Loss of single-walled carbon nanotubes selectivity by disruption of the Co-Mo interaction in the catalyst

Article abstract of DOI:10.1016/j.jcat.2003.08.005

The Co-Mo system is a catalyst commonly employed in petroleum refining for hydrotreating processes. A study on the effects of the presence of sodium in the Co-Mo catalytic system for single-walled carbon nanotubes (SWNT) production was carried out. For the SWNT production, 0.5 g of calcined catalyst was placed in a horizontal tubular reactor, heated in H₂ flow < 500°C, and then in He flow ≤ 850°C. Subsequently, CO was introduced at 850 cc/min at 80 psi and kept under these conditions for 2 hr. At the end of each run, the system was cooled down in He flow. Significant differences were observed between the undoped and Na-doped catalytic systems. The presence of sodium favored the dispersion of molybdenum, which formed sodium molybdate species over the silica support. Simultaneously, the formation of these sodium molybdate structures hindered the interaction between cobalt and molybdenum, preventing the formation of the cobalt molybdate species. This disruption was detrimental to the SWNT selectivity of the catalyst, giving further proof to the link between the presence of this surface species and the selective SWNT growth.

Full text of DOI:10.1016/j.jcat.2003.08.005

Journal of Catalysis 221 (2004) 354–364
www.elsevier.com/locate/jcat
Loss of single-walled carbon nanotubes selectivity by disruption
of the Co–Mo interaction in the catalyst
José E. Herrera and Daniel E. Resasco ∗
School of Chemical Engineering and Materials Science, University of Oklahoma, 100 East Boyd Street, Norman, OK 73019, USA
Received 24 June 2003; revised 8 August 2003; accepted 11 August 2003
Abstract
In previous work, we reported the growth of well-structured single-walled carbon nanotubes (SWNT) by CO disproportionation on
Co:Mo/SiO catalysts. We have proposed that high SWNT selectivity can only be obtained when most of the Co in the sample is interacting
2
with Mo and forming a cobalt–molybdate surface layer. In this contribution, we report further evidence to the hypothesis by an investiga-
tion of Na-doped Co:Mo/SiO catalysts. The effect of the sodium addition has been investigated by Raman spectroscopy, ultraviolet–visible
2
diffuse reflectance spectroscopy, H temperature-programmed reduction, and X-ray photoelectron spectroscopy. The results are compared to
2
those obtained in a previous study on undoped CoMo/SiO catalysts. Significant differences are observed between the undoped and Na-doped
2
catalytic systems. It has been found that the presence of sodium favors the dispersion of molybdenum, which tends to form sodium molybdate
species over the silica support. At the same time, the formation of these sodium molybdate structures hinders the interaction between cobalt
and molybdenum, preventing the formation of the cobalt molybdate species. This disruption has been found to be detrimental to the SWNT
selectivity of the catalyst, giving further evidence to the link between the presence of this surface species and the selective SWNT growth.
2003 Elsevier Inc. All rights reserved.
Keywords: Single-walled carbon nanotubes; Raman; XPS; TPR; UV–vis/DRS; TPO; Co Mo catalyst; Na
1. Introduction
that among the various formulations investigated, Co–Mo
catalysts supported on silica gel and having low Co:Mo ra-
tios exhibited the best performance [6,7]. We have explained
this performance on the basis of a Co–Mo interaction that
is critical for the performance of the catalyst [8]. Separated,
these metals are not effective; they are either inactive (Mo
alone) or unselective (Co alone). The catalyst is only ef-
fective when both metals are simultaneously present with
a low Co:Mo ratio. Through a detailed characterization that
involved a variety of characterization techniques such as EX-
AFS, XANES, UV–vis/DRS, XPS, and DRIFTS of adsorbed
NO, we were able to explain the reasons for the high selec-
tivity of these catalysts [8].
Single-wall carbon nanotubes (SWNT) can be consid-
ered as one of the building blocks for nanoscale science
and nanotechnology. They exhibit exceptional chemical and
physical properties that have opened a vast number of po-
tential applications [1]. Extensive research work is now fo-
cused on the large-scale manufacture of these materials at
low cost since there is not yet an established method that
provides large quantities of SWNT of well-determined and
reproducible characteristics. A clear understanding of the
formation mechanism of SWNT is therefore a key issue for
the development of further advances in this topic.
Several SWNT growth mechanisms have been pro-
posed [2–5]; most of them involve the interaction of carbon
species with a metal (atom, cluster, solution, carbide, etc.),
which in turn depends on the method used to synthesize the
SWNT. Our group has focused on the disproportionation
of CO on bimetallic catalysts. We have previously reported
Interestingly, the Co–Mo system is a catalyst commonly
employed in petroleum refining for hydrotreating processes.
However, most studies on these catalysts have focused on
alumina-supported systems since alumina interacts with the
metals with the appropriate strength to generate the species
with the optimum hydrotreating activity in the sulfided state.
After many years of intense research by several groups the
structure of the sulfide Co–Mo catalysts is now known at
the atomic level [9–12]. Other characteristics of these cata-
*
Corresponding author.
E-mail address: resasco@ou.edu (D.E. Resasco).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2003.08.005
2003 Elsevier Inc. All rights reserved.

J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
355
lysts have also been investigated, such as the influence of
alkali dopants on their structure and activity [13–16]. There
is still some controversy about this issue in the literature,
although it seems clear that the presence of these dopants
predominantly affects the dispersion of molybdenum species
[16–20].
The UV–vis spectra of the calcined catalysts were record-
ed using a Shimadzu double beam spectrometer UV-2101
with an integrating sphere for diffuse reflectance. Barium
sulfate was used as reflectance standard. Several Mo and Co
compounds, including MoO3, Na2MoO4, (NH4)6Mo7O24,
and CoMoO4 were used as references. Before analysis, the
samples were dried in air at 120 ◦C.
