Hydrogenation of nitrobenzene over Pd/C catalysts prepared from molecular carbonyl-phosphine palladium clusters

Article abstract of DOI:10.1016/j.molcata.2012.09.001

Molecular phosphine-carbonyl palladium clusters were anchored onto a phosphine-functionalized carbon support (C_PPh2) by ligand exchange. This support was characterized by solid-state NMR prior to reaction with the clusters. The same clusters we

Full text of DOI:10.1016/j.molcata.2012.09.001

Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
Contents lists available at SciVerse ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Hydrogenation of nitrobenzene over Pd/C catalysts prepared from molecular
carbonyl–phosphine palladium clusters
Christopher Willocqa, Vincent Duboisb, Yaroslav Z. Khimyakc,1, Michel Devillersa, Sophie Hermansa,∗
a Institute of Condensed Matter and Nanosciences (IMCN), Université catholique de Louvain, Place Louis Pasteur 1/3, B-1348 Louvain-la-Neuve, Belgium
b Laboratoire de Chimie Physique et Catalyse, Institut Meurice, Avenue Emile Gryzon 1, 1070 Bruxelles, Belgium
c Department of Chemistry, University of Liverpool, Crown Street, Liverpool L69 7ZD, United Kingdom
a r t i c l e i n f o
a b s t r a c t
Article history:
Molecular phosphine–carbonyl palladium clusters were anchored onto a phosphine-functionalized car-
bon support (CPPh2) by ligand exchange. This support was characterized by solid-state NMR prior to
reaction with the clusters. The same clusters were also deposited on the non-functionalized support
(CSX+). Characterization by SEM and XPS showed that the surface was more uniform when using the
functionalized support. After thermal activation, the Pd/C materials obtained were characterized by
CO chemisorption, XPS, SEM, TEM, powder XRD and analyzed for metal loading by ICP-OES. Again, the
solids prepared with the functionalized support presented a more uniform surface. TEM indicated that
small nanoparticles (1–10 nm) were present on the surface with narrow size distributions. The activated
Pd/C materials proved to be efficient catalysts for the hydrogenation of nitrobenzene into aniline. They
were competitive with commercial and literature catalysts. Moreover, the catalysts prepared using the
Received 2 December 2011
Received in revised form 4 July 2012
Accepted 1 September 2012
Available online 7 September 2012
Keywords:
Clusters
Carbon
Phosphines
Functionalization
Nitrobenzene
functionalized support were more active and more stable than when using CSX+
.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
In heterogeneous catalysis, the use of transition metal clusters as
precursors is also interesting due to their high potential to control
(i) the size of the obtained supported particles via their nuclear-
ity and (ii) the chemical composition of these particles via their
stoichiometry[11]. Moreover, they generally lead to a better disper-
sion than mononuclear complexes [11]. For instance, clusters have
been used as precursors on oxide supports for hydrogenation reac-
tions [3,12–17]. They were mostly deposited on the intact support
without particular pretreatment, but sometimes the occurrence
of covalent interactions with the support was reported. Clusters
have also already been incorporated on carbon [18–22] but with-
out specific interactions. Nevertheless, to obtain active catalysts,
the supported clusters have to be activated to remove the ligand
shell and to obtain supported metallic particles. This decarbonyla-
tion step, as most of the clusters used as precursors are carbonyl
clusters, is known to lead to agglomeration and fragmentation of
the supported clusters [12]. To avoid losing the advantage of using
clusters as precursors and to obtain a high dispersion of these enti-
ties on a carbonaceous support even after activation, it would be
useful to incorporate on a carbon surface suitable functions that
may act as anchoring ligands for clusters.
The major drawback of heterogeneous catalysts in general is
their inhomogeneous surface comprising different types of cat-
alytic sites, with for example metal centers being in different
chemical, electronic and/or geometrical states [1]. This might lead
to poor selectivity, especially when compared to homogeneous sys-
tems. Numerous studies have been carried out in order to gain more
control on the construction of heterogeneous catalysts active sites.
In this context, an organometallic chemistry approach can become
important with particular attention paid to the interaction between
the support and organometallic precursors [2]. By using functions
that are already present on the support, or by grafting first suit-
able ligands on its surface, organometallic complexes could react
selectively with these groups, leading to a uniform and controlled
distribution of active species on the support [2,3]. This strategy
has been applied to active carbon as support by building suitable
surface functions [4–10]. The presence of carboxylate groups in
particular on the carbon surface allowed the incorporation of func-
tions playing the role of ligands for organometallic complexes by
derivatization.
The hydrogenation of nitro-compounds is of interest for a
range of applications, as amines are used as basic raw materi-
als for polyurethanes, rubber chemicals, dyes or pharmaceutical
compounds for example [23]. In particular, the hydrogenation of
nitrobenzene (NB) into aniline (AN) (Scheme 1) has been well stud-
ied [24–27]. The catalysts usually efficient for this reaction are
∗
Corresponding author. Tel.: +32 10 47 28 10; fax: +32 10 47 23 30.
E-mail address: Sophie.Hermans@uclouvain.be (S. Hermans).
Present address: University of East Anglia, Norwich Research Park, Norwich NR4
1
7TJ, United Kingdom.
1381-1169/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.molcata.2012.09.001

C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
173
NH2
NO2
H2
NHOH
H2
NO
H2
H2O
H2O
hydroxylamine
intermediate
nitroso
intermediate
aniline
(AN)
nitrobenzene
(NB)
Scheme 1. Hydrogenation of nitrobenzene (NB) into aniline (AN).
based on noble metal (mainly Pd) supported on carbon. In this case,
the two intermediates are not quantifiable because they are too
reactive and do not accumulate. The structure–sensitivity of this
reaction is well-known [25], with reaction rates depending both on
dispersion and structure of metallic particles. Hence it is an ideal
reaction to unravel structure–activity relationships.
