Performance, structure, and mechanism of CeO2 in HCl oxidation to Cl2

Article abstract of DOI:10.1016/j.jcat.2011.11.016

Experimental and theoretical studies reveal performance descriptors and provide molecular-level understanding of HCl oxidation over CeO₂. Steady-state kinetics and characterization indicate that CeO₂ attains a significant activity level, which is associated with the presence of oxygen vacancies. Calcination of CeO₂ at 1173 K prior to reaction maximizes both the number of vacancies and the structural stability of the catalyst. X-ray diffraction and electron microscopy of samples exposed to reaction feeds with different O₂/HCl ratios provide evidence that CeO₂ does not suffer from bulk chlorination in O₂-rich feeds (O₂/HCl ≥ 0.75), while it does form chlorinated phases in stoichiometric or sub-stoichiometric feeds (O₂/HCl ≤ 0.25). Quantitative analysis of the chlorine uptake by thermogravimetry and X-ray photoelectron spectroscopy indicates that chlorination under O₂-rich conditions is limited to few surface and subsurface layers of CeO₂ particles, in line with the high energy computed for the transfer of Cl from surface to subsurface positions. Exposure of chlorinated samples to a Deacon mixture with excess oxygen rapidly restores the original activity levels, highlighting the dynamic response of CeO₂ outermost layers to feeds of different composition. Density functional theory simulations reveal that Cl activation from vacancy positions to surface Ce atoms is the most energy-demanding step, although chlorine-oxygen competition for the available active sites may render re-oxidation as the rate-determining step. The substantial and remarkably stable Cl₂ production and the lower cost of CeO₂ make it an attractive alternative to RuO₂ for catalytic chlorine recycling in industry.

Full text of DOI:10.1016/j.jcat.2011.11.016

Journal of Catalysis 286 (2012) 287–297
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Performance, structure, and mechanism of CeO2 in HCl oxidation to Cl2
Amol P. Amrute a, Cecilia Mondelli a, Maximilian Moser a, Gerard Novell-Leruth b, Núria López b,
Dirk Rosenthal c, Ramzi Farra c, Manfred E. Schuster c, Detre Teschner c,d, Timm Schmidt e,
Javier Pérez-Ramírez a,
⇑
a Institute for Chemical and Bioengineering, Department of Chemistry and Applied Biosciences, ETH Zurich, Wolfgang-Pauli-Strasse 10, CH-8093 Zurich, Switzerland
b Institute of Chemical Research of Catalonia (ICIQ), Av. Països Catalans 16, 43007 Tarragona, Spain
c Fritz-Haber Institute of the Max Planck Society, D-14159 Berlin, Germany
d Institute of Isotopes, Hungarian Academy of Science, H-1525 Budapest, Hungary
e Bayer MaterialScience AG, PUR-PTI-PRI, Chempark B598, D-41538 Dormagen, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Experimental and theoretical studies reveal performance descriptors and provide molecular-level under-
standing of HCl oxidation over CeO2. Steady-state kinetics and characterization indicate that CeO2 attains
a significant activity level, which is associated with the presence of oxygen vacancies. Calcination of CeO2
at 1173 K prior to reaction maximizes both the number of vacancies and the structural stability of the
catalyst. X-ray diffraction and electron microscopy of samples exposed to reaction feeds with different
O2/HCl ratios provide evidence that CeO2 does not suffer from bulk chlorination in O2-rich feeds (O2/
HCl P 0.75), while it does form chlorinated phases in stoichiometric or sub-stoichiometric feeds (O2/
HCl 6 0.25). Quantitative analysis of the chlorine uptake by thermogravimetry and X-ray photoelectron
spectroscopy indicates that chlorination under O2-rich conditions is limited to few surface and subsurface
layers of CeO2 particles, in line with the high energy computed for the transfer of Cl from surface to sub-
surface positions. Exposure of chlorinated samples to a Deacon mixture with excess oxygen rapidly
restores the original activity levels, highlighting the dynamic response of CeO2 outermost layers to feeds
of different composition. Density functional theory simulations reveal that Cl activation from vacancy
positions to surface Ce atoms is the most energy-demanding step, although chlorine–oxygen competition
for the available active sites may render re-oxidation as the rate-determining step. The substantial and
remarkably stable Cl2 production and the lower cost of CeO2 make it an attractive alternative to RuO2
for catalytic chlorine recycling in industry.
Received 9 August 2011
Revised 20 November 2011
Accepted 22 November 2011
Available online 26 December 2011
Keywords:
Chlorine recycling
HCl oxidation
CeO2
Oxygen vacancies
Chlorination
Ó 2011 Elsevier Inc. All rights reserved.
1. Introduction
ficiently active and durable catalysts. The CuCl2/pumice catalyst
patented by Henry Deacon in 1868 [3] exhibited fast deactivation
The heterogeneously catalyzed gas-phase oxidation of HCl to
Cl2 (Eq. (1)) is a sustainable route to recycle chlorine from by-prod-
uct HCl streams derived from the manufacture of polyurethanes
and polycarbonates with low energy input [1,2]. The amount of
waste HCl raises due to the growing demand for these versatile
plastics. As selling the HCl excess is unfeasible, and the neutraliza-
tion option is unattractive, its catalytic conversion to Cl2 is receiv-
ing an increasing interest.
due to volatilization of the active phase in the form of copper chlo-
rides. Other examples of processes of limited success are the Shell-
Chlor process (based on CuCl2–KCl/SiO2) [4] and the Mitsui-Toatsu
process (based on Cr2O3/SiO2) [5]. Only recently, ruthenium sup-
ported on specific carriers was successfully developed for large-
scale chlorine recycling [1]. In particular, Sumitomo has applied
RuO2/SiO2/TiO2-rutile in a plant producing 100 kton Cl2 per year
[2,6], while Bayer’s RuO2/SnO2–Al2O3 catalyst [7,8] has been piloted
and is ready for industrial use. The distinctive features of RuO2-
based catalysts are the high activity at low temperature and the
remarkable stability against bulk chlorination [9–11].
4HCl þ O2 $ 2Cl2 þ 2H2O
ð1Þ
Since its introduction and till recent times, the industrialization of
HCl oxidation has suffered from many failed attempts to obtain suf-
The implementation of the catalytic HCl oxidation technology
would expand if cheaper, but comparably stable, alternatives to
RuO2-based catalysts were developed. The high and fluctuating
market price of ruthenium indeed reflects in large costs for new
plants [12]. These considerations have constituted the driving force
⇑
Corresponding author. Fax: +41 44 633 1405.
E-mail address: jpr@chem.ethz.ch (J. Pérez-Ramírez).
0021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.11.016

288
A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
Table 1
to search for alternative cost-effective catalytic materials. A first
step in this direction is the recent design of a copper delafossite
(CuAlO2) catalyst, which exhibited stable performance for more
than 1200 h on stream [13]. However, this catalyst experienced
significant copper loss.
Description of catalysts’ codes.
Catalyst code
Calcination
temperature (K)
Catalyst
state
Feed O2/HCl
ratio (mol molꢀ1
)
CeO2-573-F
573
Fresh
Fresh
Fresh
Fresh
Fresh
Fresh
Used
Used
Used
Used
Used
Used
Used
Used
Used
Used
Used
Used
Used
–
–
–
–
–
–
2
2
2
2
2
2
0
0
0.25
0.75
CeO2-773-F
773
CeO2 (fluorite structure [14]) attracted our interest in view of its
wide use in redox processes in many research fields [15–21] and
oxidation reactions in particular [22–26]. In oxidations, it has been
employed both as catalyst and co-catalyst. The success of CeO2
stems from the easy generation of oxygen vacancies [22,27,28] that
facilitate activation and transport of oxygen species. Nevertheless,
only a few works have derived quantitative correlations between
its oxidation activity and this structural peculiarity, typically mea-
sured in terms of oxygen storage capacity (OSC). Furthermore,
comparisons have been preferably reported between CeO2 and so-
lid solutions of CeO2 with ZrO2 or SiO2 [29], or between ceria-con-
taining catalysts including supported (active) metal phases [30].
CeO2-based materials have been claimed in the patent literature
as catalysts potentially suited for HCl oxidation [31–33]. However,
there are no studies gathering a fundamental understanding of the
Deacon chemistry on CeO2. Only the detailed assessment of CeO2
activity and stability will determine its potential as an alternative
to RuO2-based catalysts for chlorine recycling.
