Gas-phase catalytic growth of single-walled carbon nanotubes from carbon monoxide

Article abstract of DOI:10.1016/S0009-2614(99)01029-5

Single-walled carbon nanotubes (SWNTs) have been produced in a gas-phase catalytic process. Catalysts for SWNT growth form in situ by thermal decomposition of iron pentacarbonyl in a heated flow of carbon monoxide at pressures of 1-10 atm and temperatures of 800-1200°C. The SWNT yield and diameter distribution can be varied by controlling the process parameters, and SWNTs as small as 0.7 nm in diameter, the same as that of a C₆₀ molecule, have been generated. This process shows great promise for bulk production of carbon nanotubes.

Full text of DOI:10.1016/S0009-2614(99)01029-5

5 November 1999
Ž
.
Chemical Physics Letters 313 1999 91–97
www.elsevier.nlrlocatercplett
Gas-phase catalytic growth of single-walled carbon nanotubes
from carbon monoxide
Pavel Nikolaev, Michael J. Bronikowski, R. Kelley Bradley, Frank Rohmund,
)
Daniel T. Colbert, K.A. Smith, Richard E. Smalley
Rice UniÕersity, The Center for Nanoscale Science and Technology - MS 100, 6100 Main, Houston, TX 77005-1892, USA
Received 8 July 1999
Abstract
Ž
.
Single-walled carbon nanotubes SWNTs have been produced in a gas-phase catalytic process. Catalysts for SWNT
growth form in situ by thermal decomposition of iron pentacarbonyl in a heated flow of carbon monoxide at pressures of
1–10 atm and temperatures of 800–12008C. The SWNT yield and diameter distribution can be varied by controlling the
process parameters, and SWNTs as small as 0.7 nm in diameter, the same as that of a C60 molecule, have been generated.
This process shows great promise for bulk production of carbon nanotubes. q 1999 Elsevier Science B.V. All rights
reserved.
1. Introduction
A continuous-flow synthetic method, in which
Ž
.
SWNTs are grown and separated in a flowing
gaseous feedstock mixture, could produce SWNTs in
large quantities. Such schemes would involve intro-
ducing into the feedstock flow stream nanometer-size
catalyst particles on which the individual tubes nu-
cleate and grow. The catalyst particles could be
either pre-made or produced in situ by the introduc-
tion of metal-containing species into the flow and
their subsequent condensation into appropriately
sized clusters. The latter technique is especially con-
venient, because metal can be introduced in the form
of volatile organometallic molecules.
Ž
.
Single-walled carbon nanotubes SWNTs exhibit
many unique and useful physical and chemical prop-
w x
erties 1 . Demonstrated methods for producing
SWNTs involve laser vaporization of metal-doped
w x
carbon targets 2 , arc evaporation of metal-doped
w x
carbon electrodes 3 , and decomposition of carbon-
w x
containing molecules such as C2 H4 and CO 4 and
CH4 on supported nanometer-sized metal particles
that serve as catalysts for SWNT growth 5 . All of
these methods produce SWNTs in milligram to gram
quantities in a few hours. However, many potential
applications of SWNTs require kilogram to ton quan-
tities.
w x
Many groups have investigated gas-phase contin-
uous-flow production of carbon fibers. These studies
typically involve passing a mixture of carbon source
gas and organometallic catalyst precursor molecules
through a heated furnace. The organometallics de-
compose and react, forming clusters on which carbon
)
Corresponding author. Fax: q1-713-285-5320; e-mail:
res@cnst.rice.edu, web: http:rrcnst.rice.edurreshome.html
0009-2614r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
Ž
.
PII: S0009-2614 99 01029-5

(
)
92
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
Ž
.
Ž
.
5
fibers nucleate and grow. Tibbets et al. have reported
gas-phase synthesis of carbon fibers in heated flow-
ing mixtures of methane or hexane with orga-
reactor Fig. 1 . The products of Fe CO thermal
Ž
Ž
.
.
decomposition probably Fe CO , ns0–4 react to
produce iron clusters in gas phase. These clusters act
as nuclei upon which SWNTs nucleate and grow:
solid carbon is produced by CO disproportionation
n
Ž
Ž
. . w x
2
nometallics such as iron pentacarbonyl Fe CO
6
5
Ž
Ž
. . w x
and ferrocene dicyclopentadienyl iron, Fe Cp 7 .
w
x
Ž
.
