An integrated approach to Deacon chemistry on RuO2-based catalysts

Article abstract of DOI:10.1016/j.jcat.2011.09.039

Rationally designed RuO₂-based Deacon catalysts can contribute to massive energy saving compared to the current electrolysis process in chemically recycling HCl to produce molecular chlorine. Here, we report on our integrated approach between state-of-the-art experiments and calculations. The aim is to understand industrial Deacon catalyst in its realistic surface state and to derive mechanistic insights into this sustainable reaction. We show that the practically relevant RuO₂/SnO₂ consists of two major RuO₂ morphologies, namely 2-4 nm-sized particles and 1-3-ML-thick epitaxial RuO₂ films attached to the SnO₂ support particles. A large fraction of the small nanoparticles expose {1 1 0} and {1 0 1} facets, whereas the film grows with the same orientations, due to the preferential surface orientation of the rutile-type support. Steady-state Deacon kinetics indicate a medium-to-strong positive effect of the partial pressures of reactants and deep inhibition by both water and chlorine products. Temporal Analysis of Products and in situ Prompt Gamma Activation Analysis strongly suggest a Langmuir-Hinshelwood mechanism and that adsorbed Cl poisons the surface. Under relevant operation conditions, the reactivity is proportional to the coverage of a specific atomic oxygen species. On the extensively chlorinated surface that can be described as surface oxy-chloride, oxygen activation is the rate-determining step. DFT-based micro-kinetic modeling reproduced all experimental observations and additionally suggested that the reaction is structure sensitive. Out of the investigated models, the 2 ML RuO₂ film-covered SnO₂ gives rise to significantly higher reactivity than the (1 0 1) surface, whereas the 1 ML film seems to be inactive.

Full text of DOI:10.1016/j.jcat.2011.09.039

Journal of Catalysis 285 (2012) 273–284
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
An integrated approach to Deacon chemistry on RuO -based catalysts
2
a,d,
⇑
a
a,1
a
b
b
Detre Teschnerc , Ramzi Farra , Lide Yao , Robert Schlögl , Hary Soerijanto , Reinhard Schomäcker ,
d
e
e
e
Timm Schmidt , László Szentmiklósi , Amol P. Amrute , Cecilia Mondelli , Javier Pérez-Ramírez ,
Gerard Novell-Leruth , Núria López
f
f,
⇑
a
Fritz-Haber Institute of the Max Planck Society, Berlin D-14195, Germany
b
c
d
e
f
Department of Chemistry, Technical University of Berlin, Berlin D-10623, Germany
Bayer MaterialScience AG, PUR-PTI-PRI New Processes Isocyanates, Dormagen D-41538, Germany
Institute of Isotopes, Hungarian Academy of Sciences, Budapest H-1525, Hungary
Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, HCI E 125, Wolfgang-Pauli-Strasse 10, CH-8093 Zurich, Switzerland
Institute of Chemical Research of Catalonia (ICIQ), Av. Paisos Catalans 16, Tarragona 43007, Spain
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 9 May 2011
Revised 3 August 2011
Accepted 29 September 2011
Available online 8 November 2011
Rationally designed RuO -based Deacon catalysts can contribute to massive energy saving compared to
the current electrolysis process in chemically recycling HCl to produce molecular chlorine. Here, we
report on our integrated approach between state-of-the-art experiments and calculations. The aim is
to understand industrial Deacon catalyst in its realistic surface state and to derive mechanistic insights
2
into this sustainable reaction. We show that the practically relevant RuO
RuO morphologies, namely 2–4 nm-sized particles and 1–3-ML-thick epitaxial RuO
the SnO support particles. A large fraction of the small nanoparticles expose {110} and {101} facets,
2
/SnO
2
consists of two major
2
2
films attached to
Keywords:
HCl oxidation
2
whereas the film grows with the same orientations, due to the preferential surface orientation of the
rutile-type support. Steady-state Deacon kinetics indicate a medium-to-strong positive effect of the par-
tial pressures of reactants and deep inhibition by both water and chlorine products. Temporal Analysis of
Products and in situ Prompt Gamma Activation Analysis strongly suggest a Langmuir–Hinshelwood
mechanism and that adsorbed Cl poisons the surface. Under relevant operation conditions, the reactivity
is proportional to the coverage of a specific atomic oxygen species. On the extensively chlorinated surface
that can be described as surface oxy-chloride, oxygen activation is the rate-determining step. DFT-based
micro-kinetic modeling reproduced all experimental observations and additionally suggested that the
2 2
RuO /SnO
HRTEM
PGAA
TAP
DFT
Kinetics
Deacon reaction
2 2
reaction is structure sensitive. Out of the investigated models, the 2 ML RuO film-covered SnO gives rise
to significantly higher reactivity than the (101) surface, whereas the 1 ML film seems to be inactive.
Ó 2011 Elsevier Inc. All rights reserved.
1
. Introduction
produce hydrochloric acid, is typically used in other processes,
e.g., as acid catalyst, for the neutralization of alkaline streams, as
Chlorine is used as reactive intermediate in many processes to
create thousands of often indispensable products of the chemical
industry. The annual production of Cl is ca. 50 Mton and has an
expected demand growth of 4.4% per year in the period 2010–
015 [1]. Upon use, large amounts of Cl are reduced to HCl or
chloride salts. The most prominent example is the phosgene-
mediated isocyanate production (e.g., 4 mol HCl per mol of toluene
diisocyanate). By-product HCl, pure or absorbed in water to
chlorine source in PVC production, or is recycled by electrolysis
to chlorine; this latter is, however, a highly energy-demanding pro-
2
2
cess. The catalytic oxidation of by-product HCl to Cl , the Deacon
process (2HCl + 1/2O ? Cl + H O), is an energetically efficient
2
2
2
2
2
yet environmentally friendly solution to create a complete recycle
process. However, the long-term implementation of all processes
developed since the original Deacon concept has failed namely
due to the short catalyst lifetime [2–4]. Only recently, Sumitomo
and Bayer have succeeded in turning HCl oxidation into industrial
2
reality with the use of RuO -based catalysts [5–12]. As Sumitomo
⇑
Corresponding authors. Address: Fritz-Haber Institute of the Max Planck
Society, Faradayweg 4-6, Berlin D-14195, Germany. Fax: +49 30 84134676 (D.
Teschner), fax: +34 97792231 (N. López).
has shown, the choice of the support is particularly crucial [8].
The ‘‘trick’’ to stabilize the catalyst is, in fact, to apply a support
stabilizing epitaxially the active phase. In the 1960s, Shell had
already introduced a silica-supported Ru-based system in the HCl
oxidation process, but a much less active and stable catalyst was
E-mail addresses: teschner@fhi-berlin.mpg.de (D. Teschner), nlopez@iciq.es (N.
López).
1
Present address: Department of Applied Physics, Aalto University School of
obtained [13]. Despite
a number of studies addressing the
Science and Technology, Aalto, Finland.
0
021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.09.039

2
74
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
mechanism of HCl oxidation on RuO
2
single crystals [14–17], little
Knudsen diffusion coefficient and the Wheeler–Weisz modulus, it
was verified that the reaction was not limited by intraparticle
diffusion.
is known about structure and function of the polycrystalline sup-
ported catalysts [18]. Furthermore, the approaches have generally
been based on limited experimental or computational methods.
We present here an integrated approach to unravel HCl oxida-
2.3. Catalyst characterization
2 2
tion on the industrially relevant RuO /SnO catalyst. Synthesis
and basic characterization have been combined with steady-state
tests in a dedicated plug-flow reactor. In situ studies using Prompt
Gamma Activation Analysis (PGAA), Temporal Analysis of Products
2 2 2
Standard characterization of RuO /SnO was carried out by N
adsorption at 77 K, inductively coupled plasma mass spectrometry
(ICP-MS), X-ray diffraction (XRD), and temperature-programmed
(
TAP), and Density Functional Theory (DFT) simulations coupled to
2
reduction with hydrogen (H -TPR). Micro-structural and chemical
micro-kinetic (MK) analysis were used to obtain complementary
insights into the structure and function of this current Bayer sys-
tem. This improved and deeper knowledge may enable designing
new and cheaper materials capable to act as catalysts in this
demanding reaction.
characterizations (elemental mapping) of the binder-free catalyst
were performed on a Titan 80–300 kV aberration-corrected Trans-
mission Electron Microscope (TEM). For experimental details to
these methods, see the Supplementary information.
