Confinement of metal nanoparticles in carbon nanotubes

Article abstract of DOI:10.1002/cctc.201300527

The effect of several parameters that include carbon nanotube (CNT) pretreatment and diameter, and the nature of the metal (Co, Ru, Pd), the metal precursor (nitrate, chloride, organometallic complexes), and the solvent on the filling yield of metallic nanoparticles in CNT channels is reported. The obtained results show that it is possible to modulate the filling yield between 10 and 80 % by controlling the CNT opening and playing on the molecular recognition of the inner/outer surfaces by the metal molecular precursor. Interestingly, the best filling yields have been obtained on nitric acid oxidized nanotubes; a treatment often used for the preparation of most CNT-supported metal catalysts. The confined nanoparticles systematically show a smaller particle size than those supported on the external surface. All the prepared samples were tested for the selective hydrogenation of cinnamaldehyde, and clear correlations were established between the catalytic performances and the filling yields. Are you in? Most of the studies that deal with carbon nanotubes (CNTs) for catalysis neglect the possibility of confinement of the active phase in the CNT inner cavity. This study should prompt researchers who study CNT-supported metal catalysts to integrate the possibility of confinement effects to rationalize catalytic results. Copyright

Full text of DOI:10.1002/cctc.201300527

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201300527
Confinement of Metal Nanoparticles in Carbon Nanotubes
T. Trang Nguyen and Philippe Serp*[a]
The effect of several parameters that include carbon nanotube
(CNT) pretreatment and diameter, and the nature of the metal
(Co, Ru, Pd), the metal precursor (nitrate, chloride, organome-
tallic complexes), and the solvent on the filling yield of metallic
nanoparticles in CNT channels is reported. The obtained results
show that it is possible to modulate the filling yield between
10 and 80% by controlling the CNT opening and playing on
the molecular recognition of the inner/outer surfaces by the
metal molecular precursor. Interestingly, the best filling yields
have been obtained on nitric acid oxidized nanotubes; a treat-
ment often used for the preparation of most CNT-supported
metal catalysts. The confined nanoparticles systematically
show a smaller particle size than those supported on the exter-
nal surface. All the prepared samples were tested for the selec-
tive hydrogenation of cinnamaldehyde, and clear correlations
were established between the catalytic performances and the
filling yields.
Introduction
The confinement of metal nanoparticles (NPs) in carbon nano-
tubes (CNTs) has gained an increasing amount of attention be-
cause of potential applications in catalysis, data storage, and
electronic devices. By using the spatial restriction effect of CNT
channels, a variety of nanomaterials, even of sub-nanometer
size, can be synthesized. These materials would be usually in-
trinsically unstable, particularly under elevated temperature
and pressure, or difficult to obtain under mild conditions.[1] Ad-
ditionally, modified adsorption, diffusion, structural, and chemi-
cal properties have been reported for a variety of species con-
fined in CNTs compared to their counterparts either in the bulk
or deposited on the outer CNT walls.[1a,2] A stronger adsorption
onto the CNT inner surface was reported for CO, H2, alkanes,
and alkenes,[3] which has opened the way to selective adsorp-
tion/separation processes.[4] The diffusion of various molecules,
walls, selective gas adsorption, and geometrical constraints
that affect the reaction mechanism.[11] As these effects are
expected to be enhanced if the CNT inner diameter is re-
duced,[4a,9b,12] the preparation of NPs selectively localized inside
small-diameter CNTs is a prerequisite to study this phe-
nomenon.
To achieve maximum confinement, wet chemistry appears
to be the most simple, versatile, and up-scalable method.[13]
However, the capillary effect of CNTs depends on surface func-
tionalization and CNT diameters. CNTs with internal diameters
<10 nm are usually filled to a lesser extent or even remain
empty after wet impregnation.[14] Reported methods for filling
include: 1) wet impregnation assisted by ultrasound by using
cut CNTs,[15] 2) two-step biphasic impregnation,[16] 3) impregna-
tion and selective washing,[17] 4) the use of supercritical CO2,[18]
and 5) molecular recognition.[19] The reported filling yields,
which are commonly measured either by conventional 2D
TEM, performed with[20] or without sample tilting, or by elec-
tron tomography[21] ranged between 70–90%. Among the
preparation methods cited, only the first one, originally devel-
oped by the group of Bao, has been applied for the filling of
small-diameter (<10 nm) CNTs with Ru,[22] Fe,[9,23] Cu,[12a]
MnO2,[24] TiO2,[25] and Rh[26] NPs. The first step of this simple pro-
cedure involves CNT oxidation with HNO3 to open the CNT tip
and to create oxygen surface functional groups.[27] This surface
oxidation reaction is also used for the preparation of most
CNT-supported catalysts,[1b] which often show better perform-
ances than their counterparts supported on other supports,
which include carbonaceous ones.[28] As most of the studies
that deal with CNTs for catalysis have neglected the possibility
of confinement of the active phase in the CNT inner cavity, we
wonder about the possible role of unidentified confinement ef-
fects on the catalytic results.
which include N2 and H2O,[6] inside individual CNTs was re-
[5]
ported to be faster than their bulk diffusion. g-Fe NPs with
a face-centered cubic (fcc) crystal structure, known to be
stable between 1185–1667 K in the bulk, retained their stability
at room temperature if confined in CNTs.[7] Confined cobalt
nanorods showed an fcc instead of a stable hexagonal struc-
ture.[8] The reduction of iron oxide NPs confined in CNT chan-
nels was found to be facilitated with respect to the particles lo-
cated on the outer walls.[9] As a result of these confinement ef-
fects, metal NPs filled inside CNTs usually show better catalytic
performances than those loaded on the outer surface.[1a,c,d,10]
Indeed, confinement effects can affect chemical reactions
through a host of aspects, such as changes in the thermody-
namic state of the system because of interactions with inside
[a] Dr. T. Trang Nguyen, Prof. P. Serp
Laboratoire de Chimie de Coordination UPR CNRS 8241
composante ENSIACET, Universitꢀ de Toulouse UPS-INP-LCC
4 allꢀe Emile Monso BP 44362, 31030 Toulouse Cedex 4 (France)
E-mail: philippe.serp@ensiacet.fr
We report herein the first parametric study that deals with
the confinement of metal NPs in CNTs. We studied the influ-
ence of several parameters such as CNT pretreatment and di-
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201300527.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3595

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ameter, and the nature of the metal (Co, Ru, Pd),
metal precursor (nitrate, chloride, organometallic
complexes), and solvent on the filling yield. We also
investigated the influence of NP confinement on the
performance for cinnamaldehyde hydrogenation, a re-
action for which confinement effects have been al-
ready reported.
