Reduction and reconstruction of Co3O4 nanocubes upon carbon deposition

Article abstract of DOI:10.1021/jp052551n

We describe a synthetic investigation on the formation of carbon nanofibers using a preshaped free-standing metal-oxide catalyst (single-crystal-cobalt oxide (Co₃O₄) nanocubes). In reacting with acetylene (C₂H₂) vapor, Co₃O₄ nanocubes are reduced and reconstructed into metallic cobalt. The resultant metal catalyst with a 2-fold symmetry leads to a bilateral base growth for carbon nanofibers. Our findings indicate that an understanding of catalyst-assisted chemical vapor deposition (CVD) mechanisms can be acquired, when the shape, size, and crystal orientation of pristine metal catalysts are made known to the CVD process. By tracing their evolutional changes in structure and composition, the shape-designed model catalysts may offer new opportunities for mechanistic investigations on the chemical reactivity of nanoparticles, general catalyst-assisted material synthesis, and metal intercalation chemistry. ? 2005 American Chemical Society.

Full text of DOI:10.1021/jp052551n

J. Phys. Chem. B 2005, 109, 17113-17119
17113
Reduction and Reconstruction of Co3O4 Nanocubes upon Carbon Deposition
Ji Feng and Hua Chun Zeng*
Department of Chemical and Biomolecular Engineering, Faculty of Engineering, National UniVersity of
Singapore, 10 Kent Ridge Crescent, Singapore 119260
ReceiVed: May 15, 2005; In Final Form: July 2, 2005
We describe a synthetic investigation on the formation of carbon nanofibers using a preshaped free-standing
metal-oxide catalyst (single-crystal cobalt oxide (Co ) nanocubes). In reacting with acetylene (C ) vapor,
Co nanocubes are reduced and reconstructed into metallic cobalt. The resultant metal catalyst with a 2-fold
O
3 4
2 2
H
3 4
O
symmetry leads to a bilateral base growth for carbon nanofibers. Our findings indicate that an understanding
of catalyst-assisted chemical vapor deposition (CVD) mechanisms can be acquired, when the shape, size, and
crystal orientation of pristine metal catalysts are made known to the CVD process. By tracing their evolutional
changes in structure and composition, the shape-designed model catalysts may offer new opportunities for
mechanistic investigations on the chemical reactivity of nanoparticles, general catalyst-assisted material
synthesis, and metal intercalation chemistry.
Introduction
The past decade has witnessed an explosive increase in
process in the unilateral growth of CNTs was demonstrated
successfully with in situ high-resolution transmission electron
13
microscopy (HRTEM) observation. In this connection, we
nanomaterial research, triggered by the discovery of carbon
notice that the metal-inclusion mechanism in the bilateral
growth of nanocarbons could be further pursued by taking
1
nanotubes (CNTs) in 1991. The graphene layers in CNTs are
parallel to the axis of the nanotube, which is different from the
advantage of the synthetic architecture of transition metal
2
cone-shaped carbon nanotubes known much earlier. Among
19-25
oxides.
Herein, we describe an investigation on the forma-
various studies on novel nanostructured materials, investigation
on CNTs is still taking the lead owing to their unique structural,
tion mechanism of carbon nanofibers (i.e., bi-cone-shaped
CNTs) with a preshaped free-standing model catalyst (Co3O4
nanocubes) concerning metal-oxide reduction, reconstruction,
and metal-carbon interactions under CVD conditions. Our
findings indicate that using shape-defined nanocatalysts may
provide new opportunities for mechanistic investigations on the
chemical reactivity of nanoparticles, general catalyst-assisted
material synthesis, and metal intercalation chemistry.
1
-5
electronic, and mechanical properties.
There have been a
number of mechanisms,6
-18
often in controversy, to describe
the nucleation and growth of carbon nanotubes/filaments/fibers
under catalytic chemical vapor deposition (CVD) conditions.
