Journal of Catalysis 255 (2008) 29–39

www.elsevier.com/locate/jcat

Mechanism of HCl oxidation (Deacon process) over RuO2

Núria López a,∗ , Jordi Gómez-Segura a , Raimon P. Marín a , Javier Pérez-Ramírez a,b,∗
a Institute of Chemical Research of Catalonia (ICIQ), Avinguda Països Catalans 16, 43007 Tarragona, Spain
b Catalan Institution for Research and Advanced Studies (ICREA), Passeig Lluís Companys 23, 08010 Barcelona, Spain

Received 20 August 2007; revised 16 January 2008; accepted 21 January 2008

Abstract
The catalytic oxidation of hydrogen chloride to chlorine (Deacon reaction) is an attractive means to recover Cl2 from HCl-containing waste
streams in the chemical industry. Recently, RuO2 supported on TiO2 rutile has been industrially implemented for large-scale chlorine manufacture.
However, detailed understanding of the reaction mechanism over claimed catalysts has not been acquired. Our tests in a fixed-bed reactor at
ambient pressure concluded that RuO2 powder is highly active for Cl2 production. Characterization of the fresh and used RuO2 samples by ex
situ XRD and XPS revealed no appreciable alteration of the bulk structure and limited chlorination of the catalyst. Ab initio thermodynamics
predicted that the initial state of the RuO2 (110) surface is partially over-oxidized, while the surface after reaction contains both oxygen and
chlorine. DFT described the Deacon reaction with a Mars–van Krevelen type mechanism consisting of five steps: hydrogen abstraction from
HCl, recombination of atomic chlorine, hydroxyl recombination, water desorption, and dissociative oxygen adsorption. An increased chlorine
production was observed experimentally when increasing the feed O2 -to-HCl ratio. This can be ascribed both to the lower recombination energy
of chlorine atoms to gas-phase Cl2 at high coverages and the faster surface reoxidation due to higher partial oxygen pressures.
© 2008 Elsevier Inc. All rights reserved.

Keywords: HCl oxidation; Chlorine; Deacon reaction; RuO2 ; Mechanism; Density Functional Theory; On-line Cl2 analysis; Ab initio thermodynamics

1. Introduction the main Cl2 production method, salt electrolysis, is energy in-
tensive and encompasses the disadvantage of producing caustic
Elemental chlorine (Cl2 ) is widely used in the production soda. Thus, large-scale manufacturers aim at the development
of a wide range of industrial and consumer products. Approxi- of a chlorine recycling technology due to environmental and
mately 50% of the chlorine produced worldwide (ca. 50 Mton economic reasons.
per annum) is reduced through its use to HCl or chloride An attractive route for chlorine recovery is the gas-phase
salts [1]. An example of particular relevance is the phosgenation catalytic oxidation of HCl with air or oxygen (Eq. (2)), the
of diamines to diisocyanates (Eq. (1)) as the first step to manu- so-called Deacon reaction [2–4], due to the relative ease of ap-
facture polyurethane resins and rubbers. This reaction leads to plication and low electrical power and thermal requirements
4 mole of HCl per mole of diisocyanate. when compared with electrolytic processes. The reaction is re-
R(NH2 )2 + COCl2 → R(NCO)2 + 4HCl. (1) versible and exothermic (H 0 = −28.4 kJ mol−1 HCl):

The by-product HCl can be generally marketed as hydrochlo- 4HCl + O2 ↔ 2Cl2 + 2H2 O. (2)
ric acid or used as a raw material in vinyl chloride monomer From the various systems protected by Henry Deacon in the
(VCM) production. However, as the VCM production growth late 19th century, CuCl2 supported on pumice was the pre-
is slow, excess supply of HCl is foreseen. In these cases, this ferred one. However, catalytic tests in a fixed-bed reactor at
highly toxic and corrosive acid must be neutralized. Besides, 703–773 K evidenced serious drawbacks for implementation:
low single-pass HCl conversion due to limited activity, rapid de-
* Corresponding authors. Fax: +34 977 920 224. activation due to volatilization of copper chloride above 673 K,
E-mail addresses: [email protected] (N. López), [email protected] and corrosion problems due to the presence of unreacted HCl
(J. Pérez-Ramírez). and product H2 O [4]. Many efforts have been undertaken since
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.01.020
30 N. López et al. / Journal of Catalysis 255 (2008) 29–39

