Electrochimica Acta 120 (2014) 460466

Contents lists available at ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

Chlorine Evolution Reaction on RuO2 (110): Ab initio Atomistic

Thermodynamics Study - Pourbaix Diagrams
Kai S. Exner a , Josef Anton b , Timo Jacob b , Herbert Over a,
a
Physical Chemistry Department, Justus-Liebig-University Giessen, Heinrich-Buff-Ring 58, 35392 Giessen, Germany
b
Institute of Electrochemistry, Ulm University, Albert-Einstein-Allee 47, 89069 Ulm, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Constrained ab initio thermodynamics in the form of a Pourbaix diagram can greatly assist kinetic model-
Received 25 September 2013 ing of a particular electrochemical reaction such as the chlorine evolution reaction (CER) over RuO2 (110).
Received in revised form 29 October 2013 Pourbaix diagrams reveal stable surface structures, as a function of pH and the potential. The present
Accepted 1 November 2013
DFT study indicates that the Pourbaix diagram in the CER potential region above 1.36 V and pH values
Available online 18 November 2013
around zero is dominated by a stable surface structure in which all coordinatively undercoordinated Ru
sites (Rucus ) are capped by on-top oxygen (Oot ). This oxygen saturated RuO2 (110) surface is considered to
Keywords:
serve as the catalytically active phase in the CER, quite in contrast to the heterogeneously catalyzed HCl
Chlorine evolution reaction
Pourbaix diagram
oxidation (Deacon process), for which the active RuO2 (110) surface is mainly covered by on-top chlorine.
Constrained thermodynamics The active sites in the CER are suggested to be Rucus Oot surface complexes, while in the Deacon process
Deacon process both undercoordinated surface Ru and oxygen sites must be available for the activation of HCl molecules.

2013 Published by Elsevier Ltd.

1. Introduction in order to aid unraveling the reaction mechanism. However, the

One of the greatest challenges in electrocatalysis constitutes
disclosing and unraveling the complete reaction mechanism of
an electrochemical reaction at the electrode/electrolyte interface
and identifying the catalytically active phase or active sites on
the electrode surface [1]. From pure macrokinetic experiments
(i.e. Tafel slope, reaction order, pH dependence) this ultimate
goal cannot be accomplished, but rather this challenging enter-
prise needs to identify reaction intermediates, which are the most
important bit of information to construct the reaction mecha-
nism [2]. In electrocatalysis there are only very few spectroscopic
methods available to identify reaction intermediates at the elec-
trode/electrolyte interface, most notable infrared spectroscopy
[3,4]. Electron spectroscopy, which has been demonstrated to be
pivotal to identify reaction intermediates in heterogeneous cataly-
sis of gas phase reactions [5], cannot be applied due to the presence
of the electrolyte solution and the requirement of a sufciently long
mean free path of electrons to reach the detector of the analyzer.
In heterogeneous catalysis of gas phase reactions ab initio theory
has become a powerful tool [68], in particular in combination with
surface chemical experiments at single crystalline model catalysts,
Corresponding author.
E-mail addresses: timo.jacob@uni-ulm.de (T. Jacob),
Herbert.Over@phys.Chemie.uni-giessen.de (H. Over).
full ab initio kinetic simulation of a reaction at the catalyst surface
is still a demanding enterprise, which needs for instance kinetic
Monte Carlo modeling to treat adequately the intricate interplay
of the elementary reaction steps at the surface and the exchange
of the molecules with the gas phase. Frequently the actual reac-
tion mechanism and the identication of the active surface phase
can be decoupled using rst principles thermodynamics. Ab initio
thermodynamics is able to evaluate the stability of a catalyst sur-
face structure under (constrained) reaction conditions as a function
of the reaction temperature and reactant pressures by allowing the
exposure of the surface to the reactant mixture, but suppressing the
actual surface reaction and also a possible solid state transforma-
tion of the catalyst (hereafter referred to as constrained TD). This
active surface phase is then considered a promising starting point
of a full kinetic study including surface reactions. For the case of
CO oxidation over RuO2 (110) this simple approach has succeeded
to identify the active surface phases under reaction conditions
[9], which are compatible with those found by a full (i.e. including
the actual reaction) analysis using ab initio based kMC simulations
[1012]. For the HCl oxidation reaction (so-called Deacon process)
over RuO2 (110) Studt et al. performed constrained TD calculations
[13].
