CVD synthesis of carbon nanotubes using a finely dispersed cobalt catalyst and their use in double layer electrochemical capacitors

Article abstract of DOI:10.1016/S0013-4686(03)00427-4

Carbon nanotubes (CNT) were obtained by chemical vapour deposition (CVD), decomposing turpentine oil over finely dispersed Co metal as a catalyst at 675°C. Scanning electron microscope (SEM) and transmission electron microscope (TEM) images reveal that the nanotubes are densely packed and of 10-50 nm in diameter. The XRD pattern of purified CNT shows that they are graphitic in nature. Resistivity measurements of these CNT indicate that they are highly conducting. Hall measurements of CNT reveal that electrons are the majority carriers with a carrier concentration of 1.35×10²⁰ cm^-3. Cyclic voltammetry (CV) and constant current charging/discharging was used to characterise the behaviour of electrochemical double layer capacitors of purified CNT with H₂SO₄. For CNT/2 M H₂SO₄/CNT, a capacitance of 12 F g^-1 (based on the weight of the active material) was obtained.

Full text of DOI:10.1016/S0013-4686(03)00427-4

Electrochimica Acta 48 (2003) 3439ꢀ3446
/
www.elsevier.com/locate/electacta
CVD synthesis of carbon nanotubes using a finely dispersed cobalt
catalyst and their use in double layer electrochemical capacitors
a
A.K. Chatterjee , Maheshwar Sharon *, Rangan Banerjee ,
b,
a
c
Michael Neumann-Spallart
a
Energy Systems Engineering, Indian Institute of Technology Bombay, Powai, 400076 Mumbai, India
b
Department of Chemistry, Indian Institute of Technology Bombay, Powai, 400076 Mumbai, India
Laboratoire de Physique des Solides et de Cristallog e´ n e` se, C.N.R.S., 1, place Aristide Briand, F-92195 Meudon Cedex, France
c
Received 19 March 2003; received in revised form 26 May 2003; accepted 26 May 2003
Abstract
Carbon nanotubes (CNT) were obtained by chemical vapour deposition (CVD), decomposing turpentine oil over finely dispersed
Co metal as a catalyst at 675 8C. Scanning electron microscope (SEM) and transmission electron microscope (TEM) images reveal
that the nanotubes are densely packed and of 10ꢀ/50 nm in diameter. The XRD pattern of purified CNT shows that they are
graphitic in nature. Resistivity measurements of these CNT indicate that they are highly conducting. Hall measurements of CNT
2
10 cm . Cyclic voltammetry (CV) and
0
ꢂ3
reveal that electrons are the majority carriers with a carrier concentration of 1.35ꢁ
/
constant current charging/discharging was used to characterise the behaviour of electrochemical double layer capacitors of purified
ꢂ1
CNT with H
#
2 4 2 4
SO . For CNT/2 M H SO /CNT, a capacitance of 12 F g (based on the weight of the active material) was obtained.
2003 Elsevier Ltd. All rights reserved.
Keywords: Carbon nanotube; CVD; Synthesis; Double layer capacitor; Turpentine oil
1
. Introduction
with high purity and in high yield, including textured
assemblies, if catalysts are used (e.g. vertical alignment).
Hydrocarbons such as methane [14], acetylene [15],
benzene [16], polyethene [17] etc., being derived from
fossil fuels, have been used as precursors for CNT. If a
high production volume is envisaged (e.g. for reinforced
composite building materials), the use of natural pre-
cursor sources might be an advantage. Sharon et al.
have obtained various forms of graphitic carbon by
pyrolysis of camphor, like fullerenes [18], nanotubes
[19], glassy carbon [20], spongy beads [21], and diamond
like carbon [22].
Preparation of carbon nanotubes (CNT) from C60 by
Curl, Kroto and Smalley [1] gave a great appraisal in the
field of development of CNT. In recent years, they have
generated tremendous interest from both fundamental
and applied perspectives [2,3]. CNT possess many
unique properties such as high mechanical strength
[
3,4], capillary properties [5] and remarkable electrical
conductivity [6,7], suggesting a wide range of potential
applications. CNT has been proposed as material for
storage of hydrogen [8], supercapacitors [9], etc. CNT
have been synthesised by various methods like arc
discharge [10], laser vaporisation [11], and plasma-
enhanced [12] or thermal chemical vapour deposition
Various carbonaceous materials have been discussed
as electrode materials in electrochemical double layer
capacitors. With certain carbon preparations of surface
2
ꢂ1
(CVD) [13]. Among these methods, CVD has been more
popular because by this method CNT can be synthesised
area up to 2000 m g , capacities as high as a few
ꢂ1
hundred F g have been obtained [23,24]. CNT, due to
their high specific surface area, are candidates for such
‘‘supercapacitors’’ [25,26], but also due to fact that
electrodes made of CNT can have a pore structure
determined by the open space between entangled fibres
*
Corresponding author. Tel.: ꢃ
152.
E-mail address: sharon@chem.iitb.ac.in (M. Sharon).
/
91-22-576-7174; fax: ꢃ91-22-576-
/
7
0
013-4686/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0013-4686(03)00427-4

