Covalently and non-covalently immobilized clusters onto nanocarbons as catalysts precursors for cinnamaldehyde selective hydrogenation

Article abstract of DOI:10.1016/j.jcat.2015.06.003

Abstract Ru-based nanoparticles were deposited on carbon nanotubes and graphene via an organometallic approach involving mixed-metal clusters modified with appropriate ligands. These ligands allowed either covalent or non-covalent π-π interactions with the carbonaceous surfaces. The immobilized clusters were then coalesced at different temperatures to give carbon-supported nanoparticles of different sizes. The obtained catalysts were tested in the selective hydrogenation of cinnamaldehyde. It was found that the nature of the metal(s), support nature, incorporation method and activation temperature all had a profound influence on activity and selectivity. Interestingly, the selectivity could be shifted from cinnamyl alcohol (COL) to hydrocinnamaldehyde (HCAL) by changing the reaction solvent. The best catalysts gave a very high selectivity in cinnamyl alcohol, which is not the more thermodynamically favored product, and could be reused several times without loss of activity or selectivity.

Full text of DOI:10.1016/j.jcat.2015.06.003

Journal of Catalysis 329 (2015) 389–400
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Covalently and non-covalently immobilized clusters onto nanocarbons
as catalysts precursors for cinnamaldehyde selective hydrogenation
Charles Vriamont 1,2, Tommy Haynes 1, Emily McCague-Murphy, Florence Pennetreau, Olivier Riant,
⇑
Sophie Hermans
IMCN Institute, Université catholique de Louvain, Place L. Pasteur 1, 1348 Louvain-la-Neuve, Belgium
a r t i c l e i n f o
a b s t r a c t
Article history:
Ru-based nanoparticles were deposited on carbon nanotubes and graphene via an organometallic
approach involving mixed-metal clusters modified with appropriate ligands. These ligands allowed either
Received 9 March 2015
Revised 18 May 2015
Accepted 1 June 2015
covalent or non-covalent
p–p interactions with the carbonaceous surfaces. The immobilized clusters
were then coalesced at different temperatures to give carbon-supported nanoparticles of different sizes.
The obtained catalysts were tested in the selective hydrogenation of cinnamaldehyde. It was found that
the nature of the metal(s), support nature, incorporation method and activation temperature all had a
profound influence on activity and selectivity. Interestingly, the selectivity could be shifted from cin-
namyl alcohol (COL) to hydrocinnamaldehyde (HCAL) by changing the reaction solvent. The best catalysts
gave a very high selectivity in cinnamyl alcohol, which is not the more thermodynamically favored
product, and could be reused several times without loss of activity or selectivity.
Keywords:
Clusters
Cinnamaldehyde
Nanoparticles
Bimetallic catalysts
Hydrogenation
Graphene
Ó 2015 Elsevier Inc. All rights reserved.
Carbon nanotubes
Non-covalent immobilization
Grafting
1. Introduction
costly chemicals [2,6,7]. The development of heterogeneous cata-
lysts on various supports, environmentally friendlier and able to
Supported catalysts to be competitive should simultaneously
fulfill the following criteria: a high activity and selectivity coupled
to stability and ease of recovery over time [1]. The chemoselective
selectively hydrogenate both functions separately was thus consid-
ered as a valuable option by many authors [1–3,5,8–17].
Since the transformation in the two hydrogenated products,
that is hydrocinnamaldehyde and cinnamyl alcohol, is of chemical
interest, cinnamaldehyde was chosen as benchmark substrate. The
abbreviations used, namely cinnamaldehyde (CAL), hydrocin-
namaldehyde (HCAL), cinnamyl alcohol (COL) and hydrocinnamyl
alcohol (HCOL), as well as the structure of the molecules, are
shown in Fig. 1. COL is generally described as the most challenging
product to obtain and is moreover the most valuable: It is a critical
intermediate in the production of chemicals such as flavors, phar-
maceuticals and perfumes [1,18]. Nevertheless C@C double bond
hydrogenation to obtain HCAL as a product was recently reported
as intermediate in the synthesis of medicines for the treatment of
HIV, which makes it also a challenging product [19].
The chemoselectivity toward both products is affected by a con-
siderable number of parameters such as particles metal(s) nature
and type of faces exposed, steric and electronic effects, catalyst
preparation, activation procedure, selected support or even reac-
tion conditions [1,2,7,21]. A comprehensive review written by
Gallezot et al. deals with this topic [2]. Among these parameters,
the most explored is undoubtedly the metal nature. Studies
hydrogenation of
a,b-unsaturated aldehydes has been often
selected to evaluate the above-mentioned criteria [1–5]. These
compounds present two hydrogenable sites which means that
two competitive reaction pathways can be followed: the selective
transformation of the aldehyde into the corresponding unsaturated
alcohol or the thermodynamically more favoured reduction of the
olefin bond. These reactions pathways determine the chemoselec-
tivity of the transformation and can be moreover complicated by
side reactions occurring either on the metal active phase or on
the support [3]. The Meerwein–Pondorf–Verley reaction applied
to these aldehydes in homogeneous catalysis allows to selectively
obtain the unsaturated alcohol with a high conversion but a large
amount of waste: This methodology is therefore prescribed for
⇑
Corresponding author. Fax: +32 10 47 2330.
E-mail address: sophie.hermans@uclouvain.be (S. Hermans).
These authors contributed equally.
Present address: Imperial College London, South Kensington Campus, London
1
2
SW7 2AZ, UK.
http://dx.doi.org/10.1016/j.jcat.2015.06.003
0021-9517/Ó 2015 Elsevier Inc. All rights reserved.

390
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
Fig. 1. The structure and conversion routes of a,b-cinnamaldehyde.
conducted on metals deposited on carbon supports revealed that Pt
and Ru, because of their acceptable activity, selectivity and stabil-
ity, feature among the best considered metals to obtain the desired
unsaturated alcohol [21]. In addition, a combination of several
metals can greatly improve the selectivity toward the desired pro-
duct [12,18,20,22,23]: PtARu bimetallic catalysts gave impressive
activity and selectivity [1,24–27]. This activity/selectivity enhance-
ment is a result of synergism effect [28] which can be explained
mainly by electronic and geometric effects: The former is due to
an alloy formation or intimate contacts between metals [29] while
the latter comes from effects such as a modification of the metal
dispersion [30,31] or a surface enrichment [32–34]. In general,
both effects go hand in hand. Synthetic routes to obtain supported
bimetallic nanoparticles are well documented and range from sim-
ple to more sophisticated methods such as ion exchange, impreg-
nations methods, homogeneous deposition precipitation or even
colloidal microwave process [22,35]. In almost all cases, these
multi-metallic nanoparticles arise from separate monometallic
precursors. The different metals are thus not necessarily in a close
contact which is assumed to be of importance for catalysis [3]. The
use of clusters [36] as nanoparticles precursors can de facto circum-
vent this problem. Relatively few studies deal with their use as sin-
gle precursors for high performance nanocatalysts [37–42].
oxide/reduced graphene oxide (GO/rGO) are the most commonly
used supports [1–3,5,8–17,43]. Among them, nanocarbons (CNTs,
GO, rGO) have been selected for this work. It seems interesting
to evaluate in terms of activity and selectivity the possible differ-
ence between these supports arising from metal dispersion and
electronic effects. Indeed, graphene is a 2D system compared to
the 1D CNTs [2]. Moreover, oxygenated functions, present in high
amounts in GO (and rGO on a smaller scale), could also be of
importance for the selectivity [44,45].
The present work reports an unconventional methodology to
covalently and non-covalently anchor Ru-based mono- and
RuAPt, RuAAu and RuAPtAAu multi-metallic precursors onto
nanocarbons. The undertaken route is depicted in Fig. 2, which
shows schematically the successive steps envisaged. To this end,
multi-walled carbon nanotubes (MWNTs) and graphene were
selected as supports. Two pathways were attempted to obtain
the target materials: a non-covalent and a covalent methodology.
When the covalent pathway was undertaken, a pre-functionalization
step is required (I): Acyl chloride functions were formed on GO
or MWNTs on oxygenated functions located at defects- and
end-sites. A recently reported radical approach involving xanthate
chemistry applied to MWNTs was also selected to pre-
functionalize nanocarbons [46]. In the next step, mono- and
multi-metallic clusters were selected as single source metal pre-
cursors and immobilized onto nanocarbons by a covalent or a
non-covalent strategy (II). In the latter case, a polar solvent and
polyaromatics, that is pyrenes moieties, are required to perform
The catalyst support, whose main task consists in dispersing the
metal on its surface, displays also a significant and sometimes crit-
ical role on both the activity and the selectivity in the targeted
selective hydrogenation [3,4]. Oxides, such as silica and alumina,
and carbon based supports including activated carbon, carbon
nanofibers, carbon nanotubes (CNTs) and recently graphene
this non-covalent immobilization involving
p–p stacking [47]. All
these catalysts, once thermally activated at different temperatures
Fig. 2. General methodology to obtain mono- or multi-metallic nanoparticles arising from a single source metal precursor.

