Ligand ordering determines the catalytic response of hybrid palladium nanoparticles in hydrogenation

Article abstract of DOI:10.1039/c5cy01921d

Supported palladium nanoparticles, prepared by reducing the active metal in the presence of the hexadecyl(2-hydroxyethyl)dimethylammonium dihydrogen-phosphate (HHDMA) ligand and depositing the resulting colloids on titanium silicate (TiSi₂O₆), represent a proven alternative to the archetypal poisoned catalysts in industrially-relevant selective hydrogenations. To date, a key aspect in the design of these hybrid nanocatalysts remains unaddressed, namely the impact of the ligand content on the catalytic behaviour. In order to assess the structural and associated catalytic implications of this variable, we have prepared a series of Pd-HHDMA/TiSi₂O₆ catalysts with different HHDMA content (0.3-16.8 wt%), keeping the average particle size (5 nm) and Pd content (0.3 wt%) constant. The materials are characterised with a toolbox of methods, including advanced microscopy and solid-state nuclear magnetic resonance, in order to assess the structure metal-ligand interface and the mobility of the alkyl chain. Continuous-flow three-phase hydrogenations of short-chain acetylenic compounds, nitriles, and carbonyls reveal an increase in the catalytic activity with the ligand content. Density Functional Theory indicates that the ligand behaves as a self-assembled monolayer, changing its adsorption configuration as a function of the HHDMA concentration. At low coverage, the organic layer lies almost flat on the surface of the metal nanoparticle blocking a large number of metal sites and resembling a two-dimensional catalyst; high HHDMA coverages favour an extended three-dimensional configuration of the alkyl chain, and consequently a lower fraction of Pd sites are poisoned. These results provide new fundamental insights into the role of the ligand on the catalytic activity and can enable the design of hybrid nanocatalysts with optimised performance.

Full text of DOI:10.1039/c5cy01921d

Catalysis
Science &
Technology
View Article Online
View Journal
PAPER
Ligand ordering determines the catalytic response
of hybrid palladium nanoparticles in
hydrogenation†
Cite this: DOI: 10.1039/c5cy01921d
Davide Albani,‡a Gianvito Vilé,‡a Sharon Mitchell,a Peter T. Witte,b
c
a
c
a
*
*
Neyvis Almora-Barrios, René Verel, Núria López and Javier Pérez-Ramírez
Supported palladium nanoparticles, prepared by reducing the active metal in the presence of the
hexadecylIJ2-hydroxyethyl)dimethylammonium dihydrogen-phosphate (HHDMA) ligand and depositing the
resulting colloids on titanium silicate (TiSi2O6), represent a proven alternative to the archetypal poisoned
catalysts in industrially-relevant selective hydrogenations. To date, a key aspect in the design of these
hybrid nanocatalysts remains unaddressed, namely the impact of the ligand content on the catalytic
behaviour. In order to assess the structural and associated catalytic implications of this variable, we have
prepared a series of Pd-HHDMA/TiSi2O6 catalysts with different HHDMA content (0.3–16.8 wt%), keeping
the average particle size (5 nm) and Pd content (0.3 wt%) constant. The materials are characterised with a
toolbox of methods, including advanced microscopy and solid-state nuclear magnetic resonance, in order
to assess the structure metal–ligand interface and the mobility of the alkyl chain. Continuous-flow three-
phase hydrogenations of short-chain acetylenic compounds, nitriles, and carbonyls reveal an increase in
the catalytic activity with the ligand content. Density Functional Theory indicates that the ligand behaves as
a self-assembled monolayer, changing its adsorption configuration as a function of the HHDMA concentra-
tion. At low coverage, the organic layer lies almost flat on the surface of the metal nanoparticle blocking a
large number of metal sites and resembling a two-dimensional catalyst; high HHDMA coverages favour an
extended three-dimensional configuration of the alkyl chain, and consequently a lower fraction of Pd sites
are poisoned. These results provide new fundamental insights into the role of the ligand on the catalytic
activity and can enable the design of hybrid nanocatalysts with optimised performance.
Received 9th November 2015,
Accepted 13th December 2015
DOI: 10.1039/c5cy01921d
www.rsc.org/catalysis
deposited on a suitable support. Under the appropriate condi-
tions, it is possible to obtain metal crystallites of variable par-
Introduction
Nanostructured materials combining both inorganic and or-
ganic functionalities find increasing applications as heteroge-
neous catalysts1,2 due to the unique properties that result
from complementary interactions at the inorganic–organic
interface.3–6 These materials are typically prepared by liquid-
phase reduction–deposition techniques, in which the metal
precursor is first reduced in the presence of an organic ligand
to form colloidal metal nanoparticles, that are subsequently
ticle size and shape.3–6 However, these synthetic methods
have crucial disadvantages (i.e., the need for low-boiling point
organic solvents, the fast addition of expensive or toxic
reductants) that have long hampered their industrial exploi-
tation.8 The identification of hexadecyl-2-hydroxyethyl-
dimethyl ammonium dihydrogen phosphate (HHDMA) as a
ligand that combines both reducing and stabilising func-
tionalities in a single, water-soluble molecule, has recently
enabled the synthesis at industrial level of colloidally-
prepared Pd, Pt, and Pd–Pt nanoparticles supported on inert
carriers such as titanium silicate (TiSi2O6) and activated car-
bon.7,8 The versatility of this new generation of catalysts
has been demonstrated for the selective hydrogenation of
functionalised alkynes and nitroaromatics,9–11 primary reac-
tions in the fine chemical and pharmaceutical industries
for the manufacture of polymers, vitamins, agrochemicals,
and fragrances.12
a Institute for Chemical and Bioengineering, Department of Chemistry and Applied
Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, 8093 Zurich, Switzerland.
E-mail: jpr@chem.ethz.ch
b BASF Nederland BV, Strijkviertel 67, 3454PK De Meern, The Netherlands
c Institute of Chemical Research of Catalonia (ICIQ) and Barcelona Institute of
Technology (BIST), Av. Països Catalans 16, 43007 Tarragona, Spain.
E-mail: nlopez@iciq.es
† Electronic supplementary information (ESI) available. The material contains
additional characterisation and catalytic data and theoretical results. See DOI:
10.1039/c5cy01921d
Kinetic tests in flow mode, in-depth characterisation, and
Density Functional Theory (DFT) calculations have
‡ These authors contributed equally to the work.