3440
A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ3446
/
and therefore a high accessible surface area, unobtain-
able with other carbon materials [25].
amorphous carbon. After that, the CNT were filtered
off and washed with distilled water. This gave Co
particle free nanotubes which were finally heated in air
at 400 8C to eliminate any residual amorphous carbon.
For topological characterisation, a Philips CM-200
transmission electron microscope (TEM) and a Philips
In this paper we report on the preparation of CNT by
low temperature (675 8C) CVD of turpentine oil (de-
rived from the resin of pine trees). One point of interest
was to see whether abundant and regular nanotube
growth can be brought about using a finely divided Co
catalyst, and whether such a catalyst can be prepared by
sudden thermal decomposition of a metal salt solution
containing urea. Another point concerned the charac-
terisation of CNT obtained in this way, and of an
electrochemical double layer capacitor based on such
CNT.
1
790 X-ray diffractometer, employing CuꢀK radiation,
/
a
were used. 1 mm thick pellets (1.2 cm diameter) were
prepared to study the electrical properties of the
material. For resistivity and Hall effect measurements,
the Van der Pauw configuration was used. The resistiv-
ity of the pellets was measured between room tempera-
ture and 500 8C in argon. The series resistance of
complete cells was measured with an LCR meter
(
HewlettꢀPackard). For electrochemical studies a Pine
/
2
. Experimental
AFRDE4 potentiostat was used. A cell of configuration
CNT j H SO jj H SO j CNT sandwiched between two
2
4
2
4
Co metal catalyst powder was prepared by rapid
thermal decomposition of an aqueous solution of
Co(NO ) (7.9% w/w) and urea (39.5% w/w) by sudden
pieces of carbon paper was assembled between Perspex
plates (Fig. 2).
3
2
thermal decomposition at 600 8C. As the decomposition
of urea is highly exothermic and large amounts of
ammonia and carbon dioxide are liberated, the metal
oxide is obtained in very fine form. The residue was
reduced in H at 500 8C for 3 h, yielding a very fine
2
metal powder which was used as catalyst to grow CNT
by CVD of turpentine oil. The CVD unit consisted of
two furnaces joined side-by-side with a quartz tube
traversing both of them (Fig. 1).
The first furnace (A) was used to vaporise turpentine
oil and the second (B) was used to pyrolyse the vapours
carried over from A by a stream of Ar gas. A quartz
boat (E) with 2 ml of turpentine oil was placed in
furnace A, maintained at 200 8C. The temperature of the
furnace (B), containing a quartz boat (D) with the
catalyst (Co particles), was maintained at 675 8C. After
pyrolysis, the reaction mixture was collected from the
Fig. 2. Schematic diagram of a CNT based double layer capacitor. A
.
boat and refluxed with concentrated HNO for 24 h
3
. . carbon paper, B . . . CNT, C . . . separator (filter paper soaked with
under constant stirring at 80 8C, in order to separate the
CNT from the Co catalyst powder and most of the
electrolyte), D . . . electrolyte; the assembly is held together by two
Perspex plates.
Fig. 1. Schematic diagram of the CVD setup. A . . . vaporising furnace, B . . . pyrolysing furnace, C . . . quartz tube, D . . . quartz boat with catalyst,
E . . . quartz boat with precursor, F . . . flow regulator, G . . . gas cylinder, H . . . gas bubbler.

