Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

View Article Online / Journal Homepage / Table of Contents for this issue

This article is published as part of the Dalton Transactions themed issue entitled:

Bridging the gap in catalysis via


multidisciplinary approaches
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

Guest Editors: Christophe Coperet and Rutger van Santen


Université de Lyon, France and Eindhoven University of Technology, The Netherlands

Published in issue 36, 2010 of Dalton Transactions

Image reproduced with the permission of Dieter Vogt

Articles in the issue include:


Molecular understanding of alkyne hydrogenation for the design of selective catalysts
Javier Pérez-Ramírez, Blaise Bridier and Nuria Lopez
Dalton Trans., 2010, DOI: 10.1039/C0DT00010H

Molecular weight enlargement–a molecular approach to continuous homogeneous catalysis


Michèle Janssen, Christian Müller and Dieter Vogt, Dalton Trans., 2010,
DOI: 10.1039/C0DT00175A

Structure Determination of Zeolites and Ordered Mesoporous Materials by Electron


Crystallography
Xiaodong Zou, Junliang Sun, Dalton Trans., 2010, DOI: 10.1039/C0DT00666A

Metal-Catalyzed Immortal Ring-Opening Polymerization of Lactones, Lactides and Cyclic


Carbonates
Noureddine Ajellal, Jean-François Carpentier, Clémence Guillaume, Sophie M. Guillaume, Marion
Helou, Valentin Poirier, Yann Sarazin and Alexander Trifonov, Dalton Trans., 2010, DOI:
10.1039/C001226B

Visit the Dalton Transactions website for more cutting-edge inorganic and organometallic research
www.rsc.org/dalton
PAPER www.rsc.org/dalton | Dalton Transactions

Oxygen activation on gold nanoparticles: separating the influence of particle


size, particle shape and support interaction
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

Mercedes Boronat and Avelino Corma*


Received 2nd February 2010, Accepted 16th April 2010
DOI: 10.1039/c002280b

The adsorption and dissociation of molecular O2 on extended gold surfaces, isolated gold nanoparticles
of different size and shape, and small gold clusters supported on TiO2 have been investigated by means
of DFT calculations. The influence of particle size and morphology on the mode of adsorption and
activation of O2 have been analyzed separately, and it has been found that O2 dissociation is very
sensitive to the arrangement of the gold surface atoms. The control of this variable through catalyst
preparation opens one way to improve the activity of gold catalysts.

1. Introduction One of the key steps in the mechanism of gold catalyzed


oxidation reactions, and yet one of the least understood, is the
The exceptional catalytic activity of gold nanoparticles supported activation and dissociation of O2 . While there is general consensus
on inorganic oxides initially reported by Haruta1 has attracted that dissociative adsorption of molecular O2 on gold single crystals
increased interest during the last few years. The amount of infor- is energetically unfavourable, the ability of low coordinated gold
mation published about CO oxidation by gold is impressive,2–19 atoms to adsorb and even dissociate O2 has been proposed on
and the field of gold catalysis has been opened to other re- the basis of theoretical calculations.17,50,51,69,70 The high reaction
actions of practical interest such as selective alcohol, aldehyde rate for CO oxidation observed on a highly ordered gold bilayer
and olefin oxidations,20–27 C–C coupling reactions,28–31 water gas structure in which the underlying oxide support is not accessible to
shift,32 direct synthesis of hydrogen peroxide from H2 and O2 ,33,34 reactants4,71 and the experimental confirmation by XANES that
and selective hydrogenations of olefins,35–38 aldehydes39–43 and gold nanoparticles can be oxidized by molecular O2 13 are in line
nitroaromatics.44–46 with this proposal, and suggest that particle morphology rather
There is general agreement concerning the direct relationship than particle size is a key parameter in the catalytic properties of
between the catalytic activity of gold nanoparticles and a number gold. On the other hand, the dominating role played by the gold-
of morphological and electronic effects such as particle size support interface in the adsorption and activation of O2 has been
and shape, oxidation state of gold, or charge transfer between stressed by different authors.18,19,65,66,66,72,73
the gold particle and the metal oxide support. For instance, it In this work we have investigated the interaction of molecular O2
is widely accepted that the activity for hydrogenation of gold with several models of gold catalysts, including extended surfaces,
catalysts is directly related to the number of low coordinated gold isolated nanoparticles of different size and shape and small
atoms where H2 , as well as other reactants like CO or NO, are nanoparticles supported on TiO2 . We intend to first disentangle the
more strongly adsorbed and activated.47–57 In this sense, it has influence of particle size and morphology on the way of adsorption
been suggested that the increasing activity of gold catalysts with and activation of O2 , and to analyze the changes induced by the
decreasing gold particle size is related to the higher concentration metal oxide support on several properties chosen to characterize
of low coordinated gold atoms in the smallest particles. It has also the interaction: adsorption energy, molecular geometry, nOO
been demonstrated that the morphology of gold nanoparticles stretching frequency and charge transfer between the catalyst and
and therefore the number of low coordinated gold atoms can be the adsorbate. In a second step, the complete reaction pathway
modified by interactions with CO,58–60 which could explain the for O2 dissociation on the several gold catalysts considered has
strong influence of the method of preparation and the calcination been calculated, and a relationship has been established between
or reduction pre-treatments on the activity of supported gold the degree of charge transfer in the adsorbed reactant and the
catalysts.1,2,5,6,61,62,63 There is also general agreement concerning activation energy necessary to dissociate O2 . It is demonstrated
the higher catalytic activity of gold nanoparticles supported on that O2 dissociation is sensitive to the structure of the gold surface,
reducible oxides such as TiO2 , Fe2 O3 , or CeO2 ,1,2,3,4,5,6,10,11,64 but opening a way to improve the activity of gold catalysts.
whether the role of the support is limited to the stabilization
of small particles or is involved in the formation of cationic
gold surface sites or in the activation of reactants is still under
2. Model catalysts and computational details
debate.65–68 The interaction of molecular oxygen with isolated and supported
gold nanoparticles was studied by means of periodic slab models
using super cells large enough to avoid interaction between the
periodically repeated gold particles or between the adsorbates.
Instituto de Tecnologı́a Quı́mica (UPV-CSIC), Av. de los Naranjos s/n,
46022, Valencia, Spain. E-mail: acorma@itq.upv.es; Fax: +34 96 3879444; Two isolated gold nanoparticles of different size were considered:
Tel: +34 96 387 7800 a Au38 cluster having a typical cuboctahedral shape and 1 nm