In comparison with gas phase reactions in heterogeneous catal-
ysis additional effects such as the solvent and the electrode
potential play a predominant role in electrocatalysis so that a full
ab initio based kinetic modeling of an electrochemical system is

0013-4686/$ see front matter 2013 Published by Elsevier Ltd.

http://dx.doi.org/10.1016/j.electacta.2013.11.027
 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
Fig. 1. Stoichiometric RuO2 (110) surface. Undercoordinated ruthenium atoms, on which adsorption processes occur, are the 1f-cus ruthenium atoms (denoted as Rucus ) and
2f-cus ruthenium atoms (denoted as Rubr ). Adsorbates in bridge position on Rubr atoms are denoted with the index br and adsorbates on Rucus atoms with the index ot
(on-top). In case of the stoichiometric RuO2 (110) surface the bridge sites are fully covered by bridging oxygen atoms denoted as Obr . Color code: red balls: ruthenium 1f-cus
atoms (Rucus ), pink balls: ruthenium 2f-cus atoms (Rubr ), blue balls: bulk ruthenium atoms, green balls: oxygen atoms.
quite demanding [6,14,15]. Therefore, simple methods are urgently
called for in electrochemistry to gain at least some important
information of the reaction system, such as the active surface phase
of the catalyst under reaction conditions. In electrocatalysis ab
initio TD can equally be applied to gain important information
about the active surface under reaction conditions [1619]. Here
the pH value and the electrostatic potential rather than the par-
tial pressures of the reactants and reaction temperatures are of
importance. The resulting (constrained) surface phase diagrams are
generally known as (surface) Pourbaix diagrams [20].
In this paper we will concentrate on the chlorine evolution reac-
tion (CER) over RuO2 . A detailed comparison of the CER with the
related gas phase reaction (Deacon reaction) has recently been pre-
sented [21]. In comparison to the oxygen evolution reaction (OER)
or the oxygen reduction reaction (ORR) is the CER considered to be
much simpler due to the smaller number of transferred electrons
(two instead of four electrons) [22]; the CER is almost as simple
as the prototypical hydrogen evolution reaction in electrocatalysis
[23]. Hansen et al. [24] had rst applied the constrained thermo-
dynamic approach for electrochemistry (EC) to study the reaction
mechanism of the CER over RuO2 (110) in comparison with that over
IrO2 (110). In this study solvent effects were neglected which brings
Pourbaix diagrams even closer to diagrams of related gas phase HCl
oxidation reaction (Deacon process) with the main difference that
the energies are references to water in the gas phase instead of sta-
ble elements in gas phase. However, a critical comparison with the
gas phase HCl oxidation reaction (Deacon process) revealed several
open questions [21], which inclined us to thoroughly re-investigate
the CER to deepen our understanding of the CER over RuO2 (110).
In particular, we will be studying the CER within the constraint
TD approach, including however solvent effects and addressing the
completeness of considered surface structures.
2. Methods
2.1. Calculation Details
All electronic structure calculations have been carried out
within the Density Functional Theory (DFT) approach, utilizing the
SeqQuest program package [25] and PBE as functional for correla-
tion and exchange [26,27]. The RuO2 (110) surface was modeled by
a periodic stack of asymmetric slabs containing ve O-Ru-O layers
with adsorbates only on one side of the slab. Consecutive slabs are
separated by a vacuum region of 15 A. The DFT calculations were
restricted to 2 1 surface unit cells using a 6 6 Monkhorst Pack
k-point mesh for slab calculations. In the total energy calculations
the atomic geometries of the topmost three layers including the
adsorbate layer were optimized, while the atomic positions of the
bottom two O-Ru-O layers were xed. A critical comparison of the
461
obtained adsorption energies with those derived by VASP calcu-
lations [28] indicates that the deviations are within the supposed
accuracy of DFT calculations, i.e. less than 0.15 eV per adsorbate.
