Journal of Catalysis 327 (2015) 58–64

This document version is made available version of a Published Work license http://creativecommons.org/licenses/by-nc-nd/4.0/.
This manuscriptis the Accepted ManuscriptunderContents lists available at ScienceDirect final form in Journal of Catalysis 327
the CC-BY-NC-ND 4.0 that appeared in
(2015) 58–64. version: http://dx.doi.org/10.1016/j.jcat.2015.04.016.
Link to the finalhttp://dx.doi.org/10.1016/j.jcat.2015.04.016.

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

Unraveling the structure sensitivity in methanol conversion on CeO2:
A DFT + U study
Marçal Capdevila-Cortada a,⇑, Max García-Melchor a,b, Núria López a,⇑
a
b

Institute of Chemical Research of Catalonia (ICIQ), Av. Països Catalans, 16, 43007 Tarragona, Spain
SUNCAT Center for Interface Science and Catalysis, SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA

a r t i c l e

i n f o

Article history:
Received 27 November 2014
Revised 16 March 2015
Accepted 19 April 2015

Keywords:
CeO2
Structure sensitivity
Methanol oxidation
Formaldehyde oxidation
Dehydrogenation
H2 formation
DFT
Reaction mechanisms

a b s t r a c t
Methanol decomposes on oxides, in particular CeO2, producing either formaldehyde or CO as main
products. This reaction presents structure sensitivity to the point that the major product obtained
depends on the facet exposed in the ceria nanostructures. Our density functional theory (DFT) calculations illustrate how the control of the surface facet and its inherent stoichiometry determine the sole formation of formaldehyde on the closed surfaces or the more degraded by-products on the open facets (CO
and hydrogen). In addition, we found that the regular (1 0 0) termination is the only one that allows
hydrogen evolution via a hydride–hydroxyl precursor. The fundamental insights presented for the
differential catalytic reactivity of the different facets agree with the structure sensitivity found for ceria
catalysts in several reactions and provide a better understanding on the need of shape control in selective
processes.
Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction
Cerium oxide, CeO2, is a versatile and economically attractive
oxide used in solid oxide fuel cells [1], biomedicine [2], and catalysis [3]. Either as a catalyst or as a support, it is widely used in a
vast number of oxygen-related reactions such as the water–gas
shift (WGS) [4,5], the preferential oxidation of CO [6,7], SO2 oxidation [8,9], and the HCl conversion to Cl2 [10]. All these applications
are due to the particular redox character of ceria that permits to
cycle between the Ce4+ and Ce3+ states, which involves oxygen
vacancy formation [11] and diffusion [10,12]. The oxygen storage
capacity (OSC) of CeO2 varies with the exposing surface [13], and
hence, in nanoparticles it ultimately depends on their morphology.
Recently, ceria has been also shown to present a high activity
and selectivity in hydrogenation processes [10,14]. More surprisingly, the oxidation and hydrogenation capacities of different
nanostructured CeO2 materials have been found to present a markedly structure sensitivity [15]. Thus, while hydrogenation takes
place preferentially on (1 1 1) facet orientations, oxidation reactions are favored on the more open (1 1 0) and (1 0 0) facets exposed
in rods and cubes, respectively. Ceria nanoparticles have also been
⇑ Corresponding authors.
E-mail addresses: mcapdevila@iciq.es (M. Capdevila-Cortada), nlopez@iciq.es
(N. López).
http://dx.doi.org/10.1016/j.jcat.2015.04.016
0021-9517/Ó 2015 Elsevier Inc. All rights reserved.

found to exhibit different activities in soot combustion depending
on the exposed facet [16].
Among oxidation processes, the conversion of methanol to
formaldehyde is considered an ideal test for characterizing the catalytic behavior of metal oxides [17,18]. The two main by-products
of this reaction are water and hydrogen. Therefore, the catalytic
activity and selectivity of a given material may be a result of the
interplay between its oxidation and hydrogenation abilities.
In industry, methanol is oxidized to produce formaldehyde
through the so-called Formox process [19]. There, methanol and
oxygen react at temperatures around 525–675 K in the presence
of iron oxide mixed with molybdenum or vanadium oxide according to the following equations:

CH3 OH þ 1=2O2 ! HCHO þ H2 O

ð1Þ

CH3 OH ! HCHO þ H2

ð2Þ

At higher temperatures, 875 K, silver can also act as catalyst,
probably in the form of a suboxide [20–22]. Hence, the oxidation
of methanol to formaldehyde has been extensively studied on
metal oxides, especially on titanium and vanadium oxides
[23–28]. In the case of ceria, most of the investigations have been
focused on the (1 1 1) facet [29–33]. On this stoichiometric surface,
methanol converts into formaldehyde, at ca. 550 K, via a dehydrogenation process that involves surface oxygens followed by the