2.2. Catalysts preparation
Two types of catalysts were prepared: (i) clusters anchored on
CPPh2, and (ii) clusters deposited on CSX+. The amount of cluster
engaged in each anchoring experiment corresponded to a theoreti-
cal 3 wt.% metal loading on the support after removal of the ligands.
The experimental conditions for incorporation of the different pre-
cursors on the supports are summarized in Table 1. In a typical
experiment, the appropriate quantity of cluster was introduced
with 582 mg of support and 40 mL of solvent or mixture of solvent
(20 mL/20 mL) in a round bottom Schlenk flask. Then, the suspen-
sion was stirred at room temperature for 5 days in the dark under
nitrogen (or carbon monoxide for clusters 4–6). The suspension was
filtrated and the isolated solid was washed with the used solvents
and dried at room temperature under vacuum. In order to deter-
mine the amount of incorporated cluster on the support, the filtrate
was kept for further analysis by palladium atomic absorption using
a PERKIN ELMER 3110 spectrometer fitted with a flame atomizer.
To obtain the desired activated catalysts, the supported clus-
ters were submitted to a thermal treatment in a tubular oven STF
16/450 from CARBOLITE. The samples were placed into porcelain
combustion boats and heated at 300 ◦C for 1 h for the clusters and
at 400 ◦C for 1 h for [Pd(dba)2] (heating ramp: 100 ◦C/h) under a
N2 stream. These activation conditions were obtained by thermal
studies of the clusters in their pure form, by TGA analyses on a
TGA SDTA 851e instrument from METTLER TOLEDO. These analy-
ses were carried out with a heating ramp of 10 ◦C/min and under
N2 flow (100 mL/min) and the samples (∼3 mg) were placed into
alumina containers (70 ␮L).
In this work, we report the synthesis of Pd/C catalysts from
heteroleptic carbonyl–phosphine palladium clusters. In order
to benefit from the advantage of using clusters as precursors,
we used a functionalized carbon support [28], which comprises
chelating phosphine groups on its surface (Scheme 2). These
were linked to carboxyl groups introduced by HNO3 oxidation
via ethylenediamine which was derivatized by double phos-
phinomethylation. These groups allow anchoring of clusters
on the support through covalent bonds, and this is assumed
to prevent, or at least to limit, agglomeration of the incorpo-
rated clusters and to keep them well-dispersed after activation
[29]. Indeed, carbonyl clusters are known to exchange easily
one or more ligands for chelating phosphines [30]. Here, we
describe first the anchoring of the complex [Pd(dba)2] (1) and
of the clusters [Pd10(CO)12(PBu3)6] (2), [Pd12(CO)17(PBu3)5]
(3), [Pd4(CO)5(PPh3)4] (4), [Pd10(CO)12(PPh3)6] (5) and
[Pd24(CO)32(PPh3)3] (6). All these precursors were anchored
on the functionalized support (CPPh2) by ligand exchange and
incorporated by non-specific interactions on the unmodified
carbon (CSX+) for comparison. We define anchoring as the covalent
linkage of a metallic complex (here a cluster) via an organic tether
covalently bound to the surface. Then, these supported clusters
were thermally activated to give Pd/C catalysts, which were tested
in the hydrogenation of nitrobenzene into aniline as a test-reaction
for structure effects.
Two references samples for solid-state NMR studies were pre-
pared under nitrogen as follows. First, HOCH2PPh2 was prepared by
mixing 1.08 mL HPPh2 (6.24 mmol) with 157 mg CH2O (5.23 mmol)
in 6 mL methanol, and heating 10 min at 70 ◦C. This mixture was
added to 600 mg CSX+ and stirred at room temperature for 30 min
before evaporating the solvent under reduced pressure. Second,
1.08 mL HPPh2 (6.24 mmol) were added to 600 mg of CSX+ in 10 mL
toluene, before stirring at room temperature for 30 min and evap-
orating the solvent under reduced pressure as well.
2. Materials and methods
2.1. Materials
All manipulations for the preparation of the catalysts were
carried out under oxygen-free nitrogen or carbon monoxide
atmosphere using Schlenk techniques and the samples were
stored under Ar. All solvents used for these preparations were
dried and distilled before use. The starting active carbon SXplus
(noted SX+, acid-washed steam activated carbon, ash content: 6%;
SBET = 922 m2/g; SBJH = 245 m2/g; pore volume [BJH] = 0.41 cm3/g)
was supplied by NORIT, sieved to keep only the fraction of gran-
ulometry 50–100 ␮m and noted CSX+. The complex [Pd(dba)2] (1)
was purchased from Aldrich and used as received. The 5 wt.%
commercial Pd on charcoal catalyst (type 58 paste, 49.6% H2O, John-
son Matthey, noted JM5%) was purchased from Johnson Matthey
and dried before use. The functionalized carbon support (noted
CPPh2) and the clusters [Pd4(CO)5(PPh3)4] (4), [Pd10(CO)12(PPh3)6]
(5) and [Pd24(CO)32(PPh3)3] (6) were prepared as previously
described by us [28,29,31]. The clusters [Pd10(CO)12(PBu3)6] (2) and
[Pd12(CO)17(PBu3)5] (3) were prepared as described by Mednikov
et al. [32,33].
2.3. Characterization
XPS (X-ray Photoelectron Spectroscopy) analyses were carried
out at room temperature on a SSI-X-probe (SSX-100/206) photo-
electron spectrometer from Surface Science Instruments (USA)
equipped with a monochromatized microfocus Al X-ray source.
Samples were stuck on small troughs with double-face adhesive
tape and then placed on an insulating home-made ceramic carousel
(Macor®,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 peak Au4f7/2
at 84 eV, and the binding energies were calculated with respect
to the C C, H) component of the C 1s peak fixed at 284.8 eV.