CeO2-1023-F
CeO2-1173-F
CeO2-1273-F
CeO2-1373-F
CeO2-573-2
1023
1173
1273
1373
573
CeO2-773-2
773
CeO2-1023-2
CeO2-1173-2
CeO2-1273-2
CeO2-1373-2
CeO2-773-0
CeO2-1173-0
CeO2-1173-0.25
CeO2-1173-0.75
CeO2-1173-0.25-2
CeO2-1173-0.25-7
CeO2-1173-0-7
1023
1173
1273
1373
773
1173
1173
1173
1173
1173
1173
0.25 followed by 2
0.25 followed by 7
0 followed by 7
(573 K), at which the OSC would be evaluated in relation to a few
outermost surface layers of CeO2, the H2 uptake was first separately
measured at various temperatures (523–873 K), defined on the ba-
sis of the H2-TPR profiles (Figs. S1–S2 in the ESI). The degree of bulk
chlorination of selected catalysts after HCl oxidation was deter-
mined by thermogravimetric analysis coupled to mass spectrome-
try (TGA–MS). The structure, morphology, and composition of the
catalyst particles were studied by high-resolution transmission
electron microscopy (HRTEM). X-ray photoelectron spectroscopy
(XPS) was applied to determine the degree of chlorination of the
used CeO2 catalysts. Three different layer models were employed
to calculate the number of layers in CeO2 that experienced chlorina-
tion. The first model was based on the inelastic mean free path
(IMFP) and on 100% Cl occupation, the second on 80% of IMFP for
practical effective attenuation length (EAL) and on 100% Cl occupa-
tion, and the third on 80% of IMFP for practical EAL and on 75% Cl
occupation. For details on the characterization techniques, the read-
er is referred to the Electronic supplementary information (ESI).
Herein, we report on the use of CeO2 in HCl oxidation, as well as
on the fundamental knowledge derived by means of a multi-tech-
nique approach. Performance–structure–mechanism relationships
are established, collecting information from extensive kinetic tests
in a flow reactor at ambient pressure, detailed characterization (in
relation to the catalyst activation procedure and the operating con-
ditions), and mechanistic investigations using in situ infrared spec-
troscopy and density functional theory calculations.
2. Experimental
2.1. Catalysts
CeO2 (Aldrich, nanopowder, code 544841) was calcined in static
air at various temperatures in the range of 573–1373 K using a
heating rate of 10 K minꢀ1 and a dwell time of 5 h prior to use.
These samples are referred to as ‘‘fresh.’’ A complete list of all sam-
ples codes used along the manuscript, together with their descrip-
tion, is reported in Table 1.
2.3. Catalytic tests
2.2. Characterization
The oxidation of HCl with O2 was carried out at ambient pres-
sure in a quartz fixed-bed micro-reactor (8 mm i.d.) using 0.5 g
All the samples were subjected to basic characterization prior to
and after catalytic testing. More extensive characterization was
performed on CeO2-773 and CeO2-1173. In particular, fresh sam-
ples of these two catalysts were compared to samples exposed to
reaction mixtures with feed O2/HCl ratios equal to 2, 0.75, 0.25,
or 0 (see Section 2.3 for detailed testing conditions).
of catalyst (particle size = 0.4–0.6 mm),
a
total flow of
166 cm3 STP minꢀ1, a bed temperature (Tbed) of 703 K, and reaction
times up to 3 h. The catalyst was kept at 703 K for 30 min in N2
flow before admitting a mixture containing 10 vol.% HCl (Messer,
purity 2.8, anhydrous) and 20 vol.% O2 (Pan Gas, purity 5.0), bal-
anced in N2 (Pan Gas, purity 5.0).
Standard characterization of the samples was performed by
powder X-ray diffraction (XRD), nitrogen adsorption at 77 K, and
temperature-programmed reduction with hydrogen (H2-TPR).
The oxygen storage capacity (OSC) was measured by estimating
the H2 uptake of the samples (after a similar pre-treatment in inert
gas as prior to the catalytic evaluation), as this represents an indi-
rect quantification of the oxygen that the samples can store (Eq.
(2)) [34–36].
Steady-state kinetic studies were performed on CeO2-1173-F.
First of all, it was experimentally verified that, under our experi-
mental conditions, the reaction was not limited by extra- and/or
intra-particle diffusion limitations. For this purpose, we varied (i)
the total flow rate (83–250 cm3 STP minꢀ1) at constant space time
(W/F0(HCl) = 11.2 g h molꢀ1, defined as the ratio of the catalyst
mass to the inlet molar flow of HCl) and (ii) the particle size
(0.075–0.6 mm) with other conditions constant, respectively
(Fig. S3 in the ESI). The kinetic experiments included changes of
the feed composition at different temperatures and space times.
In particular, the feed O2/HCl ratio was varied in the range of
0.5–7 at constant feed HCl concentration (10 vol.%) and in the
range 0.25–2 at constant feed O2 concentration (10 vol.%) at
Tbed = 603, 653, and 703 K and space times of 5.6, 9, and
CeO2 þ xH2 $ CeO2ꢀx þ xH2O
ð2Þ
This method was chosen among others reported in the literature
[37], as the straightforward approach of determining the OSC by
pulse chemisorption of oxygen [38] did not provide reproducible re-
sults on our samples. In order to select the appropriate temperature

A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
289
11.2 g h molꢀ1. CeO2-1173-F samples exposed to O2/HCl = 0, 0.25,
3. Results and discussion
or 0.75, and CeO2-773-F treated in O2/HCl = 0 (only HCl) at 703 K
for 3 h were collected for characterization, after switching the feed
to N2 and allowing the reactor to cool down. In addition, experi-
ments were carried out at 703 K combining a HCl-rich step, in
which the catalyst was exposed to a feed composition O2/HCl = 0
or 0.25 (causing bulk chlorination), and a O2-rich step, in which
substantial oxygen excess was supplied (O2/HCl = 2 or 7). Cl2 quan-
tification at the reactor outlet was carried out by iodometric titra-
tion in a Metter Toledo G20 Compact Titrator, using the protocol
reported elsewhere [39]. HCl conversion was determined from
the Cl2 produced, taking into account the reaction stoichiometry.
3.1. Catalytic evaluation
3.1.1. Influence of calcination temperature on activity
CeO2 samples calcined at different temperatures were tested in
the gas-phase catalytic oxidation of HCl in a continuous-flow fixed-
bed reactor operated at ambient pressure and under isothermal
conditions. XRD analysis showed that the fresh samples exhibited
the characteristic reflections of cerium(IV) oxide. The HCl conver-
sion over the samples ranged from 2% to 29% (Table 2), remaining
essentially constant in the course of the 3-h test. Interestingly, the
CeO2 samples calcined at temperatures between 573 and 1173 K
were similarly active, achieving the highest conversion levels,
while higher calcination temperatures led to a progressive and al-
most complete depletion of the activity. These findings have been
related to changes in the textural and structural properties of the
materials and are first discussed on the basis of N2 adsorption re-
sults. For the fresh catalysts, higher calcination temperatures led
to lower total surface area (SBET) values (Table 2). This result is ex-
pected, as higher temperatures favor sintering of the particles. The
dependence of the HCl conversion on the SBET of the fresh samples
is shown in Fig. 1a (open symbols). For SBET values > 25 m2 gꢀ1 (i.e.,
calcination temperature below 1173 K), the activity is independent
of the SBET, while for SBET values < 25 m2 gꢀ1 (i.e., calcination tem-
perature above 1173 K), it strongly depends on the surface area.
In particular, CeO2-1273-F (SBET = 12 m2 gꢀ1) and CeO2-1373-F
(SBET = 1 m2 gꢀ1), respectively, exhibited 14% and 2% HCl conver-
sion, i.e., 2 and ca. 15 times lower activity than the other catalysts.
Exposure of the catalysts to reaction conditions generally led to a
drop in SBET (Table 2). The change is significant for CeO2-573-2,
CeO2-773-2, and CeO2-1023-2, and evidences that substantial
sintering of the CeO2 particles occurred during HCl oxidation. In
contrast, the SBET values of CeO2-1173-2, CeO2-1273-2, and CeO2-
1373-2 were hardly affected, indicating that calcination at or above
1173 K resulted in stabilized materials in HCl oxidation. Calcina-
tion of the starting CeO2 nanopowder at 1173 K provides the best
compromise between performance and stability against sintering.