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

59

subsequent reduction of two Ce4+ ions per methanol molecule [29].
The presence of oxygen vacancies was recently reported to have a
detrimental effect on the reaction selectivity as a result of the competition between formaldehyde desorption and its further oxidation to CO [34]. In the same work, the (1 0 0) surface was also
shown to have a reduced selectivity toward formaldehyde. Both
HCHO and CO were found to be the main products on
CeO2(1 0 0), while on the reduced CeOx(1 0 0) almost all methanol
was converted to CO [34]. Temperature-programmed desorption
(TPD) experiments on nanocrystals with different nanoshapes,
namely octahedra, cubes, and rods, showed that the latter is the
most reactive shape (but non-selective to formaldehyde) one,
evolving CO and H2 at T < 583 K [35]. On the other hand, Zhou
and Mullins reported that formaldehyde does not decompose on
stoichiometric CeO2(1 1 1), but does evolve to CO and H2 on the
same reduced surface at temperatures above 500 K [36]. Overall,
ceria appears to be very sensitive to the particular choice of active
surfaces.
Despite all the above and other theoretical works reported in
the literature [37–42], up to date there is no complete study
assessing the selectivity of methanol dehydrogenation on the most
representative ceria facets and proving their different behavior as
experimentally observed. Herein, we present a thorough mechanistic study on the selective conversion of methanol to formaldehyde and its subsequent conversion to CO. To account for the
particular morphology of the three common nanoshapes above,
we have studied the three lowest index and energy surfaces
(1 1 1), (1 1 0), and (1 0 0) (Fig. 1), as these are the most exposed
facets in each nanoshape, respectively. The desorption of HCHO,
CO, and H2 from each surface is also discussed and compared to
the reported TPD results on the stoichiometric ceria surfaces
[29,33,34]. Overall, this work provides the clues behind the structure sensitivity in the aforementioned oxidation process.

acalc = 5.497 Å, in good agreement with both the experimental
value and previous computational studies [53]. To maintain the
stoichiometry of the CeO2(1 0 0) surface and avoid the dipole
moment normal to the surface, a half oxygen monolayer was transferred from the surface to the bottom of the slab [54]. The (1 1 1),
(1 1 0), and (1 0 0) surfaces were modeled as (2 Â 2) periodically
repeated slabs consisting of four, five, and three O–Ce–O layers,
respectively, separated by 15 Å of vacuum space, and were optimized using a C-centered k-point mesh denser than 0.38 ÅÀ1.
The layers allowed to relax in each surface were the five outermost
layers in CeO2(1 1 1), the three top-most layers in CeO2(1 1 0), and
the four outermost layers in CeO2(1 0 0), whereas the rest of atoms
were kept fixed to their bulk positions.
Transition states were located by means of the climbing image
nudged elastic band (CI-NEB) method [55]. The nature of all reaction minima and transition states was confirmed by means of
numerical frequency analyses.
3. Results
3.1. Dehydrogenation of methanol to formaldehyde on CeO2
We investigated the conversion of methanol to formaldehyde
on the three lowest energy surfaces corresponding to the low-index planes (1 1 1), (1 1 0), and (1 0 0). Their surface and relaxation
energies are presented in the Supplementary Material (Table S1).
The reaction consists in the sequence of the following elementary steps:

CH3 OH þ Ã ! CH3 OHÃ

ðR1Þ

CH3 OHÃ þ Ã ! CH3 OÃ þ HÃ

ðR2Þ

CH3 OÃ þ Ã ! HCHOÃ þ HÃ

ðR3Þ

2. Computational details

HCHOÃ ! HCHO þ Ã

ðR4Þ

All the calculations reported in the present study were performed at the DFT + U level using the Vienna Ab initio Simulation
Package (VASP, version 5.3.3)
[43,44].
The Perdew–
Becke–Ernzerhof (PBE) [45] functional was used in combination
with an effective U term, Ueff, of 4.5 eV (this term is defined as
the difference between the Coulomb, U, and the exchange, J, terms)
[46], as it has been previously proven to provide satisfactory
results [47–50]. The purpose of adding this Hubbard term U is to
diminish the self-interaction error and to properly localize the Ce
4f states. Projector-augmented wave (PAW) pseudopotentials
[51] were used to describe the core electrons with a plane-wave
cutoff energy of 500 eV for the valence electrons (i.e., 5s, 5p, 4f,
6s for Ce atoms, and 2s, 2p, for O and C atoms).
Ceria presents a fluorite structure, with lattice parameter
aexp = 5.411 Å [52]. We optimized the lattice parameter using a
dense C-centered 7 Â 7 Â 7 k-point mesh, leading to the value

