The virtue of defects: Stable bromine production by catalytic oxidation of hydrogen bromide on titanium oxide

Article abstract of DOI:10.1002/anie.201404022

Rutile TiO₂ is a heavily investigated oxide with, to date, scarce applications in industrial catalysis. The inactivity of this material in oxidations has been related to its inability to dissociate molecular oxygen. Herein we show how rutile catalyzes the oxidation of HBr to Br₂ through defect states that are introduced during the reaction. The identification of active, stable, and abundant materials for bromine production is key to the future implementation of Br₂-mediated alkane functionalization processes. The catalytic properties of TiO₂ are discussed in comparison to expensive rutile-type oxides, such as RuO₂ and IrO₂, on the basis of surface characterization and molecular modeling.

Full text of DOI:10.1002/anie.201404022

.
Angewandte
Communications
DOI: 10.1002/anie.201404022
Defects in Catalysis
The Virtue of Defects: Stable Bromine Production by
Catalytic Oxidation of Hydrogen Bromide on Titanium
Oxide
Maximilian Moser, Izabela Czekaj, Nfflria López,* and Javier PØrez-Ramírez*
Dedicated to the MPI für Kohlenforschung on the occasion of its centenary
Angewandte
Chemie
8628
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633

Angewandte
Chemie
Abstract: Rutile TiO2 is a heavily investigated oxide with, to
date, scarce applications in industrial catalysis. The inactivity
of this material in oxidations has been related to its inability to
dissociate molecular oxygen. Herein we show how rutile
catalyzes the oxidation of HBr to Br2 through defect states
that are introduced during the reaction. The identification of
active, stable, and abundant materials for bromine production
is key to the future implementation of Br2-mediated alkane
functionalization processes. The catalytic properties of TiO2
are discussed in comparison to expensive rutile-type oxides,
such as RuO2 and IrO2, on the basis of surface characterization
and molecular modeling.
However, ruthenium-based catalysts present a clear draw-
back for a widespread implementation owing to rutheniumꢀs
relatively high and fluctuating price. This situation has
triggered the search for more cost-effective and abundant
catalytic materials for halogen production. Unfortunately,
theoretical studies also concluded that other rutile-type
materials are comparatively inactive. For example, stoichio-
metric TiO2 was predicted to deliver a 10-order of magnitude
lower reaction rate than RuO2. The highly endergonic nature
of oxygen dissociation over defect-free TiO2 rutile was put
forward as the reason for its inactivity.[6] Indeed theoretical
studies have shown that the surface of stoichiometric TiO2 is
unable to dissociatively adsorb O2.[6,7] Several authors[7] have
identified that the adsorption of O2 can be ꢁswitched onꢀ
provided that oxygen vacancies exist in the material. How-
ever, under oxygen-rich environments, vacancies can be
healed,[8] thus their presence is metastable and TiO2 is
regarded as useless for O2 activation. In contrast, we have
now found that rutile TiO2 exhibits a remarkable activity and
stability for HBr oxidation to Br2. By means of X-photo-
electron spectroscopy and DFT, we rationalize the crucial role
of defect states to account for this unprecedented catalytic
behavior.
I
n reducible oxides, such as CeO2 and TiO2, unavoidable off-
stoichiometric compositions control the physico-chemical
properties of the materials.[1] These semiconductors can be
doped intrinsically by oxygen depletion or extrinsically by
partially replacing oxygen by higher (or lower) valence-
electron atoms such as F (or N). The extra electrons (or holes)
appear at energy levels that are forbidden (occupied) in the
stoichiometric material. As the relative position between the
active electronic levels of the solid and those of reactants and/
or products determines activity, defect states can lead to the
discovery of new catalytic properties.