Data treatment was performed with the CasaXPS program (Casa

174
C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
PPh2
PPh2
CSX+
H2N
N
HNO3
O
OH
O
NH
O
NH
1. SOCl2
HPPh2/CH2O
2. H2N(CH2)2NH2
support
CCOOH
support
CNH2
support
CPPh2
Scheme 2. Support functionalization.
Software Ltd., UK). The peaks were decomposed into a sum of Gaus-
sian/Lorentzian (85/15) after subtraction of a Shirley type baseline.
SEM images were acquired on a FEG digital scanning micro-
scope (DSM 982 Gemini from LEO) equipped with a EDXS detector
(Phoenix CDU LEAP). The samples were fixed by double-sided con-
ducting adhesive tape onto aluminum specimen stubs from Agar
Scientific.
metal dispersion (%), VNTP is the adsorbed volume under normal
temperature and pressure conditions, M is the atomic weight of
the metal (here Pd), VM is the molar volume (22.414 L mol−1) of
the active gas under normal conditions (0 ◦C, 1 atm), m is the dry
sample mass, C is the stoichiometric coefficient for CO adsorption
on Pd (a 1:1 stoichiometry was used in the present study), and P is
the wt.% of metal in the sample.
TEM images were obtained with a LEO 922 OMEGA energy filter
transmission electron microscope. The samples were suspended
in hexane under ultrasonic treatment, then allowed to settle to
discard the biggest particles. A drop of the supernatant was then
deposited on a holey carbon film supported on a copper grid, which
was dried overnight under vacuum at room temperature, before
introduction in the microscope.
The X-ray diffraction (XRD) analyses were recorded on a
SIEMENS D5000 diffractometer equipped with a copper source
(ꢀK␣ = 154.18 pm). The samples were placed on quartz monocrys-
tals and the crystalline phases were identified by reference to the
JCPDS-ICDD database.
Solid-state NMR experiments were conducted at 9.4 T using a
Bruker Avance DSX400 spectrometer equipped with a 1H/19F/X
2.5 mm probe. Samples were spun in zirconia rotors of 2.5 mm outer
1
diameter. 31P{ H} MAS NMR spectra were acquired at 161 MHz and
400.16 MHz for 31P and 1H, respectively, at a MAS rate of 10 kHz. A
31P ꢁ/3 pulse of 1.9 ␮s was used for excitation whilst TPPM decou-
pling was used during acquisition with a recycle delay of 120 s.
Chemical shifts are quoted in ppm relative to 85 wt.% H3PO4 solu-
tion.
Pd content of the obtained catalysts was determined by ICP-OES
elemental analyses carried out by Medac Ltd. after dissolution of the
samples by acid digestion.
Metal dispersion measurements were realized by CO
chemisorption on a PulseChemisorb 2700 from MICROMERIT-
ICS. The sample (about 30 mg) was first reduced under hydrogen
flow (25 mL/min) at 150 ◦C for 1 h, then swept by a helium flow
(25 mL/min) at 200 ◦C for 2 h. The sample was finally brought back
to room temperature, at which point it was subjected to a sequence
of CO pulses. The metal dispersion was calculated from the cumu-
lative amount of adsorbed CO (i.e., the sum of nondetected CO in
each pulse), following D = (VNTP·M/VM·m·C·P) × 100, where D is the
2.4. Catalytic hydrogenation of nitrobenzene (NB) into aniline
(AN)
The catalytic tests were carried out in a steel Parr 4521 reactor
fitted with a temperature control device and a mechanical stirrer
fixed at 2500 rpm. First, 350 mL of methanol (Fluka, p.a., >99.8%),
a precisely weighted out amount (∼0.075 g) of catalyst and of 1-
butanol as chromatographic internal standard (∼2.9 g ≈ 3.66 mL;
Table 1
Incorporation of the different precursors on the supports CSX+ and CPPh2
.
Solventsa
Yield (%)b
Loading (wt.%)c
XPS Pd/Cexp
XPS Pd/Ccalc
d
Precursor
Amount of precursor
Support
97.2 mg
97.2 mg
44.2 mg
44.2 mg
38.9 mg
38.9 mg
76.2 mg
76.2 mg
50.4 mg
50.4 mg
28.8 mg
28.8 mg
CSX+
CPPh2
CSX+
CPPh2
CSX+
CPPh2
CSX+
CPPh2
CSX+
CPPh2
CSX+
DCM/Tol
DCM/Tol
Hex
Hex
Hex
Hex
THF
THF
99.9
69.9
82.3
50.7
85.6
84.3
32.3
45.2
71.4
58.9
94.7
87.9
3.0
2.1
2.5
1.5
2.6
2.5
1.0
1.4
2.2
1.8
2.8
2.6
0.020
0.022
0.007
0.017
0.010
0.022
0.011
0.010
0.029
0.014
0.017
0.027
0.003
0.002
0.003
0.002
0.003
0.003
0.001
0.002
0.002
0.002
0.003
0.003
[Pd(dba)2] (1)
[Pd10(CO)12(PBu3)6] (2)
[Pd12(CO)17(PBu3)5] (3)
[Pd4(CO)5(PPh3)4] (4)
[Pd10(CO)12(PPh3)6] (5)
[Pd24(CO)32(PPh3)3] (6)
THF/Tol
THF/Tol
DCM
CPPh2
DCM
a
THF = tetrahydrofurane; Tol = toluene; DCM = dichloromethane; Hex = hexane.
The yield has been calculated by subtracting the amount of Pd detected in the filtrates from the amount of Pd engaged and by dividing this result by the amount of
b
engaged Pd.
c
The metal loading has been calculated on the basis of the corresponding yield, considering 3 wt.% = 100% yield.
Calculated Pd/C values are bulk molar ratios and are dependent on the metal loading. The amount of Pd taken into consideration for the calculations corresponds to the
d
amount of Pd actually anchored. The amount of C taken into consideration corresponds to the amount of support used, considering that it is only constituted of carbon, plus
the carbonaceous contribution of the ligands (dba, PBu3 or PPh3).