Methanol physisorbs on ceria (R1) and dissociates (R2) giving
rise to a chemisorbed methoxy and surface OH groups. The methoxy intermediate then leads to formaldehyde through a C–H bond
breaking (R3) with the concomitant hydrogenation of another surface oxygen. Finally, formaldehyde desorbs from the surface to the
gas phase (R4).
Reaction (R2) is controlled by the acid–base properties of the
surface and does not involve any change in the oxidation state of
the reactant fragments or the surface metal atoms. In contrast,
reaction (R3) can be seen as the main redox step as two Ce4+
cations are reduced to Ce3+ after the C–H scission and the subsequent formation of a second surface OH group. The discrimination
in terms of acid–base and redox steps has been recently found to
be crucial to tune the catalytic activity of oxides [56].
The calculated energy profiles for the methanol to formaldehyde conversion on the different CeO2 surfaces are presented in

Fig. 1. Top views of the three lowest index CeO2 surfaces. The outmost oxygen (in red) and cerium (in green) layers are highlighted, while the rest of atoms are shown dashed.
(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

60

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

Fig. 2. Methanol adsorption energies range between À0.5 and
À1 eV, whereas the first deprotonation requires very low energies,
<0.1 eV, rendering a more stable methoxy intermediate (I1). From
this point, the relative energy barrier for the next step (R3) on all
the surfaces is around 1 eV. More specifically, the (1 0 0) surface
exhibits the lowest energy barrier (0.89 eV), whereas the (1 1 1)
and (1 1 0) surfaces have barriers of 1.03 and 1.14 eV, respectively.
In all the cases, the transition state associated with this process
(TS2) involves the reduction of one single Ce4+ to Ce3+, whereas
the other unpaired electron resulting from the C–H scission is delocalized over these C and H atoms and the involved surface oxygen.
On the (1 1 1) surface, the C–H scission in I1 results in the physisorbed formaldehyde as intermediate, as also noted by Kropp
and Paier [39]. The next barrier from the physisorbed to chemisorbed conformation (I2) is lower than 0.05 eV (see
Supplementary Material Fig. S2). Alternatively, on the (1 1 0), the
C–H bond cleavage leads to the formation of an intermediate
where HCHO is directly chemisorbed on the hydroxylated surface
group. From this species, the following proton transfer to the adjacent surface oxygen leading to I2 only requires 0.02 eV.
Importantly, the energy barrier for the direct formation of I2 from
I1 on the (1 1 0) was previously reported [38] as 0.9 eV higher than
ours. We found that the direct mechanism from I1 to I2 via the
transition state TS2 is only favored on the (1 0 0) surface.
In I2, formaldehyde binds to the surface in a g2-like configuration, where the C atom is coordinated to an oxygen from the lattice
and the oxygen from HCHO is bound to a surface cation. This species is rather stable on all the surfaces, with binding energies of

1.23, 0.71, and 2.30 eV for the (1 1 1), (1 1 0), and (1 0 0) facets,
respectively. Hence, formaldehyde desorption from the (1 1 1)
and (1 1 0) surfaces is expected to be significantly easier compared
to (1 0 0). That is, the lifetime of HCHO on the (1 0 0) surface may be
higher than on the other surfaces, which might eventually lead to
surface poisoning.
Overall, according to the energy diagram in Fig. 2, the most
energy-demanding step in the methanol to formaldehyde conversion on the (1 1 0) surface is the C–H bond breaking, whereas on
the (1 0 0) and (1 1 1) surfaces it corresponds to formaldehyde
desorption.
3.2. Dehydrogenation of formaldehyde to carbon monoxide on CeO2
The reaction mechanism for the dehydrogenation of formaldehyde to CO on the clean (1 1 1), (1 1 0), and (1 0 0) surfaces was analyzed assuming the following elementary reaction steps:

HCHO þ Ã ! HCHOÃ

ðR4 Þ

HCHOÃ þ Ã ! CHOÃ þ HÃ

ðR5Þ

CHOÃ þ Ã ! COÃ þ HÃ

ðR6Þ

COÃ ! CO þ Ã

ðR7Þ

0

In this case, the initial step (R40 ) is the reverse process of the
above (R4) reaction. After the adsorption of formaldehyde, the
reaction then proceeds through two consecutive H abstractions

Fig. 2. Top panel: energy profile for the MeOH to HCHO conversion on CeO2(1 1 1), CeO2(1 1 0), and CeO2(1 0 0). Bottom panel: optimized geometries for the intermediates and
transition states. The most relevant distances in the transition states can be found in Fig. S1. Color code as in Fig. 1 with the addition of C (black), O from methanol (dark red),
and H (white). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