However, the geometrical structures and shapes of initial
catalysts used in many of these works were not defined and
quantified, which caused additional uncertainty for the mech-
6-18
anisms classified. Among many proposed mechanisms,
two
and (ii) tip
During the “base growth”, catalyst particles es-
Experimental Section
6-9
basic modes are attributed: (i) base growth
1
2-15
growth.
The preshaped catalyst, nanocubic Co O (in the spinel
3
4
sentially remain on the supporting substrate while CNTs grow
rising up from the catalyst base. In the “tip growth”, on the
other hand, catalyst particles are lifted off from the substrate
support, staying at the tip of the CNTs which are formed like
traces behind the catalyst particles. It has been conceived that
the two growth modes may bare some similarity in terms of
carbon diffusion/insertion on the metal catalyst surface. In both
cases, for example, the nanocarbons are grown only at one side
of the catalyst particles; that is, they are grown unilaterally. In
addition to these unilateral growths,12-15 nanocarbons can also
grow bilaterally at two opposite sides of metal particles,
resulting in metal inclusions in the nanocarbons.2,10,11 In all of
these cases, metal inclusions in either cylindrical graphene
phase), was prepared via a salt-assisted oxidation method
1
9,20
detailed in our previous studies.
Briefly, 1.2 g of NaOH
(Merck, >99.0%) was dissolved in 100.0 mL of deionized water
in a three-necked round-bottom flask, after which 120.0 g of
NaNO (Merck, >99.5%) was added. The three-necked flask,
3
mounted on top with a reflux condenser, was then immersed
into an oil bath at 105 ( 0.2 °C. A purified air stream (50 mL/
min, Soxal, O ) 21 ( 1%, H O < 2 vpm and hydrocarbons
2
2
< 5 vpm; vpm ) volume per million) was continuously bubbled
through this solution. After 30 min, 20.0 mL of 1.0 M Co-
(NO )‚6H O (Merck, >99.0%) was added dropwise within 1
3
2
min under stirring, and a blue precipitate (R-Co(OH) ) was
2
formed instantaneously. The blue slurry was kept in the oil bath
12-15
2
CNTs,
stacks and fibers
cone-shaped carbon bi-tubes, or flat graphitic planar
for further reactions with the same air stream. After the reactions,
the slurry was cooled naturally under ambient conditions to room
temperature and centrifuged at 3000 rpm for 20 min. The
obtained solid was redispersed in a HCl solution in order to
2
1
0,11
were actually dug out of the surface of
catalytic metal/alloy supports. Because the primitive forms of
metal particles before this “digging” are not known, process
history cannot be traced and related growth modes cannot be
assigned unambiguously. Very recently, the metal-inclusion
dissolve the unconverted solid precursors such as Co(OH) ,
II
II,III
19,20
Co (OH) (NO ) ‚mH O, and Co -hydrotalcite.
2
-x
3 x
2
The
mixture was centrifuged again, and the process was repeated
four times to ensure a complete removal of the precursor phases.
*
Corresponding author. E-mail: chezhc@nus.edu.sg.
1
0.1021/jp052551n CCC: $30.25 © 2005 American Chemical Society
Published on Web 08/23/2005

17114 J. Phys. Chem. B, Vol. 109, No. 36, 2005
Feng and Zeng
Figure 1. (A) TEM image of an as-prepared Co
3
O
4
nanocube and its SAED pattern (the large, wide arrows indicate the lateral elongation directions
spinel (cobalt cations are illustrated with purple
upon the C -CVD process). (B) A simplified representation of one-eighth of a unit cell of Co O
2
H
2
3 4
balls; oxygen anions are not shown) and a unit cell of fcc Co metal (cobalt atoms are depicted with purple balls). (C) Orientation of metallic Co
and the formation of conical-shaped carbon nanofibers along the 〈110〉 directions of cobalt. (D) Bilateral base growth mechanism with a symmetric
Co catalyst for carbon nanotubules: (1) upper base mode; (2) lower base mode.
