Correlating the preparation and performance of cobalt catalysts supported on carbon nanotubes and carbon spheres in the Fischer-Tropsch synthesis

Article abstract of DOI:10.1016/j.jcat.2010.11.010

A series of Co catalysts supported on carbon nanotubes (CNTs) and carbon spheres (CSs) with different cobalt particle sizes (3-45 nm) were prepared by different methods and using different cobalt precursors. The Co/CNTs and Co/CSs can be autoreduced by the supports in N₂ at ca. 480 °C; they show better Fischer-Tropsch performance than those reduced in H₂ when the reduction T > 400 °C. The turnover frequency (TOF) value for both Co/CNT and Co/CS was constant for cobalt particles above 10 nm and decreased sharply for the cobalt catalysts with smaller cobalt particles. Remarkably, the TOF for 11 catalysts prepared using different precursors and preparation methods on two different carbon supports depends only on particle size. Finally, a positive relationship was observed between cobalt particle size and C₅₊ selectivity for both Co/CNT and Co/CS catalysts.

Full text of DOI:10.1016/j.jcat.2010.11.010

Journal of Catalysis 278 (2011) 26–40
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Correlating the preparation and performance of cobalt catalysts supported
on carbon nanotubes and carbon spheres in the Fischer–Tropsch synthesis
Haifeng Xiong a,b, Myriam A.M. Motchelaho b,c, Mahluli Moyo b,c, Linda L. Jewell , Neil J. Coville
c
a,b,
⇑
a
DST/NRF Centre of Excellence in Strong Materials, University of the Witwatersrand, Johannesburg 2050, South Africa
Molecular Sciences Institute, School of Chemistry, University of the Witwatersrand, Johannesburg 2050, South Africa
b
c
School of Chemical and Metallurgical Engineering, University of the Witwatersrand, Johannesburg 2050, South Africa
a r t i c l e i n f o
a b s t r a c t
Article history:
A series of Co catalysts supported on carbon nanotubes (CNTs) and carbon spheres (CSs) with different
cobalt particle sizes (3–45 nm) were prepared by different methods and using different cobalt precursors.
Received 8 September 2010
Revised 26 October 2010
Accepted 13 November 2010
Available online 28 December 2010
The Co/CNTs and Co/CSs can be autoreduced by the supports in N
2
at ca. 480 °C; they show better
Fischer–Tropsch performance than those reduced in H when the reduction T > 400 °C. The turnover fre-
2
quency (TOF) value for both Co/CNT and Co/CS was constant for cobalt particles above 10 nm and
decreased sharply for the cobalt catalysts with smaller cobalt particles. Remarkably, the TOF for 11 cat-
alysts prepared using different precursors and preparation methods on two different carbon supports
depends only on particle size. Finally, a positive relationship was observed between cobalt particle size
and C5+ selectivity for both Co/CNT and Co/CS catalysts.
Keywords:
Fischer–Tropsch synthesis
Cobalt catalysts
Carbon nanotubes
Carbon spheres
Autoreduction
Particle size
Ó 2010 Elsevier Inc. All rights reserved.
Turnover frequency
Selectivity
1
. Introduction
etc. Recently, great attention has been paid to catalysts supported
on novel carbon nanostructured materials such as carbon nanofi-
Fischer–Tropsch synthesis (FTS) is known as a reaction that pro-
duces hydrocarbons from syngas (CO and H ) derived from coal,
natural gas, or biomass and is used for making chemicals and
transportation fuels [1–5]. The FTS catalyst is typically placed on
bers (CNFs) [11,12], multi-walled carbon nanotubes (MWCNTs)
[13], and carbon spheres (CSs) [14] for use in the FT reaction. De
Jong and coworkers have used fish-bone-like CNFs as a model sup-
port to systematically study cobalt FTS catalysts [11,12]. They
found that the turnover frequency (TOF) for CO hydrogenation
was independent of cobalt particle size for catalysts with sizes lar-
ger than 6–8 nm [11]. The lower TOF of catalysts for smaller cobalt
particles was caused by both blocking of edge/corner sites and a
lower intrinsic activity at the small terraces.
2
2 3 2 2
a support, and the supports used include Al O , SiO , TiO , and
MgO. However, the use of these supports can generate a strong me-
tal–support interaction with the classical FT catalysts, Fe and Co.
This can even lead to the formation of mixed compounds such as
2 2 2 4 4
Co SiO [6,7], Co AlO [8,9], or CoTiO [10], which are reducible
only at high reduction temperatures. To lower this interaction,
some costly metals such as Ru, Pt, and Re can be added to the Co
to assist with the reduction of the cobalt oxides. An alternative is
to use carbon as the support instead of a metal oxide. The use of
inert carbon materials as supports for cobalt FTS catalysts is an
effective method to enhance the reducibility of cobalt oxide. In-
deed, carbon has been shown to have further benefits over metal
oxides as supports. These include cost, easy removal of supports
by oxidation to recover precious metals, limited micro-porosity,
MWCNTs represent an interesting alternative to conventional
supports for reasons that include their impressive mechanical
properties, the high accessibility of the active phase, and the ab-
sence of any micro-porosity, thus eliminating intraparticle mass
transfer in the reaction medium. Further, their specific surface area
can be modulated between 50 and 500 m g , and their internal
diameter can be adjusted from 5 to 100 nm [15]. MWCNT-based
catalysts have been used for the Fischer–Tropsch reaction. A Fe/
MWCNT catalyst has been found to be a very stable and active
FTS catalyst [16]. Guczi et al. [17] reported that high catalytic activ-
ity and high selectivity toward C5+ fractions were obtained from
Fe/CNT catalysts prepared by impregnation methods. Kang et al.
[13] reported that Ru nanoparticles supported on carbon nano-
tubes, with a mean size around 7 nm, exhibited the highest
2
À1
⇑
Corresponding author at: DST/NRF Centre of Excellence in Strong Materials,
University of the Witwatersrand, Johannesburg 2050, South Africa. Fax: +27 11 717
6
749.
E-mail address: Neil.Coville@wits.ac.za (N.J. Coville).
0
021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.11.010