C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
175
Sigma–Aldrich, p.a., 99.8%) were poured into the reactor and
degassed for 15 min under nitrogen. The mixture was then sub-
mitted to a pre-treatment at 80 ◦C for 1 h, under 15–20 bars of
H2. During this time, a precisely weighted out amount of reactant
(∼7.3 g ≈ 6.13 mL of nitrobenzene; Janssen Chimica, 99%) was dis-
solved in 19 mL of methanol and degassed (under N2) for 10 min
together with 25 mL of methanol. The temperature in the reac-
tor was then fixed at 60 ◦C and the pressure was lowered to
2–2.5 bars. The degassed nitrobenzene (NB) solution was added to
the reactor mixture followed by the degassed methanol. Immedi-
ately, 5 bars of H2 were added, which initiated the reaction. During
catalytic reaction, the reactor was kept under a pressure of about
7–7.5 bars H2. Sampling of the reaction mixture took place at regu-
lar time intervals, and the product distribution was obtained by gas
chromatography. The chromatograph was an Agilent Technologies
6890 N fitted with a WCOT fused silica Chrompack CP-WAX 52 CB
capillary column (length, 50 m; internal diameter, 0.53 mm; film
thickness, 2 ␮m). The vector gas was He (50 kPa), and the tem-
perature of the injector and detector was maintained at 250 ◦C.
The samples were automatically injected via a split–splitless injec-
tor. The temperature program for analysis was as follows: start at
160 ◦C, heating to 240 ◦C (10 ◦C/min), isotherm at 240 ◦C for 2 min.
The signal received by a FID detector was directly interpreted by
a computer using the HP Chemstation software. The surface area
of each peak was converted into concentration by applying the
response factor with respect to the internal standard.
1
Fig. 1. 31P{ H} MAS NMR spectra of (a) HOCH2PPh2 deposited on CSX+, (b) HPPh2
deposited on CSX+ and (c) CPPh2. Asterisks denote spinning sidebands.
and unreacted HPPh2 (at 28.9 ppm) or from reactions between the
alcohol and surface O-containing groups known to be present on
the SX+ carbon. Finally, a very broad band is observed for the CPPh2
support (Fig. 1c), which is centered on a value different from both
references described above. The width of the band arises from a
broad distribution of chemical shifts indicating multiple chemical
environments of the P-ligand in the phosphine-functionalized sup-
port. This was to be expected, as the double phosphinomethylation
reaction is always accompanied by the single phosphinomethyla-
tion as secondary reaction, together with the fact that by-products
cannot be washed out from the solid surface if covalently bound.
However, the presence of sites with 31P chemical shifts different
from the initial compounds indicates that a reaction did occur,
forming some chelating groups on the surface of CPPh2. Model stud-
ies in solution [29] have shown that the reaction path envisaged
here does lead to a major yield in the desired chelating P-ligand,
and that it is indeed able to react with organometallic cluster com-
pounds in a non negligible way.
The different Pd precursors were incorporated on both supports
CSX+ and CPPh2 (Table 1), in engaged amounts corresponding to a
theoretical 3 wt.% metal loading. The obtained solids were ana-
lyzed by XPS (Table 1) and SEM. First, one can observe that the
amount of Pd incorporated was higher in the case of the CSX+ sup-
port (excepted for 4), indicating its highly adsorbent properties.
Secondly, the amount of Pd incorporated (loading) was in gen-
eral higher than 2 wt.%. Cluster 4 presented the lowest yields of
incorporation by opposition to 1 and 6 that presented the best
yields. Thirdly, the XPS analyses indicated that the experimental
Pd/C ratios were higher than the calculated ones. This seemed to
indicate that the precursors are well-dispersed on the supports sur-
face, at least on the external surface. In addition, one could observe
that the Pd/C ratios were generally higher in the case of CPPh2 than
CSX+ (except for 4 and 5) even though the absolute amount of incor-
porated Pd was lower. This seemed to indicate that the dispersion
of the clusters is better on CPPh2 than on CSX+, and that Pd remains
more on the external surface (rather than within the pores) in the
case of CPPh2. The dispersion difference was confirmed by SEM anal-
yses. Indeed, even if the use of clusters as precursors permitted
to obtain small particles (<50 nm) in all cases (CSX+ and CPPh2),
the registered SEM images showed that the obtained surface on
CSX+ was less uniform than the one observed on CPPh2: agglom-
erates (50–200 nm) were identified at the surface of CSX+ (Fig. 2).
The specific activity measured can be expressed by either the
rate of nitrobenzene (NB) conversion or the rate of aniline (AN) for-
mation. In this study, the catalysts are compared on the basis of the
former. The conversion rate is calculated from the slope of the linear
part of the evolution with time. Standard tests have been performed
to ensure kinetic regime and to avoid diffusional limitations.
3. Results and discussion
3.1. Incorporation of clusters on the supports
The different clusters [Pd10(CO)12(PBu3)6] (2), [Pd12(CO)17
(PBu3)5] (3), [Pd4(CO)5(PPh3)4] (4), [Pd10(CO)12(PPh3)6] (5) and
[Pd24(CO)32(PPh3)3] (6), used as precursors for the synthesized
Pd/C catalysts, were prepared from literature procedures [31–33].
Their color and IR bands guarantee their unambiguous identifica-
tion. The spectra recorded in this study gave C O bond stretching
bands between 2200 and 1600 cm−1 that fit perfectly with litera-
ture data and confirmed the success of the syntheses.