Generally, the slope–plateau relation used to depict the depen-
dence of activity on the SBET of fresh samples turned out to be a
good description for the used catalysts too (Fig. 1a, solid symbols).
An experiment showing practically constant HCl conversion over
CeO2-773-F in contrast to a strongly decreasing SBET with time-
on-stream further supports the observed independence of the
activity on SBET for values of the latter greater than 25 m2 gꢀ1
(Fig. S4 in the ESI). It was accordingly concluded that another
descriptor is needed to rationalize the HCl oxidation activity of
CeO2.
2.4. In situ Fourier transform infrared spectroscopy
A home-made transmission cell, especially designed to with-
stand the corrosive HCl oxidation mixture, was used for in situ Fou-
rier transform infrared (FTIR) spectroscopy measurements. The
sample in powder form was pressed into a self-supporting disc
(15 mg cmꢀ2) under 30 bar for 10 s. The wafer, having a thickness
of 50 lm, was placed in the sample holder, which also serves as
the heating system. Inlet gases were analytical grade and con-
trolled by mass flow controllers. The sample was heated in
20 vol.% O2 in N2 (100 cm3 STP minꢀ1) up to 653 K (5 K minꢀ1
)
and allowed to stabilize. After ca. 1 h, 1 vol.% HCl was introduced
into the O2/N2 flow to mimic a diluted Deacon feed condition.
Spectra were recorded with a Varian-670 FTIR spectrometer at
653 K and with 4 cmꢀ1 resolution.
2.5. Computational details
Density functional theory (DFT) simulations were applied to
CeO2 slabs. The calculations were performed with the 5.2 version
of the VASP code [40]. The functional of choice was a GGA + U
scheme to account in an approximate way for the complexity aris-
ing from the presence of f-electrons in cerium. The GGA was PBE
[41], the U parameter chosen was set to 4.5 [42,43], and the inner
electrons were replaced by PAW pseudopotentials [44]. The 12 va-
lence electrons of Ce atoms in the 5s, 5p, 6s, 4f, and 5d states and
the 6 valence electrons of O atoms in the 2s and 2p states were con-
sidered explicitly. The valence electrons were expanded in plane
waves with a cutoff energy of 400 eV. For the bulk, the k-point
sampling was 9 ꢁ 9 ꢁ 9 [45]. The cell parameter obtained was
5.486 Å, in good agreement with experimental results and previous
calculations [42,43]. The (111) facet was selected as it is known to
be the lowest energy surface [43]. The chosen slab corresponds to a
p(2 ꢁ 2) reconstruction and contains three layers. The k-point sam-
pling was set to 3 ꢁ 3 ꢁ 1 in this case. The calculated surface en-
ergy for this structure was 0.013 eV Åꢀ2. For this surface, several
vacancy structures were considered and benchmarked against
more accurate calculations in the literature employing hybrid
methods [43]. When forming a vacancy or adsorbing on this sur-
face, the systems were allowed to relax in all directions, except
for the last layer of the slab. In all cases, the slabs were interleaved
by 10 Å and decoupled from electronic interactions due to spurious
polarizations given the asymmetry of the adsorption configuration.
Spin-polarized calculations were performed when required. The
CI-NEB method [46] was employed to locate the transition states,
and the vibrational analysis of the potential transition state struc-
tures was performed to fully characterize the saddle points. In the
calculation of vibrational frequencies, only the adsorbed structures
of intermediates or transition states were considered. The numer-
ical Hessian was calculated with two steps of 0.02 Å for each
degree of freedom and then diagonalized to obtain the correspond-
ing eigenvalues and eigenmodes.
As mentioned in the introduction, the successful application of
CeO2-based materials in (electro) catalysis has been related to its
capability to generate oxygen vacancies in the lattice [19,27].
Therefore, our CeO2 samples were further investigated to assess
Table 2
Characterization and catalytic activity data.
a
b
SBET (m2 gꢀ1
)
XHCl (%)
OSCc
(
lg O g
)
ꢀ1
Catalyst
cat
Fresh
Used
CeO2-573
CeO2-773
CeO2-1023
CeO2-1173
CeO2-1273
CeO2-1373
117
106
53
30
12
1
46
27
39
25
11
1
29
25
25
27
14
2.1
–
390
382
376
237
78
a
b
c
Total surface area, BET method.
HCl conversion at O2/HCl = 2, Tbed = 703 K, P = 1 bar, W/F0(HCl) = 11.2 g h molꢀ1
.
Oxygen storage capacity measured at 573 K.

290
A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
temperature operation is detrimental due to thermodynamic con-
straints that limit the equilibrium HCl conversion [48]. Neverthe-
less, these restrictions can be overcome by employing higher
pressures and inlet O2/HCl ratios, so that the degree of HCl conver-
sion achieved still makes the development of a technical process
feasible. Profiles of equilibrium HCl conversion at different total
pressures and O2/HCl ratios have been included in the ESI
(Fig. S5a). Our experiments and calculations (vide infra) conclude
that oxygen vacancies determine the activity of CeO2 in HCl oxida-
tion. More active ceria-based catalysts may be obtained enhancing
its oxygen storage capacity, e.g., by the addition of dopants. How-
ever, this aspect was beyond the scope of the present manuscript.
3.1.2. Kinetic analysis
The influence of the O2/HCl ratio on the activity was examined
over CeO2-1173-F at 703 K (Fig. 2). The HCl conversion increased
upon raising the relative O2 content in the feed mixture. The reac-
tion order on O2, calculated using a power equation fitting, was
0.5. This result suggests that oxygen-assisted chlorine evolution
(re-oxidation) is the rate-limiting step, hence favored at a higher
partial O2 pressure. This point is further discussed in Section 3.4.3.
Remarkably, the HCl conversion profiles derived from measure-
ments in which the feed O2 content was step-wise changed so to
vary the O2/HCl ratio from the lowest (0.5–7, Fig. 2 black symbols)
or from the highest (7–0.5, Fig. 2, grey symbols) value evidence very
little hysteresis (Fig. 2). This outcome indicates that the catalyst
reversibly responds to variations of the feed mixture and highlights
the dynamic character of the CeO2 surface. The HCl conversion in-
creased upon raising the space time for different feed O2/HCl ratios
(Fig. S6 in the ESI). On the contrary, the Cl2 production decreased
upon increasing the relative feed HCl content (Fig. S7 in the ESI),
thus suggesting a change in the catalyst composition at the
(near-)surface level (later addressed by characterization) that leads
to activity loss. With regard to the dependency of CeO2 activity on
the reaction temperature, the HCl conversion level varied from 1%
to 18% between 603 and 703 K at O2/HCl = 2. The apparent activa-
tion energy was estimated at ca. 90 kJ mol–1. In addition, a catalytic
test using a feed mixture with O2/HCl = 5 and an increased space
time (at the limit of our current setup) was conducted. Fig. S5b in
the ESI shows that a significant single-pass HCl conversion of 40%
was reached over CeO2 at 673 K and 1 bar.
Fig. 1. HCl conversion versus (a) surface area of fresh (open symbols) and used
(solid symbols) CeO2 samples and (b) OSC, measured at 573 K, of fresh CeO2
samples. Conditions: inlet mixture of 10 vol.% HCl and 20 vol.% O2 balanced in N2,
Tbed = 703 K, W/F0(HCl) = 11.2 g h molꢀ1, P = 1 bar, and time-on-stream = 3 h.
their OSC, according to a method reported elsewhere [34]. On the
basis of the values obtained (Table 2), the dependence of HCl con-
version on OSC followed a linear trend (Fig. 1b). For CeO2-773-F,
CeO2-1023-F, and CeO2-1173-F, i.e., samples showing the highest
(and similar) activity, the OSC was estimated at ca. 390 lg O
gcatꢀ1. For CeO2-1273-F and CeO2-1373-F, which exhibit decreasing
HCl conversion levels, diminishing OSC values were measured,
respectively, equal to 237 and 78 l
g O gcatꢀ1. Based on these results,
the OSC proved to be a suitable parameter to rationalize the activ-
ity differences. Nevertheless, an influence of SBET on OSC and/or a
combined role of OSC and SBET, when the latter is below
20 m2 gꢀ1, on the catalytic activity could be considered. In this re-
gard, Terribile et al. [47] showed that OSC, as measured for the
bulk, was independent of SBET. The outcome of this study might
not be directly applicable to our case, in which OSC has to be con-
sidered limitedly to the outermost surface and subsurface layers of
CeO2, where the catalytic process takes place (vide infra). More
generally, the OSC could be described as OSC = SBETꢁq, where
q cor-
responds to the density of surface or near-surface oxygen vacan-
cies. At higher temperatures, the Gibbs energy for vacancy
formation decreases (see DFT results), thus favoring an increased
generation of vacancies in the corresponding equilibrium condi-
tions, but SBET drops. This explains why a larger OSC, i.e., more ac-
tive CeO2, is found for samples calcined at 573–1173 K.