2.4. In situ Prompt Gamma Activation Analysis
2
. Experimental section
In situ Prompt Gamma Activation Analysis (PGAA) is a tech-
nique recently implemented for studying catalysts in action [20].
It is based on the radiative neutron capture of nuclei, that is, with
using cold neutrons excitation element specific gamma rays are
emitted, enabling the quantification of elements in the investi-
gated volume, in our case inside a Deacon micro-reactor. The aim
of our experiments was to quantify in situ the Cl uptake of Deacon
catalysts under various steady-state reaction conditions and corre-
late this information with the catalytic performance. PGAA at
atmospheric pressure condition was carried out at the cold neutron
beam of the Budapest Neutron Centre [21]. A Compton-suppressed
high-purity germanium crystal was used to detect the prompt
gamma photons. Molar ratios (Cl/Ru) were determined from the
characteristic peak areas corrected by the detector efficiency and
the nuclear data of the observed elements [22]. For the experi-
ments, the same quartz tube reactor as for the steady-state kinetic
study was placed into the neutron beam and the reactor tube was
surrounded by a specially designed oven having openings for the
incoming and outgoing neutrons and for the emitted gamma rays.
These openings are covered by thin aluminum foils for decreasing
2
.1. Catalyst preparation
RuO
2
/SnO
2
was prepared by dry impregnation. Commercially
-cassiterite (Keeling & Walker, Superlite C) was cal-
available SnO
2
cined for 2 h at 1273 K prior to use in order to ensure the formation
of the rutile-type structure also at a surface level. Impregnation
was performed with a RuCl aqueous solution (nominal 2 wt%
3
Ru). The obtained material was dried at 333 K for 5 h and calcined
at 523 K under air for 16 h. Low-temperature calcination was used
to prevent extensive agglomeration of the Ru-phase, thus achiev-
ing a higher dispersion. A parallel preparation was carried out
2
ꢀ1
2 3
using additional alumina binder (c-Al O , 200 m g ; Saint-Gob-
ain NorPro) prior to impregnation to further stabilize the catalytic
performance and hence reach industrially relevant long-term (over
a year) stability. The longevity of this catalyst was very recently
demonstrated in mini-plant tests using 2 mm spherical bodies
[
19]. The two catalysts display similar performance (e.g., compara-
ble O dependence in in situ PGAA and kinetic experiment), except
for the long-term stability. Hence, to enable structure–function
correlations, characterization was performed on the binder-free
2
heat losses. 0.9 g of RuO
loaded into the reactor. An experimental series with RuO
Al was also carried out with similar results as will be detailed
here later. These latter data are included in Supplementary infor-
2
/SnO
2
of sieve fraction 0.1–0.2 mm was
2
2
/SnO –
(
RuO
and steady-state), for which stability is of utmost importance, will
be discussed using the data for the binder-stabilized (RuO /SnO
Al ) catalyst. Bulk RuO (Aldrich, 99.9%) was employed as
2 2
/SnO ) sample. However, kinetic experiments (transient
2 3
O
2
2
–
3
ꢀ1
mation. Reaction feed, at constant 166.6 cm min
was supplied by mass flow controllers, and the feed composition
was varied between O :HCl:N = 0.25:1:3.75 and 4:1:0. The reac-
tion was monitored by iodometric titration. The reaction tempera-
ture was set to 573 K for the p(O )-dependent experiment, which
total flow,
2
O
3
2
reference material in transient kinetic studies. Prior to use, the
as-received oxide powder was calcined in static air at 773 K
2
2
ꢀ1
(
10 K min ) for 5 h.
2
was performed after a HCl treatment at the same temperature to
saturate the catalyst surface with chlorine. In a separate experi-
ment at 523 K, the Cl uptake was acquired with time on stream.
2.2. Catalytic tests
For steady-state kinetic measurements, the binder-containing
catalyst was first powdered by ball milling and diluted with a five-
fold amount of SnO support material. Water was added to the
2.5. Temporal Analysis of Products
2
powder mixture and then it was dried at 353 K. A sieve fraction
of 0.2–0.45 mm-sized particles was prepared, and 1 g of it was di-
luted with 4 g glass beads and mixed well in order to minimize the
occurrence of hot spots in the reactor. An atmospheric pressure
reactor operated at 573 K was fed with a range of reactant gas mix-
Transient mechanistic studies of HCl oxidation over RuO
2 2
/SnO ,
RuO /SnO –Al , and RuO were carried out in the TAP-2 reactor
2
2
2
O
3
2
[23,24]. This technique has been successfully applied to derive
mechanistic fingerprints of the Deacon reaction over metal oxides
[18,25,26]. RuO
form, while RuO
ditions (HCl:O :He = 2:4:4, 1 bar, 623 K, W/F = 4.5 g h mol ) prior
2
/SnO
2
2 3 2
–Al O and RuO were investigated in fresh
tures by using mass flow controllers, and Cl
2
formation was de-
2
/SnO
2
was equilibrated for 50 h under Deacon con-
0
ꢀ1
HCl
tected by iodometric titration. The catalytic tube reactor was
made of quartz with 8 mm inner diameter. Various reaction feed
2
to the TAP study. The equilibration protocol has been detailed else-
where [19]. The samples (10 mg, particle size = 0.2–0.3 mm) were
loaded in the isothermal (central) zone of a quartz micro-reactor
(4.6 mm i.d., 71 mm long), between two layers of quartz particles
3
ꢀ1
compositions with a total flow of 80 cm min were used to eval-
uate the formal kinetic dependence of the reaction rate to the reac-
tants and products. The partial pressures of the reactants were
varied in the range of 0.0625–0.5 bar, and water was set between
of the same particle size. The samples were pretreated in O
2
3
ꢀ1
0
–0.5 bar, whereas Cl
2
was co-fed up to 0.2 bar. Conversion level
(20 cm STP min ) at 623 K and ambient pressure for 1 h. Thereaf-
ꢀ10
was below 10% to allow kinetic evaluation. By calculating the
ter, they were evacuated to 10
bar and TAP experiments were

D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
275
carried out at 623 K using a pulse size of ca. 1016 molecules. This
large pulse size, exceeding the Knudsen diffusion regime, was re-
quired in order to properly detect reaction products, mainly Cl ,
2
modated (Fig. S.I. 1). The total thickness of these slabs corresponds
to five layers where the upper three layers were relaxed in all
directions.
by mass spectrometry. Since current TAP theories for modeling pur-
poses are limited to operation in the Knudsen regime (pulses), quan-
titative micro-kinetic analysis was not attempted from the present
With the surface energies of the low-index facets, the structure
of equilibrium nanoparticles can be calculated through the Wulff
construction [31]. Calculations on pure RuO
agreement with XRD estimates [16]. The Wulff structure provides
about a 43% of RuO (110), 42% of RuO (101), and approximately
15% of RuO (100). Other facets as RuO (001) show a high surface
energy, and thus, in this study, we consider only those mentioned
above. In order to analyze the complete Deacon process on RuO
2
have shown a good
TAP results. Pulses of a HCl–O
lowed by pump–probe experiments of O
from two separate high-speed valves. In the latter measurements,
the O pulse is separated from the HCl pulse by a time delay ( t)
2
2
mixture (HCl:O :Kr = 2:1:1) were fol-
2
:Ar = 2:1 and HCl:Kr = 5:1
2
2
2
2
2
D
of 1–12 s; 10 s after pulsing the probe molecule, a new cycle starts
by pulsing the pump molecule and so on. In the TAP experiments,
2
nanoparticles, we have considered the individual reaction paths
on the most representative structures ((110) and (101)) indicated
above. Then, the rate was obtained through micro-kinetic simula-
tions, MK. A detailed description of the procedure is presented in
Supplementary information and can be summarized as follows:
the rate is calculated as the sum of the individual rates of the dif-
ferent facets weighted by their contribution to the nanoparticle ob-
tained from the Wulff construction based on the DFT values. In
order to do this, the rate coefficients for the elementary steps were
evaluated using the thermodynamic, kinetic, and partition func-
tions obtained from the DFT calculations and according to Transi-
tion State Theory [32]. The MK simulations were performed
through a batch-model reactor with the setup previously em-
ployed by us for ammonia oxidation on platinum [33]. The set of
differential–algebraic equations in the MK model has been re-
2
Kr (Linde, purity 5.0), Ar (Linde, purity 5.0), O (Air Products, purity
5
.2), and HCl (Praxair, purity 2.5) were used. A quadrupole mass
spectrometer (RGA 300, Stanford Research Systems) was used for
monitoring the transient responses at the reactor outlet of the fol-
lowing atomic mass units (AMUs): 84 (Kr), 70 (Cl
2
), 40 (Ar), 36
(
HCl), 32 (O ), and 18 (H O). The transient responses displayed in
2
2
this article correspond to an average of 10 pulses per AMU in order
to improve the signal-to-noise ratio. Prior to that, it was checked
that the responses were stable, that is, with invariable intensity
and shape during an interval of at least 20 pulses.