C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
391
(III), were tested in the selective hydrogenation of cinnamaldehyde
to assess and compare their activity and selectivity in order to find
the best route to obtain selectively either COL or HCAL products.
The recycling was finally tested for the best catalysts found.
If ruthenium and platinum have been selected as metals for
their well-known properties toward the selective hydrogenation
photoelectron spectrometer from Surface Science Instruments
(USA) equipped with a monochromatized microfocus Al X-ray
source. Samples were stuck onto small sample holders with
double-face adhesive tape and then placed on an insulating
home-made ceramic carousel (MacorÒ, Switzerland). Charge
effects were avoided by placing a nickel grid above the samples
and using a flood gun set at 8 eV. The energy scale was calibrated
with reference to the Au4f7/2 peak at 84 eV and the binding energies
were calculated with respect to the CA(C,H) component of the
C1s peak fixed at 284.8 eV. Data treatment was performed with
the CasaXPS program (Casa Software Ltd., UK). The peaks were
decomposed into a sum of Gaussian/Lorentzian (85/15) after
subtraction of a Shirley type baseline.
The elemental analyses (C, H, N, O, F, S, P, Ru, Pt, Au) were real-
ized by MEDAC Ltd., UK. Nuclear magnetic resonance (NMR) spec-
tra were recorded on BRUKER spectrometers (121 MHz for 31P).
Mass spectra (MS) were recorded on a Q-Exactive orbitrap from
ThermoFisher. Samples were ionized by ESI (capillary temperature:
250 °C, vaporizer temperature: 250 °C, sheath gas flow rate: 20).
The clusters and their corresponding adducts were analyzed by
infrared spectroscopy on a Bruker Equinox 55 spectrometer with
a solution cell from Perkin Elmer. The powder catalysts were ana-
lyzed by transmission electron microscopy (TEM) with a LEO922
energy filter transmission electron microscope. The powder sam-
ples were suspended in hexane under ultrasonic treatment, then
a drop of the supernatant was deposited on a holey carbon film
supported on a copper grid, which was dried overnight under vac-
uum at room temperature before analysis. Gas Chromatography
(GC) was performed on a GC trace, Finnigan Mat equipped with
an AS-3000 Autosampler (FID detector). The GC column is a
of
a,b-unsaturated aldehydes, researches conducted on gold in
these hydrogenation reactions have been appearing only in the last
few years [48–51]. Au possesses remarkable activity in the tar-
geted reaction. Nevertheless, the high selectivity obtained toward
the desired COL product was found to be correlated with the use
of oxide supports. Gold on nanocarbons as catalysts have been
far less studied and an opposite trend, namely a high selectivity
toward HCAL product, was reported in the only paper dealing with
gold deposited on nanotubes [52].
2. Experimental section
2.1. Generalities and instrumental
MWNTs were obtained from Nanocyl (Belgium) (Thin MWNTs,
95+% C purity). Graphene Oxide (GO) and Reduced Graphene Oxide
(rGO) were purchased from NanoInnova technologies (Spain).
Unless otherwise stated, all the manipulations were carried out
under an atmosphere of Ar using standard Schlenk techniques
and with anhydrous solvents. Hexane was distilled from sodium
benzophenone, dichloromethane was distilled from CaH2 and
anhydrous acetone was used as received (Fisher Chemicals).
Samples before and after the thermal activation were analyzed
by X-ray photoelectron spectroscopy (XPS) which was carried
out at room temperature with a SSI-X-probe (SSX 100/206)
Chirasil-Dex CB, 25 m ꢀ 25 mm ꢀ 0.25
lm (Agilent). Gas vector:
Helium, flow: 1.2 ml/min, injector split/split less ratio 1/30. The
PPh2
PPh2
Non-covalent
Thermal activation
O
NH
O
NH
600°C
+
No treatment
Nanocarbons
Covalent
900°C
Defects
H2N
O
PPh2
O
Δ
N
H
1.HNO3cc
2.SOCl2
PPh2
Cl
Nanocarbons
+
1300°C
Sidewall
Nanocarbons
O
Covalent
PPh2
O
N
OC6F5
S
H
Particles size thermally dependent
OC6F5
S
PPh2
O
O
+
H2N
S
S
Dilauroyl peroxide
S
O
S
O
Selective hydrogenation of
unsatured cinnamaldehyde
α,β
Fig. 3. Overview of the different pathways used to form multi-metallic nanoparticles on nanocarbons (gray spheres = cluster precursors; black spheres = metal
nanoparticles).