The steady-state activity of bulk rutile-type metal oxide
catalysts in HBr and HCl oxidation was studied at different
temperatures (Figure 1a). The lower operating temperature
Titania (TiO2) is one of the most studied oxides both in
experimental and theoretical grounds.[2] Rutile, the most
common polymorph, shows a relatively low surface area and
inert character, and thus its use in industrial catalysis is
limited. The recently commercialized RuO2/TiO2 catalyst for
chlorine recovery by HCl oxidation (Deacon reaction) is the
exception. In this catalyst, rutile TiO2 plays a decisive role as
a carrier, templating the epitaxial growth of the active phase
RuO2.[3] Herein we show that the generation of defect states
in rutile TiO2 turns an inactive semiconductor into an active
and stable catalyst for the production of bromine by HBr
oxidation (2HBr+ 1/2O2!Br2 + H2O, DH0 = ꢀ1.43 eV).
This reaction comprises an essential step for the realization
of a bromine-mediated alkane-functionalization process,
[4]
ꢀ
which is a promising technology to activate C H bonds
leading to large amounts of HBr byproduct that for sustain-
ability purposes needs to be effectively recycled. The
catalyzed oxidation of HBr is tightly related to the Deacon
reaction.[5,6] Experiments have identified the outstanding
performance of RuO2 for the oxidation of HX (X = Cl,
Br).[5] In fact, computational screening based on density
functional theory (DFT) have indicated that the closest point
to the maximal activity for rutile materials is RuO2.[6]
[*] M. Moser, Dr. I. Czekaj, Prof. J. Pꢀrez-Ramꢁrez
Institute for Chemical and Bioengineering, Department of Chemistry
and Applied Biosciences, ETH Zurich
Vladimir-Prelog-Weg 1, 8093 Zurich (Switzerland)
E-mail: jpr@chem.ethz.ch
Figure 1. a) Conversion of HX into X2 (HBr: solid symbols, HCl: open
symbols) versus temperature over rutile-type oxides at O2:HX=2:1.
b) The bromine production measured over TiO2 extrudates, expressed
as the space-time yield in a long-term test. Inset: photographs
showing the clear color change of the solid from the characteristic
white of fresh TiO2 to pale yellow upon use in HBr oxidation (the color
change occurred instantaneously after exposing the material to the
reaction mixture and is a result of the generation of defect states.
Prof. N. Lꢂpez
Institute of Chemical Research of Catalonia (ICIQ)
Av. Paꢃsos Catalans 16, 43007 Tarragona (Spain)
E-mail: nlopez@iciq.es
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201404022.
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 8629

.
Angewandte
Communications
of IrO2 and RuO2 for HBr oxidation compared to HCl
oxidation is in agreement with a previous study on supported
catalysts.[5] As shown in the energy profiles of the reactions
over RuO2 (Supporting Information, Figure S1), the lower
energy requirement for Br2 evolution by recombination of
surface bromine atoms, in comparison with that of Cl2,
explains the shift to lower temperature in HBr oxidation.
More importantly, the inactivity of TiO2 in HCl oxidation (no
conversion at 723 K) greatly contrasts with its activity in HBr
oxidation (over 90% conversion at 673 K). In fact, the
conversion profiles of TiO2 and IrO2 in HBr oxidation almost
coincide. A long-term test over TiO2 extrudates was per-
formed to assess the stability of the system, a crucial aspect for
the prospective application of this material. As shown in
Figure 1b, the catalyst in a technical form showed a stable and
high space-time yield of bromine over 30 h on stream. The
insets of Figure 1b show a significant color change of the solid
from the original white to pale yellow after reaction. The
coloration can be related to the presence of defect states,
which were rapidly introduced by contacting TiO2 with the
HBr oxidation atmosphere. All the changes occurring during
reaction affect exclusively the surface of the material. X-ray
diffraction shows that the rutile structure of the catalysts was
unaltered after exposure to HBr (Figure S2), proving that
bulk bromination does not occur.