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.

View Article Online
Paper
Catalysis Science & Technology
significantly contributed to the structural understanding of
these materials, showing that the ligand shell is a main
player in determining their efficient and selective catalytic be-
haviour. Despite the pioneering work of Schwartz and
co-workers13 on the effect of the ligand concentration and
performance of thiol-based materials, those results cannot be
extrapolated to much more complex ligands composed of
multiple chemical units. In fact, a literature analysis concluded
that these points have not yet been addressed. Our contribu-
tion accomplishes all the challenges faced in the controlled
synthesis of Pd-HHDMA/TiSi2O6 catalysts with variable
ligand content, but equivalent metal concentration (0.3 wt%)
and average particle size (5 nm), in the characterisation
of the ligand organisation and in the rationalisation of
its catalytic impact of these tasks, demonstrating the critical
and nontrivial impact of ligand ordering. Catalytic tests were
conducted in a continuous-flow three-phase micro-reactor,
showing an increased catalytic activity with increasing ligand
concentration for the hydrogenation of alkynes, nitriles, and
carbonyls. DFT calculations and electron microscopy provide
fundamental insights into this counterintuitive structure–
performance relationship, which is explained based on the
relative assembly of ligands on the nanoparticle surface.
and stirred for 1 h. The mother liquor containing excess col-
loidal nanoparticles from the preparation of Pd-HHDMA-4
was divided into two equal portions and one of these por-
tions was added to 27 g of TiSi2O6 and stirred for 1 h
obtaining Pd-HHDMA-1. For the preparation of Pd-HHDMA-2,
titanium silicate (10 g) was added to 300 cm3 of colloidal sus-
pension and stirred for 1 h. The resulting mother liquor was
divided into two equal portions; titanium silicate (20 g) was
added to one portion, and the resulting suspension was
stirred for 1 h, obtaining Pd-HHDMA-2. Prior to use, the cata-
lysts were separated by filtration and extensively washed with
deionised water.
Catalyst characterisation
The palladium content in the supported catalysts was deter-
mined by inductively coupled plasma-optical emission
spectrometry (ICP-OES) using a Horiba Ultra 2 instrument
equipped with photomultiplier tube detection. The C, H, N,
and P contents were determined by infrared spectroscopy
using a LECO CHN-900 combustion furnace. Nitrogen iso-
therms were measured at 77 K in a Micrometrics ASAP 2020
instrument, after evacuation of the samples at 393 K for 6 h.
Transmission electron microscopy (TEM) was undertaken
using an FEI Talos S200 microscope operating at 200 kV with
a 70 μm C2 aperture, 0.4 nA beam current. The colloidal sam-
ples were supported on continuous carbon-coated Cu grids
and negatively stained with a 1 vol% aqueous uranyl acetate
solution in order to increase the contrast between the ligand
tail and the background. Images were recorded in low-dose
mode to minimise radiation damage to the sample. The
supported catalysts were dispersed as dry powders on lacey
carbon coated copper grids. Scanning electron microscopy
(SEM) was undertaken in an FEI Magellan 400 microscope
working at 30 kV and 50 pA. The particle size distribution
was assessed by counting more than 150 individual Pd nano-
particles. CO chemisorption was performed using a Thermo
TPDRO 1100 unit. The samples were pre-treated in situ at 393
K under He flow (20 cm3 min−1) for 60 min, and reduced at
348 K under flowing 5 vol% H2/He (20 cm3 min−1) for 30
min. Thus, 0.344 cm3 of 1 vol% CO/He was pulsed over the
catalyst bed at 308 K every 4 min. The palladium dispersion
was calculated from the amount of chemisorbed CO, consid-
ering an atomic surface density of 1.26 × 1019 atoms m−2 and
Experimental
Catalyst preparation
Hybrid palladium nanocatalysts with different ligand content
(0.3–16.8 wt%) were synthesised by reduction–deposition,
using HHDMA as the stabiliser. The materials are coded as
Pd-HHDMA-n, where n = 1–4 and represents the relative
ligand concentration (see Table 1). As depicted in Fig. 1, an
aqueous solution of HHDMA (100 cm3, 30%) was diluted in
deionised water (3000 cm3) and mixed with an aqueous solu-
tion of Na2PdCl4 (0.1 M, 50 cm3). The pH of the mixture was
adjusted to 5 by the dropwise addition of NaOH (12 M). The
resulting solution, which contains 0.38 mg Pd cm−3, was
heated to 358 K and stirred at this temperature for 2 h to
form the colloids, obtaining metal nanoparticles of approxi-
mately 5 nm. Note that, the metal particle size is solely con-
trolled by the addition of HHDMA in the initial solution. For
the preparation of Pd-HHDMA-3 and Pd-HHDMA-4, titanium
silicate (30 or 40 g, respectively) was added to a defined
amount of colloidal suspension (200 or 600 cm3, respectively)
Table 1 Characterisation data of the catalysts
Pda
(wt%)
Ca
(wt%)
Ha
(wt%)
HHDMAa C/Pdb Weight lossc SBET
Vpore
DCO
(%)
DSEM
(%)
d
e
f
g
Catalyst
Na (wt%)
Pa (wt%)
(wt%)
(−)
(%)
(m2 g−1
)
(cm3 g−1
)
Pd-HHDMA-1 0.3(0.3)h 0.2(0.3) 1.6(1.4) 0.01(0.03) 0.02(0.1)
Pd-HHDMA-2 0.3
Pd-HHDMA-3 0.3
Pd-HHDMA-4 0.3(0.3)
0.3
4.3
8.2
1
9
15
31
4
8
10
17
426
330
273
105
0.30
0.21
0.20
0.07
23
22
13
11
24
23
22
24
2.5
4.8
1.8
2.1
0.13
0.23
0.02
0.03
9.8(9.4) 2.7(2.7) 0.45(0.53) 0.06(0.12) 16.8
a
g
b
c
d
e
f
Elemental analysis. Weight ratio. Thermogravimetric analysis. BET method. Volume of N2 at p/p0 = 0.95. CO chemisorption.
h
Scanning electron microscopy. Elemental analysis of the catalyst after hydrogenation. Conditions: Wcat = 0.2 g, T = 343 K, P = 10 bar,
FL (benzaldehyde + methanol) = 1 cm3 min1, and FG(H2) = 48 cm3 min−1
.