A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ
/
3446
3441
3
. Results and discussion
CNT growth is known to be catalysed by very small
particles of metals such as Ni, Co, Fe [27]. We therefore
devised a new method for obtaining a catalyst prepara-
tion with a high content of such particles. The rapid
thermal decomposition of metal salt solutions in the
presence of urea yielded a very fine oxide powder. In the
present case, this residue was identified as Co O by
3
4
XRD analysis and led, upon reduction, to the finely
divided Co powder. CVD growth of CNT occurred with
a high yield using this catalyst.
TEM images of CNT films are shown in Fig. 3. Fig.
a shows the CNT before purification. Here, globular
3
particles at the end of the CNT are the Co catalyst,
being absent in the purified material (Fig. 3b).
It can be seen that the diameter of a CNT is around
1
0ꢀ50 nm. All results described below were obtained for
/
purified CNT as described above. An X-ray diffraction
pattern of purified CNT is shown in Fig. 4, revealing
characteristic graphitic peaks at 2u values of 26.668 (002)
along with other planes (102) at 50.130, (100) at 42.530,
(
101) at 43.315 and (004) at 54.8858. In addition, the
reflection at 2uꢄ5.788 suggests the presence of a small
/
amount of single wall nanotubes as stated by Yosida et
al. [28]. The small peaks at 33.5 and 39.998 can be
attributed to (110) and (112) graphitic planes. Although
there may be small impurities remaining (unidentified
minor broad peak at 368), the quality of the material was
considered good enough for employing it in a capacitor:
prolonged purification leads to extensive loss of mate-
rial, which would not be practical for an application. On
the other hand, unpurified material gave quite unsatis-
factory results (more than ten times lower capacities).
Hall measurements of CNT pellets were carried out to
ascertain carrier type, mobility and carrier concentra-
tion. The material itself cannot precisely characterised in
this way, as individual CNT cannot be measured.
However, a rough estimation for pellets made of CNT
can give some useful information: electrons were found
to be the majority carriers with a concentration of
2
0
ꢂ3
around 1.4ꢁ10 cm taking into account the macro-
/
scopic dimensions of the pellets. The temperature
dependence of the resistivity between room temperature
and 700 8C shows a small linear decrease in the
conductivity with increase of temperature indicating
semiconducting nature of the prepared carbon. The
room temperature resistivity was found 0.024 V cm
Fig. 3. TEM images of the CNT before (a) and after (b) purification.
liquid pellet. For initial characterisation of this new
CNT preparation, we gave preference to this configura-
tion instead of using a solid pellet including filler.
Nevertheless, some experiments with PVA as filler
(
about 20 times higher than graphite, but low enough
for CNT to be used in an electrochemical device).
Junctions of CNT with aqueous H SO solution were
2
4
prepared in order to study the electrochemical proper-
ties of the material. The device used is outlined in Fig. 2.
The CNT, contained in a small compartment, are
soaked throughout with the electrolyte before mounting
of the device leading to a slightly compressed solid/
were carried out*
/
results not shown*and found pro-
/
mising, although showing a change of characteristics
with time (for a period of several weeks). We tried to
obtain information about the capacitance of the junc-