a second peak at BE = 459.8 eV ascribed to TiOH. This result
indicates that some surface oxygen was removed, thus, some
of the titanium atoms are in a Ti3+ oxidation state. The
calculated shifts for XPS (Table S4) further support the
appearance of Ti3+ species with lower binding energies either
by formation of vacancies (1 eV) or by bromination (0.7 eV)
both with respect to Ti4+. This is further supported by the O 1s
and Br 3d5/2 spectra (Figure S3, Table S1–S4). Additional
information on the surface state can be retrieved from the
valence band (VB) of the spectra (Figure 2b). At the top of
the VB, the fresh catalyst presents a characteristic feature,
around 3 eV. According to the density of states (DOS)
obtained with the cluster model representing the stoichio-
metric surface (Figure 2c, bottom), this feature corresponds
to the contribution of two-fold coordinated (bridging) Ob
atoms. Upon HBr treatment, this fingerprint smoothens out,
Figure 2b. This result is substantiated by the DOS from the
oxygen-deficient Ovac and the Brb models represented (Fig-
ure 2b blue and red lines and models in Figure 2c middle and
top, respectively). Both Ovac and Brb models (Figure 2b blue,
green and red dashed lines and models in Figure 2c middle
and top, respectively) diminish the intensity of the peak at
3 eV, accounting for the origin of the spectral differences in
Figure 2b. Thus, XPS analyses of used rutile provide valuable
indications to construct a proper computational model to
assess the activity patterns observed in Figure 1a.
The mechanism and energy profile of HBr oxidation over
various oxides are shown in Figure 3a. The reaction proceeds
as follows: O2 is activated (Steps 1,2) and, in parallel, HBr is
dissociatively adsorbed (Step 3) generating hydroxy groups.
These can recombine to form water (Step 4) followed by its
desorption (Step 5). Recombination of bromine atoms lead-
ing to Br2 evolution (Step 6) completes the catalytic cycle.
HBr adsorption and OH recombination are acid–base
reactions, mainly controlled by the
For an in-depth analysis of the TiO2 surface, two samples
treated in HBr oxidation (code HBr+ O2) and in HBr without
gas-phase oxygen (code HBr) were investigated by X-ray
photoelectron spectroscopy (XPS; Tables S1–S3). The XPS
spectra in Figure 2a show the Ti 2p3/2 region of the fresh and
treated samples. All three spectra are very similar and contain
a main peak for Ti4+ (binding energy, BE = 458.5 eV). The
HBr-treated sample shows an approximately 5% contribu-
tion of a peak at BE = 457.5 eVattributed to Ti3+ and 5–9% of
total charge of the ions on the surface
and thus they are weakly dependent on
the particular rutile oxide considered.
Differently, O2 adsorption and Br2
recombination have a marked redox
character as the main species upon
adsorption (O2 or Br) withdraw part
of the electronic density from the oxide.
Accordingly, it is the specific redox
ability of each rutile-type oxide what
determines its activity in the oxidation
of hydrogen halides. For metallic
oxides, like RuO2 and IrO2, the redox
properties are rather constant but in
semiconductors they can be modified to
Figure 2. X-ray photoelectron spectra of rutile TiO2 before and after exposure to different
atmospheres. a) Upon treatment in HBr, Ti3+ was detected on the catalyst surface in the Ti 2p3/2
core level spectra. b) In the valence band (solid lines), the signal of oxygen in bridge positions
(Ob) was weaker for the samples treated in HBr and HBr+O2. The interpretation of the spectra
in (b) by cluster simulations of the TiO2(110) surface (dashed lines) suggest the formation of
oxygen vacancies (Ovac) or of the replacement of Ob by bromine atoms in bridge positions (Brb).
c) Top views of various TiO2(110) clusters. The fresh sample shows a stoichiometric surface, that
is, free of defects (bottom). Dissociative adsorption of HBr with subsequent water evolution
leads to the formation of an oxygen vacancy (Ovac), changing the electronic structure of
neighboring titanium atoms (middle). These vacancies are immediately brominated (top). The
energy for the bromination process is presented in Figure 3b.
a large extent by doping. Understand-
ing the catalytic role of TiO2 in HBr
oxidation requires the consideration of
its complexity. In agreement with pre-
vious results,[6,7] O2 activation on the
stoichiometric surface (Steps 1,2) is
endothermic by more than 4 eV and
thus unfeasible (Figure 3a, green line).