The functionalized carbon support CPPh2, prepared as previ-
ously described by us [28,29], was designed to anchor molecular
clusters on its surface. Indeed, chelating phosphine groups were
incorporated on its surface to act as ligands for noble metal molec-
ular clusters. Within the framework of the present study, this
solid product has been characterized by solid-state NMR prior to
cluster incorporation. Fig. 1a and b displays the reference spectra
and Fig. 1c is the spectrum obtained for CPPh2. During the func-
tionalization procedure, HPPh2 is used as starting material, then
transformed into the HOCH2PPh2 alcohol to be finally grafted as
chelating phosphines on the support by derivatization of pend-
ing amines introduced beforehand. Reference spectra were taken
for HPPh2 (commercial) and the HOCH2PPh2 alcohol (prepared as
described in [29] and used as such without purification) brought in
contact with CSX+ for comparison (see Section 2). As expected, the
spectrum of HPPh2 adsorbed on CSX+ gives only one single narrow
band at 28.8 ppm (Fig. 1b), as there could be only one P-containing
species. The material obtained from the HOCH2PPh2 alcohol leads
to several bands in the NMR spectrum (Fig. 1a), which could arise
either from by-products of the reaction of formation of this alcohol

176
C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
Fig. 2. Representative SEM images of clusters incorporated on non-functionalized support CSX+ (before thermal treatment): (a) [Pd10(CO)12(PBu3)6], (b) [Pd4(CO)5(PPh3)4],
(c) [Pd10(CO)12(PPh3)6], and (d) [Pd24(CO)32(PPh3)3].
These results show that the use of a functionalized support for the
anchoring of molecular clusters helps to obtain more evenly dis-
persed nanoparticles [11]. The advantage of using clusters is to
allow a ligand exchange process for anchoring onto the function-
alized surface, and to offer nanoscopic building blocks for the final
catalyst.
ratios after activation were generally similar to those before acti-
vation but slightly smaller. This decrease of the Pd/C ratios seems
to indicate that some agglomeration occurred during the activation
step. As before activation, the Pd/C ratios were generally higher for
the catalysts supported on CPPh2, indicating a better dispersion on
the external surface of the functionalized support. The metal dis-
persion after activation was determined by CO chemisorption. The
lowest values (3%) were observed for a and b, i.e. when starting
from a mononuclear complex and the highest (>20%) for c, e and
k. These results confirm the advantage of using clusters as precur-
sors as they generally permit to obtain a better dispersion than
with mononuclear precursors [11]. Surprisingly, when using clus-
ters 2, 3 and 6 as precursors, the dispersion was better on CSX+ than
on CPPh2, while the opposite was found with 4 and 5. This might
indicate that, in the former cases, the ligand-exchange reaction on
the functionalized support did not occur as well as with the latter.
The values of Pd/C ratios measured by XPS displayed the opposite
trend than the dispersion values (i.e. when Pd/C ratios measured
by XPS decreased, dispersion increased, mainly for CSX+-supported
samples; the opposite being true for most CPPh2-supported sam-
ples). This confirms that Pd is more located on the external surface
3.2. Activation and characterization of the catalysts
These supported clusters were then thermally treated in order to
prepare supported Pd nanoparticles by removing the ligand shells.
The experimental conditions were determined by TGA analyses of
the precursors in their pure form (Table 2). All the clusters lose
most of their ligands below 300 ◦C. Thus, the supported clusters
were activated at 300 ◦C for 1 h under N2. In the case of the complex
[Pd(dba)2] (1), the final decomposition temperature being at 400 ◦C,
it implied an activation at this temperature for 1 h under N2. How-
ever, in all cases, the experimental weight losses were lower than
the calculated losses based on the assumption that all ligands would
be removed. We believe that decarbonylation occurs more easily
than removal of the phosphine ligands. Some remaining fragments
belonging to the phosphine groups must therefore still be present
after thermal treatment. In the case of the clusters anchored onto
the phosphine-functionalized support, the situation might even be
worse, as the phosphines are stabilized through covalent bond with
the surface. This has been verified by XPS which shows the pres-
ence of P on CPPh2 after thermal treatment at 300 ◦C for 1 h under
nitrogen.
The activated samples were first analyzed by elemental analysis,
XPS and CO chemisorption. These results are displayed in Table 3.
The exact metal loading of the Pd/C catalysts were determined by
ICP-OES elemental analysis after dissolution. The numbers obtained
correspond closely to the loading determined before activation by
atomic absorption, meaning that the soft activation conditions used
here to remove the clusters ligand shells did not cause loss of metal
by volatilization. Catalysts a, b, c, e, k and l presented the highest
loading with values between 2.4 and 3 wt.%. The catalyst g pre-
sented the lowest loading with 1 wt.% and all others presented a
loading of around 2 wt.%. The obtained XPS Pd/C atomic surface
in the case of CPPh2, causing lower dispersion. In the case of CSX+
-
supported samples, the metal penetrates deeper within the porous
network, hence leading to higher dispersion.
The activated catalysts were also analyzed by XRD, SEM and
TEM. The XRD diffractograms generally displayed no signal, indi-
cating that the supported Pd particles were of very small sizes. Only
catalysts a, g and i gave small peaks, indicating that these cata-
lysts present the biggest particles on their surface. This is roughly
in agreement with the CO chemisorption measurements (Table 3)
as these catalysts were among those presenting the lowest dis-
persions. The recorded SEM images of the Pd/C catalysts did not
bring any more information than before activation. The catalysts
supported on CPPh2 presented generally a more uniform surface
with small particles of similar sizes. In order to gain a more pre-
cise idea of the particles sizes, the catalysts supported on CPPh2
were characterized by TEM. The recorded images are displayed
in Fig. 3 with an indication of the observed particle size. One can
observe that the supported particles were of nanometrical sizes

C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
177
Fig. 3. TEM images of the activated Pd/C catalysts supported on CPPh2
.