then for developing more efficient systems, culminating with Herein we have used experiments and Density Functional
the Shell-Chlor process (Shell) in the 1960s and the MT-Chlor Theory (DFT) to study the mechanism of HCl oxidation to Cl2
process (Mitsui Toatsu Chemicals) in the late 1980s. The Shell- over RuO2 surfaces. Previous to the simulations, evidence that
Chlor process made use of a single fluidized-bed reactor with polycrystalline RuO2 powder is active in the Deacon reaction
a CuCl2 –KCl/SiO2 catalyst [5,6], but it was abandoned due has been obtained at different feed HCl:O2 ratios by means of
to limited HCl conversion and severe corrosion problems. The catalytic tests in a continuous fixed-bed reactor at ambient pres-
MT-Chlor process utilizes a Cr2 O3 /SiO2 catalyst in a fluidized- sure. A reaction mechanism comprising five fundamental steps
bed reactor at 623–673 K [7–9]. Presently, Mitsui Toatsu is has been identified and characterized. In particular, the effect of
operating a 50 kton per annum plant based on the MT-Chlor surface coverage and the nature of the active oxygen species on
process at Omuta, Japan. In the late 1990s, Totsis et al. [10,11] RuO2 have been analyzed. The fresh and used RuO2 samples
developed a dual fluidized-bed reactor system using copper- have been characterized in order to assess eventual changes in
based catalysts in order to achieve high HCl conversion to Cl2 the bulk and surface of the catalyst upon reaction.
with minimal corrosion. In the same line as Totsis et al., Nieken
and Watzenberger [12] demonstrated that periodic operation of 2. Experimental
the Deacon process in a two-step fixed-bed reactor can over-
come corrosion problems. 2.1. Catalyst characterization
A wider scope for industrialization of the one-step hy-
drochloric acid oxidation process has been realized through Ruthenium(IV) oxide, RuO2 , was purchased from Sigma-
the development by Sumitomo Chemicals of a process using Aldrich. Prior to characterization and catalytic tests, the as-
a RuO2 supported on TiO2 rutile catalyst in a fixed-bed re- received sample was heated in static air from 298 to 773 K
actor configuration [13]. Not surprisingly, RuO2 –TiO2 coated at 5 K min−1 and kept isothermal for 5 h. Powder X-ray dif-
titanium anodes are industrially used for chlorine evolution in fraction (XRD) was measured in transmission using a Bruker
NaCl electrolytic cells [14,15]. The RuO2 /TiO2 catalyst ex- AXS D8 Advance diffractometer equipped with a Cu tube,
a Ge(111) incident beam monochromator (λ = 0.1541 nm), and
hibits high activity at low temperature and remarkable stabil-
a Vantec-1 PSD. Data were recorded in the 2θ range of 5 to
ity, leading to Cl2 with higher purity than that obtained by
70◦ with an angular step size of 0.016◦ and a counting time of
NaCl electrolysis. Besides, the chlorine manufacturing cost is
2.4 s per step. X-ray photoelectron spectroscopy (XPS) analy-
claimed to be much lower than that of electrolysis and the
ses were obtained using a PHI 5500 Multitechnique System
MT-Chlor process due to the low electric power consumption
from Physical Electronics equipped with an ultrahigh vacuum
and efficient heat recovery from the reaction. The Sumitomo
chamber (pressure between 5 × 10−9 and 2 × 10−8 Torr) and
process can produce up to 400 kton of chlorine per annum in a
monochromatic AlKα radiation at 1486.6 eV as the X-ray
single reactor [16].
source. Nitrogen adsorption–desorption isotherms at 77 K were
The vast number of patents claiming Deacon-type catalysts
measured on a Quantachrome Autosorb 1MP analyzer. Before
and reactors contrasts with the scarce number of fundamental
analysis, the sample was degassed in vacuum at 393 K for 12 h.
studies dealing with mechanistic and kinetic aspects of the re- The BET method [33] was applied to determine the total sur-
action. Hisham and Benson [17] studied the basic thermochem- face area.
istry of the Deacon reaction over a number of metal oxides. The
process was examined in two steps: (1) HCl absorption by the 2.2. Activity tests
metal oxide to form the metal (oxy)chloride and (2) the oxida-
tion of the latter by O2 to regenerate the metal oxide and free The catalytic oxidation of HCl over RuO2 was studied in a
Cl2 . CuO was concluded as the only system fulfilling the re- quartz fixed-bed reactor (8 mm i.d.) using 600 mg of sample
quirements of a complete cycle below 700 K, although RuO2 (sieve fraction 200–300 µm), a total pressure of 1 bar, and a
was not included in the study. Exceptionally, Aglulin [18] in- weight-hourly space velocity (WHSV) was 16,600 ml g−1 h−1 .
vestigated the kinetics of HCl oxidation over supported Cr2 O3 The feed mixture contained 20 vol% HCl, 20–80 vol% O2 , and
catalysts in the presence of methane, although some of the steps N2 as the balance gas. Hydrogen chloride (Praxair, purity 3.0
in the hypothesized mechanism cannot be considered as ele- and water content <50 ppm), oxygen (Carburos Metálicos, pu-
mentary. rity 3.5), and nitrogen (Carburos Metálicos, purity 3.5) were
Consequently, deeper mechanistic studies of the Deacon re- used without further purification. The individual gases were
action over ruthenium oxide are required in order to better introduced to the reactor by means of digital mass-flow con-
understand how the industrial catalyst works. This informa- trollers. The material of all the lines in the set-up was Teflon®
tion can also be beneficial for further catalyst optimization. in order to prevent corrosion problems, particularly downstream
RuO2 has received considerable attention over the past years of the reactor. Two protocols were applied for catalytic eval-
as model system for surface science and computational stud- uation. Temperature-programmed reaction (TPR) was carried
ies, and also as heterogeneous catalyst in oxidation of various out by ramping the temperature of the reactor block (Tfurnace )
substrates, among others carbon monoxide [19–23], hydrogen from 333 to 583 K at 10 K min−1 in the reaction mixture. The
[24–26], ammonia [27], methane [28], alcohols [21,29–31], and catalyst was pretreated in a 40 vol% O2 in N2 at 333 K for
soot [32]. 20 min prior to HCl admission. Isothermal tests were conducted
 N. López et al. / Journal of Catalysis 255 (2008) 29–39 31

Fig. 1. Ball and stick model of different RuO2 surfaces. In the upper panel, oxygen and ruthenium atoms are represented by red and grey spheres, respectively. In
the lower panel the notation for the RuO2 (110) surface is presented in a p(2 × 1) supercell. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