8538 | Dalton Trans., 2010, 39, 8538–8546 This journal is © The Royal Society of Chemistry 2010
diameter, and a hemispherical Au13 cluster obtained by removing Kohn–Sham orbitals used to obtain the electron density were
four atomic layers parallel to (100) from the Au38 model. The Au38 expanded in a plane wave basis set with a kinetic energy cut off of
cluster has been previously employed in our group45 and has been 500 eV, and the effect of the core electrons was taken into account
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

demonstrated by Roldán et al. that it is able to dissociate O2 .70 by means of the projected augmented wave (PAW) method.78 The
The smaller Au13 clusters have also been widely used as models Brillouin zone of the (2 ¥ 2) super cell models was described using a
for gold nanoparticles, both isolated and supported on TiO2 .54,60 (5 ¥ 5 ¥ 1) Monkhorst–Pack mesh,79 and the rest of the calculations
The isolated Au38 and Au13 clusters were placed in a 20 ¥ 20 ¥ (isolated nanoparticles and (4 ¥ 4) super cell models) were carried
20 Å cubic box, and the position of all atoms was always fully out at the C k-point of the Brillouin zone. The atomic positions
relaxed. The TiO2 support was represented by a (4 ¥ 4) super were optimized by means of a conjugate-gradient algorithm until
cell slab model of the most reactive (001) facet of the anatase atomic forces were smaller than 0.01 eV Å-1 . Transition states were
polymorph. This model contains nine atomic layers in the unit cell, located using the DIMER algorithm80,81 and stationary points
that is, 144 atoms, and a vacuum region larger than 20 Å between were characterized by pertinent frequency analysis calculations.
vertically repeated slabs. Besides the stoichiometric surface (TiO2 Vibrational frequencies were calculated by diagonalizing the block
model), a reduced surface containing a oxygen vacancy defect Hessian matrix corresponding to displacements of the O atoms of
(TiO2 –Ovac model) was also considered. The Au13 cluster was then O2 molecule and the Au, Ti or O atoms of the catalyst in direct
placed on the stoichiometric and reduced TiO2 surface models contact with O2 . Charge distributions were estimated using the
described above, and the atomic positions of all Au atoms and of theory of atoms in molecules (AIM) of Bader using the algorithm
the Ti and O atoms of the two uppermost layers of the support developed by Henkelman.82,83
were fully relaxed. Finally, extended Au(111) and Au(100) surfaces
were modelled by (2 ¥ 2) super cell slabs containing four atomic
layers and a vacuum region larger than 20 Å between vertically
repeated slabs. In the geometry optimization of these extended 3. Results and discussion
gold surfaces, the atomic positions of the two innermost layers 3.1. Adsorption of molecular O2
were kept as in the bulk and the two uppermost layers were fully
relaxed. Molecular O2 was placed on the two extended gold surfaces and on
Molecular O2 was then adsorbed on the different gold surfaces the several isolated and supported gold nanoparticles considered
and nanoparticles described above, and the atomic positions of in this work with the molecule initially adopting the three different
O2 , of all Au atoms in nanoparticles, of the two uppermost layers conformations depicted in Scheme 1. The geometry of these initial
of the gold extended surface models, and of the Ti and O atoms structures was optimized as described in previous section, and
of the two uppermost layers of the TiO2 support were in all cases the most relevant information relative to the stable structures
fully relaxed. Except when otherwise stated, adsorption energies finally obtained is summarized in Table 1. Only one minimum
E ads were calculated by subtracting the energy of the relaxed gold was localized on Au(111) surface, corresponding to a top-bridge-
catalyst model E Au and of optimized O2 in its triplet state E O2 from top conformation in which the two oxygen atoms are bonded
the total energy of the adsorption complex EAu-O2 according to: to neighbouring gold atoms. The optimized Au–O distances are
2.27 Å and the O–O bond is slightly elongated in relation to the gas
E ads = E Au-O2 - (E Au + E O2 )
phase value, that is 1.235 Å. The same conformation was obtained
Taking into account the results of previous theoretical studies, on Au(100) surface and, as shown in Table 1, the optimized
the three possible ways of adsorption of molecular O2 depicted geometries and the calculated nOO stretching frequencies are
in Scheme 1 were initially considered: a end-on mode (eo) with similar on both surfaces. However, the calculated adsorption
the O2 molecule situated approximately normal to the surface and energies are positive, indicating that these tbt structures are not
with only one oxygen atom directly bonded to a gold atom, a top- stable species. A second conformation, with each of the two oxygen
bridge-top (tbt) conformation with each oxygen atom in O2 directly atoms of O2 forming a bridge with two different gold atoms (bb),
bonded to a gold atom, and a bridge-bridge (bb) conformation in was obtained on the Au(100) surface. The Au–O distances in this
which each oxygen atom is interacting with two gold atoms. structure are ~ 2.3 Å too, but the O–O bond length increases to
1.42 Å, and consequently the nOO frequency is shifted to lower
values. Although the calculated adsorption energy is negative, the
value obtained is very small, reflecting the experimental fact that
O2 does not interact with perfect gold surfaces. The last property
analyzed in this work is the charge transfer from the metal to O2 .
The calculated values for the tbt species are similar and less than
0.5 e, while the value obtained for the bb conformer is larger, 0.8
e. This suggests that the degree of metal-adsorbate charge transfer
depends mainly on the mode of adsorption of O2 , and not so much
Scheme 1 Adsorption modes of O2 on gold. on the structure of the metal surface. To check whether these results
vary with the degree of coverage of O2 we re-optimized the most
All calculations are based on density functional theory stable structure, that is the bb adsorption complex on Au(001),
(DFT) and were carried out using the Perdew–Wang (PW91)74,75 using a 4 ¥ 2 super cell slab model. The values provided by the
exchange–correlation functional within the generalized gradient new setup, corresponding to a degree of coverage q = 0.25, are
approximation (GGA) as implemented in the VASP code.76,77 The completely equivalent to those obtained for q = 0.5, showing the

