Applied Catalysis B: Environmental 197 (2016) 299–312

Contents lists available at ScienceDirect

This manuscript the Accepted Manuscript version of a Published Work license http://creativecommons.org/licenses/by-nc-nd/4.0/.
This document isversion is made available under the CC-BY-NC-ND 4.0that appeared in final form in Applied Catalysis B:
Link to final version: http://dx.doi.org/10.1016/j.apcatb.2016.02.035.
Environmental 197 (2016) 299–312. http://dx.doi.org/10.1016/j.apcatb.2016.02.035

Applied Catalysis B: Environmental

journal homepage: www.elsevier.com/locate/apcatb

Reactivity descriptors for ceria in catalysis
Marcal Capdevila-Cortada a , Gianvito Vilé b , Detre Teschner c,∗ ∗ ∗ , Javier Pérez-Ramírez b,∗∗ ,
¸
Núria López a,∗
a
b
c

Institute of Chemical Research of Catalonia (ICIQ), The Barcelona Institute of Science and Technology, Av. Països Catalans 16, 43007 Tarragona, Spain
Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, Vladimir-Prelog-Weg 1, 8093 Zurich, Switzerland
Fritz-Haber-Institute of the Max Planck Society, Faradayweg 4-6, 14195 Berlin, Germany

a r t i c l e

i n f o

Article history:
Received 14 December 2015
Received in revised form 2 February 2016
Accepted 16 February 2016
Available online 20 February 2016
Keywords:
Ceria
Structure-activity relationships
Oxidation
Hydrogenation
Acid-base
Redox

a b s t r a c t
Ceria has been very successfully employed in oxidation catalysis, whereas its application in other reactions has been less intensively investigated. The catalytic activity of ceria can be further enhanced by the
use of dopants, and it exhibits structure sensitivity for numerous processes. The rich chemistry of cerium
oxide is gathered and discussed in the present work, where the nature of each step of the most common
reactions performed on it is assessed. Chemically intuitive computational and experimental descriptors,
namely acid-base, redox, and structural features, are put forward to correlate the observed trends among
the different doped and undoped facets. We have attempted to generate a robust framework that maps the
chemically sound descriptors to the experimental fingerprints and theoretically calculated parameters.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
Ceria has reached an iconic status among binary oxides, as it
keeps more secrets than any other due to the versatile chemical
structure, in addition to the fact that although being a rare earth
it is abundant and relatively cheap [1]. Ceria is an excellent playground to understand complex chemistry, as (i) it shows marked
structure sensitive properties [2], which can be nowadays assessed
through shape controlled synthesis [3–5]; (ii) it shows both acidbase and redox chemistry [6,7]; (iii) it presents oxygen transport
properties [8,9]; (iv) it is stable under harsh conditions [10]; (v)
it is relatively easy to modify by doping to improve its chemistry
[11,12]; (vi) it presents interesting hydrophobic features [13,14];
and (vii) it also exhibits a complex electronic structure that makes
modeling efforts challenging [15,16]. All these characteristics make
ceria as a unique material with fascinating properties and a wide
scope of applications in heterogeneous catalysis [1,17].
The exploitation of the catalytic and chemical properties of CeO2
has been commonly based on its oxygen storage capacity, OSC

∗ Corresponding author.
∗∗ Corresponding author.
∗ ∗ ∗Corresponding author.
E-mail addresses: teschner@fhi-berlin.mpi.de (D. Teschner), jpr@chem.ethz.ch
(J. Pérez-Ramírez), nlopez@iciq.es (N. López).
http://dx.doi.org/10.1016/j.apcatb.2016.02.035
0926-3373/© 2016 Elsevier B.V. All rights reserved.

[11,15,18,19]. A good OSC material requires a few contributions: (i)
the cycling between the oxidized and reduced forms is energetically
not demanding (or, equivalently, it presents low vacancy formation
energy); (ii) easy transport of the vacancies (O atoms) between the
surface and the bulk and inside the bulk; and (iii) facile transformation between the crystal structures of the oxidized (CeO2 ) and
reduced (Ce2 O3 ) forms or, at least, facile alignment of vacancies
through defect lines. The OSC has been instrumental to understand
the chemistry, either as a catalyst or a support, in many oxidation
processes [7,19]. Due to its OSC, ceria is also employed as a ceramic
electrolyte in solid oxide fuel cells, typically doped with aliovalent
cations [20–23].
Since the identification of these properties, numerous reactions
relying on the high oxidation ability of ceria-based materials have
been put forward. For instance, CeO2 can oxidize CO and soot,
especially at medium temperatures, and thus it is incorporated
in three-way catalysts in combination with metallic nanoparticles [7,19]. Ceria also performs nicely in the preferential oxidation
(PROX) of CO in CO:H2 mixtures [24–26]. This reaction is required
to clean H2 feeds from impurities that can damage fuel cells if H2
is employed as an energy vector. In addition, it can also drive the
water-gas shift reaction, i.e. the production of CO2 and hydrogen
from a mixture of CO and water, thus improving the hydrogen content in syngas mixtures [26–28]. Furthermore, oxidation properties
can be extended to the application in the Deacon process [29]. This
reaction aims at the recycling of hydrogen chloride by employing

300

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Fig. 1. Schematic representation of the sets of frequent applications of CeO2 in catalysis, manifesting common steps in the reaction networks. The related inverse steps are also
highlighted. The colors of the reactions correspond to Brønsted acid-base (red), Lewis acid-base (black), redox (blue), combined acid-base redox (brown), and condensation
(green) steps. For more details see the corresponding sections below. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article).

the following process 4HCl + O2 → 2Cl2 + 2H2 O. The reaction can be
seen as an oxidation and thus close to the traditional uses of CeO2 ,
but at the same time the competition of Cl and O for the same type
of sites limits the application of the common descriptors for oxidation processes [10,30–33]. Still, limiting the catalytic properties
of ceria to its use in oxidation is far a simplistic identification, as it
was recently shown to drive the semi-hydrogenation of alkynes to
alkenes under hydrogen-rich conditions [34–36]. The description
of this catalytic versatility sets a new landmark in the research of
these materials.
Perhaps not so widespread but still largely studied are other
reactions performed over ceria, mostly involving simple or polyfunctional organic molecules. The conversion of alcohols, e.g.
methanol [37–41], ethanol [42,43], or polyols [44–46], has been
carried out on ceria, as well as other organic compounds [47–51].
Methanol is also used in the conversion of CO2 to dimethyl carbonate (DMC) catalyzed by ceria [52–55]. Finally, NOx [56–58], SOx
[59–64], and H2 S [65–67] react on ceria leading to either oxidized
or reduced products depending on the reactant and the exposed
ceria facet.
Fig. 1 illustrates the intricate map of the most common reactions
catalyzed by CeO2 . The links between the different steps and the
identification of the chemical nature of each step are also shown.
A coherent systematization of the catalytic properties of ceria has
not been yet undertaken, although many authors have pointed out
the need of separating the different terms as acid/base, redox, or
condensation [68–73]. We herein aim at revising these applications and to identify the principles that guide the performance of
CeO2 in catalysis. For this purpose, a set of activity and selectivity
descriptors both for oxidations and hydrogenations are presented,
and the detailed analysis is then applied to describe: (i) a set of
relevant reactions; (ii) several facets that account for the strong
structure sensitivity observed; and (iii) the effect of introducing
dopant cations on the different properties. This integrative perspective identifies the descriptors and maps them to direct observables
accessible from experiments.

2. Descriptors for ceria
2.1. Previous descriptor-based oxidative chemistry
Structure-performance relationships are crucial to rationalize
catalytic phenomena [74–76]. Typically, if the rate, selectivity, or
stability correlate with a single parameter or a group of parameters these are taken as descriptors, as they can express the activity
summarizing all the information for a class of materials. Energy
descriptors have been extensively used since the work of Sabatier
[77], and are nowadays accepted as the most robust theoretical
tools in high-throughput computational screening of materials.
Reactions become more complex on oxides, and thus the assessment of their catalytic performance requires complex descriptors.
In particular, to describe the activity and selectivity of oxides, especially in oxidation processes, seven pillars were postulated in the
pioneering work by Grasselli [78]: host structure, redox, metaloxygen bond, lattice oxygen, phase cooperation, multifunctionality
of active sites, and site isolation, as illustrated in Fig. 2. These pillars, despite providing clear conceptual guidelines, are described in
a qualitative manner and can hardly be mapped directly to particular fingerprints in experiments. Thus, this list may be considered
as a set of general principles, rather than quantitative parameters
that can be adjusted to reach the desired reactivity. Furthermore,
some pillars are not orthogonal and hence they cannot be individually tailored, hindering the direct interpretation of the described
chemical process.
The oxygen vacancy formation energy has been commonly
taken as the primary, and sometimes single, descriptor for reactivity in computational approaches, restricting the pillars into a single
property. However, not all reaction schemes involving vacancy
formation can be described with this magnitude, especially for reaction environments where the surface structure differs significantly
from the pristine surface [31]. Moreover, hydrogenation reactions
do not entail vacancy formation, presenting a complementary set

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Fig. 2. Schematic illustration of Grasselli’s seven pillars, based on Ref. [78]. Notice
that although the pillars seem parallel, the descriptors overlap to some degree.

of governing factors and complex steps that couple acid-base and
redox properties.
Although the correspondence between the pillars and experimental or computational descriptors is not one-to-one, the
concepts described in the oxidation chemistry by Grasselli are fully
expanded in Table 1. In addition, one must keep in mind that the
state of the surface under reaction conditions is usually very particular in terms of stoichiometry, configuration, electronic structure,
etc. Descriptors have been traditionally proposed to describe the
pristine surface, instead of the catalyst state under working conditions, and thus they might be irrelevant to the activity if they are
not self-consistently analyzed.
2.2. Chemical descriptors: mapping experiments and theory
In the following paragraphs we present an alternative
descriptor-based formulation, which can be translated into the
pillars. The suggested descriptors are quantified contributions
retrievable often from both calculations and experiments, which
can be typically classified as electronic or geometric. However,
the sparse characterization derived from experiments, especially
related to probe molecules (formation of undesired species, dissociation, adsorption in multiple sites, etc.), advocates the use of
first-principles calculated descriptors, given their easier control of
the single desired property.
The electronic descriptors include:
The charges associated to the surface atoms (qO , qCe ). Oxides
present charge separation so that the first interaction between
a molecule and the surface is the electric field generated by the
point charge distribution. We have computationally obtained them
from a Bader analysis [79]. However, charges are not observables
and hence they cannot be retrieved from experiments. The spinpolarized density, on the other hand, is an observable that can be
measured by electron paramagnetic resonance (EPR) spectroscopy,
and sometimes might correlate with the charges.
The basicity of the surface anions (O2p ), which has two main
contributions: the total charge (as defined above) and the levels
alignment with respect to the adsorbate, or, alternatively, to the
Fermi level, which would correspond to how available (polarizable)
are the oxygen lone pairs to share their density. These definitions
include both Brønsted and Lewis basicity. The basicity term can
be traced back from temperature programmed desorption (TPD)
peaks of probe compounds or from shifts in their vibrational spectra. Equations that link the TPD peaks to adsorption energies can be
used with this purpose [80]. Computationally it is obtained by their
adsorption energy or, alternatively, as the center weighted average
in the projected density of states (PDOS) of the surface oxygen 2p
band. The Lewis acidity of the cation centers (ECO ) can be determined in a similar manner, using CO or NH3 as probe molecules