(R5) and (R6), which can be again considered as coupled redox and
acid–base steps. Finally, the generated CO desorbs from the surface
to the gas phase. The reaction energy profiles obtained for the three
CeO2 surfaces are shown in Fig. 3. The initial formaldehyde
chemisorption value (A2) on the (1 1 1), (1 1 0), and (1 0 0) surfaces
is À0.88, À1.25, and À1.84 eV, respectively. Note that the lower
binding energies for the (1 1 1) and (1 0 0) facets compared to I2
in the methanol dehydrogenation (Fig. 2) are due to the formation
of hydrogen bonds between HCHO and the surface hydroxyls [57].
The energy difference can be estimated to the formation of two
hydrogen bonds [58]. On (1 1 0), A2 binding energy is higher than
I2 because the hydroxylated surface presents intrasurface hydrogen bonds (between Olat–H and another Olat) and formaldehyde
adsorbs on the acceptor oxygen disrupting the H-bond surface network. From A2, the first H abstraction in HCHO is a highly exothermic process that releases more than -2 eV and results in the
chemisorbed formyl and one surface hydroxyl group (I3). The activation energy for this step requires only 0.66 eV on the (1 0 0) surface, but is very energy demanding on both (1 1 1) and (1 1 0), that
is, 1.53 and 1.87 eV, respectively. Similar to the H stripping in
methanol (R3), in this first transition state (TS3), one Ce4+ is
reduced to Ce3+, while the other unpaired electron is delocalized
over the formaldehyde and the reacting surface oxygen. The

61

second deprotonation step (R6) is endothermic and produces physisorbed CO and two surface OH groups. The energy barriers for
this step on the (1 1 1) and (1 0 0) surfaces are 1.60 and 1.64 eV,
respectively, and 2.00 eV on the (1 1 0). Finally, the CO desorption
from the partially hydroxylated surfaces requires only 0.1–0.3 eV.
It is worth noting that CO can also form a carbonate structure
CO2- with two surface oxygens. This structure is 3.5 eV more stable
3
than the physisorbed CO [59,60], although its formation requires
two accessible surface oxygens. On hydroxylated surfaces, HCOÀ
3
and H2CO3 could be formed, which are 1.8 eV more stable and
1.3 eV less stable, respectively, than physisorbed CO. Therefore,
highly hydroxylated surfaces are less prone to form carbonates,
and their formation (HCOÀ and H2CO3) would not introduce signif3
icant alterations to the proposed mechanism.
Alternatively, the oxidation of the chemisorbed formaldehyde
on the (1 0 0) surface can also follow the blue dashed path
shown in Fig. 3. In this case, the H atom resulting from the first
C–H scission accommodates between two surface Ce4+ cations as
a hydride. Thus, unlike the previous mechanism, this alternative
pathway does not involve the reduction of any surface cerium
cations [61]. This first H abstraction is endothermic by 0.66 eV
and requires an activation energy of 0.75 eV. Similarly, the subsequent C–H bond cleavage is also endothermic by 1.06 eV and

Fig. 3. Top panel: energy profile for the HCHO conversion to CO upon adsorption on CeO2(1 1 1), CeO2(1 1 0), and CeO2(1 0 0). The alternative hydride cleavage mechanism on
the latter surface is shown in dashed blue lines. Bottom panel: optimized geometries for the intermediates and transition states. The most relevant distances in the transition
states can be found in Fig. S3. Color code as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