This journal is © The Royal Society of Chemistry 2010 Dalton Trans., 2010, 39, 8538–8546 | 8539
Table 1 Adsorption of molecular O2 on different gold models. Optimized O–O and Au–O distances (in Å), and calculated adsorption energies (E ads ,
in eV), vibrational frequencies (n(OO), in cm-1 ) and total charge on O2 (qO2 , in e)
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

E ads r(OO) r(Au–O) n(OO) qO2

Au(111) tbt +0.24 1.319 2.268 1074 -0.451


Au(100) tbt +0.09 1.325 2.200 1055 -0.462
Au(100) bb -0.10 1.422 2.287 749 -0.813
Au(100) 4 ¥ 2 bb -0.22 1.427 2.286 751 -0.843
Au38 tbta -0.99 1.367 2.123 928 -0.624
Au38 tbtb -1.04 1.364 2.111 936 -0.639
Au38 bb -0.98 1.459 2.288 714 -0.874
Au13 eo -0.10 1.293 2.127 1223 -0.367
Au13 tbt -1.08a 1.344 2.127 1005 -0.534
Au13 bb -0.74a 1.528 2.204 617 -0.906
Au13 /TiO2 eo -0.80 1.282 2.182 1270 -0.300
Au13 /TiO2 tbt -0.94 1.348 2.191 1005 -0.540
Au13 /TiO2 msi -2.15 1.343 1.994b 1122 -0.542
Au13 /TiO2 –Ovac eo -0.49 1.280 2.172 1280 -0.279
Au13 /TiO2 –Ovac tbt -0.75 1.334 2.167 1061 -0.502
Au13 /TiO2 –Ovac msi -2.41 1.450 2.131 867 -0.901
a
Calculated as E ads = (E Au13-O2 )fully relaxed geometry - [(E Au13 )fixed geometry as relaxed for the complex + (E O2 )fully relaxed ], b r(Ti–O).

validity of the 2 ¥ 2 super cell slab model for describing the studied of similar stability, the bridge-bridge complex (bb) depicted in
phenomenon. Fig. 1. The O–O bond length in this species is larger, 1.46 Å,
The Au38 cluster chosen to simulate small gold nanoparticles has and the nOO frequency shifts to 714 cm-1 . Analysis of the net
a cuboctahedral shape and exposes a combination of Au(111) and charge on the O2 molecule after adsorption indicates that, as in the
Au(100) facets. Fig. 1 shows the structures of the three most stable case of perfect gold surfaces, the degree of metal-adsorbate charge
O2 adsorption complexes obtained after complete relaxation of all transfer depends on the way the adsorbate interacts with the metal,
atoms in the model, and the most relevant information about them and is larger for the bb than for the tbt conformations. It should be
is summarized in Table 1. There are two top-bridge-top adsorption mentioned that all these structures were also obtained by Roldan
modes of similar stability that differ in the coordination of the gold et al.,70 but only the properties of the bridge-bridge conformation
atoms to which oxygen atoms are directly bonded. In structure tbta in relation to its ability to dissociate O2 were discussed in their
the two oxygen atoms of O2 are bonded to gold atoms belonging to paper.
the Au(111) facet and having coordination number 9 and 6, while To check the effect of particle size, molecular O2 was adsorbed
in structure tbtb the O2 is bonded to two gold atoms belonging to on a small Au13 cluster that initially consisted of two layers of
the interface between Au(111) and Au(100) facets, both of them four and nine atoms parallel to (100) and with (111) side facets.
with coordination number 6. However, the different coordination This and other isomers of Au13 were described in detail in ref.
of the gold atoms has no influence on the adsorption properties 54. While adsorption of O2 in the bb conformation results in a
analyzed. The stability of both structures is the same, the optimized small deformation of the initial cluster structure, the interaction
Au–O and O–O distances are equivalent, and the nOO stretching of O2 in the eo or tbt modes causes a large geometry distortion
frequency is ~ 930 cm-1 in both cases. There is a third conformation in the gold nanoparticle, as depicted in Fig. 1. The irregular
particle obtained after full optimization without any adsorbed
O2 is 0.79 eV more stable than the hemispherical model initially
considered. If adsorption energies were calculated with respect
to the most stable isomer of the Au13 cluster, as has been done
for the rest of the models considered in this work, the values
obtained would mainly reflect the energy involved in particle
deformation, and would not provide much information about the
interaction with O2 . Therefore, the adsorption energies reported
in Table 1 have been calculated with respect to the energy of the
isolated particle in the same conformation as in the adsorption
complex, without any further geometry optimization. The end-
on adsorption mode, with only one oxygen atom directly bonded
to gold, has been found to be stable only on steps and defects
involving low coordinated gold atoms.69,84 A stable structure with
only one O atom directly bonded to a gold atom with coordination
number 5 has been obtained here on the Au13 cluster (see Fig. 1,
bottom, left). The optimized Au–O distance in this structure is
Fig. 1 Optimized structures of adsorption complexes of O2 on Au38 (top) similar to the values reported for the other complexes, but the
and Au13 (bottom) nanoparticles. adsorption energy is almost negligible, and the changes in the