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
27
C
10–C20 selectivity (ca. 65%) and a relatively higher TOF for CO con-
In this study, similarities and differences between Co/CNT and
Co/CS catalysts have been investigated. The autoreduction and
the effect of precursor and preparation methods of the catalysts
on FTS performance are investigated. Carbon nanotubes and car-
bon spheres were prepared by a chemical vapor deposition meth-
od. Cobalt FTS catalysts supported on carbon nanotubes and carbon
spheres were made by different methods (impregnation and depo-
sition precipitation) and using different cobalt precursors (cobalt
nitrate and cobalt acetate). These catalysts were characterized by
a range of techniques, and their FT activity was evaluated in a
fixed-bed reactor.
version. Recently, it has been found that MWCNT-encapsulated
iron catalysts showed higher FTS activity than catalysts with iron
nanoparticles dispersed on the outer walls of MWCNTs [3,18].
CNT-supported cobalt-based catalysts with different loadings of
cobalt, ruthenium, and potassium have also been studied in the
Fischer–Tropsch synthesis [19]. It has been found that for a
1
5 wt.%Co/CNT catalyst, most of the metal nanoparticles were
homogeneously dispersed in the internal cavity of the MWCNTs,
but increasing the cobalt loading to 30 wt.% increased the amount
of cobalt on the outer surface of the CNTs. This resulted in an in-
crease in the cobalt particle size and a decrease in the reduction
temperature and dispersion. More recently, Zhang et al. [20]
showed that the use of carbon nanotubes with different outer
diameters did not affect cobalt dispersion and that carbon nano-
tubes with different diameters exhibited similar cobalt particle
sizes with similar reducibility. It appears that cobalt particle size
in cobalt catalysts supported on CNTs is mostly influenced by
chemical interactions between the cobalt and the nanotubes,
which probably occur during decomposition of the cobalt precur-
sor, and nucleation and growth of cobalt oxide crystallites.
2
. Experimental
2.1. Sample preparation
2.1.1. The preparation of MWCNTs and CSs
3
A 5%Co/5%Fe/CaCO catalyst was prepared by wet impregnation
of a CaCO support with an aqueous cobalt nitrate and an iron ni-
3
trate solution as described in our previous study [27]. In a previous
publication [27], we performed a systematic study of the synthesis
of CNTs using Fe and Co on CaCO . In that study, the conditions
3
(temperature, flow rates, gas mixtures, and catalyst composition)
to produce very pure CNTs were established. These conditions
were used in this study. The advantage of the support used is that
the CaCO is easy to remove with dilute acid. Briefly, calculated
3
amounts of the Fe and Co nitrates were mixed and dissolved in dis-
tilled water, and then, the resulting solution was impregnated di-
Carbon spheres (CSs) are composed of random curling graphitic
flakes with a size of 1–10 nm and are similar to CNTs, in having
carbon atoms with sp hybridization [21,22]. The CSs are expected
2
to have analogous chemical characteristics to CNTs. When pre-
pared by chemical vapor deposition, the carbon spheres are formed
without a catalyst and no purification is needed. Furthermore, they
are non-porous and can be expected to be good model supports in
catalysis [23]. Carbon spheres have been known for decades, but
recent studies have led to added information on their synthesis,
chemistry, and properties. Recent preliminary studies on Co/CS
catalysts showed that the cobalt oxide can be autoreduced by the
carbon spheres and that the autoreduced catalyst shows better
rectly onto the CaCO
aged for 12 h in air at room temperature, followed by drying at
20 °C for 12 h and calcining in air at 600 °C for 17 h. A tubular
3
support. Subsequently, the material was
1
quartz reactor (510 mm  19 mm i.d.) was placed horizontally in
a furnace and was used to pyrolyze acetylene and synthesize car-
bon nanotubes by the catalytic chemical vapor deposition method
FTS performance in comparison with a catalyst reduced in H
14]. More significantly, it was observed that the Co particles could
2
[
3
using the as-prepared 5%Co/5%Fe/CaCO as catalyst.
readily be detected by electron microscopy techniques. Model iron
FTS catalysts supported on carbon spheres have also been prepared
to investigate the effect of the preparation method, iron precursor,
and promoters on FTS performance [23]. A good match of results
with previous studies on non-carbon supports has been obtained.
Recently, nickel catalysts supported on carbon spheres have been
investigated in the hydrogenation of butyronitrile [24,25].
Although CNTs are often shown to have a smooth outer wall
consisting of a single graphene layer (in structural diagrams of
models used for computational chemistry), TEM studies seldom
give evidence of these ‘ideal’ surfaces, especially images of
MWCNTs. Rather, the CNT surface is covered with flake-like aggre-
gates [26], similar to the surface of a CS. To illustrate this, a cartoon
of a typical CNT and a CS is shown in Fig. 1. Given the similarity in
their surfaces, one would expect CNTs and CSs to behave in a sim-
ilar way when used as catalyst supports. Yet, in work reported to
date, no comparative study between the effect of CNTs and CSs
has been undertaken.
The detailed experimental procedures are as follows: the cata-
lyst (0.5 g) was spread to form a thin layer in a quartz boat
(
120 mm  15 mm), and the boat was then placed in the center
À1
of the quartz tube. The furnace was then heated at 10 °C min un-
der flowing N
7
À1
2
(40 mL min ). Once the temperature had reached
À1
00 °C, the N
2
flow rate was set to 240 mL min and C
2
H
2
was
À1
introduced at a constant flow rate of 90 mL min . After 60 min
2 2
of reaction time, the C H flow was stopped and the furnace was
left to cool down to room temperature under a continuous flow
À1
of N
2
(40 mL min ). The boat was then removed from the reactor,
and the carbon deposit that formed on the surface of the catalyst
was collected.
Carbon spheres (CSs) were prepared by a chemical vapor depo-
sition method [23]. A tubular quartz reactor was placed horizon-
tally in a furnace. The acetylene was pyrolyzed in the tube and
generated the carbon spheres. The temperature in the reactor
À1
was firstly increased to 900 °C in a flow of N
2
(40 mL min ). Sub-
sequently, acetylene was introduced into the reactor at a flow rate
À1
of 100 mL min and held at 900 °C for 1 h. The pyrolysis reaction
generated a sooty deposit on the inner wall of the quartz tube in
the hot region of the reactor. Finally, the reactor was cooled down
to room temperature in flowing N , and the soot was collected,
2
weighed, and analyzed. Scanning electron microscopy (SEM) anal-
ysis revealed that the size of the prepared carbon spheres was
2
À1
around 900 ± 50 nm with a BET surface area of <5 m g
.
2
.1.2. The purification and functionalization of carbon nanotubes and
carbon spheres
The purification and functionalization of CNTs were achieved
using 55 vol.% HNO solution. Briefly, the raw (i.e., unpurified)
Fig. 1. Cartoon of (a) a carbon nanotube and (b) a carbon sphere.
3

2
8
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
carbon nanotubes (1.0 g) were mixed with 55 vol.% HNO
3
(100 mL)
The scan range was 15°–90° with 0.002° steps. A Bruker D8 powder
X-ray diffractometer with monochromatic Cu K radiation and Ni
and refluxed at 120 °C for 17 h. Then, the mixture was filtered and
washed with deionized water until the pH of the mother liquor
reached a value of 7. The resulting material was dried overnight
at 90 °C in an oven.
a
filter equipped with a VANTEC-1 detector was used for the
in situ XRD measurements. A reactor cell (Anton Paar) mounted
on a goniometer was used for in situ measurement under different
atmospheres. Prior to the XRD measurement, the reactor was
flushed with high-purity argon at 150 °C for 1 h, to drive off the
water and other impurities, and then cooled down to 100 °C. The
diffractograms were recorded from 15° to 90° with 0.016° steps.
Crystallite phases were determined by comparing the diffraction
patterns with those in the standard powder XRD files (JCPDS) pub-
lished by the International Center for Diffraction Data.
For the functionalization of CSs, about 1.0 g of CSs was function-
3
alized in 100 mL of 55 vol.% HNO at different temperatures (45,
7
0, and 90 °C) for 17 h. Then, the mixture was filtered and washed
with deionized water until the pH of the mother liquor reached a
value of 7. The resulting material was dried overnight at 100 °C
in an oven and denoted as CS-A, CS-B, and CS-C for CSs prepared
at 45, 70, and 90 °C, respectively.
2.1.3. Catalyst preparation
2.2.2. N
2
adsorption–desorption
Impregnation and homogeneous deposition precipitation meth-
N
2
adsorption–desorption experiments were conducted at
ods were employed to prepare cobalt catalysts supported on car-
bon nanotubes and carbon spheres. For the impregnation
method, the appropriate cobalt nitrate solution (15 wt.%) was
impregnated onto the functionalized carbon nanotubes directly
and then dried at 100 °C for 10 h. The catalysts prepared by
impregnation were denoted as Co/CNT-IM (IM stands for impreg-
nation). For the Co/CNT catalyst prepared by the homogeneous
deposition precipitation method, urea was used as the precipitat-
ing agent and the detailed synthesis route was as follows: Co(N-
À193 °C using a Micromeretics Tristar 3000 surface area and
porosity analyzer. Prior to an experiment, the sample was out-
gassed at 200 °C for 6 h. The BET surface areas were obtained for
adsorption data in a relative pressure range from 0.05 to 0.30.
2
The total pore volumes were calculated from the amount of N va-
por adsorbed at a relative pressure of 0.99. The pore size distribu-
tions were evaluated from the desorption branches of the
isotherms using the Barrett–Joyner–Halenda (BJH) method.
O
3
)
2
Á6H
cobalt) were dissolved in deionized water (200 mL) and added to
.5 g of the functionalized carbon nanotubes. Subsequently, the
temperature was raised to 90 °C. After allowing sufficient time
17 h) for the hydrolysis of the urea, the sample was filtered and
washed with deionized water, followed by drying at 100 °C for
0 h. The sample was denoted as Co/CNT-DP (DP stands for depo-
sition precipitation). A Co/CNT catalyst was also prepared using co-
balt acetate (Co(OOCCH ) as the cobalt precursor. The sample
2
O (0.41 g) and urea (0.27 g; 2 mol urea per mole of
2.2.3. Thermogravimetric analysis (TGA)
Thermogravimetric analysis was performed with a Perkin-
Elmer STA6000 TGA using nitrogen or air as the purge gas and a
0
À1
heating rate of the 10 °C min . The flow rate of the purge gas
À1
(
was always 20 mL min
.
1
2.2.4. Transmission electron microscopy (TEM)
TEM analysis was performed on a Tecnai spirit transmission
electron microscope at 120 kV, and high-resolution transmission
electron microscopy was carried out on a FEI Tecnai F20 FEGTEM
3 2
)
was denoted as the CoA/CNT-IM.
For the Co/CS catalysts prepared by impregnation, the appropri-
ate cobalt precursor solution (5 wt.%) was added to the functional-
ized carbon spheres directly. The material was then dried at 100 °C
for 10 h. The catalysts prepared by impregnation were denoted as
Co/CS-N-IM (where N = A, B, and C corresponds to the functionali-
zation temperatures of 45, 70, or 90 °C; IM refers to impregnation).
For the Co catalyst prepared by the homogeneous deposition pre-
cipitation method, urea was used as the precipitating agent and
2
at 200 kV. An ex situ reduction step (under H or Ar atmosphere)
was used to reduce and passivate the catalysts. The pre-reduction
and passivation were carried out in a fixed-bed reactor. Firstly, the
sample was reduced at different temperatures (300, 400, and
480 °C) for 20 h. Subsequently, the temperature was decreased to
À1
30 °C, and then, 1% O
2
/He (30 mL min ) was introduced into this
system to passivate the sample for 6 h. After these steps, the pas-
sivated catalyst was dispersed in methanol using a sonicator and
loaded onto a copper grid. The particle size of the cobalt formed
was determined by counting at least 100 ‘shaped objects’ per sam-
ple. These were randomly chosen from different TEM images. Com-
mon Gaussian statistics yields values for the average cobalt
diameter.
the detailed synthesis route was as follows: Co(NO
3
)
2
Á6H
2
O
(
0.41 g) and urea (0.18 g) in a cobalt/urea ratio of 1:2 were dis-
solved in deionized water (200 mL) and added to 0.5 g of the func-
tionalized carbon spheres (CS-A, CS-B, and CS-C). Subsequently, the
temperature was raised to 90 °C. After allowing sufficient time
(
17 h) for the hydrolysis of the urea, the sample was filtered and
dried at 100 °C for 10 h. The sample was denoted as Co/CS-N-DP
where N = A, B, and C corresponds to the functionalization tem-
2.2.5. Temperature-programmed reduction (TPR)
(
The TPR experiment was carried out with a Micromeritics Auto
Chem II unit. The catalyst (ca. 0.1 g) was placed in a quartz tubular
reactor, fitted with a thermocouple for continuous temperature
measurement. The reactor was heated with a furnace. Prior to
the temperature-programmed reduction measurement, the cal-
cined catalysts were flushed with high-purity argon at 150 °C for
1 h, to drive away the water or impurities, and then cooled down
peratures of 45, 70, or 90 °C). A 5%Co/CS catalyst was also prepared
by impregnation of acid functionalized CSs (CS-C) using cobalt ace-
tate (Co(OOCCH ) ) as the cobalt precursor. The preparation proce-
3 2
dure for the catalyst was the same as that used to make 5Co/CS-C-
IM. The obtained sample was denoted as CoA/CS-C-IM.
It should be mentioned that a portion of the obtained cobalt cat-
alyst prepared above was saved for TEM characterization as well as
for FTS study. Another portion of the supported cobalt catalyst was
to 50 °C. Then, 5% H
2
/Ar was switched on, and the temperature
À1
was raised at a rate of 10 °C min
from 50 to 800 °C (hold
calcined in a flow of N
desorption measurements.
2
at 250 °C for 4 h for XRD and N
2
adsorption
10 min). The gas flow rate through the reactor was controlled by
3
À1
three Brooks mass flow controllers and was always 50 cm min
.
2
The H consumption (TCD signal) was recorded automatically by a
2
2
.2. Catalyst characterization
PC.
.2.1. X-ray powder diffraction (XRD)
2
2.2.6. H Chemisorption
X-ray powder diffraction spectra for the calcined catalysts were
recorded with a Siemens D5 using Cu K radiation and a Ni filter.
Hydrogen chemisorption was carried out in a U-tube quartz
reactor with a Micromeritics ASAP 2020 unit to give the number
a

