Relationship between the structure/composition of Co-Mo catalysts and their ability to produce single-walled carbon nanotubes by CO disproportionation

Article abstract of DOI:10.1006/jcat.2001.3383

A series of analytical techniques have been employed to fully characterize the structure and chemical state of Co-Mo/SiO₂ catalysts used for the production of single-walled carbon nanotubes (SWNT) by CO disproportionation at 700-850°C. The state of Co and Mo on a series of silica-supported catalysts was investigated using extended X-ray absorption fine structure, X-ray absorption near-edge spectroscopy, ultraviolet-visible diffuse reflectance spectroscopy, H₂ temperature-programmed reduction, X-ray photoelectron spectroscopy, and diffuse reflectance Fourier transform infrared spectroscopy of adsorbed NO after two sequential pretreatments and after the production of SWNT under pure CO. It was found that the selectivity of the Co-Mo catalysts toward SWNT strongly depends on the stabilization of Co species in a nonmetallic state before exposure to CO, which results from an interaction with Mo. The extent of this interaction is a function of the Co:Mo ratio and has different forms during the different stages of the catalyst life. From the detailed characterization conducted over the catalyst series it is concluded that after calcination, Mo is in the form of a well-dispersed Mo(6+) oxide while Co is either interacting with Mo in a superficial Co molybdatelike structure (at low Co:Mo ratios) or as a noninteracting Co₃O₄ phase (at high Co:Mo ratios). After a subsequent treatment in hydrogen, the noninteracting phase is reduced to metallic Co, while the Co molybdate-like species remain as well-dispersed Co²⁺ ions. During the production of SWNT under pure CO, the Mo oxide species are converted into Mo carbide. This conversion disrupts the interaction between Co and Mo and results in the release of metallic Co in the form of extremely small clusters, which are responsible for the production of SWNT. By contrast, large Co clusters that are formed from the noninteracting Co phase produce the nonselective forms of carbon (multiwalled nanotubes, filaments, graphite, etc.).

Full text of DOI:10.1006/jcat.2001.3383

Journal of Catalysis 204, 129–145 (2001)
doi:10.1006/jcat.2001.3383, available online at http://www.idealibrary.com on
Relationship between the Structure/Composition of Co–Mo Catalysts
and Their Ability to Produce Single-Walled Carbon Nanotubes
by CO Disproportionation
Jose E. Herrera, Leandro Balzano, Armando Borgna,1 Walter E. Alvarez, and Daniel E. Resasco2
School of Chemical Engineering and Materials Science, University of Oklahoma, 100 E. Boyd St., Norman, Oklahoma 73019
Received March 28, 2001; revised August 10, 2001; accepted August 10, 2001
the various alternatives investigated in the last few years
for production of SWNT, the catalytic decomposition of
carbon-containing molecules appears to be a promising
technique since it has the potential to be scaled-up at rela-
tively low cost. A number of researchers have investigated
different catalyst formulations and operating conditions for
this process (2, 3). Our group has focused on the dispropor-
tionation of CO on several bimetallic catalysts, which ex-
hibited a high selectivity toward the production of SWNT
at relatively low temperatures. Among the various formu-
lations investigated, Co–Mo catalysts supported on silica
gel and having low Co : Mo ratios exhibited the best perfor-
mance (4).
The Co-Mo system is well-known in the field of catal-
ysis due to its application in hydrotreating catalytic pro-
cesses. In that case, however, silica is not the most suit-
able support, due to its weak interaction with the active
components. Most studies on the Co–Mo catalysts have
focused on alumina-supported systems, since alumina in-
teracts with Co and Mo with the appropriate strength to
generate HDS active species. For that reason, alumina-
supported Co-Mo catalysts are used in industry in the form
of sulfides. Although the structure of the sulfided Co-Mo
catalysts is known almost at the atomic level (5–8), that of
the nonsulfided oxidic precursor has received less attention
(9–12). Topsøe and Topsøe (13), using infrared (IR) spec-
troscopy of adsorbed NO showed that the interaction be-
tween Mo and the alumina in the oxidicstate wasnot greatly
affected by the presence of Co. A similar conclusion was
reached by Ratnasami and Kno¨zinger (9), who used IR and
ultraviolet–visible (uv–vis) diffuse reflectance spectroscopy
(DRS) to infer that in the oxidic state, a major portion of
the Co is inside the alumina lattice in a tetrahedral environ-
ment of oxygen ions and is not exposed to the gas phase.
Silica-supported Co-Mo, which is the system of our inter-
est, displays a different behavior from that of the alumina-
supported catalysts (13–18). The differences have been at-
tributed to the lower affinity of silica for the Co and Mo
ions, thus facilitating the Co-Mo interaction. The exact na-
ture of this interaction is still unclear. While some authors
A series of analytical techniques have been employed to fully
characterize the structure and chemical state of Co–Mo/SiO2 cata-
lysts used for the production of single-walled carbon nanotubes
(SWNT) by CO disproportionation at 700–850 C. The state of
Co and Mo on a series of silica-supported catalysts was investi-
gated using extended X-ray absorption fine structure, X-ray absorp-
tion near-edge spectroscopy, ultraviolet–visible diffuse reflectance
spectroscopy, H2 temperature-programmed reduction, X-ray pho-
toelectron spectroscopy, and diffuse reflectance Fourier transform
infrared spectroscopy of adsorbed NO after two sequential pretreat-
ments and after the production of SWNT under pure CO. It was
found that the selectivity of the Co–Mo catalysts toward SWNT
strongly depends on the stabilization of Co species in a nonmetallic
state before exposure to CO, which results from an interaction with
Mo. The extent of this interaction is a function of the Co : Mo ratio
and has different forms during the different stages of the catalyst
life. From the detailed characterization conducted over the catalyst
series it is concluded that after calcination, Mo is in the form of a
well-dispersed Mo(6+) oxide while Co is either interacting with Mo
in a superficial Co molybdatelike structure (at low Co : Mo ratios)
or as a noninteracting Co3O4 phase (at high Co : Mo ratios). After a
subsequent treatment in hydrogen, the noninteracting phase is re-
duced to metallic Co, while the Co molybdate-like species remain as
well-dispersed Co2+ ions. During the production of SWNT under
pure CO, the Mo oxide species are converted into Mo carbide. This
conversion disrupts the interaction between Co and Mo and results
in the release of metallic Co in the form of extremely small clusters,
which are responsible for the production of SWNT. By contrast,
large Co clusters that are formed from the noninteracting Co phase
produce the nonselective forms of carbon (multiwalled nanotubes,
filaments, graphite, etc.).
c
2001 Academic Press
1. INTRODUCTION
The single-wall carbon nanotubes (SWNT) exhibit ex-
ceptional chemical and physical properties that have
opened a vast number of potential applications (1). Among
1 Permanent address: INCAPE, Universidad Nacional del Litoral,
Santiago del Estero 2654, Santa Fe, Argentina.
2 To whom correspondence should be addressed. E-mail: resasco@
ou.edu.
129
0021-9517/01 $35.00
c
Copyright
2001 by Academic Press
All rights of reproduction in any form reserved.