8540 | Dalton Trans., 2010, 39, 8538–8546 This journal is © The Royal Society of Chemistry 2010
properties of O2 molecule upon adsorption are minor. The O–O
bond length is slightly shorter than in the isolated molecule, and
the nOO frequency is larger than 1200 cm-1 , not so different from
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

the corresponding gas phase value (1574 cm-1 calculated, 1556 cm-1
experimental85 ). The interaction of O2 with the Au13 cluster in both
tbt and bb modes is stronger and energetically favourable. The
changes in the geometric and vibrational properties of O2 upon
adsorption are similar to those observed on the Au38 nanoparticle,
and the calculated adsorption energies are approximately in the
same range.
The most remarkable conclusion provided by the comparative
study of gold surfaces and isolated nanoparticles of different size
is that the mode of adsorption of O2 does not depend on particle
size, but on particle morphology. Thus, while the tbt conformation
is formed, although with different stability, on any gold surface or
particle having two gold atoms, the eo adsorption mode requires
the presence of low coordinated gold atoms, and only systems
exposing (100) facets are able to stabilize the bb conformation.
The metal oxide support modifies the properties of gold
nanoparticles deposited on it, and therefore the way in which
O2 adsorbs and is activated. It is known that there is a transfer
of electron density between gold and the support, and that
gold nanoparticles become positively charged when supported
on stoichiometric and oxidized TiO2 surfaces, and negatively
charged when adsorbed on reduced surfaces. Moreover, the low
coordinated gold atoms in the nanoparticle tend to increase their
coordination number by forming Au–Ti bonds, and as a result the
particle shape may be noticeably distorted. In this work, O2 has
been adsorbed on different positions of the two Au/TiO2 catalyst
models described in section 2, and that have been obtained by
placing a Au13 cluster on either a stoichiometric (Au13 /TiO2 model)
or a reduced (Au13 /TiO2 –Ovac model) TiO2 surface. The properties Fig. 2 Optimized structures of adsorption complexes of O2 on Au13 /TiO2
of these two models have been extensively described and discussed (left) and Au13 /TiO2 –Ovac (right) models.
in previous work.54,60
The optimized structures of the adsorption complexes finally initially placing molecular O2 with one oxygen atom bonded to a
obtained on Au13 /TiO2 and Au13 /TiO2 –Ovac models are depicted gold atom of the nanoparticle and the other one directly bonded
in Fig. 2, and relevant information is summarized in Table 1. to a Ti atom of the support, forming a Ti–O–O–Au bridge. On
Both end-on and top-bridge-top adsorption modes were found, the stoichiometric support, the Au–O interaction is weak, and the
with the difference in stability between them being smaller molecule is finally placed on a fivefold coordinated Ti atom of the
than on the non-supported clusters. The optimized geometries, anatase surface, forming two equivalent Ti–O bonds. The degree
vibrational frequencies and charge transfer to O2 are however quite of O2 activation in this situation, measured by the increase in the
similar to those obtained on the non-supported Au13 nanoparticle, O–O distance and in the net charge transfer from the catalyst, is
suggesting that the direct influence of the oxide support on the similar to that obtained for the tbt conformation. However, the
activating ability of gold-only sites is limited. An indirect effect, adsorption energy is considerably higher, and therefore it can be
mainly related to the morphology of the supported particle, concluded that O2 will preferentially adsorb on the oxide support
is however remarkable. As described in detail in a previous and not on the gold nanoparticle. The situation is similar, but not
theoretical study of the active sites for H2 adsorption in Au/TiO2 completely equivalent, on the reduced Au/TiO2 catalyst, simulated
catalysts,54 when the Au13 isomer exposing (100) facets is placed by the Au13 /TiO2 –Ovac model. In this case the Au–O interaction is
on either the stoichiometric or reduced TiO2 surface, its geometry strong enough to keep the O2 molecule in the initial conformation,
is highly distorted and the final particles do not contain any forming two Ti–O bonds with calculated Ti–O distances of 1.864
arrangement of atoms that could resemble the initial (100) facet. and 2.084 Å, and with one oxygen atom still directly bonded
As a consequence, the bridge-bridge mode of adsorption of O2 to gold at a Au–O distance of 2.131 Å. In this conformation,
could not even be obtained on these supported gold nanoparticles. which is by far the most stable on the Au13 /TiO2 –Ovac model,
On the other hand, new sites appear at the metal-support interface, the O2 molecule has received 0.9 e from the catalyst, and the
which have been labelled as msi sites. According to previous O–O distance has increased to 1.45 Å, indicating an important
theoretical studies on Au/TiO2 catalyst models,49,50,65,66,67,68 O2 is molecular activation. The calculated nOO vibration frequencies
highly activated when it is placed at the interface between the gold for the msi sites in stoichiometric and reduced Au/TiO2 catalyst
nanoparticle and the oxide support. Fig. 2c shows the structures models are quite different, reflecting the different geometry and
obtained after geometry optimization of complexes formed by degree of molecular activation on the two systems.

This journal is © The Royal Society of Chemistry 2010 Dalton Trans., 2010, 39, 8538–8546 | 8541
A comparative analysis of the data summarized in Table 1 allows adsorption of O2 on the way it is activated. Thus, adsorption
us to extract some general trends and conclusions that are valid for in an end-on conformation causes small changes in the properties
all catalyst models considered in this work, including perfect gold studied, while the larger activation of O2 is found in the bridge-
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