The activity of CeO2 is significantly lower than that of RuO2. A
40 K higher light-off temperature for Cl2 evolution was determined
over CeO2-1173 by temperature-programmed reaction according
to the protocol described in [39]. In view of a practical use, high-
Fig. 2. HCl conversion versus feed O2/HCl ratio at 703 K for CeO2-1173-F. Condi-
tions: inlet mixture of 10 vol.% HCl and 5–70 vol.% O2 balanced in N2, Tbed = 703 K,
W/F0(HCl) = 11.2 g h molꢀ1, and P = 1 bar.

A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
291
Fig. 3. HCl conversion versus time-on-stream over CeO2-1173 using sequential HCl-
rich/O2-rich feed mixtures. (a) HCl-rich step with O2/HCl = 0.25 followed by O2-rich
step with O2/HCl = 2. (b) HCl-rich step with O2/HCl = 0.25 followed by O2-rich step
with O2/HCl = 7. (c) HCl-rich step with O2/HCl = 0 followed by O2-rich step with O2/
HCl = 7. Other conditions: Tbed = 703 K, W/F0(HCl) = 11.2 g h molꢀ1, and P = 1 bar.
As operation in industry would be more economic lowering the
inlet partial pressure of O2 down to the stoichiometric amount (O2/
HCl = 0.25), we have investigated to which extent a stoichiometric
or excess amount of HCl can negatively impact the catalytic activ-
ity. Some effort was also made to explore whether activity recov-
ery is possible and to determine the re-oxidation kinetics. Fig. 3
displays the HCl conversion profiles obtained for CeO2-1173-F in
sequential HCl-rich/O2-rich experiments. The HCl-rich step was
conducted at 703 K using O2/HCl = 0 or 0.25 for 3 or 5 h, followed
by the O2-rich step using O2/HCl = 2 or 7 for 2 h. As shown in
Fig. 3a, a slight decrease in HCl conversion was observed when
the reaction was carried out with O2/HCl = 0.25. Upon increasing
the O2 content in the feed (O2/HCl = 2), a gradual increase in activ-
ity was obtained. The HCl conversion expected for an O2/HCl = 2
feed composition (ca. 17%, Fig. 2) was approached in 2 h (Fig. 3a).
In view of the slow re-oxidation kinetics, the experiment was re-
peated at a higher O2/HCl ratio for the O2-rich step. Using O2/
HCl = 7, the HCl conversion level immediately rose (Fig. 3b) to
the expected value (ca. 27%, Fig. 2). When repeating this two-step
experiment with the HCl-rich phase at O2/HCl = 0 and the O2-rich
at O2/HCl = 7, a similar profile was obtained (Fig. 3c). The data
show that upon treating the catalyst with HCl in the absence of
gas-phase O2, the HCl conversion was completely depleted due to
excessive chlorination (vide infra). Nevertheless, upon switching
to O2/HCl = 7, the original activity was fully restored within 1 h.
Fig. 4. XRD patterns of CeO2-1173 samples in fresh and used (at various O2/HCl
ratios) forms. The diffractogram of CeO2-773-0 is included for comparison. The
most intense reflections are due to CeO2 (JCPDS 73-6328), while those within boxes
belong to CeCl3ꢂ6H2O (JCPDS 01-0149).
chlorination was smaller than for O2-free runs. Samples treated in
O2/HCl = 0.75 or 2 did not suffer from detectable bulk chlorination.
Accordingly, the choice of O2/HCl feed ratio is of great importance
and should be equal or higher than 0.75 to avoid activity deterio-
ration of CeO2-based catalysts. The inhibiting effect of excess HCl
on the conversion (Fig. S7 in the ESI) can be related to bulk chlori-
nation. Deactivation does not appear to be induced by the volatil-
ization of the active phase in form of the metal chlorides produced
in situ, as in the case of copper-based catalysts [13,39]. Analysis of
the condensate at the reactor outlet by ICP-OES did not show
appreciable cerium loss. On the basis of the result of the catalytic
test on CeCl3 (HCl conversion = 2%, at 703 K and O2/HCl = 2), the
activity loss is assigned to the inactivity of cerium chloride for
HCl oxidation.
XRD analysis of the samples exposed to the HCl-rich/O2-rich
mixtures (Fig. S8 in the ESI) evidenced that activity recovery was
indeed induced by the removal of chlorine at higher partial pres-
sures of O2. In the case of the samples treated with O2/HCl = 7,
the original CeO2 phase was restored, while traces of CeCl3ꢂ6H2O
were still observed for the sample exposed to a lower oxygen ex-
cess (O2/HCl = 2). These evidences are in line with the activity pat-
tern described above.
Quantification of the chlorination degree was estimated by
TGA-MS studies of the used catalysts. The weight loss and MS pro-
files as a function of the temperature are reported in Fig. S9 in the
ESI. According to these data, the values for molar Cl/Ce ratio were
0.38, 0.30, and 0.08 for CeO2-773-0, CeO2-1173-0, and CeO2-1173-
0.25, respectively. These results support the trend qualitatively de-
rived by XRD.
In contrast to the extensive bulk alteration reported for CuO and
MnO2 already upon exposure to an O2-rich feed (O2/HCl = 2)
[39,49], chlorination of CeO2 occurs only under stoichiometric feed
conditions or in excess HCl and to a very limited extent. Conse-
quently, the behavior of CeO2 resembles more that of RuO2, and
this is directly related to the high stability of the latter two oxides
in the Deacon reaction.
3.2. Characterization
3.2.1. X-ray diffraction and thermogravimetry
In order to relate the activity loss in HCl-rich feeds to modifica-
tions of the catalyst structure and/or composition, CeO2-1173-F
samples treated at 703 K for 3 h using various O2/HCl feed ratios
were analyzed by XRD (Fig. 4). The diffractogram of CeO2-773-
HCl is also included for comparison purposes. For the samples
exposed to O2/HCl = 0 or 0.25, CeCl3ꢂ6H2O (JCPDS 01-0149) reflec-
tions were detected, thus pointing to bulk chlorination as the main
cause for the observed catalyst deactivation. From the intensity of
the cerium chloride reflections, the samples could be sequenced
according to the extent of bulk chlorination as follows: CeO2-
773-0 > CeO2-1173-0 > CeO2-1173-0.25. The fact that CeO2-1173-
0 was less altered than CeO2-773-0 is probably related to the better
stabilization achieved by high-temperature calcination, thus not
only rendering lower sintering upon use, but also a higher resis-
tance against chlorination. The presence of stoichiometric amounts
of O2 did not fully prevent structural alterations, but the degree of
3.2.2. High-resolution transmission electron microscopy
TEM micrographs of CeO2-773-F reveal a bimodal particle size
distribution (Fig. 5a). Small CeO2 particles of approximately

292
A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
Fig. 5. TEM of (a) CeO2-773-F, (b) CeO2-773-2, (c) CeO2-773-0, and HRTEM of (d) CeO2-773-2, revealing the surface of the CeO2 particles. (e) A CeO2-773-0 particle (surface
termination in the inset) and (f) a CeO2-773-0 particle showing an amorphous surface layer of CeCl3ꢂ6H2O. TEM of (g) CeO2-1173-F, (h) CeO2-1173-2, and (i) CeO2-1173-0.
Detailed imaging of (j) CeO2-1173-2, showing CeO2 particles with clean surface (inset) and of (k and l) CeO2-1173-0, exhibiting both the (k) CeO2 and (l) CeCl3ꢂ6H2O phases.

A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
293
Table 3
10 nm are scattered around larger particles reaching up to 300 nm.