Table 1. Specific surface area, pore volume, and Raman data of the CNT supports.
i.d./e.d.[a] Length SBET
Vpore
[cm3 gꢀ1
ID/IG
TPD
CO2 [mmolgꢀ1 CO [mmolgꢀ1
]
[nm]
[mm]
[m2 gꢀ1
]
]
]
CNT0
CNT1
CNT2
5.5:10.9 1–2
5.5:10.9 1–2
5.5:10.9 0.5–0.8 280
390
CNT2HT 5.5:10.9 0.5–0.8 288
CNT3 40:70 0.5–4 38
240
365
2.55
0.89
1.14
0.86
1.04
0.15
1.49
1.53 1296
1.52 437
1.74
1.52
0.87 295
62
211
2132
871
–
–
642
CNT1HT 5.5:10.9 1–2
–
–
Results and Discussion
[a] Ratio of internal to external diameter.
Characterization of catalyst supports
The purified CNTs (CNT0) present a specific surface
oxygen-containing surface functional groups.[30] CNT1 and
CNT2 were then treated at 9008C under Ar for CNT1 or Ar/H2
for CNT2 to remove the oxygen surface functional groups to
produce CNT1HT and CNT2HT, respectively. Selected character-
istics of the prepared supports are shown in Table 1.
The N2 adsorption isotherms and pore size distribution for
the five samples are given in Figure S3, and the TGA profiles
are presented in Figure S2. Ball-
area of 240 m2 gꢀ1, a mean internal diameter of 5.5 nm, and
a mean external diameter of 10.9 nm (Figure S1, Supporting In-
formation). Their thermogravimetric analysis (TGA) profile
under air shows a decomposition temperature of 6408C, and
a purity>98% (Figure S2). CNT0 has been subjected to differ-
ent treatments (Scheme 1) to evaluate the influence of CNT
opening and surface chemistry on the filling yield.
milling in air permits the open-
ing of a few CNTs (a moderate
increase of SBET), introduces low
amounts of oxygenated groups
(temperature-programmed de-
sorption,
TPD),
and
has
a marked influence on the multi-
walled CNT (MWCNT) length. Al-
though no statistical analysis
was performed, it can be stated
that the majority of the tube
ends in CNT2 are not open and
collapsed; these ends are closed
by bending one side of the wall
to the opposing one. Such re-
Scheme 1. Different treatments performed on the pristine CNTs.
sults are in accordance with re-
ported data.[30,31] The decrease of
CNT1 was obtained by nitric acid oxidation of CNT0. Details
of the mechanism of HNO3 oxidation of CNTs have been re-
ported elsewhere.[27] Briefly, this reaction involves the initial
rapid formation of carbonyl groups, which are consecutively
transformed into phenol and carboxylic groups. HNO3 oxida-
tion also allows an increase of the BET surface area (SBET) be-
cause of CNT tip opening.
the pore volume should be connected to the decrease of the
CNT length. Such shortening would contribute to the forma-
tion of MWCNT agglomerates with a densely packed structure
(Figure S1). Thus, a study by Kim et al. showed that the grind-
ing of CNTs resulted in shortened and densely packed CNTs
with increased cleavage as a result of increased grinding
time.[32]
With large-diameter CNTs such as CNT3, which present ex-
posed edges along the entire interior and exterior surfaces, it
has been proposed that prolonged reaction with HNO3 con-
tributes to both endohedral and exohedral functionalization.[16]
In the case of small-diameter CNTs, we recently evidenced that
3 h contact in boiling HNO3 is not enough for endohedral func-
Nitric acid oxidation allows a better opening of CNT0, as evi-
denced by the marked increase of the SBET value of CNT1 and
by the increase of the contribution of the pores that range be-
tween 2–9 nm (Figure S3). The decrease of the pore volume is
connected to the formation of a hydrogen bond network be-
tween the ꢀCOOH surface functional groups, which contrib-
tionalization.[29] Thus, we can expect both endo- and exohedral utes to the formation of dense MWCNT agglomerates. The
ꢀCOOH functionalization for the large-diameter, open-ended
CNT3,[19] and only exohedral ꢀCOOH functionalization for the
small-diameter CNT1. CNT2 was produced by ball-milling of
CNT0 in air. Ball-milling reduces the CNT length from 1–2 mm
to 500–800 nm (Figure S1) and introduces a limited amount of
high-temperature treatments performed on CNT1 and CNT2
did not significantly modify CNT opening as reflected by the
SBET values.
The Raman spectroscopic data, which yield information
about the purity and defects of MWCNTs,[33] show that the
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3596

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
high-temperature treatment performed under Ar (CNT1HT)
contributes to the introduction of defects, presumably vacan-
cies, in the carbon structure. Such an effect of high-tempera-
ture treatment on oxidized CNTs has been reported.[19]
Table 2. Metal loading, NP size, and filling yields for the Ru, Co, and Pd
systems on the different CNT supports.