A black paste product (i.e., Co3O4 nanocubes in 100% mor-
phological yield) was then resuspended in deionized water and
recentrifuged; this latter washing procedure was also repeated
four times. Chemical vapor deposition (CVD) synthesis of
metal-carbon-nanofiber composites was carried out in a
horizontal quartz-tube reactor (inner diameter of 7 mm) with
acetylene (C2H2) diluted by purified argon gas. The cubic
nanocatalyst Co3O4 was supported on glass wool and pressed
into a small packed bed. The catalyst was preheated in an electric
furnace to 400-500 °C at 10 °C/min and purged with an argon
stream. In most cases, the argon was fed as a background gas
at a constant feed rate of 40 mL/min, and the carbon source
gas acetylene was introduced at a variable rate of 5-10 mL/
min when the temperature reached a desired value. This CVD
process was kept going for 3-20 min, and then, the C2H2 stream
was shut down. The reactor was moved out of the heating zone
of the furnace and brought down to room temperature naturally
with the same flow of argon. The black powder product was
then stored in clean glass bottles for further material charac-
terization. In particular, the crystal phase was determined with
powder X-ray diffraction (XRD, Shimadzu XRD-6000, Cu KR
radiation). The structural and compositional information of the
product materials were obtained with transmission electron
microscopy and selected area electron diffraction (TEM/SAED,
JEM-2010F, 200 kV), high-resolution TEM and energy-
dispersive X-ray spectroscopy (HRTEM/EDX, JEM-300, 300
kV), and X-ray photoelectron spectroscopy (XPS, AXIS-HSi,
Kratos Analytical).
electron diffraction (SAED) and TEM-tilting experiments (Sup-
porting Information SI-1), it is known that each nanocube is
bordered with six crystallographic planes of {100} (Co3O4;
space group Fdιm; ao ) 8.084 Å, JCPDS file no. 43-1003). In
this work, the conversion of the Co O nanocubes to metallic
3
4
Co was carried out in a flowing stream of acetylene (C H )
2
2
upon the graphitization of the organic molecules (Figure 1B
and C). Considering the isotropic nature of the prepared Co O ,
3
4
a number of scenarios for carbon deposition can be proposed:
(i) on 6 equal crystal planes; (ii) at 8 equal corners; and (iii) on
12 equal edges of a cubic structure. Quite surprisingly, the
experimental results reveal that due to self-induced growth
anisotropy only one-dimensional carbon deposition proceeds,
as depicted in Figure 1C and D.
The specific volumes of atomic cobalt for Co3O4 and Co are
3
1
4.68 and 11.22 Å /atom, respectively. Because the volume
shrinkage from Co3O4 to Co is only moderate and similar crystal
symmetry is maintained, it is expected that the resultant Co
crystallites would preserve the original cubic shape of Co3O4
upon the volume contraction. Figure 2 displays a few repre-
sentative TEM images of cobalt-containing carbon nanofibers
grown in the above process, each with a black nodule (cobalt
catalyst) in the middle. Indeed, some of the cobalt centers are
still cubelike, as shown in Figure 3A, although the majority
shows an elongated morphology (Figure 2A-C). The energy-
dispersive X-ray analysis (EDX) affirms that cobalt and carbon
are present in these nanofibers (Supporting Information SI-2).
In this agreement, SAED investigation (Figure 2B and C) further
confirms that all of the Co3O4 nanocubes have been transformed
to metallic cobalt and the included cobalt nanoparticles are
single-crystalline. On the basis of observed diffraction spots of
the [110] and [111] zones, it is revealed unambiguously that
all resultant bi-cone-shaped cobalt nodules are elongated along
the 〈110〉 directions, which is in good agreement with an
Results and Discussion
A representative Co3O4 nanocube prepared from the above
procedure is shown in Figure 1A. In the present synthesis, the
mean length of cube edges was controlled at 86 ( 6 nm
(Supporting Information SI-1). On the basis of our selected area

Reduction and Reconstruction of Co3O4 Nanocubes
J. Phys. Chem. B, Vol. 109, No. 36, 2005 17115
3 4
Figure 2. (A-C) TEM images of carbon nanotubules formed with rhombic cobalt catalysts (converted from Co O nanocubes) in the center and
their respective SAED patterns of [111] and [110] zone spots (insets). (D) Lattice-resolved HRTEM image perpendicular to {111} planes (illustrated
with a colored model of the (111) surface; a surface unit cell of 3 × {422} is indicated with a dashed hexagon). (E) HRTEM image of d002 lattice
fringes (illustrated with a colored model of the (110) surface). (F) HRTEM image on the interfacial area between a conical Co crystal tip and grown
graphene layers surrounding it.
