Study of cobalt-filled carbon nanoflasks

Article abstract of DOI:10.1021/jp010083l

A new form of fullerene-type carbon, named carbon nanoflasks, is described in this paper, using Co(CO)₃NO, a special precursor, in their synthesis. Upon its decomposition, the Co(CO)₃NO is not only a source of carbon, but also gives rise to fcc cobalt particles, whose size can grow from several nanometers to hundreds of nm in the course of the decomposition. After a careful purification process, the percentage of cobalt-filled carbon flasks can be as high as 30%. The width of the flask tube-necks ranges from 50 nm to over 100 nm, while the body of the flask ranges from 100 nm to 500 nm, and in some cases reaches over 1 ??m. HRTEM reveals that the graphitic layers of the flask walls are usually over 100 nm thick, and are much thicker than the flask cap. After an acid treatment of the sample, opened and empty carbon flasks can be easily obtained. XRD, TEM, SEM, XPS, AFM, and SQUID measurements were employed in the characterization of the flasks. An explanation of the formation of the carbon nanoflasks is proposed in this paper.

Full text of DOI:10.1021/jp010083l

7606
J. Phys. Chem. B 2001, 105, 7606-7611
ARTICLES
Study of Cobalt-Filled Carbon Nanoflasks
Suwen Liu,†, | Seth Boeshore,‡ A. Fernandez,§ M. J. Sayague´s,§ J. E. Fischer,‡ and
Aharon Gedanken*,†
Department of Chemistry, Bar-Ilan UniVersity, Ramat-Gan 52900, Israel, Laboratory for Research on the
Structure of Matter, UniVersity of PennsylVania, Instituto de Ciensia de Materiales de SeVilla and
Departemento Quimica Inorganica, Centro de InVestigaciones Cientificas Isla de la Cartuja, Spain, and
Department of Applied Chemistry, UniVersity of Science and Technology of China, Hefei, Anhui 230026,
People’s Republic of China
ReceiVed: January 7, 2001; In Final Form: May 16, 2001
A new form of fullerene-type carbon, named carbon nanoflasks, is described in this paper, using Co(CO)3NO,
a special precursor, in their synthesis. Upon its decomposition, the Co(CO)3NO is not only a source of carbon,
but also gives rise to fcc cobalt particles, whose size can grow from several nanometers to hundreds of nm
in the course of the decomposition. After a careful purification process, the percentage of cobalt-filled carbon
flasks can be as high as 30%. The width of the flask tube-necks ranges from 50 nm to over 100 nm, while
the body of the flask ranges from 100 nm to 500 nm, and in some cases reaches over 1 µm. HRTEM reveals
that the graphitic layers of the flask walls are usually over 100 nm thick, and are much thicker than the flask
cap. After an acid treatment of the sample, opened and empty carbon flasks can be easily obtained. XRD,
TEM, SEM, XPS, AFM, and SQUID measurements were employed in the characterization of the flasks. An
explanation of the formation of the carbon nanoflasks is proposed in this paper.
Introduction
however, were short, having only a 200 nm length. Organo-
metallic precursors were successfully used as sources for the
Phenomenal progress has been made in the science of carbon
nanotubes (CNT) since the publication of Iijima’s landmark
paper in 1991.1 A variety of novel structures with trigonally
bonded carbons have been discovered: single-walled nanotubes2
(SWCNT), nested spheroidal shell (onions),3 giant fullerene
shells,4 interconnected fullerene-like cages,5 cross-linked graphic
cages,6 flattened nanotubes, and flattened carbon nanoshells.7,8
Strong interest in these materials has been aroused in the field
of nanotechnology9 when it became possible to insert different
kinds of materials with potentially interesting properties into
the core volume of the CNT or carbon nanoparticles.
carbon and the metal, and yielded carbon nanotubes; some of
them filled with materials.20-22
In a previous paper we reported the existence of multishelled,
flask-shaped carbon nanostructures whose overall geometry
differs from that of carbon nanotubes or carbon nanoparticles.