Support Metal loading [wt%] NP size [nm]
Filling yield [%]
Co
Ru Co Pd Ru Co Pd Ru
Pd
TPD profile deconvolution (Figure S4) shows that the groups
present on the CNT1 external surface are carboxylic acids
(1020 mmolgꢀ1), lactone (110 mmolgꢀ1), carboxylic anhydride
(165 mmolgꢀ1), phenol (918 mmolgꢀ1), and carbonyl/quinone
groups (920 mmolgꢀ1). At 9008C, most of the oxygen surface
functional groups have disappeared (Figure S4). For CNT1HT,
the exposure to air will not contribute to significant reoxida-
tion as checked independently in our laboratory, and for
CNT2HT, a surface passivation with CꢀH bond formation is ex-
pected.[34] The TGA profiles (Figure S2) are in accordance with
these observations.
CNT0
CNT1
CNT2
2.17 3.17 3.25 2.21 3.48 4.15 19ꢃ7 15ꢃ8 13ꢃ8
0.99 2.72 1.24 1.02 2.72 1.15 73ꢃ9 64ꢃ9 72ꢃ10
2.01 3.28 3.08 1.87 3.25 4.01 48ꢃ5 38ꢃ7 35ꢃ8
CNT1HT 1.41 2.94 1.64 1.25 2.88 2.48 60ꢃ9 53ꢃ8 58ꢃ9
CNT2HT 1.23 3.02 2.48 1.47 2.95 3.85 66ꢃ8 62ꢃ10 60ꢃ7
ment on metal loading, NP size, and filling yield are reported
in Table 2. The TEM analyses and particles size distribution are
given in Figures S5–S7.
To summarize, we produced different CNTs with differences
in their opening and surface chemistry. The confinement of
metal NPs in CNTs will be mainly governed by two parameters:
1) CNT opening (tip opening but also the removal of graphene
layers that could have grown perpendicularly to the tube axes
(bamboo-type structure)), and 2) molecular surface recognition,
indeed during the impregnation procedure the metallic precur-
sor should present a higher affinity for the inner than the
outer surface. The channels of CNT1 and CNT3 are well
opened and should allow a higher filling yield than the other
samples. A difference in the inner/outer surface chemistry that
may impact the filling yield can be expected for CNT1 (only
endohedral chemical functionalization), CNT1T (higher concen-
tration of defects on the external surface induced by group re-
moval), and CNT2HT (CꢀH external surface passivation).
Metal loading
The measured metal loading was lower than the theoretical
value and followed the order CoꢂPd>Ru. For a given metal,
the loading also depends on the CNT pretreatment. Surprising-
ly, regardless of the metal used, we find a decrease of metal
loading with the increase of the SBET value of the support (Fig-
ure S8), and a decrease of the metal loading with the increase
of the filling yield (Figure 1). This phenomenon is more pro-
nounced for Ru and Pd precursors than for Co. For the higher
filling yields obtained on CNT1, the two neutral Ru and Pd pre-
cursors ([Ru(NO)(NO3)3] (Ru1) and [Pd(NO3)2(H2O)2] (Pd1))[35]
should be less reactive towards the external, ꢀCOOH-decorat-
ed surface of CNTs compared to the cationic Co(NO3)2·6H2O
(Co1) species. If the presence of ꢀCOOH groups on carbon sur-
faces have been often cited as possible anchoring sites for the
metal precursor, it should depend on the reactivity of the pre-
cursor. This can explain the higher metal loading and the
lower filling yield obtained for CNT1 with respect to Ru and
Pd.
Influence of CNT pretreatment and choice of the metal on
confinement
Owing to the large number of samples prepared, and to the
fact that electron tomography and conventional 2D-TEM with
sample tilting are time-consuming methods, we measured the
filling yields from conventional 2D-TEM micrographs. If we con-
sider that the internal diameter of Nanocyl nanotubes is half
their external diameter, the filling yield was calculated as fol-
lows [Eq. (1)]:
Nanoparticle size
The mean NP size follows the order Ru<PdꢂCo. This order is
consistent with the adsorption energy of metal adatoms on
graphite: Ru (4.43 eV)>Co (3.64 eV)>Pd (1.90 eV).[36] For
a given metal, the mean NP size decreases with the increase of
the filling yield (Figure S9).
ꢀ
ꢁ
number of particles inside
total number of particles
Yield ¼ 2
ꢁ 100 ꢀ100
ð1Þ
This is because the mean size of NPs located inside the CNT
channel (NPin) size is systematically smaller than the mean size
of NPs located on the external surface (NPout). We measured
these diameters by conventional TEM for the catalysts pre-
pared on CNT1 and CNT2 (Figures 2 and 3). The phenomenon
is more pronounced for CNT2 than CNT1, which indicates a dif-
ferent external surface chemistry. We also confirmed the ob-
tained data by electron tomography analyses on Ru/CNT1 (Fig-
ure S10). Although all our efforts to find a facile correlation be-
tween this divergence of diameters and the filling yield failed,
a specific study devoted to this aspect of our work deserves to
be performed.
The yield was calculated for more than 300 particles deposit-
ed on/in 50 different CNTs. Although this method is not very
accurate, it is convenient for comparison purposes. Additional-
ly, we checked one sample independently (see below) to see
that the results are consistent with that obtained by electron
tomography. The influence of CNT pretreatment on confine-
ment was evaluated for three metals, Co, Ru, and Pd, by using
wet impregnation and nitrate salts as the NP precursor (5 wt%
theoretical loading). Acetone was chosen as the solvent be-
cause its surface tension (25.2 mNmꢀ1) is far below the cut-off
value for wetting (180 mNmꢀ1) and can fill the CNT channels
by capillary force. The results of the influence of CNT pretreat-
This effect of confinement on NP size could have two ori-
gins: 1) changes of concentration of the precursor or H2 partial
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3597

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
bute this effect to the presence of a larger amount of external
defects on the heat-treated sample that act as anchoring sites
for NPs. In our case, this could contribute to explain the lower
filling yield obtained on CNT1HT (ID/IG Raman ratio=1.74) than
on CNT1 (ID/IG Raman ratio=1.53). In the case of CNT2, the
effect of the high-temperature treatment is the opposite. The
difference should arise from the treatment performed under
an H2/Ar mixture. Under such conditions, the passivation of
the external surface by the formation of CꢀH bonds should de-
crease the amount of anchoring sites on the external surface.