Exposure to O2/HCl = 2 induced a significant change in particle size
(CeO2-773-2), which attained a range of 20–40 nm (Fig. 5b). Under
Deacon conditions, the smaller particles agglomerated and the big-
ger particles ‘‘cracked’’, thus not being detectable in the used sam-
ple. While sintering is expected to occur upon use, the latter
phenomenon is tentatively explained on the basis of a different
activity/stability of the crystal facets, where one (or more) might
be unstable under reaction conditions. It cannot be excluded that
the mechanical strain applied in the pressing step during prepara-
tion of the sieve fraction may be the origin of the disruption of the
bigger grains. Nevertheless, the significantly reduced SBET of CeO2-
773-2 implies that the loss in surface area determined by the
agglomeration of the small particles clearly surpasses the increase
due to the disruption of few big grains. High-resolution imaging of
the surface of CeO2 nanoparticles after exposure to Deacon condi-
tions shows clean surfaces exhibiting several steps (Fig. 5d). The
analysis of CeO2-773-0 reveals some clear changes due to the treat-
ment in HCl only. While the size of the nanoparticles does not
seem to differ significantly from CeO2-773-2 (Fig. 5c), the treat-
ment had a considerable effect on the particle structure. Besides
rounded nanoparticles of CeO2-773-0 (Fig. 5e), revealing numerous
‘‘clean’’ atomic steps on the surface (magnified in the inset of
Fig. 5e), nanoparticles covered with an amorphous layer are now
evidenced (Fig. 5f). SAED (selected area electron diffraction)
analysis of such particles suggests that the structure can fit to
CeCl3ꢂ6H2O, in agreement with the phase assigned by XRD
(Fig. 4). The high vacuum conditions of the TEM could result in
the removal of crystal water from the CeCl3ꢂ6H2O particles, leading
to a collapse of the structural order in the near-surface region. This
would explain the amorphous layer covering the particles. The
coexistence of clean CeO2 particles and particles made of
CeCl3ꢂ6H2O substantiates that certain crystallographic planes of
CeO2 are less prone to chlorination while others favor the forma-
tion of the chloride phase.
Quantification of the surface Cl uptake from XPS measurements, using different
models.
Catalyst
Cl/Ce
Number of layers occupied by Cl
stoichiometrya
Model 1
IMFPb
Model 2
EALc
Model 3
EAL, 75% Cld
CeO2-773-F
CeO2-773-2
CeO2-773-0
CeO2-1173-F
CeO2-1173-2
CeO2-1373-0
–
–
–
–
0.19
0.55
–
0.14
0.29
1.5
5.7
–
1.2
4.6
–
1.6
6.1
–
1.0
2.4
0.8
1.9
1.1
2.6
a
b
c
Based on homogeneous distribution of elements.
Inelastic mean free path (by TPP-2 M [50]): 22 Å.
Effective attenuation length, 0.8ꢁIMFP was used.
d
Additionally, to the EAL model, 75% Cl occupation of possible sites was con-
sidered, in line with the DFT calculations.
are not counted in XPS. This assumption is in line with DFT simu-
lations (Section 3.4.3). Based on the XPS data, it can be concluded
that during HCL oxidation, Cl atoms can occupy surface sites as
well as lattice oxygen vacancies. This is in contrast to the RuO2
case, where chlorination was confined to the surface layer
[9,10,51]. Since XRD after HCl treatment detected traces of
CeCl3ꢂ6H2O, the higher Cl uptake in the near-surface region for
CeO2-1173-0 and CeO2-773-0 can be explained by the presence
of the chloride phase. The Ce 3d difference spectrum (Fig. S10 in
the ESI) between 773-0 and 773-F confirms the presence of
CeCl3ꢂ6H2O. The calculation, indicating 8% of chloride phase (corre-
sponding to an increase in Cl/Ce by ca. 0.24), is in reasonable agree-
ment with the observed increase in Cl/Ce by 0.36, and the
difference might be due to additional subsurface Cl occupation.
In addition, Fig. S10 in the ESI suggests that calcination at 773 K
is not sufficient to produce fully oxidized cerium (CeO2), an extra
10% of Ce3+ state being estimated after this milder calcination.
The catalyst morphology within the complete ‘‘1173’’-series
(Fig. 5g–i) looks substantially similar, with the particle size ranging
between 20 and 40 nm in all the samples. The fresh catalyst does
not contain considerable amounts of bigger particles as in CeO2-
773-F. Thus, the particle size was stabilized already during calcina-
tion at the higher temperature, an effect that apparently occurred
for the CeO2-773-F sample only after exposure to Deacon condi-
tions. The only assigned phase in CeO2-1173-2 was CeO2 (Fig. 5j).
In CeO2-1173-0, both CeO2 (Fig. 5k) and CeCl3ꢂ6H2O (Fig. 5l) were
present. SAED analyses showed that CeO2 is the dominant struc-
ture in the CeO2-1173-0 sample with a minor CeCl3ꢂ6H2O contribu-
tion. This result also agrees with XRD (Fig. 4).
3.3. In situ Fourier transform infrared spectroscopy
In situ FTIR experiments were carried out to study the chlorina-
tion mechanism indirectly, that is, following the evolution of the
OH stretching bands upon introducing a diluted Deacon feed. A de-
tailed description based on our DFT analysis of different configura-
tions of OH/H2O-related species adsorbed on CeO2(111) and of the
corresponding vibrational frequencies is presented in the ESI
(Table S1). The CeO2-773-F sample was activated by heating in
O2/N2 up to 653 K and holding at this temperature till no spectral
changes were observed. The spectrum at the end of this stage
(Fig. 6, dark blue) shows two main bands at 3713 and
3632 cmꢀ1, which are, respectively, assigned to mono-coordinated
(OH-I) and bridging (OH-IIB) hydroxyls [52–54]. Another absorp-
tion band at 3684 cmꢀ1, due to water, as well as a band at
3660 cmꢀ1, attributed to a different bridging hydroxyl species
(OH-IIA), vanished during the activation process. The existence of
two types of bridged hydroxyl groups was first reported by Badri
et al. [52], who assigned the red-shifted band to bridged hydroxyl
groups next to oxygen vacancies created by water elimination.
According to our DFT calculations on CeO2(111), the mono-coordi-
nate OH band should appear at 3729 cmꢀ1, but threefold coordi-
nated OH species with surface O being part of the lattice can also
give rise to bands in the range of 3750–3732 cmꢀ1. Note that the
calculation tends to overestimate the stretching frequency by
15–30 cmꢀ1 when benchmarked against gas-phase water mole-
cules. Additionally, OH-II species seem to be unstable on the
(111) and (100) facets. On the other hand, bands related to water
with a neighboring hydroxyl group are expected to fall around
3.2.3. X-ray photoelectron spectroscopy
XPS was applied to selected catalysts to assess the degree of
surface chlorination. Table 3 compiles the experimentally deter-
mined Cl/Ce ratios. In general, CeO2-773-F was more prone to chlo-
rination than CeO2-1173-F. Moreover, exposure to HCl gave rise to
a more extended chlorination than the exposure to HCl + O2. Both
findings are in line with XRD analyses. To translate the Cl/Ce stoi-
chiometry numbers that assume homogeneous distribution of Cl in
the information depth into the number of near-surface layers con-
taining Cl, various models were applied (see details in the ESI).
Thus, a more realistic layered structure was constructed (Cl occu-
pation preference from top inwards) giving rise to the same signal
ratio as determined by XPS. As Table 3 indicates, Cl occupies
approximately 1–1.5 layers in the samples used under Deacon con-
ditions and 2.5 to 6 layers after HCl conditioning. Nevertheless, one
should remember that after exposure to the mixture, the samples
were quenched in N2; thus, it is reasonable to assume that the Cl
atoms adsorbed on external Ce atoms could desorb and hence

294
A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
sumption of OH groups, and thus through the displacement of
surface oxygen species of different coordination by Cl.
3.4. Molecular modeling
3.4.1. Vacancy formation, diffusion, and healing
The surface energy to cleave the (111) plane of CeO2(111) is
0.013 eV Åꢀ2. The inward relaxation of the external atoms is
0.090 and 0.097 Å for Ce and O atoms, respectively. The material
is known to be easily reducible and, thus, can accommodate oxy-
gen vacancies [55]. The formation of surface vacancies, thereafter
denoted as h, in 0.25 ML (monolayers) in the p(2 ꢁ 2) cell is endo-
thermic by 3.25 eV. This is a very small value when compared to
ionic materials [56] or to other reducible oxides like SrTiO3 [57].