We called these carbon vessels “carbon nanoflasks”.24 The
carbon nanoflasks were described as a new form of fullerene-
type carbon that gave a fascinating image.25 To better understand
the properties of the carbon nanoflasks, they have been purified
carefully and repeatedly. In the purified products, the percentage
of the cobalt-filled nanoflasks of all the observed TEM structures
is now over 30%, which is much higher than the 10-15% in
the original paper. Some new results on these cobalt-filled
carbon nanoflasks (CoCNF) are reported in the current paper.
Different methods of synthesis have been attempted and
various encapsulated materials have been tested,10 in particular
those involving magnetic or ferromagnetic materials.
Cobalt-filled carbon nanoparticles or nanotubes were recently
prepared and studied by several groups.11-19 Guerret-Plecourt
and co-workers reported the preparation of cobalt nanowires
encapsulated in nanotubes.11 They used the arc-discharge method
for their fabrication. The same group later demonstrated that a
sulfur impurity (of approximately 0.25%), caused the filling of
nanotubes with various materials.19 Their cobalt nanowires,
Experimental Section
The synthesis is carried out in a 2 mL closed vessel cell,
which is assembled from a stainless steel Swagelok parts. A 3/8
in. union part was capped from one side by a standard plug.
For this synthesis, 400 mg of magnesium powder and 700 mg
of Co(CO)3NO were placed in a cell at room temperature. The
vessel is then immediately closed tightly because Co(CO)3NO
is an air-sensitive material, and is heated at 900 °C for 3 h. The
reaction takes place at the autogenic pressure of the precursors.
After cooling to room temperature, we do not detect a release
of pressure from the vessel. The products are treated with 50
mL of 8 M HCl at 70 °C for 1 h, and then left overnight in the
* Corresponding author. E-mail: gedanken@mail.biu.ac.il. Phone: 0972-
3-5318315. FAX: 0972-3-5351250.
† Department of Chemistry.
‡ Laboratory for Research on the Structure of Matter.
§ Instituto de Ciensia de Materiales de Sevilla and Departemento Quimica
Inorganica.
| Department of Applied Chemistry.
10.1021/jp010083l CCC: $20.00 © 2001 American Chemical Society
Published on Web 07/19/2001

Cobalt-Filled Carbon Nanoflasks
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7607
Figure 1. X-ray diffraction pattern of the sample: (a) as-prepared sample; (b) after treatment with concentrated HCl.
Figure 2. XRD data for a sample after purification and in the air-
opened condition for three months. The peaks at 44.0695°, 51.1958°,
and 75.564° are due to fcc cobalt. The peaks at 41.515° and 47.16° are
caused by hcp cobalt. The peak at 26.3895° is graphite.
acid at room temperature. Finally, a black, soot-like product is
obtained after filtering and washing processes.
This product contained unfilled nanotubes, amorphous carbon,
and cobalt-filled carbon nanoparticles.23 To purify this blend,
the material is placed in a vial filled with ethanol and is
sonicated for 30 min. This procedure is necessary to untangle
the unpurified material. The sonicated sample is placed in a
strong magnetic field using an SmCo5 magnet. This is done
because the unpurified material responded to a magnetic field;
it is assumed that amorphous carbon and unfilled nanotubes, if
untangled, would not be attracted to a magnet. In the presence
of the field, the magnetic Co-filled materials aggregate along
field lines. These materials are removed from the vial and placed
in a second vial filled with ethanol. The procedure of sonication
followed by magnetic separation is repeated on the second vial.
This process yielded the purified materials that were subjected
to X-ray diffraction (XRD), high-resolution electron microscopy
(HREM), scanning electron microscopy (SEM), atomic force
microscopy (AFM), X-ray photoelectron spectroscopy (XPS),
magnetic force microscopy (MFM), and superconducting quan-
tum interference device (SQUID).
Figure 3. (a) SEM image of a single CoCNF resting on top of an
assortment of other CoCNF and unfilled nanotubes. Note that the cobalt
core (lighter gray) is visible within the carbon shell. (b) SEM image
of bulk material at 800x magnification. The CoCNF can be seen resting
on top of the amorphous carbon.