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
29
of active surface metal atoms. The sample weight used was ca.
Table 1
À6
N2 adsorption–desorption data of the carbon supports and cobalt catalysts.
0
.200 g. The catalyst was first evacuated to 10 mm Hg at
1
00 °C for 1 h. The catalyst was then reduced at 400 °C for 16 h
Sample
BET surface
area (m /g)
Total pore
volume (cm /g)
Average pore
size (nm)
a
2
3
using a flow of high-purity hydrogen and evacuated at 400 °C for
h to desorb any H . Subsequently, H was introduced to the sys-
tem, and the chemisorption was performed at 30 °C.
1
2
2
Raw CNTs
CNTs
Co/CNT-IM
CoA/CNT-IM
Co/CNT-DP
CS-A
46.4
79.1
73.4
103.9
91.7
1.5
1.6
2.6
2.5
2.4
0.18
0.30
0.28
0.36
16.1
14.9
15.2
13.6
17.3
–
–
–
–
–
2.3. Catalytic evaluation
0.39
0.0025
0.0033
0.0046
0.0095
0.0093
0.0048
0.0063
0.0094
CS-B
CS-C
The Fischer–Tropsch synthesis was performed in a fixed-bed
micro-reactor. Gas cylinders containing H
3
stream to the catalyst with a space velocity of 3840 h . N
used as an internal standard in order to ensure accurate mass bal-
ances. Catalyst (0.1 g) was added to the reactor and reduced in situ
at 400 °C for 20 h under a stream of H
2 2
/CO/N mixtures (60/
Co/CS-C-IM
CoA/CS-C-IM
Co/CS-A-DP
Co/CS-B-DP
Co/CS-C-DP
0/10 vol.% purity: 99.99) were used to supply the reactant gas
À1
2
was
2.3
3.0
3.4
–
–
–
À1
a
2
3
2
(2 bar, 30 mL min ). After
The uncertainties in the BET surface area <5 m /g (for density <1 g/cm ) are
large.
reduction, the temperature was decreased to 225 °C, synthesis gas
was introduced, and the pressure was increased gradually to 8 bar.
All gas lines after the reactor were kept at 150 °C, and a hot trap
placed immediately after the reactor was held at this temperature
in order to collect wax. A second trap kept at ambient temperature
was used to collect the oil and water mixture. The flow was con-
trolled using a metering valve and measured by a bubble meter.
The product stream was analyzed online using two gas chromato-
graphs. A thermal conductivity detector (TCD), equipped with a
Porapak Q (1.50 m  3 mm) packed column, was used to analyze
BET surface areas and pore volumes of the functionalized car-
bon nanotubes and cobalt catalysts are displayed in Table 1. As
can be seen, a relatively low pore volume and BET surface area
for raw CNTs are observed. After the raw CNTs were purified with
HNO
indicated that the treatment of the raw CNTs in nitric acid at
20 °C for 17 h has removed the residual Fe–Co/CaO. Compared
3
, the BET surface area and total pore volume increased. This
1
to the respective supports, the cobalt-loaded catalysts showed sig-
nificantly higher BET surface area and pore volume.
Fig. 3 displays the TGA/DTG curves of the unpurified CNTs, the
purified CNTs, and the Co/CNT catalysts prepared under both air
2 2
N , CO, and CO , and a flame ionization detector (FID), equipped
with a Porapak Q packed column, was used for the analysis of
hydrocarbons.
2
and N atmospheres. The peak position of weight loss is given in
3
. Results and discussion
Table S1. For unpurified carbon nanotubes, the sample is stable be-
low 500 °C and a ca. 80% weight loss can be detected when the
pristine CNTs are heated to 900 °C under air (Fig. 3a) due to the loss
of the carbon by oxidation. The residual weight is attributed to the
Fe-Co/CaO catalyst used during the preparation of CNTs. The rela-
tive DTG curve shows two distinct weight loss peaks between 500
and 750 °C (Table S1). After the raw CNTs were refluxed in 55%
3
3
.1. Support and cobalt catalyst characterization
.1.1. Carbon nanotubes and Co/CNT catalysts
The XRD patterns of some of the prepared Co/CNT catalysts
2
thermally treated in N are displayed in Fig. 2. For both Co/CNT-
IM and CoA/CNT-IM, the XRD patterns (Fig. 2a and b) show diffrac-
tion peaks at 31.5°, 37.0°, 44.9°, 59.6°, and 65.5°, corresponding to
the different crystal planes of the Co
3
HNO for 17 h, the TGA curve of the purified CNTs shows that
the sample is entirely oxidized at 700 °C under air (Fig. 3b). The
relative DTG curve indicated that three weight loss peaks exist,
located at 60 °C, 150–350 °C, and 650 °C, respectively. TGA experi-
ments in a nitrogen atmosphere were performed to assess the ther-
mal stability of the samples. The TGA/DTG curves of purified CNTs
3 4
O phase, as well as a rela-
tively sharp peak (b) between 20° and 30° due to the carbon
(Fig. S1). No XRD pattern for Co
3 4
O was observed for Co/CNT-DP
(Fig. 2c). This is probably because the cobalt particles on this cata-
lyst are too small to be detected by XRD (<4 nm) [28], as revealed
by TEM images (see below).
2
under N are shown in Fig. 3c. It can be seen that a ca. 20% weight
loss can be detected when the sample is heated to 900 °C. Three
weight loss peaks at 60 °C, 250 °C, and 700 °C were found in the
DTG curve.
Generally, a weight loss below 100 °C is considered to be due to
the release of absorbed water in the sample. A weight loss between
1
50 and 350 °C for the purified CNTs is attributed to the removal of
functional groups (–COOH and –OH) chemically bonded on the car-
bon nanotube surface. The weight loss observed for CNTs above
5
00 °C is ascribed to the loss of the non-graphitic carbon. These
TGA data under N show that the carbon nanotubes pretreated in
2
nitric acid contain functionalized groups [3,23].
For the uncalcined Co/CNT-IM catalyst using cobalt nitrate as a
precursor, TGA data under N
weight loss events take place at 50–150 °C, 200–400 °C, and 500–
00 °C. The first weight loss (50–150 °C) is attributed to the re-
moval of water, and the two weight loss peaks between 200 and
00 °C are attributed to the nitrate group decomposition of the co-
2
(Fig. 3d) reveal that four major
6
4
balt nitrate precursor. The fourth one (500–600 °C) is attributed to
the loss of non-graphitic carbon. The TGA curve of the uncalcined
Fig. 2. XRD patterns of calcined Co/CNT and Co/CS catalysts: (a) Co/CNT-IM; (b)
3 2
cobalt catalyst prepared from Co(OOCCH ) is shown in Fig. 3e.
CoA/CNT-IM; (c) Co/CNT-DP; (d) Co/CS-C-DP; (e) Co/CS-C-IM; (f) CoA/CS-C-IM,
CSs; b: CNTs; Ã: Co
a:
3 4
O .
As can be seen, there are two weight loss peaks between 50 and