The electronic structure of the surface vacancy corresponds to an
anti-ferromagnetic solution, where the two electrons left end up
in the next and next nearest surface Ce atoms, as described by
Ganduglia-Pirovano et al. [43]. The electronic structure, electron
localization, and formation energy are in good agreement with pre-
vious calculations with both PBE + U and HSE06 functionals [43].
The latter is supposed to be more accurate for strongly localized
f-electrons. However, the energy difference with other configura-
tions is rather small and dynamic re-ordering of the electronic
structure through polaronic effects is likely to occur [58,59].
Oxygen vacancies can be present in near-surface layers (Fig. 7c).
Indeed, the formation energy for subsurface defects turns out to be
lower than for surface defects, 2.93 eV, in agreement with previous
theoretical estimates [43]. The energy requirement for surface to
subsurface vacancy diffusion is small, 0.30 eV, in line with reported
values for bulk diffusion, 0.50 eV [60]. Vacancy agglomeration has
also been evidenced by scanning tunneling microscopy (STM)
investigations [28]. From our calculations on a p(4 ꢁ 2) supercell,
surface vacancy agglomeration at a constant concentration of
0.25 ML is weakly exothermic, by 0.13 eV, in agreement with
STM images [28].
As for the surface re-oxidation, oxygen adsorption does not take
place on a non-defective CeO2(111) surface, but occurs on surface
oxygen vacancies, h, leading to a superoxo-like species, in which
one of the two O atoms fills the hole, and the O–O distance is
1.453 Å. Dissociation of this bond heals the defect and leaves a sur-
face O atom, but it is uphill by 1.80 eV. Nevertheless, the dynamics
of oxygen vacancies in oxides can be enhanced by the presence of
gas-phase O2 [61]. Aggregation of oxygen defects at the surface or
near-surface is energetically favored. Indeed, surface vacancy diffu-
sion takes place by penetration to the subsurface layer and ejection
toward the surface in neighboring lattice position. The first step is
exothermic by 0.72 eV with a barrier of 0.24 eV, while the second is
mildly endothermic (0.59 eV) with a barrier of 0.64 eV. Once the
defect dimer is present on the surface, gas-phase oxygen molecules
can interact healing both vacancies and liberating about 6 eV.
Fig. 6. Evolution of the stretching OH band region of CeO2-773-F upon introducing
1 vol.% HCl into an O2/N2 (20/80) feed at 653 K. The dark blue curve represents the
starting state prior to HCl introduction, whereas the black spectrum shows the
stabilized state after HCl has been turned off (both in O2/N2). The temporal
evolution in the Deacon feed was as follows: light blue/brown/green/orange/red
corresponding to time-on-stream of 85/170/220/270/1055 s, respectively. (For
interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
3682–3646 cmꢀ1; thus, they may give rise to the low-frequency
absorption.
Two shoulders at 3732 cmꢀ1 and 3702 cmꢀ1 are visible on both
sides of the high-frequency OH band. Upon introducing HCl in the
O2/N2 feed, these shoulders became more resolved. According to
Table S1, they can be attributed to threefold coordinated hydroxyl
groups, as well as to mono-coordinated OH groups with different
neighboring species coexisting on the surface. During HCl oxida-
tion, significant changes were observed in the spectra, as the inten-
sity of the peak at 3632 cmꢀ1 decreased gradually till it was
completely depleted at later stages (Fig. 6, light blue till red).
According to the bands assignment in the OH stretching region sug-
gested by our DFT-calculated vibrational frequencies, this would
indicate that the initial H2O–OH pairs progressively disappeared,
likely due to the partial occupation of the vacancies by Cl. OH–Cl
pairs showed OH vibration at 3739 cmꢀ1, just in the middle of the
OH band of threefold and adsorbed OH. In parallel, the baseline in
the whole spectral region first increased (light blue/brown/green),
and, then declined to reach the level shown by the red spectrum.
This modulation of the baseline reflects the rapid evolution, desorp-
tion, and stabilization of the hydrogen-bonded water structure as
HCl dissociatively adsorbs on the ceria surface. The broad signal
at 3575 cmꢀ1 (red spectrum), developed when the hydrogen-
bonded water structure reached low enough concentration, can
be assigned to isolated adsorbed water species produced during
the reaction (in line with the theoretical value at 3591 cmꢀ1), since
it diminished after stopping the HCl flow (black line). To visualize
the changes occurring in the high-frequency OH band, the ‘‘HCl
off’’ curve was subtracted from all other spectra. This illustration
suggests that the changes in this band were not due to band shift,
but rather related to the gradual disappearance of one OH species
at 3717 cmꢀ1 in parallel with the depletion of the band at
3632 cmꢀ1. This experiment suggests that surface chlorination oc-
curs via dissociative HCl adsorption (baseline modulation) and con-
3.4.2. Chlorination
In order to assess the stability of CeO2(111), we have employed
first-principles thermodynamics [62] to consider the effect of par-
tial oxygen and chlorine pressures on the state of the surface. The
stability problem is considerable and was simplified by exclusively
considering a reactant (O2) and a product (Cl2), for which no cross-
ing between the associated chemical potentials appears in the
equations. The excess surface energy can be calculated from the
following equation:
D
GX ꢃ EX ꢀ ECeO2 þ NO2=2lO2 ꢀ NCl2=2lCl2
ð3Þ
where GX is the Gibbs energy associated with the configuration X
D
with respect to the energy of regularly terminated CeO2 ðECeO ) NO
2
2
and NCl are the number of oxygen and chlorine molecules, and
l is
2
the corresponding gas-phase chemical potential. The ability of HCl

A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
295
Fig. 7. Structural model of a p(2 ꢁ (2) supercell of CeO2(111): (a) regular surface, (b) surface vacancy, and (c) subsurface vacancy. Ce atoms are depicted in blue, surface O
atoms in red, and subsurface O atoms in scarlet and small balls.
to form chlorides is somehow higher than that of Cl2, but the pres-
ent model serves as a guide to evaluate the degree of chlorination of
the material. The Gibbs potentials of solids including 40 different X
configurations, EX, and of the regular surface, ECeO , were taken from
2
DFT calculations, and the chemical potentials of Cl2 and O2 were cal-
culated through statistical thermodynamics [63]. Fig. S11 in the ESI
shows that, at extremely low oxygen and chlorine chemical poten-
tials, the surface is regularly terminated. When increasing the Cl2
pressure, partial chlorination can take place by substitution of an
oxygen atom from the surface (Fig. S11 in the ESI, green configura-
tion). At relatively low oxygen pressures (for log(pO2) and log(pCl2)
in the range ꢀ1 to 2), the clean surface, mono-substituted Cl, and
multiple substituted Cl structures lie very close in energy and are
likely present on the catalyst surface. If the chlorine pressure is
much higher (log(pCl2) = 0), lattices with Cl in lattice oxygen posi-
tions become more stable. The maximum oxygen substitution
achievable in this way corresponds to 50% of the surface oxygen
atoms. If the oxygen pressure is raised up to the 0.1 bar regime,
as in our catalytic tests, then the most likely configuration corre-
sponds to the clean surface. Finally, the thermodynamic penalty
Fig. 8. Reaction energy profile for the Deacon process on CeO2(111). The initial
state in the profile is CeO2(111) with a surface oxygen vacancy (Fig. 7b). The color
to introduce a Cl atom occupying a vacancy site (Clh) to inner layers
is about 3 eV. Accordingly, it is implied that Cl preferentially stays
in the near-surface region, and that high HCl or Cl2 pressures might
account for lattice disruption and formation of CeCl3-related phases
(Section 3.2), as Cl cannot be accommodated in the bulk of CeO2.
codes in the bottom panels are as described in the caption of Fig. 7.
Clh atom toward the outer surface, i.e., converting Clh in Clꢄ
(Fig. 8, step D). The energy required for this elementary step is
2.15 eV and forms an oxygen vacancy at a subsurface position,
hss. Re-oxidation can take place by the complex diffusion–reaction
mechanism described in Section 3.4.1 to release nearly 3.4 eV
(Fig. 8, step E). Still, HCl can adsorb on this surface releasing
0.42 eV to form a OlatH and a Clꢄ (Fig. 8, step F). Three chlorine
atoms are then adsorbed on the surface, one Clꢄ and two Clh,
respectively. Cl2 evolution toward the gas phase takes place from
this structure; the energy required is 1.42 eV (Fig. 8, step G). As
it can be noticed in Fig. 8, states A and G correspond to the active
states of the catalyst. Therefore, the catalytic cycle runs between
steps A and G.
