Catalytic growth of single-wall carbon nanotubes from metal particles

Article abstract of DOI:10.1016/S0009-2614(98)01024-0

Single-walled carbon nanotubes (SWNTs) have been synthesized at milligram per hour rates by the catalytic decomposition of both carbon monoxide and ethylene over a supported metal catalyst known to produce larger multiwalled nanotubes. Under certain conditions, there is no termination of nanotube growth, and production appears to be limited only by the diffusion of reactant gas through the product nanotube mat that covers the catalyst. Further development of the catalyst geometry to overcome the diffusion limitation may allow bulk catalytic production of SWNTs by supported metal catalysts.

Full text of DOI:10.1016/S0009-2614(98)01024-0

30 October 1998
Ž
.
Chemical Physics Letters 296 1998 195–202
Catalytic growth of single-wall carbon nanotubes from metal
particles
Jason H. Hafner, Michael J. Bronikowski, Bobak R. Azamian, Pavel Nikolaev,
)
Andrew G. Rinzler, Daniel T. Colbert , Ken A. Smith, Richard E. Smalley
Center for Nanoscale Science and Technology, Rice Quantum Institute, and Departments of Chemistry and Physics, M.S. 100, Rice
UniÕersity, P.O. Box 1892, Houston, TX 77251, USA
Received 25 June 1998
Abstract
Ž
.
Single-walled carbon nanotubes SWNTs have been synthesized at milligram per hour rates by the catalytic decomposi-
tion of both carbon monoxide and ethylene over a supported metal catalyst known to produce larger multiwalled nanotubes.
Under certain conditions, there is no termination of nanotube growth, and production appears to be limited only by the
diffusion of reactant gas through the product nanotube mat that covers the catalyst. Further development of the catalyst
geometry to overcome the diffusion limitation may allow bulk catalytic production of SWNTs by supported metal catalysts.
q 1998 Elsevier Science B.V. All rights reserved.
1. Introduction
particles supported on a substrate. The carbon feed-
stock molecules decompose on the particle surface,
and the resulting carbon atoms then diffuse through
the particle and precipitate as a part of nanotube
from one side of the particle. This procedure typi-
cally produces imperfect multiwalled nanotubes in
Ž
.
Single-wall carbon nanotubes SWNTs have been
previously produced in high yield by laser vaporiza-
tion of a graphite rod doped with Co and Ni 1 .
These high-quality samples have for the first time
enabled experimental confirmation of the structurally
dependent properties predicted for carbon nanotubes
w x
w x
high yield 5 . Its advantage is that it is relatively
simple and can be scaled to produce nanotubes by
w
x
w x
2,3 . High-quality SWNTs have also been generated
by arc evaporation of a graphite rod doped with Y
the kilogram 6 . Clearly, a method of making high-
quality SWNTs by catalytic decomposition could
lead to economic production of SWNTs in bulk.
There are two reports of SWNTs produced by
catalytic decomposition of stable gas-phase carbon-
containing molecules on pre-formed catalytic parti-
cles. Peigney et al. reported a mixture of single- and
multiwalled nanotubes resulting from decomposition
of CH4 at 10508C on an alumina-supported Fe cata-
w x
and Ni 4 . These techniques allow production of
only gram quantities of SWNTs. Another way to
synthesize nanotubes is by catalytic decomposition
of a carbon-containing gas by nanometer-scale metal
)
Corresponding author. E-mail: colbert@cnst.rice.edu
0009-2614r98r$ - see front matter q 1998 Elsevier Science B.V. All rights reserved.
Ž
.
PII: S0009-2614 98 01024-0