Catal. Sci. Technol.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Catalysis Science & Technology
Paper
Fig. 1 Schematic illustration of the synthetic steps involved in the preparation of the supported catalysts.
an adsorption stoichiometry of Pd/CO = 2. The stoichiometry
is the most suited for chemisorption measurements
conducted over palladium catalysts at room temperature and
in dynamic mode.14 X-ray photoelectron spectroscopy (XPS)
was measured in a Physical Electronics Instruments Quan-
tum 2000 spectrometer using monochromatic Al Kα radiation
generated from an electron beam operated at 15 kV and 32.3 W.
The spectra were collected under ultra-high vacuum condi-
tions (residual pressure = 5 × 10−8 Pa) at a pass energy
of 50 eV. All binding energies were referenced to that of C
1s at 288.2 eV in order to compensate for charging effects.
Thermogravimetric analysis was performed in a Mettler
Toledo TGA/DSC 1 Star system. Prior to measurement, the
samples were dried in N2 (40 cm3 min−1) at 393 K for 1 h.
The analysis was performed in air (40 cm3 min−1), heating
through a fixed-bed reactor (3.5 mm i.d.) with catalyst parti-
cles of 0.2–0.4 mm. For the hydrogenation of alkynes, the
catalyst (0.10 g) was diluted with titanium silicate (0.12 g,
Aldrich, 99.8%). For the hydrogenation of nitriles and
carbonyls, the undiluted catalyst (0.2 g) was used. The liquid
feed contained 5 vol% of substrate and toluene (Fischer
Chemicals, 99.95%), cyclohexane (Sigma-Aldrich, 99%), or
methanol (Sigma-Aldrich, 99%) as solvents. The reactions
were conducted at various temperatures (303–363 K), total
pressures (1–40 bar), and liquid (0.3–1 cm3 min−1) and H2
(36–48 cm3 min−1) flow rates. The reaction products were col-
lected employing a liquid handler (Gilson GX-271), after
10–15 min of steady-state operation, and analysed offline
using a gas chromatograph (HP-6890) equipped with a HP-5
capillary column and a flame ionisation detector. The hydro-
genation of 3-hexyn-1-ol was conducted in batch mode using
a 250 cm3 stainless-steel autoclave. The reactor was charged
with catalyst (0.4 g) and 100 g of a solution of 3-hexyn-1-ol in
ethanol (4.5 wt%). The autoclave was flushed with
hydrogen and pressurised with 3 bar of hydrogen. The con-
version (X) of a given substrate was determined as the
amount of reacted substrate divided by the amount of sub-
strate at the reactor inlet. The selectivity (S) to each product
was quantified as the amount of the particular product
divided by the amount of reacted substrate. The reaction rate
(r) was expressed as the number of moles of product per
mole of Pd and unit time.
the sample from 298 K to 1173 K at a rate of 5 K min−1 13C,
.
1H, and 31P magic angle spinning nuclear magnetic reso-
nance spectra were recorded on a Bruker AVANCE III HD
NMR spectrometer at a magnetic field of 16.4 T correspond-
1
ing to a H Larmor frequency of 700.13 MHz. A 4 mm double
resonance probe head at a spinning speed of 10 kHz was
used for all experiments. The 13C spectra were acquired using
a cross polarization experiment with a contact time of 2 ms
and a recycle delay of 1 s. A total of 64 × 103 scans were
added for each sample. The 31P experiments used a single
pulse excitation sequence with a recycle delay of 1 s. Between
39 × 103 and 96 × 103 scans were acquired depending on the
sample. Both 13C and 31P experiments used high-power 1H
decoupling during acquisition using a SPINAL-64 sequence.15
Computational details
Density Functional Theory (DFT) calculations were performed
using the Vienna ab initio Simulation Package (VASP),16–18
employing the generalised gradient approximation with the
revised Perdew–Burke–Ernzerhof exchange–correlation func-
tional.19 The interaction between the valence electrons and
the core was described with the projected augmented wave
Catalyst evaluation
The hydrogenation of 1-hexyne (Acros Organics, 98%),
2-methyl-3-butyn-2-ol (Acros Organics, 99.9%), 3-hexyne (Acros
Organics, 99%), 4-octyne (Acros Organics, 99%), benzonitrile
(TCI, 99%), and benzaldehyde (Fluka, 99%) was carried out
in a fully-automated flooded-bed micro-reactor (ThalesNano
H-Cube Pro™), in which the gaseous hydrogen produced in
situ and the liquid reactant flow concurrently upward
method in the implementation of Kresse and Joubert.18
A
kinetic cut-off energy of 450 eV was used for the plane wave
expansion of the wave functions. Due to the size of the
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.

View Article Online
Paper
Catalysis Science & Technology
simulation cell (the shortest cell vector is ca. 28 Å), during
ion relaxations, the reciprocal space integration was
performed by considering only the Γ-point of the Brillouin
zone. The van der Waals contributions were described else-
where, using the semi-empirical DFT-D2 (ref. 20) approach
and the parameters optimised for this metal.21 The palla-
dium (111) surface was simulated using periodic slabs with a
thickness of five layers, which were separated from the neigh-
bour by a vacuum gap of 20 Å to provide sufficient space for
the adsorbates, avoiding in this way potentials interactions
with neighbouring (periodic image) slabs. In a large p(10 ×
10) unit cell, 5, 10, 18 and 25 HHDMA chains were adsorbed
on the surface, obtaining coverages of 0.5, 0.10, 0.18, 0.25
monolayers (ML). In particular, the structures with lower
HHDMA content were obtained by using the fully saturated
surface as the initial step, and removing a certain fraction of
ligand. Prior to performing DFT-D2 calculation, the different
configurations were equilibrated using the classic molecular
dynamics (MD) simulations. The MD simulation allows the
HHDMA chains to relax fast and occupy the empty volume of
the missing ligands. The DFT-D2 methodology included all
the required interactions: electrostatic (between the ions in
the ligand and the metal), covalent (anion–metal), and van
der Waals (between the metal and the ligand, and between
the ligand tails). Since the ligand interacts both with the
metal surface and with the organic tail of neighbouring li-
gands, the dominant interactions depend on the coverage. At
low concentrations, van der Waals forces between the surface
and the aliphatic chains are prevalent. As the ligand concen-
tration increases, the electrostatic interaction between the
phosphate ion of the ligand and the metal dominates over
both the metal–alkyl chain interaction (−0.05 eV per CH2
group),22 and that between the alkyl chains and the solvent.