at Tfurnace = 573 K using molar HCl:O2 ratios in the feed of c = 3.140 Å and u = 0.305) are in good agreement with the ex-
1:1, 1:2, 1:3, and 1:4. The temperature was ramped from 333 perimentally determined values using single-crystal X-ray dif-
to 573 K at 10 K min−1 in 20 vol% O2 in N2 and HCl was fraction [39]. The good accuracy of a similar approach when
subsequently introduced in the feed. The HCl:O2 ratio was var- compared to all electron calculations was tested in Ref. [40].
ied by increasing the inlet O2 concentration and the inlet HCl In a first step, the low-index (110), (101), (001), and (100)
concentration was fixed at 20 vol%. The flow of N2 was ad- surfaces were modeled (Fig. 1). The slabs contained at least
justed in order to keep a constant total flow (10 l STP h−1 ). 5 layers of RuO2 and the vacuum space between the different
During the tests, the catalyst bed temperature was continuously slabs was set to >10 Å, with a k-point sampling density of at
recorded with a Fluke Digital Multimeter (model 187) equipped least 0.4 Å−1 . The outermost 2 layers in the slabs were allowed
with a K-type thermocouple. to relax. The surface energies, calculated for the stoichiometric
Chlorine analysis was carried out using an Ocean Optics termination of each of the facets, were in agreement with pre-
miniature fiber optic spectrophotometer (model USB2000-UV- vious estimates [41], leading to (110) as the most stable surface
VIS). Spectra were collected in the broad wavelength range of with γ = 0.90 J m−2 .
200–1000 nm every 10 s using a DT-MINI-2-GS Deuterium The clean RuO2 (110) structure consists on alternating rows
Tungsten Halogen Light Source and a high-sensitivity Sony of protruding oxygen atoms, usually denoted as Ob (bridge oxy-
ILX511 2048-element linear silicon CCD-array detector up- gen) and rows of coordinatively unsaturated ruthenium atoms,
graded for working in the UV spectral region. On-line Cl2 Rucus , along the [001] direction. For mechanistic studies, the
detection via optical absorbance measurements was achieved (110) surface of RuO2 has been represented by a 3-layer slab
by the continuous flow of product gases through a Z-flow cell separated by 15.7 Å, where each of the layers contains an
adopted as flow injection analysis (FIA) type assembly having ORuO2 O sub-unit. Relaxation was only allowed in the two
a 10-mm optical path length (FIAlab instruments). Water pro- first layers of the slab keeping the lower layer fixed. In order
duced in the reaction was condensed using a reflux unit prior to avoid spurious electrostatic interactions, the dipole arising
to the detection cell. The carrier gas (N2 ) provided a reference from the formation of an asymmetric surface was eliminated
spectrum in optical absorbance processing. from the vacuum. To study adsorption and reaction processes
the supercell employed was a p(2 × 1) reconstruction, lead-
3. Computational details ing to a coverage by H, Cl, or Ocus of 0.50 ML. Tests with a
larger supercell p(4 × 1) were performed to study the atomic
Density Functional Theory (DFT) was applied to study the chlorine recombination to gas-phase Cl2 and the O2 adsorption.
mechanism of HCl oxidation to Cl2 on RuO2 . Calculations A Monkhorst–Pack [42] mesh was employed to perform the nu-
have been performed with the VASP code [34,35], employing merical integration; the meshes contain 4 × 4 × 1 points for the
the RPBE functional [36], the inner electrons have been rep- small supercell and reduced to 2 × 4 × 1 for the larger cell. The
resented by PAW pseudopotentials [37], and the valence mono- barriers of relevant elementary steps have been calculated using
electronic states have been expanded in plane-waves with a cut- the CI-NEB [43].
off energy of 400 eV. RuO2 has a rutile structure [38] and is con- RuO2 is a partially covalent oxide and thus formal charges
sequently isostructural to the most stable TiO2 polymorph. The on Ru and O differ from the real charges of the atoms. RuO2
unit cell parameters obtained with our set up (a = b = 4.569 Å, is thus reducible via O2 elimination but also the surface can be
32 N. López et al. / Journal of Catalysis 255 (2008) 29–39

Fig. 3. Absorbance due to Cl2 production in isothermal tests over RuO2 at

different molar HCl:O2 ratios. Conditions: feed mixture of 20 vol% HCl and
Fig. 2. Absorbance due to Cl2 production during temperature-programmed re-
20–80 vol% O2 in N2 , Tfurnace = 573 K, GHSV = 16,600 h−1 , P = 1 bar.
action of HCl with O2 over RuO2 . Conditions: feed mixture of 20 vol% HCl
and 40 vol% O2 in N2 , GHSV = 16,600 h−1 , P = 1 bar. Inset: UV–vis spec-
trum at a particular temperature, showing the Cl2 band centered at 330 nm.
feed HCl:O2 ratio of 1:2 and atmospheric pressure. The onset
temperature is around 500 K and the absorbance progressively
further oxidized in the Rucus centers. Under relevant pressures, increases until reaching a plateau around 625 K with a value of
the surface Rucus centers on RuO2 (110) can be partially cov- absorbance of 0.35. Acquisition data for absorption spectrum
ered [44]. We have employed ab initio thermodynamics [45,46] using the combined-spectrum source ranging between 200–
to obtain the most likely composition of the RuO2 (110) surface 410 nm (deuterium lamp) and 360–1000 nm (tungsten halogen
at 573 K in the presence of O2 and Cl2 . The thermodynamic lamp) is unambiguously indicative of the selective detection of
parameters for gas-phase oxygen or chlorine were taken from pure chlorine in the UV spectral region, exhibiting the maxi-
the NIST reference tables [47]. In the initial state, only O2 is mum absorbance at 330 nm (inset in Fig. 2) [48–50]. The satu-
in contact with the surface, thus several Ob and Ocus cover- ration of the detector was determined at 1.1 units of absorbance,
ages were calculated in a p(4 × 1) supercell and represented below which a linear relationship between the absorbance and
as a function of the oxygen pressure. To represent the state of the Cl2 concentration was observed.
the surface after reaction, a p(2 × 1) supercell was employed The influence of the HCl:O2 ratio on the Deacon activity of
where both Ob and Rucus positions can be occupied by either RuO2 was investigated at Tfurnace = 573 K. As shown in Fig. 3,
O or Cl. In fact, Ob substitution by Cl is energetically possi- the absorbance increases upon increasing the relative amount
ble, the reaction being exothermic by −0.15 eV. Moreover, the of oxygen in the feed mixture. Interruption of O2 causes a
Clb containing structure constitutes an energy minimum that rapid drop of absorbance due to the diminished Cl2 produc-
can be characterized by the lack of imaginary vibrational fre- tion. The beneficial effect of oxygen on the Cl2 production is
quencies. The electronic structure of Ru–Clb –Ru can be seen further discussed in Section 4.3 attending to simulation results
as a three center-three electron bond with partially oxidized Ru at different oxygen coverages. The single-pass HCl conversion
atoms although with a different atomic charge than that of the in the isothermal experiment at HCl:O2 = 1:2 was quantified
clean RuO2 . This is possible due to the redox character of Ru at ca. 70% by classical titration. To this end, the Cl2 concentra-
that can accept formal charges of +2, +3, and +4. In contrast, tion of the outlet gas was determined after passing through two
double Cl substitution (i.e. 2Cl in the position of Ob ) is an en- serial impingers over 10 min, each of which equipped with a
dothermic process (by 2 eV) and thus not likely. This is similar porous frit immersed into an aqueous solution of potassium io-
to what has been shown for the phase diagram of RuO2 in the dide (20–30 wt%). The iodine thus obtained was titrated with
presence of CO and O, where CO and O can occupy both bridge sodium thiosulfate (0.1 M).
and cus positions [44]. The large number of configurations aris- Consequently, the high activity of RuO2 -based catalysts in
ing (21 for the p(2 × 1) supercell) when compared to the case the Deacon process claimed by Sumitomo [13] can be con-
of oxygen (5 for p(4 × 1)) prevented us from using a more re- firmed on the basis of the low onset temperature measured
alistic supercell. in temperature-programmed reaction experiments and the rela-
tively high HCl conversion determined in steady-state tests over
4. Results and discussion unsupported ruthenium oxide.