301

[81]. Yet, another point regards the presence of hydroxyl groups on
the surface, which correspond to Brønsted acid sites. Experimentally, the degree of control of the surface hydroxyl groups is a crucial
parameter that depends on the sample preparation and activation.
However, this is typically not considered in theoretical simulations,
as the clean surface is taken as the common initial structure.
The redox term (Ered ), which for the particular case of ceria is
based on the ability of the cations to accept one electron [82]. This
term is related to the OSC and can be obtained from experiments,
but it should not be considered as a standalone Ered contribution.
The Ered term is computationally obtained by adding one extra electron to the unit cell, ensuring the proper localization [83]. Notice
that this is equivalent to the reduction potentials associated to the
half-reaction of the total redox process. It is worth noting that
whenever aliovalent dopants are present, hole or particle states
appear. This may drastically change the Ered contribution, especially
for hole states generated within the band gap. The hole/particle
states can be identified by their charge, electron density, and the
corresponding energy.
The geometric terms related to the host structure contain:
The distances between surface atoms (di,j , where i and j are
atoms in the lattice). Distances are related to the bulk structure
and, for the most common terminations, surface effects do not alter
them significantly. Dopants might induce uneven and sharp expansions/contractions, the higher the difference is in the radii of the
parent and dopant atoms, the larger this volume change.
The local coordination of surface anions and cations (NO , NCe ).
This term can be retrieved from experiments by scanning tunneling microscopy (STM) or extended X-ray absorption fine structure
(EXAFS). Structure sensitivity is typically a direct consequence of
the changes in the distances or the coordination patterns. Notice
that, in some cases such as polar surfaces (e.g. the (100) facet of
ceria), the fraction of intrinsic vacancies, xᮀ , due to surface reconstruction will have a major impact on the local coordination.
The previous two properties and the symmetry patterns can be
combined into the term ensemble, this is the geometry of the active
site. On oxides, these effects are typically linked to multifunctional
positions, where both contributions from the anion and the cation
exist and are linked to their relative topology.
Two final aspects discussed by Grasselli are phase cooperation
and site isolation. The former term was said to appear when the
key catalytic functions cannot be incorporated into one single catalytic phase, and thus two or more phases are required in close
proximity. The appearance of different phases upon reaction is
typically reported by X-ray diffraction (XRD). In some cases, the
tridimensional structure does not change but the material takes
some of the reactants up, converting a pure oxide phase for instance
into an oxyhalide, as seen for the oxidation of hydrogen halides
by operando prompt gamma-ray activation analysis (PGAA). In a
broader sense, reaction fronts on metals might be seen as phase
cooperation, where under some conditions the reaction takes place
in areas where two different compounds or surface states are in
close proximity [84,85]. This has been found to a certain extent
in theoretical simulations, where the areas of maximum activity
are linked to positions where the two reactants are more likely to
be mixed [86,87]. An alternative phase cooperation can be found
for ceria supported metal catalysts, where bifunctional mechanism
with one part of the reaction network taking place on the metal
and another set of reactions occurring in the oxide would also
belong to this category. Site isolation contributions are typically
crucial to assess selectivity as they govern the activity of the sites.
It can be experimentally inferred by either analyzing side products
in catalytic tests or performing adsorption tests of molecules that
simultaneously probe two centers. Computationally, it can usually
be described as a combination of the coordination number and the
surface distances mentioned above. The descriptors corresponding

302

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Table 1
Experimental and theoretical mapping of the seven pillars suggested by Grasselli [78]. The abbreviations used can be found in the text.
Pillar

Property

Experimental technique

Computational descriptor

Host structure

Tridimensional structure
Local coordination
Energy

XRD, neutron scattering
EXAFS, STM
Raman

Geometric

EXAFS
TPR, XPS, UPS, EPR
cyclic voltammetry
OSC
OSC
TPD (CO2 ) + IR
TPD (CO, NH3 )
Selectivity assessment by
catalytic testinga
TPD of multifunctionalized
molecules, FTIR of probe
molecules

Distances, topology
Local coordination
Cohesive energy
Vacancy formation energy
Geometry
Ionization potential
Electron affinity
Vacancy formation energy
Ea diffusion
O2p , ECO2
ECO
Lateral interactions

M O bond

Redox
Lattice oxygen

Vacancy formation
Diffusion
Basicity
M-related Acidity

Site isolation
Multifunctionality of active sites

Phase cooperation

a

Different phases
Uptake

XRD
PGAA

Adsorption energy for different
configurations sampling sites
Acid-base/geometry descriptor
combinations
Equilibrium diagrams
Resting state first principles
thermodynamics

Site isolation at the catalytic surfaces might be proposed based on selectivity assessment in partial hydrogenation and oxidation reactions.

to phase cooperation and site isolation will be only briefly touched
upon in the present work.
Additionally, the oxygen vacancy formation energy (Eᮀ ) has
been usually considered as the simplest descriptor for the activity
of ceria-based catalysts, as previously stated. We have therefore included it in our approach. Vacancy formation leaves two
electrons in the hole left by oxygen removal, which must be
accommodated by the surface [88]. In ceria, the electrons are
hosted by the reducible cations, which tend to be localized in
the next-nearest neighbors to the vacancy [89]. This induces
lattice distortions due to the larger radius of the Ce3+ cations.
The vacancy formation energy is obtained from calculations as
Eᮀ = E(CeO2 - x ) + 1/2E(O2 ) − E(CeO2 ).
An alternative descriptor related to oxygen vacancies is the
experimental oxygen storage capacity (OSC). This magnitude is
related to the formation of oxygen vacancies and their diffusion
in the material toward/from the surface. The OSC is measured as
the weight loss of the sample due to oxygen removal, commonly
using CO or H2 , and it depends on the history of the sample (i.e.
on how the sample was prepared and its subsequent treatments).
Therefore, the sample can be tailored to reach desired OSC values. However, the experimental measurements strongly depend
on the protocols applied, and thus reproducibility requires very
strict degree of control and a strict standardization in order to compare data among laboratories. The computational assessment of
the OSC is much more complex. In principle, an extensive investigation for the equilibrium oxygen loss could be done by coupling
DFT to the environment through first-principles thermodynamics
[90]. However, the intrinsic approximations included in the latter
methodology prevent from obtaining quantitative correct values
even for the equilibrium oxygen vacancies [91].
Finally, it is worth noting that the state of the surface, in particular under reaction conditions, could affect both the pillars and
the descriptors. The analysis presented here is devoted to the evaluation of the clean or slightly modified surface, i.e. not allowing
large coverage effects or significant surface reconstructions. Coverage effects, however, might influence the values of the descriptors
to a large extent. Nevertheless, our aim in this work is to present the
proof of concept of the methodology, and this is easier to carry out
under clean surface conditions. Coverage-dependent descriptors,
for the most common coverage situations, could be introduced in
a simple yet efficient manner to study the role of coadsorbates on
the reactivity. Furthermore, the only surface defects considered in

this study are point defects, namely oxygen vacancies and dopant
cations. Other defects such as steps or low-coordinated centers
(e.g. edges or corners) are not considered, given the relatively large
nanoparticles on which the following reactions are typically performed. Low-coordinated sites can however be treated in the same
methodology described here.
Descriptors are related to the characteristics of the particular
reaction networks. Therefore, a detailed knowledge of the elementary steps provides an easier identification. In the following, we
have employed reaction networks reported in the literature, exception made for the DMC synthesis case, where a potential reaction
network is proposed. The main reaction pathways in complex reaction networks depend on the reaction conditions, which affect the
relative stability of the difference phases present and in some cases
the dynamics of these phases. From a theoretical point of view,
these phases need to be evaluated separately and if they cooperate then the model shall include the boundary. In the cases studied
herein, ceria maintains its properties unless indicated.
2.3. Descriptors for cerium oxide reactivity
Most of the reactions discussed in this work are performed over
ceria nanoparticles, Fig. 3, although some examples over extended
films are also examined. The (1 1 1) termination is the most stable ceria surface, and it is the exposed facet of well-defined ceria
nanoctahedra. In addition to this nanoshape, ceria nanorods and
nanocubes are also routinely synthesized [3–5]. The former mainly
expose (1 1 0) facets, with (1 0 0) and (1 1 1) to a lesser extent,
whereas the latter present (1 0 0) termination planes. The (1 0 0)
surface is polar and thus requires polarity compensation reconstruction that may lead to Ce- or O-terminations [93]. Commonly
the latter termination is considered as the most stable (1 0 0) reconstruction and thus it is the one assumed in this study [94–98].
Although the environment might particularly affect polar surfaces,
where adsorbates can promote certain reconstructions [99,100],
our approach is limited to the pristine oxygen terminated reconstruction. It is worth noting that the different exposed facets are of
particular importance given the structure sensitivity observed for
many processes [2].
The descriptors listed in the previous section are collected in
Table 2, for the (1 1 1), (1 1 0), and (1 0 0) ceria surfaces. The most
direct descriptor is the local coordination of the topmost cerium
and oxygen atoms. The (1 1 1) surface presents higher coordination