3442
A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ
/
3446
Fig. 4. XRD pattern of CNT.
tion from cyclic voltammetry (CV). From a potentiody-
namic sweep, the capacitance (C) of any capacitor can
be obtained by considering that C equals dQ/dU and
the scan rate (s) is dU/dt. As the charging current, i, is
dQ/dt, C can be expressed as
rates (in fact, a scan rate of 100 mV sꢂ1 formally
corresponds to operation of the device at 1/16 Hz). A
value of 0.12 F was obtained under these conditions,
ꢂ1
corresponding to a specific capacitance of 12 F g
based on the total weight of active material in the two
electrodes. Peaks observed in CVs with higher scan
reversal voltages (Fig. 6) hint to electrochemical oxida-
tion of the CNT, a fact well known for carbon materials
C ꢄi=(dU=dt)ꢄi=s
(1)
For a cyclic scan, one half of the difference between
forward scan and backscan currents is taken. For an
electrode, i.e. a conductor/electrolyte junction, the
[
23], or redox reactions of yet unidentified surface active
centres already present in the here presented CNT
preparation, which might be suppressed by further
pretreatment steps.
double layer capacitance, C , can be estimated in this
dl
way, if Faradaic currents, or pseudocapacitance, are
absent. Otherwise, the dependence of i on scan rate
shows a more complex behaviour, which is actually
observed for cell voltage excursions beyond 0.4 V. A
serial of typical CVs of CNT/2 M H SO at different
To confirm the capacity measurement through CV
and to get information of the temporal behaviour, using
a generally adopted test protocol, we also performed
galvanostatic charging experiments. From Eq. (1), the
cell voltage is expected to be a linear function of
charging time. Constant currents of 1, 2, 4, and 8 mA
yielded a typical charging/discharging behaviour of a
capacitor (ideally a ‘‘/\/\/\. . .’’ shaped voltage response).
An example with current inversion voltages of 0.4 and
2
4
scan rates, are shown in Fig. 5.
In the inset, the currents at 0 V are plotted as a
function of scan rate, assuming there is least contribu-
tion of Faradaic reactions at this voltage, however
present. The plot shows a linear dependence of current
on scan rate for high scan rates. At low scan rates, a
sublinear dependence is observed. This is the expected
behaviour for Faradaic currents when diffusion kinetics
are involved, but also for the distributed nature of a
porous electrode. In the here investigated samples, the
pores in the electrode are interconnected spaces filled
with electrolyte. These spaces are large compared with
the dimension of the nanotubes, so that electrolyte
access will be easier than in the case of activated carbon
containing extremely small pores which are difficult to
access. Nevertheless, for the present case, both, porosity
and Faradaic reactions, can contribute to the observed
behaviour (below, we shall show further evidence for the
presence of the latter). If the high frequency response is
mainly representative for the double layer capacitance of
the device, the lower limit of capacitance may thus be
estimated from the constant slope at the higher scan
ꢂ
/
0.4 V is shown in Fig. 7.
From the absolute value of the slopes of the voltageꢀ
/
time behaviour a capacitance of 0.1 F was again
obtained. However, the slope decreased when approach-
ing the reversal potentials. At higher current inversion
voltages or at lower charging currents, ‘‘slower’’ com-
ponents were gaining importance. The former is related
to pseudocapacitances resulting from Faradaic reac-
tions, while the latter is related to frequency dispersion
of the capacitance due to the distributed nature of the
electrodes. Such slow electron transfer reactions or
Faradaic leakage currents represent a shunt to the
double layer capacitance which also leads to the
observed rapid self-discharge. As double layer capaci-
tance and pseudo-capacitance are in parallel, use of this
additional pseudo-capacitance as charge backup of the
fast double layer capacitance can only be made if these

A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ
/
3446
3443
Fig. 5. CV of CNT based capacitor (5 mg per electrode) in 2 M sulphuric acid at scan rates of 2, 5, 10, 20, 50, and 100 mV sꢂ1 (increasing currents);
inset: dependence of current at 0 V on scan rate.
processes are reversible (a fact that has also been stated
earlier [26]), which did not seem to be the case here. This
can be seen from voltage step experiments: total charges
of up to 0.8 C were found by integration of the iꢀt
behaviour upon a voltage step from 0 to 0.4 V. The
corresponding capacitance exceeds the one estimated
/
Fig. 6. CV of CNT in 2 M sulphuric acid with different scan reversal potentials; weight of electroactive material . . . 2ꢁ5 mg; scan rate . . . 200 mV
/
ꢂ1
s
.