3
0
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
Fig. 3. TGA curves of pristine and functionalized carbon nanotubes and cobalt catalysts under N
2
or air: (a) pristine CNTs under air; (b) functionalized CNTs under air; (c)
; (f) uncalcined Co/CNT-DP in N
functionalized CNTs under N ; (d) uncalcined Co/CNT-IM under N ; (e) uncalcined CoA/CNT-IM under N
2
2
2
2
.
4
00 °C as well as a weight loss peak located at ca. 530 °C. The peaks
be encapsulated in the unpurified carbon nanotubes (circles in
Fig. 4a). The average pore size and outer diameter of the raw car-
bon nanotube are ca. 9 and 29 nm, respectively. Further, it is noted
that some of the tips of the raw CNTs are opened (arrows in
Fig. 4b). Fig. 4c and d shows the low-resolution and high-resolution
TEM images of purified CNTs, respectively. As can be seen in
Fig. 4d, the high-resolution TEM image of purified CNTs reveals
that a tubular graphene structure is present parallel to the pore
direction and a curved graphene structure is present perpendicular
to the pore direction.
between 50 and 400 °C can be assigned to the removal of water
and the decomposition of acetate ions. The weight loss of the pre-
pared Co/CNT-DP under N is shown in Fig. 3f. Three significant
2
weight loss peaks are found in the DTG curve, located at 70, 310,
and 540 °C, respectively. Similarly, these can be attributed to the
loss of water, nitrate decomposition, and the loss of non-graphitic
carbon.
Typical TEM images of unpurified and purified CNTs are shown
in Fig. 4. As can be seen, some particles (Fe–Co/CaO) were found to

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
31
TEM images were recorded after Co had been added to the
prepared CNTs. The TEM images and histograms of the relative
cobalt particle size of passivated Co/CNT catalyst (pretreated in
heating rate has a significant effect on the cobalt particle size of
the catalysts. The passivated Co/CNT-IM pre-reduced at 300 °C in
2
H exhibited cobalt particles with an average particle size of
H
2
or N
precursors are shown in Figs. 5–9. The Co/CNT-IM catalyst was
reduced in either H or N at different temperatures to explore
2
) using different preparation methods and different cobalt
6.4 nm (Fig. S2a). Increasing the pre-reduction temperature to
400 °C, the average particle size of the cobalt declined to 3.9 nm
(Fig. 5). This change is due to a different temperature ramping rate.
It is interesting to note that the cobalt particle size increased mark-
2
2
the effect of pretreatment method and temperature on the Co
particle size.
2
edly when the Co/CNT-IM catalyst was reduced in H at 480 °C. As
Typical TEM images of Co/CNT-IM reduced in H
2
at different
observed in Fig. S2b, the cobalt particle sizes are >15 nm and it
temperatures (300, 400, and 480 °C) are shown in Fig. 5 (400 °C)
and in Fig. S2 (300 and 480 °C). The pre-reduction temperature/
seems that the cobalt particles have sintered due to the high-
temperature reduction in H .
2
Fig. 4. Typical TEM images of carbon nanotubes: (a) showing the residual metal particle in CNTs (red circles); (b) showing some opened tips (red arrows); (c) purified carbon
nanotubes; (d) high-resolution image of purified carbon nanotubes showing the structure of both the wall and edge end. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
0
d (Co )=3.9 nm
25
20
15
10
5
0
0
1
2
3
4
5
6
7
8
9
10
Cobalt particle size (nm)
0
2
Fig. 5. Typical TEM image and cobalt particle size histogram of passivated Co/CNT-IM catalyst after reduction in H at 400 °C for 20 h (d(Co ) = 3.9 nm).

3
2
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
20
15
10
5
0
0
d (Co )=12.4 (nm)
0
2
4
6
8
10 12 14 16 18 20
Cobalt particle size (n)
0
Fig. 6. Typical TEM image and cobalt particle size histogram of passivated Co/CNT-IM pretreated in N
2
at 480 °C for 20 h (d(Co ) = 12.4 nm).
30
25
20
15
10
5
0
0
d (Co )=3.0 nm
0
1
2
3
4
5
6
7
8
9
10
Cobalt particle size (nm)
0
Fig. 7. Typical TEM image and cobalt particle size histogram of passivated Co/CNT-DP after reduction in H
2
at 400 °C for 20 h (d(Co ) = 3.0 nm).
0
d (Co )=4.4 (nm)
2
2
1
1
5
0
5
0
5
0
0
1
2
3
4
5
6
7
8
9
10
Cobalt particle size (nm)
0
2
Fig. 8. Typical TEM image and cobalt particle size histogram of passivated CoA/CNT-IM after reduction in H at 400 °C for 20 h (d(Co ) = 4.4 nm).
The morphology of cobalt particles of the passivated Co/CNT-IM
pretreated in N at different temperatures (400 and 480 °C) has
also been investigated, as shown in Fig. S3 and Fig. 6. From
Fig. S3, only a small number of cobalt particles could be detected
on the CNTs and this suggested that not all the cobalt oxide has
been completely reduced by the carbon. When the Co/CNT-IM
2
was pretreated in N at 480 °C for 20 h (Fig. 6), cobalt particles with
2
an average size of 12.4 nm were present on the CNT surface.
The effect of catalyst preparation method (impregnation and
deposition precipitation) on the cobalt particle size of Co/CNT
was studied. A typical TEM image of the passivated Co/CNT-DP pre-
2
treated in H at 400 °C is shown in Fig. 7. From the histogram, an