However, the surface still maintains its
8630 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633

Angewandte
Chemie
rutile-type
oxides:
RuO2
(0.46 eV) < TiO2
(0.56 eV) <
IrO2 (0.81 eV), see Figure S5a.
Furthermore, all the catalysts
exhibit a strong dependence on
the oxygen partial pressure with
reaction orders of 0.6 (TiO2), 0.7
(RuO2), and 0.8 (IrO2; Fig-
ure S5b). This observation, in
good agreement with the results
on HCl oxidation, arises from the
large barrier for halide evolution
that limits the number of active
empty positions for surface re-
oxidation.[5,9] The competition
between oxygen and the halides
for the active sites implies that
both parameters are relevant to
retrieve activity. The catalytic
activity of TiO2(Br) in HBr oxi-
Figure 3. a) Reaction profiles of HBr oxidation over the (110) facet of RuO2, IrO2, TiO2, and Br-
containing TiO2(Brb). The numbers (1) to (6) refer to the elementary steps depicted in the inset and *
to an empty site. The catalytic cycle presented corresponds to 2HBr+1/2O2!Br2 +H2O. b) Bromina-
tion path on the TiO2(110) p(2ꢄ1) model: red O, gray Ti, white H, and brown Br atoms. c) Schematic
representation of the band diagrams of metal oxides that behave as metals (IrO2 and RuO2) and the
defect-state containing semiconductor TiO2(Brb). The possible injection mechanisms are shown by
arrows, where the relative position of the reactant levels highlights the scale of stronger (long arrows) to
weaker (small arrows) binding energies.
acid–base properties. Therefore, the heterolytic dissociation
of HBr represented by Step 3 can be driven by the basic
surface lattice oxygen atoms and the acidic empty Ticus (cus =
coordinatively unsaturated site). The reaction leads to ObH
and Brcus and opens a path to the activation of the surface with
the formation of defect states, as illustrated in Figure 3b.
After a second HBr molecule dissociates on the surface the
two hydroxy groups can recombine to form a water molecule
at a bridge position (H2Ob) with two bromine atoms sitting at
Ticus positions (Brcus). Starting from the clean surface this set
of reactions is exothermic by 2 eV. Water can then evolve to
the gas phase, leaving a vacancy behind at the bridge position
that is rapidly occupied by one of the Brcus atoms that
transforms into Brb. The overall doping process is exothermic
by 0.73 eV (Figure 3b) and explains sample coloration (inset
of Figure 1b) and the characteristic fingerprints in XPS.
Notice that similar chlorination was reported for RuO2 in HCl
oxidation.[9,10] However in TiO2 and as Br contains one
electron more than O, when it replaces the O atom, a defect
state with one extra electron is populated. Once doped,
oxygen adsorption can occur (Figure 3a, TiO2(Brb) black
line). Indeed, Step 1 leading to the monohapto molecular
adsorption of oxygen is exothermic, by 0.70 eV (1/2 of this
value in the reaction profile shown). In parallel, O2 dissoci-
ation, HBr adsorption, OH recombination, and H2O desorp-
tion are relatively easy on the doped titanium oxide sample.
The gas-phase evolution of Br2 by recombination of adsorbed
bromine atoms is, as in all the investigated oxides, the most
energy-demanding step. The computed energy need for
Step 6 on TiO2(Brb) (2.36 eV) is higher than that of RuO2
(1.76 eV) but smaller than that of IrO2 (3.02 eV).