3444
A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ3446
/
Fig. 7. Voltageꢀ
electroactive material . . . 2ꢁ
/
time dependence of the device shown in Fig. 2 for charging currents of (a) 8 and 4 mA, (b) expanded view for 4 mA; weight of
5 mg, 2 M sulphuric acid.
/
from CV or current cycling by a factor of ten. Upon
stepping back from 0.4 to 0 V, however, only 0.06 C
were recovered, corresponding to the double layer
capacitance calculated from CV.
Table 1 resumes experiments carried out with cells
made with different weights of electrode material. The
capacitance depends on the amount of active material,
as expected, and the specific capacitance is around 12 F
ꢂ1
g
.
The table shows fairly good agreement between
capacitance measurements by CV and by galvanostatic
step experiments, with the values obtained from CV
being slightly lower (as expected from the discussion of
the CV results above). Cells with 2 M sulphuric acid

A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ
/
3446
3445
Table 1
Capacitance of cells with CNT electrodes of 2.5 and 5 g in different media
Method
[H
2
SO
4
] (M)
CNT weight (total g)
0.01
Capacitance (F)
Specific capacitance (F gꢂ1
)
CV
Galv.
CV
0.1
2.0
2.0
0.085
0.125
0.118
0.139
0.060
0.063
8.5
12.5
11.8
13.9
12.0
12.6
0.01
Galv.
CV
Galv.
0.005
CV . . . by cyclic voltammetry, galv. . . . by galvanostatic step cycling.
showed lower resistance (5 V) compared with cells with
presented example, an electrochemical double layer
capacitor was assembled and tested. Specific capacities
0
.1 M acid (9 V). However, the self discharge was found
ꢂ1
to be higher for the higher acid concentration, presum-
ably due to more efficient catalysis of the Faradaic
reactions. Complete self discharge occurred in about 100
s for this cell, and even led to polarity reversal of the
device, depending on its history (previous scans, etc.).
of up to 12 F g
(based on the total weight of active
material) were obtained for this device. Slow response at
higher voltages and fast self-discharge hints to the
occurrence of Faradaic pseudo-capacitance.
Somewhat higher capacitances have previously been
ꢂ1
found for CNT base capacitors (81 F g at 10 Hz),
however for much thinner electrodes, having undergone
Acknowledgements
an unspecified process of ‘‘thermal crosslinking’’ [25],
and for acetylene black containing, PVDF binder
A.K.C. acknowledges the financial support provided
by the Ministry of Non-Conventional Energy Sources
ꢂ1
supported CNT electrodes (36ꢀ
/80 F g
[26]). How-
(MNES, India) for carrying out the research. The help
ever, our aim here was to characterise a new CNT
material first in the absence of any additives.
rendered by the Department of Earth Science and RSIC
of I.I.T. Bombay regarding TEM and XRD analysis is
acknowledged.
ꢂ2
From typical capacitance values around 20 mF cm
observed for carbon electrodes [24], a surface area
2
ꢂ1
around 100 m g can be derived for the here evaluated
CNT. Although ‘‘activated’’ carbon, having a large
2
number of pores, shows values up to 2000 m g
ꢂ1
,
References
some of the very small pores cannot be reached (filled)
by the electrolyte, leading to much lower actual values
of specific capacitance than estimated from the specific
surface area. As CNT do not have such pores, it can be
imagined that basically the total surface area can be
made available for the double layer to develop. Still,
some spreading of the material by an electrolyte
impregnated filler may increase accessibility. However,
interconnectivity of the CNT must not be lost. Experi-
ments in this direction are presently under way, also
with regard to a more practical (solid/semi-solid)
implementation of the device.
[1] H.W. Kroto, J.R. Heath, S.C. O’Brien, R.F. Smalley, Nature 318
(
1985) 162.
2] B. Susan, A.S. Rodney, Crit. Rev. Solid State Mater. Sci. 26
2001) 145.
3] R.M. Braun, Chem. Eng. News 75 (1997) 39.
[
(
[
[4] M.M.J. Treacy, T.W. Ebbesen, J.M. Gibson, Nature 381 (1996)
678.
[
5] M.R. Pederson, J.Q. Broughton, Phys. Rev. Lett. 69 (1992) 2689.
6] T.W. Ebbesen, H.J. Lezec, H. Hiura, J.W. Bennett, H.F. Ghaemi,
T. Thio, Nature 382 (1986) 54.
[
[7] J.W. Mintmire, B.I. Dunlap, C.T. White, Phys. Rev. Lett. 68
(1992) 631.
[8] H.-M. Cheng, Q.-H. Yang, C. Liu, Carbon 39 (2001) 1447.
9] Q. Xiao, X. Zhou, Electrochim. Acta 48 (2003) 575.
[
[
[
[
10] C. Journet, W.K. Maser, P. Bernier, A. Loiseau, M.L. Chapelle,
S. Lefrant, P. Deniard, R. Lee, J.E. Fischer, Nature 388 (1997)
7
56.
4
. Conclusion
11] 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. Tombnek, J.E. Fischer, R.E. Smalley, Science 273 (1996) 483.
12] M. Chen, C.-M. Chen, C.-F. Chen, J. Mater. Sci. 37 (2003) 561.
A finely divided Co metal catalyst, synthesised by
decomposition of cobalt nitrate in the presence of urea
supported the abundant growth of CNT during pyr-
olysis of turpentine vapours. These CNT are highly
dense and intrinsically n-type. They have a low resis-
tance due to high carrier concentration (1.4ꢁ10
cm ). Such low resistance CNT are attractive for
electrochemical and electronic applications. In the here
[13] M.P. Siegel, D.L. Overmyer, P.P. Provencio, Appl. Phys. Lett. 80
2002) 2171.
(
[
[
14] E. Ruckenstein, H.H. Yun, Carbon 36 (1998) 269.
15] W.Z. Li, S.S. Xie, L.X. Qian, B.H. Chang, B.S. Zou, W.Y. Zhou,
R.A. Wang, Science 274 (1996) 1701.
2
0
/
ꢂ3
[16] A.M. Benito, Y. Maniette, E. Mu n˜ oz, M.T. Mart ´ı nez, Carbon 36
(1998) 681.