The optimised layers are not completely ordered and, thus,
an average measure of their orientation with respect to the
surface is required. This dynamic behaviour of the carbon
atoms tail of the HHDMA molecules can be characterised by
an order parameter, SHHDMA, which measures the average ori-
entational order of each methylene chain segment within the
hydrocarbon tail (details in the ESI†). The order parameter is
defined as SHHDMA = 0.5 <3 cos θi cos θj–δij>, where θi (θj)
is the angle formed between the ith (jth) molecular axis
and the normal to the surface and δij is the Kronecker
delta function.23 The order parameter is finally averaged
over the number of ligand molecules in the simulation box.
To reduce the calculation time, we have only averaged the lat-
ter over the chains resulting from the DFT-D2 optimised
structure. To evaluate the influence of the ligand content on
the activity and selectivity in the hydrogenation of alkynes,
we have calculated the adsorption energy, Eads, of specific re-
actant, product, and solvent (i.e., 1-hexyne, 1-hexene and tol-
uene) on the Pd-(HHDMA)x materials. Adsorption energies
the same system in the absence of the adsorbate; and
Emolecule is the energy of the isolated gas-phase molecule.
Results and discussion
Catalyst characterisation
To assess the effects exerted by the HHDMA ligand on the
catalytic performance, four catalysts have been synthesised
with equivalent Pd loading (0.3 wt%), but different ligand
content. Table 1 shows the bulk composition and morpholog-
ical characteristics of the resulting materials. Elemental anal-
ysis of C, H, N, and P confirms the increasing ligand content
of the catalysts from Pd-HHDMA-1 to Pd-HHDMA-4.
were found using the equation: Eads = EPd-(HHDMA)x-molecule
−
(EPd-(HHDMA)x + Emolecule), where EPd-(HHDMA)x-molecule is the en-
ergy of the complete system including the metal, the HHDMA
molecules, and the adsorbate; EPd-(HHDMA)x is the energy of
Fig. 2 Scanning electron micrographs and particle size distributions of
Pd-HHDMA-1 (a), Pd-HHDMA-2 (b), Pd-HHDMA-3 (c), Pd-HHDMA-4
(d). The scale bar in (a) applies to all micrographs.
Catal. Sci. Technol.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Catalysis Science & Technology
Paper
Inspection of the scanning electron micrographs reveals that
the palladium nanoparticles possess a uniform spherical
shape (Fig. 2). Estimation by image analysis evidences an av-
erage particle diameter of around 5 nm (Fig. 2). The particles
are unevenly distributed over the surface of the support,
appearing to be self-assembled into islands in certain
regions. However no evidence of particle agglomeration is
observed despite the high concentration of Pd in the sam-
ples. This can be related to the function of the ligand shell in
separating the metal phase during synthesis. XPS spectra
show in all cases a single Pd 3d5/2 core level peak at 335.3 eV
assigned to Pd0, confirming that the palladium surface is
fully metallic, independent of the amount of HHDMA avail-
able and corresponding reducing capability (Fig. S1†).
To determine the thermal stability of the ligand shell,
thermogravimetric analyses in air were conducted (Fig. 3). In
all cases, the ligand decomposition started at 500–550 K, and
the relative weight loss was consistent with the ligand con-
tent derived from C, H, and N analysis (Table 1). The impact
of the ligand content on the metal dispersion was examined
by CO chemisorption a technique that has been also applied
in the past to probe the number of free Pd sites available
for catalysis in Pd-HHDMA/TiSi2O6, Pd-HHDMA/C, and
Pt-HHDMA/C.9,10 Consistent with the relative amount of or-
ganic moiety, the maximum degree of dispersion (23%) is
evidenced in Pd-HHDMA-1 and decreases to 11% in
Pd-HHDMA-4. It should be noted that this trend cannot be
predicted based on the examination of the metal size distri-
bution alone, from which negligible differences are evidenced
between samples. Similarly, the sample exhibiting the
highest ligand content (Pd-HHDMA-4) displays a surface area
(105 m2 g−1), four times smaller than that of the catalyst with
the lowest one (426 m2 g−1, Table 1). This is attributed to the
presence of HHDMA ligand on both the metal nanoparticle
and surface of the titanium silicate support. We have tried to
determine how the ligand distributes between the metal and
the carrier, employing thermogravimetric analysis (based on
the possible different decomposition patterns for HHDMA at-
tached to the support and to the metal), and nuclear mag-
netic resonance spectroscopy (based on the local environ-
ment experienced by the ligand on the metal and on the
support). In both case, we have not been able to conclusively
quantify the amount of ligand per nanoparticle because of
the absence of differences in the patterns. Besides, the C : P
and C : N values could not be used to obtain a preliminary
estimation of the separation of the ligand. In fact, by com-
paring the atomic C : P and C : N ratios (i.e., 422 and 25, re-
spectively, for Pd-HHDMA-4) with the theoretical values
based on the HHDMA stoichiometry (C : P = 20 and C : N =
20), a discrepancy is found. A published paper8 hypothesised
that this can be attributed to the fact that HHDMA
ligand binds on the TiSi2O6 by the amino group, releasing
orthophosphoric acid. This hypothesis, however, still needs
to be supported experimentally. In this context, our prelimi-
nary results indicate that the adsorption configuration of the
ligand on the carrier is much more complex, since we have
detected P-containing species by elemental mapping in areas
where metal nanoparticles are not present. For this reason,
we believe that HHDMA adsorbs on the support in different
configurations, with and without the orthophosphate group. As a
consequence, all the correlations derived in this manuscript
have always been referred to the total ligand content in the
catalyst. The samples were further characterised by solid-
state 13C and 31P MAS-NMR. The peaks in the 13C MAS NMR
spectra (Fig. 4a) can be assigned in analogy with the signals
observed in the solution state NMR spectra. In particular, the
peaks above 40 ppm all originate from carbon positions in
close proximity to the nitrogen. At the same time, these reso-
nances display a markedly increased linewidth when com-
pared to many of the peaks below 40 ppm which mainly
correspond to positions in the aliphatic tail of the molecule.