62

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

has an energy barrier of 1.82 eV. Although this alternative mechanism is thermodynamically less favored than the previous one
(solid lines in Fig. 3), it might play a role under high coverage
regimes, where all surface oxygens would be poisoned by chemisorbed formaldehyde. It is worth mentioning that this mechanism is enabled only on the particularly defective (1 0 0)
surface, and is unlikely to occur on the other two surfaces
(Fig. S4).
On the whole, Fig. 3 shows that the most energy-demanding
step for the oxidation of formaldehyde to CO is the same for the
three ceria facets and corresponds to the cleavage of the second
C–H bond (R6). The activation energies for the (R5) and (R6) steps
range between 1.5 and 2 eV, except for (R5) on the (1 0 0), which is
only 0.66 eV. On (1 1 1) and (1 1 0), the energies required to reach
TS3 are ca. 0.6 eV higher than the desorption of formaldehyde from
these surfaces.
3.3. H2 evolution from a hydroxylated CeO2 surface
Once methanol decomposes to either HCHO or CO, the CeO2
surfaces become partially hydroxylated, which can eventually lead
to hydrogen evolution under certain reaction conditions. In a previous work, we showed that H2 dissociation on the (1 1 1) surface
affords the homolytic product although through an heterolytic
path involving the formation of an ion pair as precursor [50]. The
reaction profile for that reverse process on the three CeO2 surfaces
is presented in Fig. 4. In the first step, one hydrogen of the two surface hydroxyl groups is transferred to a neighboring cerium atom
with the concomitant oxidation of one Ce3+ to Ce4+. This process
is highly endothermic, the reaction energies being 3.22 eV on
(1 1 1), 3.61 eV on the (1 1 0), and 2.71 eV on the (1 0 0), and the
associated barriers 3.55, 4.30, and 3.53 eV, respectively.
Interestingly, the resulting product I5 features a completely different nature depending on the considered facet, see Fig. 4. While on
the (1 1 1) and (1 1 0) surfaces, it can be described as an ion pair
with a H–H distance of 1.29 and 1.65 Å, respectively, on the
(1 0 0) surface, this ion pair is not stable and the proton and the
hydride fall apart so that the latter accommodates between two
Ce4+. This is a consequence of the intrinsic oxygen depletion of
the (1 0 0) surface. These significant structural differences are also
reflected in substantially different stabilities. More specifically,
the ion pairs formed on the (1 1 1) and (1 1 0) surfaces have energies
of 0.75 and 0.45 eV above the gas-phase reference, respectively,
whereas the charge separated form on (1 0 0) is À0.29 eV more
stable. From these precursors, the activation energy for the following H–H bond formation on (1 1 1) and (1 1 0) is very low, 0.1 eV,
but it increases to 0.62 eV on the (1 0 0).
Thus, hydrogen evolution on the (1 1 1) and (1 1 0) surfaces is
very difficult as H is always stored as hydroxyl on the regular stoichiometric surfaces. In contrast, as mentioned above, the situation on the (1 0 0) surface might be different. In fact, when
methanol is adsorbed on this surface, a large coverage of hydroxyls is likely; therefore, the hydrogen stripping from A2 can divert
toward the production of the intermediate I3 (dashed blue line
mechanism), where one of the hydrogens is transferred as a
hydride. This situation is equivalent to that presented in structure
I5, from which we have shown that H2 can be evolved with a
small energy barrier.
The evolution of hydrogen from the hydroxylated surface might
also compete with the formation of water and its subsequent desorption, thus creating an oxygen vacancy. This overall process is
endothermic on all three surfaces, by 2.22, 2.11, and 2.34 eV on
(1 1 1), (1 1 0), and (1 0 0), respectively. Despite being a highly
endothermic process, it is more likely to occur than the hydrogen
evolution from the hydroxylated surface.

4. Discussion
In this section, we present a detailed comparison between our
DFT simulations and the available experiments in the literature.
The theoretical results exposed in the previous sections are perfectly consistent with the reported TPD experiments on stoichiometric (1 1 1) and (1 0 0) CeO2 thin films [29,33,34]. In the case of
the (1 1 1) surface, the only desorption products observed are
formaldehyde and methanol. Methanol desorbs at 220 K (main
peak) and 560 K, and formaldehyde at 570 K. Similarly, TPD experiments on the (1 0 0) surface show the same product distribution
but with the additional observation of CO, CO2, H2O, and H2. In this
case, methanol appears at 560–580 K, with a large peak of
formaldehyde and CO coupled with a smaller peak assigned to
H2 and H2O. Moreover, the reaction on this surface was reported
to occur rapidly and without detecting any formyl or formate intermediates. The desorption peaks for HCHO, CO, and H2 for both the
(1 1 1) and (1 0 0) facets proceed prior to any indication of surface
reduction (related to CO2 or H2O desorption peaks).
The above structure sensitivity observed in experiments can be
rationalized as follows on the basis of our theoretical calculations.
Even though we showed that the sequential removal of H from
methanol on the (1 1 1) and (1 1 0) surfaces is energetically costly,
it derives on adsorbed formaldehyde, which can be desorbed more
easily than the further removal of H from it. That is, the adsorption
energy of A2 in Fig. 3 is significantly smaller than the energy
required to reach TS3. As a result, methanol selectively reacts
toward formaldehyde on the (1 1 1) and (1 1 0) surfaces, the latter
being more reactive (lower TS2) than the former. Major differences, however, can be found in reaction conversions. On the
(1 1 1) surface, methanol desorption and its transformation to
formaldehyde are competing processes, but the weak interaction
between methanol and the surface results in a large peak in the
TPD at low temperatures. Still, some methanol molecules survive
on the surface long enough to react, possibly pinned by undersurface vacancies [62,63]. This is clearly evidenced in the TPD experiments where the low-temperature peak (methanol desorption)
is larger than the others for the (1 1 1) surface under stoichiometric
conditions. In contrast, on the (1 0 0) surface, the energy demand to
reach TS3 is much lower than the adsorption energy of A2, and
therefore, the conversion of formaldehyde to CO is favored compared to formaldehyde desorption. These results also account for
the unique experimental observation of CO on the most oxidized
CeO2(1 0 0) surfaces [34]. It was also shown that CO is more likely
to appear on the more defective ceria surfaces. This can be also
explained with the expected increase of the A2 adsorption energy
on the O-vacancy sites, as these may trap HCHO and impede its
desorption, thus favoring CO evolution.
This is in fact the case for the (1 1 1) surface, where methanol
adsorbs dissociatively on the defective site with an adsorption
energy of À2.12 eV [64]. The subsequent C–H cleavage to form
the chemisorbed formaldehyde demands 1.32 eV and is an
exothermic process by 0.34 eV. Formaldehyde desorption from this
surface requires 1.04 eV, and the energy barrier for the subsequent
C–H bond breaking is only 0.58 eV. Hence, the oxygen deficiency
(1 1 1) traps formaldehyde and enables methanol conversion to
CO, as suggested before and shown in the reported TPD experiments [34].
When it comes to hydrogen, experiments show that only
CeO2(1 0 0) is able to evolve molecular hydrogen at about 600 K
[34]. Indeed, from Fig. 4, it can be observed that the surface OH
groups are too robust to produce H2 from any of the three surfaces.
However, on the (1 0 0), the intrinsic lack of surface oxygen makes
the situation less compromised as both hydrides and protons can
be stabilized by the surface Ce and O atoms. Hence, populating this