The 2 1 surface unit cell of RuO2 (110) contains two bridge
(Rubr ) and two cus sites (Rucus ) (cf. Fig. 1). We consider adsorption
of Obr , OHbr , OClbr , and Clbr at bridge sites (Rubr ) and adsorp-
tion of Oot , OHot , OClot , Clot , Cl(Oot )2 , (Oot )2 , (O2 )ot , and (O3 )ot ,
at coordinatively unsaturated sites (Rucus ). All possible combina-
tions of adsorption sites and adsorbates were taken into account
in this study, giving rise to more than 100 adsorbate structures (cf.
supporting information-SI). The adsorption energies of the consid-
ered adsorbates are calculated with reference to gaseous chlorine,
gaseous hydrogen and water molecules using the chemical poten-
tial from water vapor, which is in equilibrium with liquid water at
298 K and 0.035 bar [29].
The changes in the adsorption energy due to the presence of
the solvent are estimated by cluster calculations using the Jaguar
package [30]. The corresponding cluster model is constructed from
61 atoms and the open valences are saturated by hydrogen. The
solvent water is involved implicitly via the self-consistent reac-
tion eld (SCRF) approach that describes the solvent as continuum
dened by its dielectric constant and the probe radius of the solvent
[31]. First, the geometry of the adsorbate structure is optimized for
the gas phase, which is associated with an adsorption energy Egas .
In a next step, this geometry is xed and the adsorption energy
Eliq in presence of water is determined. The solvent effect Esolv
is then given by:
Esolv = Eliq Egas (1)
Finally, the total adsorption energy Etot is the sum of the
adsorption energy ESeq calculated (for the periodic system) by
SeqQuest for the gas phase and the previously determined solvent
effect term Esolv (see eq. (1)):
Etot = Eseq + Esolv (2)
2.2. Construction of Surface Pourbaix Diagrams
A Pourbaix diagram reveals the thermodynamically most sta-
ble structure of the catalyst surface among the set of considered
structure as a function of applied electrode potential U and pH
[20]. The thermodynamically favoured structure is determined by
minimization of the Gibbs enthalpy G among the set of
considered surface structures. The Gibbs free enthalpy is given by
[24]:
G = Etot + EZPE T S + Gref . (3)
Etot denotes the adsorption energy in the presence of the sol-
vent (water) calculated by DFT, EZPE the change in zero-point
462 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
energy, T the temperature in Kelvin and S the change in entropy
upon adsorption. If solvent effects are neglected, then ESeq is
used instead of Etot . The change in zero-point energy is calcu-
lated within the harmonic approximation. The change of entropy
is taken from tables of thermodynamic properties [32] in the case of
gas-phase species. Congurational entropy effects of the RuO2 (110)
surface were estimated to be negligible. For all surface structures
considered the calculated values of adsorption energies, change
of zero-point energies and change in entropy can be found in the
Supplementary Information (SI). The correction Gref in eq. (3) con-
tains potential U and pH effects, which depend on the reversible
potentials of the chlorine UCl , the standard hydrogen electrodes
(SHE), and the reaction equations (stoichiometric coefcients) of
the particular adsorption process. We obtain the following equa-
tion for room temperature (T = 298 K) for the Gibbs free enthalpy as
a function of pH and U:
G(pH, U) = Etot + EZPE T S (H+ ) 0.059[eV]
pH (e ) e U (Cl ) e UCl. (4)
The value 0.059 eV results from kTln10 evaluated at room tem-
perature. Values for the stoichiometric coefcients of transferred
electrons (e ), protons (H+ ) and chloride (Cl ) of considered
adsorption processes are listed in the SI for all studied surface
structures. UCl denotes the standard potential of the reversible
chlorine electrode (1.36 eV at 298 K and 1 bar) and U is the
applied electrode potential with respect to standard hydrogen
electrode.
Eq. (4) is valid for the case a(Cl ) = 1, a condition which can be
fullled by using a supporting electrolyte so that the concentra-
tion of chloride ions is nearly constant under reaction conditions.
However, under HCl oxidation conditions the activity of chloride is
coupled to the proton activity, a(Cl ) = a(H+ ), so that an additional
term (Cl ) enters in the pH term:
G(pH, U) = Etot + EZPE T S [(H+ ) + (Cl )]
0.059[eV] pH (e ) e U (Cl ) e UCl (5)
2.3. Critical Remarks about Pourbaix Diagrams
A Pourbaix diagram for the CER reveals the thermodynamically
most stable structure among the considered surface structures of
the catalyst surface in contact with an aqueous HCl solution as a
function of applied potential U and pH. In this thermodynamic con-
sideration the actual reaction 2 H+ + 2 Cl H2 + Cl2 is suppressed
so that Pourbaix diagrams belong to the category of so-called con-
strained thermodynamics. In general solvent effects have to be
included.