Results and Discussion
1. X-ray Diffraction and Chemical Reactions. The as-
prepared powder is characterized by X-ray powder diffraction

7608 J. Phys. Chem. B, Vol. 105, No. 32, 2001
Liu et al.
as described above, fcc cobalt patterns became the dominant
peaks (See Figure 2), which are much higher than those in
Figure 1b. This reveals that a large percentage of cobalt is
encapsulated in the carbon nanoflasks after the careful purifica-
tion process. The result, shown in Figure 2, was measured after
the sample was left in the open air for three months at room
temperature. Only very weak hcp cobalt diffraction peaks are
detected. We consider the walls of the nanoflasks and nanoubes
as hindering the conversion of the fcc Co to the stable hexagonal
phase. It further proved that fcc structured cobalt, which exists
only in high temperature (above 450 °C), can exist and be stable
inside the carbon nanoflasks stably.
On the basis of the XRD patterns, the following mechanism
is proposed. In the first step, Co(CO)3NO is decomposed into
Co, CO, and NO. Subsequently, CO reacts with Mg to yield
MgO and carbon. The reactions are as follows:
Figure 4. MFM image of four carbon nanoflasks.
Co(CO)3NO(g) f Co(s) + 3CO(g) + NO(g)
CO(g) + Mg(l) f MgO(s) + C(s)
(1)
(2)
(Figure 1a), and the main peaks are identified as those of MgO
(JCPDS 45-946) and fcc metallic cobalt (JCPDS 15-806). After
dissolving the powder in concentrated HCl (8 M), MgO and
most of metallic cobalt features disappeared. Thereby, the
graphite peak became the dominant peak (Figure 1b), while the
fcc cobalt patterns reveal weaker diffraction peaks (hcp cobalt
was not found). However, after carefully purifying the products
During carbon formation, Co plays the role of a catalyst for the
creation of nanotubes. At the same time, part of the metallic
Co produced by the decomposition of the precursor is encap-
sulated in the carbon shells. Briefly, these Co(CO)3NO mol-
ecules, on decomposition, act not only as a source of carbon,
Figure 5. (a) An individual carbon nanoflasks filled with Co. The inset is a close-up of the top part of this flask. Scale bar 200 nm. (b) Another
individual carbon nanoflasks filled with Co, there is only one cap in the top. (c) A HRTEM image of the cap. It shows that there are about 20 layers
of graphite in the cap. (d) One flask with only an inner cap.

Cobalt-Filled Carbon Nanoflasks
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7609
but also give rise to small metal clusters or particles which
function act as catalysts for the formation of the carbon
nanotubes or nanoflasks.
2. SEM, AFM, and XPS. JEOL 6300F and 6400 scanning
electron microscopes were used for the observation of bulk
samples and individual carbon nanoflask. SEM images reveal
that the sizes and shapes of the carbon materials are quite
different from the usual carbon nanotubes. In fact, the only
feature that all the carbon materials have in common is an
enlarged head, making the tubes appear similar to baby tadpoles
(Figure 3a). It is clear in this picture that the cobalt core (lighter
gray) is visible within the carbon shell. XRD results reveal that
these nanovessels were filled with fcc cobalt. The nanoflasks
have an ovoid or a spherical bulbous base with a “plus end”
from which the nanotube grows, lengthening one-dimensionally.
In some cases, the cobalt was only in the bottom of the flask
and the tube-neck is hollow. In others, the cobalt core filled
the entire length of the nanoflask. Figure 3b shows a large
amount of flask-shaped carbon in the products. The percentage
of flask-shaped materials is much higher than that in our
previous report.24 We roughly estimate that the flask-shaped
materials are over 30% of the product. Obviously, the increase
should be attributed to the better purifications methods described
in the Experimental Section. X-ray spectroscopy at 15 keV
centered on an individual flask yielded the atomic percentages
of 83.45 and 16.57 for carbon and cobalt, respectively.
The width of the tube-necks of the flasks ranged from 50
nm to over 100 nm. The body of the flask is often more than
twice the width of the tube, but in some cases it reached over
1 µm in size. It is common for the cobalt core to end and then
resume farther down along the tube, as seen in Figure 3a.
Usually, when a cobalt core is present the tube is very straight.