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

63

Fig. 4. Top panel: energy profile for H2 evolution from partially hydroxylated (0.5 ML) CeO2(1 1 1), CeO2(1 1 0), and CeO2(1 0 0). Bottom panel: optimized geometries for the
intermediates and transition states. Color code as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)

Fig. 5. Left: energy profile for the oxygen exchange between HCHO and a lattice oxygen from CeO2(1 1 1), CeO2(1 1 0), and CeO2(1 0 0). Right: optimized geometries for the
intermediates and transition states. Color codes as in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)

state is crucial for CeO2 to be able to generate H2. This also explains
why H2 evolution is also observed experimentally on defective
CeO2(1 1 1) surfaces [34]. Our results also point out that formaldehyde oxidation on the (1 0 0) surface can take place through a less
thermodynamically favored alternative mechanism and without
involving the reduction of surface Ce ions (blue dashed path in
Fig. 3). Upon CO desorption, this reaction path leads to the partially
hydroxylated surface I5, which might enhance H2 evolution on the

(1 0 0) surface, especially under high coverage regimes with all the
surface oxygens occupied by chemisorbed formaldehyde. This
mechanism, accessible for the intrinsically O-deficient (1 0 0) surface, also explains why reduced (1 1 1) surfaces can also produce
H2. These reduced surfaces can trap hydride atoms and follow a
similar mechanism as the one we proposed for the (1 0 0) surface.
Finally, the reported TPD spectrum using isotope-labeled ceria
and methanol, Ce18O2(1 1 1) and CH16OH, shows that an equal
3

64

M. Capdevila-Cortada et al. / Journal of Catalysis 327 (2015) 58–64

amount of HCH16O and HCH18O is desorbed [29]. The same observation is made upon HCH16O adsorption on reduced
Ce18O1.75(1 1 1) [36]. The latter was attributed by the authors to
the adsorption on a vacancy site.
In order to shed light on these experimental findings, we also
investigated a reaction mechanism in which an O atom from the
chemisorbed HCHO is exchanged with one lattice oxygen from
the stoichiometric CeO2 surfaces (Fig. 5). On (1 1 1) and (1 1 0), this
mechanism occurs in an elementary step with an activation energy
of 0.40 and 0.17 eV. In contrast, on the (1 0 0) termination, the chemisorbed formaldehyde A2 gives rise to an intermediate (A3) that
is 0.07 eV more stable and where the lattice oxygen has diffused
and accommodates on a defective surface site. The energy barrier
required to afford A3 is also very low (0.08 eV). The low energy difference and activation energy between both A2 and A3 indicate
that these two intermediates may be in thermodynamic equilibrium. Therefore, the low activation energies found for this process
on the different CeO2 surfaces prove that no oxygen vacancies are
needed to desorb HCH18O upon HCH16O adsorption or CH16OH
3
dehydrogenation.
5. Conclusions
A detailed mechanistic study on the reported structure sensitivity of different CeO2 surfaces in the methanol conversion to
formaldehyde or syn-gas (CO + H2 mixtures) has been performed.
The reaction proceeds via the activation of the hydroxyl group by
the basic sites on the surface and followed by the subsequent stripping of hydrogen atoms from the methyl group. Our DFT + U calculations show that while close-packed surfaces mainly produce
formaldehyde, this intermediate is trapped in the O-defective
(1 0 0) surface, which finally produces CO and H2. Hydrogen evolution is allowed on the (1 0 0) facet, but not on the close-packed
(1 1 1) and (1 1 0) terminations. This is due to the inherent defective
sites in the former that stabilize two hydrogen atoms as hydroxyls
and hydrides, which eventually lead to H–H bond formation. Only
reduced (1 1 1) and (1 1 0) surfaces could present enough vacancy
sites to allow a similar chemistry on the close-packed facets.
These active sites in turn extend the lifetime of adsorbed HCHO
enabling its decomposition, and are ultimate responsible for CO
formation.
The present work sheds light on the structure sensitivity found
for the reaction and that is related to the inherent stoichiometries
of the different surfaces. This is more acute on oxides, in particular
reducible ones, as there are more parameters that are modified
when changing the facet orientations. Moreover, shape control is
a key to obtain desired selectivities, and thus, methods devoted
to a detailed shape selection are crucial to obtain a more selective
and greener chemistry.
Acknowledgments
This research has been supported by the ERC Starting Grant
(ERC-2010-StG-258406). We acknowledge BSC-RES for providing
generous computational resources.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2015.04.016.
References
[1] Z. Shao, S.M. Haile, Nature 431 (2004) 170.
[2] A.S. Karakoti, S. Singh, A. Kumar, M. Malinska, S.V.N.T. Kuchibhatla, K. Wozniak,
et al., J. Am. Chem. Soc. 131 (2009) 14144.