3.4.3. Reaction mechanism
The reaction profile of the HCl oxidation on CeO2 is presented in
Fig. 8. The main elementary steps derived for the Deacon process
comprise (i) hydrogen abstraction from HCl by basic surface O
atoms, to form hydroxyl groups and leave chlorine atoms on the
surface, (ii) reaction of the hydroxyl groups with new incoming
HCl molecules and/or hydroxyl group recombination on the sur-
face to form water, (iii) water removal, (iv) re-oxidation, and (v)
recombination of chlorine atoms.
The list of elementary steps in the mechanism can be summa-
rized by the following reactions:
The Deacon reaction starts by the adsorption of HCl near the ba-
sic centers on the surface (lattice oxygen atoms, Olat). The reaction
is energetically favored provided that a surface vacancy exists,
where the Cl atom can be accommodated (Fig. 8, step A). In this
process, HCl adsorption is exothermic by 2.84 eV and leads to a
surface OlatH group and a Cl atom at the surface vacancy. Care
should be taken as both Olat and adsorbed O atoms can give rise
to the OH-I and OH-II vibrational bands described in Section 3.3.
At the transition state, the Cl–H and Olat–H distances are 3.590
and 2.121 Å, respectively. A second HCl molecule can adsorb, form-
ing a water molecule and leaving a Cl atom on top of a Ce atom on
the surface, Clꢄ (Fig. 8, step B). This second process is also exother-
mic, but only by 0.29 eV. The external Clꢄ atom can push the al-
ready formed water molecule H2Olat out of the lattice and fill the
nascent vacancy, thus becoming Clh. This process is almost ther-
mo-neutral (Fig. 8, step C). Then, one of the oxygen atoms from
the subsurface layer can diffuse toward the surface, pushing a
HCl þ Olat þ ꢀ ! OlatH þ Clꢀ
HCl þ OlatH þ ꢄ ! H2Olat þ Clꢄ
Clꢄ þ H2Olat ! Clꢀ þ H2O
Clꢀ þ ꢄ ! ꢀ þ Clꢄ
ð4Þ
ð5Þ
ð6Þ
ð7Þ
ð8Þ
ð9Þ
ꢀ þ 1=2O2 ! Olat
Clꢄ þ Clꢄ ! Cl2 þ 2ꢄ
It is important to notice that, although the reaction scheme on CeO2
resembles that of RuO2, some important differences are found. First,
the reaction on RuO2 occurs at almost full coverage of the under-

296
A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
Table 4
to highly expensive RuO2-based catalysts for industrial chlorine
Energy (DE) in eV and Gibbs free energy (D
G0) in kJ mol–1 M (M = Ce, Ru) atom for the
recycling. The activity is related to the presence of oxygen vacan-
cies in the material. The stability arises from the remarkable resis-
tance of cerium oxide against chlorination. Limited bulk
chlorination, that is detection of CeCl3 by XRD, takes place under
HCl-rich conditions (O2/HCl 6 0.25). The bulk chloride phase rap-
idly and completely disappears when the catalyst is exposed to
O2-rich conditions. Under O2/HCl P 0.75, only the outermost sur-
face layers of CeO2 contain chlorine. Density functional theory sim-
ulations reveal that Cl activation from vacancy positions to surface
Ce atoms is the most energy-demanding step, although chorine–
oxygen competition for the available active sites may render re-
oxidation as the rate-determining step. Current studies focus on
developing a proper strategy to support the active phase on a suit-
able carrier preserving the catalytic proprerties of bulk CeO2.
complete chlorination of CeO2 and RuO2 by Cl2 or HCl.a
Chlorination equation
D
E
DG°
573 K
703 K
CeO2 + 3/2Cl2 ? CeCl3 + O2
CeO2 + 4HCl ? CeCl3 + 2H2O + 1/2Cl2
RuO2 + 3/2Cl2 ? RuCl3 + O2
15.82
92.95
60.61
178.98
146.65
112.18
97.95
198.22
183.98
ꢀ115.99
101.86
RuO2 + 4HCl ? RuCl3 + 2H2O + 1/2Cl2
ꢀ29.95
a
Calculated using the PBE (RuO2) and PBE + U (CeO2) functionals.
coordinated Rucus positions, as nearly all available sites are occu-
pied by Cl [9]. For CeO2, under-coordinated cerium atoms only exist
when oxygen vacancies are present in the surface or near-surface
regions. These are the only active sites for the reaction, in agree-
ment with the linearity found between activity and OSC (Fig. 1b).
As a consequence, the reaction profile for CeO2 is much more abrupt
(involves higher energy requirements) than that of RuO2. This cor-
relates with the higher temperatures needed to run the Deacon
reaction on CeO2 (vide supra). The most energy-demanding step
of HCl oxidation on RuO2 is related to the formation and evolution
of Cl2. Due to the high energy required for Cl2 elimination on RuO2,
Cl self-poisoning is observed, and thus, re-oxidation turns out to be
the rate-determining step under relevant conditions, as shown by
the positive dependence of the activity on the partial pressure of
O2 [51]. For CeO2, the activation of Cl atoms from lattice vacancies
to surface positions is the most energy-demanding step.
The similar activity enhancement observed at higher partial O2
pressures (Fig. 2) can be rationalized on the basis of the chlorine
and oxygen competition for the same active sites, which tightly
couples chlorine elimination to oxygen re-adsorption. When Cl
atoms occupy most of the active positions, very few active sites ex-
ist for re-oxidation, thus producing less Cl2. In extreme cases, this
leads to catalyst deactivation.
Regarding the stability against the harsh reaction conditions
employed, while the experimental data indicate that RuO2 and
CeO2 are quite similar, the DFT analysis points out some differ-
ences. The RuO2 surface is known to be partially chlorinated, but
this chlorination is confined to the under-coordinated O and Ru
positions at the external surface. Comparing surface Cl contents,
the situation is such that the amount of Cl on the surface of RuO2
is very large, as the under-coordinated positions in the lattice are
very prone to adsorb reactants (Cl in particular). This effect is less
evident on CeO2, given the relative inertness of Ce atoms on the
surface. Indeed, most of the Cl is sitting at oxygen vacancies on
the surface and not directly on top of the active Ce sites. In addi-
tion, the penetration of Cl atoms to deeper layers is hindered in
both RuO2 and CeO2 systems by more than 2 eV. This energy is
somewhat larger for CeO2 (3 eV). Nevertheless, owing to vacancy
diffusion, and thus oxygen supply to the surface, ceria will be more
prone to subsurface and bulk chlorination in pure HCl or sub-stoi-
chiometric Deacon feeds. In line with this, we have compared the
energy required for the bulk chlorination and lattice disruption
in both Ru and Ce cases (Table 4). The energy requirement for
the chlorination of CeO2 is smaller than the corresponding value
for RuO2, in agreement with the greater chlorination detected in
the CeO2 experiments in HCl-rich feeds.
Acknowledgment
We thank Bayer MaterialScience for permission to publish these
results.
Appendix A. Supplementary material
Electronic Supplementary Information (ESI): Experimental
details of the applied characterization techniques, additional re-
sults from characterization, catalytic testing, and DFT simulations.
The above material can be found, in the online version, at
doi:10.1016/j.jcat.2011.11.016.
References
[1] J. Pérez-Ramírez, C. Mondelli, T. Schmidt, O.F.-K. Schlüter, A. Wolf, L. Mleczko,
T. Dreier, Energy Environ. Sci. 4 (2011) 4786.
[2] K. Seki, Catal. Surv. Asia 14 (2010) 168.
[3] H. Deacon, US Patent 85,370, Assigned to Gaskell, Deacon and Co., 1868.
[4] A.J. Johnson, A.J. Cherniavsky, US Patent 2542,961, Assigned to Shell
Development Company, 1951.
[5] T. Kiyoura, N. Fujimoto, M. Ajioka, T. Suzuki, Y. Kogure, K. Kanaya, T. Nagayama,
EP184413-A, Assigned to Mitsui Toatsu Chemicals, 1984.