surfaces, isolated nanoparticles of different size and shape and bridge conformations. It is interesting to remark the position of
small nanoparticles supported on TiO2 . The first clear result is the points corresponding to O2 adsorption at the metal-support
that adsorption energies roughly depend on particle size; they are interface (msi) in Au/TiO2 models, close to the tbt group for the
negligible on extended surfaces and become significant on discrete stoichiometric surface and among the most activated species in
nanoparticles, no matter the number of gold atoms they contain the bb group for the reduced surface.
nor the fact of being supported on a metal oxide surface. It can also
be concluded that on real Au/TiO2 catalysts, O2 will preferentially 3.2. O2 dissociation
adsorb on the support or at the metal–support interface, not on
the gold particles. Once it has been established which is the most activating mode
But adsorption energy is not necessarily a good predictor of of adsorption, the complete reaction path for O2 dissociation
molecular activation. Activation of adsorbed O2 occurs through on several catalyst models has been calculated, and a correlation
occupation of the p* molecular orbitals, that leads to a weakening between activation energies and adsorption mode has been found.
of the O–O bond. Therefore, the properties that could better Only the tbt and bb conformations have been considered as initial
reflect the degree of molecular activation are the total electron reactants on gold surfaces and isolated nanoparticles, while for
density transfer from the catalyst to O2 , the increase in the O–O supported nanoparticles the tbt and msi modes on the reduced
bond length caused by this transfer, and the nOO that reflects surface have been explored.
the strength of the O–O bond. Fig. 3 shows the direct relation- The energy profile for the process is depicted in Fig. 4, together
ship existing between these three variables: the nOO stretching with the optimized structures corresponding to reactant (R),
frequency linearly decreases with increasing O–O distance and transition state (TS) and reaction product (P) on the Au(100)
with increasing net charge on O2 . Moreover, and what is more surface. Starting from the optimized geometry of O2 adsorbed in
important, it puts forward the large influence of the mode of a bridge-bridge conformation, the reaction pathway involves the
breakening of the O–O bond, that is, the reaction coordinate is
clearly associated to a lengthening of the O–O bond length. The
optimized value of this r(OO) in the transition state is 2.034 Å (see
Table 2) and each oxygen atom is forming a Au–O–Au bridge with
two gold atoms. The reaction product is a structure in which the
two O atoms are placed in hollow positions on the Au(100) surface,
which is slightly distorted by the presence of adsorbed oxygen
atoms. The reaction is exothermic, and the calculated activation
energy is 0.44 eV, considerably lower than that reported for O2
dissociation on Au(111) or stepped Au(211) surfaces, 2.33 and
0.93 eV, respectively,66 and comparable to that recently reported
for O2 dissociation on the (100) facet of a Au29 nanoparticle
supported on TiC.86 The dissociation of O2 initially adsorbed
on a bb conformation on isolated Au13 or Au38 nanoparticles is
analogous to that described for the Au(100) surface. The optimized
structures of the transition states shown in Fig. 5 are equivalent
to that depicted in Fig. 4, and only the optimized value of the
r(OO)TS shows a small variation with particle size, being shorter on
the smaller particle. A similar trend is observed on the activation
barriers, that slightly decrease with decreasing particle size and
reach a really low value, 0.18 eV, on the Au13 cluster. The reaction
is exothermic in all cases, but the differences in the calculated

Table 2 Optimized value of the O–O distance in the transition states for
O2 dissociation (r(OO)TS , in Å) and calculated adsorption (E ads ), activation
(E act ) and reaction (DE) energies (in kcal mol-1 ) for O2 dissociation on gold
nanoparticles

r(OO)TS E ads E act DE

Au(100) bb 2.034 -0.22 0.44 -0.24


Au38 bb 1.968 -0.98 0.36 -0.94
Au38 tbt 1.947 -0.99 0.97 -0.43
Au13 bb 1.849 -0.98 0.18 -0.73
Fig. 3 Correlation between calculated stretching frequencies n(OO) and Au13 tbt 1.935 -1.08 1.58 -1.27
Au13 /TiO2 –Ovac tbt 2.412 -0.75 1.30 -0.36
(a) r(OO) distances and (b) total charge on dioxygen qO2 in the adsorption Au13 /TiO2 –Ovac msi 2.401 -2.41 1.10 -0.71
complexes studied in this work.

8542 | Dalton Trans., 2010, 39, 8538–8546 This journal is © The Royal Society of Chemistry 2010
to the dissociation of the O–O bond, but also involves breaking
and formation of Au–Au and Au–O bonds. It can be seen in the
optimized structures of the transition states shown in Fig. 5 that,
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

like in the mechanism starting from the bb conformation, the two


oxygen atoms are separated by less than 2 Å, and are forming two
Au–O–Au bridges. But in this case the two bridges are sharing
one gold atom, that is directly bonded to the two oxygen atoms.
After dissociation, the two oxygen atoms occupy stable 3-fold
coordinated hollow positions on the two particles considered, and
therefore the reaction is again exothermic.
The complete mechanism of O2 dissociation on supported gold
nanoparticles was calculated using the Au13 /TiO2 –Ovac model
to simulate the catalyst, and considering two different sites for
initial O2 adsorption: on the gold nanoparticle adopting a tbt
conformation, and at the metal–support interface (msi). The
reaction mechanism starting from the tbt conformer on the
Au13 /TiO2 –Ovac model is similar to that described for the isolated
Au13 cluster, and involves an activation barrier of 1.3 eV. Not
only the O–O bond but also several Au–O and Au–Au bonds are
Fig. 4 Schematic energy profile for O2 dissociation on Au(100).
modified in the process, and some restructuring of the particle
shape is observed in the optimized structure of the transition state
depicted in Fig. 6. The reaction is exothermic and after dissociation
the two oxygen atoms are situated on the gold nanoparticle, one
of them occupying a bridge position and the other one bonded
to three different gold atoms. It has been proposed that the most
active sites for O2 dissociation on supported gold nanoparticles are
located at the metal-support interface, and the results presented in
Table 2 confirm this proposal. The activation energy obtained
for the msi site is 0.20 eV lower than that calculated for the
reaction happening on the gold particle. Since the same model
and computational setup have been employed, direct comparison
of the calculated values is possible and meaningful. The reaction
path at the msi site does not only involve oxygen and gold atoms,
but also a titanium atom of the support. As described in previous
section, the two oxygen atoms of O2 molecule in the reactant
complex are interacting with a 5-fold coordinated Ti atom of the
oxide surface, and only one of them is directly bonded to a Au
atom of the particle. In the transition state, the intramolecular
O–O and one of the two Ti–O bonds are broken, while the Au–O
bond remains intact and no important restructuring is observed on