[3] A. Trovarelli, Catalysis by Ceria and Related Materials, Imperial College Press,
London, 2002.
[4] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.
[5] S. Hilaire, X. Wang, T. Luo, R. Gorte, J. Wagner, Appl. Catal. A Gen. 215 (2001)
271.
[6] M.M. Schubert, V. Plzak, J. Garche, R.J. Behm, Catal. Lett. 76 (2001) 143.
[7] O. Pozdnyakova, D. Teschner, A. Wootsch, J. Kröhnert, B. Steinhauer, H. Sauer,
et al., J. Catal. 237 (2006) 1.
[8] J.S. Yoo, A.A. Bhattacharyya, C.A. Radlowski, Ind. Eng. Chem. Res. 30 (1991)
1444.
[9] J.S. Yoo, A.A. Bhattacharyya, C.A. Radlowski, J.A. Karch, Appl. Catal. B Environ. 1
(1992) 169.
[10] A.P. Amrute, C. Mondelli, M. Moser, G. Novell-Leruth, N. López, D. Rosenthal,
et al., J. Catal. 286 (2012) 287.
[11] M. Nolan, S.C. Parker, G.W. Watson, Surf. Sci. 595 (2005) 223.
[12] D.A. Andersson, S.I. Simak, N.V. Skorodumova, I.A. Abrikosov, B. Johansson,
Proc. Natl. Acad. Sci. USA 103 (2006) 3518.
[13] J. Paier, C. Penschke, J. Sauer, Chem. Rev. 113 (2013) 3949.
[14] G. Vilé, B. Bridier, J. Wichert, J. Pérez-Ramírez, Angew. Chem. Int. Ed. 51 (2012)
8620.
[15] G. Vilé, S. Colussi, F. Krumeich, A. Trovarelli, J. Pérez-Ramírez, Angew. Chem.
Int. Ed. 53 (2014) 12069.
[16] E. Aneggi, D. Wiater, C. de Leitenburg, J. Llorca, A. Trovarelli, ACS Catal. 4 (2014)
172.
[17] J.M. Tatibouët, Appl. Catal. A Gen. 148 (1997) 213.
[18] M. Badlani, I.E. Wachs, Catal. Lett. 75 (2001) 137.
[19] http://www.formox.com/, November 2014.
[20] M. Schmid, A. Reicho, A. Stierle, I. Costina, J. Klikovits, P. Kostelnik, et al., Phys.
Rev. Lett. 96 (2006) 146102.
[21] J. Schnadt, A. Michaelides, J. Knudsen, R. Vang, K. Reuter, E. Lægsgaard, et al.,
Phys. Rev. Lett. 96 (2006) 146101.
[22] I. Karakaya, W.T. Thompson, J. Phase Equilibria 13 (1992) 137.
[23] J.L. Bronkema, D.C. Leo, A.T. Bell, J. Phys. Chem. C. 111 (2007) 14530.
[24] L.J. Burcham, I.E. Wachs, Catal. Today 49 (1999) 467.
[25] G. Deo, I.E. Wachs, J. Catal. 146 (1994) 323.
[26] T. Feng, J.M. Vohs, J. Catal. 208 (2002) 301.
[27] Y. Romanyshyn, S. Guimond, H. Kuhlenbeck, S. Kaya, R.P. Blum, H. Niehus,
et al., Top. Catal. 50 (2008) 106.
[28] Q. Wang, R.J. Madix, Surf. Sci. 496 (2002) 51.
[29] D.R. Mullins, M.D. Robbins, J. Zhou, Surf. Sci. 600 (2006) 1547.
[30] A. Siokou, R.M. Nix, J. Phys. Chem. B 103 (1999) 6984.
[31] J.J. Uhlrich, B. Yang, S. Shaikhutdinov, Top. Catal. 57 (2014) 1229.
[32] V. Matolín, J. Libra, M. Škoda, N. Tsud, K.C. Prince, T. Skála, Surf. Sci. 603 (2009)
1087.
[33] R.M. Ferrizz, G.S. Wong, T. Egami, J.M. Vohs, Langmuir 17 (2001) 2464.