130
HERRERA ET AL.
have postulated that the interacting phase is different from pregnation of aqueous ammonium heptamolybdate and Co
cobalt molybdate (13, 15, 16), others have argued that a nitrate solutions, had Co : Mo molar ratios of 2 : 1, 3 : 4, 1 : 2,
CoMoO4 phase is indeed present (14, 17, 18).
and 1 : 3. In this series, the amount of Mo was kept constant
As mentioned above, we have found that the silica- for all catalysts at 4.6 wt% , while the amount of Co was var-
supported Co-Mo system displays very high selectivity in ied accordingly. Monometallic catalysts were prepared with
the production of SWNT by CO disproportionation (19). It loadings of 1.4 wt% Co and 4.6 wt% Mo, respectively. The
is interesting to note that the high yields and selectivities SiO2 support obtained from Aldrich had an average pore
to SWNT were only obtained after a specific sequence of size of 6 nm, BET area of 480 m2/g, pore volume of
calcination at 500 C in air and reduction in H2 at 500 C, 0.75 cm3/g, and particle sizes in the range 70–230 mesh.
followed by heating in He to 700 C. Therefore, we have After impregnation, the solids were dried overnight at
focused our attention on the characterization of the differ- 120 C and then calcined for 3 h at 500 C in flowing dry
ent phases and surface species present in this system af- air.
ter each of these two pretreatments and after the pro-
The catalysts were investigated in three different forms,
duction of SWNT. In this contribution, we have used a the oxidic state, the reduced state, and after reaction. In the
variety of characterization techniques, including extended oxidic state, the catalysts were investigated directly after the
X-ray absorption fine structure (EXAFS), X-ray absorp- calcination in air at 500 C. Those in the reduced state were
tion near-edge spectroscopy (XANES), UV–vis DRS, first calcined in air at 500 C, then reduced step in H2 flow
H2-TPR (temperature-programmed reduction) X-ray pho- for 1 h at 500 C, and finally heated in He flow to 700 C.
toelectron spectroscopy (XPS), and diffuse reflectance This reduction/heating pretreatment is the one used here
Fourier transform infrared (DRIFTS) of NO adsorbed, for the production of SWNT, which has proven to be the
which provide complementary information.
most effective (4).
We have designed our characterization strategy on the
basis of the abundant precedent literature on the appli-
cation of these techniques to investigate the Co-Mo sys-
tem. For example, UV–visspectroscopyhasbeen previously
2.2. Catalyst Characterization
The UV–vis spectra of the solid samples were recorded
used to estimate the domain size of transition metal oxides using a Shimadzu double beam spectrometer UV-2101
(mostly d0 ions) by analyzing the absorption edge energy with an integrating sphere for diffuse reflectance. Bar-
according to the method proposed by Weber (20). Simi- ium sulfate was used as reflectance standard. Several
larly, the IR spectroscopy of adsorbed NO has been widely Mo and Co compounds, including MoO3, Na2MoO4,
used to characterize the Co-Mo catalytic system because (NH4)6Mo7O24 4H2O, and -CoMoO4, were used as ref-
the spectra of adsorbed NO exhibit prominent absorption erences. Before each analysis, the samples were dried in air
bands, which appear at different frequencies, depending on at 120 C.
whether the adsorption is on Mo or Co sites (11, 12, 21–
The IR spectra of adsorbed NO were obtained on a
26). Therefore, this technique can be used to determine Bio-Rad FTS 40 spectrometer, equipped with a diffuse re-
whether one of the components is blocking the other. TPR, flectance cell (Harrick Scientific Co., Praying Mantis) with
in combination with XPS and X-ray absorption spectro- an in situ reaction chamber. Before the spectrum was ac-
scopies can be used to determine the oxidation state of the quired, the catalysts were reduced ex situ under H2 flow
components and to fully describe the structure of the cata- at 500 C for 1 h and then heated in He to 700 C using a
lyst after the different treatments. The use of a combination ramp temperature of 10 C/min. Then they were cooled to
of techniques has been very successful in the study of sul- room temperature in He flow and transferred to the IR cell.
fided Co : Mo catalysts, corroborating the CoMoS structure To eliminate any superficial oxidation caused by exposure
first proposed by Topsøe and co-workers (5, 7, 27–29). In to air during the transfer, each sample was re-reduced
this paper, we have used a combination of these techniques in situ for 1 h at 500 C in H2 flow, purged in He flow at
to put together a detailed picture ofthe structure ofthe cata- that temperature, and then cooled to room temperature.
lyst, which can be used to explain the variations in selectiv- Once cooled, the samples were exposed to 3% NO in He
ity toward SWNT observed when the catalyst formulations for 30 min at room temperature and purged in He for
or pretreatment conditions are changed.
30 min.
The X-ray absorption data were obtained at the National
Synchrotron Light Source (NSLS) at Brookhaven National
Laboratory, using beam line X-18B equipped with a Si(111)
crystal monochromator. The X-ray ring at the NSLS has an
energy of 2.5 GeV and ring current of 80–220 mA. The cal-
2. EXPERIMENTAL
2.1. Catalyst Preparation and Pretreatment
A series of mono and bimetallic Co-Mo catalysts cined, reduced, and spent samples were investigated by X-
supported on silica was prepared by incipient wetness ray absorption. Both reduced and spent samples were not
impregnation. The bimetallic samples, prepared by coim- exposed to air but directly transferred from the reaction

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
131
chamber to a He glove bag, where they were wrapped in temperature was monitored using a thermal conductivity
Kapton tape and stored in He-purged sealed veilsuntilanal- detector, SRI model 110 TCD.
ysis. The EXAFS experiments were conducted in a stain-
2.3. Production and Characterization
of Carbon Nanotubes
less steel sample cell at liquid nitrogen temperature. Six
scans were recorded for each sample. The average spec-
trum was obtained by adding the six scans. The pre-edge
background was subtracted by using power series curves.
Subsequently, the postedge background was removed us-
ing a cubic spline routine. The spectra were normalized
by dividing by the height of the absorption edge. To ob-
tain structural parameters, theoretical references for Co–
Co, Co–O, Mo–O, Mo–C, Mo–Mo, and Co–Mo bonds were
obtained by using the FEFF and FEFFIT fitting programs
from the University of Washington (30–32). In this routine,
the Debye–Waller factors for each bond type ( ), the edge
energy difference (1Eo), the coordination number N, and
the difference in bond distances (1R) with respect to the
theoretical reference were used as fitting parameters. The
quality of the fit was determined using the r-factor, which
gives a sum-of-squares measure of the fractional misfit.
Therefore, the smaller the r-factor, the better the fit is. For
good fits, the r-factor is always less than or about 3% . The
spectra of MoO3, Na2MoO4, (NH4)6Mo7O24, Mo2C, CoO,
Co3O4, and -CoMoO4 , as well as Co and Mo foils were
also obtained at liquid nitrogen temperature and used as
references.
In this contribution, the production of SWNT by CO dis-
proportionation is compared in a series of catalysts with
Co : Mo ratios of 2 : 1, 1 : 2, and 1 : 3. A more extensive com-
parison has been done in a previous contribution (19). In
the present work, we have selected catalysts that display
contrasting behavior in selectivity. For the SWNT produc-
tion, 0.5 g of calcined catalyst was placed in a horizontal
tubular reactor, heated in H2 flow to 500 C, and then in
He flow to 700 C. Subsequently, CO was introduced at a
flow rate of 850 cm3/min at 84 psia and kept under these
conditions for a given period of time, which ranged from 3
to 120 min. At the end of each run, the system was cooled
in He flow. The total amount of carbon deposits was de-
termined by temperature programmed oxidation (TPO)
following the method described in our previous work (4).
Transmission electron microscopy (TEM) and Raman spec-
troscopy were used for characterizing the carbon deposits
on the catalyst. The TEM images were obtained in a JEOL
JEM-2000FX TEM. For this analysis, a suspension of the
carbon-containing samples in isopropanol was achieved by
stirring the solid sample with ultrasound for 10 min. Then,
a few drops of the resulting suspension were deposited on a
grid and subsequently evacuated before the TEM analysis.
The Raman spectra were obtained at Nicolet Instruments
Corp., on a Nicolet Almega Dispersive Raman spectrome-
ter with a 532 nm laser.
XPS data were recorded on a Physical Electronics PHI
5800 ESCA System with monochromatic AlK X rays
(1486.6 eV) operated at 350 W and 15 kV with a back-
9
ground pressure of approximately 2.0 10
Torr. A
400- m spot size and 58.7 eV pass energy were typically
used for the analysis. Sample charging during the measure-
ments was compensated for by an electron flood gun. The
electron takeoff angle was 45 with respect to the sample
surface. The pretreatment of the samples was performed in
a packed bed reactor with an on/off valve at each end of the
reactor, which allowed for a quick isolation of the samples
3. RESULTS
3.1. Characterization of the Calcined Catalysts
3.1.1. UV–vis DRS. We have used UV-vis-DRS to study
after each treatment. The reactor with the sample under the state of both Mo and Co in the oxidic form, after cal-
He was transferred to a glove bag; the sample (in pow- cination in air at 500 C. In order to estimate the band en-
der form) was placed on a stainless steel holder and kept ergy gap of the Mo oxide compounds, Barnton et al. rec-
in a vacuum transfer vessel (model 04-110A from Physi- ommended using the square root of the Kubelka–Munk
cal Electronics) to avoid any exposure to the atmosphere function multiplied by the photon energy, and plotting this
before the analysis. For each sample, the binding energy re- new function versus the photon energy (33). The position
gions corresponding to Si (95–115 eV), Mo (220–245 eV), of the absorption edge can then be determined by extrap-
and Co (760–820 eV) were scanned. The binding energies olating the linear part of the rising curve to zero (20, 33).
were corrected by reference to the C(1s) line at 284.8 eV. The values thus obtained carry information about the av-
A nonlinear Shirley-type background was used for the area erage domain size of the oxide nanoparticles. It has been
analysis of each peak. The fitting of the XPS spectra was shown that the energy band gap decreases as the domain
carried out with asymmetric peaks, using the MultiPak soft- size increases (20). Therefore, a comparison can be made
ware from Physical Electronics.
between the energy of the samples under investigation and
The H2-TPR experiments were conducted passing a those of references of known domain size. This compari-
continuous flow of 5% H2/Ar over approximately 30 mg son is made in Fig. 1, which shows the absorption edges
of the calcined catalyst at a flow rate of 10 cm3/min of several MoOx species together with those of two dif-
while linearly increasing the temperature at a heating ferent Co : Mo/SiO2 catalysts. As expected, the band gap
rate of 8 C/min. The hydrogen uptake as a function of energies in the reference series decrease as the domain size