Fig. 4 13C (a) and 31P (b) magic-angle spinning nuclear magnetic reso-
nance spectroscopy of Pd-HHDMA-1 (orange), Pd-HHDMA-2 (blue),
Pd-HHDMA-3 (red), Pd-HHDMA-4 (green). The picture on the right
side depicts the mobile character of the ligand tail and the rigidity of the
phosphate group. Colour codes: Pd (blue), P (yellow), O (red), C (light
grey), and N (purple).
Fig. 3 Thermogravimetric profiles of the catalysts.
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.

View Article Online
Paper
Catalysis Science & Technology
We interpret this as being indicative of a difference in mobil-
ity (on the NMR timescale) between the head-group and the
tail, where presumably the aliphatic tail is dynamic and the
signals in the NMR spectra are motionally narrowed. Con-
trarily, the 31P MAS NMR spectra (Fig. 4b) display only a sin-
gle, broad isotropic signal around 0 ppm, which is consistent
with an orthophosphate anion. This peak is flanked on either
side by a spinning sideband giving a rough estimate of the
chemical shift anisotropy of about 50 ppm. The 31P peak with
a full width half height of approximately 2 kHz seems
to be inhomogeneous in nature and is indicative of a stati-
cally disordered environment. Overall, the broadness of the
peak in the 31P signal compared to the sharp signals
observed in the 13C spectra reflect the more rigid bonding of
the orthophosphate anion with respect to the ligand tail,
which exhibits a high degree of structural flexibility.
total ligand concentration is confirmed in the hydrogenation
of 3-hexyne, 2-methyl-3-butyn-2-ol, and 4-octyne (Fig. 6, S2
and S3†), as well as for the hydrogenation of 1-hexyne using
cyclohexane instead of toluene as a solvent (Fig. S4†). This
trend, which appears to contrast with the fact that the num-
ber of free atoms determined by CO chemisorption is re-
duced upon increasing of the HHDMA content, is explained
below. In terms of alkene selectivity, it is noteworthy to
highlight that the catalysts can only be compared at moder-
ate and similar degree of conversion (ca. 30%). This can
facilitate the detection of deactivation phenomena (which in
turn are somewhat obscure when catalytic tests are under-
taken in a batch reactor, reaching high degree of alkyne
conversion). In the window of temperatures and pressures
where these catalysts are applied industrially (T < 323 K
and P < 2 bar), the alkene selectivity approaches 100%.
Particularly, for the hydrogenation of 1-hexyne at 363 K and
1 bar, the selectivity to 1-hexene is 100% over Pd-HHDMA-1
and Pd-HHDMA-2, 96% over Pd-HHDMA-3 (4% alkane), and
90% over Pd-HHDMA-4 (5% alkane, 5% 2-hexene). The result
can be generalised for the hydrogenation of 3-hexyne,
2-methyl-3-butyn-2-ol, and 4-octyne (Fig. 6, S2 and S3†). In
all cases, a slight decrease in alkene selectivity is observed
at high temperatures and pressures (T > 343 K and P > 3
bar). This, however, is expected for any modified palladium
catalyst and is related to the tendency of palladium to form
of α and β-hydrides, catalysing isomerisation and over-
hydrogenation reactions.24
Hydrogenation of alkynes
The performance of the catalysts containing variable ligand
concentration was assessed in the hydrogenation of 1-hexyne
(Fig. 5). In general, the activity of the materials clearly in-
creases with the amount of HHDMA in the catalyst. For ex-
ample, the rate of 1-hexyne conversion at 363 K and 1 bar is
1.1 × 103 h−1 over Pd-HHDMA-1, 1.8 × 103 h−1 over Pd-
HHDMA-2, 2.5 × 103 h−1 over Pd-HHDMA-3, and 2.9 × 103 h−1
over Pd-HHDMA-4, a value which is 2.5 times higher activity
than that for Pd-HHDMA-1. The improved activity with the
Fig. 5 Reaction rate (in 103 h−1, a) and selectivity to 1-hexene (in %, b) in the hydrogenation of 1-hexyne. Conditions: Wcat = 0.1 g, FL (1-hexyne +
toluene) = 1 cm3 min−1, and FG(H2) = 36 cm3 min−1. The contour maps were obtained through spline interpolation of 16 experimental points.
Catal. Sci. Technol.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Catalysis Science & Technology
Paper
Fig. 7 Conversion and selectivity to cis-3-hexen-1-ol in the batch
semi-hydrogenation of 3-hexyn-1-ol. Conditions: Wcat = 0.4 g, T = 303
K, p (H2) = 3 bar, t = 10–15 min.
permitted to employ low irradiation doses, dedicated
supporting or staining strategies, and/or cryogenic condi-
tions, to greatly extend the scope of modern microscopy tech-
niques for the characterisation of organic–inorganic compos-
ites.25 An advanced microscopic evaluation of this
industrially-relevant catalyst can provide new fundamental in-
sights on the properties of HHDMA-modified catalysts at the
nanoscale, ultimately guiding the design of better hybrid cat-
alyst. For this reason, we have analysed the free colloidal solu-
tion obtained before or during synthesis of the supported
nanoparticles (Fig. 1). The micrographs in Fig. 8 depict the
capping role of the ligand; the lighter circles corresponding to
the HHDMA surfactant tail surround the dark spots in the
centre to the Pd nanoparticles (as illustrated in the coloured
inset). From the microscopy results, it appears that the aver-
age ligand thickness on the nanoparticles increases with the
ligand content. The thickness of the ligand shell, in particu-
lar, passes from being barely visible (ca. 0–0.5 nm) in Pd-
HHDMA-1 to 2.5 nm in Pd-HHDMA-3 (Fig. 8).
Fig. 6 Rate of reaction (a) and selectivity to the desired product (b)
versus the total ligand content (C/Pd) in the hydrogenation of 1-hexyne
(orange), 3-hexyne (green), 2-methyl-3-butyn-2-ol (blue), and
4-octyne (red). Conditions: Wcat = 0.1 g, T = 363 K, P = 1 bar, FL
(1-hexyne + toluene) = 1 cm3 min−1, and FG(H2) = 36 cm3 min−1
.
To study the activity of these materials at high con-
version levels, the hydrogenation of 3-hexyn-1-ol was inves-
tigated in batch mode (Fig. 7 and S7†). The batch experi-
ments demonstrate that, at 95% of alkyne conversion, the
alkene selectivity is approximately 90% over all samples,
confirming the outstanding performance of the ligand-
modified materials.