Influence of solvent, metal precursor, and CNT diameter on
confinement
Four Ru precursors, [Ru(NO)(NO3)3] (Ru1), [Ru3(CO)12] (Ru2),
RuCl3 (Ru3), and [Ru(COD)(COT)] ((1,5-cyclooctadiene)(1,3,5-cy-
clooctatriene)ruthenium(0), Ru4), and two Pd precursors, Pd-
(NO3)2(H2O)2 (Pd1) and [Pd2(dba)3] (tris(dibenzylideneacetone)-
dipalladium(0), Pd2), have been used in combination with the
CNT1 support. In the case of Ru1/CNT1, three different sol-
vents, acetone, isopropanol, and water have been used. These
solvents show different surface tensions and viscosities: 1) sur-
face tension at 258C [mNmꢀ1]: acetone (23), isopropanol (23.3),
and water (72.7), well below the critical surface tension of
CNTs, 2) dynamic viscosity at 258C [mNsmꢀ2]: acetone (0.31),
isopropanol (2.04), and water (0.89), and 3) kinematic viscosity
at 258C [m2 sꢀ1]: acetone (0.39), isopropanol (2.61), and water
(0.89). The metal loadings, NP size and filling yields are given
in Table 3. TEM observations and NP size histograms are given
in Figures S11–S13.
Table 3. Metal loading, NP size, and filling yields for the Ru and Pd sys-
tems on CNT1.
Figure 1. Influence of filling yield on metal loading.
Precursor/solvent
Metal loading
[wt%]
NP size
[nm]
Filling yield
[%]
pressure in the inner cavity compared to the rest of the
medium and/or 2) a different surface chemistry of the two sur-
faces (convex/concave) that could impact the nucleation step
and/or NP stability.
Ru1/acetone (0.23% H2O)
Ru1/acetone (1% H2O)
Ru1/acetone (10% H2O)
Ru1/acetone (20% H2O)
Ru1/isopropanol
Ru1/water
Ru2/acetone
Ru3/isopropanol
Ru4/acetone
0.99
0.97
1.00
0.99
1.05
1.10
2.83
1.90
1.3
1.24
2.20
2.17 (1.33 Pd)
1.02
1.01
1.02
1.03
1.04
1.16
3.80
1.42
1.12
1.15
3.48
1.12
73ꢃ9
77ꢃ10
77ꢃ10
74ꢃ10
70ꢃ9
61ꢃ8
40ꢃ6
58ꢃ10
52ꢃ8
72ꢃ10
58ꢃ7
73ꢃ9
Filling yield
Finally, the filling yield follows the same trend regardless of the
metal used: CNT1>CNT1HTꢂCNT2HT>CNT2>CNT0. No
significant influence of the nature of the metal on the filling
yield was observed. As already stated, the filling yield should
depend on the CNT opening and thus on the SBET, and on sur-
face molecular recognition and thus on CNT surface chemistry.
The higher yields obtained on CNT1 compared to CNT0 and
CNT2 has been attributed to the CNT opening (Figure 4). The
effect of the high-temperature treatment should be more pro-
nounced on the CNT surface chemistry than on their specific
surface area. A decrease of the filling yield upon high-tempera-
ture treatment of HNO3-oxidized CNTs has already been report-
ed for Ru catalysts prepared from RuCl3.[22c] The authors attri-
Pd1/acetone
Pd2/toluene
Ru1Pd1/acetone
The solvent viscosity does not impact the filling yield, at
least not in the investigated range. A lower filling yield was ob-
served in water, which is consistent with the higher surface
tension of this solvent. The choice of the solvent has a limited
impact on NP size and metal loading. The obtained results also
show that the nuclearity of the precursor and the nature of
the ligands and/or the metal oxidation state play a significant
role on metal loading, NP size, and filling yield. Generally, pre-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3598

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
vironment, induces a decrease of
the filling yield. In the coordina-
tion-compound-functionalization
of CNTs, the p orbitals of the
CNTs should overlap with the
vacant dp orbitals of the metal
to result in efficient hexahapto
(h6) complexation of the gra-
phene sheet with transition
metals; this mode of bonding is
expected to exert the minimum
perturbation of the structural in-
tegrity of the sp2-hybridized C
atoms.[33] It seems reasonable to
propose that the reactivity of
the metal(0) precursors with the
perfect inner CNT surface should
be lower than that of the other
precursors that show a +3 oxi-
dation state and easily available
coordination sites. Indeed, in the
case of Ru2 and Pd2, the reac-
tion with the graphene surface
is preceded by ligand dissocia-
tion. Instead, a higher reactivity
of the metal(0) precursors to-
wards the external ꢀCOOH-deco-
rated CNT surface is proposed,
possibly by an oxidative addi-
tion. The low filling yields ach-
ieved with Ru4, a very reactive
Ru0 compound, supports this hy-
pothesis. On the other hand
metal(II) and metal(III) complexes
are expected to be more reac-
tive towards the internal surface
as coordination will increase the
electronic density on the metal,
rather than with the external ꢀ
COOH groups, as the oxidative
addition on metal(II) or (III) is not
favorable. Thus, if the presence
of surface ꢀCOOH groups is
often associated with potential
anchoring sites for the metal
precursor, our study showed
that it depends on the precursor
reactivity. Additionally, it is possi-
ble that ꢀCOOH group hydration
(the water content of acetone is
0.23 wt%) lowers their reactivity
and displaces the thin organic
solution film from the outer to
Figure 2. TEM micrographs and NP size distribution (NPin/NPout) for a) Ru/CNT1, b) Co/CNT1, c) Pd/CNT1, and
d) the mean NP size distribution for the three catalysts.
