Journal of Catalysis 316 (2014) 260269

Contents lists available at ScienceDirect

Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat

Dense and narrowly distributed silica-supported rhodium and iridium

nanoparticles: Preparation via surface organometallic chemistry
and chemisorption stoichiometry
Florent Hroguel a, Dominique Gebert a, Michael D. Detwiler b, Dmitry Y. Zemlyanov c, David Baudouin a,
Christophe Copret a,
a

ETH Zrich, Department of Chemistry and Applied Biosciences, Vladimir-Prelog Weg 2, CH-8093 Zrich, Switzerland
Purdue University, School of Chemical Engineering, 480 Stadium Mall Drive, West Lafayette, IN 47907-2100, United States
c
Purdue University, Birck Nanotechnology Center, 1205 W. State Street, West Lafayette, IN 47907-2057, United States
b

a r t i c l e

i n f o

Article history:
Received 1 November 2013
Revised 3 April 2014
Accepted 29 May 2014
Available online 27 June 2014
Keywords:
Silica
Surface organometallic chemistry (SOMC)
Rhodium
Iridium
Single-site
Nanoparticles

a b s t r a c t
Silica supported iridium and rhodium nanoparticles were prepared via surface organometallic chemistry
(SOMC). Following the synthesis and characterization of organometallic molecular precursors
[(COD)MOSiOR]2 (M = Ir or Rh; R = Me or Si(OtBu)3), their controlled grafting on partially dehydroxylated
silica yields isolated dinuclear surface species, as determined by 1H and 13C solid state NMR as well as
infrared transmission spectroscopies and elemental analysis. The decomposition under hydrogen of the
well-dened surface species affords a narrow size distribution and a homogeneous spatial repartition
of small M(0) particles despite a high metal density (1.2 0.3 nm and 1.4 0.3 nm for Ir and Rh, respectively). A combination of transmission electronic microscopy and gas chemisorption provides the H2 and
CO adsorption stoichiometries on the metal surface, which are highly dependent on the precursor and the
preparation route, indicating the necessity to control each step and the danger to determine particle size
solely from chemisorption studies for small iridium and rhodium supported particles.
2014 Elsevier Inc. All rights reserved.

1. Introduction
Supported nanoparticles play a key role in catalysis, where the
control of their characteristics (size, shape, composition) is key for
the understanding of their properties and the improvement of their
performance [1]. Most of the supported metallic nanoparticles are
prepared by homogeneous deposition-precipitation or by impregnation of a metal precursor on a support, followed by postthermal/chemical treatments. The impregnation approach is
broadly used for the preparation of supported iridium and rhodium
nanoparticles, which are used in catalytic conversion processes,
such as the hydrogenolysis of hydrocarbons or deNOx processes.
Of various precursors, metal chlorides, nitrates and acetylacetonates (acac) are widely used. For instance, Ir(acac)3 dispersed on
silica leads to the formation of Ir particles with size ranging
between 1 and 5 nm after treatment under H2 [24]. For Rh, surface ion exchange using RhCl3 in the presence of ammonia allows
for a homogeneous distribution of [RhCl(NH3)5]2+ at the surface
through a strong interaction of this cationic species with the
Corresponding author.
E-mail address: ccoperet@inorg.chem.ethz.ch (C. Copret).
http://dx.doi.org/10.1016/j.jcat.2014.05.023
0021-9517/ 2014 Elsevier Inc. All rights reserved.

negatively charged silica surface, thus leading to the formation of

particles with a size ranging between 1 and 2 nm after decomposition [5]. However, a common issue resulting from impregnation
techniques with some inorganic derivatives is the formation of
particle aggregates and broad particle size distribution. In addition,
the decomposition of the metal precursor (its ligands) can not only
lead to the uncontrolled growth of nanoparticles, but also to the
pollution of the metal surface by the presence of residual ligands,
which cannot be readily detected by transmission electron microscopy. Such pollution has important consequences on the chemisorption of probe molecules such as H2 and CO, which are often
used to evaluate the particle size (when the probe molecule surface
stoichiometry is known). It may thus not be too surprising that to
date there is no consensus concerning H2 or CO adsorption stoichiometry for some metals like iridium and rhodium, while it is now
well-established for other metals such as ruthenium [6], osmium
[7], nickel [8,9], and platinum [10]. In fact, the H/Msurf ratio has
been reported in a range from 0.8 to 2.8 in the case of iridium
[3,11,12] and from 1.2 to 2.0 for rhodium [1113]. In view of the
importance of Rh and Ir in catalyis, it is important to develop more
controlled methods to obtain small supported Rh and Ir nanoparticles with a narrow distribution, a high particle density on the

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

support with a clean surface and to establish the adsorption

stoichiometry.
One possible and alternative approach to prepare supported
nanoparticles is to use surface organometallic chemistry [1422],
which relies on the controlled grafting of molecular precursor, typically on the hydroxyl surface functionalities of an oxide support.
This process yields well-dened surface complexes, whose metal
density is mainly controlled by the initial OH density, which can
be controlled by the pre-treatment temperature of the support
[23]. Upon post-treatment, typically under H2, this approach
allows for the controlled growth of nanoparticles with small sizes
with a narrow distribution and a homogeneous spatial distribution
combined with a high metal surface density. This concept, already
proposed by Ermakov in the early 1980s [15], has been used more
recently for preparing silica-supported Au [24] and Pt nanoparticles with narrow size distribution and controlled interfaces
[10,25,26]. In the context of Rh and Ir, the decomposition of iridium and rhodium allyl complexes was also studied on various supports, and leads to metal nanoparticles with diameters ranging
between 1 and 3 nm [2729]. In a more recent study, [Ir(COD)Cl]2
was used to prepare 2.7 nm iridium particles on alumina [3032].
Other alternative approaches have used small rhodium and iridium
clusters as metal source [33,34].
Here, we focus our study on the controlled growth of silica-supported Rh and Ir nanoparticles, starting from well-dened and
readily prepared cyclooctadiene M(I) precursors, [M(COD)(X)],
with X = Cl, OMe and OSi(OtBu)3. We investigate the effect of the
nature of the anionic ligand X on the nal supported nanoparticles,
in particular with respect to the size, the density and the hydrogen
and CO chemisorption stoichiometry.

261

and 700 MHz spectrometers with a conventional double resonance

4 mm CP-MAS probe. The samples were introduced under Ar in a zirconia rotor, which was then tightly closed. In all experiments, the
rotation frequency was set to 10 kHz unless otherwise noted. Chemical shifts are given with respect to tetramethylsilane as the external
reference for both 1H and 13C NMR. 1H and 13C liquid-state NMR
spectra were recorded on Bruker DRX 200, 250 and 300 and
400 MHz spectrometers. The spectra were recorded at room
temperature. Chemical shifts are given with respect to the residual
solvent peak chemical shift.
2.2.3. X-ray crystallography
The crystals were placed in paratone and mounted in the beam
under a ow of nitrogen at 100 K on a Bruker SMART APEX II
diffractometer equipped with a CCD area detector using Mo Ka
radiation. Empirical absorption correction was performed with
SADABS-2008/1 (Bruker). The structures were solved by direct methods (SHELXS-97) followed by least-square renement (SHELXL-97)
using WinGX suite of programs and OLEX2-1.1. The non-hydrogen
atoms were rened anisotropically. For RhOSi(OtBu)3, the hydrogen
atoms were rened anisotropically while they were placed at
calculated positions for other complexes. Projected surface areas
were calculated with ChemAxon MarvinSketch software using
structures obtained by single crystal XRD.
CCDC
969418
([(COD)Ir(OSi(OtBu)]2),
CCDC
969419
([(COD)IrOMe]2), and CCDC 969420 ([(COD)Rh(OSi(OtBu)]2) contain the supplementary crystallographic data for this paper. These
data can be obtained free of charge from The Crystallographic Data
Centre via www.ccdc.cam.ac.uk/data_request/cif. Figure Graphics
are generated using MERCURY 3.1 supplied with Cambridge
Structural Database; CCDC: Cambridge, U.K., 20012012.