To demonstrate the robustness of the catalytic system,
characterisation of the used Pd-HHDMA-1 and Pd-HHDMA-4
has been carried out. The Pd and HHDMA contents in the
used catalysts are retained upon hydrogenation (Table 1),
highlighting the absence of any leaching upon reaction.
To assess the origins of the different thickness of the or-
ganic shell in the materials, it is important to estimate the
average orientation of the ligand chain. For this reason, we
introduce the order parameter, SHHDMA, which in quantum
chemistry measures the average orientational order of the
methylene chain units within the hydrocarbon tail.23 This
concept describes the correlation between the structure of a
self-assembled monolayer (SAM) and the metal and is de-
fined as SHHDMA = 0.5 <3 cos θi cos θj–δij>, where θi is the an-
gle between the ith molecular axis and the surface normal; θj
is the angle between the jth molecular axis and the normal to
the surface; and δij is the Kronecker delta function. Notice,
that the order parameter reaches a value of 0 when the sur-
factants flatten on the surface. At the maximal HHDMA den-
sity, a saturation value of SHHDMA,max = 1.24 is obtained, cor-
responding to an angle of the self-assembled layer with
respect to the normal to the surface of 34°. In general, Pd-
HHDMA nanoparticles with a low ligand content present very
low order parameters, SHHDMA. This indicates that the chains
are almost flattened on the surface. From the combined MD/
Structure–performance relationships
In order to rationalise the catalytic performance of these ma-
terials, we have investigated the structure and organisation of
the ligand interface, firstly through microscopy analyses. The
direct observation of the hybrid nanocatalyst has often been
approached by SEM and TEM.4–8 However, the use of these
conventional techniques is not straightforward due to both
the low contrast of the organic layer and the structural
changes which may occur upon irradiation. Without special
care to preserve the structure and enhance the visibility of
the organic shell, these techniques only provide information
about the size of the metal core. More recent developments,
particularly related to the study of biological materials, have
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.

View Article Online
Paper
Catalysis Science & Technology
Fig. 8 Transmission electron micrographs of the colloidal Pd nanoparticles with different ligand content employed in the synthesis of Pd-
HHDMA-1 (a), Pd-HHDMA-2 (b), and Pd-HHDMA-3 (c). The insets present a 4-fold magnification zoom of the boxed particles (light blue = palla-
dium, yellow = HHDMA layer). The scale bars in (a) apply to all micrographs.
DFT-D2 calculations (see the ESI† for details), we have ob-
served that this catalyst has a 2D-type surface, leaving small
areas where non-poisoned Pd atoms can catalyse the reaction
(Fig. 9a). This resembles the structure of the typical Lindlar
catalyst (Fig. 10).27 When the concentration of HHDMA
increases, SHHDMA also increases and the aliphatic chains
assume a more extended configuration (exhibiting similar
behaviour to a typical self-assembled monolayer), giving rise
to a 3D-like material (Fig. 9a).26 In this configuration, the
organic shell moves cooperatively, opening channels in the
structure where adsorption and reaction can occur. By corre-
lating the hydrogenation rate with the order parameter
(Fig. 9b and c), a linear correlation is obtained. This re-
lationship indicates that SHHDMA is the intrinsic descriptor
for activity the ligand-coated nanoparticles. As depicted in
Fig. S6,† this linear dependence can be generalised to
all other substrates in this work, although different slopes
could be observed, depending on the specific interaction
Fig. 9 Schematic representation of the catalysts structures with variable ligand concentration (a). Variation of the order parameter, SHHDMA, as a
function of the ligand content (b) and rate of 1-hexyne hydrogenation as a function of the order parameter (c). The red symbols in (b) represent
experimental points and the black symbols refer to the theoretical data. Conditions: Wcat = 0.1 g, T = 363 K, P = 1 bar, FL (1-hexyne + toluene) =
1 cm3 min−1, FG(H2) = 36 cm3 min−1. Colour codes as in Fig. 4.
Catal. Sci. Technol.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Catalysis Science & Technology
Paper
Fig. 10 Top view of the surface of the classical Lindlar catalyst
poisoned with quinoline (a), and Pd-HHDMA-1 (b). Colour codes: Pd
(blue), Pb (dark grey), P (yellow), O (red), C (light grey), and N (purple).
Fig. 11 Rate of reaction and selectivity versus the order parameter
(SHHDMA) in the hydrogenation of benzonitrile to benzylamine (a), and
benzaldehyde to benzyl alcohol (b). T = 343 K (orange), 353 K (green),
363 K (blue), and 373 K (red). Other conditions: Wcat = 0.2 g, P = 40
bar (a) and 10 bar (b), FL (substrate + methanol) = 0.3 cm3 min−1 (a) and
1 cm3 min−1 (b), and FG(H2) = 48 cm3 min−1
.
between the catalyst (metal and ligand) and the substrate.
The activity trend observed in Fig. 5 and 11 could be in con-
trast with the fact that the number of free surface atoms re-
quired for the catalytic reaction decreases upon ligand cover-
age increase, as shown by the CO chemisorption values. This
effect, however, can be rationalised by considering the differ-
ent size of CO and the alkyne molecules. While CO can easily
access the to the active site, even when the ligand is flattened
on the surface, due to its small size, the adsorption of the
bulkier reactants at low ligand content is less favourable.
This results in the atypical catalytic trend observed, with high
activity only at high HHDMA content, when ‘free’ pockets on
the surface are orderly formed and are accessible to bulky
acetylenic compounds.
The variation in selectivity over the densely-packed
HHDMA ligands at high coverage also arises from a combina-
tion of both the preferential orientation of the chains and
the impeded diffusion through the surfactant layer. The al-
kyne, thus, can penetrate the surfactant layer, is transformed
into the alkene on the surface, and if the hydrogen content at
the surface is sufficiently large, can be further transformed
into the alkane or isomerise. The latter explains the loss of
selectivity at high hydrogen pressures. On the other hand, in
the catalyst containing less ligand, the flattened HHDMA
layer leaves only a few reaction pockets, where the alkyne
molecules can adsorb and eventually react. Full selectivity
can be retrieved when the comparative thermodynamic term
for the adsorption of the alkyne is favourable with respect to
that of the alkene (i.e., strong alkyne and negligible alkene
adsorption).28–30 Since the reaction is carried out in solution,
these adsorption energies have to be referenced to that of the
solvent, as toluene or (cyclohexane) can competitively adsorb
on the surface, acting as an additional selectivity modifier.