investigation on the formation of metallic bi-cones and their
related carbon bi-tubes. The high crystallinity of single-crystal
should also be attainable when multiple pods of metallic
catalysts are synthesized.
2
cobalt inclusions is also elucidated in our lattice-resolved
HRTEM images. For example, a 6-fold symmetry corresponding
to a coplanar atomic arrangement of {111} crystal planes of
face-centered cubic (fcc) cobalt can be observed in Figure 2D;
the interatomic distance is measured at 2.51 ( 0.02 Å (space
group Fmιm; ao ) 3.554 Å; JCPDS card no. 01-1259). It is not
The carbon nanofibers synthesized with this approach could
be as long as a few micrometers but usually are a few hundred
nanometers in length. Image statistics of over 300 cobalt-
containing carbon nanofibers shows that only less than 3% of
them have diameters greater than 200 nm, and a distribution
profile of the nanofibers with diameters smaller than 200 nm is
displayed in Figure 5A. It can be seen that, despite the long
tail toward the right side, diameters of more than two-thirds of
nanofibers fall between 30 and 70 nanometers (Supporting
Information SI-3). Hence, a direct link between the Co3O4
nanocubes and resultant Co in size, dimension, and crystal
orientation is exhibited with respect to both the volume
shrinkage and the 〈110〉 elongation. More importantly, this
observation reveals that a critical size exists for Co3O4
nanocubes to undergo the initial graphitization under a set of
experimental parameters (see also Figure 8F). The diameter of
the nanofiber is proportional to the dimension of the largest
cross section of cobalt particle, consistent with an observation
involving the surface diffusion of carbon on the active atomic
1
surprising to observe /3{422} forbidden reflections in the [111]
26
zone spots, since structural defects such as stacking faults along
the [111] axis are anticipated for our conversion process of
Co3O4 to fcc Co. Furthermore, the lattice fringes of d002 (1.78
(
0.03 Å) shown in Figure 2E are in clear association with the
110〉 orientations of the cobalt. Figure 2F also displays the
〈
interfacial structure between the cobalt catalyst and carbon
nanofiber, showing the formation of woven graphene layers
(
average interlayer distance ≈0.34 nm) parallel to the biconical
cobalt growth fronts. However, the graphene layers become
more parallel to the fiber axis when they are farther away from
the catalyst, as shown in Figure 3. Together with those of Figure
2
A-C, Figure 4 indicates that the orientation of the cobalt
1
3
catalyst is along the 〈110〉 directions of fcc cobalt exclusively
for all linear metal-carbon-nanofiber nanocomposites. The
droplet of cobalt confirms that the carbon deposition takes place
along the largest diameter part of cobalt catalyst; the growth
action also pushes the tip cobalt droplet away from the original
catalyst. Consistent with this biconical carbon insertion, two
steps, noting that many atomic steps exist on the reshaping
biconical cobalt catalyst for this diffusive growth of graphene
layers.