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

303

Fig. 3. Structure and composition of ceria nanoctahedra (left), nanorods (middle), and nanocubes (right). The micrograph of ceria nanoctahedra is adapted from [92]. Color
codes: O (red) and Ce (yellow). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

for both atoms, NCe = 7 and NO = 3, whereas the former decreases
to 6 in both (1 1 0) and (1 0 0), and NO lowers to 2 in the (1 0 0), but
remains 3 in the (1 1 0). A straight consequence of NO is the CeO distance (dCe,O ), which experiences a minor difference between
the (1 1 1) and the (1 1 0) surfaces but decreases significantly on the
(1 0 0) facet. The dO,O , on the other hand, is equivalent in the (1 1 1)
and (1 0 0) surfaces, but two shorter distances exist in the (1 1 0) as a
result of the morphology of this termination plane (Fig. 3). Although
the charges of surface Ce and O do not change significantly, the
other electronic descriptors show more drastic differences among
the surfaces. The basicity term is much higher and nearly identical
for the (1 1 1) and (1 0 0), and lower for the (1 1 0) surface. However, the acidity of the cerium atoms increases following the order
(1 0 0) > (1 1 0) > (1 1 1), inverse to their stability order. The Ered term
is endothermic in all cases, being 0.65 and 1.65 eV for the (1 1 1) and
(1 1 0), respectively, and almost zero for the (1 0 0). This parameter
is completely connected to how the surface can accommodate the
bigger Ce3+ cations. The low NCe and NO coordinations in (1 0 0)
favor the required rearrangement to incorporate the Ce3+ . However, in the (1 1 0) surface, since each Ce4+ is surrounded by four O in
the surface plane (Fig. 3), the accommodation of the bigger reduced

atom is less favored. Finally, the endothermic vacancy formation
energy is higher for the (1 1 1), followed by the (1 0 0) and the (1 1 0)
surfaces. The latter termination further stabilizes the vacancy formation by bridging an oxygen atom between two cerium cations
as a split vacancy [101,102]. It is worth noting that on the (1 1 1)
surface, the lowest vacancy formation energy corresponds to the
subsurface vacancy [15], whereas this is not the case for the other
two low-index surfaces. Notice also that since the (1 0 0) is a polar
surface, reconstruction is required to compensate the polarity and
render it stable [93,103]. Such reconstruction leaves a fraction of
0.5 of intrinsic oxygen vacancies, xᮀ [104]. This fact also enables
the surface to adsorb oxygen atoms on these vacancy sites, which
is an endothermic process but still more favorable than vacancy
formation.
Similar to Table 2, Table 3 contains the descriptor values for
Ga-, Zr-, and V-doped ceria(1 1 1). Ga- and V-CeO2 (1 1 1) are examples of aliovalent dopants (n- and p-type, respectively), whereas
Zr-CeO2 (1 1 1) is an example of isovalent doping, Fig. 4. The addition
of dopant cations induces symmetry breaking, which drastically
influences some specific descriptor values. On the one hand, cerium
and oxygen charges, local coordinations, and surface acidity remain

304

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Table 2
General descriptors for the lowest surface energy facets of CeO2 . The charges are in |e− |, distances in Å, and energies in eV (negative values correspond to exothermic
processes).
(1 1 1)

(1 1 0)

(1 0 0)

Calc.

Exp.

Calc.

Exp.

Calc.

Exp.

–
–
–
–

6
2
2.206
3.887

–
–
–
–

–
–
0.9a
wb
–
–

2.33
−1.12
−0.93
−0.34
0.03
1.57
(0.99)c
3.21

–
–
0.3a
wb
–
–

–

353d
6.8e
–
–

NCe
NO
dCe,O
dO,O

first neighbors
first neighbors
metal-oxygen distances
oxygen-oxygen
distances

7
3
2.375
3.886

–
–
–
–

qCe
qO
O2p
ECO
Ered
Eᮀ

ion charge
ion charge
oxygen basicity
CO adsorption
surface reduction
vacancy formation
energy surface
vacancy formation
energy subsurface
oxygen storage
capacity
intrinsic defective fraction
spin density

2.37
−1.19
−0.95
−0.19
0.65
1.85

–
–
0.1a
wb
–
–

6
3
2.342
2.604
2.898
3.887
2.31
−1.16
−1.45
−0.24
1.16
1.17

1.68

–

2.22

–

–

318d
0.4e
–
–

–

554d
6.3e
–
–

Eᮀ , ss
OSC
xᮀ
␳spin
a
b
c
d
e

0
0

0
0

0.5
0

–

CO2 adsorption rate (in CO2 nm−2 ), from [71].
w = weak, from [71].
In parentheses: energy requirement upon the addition of an oxygen atom to the stoichiometric surface.
In ␮moles O g−1 , from [109].
In ␮moles O m−2 , from [110].

Fig. 4. Electron localization of holes or particles in Ga-, Zr-, and V-doped ceria(1 1 1). The small insets illustrate the resulting band structure for each doped surface, where
the hole in the valence band and the particle in the band gap close to the conduction band are qualitatively shown for the Ga- and V-CeO2 materials, respectively.

almost unaltered. On the other hand, doping has severe impact
on interatomic distances, oxygen basicity, redox, and vacancy formation energy. The average dCe,O and dO,O distances are largely
unaffected, but the values close the dopant site can vary for dCe,O
ca. 0.1, 0.05, and 0.3 Å for Ga-, Zr-, and V-ceria, respectively, and ca.
0.5, 0.4, and 1.1 Å, respectively, for dO,O . The basicity term decreases
for the three doped surfaces compared to the pristine CeO2 (1 1 1),
strongly on the p-doped surface. The addition of n- or p-type
dopants triggers a singular change on the electronic structure of the
semiconductor material, creating a hole or a particle state, respectively. This has major consequences on Ered and Eᮀ terms. Although
for Zr-CeO2 (1 1 1) both descriptors decrease in a notable manner,
for Ga- and V-CeO2 (1 1 1) this decrease is much larger, and Ered is
even highly exothermic for the n-doped surface. Interestingly, subsurface vacancy formation is still favorable for the isovalent doping,
but turns unfavorable for Ga- and V-doping.
The descriptors listed in Tables 2 and 3 will serve as the framework to describe the most common reactions performed on ceria.
The computational set up for the calculations is summarized in the
Supplementary Material, while the details on the experiments can

be found on the particular references. Nevertheless, this list enables
us to start providing a quantitative description of the most recurrent term related to the reactivity on ceria, the vacancy formation
energy, based on the systems previously detailed, which include the
three pristine and the three doped surfaces. The vacancy formation
energy is found to be a combined balanced function of the basicity of
the surface oxygens and the redox ability of the surface, O2p and Ered
(Fig. 5). Therefore, Eᮀ depends simultaneously on the stability of the
surface oxygen and on how well the surface can accommodate the
electrons left by the oxygen removal. Both descriptors are individually unable to correlate Eᮀ , however their combination results in
a collective descriptor that provides a notable description of the
vacancy formation energy. In particular, the relative weigh of each
variable in the collective descriptor is 0.58 for Ered and 0.42 for O2p ,
once balanced by the span of each variable (Ered and O2p span 2.12
and 0.52 eV, respectively, see Tables 2 and 3).
Collective descriptors are sometimes crucial to describe certain
complex steps or adsorptions [105–108]. These descriptors are not
observables and do not have an additional physical meaning other
than the proper balance of the individual contributions. It is also

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312
Table 3
General descriptors for p-, isovalently-, and n- doped CeO2 (1 1 1) surfaces, following
the nomenclature given in Table 2. The charges are in |e− |, distances in Å, and energies in eV (negative values correspond to exothermic processes). Distance values
refer to minimum, average, and maximum distance, respectively.

a

Table 4
Adsorption energies (in eV) of O2 , CO, CO2 , C2 H2 , and C2 H4 on the clean and defective
(1 1 1), (1 1 0), and (1 0 0) CeO2 surfaces.

Zr-CeO2 (111)

V-CeO2 (111)

7
4
3a
2.277/2.379/2.519
1.940
3.345/3.892/4.434
2.38
1.62
−1.14
−1.14
−0.22
−0.04
−0.96
0.20
0.29
0
1.0 (O• )

7
4
3
2.344/2.399/2.455
2.135
3.469/3.883/4.106
2.39
2.51
−1.20
−1.04
−0.22
−0.12
0.40
1.42
1.30
0
0

7
4
3a
2.145/2.417/2.840
1.715
2.890/3.911/4.489
2.34
1.89
−1.13
−1.26
−0.21
−0.05
0.01
0.60
0.97
0
1.0 (Ce3+ )

(111)

O2
CO
CO2
C2 H2
C2 H4
a
b
c

(110)

(100)

clean

Ga-CeO2 (111)
NCe
NM
NO
dCe,O
dM,O
dO,O
qCe
qM
qO
O2p
ECO on Ce
ECO on M
Ered
Eᮀ
Eᮀ , ss
xᮀ
␳spin

305

defective clean

defective clean

defective

−0.02
−0.20
−0.47a
−0.08
−0.30b
−0.07
0.28b

−1.86
−0.19
−0.99a
–

−1.03
−0.15
−0.93a
–

−2.22
−0.32−1.58
−2.65a
–

–

−0.03
−0.24−3.61a
−1.33a
−0.23
−2.43c
−0.22
−0.47c

–

−0.08
−0.34−4.34a
−1.98a
−0.27
−4.05c
−0.14
−2.76c

–

Chemisorbed via carbonate formation.
Chemisorbed via O-[R]• radical formation.
Chemisorbed via O-[R]-O formation.

Fig. 5. Dependence of the vacancy formation energy, Eᮀ , on a collective variable
including the O2p and Ered terms, for the pristine (1 1 1), (1 1 0), (1 0 0), and doped
Ga-, Zr-, V-(1 1 1) ceria surfaces. The fitted equation is f(O2p , Ered ) = (2.33 O2p + 0.78
Ered + 3.60) eV.

worth noting that the coefficients of each variable may be expressed
in the right units to correctly describe the correlated magnitude.
3. Catalytic applications of ceria
3.1. CO oxidation and water-gas shift reaction
The simplest reaction that has been studied over CeO2 is CO
oxidation [69,111]. Notice that ceria is also a suitable material for
soot oxidation, which shares some of the characteristics of the CO
oxidation scheme [110]. This reaction can be described based on
the following list of elementary steps, where * stands for the Lewis
center and ᮀ represents the defect site:
O2 + ᮀ → O2ᮀ

(1)

O2ᮀ + ᮀ → 2Oᮀ

(2)

CO + ∗ → CO∗

(3)

(4)

CO ∗ + Oᮀ → CO2ᮀ

(5)

CO2 ∗ → CO2 + ∗

(6)

CO2ᮀ → CO2 + ᮀ

One of the surface oxygens adjacent to the dopant site has coordination 2.