2.6. First principles simulations and micro-kinetic modeling
Density Functional Theory (DFT) applied to slabs representing
the lowest-index facets of RuO was employed to determine the
2
™
solved with Maple [34]. In the simulations, the initial relative
pressures and temperatures correspond to those employed exper-
imentally and also the number of particles initially contacting the
surface has been balanced with respect to those in the experi-
ments. From the MK models, several parameters have been ob-
tained, like the apparent activation energy, the inhibition effect
by reaction products, and the correlation between the species on
the surface and the reaction rate. Further details to our theoretical
approach, together with equilibrium constants for each of the reac-
tions, calculated core-level shifts, and vibrational frequencies of
some adsorbed species can be found in Supplementary
information.
thermodynamic and kinetic parameters for HCl oxidation. The
VASP code was employed in the calculations [27,28]. The ex-
change–correlation functional was RPBE [29], and the inner elec-
trons were replaced by PAW pseudopotentials [30] while
monoelectronic valence states were expanded in plane waves with
a cutoff energy of 400 eV. For RuO
and (100) surfaces (Fig. 1), each of the slabs contains three trilay-
ers of Ru stoichiometry. On these surfaces ((110), (101) and
100)), two kinds of surface oxygen atoms can be identified. First
2
, we employed the (110), (101),
2 4
O
(
of all, threefold-coordinated O3c positions are present. This is the
same coordination as for the oxygen atoms in the bulk. Some O
atoms only show a coordination of two and are usually denoted
as bridge oxygen atoms, O
b
. Ru cations are also present at the sur-
3. Results and discussion
face with a coordination number of five. This is lower than that in
the bulk (6) and, thus, these Ru atoms are denoted as under-coor-
3.1. Primary characterization
b
dinated, Rucus sites. These O , O3c, and Rucus motives are common
to all low-index facets and are thus expected to be present even
for less well-defined surfaces. To describe the different positions
at the surfaces, we employ the subindexes ‘‘b,’’ if a given atom or
The Ru content in RuO
2
/SnO
2
, determined by ICP-MS, is 2.4 wt%,
adsorption is
2
BET = 6 m g . The specific surface of the SnO support is low
and the total surface area determined from N
2
2
ꢀ1
S
2
ꢀ1
fragment is adsorbed at a formerly O
sorbed at a formerly open Rucus site. To represent the RuO
catalyst, on top of SnO (110), either one (model RuO /SnO (1 ML))
or two (model RuO /SnO (2 ML)) epilayers of RuO were accom-
b
position, and ‘‘cus,’’ if ad-
(SBET = 9 m g ) due to its high temperature pretreatment prior
2
/SnO
2
to impregnation with the ruthenium salt. The XRD analysis of
2
2
2
RuO
2
/SnO
2
provides evidence of the SnO
2
crystalline phase only
2
2
2
(Fig. S.I. 2). The presence of RuO
2
is not observed neither as phase
Fig. 1. Schematic representation of the RuO
2
(110), (101), and (100) surfaces. Red large spheres represent O atoms, while gray ones stand for Ru. The labels for the most
relevant surface centers are indicated. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2
76
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
alone, owing to the relatively low ruthenium amount and the high
dispersion (shown later), nor as solid solution with the support, as
no peak shift of the cassiterite diffraction lines is detected. This
suggests the deposition of the active phase as a thin film and/or
long-term stable material, we have selected the binder-containing
RuO /SnO –Al catalyst for this purpose. We chose 573 K and a
feed of 20 cm min
as a reference point (r
investigated reaction condition (r) by r
2
2
2 3
O
3
ꢀ1
3
ꢀ1
3
ꢀ1
O
2
, 20 cm min HCl, and 40 cm min
), and normalized the reaction rate at every
, as shown in Fig. 3a. Within
N
2
0
nanoparticles covering the SnO
Fig. S.I. 3) indicate that a higher reduction temperature is needed
to reduce the supported ruthenium phase than RuO alone, sug-
gesting a strong interaction between RuO and SnO
2
2
particles. The H -TPR profiles
0
(
the investigated conditions, the apparent HCl reaction order is
determined from differential measurements to be 0.2, whereas that
2
2
2
.
of O
though for HCl only slight, influence on the reaction rate. On the
other hand, both products (Cl and H O) strongly inhibit the reac-
2
is 0.4 (Fig. 3a). This means that both reactants show a positive,
3.2. Electron microscopy
2
2
tion, as suggested by the strongly bended curves of chlorine pro-
duced determined as function of residence time (Fig. 3d). In
2
The morphology of SnO support particles has been revealed by
TEM observations, as shown in Fig. 2a. Most particles look smooth,
displaying nicely faceted feature with a mean diameter around
Fig. 3b, measurements with increasing co-feed of chlorine or H
are depicted, with the reaction order being near ꢀ1.0 for both, Cl
and H O. This observation implies that the rates shown in Fig. 3a
2
O
2
5
0 nm. Based on electron diffraction (ED) analysis (also see inset
2
of Fig. 2a), all diffraction rings can be indexed by rutile structure,
which is in agreement with the XRD data. A high-resolution TEM
image in Fig. 2b clearly shows two particles with {110} and
were not determined under accurate differential conditions, since
the measured conversion varies with feed composition, and thus,
this effect leads to an underestimation of the influence of the reac-
tant pressures on the reaction rate, expressed in too low reaction or-
{
{
011} facets. (We would like to remind that the families of
011} and {101} are equivalent due to the rutile symmetry.) These
ders. To circumvent this, we repeated the HCl- and O
2
-dependent
kinds of facets, in fact, preferentially form in rutile phase [35,36]. In
contrast to the SnO support particles, the RuO /SnO catalyst par-
ticles in Fig. 2c appear rougher, suggesting the formation of ruthe-
nium oxide nanoparticles on the support surface. Via selected area
series with added 0.06 bar H O and 0.06 bar Cl at the reactor inlet
2
2
2
2
2
(Fig. 3c). In this plot, the data are again normalized to the rate at
an equimolar feed of the reactants. Results show that the HCl reac-
tion order increases to 0.5, while that of oxygen is close to 1. Based
ED analysis, the RuO
2
phase was determined with the same rutile
on the stoichiometric equation, 2 HCl + 0.5 O
2 2
M H O + Cl2, the rate
structure as the support. From EDX elemental mappings (see
Fig. S.I. 4), O and Sn elements have nearly homogeneous distribu-
tions and Ru can also be detected on the support everywhere. Ru
of the reaction can be expressed with the formal kinetic rate law:
k
f
1
0:5
1
1
K
f
p
p
ꢀ
pKpꢀ0:5
p p
Cl2 H2O
O2 HCl
1:5
HCl
p
O2
r ¼
ð1Þ
is observed especially around SnO
or-mixed mapping, confirming the coverage with RuO
high-magnification TEM image in the inset of Fig. 2c further evi-
dences very densely packed RuO particles with typical size of 2–
nm over the SnO carrier. Moreover, it can be frequently seen
that some RuO nanoparticles cluster into small aggregates. Be-
sides nanoparticles, another typical morphology of thin RuO film
2
particles, as shown by the col-
1
þ KCl2pCl2 þ KH2OpH2O
2
layers. The
where k
f
is the forward rate constant, K the equilibrium constant,
and KCl2 and KH2O are the inhibition constants of chlorine and water,
respectively. The denominator represents the usual inhibition term,
whereas the numerator includes the derived rate dependences. The
complex reverse rate coefficient ensures that the rate goes to zero at
equilibrium pressure conditions. The apparent activation energy
estimated from initial rates without co-feed of product in the tem-
2
4
2
2
2
with a few atomic layers on the support is often observed, as
shown in Fig. 2d. By comparing fast Fourier transformation (FFT)
patterns from two selected areas, a streaking effect is remarkably
visible along the [110] direction in the FFT pattern from the film
ꢀ1
perature range of 400–671 K is 69 kJ mol . Stronger temperature
dependence is expected when products are co-feed.