The above C2H2-CVD process was also investigated with the
XRD method (Figure 5B). The diffraction pattern of the sample
obtained after 3 min of reaction indicates the formation of a
carbon phase, with a reflection at around 2θ ) 26.3° (i.e., inter-
graphene-sheet distance 3.35 Å). The broad asymmetric peak
observed also reveals that the carbon phase was not well
integrated. It is also indicative that the precursor spinel oxide
had been completely converted to fcc elemental cobalt even
within the first 3 min of reaction. Though the lattice of metallic
cobalt commonly adopts the hexagonal close-packed (hcp)
structure, it has been reported that the fcc packing is a common
“
diffraction arcs” of graphene layers, 0002 and 000-2, are
observed in the SAED pattern of Figure 2B. The arc curves in
fact comprise numerous diffraction spots, registering different
orientations of the graphene layers with respect to the axis of a
nanofiber. The carbon nanofibers were grown bilaterally from
both sides of a central cobalt (Figure 1C), showing a 2-fold
symmetry for the overall composite structure. Because the
symmetric Co crystallite was a catalytic substrate tailor-made
in our carbon deposition, remarkably, the unique final product
morphology reveals a bilateral base-mode mechanism (Figure
2
7
lattice configuration for nanostructured cobalt. On the basis
of the above observations, it is clear that the pristine Co3O4
nanocubes were reduced to metallic cobalt while the hydrocar-
bon feedstock C2H2 was oxidized and decomposed into graphene
1
D). In principle, on the basis of the current findings, more
complex branching (such as trilateral) from the central metal

17116 J. Phys. Chem. B, Vol. 109, No. 36, 2005
Feng and Zeng
Figure 4. TEM images and SAED patterns of Co-containing carbon
nanotubules at different tilted angles: (A) tilted angle 0°; (B) [110]
zone diffraction spots from part A (refer to Figure 2C); (C) tilted angle
+
30°; (D) [111] zone diffraction spots of part C (refer to Figure 2B).
The white arrows indicate the cobalt droplets.
detected should be carbon, oxygen, and cobalt. Apart from the
wide-scan spectrum reported in Figure 6A, a narrow-scan X-ray
photoelectron spectroscopy analysis on the binding energies of
core-level photoelectrons C 1s, O 1s, and Co 2p can be found
2
8-30
in Figure 6B-D.
The XPS spectrum shows only a very
weak signal of the existence of the cobalt element (Figure 6D).
This confirms that the cobalt particles are largely embedded in
the carbon phase so that only a small fraction of cobalt is XPS-
detectable. In this sample, there are two subpeaks for the surface
carbon species (Figure 6B). The strongest peak at 284.5 eV can
be assigned to the graphene phase carbon as well as the
adventitious hydrocarbon, and the peak at 285.6 eV is assigned
2
-
to a small amount of CO3 anions from the atmospheric
2
8-30
CO2.
The peak of O 1s photoelectrons is very weak,
indicating that there is only little surface oxygen-containing
species (Figure 6C). There are three subpeaks at 533.9, 532.6,
and 531.4 eV, respectively. The first peak can be assigned to
the oxygen in adsorbed water molecules and the second to the
2
-
28-30
oxygen of the CO3 anions.
The last peak is commonly
known to arise from the oxygen in hydroxide ions, which
suggests the existence of cobalt hydroxide on the surface. This
is confirmed by the Co 2p photoelectron spectrum of the same
sample (Figure 6D). The spectrum is noisy but shows the typical
Figure 3. TEM images of Co-containing carbon nanotubules: (A)
overall structure; (B) area 1 of part A; (C) area 2 of part A.
carbon. This redox reaction is also verified by an observation
of the generation of colorless transparent condensate (H2O) on
the downstream side of the quartz-tube reactor, as explained in
the chemical equation
well-separated branches of 2p2/3 and 2p1/2 with a separation of
1-3
about 15 eV in binding energy.
The peak with a binding
energy of 778.6 eV and its corresponding 2p1/2 branch at 793.6
eV are assigned to elemental cobalt, which is the bulk species
of the Co element as suggested by XRD results. The 2p3/2 peak
with a binding energy of 780.9 eV and its 2p1/2 counterpart
indicate that there are cobalt ions associated with hydroxide
ions (Co-OH). The peaks at 783.1 and 798.1 eV can be
assigned to the Co cations in association with carbonate ions
(Co-CO3). The existence of Co-OH and Co-CO3 species is
due to the exposure of the sample to air (containing CO2 and
Co O (s) + 4C H (g) ) 3Co(s, fcc) + 8C(s, graphene) +
3
4
2
2
4
H O(g) (1)
2
The XRD pattern of the sample obtained after 10 min of reaction
provides similar findings, except for the peak narrowing and
stronger reflections. In contrast to the pronounced XRD peaks
of metallic cobalt observed in the sample after 10 min of CVD
treatment, the X-ray photoelectron spectroscopy (XPS) study
indicates a very weak signal for the existence of surface cobalt.