CO ∗ + O2ᮀ → CO2 ∗ + O

(7)

Two types of active sites are required. In fact, CO adsorption
takes place on the Lewis acid centers of the ceria surface, the Ce4+
sites. In parallel, O2 does not adsorb on the clean surface (Table 4),
since the electron transfer to O2 is not favored but only on very particular sites of the surface, i.e. oxygen vacancies, line defects such
as steps, and open surfaces like CeO2 (100), which are intrinsically
defective.
Vacancies are thus required to perform CO oxidation. Therefore,
the vacancy formation energy (Eᮀ ) and the number of potential
adsorption sites (fraction of vacancies, xᮀ ) are the most important activity descriptors. Thus, analogously, CO oxidation activity
will depend on the Ered and O2p terms, as outlined in Fig. 5. Based
on this interpretation, the CO oxidation rate is face sensitive over
ceria. Considering that the vacancy formation energy varies in the
order 1 1 0 > 1 0 0> 1 1 1, the oxidation rate is expected to follow
the trend r110 > r100 > r111 . This theoretical activity trend is in line
with the experimental findings [69,111,112], where it has been
shown that ceria nanocubes are more active in CO oxidation compared to ceria nanoctahedra, Fig. 6a. It is worth noting that the
large number of basic surface sites on CeO2 can react with CO
and CO2 molecules, forming carbonates that behave as spectators
on the surface [69,113,114]. The presence of unidentate, bidentate, and bridged carbonates has been demonstrated by infrared
spectroscopy [69,115].
Ceria nanoctahedra and nanocubes exhibit a different behavior
upon aging [112]. Whereas the nanocubesb do not decrease their
CO oxidation performance, the aged octahedra are nearly inactive,
Fig. 6a. As shown in Table 4, O2 only adsorbs on the defective sites
of ceria. Therefore, once CO reacts with the adsorbed O2 filling the
vacancy, Eq. (4), the reaction can only proceed via (5), which is
highly disfavored, especially on the (1 1 1) facet. Step (5) is thus
described by the Eᮀ term. On the other hand, carbonate formation
upon CO adsorption takes place on the (1 1 0) and (1 0 0) surfaces,
and releases two Ce3+ . O2 adsorbs on the Ce3+ centers [116–118]
that now are extensively available, so the reaction can proceed as
in (4). This is however not possible on the (1 1 1) surface, since
carbonate formation is energetically disfavored on this facet.
Accordingly, doped ceria may enhance CO oxidation by decreasing the vacancy formation energy. Zr-doped cerium oxide increases
the OSC, Fig. 6b, leading to an increased CO conversion [119]. This
is further improved by the use of trivalent dopants, where Eᮀ is
significantly diminished [120].
Finally, various metal nanoparticles supported on ceria exhibit
high CO oxidation activity, namely Au, Cu, Pt, or Pd [26,121–129].

306

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Fig. 6. (a) CO oxidation as a function of the temperature over ceria nanocubes (NC) and nanoctahedra (NO) fresh and aged. (b) CO oxidation as a function of the oxygen
storage capacity in mixed CeZrO2 oxides. (c) Performance of an Au/CeO2 catalyst as a function of temperature. The model structure of the catalyst is shown in the inset. The
figures are adapted from [112], [132], and [133], respectively.

Fig. 6c shows how CO conversion is enhanced on Au/CeO2 compared to pristine CeO2 . Oxides with metal supported clusters are
a paramount example of phase cooperation, where the oxide and
metal phases can promote different steps of the mechanism [27].
Some steps are thermodynamically favored via oxygen spillover
from ceria to the metal [130]. Plenty of studies have been conducted
on metal/ceria catalysts, but our analysis focuses on standalone
ceria, and hence, these are beyond the scope of this work.
It is worth pointing out that CO oxidation shares some elementary steps with the water-gas shift (WGS) reaction [26,27].
The overall WGS reaction is H2 O + CO → H2 + CO2 , where the oxidation of CO to CO2 can occur as the direct process explained above
(CO + Oᮀ ), or through the COOH intermediate species, Fig. 1. CO
conversion during WGS was shown to be much higher for ceria
nanocubes compared to rod and octahedron nanoparticles [28], as
also shown above for the CO oxidation reaction. This comes either
from an easier removal of the lattice oxygen by CO, or from an
improved water activation due to the larger fraction of vacancies
on the surface. Water can get activated at acid-base pair centers,
such as Ce-O or ᮀ-O, and thus acidity and basicity control this step
[131].
3.2. Halogen production via hydrogen halide oxidation
A process for which ceria is close to industrial application is
the recycling of chlorine by the gas-phase HCl oxidation [134,135].
Chlorine is employed to functionalize organic molecules and typically one atom gets included in the organic moiety. Furthermore,
chlorine-containing compounds can be potentially hazardous and
thus in many final products they need to be removed, usually in the
form of HCl. The oxidation of HCl to Cl2 is an exothermic process carried out via a heterogeneous catalyst, typically RuO2 [29,136–138].
However, RuO2 is known to experience some stability issues at high
temperatures [139]. On the other hand, CeO2 possesses lower activity in the medium temperature regime, but its stability makes it a
suitable candidate for the process in the reactor areas, where the
temperature is known to rise [10]. The reaction network on CeO2 ,
Fig. 7a, was proposed to consist of the following steps:
HCl + Oᮀ + ᮀ → Clᮀ + Oᮀ H

(8)

HCl + Oᮀ + ∗ → Cl ∗ + Oᮀ H

(9)

Clᮀ + Cl∗ → Cl2 ∗ + ᮀ

(10)

Cl2 ∗ → Cl2 + ∗

(11)

Oᮀ H + Oᮀ H → H2 Oᮀ + Oᮀ

(12)

H2 Oᮀ → H2 O + ᮀ

(13)

O2 + ᮀ → O2ᮀ

(14)

O2ᮀ + ᮀ → 2Oᮀ

(15)

HCl adsorbs dissociatively on ceria, where the halide fills a
vacancy site leaving the system with a defect state as a Ce3+ atom.
Further HCl adsorption then proceeds as Cl− adsorbed on the surface. The first step relates to the acid-base and redox properties
of the surface, whereas the latter only involves acid-base properties. Two Cl atoms from the surface and lattice positions can evolve
toward the gas-phase, and the vacancy can be replenished by O2 .
Therefore, the latter halide evolution and oxygen vacancy healing steps mainly involve redox properties, and also depend on the
mobility of the vacancies in the material. Chlorine evolution is the
most energy-demanding step, and thus the ability of the surface to
rapidly cycle between Ce4+ and Ce3+ controls the activity. Therefore, the activity of ceria for HCl oxidation was shown to correlate
with the experimental OSC measurements, Fig. 7b, performed on
different CeO2 samples.
Although the OSC character was shown to be the main activity descriptor, by the iso and aliovalent doping of ceria we found
that the complexity of the reaction cannot be fully unveiled by
this single parameter. While the activity increased for the isovalently doped ceria, trivalent dopant cations induced the opposite
effect [31], which however have a positive impact on the OSC.
The main responsible for this unintuitive behavior is that the
redox character is better maintained through the isovalent doping, since trivalent cations generate O• centers that are healed
upon Cl adsorption, which ultimately annihilate the redox ability
of the surface. The formation of these defect pairs can be characterized by the chlorine replacement energy, Erep , of the lattice
CeO2 + 1/2Cl2 → CeO2-x Cl + 1/2O2 . For the pristine (1 1 1) surface,
this Erep is −0.40 eV, while it reaches −1.77 eV for the Ga-doped
material. This means that the further step displacing the substituting chlorine (10) is a much higher energy-demanding step for
the latter doped material compared to the native surface. A similar conclusion was recently reached by Over and coworkers [33].
Remarkably, it is worth noting that the Zr-doped CeO2 catalyst is
not only more active than the pristine material, but also presents
significantly high stability, Fig. 7c inset.
The chemistry found for HCl can be extended to heavier elements in the series, such as the recycling of HBr into Br2 [32,141].
HBr oxidation always takes place at lower temperatures than HCl
for all the studied materials, and the reaction is also more exothermic. The bigger radius of bromine compared to chlorine induces
a decrease in the energy required to replace oxygen by the halide,

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

307

Fig. 7. (a) Mechanism of HCl oxidation over a CeO2 (1 1 1) surface. (b) Space-time yield as a function of the oxygen storage capacity of CeO2 . The temperatures indicate the
temperature at which the oxide was calcined in static air, in order to alter the surface vacancy chemistry. (c) HCl conversion as a function of the temperature over CeO2 /ZrO2
catalyst. The inset represents the stability of the catalyst for 800 h on stream. (d) Experimental light-off temperatures (at the rate of X2 production, r = 3 mol X2 mol Ce−1 h−1 )
as a function of the two-variable descriptor involving Ered and Erep , for Cl (filled dots) and Br (circles) on pristine (black) and Zr-doped (blue) ceria. The fitted equation is f(Ered ,
Erep ) = (−522 Ered + 777 Erep − 1) K. The figures (b) and (c) are adapted from [140] and [10]. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

Erep , from −0.40 to −0.04 eV. These values turn more exothermic for
the Zr-doped surface, −0.72 and −0.39 eV for Cl and Br, respectively.
The lower stability of the Br-substituted surfaces makes HBr more
active for the overall reaction. However, the redox character of the
surface also plays a role, since the formation of X2 (where X is Cl or
Br) is a redox step (10). This is illustrated in Fig. 7d, where the lightoff temperatures [32], at the X2 production rate r = 3 mol X2 mol
Ce−1 h−1 , are shown to be a function of the two-variable descriptor embracing Erep and Ered . Thus, both oxygen-halide replacement
energy and surface redox ability need to be controlled to predict
the light-off temperatures, for a given production rate.