area, being absent with SnO
from stacking faults in the thin film. Considering the same rutile
structure and small lattice mismatch between RuO and SnO , a
thin film with one-two monolayer-like RuO may be epitaxially
formed under suitable chemical conditions, but contains a few
stacking faults to some extent. Additionally, RuO particles can also
be epitaxially grown on SnO , which is, in fact, often observed in
the HRTEM characterization. For instance, Fig. 2e presents a typical
rutile on rutile-type micrograph, where RuO particles and SnO
support can be easily identified by FFT analysis, with close rela-
tionships: (101)RuO //(101)SnO and (110)RuO //(110)SnO . It
is expected that there is strong interaction between RuO and
the SnO support in these cases due to epitaxy. For the purpose
of demonstrating surface orientation of RuO particles, a local en-
larged image close to the support’s edge is shown in Fig. 2f. Clearly,
RuO particles expose {110} and {011} facets, in consonance with
the Wulff construction in rutile. As a matter of fact, most RuO par-
2
only. Here, the streaking effect results
3
.4. Theoretical modeling
2
2
2 2 2
The Deacon reaction, 2HCl + 1/2O ? Cl + H O, is calculated to
2
be exothermic with an enthalpy of ꢀ54.3 and the Gibbs free energy
ꢀ1
is ꢀ17.3 kJ mol at 573.15 K. The calculated values are in very
2
good agreement with those in the databases: enthalpy
2
ꢀ
ꢀ
1
1
ꢀ
57.2 kJ mol
and
corresponding
Gibbs
free
energy
ꢀ
20.3 kJ mol [37]. The mechanism of the Deacon reaction can
2
2
be described with the following list of elementary steps taking
place on any of the surfaces mentioned in Section 2 (i.e.,
2
2
2
2
2 2 2
RuO (110), RuO (101), and the overlayers on SnO ). The mecha-
nism, first described by us [16], has been further confirmed by
other DFT studies [14,17]. Although more complex re-oxidation
paths have been described for the RuO (110) surface [38], the
2
equations (6–7) in the scheme correctly reproduce the main fea-
tures of surface re-oxidation.
2
2
2
2
2
ticles observed expose such facets. Despite the close relation in
structure, in some cases, partly due to lack of epitaxy, many other
surfaces like {120} planes are exposed as well. In addition, there
ꢁ
ꢁ
ꢁ
ꢁ
HCl þ O þ $ OH þ Cl
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
ꢁ
ꢁ
ꢁ
ꢁ
OH þ OH $ H
2
O þ O
are various types of defects found with the RuO
2
nanoparticles,
such as steps with one lattice unit, dislocations, and stacking faults.
ꢁ
Oþꢁ
H
2
O $ H
2
3.3. Atmospheric catalytic tests
ꢁ
ꢁ
þ 2ꢁ
Cl þ Cl $ Cl
2
Since we aimed at identifying the dependence of the reaction rate
on the reactants and products with high accuracy requiring thus a
ꢁ
ꢁꢁ
O
2
þ 2 $ O
2

D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
277
Fig. 2. (a) Overview of SnO
catalyst by TEM, where inset image shows densely packed RuO
2
-support particles by TEM and corresponding electron diffraction pattern in inset; (b) SnO
2
particles with {110} facets; (c) overview of RuO
film structure over the
support (rutile on rutile); (f) HRTEM image from a local edge region in (e), where RuO particles expose
110} and {011} family of facets. Note that due to the rutile symmetry (101) and (011) are equivalent facets and are thus marked as {011}.
2 2
/SnO
2 2 2
particles (typical size: 2–4 nm) over the SnO support; (d) 1–2 monolayer-like RuO
support surface; (e) RuO particles are epitaxially grown on SnO
2
2
2
{
ꢁ
ꢁ
$ 2Oꢁ
O
ð7Þ
charges exchanged during the elementary steps are not localized
in the substrate due to the metallic character of ruthenium oxide.
Therefore, no charges have been associated with the intermediates
in the description of the reactions (2)–(7).
2
This list of reactions (2)–(7) indicates only the elementary steps
taking place, but does not reflect the nature of the different oxygen
atoms involved (either from the lattice, O
b
, or from incoming O
2
In the mechanism, HCl adsorbs close to an O atom (reaction (2))
and the basic center abstracts the H atom from HCl. The source of
ꢁ
molecules, Ocus) or the nature of the empty positions, . Partly
due to various surfaces investigated, care must be taken because
the energy differences of different O species can be rather large.
Moreover, even if the Deacon reaction is a redox reaction, the
oxygen for HCl stripping can come either from the lattice (O ) or
b
from newly adsorbed oxygen on the surfaces (indicated as Ocus).
Thus, hydroxyl groups are formed on the surface, either O H or
b

2
78
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
Fig. 3. The influence of (a) HCl and O
2
and (b) H
2
O and Cl
2
on the reaction rate at 573 K and 1 bar in the Deacon reaction over RuO
2
/SnO
2
–Al
2
O
3
. The rate is normalized to the
and 0.06 bar H O in the feed.
case 20 cm3 min
Rate is normalized to the case 20 cm min
73 K in a residence time variation experiments with a feed of O
-1
O
, 20 cm3 min HCl, and 40 cm min
ꢀ1
3
ꢀ1
N
; p
is expressed in bar. In (c), the influence of HCl and O
is shown with 0.06 bar Cl
2
2
i
2
2
2
3
-1
, 20 cm3 min HCl, 30 cm min
ꢀ1
3
-1
, 5 cm3 min
ꢀ1
O, and 5 cm3 min Cl
ꢀ1
O
2
N
2
H
2
2
. In (d), the temperature is increased from 573 to
. Shown is the partial pressure of Cl produced.
6
2
and HCl in a molar ratio of 1:1 without dilution with N
2
2
O
cusH. These moieties can recombine on the surface to produce H
2
O
energy-demanding step corresponds to atomic chlorine recombi-
nation. This feature is common to the other low-index facets. In
(
reaction (3)). If water is produced at the cus centers, it can be des-
orbed from the surface (reaction (4)). In any case, hydroxyl recombi-
nation leaves an oxygen atom behind. This O atom can either belong
2
addition, the RuO (101) energy profile shows a very similar pat-
tern to that of the (110) facet. The largest difference is that the
binding energies for all the species in the (110) are somehow lar-
ger than for the (101) surface. This affects both the surface cover-
age of each species and the reaction barriers that are smaller in the
(101) case. Thus, a proper comparison of these surfaces requires
the use of micro-kinetic analysis to evaluate the differences in
activity, see the micro-kinetic modeling subsection. The (100) sur-
face binds even less strongly intermediates, and therefore, chlorine
evolution shows a much smaller energy barrier, but this is some-
to the lattice O
the surface can react to form Cl
b
or be sitting in a cus position, Ocus. The Cl atoms on
and desorb (reaction (5)); these tak-
2
ing place in one single step. The final couple of reactions correspond
to surface re-oxidation and active oxygen regeneration on the sur-
face. The dissociation of O takes place from a molecularly adsorbed
2
precursor (reaction (6)) that by interacting with two empty cus posi-
tions, finally leads to two O atoms on the surface (reaction (7)).
As indicated previously for the (110) surface, proton scavenging
and hydroxyl recombination are low-energy-demanding steps. In
how compensated by the weak O
high barrier found for O dissociation.