2. Experimental section
2.1. General procedures
All experiments were conducted under argon atmosphere using
Schlenk or glove-box techniques. Hydrated iridium trichloride,
hydrated rhodium trichloride, (acetylacetonato)(1,5-cyclooctadiene)iridium(I) and hydrated rhodium nitrate were purchased from
Strem. [(COD)IrCl]2 [35], [(COD)RhCl]2 [36], [(COD)IrOMe]2 [37],
[(COD)RhOMe)]2 [37] and [(COD)RhOSi(OtBu)3]2 [38] were prepared according to published procedures. Tris(tert-butoxy)silanol
99.999% (HOSi(OtBu)3) was purchased from SigmaAldrich and
used as received. Pentane, diethyl ether and dichloromethane were
dried using an MBraun solvent purication system, and then contacted with 4 molecular sieves (MS) and degassed under vacuum.
Benzene was dried over Na/K/benzophenone and distilled. Dihydrogen was puried over R3-11 BASF catalyst/MS 4 prior to
use. Gas phase analyses were performed on an Agilent Technologies 7890 A Gas Chromatography apparatus equipped with a ame
ionization detector (FID) and a KCl/Al2O3 column (50 m 
0.32 mm). Elemental analyses were performed at Mikroanalytisches Labor Pascher in Germany.
2.2. Characterization methods
2.2.1. FTIR spectroscopy
Infrared spectra were recorded on a Bruker Alfa-T spectrometer
(inside a glove-box under argon atmosphere) using Opus software.
Samples were pressed into a thin pellet using a 7 mm die set. Typically 24 scans were accumulated for each spectrum at a resolution
of 4 cm1.
2.2.2. Nuclear magnetic resonance (NMR) spectroscopy
1
H magic angle spinning (MAS) and 13C cross polarization (CP)MAS solid-state NMR spectra were recorded on Bruker Avance 400

2.2.4. Computational details

Optimizations carried out using density functional theory (DFT)
calculations were performed with Gaussian 03 [39], using the
hybrid functional B3LYP [4042] with SDD basis set. Geometry
optimizations were achieved either with or without symmetry
restriction.
2.2.5. Transmission electron microscopy (TEM)
Transmission Electron Micrographs were obtained on a Tecnai
F30 ST with 300 kV acceleration voltage and High-Angle Annular
Dark-Field Scanning Transmission Electron Microscopy (HAADFSTEM) was performed on a Hitachi HD2700CS with 200 kV
acceleration voltage, in mode resulting in atomic number contrast
(Z contrast). Lacey carbon grids were prepared by directly dipping the grid into the silica powder (displacement of NPs was
observed in case of preparation of the grid by sonication of the grid
in suspension of the particles in ethanol). The particle size distribution was estimated by statistical analysis of ca. 300 particles. Metal
dispersion, dened as the ratio between surface and total metal
atoms, was calculated back from particle size distribution. A simple
average of a particle size distribution is not representative of the
sample dispersion, as a large particle will carry much more weight
than a small one on the nal Msurface/Mtotal ratio. In fact, for
instance iridium particles of 1 nm are made of roughly 38 atoms,
while those of 1.7 nm contain 201 atoms. Hence, a more representative estimation of metal dispersion was calculated using the
truncated cubic octahedron model to sum the surface and bulk
metal atoms of all individual particles counted from TEM pictures.
2.2.6. Chemisorption of H2 and CO
Chemisorption experiments were carried out on a Belsorb-Max
device from BEL Japan. In a measuring cell, ca. 100 mg of catalyst
were treated at 106 mbar at 623 K for 3 h using a ramp of 1 K min1.
Adsorption isotherm measurements performed at 298 K after this

262

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

pretreatment are referred in this paper as total adsorption, i.e.

adsorption on fully degassed particles. In all cases, the pressures at
equilibrium were recorded when the pressure variation was below
0.02% per minute. The quantication of gas adsorbed on surface
metal atoms was calculated from the adsorption at saturation deriving from a double Langmuir adsorption equation model, assuming
complete reduction of the metal and truncated cubic octahedron
geometry. A H2 and CO blank on metal free support was performed
and adsorbed negligible amounts (not shown).
2.2.7. X-ray photoelectron spectroscopy (XPS)
X-ray Photoelectron Spectroscopy (XPS) measurements were
performed using a Kratos Axis Ultra DLD spectrometer with monochromatic Al Ka radiation (hm = 1486.6 eV). The high-resolution
XPS spectra were collected in constant pass energy (PE) mode with
a PE of 20 eV. A built-in commercial Kratos charge neutralizer was
used to achieve better resolution. Air exposed and reduced catalysts were analyzed after loading into a stainless-steel sample
stub-holder capable of holding up to 5 powder samples. The
reduced catalysts were treated in a standard Kratos catalytic cell
(CatCell) at 500 C for 6 h (heating ramp of 2 C/min) in owing
5%H2/Ar at 50 sccm. The CatCell was connected to the XPS chamber
via transfer arms so that the reduced catalysts were not exposed to
air before analysis.
Binding energy (BE) values refer to the Fermi edge and the
energy scale was calibrated using Au 4f7/2 at 84.0 eV and Cu 2p3/2
at 932.67 eV. The photoemission peak positions were charge corrected to the adventitious carbon signal at 284.8 eV. Spectra were
analyzed using the CasaXPS software program, version 2.3.16 PR
1.6 (Casa Software Ltd.). A Shirley background was subtracted from
each Ir 4f and Rh 3d region before curve tting were t with asymmetric GaussianLorentzian peaks with tail dampening (CasaXPS
Lineshapes = LF(1, 2.4, 30, 90) and LF(1, 1.8, 30, 50) for Ir 4f and Rh
3d, respectively). Because of the asymmetric shape of the peaks,
the presence of small amounts of oxidized species, which would
appear in the higher binding energy tails, could not be eliminated.
Spin orbit coupling doublets in the Ir 4f (4f7/2 and 4f5/2) and Rh 3d
(3d5/2 and 3d3/2) regions were subject to spacing constraints of
2.98 eV and 4.7 eV, respectively [43]. The intensity ratio of the Ir
4f (4f7/2 and 4f5/2) and Rh 3d (3d5/2 and 3d3/2) spin orbit coupling
doublets was xed to be 4:3 and 3:2, respectively. The atomic concentrations of the chemical elements on the near-surface region
were estimated after the subtraction of a Shirley type background,
taking into account the corresponding Scoeld atomic sensitivity
factors and inelastic mean free path (IMFP) of photoelectrons using
standard procedures in the CasaXPS software.
2.3. Support material
SiO2-(700) was obtained as follows: silica (Aerosil Degussa,
200 m2 g1) was compacted with distilled water, dried at 120 C
for 2 days, sieved (250450 lm), calcined at 500 C under air for
4 h, treated under secondary vacuum (105 mbar) at 500 C for
12 h and then at 700 C for 6 h. The density of surface silanols
was determined by titration with methylmagnesium bromide to
be 0.26 mmol g1 [23,44].