3.3. Hydrogen activation and alkyne semi-hydrogenation
Excellent catalytic features of ceria have been also reported for
hydrogenation reactions under hydrogen-rich conditions [35]. The
key elementary step to understand this behavior is the activation of
hydrogen, which is related to the presence of Ce-O short distances
and charge separation [142–145]. Thus, H2 is polarized and dissociates heterolytically on CeO2 generating acid-base pairs on the
surface (17) [146]. Alternatively, the homolytic dissociation over
two surface oxygens is also possible (19), although it exhibits a
slightly higher energy barrier [142]. The mechanisms are described
as follows:
H2 + ∗ → H2 ∗

(16)

H2 ∗ + Oᮀ → Oᮀ H + H∗

(17)

H ∗ + O ᮀ → Oᮀ H + ∗

(18)

H2 ∗ + 2Oᮀ → 2Oᮀ H

(19)

The heterolytic mechanism results in a proton and a hydride
attached to the basic and acid sites on the surface, indicating that
acid (ECO ) and basic (O2p ) terms become relevant. The charge of the
centers and their geometric distribution are crucial to understand
the chemistry behind [142]. The reaction then proceeds, since the
hydride can further move toward the equilibrium hydroxyl state
(18), reducing two Ce surface centers. The latter step is associated to the reducibility of the surface, Ered . The final surface state is
thus achieved via both the heterolytic and homolytic mechanisms.
Therefore, the temperature of maximum H2 consumption from the
H2 -TPR on the three pristine surface facets and the Ga-doped (1 1 1)
correlates with the two-variable descriptor that includes the O2p
and Ered terms, Fig. 8a.
Hydrogen activation has been shown to be a key step in hydrogenation reactions [147]. Then, the hydrogenation mechanism has
been reported to occur through two different pathways consisting
of Eqs. (21)–(23):
C2 H2 + ∗ → C2 H2 ∗

(20)

C2 H2 ∗ + H2 ∗ → C2 H4 ∗ + ∗

(21)

C2 H2 ∗ + Oᮀ H → C2 H3 ∗ + Oᮀ

(22)

308

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

Fig. 8. (a) Temperature of maximum H2 consumption during H2 -TPR as a function of the two-variable descriptor that includes the O2p and Ered terms, for the (1 1 1), (1 1 0),
(1 0 0), and Ga0.2 Ce0.8 O2 - and Ga0.05 Ce0.95 O2 -(1 1 1) surfaces. The fitted equation is f(O2p , Ered ) = (426 O2p + 255 Ered + 1001) K. The H2 consumption TPR for nanoctahedra,
nanorods, and nanocubes are shown in the inset. (b) Acetylene hydrogenation performance as a function of temperature over CeO2 . (c) Rate of acetylene hydrogenation
as a function of the oxygen storage capacity over ceria nanoctahedra and nanocubes. (d) Influence of temperature on the activity of CeO2 and Ga0.05 Ce0.95 O2 in acetylene
hydrogenation. The figure, in particular, depicts that Ga-doping enables a reduction of the operating temperature. Figures (b), (c), and (d) are adapted from [34], [112], and
[36], respectively.

C2 H3 ∗ + Oᮀ H → C2 H4 ∗ + Oᮀ

(23)

C2 H4 ∗ → C2 H4 + ∗

(24)

The presence of large amounts of hydrogen in the environment leads to a majorly hydroxylated surface termination [147].
The reactivity is controlled by the availability of sites on the substrate and the fact that H2 activation is difficult. For the (1 1 1)
surface, it was found that a concerted six-membered ring transition
state in (21) comprising the alkyne C C bond, the lattice OH, and
the hydrogen molecule could be responsible for both the experimentally observed activity and selectivity [147]. An alternative
mechanism was also proposed [148], where the alkyne adsorbs
on the surface forming a radical species. This intermediate can
strip the first hydrogen from the surface with a very low energy
barrier (22), although the second H transfer (23) exhibits a large
activation energy (>2.5 eV). Still some controversy on the nature
of the reaction path exists, albeit recent NMR experiments seem
to support the former mechanism [149]. Furthermore, if the reaction follows the concerted path, a stereoselectivity to the cis-alkene
is expected over CeO2 . This is confirmed experimentally, showing
full cis-alkene selectivity in the ceria-catalyzed hydrogenation of a
variety of functionalized and internal acetylenic compounds [35].
The capability of the Ce-O pair to generate a large enough electric
field that polarizes the incoming H2 molecule is the most important

point in the reactivity of the material in hydrogenations. This can be
also traced back to the basicity of the surface and the redox term,
together with the geometric ensemble that makes the transition
state for hydrogen activation possible. This is likely to be extrapolated to the hydrogenation activity found for other oxides, such as
ZrO2 or La2 O3 [150–152].
Face sensitivity of the reaction rate was also found in hydrogenation, where (1 1 1) facets exhibit higher performance than the
(1 0 0) exposing nanocubes [112]. The dependence was thus the
opposite than that found for CO oxidation, Fig. 8c. Our description can explain this observation based on surface poisoning on
the (1 0 0), Table 4. Strong adsorption is facilitated by basic centers
on the surface and their adaptability to the coordination. On the
(1 1 1) surface, alkynes adsorb on one basic center forming a radical intermediate with Eads = −0.30 eV [148], whereas on the other
surfaces the local structure allows coordination to two basic centers. Indeed, we have found an OCHCHO intermediate that sits on
the surface in a robust manner, with coordination energy of −4.05
and −2.41 on the (1 0 0) and (1 1 0) surfaces, respectively. This is
reminiscent to carbonate poisoning in CO oxidation, and thus the
basicity of the surface O2p and the local coordination of the oxygens on the surface that allow their high mobility are the ultimate
responsible for the observed poisoning.

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

The hydrogenation of alkynes is accelerated by the addition of
trivalent dopants [36]. This has been experimentally shown in the
gas-phase hydrogenation of alkynes over Ga-doped CeO2 catalysts,
Fig. 8d. It shall be noticed that for doped surfaces with trivalent
cations the reaction mechanism is electronically different. The surface provides a localized hole state, located on the oxygen center
near the doping atom, Fig. 4. Then the contribution from covalency becomes dominant and thus H2 dissociation is homolytic.
In the final state, the hole is filled with one of the electrons of the
hydrogen molecule and the remaining H• radical adsorbs on the
adjacent cerium center. The ultimate consequence of this alternative mechanism is that the activation barrier for H2 dissociation is
much lower than for the regular surface. The best descriptor for the
homolytic dissociation is then the energy of the hole on the surface,
Ered (which in this particular case is related to the hole instead of
the Ce4+ -Ce3+ cycle) [153,154]. In particular, trivalent cations were
found to be more active in the following order: In > Ga > Al. The
ordering closely follows the ionic radius of these dopants [36].
Other hydrogenation reactions over cerium oxide were also
reported. Hydrogenation of benzoic acid to benzaldehyde, also
under H2 rich conditions, was performed exhibiting high activity and selectivity [151,155,156]. The reaction was suggested to
proceed as a Mars-van-Krevelen mechanism, where the oxygen
vacancy sites acted as the active sites. Hence, the (1 1 0) and (1 0 0)
surfaces are more active than the (1 1 1). Likewise, the addition of
dopants that increase the Eᮀ and OSC promotes this reaction [151].
The hydrogenation of nitroaromatic compounds using N2 H4 as a
reducing agent was also performed on ceria nanoparticles, providing the related aniline derivative [157].
3.4. Alcohol conversion
Alcohol conversion has been extensively performed on ceria
[37–46], since these reactive compounds are traditionally used as
probe molecules to assess the catalytic properties of metal oxides
[158,159]. Alcohols typically convert to aldehydes, following the
proposed elementary steps listed below:
RCH2 OH + ∗ → RCH2 OH∗

(25)

RCH2 OH ∗ + Oᮀ → RCH2 O ∗ + Oᮀ H

(26)

RCH2 O ∗ + Oᮀ → RCHO ∗ + Oᮀ H

(27)

RCHO∗ → RCHO + ∗

(28)

Oᮀ H + Oᮀ H → H2 Oᮀ + Oᮀ

(29)

H2 Oᮀ → H2 O + ᮀ

(30)

The first acid-base deprotonation step (26) precedes the redox
C H scission (27), where two cerium cations reduce to Ce3+ . For the
acid-base step, the O2p term controls the reaction rate, whereas
the two-variable descriptor embracing O2p and redox is required
to describe the second step [108]. Oxygen vacancies however may
play a role in this process, switching the selectivity to alkene formation through the following elementary steps:
RCH2 CH2 OH + ᮀ + Oᮀ → RCH2 CH2 Oᮀ + Oᮀ H

(31)

RCH2 CH2 Oᮀ + Oᮀ → RCHCH2 Oᮀ + Oᮀ H

(32)

RCHCH2 Oᮀ + ∗ → RCH CH2 ∗ + Oᮀ

(33)

RCH = CH2 ∗ → RCH CH2 + ∗

(34)

Oᮀ H + Oᮀ H → H2 Oᮀ + Oᮀ

(35)

H2 Oᮀ → H2 O + ᮀ

(36)

Thus, in addition to the O2p and Ered terms to describe the O H
(31) and C H (32) scissions, the oxygen vacancy formation energy

309

Fig. 9. Formaldehyde first C H scission activation energy as a function of a collective variable embracing Ered and O2p , for the pristine (1 1 1), (1 1 0), and (1 0 0)
CeO2 surfaces and the doped Ga-, V-, Zr-, Hf-, and Th-(1 1 1) surfaces. The implicit
doped (1 1 1) structures obtained from applying strain to the (1 1 1) surface are also
included (grey circles). The fitted equation is f(Ered , O2p ) = (0.57 Ered − 0.83 O2p +
0.02) eV. The Hf-, Th-, and strained (1 1 1) values are adapted from [108].

determines the rate of this side reaction and thus the overall selectivity.
As demonstrated experimentally, alcohol conversion is also
structure sensitive [39,40,42]. The rate of methanol and ethanol
conversion is higher on the (1 0 0) surface than on the (1 1 1) surface. This is attributed to the fact that the activation energy for C H
cleavage is much lower on the (1 0 0) surface, due to its higher redox
ability. Furthermore, for the conversion of ethanol, a temperature
increase favors the formation of ethylene over acetaldehyde on the
(1 0 0) compared to the (1 1 1) facet [42]. This is due to the lower
oxygen vacancy formation energy for the former surface (Table 2).
For the particular case of methanol, the subsequent oxidation
of formaldehyde to CO was also observed in the more open (1 0 0)
surface. The reaction consists of the following elementary steps:
HCHO ∗ + Oᮀ → CHO ∗ + Oᮀ H

(37)

CHO ∗ + Oᮀ → CO ∗ + Oᮀ H

(38)

CO∗ → CO + ∗

(39)

Oᮀ H + Oᮀ H → H2 Oᮀ + Oᮀ

(40)

H2 Oᮀ → H2 O + ᮀ

(41)

Therefore the competition between formaldehyde desorption
and the first C H cleavage of formaldehyde triggers the selectivity
[160]. The first step of this reaction is a redox step, where formaldehyde is oxidized and surface is reduced. The second CH stripping,
O2p and Ered terms are crucial, analogous to the C H bond breaking
in methanol. Fig. 9 shows the correlation of the activation energy
of this elementary step with respect to the combined O2p and Ered
terms, for the three low-index surfaces and the isovalent and aliovalent doped-(1 1 1). Both the basicity of the recipient oxygen and
the capability of the surface to accommodate the two electrons
formaldehyde oxidation are required to describe this step. For the
second H-stripping, only the O2p is required as it is an acid-base
step. In addition, to assess the selectivity, formaldehyde desorption must be also described. Formaldehyde chemisorbs on ceria as
a bidentate structure forming Ce-O and Oᮀ C bonds. Then the acidity of the cerium cations, the basicity of the oxygens, and the site