Temperature-programmed reduction (TPR) experiments
were conducted passing a continuous flow of 5% H2/Ar over
approximately 30 mg of the calcined catalyst at a flow rate
of 10 cm3/min, while linearly increasing the temperature at
a constant heating rate of 8 ◦C/min. The hydrogen uptake
as a function of temperature was monitored using a thermal
conductivity detector, SRI Model 110 TCD.
X-ray photoelectron spectroscopy data were recorded on
a Physical Electronics PHI 5800 ESCA System with mono-
chromatic Al-Kα X-rays (1486.6 eV) operated at 350 W
and 15 kV with a background pressure of approximately
2.0 × 10−9 Torr. A 400-µm spot size and 58.7-eV pass en-
ergy were typically used for the analysis. Sample charging
during the measurements was compensated by an electron
flood gun. The electron takeoff angle was 45◦ with respect
to the sample surface. For each sample, the binding energy
regions corresponding to Si (95–115 eV), Mo (220–245 eV),
and Co (760–820 eV) were scanned. The binding energies
were corrected by reference to the C(1s) line at 284.8 eV.
A nonlinear Shirley-type background was used for the area
analysis of each peak. The fitting of the XPS spectra was
carried out with Gaussian–Lorentizian peaks, using the Mul-
tiPak software from Physical Electronics.
In the present contribution, we have turned our attention
to study the effect of the presence of sodium in the Co–
Mo catalytic system for the SWNT production. As noted
above, in our proposed model, the interaction of Co and Mo
is critical for a good catalyst performance. Therefore, the in-
troduction of an element such as sodium that will compete
with Co for the interaction with molybdenum is expected
to be detrimental for the catalytic selectivity. We have fol-
lowed a characterization strategy on the basis of the abun-
dant precedent literature and our own work to investigate this
multicomponent system. Based on the results of this charac-
terization, we have put together a schematic picture of the
doped catalyst, and built up an analogy with the undoped
Co–Mo system, which can be used to explain the variations
in selectivity toward SWNT observed when the catalyst is
doped with sodium.
2. Experimental
2.1. Catalysts preparation and pretreatment
A series of Na-doped and undoped silica-supported Co–
Mo catalysts was prepared. For the Na-doped catalysts dif-
ferent amounts of sodium were added to the SiO2 support
by incipient wetness impregnation using aqueous solutions
of NaNO3 of appropriate concentrations. The SiO2 support
obtained from Aldrich had an average pore size of 6 nm,
BET area 480 m2/g, pore volume 0.75 cm3/g, and particle
sizes in the range 70–230 mesh. After evaporation of water
at room temperature, the solid was dried at 110 ◦C overnight
and then calcined at 500 ◦C for 3 h. Subsequently, the Na-
doped and undoped supports were loaded with molybdenum
(4.6 wt%) and cobalt (1.4 wt%) by incipient wetness coim-
pregnation of a solution of cobalt nitrate and ammonium
heptamolybdate to achieve a Co:Mo molar ratio of 1:3. Af-
ter impregnation, the solids were dried overnight at 120 ◦C
and then calcined for 3 h at 500 ◦C in dry-air flow. Two
monometallic Na-doped catalysts were also prepared, con-
taining 1.3 wt% Co and 4.6 wt% Mo.
2.3. Production and characterization of carbon nanotubes
For the SWNT production, 0.5 g of calcined catalyst was
placed in a horizontal tubular reactor, heated in H2 flow
up to 500 ◦C, and then in He flow up to 850 ◦C. Subse-
quently, CO was introduced at a flow rate of 850 cm3/min
at 80 psia and kept under these conditions for 2 h. At
the end of each run, the system was cooled down in He
flow. The total amount of carbon deposits was determined
by temperature-programmed oxidation (TPO) following the
method described in our previous work [6]. Since Na has an
inherent catalytic effect on oxidation processes and might
affect the position of the peaks, before each TPO analysis
undoped catalyst samples were impregnated with a solution
of sodium nitrate to get a total loading of 4% Na and heated
in He to decompose the inorganic salt.
2.2. Catalyst characterization
Transmission electron microscopy (TEM) was used to
analyze the quality of the nanotubes obtained over the dif-
ferent catalysts. The TEM images were obtained in a JEOL
JEM-2000FX TEM. For this analysis, a suspension of the
carbon-containing samples was achieved by ultrasound stir-
ring the solid sample in isopropanol for 10 min. A few drops
of the resulting suspension were deposited on a grid and sub-
sequently evacuated before the TEM analysis.
The Raman spectra of calcined catalysts were obtained
in a Jovin Yvon-Horiba LabRam 800 equipped with an
air-cooled CCD detector equipped with a He–Ne laser
(632.8 nm). Typical laser powers ranged from 1.0 to 2.0 mW;
integration times were around 15 s for each spectrum; 10 Ra-
man spectra were averaged for each sample.

356
J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
The quality of the nanotubes was also examined by Ra-
man spectroscopy, one of the most powerful techniques
available to study the structural and electronic properties of
SWNT. The analysis of radial A1g breathing mode (below
300 cm−1) gives direct information about the tubes diame-
ter [21], while the analysis of the G band in the tangential
mode range, i.e., 1400–1700 cm−1, provides information on
the electronic properties of the nanotubes. In addition, the
analysis of the so-called D band at around 1350 cm−1 gives
an indication of the level of disordered carbon. Therefore,
we have used the intensity ratio of the D band (1350 cm−1
)
to the G band (1590 cm−1) as semiquantitative measurement
of the formation of undesirable forms of carbon [22,23].