Rao and co-workers 8–10 have investigated the
catalytic growth of carbon nanotubes in gas phase,
both by direct pyrolysis of ferrocene and other metal-
locenes and by catalytic decomposition of hydrocar-
bons or carbon monoxide in the presence of metal-
the Boudouard reaction :
COqCO ™ C s qCO ,
which occurs catalytically on the surface of the iron
particles. These particles also promote the formation
1
Ž .
Ž .
2
of the tube’s characteristic graphitic carbon lattice.
Ž
.
locenes or Fe CO . In two of these studies, this
5
Y
Ž
.
The flow cell apparatus consists of a 1 2.54 cm
group reported production of SWNTs from mixtures
OD thick-walled quartz flow tube contained within a
tube furnace, through which reactant gases are
flowed. The tube section inside the furnace was
maintained between 800 and 12008C, while the tube
entrance and exit were maintained at room tempera-
ture. We found that the rate at which the reactant
gases were heated had substantial effects on the
amount and quality of SWNTs produced. In some
Ž
.
w x
w x
of Fe CO with acetylene 8 and benzene 10 .
5
Very recently, Dresselhaus and co-workers have re-
ported the production of SWNTs in a heated flow of
benzene and ferrocene 11,12 .
To date, all investigations reporting gas-phase
continuous-flow production of SWNTs have relied
on hydrocarbons as the carbon source molecule for
tube growth. However, hydrocarbons pyrolize read-
ily on most surfaces heated above 600–7008C tem-
peratures required for growth of SWNTs , forming
graphitic deposits or amorphous carbon. Nanotubes
grown in heated flows of gaseous hydrocarbons of-
ten show substantial amorphous carbon overcoating,
which would need to be removed in subsequent
processing steps. One may anticipate that such ef-
fects would complicate efforts to scale up production
methods using hydrocarbon feedstocks.
We report here the catalytic production of SWNTs
in a continuous-flow gas-phase process using CO as
the carbon feedstock and Fe CO as the iron-con-
taining catalyst precursor. We find that both the yield
of SWNT material and the diameters of the na-
notubes produced can vary over a wide range, de-
pending on the conditions and flow-cell geometry
used. In particular, we have succeeded in producing
SWNTs with diameters of 0.7 nm, which are ex-
pected to be the smallest achievable chemically sta-
w
x
Ž
.
experiments the CO and Fe CO were introduced
Ž
5
through a water-cooled injector positioned inside the
quartz tube, which maintained the gases at low tem-
perature until they were injected into the furnace,
resulting in rapid heating. Around the exit of this
injector could be positioned a circle of narrow-gauge
needles through which preheated CO was passed at
high flow rate to mix with the cool flow emerging
from the injector, further increasing the heating rate
of the injected gas. The CO sprayed from this
‘showerhead’ mixer was preheated by passing
through a spiral heat exchanger positioned within the
furnace.
.
Ž
.
5
w x
Ž
.
ble SWNTs 1 . This process or variations of it
should be scalable for bulk SWNT production.
2. Experimental
We have produced SWNTs by flowing CO mixed
Fig. 1. Layout of CO flow-tube reactor, showing water-cooled
injector and ‘showerhead’ mixer.
Ž
.
with a small amount of Fe CO through a heated
5

(
)
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
93
Ž
Our gas mixture consisted of CO 1–10 atm
growth in the gas phase. This was confirmed in
initial experiments conducted without the cooled in-
jector. In these experiments, 6.5 atm of unpurified
CO was flowed directly through the quartz tube
heated to 8508C. We determined that the CO con-
.
flowing at 1–2 standard liters per minute dosed with
a small amount 0–25 mTorr of gaseous Fe CO .
Dosing was accomplished by passing a measured
Ž
.
Ž
.
5
Ž
.
fraction of the CO stream through a Fe CO -filled
5
Ž
.
Ž
.
bubbler maintained at 08C. This produced a partial
pressure of Fe CO equal to its vapor pressure of
tained 4–5 ppm of Fe CO , giving a Fe CO con-
5
5
Ž
.
centration of 20–25 mTorr. When the flow velocity
in the region of transition from room temperature to
5
Ž
.
;7 Torr, and the partial pressure of Fe CO was
5
Ž
.
subsequently reduced by diluting with additional CO
before its introduction into the flow cell. Commercial
SWNT growth temperature ;10 cm long was
around 10–20 cmrs, we found that SWNTs were
produced and deposited as thin mats or films on the
cold parts of the quartz tube. When the flow velocity
Ž
.