310

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

structure control how the bidentate adsorbate accommodates on
the surface [108].
The (1 0 0) surface has higher cerium acidity, oxygen basicity,
and larger site area than the (1 1 1) or the (1 1 0) surfaces, and thus
HCHO binds stronger on this facet. Furthermore, the high inherent number of vacancy sites on the (1 0 0) facet also favors the
adsorption of O-R compounds. Similarly, the higher redox ability of
the (1 0 0) surface in combination with the oxygen basicity favors
C H scission on this facet. Formaldehyde hence converts to CO on
the (1 0 0) surface, whereas it desorbs from the (1 1 1) and (1 1 0)
surfaces, as observed in the TPD experiments and the DFT energy
profiles [39,160].
3.5. Dimethyl carbonate synthesis
More sophisticated reactions to valorize CO2 [44,52,161–164]
have been recently put forward. For example, CO2 from waste effluents can react with methanol to form dimethyl carbonate (DMC),
an important intermediate for polycarbonates and fuel additives.
CeO2 catalysts have been proposed as suitable active materials for
this family of reactions. The overall reaction is as follows:
CO2 + 2MeOH

CH3 OCOOCH3 + H2 O

(42)

The equilibrium of the reaction is driven toward reactants unless
water is removed from the medium employing a sorbent. The
proof-of-principle of the higher conversion when employing a
dehydrating agent was presented by Tomishige et al. [53,54], and
recently a continuous process employing relatively high pressures
for the CO2 has been developed [55].
The list of elementary reactions potentially participating in the
process is as follows:
CO2 + ∗ → CO2 ∗

(43)

CH3 OH + ∗ → CH3 OH∗

(44)

CH3 OH ∗ + Oᮀ → CH3 O ∗ + Oᮀ H

(45)

CO2 ∗ + CH3 O∗ → CO2 OCH3 ∗ + ∗

(46)

CO2 OCH3 ∗ + Oᮀ H → HOCOOCH3 ∗ + Oᮀ

(47)

HOCOOCH3 ∗ + CH3 O∗ → CH3 OCOOCH3 ∗ + OH∗

(48)

OH ∗ + Oᮀ H → H2 O ∗ + Oᮀ

(49)

CH3 OCOOCH3 ∗ → CH3 OCOOCH3 + ∗

(50)

H2 O∗ → H2 O + ∗

(51)

The adsorption of CO2 depends on the basicity of the surface,
O2p , in agreement with our previous results and experimentally
illustrated by Lee et al. [70]. As we have shown earlier, methanol
adsorption (44) precedes an acid-base dissociation (45). On the
clean surface the Ce-O pairs act as such centers. It shall be noticed
that the basicity of O2p is involved in both CO2 trapping and CH3 OH
adsorption. As methanol reacts easily with the surface, the need
of high CO2 pressure can be explained as the competition for the
adsorption on the basic sites of the surface. The subsequent reaction
steps can be understood as a substitution where CO2 is attacked by
the methoxy group. At this stage, the carbonate gets protonated
to improve its ability as leaving group. The ligand displacement by
another methoxy group on the surface would generate a surface
hydroxyl group and the DMC, weakly bound to the surface. The
recombination of OH with another proton from the surface leads to
the secondary water product.
The mechanistic analysis indicates that the following descriptors are crucial for DMC synthesis. First, both the basicity and
the acidity of the material are required to initiate the reaction.

However, a high acidity of the material has been described as detrimental to the overall performance. The reason for that can be traced
back to the fact that steps (46) and (47) are limited by the presence of strong acid (Brønsted) sites as clean basic O centers are
needed on the surface. Still, some degree of acidity is needed to
eliminate water from the surface (51). This explains the acid-base
promotion found experimentally [70]. Moreover, the mechanism
sketched above provides two major threats. On the one hand, areas
with high density of vacancies where some of the carbonates can
be trapped will activate the internal bonds in undesired manners.
Therefore, facets or doping that improve the formation of vacancies
or contain more vacancies would be less efficient to form DMC. An
alternative way of illustrating this is by inferring the size required
for the reaction ensemble. While the reactions described before
dimethyl carbonate synthesis required only a very few adsorption
sites, the large hapticity of the fragments involved in this reaction
implies that the isolation of sites is not a good strategy in DMC
synthesis.

4. Conclusions and outlook
The applications of CeO2 in heterogeneous catalysis have grown
over the last years. The flexibility of the electronic structure able to
be reduced, the acid-base properties of the surface, and its resistance against harsh reactions conditions make it suitable for many
uses. These versatile and unique catalytic properties turn ceria an
optimal playground to make a real step forward in catalysis. In this
perspective, the synergy between careful synthetic shape control
at the nanoscale, accurate characterization, catalytic testing, and
extensive use of the computational techniques for the rationalization and prediction of the properties are gathered and described.
The oxygen activation ability has been taken as the main property of the material, thus being mainly employed in traditional
oxidation processes like CO and soot. However, these properties
have been extended to other challenging oxidation processes, like
those of hydrogen halides (HCl and HBr) for halogen recovery.
Complementary reactions including alkyne semi-hydrogenation
have been also reported recently and are being extended to other
oxides. Other reactions that are basically centered in the reduction
of methanol to formaldehyde and the condensation of methanol
with CO2 to dimethylcarbonate have also been proposed. This wide
variety of reactions can be systematized based on descriptors for
the reactivity. Starting from the classical terms described by Grasselli [78], which are non-ortogonal, we have developed a full set
of chemical descriptors that can be used to describe the reactivity of these materials and that can be either experimentally or
computationally derived. In most cases, due to some difficulties
in either shape-selected synthesis or precise characterization, it
is easier to calculate these descriptors from first-principles. Our
results show that it is possible to trace back the reactivity and stability of materials mainly through the introduction of basicity, redox,
and geometric terms.
We herein propose a systematic way to analyze reactivity of
ceria in catalytic processes. We consider that this is the first step
towards the prediction of more active and selective surfaces (or
doped systems) for the next generation of ceria-based materials.
Improving the performance via chemical descriptors is a much
more intuitive approach than other structure-activity relationships based on energies, as they are directly meaningful and might
be benchmarked against experimental descriptors. Finally, the
approach provided herein to describe the catalysis of ceria can be
extended to other families of complex compounds including oxides,
carbides, sulphides, and multicomponent materials with acid-base
and/or redox sites, and employed to oxidations, reductions, acidbase, and associative reactions. The procedure then shall enclose: (i)

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

determination of the phases under reaction conditions; (ii) experimental and first-principles thermodynamics to analyze the state of
the surface (stoichiometry, geometry, etc); and finally, (iii) reduce
the complete set of chemical descriptors to the meaningful ones for
a particular reaction, experimental conditions, and/or material.
Acknowledgements
N.L. acknowledges the Ministerio de Economía y Competitividad
− MINECO (CTQ2012-33826), and the Generalitat de Catalunya −
AGAUR (SGR-2014-SGR-145), as well as the BSC-RES for providing
generous computational resources. M.C.-C. is grateful to MINECO
for a “Juan de la Cierva − Formación” fellowship (FJCI-2014-20568).
J.P.-R. thanks the Swiss National Science Foundation (SNF project
number 200021-156107).
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.apcatb.2016.
02.035.
References
[1] A. Trovarelli, Catalysis by Ceria and Related Materials, Imperial College
Press, London, 2002.
[2] W. Huang, Y. Gao, Catal. Sci. Technol. 4 (2014) 3772–3784.
[3] C. Sun, H. Li, L. Chen, Energy Environ. Sci. 5 (2012) 8475.
[4] D. Zhang, X. Du, L. Shi, R. Gao, Dalton Trans. 41 (2012) 14455–14475.
[5] Z.-A. Qiao, Z. Wu, S. Dai, ChemSusChem 6 (2013) 1821–1833.
[6] H. Metiu, S. Chrétien, Z. Hu, B. Li, X. Sun, J. Phys. Chem. C 116 (2012)
10439–10450.
[7] A. Trovarelli, Catal. Rev. 38 (1996) 439–520.
[8] M.V. Ganduglia-Pirovano, A. Hofmann, J. Sauer, Surf. Sci. Rep. 62 (2007)
219–270.
[9] J. Paier, C. Penschke, J. Sauer, Chem. Rev. 113 (2013) 3949–3985.
[10] A.P. Amrute, C. Mondelli, M. Moser, G. Novell-Leruth, N. López, D. Rosenthal,
et al., J. Catal. 286 (2012) 287–297.
[11] D.A. Andersson, S.I. Simak, N.V. Skorodumova, I.A. Abrikosov, B. Johansson,
Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 3518–3521.
[12] J.S. Elias, M. Risch, L. Giordano, A.N. Mansour, Y. Shao-Horn, J. Am. Chem.
Soc. 136 (2014) 17193–17200.
[13] G. Azimi, R. Dhiman, H.-M. Kwon, A.T. Paxson, K.K. Varanasi, Nat. Mater. 12
(2013) 315–320.
[14] G. Carchini, M. García-Melchor, Z. Łodziana, N. López, ACS Appl. Mater.
Interfaces 8 (2016) 152–160.
[15] F. Esch, S. Fabris, L. Zhou, T. Montini, C. Africh, P. Fornasiero, et al., Science
309 (2005) 752–755.
[16] M. Nolan, S.C. Parker, G.W. Watson, Surf. Sci. 595 (2005) 223–232.
[17] D.R. Mullins, Surf. Sci. Rep. 70 (2015) 42–85.
[18] C.T. Campbell, C.H.F. Peden, Science 309 (2005) 713–714.
[19] H.C. Yao, Y.F.Y. Yao, J. Catal. 86 (1984) 254–265.
[20] E.P. Murray, T. Tsai, S.A. Barnett, 400 (1999) 649-651.
[21] S. Park, J. Vohs, R. Gorte, Nature 404 (2000) 265–267.
[22] B.C. Steele, A. Heinzel, Nature 414 (2001) 345–352.
[23] Z. Shao, S.M. Haile, Nature 431 (2004) 170–173.
[24] M.M. Schubert, V. Plzak, J. Garche, R.J. Behm, Catal. Lett. 76 (2001) 143–150.
[25] O. Pozdnyakova, D. Teschner, A. Wootsch, J. Kröhnert, B. Steinhauer, H.
Sauer, et al., J. Catal. 237 (2006) 1–16.
[26] J.B. Park, J. Graciani, J. Evans, D. Stacchiola, S. Ma, P. Liu, et al., Proc. Natl.
Acad. Sci. U. S. A. 106 (2009) 4975–4980.
[27] J.A. Rodriguez, S. Ma, P. Liu, J. Hrbek, J. Evans, M. Pérez, Science 318 (2007)
1757–1760.
[28] S. Agarwal, L. Lefferts, B.L. Mojet, D.A.J.M. Ligthart, E.J.M. Hensen, D.R.G.
Mitchell, et al., ChemSusChem 6 (2013) 1898–1906.
[29] J. Pérez-Ramírez, C. Mondelli, T. Schmidt, O.F.-K. Schlüter, A. Wolf, L.
Mleczko, et al., Energy Environ. Sci. 4 (2011) 4786–4799.
[30] R. Farra, M. Eichelbaum, R. Schlögl, L. Szentmiklósi, T. Schmidt, A.P. Amrute,
et al., J. Catal. 297 (2013) 119–127.
[31] R. Farra, M. García-Melchor, M. Eichelbaum, M. Hashagen, W. Frandsen, J.
Allan, et al., ACS Catal. 3 (2013) 2256–2268.
[32] M. Moser, G. Vilé, S. Colussi, F. Krumeich, D. Teschner, L. Szentmiklósi, et al.,
J. Catal. 331 (2015) 128–137.
[33] M. Möller, S. Urban, P. Cop, T. Weller, R. Ellinghaus, M. Kleine-Boymann,
et al., ChemCatChem 7 (2015) 3738–3747.
[34] G. Vilé, B. Bridier, J. Wichert, J. Pérez-Ramírez, Angew. Chem. Int. Ed. 51
(2012) 8620–8623.