4.1. Deacon performance of RuO2 powder 4.2. Characterization of fresh and used RuO2

Fig. 2 shows the absorbance profile due to Cl2 production The XRD pattern of the fresh RuO2 in Fig. 4 shows the char-
vs the temperature of the catalyst bed during a temperature- acteristic reflections of ruthenium(IV) oxide (JCPDS 40-1290),
programmed reaction experiment over RuO2 powder at a molar with (110) and (101) as the prevailing primary orientations.
 N. López et al. / Journal of Catalysis 255 (2008) 29–39 33

Fig. 5. Ab initio thermodynamics diagram of the surface composition of

Fig. 4. X-ray diffraction patterns of the fresh and used RuO2 samples. Inset: RuO2 (110) as a function of the O2 pressure at 573 K.
Wulff construction of the powder with the calculated surface energies for the
facets in Fig. 1. nium oxychloride such as RuOCl2 was observed after reaction.
This indicates that the bulk of the catalyst was not altered by
We have performed calculations on the (110), (100), (101), and the Deacon test. The BET surface area of the fresh catalyst
(001) facets of RuO2 (see Fig. 1). The (110) facet is the lowest (10 m2 g−1 ) decreased to 2 m2 g−1 in the used one, suggesting
in energy and thus found in the largest extension when em- some degree of sintering of the RuO2 particles. X-ray pho-
ploying the Wulff construction [51] to build the structure of toelectron spectroscopy was conducted in order to investigate
a nanoparticle (inset of Fig. 4). Moreover, in the TiO2 rutile changes in the nature of the ruthenium species and to deter-
used as the support for RuO2 in the Sumitomo catalyst [13,16], mine the extent of chlorination of the surface. The core level
the (110) surface is the most stable too [52]. Accordingly, fur- peaks of Ru 3d, O 1s, and Cl 2p are illustrated in Fig. 6 for the
ther DFT mechanistic studies concentrated on the RuO2 (110) used RuO2 catalyst, i.e. after Deacon reaction. The binding en-
surface. The choice of the (110) facet for calculations and ergy of ruthenium and oxygen (Ru 3d5/2 at 281 eV, Ru 3d3/2 at
the polycrystalline sample for catalytic tests can be justified. 285 eV, and O 1s at 529 eV) is characteristic of polycrystalline
Temperature-programmed reaction studies over RuO2 (110) in RuO2 [21] and was identical in the fresh and used samples. This
UHV and polycrystalline RuO2 at high-pressure conditions are suggests that the Deacon reaction does not induce significant
virtually identical [21], indicating that the reaction pathways changes in the oxidation state of surface ruthenium sites in the
are very much alike and therefore the active centers on RuO2 tested period. The distinctive feature of the fresh and used RuO2
are the same. In addition, it is also known that the efficiency specimens was the presence of chlorine in the latter sample, as
for CO oxidation hardly depends on the orientation of RuO2 shown in Fig. 6 by the Cl 2p3/2 (198 eV) and Cl 2p1/2 (199.5 eV)
[53], at least not when the (110), (100), and (101) facets are peaks. This demonstrates a certain degree of surface chlorina-
compared, since all three low-energy surfaces contain five-fold tion caused by HCl oxidation, although we cannot conclusively
coordinated ruthenium atoms, Rucus (see Fig. 1). In addition, distinguish the nature of chlorine, i.e. whether it is as molecu-
the results from ab initio thermodynamics shown in Fig. 5 indi- lar chlorine solvated, hydrochloric acid solvated, or in the form
cate that under the experimental conditions, a partial occupation of a surface oxychloride species. In any case, the minor amount
of the Rucus atoms on the surface is likely. The state of the ini- of chlorine is limited to the catalyst surface. Accordingly, the
tial active surface in the presence of the gas-phase O2 pressures use of an oxide, rather than a chlorinated compound, for DFT
contains all the Ob sites occupied and coverage of Ocus in the modelling is justified. In line with XPS analysis, application
range of 0.25–0.75 ML. of other characterization methods such as far-infrared and Ra-
The used RuO2 sample resulting from temperature-pro- man spectroscopies, temperature-programmed desorption cou-
grammed reaction experiments was characterized ex situ in or- pled to mass spectrometry, and chemical analysis detected vir-
der to assess changes due to the reaction. It should be recalled tually no chlorine in the used RuO2 catalyst. The practically
that the catalyst was subjected to a ramp of 5 K min−1 from 333 unaltered state of RuO2 after the Deacon reaction highlights the
to 583 K in a mixture of 20 vol% HCl, 40 vol% O2 , and 40 vol% stability of this catalyst and likely explains its successful im-
N2 and the final temperature was kept for 1 h. Thereafter, the re- plementation as compared to other oxides based on copper and
actor was rapidly cooled down to room temperature in N2 flow. chromium, where extensive volatilization upon formation of the
The X-ray diffraction pattern of the used sample did not expe- corresponding (oxy)chlorides has been observed. Future work
rience apparent changes with respect to the fresh one (Fig. 4). will consider application of operando spectroscopic techniques
The position and relative intensity of the reflections of the RuO2 to monitor the Deacon reaction. This approach proved success-
phase were very similar and in particular no crystalline ruthe- ful to assess the chlorination at both surface and bulk levels of
34 N. López et al. / Journal of Catalysis 255 (2008) 29–39