For designing a Pourbaix diagram rst a set of relevant adsorbate
structures has to be selected. The Pourbaix diagram may depend on
the number of reasonable structures considered. Structures, which
are not taken into account, do not enter the Pourbaix diagram and
may easily be missed. Therefore, it is mandatory to use a wide
range of possible but also reasonable structures. A comparison of
two Pourbaix diagrams is only meaningful if the adsorbate struc-
tures of one Pourbaix diagram represent a subset of the adsorbate
structures of the other Pourbaix diagram.
In the relevant electrode potential-pH range of CER the adsorp-
tion of oxygen is favored since liquid water and not molecular
oxygen in the gas phase is used as reference in the thermodynam-
ics calculations. The adsorption of oxygen on the catalyst surface
requires the cleavage of water and the transmission of two protons,
which are counterbalanced by two electrons. These electrons move
through the anode and the external circuit to undertake a reduc-
tion reaction at the cathode. The corresponding energy gain nally
amounts to (e )eU, therefore, in case of water splitting to 2eU.
Thus, at high electrode potentials U the Pourbaix diagram is domi-
nated by adsorption processes with large number of electrons being
transferred rather than by the actual adsorption energies. This dom-
inance of electron transfer processes at high potentials renders the
topology of Pourbaix diagrams more robust against approximations
in the applied DFT approach than the corresponding surface phase
diagrams in heterogeneous catalysis.
The oxygen evolution reaction (OER) is a competing reaction to
the CER. The OER is thermodynamically even more favorable than
CER (1.23 V versus 1.36 V with respect to SHE) but due to kinetic
limitations the exchange current for the OER is quite low. Both evo-
lution reactions produce gases. Thus, an adsorbate structure should
be considered in the Pourbaix diagram only if this structure is stable
at room temperature with respect to the molecular species in the
gas phase. Therefore, surface structures considered in the Pourbaix
diagram must also be referenced against the gas phase molecules
O2 and Cl2 . For instance, if two oxygen atoms or two chlorine atoms
bind as molecular species on the electrode surface the formation of
gaseous oxygen or gaseous chlorine is thermodynamically favored
at room temperature so that these adsorbate structures have to be
ignored in the construction of the Pourbaix diagram. This restricts
the choice of possible adsorbate structures considerably.
As an illustrative example we may inspect the formation of sev-
eral layers of oxygen on the catalysts surface, which would appear
to be thermodynamically favorable as the adsorption of additional
oxygen on the surface releases two additional electrons per O atom
if referenced to water. At higher potential U these electrons results
in a large energy gain. An example of such a Pourbaix diagram can
be found in the SI. However, such an oxygen rich structure is not
stable against the gas evolution as O2 is readily released at room
temperature.
Pourbaix diagrams, although of purely thermodynamic nature,
do provide useful information for the kinetics of the suppressed
reaction. They reveal the potentially active surface phase, which
is thermodynamically stable and which may equally serve as the
active phase under reaction conditions. This simple notion relies,
however, on the tacit assumption that the actual surface reaction
does not change the thermodynamically stable surface structure,
an assumption that may not always be fullled [33].
Another pitfall of ab initio Pourbaix diagrams is that important
surface structures for the kinetic study might be missed. It is con-
ceivable that even surface terminations, which are not the most
stable structures, are relevant as intermediates in the kinetics of
the CER. Therefore, surface structures being only slightly less sta-
ble than the preferred structure in the Pourbaix diagram need to be
taken into account in the kinetic study for the relevant potential-pH
range (U > 1.36 V, pH = 0 2) of the CER.
In constructing surface Pourbaix diagrams several points have
to be considered: It is mandatory to use a wide range of possi-
ble but also reasonable surface structures, which are also stable
against the gas phase. For the actual kinetics we need to include
also surface structures which are only slightly less stable than
the preferred structure. Summarizing, Pourbaix diagrams are the
method of choice for identifying the thermodynamically stable sur-
face phase. These stable phases may be preserved under reaction
conditions at least for the case of low current densities and may
serve as promising starting point for a full kinetic study.