Figure 3. TEM micrographs and NP size distribution (NPin/NPout) for a) Ru/CNT2, b) Co/CNT2, c) Pd/CNT2, and
d) mean NP size distribution for the three catalysts.
cursors with a higher nuclearity, such as dimers (Pd2) or trim-
ers (Ru2) allow higher metal loadings and a larger particle size
to be reached. The use of these precursors, for which the
metal is in the zero oxidation state in a pseudo-octahedral en-
the inner surface.[16] To confirm this hypothesis we performed
experiments at 1, 10, and 20 wt% water in acetone (TEM
images are shown in Figure S14). A moderate increase of the
yield was observed for 1–10 wt% H2O, which corresponds to
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3599

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Figure 4. Influence of CNT BET surface area on the filling yield (the reported
filling yields are average values for the Ru, Pd, and Co catalysts).
20–200 equivalents of H2O relative to the ꢀCOOH surface func-
tional groups.
Interestingly, mixing two precursors to produce a bimetallic
catalyst (Ru1Pd1/CNT1) has no significant effect on the filling
yield. Finally, we investigated the effect of CNT diameter on
the filling yield. The Ru1Pd1/CNT3 catalyst shows a metal
loading of 3.18% (2.0% Pd), a mean NP size of 2.35 nm, and
a filling yield of 84%. For CNT3, even if the existence of exo-/
endohedral ꢀCOOH functionalization limits the surface molecu-
lar recognition, the increase of the CNT internal diameter
allows the filling yield to be improved.[35] To further support
the importance of surface molecular recognition, we trans-
formed the ꢀCOOH groups of CNT3 into long (16 C atoms)
amide groups to produce CNT4. This was realized by reacting
CNT3 with thionyl chloride, and hexadecylamine (HDA), as de-
scribed elsewhere.[19] It is expected that this functionalization
will mainly affect the external surface because of steric hin-
drance of the long HDA molecules. In that case, the external
surface of the large-diameter CNT4 is passivated by long alkyl
chains, and the filling yield increases to 92% (Figure 5). The
metal loading of Ru/CNT4 is 3.08% (1.81% Pd), and the mean
NP size is 2.27 nm. The obtained results show that besides the
CNT opening, surface molecular recognition is an important
aspect to take into consideration in improving the filling yield
of CNTs.
Figure 5. TEM micrographs and particle size distribution of a) Pd1Ru1/CNT3
and b) Pd1Ru1/CNT4.
lytic activity compared with those located on the outer surface.
The chemoselectivity of the partially confined Au catalyst
seems unique as competitive hydrogenation at the conjugated
C=O bond into COL is avoided, which allows 90% selectivity
for HCAL.[39] Bimetallic PtRu NPs have been selectively confined
inside or deposited outside large-diameter CNTs (50 nm inter-
nal diameter), and a remarkable influence of the NP location
on catalyst performance has been evidenced.[19] The confined
PtRu NPs display a significantly higher selectivity and activity
for COL formation.
The catalytic performances of the Ru, Co, and Pd supported
catalysts reported in Table 2 have been evaluated for CAL hy-
drogenation at 1008C and 20 bar of H2. The results are report-
ed in Figure 6, in which the turnover frequency (TOF) [hꢀ1] and
selectivity at isoconversion, Ru (COL, 88% conversion), Co
(COL, 35% conversion), and Pd (HCAL, 100% conversion), are
reported.
Hydrogenation of cinnamaldehyde
One of the first reported applications of CNTs in heterogene-
ous catalysis was their use as supports for Ru NPs in the hydro-
genation of cinnamaldehyde (CAL).[37] CAL contains both a C=C
and a C=O bond in an a,b-unsaturated arrangement. Depend-
ing on which of the bonds are activated, hydrocinnamalde-
hyde (HCAL), cinnamyl alcohol (COL), or phenyl propanol
(HCOL) can be produced. CNT-confined Pd,[38] PtRu,[19] and
Au[39] NPs have shown high catalytic performances in this reac-
tion. Pd is highly selective for HCAL, but not active for COL for-
mation. Tessonnier et al. reported the confinement effect of Pd
NPs deposited on the inner walls of large-diameter CNTs
(50 nm internal diameter) for the selective hydrogenation of
CAL to HCAL.[38] Zhang et al. have shown that gold NPs partial-
ly confined in the cavity of small-diameter CNTs (2–5 or 5–
10 nm internal diameter) show a striking enhancement of cata-
Although parameters other than confinement, such as the
NP size or the CNT surface chemistry, can affect the selectivity
and the activity of these catalysts for CAL hydrogenation,[40]
Figure 6 clearly shows an influence of the filling yield on the
catalytic performances. The effect is more pronounced for the
Ru and Pd catalysts than for the Co one. Ru and Pd are the
most active catalysts. As expected, Ru is the most selective
metal for COL production, and the major product of the reac-
tion is HCOL (ꢂ60%). The Pd catalysts are the most selective
for HCAL synthesis, and the other product of the reaction is
HCOL. The use of the bimetallic Pd1Ru1/CNT catalyst allows
the COL selectivity to be increased to 38% (HCOL 40%, HCAL
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3600

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ment effects on the catalytic results obtained with CNT-sup-
ported metal NPs.