Fig. 6. XPS spectra of Ru 3d (left), O 1s (center) and Cl 2p (right) core level electrons of the RuO2 catalyst used in the Deacon reaction.

lanthanide oxide-based catalysts during the destruction of CCl4 have studied the reaction network at distinct degrees of surface
in the presence of steam [54]. oxidation and chlorination.
For the used catalyst, the RuO2 (110) structure shall be in
HCl + O* + * ↔ OH* + Cl*, (3)
equilibrium with both Cl2 and O2 . In this case the ab initio
thermodynamics are much more complex, the configurations Cl* + Cl* ↔ Cl2 + 2*, (4)
included in the search are reported in top Fig. 7 while the energy
OH* + OH* ↔ H2 O* + O*, (5)
diagram is shown in bottom Fig. 7. The most likely configura-
tions at relevant O2 and Cl2 pressures contain bridging sites H2 O* ↔ H2 O + *, (6)
occupied mostly with oxygen, the complete substitution of Ob O2 + 2* ↔ 2O*. (7)
atoms by Clb is energetically unfavored (by more than 1 eV
with respect to the clean surface), thus suggesting that com- 4.3.1. Hydrogen abstraction from hydrogen chloride
plete chlorination of the surface is indeed unlikely. The Rucus The first step in the mechanism is the hydrogen abstraction
sites are also partially occupied, however, in this position Cl from HCl leading to adsorbed atomic chlorine and hydroxyl
is more likely than O. Although partial substitution of Ob by species (Eq. (3)). Three dissociation scenarios were investi-
Cl is energetically favored under equilibrium conditions, care gated by using different oxygen coverages (Fig. 8). For the
must be taken since the elimination of Ob from the surface regular (110) surface, the bridge oxygen (Ob ) can act as the ba-
is highly endothermic (1.95 eV with respect to gas-phase O2 ) sic center to dissociate hydrogen chloride. In the initial state,
and the kinetics of replacement can not compete with other the HCl molecule is bonded through electrostatic interaction
less energy demanding processes like O2 desorption from Rucus to Ob . This adsorbed state is determined −0.48 eV below the
sites. gas-phase reactants. For this structure, the relevant geomet-
ric parameters are d(Ob –H) = 1.780, d(H–Cl) = 1.358, and
4.3. Reaction mechanism d(Cl–Rucus ) = 2.668 Å. The H–Cl distance is already longer
than that in the gas-phase molecule, i.e. d(H–Cl) = 1.274 Å.
Dissociation of the HCl is exothermic by −0.98 eV and leads
The mechanism of the overall Deacon reaction (Eq. (2)) to a bridging hydroxyl group (Ob H) and a chlorine atom ad-
on RuO2 (110) can be structured in five steps: hydrogen ab- sorbed on the Rucus position (denoted as Clcus ). The geometric
straction from hydrogen chloride by adsorbed atomic oxygen parameters in the final state are d(Ob –H) = 0.992, d(H–Cl)
(Eq. (3)), recombination of surface chlorine atoms and des- = 2.278, and d(Cl–Rucus ) = 2.361 Å. At the transition state,
orption as gas-phase Cl2 (Eq. (4)), recombination of surface Ea = 0.05 eV, the hydrogen atom is midway between the oxy-
hydroxyl groups (Eq. (5)), water desorption (Eq. (6)), and dis- gen and the chlorine atoms, i.e. d(Ob –H) = 1.702, d(H–Cl) =
sociative oxygen adsorption for surface regeneration (Eq. (7)). 1.371, and d(Cl–Rucus ) = 2.672 Å. The basic features of this
Although the above reaction scheme appears relatively simple, elementary step are not affected by the presence of Clcus atoms
the complexity of the real process is large due to the distinct in the Ru neighboring to the Rucus active site, top right Fig. 8.
nature of the oxygen species on the surface and the different The HCl adsorption is reduced to −0.12 eV and dissociation is
oxygen and chlorine coverage. Accordingly, each of the simpli- slightly less exothermic (−0.93 eV) compared to the clean sur-
fied steps in the reaction can occur in different scenarios. For face, while the dissociation barrier is even smaller than 0.05 eV.
instance, O* in Eqs. (3) and (5) can be either Ocus or Ob and When a partially O-covered Rucus exists on the surface (see
each case has to be computed separately. In consequence, we bottom Fig. 8), an alternative reaction path is feasible. In the
 N. López et al. / Journal of Catalysis 255 (2008) 29–39 35

Fig. 7. Schematic models included in the ab initio thermodynamics diagram of the surface composition of RuO2 (110) as a function of the partial O2 and Cl2
pressures at 573 K.

initial configuration, HCl can point towards the Ocus atom lead- RuO2 (110)-Ocus or Clcus and thus can easily occur at any con-
ing to a hydroxyl group on a Rucus site (Ocus –H) and a chlorine dition studied here. In good correspondence, Hisham and Ben-
atom on a neighboring Rucus site (Clcus ). The initial state is son [17] concluded that the chlorination step is fast over most
stabilized by −0.47 eV with respect to the gas-phase molecule single oxides.
and the relevant distances are d(Ocus –H) = 1.659, d(H–Cl) =
1.356, and d(Cl–Rucus ) = 2.641 Å. The Ocus -assisted H ab- 4.3.2. Chlorine recombination
straction is exothermic by −0.75 eV, slightly smaller than the Recombination of adsorbed atomic chlorine (Eq. (4)) has
Ob -assisted H abstraction (−0.98 eV). In the final state, the been studied for several conditions (Fig. 9). First, a regular sur-
d(Ocus –H) = 0.994 and d(Cl–Rucus ) = 2.358 Å. The transition face with a p(2 × 1) unit cell was employed with both Rucus
state is 0.11 eV over reactants and the distances are d(Ocus –H) sites covered by Clcus . However, a chlorine atom can substi-
= 1.684, d(H–Cl) = 1.351, and d(Rucus –Cl) = 2.640 Å. tute an Ob in the RuO2 structure (Clb ), being able to recombine
In summary, the barriers for hydrogen abstraction from with Clcus atom at a Rucus site. The recombination of two Clcus
HCl are small over both p(2 × 1)-RuO2 (110) and p(2 × 1)- starts by the approach of one of the chlorine atoms to the sec-
36 N. López et al. / Journal of Catalysis 255 (2008) 29–39