As already mentioned, it is possible to construct epitaxial RuO
layers on top of SnO . In this particular case, we have chosen only
the (110) surfaces, as they are the most abundant. The effect of
supporting RuO layers on SnO is reflected on the small changes
2
binding to the surface and the
general, we identify that bridging oxygen atoms (O
b
) are effective
2
proton scavengers. However, water desorption from bridge posi-
tion is more hindered than from cus positions (0.8–1.0 eV from
Rucus < 2 eV from isolated H O ); therefore, Ocus is the preferred
2 b
proton elimination species. For the (110) surface, the highest en-
ergy requirement in the mechanism corresponds to Cl recombina-
tion; however, if very high Cl coverages are present, this might not
2
2
2
2
induced in the electronic structure of the lattice oxygen atoms as
described by the expected XPS core-level shifts (see Table S.I. 2).
In the following, we concentrate on the perturbations induced on
be the rate-determining step, as O
neighboring Rucus sites.
2
adsorption requires two empty
the reaction energy profiles of supported RuO
Fig. 4b. From the reaction profile, it is clearly shown that the single
RuO monolayer on SnO (110) binds reactants and intermediates
far too strongly. For instance, the chlorine evolution barrier is close
2
depicted in
To investigate the effect of surface orientation, we discuss the
energy profiles for the different facets of pure RuO , as seen in
Fig. 4a. The reaction profile for RuO (110) shows that the most
2
2
2
2

D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
279
the SnO
2
(110) surface. This is clearly seen when comparing the
(110) where
Projected Density of States for 2 and 3 ML on SnO
2
only small differences are observed, Fig. S.I. 7. Then, the geometric
effect disturbing the Ru surroundings implies that the Ru levels are
further away from the Fermi position and less active in binding
reactants, intermediates, or products. Therefore, self-poisoning by
2
Cl is less effective than for the pure RuO (110) surface.
A final comment regards the analysis of the chlorination of the
surfaces employed in the calculations. We identified through ab
initio thermodynamics that partial chlorination on the bridge posi-
tions, giving rise to the formation of surface oxy-chloride, is ener-
getically favored [16]. Indeed, partial chlorination of the surface
has been experimentally detected [18]. Nevertheless, calculations
indicate that the penetration of Cl atoms inside the lattice is ener-
getically very demanding for both RuO
more than 2 eV. Therefore, we expect that chlorination takes place
exclusively in the outermost layer by replacing some O or by
2 2
(110) and RuO (101),
b
occupying empty Rucus positions. This result will enable us to use
the intrinsically bulk-sensitive PGAA as a surface-sensitive tech-
nique for this particular problem.
3.5. In situ Prompt Gamma Activation Analysis
To study chlorine uptake, and hence the Cl coverage under stea-
dy state Deacon conditions, in situ Prompt Gamma Activation
Analysis was performed. Using PGAA, this can be achieved, since
Cl, as mentioned above, preferentially occupies surface sites and
thus, if the amount of Cl and Ru in the investigated volume is quan-
tified and the constant gas-phase Cl contribution is subtracted, a
ratio of Cl/Ru can be calculated that is proportional to the actual
Cl coverage. Since we performed these measurements in different
2
reaction mixtures, that is, with variable inlet p(O ) and fixed inlet
Fig. 4. Reaction energy profile for HCl oxidation on (a) pure RuO
surfaces, (b) RuO (110) compared to one and two monolayer RuO
2 2 2
SnO (110). The profiles correspond to the reaction 2HCl + O ? Cl + H O + O
leaving an oxygen atom on the surface, hence not the whole catalytic cycle is
depicted. The missing steps are repetition of some of the shown ones; the complete
profiles can be found in Supplementary information.
2
lowest energy
p(HCl), we can evaluate the response of feed composition to the
surface coverage, and thus, correlate the reaction rate with the sur-
face Cl concentration. By changing the oxygen content of the feed
2
2
adsorbed on
ꢁ
2
gas, chlorine production increases with p(O
approximately 0.4 order formal rate dependence (Fig. S.I. 5), in line
with the results reported above in the kinetic analysis (without Cl
2
) and displays an
2
to 2 eV, in contrast to 1.54 eV found for the pure rutile case. There-
and H O co-feed) and previously reported data on RuO2 [16].
2
fore, a single RuO
reactive than the pure RuO
can be traced back to the electronic modifications induced by the
presence of the SnO (110) support, see Figs. S.I. 7 and 8. The for-
2
epitaxial monolayer is expected to be much less
Therefore, we can assure that our measurements do not suffer from
2
(110). The reasons for this behavior
any catalytic artifacts, and hence, the spectroscopic results can be
correlated with the online catalytic data. We have also verified that
RuO /SnO slowly deactivates at the very initial stage of reaction
2
2
2
mation of the bimetallic overlayer implies in this particular case
a higher energy of the bands associated with Ru, making them
more prone to adsorption and, as a consequence, the surface is
more likely to be poisoned by reaction products. The role of the
electronic modifications induced by the close contact between Ru
and Sn oxides is very clear for the monolayer when the density dif-
ference at the interface is computed and density accumulation at
the Rucus position is favouring the interaction with electronegative
with time on stream. This deactivation is accompanied by an
increasing uptake of chlorine (Fig. 5), thus giving us a first hint that
adsorbed chlorine plays a detrimental role in the reaction.
To circumvent the time-on-stream-dependent surface chlorina-
tion, in the experiments described in the following section, an ini-
tial HCl treatment was performed. In the oxygen dependent series
of experiments, as stated above, increasing oxygen feed content
gives rise to higher reaction rate. The measured Cl/Ru ratio does
not change strongly; however a clear trend – in a relatively narrow
Cl/Ru range – is observed (Fig. 6a). Higher reaction rate is reached
at lower Cl/Ru ratio and, as Cl/Ru gradually increases, the reaction
rate linearly decreases. This trend is characteristic of pure RuO2
[39]; thus, we expect the contribution from the SnO2 carrier to
be negligible to the trend observed. It is worth noting that the
change in the Cl/Ru ratio from 0.72 to 0.66 was the result of alter-
2
adsorbates, Fig. S.I. 8. For two RuO layers, the situation is, how-
ever, rather different. The binding energies of reactants and inter-
mediates are smaller than those for the pure surface, and thus a
2
promoting effect on the activity of the RuO layers induced from
the SnO substrate is possible. In this case, the electronic effect de-
2
scribed for one monolayer is already smoothed out, and instead,
the geometric constraints of small lattice mismatch induced by
the presence of the support (strain induced by the epitaxy to the
ation of HCl:O ratio in the feed from 1:0.25 to 1:4; thus, a factor of
2
SnO
2
(110) surface) seem to control the ability of the Ru atoms to
(110)
16-fold increase in p(O ) induced a drop of Cl coverage of less than
2
bind the reaction intermediates weaker than the pure RuO
2
10%. As for the binder-containing catalyst, we observe the same
surface. Therefore, for a single monolayer, the electronic interac-
tion pushes the Ru levels toward the Fermi level, thus being more
prone to adsorption but also to self-poisoning by Cl. As for the 2 ML
case, the electronic effect is already quenched and the most impor-
tant contribution comes from the strain induced by the epitaxy to
trend (Fig. S.I. 6); however, due to strong binding of Cl to the alu-
mina, the Cl/Ru ratio is almost doubled. The two catalysts per-
formed very similarly during the PGAA experiments. Due to the
negative dependence of the reaction rate to the surface chlorine,
one can extrapolate the chlorine content to zero activity. When

2
80
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
tion of all possible species is exothermic, we expect no significant
coverage of empty sites, and thus, we assign this non-Cl species to
an ‘‘active’’ surface oxygen. Note that this is not necessarily the
coverage of all possible O-based species.
3.6. Temporal Analysis of Products
The TAP technique was primarily applied (i) to qualitatively as-
sess the influence of the SnO
transient responses of Cl and H
based materials as well as (ii) to analyze the influence of coverage
on the net Cl production. Fig. 7 shows the normalized transient re-
sponses of the reaction products upon pulsing of a mixture of HCl
and O over fresh RuO /SnO –Al , fresh RuO , and equilibrated
RuO /SnO . The transients of Kr (inert gas accompanying HCl and
in each pulse) are also shown for reference purposes. The shapes
of the transients of Cl and H O over the samples are comparable.