[6] T. Hibi, H. Nishida, H. Abekawa, US Patent 5871,707, Assigned to Sumitomo
Chemical Company, 1999.
[7] C. Mondelli, A.P. Amrute, F. Krumeich, T. Schmidt, J. Pérez-Ramírez,
ChemCatChem 3 (2011) 657.
[8] A. Wolf, L. Mleczko, O.F. Schlüter, S. Schubert, EP2026905, Assigned to Bayer
MaterialScience, 2006.
[9] M.A.G. Hevia, A.P. Amrute, T. Schmidt, J. Pérez-Ramírez, J. Catal. 276 (2010)
141.
[10] D. Crihan, M. Knapp, S. Zweidinger, E. Lundgren, C.J. Weststrate, J.N. Andersen,
A.P. Seitsonen, H. Over, Angew. Chem. Int. Ed. 47 (2008) 2131.
[11] J.P. Hofmann, S. Zweidinger, A.P. Seitsonen, A. Farkas, M. Knapp, O. Balmes, E.
Lundgren, J.N. Andersen, H. Over, Phys. Chem. Chem. Phys. 12 (2010) 15358.
[12] Retrieved from <http://www.platinum.matthey.com/pgm-prices>.
[13] C. Mondelli, A.P. Amrute, T. Schmidt, J. Pérez-Ramírez, Chem. Commun. 47
(2011) 7173.
[14] R.W.G. Wyckoff, Crystal Structures, Interscience, New York, 1965.
[15] R.J. Gorte, AIChE J. 56 (2010) 1126.
[16] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.
[17] A. Primo, T. Marino, A. Corma, R. Molinari, H. Garcia, J. Am. Chem. Soc. 133
(2011) 6930.
[18] Y.S. Chaudhary, S. Panigrahi, S. Nayak, B. Sarpati, S. Bhattacharjee, N. Kulkarni,
J. Mater. Chem. 20 (2010) 2381.
[19] M. Mogensen, N.M. Sammes, G.A. Tompsett, Solid State Ion. 129 (2000) 63.
[20] B. Zhu, M.D. Mat, Int. J. Electrochem. Sci. 1 (2006) 383.
[21] F. Zhou, X. Zhao, H. Xu, C. Yuan, J. Phys. Chem. C 111 (2007) 1651.
[22] A. Trovarelli, Catal. Rev.–Sci. Eng. 38 (1996) 439.
[23] A. Gómez-Cortés, Y. Márquez, J. Arenas-Alatorre, G. Díaz, Catal. Today 133–135
(2008) 743.
[24] A. Martínez-Arias, A.B. Hungría, G. Munuera, D. Gamarra, Appl. Catal. B 65
(2006) 207.
4. Conclusions
[25] W. Shin, C. Jung, J. Han, S. Nam, T. Lim, S. Hong, H. Lee, J. Ind. Eng. Chem. 10
(2004) 302.
[26] T.X.T. Sayle, S.C. Parker, C.R.A. Catlow, Surf. Sci. 316 (1994) 329.
[27] H.C. Yao, Y.F.Y. Yao, J. Catal. 86 (1984) 254.
[28] F. Esch, S. Fabris, L. Zhou, T. Montini, C. Africh, P. Fornasiero, G. Comelli, R.
Rosei, Science 309 (2005) 752.
In this study, we present a fundamental understanding of HCl
oxidation on bulk CeO2 combining catalyst testing, steady-state
kinetics, characterization, and DFT simulations. Due to its remark-
able activity and stability, CeO2 constitutes a promising alternative

A.P. Amrute et al. / Journal of Catalysis 286 (2012) 287–297
297
[29] N. Laosiripojana, W. Sutthisripok, P. Kim-Lohsoontorn, S. Assabumrungrat, Int.
J. Hydrogen Energy 35 (2010) 6747.
[30] R. Di Monte, J. Kašpar, Top. Catal. 28 (2004) 47.
[48] M.W.M. Hisham, S.W. Benson, J. Phys. Chem. 99 (1995) 6194.
[49] A.P. Amrute, C. Mondelli, M.A.G. Hevia, J. Pérez-Ramírez, ACS Catal. 1 (2011)
583.
[31] A. Hagemeyer, P. Trübenbach, C.W. Rieker, M. Wünsch, O. Watzernberger, EP
Patent 0761594-A1, Assigned to BASF Aktiengesellschaft, 1997.
[32] A. Wolf, L. Mleczko, O.F. Schlüter, S. Schubert, WO2010133313-A1, Assigned to
Bayer Material Science, 2010.
[33] G. Lee, S.W. Lee, I. Sohn, Y.C. Kwon, J. Song, C.-S. Son, WO2009035234-A2,
Assigned to Hanwha Chemical Corporation, 2009.
[50] S. Tanuma, C.J. Powell, D.R. Penn, Surf. Interface Anal. 21 (1994) 165.
[51] D. Teschner, R. Farra, L.-D. Yao, R. Schlogl, H. Soerijanto, R. Schomaecker,
T. Schmidt, L. Szentmiklósi, A.P. Amrute, C. Mondelli, J. Pérez-Ramírez,
G. Novell-Leruth, N. Lopez, J. Catal. 285 (2012) 273.
[52] A. Badri, C. Binet, J. Lavalley, Faraday Trans. 92 (1996) 4669.
[53] A. Laachir, V. Perrichon, A. Badri, J. Lamotte, E. Catherine, J.C. Lavalley, J. El
Fallah, L. Hilaire, F. Le Normand, E.E. Quéméré, G.N. Sauvion, O. Touret, Faraday
Trans. 87 (1991) 1601.
[34] G. Dutta, U.V. Waghmare, T. Baidya, M.S. Hegde, K.R. Priolkar, P.R. Sarode,
Chem. Mater. 18 (2006) 3249.
[35] G.R. Rao, B.G. Mishra, Bull. Catal. Soc. India 2 (2003) 122.
[36] H.A. Al-Madfaa, M.M. Khader, Mater. Chem. Phys. 86 (2004) 180.
[37] B.M. Reddy, P. Bharali, P. Saikia, J. Phys. Chem. C 112 (2008) 11729.
[38] F. Deganello, A. Martorana, J. Solid State Chem. 163 (2002) 527.
[39] A.P. Amrute, C. Mondelli, M.A.G. Hevia, J. Pérez-Ramírez, J. Phys. Chem. C 115
(2011) 1056.
[54] C. Binet, M. Daturi, J. Lavalley, Catal. Today 50 (1999) 207.
[55] T.B. Lindemer, J. Brynestad, J. Am. Ceram. Soc. 69 (1986) 867.
[56] J. Carrasco, N. Lopez, F. Illas, Phys. Rev. Lett. 93 (2004) 225502.
[57] J. Carrasco, F. Illas, N. Lopez, E.A. Kotomin, Y.F. Zhukovskii, R.A. Evarestov, Y.A.
Mastrikov, S. Piskunov, J. Maier, Phys. Rev. B 73 (2006) 064106.
[58] P.M. Kowalski, M.F. Camellone, N.N. Nair, B. Meyer, D. Marx, Phys. Rev. Lett.
105 (2010) 146405.
[40] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
[41] J.-P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[42] M. Cococcioni, S. Gironcoli, Phys. Rev. B 71 (2005) 035105.
[43] M.V. Ganduglia-Pirovano, J.L.F. Da Silva, J. Sauer, Phys. Rev. Lett. 102 (2009)
026101 (references therein).
[59] S. Chretien, H. Metiu, J. Phys. Chem. C 115 (2011) 4696.
[60] P.P. Dholabhai, J.B. Adams, P. Crozier, R. Sharma, J. Chem. Phys. 132 (2010)
094104.
[61] N. Lopez, J.D. Prades, F. Hernandez-Ramirez, J.R. Morante, J. Pan, S. Mathur,
Phys. Chem. Chem. Phys. 12 (2010) 2401.
[44] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953.
[45] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
[46] G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901.
[47] D. Terribile, A. Trovarelli, C. de Leitenburg, G. Dolcetti, J. Llorca, Chem. Mater. 9
(1997) 2676.
[62] X.G. Wang, A. Chaka, M. Scheffler, Phys. Rev. Lett. 84 (2000) 3650.
[63] D.A. McQuarrie, J.D. Simon, Molecular Thermodynamics, University Science
Books, Sausalito, California, 1999.