3. Results
3.1. Characterization of the catalysts
3.1.1. Raman spectroscopy
Raman spectroscopy has shown to be an extremely pow-
erful tool for characterizing supported heterogeneous cata-
lysts containing d0 transition metal oxide species [17]. In
fact, substantial information about the dispersion of molyb-
denum oxide species over the high surface-area support has
been obtained from detailed analysis of Raman spectra [19,
24–28]. Those studies provide the basis for interpreting the
results of the present catalyst characterization. Fig. 1 shows
the Raman spectra of an undoped Co:Mo (1:3)/SiO2 and of
two Na-doped Co:Mo (1:3)/SiO2 catalyst; the Raman spec-
tra of reference samples of cobalt oxide (Co3O4), cobalt
molybdate (CoMoO4), and sodium molybdate (Na2MoO4)
are also shown in the figure.
Fig. 1. Raman spectra of an undoped and two Na-doped Co:Mo (1:3)/SiO
2
catalysts compared to (a) the spectra of two analytical samples of cobalt
molybdate (CoMoO ) and sodium molybdate (Na MoO ) and compared
4
2
4
to (b) the spectrum obtained for a reference sample of cobalt oxide Co O .
3
4
The laser excitation energy was 633 nm.
previous UV–vis data [8]. Moreover, a comparison of the
Raman spectrum of the undoped bimetallic catalyst with that
of the CoMoO4 gives strong support to the presence of this
type of species. Therefore, the Raman band at 880 cm−1, the
shoulder at 930 cm−1, and the broad band at 350 cm−1 must
all be assigned to Mo–O–Co stretching vibrations in cobalt
molybdate-like species [27,30–33].
A noticeable change in the Raman bands is observed
when sodium is present in the catalyst. First of all, when
inspecting the spectrum of the for the Na-doped Co:Mo
(1:3)/SiO2 catalysts, a new band at 889 cm−1 becomes
prominent. This band has been previously assigned to
Mo=O stretching of isolated monooxo (MoO4−2) spe-
cies [34]. Also, the band at 825 cm−1 (asymmetric Mo–O–
Mo stretching) shows a dramatic shift to lower wavenum-
bers (805 cm−1), which is consistent with a decrease in
the amount of Mo–O–Mo moieties on the catalysts when
sodium is present in the sample. It is also important to note
that the bands assigned to Co–O–Mo species present on the
undoped catalyst (350, 880, and 930 cm−1) become very
weak until they almost totally vanish for the catalyst with
the higher Na loading.
Fig. 1a compares the Raman spectrum of the undoped
and Na-doped catalysts with that of a CoMoO4 reference
sample. The peaks appearing in the 750–1000 cm−1 re-
gion should be ascribed to Mo species. This is the region
where the bands for stretching of Mo–O bonds are typically
observed [17]. Several factors such as metal loading, pH,
precursor, impregnation method, and impurities are known
to influence the exact position of the Raman bands. How-
ever, a preliminary assignment can be done based on the
preceding literature and the comparison with the reference
samples. For instance, in the case of the undoped catalyst
the peak ce−nt6ered at 943 cm−1 has been previously ascribed
to Mo7O24 species while the shoulder at 960 cm−1 has
been associated with larger clusters of MoOx species, such
−4
as Mo8O26 [17,27,28]. This assignment is in good agree-
ment with our previous UV–vis spectroscopy results [8],
which suggested the presence of well-dispersed clusters of
MoOx. According to the calculated band energy gap, the
average domain size of these species was estimated to be
slightly larger than that of Mo7O24−6. At the same time, the
presence of large polymeric MoOx species must be ruled
out when there are no peaks in the 990–1000 cm−1 region
[20,29], a conclusion that we also indicated based on the
Fig. 1b shows a comparison of the Raman spectra of the
Na-doped catalysts with the spectrum of the cobalt oxide
reference. The assignments for the Raman active phonon

J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
357
modes that are displayed in the spectrum of Co3O4 are as
follows. The peaks at 199, 621, and 681 cm−1 correspond
to F2g phonon modes, while the peaks at 477 and 525 cm−1
are Eg and A1g phonon modes, respectively [35]. As seen in
Fig. 1b, this set of peaks is undoubtedly present in the spec-
trum of the 4% Na-doped catalyst, while it is entirely absent
in the case of the undoped catalyst. From this evidence, it is
clear that segregated Co3O4 species are not present on the
undoped catalyst while the addition of Na seems to favor the
formation a new segregated cobalt oxide phase.
erence series (top panel) decrease as the domain size in-
creases. An analogous behavior is observed for the case of
the Co:Mo (1:3)/SiO2 catalysts (bottom panel). An increase
in the gap energy values is observed when the amount of
Na in the catalysts increases. While for the undoped cata-
lyst the energy gap lies between the values corresponding to
(NH4)6Mo7O24 and MoO3, for the doped catalysts the en-
ergy gap shifts to higher values, indicating a decrease in the
average size of the molybdenum domains. Indeed, for the
catalyst with the higher loading of Na this value shifts up−to2
4.2 eV, which is very close to the value obtained for MoO4
tetrahedral species. In agreement with the results obtained by
Raman spectroscopy it is clear that the dispersion of oxidic
molybdenum species is deeply influenced by the presence of
sodium.
3.1.2. Diffuse reflectance UV–visible spectroscopy
(UV–vis/DRS)
We have used UV–vis/DRS to study the state of both Mo
and Co in the fresh catalyst. In order to estimate the band en-
ergy gap of d0 oxides, it has been recommended to use the
square root of the Kubelka–Munk function multiplied by the
photon energy, and to plot this new function versus the pho-
ton energy [36]. The position of the absorption edge can then
be determined by extrapolating the linear part of the rising
curve to zero. The values thus obtained carry information
about the average domain size of the oxide nanoparticles.