Conclusions
The preparation of metal catalysts supported on carbon nano-
tubes (CNTs) often involves a CNT purification/functionalization
step with nitric acid. In most reports, the issue of nanoparticle
(NP) location (exohedral vs. endohedral grafting) is ignored. In
this study, we investigated the effects of many parameters,
which included CNT pretreatment and diameter, the nature of
the metal, the nature of the metal precursor, and the nature of
the solvent on the metal NP location. The results obtained for
Ru, Pd, PdRu, and Co NPs have shown that three factors
should be taken into consideration to achieve high filling
yields: 1) the opening of the CNT inner cavity that can be ach-
ieved by nitric acid oxidation, but other treatments are also ef-
[43]
ficient such as air[42] or CO2 oxidation and treatment with
steam[44] or sulfuric acid/potassium permanganate,[45] 2) the ap-
propriate choice of solvent(s) in terms of viscosity but mainly
surface tension, and 3) molecular surface recognition, as the
metal precursor should present a higher affinity for the perfect
inner surface than the outer, often defective one. This latter
point is by far the most difficult to optimize as it involves com-
plex chemistry between the carbon surface and coordination
compounds. We have shown that the control of these factors
can allow modulation of the filling yield of metal NPs between
10 and 80%. We have also shown the impact of the confine-
ment of the active phase on the activity and selectivity of the
produced catalysts for cinnamaldehyde hydrogenation. Thus,
we hope that this study will prompt researchers who study
CNT-supported metal catalysts to integrate the possibility of
confinement effects to rationalize the catalytic results. Indeed,
even if a few studies have reported that the confinement
effect is not helpful for all reactions, and CNT-supported cata-
lysts do not improve the activity for all reactions,[46] in many
cases it does.
Figure 6. Selectivity and TOF of the prepared a) Ru, b) Co, and c) Pd
catalysts.
22%) with a TOF of 240 hꢀ1. Regardless of the CNTs used, Co
shows a moderate activity and low selectivity towards COL,
and HCAL is the major product (ꢂ60%) followed by HCOL
(ꢂ30%).
Experimental Section
Catalyst preparation
MWCNTs treatment
The higher TOF obtained at a high filling yield could be cor-
related to the smaller NPin size compared to the NPout size. An-
other explanation could arise from the enrichment of reactants
inside the CNT channels. Molecules such as H2, alkanes, al-
kenes, and carbon tetrachloride have been reported to bind
more strongly on the interior surface of CNTs.[41] By combining
first principles calculations with Monte Carlo simulation, Bao
et al. have shown that both CO and H2 molecules are enriched
in a pressure range of 1–9 MPa inside single-walled CNT
(SWCNT) channels.[3a] As positive orders are generally reported
both for H2 and cinnamaldehyde in the kinetic law of this reac-
tion, the enrichment of these two reactants inside the CNT
channels should induce a higher TOF. This latter effect could
also induce differences in selectivity. These results highlight
the possible role of too often unidentified/neglected confine-
Small-diameter MWCNTs produced by catalytic chemical vapor
deposition were supplied by Nanocyl, Belgium (purity of as-re-
ceived CNTs above 95 wt%). Large-diameter CNTs (Pyrograph III)
were supplied by Applied Sciences. Their specific surface area,
measured by N2 adsorption at liquid N2 temperature, was
38 m2 gꢀ1. Their average outer diameter is 120–150 nm, and their
average inner diameter 50–70 nm. The as-received CNTs were treat-
ed with a 1:1 mixture of concentrated H2SO4/water at 1408C for
3 h to remove amorphous carbon and residual metal catalyst. In
our study, purified Nanocyl CNTs (2 g, CNT0) were suspended in
concentrated nitric acid (80 mL, 65%) and heated to reflux at
1408C for 3 h. After cooling, the CNTs were collected by filtration,
washed with distilled water until neutralization of the filtrate (pH
ꢂ6), and dried in air at 1108C for 2 d to obtain CNT1. In addition,
CNT0 was ball-milled for 78 h in a Pulverisette 0 Fritsch apparatus
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3601

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
to obtain CNT2. Samples CNT1HT and CNT2HT were prepared by
secondary heat treatment of CNT1 under Ar and CNT2 under Ar/
H2 (10% H2 v/v), respectively, at 9008C for 2 h. The as-received Py-
rograph III CNTs were oxidized by concentrated nitric acid (65%) at
reflux for 3 h to produce CNT3. CNT3 (1 g) was further refluxed in
a solution of thionyl chloride (30 mL) for 24 h at 708C under Ar,
then the solvent was evaporated under vacuum and the CNTs
were dried under vacuum for 24 h. Chlorinated CNTs (1 g) were re-
acted with hexadecylamine (HDA, 1 g) in THF (40 mL) at 658C for
24 h. Finally, THF was evaporated and the CNTs were washed with
ethanol (100 mL) to remove excess amine, and dried at 1108C for
24 h under vacuum to obtain CNT4.
The selectivity and the conversion were calculated after 3 h of re-
action based on GC analysis.
Acknowledgements
T.T.NG. thanks the USTH consortium and the Ministry of Educa-
tion and Training of Vietnam for her PhD grant. We thank Prof.
J. L. Faria (Porto University, Portugal) for his help in TPD experi-
ments and Prof. O. Ersen (Strasbourg University, France) for his
help in electron tomography.
Keywords: cobalt · hydrogenation · nanotubes · palladium ·
Preparation of metal-filled CNTs
ruthenium
Different metallic precursors were used and dissolved in appropri-
ate solvents, depending on their solubility, with a concentration
calculated as to give a theoretical value of 5 wt% metal/CNT cata-
lysts after the impregnation step: [Ru(NO)(NO3)3] (Ru1), [Ru3(CO)12]
(Ru2), [Ru(COD)(COT)] (Ru4), Co(NO3)2·6H2O (Co1), and Pd(NO3)2-
(H2O)2 (Pd1) in acetone, RuCl3 (Ru3) in isopropanol, and [Pd2(dba)3]
(Pd2) in toluene. CNTs (200 mg) were dispersed in a 20 mL solution
of the metallic precursor in the suitable solvent by ultrasonication
for 60 min. After ultrasonic treatment the mixture was stirred at RT
for 12 h and then filtered. For Ru1, Ru2, Ru3, Co1, and Pd1 precur-
sors, the solids were dried at 1108C for 24 h and then reduced at
3008C for 2 h in a furnace under flowing H2/Ar (10% H2 v/v). For
Pd2, the solid was dried under vacuum and then reduced at RT
overnight in a batch autoclave under 3–5 bar of H2.