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
33
average cobalt particle size of 3.0 nm was measured. This is slightly
smaller than the cobalt particle size of 3.9 nm obtained for Co/CNT-
IM prepared by impregnation. Furthermore, the particle size distri-
bution was found to be much narrower and the Co particles
smaller than the cobalt particle size (8.0 nm) of similar Co/CNF cat-
alysts prepared by deposition precipitation studied by Bezemer
et al. [29].
(Fig. 10a). The relative DTG curve exhibits two loss weight peaks,
located at 100 and 700 °C, respectively. For carbon spheres func-
tionalized by nitric acid (CS-B), the TGA curve under N is shown
2
in Fig. 10b. It can be seen that ca. 6% weight loss is revealed when
the temperature was increased to 900 °C. The relative DTG curve
shows four significant weight loss peaks, which are located at 70,
270, 400, and 700 °C, respectively.
The effect of changing cobalt nitrate to cobalt acetate on the co-
balt particle size of the Co/CNT catalyst is shown in Fig. 8. An aver-
age cobalt particle size of ca. 4.4 nm was noted for CoA/CNT-IM,
which is slightly larger than the cobalt particle size of 3.9 nm ob-
tained for Co/CNT-IM. After pretreatment of CoA/CNT-IM in
The data can readily be rationalized according to the discussion
on CNTs (above) and a previous study [23]. Generally, a weight loss
below 100 °C is considered to be due to the loss of absorbed water
in the sample. A weight loss between 150 and 450 °C is attributed
to the removal of functional groups chemically bonded on the car-
bon spheres surface. The weight loss observed above 500 °C is as-
cribed to the loss of the non-graphitic carbon. The TGA data under
5
2
%H /Ar at 450 °C, the catalyst showed cobalt particles of around
1
6.6 nm (Fig. 9).
From Figs. 4–8, it can be seen that the cobalt particles are dis-
2
N (Fig. 10b) also shows that the carbon spheres pretreated in ni-
persed on both the interior and the exterior of the purified carbon
nanotubes. For the Co/CNT catalysts prepared by impregnation,
the majority of the cobalt particles (>70%) were dispersed on
the exterior of CNT, regardless of the precursors and pretreatment
methods (Figs. 5 and 6, 8 and 9). The catalyst activity is thus pre-
dominantly associated with Co particles outside the tube. This is
different from the result reported by Trépanier et al. who showed
that most of cobalt particles were distributed inside the tubes for
a 15 wt.% Co/CNT catalyst prepared by impregnation [18]. Thus, it
is suggested that the position of cobalt particles on CNTs for Co/
CNT catalysts is closely associated with the preparation method,
regardless of the cobalt precursors and catalyst pretreatment
methods.
tric acid contain functionalized groups.
2
For the Co/CS-C-DP catalyst, the TGA curve under N revealed
four weight loss peaks (Fig. 10e). Compared to the support, the
weight loss peak of the non-graphitic carbon for Co/CS-C-DP
shifted from 680 to 550 °C (Table S1). This is consistent with that
revealed for a similar Co/CNT catalyst.
Figs. 11 and 12 show typical TEM images of pristine carbon
spheres and the Co/CS catalysts. As can be seen, the carbon spheres
are round and smooth with an average diameter of 900 nm. It
should be mentioned that from TEM characterization (Fig. S4), no
significant change takes place when the carbon spheres are func-
tionalized using nitric acid at the different temperatures used in
this study. The high-resolution image (Fig. 11b) of a carbon sphere
reveals that it is composed of random curling graphitic flakes, sim-
ilar to those found in CNT where they are seen to be parallel to the
CNT pore. A Co/CS-C-IM catalyst were prepared by impregnation,
and cobalt particles with a size of 14.2 nm was found (Fig. S5). Co-
balt catalysts supported on functionalized CSs were also prepared
by deposition precipitation using urea hydrolysis. Typical TEM
3.1.2. Carbon spheres (CSs) and Co/CS catalysts
For Co/CS-C-DP, a very weak Co signal was detected from the
3 4
O
XRD measurement (Fig. 2d). This is due to the small cobalt particles
or a low cobalt loading for this catalyst. As revealed by XRD mea-
surements, the main cobalt phase in both the Co/CS-C-IM and CoA/
2
images of the passivated Co/CS pre-reduced in H are shown in
Fig. 12. As revealed in a previous study [23], different functionali-
zation methods can affect the concentration of surface groups on
CS-C-IM catalysts thermally treated in N
This is similar to that of Co/CNT catalysts. The peak due to carbon
20–30°) associated with CSs is more broadened than that of the
2 3 4
was Co O (Fig. 2e and f).
(
carbon materials. The passivated Co/CS-A-DP pre-reduced in H
at 400 °C shows cobalt particles with d = 45 nm (Fig. 12a), while
the passivated Co/CS-B-DP pre-reduced in H at 400 °C exhibited
cobalt particles with an average particle size of 12.6 nm
Fig. 12b). The passivated Co/CS-C-DP pre-reduced in H at
00 °C produced cobalt particles with d = 3.1 nm (Fig. 12c). Fur-
2
CNT sample. It is known that a carbon sphere is composed of ran-
dom curling graphitic flakes with a size of 1–10 nm [23], while car-
bon nanotubes consist of parallel graphene sheets but are typically
covered with flake-like material, unless annealed [26]. The above
2
(
4
2
data revealed that the main Co phase of calcined Co/CS is Co
TGA/DTG data were recorded on the raw carbon spheres, func-
tionalized carbon spheres, and cobalt catalysts under N and air,
and the data are shown in Fig. 10 and Table S1. When pristine
CSs are heated to 900 °C under N , a small weight loss is exhibited
at around 100 °C and finally a ca. 5% weight loss can be detected
3 4
O .
thermore, it was found that different reduction atmospheres had
different effects on the cobalt particle size (Fig. 12d). The Co/CS-
2
2
C-DP pretreated in 5%H /Ar at 480 °C showed cobalt particles with
a size of ca. 21 nm.
2
2
1
1
1
1
1
0
8
6
4
2
0
8
6
4
2
0
0
d (Co )=16.6 (nm)
0
4
8
12
16
20
24
28
Cobalt particle size (nm)
0
2
Fig. 9. Typical TEM image and cobalt particle size histogram of passivated CoA/CNT-IM reduced in 5%H /Ar at 450 °C for 20 h (d(Co ) = 16.6 nm).

3
4
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
Fig. 10. TGA curves of functionalized carbon spheres and Co/CS catalysts under N
under air; (e) uncalcined Co/CS-C-DP in N
2 2 2
or air: (a) pristine CSs under N ; (b) CS-B under N ; (c) pristine CSs under air; (d) CS-B
2
.
3
.1.3. Reduction of Co/CNT and Co/CS
The effect of pretreatment temperature in Ar (prior to TPR test-
ing) on the reduction behavior of the Co/CNT-IM was monitored by
-TPR under 5%H /Ar (Fig. 13). As can be seen, the TPR profile
Fig. 13a) of Co/CNT-IM after pretreatment at 250 °C has three
450 °C for 2 h (Fig. 13b). The intensity of this third reduction peak
also decreased with temperature (Fig. 13c and d). As revealed by
XRD, Co
mally treated in N
3 4
the reduction of Co O
3
O
4
is the main cobalt phase when Co/CNT-IM was ther-
. Thus, the first peak at 280 °C is attributed to
to CoO, and the second reduction peak at
H
(
2
2
2
reduction peaks, located at 280, 363, and 477 °C, respectively. An
increase in the pretreatment temperature under flowing Ar to
363 °C is attributed to the reduction of CoO to metal Co [14,30].
When compared to the TPR profile (Fig. 13e) of purified CNT sup-
port, the third reduction peak at 477 °C for the Co/CNT-IM can be
attributed to the catalytic decomposition of the support. When
the pretreatment temperature P450 °C, no cobalt oxide phase
4
50 °C prior to the TPR experiment reveals the disappearance in
both the first and the second reduction peaks. The third shoulder
peak at 477 °C became dominant when pretreatment occurred at

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
35
Fig. 11. TEM images of carbon spheres with (a) low resolution and (b) high resolution (scale bar: 5 nm).
was present as shown in the TPR profiles. The cobalt oxide was
either reduced to Co metal or transformed to another phase which
The TPR profiles of different Co/CS catalysts pretreated in a flow
of high-purity Ar are displayed in Fig. 16. The TPR profile of Co/CS-
C-IM (Fig. 16a) pretreated at 120 °C for 2 h showed three distinct
peaks at 150–230 °C, 230–300 °C, and 300–500 °C, respectively.
When Co/CS-C-IM was pretreated at 250 °C for 2 h (Fig. 16b), the
TPR pattern showed that the first peak had disappeared and two
reduction peaks at 276 °C, and between 300 and 600 °C were
found. On pretreatment of Co/CS-C-IM at 350 °C for 2 h (Fig. 16c),
the peak at 276 °C disappeared. According to the Co/N-CS study
cannot be reduced in H
Ar.
2
after a treatment temperature P450 °C in
The phase change was also monitored by in situ X-ray diffrac-
tion. Fig. 14 shows the in situ X-ray diffraction patterns of Co/
CNT-IM in an N
XRD patterns obtained in flowing N
phase observed at 30 °C is Co . When the reduction temperature
reached 400 °C, the diffraction peaks due to Co disappeared
completely and the diffraction peaks of CoO were present, indicat-
ing that the phase transformation of Co to CoO was complete. It
2
atmosphere with increasing temperature. The
2
revealed that the main cobalt
3 4
O
3
O
4
3 4
[14], the first peak is attributed to the reduction of Co O to CoO.
The second reduction peak is attributed to the reduction of CoO
0
O
3 4
to Co . Compared to the TPR profile of the CS support (Fig. 16f),
can be seen that the diffraction peaks of CoO disappeared com-
pletely when the temperature reached 480 °C, while diffraction
peaks associated with cobalt metal were observed. No indication
of new phases or complexes (e.g., cobalt carbides) was observed.
Based upon the above data, it is clear that cobalt oxide sup-
ported on carbon nanotubes can be reduced by the carbon support
at T 6 480 °C in an inert atmosphere to give Co that appears to be
the third reduction peak at 477 °C for Co/CS-IM is attributed to
the catalytic decomposition of the support. For CoA/CS-C-IM
(Fig. 16d), three reduction peaks are also found in the TPR profile.
The first reduction peak is at 314 °C, and the second reduction peak
was centered at 363 °C. The third reduction peak is at 540 °C. It is
noted that the TPR profile of Co/CS-DP shows only one reduction
peak, located at 382 °C. In comparison with Co/CS-C-IM and CoA/
CS-C-IM, the TPR profile of the cobalt catalyst supported on CSs
by the deposition precipitation method did not show the reduction
2
reduced under H . Recently, we found that cobalt oxide supported
on carbon spheres could also be reduced by the carbon support at
T 6 480 °C in an inert atmosphere [14].
3 4
of Co O to CoO, similar to that of Co/CNT-DP. This suggests that
the CoO is also the main cobalt phase for the Co/CS-DP catalyst.
The
p
-orbital axis vector analysis shows that the chemistry of
fullerene-type carbon relates directly to the strain of the graphitic
network [31]. Deviation from planarity causes the partial rehybrid-
ization from sp to sp3 and the pyramidalization of the carbon
atoms in curved aromatic systems. Carbon nanotubes and car-
bon spheres have a similar curved surface and sp -hybridized
carbon atoms. The interaction between the convex surface of the
3.2. Fischer–Tropsch synthesis
2
Fischer–Tropsch synthesis performance of the Co/CNT and Co/
CS catalysts was investigated in a fixed-bed micro-reactor at
225 °C and 0.8 MPa. It is believed that the active site for the cobalt
catalysts in Fischer–Tropsch synthesis is metallic cobalt. Thus, a
reduction step is indispensable in activating the catalysts. The
autoreduction of Co/CS catalysts has been explored in a previous
study [14], while the FTS performance of the autoreduced cobalt
catalysts supported on CNTs, reported here, permits a comparison
of the behavior of the two types of catalysts. The influence of the
cobalt particle size on the activity, TOF, and C5+ selectivity for the
Co/CNT and Co/CS catalyst has been compared.
2
3 4
carbon atoms and the oxygen in cobalt oxides (Co O and CoO)
could lead to weakened bonding strength within the cobalt oxides
and hence lead to the reduction occurring at lower temperatures.
The TPR profiles of other Co/CNT catalysts pretreated in a flow
of high-purity Ar are compared in Fig. 15. The TPR profile of a
Co/CNT-DP catalyst pretreated at 250 °C for 2 h shows a reduction
peak at 380 °C and a shoulder peak at 500 °C. Increasing the pre-
treatment temperature to 550 °C leads to the disappearance of
the peak at 380 °C, i.e., the cobalt phase in the catalyst has been re-
duced during the pretreatment process in Ar at 550 °C.
3.2.1. Autoreduction of cobalt catalysts
The TPR profile of the CoA/CNT-IM catalyst also shows three
reduction peaks, located at 287, 360 °C, and between 370 and
00 °C, respectively. The first peak is attributed to the reduction
As discussed above, cobalt oxide supported on both carbon
nanotubes and carbon spheres can be reduced by the carbon sup-
ports. A previous study has reported on the autoreduction of Co/CS
and the relative FTS performances of the catalysts [14]. In the pres-
ent study, data on the FTS activity of cobalt supported on CNTs pre-
treated in H and N are given. The Fischer–Tropsch synthesis of
2 2
the as-prepared Co/CNT-IM catalyst was investigated in a fixed-
bed micro-reactor (Table 2) after pretreatment in both high-purity
6
of Co to CoO. The second reduction peak is attributed to the
3 4
O
0
reduction of CoO to Co . The intensities of both the first and the
second reduction peaks are smaller, and the third reduction peak
is broadened relative to those of Co/CNT-IM.
3 4
A comparison of Fig. 15a and c indicated that no Co O is de-
tected when the Co is generated by deposition precipitation (Co/
CNT-DP).
H
2
and high-purity N
2
at different temperatures (300, 400, and
480 °C). It should be mentioned firstly that the Co/CNT-IM catalyst