It has been shown that the energy band gap decreases as
the domain size increases [36,37]. Therefore, a comparison
can be made between the energy of the samples under in-
vestigation and those of references of known domain size.
This comparison is made in Fig. 2, which shows the absorp-
tion edges of several MoOx species together with those of
two different Na-doped and an undoped Co:Mo (1:3)/SiO2
catalysts. As expected, the band gap energies in the ref-
In addition to the charge-transfer bands due to Mo, ap-
pearing in the UV region, the visible spectra of these cat-
alysts exhibit bands in the 500–750 nm region, which are
associated with Co species and are ascribed to d–d tran-
4
4
4
sitions (4T2g to A2g and T2g to T1g (P)) of high spin
octahedral Co complexes [38,39]. Fig. 3 shows the DRS
spectra in this region for two Na-doped and an undoped
Co:Mo (1:3)/SiO2 catalysts. The spectrum for the undoped
catalyst is very similar to that of CoMoO4, in which Co is in
an octahedral environment. The identical shape of the spec-
tra of both samples indicates that the same type of ligand
and symmetry around Co are common for the undoped cata-
lyst and the cobalt molybdate. By contrast, the shape of the
spectrum of the heavily doped catalyst is markedly different
Fig. 2. Lower panel: UV absorption spectra for (a) an undoped Co:Mo
Fig. 3. Visible spectra obtained for an undoped and two Na-doped Co:Mo
(1:3)/SiO catalysts, (b) 2.0 wt% Na-doped Co:Mo (1:3)/SiO catalyst, and
(1:3)/SiO catalysts compared to the spectra of a reference sample of cobalt
2
2
2
(c) a 4.0 wt% Na-doped Co:Mo (1:3)/SiO catalyst. Upper panel: UV ab-
molybdate (CoMoO ). The spectrum obtained for a Co/SiO catalyst is also
2
4
2
sorption spectra for three molybdenum oxide species of known domain size.
included for comparison.

358
J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
and exhibits two new bands at around at 525 and 670 nm.
These bands are in turn the characteristic features of the
pure Co catalyst (Fig. 3 upper panel) and can be associated
with Co3O4 species, which, as previously shown, are present
in the fresh monometallic Co catalyst [8]. Therefore, it can
be concluded that the addition of Na to the catalyst breaks
up the interaction between Mo and Co in such a way that
free Co oxide starts appearing. This conclusion is in prefect
agreement with the Raman spectroscopy results.
sible to use the reduction profiles to identify the presence
of Co and Mo species in the absence of mutual interactions.
Accordingly, the TPR of the undoped Co:Mo (1:3)/SiO2 cat-
alyst indicates that, in this sample, the vast majority of Co
oxide species are interacting with Mo. While most of the
Co in the monometallic catalyst gets reduced below 500 ◦C,
almost no reduction takes place below this temperature in
the undoped bimetallic catalysts. It has been proposed that
the addition of Mo oxide to Co oxide inhibits the reduction
of the Co species because Mo6+ polarizes the Co–O bonds,
making them more ionic and consequently more difficult to
reduce [40]. In agreement with the Raman and UV–vis/DRS
data, TPR indicates that a high degree of Co–Mo interaction
is observed for the undoped Co:Mo (1:3)/SiO2 catalyst.
A contrasting behavior is observed in Fig. 4b. In this case,
the TPR profiles of the Na-doped Co:Mo (1:3)/SiO2 catalyst
are compared to the undoped catalyst. As shown in Fig. 4b,
a gradually increasing fraction of segregated Co species as
a function of the sodium content is apparent from the peak
growing at 400 ◦C which is associated with the reduction
of noninteracting Co oxide [41]. The observed trend agrees
with the results obtained by UV–vis/DRS and Raman spec-
troscopy on the Na-doped catalysts. It is obvious that the
fraction of cobalt interacting with Mo decreases with in-
creasing Na content.
3.1.3. Temperature-programmed reduction
The reduction profiles of the undoped and two Na-doped
Co:Mo (1:3)/SiO2 catalysts in the precalcined state, to-
gether with those of the precalcined Co/SiO2 and Na-doped
Mo/SiO2, catalysts, are shown in Fig. 4a. The TPR profile
of the precalcined Co catalyst shows two peaks at 360 and
445 ◦C, which can be ascribed to the reduction of Co oxide
species. On the other hand, the reduction profile of the Na-
doped Mo catalysts exhibits three different peaks that appear
at much higher temperatures than those of Co. It is then pos-
3.1.4. X-ray photoelectron spectroscopy (XPS)
XPS studies were performed on the air-calcined Na-
doped and undoped catalysts to get an insight into the chem-
ical state of the catalyst constituents. Fig. 5 shows the Co 2p
spectra of the undoped and doped Co:Mo (1:3)/SiO2 cata-
lysts compared to the spectra obtained for the monometallic
Co/SiO2, CoMoO4, and Co3O4. As summarized in Table 1,
the binding energies for the bimetallic catalysts seem to de-
crease as the Na content increases, which is the trend that
one would expect if CoMoO4 species in the catalyst are con-
verted into Co3O4 species [40,42]. However, the values are
too close to be conclusive. A more convincing evidence for
this conversion is obtained by analyzing the shake-up satel-
lite peaks appearing at 5–6 eV higher binding energies than
the main peak in the Co 2p spectrum. The relative intensity
of these satellite peaks has often been used for the identifi-
cation of the cobalt species [43–45] since the intensities of
the shake-up peaks satellites associated with the Co 2p3/2 of
both oxides are quite different. As shown in Fig. 5, this satel-
lite peak is very small in the spectrum of Co3O4 but very
strong in the spectrum of cobalt molybdate, as previously
observed [45,46]. Therefore, it is easy from the features of
the Co 2p spectra to determine whether the Co species are
present as cobalt molybdate or as cobalt oxide (Co3O4) on
the catalyst. A clear trend is observed in Table 1 for the rel-
ative intensities. The intensity ratio for the undoped catalyst
is high (0.6) and close to that of stoichiometric CoMoO4, but
gradually approaches the value of cobalt oxide (0.3) as the
Na content increases.