This preliminarily indicates that stabilization of RuO /SnO does
not induce prominent changes in the surface properties of the
ruthenium phase, as observed in parallel experiments on RuO
TiO [18]. Still, some differences are detected. Considering the Cl
transients (Fig. 7a), the response of RuO /SnO –Al appears shar-
per than that of pure RuO . In particular, the time at maximum
max, Fig. 7d) is slightly shorter, the width at half height (th/2
2
support and the Al
2 3
O binder on the
2
2
O during HCl oxidation over RuO
2
-
2
2
2
2
2
O
3
2
2
2
Fig. 5. Initial deactivation of RuO
HCl:O :N = 1:1:3) with the corresponding increasing Cl/Ru ratio with time-on-
stream as measured by in situ PGAA. Activity is normalized to the first data point.
2 2
/SnO at 523 K and 1 bar (feed composition:
O
2
2
2
2
2
2
2
2
/
2
2
2
2
2 3
O
2
(t
,
Fig. 7d) shifted to shorter times by 0.3 s, and the tailing is less
pronounced. Since the amount of chlorine produced by the
supported catalyst in the TAP reactor (see non-normalized re-
2
sponses in Fig. 7c) is consistently higher than for RuO , the sharper
Fig. 6. Normalized reaction rate (r
coverage (b) as measured by in situ PGAA at 573 K. Dataset corresponds to the
p(O )-dependent experiment with O :HCl:N varied between 0.25:1:3.75 and 4:1:0
i 0
/r ) versus Cl uptake (a) and of ‘‘active’’ O
2
2
2
3
ꢀ1
0
at a total flow of 166.6 cm min and 1 bar. The r reaction condition corresponds
to the feed ratio 1:1:3.
Fig. 7. Normalized transient responses of Cl
2 2
(a) and H O (b) on pulsing a mixture
normalizing the difference between the actual and the intercept
chlorine content to the intercept chlorine content, one can calcu-
late a coverage (based solely on the maximal Cl content under Dea-
con conditions) related to a non-Cl species that scales well with the
activity. Fig. 6b shows this first-order dependence. Since adsorp-
of HCl:O2 = 2:1 at 623 K over RuO /SnO –Al O , RuO , and RuO /SnO -eq. The
2
2
2
3
2
2
2
responses of the reference gas Kr are shown as dashed lines with the color of the
respective catalyst material. The insets (c and d) display the non-normalized
transient responses of Cl over the catalysts and the graphical representation of tmax
2
and th/2, respectively. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)

D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
281
Cl
2
response is related to favored chlorine evolution. This may be
3.7. Micro-kinetic modeling
tentatively attributed to both the active-phase morphology and
structure sensitivity of the reaction. As shown by HRTEM, the sup-
From the reaction energy profiles shown in Fig. 4 and the
corresponding partition functions, we have evaluated several exper-
imentalparameters discussed in above sections. In the micro-kinetic
(MK) models, a minimum of 12 elementary steps (Eqs. (2)–(7), plus
the reverse reactions) and the site balance have been taken into ac-
count for each of the studied surfaces (110), (101), and (100). With
the MK model, it is possible to self-consistently obtain the popula-
ported catalyst is constituted by a thin coat of few RuO
2
layers epi-
taxially grown onto the SnO carrier accompanied by small RuO
2
2
nanoparticles, both preferentially exposing {110} and {011} crystal
facets. This description applies also to the binder-containing cata-
lyst, as the presence of alumina has been shown not to affect the
morphology of RuO
steps of the reaction mechanism already reported for RuO
14,16] and extended above to different surface facets predicts a
2
/SnO
2
[19]. DFT analysis of the elementary
2
(110)
2 2
tions of intermediates (O , O, Cl, OH, H O) at the surface and the reac-
ꢀ1
[
tivity as either reaction rate or turnover frequency, TOFCl2 in s , at a
given time.
higher activity in HCl oxidation for the (110) and (011) planes over
the (100) plane. Further, the calculations on mono-/bilayers of
As benchmark, we have calculated the TOF for the RuO
and (100) surfaces, at the conditions reported by Over et al. [40]
(p(O ) = 0.5 mbar, p(HCl) = 2 mbar, T = 650 K), considering the con-
version after 2 h as in their experiments. The calculated value on
2
(110)
RuO
tive than the unsupported RuO
tatively related to a less effective self-poisoning of RuO
2
on SnO
2
indicate that the supported RuO
(Fig. 4). The latter outcome is ten-
by chlorine
2
bilayer is more ac-
2
2
2
ꢀ1
in view of the geometric effect caused by epitaxy, which relates well
to the easier chlorine evolution evidenced in the TAP data. The
similar transients obtained for equilibrated RuO /SnO (Fig. 7a–c)
2 2
seem to indicate that the main mechanistic features are retained,
and in spite of the strong morphological changes, this material
underwent upon use, leading to the formation of larger nanoparti-
2
RuO (110) using the formula in their work is 0.61 s , while the
ꢀ1
experimental estimate is 0.6 s . The agreement found between
our MK results and the low-pressure experimental value is, thus,
particularly remarkable, and gives a strong support to the validity
of the present approach. Low-pressure experiments were also car-
ried out for the (100) surface. In that case, the experimental esti-
ꢀ1
cles by sintering [19]. Even in this scenario, the gap to bulk RuO
still significant. Regarding the H O responses (Fig. 7b), the situation
appears reversed in the sense that the response of RuO /SnO
Al is broader than that of RuO , showing a tmax higher by 0.4 s
2
is
mate is 0.6 s , as well that contrasts slightly the calculated TOF
ꢀ1
2
value of 1.12 s . The origin of this difference might be related to
the ability of the (100) surface to reconstruct into (110)-like fac-
ets. Indeed, a similar reconstruction has been already proposed
2
2
–
2
O
3
2
and a th/2 shifted to longer times by 0.4 s as well. This difference
is not due to experimental artifacts, since the Kr responses are prac-
tically identical in all cases, but most likely relates to the presence
of alumina in the former sample. Therefore, the TAP response sug-
gests that alumina affects the adsorption/desorption equilibrium of
water, inducing a somewhat more impeded evolution of this
for TiO
surfaces are
2
(100) [41]. The surface energies for the (110) and (100)
ꢀ2
c110 = 0.041 and c100 = 0.047 eV Å and the difference
can be the driving force for reconstruction. The reconstruction
would account then for the similar reactivity of both (110) and
(100) surfaces observed experimentally. Finally, it is important
to notice that the Wulff construction and the XRD pattern of pure
product from the binder-containing catalyst. As for the RuO
catalyst, again, it shows a promoted product desorption with
respect to RuO indicating the role of structural effects, upon
2
/SnO
2
2
RuO powder seem to indicate 10–15% contribution of this (100)
facet, though X-ray diffraction is a bulk technique and thus cannot
precisely determine if the topmost layers are partially recon-
structed. Therefore, in modeling a nanoparticle structure, we ne-
glected the (100) facet. Based on the above discussion, in the MK
2
supporting ruthenium, in the reaction.
Valuable insights into the reaction mechanism on RuO
2 2
/SnO –
2
Al O
3
were obtained from the examination of the Cl
2
responses
2
modeling of case ‘‘RuO nanoparticle,’’ we have employed a relative
in pump–probe experiments. A detailed description of these exper-
iments was given in Refs. [19,26]. These measurements were sys-
surface area for the (110) facet of about 58% and 42% (101). As we
will see later, these results are robust enough as small modifica-
tions in these values do not change the trends observed.
In order to gain further insight into the mechanism, we have
investigated the inhibiting role of reaction products. For this, the
micro-kinetic model on the nanoparticle structure, MK–NP, was
tematically obtained with O
pump and HCl probe or HCl pump and O
between the pump and probe pulses ( t) was varied in the range
of 1–12 s, which enables to study the influence of the pump mole-
cule coverage on the Cl production. The consecutive cycles were
2
and HCl exchanging their roles (O
2
2
probe). The time delay
D
2
investigated at T = 573 K, p(O
variable initial pressures of either Cl
0.6 bar. The results are displayed in Fig. 9. Although the effect
shown here for H O is smaller than in Fig. 3b, the micro-kinetic
2
) = 0.2 bar, and p(HCl) = 0.2 bar with
linked in such a way that the time elapsed between the probe pulse
and the pump pulse of the next cycle was always 10 s. The cover-
age of the probe molecule at the beginning of the cycle is therefore
very low, since the probe pulse has almost entirely eluted after
2
or H O in the range of 0–
2
2
model based on the DFT-calculated parameters does tend to qual-
itative reproduce the flow experiments, and a reduction in the
overall activity is obtained when increasing the pressure of the
products at the reactor inlet.