the solvent volume to 20 mL, the reaction mixture was cooled to

30 C for crystallization. Large orange crystals suitable for XRD
were collected (602 mg, 72%). 1H NMR (300 MHz, C6D6, 20 C):
d = 4.69 (s, 8H, COD @CH), 2.52 and 2.30 (2 multiplets, 8H, COD
ACH exo), 1.62 and 1.58 (2 multiplets, 8H, COD ACH endo), 1.53
(s, 54H, OtBu). 13C NMR (300 MHz, C6D6, 20 C): 73.22 (OtBu),
58.12 and 55.03 (COD CH2), 32.80 (COD CH2), 32.42 (OtBu) and
31.81 (COD ACH2). FTIR: 1473, 1449, 1429 (COD), 1387,
1362 cm1 (OSi(OtBu)3). Elemental analysis: Ir = 33.8 wt%,
C = 41.62 wt%, H = 6.80 wt%, Si = 7 1.5 wt%.
2.5. Preparation of silica-supported nanoparticles
2.5.1. Grafting of [(COD)Ir(OSi(OtBu)]2 on SiO2-(700). Representative
procedure
[(COD)Ir(OSi(OtBu)]2 (421 mg; 0.37 mmol; 1.42 equiv.) was dissolved in benzene (15 mL) and contacted with 1.0 g of SiO2-(700)
(0.26 mmol SiOH; 1 equiv.). The reaction mixture was allowed to
react for 4 h at room temperature under gentle stirring. The solid
was washed 4 times with benzene and dried under vacuum
(<105 mbar) for 1 h to yield an orange solid. Ferrocene was added
to the collected solvents to quantify unreacted complex
(0.23 mmol; 0.88 equiv.) and released HOSi(OtBu)3 (0.13 mmol;
0.49 equiv.) by 1H liquid state NMR spectroscopy (recycle delay:
30 s). FTIR (cm1): 2978, 2939, 2922, 2891, 2876, 2841, 1475,
1451, 1432, 1390 and 1366. Elemental analysis: Ir = 4.40 wt%,
C = 4.36 wt%, H = 0.56 wt%.
2.5.2. Grafting of [(COD)Rh(OSi(OtBu)]2 on SiO2-(700)
The experimental procedure described above was carried out
using [(COD)Rh(OSi(OtBu)]2 (164 mg, 0.172 mmol, and SiO2-(700)
(0.5 g, 0.13 mmol SiOH). FTIR (cm1): 2976, 2943, 2924, 2888,
2840, 1477, 1455, 1436, 1390 and 1365. Elemental analysis:
Rh = 2.81 wt%, C = 5.04 wt%, H = 0.76 wt%.
2.5.3. Grafting of [(COD)IrOMe]2 on SiO2-(700)
The experimental procedure described above was carried out
using [(COD)IrOMe]2 (134 mg, 0.203 mmol, 1.5 equiv.), benzene
(10 mL) and SiO2-(700) (0.5 g, 0.13 mmol SiOH, 1 equiv.). FTIR
(cm1): 2980, 2965, 2945, 2923, 2892, 2877, 2842, 2818, 1477,
1453 and 1432. Elemental analysis: Ir = 5.58 wt%, C = 2.97 wt%,
H = 0.41 wt%.
2.5.4. Grafting of [(COD)IrO*Me]2 on SiO2-(700)
The experimental procedure described above was carried out
using [(COD)IrO*;Me]2 (86 mg, 0.129 mmol, 2.5 equiv.), benzene
(8 mL) and SiO2-(700) (0.193 g, 0.052 mmol SiOH, 1 equiv.). FTIR
(cm1): 2980, 2965, 2945, 2923, 2892, 2877, 2842, 2814, 1477,
1453 and 1432.
2.5.5. Grafting of [(COD)RhOMe]2 on SiO2-(700)
The experimental procedure described above was carried out
using [(COD)RhOMe]2 (98 mg, 0.203 mmol, 1.5 equiv.), benzene
(10 mL) and SiO2-(700) (0.5 g, 0.13 mmol SiOH, 1 equiv.). FTIR
(cm1): 2998, 2985, 2945, 2922, 2888, 2840, 2803, 1480, 1469,
1454 and 1435. Elemental analysis: Rh = 3.79 wt%, C = 4.09 wt%,
H = 0.62 wt%.

2.4. Preparation of organometallic precursors

2.4.1. Preparation of [(COD)Ir(OSi(OtBu)]2
A diethylether solution (40 mL) of KOSi(OtBu)3 [45] (0.900 g,
2.98 mmol, 2 equiv.) was added to a diethylether solution
(40 mL) of [(COD)IrCl]2 (1.0 g, 1.49 mmol, 1 equiv.) under inert
atmosphere, and the reaction mixture was stirred for 3 h at
25 C. All volatiles were evaporated, pentane was added, and the
reaction mixture was ltered through Celite. After reduction of

2.5.6. Grafting of [(COD)IrCl]2 on SiO2-(700)

The experimental procedure described above was carried out
using [(COD)Ir(Cl)]2 (136.0 mg; 0.405 mmol; 1.5 equiv.), benzene
(10 mL) and SiO2-(700) (0.135 mmol SiOH, 0.26 mmol SiOH/g,
1 equiv.). The solid discolored upon washing from orange to white.
After 3 washing and drying under high vacuum, FTIR spectroscopy
revealed the absence of any organic moieties in the surface.
Reaction in dichloromethane or ethyl acetate led to a similar result.

263

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

2.5.7. Grafting of [(COD)RhCl]2 on SiO2-(700)

The experimental procedure described above was carried out
using [(COD)RhCl]2 (20 mg, 0.041 mmol, 1.5 equiv.) and SiO2-(700)
(0.10 g, 0.026 mmol SiOH, 1 equiv). The solid discolored upon
washing from orange to white. After 3 washing and drying under
high vacuum, FTIR spectroscopy revealed the absence of any
organic moieties in the surface.

3. Results

should readily react with surface silanols and yield the corresponding
surface complex and HX, we prepared and characterized them with
the classical methods of molecular chemistry (elemental analysis,
IR and NMR spectroscopies and in some cases X-ray) as well as computational chemistry. MOMe and MOSi(OtBu)3 are readily prepared from
MCl in one step (Scheme 1), and the solid-state structure of IrOSi(OtBu)3,
RhOSi(OtBu)3 and IrOMe were determined by X-ray crystallography (see
Fig. 1, all bond distances and angles are presented in ESI). They all
present dinuclear structures in the solid-state according to single
crystal XRD (Table 1). All complexes, but RhOMe presenting a planar
M2O2 core [46] {the dihedral angle x (M1X1X2M2) = 180}, have
a buttery structure with dihedral angles ranging from 108.32 for
IrOSi(OtBu)3 to 168.35 for RhCl, similarly to structures observed for
analogous complexes [4749]. In 1H NMR, IrCl, RhCl, IrOMe, and RhOMe
display a single broad resonance for each type of cyclooctadiene protons (vinylic, endo and exo allylic), even when lowering the temperature to 180 K. While this is consistent with the D2h symmetry of
RhOMe in the solid state, it shows that the other complexes are quite
uxional. A contrario, IrOSi(OtBu)3 and RhOSi(OtBu)3 show 2 distinct resonances for cyclooctadiene endo and exo allylic protons, suggesting a
buttery structure in solution as found in the solid state. Finally, in
13
C NMR, note that the vinylic carbons of IrOSi(OtBu)3 are strongly
shielded compared to RhOSi(OtBu)3 (58.1 and 55.0 ppm for the former
and 76.1 and 72.0 ppm for the latter), while other peaks are observed
at similar chemical shifts (for iridium, OtBu is observed at 73.2 and
32.4 ppm and allylic carbons at 31.8 and 32.8 ppm). Those geometries are consistent with DFT calculations (details presented in ESI)
as the most stable calculated structures for IrCl, RhCl, IrOMe, and RhOMe
have a at core (x0 = 180), while folding of the M2O2 core stabilize
IrOSi(OtBu)3 (x0 = 111.49) and RhOSi(OtBu)3 (x0 = 118.90) where buttery interconversion (and so averaging of chemical shifts) is prevented because of the high energy of the at intermediate
structures (23.9 and 22.4 kcal/mol higher than the folded structure,
respectively), as determined by DFT calculations. Solid-state structures slightly differ from the calculated structures for all complexes
but RhOMe and hence present different roof angles, as shown in
Table 1. Those differences are attributed to packing effects, as energy
differences between solid-state and optimized structures are relatively low (DE < 2.7 kcal/mol). While repulsive interactions between
lled p orbitals on the bridging ligand and d orbitals on metal were
proposed to explain folding of M2O2 core [50], we suggest that the
main contribution is steric repulsion, where the presence of the bulky
siloxy ligands, in opposition to chloro or methoxy ligands, leads to
bending of the structure to stabilize the molecule, as indicated by
DFT calculations.

