce sintering of catalysts has been
used previously in catalytic reactions. For instance, Miller et
al. have reported improved textural and catalytic properties of
ZrO2/silica catalyst over pure ZrO2 catalyst for the 1-butene
isomerization reactions.40
The Nature of Support Materials. We have systematically
investigated various silica, alumina and multicomponent silica-
alumina supported catalysts for the methane CVD process
throughout this work. A general conclusion is reached that silica
supported catalysts are inferior to alumina based catalysts for
synthesizing high yield and purity bulk SWNT materials, even
when the catalysts have similar textural properties. For instance,
we prepared a pure silica support in the same manner as the
hybrid support described above without the presence of Al2O3.
The catalyst based on the pure silica material exhibits similar
surface area and pore volume as the Fe/Mo/Al2O3 and the Fe/
Mo/hybrid catalysts. As found by TEM and weight gain studies,
the CVD material made by the silica based catalyst led to much
to exhibit several types of hydroxyl groups in the range of
-
1 41-43
3400-3750 cm ,
while the silica shows a peak at 3746
-1
44
cm corresponding to terminal hydroxyl groups. The spectrum
in the -OH regime for the hybrid material qualitatively
resembles the sum of the spectra of Al O and silica. Pyridine
2
3
adsorption-desorption studies on silica found only weakly
H-bonded pyridine to surface hydroxyls (peaks at 1447 and 1597
-1
44
cm , diminished upon heating to 150 °C; data not shown).
Al2O3 exhibits Lewis acidity as shown in Figure 10b, where
-
1
two peaks (1454 and 1615 cm ) corresponding to Lewis-
coordinated pyridine persist up to 500 °C.
43-45
Figure 10c shows
pyridine desorption results for the hybrid support that exhibit
strong Lewis acidity (persistent peaks at 1492, 1597, and 1622
-
1
cm ) and some Br o¨ nsted acidity (peaks centered at 1549 and
-1
1640 cm ). The Lewis acidity of the hybrid material is stronger
than that of alumina as evidenced by the more persistent Lewis-
coordinated pyridine peaks upon heating to high temperatures.
We attribute this enhanced acidity to the inductive effects of
4
4
Si-F groups in the hybrid material. On the other hand, the
Br o¨ nsted acidity of the hybrid could be due to strongly
H-bonded H2O molecules or trapped HF molecules.

6490 J. Phys. Chem. B, Vol. 103, No. 31, 1999
Cassell et al.
Figure 7. Low magnification TEM images of SWNT materials
synthesized by CH CVD in 10 min using the Fe/Mo/hybrid catalyst.
a and b are images recorded with samples obtained by different CVD
runs, showing the high reproducibility of the synthesis process.
4
Figure 8. (a) SEM and (b-d) high magnification TEM images of
SWNT materials synthesized using the Fe/Mo/hybrid catalyst.
The IR results presented above have several implications.
First, it is confirmed that the alumina is not fully covered by
the silica phase in the hybrid support material, and the silica
phase is merely interconnecting the alumina particles. This is
consistent with our proposal that the hybrid support leads to a
better performing catalyst due to reduced sintering by the silica
“
spacers”. Second, since both the alumina and hybrid support
lead to excellent catalysts for CVD growth of SWNTs, and the
hybrid support has highly accessible alumina surface sites, we
conclude that alumina is in general better than silica as catalyst
support for the methane CVD process. A main reason seems to
be that metal-support interactions are weaker on silica supports
than on alumina supports, as found by the TEM observation of
metal particle lift-off in silica supported catalysts after CVD.
The strong metal-support interaction between alumina surface
sites and Fe2(SO4)3 can be attributed to chemical reactions that
Figure 9. N
Fe/Mo/hybrid catalysts after heating to 900 °C in Ar. The Fe/Mo/hybrid
catalyst exhibits higher surface area (higher adsorbed volume at P /P
2 2 3
adsorption and desorption isotherms of Fe/Mo/Al O and
occur in the impregnation step.3
2,33
However, it is not clear at
the present time if the Lewis acid sites on the alumina surface
are directly involved in reactions with methane via heterolytic
s
0
∼ 0.05) and larger mesopore volume (higher adsorbed volume at P /
s
46
cleavage, affording a more active catalyst material. This issue
will be investigated in future studies.
P
0
∼ 1) than those of the Fe/Mo/Al
2 3
O catalyst.
SWNT Growth Mechanism and Rate. In the SWNT
materials synthesized by Fe2(SO4)3/Al2O3, Fe/Mo/Al2O3 and Fe/
Mo/hybrid catalysts, we frequently observed closed SWNT ends
that are free of encapsulated metal particles, far away from the
catalyst support materials (Figure 11a, left part Figure 8c). This
strongly suggests that the SWNTs are formed predominantly
through the “base growth” mechanism in the methane CVD
process. That is, during the CVD process, the catalyst particles
responsible for nanotube nucleation and growth remain pinned
25,26,47-49
on the support surface.