CO will always contain a few ppm of Fe CO as a
5
contaminant, which we filtered out using an in-line
catalytic purifier Oxyzorb, Matheson prior to intro-
duction into our flow system. Alternately, this small
‘background’ concentration of Fe CO was used
directly in experiments with high CO pressure andror
low Fe CO concentration because the bubbler was
found to be unreliable under these conditions. The
1
Ž
.
was slower , 0.5–1 cmrs, no SWNTs were formed.
Fig. 2 shows TEM and SEM images of the SWNT
material from these experiments. Approximately 3.6
mg of material was produced in 16 h of gas flow at
1500 sccm. EDX and TGA showed that the material
contained 45% of iron atoms, or 75% iron by weight.
Ž
.
5
Ž
.
5
Ž
.
Ž
.
concentration of background Fe CO in the CO
Much more rapid heating of the COrFe CO
5
5
stream, typically 2–10 ppm, was determined by mea-
suring the total amount of iron deposited in the
reactor. In control experiments using no Fe CO
with the catalytic purifier in use , no iron or carbon
mixture was achieved using the cooled injector, and
this setup was used to study the dependence of
SWNT production on temperature and pressure. In
Ž
.
5
Ž
.
Ž
.
these experiments, CO containing 5 ppm of Fe CO
5
reaction products were observed.
was flowed through the heated reactor tube at 1000
sccm for 5–15 h. Nanotube containing material was
deposited on the walls of the quartz tube and on a
filter positioned downstream of the reactor. This
material was collected, weighed, and the mass frac-
tion of carbon was determined by TGA. TEM obser-
vations of the product material suggest that most of
the carbon was present as SWNTs, rather than other
forms such as amorphous carbon. TEM images also
gave the diameter distributions of the product
SWNTs. Very few tube ends were observed, imply-
ing that the tubes are long compared to their width:
we estimate average tube length of ;1 mm, based
on the number of tube ends observed.
Ž
.
Flow of COrFe CO mixtures through the heated
5
reactor resulted in black deposits on the walls of the
quartz tube outside the furnace. These deposits con-
sisted of SWNTs and iron particles apparently over-
coated with carbon. Material was collected and
weighed, and analyzed using transmission electron
Ž
.
microscopy TEM , scanning electron microscopy
Ž
.
Ž
.
SEM , energy-dispersive X-ray spectroscopy EDX ,
and thermogravimetric analysis TGA .
Ž
.
3. Results
Gas-phase iron pentacarbonyl will decompose
Table 1 lists the rate of SWNT production vs.
temperature at 10 atm. We found that the highest
production rate was achieved at our furnace’s highest
available temperature, 12008C. We next investigated
w
x
Ž .
rapidly at 2508C 13 . Reaction 1 occurs at a signif-
w
x
icant rate only at temperatures above 5008C 14 .
One may thus anticipate that the rate at which the
gas mixture is heated through the temperature range
of 200–5008C will be important in determining the
outcome of this process: too slow a heating rate
could allow the forming Fe clusters to grow too large
or encounter and stick to the quartz tube wall before
1 This experiment was performed with the furnace positioned
vertically, which forced a uniform gas flow. In the former experi-
ment the furnace was positioned horizontally, and gas was drawn
Ž .
they reach the temperature required for reaction 1
Ž
.
into the furnace much faster 10–20 cmrs due to formation of a
to proceed, thus preventing SWNT nucleation and
convection cell.

(
)
94
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
Table 2
Pressure
Production rate
SWNT yield
SWNT yield
Ž
.
Ž
.
Ž
.
Ž
.
atm
mgrh
mole%
wt.%
1
3
10
1.16
1.38
1.24
61
67
79
25
30
44
5 ppm. Again we found that production rate in-
creased with increasing pressure up to the maximum
pressure we could achieve, 10 atm. Table 2 lists the
SWNT production rate vs. CO pressure. We also
observed that the diameters of the product SWNTs
varied as reaction conditions were varied. Fig. 3
shows the diameter distributions observed at various
pressures. The product SWNTs consistently showed
smaller diameters as pressure was increased. Reac-
tion at 10 atm gave tubes as small as 0.7 nm in
diameter, which is about the diameter of a C60
molecule.