3
6
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
40
35
30
25
20
15
10
0
d (Co )=45 (nm)
5
0
0
10 20 30 40 50 60 70 80 90 100
Cobalt particle size (nm)
2
0
5
0
5
0
0
d (Co )=12.6 (nm)
1
1
0
2
4
6
8
10 12 14 16 18 20
Cobalt particle size (nm)
35
30
25
20
15
10
5
0
0
d(Co )=3.1 nm
0
1
2
3
4
5
6
7
8
9
10
Cobalt particle size (nm)
0
20
d(Co )=21.0 nm
1
5
0
5
0
1
12
16
20
24
28
32
Cobalt particle size (nm)
0
Fig. 12. Typical TEM images and cobalt particle size histograms of passivated Co/CS: (a) Co/CS-A-DP after reduction in H
2
at 400 °C for 20 h (d(Co ) = 45 nm); (b) Co/CS-B-DP
0
0
after reduction in H
2
at 400 °C for 20 h (d(Co ) = 12.6 nm); (c) Co/CS-C-DP after reduction in H
2
at 400 °C for 20 h (d(Co ) = 3.1 nm); (d) Co/CS-B-DP after reduction in 5%H /Ar
2
0
at 450 °C for 20 h (d(Co ) = 21 nm).

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
37
o
3
63 C
o
4
77 C
o
2
80 C
a
b
c
d
e
1
00 200 300 400 500 600 700 800
o
Temperature ( C)
Fig. 13. TPR profiles for Co/CNT-IM catalyst and purified CNTs using different
pretreatment temperatures in a flow of high-purity Ar for 2 h: (a) 250 °C; (b)
Fig. 16. TPR profiles for CSs and different Co/CS catalysts after different pretreat-
ment temperatures in a flow of high-purity Ar for 2 h: (a) Co/CS-C-IM pretreated at
4
50 °C; (c) 550 °C; (d) 600 °C; (e) purified CNTs pretreated at Ar at 250 °C for 2 h.
120 °C for 2 h; (b) Co/CS-C-IM pretreated at 250 °C for 2 h; (c) Co/CS-C-IM
pretreated at 350 °C for 2 h; (d) CoA/CS-C-IM pretreated at 250 °C for 2 h; (e) Co/
CS-C-DP pretreated at 250 °C for 2 h; (f) CSs pretreated at 250 °C.
o
6
00 C
alysts pretreated with H
defined smaller cobalt particles (Fig. 4 and Fig. S2a). It is noted that
the big cobalt particles are formed during the H reduction process
prior to the FTS process for the passivated Co/CNT-IM catalyst pre-
treated in H at 480 °C (Figs. S2 and S7). The big cobalt particles
formed are due to reduction-induced sintering. Thus, the high CO
activity for the catalysts treated under H can be attributed to
the small cobalt particles and high cobalt dispersion. When the
Co/CNT-IM was pretreated in N , it was observed that the activity
of the catalyst increased with the increase in pretreatment temper-
ature from 400 to 480 °C. As demonstrated above, cobalt oxide can
be reduced by the CNT support and the higher CO conversion
should be attributed to the higher degree of catalyst reduction,
2
at both 300 and 400 °C showed well-
+
o
+
480 C
2
#
o
#
#
4
00 C
2
o
2
50 C
2
α
*
o
*
* *
60
3
0 C
2
*
2
0
30
40
50
70
80
90
2
Theta (degree)
Fig. 14. In situ X-ray diffraction patterns of Co/CNT-IM catalyst with a function of
2
resulting from the higher pretreatment temperature in N . It is also
0
temperature in N
2
,
a
: carbon nanotubes, Ã: Co
3
O
4
, #: CoO, +: Co .
proposed, similar to studies with Co/CS [14], that the autoreduced
catalyst resulted in the Co particles being ‘‘trapped’’ in cavities pro-
duced by the removal of surface carbon that was in contact with
the cobalt oxide species prior to reduction (Fig. S8). This point
can also be evidenced by the in situ XRD result which showed that
the peak intensity (20–30°) of CNTs became weak and the peak be-
came narrower after an increase in pretreatment temperature
a
b
o
468
C
(
Fig. S9).
The olefin/paraffin ratios in the FTS products are illustrated in
Fig. S10. For the Co/CNT-IM catalyst pretreated in H at 400 °C, a
low olefin-to-paraffin ratio was observed. For the Co/CNT-IM cata-
lyst pretreated in H at 480 °C, a high olefin/paraffin ratio was pres-
ent. The Co/CNT-IM catalysts pretreated in N at 400 and 480 °C
o
2
87.3 C
c
2
d
2
o
2
80 C
2
e
show comparable olefin/paraffin ratios. This correlates with the
1
00 200 300 400 500 600 700 800
CO conversion data.
o
Temperature ( C)
3.2.2. The effect of cobalt precursor, preparation methods, and cobalt
Fig. 15. TPR profiles for different Co/CNT catalysts pretreated in Ar: (a) Co/CNT-DP
pretreated at 250 °C for 2 h; (b) Co/CNT-DP pretreated at 550 °C for 2 h; (c) CoA/
CNT-IM pretreated at 250 °C for 2 h; (d) CoA/CNT-IM pretreated at 400 °C for 2 h;
particle size on the FTS performance
The FTS data for Co/CNT-IM pretreated under both H
2
and Ar at
at
00 °C gave the highest CO conversion. Thus, this pretreatment
different temperatures indicated that pretreatment under H
4
2
(
e) Co/CNT-IM pretreated at 250 °C for 2 h.
condition was chosen to investigate the FTS activity of the Co/
CNT and Co/CS catalysts prepared using different methods and dif-
ferent cobalt precursors. The FTS performance of Co/CNT and Co/CS
pretreated under both atmospheres showed a good stability in the
FTS reaction (Fig. S6). As can be seen, the data for the catalyst pre-
treated in H
creased when the pretreatment temperature was increased from
00 to 400 °C, while it decreased sharply when the temperature
2
revealed that the CO conversion of the catalyst in-
catalysts pre-reduced in H at 400 °C is shown in Table 3. It is noted
that the cobalt catalyst prepared from cobalt nitrate showed
slightly higher CO conversion than the cobalt catalyst prepared
from cobalt acetate.
2
3
was increased to 480 °C. A TEM image (Fig. S7) of the spent Co/
CNT-IM catalyst (i.e., after 200 h of FTS reaction) that had been pre-
The average cobalt particle size of Co/CNT and Co/CS catalysts
studied is summarized in Table 4. The cobalt time yield
(mol CO g Co s ) was calculated and plotted as average cobalt
2
treated with H at 480 °C showed that big metallic cobalt particles
À1
À1
(
>50 nm) had formed on the surface of the support, while the cat-