However, in the CVD materials

Single-Walled Carbon Nanotubes
J. Phys. Chem. B, Vol. 103, No. 31, 1999 6491
Figure 11. TEM images of SWNTs. (a) Left part: high-resolution
image showing closed nanotube ends in the CVD material synthesized
from the Fe/Mo/hybrid catalyst. Right part: a small fraction of SWNTs
contain metal particles at their ends in the same material. (b) High-
resolution TEM images of nanotubes in the materials synthesized from
the Fe/Ru/Al O catalyst. Left part: a SWNT with closed end. Right
Figure 10. DRFTIRS of catalyst support materials. (a) Spectra of
Al
2
O
3
, silica, and the Al
2
O
3
-SiO hybrid material in the -OH stretching
2
2
3
region. (b) Results of pyridine desorption experiments on the Al
2
O
3
part: a SWNT with metal particle at the end. (c) Disconnected and
circular tubes in the CVD material synthesized by using the Fe/Ru/
Al O catalyst.
surface. (c) Results of pyridine desorption studies of the Al
hybrid.
2 3
O -SiO
2
2
3
synthesized from the bimetallic Fe/Mo/Al2O3 and Fe/Mo/hybrid
catalysts, we also noticed a small fraction (∼5%) of SWNTs
that do contain metal particles at their ends (Figure 11a, right
part). This indicates tip-growth as the secondary nanotube
growth mechanism in methane CVD using the Fe/Mo catalysts.
Similarly, in the CVD material obtained by the Fe/Ru/Al2O3
catalyst, the majority of the nanotube ends were found closed
and metal-particle free (Figure 11b, right part), while a small
fraction of metal containing SWNT ends was observed (Figure
the Ru catalyst indeed led to SWNTs in the methane CVD
process.
5
1
An obvious feature in the weight gain data for all catalyst
systems is the decrease in growth rate with time (Figure 3).
That is, the weight gain deviates from a linear relationship with
time and levels off after only a few minutes of CVD reactions.
This phenomenon seems to be common in the CVD growth of
nanotubes. For instance, in our recent work of growing regular
arrays of aligned multiwalled nanotubes on porous silicon
substrates, the nanotube length as a function of time was found
to deviate from a straight line after ∼20-30 min of CVD
11b, lower part). Also, we frequently observed crop circles as
well as “disconnected” nanotubes (Figures 11c and 6c) in the
CVD material derived from the Fe/Ru/Al2O3 catalyst, which
indicates that a gas-phase nanotube growth mechanism operates
in this system. We propose that volatilized Ru species are
responsible for catalyzing the formation of the disconnected and
1
3
reaction. In the growth of SWNTs using ethylene as carbon
feedstock, the growth rate was also found to decrease at long
reaction times and was attributed to the thickening mat of
2
8
nanotubes. We believe that at least one of the rate-limiting
steps in CVD is gas diffusion. Comparing the weight gain curves
in Figure 3 obtained with various catalysts and correlating them
with the textural properties of the catalysts in Table 1, we do
observe the trend that a catalyst exhibiting more open pore
structures (larger pore volume) improves the nanotube growth
rate and yield (weight gain). These results strongly suggest the
importance of highly porous catalysts for CVD nanotube
synthesis. A detailed mechanism is lacking to account for the
5
0
circular nanotubes. From the Fe-Ru binary phase diagram
and due to the small-size effect, it is possible that the Ru
nanophase in the Fe-Ru mixture exhibits a low melting
temperature (∼1000 °C) and sufficient vapor pressure. Detailed
mechanisms for the metal vaporization and the subsequent gas
phase nanotube formation are currently lacking. Nevertheless,
we did confirm the activity of Ru species in catalytic growth
of SWNTs. We prepared a Ru/Al2O3 catalyst and found that

6492 J. Phys. Chem. B, Vol. 103, No. 31, 1999
Cassell et al.
growth rate decrease and termination at long reaction times.
However, we believe that plausible possibilities include pore
clogging and catalyst poisoning by the grown nanotubes and
small amounts of deposited amorphous carbon.
(5) Dai, H. J.; Hafner, J. H.; Rinzler, A. G.; Colbert, D. T.; Smalley,
R. E. Nature 1996, 384, 147.
(
(
6) Dai, H.; Franklin, N.; Han, J. Appl. Phys. Lett. 1998, 73, 1508.
7) Wong, S.; Harper, J.; Lansbury, P.; Lieber, C. M. J. Am. Chem.