When the core ends, the remaining length sometimes becomes
curved.
Atomic force microscopy (AFM) was used to topographically
map the cobalt-filled nanoflasks. Magnetic force microscopy
(MFM) is also used in an attempt to find the magnetic response
to the domain structure of the cobalt core. This procedure is
similar to AFM, except that a magnetic tip is used to probe the
sample. However, the MFM measurements are not so helpful
for such CoCNFs (cobalt carbon nanoflasks). The response of
the products to the magnetic tip is not very apparent. Only after
much data manipulation was it possible to see the response.
Figure 4 is a topographical image obtained from MFM
measurements. The resolution of the image is decreased
compared to the AFM image because of a thicker magnetic tip
was used.
Figure 6. (a) An open and empty nanoflask was formed after a
treatment of 8 M HCl. A cobalt-filled carbon nanoflask. The part shown
by an arrow was empty. (b) Scale HREM image of the arrow marked
site in Figure 6a.
is only an outer cap covering the flask. The HREM image
revealed that the cap was composed of about 20 layers of
graphite (Figure 5c). In addition, we also found a small amount
of flasks with only inner caps. Figure 5d shows such a cap.
Usually, the thickness of the cap is much thinner than the walls
of the flask body. This result implied that the cap could be easier
to break or remove than the wall of the flasks. Follow-up
experiments and results confirmed that our above prediction was
correct.
We previously reported that when the as-prepared sample was
further treated with an aqueous solution of H2SO4 and H2O2
under ultrasonic irradiation for 2 h, some of flask caps were
removed.24 The bodies of the nanoflasks, however, were
unaffected. This process also caused the dissolution of the cobalt
and the flasks became empty. This result revealed that the
nanoflasks sustaining the sonication, remained intact and are
firm nanovessels. In fact, we have also found a very small
amount of empty nanoflasks in the as-prepared products (treated
only by 8 M HCl as description in the Experimental Section).
Figure 6a shows two empty flasks in the as-prepared sample,
one was opened at the cap; surprisingly, the other was opened
X-ray photoelectron spectroscopy (XPS) was employed with
an AXIS, HIS, 160, ULTRA apparatus. The XPS spectrum
detected only pure carbon (no Co signal). This result seems
not consistent with both the SEM and XRD results. The obvious
reason for the absence of the cobalt spectrum is due to the fact
the XPS monitors mostly the surface area. The outer cobalt layer
can be found as deep as several nanometers from the surface,
and therefore, it is not observed in the XPS spectrum.
3. A Close-up of Carbon Nanoflasks and Removing Their
Caps. Figure 5a shows an individual carbon nanoflask, which
is almost fully filled with Co. Only a small part near the top is
empty. The inset depicts a close-up of the top of this flask. It
is found that two caps cover this nanoflask. One is an inner
cap, which is on the end of nanowires and covers the tip tightly.
Only some of the nanoflasks have inner caps. The second is an
outer cap that covers the tube end at the top of the flask. Figure
5b shows an imaging of the top part of a flask. Obviously, there

7610 J. Phys. Chem. B, Vol. 105, No. 32, 2001
Liu et al.
Figure 7. Hysteresis loop for the magnetization of cobalt, taken at 50 K. Best-fit line created using data points. Note the magnetization at zero
field; it is almost zero.
Figure 8. Hysteresis loop for the magnetization of cobalt, taken at 50 K. It has been expanded to show the low-field response of the CoCNF.
in the neck. Most of empty flasks were opened at the cap. One
of the reasons is the cap is much thinner than the flask wall.
Another reason is that the strain at the cap and the presence of
pentagons help to initiate an oxidation process. The degree of
curvature, e.g., the strain, between the caps and body of the
nanoflasks is different. So, although pentagons might play a
role, strain must be the key factor in the onset of oxidation at
the top of the nanoflasks.