(<10 nm) and evenly dispersed on the functionalized support CPPh2
.
which is in agreement with CO chemisorption and XRD analyses,
and l displayed the least uniform surface with particle sizes from
1 nm to 10 nm. These results confirm that the use of a functional-
ized support allows to obtain small and well-dispersed supported
The catalysts d, f and j displayed the smallest particles (1–2 nm)
amongst the CPPh2–supported catalysts, with narrow sizes distri-
butions. Catalysts b and h displayed bigger particles (1–5 nm),
Table 2
Thermogravimetric analyses.
Precursor
Weight loss (%)
Final decomposition temp. (◦C)
Calcd. weight loss (%)
1
2
3
4
5
6
72.5
54.0
47.4
64.9
53.1
24.4
400
200
200
280
300
250
81.5 (2 dba)
59.3 (12 CO + 6 PBu3)
54 (17 CO + 5 PBu3)
76 (5 CO + 4 PPh3 + 2 CH2Cl2)
64.2 (12 CO + 6 PPh3)
39.7 (32 CO + 3 PPh3)

178
C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
Table 3
Characterization of the activated Pd/C catalysts and catalytic results.
Precursor
Name of
obtained catalyst
Loading before
Loading after
XPS Pd/C ratios
before activation
XPS Pd/C ratios
after activation
Dispersion
(%)c
Activity per g of catalyst
TOFd
activation (wt.%)a
activation (wt.%)b
(mol L−1 min−1 g−1
)
1
2
3
4
5
6
–
Pd/CSX+ (a)
Pd/CPPh2 (b)
2.99
2.11
2.48
1.54
2.58
2.54
0.99
1.38
2.16
1.79
2.85
2.65
–
2.97
2.44
2.35
1.68
2.78
2.11
1.02
1.81
1.94
1.98
2.61
2.93
5
0.012
0.022
0.007
0.017
0.010
0.022
0.011
0.010
0.029
0.014
0.017
0.027
n.d.
0.014
0.015
0.006
0.012
0.008
0.016
0.006
0.006
0.014
0.010
0.014
0.021
n.d.
3.0
3.0
29.1
17.4
25.6
7.8
3.3
8.1
7.4
0.0030
0.0048
0.0334
0.0115
0.0314
0.0133
0.0008
0.0090
0.0020
0.0134
0.0150
0.0190
0.0465
143
279
208
167
188
344
101
261
59
234
111
303
337
Pd10Med/CSX+ (c)
Pd10Med/CPPh2 (d)
Pd12Med/CSX+ (e)
Pd12Med/CPPh2 (f)
Pd4/CSX+ (g)
Pd4/CPPh2 (h)
Pd10/CSX+ (i)
Pd10/CPPh2 (j)
Pd24/CSX+ (k)
Pd24/CPPh2 (l)
12.3
22.1
9.1
e
JM5%
11.8
a
See Table 1.
b
The Pd loading after activation was determined by ICP-OES elemental analysis, except for JM5% commercial catalyst.
Measured by CO chemisorption.
The TOF is expressed as the number of NB molecules reduced per surface Pd atom and per minute.
n.d. = non determined.
c
d
e
Fig. 4. Activity per g of Pd: catalysts supported on CSX+ in white and those supported on CPPh2 in black.
particles when using clusters as precursors. The catalyst prepared
from mononuclear complex (b) and the catalyst prepared from the
Pd24 compound (l), which is thought to be of colloidal nature [31],
both display larger distributions of particle sizes. It should be noted
however, that even in the case of clusters thought to be anchored
on the phosphine-functionalized support, the final particle sizes
do not correspond simply to the bare metallic core of the starting
clusters. Some agglomeration still occurred, but to a lesser extent.
the different prepared catalysts, the activities have been normal-
ized to be expressed per gram of Pd. These results are illustrated
in Fig. 4. The highest activities were observed for Pd10Med/CSX+ (c)
and Pd12Med/CSX+ (e) and not for the commercial catalyst anymore.
These two catalysts presented the highest dispersions (see Table 3).
The catalysts presenting the lowest activities (Pd/CSX+ (a), Pd/CPPh2
(b), Pd4/CSX+ (g) and Pd10/CSX+ (i)) were those with the lowest
dispersions and showing signals in the X-ray diffractograms. The
other catalysts presented activities comparable to JM5% material.
As shown in Fig. 5, the activity per gram of Pd is function of the
3.3. Nitrobenzene hydrogenation
All the catalysts prepared in this work and summarized in
Table 3 were used in the hydrogenation of nitrobenzene into
aniline as a standard test-reaction for structure effects. The
resulting activity of each catalyst is also displayed in Table 3.
For comparison, a commercial Pd 5 wt.%/C catalyst (JM5%) was
tested in the same conditions. Moreover, in order to verify that the
pristine and functionalized supports do not play any role in the
observed activity, as it can be the case with Pd/C catalysts used
in nitroaliphatic compound (such as 2-methyl-2-nitropropane)
hydrogenation [34], blank tests were conducted with the supports
alone (without deposited metal). These experiments showed that
the CSX+ support by itself does not give any conversion and that
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
C
PPh2 gives rise to a negligible activity of 0.0007 mol L−1 min−1 g−1
When looking at the activity per gram of catalyst (Table 3, sec-
.
0
5
10
15
20
25
30
35
Dispersion (%)
ond to last column), the JM5% catalyst presented unsurprisingly the
highest activity while Pd4/CSX+ (g) presented the lowest one. This
is due to their respective Pd loadings. As the loading varied among
Fig. 5. Activity per g of Pd as a function of the dispersion (measured by CO
chemisorption).

C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
179
Fig. 6. Activity per g of accessible Pd: catalysts supported on CSX+ in white and those supported on CPPh2 in black.
dispersion. In this figure, plotting the activity per gram of Pd in
function of the dispersion, a sequenced evolution is observed with
a significant rise in activity when reaching 7–8% of dispersion and
tending to a linear evolution above 20%. The evolution of the activ-
ity as function of dispersion is not regular, indicating that more
subtle effects play a role in the activity differences observed.