According to the species present in this system, the elements
2
8-30
H2O) during sample handling after the synthesis.
As the
XPS technique is surface sensitive, this observation reveals that
the cobalt nanoparticles are in fact embedded underneath the
graphene carbon so that only a small fraction of cobalt was

Reduction and Reconstruction of Co3O4 Nanocubes
J. Phys. Chem. B, Vol. 109, No. 36, 2005 17117
Figure 5. (A) Diameter distribution of the carbon nanofibers (the inset
gives representative carbon tubules together with their central Co
crystallites used in this statistic study; scale bar 50 nm). (B) XRD
patterns of Co
and 10 min of CVD reactions at 450 °C (C
3
O
4
nanocubes and the formed fcc Co crystallites after 3
flow rate 10 mL/min).
2
H
2
The asterisks in the XRD patterns (3 and 10 min) denote the graphene
carbon phase formed on the two sides of a single-crystal Co center.
detectable by XPS. Indeed, our TEM observation reveals that
several layers of graphene were also deposited on the base of
the biconical cobalt catalyst, where the entrance points of carbon
are supposed to be. The TEM images in Figure 7 provide some
detailed views on graphene structures formed on the utmost
surface of the cobalt catalyst. A number of graphene layers were
still formed on the central part of the cobalt catalyst (corre-
sponding to the diameter of the carbon nanofiber). Although
the X-ray of XPS can penetrate through these graphene layers
to reach underneath cobalt, its (X-ray) intensity and the resultant
intensity of photoelectrons of Co 2p had been significantly
attenuated.
Figure 6. X-ray photoelectron spectra of the carbon composite material
synthesized in this approach: (A) a wide scan; (B) 1s photoelectrons
of carbon; (C) 1s photoelectrons of oxygen; (D) 2p photoelectrons of
cobalt. The sample was prepared at 500 °C (10 min; C H flow rate 10
2 2
mL/min).
To investigate temperature effect, the as-prepared metal-
carbon composite sample was heated inside an argon atmosphere
as shown in Figure 8A and B. The lattice fringes of the graphene
phase clearly show that the conical graphene layer stacks are
now all changed to continuous parallel multiwalled carbon
nanotubes (Figure 8A), during which the original discontinuous
graphene layers were repaired by carbon diffusion and recrys-
tallization. In Figure 8B, the metal nanoparticle is elongated
along the internal cavity of the normal multiwalled carbon
nanotubes. In particular, the two ends of the biconical cobalt
(gas flow rate 10 mL/min) with a quartz-tubular furnace at 10
°
C/min to 900 °C, and the temperature was held constant for 1
h. After this heat treatment, the sample was cooled to room
temperature under ambient conditions within the same argon
atmosphere. Interestingly, both the carbon and metal particles
in the carbon nanofibers underwent significant reconstructions,

17118 J. Phys. Chem. B, Vol. 109, No. 36, 2005
Feng and Zeng
Figure 7. TEM images of the graphene-layer-covered Co nanocrystallite center. The arrows indicate the graphene layers on the Co nanocrystallite
center.
Figure 8. (A and B) Annealing of as-prepared cobalt-containing nanofibers (in an argon atmosphere, 900 °C, 1 h). (C) An observed mechanism
for the formation of parallel graphene layers (initial Co/Co
Competitive reduction for different sizes of Co nanocubes at a lower flow rate of C
formed on the {100} surfaces of large Co nanocubes (C flow rate 5 mL/min; 10 min at 400 °C).