We have analyzed in detail the dependence of the reaction rate
to the intermediates adsorbed on the surface with the aim of
unraveling the origin for the observations in the PGAA experi-
ments. For this purpose, we have simulated the experimental con-
ditions used during PGAA (p(HCl) = 0.2 bar, 573 K) by varying the
1
0 s. The transient responses of Cl
sented in Fig. 8. Chlorine production progressively decreases as
increases both when O is the pump (Fig. 8a) and when HCl is the
pump (Fig. 8b). This trend is qualitatively identical to RuO [18]
2 2 2 2 3
over RuO /SnO –Al O are pre-
D
t
2
2
and indicates that the catalytic activity depends on the coverage
of both adsorbed species. Further details on the mechanism can
be attained by the analysis of the
tion (Fig. 8c). When O is the pump molecule, increasing the time
until HCl is pulsed ( t) from 1 to 12 s provokes a decrease in Cl
2
Dt effect on the total Cl produc-
2
D
2
2
p(HCl):p(O ) ratio from 1:4 to 1:0.25, see Fig. 10. To better under-
production of ca. 100%. This is a direct evidence of reaction with
adsorbed oxygen. Similarly, when HCl is the pump, an important
stand the data, we have proceeded as follows, first the individual
(110) and (101) surfaces have been analyzed and the rates for
these surfaces as a function of the surface coverages obtained. In
the second step, a nanoparticle containing both (110) and (101)
surfaces was considered. Finally, the role of the SnO support
2
was studied in a different set of computational experiments,
Fig. 11. To compare all reaction rates in the batch MK model, the
decrease in Cl
served, implying that the reaction occurs through adsorbed Cl spe-
cies. These data support that HCl oxidation on RuO /SnO –Al
proceeds via a Langmuir–Hinshelwood mechanism, in line with
the observations made for bulk RuO and RuO /TiO catalysts
18] and also in agreement with PGAA data.
2
production with increasing Dt (ca. 120%) is ob-
2
2
2
O
3
2
2
2
[
2
production of Cl was obtained always at the same time (t = 0.1 s).

2
82
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
Fig. 8. Transient responses of Cl
related to Cl production derived from these experiments. The total Cl
the experiment at t = 12 s.
2
over RuO
2
/SnO
2
–Al
2
O
3
in pump–probe experiments of (a) O
2
and HCl and (b) HCl and O
2
at different time delays and 623 K. (c) Normalized
) and probe (a ) pulses, was normalized to
2
2
production, defined as the sum of the areas in the pump (a
1
2
D
the negative slope found for hCl results into positive dependence
are considered (see central panel). Nevertheless,
when 1 ꢀ hCl or h
O
geometric constraints on the (101) surface seem to give rise im-
peded activity as the total coverage of all species is always lower
than on the (110) surface due to strong repulsions along the Rucus
chains.
The MK–NP model shows a mixed behavior, the reaction rate is
about a half of that of the most active surface (surface fractions are
5
8% (110) and 42% (101)), and the hCl coverage is obviously also
much higher than for the (101) facet, but smaller than that of
the (110) one. The dependence of the activity with the amount
of chlorine on the surface is again negative, indicative of the
poisoning effect of surface chlorine. As for the individual facets,
positive dependencies with the amount of oxygen on the surface
or with 1 ꢀ hCl are found. These results suggest that there is an
optimum relationship between the Cl and O coverages that pro-
vides the most active catalytic phase. More importantly, for highly
covered Cl surfaces, the difficulties in finding two neighboring
2
open Rucus sites can block O dissociative adsorption impeding
the re-oxidation and preventing the catalytic process to complete
the cycle. Thus, we conclude that though Cl recombination is the
elementary step with the highest barrier, at high Cl coverages,
Fig. 9. Micro-Kinetic results: Dependence of the relative rate,
pressures of products at the inlet, p of H O or Cl (in bar). The r
correspond to 0.2 bar for both reactants, HCl and O at 573 K and without any
products present at the inlet. All rates determined under initial rate conditions. r
and r have the same meaning as in Fig. 3.
r
i
/r
0
, with the
i
2
2
0
reaction conditions
2
the high stability of Clcus drives O dissociative chemisorption as
2
the likely rate limiting step. Additionally, geometric constrains
enhance differences in the reactivity of different facets. Finally,
we have determined the apparent activation energy for the NP
case by determining the MK-calculated activity at different
temperatures. In the region 400–673 K, the obtained estimate is
0
–
1
Starting by the RuO
2
(110) MK model, the reaction rate is found
72 kJ mol , in reasonable agreement with the results in the cata-
–
1
to be dependent on the chlorine coverage at Rucus sites, see
Fig. 10a. Under the conditions in the experiment, the values of Cl
coverage observed are high (though its absolute number is not
estimated) and thus Cl is likely the most abundant reactive inter-
lytic test section (69 kJ mol ).
The results for the RuO adlayers on SnO
been analyzed with the same methodology, see Fig. 11. The reac-
tion rate obtained for the RuO /SnO (1 ML) is close to zero for
all HCl:O ratios, showing that this model surface is not active at
573 K, and thus is the least active of all investigated surfaces. In-
stead, the activity observed for the RuO /SnO (2 ML) is slightly
higher than that observed for the pure RuO (110) at the same
2
2
(110) surface have
2
2
mediate. The largest activity is retrieved for the 1:4 p(HCl):p(O
2
)
2
ratio and decays along the series when less O is present. The
2
dependence observed shows almost a linear negative slope indica-
tive of the self-poisoning effect of Cl under such conditions. Obvi-
ously, the inverse slope is found when the dependence with the
coverage of non-chlorine species 1 ꢀ hCl, is investigated. For the
2
2
2
time. Again, the activity depends negatively on the chlorine cover-
age (hCl) of the surface. The corresponding oxygen coverage is even
smaller than for the (110) surface, but in both cases, a positive
dependence on the calculated Ocus coverage and the total fraction
of non-chlorine-containing sites (1 ꢀ hCl) is obtained.
(
101) surface, the situation is somewhat different, as the intrinsic
activity seems to be lower, and the chlorine coverage is much
smaller than for the (110) facet, about 0.6 ML. Also in this case,

D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
283
Fig. 10. Micro-Kinetic results: Chlorine, hCl, oxygen, h
O
, and 1 ꢀ hCl coverage (in rows) at cus position obtained with the micro-kinetic model at initial pressures of 0.2 bar of
at t = 0.1s) for RuO (110), RuO (101), and
HCl and varying p(O ) pressures (ratios = 1:0.25, 1:0.5, 1:1, 1:2, 1:4) at 573 K and the corresponding normalized reaction rate (r/r
2
0
2
2
the nanoparticle, NP (in columns). All data are normalized to the NP case at 1:1 feed ratio.
Fig. 11. (a) Micro-Kinetic results: Normalized reaction rate as a function of chlorine cus coverage, hCl, for the HCl:O
2
ratios of = 1:0.25, 1:0.5, 1:1, 1:2, 1:4 at 573 K and the
(2 ML). (c) Normalized reaction rate as a function of the
(2 ML). (d) Same as (c) for the 1 ꢀ hCl coverage. All data are normalized to the NP case at 1:1 feed ratio
single monolayer on the SnO
calculated cus oxygen coverage h
Fig. 10).
2
(110) surface RuO
2
/SnO
2
(1 ML). (b) Same as (a) for the RuO
/SnO
2 2 2
bilayer, RuO /SnO
O
for the bilayer model RuO
2
2
(

2
84
D. Teschner et al. / Journal of Catalysis 285 (2012) 273–284
DFT-based MK modeling was capable to reproduce all trends
observed in steady-state catalytic experiments and during PGAA.