Soc. 1998, 120, 603.
8) Wong, S.; Joselevich, E.; Woolley, A.; Cheung, C.; Lieber, C.
Nature 1998, 394, 52.
9) Rinzler, A. G.; Hafner, J. H.; Nikolaev, P.; Lou, L.; Kim, S. G.;
Tomanek, D.; Nordlander, P.; Colbert, D. T.; Smalley, R. E. Science 1995,
(
4. Conclusion and Future Direction
(
The synthesis of bulk amounts of high quality single-walled
carbon nanotubes (SWNTs) is accomplished by systematically
optimizing the chemical compositions and textural properties
of catalysts within the methane CVD synthetic approach.
Materials that contain ∼42 wt % of SWNTs are obtained by
CVD and scale-up to the gram level has been demonstrated.
The nanotube materials consist of individual and bundled
SWNTs that are nearly free of defects and amorphous carbon
coating on the nanotube walls.
2
69, 1550.
(10) de Heer, W. A.; Chatelain, A.; Ugarte, D. Science 1995, 270, 1179.
(
11) Saito, Y.; Hamaguchi, K.; Hata, K.; Tohji, K. Ultramicroscopy
998, 73, 1.
12) Wang, Q.; Setlur, A.; Lauerhaas, J.; Dai, J.; Chang, R. H. Appl.
1
(
Phys. Lett. 1998, 72, 2912.
(13) Fan, S.; Chapline, M.; Franklin, N.; Tombler, T.; Cassell, A.; Dai,
H. Science 1999, 283, 512.
(
(
14) Tans, S.; Verschueren, A.; Dekker: C. Nature 1998, 393, 49.
15) Martel, R.; Schmidt, T.; Shea, H. R.; Hertel, T.; Avouris, P. Appl.
We have taken significant steps to understand the catalyst
materials and the CVD process. The key to the synthesis of
high yield, quality, and purity SWNTs is found to be the catalyst
material. Iron sulfate is found to interact strongly with the
surface of alumina support, which leads to high metal dispersion,
virtually eliminates metal lift-off and aggregation and produces
textural mesopores. In terms of chemical compositions, the Fe/
Mo bimetallic catalyst increases SWNT yield over Fe catalysts
possibly by promoting aromatization reactions of methane. The
Fe/Ru bimetallic catalyst also leads to higher yield of SWNTs
through the formation of circular tubes and “free-floating” tubes
in the gas phase. Alumina is found to be a superior catalyst
support material over silica partly due to stronger metal-support
interactions on the alumina. A novel Al2O3-SiO2 hybrid support
material is synthesized by sol-gel to further increase the yield
of SWNTs. The alumina phase in the hybrid support retains its
chemical characteristics without being fully covered by the silica
phase. The advantages of the hybrid support over the pure
alumina support include higher surface-area and larger pore
volume at elevated temperatures due to the silica “spacing
effect”.
There is still significant room for future work to optimize
methane CVD SWNT synthesis and understand the processes
involved in the catalysis of C1 to C∞. In terms of SWNT yield
optimization, future work should include rational design and
creation of new types of catalyst support materials and catalysts.
An ideal catalyst should exhibit the desired textural properties
that are stable at elevated temperatures in order to obtain high
densities of catalytic sites and enable efficient and steady
nanotube growth. For instance, a catalyst based on an alumina
support that exhibits several level of pore structures (e.g., 1 nm
scale and 1-10 µm scale) will allow efficient catalyst particle
formations and “free” space for gas diffusion and nanotube
growth. In terms of the chemistry aspects of catalyst materials
and CVD processes, detailed understanding should be obtained
on the microscopic catalyst structures, metal-support interac-
tions and mechanisms of nanotube growth, kinetics, and ter-
mination. Further understanding and optimization could enable
the synthesis of 10 g of SWNTs by 1 g of catalyst by CVD.
Phys. Lett. 1998, 73, 2447.
(16) Soh, H.; Morpurgo, A.; Kong, J.; Marcus, C.; Quate, C.; Dai, H.
Appl. Phys. Lett. 1999. Submitted for publication.
(17) Bethune, D. S.; Kiang, C. H.; DeVries, M.; Gorman, G.; Savoy,
R.; Vazquez, J.; Beyers, R. Nature 1993, 363, 605.
(18) Journet, C.; Maser, W. K.; Bernier, P.; Loiseau, A.; Delachapelle,
M. L.; Lefrant, S.; Deniard, P.; Lee, R.; Fischer, J. E. Nature 1997, 388,
7
56.
(19) Thess, A.; Lee, R.; Nikolaev, P.; Dai, H. J.; 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.
(
20) Tibbetts, G. G. In Carbon Fibers, Filaments and Composites;
Kluwer Academic: Amsterdam, 1990; pp 73-94.