ond followed by evolution of gas-phase Cl2 . The direct process the latter process is not feasible under the experimental condi-
is endothermic by E d = 1.56 eV on the clean surface. The Clcus tions applied in the experiments.
+ Clb recombination is much more endothermic, 3.20 eV. Thus For the Clcus + Clcus reaction, the effect of excess Ocus
and Clcus on the neighboring Rucus sites has been studied in
the p(4 × 1) supercell varying the amount of extra Xcus =
Ocus and/or Clcus on the surface: θXcus = 0, 0.25, and 0.50 ML
(Fig. 10). For the clean RuO2 (110) surface, a desorption energy
E d = 1.80 eV was found. The presence of Ocus (Clcus ) signifi-
cantly lowered the energy required for Cl + Cl recombination,
from 1.80 eV at θOcus = 0 to 1.26 eV at θOcus = 0.50 ML or
1.37 eV at final θClcus = 0.50 ML. Thus the endothermicity of
the reaction can be largely affected by the quality of the surface,
i.e. the degree of oxidation or chlorination. This implies a pos-
itive effect of Ocus and/or Clcus on Cl2 production due to easier
chlorine recombination. This fact might (at least partially) ac-
count for the increased Cl2 production at higher O2 pressures
in Fig. 3.
Regarding technical issues, Cl2 recombination calculated in
the small p(2 × 1) supercell (open circle at θOcus = 0 in Fig. 10)
leads a value of E d = 1.56 eV. This is due to the presence of
spurious lateral repulsions in this cell and indicates that one
should take precautions when using small supercells to study
this process. This is critical here, as chlorine recombination is
the most energetically demanding step in the proposed Deacon
scheme. Finally, the absolute values for adsorption/desorption
energies are known to be affected by systematic errors due to
inaccuracies in the functional [36]. However, the dependence
of the recombination energy with the total coverage found in
Fig. 10 should be seen as a derivative of the energy and is thus
less affected by systematic errors. In conclusion, the trend in
Fig. 10 reflects a physical situation.

4.3.3. Water formation

Fig. 8. Schematic representation of the initial and final states for hydrogen ab- H2 O is formed by recombination of surface hydroxyl groups
straction from HCl on RuO2 (110) at two different oxygen/chlorine coverages. followed by desorption (Eqs. (5) and (6)). Firstly, H is trans-

Fig. 9. Schematic representation of the initial states for atomic chlorine recombination on RuO2 (110). In this case a p(4 × 1) cell is employed and three different
extra O coverages.
 N. López et al. / Journal of Catalysis 255 (2008) 29–39 37

ferred from Ob H to Ocus . In a second step, hydroxyl recombi- gen and oxygen atoms are d(Ob –H) = 1.264 and d(Ocus –H) =
nation was studied. These processes are visualized in Fig. 11. 1.210 Å, respectively. The Ocus –Rucus bond is distorted to assist
Since Ob atoms are difficult to remove from the RuO2 (110) sur- hydrogen transfer and the corresponding distance is 1.849 Å.
face, recombination has only been studied as Ob H + Ocus H At the final state, the d(Ob –H) is enlarged to 2.453 Å and the
to yield Ob and a water molecule sitting at the Rucus site fol- hydroxyl group distance is d(Ocus –H) = 0.978 Å, while the
lowed by water desorption. The same process can also take bonding distance in the Rucus –Ocus H group is 1.942 Å. If chlo-
place between two hydroxyl groups on the Rucus sites, hydroxyl rine is present in the neighboring Rucus site the energy for the
recombination and water desorption was also examined for this reaction is −0.03 eV while the barrier is 0.48 eV; thus the effect
case (vide infra). of Clcus is very small.
Hydrogen transfer from the Ob –H to the Ocus site is only The second hydrogen transfer can be described as follows:
exothermic by −0.01 eV. At the initial state the d(Ob –H) = Ob –H + Ocus –H → Ob + H2 Ocus (Fig. 11, bottom). The en-
0.982 Å, i.e. typical from a hydroxyl group. The d(Ocus –H) = ergy at the transition state is 0.38 eV higher than reactants and
2.149 Å and the d(Ocus –Rucus ) = 1.752 Å. The hydrogen trans- the overall process is slightly endothermic, 0.27 eV. In the final
fer is hindered by a barrier of 0.55 eV (0.56 eV for the inverse structure a water molecule is sitting at a Rucus site. The desorp-
barrier). At the transition state, the distances between the hydro- tion energy of the water molecule is 0.67 eV. If a Clcus atom is
present in the neighboring Rucus site the energies are slightly af-
fected, the hydrogen transfer process is exothermic by 0.16 eV
due to the formation of a Cl–H2 O interaction in the final state,
and the barrier is lowered to 0.08 eV. However, this is compen-
sated by the largest desorption energy of water, 0.90 eV. The
recombination of two hydroxyl groups in neighboring Rucus
sites (Ocus –H + Ocus –H → Ocus + H2 Ocus ) can be also con-
sidered (not shown in Fig. 11). This may correspond to the high
oxygen coverage regime. At the initial state the H–Ocus distance
is 1.878 Å while at the transition state the H atom is placed at
1.287 and 1.183 Å from both Ocus atoms. The transition state is
only 0.24 eV higher in energy than reactants and the process
is exothermic by −0.11 eV. Water desorption in this case is
endothermic by 0.90 eV. The larger endothermicity found for
water desorption is associated with the formation of hydrogen
bonds. Our results leading to small barriers for hydroxyl re-
Fig. 10. Recombination energy of chlorine atoms to gas-phase Cl2 as a function combination are in agreement with previous calculations and
of the coverage Xcus = O, Cl on RuO2 (110). experimental results [24–26].