392
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
internal standard used is dodecane and the temperature program
is 120 °C (120) ? 120 °C to 150 °C (50) ? 150 °C (100).
2.4. Covalent immobilization of adducts 7 and 8 onto MWNT-Cl and G-
Cl to obtain supported catalysts C7, C8, GC7 and GC8
The proposed methodology to synthesize mono- and
multi-metallic nanoparticles supported on nanocarbons is uncon-
ventional: A detailed overview is presented in Fig. 3 and will serve
as a guiding thread in the following sections. The preparation of
the functionalized supports referenced MWNT-Cl and GO-Cl is
described in SI, Section 3.2. The decoration of carbon nanotubes
with activated esters using xanthate 13 is described in SI
Section 3.3, to give MWNTs-X.
In a typical experiment, 200 mg of MWNT-Cl or G-Cl are intro-
duced in a 100 mL Schlenk flask and submitted to ultrasound
(Ultrasonic Cleaner, VWR) in dichloromethane (15 mL) for 1 h.
The cluster 7 (200 mg, 0.147 mmol) or 8 (200 mg, 0.140 mmol) in
dichloromethane (15 mL) was added and the suspension was stir-
red at room temperature for 2 h. The resulting mixtures were fil-
tered out (on PVDF 0.22 lm pore size membrane) and washed
twice with dichloromethane (2 ꢀ 20 mL) to separate the reactions
products 7 and 8 from the supported adducts. These were vacuum
dried to obtain C7 and C8 on MWNT-Cl and GC7 and GC8 on G-Cl
corresponding to the immobilization of respectively 7 and 8 on
the two different supports.
2.2. Clusters and bifunctional ligands syntheses
The two bifunctional ligands 1 and 2 were synthesized as
reported previously [53]. The mono-, bi-, and tri-metallic clusters
[Ru6C(CO)17] (3), [Ru5PtC(CO)16] (4), [Ru6C(CO)16(Au{PPh3})2] (5)
and [Ru5PtC(CO)15(Au{PPh3})2] (6) were obtained according to lit-
erature procedures [54,59–61].
2.5. Covalent immobilization of adducts 7 and 8 onto MWNT-X to
obtain supported catalysts X7-X8
In a typical experiment, 100 mg of MWNT-X was introduced in
a Schlenk round bottom flask and submitted to ultrasound for 1 h
(Ultrasonic Cleaner, VWR) in dichloromethane (15 mL). The cluster
7 (100 mg, 0.074 mmol) or 8 (100 mg, 0.070 mmol) in dichloro-
methane (15 mL) was added and the suspension was stirred at
room temperature for 8 h. The resulting mixture was then filtered
2.3. Ligands-clusters adducts (7–12) synthesis
The synthesis of adducts 7–12 was conducted using standard
schlenk techniques. The clusters [Ru6C(CO)17] (3), [Ru5PtC(CO)16
]
(4), [Ru6C(CO)16(Au{PPh3})2] (5) or [Ru5PtC(CO)15(Au{PPh3})2] (6)
(200 mg, 1 eq (1.871 ꢁ 10ꢂ4 mol (3); 1.7232 ꢁ 10ꢂ4 mol (4);
1.0075 ꢁ 10ꢂ4 mol (5); 9.7509 ꢁ 10ꢂ5 mol (6)) were dissolved in
25 ml of dichloromethane in a Schlenk round bottom flask. One
equivalent of ligand 1 (for clusters (3) and (4)) or 2 (for clusters
(3–6)) was added to the solution and then stirred for variable dura-
tion (5 min (4), overnight (3–5), 48 h (6)). The solvent was
removed under reduced pressure. A volume of 20 mL of hexane
was added and the solution was finally filtrated to recover a solid
corresponding to the ligand-cluster coordination products 7–12.
Product 7: yield: 98% (0.249 g); IR mCO (CH2Cl2) (cmꢂ1): 2056 (s),
2030 (s); 31P NMR (121 MHz, CDCl3): d = 47.71 (s); MS (ESI): m/z
calcd for C36H20NO16PRu6 1359.93 [M+1]+, found: 1359,84 [M+1]+;
1310.96 [MꢂNH2ꢂ2CO+Na]+, 1286.00 [MꢂNH2ꢂ3CO+Na]+, 1258.09
[MꢂNH2ꢂ4CO+Na]+, 1228.11 [MꢂNH2ꢂ5CO+Na]+; 1202.22 [Mꢂ
NH2ꢂ6CO+Na]+, 1171.09 [MꢂNH2ꢂ7CO+Na]+, 1146.06 [MꢂNH2ꢂ
8CO+Na]+. Product 8: yield: 97% (0.238 g); IR mCO (CH2Cl2) (cmꢂ1):
2057 (s), 2037 (s), 2003 (w, br), 1882 (w, br); 31P NMR (121 MHz,
CDCl3): d = 29.44 (t, J = 3.197 MHz); MS (ESI): m/z calcd for
out over a PVDF (0.22 lm pore size) membrane and washed twice
with dichloromethane (2 ꢀ 20 mL) to separate the reactions prod-
ucts 7 and 8 from the supported adducts. The resulting material
was dried under vacuum overnight to get the supported adducts
X7 (from 7) and X8 (from 8).
2.6. Non-covalent immobilization of adducts 9–12 to obtain supported
catalysts P9, PG9, P10, PO10, P11, P12
In a typical experiment, 200 mg of pristine MWNTs (Nanocyl) or
rGO (NanoInnova Technologies) was introduced in a 100 mL
Schlenk flask and submitted to sonication in an ultrasonic bath
(Ultrasonic Cleaner, VWR) in acetone (15 mL) for 1 h. The clusters
9 (200 mg, 0.1228 mmol), 10 (200 mg, 0.1165 mmol), 11 (200 mg,
0.0875 mmol) or 12 (200 mg, 0.0774 mmol), dissolved in acetone
(5 mL) were added to the suspension and stirred at room temper-
ature for 1 h. The resulting powders were submitted to centrifuga-
tion (6500 rpm, 10 min) and washed three times with acetone
(6500 rpm, 10 min) to separate the non-immobilized clusters 9,
10, 11 and 12 from the supported adducts. Finally, the powders
were dried under vacuum. The following supported adducts P9,
P10, P11 and P12 on MWNTs and PG9, PG10 on rGO were obtained
and corresponded to the immobilization of respectively 9, 10, 11
and 12 onto MWNTs and 9 and 10 on rGO.
C
35H19NO15PPtRu5K 1466.50 [M+K]+; found: 1466.57 [M+K]+.
Product 9: yield: 97% (0.295 g); IR mCO (CH2Cl2) (cmꢂ1): 2056 (s),
2030 (s); 31P NMR (121 MHz, CDCl3): d = 47.27 (s); MS (ESI): m/z
calcd for C56H32NO17PRu6Na 1651.56 [M+Na]+; found: 1651.93
[M+Na]+; 1622.92 [MꢂCO+Na]+, 1594.97 [Mꢂ2CO+Na]+, 1568.96
[Mꢂ3CO+Na]+. Product 10: yield: 96% (0.284 g) IR mCO (CH2Cl2)
(cmꢂ1): 2057 (s), 2037 (s), 2003 (w, br), 1882 (w, br); 31P NMR
(121 MHz, CDCl3): d = 29.55 (t, J = 3.198 MHz); MS (ESI): m/z calcd
2.7. Thermal activation
for
C
55H32NO16PPtRu5Na 1716.63 [M+Na]+; found: 1715.99
The supported adducts were all submitted to a thermal treat-
ment in a tubular oven STF 16/450 from CARBOLITE. The samples
were placed into porcelain combustion boats and heated during
1 h at the selected temperature (heating ramp and cooling ramp:
100 °C/h) under a stream of N2/H2 (95:5) at 600 °C when supported
on graphene GO-Cl and rGO and at 900 °C or 1300 °C when sup-
ported on MWNTs.
[M+Na]+; 1662.01 [Mꢂ2CO+Na]+, 1286.00 [Mꢂ11CO+K]+,
1425.14. Product 11: IR mCO (CH2Cl2) 2067 (vw), 2052 (s),
2007 (s, sh), 2001 (vs) (cmꢂ1); 31P NMR (121 MHz, CDCl3):
d = 28.90, 42.12, 63.59, 64.02, 67.21; MS (ESI): m/z calcd
for
C
69H47Au2NNaO12P2Ru6Na [Mꢂ5CO+Na+] 2167.61 found:
2167.72. Product 12: IR mCO (CH2Cl2) (cmꢂ1): 2068 (m), 2050 (m,
sh), 2037 (s), 2014 (vs), 1967 (m); 31P NMR (121 MHz, CDCl3):
d = 28.73, 29.39, 29.61, 30.76, 35.45, 39.51, 43.01, 68.59, 69.14,
69.70, 70.22, 71.46, 71.92, 73.32; MS (ESI): m/z calcd for
2.8. Hydrogenation of cinnamaldehyde: catalytic experiments
C
86H62Au2NNaO12P3PtRu5Na 2511.76 [MꢂCꢂ3CO+Na]+, found:
The catalytic experiments were realized in a 250 mL stainless
steel PARR autoclave. The experimental conditions for hydrogena-
tion depended on the solvent used and are summarized in Table S1
(SI). Two laboratory lines, N2 and H2 lines, were used with
2511.59 [M+NaꢂCꢂ3CO]+; m/z calcd for C107H79Au2N2NaO13
-
P3PtRu5 2810.90 [M+NaꢂCꢂ4COꢂPPh3 + ligand 2]+, found:
2810.73 [M+NaꢂCꢂ4COꢂPPh3 + ligand 2]+.

C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
393
regulatory high-pressure valves. In
a
250 mL autoclave, cin-
(6). The addition of platinum (4) or gold (5) or both metals (6) to
the reference cluster 3 was realized to compare and follow the
implication of these metals on the selectivity in the targeted selec-
tive hydrogenation.
namaldehyde was introduced simultaneously with the desired sol-
vent. The catalyst was then added and the autoclave was sealed.
Afterward the system was purged 4 min with a stream of nitrogen,
and then heated up to the desired temperature. Once the desired
temperature had been reached, 50 bars of hydrogen were intro-
duced and the mixture was stirred at 1500 rpm for 2 h. The hydro-
gen pressure is left to evolve naturally during the catalytic tests
(slow decrease of pressure as H2 is consumed). Hydrogen was sub-
sequently slowly removed before cooling down the system to room
temperature. The system was finally purged for 4 min with a
stream of nitrogen. The solution was filtered out over a PVDF
3.2. Clusters-bifunctional ligand adducts syntheses
Adducts 7–12 were obtained according to the synthetic proce-
dures given in Table 1 and proceeded through a quantitative
exchange of ligands. This involves the phosphine of the ligand 1
or 2 and one of the constitutive carbonyl ligands of the clusters
3–4. Two routes are possible when clusters 5 and 6 are involved:
a phosphine-CO or phosphine-phosphine exchange. Adducts 7
and 9 were obtained instantaneously (5 min) while the obten-
tion of 8, 10–12 took more time (from overnight to 48 h). All the
reactions were followed by IR spectroscopy and the final prod-
ucts were characterized by IR, 31P NMR and MS-ESI (see SI for
characterizations).
(0.22 lm pore size) membrane and the catalyst was washed with
50 mL solvent. The resulting solution was analyzed by GC.
3. Results and discussion
3.1. Clusters and bifunctional ligands
IR spectra were recorded in the CO stretching zone between
2200 and 1600 cmꢂ1 to check the ligand exchange between bifunc-
tional ligand 1 or 2 and a carbonyl/phosphine group of clusters 3–
6. A comparison made before and after adducts syntheses is
The selected bifunctional ligands 1 and 2 (Fig. 4) were syn-
thetized on
a multi-gram scale from the starting material
1-bromobenzonitrile, and were obtained following the synthetic
route outlined in Fig. 4 [53]. Two functions of paramount impor-
tance were incorporated within 1 and 2: a triphenylphosphine
group since transition metals clusters are known to exchange read-
ily their CO ligands to phosphine [54] and an amine (1) or pyrene
(2) group to respectively interact covalently or non-covalently with
nanocarbons. Pyrenes moieties were selected because it currently
represents the only viable route to bring together nanotubes and
reported in Table 1. A shift toward lower wave numbers (cmꢂ1
)
occurred due to modification of the clusters electronic environ-
ment in the cases of adducts 7–10. When gold based clusters 11
and 12 were selected, the IR values observed before and after
adducts syntheses were close (Table 2) (see SI: Section 1.2 for IR
spectra). This might indicate ligand exchange with phosphines ini-
tially present in the starting clusters. Other characterization tech-
niques, namely 31P NMR and MS have been implemented to
solve this issue.
polyaromatic moieties through p–p stacking [55–58].
Four clusters were selected as mixed-metal molecular precur-
sors to obtain nanoparticles arising from a single source. The four
clusters were obtained according to literature procedures [54,59–
61]. These are depicted in Fig. 5 and are constituted of one, two
31P NMR and Mass Spectrometry (ESI-MS) were therefore
recorded for all these compounds to give a deeper understanding
of adducts formations. 31P NMR of products 7–10 (SI Section1.3)
showed a total disappearance of free ligands, no presence of oxi-
dized ligands and their quantitative conversion into the
or even three different metals: [Ru6C(CO)17] (3), [Ru5PtC(CO)16
]
(4), [Ru6C(CO)16(Au{PPh3})2] (5) and [Ru5PtC(CO)15(Au{PPh3})2]
Fig. 4. Synthesis of the covalent 1 and non-covalent 2 bifunctional ligands.
Fig. 5. The four selected mono-, bi- and tri-metallic clusters. CO ligands are omitted for clarity.