dation can be understood by the level alignment between the
material and the reactants. A schematic band diagram is
shown in Figure 3c. For the defect-free TiO2, occupied levels
(marked gray) are too deep for charge transfer to O2
interacting with the surface thus rendering it inactive. Instead,
defect states associated with the bromine-doped TiO2(Brb)
are placed in the band gap close to the conduction band and
thus at much higher energies (marked black). These high-
energy electrons can then be effectively transferred to O2,
turning oxygen adsorption exothermic (Figure 3a). However,
charge transfer also takes place when halides desorb. In Step 6
of the reaction scheme, X2 desorption implies that the
electrons stored in the isolated halide, which is negatively
charged, are transferred back to the material. As the electron
affinity of Cl is about 0.2 eV higher than that of Br,[11] it forms
stronger bonds to TiO2(X) and requires more energy to
evolve Cl2 (DE = 3.17 eV) than Br2. Cl2 evolution is much
more energy demanding than Br2 and O2 adsorption cannot
effectively take place at the operating temperature at which
formation of gas-phase Cl2 is effective. This situation results in
the zero activity of TiO2 in HCl oxidation (Figure 1a). Thus,
the energy alignment of TiO2(X) defect states is reasonable
for HBr but not adequate for HCl oxidation. This explanation
also reveals why SnO2, also with rutile structure, is inactive for
HBr oxidation, as its defect levels are too low in energy to be
efficiently transferred to O2. The O2 calculated adsorption
energy is ꢀ0.40 eV, not enough to occur at the temperatures
required for Br2 evolution.
There is a final confirmation of the different origin of
activity for TiO2 when compared to RuO2 and IrO2. The
variation of the activity for Cl and Br on RuO2 (or IrO2) at
60% conversion is approximately 125 K. This is related to the
halide evolution (Step 6) that controls the activity through the
number of empty sites available for O2 adsorption.[9a] As both
are linked, in principle there is a single electronic parameter,
the metal oxide work function, that provides the energy from
which electrons are injected and back-donated that deter-
mines activity (see Supporting Information, Section S2).
However, this behavior is not retrieved for TiO2, over which
These energies are decreased by the gas-phase entropy
gain (between 1.1–1.6 eV depending on the temperature). For
TiO2(Brb), Br2 evolves leaving the self-doped surface behind
and completing the catalytic cycle. Upon taking into account
competitive adsorption and gas-phase entropy, the DFT
results indicate that the activity of TiO2(Brb) falls in between
that of RuO2 and IrO2. Therefore, in line with the exper-
imentally determined apparent activation energy of the
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
8631

.
Angewandte
Communications
only HBr can be oxidized. For TiO2, the adsorption depend-
ence between molecular oxygen and halide adsorption is
different than for the metallic oxides (RuO2 and IrO2) as
a result of different electrostatic contributions. Thus, it is an
outlier in the corresponding linear scaling relationships.
Indeed, the adsorption energy of O2 is smaller than it would
be for a metallic rutile. This explains why IrO2 is as active as
TiO2 because the larger barrier for halide evolution is
compensated by a very effective O2 trapping; the molecular
Information illustrates the temperature profile, HBr conver-
sion, and the O2:HBr ratio along the six reactors filled with
TiO2. The modular design allows the optimization of the
bromine production by manipulation of the reactor inlet
temperatures and/or the HBr feed streams. The simulated
case demonstrates the production of approximately 1.7 tons
of bromine per day using only 9 kg of rutile TiO2.
monohapto adsorption on IrO2 is exothermic by 1.32 eV (see Experimental Section
Catalyst testing and characterization: RuO2 and IrO2, were prepared
Supporting Information Section S3).
by calcination of anhydrous RuCl3 (Alfa Aesar, 99.99%) and IrCl3
(ABCR, 99.9%), respectively, at 823 K for 5 h (heating ramp =
5 Kminꢀ1). Rutile TiO2 (Aldrich, nanopowder, 99.5%) was calcined
identically prior to its use. TiO2 extrudates (diameter= 2 mm,
length = 4 mm) containing 80 wt.% rutile and 20 wt.% anatase
were prepared following the method described elsewhere.[9b] The
rutile-type metal oxides in particulate form (particle size = 0.4–
0.6 mm, 0.25 g) and the extrudates (1 g) were evaluated in the gas-
phase oxidation of HX (X = Br, Cl) in a continuous-flow fixed-bed
reactor with 8 mm or 17 mm i.d., respectively. Steady-state tests were
The analysis shows when oxidation mechanisms could
potentially occur on these materials and how the process is
linked to photo-electro-chemical uses of TiO2 compounds. In
the photo-electro-chemical applications, the population of
high-chemical-potential electrons is achieved by using pho-
tons or a voltage, these high-energy electrons have similar
properties to those coming from the self-doping found for the
HBr reaction on TiO2. But this mechanism has some
limitations. It does not work if there is a single redox pair
(i.e. only oxygen) in the reaction network. The example is CO
oxidation, which on TiO2 occurs by the activation of oxygen at
vacancies, thus when the reaction is carried out it completely
heals the defects. Once the available surface and near surface
vacancies are extracted and filled,[8b] the reaction stops
revealing its stoichiometric (non-catalytic) character. There-
fore, for a catalytic oxidation to occur on the semiconductor,
the doping levels need to be stable under reaction conditions.