Fig. 4. (a) TPR profiles of a Co/SiO and Na-doped Mo/SiO catalysts com-
2
2
pared to the undoped Co:Mo (1:3)/SiO bimetallic catalyst. (b) TPR profiles
2
obtained for an undoped and two Na-doped Co:Mo (1:3)/SiO catalysts. In
2
all cases the reduction was conducted under 5% H /Ar, using a heating lin-
2
◦
ear ramp of 8 C/min.

J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
359
Fig. 5. XPS Co 2p spectra of an undoped and two Na-doped Co:Mo (1:3)/SiO catalysts. The spectrum obtained for a monometallic Co/SiO catalyst is also
2
2
included for comparison. The spectra obtained for two reference samples of cobalt oxide Co O and CoMoO are also presented.
3
4
4
Table 1
Co 2p binding energies for undoped and Na-doped catalysts
Catalyst
Co 2p
peak
Co 3p
satellite peak
Intensity ratio
3/2
3/2
position (eV)
position (eV)
(Co 2p /satellite)
3/2
0% Na–Co:Mo (1:3)/SiO
2% Na–Co:Mo (1:3)/SiO
4% Na–Co:Mo (1:3)/SiO
780.9
780.6
779.5
780.5
779.6
780.7
785.5
786.5
786.5
787.3
788.2
785.9
0.60
0.58
0.33
0.34
0.30
0.79
2
2
2
Co/SiO
2
Co O (reference)
3
4
CoMoO (reference)
4
The position of the Co 2p
satellite peak and its relative intensity is also indicated. The values obtained for two analytical samples of cobalt oxide Co O
3
1/2
4
and CoMoO are also presented.
4
Fig. 6 shows the Mo 3d spectra of the doped and un-
doped Co:Mo (1:3)/SiO2 catalysts. The values of the bind-
observed on the linewidth of the spectra. The linewidth of
the Mo3d peaks obtained for the sample without sodium is
wider than that for the Na-doped sample. This effect has
been previously reported in the literature and has been at-
tributed to the presence of monomeric MoO4−2 species [45].
ing energies of the two spin orbit components (3d5/2
232.0 eV; 3d3/2 = 234.7 eV) are characteristic of Mo(VI)
=
species [47–49]. However an interesting difference can be

360
J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
Fig. 7. Temperature-programmed oxidation profiles of all the carbonaceous
species present in an undoped and two Na-doped Co:Mo (1:3)/SiO cata-
2
◦
lysts after CO disproportionation at 850 C for 2 h. The arrow indicates the
center of the peak corresponding to the oxidation of SWNT. The total yield
of carbon is also indicated.
doped Co:Mo (1:3)/SiO2 catalysts for SWNT production.
The reaction temperature employed for the CO dispropor-
tionation after a prereduction step was kept at 850 ◦C. At
the end of a 2-h reaction period, the spent catalyst contain-
ing the carbon deposits was cooled down in He flow. The
characterization of the carbon deposits was done by way of
temperature-programmed oxidation, transmission electron
microscopy, and Raman spectroscopy with varying excita-
tion energy, three techniques that we have previously used
and tested [6–8,52].
We have shown that from the TPO analysis one can obtain
a quantitative measurement of the carbon yield and selectiv-
ity toward SWNT [6]. We have previously shown that under
the TPO conditions and while still immersed on this particu-
lar catalyst, the SWNT are oxidized in a relatively narrow
temperature range (maximum around 570 ◦C), which lies
below the temperature in which MWNT, graphite, and car-
bon fibers are oxidized, but above the temperature at which
amorphous and chemically impure carbon species are oxi-
dized [7].
Fig. 6. XPS Mo 3d spectra of an undoped and two Na-doped Co:Mo
(1:3)/SiO catalysts. The spectra were fitted using Gaussian–Lorentizian
2
functions to resolve the individual 3d
and 3d
contributions.
5/2
3/2
On the other hand, it is well known that a number of
alkali metal oxides act as catalysts for the oxidation of car-
bon by gaseous oxidants [50]. These processes are generally
thought to involve oxygen transfer steps in which the cat-
alyst participates in a redox cycle between stoichiometric
and substoichiometric oxide species [51]. As a consequence,
one should expect that the presence of sodium can affect
the oxidation process. Fig. 7 shows the TPO profile of the
undoped CoMo catalyst in which sodium was added after
SWNT synthesis in order to account for the presence of the
alkali metal in the TPO. The postaddition of Na only results
in a shift from 570 ◦C as observed without Na [6] to 450 ◦C.
This shift clearly demonstrates the catalytic influence of Na
in the oxidation process. However, it is important to point
out that the shape or number of peaks has not been affected.
Therefore, it is still possible to obtain information about the
Even though this observation cannot be conclusive by itself,
it perfectly agrees with the results obtained by Raman and
UV–vis spectroscopy, which indicate that the dispersion of
oxidic molybdenum species is deeply influenced by the pres-
ence of sodium.