21) Tibbetts, G. G. Appl. Phys. Lett. 1983, 42, 666.
(
(22) Endo, M. Chemtech 1988, 568-576.
(23) Snyder, C. E.; Mandeville, W. H.; Tennent, H. G.; Truesdale, L.
K. US Patent, 1989.
(24) Baker, R. T. K.; Rodriguez, N. M. In Symposium of the Materials
Research Society; Materials Research Society: 1994; Vol. 349, p 251.
(25) Kong, J.; Cassell, A. M.; Dai, H. Chem. Phys. Lett. 1998, 292,
5
67.
(26) Kong, J.; Soh, H.; Cassell, A.; Quate, C. F.; Dai, H. Nature 1998,
3
95, 878.
(
27) Cheng, H.; Li, F.; Su, G.; Pan, H.; Dresselhaus, M. Appl. Phys.
Lett. 1998, 72, 3282.
28) Hafner, J.; Bronikowski, M.; Azamian, B.; Nikolaev, P.; Colbert,
D.; Smalley, R. Chem. Phys. Lett. 1998, 296, 195.
29) Flahaut, E.; Govindaraj, A.; Peigney, A.; Laurent, C.; Rao, C. N.
(
(
Chem. Phys. Lett. 1999, 300, 236.
(30) Brunnauer, S.; Emmett, P.; Teller, E. J. Am. Chem. Soc. 1938, 60,
09.
3
(
73.
31) Barrett, E.; Joyner, L.; Halenda, P. J. Am. Chem. Soc. 1951, 73,
3
(32) Rajan, S. S. S. Soil Sci. Soc. Am. J. 1978, 42, 39.
(33) Okamoto, Y.; Imanaka, T. J. Phys. Chem. 1988, 92, 7102.
(
34) Wang, L.; Tao, L.; Xie, M.; Xu, G. Catal. Lett. 1993, 21, 35.
(35) Liu, J.; Dai, H. J.; Hafner, J. H.; Colbert, D. T.; Smalley, R. E.;
Tans, S. J.; Dekker: C. Nature 1997, 385, 780.
(36) Kong, J.; Dai, H. Unpublished results.
(37) Brinker, C. J.; Scherer, G. W. Sol-Gel Science: The Physics and
Chemistry of Sol-Gel Processing; Academic Press: San Diego, 1990.
(38) Nassau, K.; Rabinovich, E.; Miller, A.; Gallagher, P. J. Non-Cryst.
Solids 1986, 82, 78.
(39) Winter, W.; Chan, J.; Frattini, R.; Jonas, J. J. Noncryst. Solids 1988,
05, 214.
1
(
(
(
40) Miller, J.; Rankin, S.; Ko, E. J. Catal. 1994, 148, 673.
41) Morterra, C.; Magnacca, G. Catal. Today 1996, 27, 497.
42) Tsyganenko, A.; Mardilovich, P. J. Chem. Soc., Faraday Trans.
1996, 92, 4843.
(43) Liu, X.; Truitt, R. J. Am. Chem. Soc. 1998, 119, 99856.
Acknowledgment. This work is supported by National
Science Foundation (NSF Grant ECS-9871947), the Semicon-
ductor Research Corporation (SRC Grant 98-MJ-630), the
Camille Henry Dreyfus Fondation, and the American Chemical
Society.
(
44) Hair, M. L. Infrared Spectroscopy in Surface Chemistry; Marcel
Dekker: New York, 1967.
(45) Parry, E. J. Catal. 1963, 2, 371.
(
46) Schobert, H. The Chemistry of Hydrocarbon Fuels; Butterworth
&
Co.: Boston, 1990.
(47) Baker, R. T. K. Carbon 1989, 27, 315.
(
67.
48) Tibbetts, G. G.; Devour, M. G.; Rodda, E. J. Carbon 1987, 25,
References and Notes
3
(
1) Iijima, S. Nature 1991, 354, 56.
(49) Amelinckx, S.; Bernaerts, D.; Zhang, X. B.; Van Tendeloo, G.;
Van Landuyt, J. Science 1995, 267, 1334.
(2) Dresselhaus, M. S.; Dresselhaus, G.; Eklund, P. C. Science of
Fullerenes and Carbon Nanotubes; Academic Press: San Diego, 1996.
(50) Massalski, T. B. Binary Alloy Phase Diagrams; William W. Scott,
Jr., Ed.; American Society for Metals: Metals Park, OH, 1990.
(51) Cassell, A.; Dai, H. Unpublished results.
(3) Yakobson, B. I.; Smalley, R. E. Am. Sci. 1997, 85, 324.
4) Ebbesen, T. W. Phys. Today 1996, 49, 26.
(