311

[35] G. Vilé, S. Wrabetz, L. Floryan, M.E. Schuster, F. Girgsdies, D. Teschner, et al.,
ChemCatChem 6 (2014) 1928–1934.
[36] G. Vilé, P. Dähler, J. Vecchietti, M. Baltanás, S. Collins, M. Calatayud, et al., J.
Catal. 324 (2015) 69–78.
[37] R.M. Ferrizz, G.S. Wong, T. Egami, J.M. Vohs, Langmuir 17 (2001) 2464–2470.
[38] D.R. Mullins, M.D. Robbins, J. Zhou, Surf. Sci. 600 (2006) 1547–1558.
[39] P.M. Albrecht, D.R. Mullins, Langmuir 29 (2013) 4559–4567.
[40] Z. Wu, M. Li, D.R. Mullins, S.H. Overbury, ACS Catal. 2 (2012) 2224–2234.
ˇ
[41] V. Matolín, J. Libra, M. Skoda, N. Tsud, K.C. Prince, T. Skála, Surf. Sci. 603
(2009) 1087–1092.
[42] M. Li, Z. Wu, S.H. Overbury, J. Catal. 306 (2013) 164–176.
ˇ c
[43] Y. Lykhach, A. Neitzel, K. Sevˇ íková, V. Johánek, N. Tsud, T. Skála, et al.,
ChemSusChem 7 (2014) 77–81.
[44] M.H. Haider, N.F. Dummer, D.W. Knight, R.L. Jenkins, M. Howard, J. Moulijn,
et al., Nat. Chem. 7 (2015) 1028–1032.
[45] D.R. Mullins, S.D. Senanayake, T.-L. Chen, J. Phys. Chem. C 114 (2010)
17112–17119.
[46] T.-L. Chen, D.R. Mullins, J. Phys. Chem. C 115 (2011) 13725–13733.
[47] T.-L. Chen, D.R. Mullins, J. Phys. Chem. C 115 (2011) 3385–3392.
[48] F.C. Calaza, Y. Xu, D.R. Mullins, S.H. Overbury, J. Am. Chem. Soc. 134 (2012)
18034–18045.
[49] A.K.P. Mann, Z. Wu, F.C. Calaza, S.H. Overbury, ACS Catal. 4 (2014)
2437–2448.
[50] S.M. Schimming, O.D. LaMont, M. König, A.K. Rogers, A.D. D’Amico, M.M.
Yung, et al., ChemSusChem 8 (2015) 2073–2083.
[51] S.D. Senanayake, W.O. Gordon, S.H. Overbury, D.R. Mullins, J. Phys. Chem. C
113 (2009) 6208–6214.
[52] M. Honda, M. Tamura, Y. Nakagawa, S. Sonehara, K. Suzuki, K.-I. Fujimoto,
et al., ChemSusChem 6 (2013) 1341–1344.
[53] K. Tomishige, Y. Furusawa, Y. Ikeda, M. Asadullah, K. Fujimoto, Catal. Lett. 76
(2001) 71–74.
[54] K. Tomishige, K. Kunimori, Appl. Catal. A Gen. 237 (2002) 103–109.
[55] A. Bansode, A. Urakawa, ACS Catal. 4 (2014) 3877–3880.
[56] J.A. Rodriguez, T. Jirsak, S. Sambasivan, D. Fischer, A. Maiti, J. Chem. Phys.
112 (2000) 9929.
[57] R.M. Ferrizz, T. Egami, G.S. Wong, J.M. Vohs, Surf. Sci. 476 (2001) 9–21.
[58] M. Daturi, N. Bion, J. Saussey, J.-C. Lavalley, C. Hedouin, T. Seguelong, et al.,
Phys. Chem. Chem. Phys. 3 (2001) 252–255.
[59] M. Waqif, P. Bazin, O. Saur, J.C. Lavalley, G. Blanchard, O. Touret, Appl. Catal.
B Environ. 11 (1997) 193–205.
[60] J.A. Rodriguez, T. Jirsak, A. Freitag, J.C. Hanson, J.Z. Larese, S. Chaturvedi,
Catal. Lett. 62 (1999) 113–119.
[61] S.H. Overbury, D.R. Mullins, D.R. Huntley, L. Kundakovic, J. Phys. Chem. B 103
(1999) 11308–11317.
[62] R.M. Ferrizz, R.J. Gorte, J.M. Vohs, Catal. Lett. 82 (2002) 123–129.
[63] M.Y. Smirnov, A.V. Kalinkin, A.V. Pashis, A.M. Sorokin, A.S. Noskov, K.C.
Kharas, et al., J. Phys. Chem. B 109 (2005) 11712–11719.
[64] M. Happel, Y. Lykhach, N. Tsud, T. Skála, K.C. Prince, V. Matolín, et al., J. Phys.
Chem. C 115 (2011) 19872–19882.
[65] M. Flytzani-Stephanopoulos, M. Sakbodin, Z. Wang, Science 312 (2006)
1508–1510.
[66] D.R. Mullins, T.S. McDonald, Surf. Sci. 601 (2007) 4931–4938.
[67] B. Liu, H. Xu, Z. Zhang, Chin. J. Catal. 33 (2012) 1631–1635.
[68] C. Binet, M. Daturi, J.-C. Lavalley, Catal. Today 50 (1999) 207–225.
[69] Z. Wu, M. Li, S.H. Overbury, J. Catal. 285 (2012) 61–73.
[70] H.J. Lee, S. Park, I.K. Song, J.C. Jung, Catal. Lett. 141 (2011) 531–537.
[71] Z. Wu, A.K.P. Mann, M. Li, S.H. Overbury, J. Phys. Chem. C 119 (2015)
7340–7350.
[72] M. Boronat, T. López-Ausens, A. Corma, Surf. Sci. (2015), http://dx.doi.org/
10.1016/j.susc.2015.10.047.
[73] L. Vivier, D. Duprez, ChemSusChem 3 (2010) 654–678.
[74] J.K. Nørskov, T. Bligaard, B. Hvolbæk, F. Abild-Pedersen, I. Chorkendorff, C.H.
Christensen, Chem. Soc. Rev. 37 (2008) 2163–2171.
[75] F. Studt, F. Abild-Pedersen, T. Bligaard, R.Z. Sørensen, C.H. Christensen, J.K.
Nørskov, Science 320 (2008) 1320–1322.
[76] F. Calle-Vallejo, D. Loffreda, M.T.M. Koper, P. Sautet, Nat. Chem. 7 (2015)
403–410.
[77] P. Sabatier, La Catalyse En Chimie Organique, Librairie Polytechnique, Paris,
1920.
[78] R.K. Grasselli, Top. Catal. 21 (2002) 79–88.
[79] W. Tang, E. Sanville, G. Henkelman, J. Phys. Condens. Matter 21 (2009)
084204.
[80] P.A. Redhead, Vacuum 12 (1962) 203–211.
[81] Although many clear probes for basic/acid sites are commonly available,
polar reactants may dissociate upon interaction with the surface (e.g. acid
molecules), making impossible to disentangle the contributions.
[82] Care shall be taken though as probably the number of potential acceptors
(i.e. in high symmetry positions) that can stabilize by polarons is relevant to
the overall redox ability.
[83] The electron added was localized on one single Ce(III) cation.;1; Charge
compensation is applied on these non-neutral unit cells by the addition of a
homogeneous background charge. This is sensitive to the vacuum width,
which must be kept constant on the different calculations.
[84] C. Sachs, M. Hildebrand, S. Volkening, J. Wintterlin, G. Ertl, Science 293
(2001) 1635–1638.