132
HERRERA ET AL.
catalyst in the calcined state. Therefore, it can be concluded
that the catalysts with low Co : Mo ratios exhibit most of the
Co interacting with Mo. However, as the Co : Mo ratio in-
creases, free Co oxide beginsto appear. In fact, the spectrum
of the Co : Mo (3 : 4)/SiO2 catalyst can be rationalized as a
sum ofcontributionsfrom two typesofspecies, one interact-
ingwith Mo (main band at around 600nm) and a second one
in which the Co oxidic species are segregated and not inter-
acting with Mo (main band at 680 nm). Similar conclusions
have been previously drawn from Raman spectroscopy and
X-ray diffraction (XRD) data, which indicated that a non-
interacting Co phase is formed on Co-Mo/SiO2 catalysts at
high Co : Mo ratios (17, 34).
3.1.2. EXAFS/XANES. Figure 3 shows the K-edges of
Mo (E = 20,000eV) for two different calcined catalystswith
Co : Mo molar ratios of 2 : 1 and 1 : 3. Absorption spectra for
the two reference compounds, -CoMoO4 and ammonium
heptamolybdate, have been included for comparison. The
absorption edges for both catalysts are remarkably similar.
They both exhibit a pre-edge feature, which is also observed
in the spectrum of ammonium heptamolybdate. This pre-
edge feature is typically observed in distorted octahedral
environments, such as that found in the heptamolybdate.
It is due to a 1s → 4d bound-state transition that in the
case of a perfect octahedral geometry is formally forbid-
den. Accordingly, it is barely present in compounds such
as MoO3 and -CoMoO4. However, it becomes allowed
when the d-states of the metal mix with the p orbitals of
FIG. 1. Lower panel: UV absorption spectra for two Co : Mo/SiO2 cal-
cined catalysts. Upper panel: (a) MoO3, (b) (NH4)6Mo7O24, (c) NaMoO4
references.
increases. Those of the Co : Mo/SiO2 catalysts lie between
those of (NH4)6Mo7O24 and MoO3. From this comparison,
it can be inferred that the Mo species in the calcined cata-
lysts have relatively small domain sizes. The presence of a
small contribution of MoO3 species could not be ruled out
since a small tail can be observed below 3.0 eV, but most of
the Mo is in a high state of dispersion.
In addition to the charge-transfer bands due to Mo ap-
pearing in the UV region, the visible spectra of the bimetal-
lic catalysts present bands in the 500- to 750-nm region,
which did not appear for the Mo/SiO2 catalyst. These bands
are associated with Co species and have previously been as-
cribed to d–d transitions (4T2g → A2g and 4T2g → T1g(P))
of high spin octahedral Co complexes (14). Figure 2 shows
the DRS spectra in this region for three calcined catalysts,
Co : Mo (3 : 4)/SiO2, Co : Mo (1 : 3)/SiO2, and Co/SiO2. The
spectrum for the Co : Mo (1 : 3) catalyst isverysimilar to that
of -CoMoO4, which is typical of Co in an octahedral en-
vironment. The spectrum for the Co : Mo (1 : 2) was almost
identical to that of the Co : Mo (1 : 3) catalyst, so only one of
them is included in the graph. By contrast, the shape of the
spectrum of the Co : Mo (3 : 4)/SiO2 catalyst was markedly
different and exhibited the appearance of a band at around
680 nm. This band was in turn the dominant feature in
the pure Co catalyst and should be associated with Co3O4
species, which, as shown below, are present in the pure Co
4
4
FIG. 2. Visible spectra for two calcined bimetallic Co : Mo/SiO2 cata-
lysts with different Co : Mo ratios (1 : 3 and 3 : 4) and that for a monometal-
lic Co (0.02 wt% )/SiO2. The spectrum of an -CoMoO4 reference is in-
cluded for comparison.

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
133
that this shoulder coincides with the white line of CoMoO4
and CoO species. A first approximation of a XANES com-
posed of two different phases can be obtained by simple ad-
dition of the XANES of the individual components. A sim-
ple fitting with a linear combination of contributions from
Co3O4 and CoMoO4 reproduces the XANES spectrum of
the Co : Mo (2 : 1) catalyst (see Fig. 5B). This comparison in-
dicates that in the catalyst with Co excess, most of the Co is
in the form of Co3O4 and a small fraction is CoMoO4. When
the fitting was attempted using CoO as a third component,
the best fit did not include any contribution of this oxide.
The EXAFS data were in good agreement with the con-
clusions reached from the XANES analysis. As shown, in
Fig. 6, the Fourier transform for the calcined Co : Mo (2 : 1)
catalyst is very similar to that of Co3O4, indicating that this
oxide is the predominant form present when Co is in excess.
The results on the sample with low Co : Mo ratio require
some further consideration. It is interesting to note that
while the XANES spectra for these catalysts look simi-
lar to that of -CoMoO4 from the Co side, they bear no
FIG. 3. Mo-edge (20,000 eV) XANES of two CoMo calcined catalyst
compared to -CoMoO4 and (NH4)6Mo7O24 references.
the ligand, as in compounds with distorted octahedral sym-
metries (19, 35, 36). Of course, a pre-edge feature is always
observed in Mo species with tetrahedral symmetry, such as
in Na molybdate. However, in such casesthe feature ismuch
more pronounced than that observed here. Therefore, it can
be concluded that in the bimetallic catalysts, Mo is mostly
in a structure similar to that of the heptamolybdate. It is
interesting that this is the same for both the Co : Mo (2 : 1)
and Co : Mo (1 : 3) samples.
Next, it isimportant to investigate the same set of samples
at the Co edge. Figure 4 compares the XANES spectra for
the K-edge of Co (Eo = 7709 eV) in the calcined Co : Mo
(1 : 3)/SiO2 catalyst and that in the -CoMoO4 reference.
Except for some difference in the size and shape of the first
peak in the edge, both spectra look remarkably similar. By
contrast, the Co edge for the Co : Mo (2 : 1) catalyst, contain-
ing excess Co, is very different from that of the -CoMoO4
reference. As shown in Fig. 5A, the XANES of the catalyst
is in fact very similar to that of Co3O4, although a small
shoulder appearing at around 7726 eV is more pronounced
for the catalyst than for the oxide. It is interesting to note
FIG. 5. (A) Co K-edge XANES of a calcined Co : Mo (2 : 1)/SiO2 cata-
lyst compared to those of -CoMoO4, CoO, and Co3O4 references. (B) Co
K-edge XANES of a calcined Co : Mo (2 : 1)/SiO2 catalyst (curve a) and
a weighted linear combination of XANES from -CoMoO4 and Co3O4
(curve b). The contributions resulting from the best fit were 82% Co3O4
and 18% -CoMoO4.
FIG. 4. Co K-edge (7719 eV) XANES of calcined Co : Mo (1 : 3)/SiO2
catalyst compared to -CoMoO4 used as a reference.

134
HERRERA ET AL.
3.1.3. TPR. The reduction profiles of calcined mono-
metallic Co/SiO2 and Mo/SiO2 catalysts together with that
of the bimetallic Co : Mo (1 : 3)/SiO2 catalyst are shown in
Fig. 8a. The TPR profile of the Co monometallic cata-
lyst shows two peaks at 360 and 445 C, which can be as-
cribed to the reduction of Co oxide species. The reduction
of the monometallic Mo catalyst also exhibits two peaks,
but they appear at much higher temperatures than those
of Co. Therefore, from the reduction profiles it is possible
to identify the presence of Co and Mo species in the ab-
sence of interactions. Accordingly, the TPR of the bimetal-
lic Co : Mo (1 : 3)/SiO2 catalyst indicates that in this sample,
the vast majority of Co oxide species are interacting with
Mo. It is clear that while most of the Co in the monometal-
lic catalyst gets reduced below 500 C, almost no reduction
takes place below that temperature in the bimetallic cata-
lyst. It has been reported that the reduction of interacting
Co–Mo-supported species occurs at similar temperatures as
those assigned for the reduction of free Mo species (17). It
has also been proposed that the addition of Mo oxide to Co
oxide inhibits the reduction of the Co species because Mo6+
FIG. 6. Fourier transforms of the k3 EXAFS data of the Co K-edge
obtained on the calcined Co : Mo (2 : 1)/SiO2 catalyst (solid line) and for
Co3O4 reference (dotted line).
resemblance to this compound from the Mo side. One
may rationalize this contrasting behavior by proposing that
while most of the Co in the catalyst is forming a CoMoO4-
like phase, only a fraction of Mo participates in this com-
pound. The rest of the Mo would be in a dispersed Mo oxide.
We propose that the Co is in a CoMoO4-like phase because
the XANES indicates that the local environment of Co in
the Co : Mo (1 : 3) catalyst is very similar to that in CoMoO4,
but the EXAFS data are significantly different from that of
the compound. This comparison is made in Fig. 7, which
shows the Fourier transforms for the K-edge of Co in the
calcined Co : Mo (1 : 3)/SiO2 catalyst, together with that of
-CoMoO4. The low intensity observed in the catalyst for
˚
the peaks between 2.5 and 4 A, clearly observable for the
Co molybdate, would indicate that bulk -CoMoO4 is not
present, but rather a highly dispersed CoMoO4-like struc-
ture, coexisting with Mo oxide species.
FIG. 8. (a) TPR profiles of monometallic Co/SiO2 and Mo/SiO2 catal-
ysts compared to the Co : Mo (1 : 3)/SiO2 bimetallic catalyst. (b) TPR pro-
files of Co-Mo/SiO2 bimetallic catalysts with varying Co : Mo ratios. In all
cases the reduction was conducted under 5% H2/Ar, using a heating linear
ramp of 8 C/min.
FIG. 7. Fourier transforms of the k3 EXAFS data of the Co K-edge,
obtained on the calcined Co : Mo (1 : 3)/SiO2 catalyst (dotted line) and on
a
-CoMoO4 reference (solid line).