3
8
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
Table 2
Fischer–Tropsch synthesis performances of Co/CNT catalyst pretreated in high-purity H
and high-purity N
2
for 20 h.a
2
Sample
Pretreatment
CO conversion (%)
Product selectivity (mol%)
CH
4
C
2
C
3
C
4
C
5+
Co/CNT-IM
H
2
H
2
H
2
N
2
N
2
at 300 °C
at 400 °C
at 480 °C
at 400 °C
at 480 °C
12.5
25.9
3.7
3.4
13.3
11.1
30.6
2.9
3.5
8.9
0.5
1.4
0.2
0.1
0.6
1.3
2.9
0.5
0.3
1.7
1.4
3.1
0.5
0.3
1.8
85.8
62.1
95.9
95.8
87.0
a
À1
Reaction conditions: 225 °C, 0.8 MPa, CO/H
2
= 1:2, GHSV = 3840 h , FTS data were collected at steady state (200 h).
Table 3
Fischer–Tropsch synthesis performances of Co/CNT and Co/CS catalysts pre-reduced in high-purity H
a
2
at 400 °C for 20 h.
Sample
Cobalt loading (wt.%)
CO conversion (%)
Product selectivity (mol%)
CH
4
C
2
C
3
C
4
C
5+
Co/CNT-IM
CoA/CNT-IM
Co/CNT-DP
Co/CS-A-DP
Co/CS-B-DP
Co/CS-C-DP
CoA/CS-C-IM
Co/CS-C-IM
13.23
12.74
4.31
0.98
1.12
1.47
5.11
5.21
25.9
22.6
9.9
1.8
2.5
3.7
2.4
2.6
30.6
28.3
23.5
2.4
3.1
4.0
1.4
1.3
0.9
0.1
0.1
0.2
0.1
0.1
2.9
3.2
1.8
0.2
0.3
0.5
0.2
0.1
3.1
3.5
1.7
0.4
0.2
0.5
0.2
0.1
62.1
63.8
72.1
97.0
96.2
94.8
98.8
99.1
0.7
0.6
a
À1
Reaction conditions: 225 °C, 0.8 MPa, CO/H
2
= 1:2, GHSV = 3840 h , FTS data were collected at steady state (200 h).
Table 4
Cobalt particle sizes and Fischer–Tropsch synthesis selectivities of Co/CNT and Co/CS catalysts at 4% CO conversion.
a
Sample
Cobalt particle size (nm)
Product selectivity (mol%)
CH
TEM (Co0)
H
-chemisorption (Co0)b
C
C
C
4
C
5+
2
4
2
3
Co/CNT-IM
Co/CNT-IMc
Co/CNT-IM
3.9
6.4
12.4
4.4
16.6
3.0
45
12.6
21
3.1
14.2
4.5
7.1
14.3
5.6
19.0
4.8
62
14.7
28.3
4.8
4.1
3.4
3.5
3.9
3.0
6.0
1.9
2.8
2.7
4.0
1.5
0.3
0.2
0.1
0.1
0.2
0.3
0.1
0.2
0.3
0.2
0.1
0.4
0.4
0.3
0.2
0.2
0.6
0.2
0.3
0.3
0.5
0.1
0.3
0.4
0.3
0.2
0.1
0.5
0.1
0.2
0.3
0.5
0.1
94.9
95.6
95.8
95.6
96.5
92.6
97.7
96.5
96.4
94.8
98.2
d
CoA/CNT-IM
CoA/CNT-IMe
Co/CNT-DP
Co/CS-A-DP
Co/CS-B-DP
Co/CS-B-DPf
Co/CS-C-DP
Co/CS-C-IM
16.3
a
2 2
Reaction conditions: 225 °C; 0.8 MPa; CO/H = 1:2; FTS data were collected at steady state (200 h); catalyst was pre-reduced in high-purity H at 400 °C for 20 h unless
noted.
b
Co particle size was obtained assuming hemispherical crystallites.
c
d
e
f
2
Catalyst was pre-reduced in high-purity H at 300 °C for 20 h.
Catalyst was pretreated in high-purity N at 480 °C for 20 h.
2
Catalyst was pre-reduced in 5%H
Catalyst was pre-reduced in 5%H
2
/Ar at 450 °C for 20 h.
/Ar at 450 °C for 20 h.
2
particle size for all the cobalt catalysts. As shown in Fig. 17a, a clear
relationship between the cobalt time yield and cobalt particle size
is observed such that smaller cobalt particles result in a higher co-
balt time yield. A similar trend has been reported in a previous
cobalt particle sizes (Fig. 17b). As can be seen, the TOF is constant
for cobalt particles above 10 nm but decreased sharply for the co-
balt catalysts with smaller cobalt particles. Indeed, the impact of
cobalt particle size on turnover frequency of FTS has been investi-
gated by others [11,33–37]. Iglesia et al. [33] found that the intrin-
sic site reaction rate does not depend on the Co particle size in the
range of ca. 10–200 nm. Bezemer et al. [11] have suggested a non-
classical structure sensitivity for Co catalysts supported on inert
carbon nanofibers (Co/CNF). The TOF increases for particles that
are smaller than 6–8 nm continuously and remains invariant for
larger particles. This behavior differs from classical structure sensi-
tivity which is related to the metal surface crystallographic anisot-
ropy and becomes apparent only when the particle size is reduced
typically below 3–4 nm [38]. More recently, den Breejen et al. re-
ported that cobalt particles of 4.7 ± 0.2 nm are the most active in
2
study for two Co/SiO catalysts with different particle sizes [32].
Bezemer et al. however observed a volcano-like curve for Co/CNF
catalysts when the cobalt time yield was plotted as a function of
the cobalt particle size [11]. A possible explanation for the differ-
0
ence is that the aggregation of the Co crystallites occurred during
their FTS study and that the small cobalt particles aggregated more
significantly than the large cobalt particles, leading to the decrease
in cobalt time yield for the catalyst with small cobalt particles [32].
This did not occur in the present study.
The turnover frequency (TOF) of these cobalt catalysts based on
2
H chemisorption has also been calculated and plotted against

H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
39
2
5
2
2
2
1
1
8
4
0
6
2
8
4
Co/CNT
Co/CS
20
1
5
0
5
0
1
Co/CNT
Co/CS
0
5
10 15 20 25 30 35 40 45 50
0
5
10 15 20 25 30 35 40 45 50
Cobalt particle size (nm)
Cobalt particle size (nm)
Fig. 17. The influence of cobalt particle size on the activity and turnover frequency of both Co/CNT and Co/CS (Reaction conditions: CO/H
GHSV = 3840 h ).
2
= 1:2, 225 °C, 0.8 MPa,
À1
monoxide within the tube. The gas reaching the catalyst inside the
tube is thus richer in hydrogen than in the synthesis gas [40]. For
Co catalysts supported on non-porous carbon spheres, all cobalt
particles are dispersed on the outer surface of the support. The hydro-
gen concentration around the active metal cobalt is the same as
that in the reactor. Thus, less light hydrocarbons will be produced
for Co/CS during FTS, as observed.
A previous study on Co/
product selectivity of FTS can be affected by cobalt particle size
41]. A positive correlation between cobalt particle size (above
0 nm) and C5+ selectivity was found by investigating the influence
2 3
c-Al O catalysts reported that the
[
1
of cobalt particle size in the range 3–18 nm. The authors reported
that the maximum C5+ selectivity was located at 7–8 nm [42] and
that the C5+ selectivity was not only related to the cobalt particle
size but also to the choice of support.
Fig. 18. The influence of cobalt particle size on C5+ selectivity of Co/CNT and Co/CS
at 4% CO conversion by changing space velocity (Reaction conditions: CO/H = 1:2,
25 °C, 0.8 MPa). Black dotted line is the trendline for all catalysts.
4
. Conclusions
2
2
Carbon nanotubes and carbon spheres were prepared by chem-
ical vapor deposition. Co/CNT and Co/CS catalysts with cobalt par-
ticle size ranging from 3 to 45 nm were prepared by different
methods using different cobalt precursors. It was found that the
cobalt catalysts supported on CNTs and CSs can be autoreduced
by the supports in an inert atmosphere at ca. 480 °C. In both cases,
the cobalt catalysts that are autoreduced in an inert atmosphere
show better Fischer–Tropsch performance than those catalysts re-
the FT reaction (1 bar, 220 °C) [39]. Prieto and coworkers [37] stud-
ied the effect of cobalt particle size effect on the TOF of FTS
employing Co/ITQ-2 as model catalysts, and a similar relationship
between cobalt particle size and TOF was obtained.
The effect of cobalt particle size on the hydrocarbon selectivity
of FTS was also investigated. It has been reported that C5+ selectiv-
ity can be affected by the amount of water produced at different CO
conversions. To overcome this, the present study compared the ef-
fect of cobalt particle size on the C5+ at a low CO conversion (4.0%)
to minimize the effect of water. The cobalt particle size and C5+
selectivity are displayed in Table 4 and plotted in Fig. 18. As can
be seen, a positive relationship was observed between cobalt par-
ticle size and C5+ selectivity for both Co/CNT and Co/CS catalysts. The
change in C5+ selectivity with particle size seems reasonable be-
cause it is believed that bridge-type adsorbed CO is more easily
formed on large Co particles [28]. The bridge-type CO is much
more active than linear-type CO because it has a weaker C–O bond
and thus can be more easily dissociated to carbon and oxygen. An
duced in H at the same temperature. This lower activity of the co-
2
balt catalysts reduced in H2 when the temperature is higher than
400 °C is due to enhanced sintering. It is noted that the FT activity
using H reduction is better at T = 400 °C but reduction under N is
2
2
better at T = 480 °C.
Cobalt particle size has been found to have an impact on the co-
balt time yield, turnover frequency (TOF), and C5+ selectivity for
both Co/CNT and Co/CS catalysts. A direct relation between the co-
balt time yield and cobalt particle size is observed such that a
smaller cobalt particle results in higher cobalt time yield. The
TOF value for both Co/CNT and Co/CS was constant for cobalt par-
ticles above 10 nm and decreased sharply for the cobalt catalysts
with smaller cobalt particles. Finally, a positive relationship was
observed between cobalt particle size and C5+ selectivity for both
Co/CNT and Co/CS catalysts.
It is clear that the Co/CS and Co/CNT catalysts behave similarly
and any differences relate to the surface area and/or dispersion ef-
fects associated with the two different carbon supports. Since CSs
are easier to prepare than CNTs and contain no metal impurities,
they provide better model supports to explore cobalt–carbon
increased CH (1 6 x 6 3) concentration on a large Co particle can
x
enhance the chain growth and thus increase the C5+ selectivity.
Compared to the C5+ selectivity obtained for Co/CNT, the Co/CS
exhibited a slightly higher C5+ selectivity for a similar sized parti-
cle. For our CNT catalysts, active sites are situated predominantly
not exclusively outside the tube. Thus, a proportion of light hydro-
carbons obtained from the Co/CNT catalyst may be due to the
greater rate of diffusion of hydrogen compared with that of carbon