Given this, the activities per gram of Pd were divided by the dis-
persion figures in order to obtain activities per gram of accessible
Pd. In this way, the activities would represent the intrinsic per-
formances of the surface Pd atoms playing a catalytic role. These
results are displayed in Fig. 6. First, the activities are different for
each catalyst, which confirms the “structure sensitive” nature of the
reaction, as already mentioned in the literature [25]. Indeed, if the
reaction was only dependent on the dispersion, all catalysts would
present the same activity in Fig. 6. Secondly, the activities obtained
for the catalysts supported on CPPh2 were generally higher (except
for Pd10Med/C) than those supported on CSX+. Moreover, these activ-
ities were competitive with that of the commercial catalyst. This
shows the advantage of using a functionalized support. The pres-
ence of a covalent link to a phosphine ligand probably alters the
activation mechanism, retarding coalescence of metal atoms and
crystallization of metallic phases. The obtained nanoparticles con-
tain therefore more defects that are known to be very active sites.
In addition, as XPS characterization unraveled, Pd is more situated
on the external surface in the case of CPPh2, making it more accessi-
ble for a liquid phase process. Indeed, internal diffusion of reactants
within the pores of a microporous solid cannot be under-evaluated
[35]. Moreover, we believe that, even if CO ligands are easily driven
off by the activation procedure, some residue of the phosphine arm
might remain underneath the metallic nanoparticles (see above
comments in Table 2), acting as ‘glue’ with the support, but play-
ing also electronic effects. It cannot be excluded either that some P
atoms migrate within the nanoparticles.
Finally, for a better comparison with the results of the litera-
ture, TOFs, expressed as the number of NB molecules reduced per
surface Pd atom and per minute, were also calculated (Table 3, last
column). The catalysts obtained in our case presented activities
similar to those described in the literature and prepared from
H2PdCl4 [24]. No direct linear correlation between these activities
and the dispersion had been observed either. However, it had been
shown that when the dispersion decreases, the TOF increases, indi-
cating that NB prefers plane catalytic sites than edges or corners
[24,25]. This was also observed in our case, as the catalysts present-
ing the smallest particles in TEM images are not giving the highest
TOFs values. On the contrary, the best TOF values are obtained for
the catalysts b, f and l (Table 3, last column), which do not present
particles smaller than 2 nm (see Fig. 3). It should also be noted that
the TOF values are higher for catalysts supported on CPPh2 than on
CSX+, indicating again that the Pd atoms are somehow ‘activated’
by the preparation procedure involving a functionalized support.
The filtrates of the catalytic reaction media were analyzed by
palladium atomic absorption in order to assess the metal losses of
the catalysts during the tests. For example, catalyst i lost 82% of
its metal content during catalytic test, while catalyst j retained all
the palladium on its surface. Thus, functionalization of the support
plays a significant role on the retention of the metallic phase dur-
ing the catalytic tests. However, catalyst l presented a loss of 17%
of Pd, indicating that all the metallic phase is not, each time, totally
retained on CPPh2, but that the loss is limited compared to the one
observed on CSX+. The residue of thermally-treated phosphines act-
ing as ‘glue’ might explain this effect. Thus the major benefit of using
a functionalized support resides in inhibition of metal leaching.
4. Conclusion
In this work, we have shown that molecular palladium clus-
ters could be anchored onto a functionalized support in order to
synthesize heterogeneous catalysts presenting well-dispersed sup-
ported nanoparticles. Five different carbonyl palladium clusters
were incorporated on a phosphine-functionalized carbon support
(CPPh2) and on an unmodified one (CSX+) for comparison. A mono-
nuclear complex was also used as precursor as reference. The CPPh2
support was characterized by solid-state NMR in order to confirm
the presence on its surface of covalently-grafted chelating phos-
phine groups. SEM images were recorded before and after thermal
treatment and they showed that the surface of the functionalized
support-derived samples was more uniform than with the unmo-
dified CSX+, i.e. presented less agglomerates. Surface Pd/C atomic
ratios measured by XPS were generally higher on CPPh2 than on
CSX+. The size of the obtained activated particles on the functional-
ized support was confirmed by TEM and XRD analyses. TEM images
showed well-dispersed particles with sizes comprised between 1
and 10 nm with a narrow size distribution (except for Pd24/CPPh2).
The XRD patterns displayed most of the time no signal which
confirms the formation of small particles. Metal dispersions were
measured by CO chemisorption and were in agreement with the
XRD measurements as the few catalysts presenting the lowest dis-
persions were those presenting small signals in the diffractograms.
The lowest dispersions were obtained for [Pd(dba)2] on CPPh2 as
well as on CSX+, indicating that the use of clusters allows to obtain
more efficiently well-dispersed particles than mononuclear com-
plexes.
The Pd/C catalysts obtained displayed a catalytic activity in the
hydrogenation of nitrobenzene into aniline. When the activities
were normalized and expressed per gram of accessible Pd, the
catalysts supported on CPPh2 presented a higher activity than

180
C. Willocq et al. / Journal of Molecular Catalysis A: Chemical 365 (2012) 172–180
those supported on CSX+. Moreover, they were competitive with
a commercial catalyst and reports of the literature. These higher
activities were ascribed to retarded formation of metallic nanopar-
ticles during activation, enrichment in Pd on the external surface,
and residual P from the tethering arm acting as glue. Additional
catalytic tests showed that the use of a functionalized support
permits to retain more efficiently the active phase at the surface
of the catalysts during testing.
[10] M.C. Roman-Martinez, J.A. Diaz-Aunon, C.S.M. de Lecea, H. Alper, J. Mol. Catal.
A: Chem. 213 (2004) 177.
[11] P.J. Braunstein, J. Rose, in: P. Braunstein, L.A. Oro, P.R. Raithby (Eds.), Metal
Clusters in Chemistry, Wiley-VCH, Weinheim, 1999, p. 616.