394
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
Table 1
Finally and in light of the obtained results, the required time to
synthesize the different adducts can be explained and greatly
depends on which metals are present in the clusters used. The cou-
pling took place instantly on the Pt site when PtARu bimetallic
clusters were chosen which is probably due to a lower dissociative
energy of the PtACO bond (578 kJ/mol) compared to RuACO bond
(648 kJ/mol) [62–63]. It could therefore explain the time difference
observed between adducts formations 7–8 and 9–10. The synthesis
of 11 is based on a phosphine–phosphine ligand exchange and can-
not be compared with the four previous adducts. The required time
to obtain adduct 12, which contains platinum, does not follow the
same model. The carbonyl group bonded to platinum in the
tri-metallic cluster 6, sterically hindered, is probably difficult to
reach. Phosphine vs. CO (except for that of Pt) ligand exchange
explains the required long vs. short time to complete these cou-
pling reactions (see SI Section 2).
Synthesis of adducts 7–12 by reaction between 1 or 2 as bifunctional ligand and 3–6
as cluster.
Cluster
Ligand
Adduct
3 = [Ru6C(CO)17
]
+
+
+
+
+
+
1
1
2
2
2
2
=
=
=
=
=
=
7
8
9
10
11
12
4 = [Ru5PtC(CO)15
3 = [Ru6C(CO)17
4 = [Ru5PtC(CO)15
5 = [Ru6C(CO)16(Au{PPh3})2]
6 = [Ru5PtC(CO)15(Au{PPh3})2]
]
]
]
Table 2
IR
mCO (cmꢂ1
bifunctional ligands.
–
) before (clusters) and after coupling reaction (adducts) with
mCO (cmꢂ1) clusters
mCO (cmꢂ1) adducts
7
8
2066 (s), 2045 (s)
2056 (s), 2030 (s)
2065 (s), 2050 (vs), 2005 (w,br), 1869 2057 (s), 2037 (s), 2003 (w,br),
3.3. Adducts incorporations onto nanocarbons
(vw,br)
1882 (w,br)
9
2066 (s), 2045 (s)
2056 (s), 2030 (s)
10 2065 (s), 2050 (vs), 2005 (w,br), 1869 2057 (s), 2037 (s), 2003 (w,br),
Incorporations of adducts 7–12 were performed on nanocar-
bons (MWNTs and graphene) by non-covalent and covalent immo-
bilizations, including sidewall and defect sites anchoring, as
outlined in Fig. 2. If the non-covalent strategy does not require
any pre-treatment, the opposite is true in the covalent cases.
(vw,br)
1882 (w,br)
11 2067 (w), 2049 (s), 2017 (vs), 1965
(w), 1821 (m)
067 (vw), 2052 (s), 2007 (s,sh),
2001 (vs)
12 2068 (m), 2038 (s), 2015 (vs), 1968
(m), 1859 (m), 1834 (m)
2068 (m), 2050 (m,sh), 2037 (s),
2014 (vs), 1967 (m)
3.3.1. Surface functionalization of nanocarbons
Given the amine pending groups present in adducts, we choose
to incorporate acyl chloride and activated ester groups at the
nanocarbon surface since these are known to readily react with
amine groups to form robust amide bonds.
corresponding adducts. When bimetallic RuAPt clusters were
selected, pseudo-triplet peaks were observed which corresponded
to satellite peaks characteristic of the 195PtA31P coupling. This indi-
cates the exact ligand exchange position, namely at the platinum
site. When gold clusters 11 and 12 were selected, the 31P NMR
analyses showed a total disappearance of both free ligand 2 and
clusters 5 and 6. The peak assignments were nonetheless not so
straightforward due to the large number of 31P NMR peaks in both.
The suggested mechanism involves an exchange of ligands
between 2 and one of the initially present triphenylphosphine
3.3.1.1. Surface modification (MWNTs and GO) to graft acyl chloride
functions. An oxidative treatment to create acidic functions at the
tubes surface was conducted on pristine MWNTs (95%+ purity),
based on a previous study [64]. The selected treatment was chosen
with the aim to oxidize the nanotubes without degrading them,
namely the use of concentrated nitric acid under reflux for 2 h.
This specific treatment is not requested when GO was used due
to the initially existing oxygenated functions on this support (see
Table 3). The acylation step, involving the use of SOCl2, was applied
in both cases and transformed carboxylic acid groups in acyl chlo-
rides functions to furnish the functionalized supports referenced
MWNT-Cl and GO-Cl [65] (see SI Section 3.2 for experimental pro-
cedures). All these samples were characterized by XPS before and
after functionalization as shown in Table 3. As expected, the ratio
O/C increased during the oxidative treatment of MWNTs. The acy-
lation step was successfully performed with SOCl2 as a reagent
regardless of the support. The initial O/C ratio was considerably
higher when GO was used (10 times higher than MWNTs-ox)
and logically the chlorine atomic percentage obtained by XPS fol-
lowed the same trend to afford a considerably higher number of
acyl chlorides functions at the surface of graphene than nanotubes.
ligand within the clusters
5 and 6 to obtain 11 and 12.
Nevertheless a CO-ligand exchange is also possible and could
explain the high number of observed peaks in these two last
adducts.
Finally, mass spectrometry gave further evidence of the synthe-
ses of adducts 7–12. The observed signals (m/z) for compounds 7–
10 gave not only the pseudo molecular ion (M+H+ for 7, M+K+ for 8
and M+Na+ for 9 and 10) but also some peaks from the gradual loss
of CO ligands. If similar evidence was not observed for adducts 11
and 12, that is no pseudo-molecular peaks, some fragments were
obtained that confirmed the bond between bifunctional ligands
and clusters 5 and 6. Furthermore, the obtained fragments con-
firmed the proposed exchange between a triphenylphosphine
ligand of 5 and 2 to get 11 (see SI Section 1.4). In addition, these
analyses indicated the presence of two different possible adducts
in the case of 12 (Phosphine–CO exchange and a double exchange:
Phosphine–CO and Phosphine–phosphine within the same
adduct). Their presence does not hamper the desired strategy
and they can therefore be used to be immobilized on nanotubes
and graphene.
3.3.1.2. Surface modification of MWNTs through the radical xanthate
chemistry. The xanthate 13 (see SI Section 3.3 for experimental)
has already been proven reliable [46] and was thus selected to
Table 3
XPS analyses of the supports used for adducts immobilization.
XPS atomic ratio
Pristine MWNT
MWNT-ox
MWNT-Cl
MWNT-X
GO
GO-Cl
rGO
O/C
Cl/C
F/C
0.0125
/
/
0.0480
0.0472
0.0030
/
/
0.0180
/
0.0227
0.0021
0.4863
/
/
0.3061
0.0194
/
/
0.0900
/
/
/
/
/
/
S/C
0.0014
0.0082