Fig. 11. Schematic representation of the initial and final states for hydrogen transfer for water formation at the Rucus site on RuO2 (110). Top: Ob –H + Ocus → Ob
+ Ocus H; and bottom Ob –H + Ocus H → Ob + H2 Ocus .
38 N. López et al. / Journal of Catalysis 255 (2008) 29–39

HCl molecule and as a hydroxyl reservoir in water formation.

However, structural oxygen in ruthenium oxide (Ob ) cannot
be removed as water (H2 Ob ) and therefore cannot oxidize by
itself hydrogen chloride. In contrast, Ocus can act as proton
scavenger, forming hydroxyl groups on the surface and water
molecules. The binding energy of the latter is low enough so
they can desorb under reaction conditions.

5. Conclusions

DFT calculations in combination with experiments have

Fig. 12. Schematic representation of the initial and final states for O2 dissocia- been used to study the HCl oxidation to Cl2 (Deacon process)
tion on RuO2 (110).
over RuO2 . Activity tests in a fixed-bed reactor at ambient pres-
sure demonstrated that RuO2 powder is a highly active catalyst
4.3.4. Surface reoxidation for Cl2 production. Characterization of the fresh and used RuO2
In order to complete the catalytic cycle, the surface Ocus revealed no appreciable alteration of the bulk structure and lim-
atoms on the surface need to be regenerated by dissociative ited chlorination of the sample, in agreement with ab initio
adsorption of gas-phase O2 (Eq. (7)). This is similar to what thermodynamics. The Deacon reaction has been described by
happens for CO oxidation on this surface and has been quoted a Mars–van Krevelen type reaction mechanism consisting of
as Mars–van Krevelen mechanism [19]. We have experimen- five elementary steps: hydrogen abstraction from HCl, recom-
tally observed that the Cl2 production fades away when no O2 bination and desorption of atomic chlorine, hydroxyl recombi-
is available in the feed gas (see Fig. 3). The O2 dissociation nation, water desorption, and oxygen adsorption. Atomic chlo-
on the Rucus sites leading to two Ocus atoms has been studied. rine recombination to yield gas-phase Cl2 is the most energy-
The O2 molecule has been adsorbed on different configurations demanding step. In addition, surface reoxidation can be blocked
over the Rucus rows of the RuO2 surface. The O2cus tilted is by high surface coverages. Reaction conditions (mainly partial
adsorbed by −0.47 eV and can be considered as the precursor
pressures and temperature) will determine whether the rate-
for dissociation (Fig. 12, left). In the O2cus configuration, the
determining step of the Deacon process is chlorine recombina-
O–O distance is 1.307 Å, thus close to a superoxo O− 2 species, tion or oxygen readsorption. Further clarification of this aspect
the shortest O–Rucus distance is 2.044 and the longest 2.498 Å, requires more specific mechanistic and kinetic studies. Finally,
from the second O to the free Rucus . The transition state struc- we have experimentally observed a larger Cl2 production upon
ture is located 0.38 eV higher in energy than the initial state. increasing the feed O2 :HCl ratio. Based on our calculations, this
In the structure, the O–Rucus distance is 1.891 Å while the sec- trend is tentatively attributed to both the increase of O coverage
ond O–Rucus is 1.883 Å and the O–O distance is enlarged to on the surface at higher partial O2 pressures and the reduced
1.670 Å. The final structure is 0.76 eV more stable than the energy for chlorine formation at high coverages.
O2cus and consists in 2Ocus . The effect of surface coverage has
been explored in a similar way as for Cl2 formation, i.e. in a
Acknowledgments
p(4 × 1) supercell and with various Clcus and Ocus coverages.
The adsorption energy of the precursor state O2cus molecule
is reduced with respect to the clean surface, < −0.20 eV. The Financial support from the Spanish MEC (projects CTQ-
transition state lays about 0.4 eV higher in energy than the pre- 2006-01562/PPQ, CTQ2006-00464BQU, and Consolider-
cursor state, while the final state is about −0.65 eV lower in Ingenio 2010, grant CSD2006-003) and the Generalitat de
energy than the reactants. The surface coverage effect on O2 Catalunya is acknowledged. The authors thank BSC and RES-
dissociation is to reduce the binding energy of the precursor to CesViMa for computing time.
the surface.
Finally, in the reaction network, O2 readsorption is the only References
step requiring two empty Rucus sites. As we have seen in the
computational characterization of the surface (Section 3), two [1] S. Motupally, D.T. Mah, F.J. Freire, J.W. Weidner, Electrochem. Soc. In-
terface 7 (3) (1998) 32.
neighboring empty Rucus sites might not be abundant at high
[2] H. Deacon, U.S. Patent 165,802, 1875.
Cl2 and O2 pressures. Consequently, surface oxidation may be- [3] H.F. Johnstone, Chem. Eng. Prog. 44 (1948) 657.
come a difficult step in the reaction. This latter aspect might be [4] H.Y. Pan, R.G. Minet, S.W. Benson, T.T. Tsotsis, Ind. Eng. Chem. Res. 33
partially responsible for the dependence observed between the (1994) 2996, and references therein.
Cl2 production over RuO2 and the partial O2 pressure in the [5] A.J. Johnson, A.J. Cherniavsky, U.S. Patent 2,542,961, 1951.
feed mixture (Fig. 3). [6] F. Wattimena, W.M.H. Sachtler, Stud. Surf. Sci. Catal. 7 (1981) 816.
[7] T. Yasuaki, Stud. Surf. Sci. Catal. 92 (1995) 41.
The nature of the active oxygen on the RuO2 (110) surface
[8] H. Itoh, Y. Kono, M. Ajioka, S. Takezaka, M. Katzita, U.S. Patent
has been a matter of extensive discussion [23,55]. Our calcu- 4,803,065, 1989.
lations point out to the dual role of the oxygen on the surface. [9] T. Kiyoura, Y. Kogure, T. Nagayama, K. Kanaya, U.S. Patent 4,822,589,
The lattice oxygen Ob can act as a proton scavenger for the 1989.
 N. López et al. / Journal of Catalysis 255 (2008) 29–39 39