3.1. Preparation of the molecular complexes

3.2. Reactivity of MX with SiO2-(700)

Prior to the grafting of [(COD)M(X)]2 (M = Ir and Rh) MX

with different anionic ligands (X = Cl, OMe and OSi(OtBu)3), which

First, the IR spectrum of a silica dehydroxylated at 700 C,

SiO2-(700) (0.26 mmol OH g1 ca. 0.8 OH nm2) [23,44], contacted

2.5.8. Impregnation of (COD)Ir(acac) on SiO2-(700)

An dichloromethane solution of (COD)Ir(acac) (the concentration was adjusted to achieve the desired loading) was added dropwise to SiO2-(700) to achieve incipient wetness impregnation, under
argon atmosphere with vigorous stirring to ensure homogeneous
repartition of the precursor onto the support. The powder was then
transferred to a glass ow reactor and dried at 80 C (0.5 C/min)
for 12 h.
2.5.9. Impregnation of Rh(NO3)3on SiO2-(700)
An aqueous solution of Rh(NO3)3 (the concentration was
adjusted to achieve the desired loading) was added dropwise to
SiO2-(700) to achieve incipient wetness impregnation, under argon
atmosphere with vigorous stirring to ensure homogeneous
repartition of the precursor onto the support. The powder was then
transferred to a glass ow reactor and dried at 100 C (0.5 C/min)
for 12 h.
2.5.10. Formation of supported nanoparticles under static conditions.
Representative procedure
500 mbar of dihydrogen were added to a 500 mL glass reactor
containing 200 mg of supported species, and the powder turned
from yellow/orange to black/dark grey within 5 min. The mixture
was heated up to 300 C (0.5 C/min) for 12 h, and the reactor
was evacuated under high vacuum for 30 min at 300 C and
transferred to a glove-box.
2.5.11. Formation of supported nanoparticles under ow conditions.
Representative procedure
A ow of dihydrogen was passed through a U shaped glass frit
reactor (20 mL/min) containing 200 mg of supported species, and
the mixture was heated up to 500 C (0.5 C/min) for 12 h under
ow. The solid turned from yellow/orange to black/dark grey.
The reactor was cooled to 300 C, evacuated for 30 min at 300 C
and transferred into a glove-box.

(a)

Si(OtBu)3
Cl
M

M
Cl

2 KOSi(OtBu)3
Et2O
- 2 KCl

O
M

M
O
Si(OtBu)3
Me

(b)

Cl
M

M
Cl

2 KOH, 2 MeOH
- 2 KCl, - 2 H2O

M
O
Me

Scheme 1. Preparation of (a) MOSi(OtBu)3 and (b) MOMe (M = Ir or Rh).

264

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

Table 1
Selected data for precursors.

a
b
c
d

M = Rh

M = Ir

X = Cl

X = OMe

X = OSi(OtBu)3

X = Cl

X = OMe

X = OSi(OtBu)3

M1M2
X1X2
M1X1M2
M1X2M2
xa (XRD)
x0b (calculation)
DEc (kcal/mol)
Refs.

3.517
3.272
94.11
93.32
168.35
180
0.0
[51]

3.231
2.546
103.52
103.52
180
180
0.0
[46]

2.757
2.626
81.70
81.06
109.97
118.90
1.5

2.910
3.216
74.53
74.72
114.75
180
2.7
[52]

3.101
2.498
75.05
74.12
141.97
180
0.9

2.776
2.590
81.85
82.09
108.32
120.75
1.8

x is the dihedral angle M1X1X2M2 measured in the solid state.

x0 is the dihedral angle M1X1X2M2 for the most stable calculated structure.
DE is the energy difference between the solid state structure and the most stable calculated structure.
Present work.

with MCl (M = Ir or Rh) at 25 C in dichloromethane for 4 h

followed by 4 washing steps shows that no or little grafting took
place as indicated by the persistence of the band characteristic of
free surface silanol at 3747 cm1 and the absence of any band in
the m(CAH) region (Fig. S4); this observation is in sharp contrast
to what has been observed for the grafting of the same complex
on c-alumina [30]. In opposition, IrOSi(OtBu)3 (1.39 equiv. per surface silanols) reacts at room temperature with SiO2-(700) as indicated by the disappearance of the band associated with the
surface free silanols at 3747 cm1, the appearance of a broad band
centered at 3713 cm1 associated with OAH in interaction with
adjacent hydrocarbyl ligands and the appearance of the characteristic bands of the OSi(OtBu)3 (1390 and 1365 cm1) and COD
(1477, 1455 and 1436 cm1) ligands, which are also present in
the corresponding molecular complexes (Fig. 2). Elemental analysis shows an Ir loading of 4.40 wt%, which corresponds to
0.25 mmol Ir g1 or 0.75 Ir nm2. While this value is close to the
amount of initial surface silanols (0.26 mmol g1), a large quantity
of molecular complexes remains unreacted according to the mass
balance: 0.88 equiv. of starting material is recovered after grafting,
corresponding to a grafting yield of 54%. The amount of
HOSi(OtBu)3 released is 0.49 equiv. per initial surface silanols or
ca. 0.93 equiv. of ligand released per reacted complex. These data
are in good agreement with the grafting of the molecular dinuclear
complex via the cleavage of only one of the two IrAOR bonds, and
the data speak for the formation of a dinuclear grafted species
(Scheme 2). Carbon and hydrogen elemental analyses (Table 2)
support this analysis with 16 C per Ir (14 expected) and 24.5 H
per Ir (25.5 expected). We suggest that the large steric hindrance
of the dinuclear structure accounts for non-quantitative grafting:
in fact, the projected surface area of IrSi(OtBu)3 is 1.18 nm2, which
is close to the density of silanol in SiO2-(700), 1 SiOH per 1.25 nm2
(0.8 SiOH nm2). While all silanols could in principle be occupied

(a)

Fig. 2. FTIR spectra of (a) SiO2-(700); (b) IrOSi(OtBu)3; (c) IrOSi(OtBu)3/SiO2-(700); (d)
IrOSi(OtBu)3/SiO2-(700) after static H2 treatment at 300 C and (e) IrOSi(OtBu)3/SiO2-(700)
after ow H2 treatment at 500 C.

according to the aforementioned van der Waals radius, the system

would be very packed and require a perfect non-realistic and ideal
distribution of OH groups [44]. Similar observations including
spectroscopic data (Fig. S5) are found upon grafting of the corresponding Rh complex under the same conditions (1.28 equiv./
SiOH): a grafting yield of 53% (0.86 equiv. of starting material
recovered), the release of ca. 1.1 equiv of HOSi(OtBu)3 per grafted
Rh complex (0.60 equiv. per surface silanol) and a surface density
of 0.28 mmol Rh g1 or 0.84 Rh nm1 (2.81 wt% Rh). Similarly to

(b)
Ir1
O1

Ir2
O2

(c)

O1

IrOSi(OtBu)3

Rh1

Ir1 O2

IrOMe

O1

Rh2
O2

Ir2

RhOSi(OtBu)3

Fig. 1. ORTEP drawing of (a) IrOSi(OtBu)3; (b) IrOMe; (c) RhOSi(OtBu)3. Thermal ellipsoids are shown at the 50% probability, H atom are omitted for clarity.