C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
395
decorate the tubes with activated esters (MWNTs-X). The XPS
3.3.3. Immobilization and characterization of adducts 9–12
incorporated on nanocarbons by a non-covalent methodology to
obtain P9 (MWNTs + 9), PG9 (rGO + 9), P10 (MWNTs + 10), PG10
(rGO + 10), P11 (MWNTs + 11) and P12 (MWNTs + 12)
The selection of pyrene moieties in a polar solvent to immobi-
lize these adducts onto nanocarbons was guided by some studies
results obtained before and after functionalization are also
reported in Table 3. The obtained F/C ratio was in agreement with
previously reported values [46], and corresponded to a functional-
ization degree of 1 function every 22 carbons of the tubes given the
average number of concentric tubes in a MWNT observed by TEM,
that is approximately 10 (see SI Section 3.1).
devoted to
p–p stacking [47,48]. Metallic ratios (Pt/Ru and
Au/Ru) determined by XPS after non-covalent immobilization of
cluster/ligands adducts were close to the calculated values (see
Table 4). The situation was also comparable regarding the Ru/P
ratios and in both cases slightly lower than expected.
Nevertheless, much lower Ru/P values were obtained when gold
was present within the clusters (P11–P12), which cannot be fully
explained. For all samples, the high increase of O/C values (to com-
pare with pristine MWNTs and rGO in Table 3) clearly demon-
strated the immobilization of clusters, with their substantial
oxygen content within the CO ligands.
3.3.2. Immobilization and characterization of the adducts 7 and 8
incorporated on nanocarbons by a covalent methodology to obtain C7
(MWNT-Cl + 7), GC7 (GO-Cl + 7), X7 (MWNT-X + 7), C8 (MWNT-
Cl + 7), GC8 (GO-Cl + 8) and X8 (MWNT-X + 8)
The suggested mechanism for covalent immobilization involves
an amine attack (from adducts 7 and 8) on acyl chloride or acti-
vated ester functions born by nanocarbons, followed by departure
of chloride or fluorinated ester and formation of a covalent amide
bond. The values obtained by XPS after covalent anchoring are
given in Table 4. The presence of the expected metals, namely Ru
and Pt, was confirmed by XPS. Moreover, when both metals are
present, their Ru/Pt ratios were in agreement with the theoretical
values (i.e. 5). The obtained Ru/P ratios were also generally close
to the calculated values but nevertheless slightly lower. The F/C
and Cl/C XPS ratios both decrease after functionalization (compare
Tables 3 and 4), which indicates the expected covalent bond for-
mation. The amount of metal (metal atomic percentage) covalently
bonded to MWNTs was similar irrespective of the methodology of
incorporation selected, as demonstrated by a simple Ru/C compar-
ison between C7-X7 and C8-X8. The support used greatly influ-
enced the quantity of grafted metal since the atomic percentage
was two or three times higher when immobilization was
performed onto graphene. This undoubtedly came as a result of
an initially more functionalized surface (see Table 3).
3.4. Thermal treatments
Adducts incorporated by covalent (C7, X7, GC7, C8, X8 and GC8)
or non-covalent (P9, PG9, P10, PG10, P11 and P12) methodology
onto nanocarbons were submitted to a thermal treatment. This
treatment was required to remove both carbonyl ligands surround-
ing the clusters and the bifunctional ligand used to immobilize
them on the supports in order to obtain naked metallic particles
on the surface. Some related studies have demonstrated the benefit
of such anchoring/thermoactivation techniques allowing a homo-
geneous dispersion of the resulting particles [65].
The selected activation temperature also influences not only the
average particle sizes but also the number of remaining oxygen
surface groups and both factors can influence the selective
Table 4
XPS analyses of the adducts incorporated onto nanocarbons.
Sample Code
C7
X7
GC7
C8
X8
GC8
P9
PG9
P10
PG10
P11
P12
0.045
O/C
Cl/C
F/C
0.092
0.002
/
0.005
0.021
/
0.069
/
0.016
0.004
0.024
/
0.184
0.020
/
0.008
0.042
/
0.079
0.003
/
0.004
0.010
3.9a
2.5
0.055
/
0.015
0.003
0.011
4.1a
0.0184
0.017
/
0.010
0.030
3.7a
0.043
/
/
0.002
0.011
/
0.157
/
/
0.006
0.036
/
0.036
/
/
0.001
0.003
3.6a
3.3
0.154
/
/
0.008
0.035
4.1a
4.3
0.043
/
/
0.003
0.004
2.4b
1.3
/
/
P/C
0.004
0.003
1.6b 3.1a
0.8
Ru/C
Ru/Ma,b
Ru/P
4.3
6.3
5.2
3.8
3.0
4.6
5.6
a
Pt.
Au.
b
Table 5
The fifteen synthesized catalysts: sample code and preparation methodology.

396
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
hydrogenation of
a,b-unsaturated aldehydes: The higher the tem-
distribution. Indeed, p–p stacking and covalent anchoring method-
perature, the bigger the particles. But also, the higher the temper-
ature, the least acidic and oxygen surface groups. If the increase in
particle size was generally admitted to improve the selectivity
toward the unsaturated alcohol, the oxygen surface groups effect
is still a matter of debate [44,45].
To gain an overview of the temperature effect, three tempera-
tures were selected: 600 °C, 900 °C and 1300 °C. The thermal acti-
vations were realized under reducing atmosphere with a mixture
of nitrogen and hydrogen in a ratio 95/5. The two highest temper-
atures were applied to MWNTs while only the lowest value (i.e.
600 °C) was used for graphene: A higher temperature leads to a
large weight loss. This can be explained by the very high amount
of oxygen initially contained in graphene (see Table 3) leading to
losses as CO and CO2 at high temperature, but also of its exfoliated
structure.
Fifteen catalysts were obtained by combining four clusters,
three temperatures, three different types of immobilization proce-
dures and two different nanocarbon supports. They are presented
and codified for more clarity in Table 5. This includes the following
information: metal(s) nature, support used, functionalization
methodology and thermal activation temperature.
To get more information about the influence of the parameters
tested on the catalysts obtained, TEM images were recorded and
some of them are presented in Fig. 6. As expected, average particle
sizes are indeed growing up when heating up. The methodology of
incorporation also had an influence on the particle size
ology through acyl chlorides gave, for a given temperature and
support, bigger particles than the sidewall radical functionalization
methodology (see Fig. 6). Histograms (at the right of TEM images)
were built by measuring in each case more than a hundred metal
particles on several images of the selected sample using the
AnalySIS Auto 5.0 software (Olympus Soft Imaging Solutions
GmbH, Germany). A greater number of particles are indeed cen-
tered on 3–4 nm for p–p stacking and defects/end sites functional-
ization than for xanthate chemistry. These phenomena can be
explained by the nature of the link established with nanocarbons.
The p–p stacking is a weaker ‘‘link’’ than the covalent ones and log-
ically led to bigger particles. The location of the covalent anchoring
points, that is a uniform distribution for activated esters versus
defect/end sites for acyl chlorides could play on the sintering:
The latter are closer together, which can facilitate the coalescence
and then increase the particle size. In addition, oxygenated func-
tions are more easily cleaved. These considerations can explain
the different particle sizes observed.
XPS analyses (Table 6) were also conducted after thermal treat-
ment and revealed in most cases a decrease or disappearance of
phosphorous peaks. A slight change in the Ru/Pt and a significant
change in the Ru/Au ratios were observed after heating. These dif-
ferences indicate some modifications in the clusters structures.
These observations, at least in the case of RuPt nanoparticles, are
in agreement with the selected technique (XPS) giving a surface,
and not a bulk analysis. The coalesced nanoparticles might have
Fig. 6. TEM images of catalysts (a) C8-900, (b) X8-900, (c) P10-900, (d) P10-1300 and (e) GC8-600.