3.2. Production and characterization of single-walled
carbon nanotubes by catalytic disproportionation of CO
As we have reported in previous papers [6,8], the silica-
supported Co–Mo system displays a very high selectivity in
the production of single-wall nanotubes by CO dispropor-
tionation. Interestingly, the high yields and selectivities to
SWNT were only obtained after a specific sequence of cal-
cination at 500 ◦C in air and reduction in H2 at 500 ◦C. In the
present work, we compare the activity of undoped and Na-

J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
361
ence is observed on the carbon structures produced with
an undoped CoMo catalyst and with a heavily Na-doped
sample. While the sample obtained over the undoped cata-
lyst exhibited a high density of SWNT, the sample obtained
with the 4% Na-doped catalyst mainly produced defective
MWNT and graphitic carbon deposits. In the first case, one
can observe the typical parallel lattice fringes characteristic
of SWNT, while in the second case, a spacing of 0.34 nm
can be observed between the fringes, which is the character-
istic value for the lattice space between individual walls of
graphitic fibers [53].
4. Discussion
Fig. 8. (a) Raman spectra of the carbon deposits obtained by disproportiona-
In previous reports we have discussed the characteriza-
tion results obtained on the undoped Co:Mo system used for
SWNT production [8]. The most important conclusions of
such studies were that the extent of the Co–Mo interaction
strongly depended on the Co:Mo molar ratio in the catalyst
and that this interaction had diverse forms during the vari-
ous stages of the catalyst life. For instance, we were able to
establish that in the calcined state, Mo was in the form of a
well-dispersed Mo(VI) oxide while the state of Co strongly
depended on the Co:Mo ratio. At low Co:Mo ratios, it in-
teracted with Mo forming a superficial Co molybdate-like
structure. At high Co:Mo ratios, it forms a noninteracting
Co3O4 phase. During the subsequent reduction treatment in
hydrogen, the noninteracting Co phase reduced to metallic
Co, while the Co molybdate-like species remained as well-
dispersed Co2+ ions.
We demonstrated that the effect of having Co stabilized
in the Co-molybdate environment was critical for a good
performance during the nanotube growth. First of all, this
stabilization avoided the sintering of Co into large metal-
lic aggregates, which generate undesired forms of carbon,
as it occurs when the noninteracting cobalt oxide phase gets
reduced before nanotube growth. By contrast, in the selec-
tive catalyst, Co clusters are only formed after the Mo oxide
has become Mo carbide, thus breaking the Co–Mo interac-
tion. Under the reaction conditions (high temperature and
CO pressure), CO dissociates on the nascent Co clusters and
carbon atoms begin to nucleate over Co and generate the
nanotube cap that is the precursor of the single-walled nan-
otube.
◦
tion of CO at 850 C over an undoped and two Na-doped Co:Mo (1:3)/SiO
catalysts after a pretreatment at 500 C in hydrogen. The laser excitation en-
ergy was 633 nm.
2
◦
different kinds of carbon species from the TPO profiles. For
instance, Fig. 7 illustrates the strong influence of the pres-
ence of sodium in yield and selectivity during the SWNT
growth. As noted above, the TPO profile of the product ob-
tained over the undoped Co:Mo (1:3)/SiO2 catalyst displays
a single TPO peak centered at 450 ◦C. A contrasting behav-
ior is observed in the TPO profile of the material obtained
over the 4% Na-doped catalysts (see Fig. 7). Instead of the
single TPO peak observed in the previous case, three oxi-
dation peaks are obtained on this sample. Although it is not
possible to make a direct assignment of the different carbon
species that originate each of the different peaks in the TPO
profile based solely on the TPO data [52], this result clearly
indicates a decrease in the selectivity to SWNT originated
by the presence of sodium in the catalyst.
Raman spectroscopy is another useful technique for eval-
uating the structure of carbon nanotubes. As noted above,
from the position of the radial breathing mode band, direct
information about the tubes diameter can be obtained [21]
while the relative intensity of the D band relative to the
G band is widely used as qualitative measurement of the
formation of undesirable forms of carbon [22,23]. Fig. 8
shows typical Raman spectra obtained on the carbon de-
posits formed on the undoped and doped Co:Mo (1:3)/SiO2
catalysts. Depending on the amount of sodium present on the
catalysts a pronounced variation is observed in the relative
intensity of the D band. As a matter of fact Fig. 8 shows that
the SWNT material obtained over the undoped catalyst is of
high quality, while the samples obtained in both Na-doped
catalysts show much lower quality, which undoubtedly indi-
cates the presence of undesirable forms of carbon. In agree-
ment with the results obtained by TPO analysis, it is evident
that the presence of Na has a negative impact on the selec-
tivity of the Co:Mo (1:3)/SiO2 catalysts.
The characterization results for the Na-doped Co–Mo
catalyst series reported in this work can be put together in
an analogous picture. For instance the Raman, UV–vis/DRS,
and XPS data clearly indicate that the dispersion of molyb-
denum has been strongly affected by the presence of sodium
in the catalyst. Furthermore, the development of new bands
in the Raman spectra as well as the values of the gap ener-
gies (Fig. 2) clearly indicates that molybdenum is interacting
with the Na dopant forming sodium molybdate moieties in
the catalyst. In fact, it has been argued that alkali ion impu-
rities such Na+ or K+ have higher affinity for molybdenum
The TEM observations fully support the results of the
other techniques. As shown in Fig. 9, a contrasting differ-

362
J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
(a)
(b)
Fig. 9. (a) TEM images showing SWNT produced by CO disproportionation over an undoped Co:Mo (1:3)/SiO catalyst. (b) TEM images showing a mixture
2
of MWNT and graphite produced by CO disproportionation over a 4% Na-doped Co:Mo (1:3)/SiO catalyst.
2
species than the silica support [19,54], although still some
disagreement remains on this particular topic [17]. Another
plausible explanation could be that the presence of alka-
line ions results in an increase of the surface pH values at
the point of zero charge (PZC). Thus when Na is present,
large molybdenum oxide polyanions become less favorable
to monomeric species [55]. In either case, it is clear that the
presence of sodium deeply changes the dispersion of molyb-
denum species.