Experiments with the cooled injector alone using
Ž
.
higher Fe CO concentrations did not give addi-
5
tional nanotube material. Rather, iron in excess of 5
ppm merely formed additional iron particles mixed
in with the product material. This inefficient use of
available iron could result from too small a heating
Ž
.
rate of the COrFe CO mixture. Therefore the
5
Ž
.
‘showerhead’ mixer Fig. 1 was implemented to
achieve even faster heating of the injected gases. In
experiments with this mixer the CO pressure was
maintained at 3 atm rather than 10 atm in order to
Ž .
Fig. 2. a SEM image of SWNT ropes produced in a flow of CO
Ž
.
achieve higher gas flow velocities. The COrFe CO
5
Ž
.
Ž
Ž .
.
Ž
.
6.5 atm and Fe CO 5 ppm through the flow-tube apparatus
5
mixture was flowed through the cooled injector at
heated to 8508C. b TEM image of nanotubes produced in a flow
Ž
.
1000 sccm, while hot gas pure CO was sprayed
from the mixer at 5000 sccm. All showerhead mixer
experiments were carried out at 12008C.
Ž
.
Ž
.
Ž
.
of CO 6.5 atm and Fe CO 5 ppm heated to 8508C, showing
5
nanotube rope cross-section.
Using this setup we were able to utilize somewhat
SWNT production rate vs. pressure at 12008C, where
the CO pressure was varied from 1 to 10 atm and the
Ž
.
Ž
.
higher Fe CO concentrations in the CO. Fe CO
5
5
concentrations of 8 ppm gave SWNT production
rates of up to 15 mgrh. The showerhead also led to
more efficient use of the available iron: 8 ppm of
Ž
.
relative concentration of Fe CO was maintained at
5
Ž
.
Fe CO gave product material containing only 7
5
Table 1
mole% iron, compared to 31 mole% obtained using 5
ppm Fe CO with the cooled injector alone. The
diameter distributions of the SWNTs from the show-
erhead configuration were similar to those from the
cooled injector alone Fig. 3 . Clearly, the rate of
heating of the COrFe CO mixture has a substan-
Temperature Production rate SWNT yield SWNT yield
Ž
.
5
Ž
8C
.
Ž
.
Ž
.
Ž
.
mgrh
mole%
wt.%
850
1000
1200
0.68
1.00
1.24
73
71
79
37
34
44
Ž
Ž
.
.
5

(
)
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
95
Ž
growth. Small amounts of methane 0.7% by vol-
.
ume produced clean nanotubes in increased yields:
the yield with methane was 20 mgrh of material
containing 5 at.% iron, vs. 15 mgrh of 7% iron
material produced under identical conditions without
methane. However, higher CH4 concentrations
Ž
.
1.4% gave nanotubes with significant amorphous
Fig. 3. Diameter distributions of SWNT produced at 12008C, at
various CO pressures.
tial effect on the SWNT material produced. Fig. 4
shows TEM images of the SWNT material produced
using the cooled injector with the showerhead mixer.
Note from the TEM images in both Figs. 2 and 4 that
SWNTs produced by this process are essentially free
of amorphous carbon overcoating.
One disadvantage of CO over hydrocarbons as a
carbon feedstock is that, at a given temperature and
pressure, CO disproportionation is much slower than
hydrocarbon thermal decomposition. We therefore
performed additional experiments in which a small
amount of methane was added to the CO flow to
increase the amount of carbon available for SWNT
Ž .
a Low-magnification TEM image of SWNT material
produced using the cooled injector with showerhead mixer. Condi-
Fig. 4.
Ž
.
Ž .
tions: 3 atm CO with 8 ppm of Fe CO 5, 12008C. b High-mag-
Ž .
nification TEM image of SWNT material in a .

(
)
96
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
Ž .
increased. Reaction 1 should be faster as these
parameters are increased.
overcoating unless the reaction temperature was low-
ered from 1200 to 11008C.
Of central importance is the mechanism by which
nanotube growth ceases. The particles in the product
material are 5–10 nm in diameter, much larger than
SWNT diameters, suggesting that the particles con-
tinue to grow even after nucleating a tube. The
additional accreting iron atoms could come from
either direct gas-phase collisions with other atoms or
clusters, or by adsorption of Fe atoms onto the
growing tube and their diffusion to the particle at the
end. Note that we never see nanotubes with non-con-
stant diameter: apparently, once the SWNT is nucle-
ated, additional carbon atoms can only add to the
existing structure with its fixed diameter. As long as
the catalytic particle is about the same size as the
nanotube growing from it, it is energetically more
favorable for carbon to add to the nanotube than to
form a layer overcoating the catalyst particle. As the
catalyst particle becomes larger, however, the strain
energy of the overcoating layer becomes smaller, and
4. Discussion
We propose a model for gas-phase nanotube
growth in which metal clusters form first, then nucle-
ate and grow SWNTs. Metal clusters initially form
by aggregation of iron atoms from the decomposition
Ž
.
of Fe CO . Clusters grow by collision with addi-
5
tional metal atoms and other clusters, eventually
reaching a diameter comparable to that of a SWNT,
0.7–1.4 nm, corresponding to 50–200 iron atoms.