C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
397
Table 6
XPS results obtained after thermal activation at different temperatures (600, 900 and 1300 °C).
Sample
Code
C7-
900
C7-
1300
X7-
900
GC7-
600
C8-
900
X8-
900
GC8-
600
P9-
900
P9-
1300
PG9-
600
P10-
900
P10-
1300
PG10-
600
P11-
900
P12-
900
O/C
P/C
0.011
0.002
0.021
/
0.501
0.002
0.024
/
0.025
0.089
0.002
0.010
/
0.048
0.056
0.011
6.9a
0.019
/
0.168
0.016
0.011
3.9a
0.011
0.002
0.036
/
0.028
0.002
0.003
/
0.100
0.014
0.035
/
0.015
/
0.026
/
0.083
0.006
0.018
0.002
0.017
0.002
/
0.042
/
Ru/C
0.030
0.004
0.003
Ru/Ma,b
4.1a
4.7b
2.7a
5.6a
4.1
7.5b
0.9
6.5b
3.3a
0.6
Ru/P
4.4
5
/
2.9
1.5
/
2.0
2.7
5.3
3.6
/
/
a
Pt.
Au.
b
surface enrichment in one or the other metal. Moreover, the big-
gest particles will only partially contribute to the XPS signal.
Similar results were already obtained in a related study for Ru, Pt
and Au based clusters [66].
calculated from the ICP values for Ru content obtained by
MEDAC (Tables in Sections 4.1 and 4.2 in Supporting
Information). Some illustrative curves of conversion as function
of time are also given in Supplementary Information.
ICP analyses (Ru) were also realized after thermal activation by
the firm MEDAC LTD (UK) to determine metal loading of all the cat-
alysts (see Tables in Sections 4.1 and 4.2 in Supporting
Information). The obtained results sometimes differ from those
obtained by XPS, as expected given the nano-supports and large
particle sizes in some cases.
3.5.2. Influence of various parameters on the selectivity
All the catalysts tested are plotted in Fig. 7 according to their
increasing selectivity toward COL. Nature of the metal(s), incorpo-
ration methodology, nature of the support and temperatures of
thermal activation have been analyzed separately in order to
understand in more details the parameters influencing the reaction
selectivity. Finally, some catalysts were also tested in a non-polar
solvent to determine whether the solvent had any influence on
the selectivity.
3.5. Selective hydrogenation of cinnamaldehyde
Activity and selectivity of ruthenium, ruthenium–platinum,
ruthenium–gold and ruthenium–platinum–gold catalysts will be
studied in this section. These greatly depend on a lot of parameters
such as the metal incorporation methodology, the support used,
the thermal activation temperature and the nanoparticles metallic
nature. The effects of these factors will be discussed below. The
experimental procedure for hydrogenation of cinnamaldehyde is
given in SI, Section 4.
3.5.2.1. Nature of the metal. The selectivity was influenced to a con-
siderable extent by the nature of the metal(s) used within the
selected clusters. This effect is depicted in Fig. 8 where the only
variable is the metal nature. Indeed all the presented catalysts
were prepared by an identical pathway, namely non-covalently
immobilization on pristine MWNTs and subsequent activation at
900 °C. The Ru–Pt partnership (P10-900) strongly improved the
selectivity toward COL and gave, as expected, the best values
[1,27,67] while the presence of gold within the bi-metallic Ru–Au
nanoparticles (P11-900) shifted the selectivity toward HCAL (as
previously observed [52]). Finally, the trimetallic catalyst
P12-900, gave a better selectivity toward COL than P9-900 and
P11-900 most probably due to the presence of platinum within
the nanoparticles.
3.5.1. Activity of the catalysts
All the catalysts were tested in the selective hydrogenation of
cinnamaldehyde and their activities were compared after 2 h at
120 °C under 50 bars of H2 in 2-propanol/water (5/1 volume) as
solvent (see SI Section 4, Table S1 and Experimental above). Note
that the presence of water was needed to avoid alcohol
by-products formation (acetal formation) [9]. The conversions ran-
ged from 26% to 83% and are listed in Tables in Sections 4.1 and 4.2
in Supporting Information. For greater clarity, catalysts activities
are split according to the support used: Table in Section 4.1 in SI
includes the catalysts immobilized on graphene while
nanotubes-based catalysts are presented in Table in Section 4.2
in SI. It is striking that the best conversions are always obtained
3.5.2.2. Effects of temperature, methodology of incorporation and
nature of the support. It has been reported that the selectivity
toward COL increases concomitantly with a growing particle size
[2]. This particle size effect can be explained by a directing effect
of the phenyl group [44]. The metal surface prevents access to
the phenyl groups and thus repels the close C@C bond from the
surface in the case of big nanoparticles, the C@O bonds being then
with catalysts prepared by
p–p stacking. TON and TOF (both calcu-
lated on the basis of the conversion to all products) values were
Fig. 7. Conversion and selectivity toward COL of the fifteen catalysts (t = 2 h).

398
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
These results are related to the different cluster anchoring routes,
which affect the particle size and consequently the selectivity. As
depicted in Fig. 6, the particle size distribution was centered
between 3 and 4 nm when a non-covalent anchoring was used
while an average smaller particle size and broad distribution after
the heating process were encountered in covalent cases. In addi-
tion, covalent routes might lead to remaining functions on the sur-
faces which also might be deleterious for the selectivity.
In order to selectively obtain the COL product, thermoactiva-
tions at 900 and 1300 °C were investigated when MWNTs were
used. These temperatures were inadequate with graphene as sup-
port due to large weight loss. However, a comparison between
GC7-600 and C7-900 or P10-900 and PG10-600 shows that the
lower thermal activation is compensated by the use of graphene
instead of MWNTs and revealed the positive effect of this support
on the targeted product. Nevertheless, the possibility of heating
MWNTs at temperatures higher than 600 °C is a decisive factor in
view of the correlation between selectivity and large particles:
This provides the best catalyst of this work (P10-1300).
The methodologies described here, nanocarbons functionaliza-
tion following three different routes to anchor mono- or
multi-metallic clusters are at an early stage. This first study in cin-
namaldehyde hydrogenation was capable of rapidly increasing the
selectivity toward the targeted product through a fine-tuning of
the used conditions. On these bases, the best catalyst found gave
good activity and selectivity, namely 68% conversion and 62%
selectivity respectively. Compared to literature [1,27,45,67], there
is still room for improvement compared to the best results
reported so far, namely 95% [1,67] and 93% [27] selectivity toward
COL using Ru–Pt catalysts. Nevertheless, in these two cases, the
metal weight ratios are 1/1 (Pt/Ru), that is higher than in our case.
Fig. 8. Metal(s) effect in the selective hydrogenation of cinnamaldehyde.
able to approach the surface in priority in order to proceed to
hydrogenation. These considerations do not apply in the case of
small particles since no repelling between phenyl rings and parti-
cles can occur as depicted in Fig. 9.
The depicted effect was clearly observed here with all the cata-
lysts. As shown by TEM, particle sizes are thermally dependent and
the selectivity toward COL was strongly increased for catalysts
activated at higher temperature, regardless of the selected incorpo-
ration methodology. A comparison between the catalysts C7-900
and C7-1300, P9-900 and P9-1300 or P10-900 and P10-1300
clearly indicates an increase in the selectivity toward the desired
COL product due to higher temperature of activation. The out-
standing importance of the particle size effect can be proven when
comparing bimetallic P10-900 and monometallic P9-1300 cata-
lysts: The latter shows indeed exactly the same selectivity than
the bimetallic one and was therefore able to compensate the lack
of platinum. The particle size difference between catalysts ther-
moactivated at 900 and 1300 °C is important. Indeed, a comparison
between P10-900 and P10-1300 shows a shift from particles
broadly centered around 3–4 nm for the former to particle sizes
between 5 and 9 nm for the latter (see Fig. 6).
As presented in Fig. 3, the methodology of immobilization had a
real impact on catalytic properties. The non-covalent methodology
was the best route to selectively hydrogenate the C@O bond since
six of the eight best catalysts were prepared by this route. Defect
sites functionalization and sidewall functionalization came after
and in that order. A comparison between RuPt bimetallic catalysts
onto nanotubes thermally activated at 900 °C X8-900, C8-900 and
P10-900 clearly demonstrates this effect. It is worth noting that
the routes involving radicals lead to catalysts more selective
toward HCAL than COL, even if platinum was present (see Fig. 7).
3.5.2.3. Solvent effect. It is generally considered that the solvent
greatly affects the activity and moderately the selectivity of the
catalyst in the selective hydrogenation of cinnamaldehyde [2].
The mix 2-propanol/water previously used was replaced by one
of the best non-polar solvent to perform the selective hydrogena-
tion of cinnamaldehyde: toluene [2]. The experimental conditions
used are the following: 50 bars H2, 110 °C and 1500 rpm. The
results obtained in this case are depicted in Fig. 10. Compared with
the previous results achieved when 2-propanol/water was chosen
as a solvent, the activity was plummeted to barely reach 32% con-
version after 2 h in the best case. Surprisingly the selectivity was
completely modified to selectively obtain the HCAL product in all
cases with selectivities between 71% and 85%. Even if a slight
decrease in the selectivity toward COL was awaited, this reversal
was not expected and shows the importance of solvent selection
in the selective hydrogenation of cinnamaldehyde. Nevertheless,
and because both HCAL and COL products are of synthetic
Fig. 9. Adsorption of cinnamaldehyde on a small metal particle (left) and a large
particle (right) (adapted from Ref. [44]).
Fig. 10. Conversion and selectivity toward HCAL, HCOL and COL of some catalysts
tested in toluene as solvent (t = 2 h).