Figure 6a shows an interesting individual carbon nanoflask,
with filled cobalt in the both of bottom and tube-neck of the
flask, while there is no Co in the connection part between the
bottom and the tube-neck. The wall shown by an arrow is
observed by HREM. Figure 6b is the HREM image of the
connection between the flask body and the tube-neck. There
are about 100 layers of graphite in the area. As can be seen in
Figure 6a, the wall thickness in the lowest bottom of this flask
is much thicker than that in other places. It should be pointed
out that empty nanoflasks have never been found in any of the
crude products (before HCl treatment). The implication of this
result is that the as-prepared carbon nanoflasks are always filled
with cobalt.
4. Superconducting Quantum Interference Device Mea-
surements. One of the main goals of this research is learning
more about the cobalt core. It has been theorized that cobalt in
its fcc phase becomes nonmagnetic.26 Because MFM was unable
to yield viable data for the magnetization of cobalt, another
method was needed. It is assumed that the MFM tip is not
sensitive enough to detect the magnetic force. The magnetic
measurements of the cobalt core were conducted using a SQUID
magnetometer. The sample cannot be free to move within the
SQUID chamber. Because of this, a pellet of Co-filled carbon
nanoflasks and paraffin is made by mixing approximately 3 mg
of the sample and 5 mg of paraffin. The data are acquired for
the paraffin pellet of unaligned, unpurified CoCNT.

Cobalt-Filled Carbon Nanoflasks
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7611
Six hysteresis loops were obtained; one was measured for
temperatures ranging from 50 to 300 K at 50 K intervals. The
results of the SQUID measurements clearly show that the sample
is being magnetized (Figure 7). The loops for each temperature
were almost identically similar.
After plotting the data for the hysteresis loops, it became
evident that the area around zero field should be expanded to
find the magnetization at zero field (Figure 8). The magnetiza-
tion was found to be slightly less than 0.001 emu and practically
independent of temperature. The interpretation of this result is
that the slight magnetization at zero field is caused by the
presence of ferromagnetic hcp cobalt particles. The bulk of the
magnetic material, however, is fcc cobalt, which behaves like
a paramagnetic material.
CoCNFs are removed by acid treatment, the empty carbon
nanoflasks can be recovered. SQUID measurements clearly
show that the fcc cobalt within the tubes is paramagnetic. Also,
many of the CoCNFs have cores that end far before the length
of the nanotube. This core space is much larger than normal
carbon nanotubes and could potentially be used for the storage
of materials.
Acknowledgment. Dr. Suwen Liu thanks The Fred and
Barbara Kort Sino-Israel Postdoctoral Fellowships Foundation
for financial support and the China Scholarship Council for their
support. Profs. J. E. Fischer and A. Gedanken gratefully
acknowledge the receipt of a NEDO International Joint Research
Grant. We thank Dr. Shifra Hochberg for editorial assistance.
5. Growth Mechanism of the Carbon Nanoflasks. There
is widespread agreement about how a nanotube lengthens, once
it has nucleated from a metal particle. It seems to be generally
accepted that metallic particles act as catalysts for the graphi-
tization of carbon present in the vapor phase. The carbon then
forms a hemispherical graphene cap on the metal particle, and
the nanotubes grow from such a graphene cap.27
References and Notes
(1) Iijima, S. Nature 1991, 354, 56.
(2) Iijima, S.; Ichihashi, T. Nature 1993, 363, 603.
(3) Ugarte, D. Nature 1992, 359, 707.
(4) Bourgeois, L. N.; Bursill, L. A. Philos. Mag. A 1997, 76, 753.
(5) Iijima, S.; Wakabayashi, T.; Achiba, Y. J. Phys. Chem. 1996, 100,
5839.
(6) Amaratunga, G. A. J.; Chhowalla, M.; Kiely, C. J.; Alexandrou,
I.; Aharonow, R.; Devenish, R. M. Nature 1996, 383, 321.
(7) Chopra, N. G.; Benedict, L. X.; Crespi, V. H.; Cohen, M. L.; Louie,
S. G.; Zettl, A. Nature 1995, 377, 135.
(8) Bourgeois, L. N.; Bursill, L. A. Chem. Phys. Lett. 1997, 277, 571.
(9) Ruoff, R. S. Nature 1994, 372, 731.
(10) Sloan, J.; Cook, J.; Green, M. L. H.; Hutchison, J. L.; Tenne, R. J.