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
135
polarizes the Co–O bonds, making them more ionic and
consequently more difficult to reduce (37). In agreement
with the DRS and EXAFS/XANES data, TPR indicates
that a high degree of Co-Mo interaction is only observed
for the catalyst with a low Co : Mo ratio. As shown in Fig. 8b,
as the Co : Mo ratio increases, a gradually increasing frac-
tion of segregated Co species is apparent from the peaks at
360 and 445 C, which are associated with the reduction of
noninteracting Co oxide.
3.2. Characterization of the Reduced Catalysts
3.2.1. DRIFTS. The vibrational spectrum of adsorbed
NO was used to investigate the Co : Mo/SiO2 catalysts after
the reduction pretreatment in H2 at 500 C. As mentioned
above, both Co and Mo are able to adsorb NO at room
temperature, exhibiting characteristic IR absorption bands
that can be used to identify the NO adsorption on each
metal. It is generally agreed that MoVI does not adsorb NO
(22), but MoII, MoIII, and MoIV have all been suggested as
potential NO adsorption sites (11, 24). Similarly, both cal-
cined and reduced Co catalysts are able to adsorb NO (26).
It has been reported that adsorption on reduced Co results
in bands at slightly lower frequencies than on oxidized Co
(25). However, it is not always possible to determine the
chemical state of Co based on NO adsorption. Therefore,
although DRIFTS of adsorbed NO may not be the best
technique to characterize the chemical state of Co and Mo,
it is certainly a powerful tool to quantify the degree of site
blocking of one of the two components by the other.
FIG. 9. IR spectra of NO adsorbed on monometallic (Mo/SiO2 and
Co/SiO2) and bimetallic (Co : Mo/SiO2) catalysts. The catalysts were re-
duced under hydrogen at 500 C.
band, which mayinclude severalindividualbands, appeared
Figure 9 shows the DRIFTS spectrum of NO adsorbed in the region of 1730–1650 cm 1 and can only be associated
on two monometallic catalysts with 4.6 wt% Mo and with adsorption on Mo. The appearance of different com-
1.4 wt% Co, and a bimetallic catalyst with 4.6 wt% Mo ponents associated with this asymmetric-mode band in the
and 1.4 wt% Co (Co : Mo = 1 : 2). On the monometallic bimetallic catalyst contrasting to that on Mo/SiO2 catalyst
Mo/SiO2 catalyst, the bands corresponding to the sym- could be explained in terms of Mo sites with different de-
metric and antisymmetric stretching modes of dinitrosyl grees of coordinative unsaturation (12), or alternatively, in
speciesare clearlyobserved. The symmetricmode exhibited terms of sites where Mo is influenced by Co to various de-
a band at 1814 cm 1, while the antisymmetric mode gene- grees. It is also interesting to note that the overall intensity
rated a broad band centered at around 1714 cm 1. Yao and of the bands of the bimetallic Co-Mo catalysts is consis-
Rothschild (23) have attributed the broadening of the an- tently much lower than that of the monometallic Co and
tisymmetric band to inhomogeneities on the surface that Mo catalyst. This effect has been previously observed for
influence the vibrational transition moment of the asym- alumina-supported Co-Mo catalysts. In that case, a lower
metric mode more than the moment of the symmetric dispersion of the active species on the bimetallic catalysts,
mode. For the monometallic Co/SiO2 catalyst, the bands compared to that of the monometallic ones, was considered
of the dinitrosyl species appeared at significantly higher responsible for the observed loss in intensity (22).
wavenumbers (1880 and 1803 cm 1) than those on Mo. In
Figure 10 shows the Fourier transform infrared (FTIR)
this case, the antisymmetric mode is the dominant band spectra of NO adsorbed on different reduced bimetallic
in the spectrum. Finally, the bimetallic Co : Mo (1 : 2)/SiO2 catalysts, in which the Mo content was kept constant while
catalyst showed three absorption bands that roughly cor- the Co content was increased. It is clear that as the Co
respond to those of the individual components. The domi- content increases, the adsorption of NO over the Mo sites
1
nant band appearing at 1806 cm obviously has contribu- is inhibited. A similar effect has been previously reported
tions from both the symmetric band of NO adsorbed on for the sulfided Co : Mo/Al2O3 catalyst and interpreted as
Mo and the antisymmetric band of NO adsorbed on Co. A a blockage of Mo sites by Co (21). Since the low-frequency
smaller band appeared at 1883 cm 1, which can be associa- antisymmetric NO stretching on Mo does not have con-
ted with the symmetric mode on Co sites, and a broader tributions from species adsorbed on Co, it can be used as

136
HERRERA ET AL.
TABLE 1
Binding Energies for Reduced Catalysts and
Reference Compounds
Binding energies (eV)
Determinant
Mo 3d5/2
Co 2p3/2
Reference Samples
Co
CoO
Co3O4
-CoMoO4
MoO3
778.2
780.5
780.1
781.4
232.5
232.4
229.1
232.3
MoO2
(NH4)6Mo7O24 4H2O
Reduced Catalysts
Mo/SiO2
229.2
231.0
233.4
228.9
231.1
233.9
228.3
230.8
233.3
228.9
230.8
233.1
228.7
230.9
233.2
Co : Mo (1 : 3)/SiO2
Co : Mo (1 : 2)/SiO2
Co : Mo (3 : 4)/SiO2
Co : Mo (2 : 1)/SiO2
778.1
781.5
777.9
780.7
778.5
781.5
778.4
781.4
FIG. 10. IR bands of NO adsorbed on Co : Mo (2 : 1)/SiO2, Co :Mo
(3 : 4)/SiO2, and Co : Mo (1 : 3)/SiO2. The catalysts were reduced under
hydrogen at 500 C.
(781.4 eV) references, respectively. Similar values for these
reference materials have been previously reported in the
literature (14, 26, 37, 38). The respective surface fractions
of Co in the two chemical states were obtained by fitting the
spectra with asymmetric curves centered at the correspond-
ing binding energies. The fraction of Co in the metallic state
after reduction at 500 C, as determined from this analysis,
is shown in Fig. 11 as a function of the Co : Mo ratio. At low
an indication of the density of Mo sites covered by Co. A
clear decrease in the intensity of this band is observed as
the Co : Mo ratio increases, becoming practically negligible
when Co is in excess (e.g., for a Co : Mo ratio of 2 : 1). At the
same time, the bands for this catalyst appear at the same
wavelengths as those on the monometallic Co catalyst (i.e.,
1880 and 1803 cm 1). A straightforward conclusion drawn
from these observations is that when Co is in excess, it al-
most completely covers the Mo sites, while the excess Co
forms a noninteracting species.
3.2.2. XPS. XPS can be used to determine the chemical
nature of the catalyst constituents and to roughly estimate
their distribution on the surface. For the first purpose, the
binding energy of the catalysts can be compared to those of
reference compounds. In Table 1, the binding energies of
the Co 2p3/2 and Mo 3d5/2 levels obtained for the catalysts
in the reduced state are compared to those of the reference
compounds. The spectra corresponding to the Co 2p3/2 lev-
els for all the catalysts can be described in terms of two
contributions, one appearing at about 778 eV and other at
781.5 eV. As shown in the table, these binding energies are
in good correspondence with the binding energies that were
FIG. 11. Fraction of cobalt reduced to the metallic state as a function
of Co : Mo nominal ratio, as determined by XPS on samples pretreated
obtained for the metallic Co (778.2 eV) and the CoMoO4 under H2 at 500 C and heated in He at 700 C, without exposure to air.