3. State of the Art: Pourbaix Diagrams of Hansen et al.
The rst ab initio study of the Chlorine Evolution Reaction (CER)
on RuO2 (110) catalyst based on ab initio DFT calculations were pub-
lished by Hansen et al. in 2010 [24] constructing a surface Pourbaix
 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
Fig. 2. Pourbaix diagram for RuO2 (110) in equilibrium with Cl , H+ and H2 O at 298 K
and a(Cl ) = 1. The calculations were performed in a (2 1) unit cell of RuO2 (110)
with two cus and two bridge sites. Solvent effects were neglected. The regions where
chlorine evolution (CER) or oxygen evolution (OER) is expected to become signi-
cant have been marked. The gure is modied from Ref. [24]. Color code: red balls:
ruthenium 1f-cus atoms, blue balls: ruthenium 2f-cus atoms and bulk ruthenium,
green balls: oxygen atoms, grey balls: chlorine atoms, small white balls: hydrogen
atoms.
diagram. The activity of chloride was set to unity and solvent effects
were neglected. Hansens study considered the following adsorbate
structures: 2 Obr and 2 OHbr at bridge sites as well as Oot , OHot , OClot ,
Clot , Cl(Oot )2 , (Oot )2 , and (O2 )ot at cus sites. The resulting Pourbaix
diagram is depicted in Fig. 2. Adsorbates in bridge position on 2f-cus
ruthenium atoms are denoted with the index br and adsorbates
on 1f-cus ruthenium atoms with the index ot (on-top).
In the CER region (U > 1.36 V and pH values around 0) the most
stable surface consists of the RuO2 (110) which is covered by at
lying molecular oxygen (Oot )2 bound on two adjacent 1f-cus ruthe-
nium atoms. For potentials higher than 1.55 V the precursor species
Cl(Oot )2 is formed. According to Hansen et al. [24] this precursor
species represents the molecular precursor for the CER as chloride
Cl from the solution react with Cl from Cl(Oot )2 to produce gaseous
chlorine:
(Oot )2 + Cl Cl(Oot )2 + e
Cl(Oot )2 [Cl(Oot )2 ]+ + e
[Cl(Oot )2 ]+ + Cl (Oot )2 + Cl2
From calculation of the work function, Cl(Oot )2 was deduced to
be more positively charged than that of (Oot )2 . Therefore, the indef-
inite chloronium ion species may be ascribed to Cl(Oot )2 , thus being
compatible with the Krishtalik mechanism [34]. Furthermore, the
active site on the catalyst surface is attributed to the molecular
oxygen (Oot )2 covered RuO2 (110) surface.
This study of Hansen et al. represents conceptually a step for-
ward in modern electrocatalysis research. Nevertheless, we have
to denote some critical remarks [20]. Hansen et al. performed
their calculations without including solvent effects. Accordingly
their Pourbaix diagram relies on a gas phase diagram, which can
directly be compared with the heterogeneously catalyzed gas phase
oxidation of HCl, the so-called Deacon process. The active cat-
alyst of the Deacon process is partly chlorinated at the bridge
sites [13,3537]. This partial chlorination was not considered
in Hansens calculations. Additionally, the stability of molecular
oxygen bridging two 1f-cus ruthenium atoms is questionable if
compared with the gas phase oxidation of HCl. It was shown by
463
Fig. 3. Pourbaix diagram for RuO2 (110) in equilibrium with Cl , H+ and H2 O at 298 K
and a(Cl ) = 1. The calculations were performed in a (2 1) unit cell of RuO2 (110)
with two cus and two bridge sites. Solvent effects are considered. The regions where
chlorine evolution (CER) and oxygen evolution (OER) are expected to become signif-
icant have been marked. Color code: red balls: ruthenium 1f-cus atoms, blue balls:
ruthenium 2f-cus atoms and bulk ruthenium, green balls: oxygen atoms, grey balls:
chlorine atoms.
Schneider et al. that the dissociation of lying down (Oot )2 species,
forming two Oot species, is thermodynamically favored [38].
In addition to the lying down (Oot )2 species the stability of the
upright (O2 )ot species and the existence of OHot in a potential range
above 1.2 V are debatable. In this potential range hydrogen leaves
already the surface as a proton and releases one electron. Con-
sequently an Oot adsorbate is thermodynamically by about 1.2 eV
more stable than an OHot adsorbate under these potential condi-
tions.
4. Results and Discussion
4.1. Construction of improved Pourbaix Diagrams based on
Hansen et al.
On the basis of the discussion in section 3 we constructed an
improved surface Pourbaix diagrams, shown in Fig. 3. This Pour-
baix diagram takes solvation effects into account and considers the
following adsorbate structures: Obr , OHbr , OClbr and Clbr at bridge
sites as well as Oot , OHot , OClot , Clot , Cl(Oot )2 , (Oot )2 , and (O2 )ot at cus
sites and all possible combinations of these adsorbates. All adsor-
bate structures accounted for by Hansen et al. are also included in
our study so that the Pourbaix diagrams in Figs. 2 and 3 can directly
be compared.