4
0
H. Xiong et al. / Journal of Catalysis 278 (2011) 26–40
interactions than CNTs. The disadvantage of using our CSs relates
to their low surface areas. Current studies are centered on generat-
ing CSs with larger surface areas to overcome this limitation.
[15] P. Serp, E. Castillejos, ChemCatChem 2 (2010) 41.
[
[
[
[
[
[
16] M.C. Bahome, L.L. Jewell, D. Hildebrandt, D. Glasser, N.J. Coville, Appl. Catal. A
87 (2005) 60.
17] L. Guczi, G. Stefler, O. Geszti, Zs. Koppány, Z. Kónya, E. Molnár, M. Urbán, I.
Kiricsi, J. Catal. 244 (2006) 24.
2
18] R.M. Malek Abbaslou, A. Tavassoli, J. Soltan, A.K. Dalai, Appl. Catal. A 367
Acknowledgments
(
2009) 47.
19] M. Trépanier, A. Tavasoli, A.K. Dalai, N. Abatzoglou, Appl. Catal. A 353 (2009)
193.
20] H. Zhang, C. Lancelot, W. Chu, J.P. Hong, A.Y. Khodakov, P.A. Chernavskii, J.
Zheng, D.G. Tong, J. Mater. Chem. 19 (2009) 9241.
This work was supported financially by the DST/NRF Centre of
Excellence in Strong Materials, a SPARC fellowship (HX) and the
University of the Witwatersrand. We thank M.K. Rayner and Prof.
D.G. Billing for XRD measurements.
21] Z.L. Wang, Z.C. Kang, J. Phys. Chem. 100 (1996) 17725.
[22] Z.C. Kang, Z.L. Wang, J. Mol. Catal. A: Chem. 118 (1997) 215.
[
[
[
[
[
23] H.F. Xiong, M. Moyo, M.A.M. Motchelaho, L.L. Jewell, N.J. Coville, Appl. Catal. A
88 (2010) 168.
24] A. Nieto-Márquez, D. Toledano, J.C. Lazo, A. Romero, J.L. Valverde, Appl. Catal. A
73 (2010) 192.
25] A. Nieto-Márquez, D. Toledano, P. Sánchez, A. Romero, J.L. Valverde, J. Catal.
69 (2010) 242.
26] J.J. Delgado, X. Chen, J.P. Tessonnier, M.E. Schuster, E. Del Rio, R. Schlögl, D.S.
Su, Catal. Today 150 (2010) 49.
27] S.D. Mhlanga, K.C. Mondal, R. Carter, M.J. Witcomb, N.J. Coville, S. Afr. J. Chem.
3
Appendix A. Supplementary material
3
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2010.11.010.
2
References
62 (2009) 67.
[
[
28] A. Moen, D.G. Nicholson, Chem. Mater. 9 (1997) 1241.
29] G.L. Bezemer, P.B. Radstake, V. Koot, A.J. van Dillen, J.W. Geus, K.P. de Jong, J.
Catal. 237 (2006) 291.
30] H.F. Xiong, Y.H. Zhang, K.Y. Liew, J.L. Li, J. Mol. Catal. A: Chem. 295 (2008) 68.
31] R.C. Haddon, Science 261 (1993) 1545.
[
[
[
[
[
1] M.S. Luo, H. Hamdeh, B.H. Davis, Catal. Today 140 (2009) 127.
2] A. Martínez, G. Prieto, J. Rollán, J. Catal. 263 (2009) 292.
3] W. Chen, Z.L. Fan, X.L. Pan, X.H. Bao, J. Am. Chem. Soc. 130 (2008) 9414.
4] A.K. Dalai, B.H. Davis, Appl. Catal. A 348 (2008) 1.
5] E. de Smit, I. Swart, J.F. Creemer, G.H. Hoveling, M.K. Gilles, T. Tyliszczak, P.J.
Kooyman, H.W. Zandbergen, C. Morin, B.M. Weckhuysen, F.M.F. de Groot,
Nature 456 (2008) 222.
6] T. Mochizuki, T. Hara, N. Koizumi, M. Yamada, Appl. Catal. A 317 (2007) 97.
7] A.Y. Khodakov, W. Chu, P. Fongarland, Chem. Rev. 107 (2007) 1692.
8] G. Jacobs, J.A. Chaney, P.M. Patterson, T.K. Das, B.H. Davis, Appl. Catal. A 264
[
[
[
32] G.Z. Bian, N. Fujishita, T. Mochizuki, W.S. Ning, M. Yamada, Appl. Catal. A 252
(
2003) 251.
[
[
[
[
33] E. Iglesia, S.L. Soled, R.A. Fiato, J. Catal. 137 (1992) 212.
[
[
[
34] S.W. Ho, M. Houlalla, D.M. Hercules, J. Phys. Chem. 94 (1990) 6396.
35] A. Barbier, A. Tuel, I. Arcon, A. Kodre, G.A. Martin, J. Catal. 200 (2001) 106.
36] T. Herranz, X.Y. Deng, A. Cabot, J.G. Guo, M. Salmeron, J. Phys. Chem. B 113
(
2004) 203.
(
2009) 10721.
[
9] B. Jongsomjit, J. Panpranot, J.G. Goodwin Jr, J. Catal. 215 (2003) 66.
[
[
[
37] G. Prieto, A. Martínez, P. Concepción, R. Moreno-Tost, J. Catal. 266 (2009) 129.
38] M. Boudart, Chem. Rev. 95 (1995) 661.
39] J.P. den Breejen, J.R.A. Sietsma, H. Friedrich, J.H. Bitter, K.P. de Jong, J. Catal. 270
[
[
10] J.L. Li, N.J. Coville, Appl. Catal. A 181 (1999) 201.
11] G.L. Bezemer, J.H. Bitter, H.P.C.E. Kuipers, H. Oosterbeek, J.E. Holewijn, X. Xu, F.
Kapteijn, A.J. van Dillen, K.P. de Jong, J. Am. Chem. Soc. 128 (2006) 3956.
12] J.P. den Breejen, P.B. Radstake, G.L. Bezemer, J.H. Bitter, V. Frøseth, A. Holmen,
K.P. de Jong, J. Am. Chem. Soc. 131 (2009) 7197.
(
2010) 146.
[
[
[
[
[
40] R.B. Anderson, W.K. Hall, A. Krieg, B. Seligman, J. Am. Chem. Soc. 71 (1949) 183.
41] Ø. Borg, S. Eri, E.A. Blekkan, S. Storsæter, H. Wigum, E. Rytter, A. Holmen, J.
Catal. 248 (2007) 89.
42] Ø. Borg, P.D.C. Dietzel, A.I. Spjelkavik, E.Z. Tveten, J.C. Walmsley, S. Diplas, S.
Eri, A. Holmen, E. Rytter, J. Catal. 259 (2008) 161.
13] J.C. Kang, S.L. Zhang, Q.H. Zhang, Y. Wang, Angew. Chem. Int. Ed. 48 (2009)
2
565.
[
14] H.F. Xiong, M. Moyo, M.K. Rayner, L.L. Jewell, D.G. Billing, N.J. Coville,
ChemCatChem 2 (2010) 514.