We have computed these adsorption energies for 1-hexyne,
1-hexene, and toluene, employing the Pd-HHDMA model
structure with the lowest ligand content. The resulting values
(Eads = −1.43, −0.45, −0.82 eV for 1-hexyne, 1-hexene, and tolu-
ene, respectively) indicate that, while 1-hexyne is strongly
bound to the surface, 1-hexene can be easily displaced by the
solvent, which further justifies the high alkene selectivity
observed.
The catalytic results collected here apparently contrast
with earlier works. Schwartz and co-workers, for example, ob-
served, for thiol-decorated palladium and platinum nano-
particles, a decrease in activity with the increased thiol con-
centration.13 This result, however, can be rationalised by
comparing the HHDMA-covered nanoparticles with thiol- and
amine-covered ones. While thiols and amines flatten on the
surface independently from the ligand content,31,32 the pres-
ence of anionic and cationic counterparts in the HHDMA li-
gand is responsible for the different ligand conformation at
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.

View Article Online
Paper
Catalysis Science & Technology
low and high content. Thus, while thiols and amines can be
displaced by the reactants,33 which are strongly adsorbed on
the surface, the same does not occur over HHDMA-modified
nanoparticles. Similarly, Kiwi-Minsker and co-workers ob-
served a 4-fold increase in the TOF of acetylene hydrogena-
tion and a drop in olefin selectivity upon the removal of the
ligand shell in PVP-protected Pd particles by UV-ozone irradi-
ation.6 This was attributed to the better accessibility of the re-
actants to the active sites. Also in this case, the nature of the
interaction between these two classes of stabilisers (PVP and
HHDMA) with the nanoparticles should be considered. PVP
and HHDMA are chemically very different. PVP is a rather
homogeneous polymer with internal covalent bonds, low
polarity, two anchoring points (N and CO bonds) and
relatively small van der Waals contributions.34 In contrast,
HHDMA is a surfactant with charge separation (hydro-
phosphate anions and quaternary N cation heads), and a
large van der Waals contributions due to the long aliphatic
tails.9 This produces an extra degree of freedom in the
HHDMA-containing system, that allows the material to self-
organise, behaving as a 2D material at low ligand coverages
and as a 3D material at higher ligand coverages (Fig. 9a).
This extra source of flexibility is the key factor that explains
the differences between PVP and Pd-HHDMA catalyst.
evidencing an increased catalytic activity at higher ligand
concentrations. DFT calculations enable to rationalise of this
trend, showing that the adsorption geometry of HHDMA on
Pd change upon decreasing the ligand amount. Particularly,
the system passes from a 3D structure with an almost com-
plete coverage of ligand, to a 2D structure at the lowest con-
centration, resembling that of the Lindlar catalyst. Moreover,
it was found that the order parameter of the organic layer
represents an important descriptor that governs the struc-
ture–activity relationships of these catalysts. These results
were also generalised to the hydrogenation of other func-
tional groups such as nitriles and carbonyls. Overall, our con-
tribution has provided a deeper understanding of the influ-
ence of the HHDMA ligand on the catalytic performance of
the hybrid catalysts. Despite having developed a new syn-
thetic methodology to tune the ligand content in the mate-
rial, the accurate quantification of the ligand coverage per
nanoparticle remains an open question for future studies.
Acknowledgements
Financial support from ETH Zurich, ICIQ Foundation, and
the Spanish Ministerio de Economía
y Competitividad
(CTQ2012-33826 and ‘Ayuda formación Posdoctoral’ Fellow-
ship (N. A.-B.)) are acknowledged. Dr Roland Hauert (EMPA,
Dubendorf) is acknowledged for the XPS analyses and Mr. Pe-
ter Tittmann for the help with catalyst preparation. The Scien-
tific Centre for Optical and Electron Microscopy (ScopeM) of
ETH is acknowledged for use of its facilities. BSC-RES is
thanked for generous computational resources.
Generalisation to other substrates
We have finally extrapolated the original findings to the hy-
drogenation of nitriles to amines and of carbonyls to alco-
hols. None of these applications has been previously studied
over Pd-HHDMA/TiSi2O6. As shown in Fig. 11, the same activ-
ity pattern encountered in the case of alkynes has been
obtained for these other functional groups. In particular, by References
increasing the ligand amount, an increased catalytic activity
1 (a) A. Roucoux, J. Schulz and H. Patin, Chem. Rev.,
and a slightly decreased selectivity are observed. These find-
ings further elucidate how hybrid catalysts can achieve full
product selectivity for the hydrogenation of nitriles and car-
bonyls without typical additions of corrosive acids and/or
toxic bases (e.g., HCl and NH3) in stoichiometric amounts
and additional downstream separation steps. Thus, the
electronic and geometrical modifications imparted by the li-
gand can efficiently limit the kinetics of undesired products
formation.
2002, 102, 3757; (b) H.-U. Blaser, C. Malan, B. Pugin, F.
Spindler, H. Steiner and M. Studer, Adv. Synth. Catal.,
2003, 345, 1.
2 C. Burda, X. Chen, R. Narayanan and M. A. El-Sayed, Chem.
Rev., 2005, 105, 1025.
3 (a) N. Semagina and L. Kiwi-Minsker, Catal. Rev.: Sci. Eng.,
2009, 51, 147; (b) A. J. McCue, F. McKenna and J. A.
Anderson, Catal. Sci. Technol., 2015, 5, 2449.
4 R. Venkatesan, M. H. G. Prechtl, J. D. Scholten, R. P. Pezzi,
G. Machado and J. Dupont, J. Mater. Chem., 2011, 21, 3030.
5 H. Bönnemann, W. Brijoux, R. Brinkmann, E. Dinjus, T.
Joussen and B. Korall, Angew. Chem., Int. Ed. Engl., 1991, 30,
1312.
6 M. Crespo-Quesada, J. M. Andanson, A. Yarulin, B. Lim, Y.
Xia and L. Kiwi-Minsker, Langmuir, 2011, 27, 7909.