The long-term stability of the catalyst is difficult to maintain
because of the decreasing number of defect states in oxygen-
rich environments.[8] However, in TiO2(Brb), bromine sub-
stitution leaves active defect states in the material but
simultaneously blocks the paths for bulk re-oxidation. There-
fore, the outstanding stability of titanium oxide in HBr
oxidation (Figure 1b) is based on the self-doping character of
the reaction. Extended geometrically defective areas, such as
steps,[8c] could potentially render defect states for the reaction
in a similar manner opening new parallel channels.
In summary, the generation of defect related states in
a semiconductor, such as rutile TiO2, originates an active and
stable catalyst for bromine production through HBr oxidation
and self-doping. Self-doping can be achieved provided that
1) the determining reaction steps belong to redox couples,
2) the alignment between the levels of the active species and
the defect states favor the adsorption/desorption process, and
3) the defect states are preserved during the reaction. The
preservation of defect states guarantees the long-term
stability of the material and the scope for practical applica-
tion. Our results warn against considering theoretical models
that are too simple when addressing the properties of oxides
and the possibilities of doping in semiconducting oxides. To
our knowledge, this is the first use of rutile TiO2 as an active
phase in heterogeneously catalyzed processes, and besides the
remarkable HBr oxidation performance, the low cost and
abundance of the material should be highlighted. The large-
scale implementation of TiO2 can be done using recently
developed fixed-bed technology for HCl oxidation, an
adiabatic reactor cascade with intermediate heat exchange
and HBr feeding (Figure S6a). Figure S6b in the Supporting
conducted at ambient pressure using
a total flow of FT =
166 cm3 STP minꢀ1, feed mixtures with 10 vol.% HX, 5–60 vol.%
O2, balanced in N2, and temperatures of T= 420–740 K. X2 at the
reactor outlet was quantified by iodometric titration using a Mettler
Toledo G20 compact titrator. The conversion of HX was determined
as XHX = 2 mole X2 at the reactor outlet/1 mole of HX at the reactor
inlet ꢂ 100. The space-time yield is defined as STY= grams of Br2/
(hourꢂgram of catalyst). The catalysts were collected after rapid
quenching of the reactor to room temperature in N2 flow and
characterized by X-ray diffraction (PANalytical X’Pert PRO-MPD).
X-ray photoelectron spectroscopy (VG Thermo Escalab 220i-XL,
Al Ka source, UHV 10ꢀ9 mbar) was carried out over TiO2 samples
after reaction for 5 h at 623 K in O2:HBr= 0.25, 1, 2 (HBr+ O2), and 0
(pure HBr). The samples were prepared and transported in the
nitrogen atmosphere. The XPS spectra were charge corrected with
respect to the C 1s signal (i.e. 284.5 eV).
Computational details: Density functional theory as implemented
in the VASP was applied to slabs representing the (110) facets of
RuO2, IrO2, and TiO2.[12] The exchange-correlation functional was
PBE,[13] and for TiO2 PBE + U, Ueff = 4.5 eV. Inner electrons were
replaced by PAW,[14] and monoelectronic valence states were
expanded in plane waves with a cutoff energy of 450 eV, spin
polarized calculations were performed when needed. The supercell
employed was a (2 ꢂ 2) containing five trilayers (OTiO2O motif). In
optimization the two bottom layers were fixed. The k-point sampling
was 5 ꢂ 3 ꢂ 1. Transition states were identified by CI-NEB.[15] Cluster
calculations were performed using StoBe and Ti21O64H44 model[16]
only neighboring centers to Brb or oxygen vacancies were reopti-
mized.