[1] a) X. Pan, X. Bao in Nanomaterials in Catalysis, Wiley-VCH, Weinheim,
2013, pp. 415–441; b) R. R. Bacsa, P. Serp, in Carbon Meta-Nanotubes,
Wiley, 2011, pp. 163–221; c) X. Pan, X. Bao, Acc. Chem. Res. 2011, 44,
553–562; d) P. Serp, E. Castillejos, ChemCatChem 2010, 2, 41–47;
e) D. S. Su, S. Perathoner, G. Centi, Chem. Rev. 2013, 113, 5782–5816;
f) T. W. Chamberlain, T. Zoberbier, J. Biskupek, A. Botos, U. Kaiser, A. N.
Khlobystov, Chem. Sci. 2012, 3, 1919–1924.
[2] A. La Torre, M. d. C. Gimꢁnez-Lꢂpez, M. W. Fay, G. A. Rance, W. A. Solo-
monsz, T. W. Chamberlain, P. D. Brown, A. N. Khlobystov, ACS Nano
2012, 6, 2000–2007.
[3] a) J. Guan, X. Pan, X. Liu, X. Bao, J. Phys. Chem. C 2009, 113, 21687–
21692; b) P. Kondratyuk, Y. Wang, J. K. Johnson, J. T. Yates, J. Phys. Chem.
B 2005, 109, 20999–21005.
[4] a) W.-H. Zhao, B. Shang, S.-P. Du, L.-F. Yuan, J. Yang, X. C. Zeng, J. Chem.
Phys. 2012, 137, 034501–034508; b) Y. Jiao, A. Du, M. Hankel, S. C.
Smith, Phys. Chem. Chem. Phys. 2013, 15, 4832–4843.
[5] G. Arora, N. J. Wagner, S. I. Sandler, Langmuir 2004, 20, 6268–6277.
[6] A. Alexiadis, S. Kassinos, Mol. Simul. 2008, 34, 671–678.
[7] H. Kim, W. Sigmund, J. Cryst. Growth 2005, 276, 594–605.
[8] S. Liu, J. Zhu, Y. Mastai, I. Felner, A. Gedanken, Chem. Mater. 2000, 12,
2205–2211.
[9] a) W. Chen, X. Pan, M.-G. Willinger, D. S. Su, X. Bao, J. Am. Chem. Soc.
2006, 128, 3136–3137; b) W. Chen, X. Pan, X. Bao, J. Am. Chem. Soc.
2007, 129, 7421–7426.
[10] G. A. Rance, W. A. Solomonsz, A. N. Khlobystov, Chem. Commun. 2013,
49, 1067–1069.
[11] F. Goettmann, C. Sanchez, J. Mater. Chem. 2007, 17, 24–30.
[12] a) D. Wang, G. Yang, Q. Ma, M. Wu, Y. Tan, Y. Yoneyama, N. Tsubaki, ACS
Catal. 2012, 2, 1958–1966; b) Q. Shao, L. Huang, J. Zhou, L. Lu, L.
Zhang, X. Lu, S. Jiang, K. E. Gubbins, Y. Zhu, W. Shen, J. Phys. Chem. C
2007, 111, 15677–15685; c) H. Kyakuno, K. Matsuda, H. Yahiro, Y. Inami,
T. Fukuoka, Y. Miyata, K. Yanagi, Y. Maniwa, H. Kataura, T. Saito, M.
Yumura, S. Iijima, J. Chem. Phys. 2011, 134, 244501–244514.
[13] D. S. Su in Nanomaterials in Catalysis, Wiley-VCH, Weinheim, 2013,
pp. 331–374.
[14] H. Ma, L. Wang, L. Chen, C. Dong, W. Yu, T. Huang, Y. Qian, Catal.
Commun. 2007, 8, 452–456.
[15] C. Wang, S. Guo, X. Pan, W. Chen, X. Bao, J. Mater. Chem. 2008, 18,
5782–5786.
Catalyst characterization
TEM was performed by using a JEOL JEM-1011 microscope coupled
with an energy dispersive X-ray spectrometer (EDS), using acceler-
ating voltages of 100 kV. The samples were dispersed by sonication
in ethanol, then deposited on a holey carbon TEM grid and dried.
TGA in air was conducted by using a Setaram apparatus using
a temperature program of 25–10008C with a heating rate of
108Cminꢀ1. The specific surface area, pore volume, and pore size
were measured by N2 physisorption at P/P0 =0.99 at liquid nitrogen
temperature by using a Belsorpmini (BEL Japan Inc.) apparatus.
The specific surface area was calculated according to BET theory,
and the pore size and the pore volume were calculated by the Bar-
rett–Joyner–Halenda (BJH) method based on the desorption
branch of the isotherm. Raman spectra were recorded by using
a Labram HR800 of Jobin Yvon Raman spectrometer (excitation
wavelength=632.82 nm). TPD profiles of CO and CO2 were ob-
tained by using an AMI-200 (Altamira Instruments) apparatus. In
a typical experiment, the sample (100 mg) was subjected to
a 58Cminꢀ1 linear temperature rise up to 11008C under a He flow
of 25 sccm. The amounts of CO (m/z=28) and CO2 (m/z=44) des-
orbed from the samples were monitored with a mass spectrometer
(Dymaxion 200 amu, Ametek).
[16] J.-P. Tessonnier, O. Ersen, G. Weinberg, C. Pham-Huu, D. S. Su, R. Schlçgl,
ACS Nano 2009, 3, 2081–2089.
[17] a) Q. Fu, W. Gisela, D.-s. Su, New Carbon Mater. 2008, 23, 17–20; b) A. La
Torre, M. W. Fay, G. A. Rance, M. del Carmen Gimenez-Lopez, W. A. Solo-
monsz, P. D. Brown, A. N. Khlobystov, Small 2012, 8, 1222–1228.