265

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

R
R

OH
O

O Si

Si
O

O
O O

OH

Si
O

R
grafting, - ROH
R = Si(OtBu)3 or Me

O
OH

O Si

Si

O
O O

OH

OH

Si
O

H2
decomposition

O Si

Si
O

O
O O

Si
O

Scheme 2. SOMC approach for the preparation of Ir and Rh surface species and NPs.

Table 2
M, C, and H wt%, as determined by mass balance and elemental analysis for
MOSi(OtBu)3/SiO2-(700).
M = Ir

wt% M
wt% C
wt% H

M = Rh

Mass balance

EA

Mass balance

EA

4.80
4.20
0.64

4.40
4.36
0.56

2.78
4.54
0.69

2.81
5.04
0.76

Ir, the relatively large projected surface area of RhOSi(OtBu)3

(1.16 nm2) accounts for non-quantitative grafting.
1
H MAS solid-state NMR spectroscopy of IrOSi(OtBu)3/SiO2-(700)
conrms the presence of -OSi(OtBu)3 ligand with an intense peak
at 1.3 ppm (Fig. S6). However, because of the large intensity and
width of the signal, it is not possible to observe any other signal
associated with the COD ligand, even at high spinning rate
(60 kHz). On the other hand, carbon-13 solid-state NMR spectroscopy shows that the coordinated cyclooctadiene moiety (COD) is
conserved as indicated by the broad peak at 58 ppm associated
with the olenic carbon bonded to Ir, while the allylic carbon signal is hidden by the intense peak of the tert-butyl groups at
31 ppm. Finally, the quaternary carbon of the tert-butyl group is
observed at 75 ppm. Overall, the good agreement between 13C
NMR data of the precursor and the surface species suggests similar
structures between molecular and surface species in line with the
proposed surface dinuclear structure (Fig. 3). Similar observations
are obtained for the Rh complex, where tert-butyl and allylic carbons overlap at 31 ppm and the quaternary carbon from OtBu
and vinylic carbons from COD are observed with 3 overlapping
peaks at 73, 75 and 77 ppm (Fig. S7). Note that rhodiumcarbon
coupling are too small (JRhC  15 Hz) to be observed with solidstate NMR.

Fig. 3. (a) 13C NMR spectrum of IrOSi(OtBu)3 in benzene solution (residual solvent
peak is truncated) and (b) 13C CPMAS ssNMR spectrum of IrOSi(OtBu)3/SiO2-(700);
400 MHz, 10 kHz, d1 = 1s, 55751 scans.

For MOMe (M = Ir and Rh), grafting is also not quantitative, as

indicated by the persistence of some unreacted surface silanol
according to IR spectroscopy (Figs. S8 and S9). IR spectroscopy of
IrOMe/SiO2-(700) shows the appearance of methoxy ligand with a
band at 2815 cm1 as well as COD ligand with the presence of 3
bands at 1478, 1454 and 1432 cm1, by correspondence with the
IR signature of IrOMe (Fig. S8). Elemental analysis shows the presence of 5.58 wt% of iridium, corresponding to 0.29 mmol of Ir per
g of silica (0.87 Ir nm2) or 61% grafting yield. The 1H solid-state
NMR spectrum displays 2 broad peaks at 2.5 and 1 ppm, and the
signals from COD and methoxy ligands as well as unreacted silanols overlap (Fig. S10). Once again 13C CPMAS ssNMR spectroscopy
is more informative as shown in Fig. S11 for IrOMe/SiO2-(700): the
COD ligand is observed with broad peaks at 59 and 33 ppm while
the methoxy ligand resonates at 52 ppm. Similarly to the siloxy
precursor, spectroscopy indicates the formation of a supported
iridium dimer during grafting of IrOMe on silica, as conrmed by
carbon and hydrogen elemental analyses with 8.5 C per Ir (8.5
expected) and 14 H per Ir (13.5 expected). The data for RhOMe/
SiO2-(700) are similar: 77% grafting yield (3.79 wt% of rhodium,
1.11 Rh.nm2 or 0.37 mmol of Rh per gram of silica), identical IR
spectra, 2 broad peaks at 2.3 and 0.9 ppm in the 1H ssNMR spectrum associated with cyclooctadiene, tert-butoxy ligands and
unreacted silanols and 3 peaks at 30, 50 and 76 ppm in 13C CPMAS
ssNMR spectra associated with cyclooctadiene and tert-butoxy
ligands. The higher grafting yield is in line with the lower calculated projected surface area than the siloxy surface species
(0.84 nm2 for both Ir and Rh).
3.3. Nanoparticles formation
Addition of 300 mbar of H2 to IrOSi(OtBu)3/SiO2-(700) in a batch
reactor leads to a change of color within a few minutes at room
temperature from orange to black. 2-Methylpropane is observed
in the gas phase, resulting from the decomposition and hydrogenation of the tris(tBuO)siloxy ligand. After heating at 300 C to ensure
complete reduction of the metal, all ligands undergo deep hydrogenolysis to form 2.7 mmol of methane per gram of compound (corresponding to 3.3 wt% C), slightly under the total amount of carbon
determined by elemental analysis before decomposition (4.4 wt%
C). The decomposition of organic moieties is conrmed by FTIR
spectroscopy by a strong decrease of bands intensities in the
m(CAH) region, between 2800 and 3000 cm1. Nevertheless, a
small residual CAH band is observed, in agreement with the persistence of some organic groups on the surface [53], while the reappearance of surface free silanols is shown by the regeneration of
its characteristic band at 3747 cm1. As determined by transmission electron microscopy (TEM), silica-supported iridium nanoparticles with a mean size of 1.9 0.2 nm (Fig. S13) are obtained.
Similar observations are obtained for the decomposition of
RhOSi(OtBu)3/SiO2-(700) under H2, leading to similar NPs size
(1.9 0.2 nm). Particle size distributions were determined
independently using regular TEM and high-angle annular
dark-eld scanning transmission microscopy (HAADF-STEM), leading