7 P. T. Witte, P. H. Berben, S. Boland, E. H. Boymans, D. Vogt,
J. W. Geus and J. G. Donkervoort, Top. Catal., 2012, 55, 505.
8 P. T. Witte, S. Boland, F. Kirby, R. van Maanen, B. F.
Bleeker, D. A. M. de Winter, J. A. Post, J. W. Geus and P. H.
Berben, ChemCatChem, 2013, 5, 582.
Conclusions
We have prepared and characterised a series of HHDMA-
modified Pd catalysts with different amounts of ligand. Ad-
vanced techniques going beyond state-of-the-art approaches
were essential not only to visualise the changes in the thick-
ness of the ligand shell on the Pd nanoparticles, but also to
understand and describe its flexibility and structure. These
techniques include high-resolution magic-angle spinning nu-
clear magnetic resonance spectroscopy, and transmission
electron microscopy of the colloids after negatively staining
with uranyl acetate. The materials were tested in the
continuous-flow three-phase semi-hydrogenation of alkynes,
9 G. Vilé, N. Almora-Barrios, S. Mitchell, N. López and J. Pérez-
Ramírez, Chem. – Eur. J., 2014, 20, 5926.
Catal. Sci. Technol.
This journal is © The Royal Society of Chemistry 2015

View Article Online
Catalysis Science & Technology
Paper
10 G. Vilé, N. Almora-Barrios, N. López and J. Pérez-Ramírez,
ACS Catal., 2015, 5, 3767.
21 N. Almora-Barrios, G. Carchini, P. Błoński and N. López,
J. Chem. Theory Comput., 2014, 10, 5002.
11 E. H. Boymans, P. T. Witte and D. Vogt, Catal. Sci. Technol.,
2015, 5, 176.
22 R. García-Muelas and N. López, J. Phys. Chem. C, 2014, 118,
17531.
12 (a) P. A. Sheth, M. Neurock and C. M. Smith, J. Phys. Chem.
B, 1996, 30, 33; (b) P. A. Sheth, M. Neurock and C. M. Smith,
J. Phys. Chem. B, 2005, 109, 12449; (c) J. Greeley and M.
Mavrikakis, J. Phys. Chem. B, 2005, 109, 3460; (d) D.
Teschner, J. Borsodi, A. Wootsch, Z. Révay, M. Hävecker, A.
Knop-Gericke, S. D. Jackson and R. Schlögl, Science,
2008, 320, 86; (e) A. J. McCue, C. J. McRitchie, A. M.
Shepherd and J. A. Anderson, J. Catal., 2014, 319, 127; ( f )
A. J. McCue and J. A. Anderson, Catal. Sci. Technol., 2014, 4,
272; (g) A. J. McCue and J. A. Anderson, Front. Chem. Sci.
Eng., 2015, 9, 142.
23 A. R. Leach, Molecular Modelling Principles and Applications,
2nd edn., 2001, p. 395.
24 F.-M. McKenna and J. A. Anderson, J. Catal., 2011, 281, 231.
25 (a) G. Nizri, S. Magdessi, J. Schmidt, Y. Cohen and Y.
Talman, Langmuir, 2004, 20, 4380; (b) Z. Lee, K.-J. Jeon, A.
Date, R. Erni, T. J. Richardson, M. Frenklach and V.
Radmilovic, Nano Lett., 2009, 9, 3365; (c) F. Schacher, E.
Betthausen, A. Walther, H. Schmalz, D. V. Pevgushov and
A. H. E. Müller, ACS Nano, 2009, 3, 2095; (d) C. V.
Synatschke, F. H. Schacher, M. Förtsch, M. Dreschsler and
A. H. E. Müller, Soft Matter, 2011, 7, 174.
13 S. T. Marshall, M. O'Brien, B. Oetter, A. Corpuz, R. M.
Richards, D. K. Schwartz and J. W. Medlin, Nat. Mater.,
2010, 9, 853.
14 G. Fagherazzi, P. Canton, P. Riello, N. Pernicone, F. Pinna
and M. Battagliarin, Langmuir, 2000, 16, 4539.
26 J. C. Love, L. A. Estroff, J. K. Kriebel, R. G. Nuzzo and G. M.
Whitesides, Chem. Rev., 2005, 105, 1103.
27 M. García-Mota, J. Gómez-Díaz, G. Novell-Leruth, C. Vargas-
Fuentes, L. Bellarosa, B. Bridier, J. Pérez-Ramírez and N.
López, Theor. Chem. Acc., 2011, 128, 663.
15 (a) J. Schaefer, E. O. Stejskal and R. Buchdahl,
Macromolecules, 1975, 8, 291; (b) A. Pines, M. G. Gibby
and J. S. Waugh, Chem. Phys. Lett., 1972, 15, 373; (c) B. M.
Fung, A. K. Khitrin and K. Ermolaev, J. Magn. Reson.,
2000, 142, 97.
16 G. Kresse and J. Furthmuller, Phys. Rev. B: Condens. Matter
Mater. Phys., 1996, 54, 11169.
17 G. Kresse and J. Furthmuller, Comput. Mater. Sci., 1996, 6,
15.
18 G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter
Mater. Phys., 1999, 59, 1758.
19 B. Hammer, L. B. Hansen and J. K. Nørskov, Phys. Rev. B:
Condens. Matter Mater. Phys., 1999, 59, 7413.
28 Y. Segura, N. López and J. Pérez-Ramírez, J. Catal.,
2007, 247, 383.
29 F. Studt, F. Abild-Pedersen, T. Bligaard, R. Z. Sørensen, C. H.
Christensen and J. K. Nørskov, Science, 2008, 320, 1320.
30 J. A. Anderson, J. Mellor and R. P. K. Wells, J. Catal.,
2009, 261, 208.
31 C. A. Schoenbaum, D. K. Schwartz and J. W. Medlin,
J. Catal., 2013, 303, 92.
32 L. Altmann, S. Kunz and M. Bäumer, J. Phys. Chem. C,
2014, 118, 8925.
33 M. M. Green, M. P. Reidy, R. D. Johnson, G. Darling, D. J.
O'Leary and G. Willson, J. Am. Chem. Soc., 1989, 111, 6452.
34 M. Okumura, Y. Kitagawa, T. Kawakami and M. Haruta,
Chem. Phys. Lett., 2008, 459, 133.
20 S. Grimme, J. Comput. Chem., 2006, 27, 1787.
This journal is © The Royal Society of Chemistry 2015
Catal. Sci. Technol.