3446
A.K. Chatterjee et al. / Electrochimica Acta 48 (2003) 3439ꢀ3446
/
[
[
17] N.A. Kiselev, D.N. Zakharov, J. Sloan, E.F. Kukovitskii, L.J.
Hutchinson, A.S. Kotosonov, Mol. Mater. 10 (1998) 155.
18] K. Mukopadhyay, K.M. Krishna, M. Sharon, Phys. Rev. Lett. 72
[23] S. Sarangapani, B.V. Tilak, C.-P. Chen, J. Electrochem. Soc. 143
(1996) 3791.
[24] A. Burke, J. Power Sources 91 (2000) 37.
[25] C. Niu, E.K. Sichel, R. Hoch, D. Moy, H. Tennent, Appl. Phys.
Lett. 70 (1997) 1480.
(1994) 3182.
[
[
19] D. Pradhan, M. Sharon, Mater. Sci. Eng. B 96 (2002) 24.
20] M. Sharon, K. Mukhopadhyay, Mater. Chem. Phys. 49 (1997)
[26] E. Frackowiak, K. Metenier, V. Bertagna, F. Beguin, Appl. Phys.
Lett. 77 (2000) 2421.
105.
[
[
21] M. Sharon, K. Mukhopadhyay, K. Yase, S. Iijima, Y. Ando, X.
Zhao, Carbon 36 (1998) 507.
22] K. Mukhopadhyay, M. Sharon, Mater. Manufact. Processes 12
[27] E. Flahaut, A. Govindaraj, A. Peigney, C. Laurent, A. Rousset,
C.N.R. Rao, Chem. Phys. Lett. 300 (1999) 236.
[28] Y. Yosida, J. Appl. Phys. 87 (2000) 3338.
(1997) 541.