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
137
FIG. 12. Fraction of reduced Mo (i.e., Mo4+ + Mo5+) as a func-
tion of Co : Mo nominal ratio, as determined by XPS on samples pre-
treated under H2 at 500 C and heated in He at 700 C, without exposure
to air.
FIG. 13. Surface atomic percentage of molybdenum (filled circles)
and cobalt (open circles) as function of the nominal Co : Mo ratio, as de-
termined by XPS on samples pretreated under H2 at 500 C and heated in
He at 700 C, without exposure to air.
Co : Mo ratios, when the majority of the Co oxide species
are interacting with Mo, most of the Co remains in the oxi-
dic form, but as the Co : Mo ratio increases, a larger fraction
of Co gets reduced. In agreement with this trend and with
the TPR data shown above, previous reports have indicated
that the Co–Mo interacting phase is, in fact, more difficult
to reduce than Co oxide alone (17).
The assignment of the Mo 3d5/2 levels to different chem-
ical states of Mo is not as straightforward as that of Co. As
shown in Table 1, the spectra of the reduced catalysts can be
described in terms of three contributing peaks. The one at
the lowest binding energy appears in the same region as that
of Mo4+ in MoO2. The one at the highest binding energy
(233.1–233.9 eV) appears at slightly higher energy than that
of Mo6+ in MoO3. This peak has been previously assigned
to Mo6+ species in an oxidic environment (37, 39), although
some authors have indicated that these species should have
binding energies in the region of 232.5–232.7 eV (40, 41).
The remaining peak appearing at around 231 eV can be
attributed to Mo in an intermediate state such as Mo+5, as
previously proposed (37, 42). From the fitting of the spectra
of the different catalysts with asymmetric curves centered
at the indicated binding energies, the fraction of reduced
Mo (i.e., Mo4+ + Mo5+) has been calculated and plotted in
Fig. 12 as a function of the Co : Mo ratio. It is observed
that except for the monometallic catalyst, there is a slight
decrease on the reducibility of Mo as the Co : Mo ratio in-
creases. That is, the Co-Mo interaction is also evident from
the Mo analysis, although the effect is not as pronounced
as for Co.
In Table 2, we report the surface atomic fractions of Co
and Mo on the catalyst after the reduction pretreatment.
The intensity of the Si peak remained almost constant for
all samples, except for the catalyst with a Co : Mo ratio of
2 : 1, in which it decreased. Figure 13 shows the fractions of
Co/Si and Mo/Si as a function of bulk Co : Mo ratio. It is
seen that even though the Mo concentration in the catalyst
was kept constant, the Mo/Si ratio clearly decreased as the
amount of Co increased. In agreement with the DRIFTS
TABLE 2
Surface Atomic Concentrations for Reduced Catalysts as
Determined by XPS
Atomic concentration (% )
Reduced catalysts
Mo
Co
Si
Mo/SiO2
1
0
30.41
31.08
31.04
31.42
29.77
Co : Mo (1 : 3)/SiO2
Co : Mo (1 : 2)/SiO2
Co : Mo (3 : 4)/SiO2
Co : Mo (2 : 1)/SiO2
0.99
0.94
0.58
0.33
0.27
0.29
0.43
0.35
FIG. 14. Fourier transforms of the k3 EXAFS data of the Co K-edge
obtained for several Co : Mo/SiO2 reduced catalysts with different Co/Mo
ratios. The data for a Co foil are included for comparison.

138
HERRERA ET AL.
TABLE 3
Structural Parameters Resulting from the Fitting of the Co-Edge EXAFS Data Obtained
for the Co : Mo (2 : 1)/SiO2 Catalyst
Fitting (E0 = 8.2)
Reference
phase
Coordination
number
Bond
Distances
Sigma
r-factor
Co
Co–Co (first shell)
Co–Co (second shell)
Co–Co (third shell)
7.8
2.4
9.6
2.492
3.523
4.327
0.0046
0.0047
0.0059
0.0014
Note. E0 represents the energy shift, sigma is the Debye–Waller factor, and the r-factor is a measurement
of the fit quality.
data, these results show that the addition of Co results in a ratios, filtered EXAFSdata were fitted for the Co : Mo (1 : 3)
gradual coverage of Mo.
catalyst, as previously described. In the first attempt, the
EXAFS data were fitted using a single phase, which was
either metallic or oxidized Co. In both cases, the results
were unsatisfactory. A more complex model was needed to
obtain a good fit. In the second model used for fitting, the
simultaneous presence of metallic Co clusters and oxidized
Co species was considered. Satisfactory fits were only ob-
tained with the simultaneous presence of Co metallic clus-
ters and a CoMoO4 phase. Figure 16 illustrates the quality
of the fit, while Table 4 summarizes the structural param-
eters determined from the analysis. The resulting parame-
ters indicate that in this catalyst, cobalt is mainly forming a
CoMoO4-like phase and a small fraction is forming metal-
lic Co.
3.2.3. EXAFS. The EXAFS results for the reduced
catalysts, shown in Fig. 14, reveal a definite trend, which
ties in nicely with the XPS data. The data indicates that,
keeping the amount of Mo fixed, the fraction of metal-
lic Co increases with the Co : Mo ratio. The appearance
of clearly observable peaks in the range corresponding to
the second and thrid coordination spheres (0.25–0.45 nm)
indicates that larger Co clusters are present on the cat-
alyst with the higher Co content. To quantify this trend,
the data for the first, second, and third coordination shells
of Co on the Co : Mo (2 : 1)/SiO2 were isolated by applying
an inverse Fourier transform over a restricted range of r
(0.13–0.45 nm). The filtered data were then fitted using the
FEFFIT program and assuming that only Co was present.
As shown in Fig. 15, the quality of this fit was excellent. The
structural parameters resulting from the fit are summarized
in Table 3. From the above results, we can conclude that in
the reduced Co : Mo/SiO2 with high Co : Mo ratios, a large
fraction of Co is in the form of metallic Co clusters.
3.3. Production of Carbon Nanotubes by Catalytic
Disproportionation of CO
The ability of the different catalysts to produce SWNT by
CO disproportionation was tested by passing pure CO over
the catalyst at 700 C. Before the reaction, the catalysts were
calcined in air at 500 C, then reduced in H2 at 500 C, and
then heated in He flow to the reaction temperature. At the
To compare the structures of the Co phases in the cata-
lysts with low Co : Mo ratios with those of high Co : Mo
TABLE 4
Structural Parameters Resulting from the Fitting of the Co-Edge EXAFS Data
Obtained for the Co : Mo (1 : 3)/SiO2 Catalyst
Fitting (E0 = 9.88)
Coordination
number
Distances
˚
Phase
Bond
(A)
Sigma
r-factor
-CoMoO4
Co–O (first shell)
3
1
2.1 0.18
3.03
3.15
3.7 0.05
3.9 0.13
2.50
0.0029
0.0035
0.0323
Co–Co (first shell) &
Co–Mo (first shell)
Co–O (second shell)
Co–Mo (second shell)
Co–Co (first shell)
Co–Co (second shell)
Co–Co (third shell)
3
0.0019
0.0027
0.0035
0.0023
0.0033
3.2
1.5
2.2
1.6
Co
3.65
4.35
Note. Six different scattering paths were used for the Co–O (first and second shells) and Co–Mo (second
shell) pairs.

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
139
TABLE 5
Carbon Yield and Selectivity for Two Catalysts with Different
Co : Mo Ratios
MWNT
and
Total
carbon
Amorphus
Catalysts
carbon (% ) SWNT (% ) graphite (% ) yield (% )
Co : Mo (1 : 3)/SiO2
Co : Mo (2 : 1)/SiO2
18.7
14.8
70.7
15.2
10.6
70.0
18.25
16.35
Note. The yield is defined as mass of total deposited carbon per mass
of catalyst. The selectivity to SWNT is the mass of SWNT per total mass
of carbon deposits.
bon deposits. In the first case, one can observe the typical
parallel lattice fringes characteristic of SWNT. In the sec-
ond case, a spacing of 0.34 nm can be observed between the
fringes, which is characteristic of the space between indi-
vidual walls in MWNT (43).
Raman spectroscopy provides valuable information
about the structure of carbon nanotubes. The analysis of the
radial A1g breathing mode (below 300 cm 1) gives direct
information about the tubes diameter, while the analysis
FIG. 15. Fit in k-space for Co : Mo (2 : 1)/SiO2. Experimental (trian-
gles) and modeled EXAFS contribution around Co (full line).
end of a 2-h reaction period, the spent catalyst containing
the carbon deposits was cooled in He flow. The character-
ization of the carbon deposits was done using three main
techniques that we have previously used and tested. They
are temperature programmed oxidation (TPO), transmis-
sion electron microscopy (TEM), and Raman spectroscopy.
We have previously shown that from the TPO results we can
obtain a quantitative measure of the carbon yield and se-
lectivity toward SWNT (4). These results are summarized
in Table 5 and illustrate the strong influence of the Co : Mo
ratio on SWNT selectivity. The TEM observations totally
support the TPO results. As shown in Figs. 17 and 18, a con-
trasting difference is observed in the carbon structures pro-
duced with the two different samples. While the sample with
a Co : Mo ratio of 1 : 3 exhibited a high density of SWNT,
the sample with a Co : Mo ratio of 2 : 1 mainly produced
multiwalled carbon nanotubes (MWNT) and graphitic car-
FIG. 16. Fit in k-space for Co : Mo (1 : 3)/SiO2. Experimental (trian-
FIG. 17. TEM images showing SWNT produced by CO dispropor-
gles) and modeled EXAFS contribution around Co (full line).
tionation on a Co : Mo (1 : 3)/SiO2 catalyst.