(
)
196
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
w x
lyst 7 . They do not report the relative amounts of
single- and multiwalled nanotubes. Dai et al. re-
ported the growth of SWNT and a small amount of
nanotubes are produced. This allows the growth
exclusively of SWNTs from catalyst systems previ-
ously thought to produce only larger-diameter multi-
walled nanotubes.
Ž
.
double-wall nanotubes DWNTs by disproportiona-
tion of CO on alumina-supported Mo particles at
w x
12008C 8 . The nanotubes had diameters ranging
from 1 to 5 nm and were 100 nm to microns in
2. Experimental
length, but grew in low yield and appeared by trans-
Ž
.
mission electron microscopy TEM observation to
be more defective than SWNT formed by the laser
vaporization or arc techniques.
SWNTs are grown by passing carbon-containing
Ž
.
gases CO or C2 H4 at elevated temperatures over
nanometer-size metal particles supported on larger
Ž
.
We report here the production of high-quality
SWNTs, in some cases including DWNT, in yields
much larger than previously reported by catalytic
decomposition of carbon-containing precursor gases.
Under certain conditions, it appears that SWNTs can
grow continuously to arbitrary length. Our results
demonstrate a means for nucleating and growing
nanotubes only from the smallest of the supported
catalyst particles, which produce SWNTs, while de-
activating the larger particles so that no multiwalled
10–20 nm alumina particles. Two different metal
catalysts have been used, one containing pure Mo,
the other containing Fe and Mo in a ratio of 9:1.
Both catalysts are made using a method described
previously 5 . Briefly, fumed alumina Degussa is
stirred with methanol, and to the resulting slurry is
added a methanol solution of metal salts ferric
nitrate andror bis acetylacetonato -dioxomolyb-
denum VI . The combined slurry is stirred for sev-
w x
Ž
.
Ž
Ž
.
Ž
..
eral hours, dried in a rotary evaporator, baked in
Fig. 1. Individual SWNT grown by reacting 1200 sccm CO at 8508C over an alumina-supported Mo particle system.

(
)
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
197
vacuum at 1808C for 12–15 h, and ground into a fine
powder.
been greatly improved since an initial report by this
w x
laboratory 8 . The catalyst is 34:1 alumina:Mo by
For each growth experiment, a quartz boat con-
mass. The reaction is carried out at 8508C under a
flow of 1200 sccm of CO at 900 Torr. The resulting
material, which consists of SWNT very mono-
Ž
.
taining a carefully weighed amount typically 20 mg
of the catalyst powder was placed in the center of a 1
inch quartz tube furnace. The system was purged
with Ar, then heated under flowing reactant gases to
an elevated temperature for a controlled time. The
resulting catalyst material, which now also contains
reaction products dominated by SWNTs, was re-
moved from the boat and weighed again. The yield is
defined as the mass increase divided by the original
catalyst mass. Samples were prepared for TEM
imaging by sonicating this material in methanol and
drop-drying the resulting suspension onto TEM grids.
Ž
.
disperse in diameter 0.8–0.9 nm , is shown in Fig.
1. Particles of the fumed alumina support, 10–20 nm
in size, are also visible in this and subsequent TEM
images. The yield of nanotubes is plotted as a func-
tion of reaction time in Fig. 2. Note that yield
continues to increase even for very long reaction
times.
CO also forms nanotubes with a second catalyst,
this one prepared with 90:9:1 alumina:Fe:Mo by
mass. The reaction, when carried out exactly as
described above for the alumina:Mo catalyst, yields
nanotubes of a wider diameter distribution, 0.5–3
nm, with SWNTs and some DWNTs. A representa-
tive TEM image is shown in Fig. 3. For this catalyst,
the yield increases with time initially, but is limited
to ;40% after 1 h of exposure. No additional mass
3. Results
The production of SWNTs by the disproportiona-
tion of CO over alumina-supported Mo particles has
Fig. 2. Yield as a function of time for CO over Mo particles and C2 H4 over Fe:Mo particles. The fits give a time dependence of t0.5 and
t0.4, respectively.

(
)
198
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
Fig. 3. TEM image of SWNT grown at 8508C by 1200 sccm CO over an alumina:Fe:Mo catalyst.
increase is observed even for much longer exposures
the termination is temperature dependent. For reac-
tions at 8508C, the yield increases until it reaches
7%, at which point the growth terminates. As the
reaction temperature is lowered, the yield reaches
higher levels before growth termination. At 7008C,
the growth does not terminate, but its rate decreases
as shown in Fig. 2.
Ž
.
up to 20 h .
We have also grown SWNTs from C2 H4 using
this technique. The 90:9:1 alumina:Fe:Mo catalyst is
first reduced by exposing to 1000 sccm Ar and 0.33
sccm H2 at 8008C for 30 min. The growth reaction
then proceeds at the reaction temperature by adding
0.66 sccm C2 H4 to the gas flow. The resulting
product is nanotube bundles containing SWNTs and
Ž
.
DWNTs Fig. 4 . One hundred nanotube cross-sec-
tions were observed at several reaction temperatures
to count the relative number of single- to double-
walled nanotubes. The amount of DWNTs increases
from 30% at 7008C to 70% at 8508C. Outer diame-
ters of the individual tubes in a bundle range from
0.5 to 3 nm. There appears to be no correlation
between outer diameter and number of walls, other
4. Discussion
The most remarkable aspect of this series of
experiments is that, in every case, only SWNTs and
sometimes DWNTs are produced with diameters in
the range 0.5–3 nm. We see no 5–20 nm diameter
multiwalled nanotubes that are typically produced by
supported catalyst particles. We believe that the key
difference responsible for these effects is that we
have arranged for the growth reaction rate to be
limited by the supply of carbon to the catalyst parti-
cles, whereas the multiwalled nanotube growth is
Ž
.
than that the smallest nanotubes -1 nm diameter
are never double-walled.
The mass yield of nanotubes increases at a similar
rate for reaction temperatures from 700 to 8508C, but