When these sodium molybdate monomeric species are
formed, the interaction of molybdenum and cobalt is hin-
dered. In fact the Raman, UV–vis/DRS, TPR, and XPS re-
sults clearly show that a segregated phase of cobalt oxide
(Co3O4) is formed. In this context, it is important to note
that a Na excess is needed to break the Co–Mo interaction in
the cobalt molybdate phase. For example, the UV–vis/DRS
data clearly show that a relative large amount of sodium
(i.e., 4 wt%) must be added to the catalyst before a nonin-
teracting Co oxide phase begins to form. The Raman spectra
(Fig. 1b) together with the results obtained from X-ray pho-
toelectron spectroscopy (Fig. 5) undoubtedly show that this
noninteracting phase is Co3O4. These results are schemati-
cally represented in Fig. 10.
dency to generate defective MWNT, carbon filaments, and
graphite nanofibers. As noted above, when Mo is present
in the catalyst and there is no excess of free Co, a well-
dispersed Co+2 species in the form of a Co molybdate-like
phase is stabilized. From the detailed studies conducted over
the Na-doped Co–Mo system reported in this contribution,
we can propose that the disruption by sodium of the cobalt–
molybdenum interaction is responsible for the loss of selec-
tivity toward the formation of SWNT in the catalyst. The
driving force for this disruption is either the tendency of
Na to form sodium molybdate or the change of the surface
pH values. The disruption of the Co–Mo interaction causes
the segregation of cobalt into a new Co3O4 phase. As a
consequence, the Na-doped catalyst shows a low selectiv-
ity because the ability of a Co–Mo catalyst to form SWNT
depends on the formation of the surface cobalt molybdate
species.
5. Conclusions
We can summarize the main findings of the present work
in the following conclusions:
In previous reports [8], we have proposed that the role of
Co is the activation of CO, while the role of Mo is the sta-
bilization of Co(II) ions. When Co is not interacting with
Mo, in the reduced state it sinters and forms large metal
aggregates. These large metallic Co aggregates have the ten-
1. The presence of sodium deeply affects the dispersion
of molybdenum. In the undoped catalyst, molybdenum
forms small aggregates with domain size close to that of

J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
363
Fig. 10. Schematic description of the structure of the Co:Mo (1:3)/SiO catalyst with and without sodium as derived from the characterization methods.
2
heptamolybdate species while in the case the doped cat-
alyst monomeric sodium molybdate species are formed.
2. When sodium molybdate species are formed, the inter-
action between molybdenum and cobalt is disrupted and
a new phase of segregated Co3O4 is formed.
[11] S.M.A.M. Bouwens, R. Prins, V.H.J. de Beer, D.C. Koningsberger,
J. Phys. Chem. 94 (1990) 3711.
[12] S.M.A.M. Bouwens, J.A.R. van Veen, D.C. Koningsberger, V.H.J. de
Beer, R. Prins, J. Phys. Chem. 95 (1991) 123.
[13] A.M. Venezia, V. La Parola, A. Longo, A. Martorana, J. Solid State
Chem. 161 (2001) 373.
3. During the reduction pretreatment, the noninteracting
Co3O4 phase is reduced to metallic Co, which easily
sinters at the nanotube growth conditions. Large Co par-
ticles generate defective MWNT, carbon filaments, and
graphite nanofibers under reaction conditions, originat-
ing a loss in selectivity of the Co–Mo catalyst.
[14] A.M. Venezia, F. Raimondi, V. La Parola, G. Deganello, J. Catal. 194
(2000) 393.
[15] G. Muralidhar, F.E. Massoth, J. Shabtai, J. Catal. 85 (1984) 44.
[16] G. Xiong, Z. Feng, J. Li, Q. Yang, P. Ying, Q. Xin, C. Li, J. Phys.
Chem. B 104 (2000) 3581.
[17] G. Mestl, T.K.K. Srinivasan, Catal. Rev.-Sci. Eng. 40 (1998) 451.
[18] N.F.D. Verbruggen, L.M.J. von Hippel, G. Mestl, B. Lengeler, H.
Knoezinger, Langmuir 10 (1994) 3073.
[19] M.A. Bañares, N.D. Spencer, M.D. Jones, I.E. Wachs, J. Catal. 146
(1994) 204.
[20] C.C. Williams, J.G. Ekerdt, J.M. Jehng, F.D. Hardcastle, A.M. Turek,
I.E. Wachs, J. Phys. Chem. 95 (1991) 8781.
Acknowledgments
[21] A.M. Rao, E. Richter, S. Bandow, B. Chase, P.C. Eklund, K.A. Wil-
liams, S. Fang, K. Subbaswamy, M. Menon, A. Thess, R.E. Smalley,
G. Dresselhaus, M.S. Dresselhaus, Science 275 (1997) 187.
[22] S. Rols, A. Righi, L. Alvarez, E. Anglaret, R. Almairac, C. Journet,
P. Bernier, J.L. Sauvajol, A.M. Benito, W.K. Maser, E. Munoz, M.T.
Martinez, G.F. de la Fuente, A. Girard, J.C. Ameline, Eur. Phys. J.
B 18 (2000) 201.
This research was conducted with financial support from
the Department of Energy, Office of Basic Energy Sciences
(Grant DE-FG03-02ER15345).One of us (J.E.H.) thanks the
Fulbright-CAREC program for a scholarship. We acknowl-
edge Walter E. Alvarez and Leandro Balzano for helpful
discussions.