Deacon reaction over RuO catalysts thus follows the Langmuir–
Hinshelwood mechanism with oxygen activation being rate deter-
mining under typical Deacon conditions.
Appendix A. Supplementary material
2
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2011.09.039.
References
[
[
[
[
1] www.eurochlor.org.
4
. Conclusions
2] A.J. Johnson, A.J. Cherniavsky, US Patent 2542,961, 1951.
3] F. Wattimena, W.M.H. Sachtler, Stud. Surf. Sci. Catal. 7 (1981) 816.
4] T. Kiyoura, N. Fujimoto, M. Ajioka, T. Suzuki, Y. Kogure, K. Kanaya, T. Nagayama,
EP184413-A, 1984.
This article describes experimental and computational results
that helped to assess realistic aspects as well as mechanistic details
of Deacon reaction over a technical RuO -based catalyst. We ap-
plied, beyond standard characterization techniques, state-of-the-
art experimental methods (aberration-corrected HRTEM, TAP,
in situ PGAA) in concert with DFT calculation and micro-kinetic
modeling. HRTEM shows that two major RuO
present in RuO /SnO catalysts, namely 2–4-nm-sized particles
and 1–3-ML-thick epitaxial RuO films attached to the SnO
[
5] T. Hibi, H. Nishida, H. Abekawa, US Patent 5871,707, 1999.
2
[6] T. Hibi, H. Abekawa, K. Seki, T. Suzuki, T. Suzuta, K. Iwanaga, T. Oizumi,
EP936184, 1999.
[
[
[
7] T. Hibi, T. Okuhara, K. Seki, H Abekawa, H. Hamamatsu, WO200110550-A1,
001.
2
8] K. Iwanaga, K. Seki, T. Hibi, K. Issoh, T. Suzuta, M. Nakada, Y. Mori, T. Abe,
Sumitomo Kagaku (2004).
9] K. Seki, Catal. Surv. Asia 14 (2010) 168.
2
morphologies are
2
2
[
[
10] A. Wolf, L. Mleczko, O.F. Schlüter, S. Schubert, EP2026905, 2006.
11] A. Wolf, J. Kintrup, O.F. Schlüter, L. Mleczko, EP2027062, 2006.
2
2
support particles. A large fraction of nanoparticles exposes {110}
and {101} type facets. Steady-state kinetic experiments indicate
that both reactants display positive rate dependence (HCl: 0.2;
[12] A. Wolf, L. Mleczko, S. Schubert, O.F. Schlüter, EP2027063, 2006.
[
[
13] J. Heemskerk, J.C.M. Stuiver, DE1567788, 1964.
14] S. Zweidinger, D. Crihan, M. Knapp, J.P. Hofmann, A.P. Seitsonen, C.J.
Weststrate, E. Lundgren, J.N. Andersen, H. Over, J. Phys. Chem. C 112 (2008)
9966.
O
2
: 0.4 order), whereas products cause strong inhibition. When
[
[
15] D. Crihan, M. Knapp, S. Zweidinger, E. Lundgren, C.J. Weststrate, J.N. Andersen,
A.P. Seitsonen, H. Over, Angew. Chem. 120 (2008) 2161.
16] N. López, J. Gómez-Segura, R.P. Marín, J. Pérez-Ramírez, J. Catal. 255 (2008) 29.
both products are co-fed, the reaction order of HCl and O gets 0.5
2
and 1, respectively. Pump–probe experiments conducted in
the TAP reactor clearly indicated a reaction mechanism within
adsorbed species, without the influence of bulk oxygen
[17] F. Studt, F. Abild-Pedersen, H.A. Hansen, I.C. Man, J. Rossmeisl, T. Bligaard,
ChemCatChem 2 (2010) 1.
[
18] M.A.G. Hevia, A.P. Amrute, T. Schmidt, J. Pérez-Ramírez, J. Catal. 276 (2010)
41.
(
Langmuir–Hinshelwood). By using in situ PGAA, we demonstrated
1
that adsorbed Cl poisons the surface and the reactivity is propor-
tional to the coverage of a ‘‘non-Cl,’’ likely adsorbed oxygen spe-
cies. All experimental observations (kinetic trends, coverage
effects) were conciliated with DFT-based micro-kinetic modeling.
Calculations additionally indicated that the reaction, despite previ-
ous claims, is structure sensitive. The (100) and (110) facets, as
[19] C. Mondelli, A.P. Amrute, F. Krumeich, T. Schmidt, J. Pérez-Ramírez,
ChemCatChem 3 (2011) 657.
[
20] Z. Revay, T. Belgya, L. Szentmiklosi, Z. Kis, A. Wootsch, D. Teschner, M.
Swoboda, R. Schlogl, J. Borsodi, R. Zepernick, Anal. Chem. 80 (2008) 6066.
21] D. Teschner, J. Borsodi, A. Wootsch, Z. Revay, M. Havecker, A. Knop-Gericke,
S.D. Jackson, R. Schlogl, Science 320 (2008) 86.
[
[
[
22] Z. Revay, Anal. Chem. 81 (2009) 6851.
23] J.T. Gleaves, G.S. Yablonsky, P. Phanawadee, Y. Schuurman, Appl. Catal. A 160
well as the 2 ML RuO
higher reactivity as compared to the (101) surface, whereas the
ML RuO film seems to be inactive. At typical reaction conditions,
2 2
film-covered SnO give rise to significantly
(
1997) 55.
[24] J. Pérez-Ramírez, E.V. Kondratenko, Catal. Today 121 (2007) 160.
[
[
25] A.P. Amrute, C. Mondelli, M.A.G. Hevia, J. Pérez-Ramírez, J. Phys. Chem. C 115
2011) 1056.
26] A.P. Amrute, C. Mondelli, M.A.G. Hevia, J. Pérez-Ramírez, ACS Catal. 1 (2011)
583.
1
2
(
the surface is extensively chlorinated; thus, the active material is
better described as a surface oxy-chloride. Despite the highest bar-
rier of Cl recombination, under realistic Deacon conditions, oxygen
activation is rate determining.
[
[
[
[
[
[
27] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
28] G. Kresse, J.M. Furthmüller, Phys. Rev. B 54 (1996) 11169.
29] B. Hammer, L.B. Hansen, J.K. Nørskov, Phys. Rev. B 59 (1999) 7413.
30] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
31] G. Wulff, Z. Krystallogr. Min. 34 (1901) 449.
32] I. Chorkendorff, J.W. Niemantsverdriet, Concepts of Modern Catalysis and
Kinetics, Wiley–VCH, Weinheim, 2003.
Acknowledgments
[
33] G. Novell-Leruth, J.M. Ricart, J. Pérez-Ramírez, J. Phys. Chem. C 112 (2008)
1
3554.
We thank Bayer MaterialScience AG for permission to publish
these results. Drs. M.A.G. Hevia and L. Durán-Pachón (ICIQ Tarrag-
ona) are acknowledged for experimental input. ETH Zurich, the
BMBF Project 033R018A, and MICINN (CTQ2009-07753/BQU,
CSD2006-0003) are thanked for financial support and BSC-RES
for providing generous computational resources. For the in situ
PGAA measurements, we acknowledge the financial support of
the EU FP7 NMI3 Access Programme and the NKTH NAP VENEUS
Grant (OMFB-00184/2006).
[
[
34] http://www.maplesoft.com/products/maple/.
35] N. Lu, Q. Wan, J. Zhu, J. Phys. Chem. Lett. 1 (2010) 1468.
[36] M. Batzill, K. Katsiev, J.M. Burst, U. Diebold, A.M. Chaka, B. Delley, Phys. Rev. B
2 (2005) 165414.
7
[
[
37] http://webbook.nist.gov/chemistry (retrieved December 2010).
38] H. Wang, W.F. Schneider, D. Schmidt, J. Phys. Chem. C 113 (2009) 15266.
[39] D. Teschner, et al., Unpublished results.
[40] S. Zweidinger, J.P. Hofmann, O. Balmes, E. Lundgren, H. Over, J. Catal. 272
(
2010) 169.
[
41] P.D.J. Lindan, N.M. Harrison, Surf. Sci. 479 (2001). L375.