[10] R.G. Minet, S.W. Benson, M.K. Mortensen, T.T. Tsotsis, WO 97/11026. [31] S. Yang, Y. Feng, J. Wan, W. Zhu, Z. Jiang, Appl. Surf. Sci. 246 (2005)
[11] M. Mortensen, R.G. Minet, T.T. Tsotsis, S.W. Benson, Chem. Eng. Sci. 54 222.
(1999) 2131. [32] K. Villani, C.E.A. Kirschhock, D. Liang, G. Van Tendeloo, J.A. Martens,
[12] U. Nieken, O. Watzenberger, Chem. Eng. Sci. 54 (1999) 2619. Angew. Chem. Int. Ed. 45 (2006) 3106.
[13] T. Hibi, T. Okuhara, K. Seki, H. Abekawa, H. Hamamatsu, WO 01/10550, [33] S. Brunauer, P.H. Emmet, E. Teller, J. Am. Chem. Soc. 60 (1938) 309.
2001. [34] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
[14] T. Arikawa, Y. Murakami, Y. Takasu, J. Appl. Electrochem. 28 (1998) 511. [35] G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169.
[15] V.V. Panic, A. Dekanski, S.K. Milonjic, R.T. Atanasoski, B.Z. Nikolic, [36] B. Hammer, L.B. Hansen, J.K. Nørskov, Phys. Rev. B 59 (1999) 7413.
Colloids Surf. A 157 (1999) 269. [37] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
[16] K. Iwanaga, K. Seki, T. Hibi, K. Issoh, T. Suzuta, M. Nakada, Y. Mori, [38] R.W.G. Wyckoff, Crystal Structures, vol. 1, second ed., Wiley, New York,
T. Abe, Sumitomo Kagaku (2004) 4. 1963.
[17] M.W.M. Hisham, S.W. Benson, J. Phys. Chem. 99 (1995) 6194. [39] C.E. Boman, Acta Chem. Scand. 24 (1970) 116.
[18] A.G. Aglulin, Kinet. Catal. 39 (1998) 521. [40] A. Kiejna, G. Kresse, J. Rogal, A. De Sarkar, K. Reuter, M. Scheffler,
Phys. Rev. B 73 (2006) 035404.
[19] H. Over, Y.D. Kim, A.P. Seitsonen, S. Wendt, E. Lundgren, M. Schmid,
[41] Y.D. Kim, S. Schwegmann, A.P. Seitsonen, H. Over, J. Phys. Chem. B 105
P. Varga, A. Morgante, G. Ertl, Science 287 (2000) 1474.
(2001) 2205.
[20] L. Wang, H. Kisch, Angew. Chem. Int. Ed. 39 (2000) 3921.
[42] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
[21] H. Madhavaram, H. Idriss, S. Wendt, Y.D. Kim, M. Knapp, H. Over, J. Aß-
[43] G. Henkelman, B.P. Uberuaga, H. Jónsson, J. Chem. Phys. 113 (2000)
mann, E. Löffler, M. Muhler, J. Catal. 202 (2001) 296.
9901.
[22] J. Assmann, E. Löffler, A. Birkner, M. Mühler, Catal. Today 85 (2003)
[44] K. Reuter, M. Scheffler, Phys. Rev. B 68 (2003) 045407.
235.
[45] X.G. Wang, A. Chaka, M. Scheffler, Phys. Rev. Lett. 84 (2000) 3650.
[23] K. Reuter, M. Scheffler, Phys. Rev. B 73 (2006) 045433. [46] Z. Łodziana, J.K. Nørskov, P. Stoltze, J. Chem. Phys. 118 (2003) 11179.
[24] J.H. Wang, C.Y. Fan, Q. Sun, K. Reuter, K. Jacobi, M. Scheffler, G. Ertl, [47] NIST webpage http://webbook.nist.gov/chemistry, September 2007.
Angew. Chem. Int. Ed. 42 (2003) 2151. [48] G.E. Gibson, N.S. Bayliss, Phys. Rev. 44 (1933) 188.
[25] M. Knapp, D. Crihan, A.P. Seitsonen, E. Lundgren, A. Resta, J.N. Ander- [49] S. Hubinger, J.B. Nee, J. Photochem. Photobiol. A 86 (1995) 1.
sen, H. Over, J. Phys. Chem. C 111 (2007) 5363. [50] J.B. Burkholder, E.J. Bair, J. Phys. Chem. 87 (1983) 1859.
[26] M. Knapp, D. Crihan, A.P. Seitsonen, H. Over, J. Am. Chem. Soc. 127 [51] G. Wulff, Z. Krystallogr. Miner. 34 (1901) 449.
(2005) 3236. [52] M. Ramamoorthy, D. Vanderbilt, R.D. King-Smith, Phys. Rev. B 46
[27] Y. Wang, K. Jacobi, W.-D. Schoene, G. Ertl, J. Phys. Chem. B 109 (2005) (1994) 16721.
7883. [53] H. Over, M. Gierer, H. Bludau, G. Ertl, S.Y. Tong, Surf. Sci. 314 (1994)
[28] I. Balint, A. Miyazaki, K.-I. Aika, J. Catal. 220 (2003) 74. 243.
[29] B.-Z. Zhan, M.A. White, T.-S. Sham, J.A. Pincock, R.J. Doucet, K.V.R. [54] P. van der Avert, S.G. Podkolzin, O. Manoilova, H. de Winne, B.M. Weck-
Rao, K.N. Robertson, T.S. Cameron, J. Am. Chem. Soc. 125 (2003) 2195. huysen, Chem. Eur. J. 10 (2004) 1637.
[30] H. Liu, E. Iglesia, J. Phys. Chem. B 109 (2005) 2155. [55] S. Wendt, M. Knapp, H. Over, J. Am. Chem. Soc. 126 (2004) 1537.