O
3 4
cube in black, fcc Co in purple, and graphene layers in black lines). (D and E)
3
O
4
2 2
H
(5 mL/min; 10 min at 450 °C). (F) Graphene layers
3
O
4
2 2
H
are no longer pointed but rounded with the atomically compact
111} facets. Once again, the SAED investigation (not shown)
indicates that the confined cobalt nanorod is still single-
crystalline, maintaining the same propagating orientations (i.e.,
nanocubes were readily covered with a few graphene layers
(Figure 8F), although they were only partially reduced (con-
firmed with XRD). The reactions in this low-temperature
environment in fact shed light on the initial state of Co3O4
nanocube reduction and related conversion to metal cobalt. Upon
the partial reduction in C2H2 (i.e., initial nucleation of graphene
layers), indeed, the Co/Co3O4 solid mixture still keeps an overall
cubic structure (Figure 8D-F). With the more favorable reacting
conditions (higher flow rate of C2H2 and/or higher temperatures),
all graphene-covered cubical solid mixtures would then be
reshaped into the biconical cobalt core. As have been seen in
Figures 2 and 3, the graphene layers wrapped on a reduced Co/
Co3O4 nanocube would exert a force onto it due to the strong
tendency of tubular formation (Figure 8C), inducing the
elongation along the 〈110〉 directions of cobalt and subsequent
growth of nanofibers.
{
〈
110〉 directions) as those in unannealed samples. Figure 8C
gives a schematic illustration for the above process. The above
investigation also suggests that extra precautions must be taken
when interpreting a CVD mechanism, since the original shape
of the metal catalyst and its actual growth interface might not
be necessarily preserved at a reported nominal temperature due
to additional reconstructions.
The reactions between Co3O4 nanocubes and C2H2 (flow rate
10 mL/min) were quite fast (Figure 5B). By decreasing the flow
rate of C2H2 down to 5 mL/min, nevertheless, the reduction
rate of the spinel oxide could be greatly lowered at the same
temperature of 450 °C (Figure 8D and E). Under these less
kinetically favorable conditions, some carbon nanofibers con-
taining small cobalt nanocrystallites could still be formed but
larger particles would just retain their pristine cubic morphology,
though they had been partially etched (Co3O4 to Co) by the
gaseous reducing agent acetylene. Compared to smaller particles,
the large Co3O4 nanocubes are less reactive and were not able
to compete with their smaller counterparts for the limited C2H2
supply. At as low as 400 °C, however, the large Co3O4
Conclusions
In summary, we have demonstrated that the original shape,
size, and symmetry (crystal orientation) of metal-oxide catalysts
play paramount roles in determining a nominal formation mode
of carbon nanofibers under hydrocarbon-CVD conditions. With
our preshaped discrete metal-oxide nanocubes, a 2-fold sym-

Reduction and Reconstruction of Co3O4 Nanocubes
J. Phys. Chem. B, Vol. 109, No. 36, 2005 17119
metrical reconstructing metal catalyst is attained throughout the
course of carbon deposition, revealing a bilateral base growth
mode. In the same analogy, the synthetic approach with
preshaped model catalysts may also be extendable to other
mechanistic investigations on the reactivity of nanoparticles,
general catalyst-assisted material synthesis, and intercalation
chemistry. In principle, other metal-carbon nanocomposites
with a complex branching morphology should also be attainable
in the future when multiple-branched metallic catalysts can be
prepared.
(9) Hata, K.; Futaba, D. N.; Mizuno, K.; Namai, T.; Yumura, M.; Iijima,
S. Science 2004, 306, 1362-1364.
10) Rodriguez, N. M.; Chambers, A.; Baker, R. T. K. Langmuir 1995,
1, 3862-3866.
11) Chambers, A.; Rodriguez, N. M.; Baker, R. T. K. J. Phys. Chem.
1996, 100, 4229-4236.
(12) Okai, M.; Muneyoshi, T.; Yaguchi, T.; Sasaki, S. Appl. Phys. Lett.
000 77, 3468-3470.
13) Helveg, S.; L o´ pez-Cartes, C.; Sehested, J.; Hansen, P. L.; Clausen,
(
1
(
2
(
B. S.; Rostrup-Nielsen, J. R.; Abild-Pedersen, F.; Nørskov, J. K. Nature
2004, 427, 426-429.