Fig. 5 Optimized structures of reactants (R), transition states (TS)


and products (P) involved in the dissociation of O2 on isolated gold
nanoparticles.

reaction energies are pronounced, and can be related to the


different geometrical arrangement of the O atoms in the reaction
product. While after dissociation the two O atoms occupy bridge
positions on the Au13 cluster, they are situated in the most stable
3-fold coordinated hollow positions on the Au38 nanoparticle, as
shown in Fig. 5.
The reaction path for dissociation of O2 initially adsorbed on
a tbt conformation is energetically less favourable, with calculated Fig. 6 Optimized structures of reactants (R), transition states (TS)
activation barriers equal to or larger than 1 eV in all cases. The and products (P) involved in the dissociation of O2 on supported gold
reaction coordinate describing this pathway is not only associated nanoparticles.

This journal is © The Royal Society of Chemistry 2010 Dalton Trans., 2010, 39, 8538–8546 | 8543
the gold nanoparticle. In the final state, a Ti–O–Au bond has been
formed at the interface, and only one oxygen atom is adsorbed on
a hollow position at the gold nanoparticle.
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

From these results it can be initially concluded that dissociation


of O2 involves low activation barriers (lower than 0.5 eV) when
the reactant molecule is adsorbed on (100) facets of gold in a
bridge-bridge mode. In this conformation, the charge transfer
from the metal surface to O2 and therefore the degree of molecular
activation is the highest. While the influence of particle size
on O2 dissociation by gold nanoparticles and the existence of
a critical size has been recently discussed,70 to the best of our
knowledge this is the first time that the structure-sensitivity of
this reaction is proposed. According to our results it seems that
O2 should easily dissociate on Au(100) single crystals, what is
contrary to experiment. However, it should be taken into account
that O2 dissociation and O2 desorption from the metal surface are
competitive processes, and the first one will only occur when the
activation energy is lower than the desorption energy. Analysis
of data in Table 2 allows us to conclude that dissociation is
energetically more favourable than desorption on isolated Au13
and Au38 nanoparticles, but not on Au(100), in agreement with the
experimental evidence that bulk gold is catalytically inactive. If the
same analysis is applied to the supported nanoparticles it is found
that, although the activation energy for O2 dissociation is higher
than 1 eV on the two cases studied, the adsorption/desorption
energy at the metal–support interface is considerably larger,
2.41 eV, and therefore the reaction will take place.
In order to find some general trends that could work both
for isolated and supported gold nanoparticles, the calculated
activation energies were plotted against a number of different
variables, and the most interesting relationships are shown in
Fig. 7. It should be mentioned that we do not observe a dependence
between activation and adsorption energies. Instead, there is a
linear relationship between the activation energy and the net
charge on adsorbed O2 , thus validating the hypothesis that the
degree of charge transfer determines the molecular activation.
The situation is different only in the case of O2 adsorbed at the
metal-support interface, where other interactions, such as Ti–O
bonds, exist. Finally, and most interesting from the point of view
of experimental characterization, a linear relationship has been
found between activation energy and nOO stretching frequency
of the adsorbed reactant, that is valid for all the catalyst models
studied in this work. The plot in Fig. 7 suggests that O2 molecules
whose nOO stretching frequency appears at values lower than
~900 cm-1 are sufficiently activated to dissociate. It can be argued
that in some cases, as for instance O2 adsorbed on Au(100), the
calculated nOO frequency is low, 750 cm-1 , but O2 desorption takes
place before dissociation. In these cases the condition is still valid,
because if O2 adsorption is weak and the molecule desorbs very
rapidly, the band corresponding to this vibrational mode will not
appear in the IR spectra.
Fig. 7 Correlation between calculated activation energies for O2 dissoci-
ation E act and adsorption energy E ads (top), total charge qO2 (middle) and
4. Conclusions
stretching frequency n(OO) (bottom) of the corresponding O2 adsorbed
Adsorption and dissociation of molecular O2 on different gold reactant species.
surfaces and nanoparticles has been theoretically investigated
in order to separate the effect of particle size, morphology and only existing on small particles with low-coordinated gold atoms
support on the mechanism of O2 activation. Three modes of O2 and which is not interesting from the catalytic point of view; (b) a
adsorption have been found: (a) a weakly interacting end-on mode top-bridge-top mode stable on all gold surfaces and particles, and

8544 | Dalton Trans., 2010, 39, 8538–8546 This journal is © The Royal Society of Chemistry 2010
(c) a bridge-bridge mode that requires the presence of four gold 23 A. Abad, A. Corma and H. Garcı́a, Chem.–Eur. J., 2008, 14, 212.
atoms arranged as in Au(100) surface. Adsorption energies roughly 24 M. D. Hughes, Y. J. Xu, P. Jenkins, P. McMorn, P. Landon, D. I.
Enache, A. F. Carley, G. A. Attard, G. J. Hutchings, F. King, E. H.
depend on particle size. The properties that could better reflect the
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

Stitt, P. Johnston, K. Griffin and C. J. Kiely, Nature, 2005, 437, 1132.