Mater. Chem. 1997, 7, 1089.
(11) Guerret-Plecourt, C.; Le Bouar, Y.; Loiseau, A.; Pascard, H. Nature
1994, 372, 761.
(12) Seraphin, S.; Zhou, D.; Jiao, J. J. Appl. Phys. 1996, 80, 2097.
(13) Lafdi, K.; Chin, A.; Ali, N.; Despres, J. F. J Appl. Phys. 1996, 79,
6007.
(14) Jiao, J.; Seraphin, S.; Wang, X.; Whiters, J. C. J Appl. Phys. 1996,
80, 103.
The results presented here show that Co(CO)3NO is a quite
special precursor. These molecules, on decomposition, act not
only as a source of carbon but also give rise to fcc cobalt
particles whose size can grow from several nanometers to
hundreds of nm in the course of the Co(CO)3NO decomposition.
These cobalt particles, serve as nucleation centers for the carbon,
which is formed by the reduction of the CO residue by the Mg
atoms under the reaction conditions. When this process occurs,
the growth proceeds by the addition of carbon as atoms, chains
and rings.16 The formation of the filled nanoflasks happens
because the cobalt clusters coagulate very fast at the beginning.
However, midway through the process the coagulation slows
down and the carbon wraps most of the base. At this moment,
only a slight edge of the cobalt cluster that is not wrapped by
carbon protrudes from the coated particle, so that when cobalt
is further deposited, it causes a wire to grow in one-dimension.
The carbon follows the wire, and the tube continues growing
onto the base along the cylindrical axis. Furthermore, forming
a nanotube-neck and, in fact, also forming a multiwalled carbon
nanotube seem to require a suitable diameter of catalytic
particles. Only when the diameter of the Co core is smaller than
a certain value can the nanotube-necks be formed. Usually, the
diameter of cobalt nanowires in the necks is smaller than 100
nm.
(15) Kiang, C-H.; Goddard, W. A.; Beyers, R.; Salem, J. R.; Bethune,
D. S. J. Phys. Chem. 1994, 98, 6612.
(16) Harris, P. J. F.; Tsang, S. C. Chem. Phys. Lett. 1998, 293, 53.
(17) Rao, C. N. R.; Govindaraj, A.; Sen, R.; Satishkumar, B. C. Mater.
Res InnoVat. 1998, 2, 128.
(18) Terrones, M.; Grobert, N.; Zhang, J. P.; Terrones, H.; Olivares, J.;
Hsu, W. K.; Hare, J. P.; Cheetham, A. K.; Kroto, H. W.; Walton, D. R. M.
Chem. Phys. Lett. 1998, 285, 299.
(19) Demoncy, N.; Stephan, O.; Brun, N.; Colliex, C.; Loiseau, A.;
Pascard, H. Eur. Phys. J. B 1998, 4, 147.
(20) Rao, C. N. R.; Govindaraj, A.; Sen, R.; Satishkumar, B. C. Mater.
Res. InnoVations 1998, 2, 128.
(21) Sen, R.; Govindaraj, A.; Rao, C. N. R. Chem. Mater. 1997, 9, 2078.
(22) Sen, R.; Govindaraj, A.; Rao, C. N. R. Chem. Phys. Lett. 1997,
267, 276.
(23) Liu, S.; Zhu, J.; Mastai, Y.; Felner, I.; Gedanken A. Chem. Mater.
2000, 12, 2205.
(24) Liu, S.; Tang, X.; Yin, L.; Koltypin, Y.; Gedanken, A. J. Mater.
Chem. 2000, 10, 1271.
(25) A private communication with Prof. Eiji Osawa.
(26) Yoo, C. S.; Cynn, H.; So¨derlind, P.; Iota, V. Phys. ReV. Lett. 2000,
84, 4132.
(27) Dai, H.; Rinzler, A. G.; Nikolaev, P.; Thess, A.; Colbert, D. T.;
Smalley, R. E. Chem. Phys. Lett. 1996, 260, 471.
Conclusion
High percentages of CoCNFs were prepared and characterized
in this paper. The shapes of the carbon flasks are quite different
from the usual carbon nanotubes. After the thinner caps of the