By the time they reach that size, CO can dispropor-
tionate on the surface of such clusters via reaction
Ž .
1 to yield solid carbon, and SWNTs will nucleate
and grow from them by the ‘Yarmulke’ mechanism
w
x
15 . That is, a hemifullerene cap forms on the
partially carbon-coated particle, lifts off, and addi-
tional carbon atoms are continuously added to the
edge of the cap, forming a hollow tube of constant
diameter which grows away from the particle. As
w x
formation of a graphitic overcoat is favored 4 . Thus
the catalyst particle ultimately overcoats, terminating
further nanotube growth. The particle might separate
w
x
with supported catalysts 4,15 , one expects that the
diameter of the nanotube will reflect the size of the
catalytic particle at the time of tube nucleation. The
observed SWNT diameter distribution should thus
correspond to the size distribution of the catalyst
particles that spawn nanotubes. The smallest SWNTs
we ever see are ;0.7 nm in diameter, the size of a
C60 molecule. Since in this model growing tubes
Ž
.
from the nanotube end which would seal over ,
but most likely the particlernanotube pair remain
physisorbed together.
We estimate a product SWNT length of ;1 mm
5
Ž
.
;10 carbon atoms , based on the frequency with
which tube ends are observed in TEM images. The
particles in our product material, 5–10 nm wide, will
contain 104–105 iron atoms. Our observed C:Fe
mole ratios of between 5:1 and 10:1 thus imply that
the number of nanotubes in our product material is
approximately equal to the number of iron particles,
suggesting that the particles in the product material
are the same ones that nucleated the nanotubes when
they were smaller.
Even under our best conditions the ratio of SWNTs
produced to iron consumed is not very high. Ideally,
every iron atom would be part of a cluster of 50–200
atoms that spawns its own nanotube. If, on average,
a 100-atom iron cluster spawns and grows a 1 mm
tube, we would see product material with a C:Fe
mole ratio of 1000:1. Our observed C:Fe ratio of
about 10:1, thus implies that most of the iron is
consumed not in forming nuclei, but in the continued
growth of particles after SWNT nucleation: ongoing
Ž
will be capped with a hemifullerene one half of a
spherical fullerene molecule , our observation is con-
.
sistent with the fact that C60 is the smallest stable
fullerene: SWNTs cannot nucleate on smaller Fe
particles because no stable end cap can form. Delays
in cap formation and lift-off will allow the Fe parti-
cle to accumulate more iron and grow larger so that,
if a SWNT does eventually nucleate it will have a
larger diameter. This explains why we observe
smaller SWNT diameters and narrower diameter dis-
tributions as CO pressure is increased: higher CO
pressure leads to faster CO disproportionation, giv-
ing more C atoms on each Fe particle and faster
Ž
.
earlier SWNT nucleation relative to continued par-
ticle growth. Also, more SWNTs are produced, both
in absolute terms and relative to the mass of iron
catalyst consumed, as temperature and pressure are

(
)
P. NikolaeÕ et al.rChemical Physics Letters 313 1999 91–97
97
Fe accretion is faster than formation of initial tube-
nucleating clusters. Note that the 10:1 carbon-to-iron
ratio is maintained over a wide range of experimen-
tal parameters and is approximately the same as that
production methods: it is a continuous-flow process
rather than a batch process, and should be able to be
scaled up to produce SWNTs in much larger quanti-
ties.
w x
observed in the laser-oven apparatus 2 .
The yield of carbon to iron can be increased in
either of two ways. If a growing SWNT ‘lives’ for
only a certain time before growth ceases due to
excess iron accretion followed by overcoating, one
may be able to increase yield simply by operating at
higher CO pressures, which reduce the relative
amount of catalyst present and may increase the rate
Acknowledgements
This work was supported by the National Aero-
nautics and Space Administration, the Texas Ad-
vanced Technology Program, and the Robert A.