degree of molecular activation towards dissociation, such as total 25 T. A. Nijhuis, T. Visser and B. M. Weckhuysen, Angew. Chem., Int. Ed.,
electron density transfer from the catalyst to O2 , lengthening of the 2005, 44, 1115.
O–O bond and shift in the nOO stretching frequency, are clearly 26 A. M. Joshi, W. N. Delgass and K. T. Thomson, J. Phys. Chem. C, 2007,
111, 7841.
related among them and with the geometrical conformation of 27 J. J. Bravo-Juárez, K. K. Bando, J. Lu, M. Haruta, T. Fujitani and T.
adsorbed O2 . Thus, the highest degree of molecular activation is Oyama, J. Phys. Chem. C, 2008, 112, 1115.
always found in bridge-bridge conformers and, accordingly, the 28 A. S. K. Hashmi, J. P. Weyrauch, M. Rudolph and E. Kurpejovic,
Angew. Chem., Int. Ed., 2004, 43, 6545.
lowest activation barriers for O2 dissociation are also obtained 29 S. Carrettin, J. Guzman and A. Corma, Angew. Chem., Int. Ed., 2005,
for these species. Since bridge-bridge conformers are only stable 44, 2242.
on Au(100) facets, this indicates that particle morphology is a 30 C. González-Arellano, A. Corma, M. Iglesias and F. Sánchez, Chem.
key factor influencing O2 dissociation. On the other hand, when Commun., 2005, 3451.
31 C. González-Arellano, A. Abad, A. Corma, H. Garcı́a, M. Iglesias and
the gold nanoparticles are supported on TiO2 , the most stable F. Sánchez, Angew. Chem., Int. Ed., 2007, 46, 1536.
positions for O2 adsorption and the most active sites for O2 32 Q. Fu, H. Saltsburg and M. Flytzani-Stephanopoulos, Science, 2003,
dissociation are found at the metal-support interface, and not on 301, 935.
33 J. K. Edwards, B. Solsona, N. N. Edwin, A. F. Carley, A. A. Herzing,
the gold particle.
C. J. Kiely and G. J. Hutchings, Science, 2009, 323, 1037.
34 G. J. Hutchings, Chem. Commun., 2008, 1148.
35 J. Jia, K. Haraki, J. N. Kondo, K. Domen and K. Tamaru, J. Phys.
Acknowledgements Chem. B, 2000, 104, 11153.
36 T. V. Choudhary, C. Sivadinarayana, A. K. Dayte, D. Kumar and D. W.
The authors thank CICYT (MAT 2006-14274-C02-01), Goodman, Catal. Lett., 2003, 86, 1.
Consolider-Ingenio-2010 (project MULTICAT) and Fundación 37 M. Neurock and D. H. Mei, Top. Catal., 2002, 20, 5.
Areces for financial support, and Red Española de Supercom- 38 D. H. Mei, E. W. Hansen and M. Neurock, J. Phys. Chem. B, 2003,
107, 798.
putación (RES) and Centre de Càlcul de la Universitat de València 39 S. Schimpf, M. Lucas, C. Mohr, U. Rodemerck, A. Brückner, J. Radnik,
for computational resources and technical assistance. H. Hofmeister and P. Claus, Catal. Today, 2002, 72, 63.
40 C. Mohr, H. Hofmeister, J. Radnik and P. Claus, J. Am. Chem. Soc.,
2003, 125, 1905.
References 41 J. E. Bailie and G. J. Hutchings, Chem. Commun., 1999, 2151.
42 C. Milone, M. L. Tropeano, G. Guline, G. Neri, R. Ingoglia and S.
1 M. Haruta, T. Kobayashi, H. Sano and N. Yamada, Chem. Lett., 1987, Galvagno, Chem. Commun., 2002, 868.
405. 43 R. Zanella, C. Louis, S. Giorgio and R. Touroude, J. Catal., 2004, 223,
2 M. Haruta, Catal. Today, 1997, 36, 153. 328.
3 M. Valden, X. Lai and D. W. Goodman, Science, 1998, 281, 1647. 44 A. Corma and P. Serna, Science, 2006, 313, 332.
4 M. S. Chen and D. W. Goodman, Science, 2004, 306, 252. 45 M. Boronat, P. Concepción, A. Corma, S. González, F. Illas and P.
5 G. C. Bond and D. T. Thomson, Catal. Rev. Sci. Eng., 1999, 41, 319. Serna, J. Am. Chem. Soc., 2007, 129, 16230.
6 G. C. Bond and D. T. Thomson, Gold Bull., 2000, 33, 41. 46 A. Corma, P. Concepción and P. Serna, Angew. Chem., Int. Ed., 2007,
7 G. J. Hutchings, Gold Bull., 2004, 37, 3. 46, 7266.
8 G. J. Hutchings, Catal. Today, 2005, 100, 55. 47 Sh. K. Shaikhutdinov, R. Meyer, M. Naschitzki, M. Bäumer and H. J.
9 A. S. K. Hashmi and G. J. Hutchings, Angew. Chem., Int. Ed., 2006, Freund, Catal. Lett., 2003, 86, 211.
45, 7896. 48 C. Lemire, R. Meyer, Sh. K. Shaikhutdinov and H. J. Freund, Angew.
10 S. Carretin, P. Concepción, A. Corma, J. M. López-Nieto and V. F. Chem., Int. Ed., 2004, 43, 118.
Puntes, Angew. Chem., Int. Ed., 2004, 43, 2538. 49 M. Mavrikakis, P. Stoltze and J. K. Nørskov, Catal. Lett., 2000, 64,
11 J. Guzman, S. Carrettin and A. Corma, J. Am. Chem. Soc., 2005, 127, 101.
3286. 50 N. López and J. K. Nørskov, J. Am. Chem. Soc., 2002, 124, 11262.
12 B. Schumacher, Y. Denkwitz, V. Plzak, M. Kinne and R. J. Behm, 51 T. Jiang, D. J. Mowbray, S. Dobrin, H. Falsig, B. Hvolbæk, T. Bligaard
J. Catal., 2004, 224, 449. and J. K. Nørskov, J. Phys. Chem. C, 2009, 113, 10548.
13 N. Weiher, A. M. Beesley, N. Tsapatsaris, L. Delannoy, C. Louis, J. A. 52 E. Bus, J. T. Miller and J. A. van Bokhoven, J. Phys. Chem. B, 2005,
van Bokhoven and S. L. M. Schroeder, J. Am. Chem. Soc., 2007, 129, 109, 14581.
2240. 53 A. Corma, M. Boronat, S. González and F. Illas, Chem. Commun.,
14 H. Hakkinen, S. Abbet, A. Sanchez, U. Heiz and U. Landman, Angew. 2007, 3371.
Chem., Int. Ed., 2003, 42, 1297. 54 M. Boronat, F. Illas and A. Corma, J. Phys. Chem. A, 2009, 113, 3750.
15 G. J. Wang and B. Hammer, Phys. Rev. Lett., 2006, 97, 136107. 55 P. Serna, P. Concepción and A. Corma, J. Catal., 2009, 265, 19.
16 N. López, T. V. J. Janssens, B. S. Clausen, Y. Xu, M. Mavrikakis, T. 56 C. Mohr, H. Hofmeister and P. Claus, J. Catal., 2003, 213, 86.
Bligaard and J. K. Nørskov, J. Catal., 2004, 223, 232. 57 P. Claus, Appl. Catal., A, 2005, 291, 222.
17 I. N. Remediakis, N. Lopez and J. K. Nørskov, Angew. Chem., Int. Ed., 58 T. Diemant, Z. Zhao, H. Rauscher, J. Bansmann and R. J. Behm, Top.
2005, 44, 1824. Catal., 2007, 44, 83.
18 N. C. Hernández, J. F. Sanz and J. A. Rodriguez, J. Am. Chem. Soc., 59 K. P. McKenna and A. L. Shluger, J. Phys. Chem. C, 2007, 111,
2006, 128, 15600. 18848.
19 T. V. W. Janssens, B. S. Clausen, B. Hvolbaek, H. Falsig, C. H. 60 M. Boronat, P. Concepción and A. Corma, J. Phys. Chem. C, 2009,
Christensen, T. Bligaard and J. K. Nørskov, Top. Catal., 2007, 44, 113, 16772.
15, and references therein. 61 J. D. Grunwaldt, M. Maciejewski, O. S. Becker, P. Fabrizioli and A.
20 M. Comotti, C. Della Pina, R. Matarrese and M. Rossi, Angew. Chem., Baiker, J. Catal., 1999, 186, 458.
Int. Ed., 2004, 43, 5812. 62 M. Maciejewski, P. Fabrizioli, J. D. Grunwaldt, O. S. Becker and A.
21 D. I. Enache, J. K. Edwards, P. Landon, B. Solsona-Espriu, A. F. Carley, Baiker, Phys. Chem. Chem. Phys., 2001, 3, 3846.
A. A. Herzing, M. Watanabe, C. J. Kiely, D. W. Knight and G. J. 63 F. Bocuzzi, A. Chiorino, M. Manzoli, P. Lu, T. Akita, S. Ichikawa and
Hutching, Science, 2006, 311, 362. M. Haruta, J. Catal., 2001, 202, 256.
22 A. Abad, C. Almela, A. Corma and H. Garcı́a, Chem. Commun., 2006, 64 Th. Risse, Sh. K. Shaikhutdinov, N. Nilius, M. Sterrer and H. J. Freund,
3178. Acc. Chem. Res., 2008, 41, 949.