266

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

to identical results. Note that microscopy analysis was carried out

after exposure of the samples to air. This can lead to a slight particle
volume increase due to partial oxidation of the particle surface and
hence to a particle size overestimation.
X-ray photoelectron spectroscopy (XPS) was carried out to
identify the metal state after exposure to air and after reduction.
The spectra of Ir NPs formed from IrOSi(OtBu)3/SiO2-(700) display an
Ir 4f7/2 peak at 61.0 and 61.3 eV after exposure to air and after
reduction, respectively, assigned to metallic iridium (Fig. S18).
The peak shifts from a reference Ir 4f7/2 peak for Ir(0) (observed
at 60.9 eV [54]) could be attributed to either a particle size effect
or a charge reference uncertainty. Typically, NPs demonstrate
higher binding energy (BE) of core levels than the corresponding
values for bulk material. The BE shift also varies depending on
the NP size. Additionally, because the charge was corrected to
the adventitious carbon C 1s peak at 284.8 eV, the chemical state
of carbon upon reduction might change, and this might cause
uncertainty in BEs of a few tenths of an eV. Regardless, the
chemical state of the Ir NPs can be unambiguously assigned to
metallic Ir.
Similarly, metallic Ir is observed after exposure to air and after
reduction for Ir NPs from (COD)Ir(acac) (60.8 and 60.7 eV, respectively, Fig. S19) and from [(COD)IrOMe]2 (60.9 and 61.0 eV, respectively, Fig. S20). XPS analysis of Rh NPs obtained from RhOSi(OtBu)3/
SiO2-(700) leads to analogous observations: the Rh 3f5/2 peaks at
307.9 and 307.6 eV (after exposure to air and after reduction,
respectively, Fig. S21) are attributed to Rh(0). The shifts to higher
BE relative to the Rh(0) reference at 307.2 eV [54] is possibly due
to a particle size effect. However, since Rh 3f5/2 peaks of the airexposed Rh NPs appeared at higher BEs than the reduced NPs,
the existence of the surface oxide cannot be ruled out completely.
Overall, M(0) is the major species after exposure to air and after
reduction. This conclusion is based on the measured BEs and asymmetry of the photoemission peaks, which are characteristic for
metals. As it was said above, the presence of a thin oxide layer at
the particles surface cannot be excluded. However, since the full
width at half maximum (FWHM) of the Ir 4f and Rh 3d peaks following reduction does not change, the oxidation is likely negligible. Particle sintering upon reduction would be accompanied by
shifts of the Ir 4f and Rh 3d peaks towards lower BEs; this is not
the case except for the 0.3 eV Rh 3d peak shifts for the Rh NPs.
The XPS data point to the conclusion that the reduction does not
cause sintering, and the particle volume increase due to oxidation
is negligible, and we approximate the particle size of the reduced
samples to the particle size observed by TEM after exposure to
air. Assuming truncated cuboctahedral geometry, the mean
number of metal atoms per particles is 255 and 261 for iridium
and rhodium, respectively. The metal dispersion, as determined
by the ratio of the number of surface atoms to the total number
of atoms in the sample, is 55% for iridium and 57% for rhodium.
It is noteworthy that the dispersion is calculated as the sum of
the surface atoms divided by the sum of the total atoms of all
counted particles, hence it is independent from the mean particle
size. Nevertheless, the particle size distribution can be truncated
in the subnanometric region, as these particles/clusters are hard
to distinguish, which lead to an underestimation of the metal
dispersion.
The use of a ow of H2 also leads to the presence of residual
organic ligands at 300 C as observed by FTIR spectroscopy, but
they totally disappear upon treatment under H2 gas ow at
500 C. The use of these gas ow conditions at 500 C leads to
the formation of smaller particles with a mean diameter of
1.2 0.3 nm (64 atoms, 68% dispersion) and 1.4 0.3 nm
(104 atoms, 66% dispersion) for the Ir and Rh particles, respectively
(Fig. 4).

Decomposition of MOMe/SiO2-(700) under static H2 treatment at

300 C leads to similar results: regeneration of free silanol with
the reappearance of the characteristic band at 3746 cm1 and a
strong decrease of the vibration in the CAH region and formation
of slightly larger NPs (2.8 0.4 nm and 2.6 0.6 nm for Ir and Rh,
respectively). Note that it is critical to carefully wash methanol
released during the grafting step, as residual methanol leads to
the formation of CO, which remains strongly adsorbed at the surface of the nanoparticles (data not shown). In contrast to the siloxy
precursors, treatment under H2 ow at 500 C does not allow the
complete disappearance of CAH contributions in FTIR spectra,
but it still leads to the reduction of the mean particle size to
1.3 0.4 nm and 1.7 0.4 nm for Ir and Rh, respectively. For
comparison, this study is also conducted for iridium and rhodium
particles prepared by a classical impregnation approach using
[(COD)Ir(acac)] and Rh(NO3)3 as precursors. This approach
provided particles of 2.4 1.3 nm and 2.6 0.7 nm for Ir and Rh,
respectively (Figs. S14S16). Note that trichloride molecular
precursors yield larger particles with a broad distribution in our
hands (2.3 0.7 nm and 2.7 2.0 nm for Ir and Rh, respectively,
with some NPs > 4 nm, Fig. S17).
While the standard deviation for NPs size population can testify
the narrowness of the size distribution, assessment of the spatial
uniformity is also an important parameter. Quadrat-based metrics
are usually preferred over distance-based metrics for the study of
particles distribution [55]. In this method, the surface is divided
into identical squares called quadrats and the number of nanoparticles in each quadrat is counted on TEM pictures. To quantify the
spatial uniformity, the index of dispersion (ID) [56] is calculated
according to the formula:

ID q  1s2 =x0
where q is the number of quadrats, x0 and s2 are the mean and sample variance of the number of points per quadrat. A large value of ID
indicates clustering while low values are observed for a complete
spatial randomness (CSR).
Using 15  15 nm quadrat, we determined ID = 15 for the sample prepared from IrOSi(OtBu)3/SiO2-(700), while ID = 78 is obtained
for the sample obtained from decomposition of (COD)Ir(acac),
showing a better spatial uniformity for the sample prepared by
the grafting approach.
Measuring the hydrogen and carbon monoxide chemisorptions
combined with TEM allows hydrogen and carbon monoxide
adsorption stoichiometry on surface metal atoms to be determined. While the samples were exposed to air prior to TEM measurement leading to a negligible particle size increase, the
reduced particles were handled under inert atmosphere until the
adsorption measurement, as even a slight surface oxidation of
the metal surface would have dramatic consequences on the
amount of gas adsorbed. Assuming the absence of particle size
effect and a negligible spill over phenomena on the support surface, silica in the present study, it is possible to gather insights in
the metallic surface state. The number of surface metal atoms is
determined from the calculated size of the reduced nanoparticles
according to TEM results. The adsorption factors (H/Ms and CO/
Ms) discussed below are assigned to the adsorption at saturation
at 25 C obtained from dissociative and non-dissociative Langmuir
equations for H2 and CO, respectively [7]. To model the strongly
and weakly adsorbed H&CO (so called reversible and irreversible
H&CO), two Langmuir equations with high and low adsorption
equilibrium constant were summed and the tting optimized.
The sums of the two maximal adsorption values Qmax are used
for determining the H/Ms and CO/Ms (Ms: surface metal). All samples display a fast adsorption of H2 and CO in the 010 mbar pressure range, suggesting a strong chemisorption. A contrario, weak

267

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

(a)

(c)

IrOSi(OtBu)3

IrOMe

(b)

(d)

RhOSi(OtBu)3

RhOMe

Fig. 4. HAADF-STEM picture and NPs size distribution from H2 ow treatment at 500 C of 4 different precursors; (a) IrOSi(OtBu)3; (b) IrOMe; (c) RhOSi(OtBu)3; (d) RhOMe.

adsorption is observed at higher pressures (10200 mbar), as

shown in Fig. 5 for the H2 adsorption isotherm on Ir NPs prepared
from IrOSi(OtBu)3. This two-step adsorption can be attributed to the
adsorption of H and CO rst to free sites leading to rst l3 species,
that slowly give way to l2 and l1 species with increasing pressure
and coverage [57]. Adsorption stoichiometries and dispersion
from TEM measurements are summarized in Table 3. They are also
presented graphically in Fig. 6 for all catalysts obtained by H2
ow treatment at 500 C. For all measured samples, the CO
adsorption stoichiometry was lower than the hydrogen adsorption
stoichiometry.

Fig. 5. H2 chemisorption isotherm at 25 C for Ir NPs formed from [(COD)IrOSi(OtBu)3]2: experimental points (black squares) and double Langmuir t (black
curve).