312

M. Capdevila-Cortada et al. / Applied Catalysis B: Environmental 197 (2016) 299–312

¨
[85] C. Sachs, M. Hildebrand, S. Volkening, J. Wintterlin, G. Ertl, J. Chem. Phys. 116
(2002) 5759.
[86] K. Reuter, M. Scheffler, Phys. Rev. Lett. 90 (2003) (046103).
[87] D. Teschner, G. Novell-Leruth, R. Farra, A. Knop-Gericke, R. Schlögl, L.
Szentmiklósi, et al., Nat. Chem. 4 (2012) 739–745.
[88] J. Carrasco, N. Lopez, F. Illas, Phys. Rev. Lett. 93 (2004) 225502.
[89] M. Ganduglia-Pirovano, J. Da Silva, J. Sauer, Phys. Rev. Lett. 102 (2009)
(026101).
[90] X. Wang, A. Chaka, M. Scheffler, Phys. Rev. Lett. 84 (2000) 3650–3653.
[91] M.V. Bollinger, K.W. Jacobsen, J.K. Nørskov, Phys. Rev. B 67 (2003), 085410.
[92] Z.-G. Yan, C.-H. Yan, J. Mater. Chem. 18 (2008), 5046.
[93] C. Noguera, J. Phys. Condens. Matter 12 (2000) R367–R410.
[94] G.S. Herman, Phys. Rev. B 59 (1999) 14899–14902.
[95] Y.J. Kim, Y. Gao, G.S. Herman, S. Thevuthasan, W. Jiang, D.E. McCready, et al.,
J. Vac. Sci. Technol. A Vac. Surf. Film 17 (1999) 926.
[96] H. Nörenberg, J.H. Harding, Surf. Sci. 477 (2001) 17–24.
[97] M. Nolan, S. Grigoleit, D.C. Sayle, S.C. Parker, G.W. Watson, Surf. Sci. 576
(2005) 217–229.
[98] P.M. Albrecht, D. Jiang, D.R. Mullins, J. Phys. Chem. C 118 (2014) 9042–9050.
[99] P. Geysermans, F. Finocchi, J. Goniakowski, R. Hacquart, J. Jupille, Phys.
Chem. Chem. Phys. 11 (2009) 2228–2233.
[100] J.V. Lauritsen, S. Porsgaard, M.K. Rasmussen, M.C.R. Jensen, R. Bechstein, K.
Meinander, et al., ACS Nano 5 (2011) 5987–5994.
[101] D.O. Scanlon, N.M. Galea, B.J. Morgan, G.W. Watson, J. Phys. Chem. C 113
(2009) 11095–11103.
[102] M. Nolan, Chem. Phys. Lett. 499 (2010) 126–130.
[103] J. Goniakowski, F. Finocchi, C. Noguera, Rep. Prog. Phys. 71 (2008) 016501.
[104] P. Oliver, G. Watson, S. Parker, Phys. Rev. B 52 (1995) 5323–5329.
[105] M. Salciccioli, Y. Chen, D.G. Vlachos, J. Phys. Chem. C 114 (2010)
20155–20166.
[106] M. Salciccioli, S.M. Edie, D.G. Vlachos, J. Phys. Chem. C 116 (2012)
1873–1886.
[107] R. García-Muelas, N. López, J. Phys. Chem. C 118 (2014) 17531–17537.
[108] M. Capdevila-Cortada, N. López, ACS Catal. (2015) 6473–6480.
[109] H.-X. Mai, L.-D. Sun, Y.-W. Zhang, R. Si, W. Feng, H.-P. Zhang, et al., J. Phys.
Chem. B 109 (2005) 24380–24385.
[110] E. Aneggi, D. Wiater, C. de Leitenburg, J. Llorca, A. Trovarelli, ACS Catal. 4
(2014) 172–181.
[111] E. Aneggi, J. Llorca, M. Boaro, A. Trovarelli, J. Catal. 234 (2005) 88–95.
[112] G. Vilé, S. Colussi, F. Krumeich, A. Trovarelli, J. Pérez-Ramírez, Angew. Chem.
Int. Ed. 53 (2014) 12069–12072.
[113] M. Nolan, G.W. Watson, J. Phys Chem B 110 (2006) 16600–16606.
[114] M. Nolan, S.C. Parker, G.W. Watson, Surf. Sci. 600 (2006) 175–178.
[115] R. Farra, S. Wrabetz, M.E. Schuster, E. Stotz, N.G. Hamilton, A.P. Amrute,
et al., Phys. Chem. Chem. Phys. 15 (2013) 3454–3465.
[116] M. Huang, S. Fabris, Phys. Rev. B 75 (2007) 081404.
[117] H.-T. Chen, J.-G. Chang, H.-L. Chen, S.-P. Ju, J. Comput. Chem. 30 (2009)
2433–2442.
[118] Z. Wu, M. Li, J. Howe, H.M. Meyer, S.H. Overbury, Langmuir 26 (2010)
16595–16606.
[119] O.H. Laguna, A. Pérez, M.A. Centeno, J.A. Odriozola, Appl. Catal. B Environ.
176–177 (2015) 385–395.
[120] T. Vinodkumar, B.G. Rao, B.M. Reddy, Catal. Today 253 (2015) 57–64.
[121] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003)
935–938.
[122] R. Si, M. Flytzani-Stephanopoulos, Angew. Chem. Int. Ed. 47 (2008)
2884–2887.
[123] Z. Zhou, S. Kooi, M. Flytzani-Stephanopoulos, H. Saltsburg, Adv. Funct.
Mater. 18 (2008) 2801–2807.
[124] J.A. Rodriguez, P. Liu, J. Hrbek, J. Evans, M. Pérez, Angew. Chem. Int. Ed. 46
(2007) 1329–1332.

[125] H.Y. Kim, G. Henkelman, J. Phys. Chem. Lett. 4 (2013) 216–221.
[126] M. Fernández-Garcı´a, A. Martı´nez-Arias, L.N. Salamanca, J.M. Coronado, J.A.
ı
ı
Anderson, J.C. Conesa, et al., J. Catal. 187 (1999) 474–485.
[127] S. Hinokuma, H. Fujii, M. Okamoto, K. Ikeue, M. Machida, Chem. Mater. 22
(2010) 6183–6190.
[128] M. Kurnatowska, L. Kepinski, W. Mista, Appl. Catal. B Environ. 117 (2012)
135–147.
[129] R.V. Gulyaev, E.M. Slavinskaya, S.A. Novopashin, D.V. Smovzh, A.V.
Zaikovskii, D.Y. Osadchii, et al., Appl. Catal. B Environ. 147 (2014) 132–143.
[130] G.N. Vayssilov, Y. Lykhach, A. Migani, T. Staudt, G.P. Petrova, N. Tsud, et al.,
Nat. Mater. 10 (2011) 310–315.
[131] D. Fernández-Torre, K. Ko´ mider, J. Carrasco, M.V. Ganduglia-Pirovano, R.
s
Pérez, J. Phys. Chem. C 116 (2012) 13584–13593.
[132] M. Boaro, C. de Leitenburg, G. Dolcetti, A. Trovarelli, J. Catal. 193 (2000)
338–347.
[133] J. Qi, J. Chen, G. Li, S. Li, Y. Gao, Z. Tang, Energy Environ. Sci. 5 (2012) 8937.
[134] T. Schmidt, A. Wolf, O.F.-K. Sch T., Westermann C., Mondelli J. Perez-Ramirez
et al., (2013) WO2013004651.
[135] T. Schmidt, A., Wolf, O.F.-K. Schluter, T., Westermann, C., Mondelli, J.
Perez-Ramirez, et al., (2013) WO2013004649.
[136] N. López, J. Gómez-Segura, R.P. Marín, J. Pérez-Ramírez, J. Catal. 255 (2008)
29–39.
[137] D. Crihan, M. Knapp, S. Zweidinger, E. Lundgren, C.J. Weststrate, J.N.
Andersen, et al., Angew. Chem. Int. Ed. 47 (2008) 2131–2134.
[138] H. Over, J. Phys. Chem. C 116 (2012) 6779–6792.
[139] H. Over, Chem. Rev. 112 (2012) 3356–3426.
[140] M. Moser, C. Mondelli, T. Schmidt, F. Girgsdies, M.E. Schuster, R. Farra, et al.,
Appl. Catal. B Environ. 132–133 (2013) 123–131.
[141] M. Moser, I. Czekaj, N. López, J. Pérez-Ramírez, Angew. Chem. Int. Ed. 53
(2014) 8628–8633.
[142] M. García-Melchor, N. López, J. Phys Chem. C 118 (2014) 10921–10926.
[143] D. Fernández-Torre, J. Carrasco, M.V. Ganduglia-Pirovano, R. Pérez, J. Chem.
Phys. 141 (2014) 014703.
[144] F.R. Negreiros, M.F. Camellone, S. Fabris, J. Phys. Chem. C. 119 (2015)
21567–21573.
[145] J. Höcker, T.O. Mentes, A. Sala, A. Locatelli, T. Schmidt, J. Falta, et al., Adv.
¸
Mater. Interfaces 2 (2015) 1500314.
[146] Z. Hu, H. Metiu, J. Phys. Chem. C. 116 (2012) 6664–6671.
[147] M. García-Melchor, L. Bellarosa, N. López, ACS Catal. 4 (2014) 4015–4020.
[148] J. Carrasco, G. Vilé, D. Fernández-Torre, R. Pérez, J. Pérez-Ramírez, M.V.
Ganduglia-Pirovano, J. Phys. Chem. C 118 (2014) 5352–5360.
[149] E.W. Zhao, H. Zheng, R. Zhou, H.E. Hagelin-Weaver, C.R. Bowers, Angew.
Chem. Int. Ed. 54 (2015) 14270–14275.
[150] E.J. Grootendorst, R. Pestman, R.M. Koster, V. Ponec, J. Catal. 148 (1994)
261–269.
[151] Y. Sakata, V. Ponec, Appl. Catal. A Gen. 166 (1998) 173–184.
[152] S. Chrétien, H. Metiu, J. Phys. Chem. C 119 (2015) 19876–19882.
[153] N. López, C. Vargas-Fuentes, Chem. Commun. 48 (2012) 1379–1391.
[154] G. Novell-Leruth, G. Carchini, N. López, J. Chem. Phys. 138 (2013), 194706.
[155] T. Yokoyama, N. Yamagata, Appl. Catal. A Gen. 221 (2001) 227–239.
[156] M. Chong, D. Cheng, L. Liu, F. Chen, X. Zhan, Catal. Lett. 114 (2007) 198–201.
[157] H.-Z. Zhu, Y.-M. Lu, F.-J. Fan, S.-H. Yu, Nanoscale 5 (2013) 7219–7223.
[158] M. Badlani, I.E. Wachs, Catal. Lett. 75 (2001) 137–149.
[159] D. Kulkarni, I.E. Wachs, Appl. Catal. A Gen. 237 (2002) 121–137.
[160] M. Capdevila-Cortada, M. García-Melchor, N. López, J. Catal. 327 (2015)
58–64.
[161] L. Zhang, Z. Wu, N.C. Nelson, A.D. Sadow, I.I. Slowing, S.H. Overbury, ACS
Catal. (2015) 6426–6435.
[162] M.A. Pacheco, C.L. Marshall, Energy Fuels 11 (1997) 2–29.
[163] P. Tundo, M. Selva, Acc. Chem. Res. 35 (2002) 706–716.
[164] M. Aresta, A. Dibenedetto, Dalton Trans. (2007) 2975–2992.