Most notably in Fig. 3 is the fact that only a few of structures turn
out to be thermodynamically stable in the considered potential-pH
range, although more than 100 adsorbate structures were taken
into account. The Pourbaix diagram reveals that oxygen bound on-
top on 1f-cus ruthenium atoms is stable over a wide potential-
and pH range, in particular for U > 1.36 V. Therefore, the adsorbate
structure 2Obr + 2Oot may serve as the active surface phase under
reaction conditions. At slightly higher potentials (U > 1.49 V) the
terminal adsorption of chlorine on on-top O is 0.7 eV energetically
more favorable than adsorption of chlorine on bridging oxygen.
Therefore, the adsorbate structure (OCl)ot may be identied as pre-
cursor for the CER.
Our DFT calculations demonstrate that (Oot )2 is thermodynami-
cally unstable against dissociation to form 2Oot . Taking solvent
effects into account the 2Oot phase turned out to be 1 eV more
stable than the (Oot )2 phase. A similar result is obtained for pre-
cursor Cl(Oot )2 of Hansen et al. [24]: The 1OClot 1Oot species turn
464 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
Fig. 4. Pourbaix diagram for RuO2 (110) in equilibrium with Cl- , H+ and H2 O at 298 K
and a(Cl ) = 1 without (O2 )ot adsorbate structures. The calculations were performed
in a (2 1) unit cell of RuO2 (110) with two cus and two bridge sites. Solvent effects
are considered. The region where chlorine evolution (CER) is expected to become
signicant has been marked. Colour code: red balls: ruthenium 1f-cus atoms, blue
balls: ruthenium 2f-cus atoms and bulk ruthenium, green balls: oxygen atoms, grey
balls: chlorine atoms.
out to be around 1 eV more stable than Cl(Oot )2 . Neglecting solvent
effects results in an energy difference of merely 0.3 eV instead of
1 eV, thus being within the error bar range of present DFT calcula-
tions. This nding underlines the importance of solvent effects on
adsorption energies. We will discuss this point more thoroughly in
a subsequent paper [39]. We may recall that the DFT calculations
of Hansen et al. used the RPBE functional which is well-known to
yield smaller adsorption energies compared to the PBE functional
used in our study [40].
In contrast to Hansens study, the Pourbaix diagram in Fig. 3
indicates that for potentials above 1.59 V a second chlorine atom
is bound to the Oot -saturated RuO2 (110). The adsorption of a third
chlorine atom on oxygen bound at bridge sites is only observed for
potentials higher than 2 V.
In analogy to Hansens Pourbaix diagram in Fig. 2 2Obr + 2(O2 )ot
and 2Obr + 1OClot 1(O2 )ot adsorbates are identied as thermody-
namically stable phases. The (O2 )ot species can be considered as
precursor for the OER. However, our DFT calculations indicate
that (O2 )ot is thermodynamically unstable against desorption into
the gas phase. Therefore, (O2 )ot adsorbate structures have to be
excluded in the construction of Pourbaix diagrams of CER and OER.
The nal Pourbaix diagram is depicted in Fig. 4.
In the Pourbaix diagram of Fig. 4 only four adsorbate struc-
tures are thermodynamically stable against both references water
and gaseous O2 , Cl2 . A clear precursor state for the OER cannot be
deduced form the Pourbaix diagram. The 2Obr + 2Oot structure is
stable against O2 desorption at room temperature. The activation
energy for recombination of two neighboring Oot species is about
2 eV, which cannot be overcome at room temperature but would
rather need temperatures as high as 400 500 K. However, one can
extract that the 2Obr + 2Oot structure will serve as the active phase
of the catalyst under OER conditions.