For both iridium and rhodium, we observed variations of adsorbate to metal ratios depending on the precursor and the thermal
treatment. In addition, no general trend such as a global decrease
of the adsorption coefcient with decreasing dispersion could be
observed. Hence, results for each species are discussed consecutively. Ratios as high as 2.4 H and 1.2 CO/Irs are obtained for
IrOSi(OtBu)3/SiO2-(700) prepared under H2 ow at 500 C, which attest
for an excellent ability of IrOSi(OtBu)3 to decompose into a clean
metal surface. Similar results are found for RhOSi(OtBu)3/SiO2-(700)
under H2 ow at 500 C for H2 adsorption with 2.5 H/Rhs, while
CO adsorption is much greater reaching 2.4 CO/Rhs. While a lower
dispersion (from TEM) is observed for particles obtained under static conditions at 300 C, the adsorption stoichiometries are similar,
albeit a bit lower, showing that higher temperature does not
greatly inuence the surface state. For IrOMe and RhOMe, H2 (and
CO) stoichiometries are lower than values obtained on nanoparticles prepared from MOSi(OtBu)3 for both metals: 1.5 H/Irs (1.0 CO)
and 1.5 H/Rhs (1.3 CO) for particles obtained after H2 treatment
at 500 C. The values are even lower for the samples treated under
static conditions at 300 C {1.2 H/Irs (0.9 CO) and 1.3 H (1.0 CO)/
Rhs}, and the difference with the high temperature treatment is
proportionally greater than on particles prepared from siloxy
precursors. Those data probably indicate that the surface is
contaminated, probably by the presence of residual carbon or
hydrocarbon species at the surface. It is important to precise that,
while the OSi(OtBu)3 ligand will decompose under H2 to SiO2, H2O
and CH4, OMe ligands will lead to water and CO as an intermediate
product (observed), and such decomposition products could play a
role in the decrease of the H2&CO adsorption properties, as a
polluted surface adsorbs to a lower extent. Note that the sample
obtained from impregnation (COD)Ir(acac) shows an adsorption
stoichiometry intermediate between the values obtained from

268

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

Table 3
Main characteristic of supported metallic NPs for various precursors.

IrOSi(OtBu)3
IrOSi(OtBu)3
IrOMe
IrOMe
(COD)Ir(acac)
RhOSi(OtBu)3
RhOSi(OtBu)3
RhOMe
RhOMe
Rh(NO3)3
Rh(NO3)3

Decomposition method

Metal loading (EA) (%)

NPs diameter standard

deviation (TEM) (nm)

Dispersion (TEM) (%)

H/Ms

CO/Ms

Static, 300 C
Flow, 500 C
Static, 300 C
Flow, 500 C
Flow, 500 C
Static, 300 C
Flow, 500 C
Static, 300 C
Flow, 500 C
Flow, 500 C
Flow, 500 C

4.4
4.4
5.6
5.6
1.0
3.0
3.0
3.8
3.8
1.0
2.7

1.9 0.2
1.2 0.3
2.8 0.4
1.3 0.4
2.4 1.3
1.9 0.2
1.4 0.3
2.6 0.6
1.7 0.4
1.8 0.7
2.6 0.7

55
68
55
62
32
57
66
41
56
45
35

1.9
2.4
1.2
1.5
1.7
2.3
2.5
1.3
1.5
2.5
2.0

1.1
1.2
0.9
1.0
1.5
1.3
2.4
1.0
1.3
2.1
1.6

dispersion
H / Ms
CO / Ms

Fig. 6. H2 and CO adsorption stoichiometry and dispersion for various precursors treated at 500 C under H2 ow.

IrOSi(OtBu)3 and IrOMe. Finally, the Rh NPs obtained by the decomposition of impregnated Rh(NO3)3 show an adsorption stoichiometry
similar to this of RhOSi(OtBu)3 sample (2.5 H/Rhs, 1 wt% Rh) with the
same particle size (1.4 nm). However, it dramatically drops to 2.0
H/Rhs when the average particle size and metal loading increase
to 2.6 0.7 nm and 2.7 wt% Rh, a metal loading in the range of
what is obtained via the SOMC route. The metal loading dependence of H2 adsorption stoichiometry [11,58,59] coupled with
the direct relation between metal loading and mean NPs size can
at least partially explain the decrease of H&CO to Ms ratios
observed. Overall, these results show that direct determination of
highly dispersed supported nanoparticle size from chemisorption
measurement is not straightforward. The measured stoichiometries are greater than in single metal crystals where stoichiometries close to unity are admitted [57] due to the high proportion
of low coordination number atoms for supported nanoparticles.

4. Conclusions
Grafting of dinuclear rhodium(I) and iridium(I) cyclooctadiene
molecular complexes bearing one alkoxide ligand [(COD)MX]2
(X = OMe or OSi(OtBu)3) on partially dehydroxylated silica at
700 C yields the corresponding well-dened dinuclear surface
complexes, [(COD)M(X)(SiO)M(COD)], where one of the X anionic
ligand is replaced by a surface siloxy ligands. In sharp contrast
the chloride derivatives did not react with surface silanols. Grafting
provides silica nanoparticles with high density of Rh and Ir (0.75
1.11 nm2), the higher density being attained with the smaller
OMe ligand. Treatment under H2 at 300 or 500 C of these surface

complexes leads to the formation of small and narrowly distributed metallic nanoparticles, in the 12 nm range, which are randomly and densely dispersed at the surface of the silica
nanoparticles (up to 8000 NPs lm2), the smaller nanoparticles
being obtained at 500 C under gas ow conditions. Note that
obtaining a high density of supported metal nanoparticles by incipient wetness impregnation is more difcult, showing the advantage of the surface organometallic approach. H2 and CO
chemisorption also shows that the ligands of the metal molecular
precursors have a great importance on the nal surface states. Rh
and Ir can adsorb 2.5 and 2.4 H per surface metal atoms, respectively, when using the surface complex with X = OSi(OtBu)3. In
contrast, other molecular precursors lead to particle with much
smaller adsorption capacities (down to ca. 1.2 H per surface metal
atoms) despite similar particle size, thus illustrating the importance of the preparation methods and the structure of the molecular precursors. Similar behaviors are obtained for the adsorption of
CO. This study also illustrates how difcult it is and how cautious
one should be to evaluate the size of particle solely based on
chemisorption. We are currently further exploring the generality
of surface organometallic chemistry towards the preparation of
supported nanoparticles and measuring the intrinsic chemisorption of supported metallic nanoparticles.

Acknowledgments
The authors thank F. Krumeich, E. Oatkon and G. Siddiqi for TEM
measurements, G. Lapadula, V. Mougel and M. Wrle for single
crystal XRD measurement and structure determination, R. Vrel