C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
399
pathway with nanocarbons. A pre-functionalization step was
required in the covalent case: The well-known oxidative treatment
and its subsequent acylation step to provide acyl chloride at the
support surface (MWNTs-Cl and G-Cl) and the xanthate chemistry
applied to MWNTs to obtain activated esters at their surface
(MWNTs-X) were applied. Covalent anchoring and non-covalent
anchoring were carried out and the solids were subsequently acti-
vated at different temperatures (600 °C for graphene, 900 and
1300 °C for MWNTs). These were analyzed by XPS before and after
thermal treatments and in general were characterized by metallic
ratios close to the calculated values. The fifteen catalysts obtained
were used in the selective hydrogenation of cinnamaldehyde and
the influence of various parameters was studied. The presence of
Pt within the clusters greatly improves the selectivity toward
COL, as well as bigger particle size that are dependent on the ther-
mal activation: A high activation temperature (1300 °C) was
required in this work to reach the best selectivities. The methodol-
ogy of incorporation also had an impact since a weak bond
Fig. 11. Recycling of catalyst P10-900.
80
70
60
50
40
30
20
10
0
between metal precursor and carbon surface, such as p–p stacking,
will lead to a more significant sintering hence bigger particles, thus
influencing positively the selectivity of the reaction. Compared to
platinum, gold within clusters shows an opposite tendency and
enhances the selectivity toward HCAL. In addition, a modification
of solvent in the catalytic reaction changes significantly the selec-
tivity: A same catalyst, for instance P10-900, can selectively give
COL (51%) in isopropanol/water and HCAL (71%) in toluene. A ther-
mal treatment of graphene beyond 600 °C leads to a high weight
loss which limits the activation temperature. Despite this limita-
tion, good results were obtained with this support demonstrating
its promising future in this selective hydrogenation reaction.
When the best conditions are in place, namely [Ru5PtC(CO)16] as
precursor non-covalently immobilized onto MWNTs and thermally
treated at 1300 °C, the selectivity toward COL product may reach
75% with good activity (between 70% and 80% conversion for the
best catalysts). These catalysts are moreover recyclable without
any loss of selectivity and/or activity. Such peculiar synthetic strat-
egy can thus have an attractive future to finely tune the ratio
between the different elements by varying the initial ratio between
metals within clusters as single source precursors. Moreover, these
strategies led to well dispersed particles onto the support and with
a narrow size distribution in addition to composition control.
Run 1
25
Run 2
19
Run 3
19
Run 4
22
Run 5
17
Run 6
16
HCAL
HCOL
COL
13
62
68
12
69
72
12
69
74
12
66
71
8
12
72
76
75
55
Conversion
Fig. 12. Recycling of catalyst P10-1300.
importance, this solvent dependence is of real interest. Two of the
best catalysts described in this study, namely P10-900 and
P9-1300 can therefore selectively transform CAL into COL or
HCAL by a simple solvent change.
3.5.3. Recycling
Catalyst reusability is of paramount importance. It was there-
fore attempted with two of the best catalysts found, that is
P10-900 and P10-1300. The results are given in Figs. 11 and 12.
In the first case, we obtained a constant activity, around 80% con-
version, coupled with a selectivity that remains constant. After the
two first cycles, the selectivity even marginally increased to reach
57% selectivity toward COL. This slight increase was probably due
to a small modification of the catalyst after the two first catalytic
cycles to give it its final characteristics. The same tendency was
observed with the catalyst P10-1300 to reach, after the first run,
a selectivity approaching 70% toward COL (with a peak at 75%
selectivity) and a conversion of about 70% along the runs.
Acknowledgments
We
acknowledge
the
FRS-FNRS,
FRIA,
Fédération
Wallonie-Bruxelles, Loterie Nationale, Belgian State (IAP Project
N° P6/17) and Université catholique de Louvain for funding. We
are also grateful to Jean-François Statsijns, Cécile Leduff and
Michel Genet for technical assistance and fruitful discussions. We
also thank Deborah Vidick for some cluster syntheses, Geoffrey
Ledent for developments of pyrene strategy and the Nanocyl firm
for generous donation of carbon nanotubes.
Appendix A. Supplementary material
4. Conclusion
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2015.06.003.
We have reported unusual methodologies to form mono- or
multi-metallic nanoparticles onto nanocarbons (nanotubes and
graphene) and these were used as catalysts in the selective hydro-
genation of cinnamaldehyde. Four clusters, that is [Ru6C(CO)17] (3),
References
[Ru5PtC(CO)16
]
(4),
[Ru6C(CO)16(Au(PPh3))2]
(5)
and
[1] J. Teddy, A. Falqui, A. Corrias, D. Carta, P. Lecante, I. Gerber, P. Serp, J. Catal. 278
(2011) 59–70.
[2] P. Gallezot, D. Richard, Catal. Rev. Sci. Eng. 40 (1&2) (1998) 81–126.
[3] P. Mäki-Arvela, J. Hájek, T. Salmi, D.Y. Murzin, Appl. Catal. A.: Gen. 292 (2005)
1–49.
[Ru5PtC(CO)15(Au(PPh3))2] (6) were selected as nanoparticles pre-
cursors with the aim to retain their metallic stoichiometries and
intimate contacts in the final catalysts. These cluster complexes
were reacted with bifunctional ligands to quantitatively form
adducts (7–12) able to interact by a covalent or a non-covalent
[4] Y. Yuan, S. Yao, M. Wang, S. Lou, N. Yan, Curr. Org. Chem. 17 (2013) 400–413.
[5] U.K. Singh, M. Vannice, A. Appl, Catal. A: Gen. 213 (2001) 1–24.