140
HERRERA ET AL.
FIG. 18. TEM images showing a mixture of SWNT, MWNT, and graphite produced by CO disproportionation on a Co : Mo (2 : 1)/SiO2 catalyst.
1
of the tangential mode (TM) range (i.e., 1400–1700 cm
)
high selectivity toward SWNT and a dominant band in the
provides information on the electronic properties of the breathing mode region, centered at 265 cm 1. A nonlin-
tubes. In addition, the analysis of the so-called D band at ear relationship has been recently developed, by which the
1
around 1350 cm gives an indication of the level of dis- diameter of a tube in a bundle is directly given by the ex-
ordered carbon. The size of the D band relative to the TM pression (cm 1) = 238/(d(nm))0.93 (46). Accordingly, the
band at around 1590 cm 1has been used as qualitative mea- band observed at around 265 cm would correspond to a
1
surement of the formation of undesirable forms of carbon tube diameter of 0.89 nm.
(44, 45). Figure 19 shows the Raman spectrum obtained
It is noteworthy mentioning that when the Co : Mo
with an excitation source of 532 nm on the carbon deposits (1 : 3)/SiO2 catalyst that exhibited a high yield and selec-
formed on the catalyst with a low Co : Mo ratio. In agree- tivity toward SWNT was employed without the reduction
ment with the TPO and TEM results, the spectrum shows a step or with an exceedingly high reduction temperature,

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
141
FIG. 20. Fourier transforms of the k3EXAFS data obtained for the
K-edge of Co for a fresh Co : Mo (1 : 2)/SiO2 catalyst reduced in hydrogen
(500 C), and after the growth of carbon nanotubes for reaction periods of
3 and 30 min. The EXAFS data of a Co foil are included for comparison.
FIG. 19. Raman spectrum of the carbon nanotubes produced by cata-
lytic disproportionation of CO on Co : Mo (1 : 3)/SiO2 at 700 C. Excitation
laser energy at 532 nm.
poor SWNT yields were attained. It is suggestive that the
TPR indicates that the Co reduction in the Co : Mo (1 : 3)
does not start below 500 C. One may surmise that some
degree of reduction, perhaps creation of oxygen vacancies
around Mo, is necessary, since a totally oxidized catalyst is
not efficient. On the other hand, a high degree of reduction
of the catalyst, causing the appearance of metallic Co, is
also detrimental for selectivity toward SWNT.
pure Co metal. Undoubtedly, during the formation of car-
bon nanotubes over the selective catalyst, Co is gradually
reduced to the metallic state. The detailed structural anal-
ysis of the Co EXAFS data indicated that as the reaction
proceeds, the Co reduction is accompanied by an increase
in metal particle size. Evidence for this process is presented
in Table 6, where an increase in the Co–Co coordination
number as a function of reaction time is observed.
3.4. Characterization of the Spent Catalysts
In the same way, as illustrated in Fig. 21, after a 30-min
The EXAFS results for the spent catalysts are shown in reaction period, the Fourier transforms of the EXAFS data
Fig. 20. The Fourier transform of the EXAFS data for the of Mo (Eo = 20,000 eV) reveal a drastic change, developing
K-edge of Co (Eo = 7709 eV) for the Co : Mo (1 : 2) cata- peaks that correspond exactly to those of the Mo2C refer-
lyst shows that after the pretreatment and before the reac- ence. The transformation of oxidized Mo species into Mo
tion with CO, a significant fraction of Co remains oxidized. carbide during the reaction is indeed very clear, with a small
However, as the reaction proceeds, metallic Co begins to fraction of unconverted Mo oxide remaining in the cata-
form, and after 30 min, the Fourier transform is indicative of lyst. This conversion can be clearly seen from the EXAFS
TABLE 6
Structural Parameters for the Co : Mo (1 : 2)/SiO2 Catalyst after Different Reaction Periods, as Obtained from the EXAFS Data Analysis
Coordination
˚
Sample
Sample treatment
Spent 3 min
Edge (M)
Co
E0
Sigma
NM
N M
N M
C
Distances (A)
r-factor
M
O
-
-
-
Co : Mo (1 : 2)/SiO2
4.2
0.0068
0.0025
0.0065
0.0003
0.0063
0.0106
0.0075
0.0075
0.0039
5.8
—
9.2
—
11.6
—
6.7
—
—
2.2
—
1.1
—
—
2.47
2.05
2.50
2.07
2.97
2.08
2.98
2.09
1.68
0.043
—
—
—
—
5.8
—
—
3.3
Co : Mo (1 : 2)/SiO2
Mo2C
Spent 30 min
Reference
Co
9.4
6.6
7.1
0.020
0.022
0.038
Mo
Mo
Co : Mo (1 : 2)/SiO2
Spent 30 min
—
0.9
—
—
Note. E0 represents the energy shift; is the Debye–Waller factor; NM M , N M C , and NM are the coordination numbers between the metals (Mo
O
-
-
–
˚
or Co) with another metal atom, a C atom, and an O atom, respectively; R (A) is the coordination distance for each respective bond; and the r-factor
is a measurement of the fit quality.

142
HERRERA ET AL.
supported catalysts at high metal contents (14). All these
results can be put together in a simple model for the cal-
cined catalysts with low Co : Mo ratios. It can be proposed
that Co is in the form of a CoMoO4-like layer on top of Mo
oxide species. Therefore, most of the Co is in an interacting
phase, while only a fraction of Mo is in such a phase. There-
fore, the characterization of Mo does not reflect a strong
interaction with Co, while the characterization of Co shows
in fact a high degree of interaction with Mo.
In contrast, when the Co is in excess, more than one
type of Co species is present in the catalyst. The UV–vis
DRS data show that in addition to the interacting phase of
Co, which dominates at low Co : Mo ratios, a noninteract-
ing phase begins to form as the Co content increases. The
X-ray absorption data (Fig. 6) undoubtedly show that this
noninteracting species is Co3O4, which has been previously
detected by Raman spectroscopy and X-ray diffraction in
Co : Mo/SiO2 catalysts (17, 34).
FIG. 21. Fourier transforms of the k3EXAFS data obtained for the
K-edge of Mo for a fresh Co : Mo (1 : 2)/SiO2 catalyst reduced in hydrogen
(500 C), and after the growth of carbon nanotubes for reaction periods of
3 and 30 min. The EXAFS data of Mo2C are included for comparison.
4.2. Characteristics of the Reduced Catalysts
analysis data of the spent catalyst summarized in Table 6.
The best fit of the experimental data was obtained when the
coordination sphere of the Mo was described as composed
of Mo–C, Mo–Mo, and Mo–O. This indicates that the con-
version to Mo carbide is not complete and a fraction of Mo
remains in the spent catalyst as an oxidized species.
The DRIFTS data of adsorbed NO give further evidence
of the interaction between Co and Mo in the reduced cata-
lysts with low Co : Mo ratios. When the position of the
symmetric-stretching band of dinitrosyl adsorbed on Co is
compared for the monometallic Co and bimetallic Co-Mo
1
catalysts, a shift of about 8 cm to higher wavenumbers is
observed with the bimetallic catalyst (Fig. 9). Furthermore,
the binding energy obtained by XPS for the fraction of Co
that remains unreduced after the reduction treatment in the
bimetallic catalyst (around 781.4 eV) corresponds to that
4. DISCUSSION
4.1. Characteristics of the Calcined Catalysts
The UV–vis and the X-ray absorption data indicate that of Co interacting with Mo oxide, rather than of a noninter-
in all the calcined samples, Mo is in an octahedral envi- acting Co oxide species (780.1–780.5 eV).
ronment in the 6+ oxidation state. Furthermore, from the
However, as the amount of Co in the reduced catalyst in-
values of the adsorption edge energy it can be concluded creases, the shift in the IR main band to higher wavenum-
that the molybdenum oxide species form small clusters bers becomes gradually smaller, until the position of the
with an average domain size somewhat larger than that of band coincides with that of the monometallic Co/SiO2 cata-
the heptamolybdate ion, but not larger enough to form bulk lyst (i.e., 1803 cm 1). At the same time, the intensity of the
MoO3. Regardless of the Co : Mo ratio in the catalysts, the band ascribed to the Mo sites gradually decreases, until it
majority of Mo forms these oxidic species. It is apparent practically disappears for the catalyst with a Co : Mo ratio
that the degree of dispersion of Mo is more a result of the of 2 : 1. This trend is consistent with the XPS data, which in-
loading and degree of interaction with the support, rather dicate that the Mo sites are progressively covered by Co as
than a consequence of the extent of interaction with Co. the Co : Mo ratio increases. Delmon and co-workers have
A contrasting picture is obtained from the characterization proposed that cobalt molybdate is present in the silica-
of Co, particularly in the catalysts with low Co : Mo ratio. supported catalyst, and that the noninteracting Co or Mo
In this case, both, the UV–vis DRS and the XANES data species (depending on which metal is in excess) agglomer-
obtained on the catalyst with a Co : Mo ratio of 1 : 3 demon- ate over CoMoO4,forming a geodelike structure (14). Our
strated that most of the Co is in an environment similar to results clearly show that when Mo is in excess, this is not the
that in -CoMoO4 (i.e., closely interacting with Mo). How- case. The DRIFTS data show that the Co sites are strongly
ever, the slight but clear differences between the spectra influenced by the presence of Mo but they are not covered
of the catalyst and the -CoMoO4 reference suggest that by it. On the other hand, a geodelike structure could be
the similarity is limited to the local environment and na- postulated for the catalysts in which Co is in excess. That is,
ture of ligands. In fact, the EXAFS data demonstrate that Co oxide could partially cover the Co molybdate species.
bulk -CoMoO4 is not present in the bimetallic catalysts.
From the combined analysis of the XPS, TPR, and
Bulk -CoMoO4 has only been observed by XRD in silica- EXAFS data, one can infer that the reducibility of the Co