Welch Foundation.
Ž .
of reaction 1 . Another strategy is to increase the
fraction of the metal that goes into small, tube-
nucleating clusters relative to the fraction that ac-
cretes onto active catalyst clusters. If this can be
achieved, more of the iron will contribute to initiat-
ing tube growth, and less to ending it. If initial
References
w x
Ž
.
1
w x
2
B.I. Yakobson, R.E. Smalley, Am. Sci. 85 1997 324.
A. Thess, P. Nikolaev, H.J. Dai, P. Petit, J. Robert, C.H. Xu,
Y.H. Lee, S.G. Kim, A.G. Rinzler, D.T. Colbert, G.E. Scuse-
ria, D. Tomanek, J.E. Fischer, R.E. Smalley et al., Science
Ž
.
273 1996 483.
Ž
.
thermal dissociation of Fe CO is the rate-limiting
5
w x
3
C. Journet, W.K. Maser, P. Bernier, A. Loiseau, M.L.
Delachapelle, S. Lefrant, P. Deniard, R. Lee, J.E. Fischer,
step in producing Fe clusters, one might speed up the
process using higher temperatures or using non-ther-
Ž
.
Ž
.
Nature London 388 1997 756.
Ž
.
mal means to decompose the Fe CO , e.g. laser
w x
4
J.H. Hafner, M.J. Bronikowski, B.R. Azamian, P. Nikolaev,
A.G. Rinzler, D.T. Colbert, K.A. Smith, R.E. Smalley, Chem.
5
photolysis. All of these possibilities are currently
being explored.
Ž
.
Phys. Lett. 296 1998 195.
w x
5
Ž
.
J. Kong, A.M. Cassel, H. Dai, Chem. Phys. Lett. 292 1998
567.
w x
6
G.G. Tibbetts, C.A. Bernardo, D.W. Gorkiewitz, R.L. Alig,
5. Conclusions
Ž
.
Carbon 32 1994 569.
w x
7
G.G. Tibbetts, D.W. Gorkiewicz, R.L. Alig, Carbon 31
Ž
.
1993 809.
We have demonstrated a technique for catalytic
production of single-walled nanotubes in gas phase.
Our flow reactor uses carbon monoxide as the car-
bon feedstock and gaseous iron pentacarbonyl as the
metal-particle catalyst precursor. Unlike methods in-
volving hydrocarbon feedstocks, our method results
in no amorphous carbon overcoating on the product
nanotubes. Nanotubes as small as 0.7 nm in diameter
have been produced, and the size and diameter distri-
bution of the nanotubes produced can be roughly
selected by controlling the pressure of CO in which
w x
8
B.C. Satishkumar, A. Govindaraj, R. Sen, C.N.R. Rao, Chem.
Ž
.
Phys. Lett. 293 1998 47.
w x
9
Ž
.
R. Sen, A. Govindaraj et al., Chem. Phys. Lett. 267 1998
276.
w
w
x
Ž
.
10 R. Sen, A. Govindaraj et al., Chem. Mater. 9 1997 2078.
x
11 H.M. Cheng, F. Li, G. Su, H.Y. Pan, L.L. He, X. Sun, M.S.
Ž
.
Dresselhaus, Appl. Phys. Lett. 72 1998 3282.
w
x
12 H.M. Cheng, F. Li, X. Sun, S.D.M. Brown, M.A. Pimenta,
A. Marucci, G. Dresselhaus, M.S. Dresselhaus, Chem. Phys.
Ž
.
Lett. 289 1998 602.
w
x
13 D. Nicholls, in: J.C. Bailar, Jr., H.J. Emeleus, R. Nyholm,
Ž
.
A.F. Trotman-Dickenson Eds. , Comprehensive Inorganic
Chemistry, vol. 3, Pergamon, Oxford, 1973, p. 990.
Ž
the reaction occurs. Highest yields and narrowest
w
w
x
14 G.D. Renshaw, C. Roscoe, P.L. Walker Jr., J. Catal. 18
.
tubes were produced at the highest accessible tem-
Ž
.
1970 164.
perature and pressure, 12008C and 10 atm, respec-
tively. This process has advantages over other SWNT
x
15 H.J. Dai, A.G. Rinzler, P. Nikolaev, A. Thess, D.T. Colbert,
Ž
.
R.E. Smalley, Chem. Phys. Lett. 260 1996 471.