4.2. Chloride Activity
The activity of chloride is another point, which we want to
address in this section oder gar subsection. The assumption of
a(Cl ) = 1 in the CER is valid only if using a supporting elec-
trolyte so that the concentration of chloride is nearly invariant
under reaction conditions. However, for the electrochemical HCl
oxidation this assumption is not met as the activity of chloride
is strictly coupled to the proton activity. The resulting Pourbaix
Fig. 5. Pourbaix diagram for RuO2 (110) in equilibrium with Cl , H+ and H2 O at
298 K and a(Cl ) = a(H+ ). The calculations were performed in a (2 1) unit cell of
RuO2 (110) with two cus and two bridge sites. Solvent effects are considered. The
regions where chlorine evolution (CER) and oxygen evolution (OER) are expected
to become signicant have been marked. Color code: red balls: ruthenium 1f-cus
atoms, blue balls: ruthenium 2f-cus atoms and bulk ruthenium, green balls: oxygen
atoms, grey balls: chlorine atoms.
diagram (cf. Fig. 5) is qualitatively equal to the Pourbaix dia-
grams using a(Cl ) = 1 (cf. Fig. 3). Quantitatively, the slope of the
boundary lines is different if chlorine is bound on the catalyst
surface.
4.3. Remarks about the Reaction Mechanism of CER
The discussion about the reaction mechanism of the CER over
RuO2 (110) is reopened. The Krishtalik mechanism proposed by
Hansen et al. [24] is refuted because the lying molecular oxy-
gen (Oot )2 phase is not stable against dissociation so that the
proposed precursor Cl(Oot )2 is not stable in comparison with
OClot adsorbates. The three most frequently discussed reaction
mechanism for the CER, Volmer Tafel [41], Volmer Heyrovsky
[22,42] and Krishtalik [34] appear feasible and none of them
can clearly be ruled out on the basis of the present Pourbaix
diagram. Even the Volmer Tafel mechanism, which implies the
recombination of two adjacent chlorine species on the catalyst
surface, has to be revisited due to the presence of a 2Obr + 2OClot
adsorbate structure in the Pourbaix diagram. The argument
against the Volmer Tafel mechanism is based on the fact that
a reaction order of +1 in chloride was found in kinetic mea-
surements [43,44]. However, if is not the recombination of two
adjacent chlorine species on the catalyst surface but rather the
adsorption of chloride on the 2Obr + 2Oot constitutes the rate-
determining step then the Volmer-Tafel mechanism appears to be
possible.
Our DFT calculations demonstrate that the adsorption of chlo-
rine over on-top oxygen is by 0.7 eV more favorable than adsorption
on oxygen bound at bridge sites. This nding can be explained
qualitatively on the basis of chemical redox equations using formal
oxidation states. Assuming that oxygen of RuO2 is always in II oxi-
dation state and assuming that each O atom in the RuO6 octahedra
attract 2/3 e , the 1f-cus Ru site (denoted as Rucus ) has an oxidation
state of +III 2/3, while this valence state changes to +V 1/3 when an
on-top O is adsorbed. Adsorption of chlorine on Oot reduces the
then oxidation number of Rucus from +V 1/3 to +III 1/3. The elec-
tron transfer occurs via electron hopping as oxygen does not change
its oxidation state. The Rucus atom with the formal oxidation state
of +III 1/3 releases an electron by electron hopping to the inside
of the RuO2 electrode thus switching the oxidation state of Rucus
back to +IV 1/3. The released electron moves through the external
 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
circuit to the cathode where it performs a H+ reduction to molec-
ular hydrogen. This process can be summarized by the following
reaction equation:
Rucus (+V1/3)Oot (II) + Cl Rucus (+III1/3)[O(II)Cl(+I)]ot
Rucus (+IV1/3)[O(II)Cl(+I)]ot + e
For the adsorption of chlorine on bridging oxygen Obr the under-
lying electron processes are similar to those of the adsorption on
Oot . However, the oxidation states of the 2f-cus ruthenium atoms
(denoted as Rubr ) differ from the oxidation states of Rucus atoms:
Rubr (+IV2/3)Obr (II) + Cl Rubr (+II2/3)[O(II)Cl(+I)]br
Rubr (+III2/3)[O(II)Cl(+I)]br + e
Considering the present DFT calculations, the formation of a
Rubr (+II 2/3) species seems to be thermodynamically unfavourable
in comparison with the formation of a Rucus (+III 1/3) species as
the oxidation state +III 1/3 is closer to the oxidation number +IV
of bulk ruthenium dioxide. Altogether, this stabilizing effect on the
RuO2 (110) surface may explain why adsorption of chlorine takes
place selectively on on-top oxygen atoms.
This simple examination of redox properties discloses, however,
an important feature of the chemical nature of the active site. The
catalytically active site in the CER over RuO2 (110) is not Oot alone
but rather the surface complex consisting of Rucus Oot . From the
present study we think that the actual reaction mechanism of the
CER takes place on the Oot saturated RuO2 (110) surface and the
Oot -Cl species will be the reaction intermediate in the CER. For a
further elucidation of the reaction mechanism of the CER we see
the need for in situ spectroscopic studies in order to identify this
reaction intermediates but also the need to perform kinetic studies
using ab initio methods.