(14) Ngo, Q.; Cruden, B. A.; Cassell, A. M.; Sims, G.; Meyyappan,
M.; Li, J.; Yang, C. Y. Nano Lett. 2004, 4, 2403-2407.
(
15) Tanaka, A.; Yoon, S.-H.; Mochida, I. Carbon 2004, 42, 1291-
Acknowledgment. The authors gratefully acknowledge the
financial support of the Ministry of Education, Singapore
1
298.
(
16) Andriotis, A. N.; Menon, M.; Froudakis, G. Phys. ReV. Lett. 2000,
8
5, 3193-3196.
Supporting Information Available: TEM images of as-
prepared Co3O4 nanocubes, EDX spectrum of Co-containing
nanocarbons, and the size statistics of nanocubes. This material
is available free of charge via the Internet at http://pubs.acs.org.
(
17) Mann, D. J.; Halls, M. D.; Hase, W. L. J. Phys. Chem. B 2002,
106, 12418-12425.
18) Tsai, C. L.; Hsu, J. H.; Chen, C. F. Appl. Phys. Lett. 2003, 82,
337-4339.
(
4
(
(
(
19) Xu, R.; Zeng, H. C. J. Phys. Chem. B 2003, 107, 926-930.
20) Feng, J.; Zeng, H. C. Chem. Mater. 2003, 15, 2829-2835.
21) Lou, X. W.; Zeng, H. C. J. Am. Chem. Soc. 2003, 125, 2697-
References and Notes
2
704.
(
1) (a) Iijima, S. Nature 1991, 354, 56-58. (b) Ebbesen, T. W.; Ajayan,
P. M. Nature 1992, 358, 220-222.
2) Audier, M.; Oberlin, A.; Coulon, M. J. Cryst. Growth 1982, 57,
(22) Liu, B.; Zeng, H. C. J. Am. Chem. Soc. 2004, 126, 8124-8125.
(
(23) Yang, H. G.; Zeng, H. C. Angew. Chem., Int. Ed. 2004, 43, 5930-
5933.
(24) Liu, B.; Zeng, H. C. J. Am. Chem. Soc. 2004, 126, 16744-16746.
5
24-534.
(3) Charlier, J.-C.; De Vita, A.; Blase, X.; Car, R. Science 1997, 275,
6
47-649.
(
(
25) Yang, H. G.; Zeng, H. C. J. Am. Chem. Soc. 2005, 127, 270-278.
26) Germain, V.; Li, J.; Ingert, D.; Wang, Z. L.; Pileni, M. P. J. Phys.
(
4) Odom, T. W.; Huang, J.-L.; Kim, P.; Lieber, C. M. J. Phys. Chem.
B 2000, 104, 2794-2809.
Chem. B 2003, 107, 8717-8720.
27) Dinega, D. P.; Bawendi, M. G. Angew. Chem., Int. Ed. 1999, 38,
788-1791.
28) Sampanthar, J. T.; Zeng, H. C. Chem. Mater. 2001, 13, 4722-
730.
(29) Xu, R.; Zeng, H. C. Chem. Mater. 2003, 15, 2040-2048.
(30) Xu, R.; Zeng, H. C. Langmuir 2004, 20, 9780-9790.
(
5) Zheng, Q.; Jiang, Q. Phys. ReV. Lett. 2002, 88, 045503.
(
(6) Fan, S.; Chapline, M. G.; Franklin, N. R.; Tombler, T. W.; Cassell,
1
4
A. M.; Dai, H. Science 1999, 283, 512-514.
7) Gavillet, J.; Loiseau, A.; Journet, C.; Willaime, F.; Ducastelle, F.;
Charlier, J.-C. Phys. ReV. Lett. 2001, 87, 275504.
8) Gavillet, J.; Loiseau, A.; Ducastelle, F.; Thair, S.; Bernier, P.;
St e´ phan, O.; Thibault, J.; Charlier, J.-C. Carbon 2002, 40, 1649-1663.
(
(
(