(
)
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
199
Ž .
Ž .
Fig. 4. a SEM image of nanotube ropes grown at 8008C in 1000 sccm Ar, 0.66 sccm C2 H4, 0.33 sccm H2 , b TEM shows that the ropes
consist of nanotubes of diameter 0.5–3 nm and 1 or 2 walls.

(
)
200
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
thought to be limited by the diffusion of carbon
the energy per carbon atom as a function of structure
size. For SWNTs, the caps were neglected in favor
of the vastly greater number of carbon atoms in the
side walls. The energies of carbon atoms at the
nanotube–metal interface and nanotube ends were
neglected since we are only considering the final
product energies and not nucleation or growth mech-
anisms. We also considered a graphene capsule en-
tirely surrounding the catalyst particle. For all cata-
lyst particle diameters, one expects the graphene
cylinder to be lower in energy than the capsule since
the cylinder has only simple curvature compared
with the complex curvature of the capsule. However,
the attractive interaction between the graphene cap-
sule and the metal particle will lower the energy per
atom of the capsule. We used simple formulas for
the energies of curved graphene sheets for the nan-
w x
through the catalyst particles 9 .
In the reaction of CO to produce nanotubes, the
slow carbon supply rate arises because CO decompo-
sition is a bimolecular disproportionation that in-
volves the breaking of two strong CO triple bonds:
such a reaction is expected to proceed very slowly
except at very high pressures, much greater than our
reaction pressure of ;1 atm. The catalytic decom-
position of C2 H4 proceeds quickly at 1 atm, so in
this case we slow the reaction down by limiting the
partial pressure of C2 H4 to 0.5 Torr.
Evidence that we have successfully changed the
rate-limiting step from carbon diffusion through the
catalytic particle to carbon supply to the catalytic
particle can be found in three aspects of the C2 H4
system. First, it has been found that the product mass
increase rate varies linearly with the C2 H4 partial
pressure. Second, ignoring termination, the mass
growth rate is independent of the reaction tempera-
ture from 700 to 8508C. If the reaction were limited
by diffusion of carbon through the metal, the rate
would double from 700 to 8508C assuming an Arrhe-
nius temperature dependence and the activation en-
w
x
w x
otubes 10 and large fullerenes for the capsules 11 .
An estimate for the graphene–metal interaction was
taken from an experimental measurement of the en-
w
x
ergy of the graphite–ferrite interface 12 . The result,
displayed in Fig. 5, shows that the nanotube energy
is lower than that of the capsule in a diameter range
similar to the SWNT diameters we find in our
experiments. These calculations lend support to our
hypothesis that in our experiments supply-limited
growth allows more time to anneal to the lowest-en-
ergy structure so that smaller particles produce nan-
otubes while larger particles are encapsulated. This
model could give further insight into the presence or
absence of DWNTs and multiwalled nanotubes if the
relative graphene–graphene and graphene–metal in-
teraction strengths were well known.
w
x
ergy of carbon diffusion through iron 10 . Admit-
tedly, the current experiments only measure a bulk
growth rate as opposed to the microscopic growth
rate of an individual nanotube. However, assuming
that the same number of nanotubes nucleate per unit
mass of catalyst, the two rates are proportional.
Finally, we find that the bulk growth rate of carbon
on the catalyst equals 5% of the mass of carbon in
ethylene that flows over the catalyst. Although this is
not 100% as would be expected of a supply limited
reaction, a simple model assuming laminar flow
suggests that only 5% of the ethylene molecules
strike the catalyst bed.
For experiments in which the reaction time is
short, we have observed that the SWNTs grow with
a particle at one end and closed at the other. This
supports nucleation of these nanotubes by the
w x
Changing from a diffusion-limited to a supply-
limited reaction by lowering the carbon supply to the
catalyst reduces the carbon concentration in the cata-
lyst particles as they form nanotubes. A lower carbon
concentration will likely allow the structures to form
more slowly, giving each carbon atom more time to
anneal to its lowest energetic configuration. There-
fore, the observed products should match those which
are predicted to have the lowest energy.
yarmulke mechanism, described previously 8 , in
which a hemispherical graphene cap forms on the
catalyst particles and lifts off to nucleate closed
nanotubes.
As shown in Fig. 2, both CO disproportionation
over Mo catalyst particles at 8508C and the reaction
of C2 H4 with FerMo particles at 7008C appear to
generate SWNTs that grow continuously without
termination of the growth reaction. These results
constitute the first demonstration of continuous gen-
eration of SWNTs with lengths that are, in principle,
We considered the energetics of structures that
could grow off of the catalyst particles by calculating