Received: April 5, 2014
Published online: June 2, 2014
Keywords: defects · heterogeneous catalysis ·
.
hydrogen bromide · oxidation · titanium oxide
[1] G. Pacchioni, ChemPhysChem 2003, 4, 1041 – 1047.
[2] a) U. Diebold, Surf. Sci. Rep. 2003, 48, 53 – 229; b) M. Grꢃtzel,
Nature 2001, 414, 338 – 344; c) X. Chen, S. S. Mao, Chem. Rev.
2007, 107, 2891 – 2959.
[3] a) K. Seki, Catal. Surv. Asia 2010, 14, 168 – 175; b) J. Pꢄrez-
Ramꢅrez, C. Mondelli, T. Schmidt, O.-F. Schlꢆter, A. Wolf, L.
Mleczko, T. Dreier, Energy Environ. Sci. 2011, 4, 4786 – 4799.
8632
www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633

Angewandte
Chemie
[4] a) E. McFarland, Science 2012, 338, 340 – 342; b) K. Ding, H.
Metiu, G. D. Stucky, ChemCatChem 2013, 5, 1906 – 1910.
[5] M. Moser, L. Rodrꢅguez-Garcꢅa, A. P. Amrute, J. Pꢄrez-Ram-
ꢅrez, ChemCatChem 2013, 5, 3520 – 3523.
[10] a) D. Crihan et al., Angew. Chem. 2008, 120, 2161 – 2164; Angew.
Chem. Int. Ed. 2008, 47, 2131 – 2134.
[11] N. N. Greenwood, A. Earnshaw, Chemistry of the Elements, 2nd
ed., Elsevier, Dordrecht, 2008, p. 800.
[6] A. Toftelund, I. C. Man, F. Abild-Pedersen, T. Bligaard, J.
Rossmeisl, F. Studt, ChemCatChem 2012, 4, 1856 – 1861.
[7] H. Metiu, S. Chrꢄtien, Z. Hu, B. Li, X. Sun, J. Phys. Chem. C
2012, 116, 10439 – 10450, and references therein.
[8] a) M. Setvin, U. Aschauer, P. Sheiber, Y.-F. Li, W. Hou, M.
Schmid, A. Selloni, U. Diebold, Science 2013, 341, 988 – 991;
b) N. Lꢇpez, J. D. Prades, F. Hernandez-Ramirez, J. R. Morante,
J. Pan, S. Mathur, Phys. Chem. Chem. Phys. 2010, 12, 2401 – 2406;
c) R. Bechstein, H. H. Kristoffersen, L. B. Vilhelmsen, F.
Rieboldt, J. Stausholm-Møller, S. Wendt, B. Hammer, F.
Besenbacher, Phys. Rev. Lett. 2012, 108, 236103.
[12] a) G. Kresse, J. Furthmꢆller, J. Comput. Mater. Sci. 1996, 6, 15 –
50; b) G. Kresse, J. Furthmꢆller, Phys. Rev. B 1996, 54, 11169 –
11186.
[13] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77,
3865 – 3868.
[14]
P. E. Blçchl, Phys. Rev. B 1994, 50, 17953 – 17979.
[15] G. Henkelman, B. P. Uberuaga, H. J. Jꢇnsson, J. Chem. Phys.
2000, 113, 9901 – 9904.
[16] a) I. Czekaj, J. Wambach, O. Krçcher, Int. J. Mol. Sci. 2009, 10,
4310 – 4329; b) K. Hermann et al., StoBe-deMon, deMon Soft-
ware, Stockholm, Berlin, 2005, see Supporting Information.
[9] a) D. Teschner et al., J. Catal. 2012, 285, 273 – 284; b) M. Moser
et al., ACS Catal. 2013, 3, 2813 – 2822, see Supporting Informa-
tion.
Angew. Chem. Int. Ed. 2014, 53, 8628 –8633
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
8633