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
143
FIG. 22. Schematic description of the structure of the different catalysts in the reduced state before reaction, as derived from the characterization
methods.
species is strongly influenced by the presence of Mo in the with high selectivity. The results obtained by EXAFS on
catalyst. The amount of reducible Co clearly increases with the spent Co : Mo (1 : 2)/SiO2 catalyst clearly show that Co
the Co : Mo ratio. This trend can be explained in terms of suffers a severe transformation under reaction conditions.
the formation of an interacting Co molybdatelike species in Before reaction, but after pretreatment in H2 at 500 C and
the oxidized catalyst, which should be more difficult to re- then in He at 700 C, a considerable fraction of Co is still in
duce than noninteracting Co oxide (17). At the same time, the oxide state closely interacting with Mo on the surface,
as shown in Fig. 12, the reducibility of the Mo species is also as shown by FTIR, XPS, and EXAFS/XANES. After 3 min
affected by the presence of Co, indicating that although the under reaction conditions, a significant growth in metal-
techniques that probe bulk Mo do not show an interaction lic Co was observed, although some oxidized Co was still
with Co (XANES, UV–vis DRS), an interaction indeed ex- present. After 30 min, the particles were even larger and
ists at the catalyst surface and it is more pronounced at low essentially all the Co became metallic. Simultaneously, the
Co : Mo ratios.
Combiningallthe information obtained from FTIR, XPS, the carbidic form, as clearly demonstrated by EXAFS.
UV–vis DRS, H2-TPR, and EXAFS/XANES, a coherent As previously proposed (4, 19), the role of Co is the acti-
Mo that was initially in the oxidized state is converted to
representation of the catalyst structure, as it stands right vation of CO. However, when it is in the form of large metal
before the beginning of the SWNT reaction, can be put to- aggregates, it has the tendency to generate mostly MWNT,
gether. This picture is schematically represented in Fig. 22. carbon filaments, and graphite. When Mo is present in the
Three main phases have been identified in the reduced cata- catalyst and there is no Co in excess, a well-dispersed Co+2
lyst just before reaction: molybdenum oxide clusters cobalt species in the form of a Co–molybdate-like phase is stabi-
interacting with molybdenum oxide in a cobalt–molybdate- lized. From the detailed studies conducted in this contribu-
like structure, partially covering the Mo oxide clusters and tion, we can confirm that the formation of this interacting
segregated metallic cobalt particles. The fraction of each of Co–molybdate-like species plays a determinant role in the
these species depends on the Co : Mo ratio.
catalyst selectivity toward the formation of SWNT.
The effect of having Co stabilized on this cobalt-
molybdate-like environment is dramatic. First of all, it
avoids the reduction and formation of large metallic aggre-
gates. The importance of preventing the reduction of Co is
evident when the selective catalyst is compared with a nons-
elective one (Co : Mo ratio of 2 : 1). In the nonselective cata-
lyst, most of the Co in the oxidic state after calcination is
in the form of Co3O4 and, as in the monometallic Co/SiO2
catalyst, this oxide is converted into large metallic Co clus-
ters upon reduction, which produced the unwanted forms
of carbon during reaction at 700–800 C. By contrast, on the
selective catalyst, a concerted mechanism should take place
4.3. Relationship between Catalyst Morphology
and Selectivity toward SWNT
In previous works (4, 19), we have reported that although
MWNT and other carbonaceous forms can be produced
from CO disproportionation on monometallic Co cata-
lysts, these catalysts are not selective for the production of
SWNT. On the other hand, monometallic Mo catalysts are
inactive at the reaction temperatures that we used (around
700–800 C). However, when Co and Mo coexist in the cata-
lysts, especially when Mo is in excess, SWNTs are formed

144
HERRERA ET AL.
during the reaction. As the CO disproportionation starts,
(ii) The extent of the Co–Mo interaction is inversely pro-
Mo oxide is converted into Mo carbide. This transformation portionalto the Co : Mo ratio and hasdifferent formsduring
breaks up the Co–molybdate-like structure, allowing for the different stages of the catalyst life.
the reduction of Co by CO. However, the Co ions are now
(iii) In the oxidic state after calcination, Mo forms small
highly dispersed and in the presence of high concentration clusters in the form of well-dispersed Mo (6+) species and
of CO in the gas phase. This environment seems to be favor- Co either interacts with Mo in a superficial Co-molybdate-
able for the production of SWNT instead of the sintering like structure (at low Co : Mo ratios) or as a noninteracting
that would normally occur during a high-temperature re- Co3O4 phase (at high Co : Mo ratios).
duction process. We may envision the release of extremely
(iv) During the subsequent reduction treatment, the non-
small metal Co clusters into the gaseous CO environment in interacting phase is reduced to metallic Co, while the Co–
a way that resembles the vapor-phase production of SWNT molybdate-like species remain as well-dispersed Co2+ ions.
by the CO disproportionation coupled with decomposition
(v) Under reaction conditions, the Mo oxide species are
of Fe pentacarbonyl that has been recently developed (47). converted into Mo carbide, thus breaking the interaction
In that case, like in ours, the production of large metal clus- and releasing the metallic Co in a state of high dispersion,
ters causes losses in SWNT yield and selectivity.
which is responsible for the production of SWNT. By con-
On the very small Co clusters, the production of MWNT, trast, the large Co clusters are responsible for the nonselec-
carbon filaments, and graphite does not occur. Such produc- tive forms of carbon (MWNT, filaments, graphite, etc.).
tion normally occurs on larger Co clusters, following the
well-known mechanism for carbon filament growth (48).
That is, the metal clusters begin to decompose CO, pro-
ducing Co carbide particles, which then tend to precipitate
graphite at their end in the form of cylindrical filaments.
When the Co clusters are so small that they are only com-
posed of a few atoms, or even isolated Co atoms, a dif-
ferent mechanism must be considered (49). For instance,
a so-called scooter mechanism has been proposed, which
takes into consideration the known fact that without a cat-
alyst present, the carbon precursors would tend to grow
into C60–fullerenes. However, in the presence of a catalyst,
the structure does not close into a hemisphere but is kept
open. It has been proposed that metal atoms would diffuse
(scoot) along the open edge and prevent the closure of the
carbon rings while catalyzing the incorporation of carbon
to the growing tube (50). Another model that does not re-
gard the catalyst as isolated atoms, but rather as a metal
cluster, has been proposed. In this model, the addition of
carbon to the tube occurs at the end of the tube over a metal
cluster that is suspended by the growing tube. In this mode,
the metal cluster is of the same size or slightly smaller than
the diameter of the tube. The metal cluster allows for the
stabilization of reactive bonds dangling at the tube tip and
prevents the tube closure by moving up with the growing
tip (51, 52). When the catalyst particle increases in size, the
strain energy of the overcoating becomes smaller, and as a
result, the formation of a graphitic overcoat is favored (53).
ACKNOWLEDGMENTS
This research was conducted with financial support from the Oklahoma
Center for Advancement of Science and Technology (OCAST) and the
College of Engineering at the University of Oklahoma. Technical support
from the personnel at NSLS, Brookhaven National Lab, for the EXAFS
experiments and from the Noble Electron Microscopy Laboratory at OU
for TEM measurements is gratefully acknowledged. The generous tech-
nical support of Nicolet Instruments in obtaining the Raman spectra is
greatly appreciated. J.E.H. thanks the Fulbright–CAREC program and
A.B. thanks Fundacion Antorchas and Fulbright for scholarships.
REFERENCES
1. Yakpbson, B. I., and Smalley, R. E., Am. Sci. 85, 324 (1997).
2. Kong, J. A., Cassel, M., and Dai, H., Chem. Phys. Lett. 292, 567 (1998).
3. Nikolaev, P., Bronikowski, M., Bradley, R., Rohmund, F., Colbert, D.,
Smith, K., and Smalley, R., Chem. Phys. Lett. 313, 91 (1999).
4. Kitiyanan, B., Alvarez, W. E., Harwell, J. H., and Resasco, D. E., Chem.
Phys. Lett. 317, 497 (2000).
5. Lauritsen, J. V., Helveg, S., Lægsgaard, E., Stensgaard, I., Clausen,
B. S., Topsøe, H., and Besenbacher, F., J. Catal. 197, 1 (2001).
6. Topsoe, H., and Clausen, B. S., Catal. Rev. Sci. Eng. 26, 395 (1984).
7. Bouwens, S. M. A. M., Prins, R., de Beer, V. H. J., and Koningsberger,
D. C., J. Phys. Chem. 94, 3711 (1990).
8. Bouwens, S. M. A. M., van Veen, J. A. R., Koningsberger, D. C., de
Beer, V. H. J., and Prins, R., J. Phys. Chem. 95, 123 (1991).
9. Ratnasamy, P., and Kno¨zinger, H., J. Catal. 54, 155 (1978).
10. Wang, L., and Hall, W. K., J. Catal. 66, 251 (1980).
11. Millman, W. S., and Hall, W. K., J. Phys. Chem. 83, 427 (1979).
12. Segawa, K., and Hall, W. K., J. Catal. 77, 221 (1982).
13. Topsøe, N. Y., and Topsøe, H., J. Bull. Chim. Belg. 90, 1311 (1981).
14. Gajardo, P., Grange, P., and Delmon, B., J. Phys. Chem. 83, 1771(1979).
15. Topsøe, H., Clausen, B. S., Candia, R., Wivel, C., and Mørup, S., Bull.
Soc. Chim. Belg. 12, 1187 (1981).
5. CONCLUSIONS
16. Topsøe, H., Clausen, B. S., Burriesci, N., Candia, R., and Mørup, S., in
“Preparation of Catalysts II” (B. Delmon, P. Grange, P. A. Jacobs, and
G. Poncelet, Eds.), p. 479. Elsevier, Amsterdam, 1979.
We can summarize the main findings of the present work
in the following conclusions:
17. de Boer, M., Koch, E. P. F. M., Blaauw, R. J., Stobbe, E. R., Hoffmann,
A. N. J. M., Boot, L. A., van Dillen, A. J., and Geus, J. W., Solid State
Ionics 63–65, 736 (1993).
18. Gajardo, P., Pirotte, D., Grange, P., and Delmon, B., J. Phys. Chem.
83, 1780 (1979).
(i) The selectivity of the Co-Mo catalysts toward SWNT
production by CO disproportionation strongly depends on
the stabilization of Co2+ species, which result from an in-
teraction with Mo.