[12] O. Alexeev, B.C. Gates, Top. Catal. 10 (2000) 273.
[13] R. Raja, G. Sankar, S. Hermans, D.S. Shephard, S. Bromley, J.M. Thomas, B.F.G.
Johnson, Chem. Commun. (1999) 1571.
[14] R. Raja, T. Khimyak, J.M. Thomas, S. Hermans, B.F.G. Johnson, Angew. Chem. Int.
Ed. 40 (2001) 4638.
[15] S. Hermans, R. Raja, J.M. Thomas, B.F.G. Johnson, G. Sankar, D. Gleeson, Angew.
Chem. Int. Ed. 40 (2001) 1211.
[16] O. Alexeev, G. Panjabi, B.C. Gates, J. Catal. 173 (1998) 196.
[17] A. Zhao, B.C. Gates, J. Catal. 168 (1997) 60.
Acknowledgements
[18] C. Moreno-Castilla, M.A. Salas-Peregrin, F.J. Lopez-Garzon, Fuel 74 (1995) 830.
[19] B.L. Garcia, B. Captain, R.D. Adams, A.B. Hungria, P.A. Midgley, J.M. Thomas, J.W.
Weidner, J. Clust. Sci. 18 (2007) 121.
[20] J. Venter, M. Kaminsky, G.L. Geoffroy, M.A. Vannice, J. Catal. 105 (1987) 155.
[21] J. Venter, M. Kaminsky, G.L. Geoffroy, M.A. Vannice, J. Catal. 103 (1987) 450.
[22] S.M. Yunusov, E.S. Kalyuzhnaya, H. Mahapatra, V.K. Puri, V.A. Likholobov, V.B.
Shur, J. Mol. Catal. A: Chem. 139 (1999) 219.
[23] E. Auer, A. Freund, J. Pietsch, T. Tacke, Appl. Catal. A: Gen. 173 (1998) 259.
[24] N. Bouchenafa-Saib, P. Grange, P. Verhasselt, F. Addoun, V. Dubois, Appl. Catal.
A: Gen. 286 (2005) 167.
[25] G. Carturan, G. Facchin, G. Cocco, G. Navazio, G. Gubitosa, J. Catal. 82 (1983) 56.
[26] J. Relvas, R. Andrade, F.G. Freire, F. Lemos, P. Araujo, M.J. Pinho, C.P. Nunes, F.R.
Ribeiro, Catal. Today 133–135 (2008) 828.
The authors gratefully acknowledge the “Fonds de la Recherche
Scientifique” (FRS-FNRS) and the “Fonds pour la Formation à la
Recherche dans l’Industrie et dans l’Agriculture” (FRIA) for financial
support, the NORIT firm for supplying the SXplus carbon support, as
well as J.-F. Statsijns, P. Lipnik and D. Desmecht for their technical
assistance. This work has been performed within the framework
of the Interuniversity Attraction Poles Programme of the Belgian
State, Belgian Science Policy (Project INANOMAT P6/17).
[27] Y. Zhao, C.-H. Li, Z.-X. Yu, K.-F. Yao, S.-F. Ji, J. Liang, Mater. Chem. Phys. 103
(2007) 225.
References
[28] C. Willocq, S. Hermans, M. Devillers, J. Phys. Chem. C 112 (2008) 5533.
[29] C. Willocq, D. Vidick, B. Tinant, A. Delcorte, P. Bertrand, M. Devillers, S. Hermans,
Eur. J. Inorg. Chem. (2011) 4721–4729.
[30] S. Hermans, T. Khimyak, N. Feeder, S.J. Teat, B.F.G. Johnson, Dalton Trans. (2003)
672.
[31] C. Willocq, B. Tinant, F. Aubriet, V. Carré, M. Devillers, S. Hermans, Inorg. Chim.
Acta 373 (2011) 233.
[32] E.G. Mednikov, N.K. Eremenko, S.P. Gubin, Yu L. Slovokhotov, Yu T. Struchkov,
J. Organomet. Chem. 239 (1982) 401.
[33] E.G. Mednikov, N.K. Eremenko, S.P. Gubin, J. Organomet. Chem. 202 (1980)
C102.
[34] V. Dubois, G. Jannes, P. Verhasselt, Stud. Surf. Sci. Catal. 108 (1997) 263.
[35] S.L. Pirard, C. Diverchy, S. Hermans, M. Devillers, J.-P. Pirard, N. Job, Catal. Com-
mun. 12 (2011) 441–445.
[1] A. Choplin, F. Quignard, Coord. Chem. Rev. 178–180 (1998) 1679.
[2] J.M. Basset, F. Lefebvre, C. Santini, Coord. Chem. Rev. 180 (1998) 1703.
[3] J. Guzman, B.C. Gates, Dalton Trans. (2003) 3303.
[4] A.R. Silva, M. Martins, M.M.A. Freitas, A. Valente, C. Freire, B. de Castro, J.L.
Figueiredo, Microporous Mesoporous Mater. 55 (2002) 275.
[5] A.R. Silva, M. Martins, M.M.A. Freitas, J.L. Figueiredo, C. Freire, B. de Castro, Eur.
J. Inorg. Chem. (2004) 2027.
[6] A.R. Silva, V. Budarin, J.H. Clark, B. de Castro, C. Freire, Carbon 43 (2005)
2096.
[7] A.R. Silva, J.L. Figueiredo, C. Freire, B. de Castro, Catal. Today 102 (2005)
154.
[8] B. Jarrais, A.R. Silva, C. Freire, Eur. J. Inorg. Chem. (2005) 4582.
[9] J.A. Diaz-Aunon, M.C. Roman-Martinez, C.S.M. de Lecea, H. Alper, Stud. Surf. Sci.
Catal. 143 (2002) 295.