This journal is © The Royal Society of Chemistry 2010 Dalton Trans., 2010, 39, 8538–8546 | 8545
65 M. M. Schubert, S. Hackenbertg, A. C. van Veen, M. Muhler, V. Plzak 76 G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter, 1996, 54,
and J. Behm, J. Catal., 2001, 197, 113. 11169.
66 Z. P. Liu, P. Hu and A. Alavi, J. Am. Chem. Soc., 2002, 124, 14470. 77 G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter, 1993, 47, 558.
Published on 29 May 2010. Downloaded by IInstituto Ciencia y Tecnologia del Carbon on 4/30/2024 9:46:01 AM.

67 G. J. Wang and B. Hammer, Top. Catal., 2007, 44, 49. 78 P. E. Blöchl, Phys. Rev. B: Condens. Matter, 1994, 50, 17953.
68 S. Laursen and S. Linic, J. Phys. Chem. C, 2009, 113, 6689. 79 H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Solid State, 1976, 13,
69 G. Mills, M. S. Gordon and H. Metiu, J. Chem. Phys., 2003, 118, 5188.
4198. 80 A. Heyden, A. T. Bell and F. J. Keil, J. Chem. Phys., 2005, 123, 224101.
70 A. Roldan, S. González, J. M. Ricart and F. Illas, ChemPhysChem, 81 G. Henkelman and H. Jónsson, J. Chem. Phys., 1999, 111, 7010.
2009, 10, 348. 82 E. Sanville, S. D. Kenny, R. Smith and G. Henkelman, J. Comput.
71 M. Chen, Y. Cai, Z. Yan and D. W. Goodman, J. Am. Chem. Soc., 2006, Chem., 2007, 28, 899.
128, 6341. 83 G. Henkelman, A. Arnaldsson and H. Jónsson, Comput. Mater. Sci.,
72 L. M. Molina and B. Hammer, Appl. Catal., A, 2005, 291, 21. 2006, 36, 354.
73 M. Kotobuki, R. Leppelt, D. A. Hangsgen, D. Widmann and R. J. 84 F. Tielens, J. Andrés, M. Van Brusssel, C. Buess-Herman and P.
Behm, J. Catal., 2009, 264, 67. Geerlings, J. Phys. Chem. B, 2005, 109, 7624.
74 J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, 85 K. P. Huber, G. Herzberg, in Molecular Spectra and Molecular Structure
D. J. Singh and C. Fiolhais, Phys. Rev. B: Condens. Matter, 1992, 46, IV. Constants, of Diatomic Molecules, Reinhold, Van Nostrand, New
6671. York, 1987.
75 J. P. Perdew and Y. Wang, Phys. Rev. B: Condens. Matter, 1992, 45, 86 J. A. Rodriguez, L. Feria, T. Jirsak, Y. Takahashi, K. Nakamura and
13244. F. Illas, J. Am. Chem. Soc., 2010, 132, 3177.

8546 | Dalton Trans., 2010, 39, 8538–8546 This journal is © The Royal Society of Chemistry 2010

You might also like