[34] P.M. Albrecht, D.R. Mullins, Langmuir 29 (2013) 4559.
[35] Z. Wu, M. Li, D.R. Mullins, S.H. Overbury, ACS Catal. 2 (2012) 2224.
[36] J. Zhou, D.R. Mullins, Surf. Sci. 600 (2006) 1540.
[37] D. Mei, N.A. Deskins, M. Dupuis, Q. Ge, J. Phys. Chem. C 111 (2007) 10514.
[38] D. Mei, N.A. Deskins, M. Dupuis, Q. Ge, J. Phys. Chem. C 112 (2008) 4257.
[39] T. Kropp, J. Paier, J. Phys. Chem. C 118 (2014) 23690.
[40] M. Nolan, J. Chem. Phys. 139 (2013) 184710.
[41] T. Kropp, J. Paier, J. Sauer, J. Am. Chem. Soc. 136 (2014) 14616.
[42] B.-T. Teng, S.-Y. Jiang, Z.-X. Yang, M.-F. Luo, Y.-Z. Lan, Surf. Sci. 604 (2010) 68.
[43] G. Kresse, Phys. Rev. B 54 (1996) 11169.
[44] G. Kresse, J. Furthmüller, Comput. Mater. Sci. 6 (1996) 15.
[45] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[46] S.L. Dudarev, S.Y. Savrasov, C.J. Humphreys, A.P. Sutton, Phys. Rev. B 57 (1998)
1505.
[47] S. Fabris, S. de Gironcoli, S. Baroni, G. Vicario, G. Balducci, Phys. Rev. B 72
(2005) 237102.
[48] M. Ganduglia-Pirovano, J. Da Silva, J. Sauer, Phys. Rev. Lett. 102 (2009) 026101.
[49] R. Farra, M. García-Melchor, M. Eichelbaum, M. Hashagen, W. Frandsen, J.
Allan, et al., ACS Catal. 3 (2013) 2256.
[50] M. García-Melchor, N. López, J. Phys. Chem. C 118 (2014) 10921.
[51] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953.
[52] E. Kümmerle, G. Heger, J. Solid State Chem. 147 (1999) 485.
[53] J. Da Silva, M. Ganduglia-Pirovano, J. Sauer, V. Bayer, G. Kresse, Phys. Rev. B 75
(2007) 045121.
[54] P. Oliver, G. Watson, S. Parker, Phys. Rev. B 52 (1995) 5323.
[55] G. Henkelman, B.P. Uberuaga, H. Jónsson, J. Chem. Phys. 113 (2000) 9901.
[56] M. Moser, I. Czekaj, N. López, J. Pérez-Ramírez, Angew. Chem. Int. Ed. 53 (2014)
8628.
´
[57] D. Fernández-Torre, K. Kosmider, J. Carrasco, M.V. Ganduglia-Pirovano, R.
Pérez, J. Phys. Chem. C 116 (2012) 13584.
[58] G. Revilla-López, N. López, Phys. Chem. Chem. Phys. 16 (2014) 18933.
[59] M. Nolan, G.W. Watson, J. Phys. Chem. B 110 (2006) 16600.
[60] M. Nolan, S.C. Parker, G.W. Watson, Surf. Sci. 600 (2006) 175.
[61] This fact was confirmed by the obtained spin densities at this surface.
[62] F. Esch, S. Fabris, L. Zhou, T. Montini, C. Africh, P. Fornasiero, et al., Science 309
(2005) 752.
[63] G.E. Murgida, M.V. Ganduglia-Pirovano, Phys. Rev. Lett. 110 (2013) 246101.
[64] Calculations on defective ceria (1 1 1) were performed using the p(2 Â 2)
model slab described in the Computational details section. Nevertheless, the
results are in close agreement with those in Ref. [41].