[23] S. Bandow, S. Asaka, Y. Saito, A.M. Rao, L. Grigorian, E. Richter,
P.C. Eklund, Phys. Rev. Lett. 80 (1998) 3779.
[24] M.A. Vuurman, I.E. Wachs, J. Mol. Catal. 77 (1992) 29.
[25] A.R. Saini, B.G. Johnson, F.E. Massoth, Appl. Catal. 40 (1988) 157.
[26] B. Vielhaber, H. Knözinger, Appl. Catal. 26 (1986) 375.
[27] F.R. Brown, L.E. Makovsky, K.H. Rhee, J. Catal. 50 (1977) 162.
[28] V. La Parola, G. Deganello, C.R. Tewell, A.M. Venezia, Appl. Catal.
A 235 (2002) 171.
[29] R. Radhakrishnan, C. Reed, S.T. Oyama, M. Seman, J.N. Kondo, K.
Domen, Y. Ohminami, K. Asakura, J. Phys. Chem. B 105 (2001) 8519.
[30] Z. Li, Y. Fu, J. Bao, M. Jiang, T. Hu, T. Liu, Y.-N. Xie, Appl. Catal.
A 220 (2001) 21.
[31] G.M. Clark, W.P. Doyle, Spectrochim. Acta 22 (1966) 1441.
[32] F. Trifiro, P. Centola, I. Pasquon, J. Catal. 10 (1968) 86.
[33] W. Kuang, Y. Fan, K. Chen, Y. Chen, J. Catal. 186 (1999) 310.
[34] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordi-
nation Compounds, third ed., Wiley, New York, 1986.
[35] V.G. Hadjiev, M.N. Iliev, I.V. Vergilov, J. Phys. C 21 (1988) L199.
[36] D.G. Barnton, M. Shtein, R.D. Wilson, S.L. Soled, E. Iglesia, J. Phys.
Chem. B 103 (1999) 630.
References
[1] B.I. Yakobson, R.E. Smalley, Am. Scientist 85 (1997) 324.
[2] Y. Saito, T. Nakahira, S. Uemura, J. Phys. Chem. B 107 (2003) 931.
[3] B.Q. Wei, R. Vajtai, P.M. Ajayan, Carbon 41 (2003) 185.
[4] J. Gavillet, A. Loiseau, F. Ducastelle, S. Thair, P. Bernier, O. Stephan,
J. Thibault, J.-C. Charlier, Carbon 40 (2002) 1649.
[5] J.-C. Charlier, S. Iijima, Top. Appl. Phys. 80 (2001) 55.
[6] B. Kitiyanan, W.E. Alvarez, J.H. Harwell, D.E. Resasco, Chem. Phys.
Lett. 317 (2000) 497.
[7] W.E. Alvarez, B. Kitiyanan, A. Borgna, D.E. Resasco, Carbon 39
(2001) 547.
[8] J.E. Herrera, L. Balzano, A. Borgna, W.E. Alvarez, D.E. Resasco,
J. Catal. 204 (2001) 129.
[9] J.V. Lauritsen, S. Helveg, E. Lægsgaard, I. Stensgaard, B.S. Clausen,
H. Topsøe, F. Besenbacher, J. Catal. 197 (2001) 1.
[10] H. Topsøe, B.S. Clausen, Catal. Rev. Sci. Eng. 26 (1984) 395.

364
J.E. Herrera, D.E. Resasco / Journal of Catalysis 221 (2004) 354–364
[37] R.S. Weber, J. Catal. 151 (1995) 470.
[46] Y. Okamoto, T. Imanaka, S. Teranishi, J. Catal. 65 (1980) 448.
[47] F. Solymosi, J. Cserényi, A. Szöke, T. Bánsági, A. Oszkó, J. Catal. 165
(1997) 150.
[48] A. Katrib, P. Leflaive, L. Hilaire, G. Maire, Catal. Lett. 38 (1996) 95.
[49] L. Coulier, V.H.J. de Beer, J.A. R van Veen, J.W. Niemantsverdreit,
Top. Catal. 13 (2000) 99.
[38] H. Kasper, Monatsh. Chem. 98 (1967) 2104.
[39] P. Gajardo, P. Grange, B. Delmon, J. Phys. Chem. 83 (1979) 1771.
[40] S. Halawy, M. Mohamed, G. Bond, J. Chem. Tech. Biotechnol. 58
(1993) 237.
[41] J. Li, G. Jacobs, Y. Zhang, T. Das, B.H. Davis, Appl. Catal. A 223
(2002) 195.
[50] D.W. McKee, Carbon 25 (1987) 587.
[51] H.Y. Huang, R.T. Yang, J. Catal. 185 (1999) 286.
[52] J.E. Herrera, D.E. Resasco, Chem. Phys. Lett. 376 (2003) 302.
[53] T.W. Ebbesen, P.M. Ajayan, Nature 358 (1992) 220.
[54] M.A. Bañares, H. Hu, I.E. Wachs, J. Catal. 150 (1994) 407.
[55] R.J. Hunter, in: R.H. Ottewill, R.L. Rowell (Eds.), Zeta Potential in
Colloid Science, Principle and Application in Colloid Science, Acad-
emic Press, London, 1988, p. 233.
[42] R. Kleyna, H. Mex, M. Voß, D. Borgmann, L. Viscido, J.M. Heras,
Surf. Sci. 433 (1999) 723.
[43] R.L. Chin, D.M. Hercules, J. Phys. Chem. 86 (1982) 3079.
[44] F.B. Noronha, C.A. Perez, R. Frety, M. Schmal, Phys. Chem. Chem.
Phys. 1 (1999) 2861.
[45] A.M. Venezia, V. La Parola, G. Deganello, D. Cauzzi, C. Leonardi,
G. Predieri, Appl. Catal. A 229 (2002) 261.