400
C. Vriamont et al. / Journal of Catalysis 329 (2015) 389–400
[6] H. Ma, L. Wang, L. Chen, C. Dong, W. Yu, T. Huang, Y. Qian, Catal. Commun. 8
(2007) 452–456.
[38] R. Raja, T. Khimyak, J.M. Thomas, S. Hermans, B.F.G. Johnson, Angew. Chem. Int.
Ed. 40 (2001) 4638–4642.
[7] N. Mahata, F. Gonçalves, M.F.R. Pereira, J.L. Figueiredo, Appl. Catal. A: Gen. 339
(2008) 159–168.
[39] R. Raja, G. Sankar, S. Hermans, D.S. Shepahard, S. Bromley, J.M. Thomas, B.F.G.
Johnson, Chem. Commun. (1999) 1571–1572.
[8] G.R. Bertolini, C.I. Cabello, M. Munnoz, M. Casella, D. Gazzoli, I. Pettiti, G.J.
Ferraris, G.J. Mol, Catal. A.: Chem. 366 (2013) 109–115.
[40] M.S. Nasher, D.M. Somerville, P.D. Lane, D.L. Adler, J.R. Shapley, R.G. Nuzzo, J.
Am. Chem. Soc. 118 (1996) 12964–12974.
[9] B.F. Machado, S. Morales-Torres, A.F. Pérez-Cadenas, F.J. Maldonado-Hodar, F.
Carrasco-Marin, A.M.T. Silva, J.L. Figueiredo, J.L. Faria, Appl. Catal. A: Gen. 425–
426 (2012) 161–169.
[41] J.M. Thomas, R. Raja, D.W. Lewis, Angew. Chem. Int. Ed. 44 (2005) 6456–6482.
[42] R. Ferrando, J. Jellinek, R.L. Jhonston, Chem. Rev. 108 (2008) 845–910.
[43] J. Shi, R. Nie, P. Chen, Z. Hou, Catal. Commun. 41 (2013) 101–105.
[44] A.J. Plomp, H. Vuori, A.O.I. Krause, K.P. de Jong, J.H. Bitter, Appl. Catel. A: Gen.
351 (2008) 9–15.
[45] M.L. Toebes, Y. Zhang, J. Hájek, T.A. Nijhuis, J.H. Bitter, A.J. van Dillen, D.Y.
Murzin, D.C. Koningsberger, K.P. de Jong, J. Catal. 226 (2004) 215–225.
[46] B. Vanhorenbeke, C. Vriamont, F. Pennetreau, M. Devillers, O. Riant, S.
Hermans, Chem. Eur. J. 19 (2013) 852–856.
[47] C. Janiak, J. Chem. Soc., Dalton Trans. (2000) 3885–3896.
[48] E. Bus, R. Prins, J.A. van Bokhoven, Catal. Commun. 8 (2007) 1397–1402.
[49] C. Milone, C. Crisafulli, R. Ingoglia, L. Schipilliti, S. Galvagno, Catal. Today 122
(2007) 341–351.
[10] A.K. Prashar, S. Mayadevi, R.N. Devi, Catal. Commun. 28 (2012) 42–46.
[11] E. Castillejos, E. Gallegos-Suares, B. Bachiller-Baeza, R. Basca, P. Serp, A.
Guerrero-Ruiz, I. Rodriguez-Ramos, Catal. Commun. 22 (2012) 79–82.
[12] A.J. Plomp, D.M.P. van Asten, A.M.J. van der Eerden, P. Mäki-Arvela, D.Y.
Murzin, K.P. de Jong, J.H. Bitter, J. Catal. 263 (2009) 146–154.
[13] J. Hajek, N. Kumar, T. Salmi, D.Y. Murzin, Catal. Today 100 (2005) 349–353.
[14] B.M. Reddy, G.M. Kumar, I. Ganesh, A. Khan, J. Mol. Catal. A: Chem. 247 (2006)
80–87.
[15] M. Lashdaf, V.V. Nieminen, M. Tiitta, T. Venäläinen, H. Osterholm, O. Krause,
Microporous Mesoporous Mater. 75 (2004) 149–158.
[16] M. Lashdaf, J. Lahtinen, M. Lindblad, T. Venäläinen, A.O.I. Krause, Appl. Catal. A:
Gen. 276 (2004) 129–137.
[50] H. Rojas, G. Diaz, J.J. Martinez, C. Castaneda, A. Gomez-Cortés, J. Arenas-
Alatorre, J. Mol. Catel. A: Chem. 363–364 (2012) 122–128.
[51] J. Lenz, B.C. Campo, M. Alvarez, M.A. Volpe, J. Catal. 267 (2009) 50–56.
[52] X. Zhang, Y.C. Guo, Z.C. Zhang, J.S. Gao, C.M. Xu, J. Catal. 292 (2012) 213–226.
[53] C. Vriamont, M. Devillers, O. Riant, S. Hermans, Chem. Eur. J. 19 (2013) 12009–
12017.
[17] J. Hajek, N. Kumar, P. Mäki-Arvela, T. Salmi, D.Y. Murzin, I. Paseka, T. Heikkilä,
E. Laine, P. Laukkanen, J. Väyrynen, Appl. Catal.A: Gen 251 (2003) 385–396.
[18] J.P. Stassi, P.D. Zgolicz, S.R. de Miguel, O.A. Scelza, J. Catal. 306 (2013) 11–29.
[19] G.R. Bertolini, C.I. Cabello, M. Munnoz, M. Casella, D. Gazzoli, I. Pettiti, G.
Ferraris, J. Mol. Catal. A: Chem. 366 (2013) 109–115.
[54] S. Hermans, T. Khimyak, N. Feeder, S.J. Teat, B.F.G. Jonhson, Dalton Trans.
(2003) 672–684.
[20] P. Claus, Top. Catal. 5 (1998) 51–62.
[21] N. Job, R. Pirard, J. Marien, J.-P. Pirard, Carbon 42 (2004) 619–628.
[22] Z. Guo, Y. Chen, L. Li, X. Wang, G.L. Haller, Y. Yang, J. Catal. 276 (2010) 314–326.
[23] Y. Li, P.-F. Zhu, R.-X. Zhou, Appl. Surf. Sci. 254 (2008) 2609–2614.
[24] M. Chatterjee, Y. Ikushima, F. Zhao, New J. Chem. 27 (2003) 510–513.
[25] C. Li, Z. Shao, M. Pang, C.T. Williams, X. Zhang, C. Liang, Ind. Eng. Chem. Res. 51
(2012) 4934–4941.
[55] G. Liu, B. Wu, J. Zhang, X. Wang, M. Shao, J. Wang, Inorg. Chem. 48 (2009)
2383–2390.
[56] F. Li, B. Zhang, X. Li, Y. Jiang, L. Chen, Y. Li, L. Sun, Angew. Chem. 123 (2011)
12484–12487.
[57] L. Xing, J.-H. Xie, Y.-S. Chen, L.-X. Wang, Q.-L. Zhou, Adv. Synth. Catal. 350
(2008) 1013–1016.
[26] E. Castillejos, M. Jahjah, I. Favier, A. Orejón, C. Pradel, E. Teuma, A.M. Madseu-
Bulto, P. Serp, M. Gomez, ChemCatChem 4 (2012) 118–122.
[27] H. Vu, F. Gonçalves, R. Philippe, E. Lamouroux, M. Corrias, Y. Khin, D. Plee, P.
Kalck, P. Serp, J. Catal. 240 (2006) 18–22.
[58] P.D. Tran, A. Le Goff, J. Heidkamp, B. Jousselme, N. Guillet, S. Palacin, H. Dau, M.
Fontecave, V. Artero, Angew. Chem. Int. Ed. 50 (2011) 1371–1374.
[59] S. Hermans, T. Khimyak, B.F.G. Johnson, J. Chem. Soc., Dalton Trans. (2001)
3295–3302.
[28] M. Sankar, N. Dimitros, P.J. Miedziak, P.P. Wells, C.J. Kiely, G. Hutchings, J.
Chem. Soc. Rev. 41 (2012) 8099–8139.
[60] T. Khimyak, B.F.G. Johnson, S. Hermans, A.D. Bond, Dalton Trans. (2003) 2651–
2657.
[29] K. Liberkova, R. Touroude, J. Mol. Catel. A: Chem. 180 (2002) 221–230.
[30] S. Recchia, C. Dossi, N. Poli, A. Fusi, L. Sordelli, R. Psaro, J. Catal. 184 (1999) 1–4.
[31] P. Reyes, G. Pecchi, J.L.G. Fierro, Langmuir 17 (2000) 522–527.
[32] P. Reyes, M.C. Aguirre, J.L.G. Fierro, G. Santoni, O. Ferretti, J. Mol. Catal. A:
Chem. 184 (2002) 431–441.
[33] P. Reyes, C. Rodrigues, G. Pecchi, J.L.G. Fierro, Catal. Lett. 69 (2000) 27–32.
[34] P. Mäki-Arvela, L.-P. Tiainen, M. Lindblad, K. Demirkan, N. Kumar, R. Sjöholm,
T. Ollonqvist, J. Väyrynen, T. Salmi, D.Y. Murzin, Appl. Catal. A 241 (2003) 271–
288.
[61] M.I. Bruce, E. Horn, P.A. Humphrey, E.R.T. Tiekink, J. Organomet. Chem. 518
(1996) 121–138.
[62] G.W. Smith, E.A. Carter, J. Phys. Chem. 95 (1991) 2327–2339.
[63] J.A. Martinho Simoes, J.L. Beauchamp, Chem. Rev. 90 (1990) 629–688.
[64] S. Hermans, V. Bruyr, M. Devillers, J. Mater. Chem. 22 (2012) 14479–14486.
[65] D. Vidick, M. Herlitschke, C. Poleunis, A. Delcorte, R.P. Hermann, M. Devillers, S.
Hermans, J. Mater. Chem. A 1 (2013) 2050–2063.
[66] C. Willocq, D. Vidick, B. Tinant, A. Delcorte, P. Bertrand, M. Devillers, S.
Hermans, Eur. J. Inorg. Chem. (2011) 4721–4729.
[35] A. Jung, A. Jess, T. Schubert, W. Schütz, Appl. Catal. A: Gen. 362 (2009) 95–105.
[36] F.A. Cotton, Inorg. Chem. 3 (1964) 1217–1220.
[67] E. Castillejos, P.J. Debouttiere, L. Roiban, A. Solhy, V. Martinez, Y. Kihn, O. Ersen,
K. Philippot, B. Chaudret, P. Serp, Angew. Chem. 48 (2009) 2529–2533.
[37] J.M. Thomas, B.F.G. Johnson, R. Raja, G. Sankar, P.A. Midgley, Acc. Chem. Res. 36
(2003) 20–30.