[18] W. A. Solomonsz, G. A. Rance, M. Suyetin, A. La Torre, E. Bichoutskaia,
A. N. Khlobystov, Chem. Eur. J. 2012, 18, 13180–13187.
[19] E. Castillejos, P.-J. Debouttiꢃre, L. Roiban, A. Solhy, V. Martinez, Y. Kihn,
O. Ersen, K. Philippot, B. Chaudret, P. Serp, Angew. Chem. 2009, 121,
2567–2571; Angew. Chem. Int. Ed. 2009, 48, 2529–2533.
[20] F. Winter, G. Leendert Bezemer, C. van der Spek, J. D. Meeldijk, A. Jos
van Dillen, J. W. Geus, K. P. de Jong, Carbon 2005, 43, 327–332.
Cinnamaldehyde hydrogenation
The reactions were performed in a stirred (900 rpm) autoclave (Top
Industrie) under 20 bar of H2 at 1008C for 3 h. A solution of CAL
(2 g) in dioxane (40 mL) and catalyst (0.1 g) were transferred to the
reactor. The mixture was purged five times with H2. At 1008C, the
system was pressurized under 20 bar of H2. The variation of H2
pressure was monitored and the liquid phase was analyzed by GC.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3602

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
[21] H. Friedrich, P. E. de Jongh, A. J. Verkleij, K. P. de Jong, Chem. Rev. 2009,
109, 1613–1629.
[33] E. Bekyarova, S. Sarkar, S. Niyogi, M. E. Itkis, R. C. Haddon, J. Phys. D
2012, 45, 154009.
[22] a) S. Guo, X. Pan, H. Gao, Z. Yang, J. Zhao, X. Bao, Chem. Eur. J. 2010, 16,
5379–5384; b) H. Friedrich, S. Guo, P. E. de Jongh, X. Pan, X. Bao, K. P.
de Jong, ChemSusChem 2011, 4, 957–963; c) L. Wang, L. Ge, T. E. Ruf-
ford, J. Chen, W. Zhou, Z. Zhu, V. Rudolph, Carbon 2011, 49, 2022–
2032.
[23] Z. Yang, S. Guo, X. Pan, J. Wang, X. Bao, Energy Environ. Sci. 2011, 4,
4500–4503.
[24] W. Chen, Z. Fan, L. Gu, X. Bao, C. Wang, Chem. Commun. 2010, 46,
3905–3907.
[25] W. Chen, Z. Fan, B. Zhang, G. Ma, K. Takanabe, X. Zhang, Z. Lai, J. Am.
Chem. Soc. 2011, 133, 14896–14899.
[26] X. Pan, Z. Fan, W. Chen, Y. Ding, H. Luo, X. Bao, Nat. Mater. 2007, 6,
507–511.
[27] I. Gerber, M. Oubenali, R. Bacsa, J. Durand, A. GonÅalves, M. F. R. Pereira,
F. Jolibois, L. Perrin, R. Poteau, P. Serp, Chem. Eur. J. 2011, 17, 11467–
11477.
[28] P. Serp in Carbon Materials for Catalysis, Wiley, 2008, pp. 309–372.
[29] M. O. Bruno Machado, R. M. Axet, T. Trang Nguyen, M. Tunckol, O. Ersen,
I. C. Gerber, P. Serp, ACS Catal., in press.
[34] J. A. Menꢁndez, L. R. Radovic, B. Xia, J. Phillips, J. Phys. Chem. 1996, 100,
17243–17248.
[35] O. Ersen, J. Werckmann, M. Houllꢁ, M.-J. Ledoux, C. Pham-Huu, Nano
Lett. 2007, 7, 1898–1907.
[36] K. Nakada, A. Ishii, Solid State Commun. 2011, 151, 13–16.
[37] J. M. Planeix, N. Coustel, B. Coq, V. Brotons, P. S. Kumbhar, R. Dutartre, P.
Geneste, P. Bernier, P. M. Ajayan, J. Am. Chem. Soc. 1994, 116, 7935–
7936.
[38] J.-P. Tessonnier, L. Pesant, G. Ehret, M. J. Ledoux, C. Pham-Huu, Appl.
Catal. A 2005, 288, 203–210.
[39] X. Zhang, Y. C. Guo, Z. Cheng Zhang, J. S. Gao, C. M. Xu, J. Catal. 2012,
292, 213–226.
[40] a) P. Gallezot, D. Richard, Catal. Rev. Sci. Eng. 1998, 40, 81–126; b) P.
Claus, Top. Catal. 1998, 5, 51–62.
[41] P. Kondratyuk, J. T. Yates, Acc. Chem. Res. 2007, 40, 995–1004.
[42] P. M. Ajayan, T. W. Ebbesen, T. Ichihashi, S. Iijima, K. Tanigaki, H. Hiura,
Nature 1993, 362, 522–525.
[43] S. C. Tsang, P. J. F. Harris, M. L. H. Green, Nature 1993, 362, 520–522.
[44] G. Tobias, L. Shao, C. G. Salzmann, Y. Huh, M. L. H. Green, J. Phys. Chem.
B 2006, 110, 22318–22322.
[30] A. Solhy, B. F. Machado, J. Beausoleil, Y. Kihn, F. GonÅalves, M. F. R. Per-
eira, J. J. M. ꢄrf¼o, J. L. Figueiredo, J. L. Faria, P. Serp, Carbon 2008, 46,
1194–1207.
[45] H. Hiura, T. W. Ebbesen, K. Tanigaki, Adv. Mater. 1995, 7, 275–276.
[46] Z. Li, Z.-X. Chen, G.-J. Kang, X. He, Catal. Today 2011, 165, 25–31.
[31] Z. Kꢂnya, J. Zhu, K. Niesz, D. Mehn, I. Kiricsi, Carbon 2004, 42, 2001–
2008.
Received: July 3, 2013
Published online on October 7, 2013
[32] Y. A. Kim, T. Hayashi, Y. Fukai, M. Endo, T. Yanagisawa, M. S. Dresselhaus,
Chem. Phys. Lett. 2002, 355, 279–284.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 3595 – 3603 3603