F. Hroguel et al. / Journal of Catalysis 316 (2014) 260269

for assistance with NMR and A. Comas Vives for guidance with DFT
calculations. Prof. Fabio Ribeiro is acknowledged for fruitful
discussions.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2014.05.023.
References
[1] K.P. De Jong, Synthesis of Solid Catalysts, Wiley-VCH Verlag GmbH & Co. KGaA,
2009.
[2] S. Nassreddine, G. Bergeret, B. Jouguet, C. Geantet, L. Piccolo, Phys. Chem.
Chem. Phys. 12 (2010) 78127820.
[3] F. Locatelli, B. Didillon, D. Uzio, G. Niccolai, J.P. Candy, J.M. Basset, J. Catal. 193
(2000) 154160.
[4] F. Locatelli, J.P. Candy, B. Didillon, G.P. Niccolai, D. Uzio, J.M. Basset, J. Am.
Chem. Soc. 123 (2001) 16581663.
[5] J.P. Candy, A. Elmansour, O.A. Ferretti, G. Mabilon, J.P. Bournonville, J.M. Basset,
G. Martino, J. Catal. 112 (1988) 201209.
[6] R. Berthoud, P. Delichere, D. Gajan, W. Lukens, K. Pelzer, J.M. Basset, J.P. Candy,
C. Copret, J. Catal. 260 (2008) 387391.
[7] J.E. Low, A. Foelske-Schmitz, F. Krumeich, M. Wrle, D. Baudouin, F. Rascn, C.
Copret, Dalton Trans. 42 (2013) 1262012625.
[8] D. Baudouin, U. Rodemerck, F. Krumeich, A. de Mallmann, K.C. Szeto, H.
Menard, L. Veyre, J.P. Candy, P.B. Webb, C. Thieuleux, C. Coperet, J. Catal. 297
(2013) 2734.
[9] C.H. Bartholomew, R.B. Pannell, J. Catal. 65 (1980) 390401.
[10] P. Laurent, L. Veyre, C. Thieuleux, S. Donet, C. Coperet, Dalton Trans. 42 (2013)
238248.
[11] B.J. Kip, F.B.M. Duivenvoorden, D.C. Koningsberger, R. Prins, J. Catal. 105 (1987)
2638.
[12] J.H. Sinfelt, D.J.C. Yates, J. Catal. 8 (1967). pp. 82.
[13] B.J. Kip, F.B.M. Duivenvoorden, D.C. Koningsberger, R. Prins, J. Am. Chem. Soc.
108 (1986) 56335634.
[14] D.G.H. Ballard, Adv. Catal. 23 (1973) 263.
[15] Y.I. Yermakov, B.N. Kuznetsov, V.A. Zakharov, Studies in Surface Science and
Catalysis, Elsevier, Amsterdam, 1981.
[16] T.J. Marks, Acc. Chem. Res. 25 (1992) 5765.
[17] J.M. Thomas, R. Raja, D.W. Lewis, Angew. Chem. Int. Ed. 44 (2005) 64566482.
[18] M. Tada, Y. Iwasawa, Coord. Chem. Rev. 251 (2007) 27022716.
[19] J.M. Basset, J.P. Candy, C. Copret, Surface organometallic chemistry, in: R.H.
Crabtree, D.M.P. Mingos (Eds.), Comprehensive Organometallic Chemistry III,
Elsevier, Oxford, 2007, pp. 499553.
[20] C. Copret, M. Chabanas, R. Petroff Saint-Arroman, J.M. Basset, Angew. Chem.
Int. Ed. 42 (2003) 156181.
[21] S. Hermans, Clusters and immobilization, in: K.P.d. Jong (Ed.), Synthesis of
Solid Catalysts, WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim, 2009.
[22] C. Coperet, Pure Appl. Chem. 81 (2009) 585596.
[23] F. Rascon, R. Wischert, C. Coperet, Chem. Sci. 2 (2011) 14491456.
[24] D. Gajan, K. Guillois, P. Delichere, J.M. Basset, J.P. Candy, V. Caps, C. Copret, A.
Lesage, L. Emsley, J. Am. Chem. Soc. 131 (2009) 1466714668.
[25] D.A. Ruddy, J. Jarupatrakorn, R.M. Rioux, J.T. Miller, M.J. McMurdo, J.L. Mcbee,
K.A. Tupper, T.D. Tilley, Chem. Mater. 20 (2008) 65176527.

269

[26] M.L.M. Bonati, T.M. Douglas, S. Gaemers, N. Guo, Organometallics 31 (2012)

52435251.
[27] H.C. Foley, S.J. DeCanio, K.D. Tau, K.J. Chao, J.H. Onuferko, C. Dybowski, B.C.
Gates, J. Am. Chem. Soc. 105 (1983) 30743082.
[28] P. Dufour, C. Houtman, C.C. Santini, J.M. Basset, J. Mol. Catal. 77 (1992) 257
272.
[29] R.J. Trovitch, N. Guo, M.T. Janicke, H.B. Li, C.L. Marshall, J.T. Miller, A.P.
Sattelberger, K.D. John, R.T. Baker, Inorg. Chem. 49 (2010) 22472258.
[30] J.E. Mondloch, Q. Wang, A.I. Frenkel, R.G. Finke, J. Am. Chem. Soc. 132 (2010)
97019714.
[31] J.E. Mondloch, R.G. Finke, J. Am. Chem. Soc. 133 (2011) 77447756.
[32] J.E. Mondloch, R.G. Finke, Abstr. Pap. Am. Chem. Soc. 242 (2011).
[33] J. Guzman, B.C. Gates, Dalton Trans. (2003) 33033318.
[34] A. Kulkarni, R.J. Lobo-Lapidus, B.C. Gates, Chem. Commun. (Cambridge, U.K.)
46 (2010) 59976015.
[35] J.L. Herde, J.C. Lambert, C.V. Senoff, Inorg. Synth. 15 (1974).
[36] G. Giordano, R.H. Crabtree, Inorg. Synth. 28 (1990) 8890.
[37] R. Uson, L.A. Oro, J.A. Cabeza, H.E. Bryndza, M.P. Stepro, Inorg. Synth. 23 (1985)
126127.
[38] J. Jarupatrakorn, T.D. Tilley, Dalton Trans. (2004) 28082813.
[39] M.J. Frisch, Gaussian (Revision B.05, see ESI for the full list of authors), 2003.
[40] A.D. Becke, J. Chem. Phys. 98 (1993) 56485652.
[41] C.T. Lee, W.T. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785789.
[42] P.J. Stephens, F.J. Devlin, C.F. Chabalowski, M.J. Frisch, J. Phys. Chem. 98 (1994)
1162311627.
[43] J.F. Moulder, J. Chastain, R.C. King, Handbook of X-ray Photoelectron
Spectroscopy: A Reference Book of Standard Spectra for Identication and
Interpretation of XPS Data, Physical Electronics Eden Prairie, MN, 1995.
[44] L.T. Zhuravlev, Colloid Surf. A 173 (2000) 138.
[45] A.K. Mcmullen, T.D. Tilley, A.L. Rheingold, S.J. Geib, Inorg. Chem. 28 (1989)
37723774.
[46] I. Tanaka, N. Jinno, T. Kushida, N. Tsutsui, T. Ashida, H. Suzuki, H. Sakurai, Y.
Morooka, T. Ikawa, Bull. Chem. Soc. Jpn. 56 (1983) 657661.
[47] A. Viziorosz, R. Ugo, R. Psaro, A. Sironi, M. Moret, C. Zucchi, F. Ghel, G. Palyi,
Inorg. Chem. 33 (1994) 46004603.
[48] P. Krzyzanowski, M. Kubicki, B. Marciniec, Polyhedron 15 (1996) 14.
[49] B. Marciniec, I. Kownacki, M. Kubicki, Organometallics 21 (2002) 3263
3270.
[50] E. Samuel, J.F. Harrod, M.J. Mcglinchey, C. Cabestaing, F. Robert, Inorg. Chem.
33 (1994) 12921296.
[51] D.J.A. Deridder, P. Imhoff, Acta Crystallogr. Sect. C-Cryst. Struct. Commun. 50
(1994) 15691572.
[52] F.A. Cotton, P. Lahuerta, M. Sanau, W. Schwotzer, Inorg. Chim. Acta 120 (1986)
153157.
[53] The Latter Residual CAH Bands are not Observed After H2 Batch Treatment at
500 C.
[54] Handbook of X-ray Photoelectron Spectroscopy: A Reference Book of Standard
Spectra for Identication and Interpretation of XPS data, eds. Physical
Electronics Division, Perkin-Elmer Corporation, 1995.
[55] K.M. Kam, L. Zeng, Q. Zhou, R. Tran, J. Yang, J. Manuf. Syst. 32 (2013) 154
166.
[56] While those values clearly show a better spatial distribution for the grafting
route, note that the number of NPs per quadrat strongly depends on the
number of over layered silica particles in the TEM images, which could lead to
a slight overestimation.
[57] G.C. Bond, Metal-Catalysed Reactions of Hydrocarbons, Springer, 2005.
[58] B.J. Kip, J. Vangrondelle, J.H.A. Martens, R. Prins, Appl. Catal. 26 (1986) 353
373.
[59] O. Alexeev, B.C. Gates, J. Catal. 176 (1998) 310320.