(
)
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
201
Fig. 5. Energies of capsules an single-wall nanotubes relative to an infinite graphene sheet.
arbitrarily long. In practice, however, the mass of the
grown nanotubes exhibits a time dependence that is
less than linear, so that growth slows more and more
proportion of DWNTs increases with temperature as
well.
Ž
with increasing time a fit of the data sets in Fig. 2
give roughly square root dependencies of yield on
5. Conclusions
.
time . We attribute this slowing growth to the in-
creasing diffusion time of the carbon feed-stock
molecules through the thickening mat of nanotubes
surrounding the catalyst particles.
We have demonstrated the ability to grow nan-
otubes by catalytic decomposition of C2 H4 and CO
only from the small particles in a supported catalyst
system, leading to the growth of SWNTs and deacti-
vation of multiwalled nanotube growth by encapsula-
tion of larger particles. For certain conditions, nan-
otubes can apparently be grown to arbitrary length,
but become limited by the diffusion of reactants to
the catalyst particles. This problem has been solved
for the production of multiwalled nanotubes from
this catalyst by using flat alumina flakes, as opposed
to fumed alumina particles, so that the nanotubes
grow aligned in large bundles, keeping their growing
ends exposed to the gaseous feedstock. Similar mod-
ifications to the current technique may allow the
bulk production of SWNTs.
The growth termination mechanism for the FerMo
catalyst at higher temperatures is not clear. One
possibility is that a new graphene cap forms on the
surface of a particle which has already nucleated a
nanotube and encapsulates the particle, thus terminat-
ing the nanotube growth. That the termination is
more effective at higher temperatures could be ex-
plained in terms of a higher carbon concentration in
the particle at higher temperature leading to more
frequent nucleation of a second graphene cap. This
mechanism could also explain the presence of
DWNTs. If the second cap forms on the catalytic
particle inside the first nanotube, it could lift off to
nucleate a second nanotube. As stated earlier, the

(
)
202
J.H. Hafner et al.rChemical Physics Letters 296 1998 195–202
w x
5
C.E. Snyder, W.H. Mandeville, H.G. Tennent, L.K. Trues-
dale, J.J. Barber, Int. Pat. WO 9r07163, 1989.
References
w x
1
w x
A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C.
Xu, Y.H. Lee, S.G. Kim, A.G. Rinzler, D.T. Colbert, G.E.
Scuseria, D. Tomanek, J.E. Fischer, R.E. Smalley, Science
6
Hyperion Catalysis International, Cambridge, MA.
A. Peigney, Ch. Laurent, F. Dobigeon, A. Rousset, J. Mater.
w x
7
Ž
.
Res. 12 1997 613.
Ž
.
w x
8
273 1996 483.
H. Dai, A.G. Rinzler, P. Nikolaev, A. Thess, D.T. Colbert,
w x
2
Ž
.
J.W.G. Wildoer, L.C. Venema, A.G. Rinzler, R.E. Smalley,
C. Dekker, Nature London 391 1998 59.
R.E. Smalley, Chem. Phys. Lett. 260 1996 471.
Ž
.
Ž
.
w x
9
R.T.K. Baker, P.S. Harris, R.B. Thomas, R.J. Waite, J. Catal.
w x
3
Ž
.
T.W. Odom, J.L. Huang, P. Kim, C.M. Lieber, Nature
30 1973 86.
Ž
.
Ž
.
w
w
x
Ž
.
London 391 1998 62.
10 S. Sawada, N. Hamada, Solid State Commun. 83 1992 917.
w x
4
x
C. Journet, W.K. Maser, P. Bernier, A. Loiseau, M. Lamy de
la Chapelle, S. Lefrant, P. Deniard, R. Lee, J.E. Fischer,
11 G.B. Adams, O.F. Sankey, J.B. Page, M. O’Keeffe, D.A.
Ž
.
Drabold, Science 256 1992 1792.
Ž
.
Ž
.
w
x
Ž
.
Nature London 388 1997 756.
12 V.N. L’nyanoy, Russ. Metall. 8 1979 91.