STRUCTURE/COMPOSITION OF CO–MO CATALYSTS
145
19. Alvarez, W. E., Kitiyanan, B., Borgna, A., and Resasco, D. E., Carbon
39, 547 (2001).
37. Halawy, S., Mohamed, M., and Bond, G., J. Chem Tech. Biotechnol.
58, 237 (1993).
20. Weber, R. S., J. Catal. 151, 470 (1995).
21. Topsøe, N. Y., and Topsøe, H., J. Catal. 84, 386 (1983).
38. Kleyna, R., Mex, H., Voß, M., Borgmann, D., Viscido, L., and Heras,
J. M., Surf. Sci. 433, 723 (1999).
22. Portela, L., Grange, P., and Delmon, B., Catal. Rev. Sci. Eng. 37, 699 39. Solymosi, F., Csere´nyi, J., Szo¨ke, A., Ba´nsa´gi, T., and Oszko´, A.,
(1995). J. Catal. 165, 150 (1997).
23. Yao, H. C., and Rothschild, G. W., in “Proceedings Climax 4th. Inter- 40. Katrib, A., Leflaive, P., Hilaire, L., and Maire, G., Catal. Lett. 38, 95
national Conference on Chemical Uses of Molybdenum” (H. F. Barry
and P. H. Mitchell, Eds.), p. 31. Climax Molybd. Co., Ann Arbor,
Michigan, 1982.
(1996).
41. Coulier, L., de Beer, V. H. J., van Veen, J. A. R., and Niemantsverdreit,
J. W., Top. Catal. 13, 99 (2000).
24. Kazusaka, A., and Howe, R. F., J. Catal. 63, 447 (1980).
25. Fierro, J. L. G. , in “Spectroscopy Characterization of Heterogeneous
Catalysis” (J. L. G. Fierro, Ed.), p. B119. Elsevier, Amsterdam, 1990.
26. Prada, R., Fierro, J. L. G., Grange, P., and Delmon, B., in “Preparation
of Catalysts IV” (B. Delmon, P. Grange, P. A. Jacobs, and G. Poncelet,
Eds.), p. 605. Elsevier, Amsterdam, 1987.
27. Clausen, B. S., Lengeler, B., Candia, R., Als-Nielsen, J., and Topsøe,
H., Bull. Soc. Chim. Belg. 90, 1249 (1981).
28. Topsøe, H., Clausen, B. S., Topsøe, N. Y., Pedersen, E., Niemann, W.,
42. Patterson, T. A., Carver, J. C., Leyden, D. E., and Hercules, D. M.,
J. Phys. Chem. 80, 1700 (1976).
43. Ebbesen, T. W., and Ajayan, P. M., Nature 358, 220 (1992).
44. Bandow, S. Asaka, S., Saito, Y., Rao, A. M., Grigorian, L., Richter, E.,
and Eklund, P. L., Phys. Rev. Lett. 80, 3779 (1998).
45. Rols, S., Righi, A., Alvarez, L., Anglaret, E., Almairac, R., Journet,
C., Bernier, P., Sauvajol, J. L., Benito, A. M., Maser, W. K., Munoz,
E., Martinez, M. T., de la Fuente, G. F., Girard, A., and Ameline, J. C.,
Eur. Phys. J. B 18, 201 (2000).
Mu¨ller, A., Bo¨gge, H., and Lengeler, B., J. Chem. Soc., Faraday Trans. 46. Rols, S., Righi, A., Alvarez, L., Anglaret, E., Almairac, R., Journet,
83, 2157 (1987).
C., Bernier, P., Sauvajol, J. L., Benito, A. M., Maser, W. K., Munoz,
E., Martinez, M. T., de la Fuente, G. F., Girard, A., and Ameline, J. C.,
Eur. Phys. J. B 18, 201 (2000).
29. Clausen, B. S., Topsøe, H., Candia, R., Villadsen, J., Lengeler, B.,
Als-Nielsen, J., and Christensen, F., J. Phys Chem. 85, 3868
(1981).
30. Rehr, J. J., Zabinsky, S. I., and Albers, R. C., Phys. Rev. Lett. 69, 3397
(1992).
31. Rehr. J. J., Mustre de Leon, J., Zabinsky, S. I., and Albers, R. C.,
J. Am. Chem. Soc. 113, 5135 (1991).
32. Mustre de Leon, J., Rehr, J. J., Zabinsky, S. I., and Albers, R. C., Phys.
Rev. B 44, 4146 (1991).
47. Nikolaev, P., Bronikowski, M. J., Bradley, R. K., Rohmund, F, Colbert,
D. T., Smith, K. A., and Smalley, R. E., Chem. Phys. Lett. 313, 91 (1999).
48. Tibbetts, G. G., Gorkiewicz, D. W., and Alig, R. L., Carbon 31, 809
(1993).
49. Andriotis, A., Menon, M., and Froudakis, G., Phys. Rev. Lett. 85, 3193
(2000).
50. Thess, A., Lee, R., Nikolaev, P., Dai, H., Petit, P., Robert, J., Xu, C.,
Lee, Y. H., Kim, S. G., Rinzler, A. G., Colbert, D. T., Scuseria, G. E.,
Tomanek, D., Fischer, J. E., and Smalley, R. E., Science 273, 483 (1996).
33. Barnton, D. G., Shtein, M., Wilson, R. D., Soled, S. L., and Iglesia, E.,
J. Phys. Chem. B 103, 630 (1999).
34. Jeziorowski, H., Kno¨zinger, H., Grange, P., and Gajardo, P., J. Phys. 51. Guo, T., Nikolaev, P., Thess, A., Colbert, D. T., and Smalley, R. E.,
Chem. 84, 1825 (1980).
Chem. Phys. Lett. 243, 49 (1995).
35. Leliveld, R. G., van Dillen, A. J., Geus, J. W., and Koningsberger, D. C.,
J. Catal. 165, 184 (1997).
52. Maiti, A., Brabec, C., . Roland, C., and Bernholc, J., Phys. Rev. B 52,
14,850 (1995).
36. Xie, S., Chen, K., Bell, A. T., and Iglesia, E., J. Phys. Chem. B 104, 53. Dai, H. J., Rinzler, A. G., Nikolaev, P., Thess, A., Colbert, D. T., and
10059 (2000). Smalley, R. E., Chem. Phys. Lett. 260, 471 (1996).