4.4. Comparison with gas phase HCl oxidation
A detailed discussion of the heterogeneously catalyzed HCl oxi-
dation (Deacon process) in comparison with the electrochemical
HCl oxidation can be found in Ref. [21]. Therefore, we emphasize
here only those similarities and differences, which are relevant for
the discussion of the Pourbaix diagram.
Hansens calculations did not consider a partial chlorination
of the catalyst in bridge position for the CER, which is, how-
ever, mandatory for the gas phase HCl oxidation. Our calculations
demonstrate that a partial chlorination at the bridge sites during
CER does not occur for potentials above 1 V.
In the potential region where the CER is signicant (above
1.36 V) adsorption on the catalyst surface may be driven by the
applied electrode potential U through the term (e )eU in eq.
(4). Adsorbates, which need more electrons for their formation, are
preferred over adsorbates with a smaller charge transfer, regardless
of the actual adsorption energy difference at zero potential. There-
fore, the absence of partial surface chlorination of the catalyst in
bridge position in the CER region can readily be explained by the
fact that the adsorption of oxygen is a two-electron process, while
adsorption of chlorine displays a one-electron process.
The active catalyst surface for the electrochemical HCl oxida-
tion under reaction conditions constitutes the structure 2Obr + 2Oot .
For the Deacon process the active catalyst is heavily chlorinated
both at bridge and on cus sites. The 1f-cus ruthenium sites and the
undercoordinated surface O atoms constitute the active sites for
HCl activation of the catalyst as HCl dissociation results in chlorine
atoms directly bound to the metal atoms and H atoms are bound
to undercoordinated O atoms [21]. Under electrochemical condi-
tions chlorine cannot bind directly to the metal in the potential
465
region above 1.36 V where the CER happens to occur because the
ruthenium atoms are already fully covered by oxygen.
5. Conclusions
Based on DFT calculations, we constructed surface Pourbaix dia-
grams for the CER over RuO2 (110) using the constrained ab initio
thermodynamics approach. We demonstrated that the topology
of the Pourbaix diagram is determined by the term (e )eU
of eq. (4) in the potential-pH region of the CER. In a typical
Pourbaix diagram the stability of specic surface structures is refer-
enced against water (solvent). For gas evolution reactions such as
OER and CER this reference is not sufcient but rather the con-
sidered surface structure must also be stable against desorption
into the gas phase. From our Pourbaix diagram we infer that the
active electrocatalyst for the CER constitutes an oxygen-terminated
RuO2 (110)-(2Obr + 2Oot ) surface and OClot adsorbates are proposed
to be the precursor for the CER. In order to corroborate the OClot pre-
cursor species for CER in situ spectroscopic studies are needed. The
active surface phase 2Obr + 2Oot is stable under reaction conditions
with respect to the OER as desorption of molecular oxygen from
this surface termination requires temperatures as high as 450 K.
In addition, the 2Obr + 2Oot phase is stable under CER conditions
since chloride ions are adsorbing onto this surface and the release
of molecular Cl2 occurs via direct recombination of neighboring
surface Cl species adsorbed on on-top O or by direct abstraction
of surface chlorine via chloride species from the solution keeping
the underlying Oot saturated RuO2 (110) surface intact. On the basis
of simple redox considerations we suggest that the Rucus Oot com-
plex is the catalytically active site for the electro-oxidation of HCl
over RuO2 (110) consistent with calculated adsorption energies of
chloride ions being 0.7 eV more preferred on on-top oxygen than
on bridging oxygen. These ndings for the CER are substantially
different from those for the Deacon process, the heterogeneously
catalyzed gas phase oxidation of HCl. In the Deacon process chlorine
is directly bound to 1f-cus ruthenium atoms.
Acknowledgements
We appreciated fruitful discussions with Jan Rossmeisl, Marc
Koper, and Jose Conesa. Additionally, we would like to thank Petr
Krtil for organizing the 13th ICEI conference which provides a
stimulating and inspiring environment for discussing in-depth the
reaction mechanism of the CER. Particular thanks go to Franziska
Hess for her great and valuable help in gure design.
Appendix A. Supplementary data
Supplementary material related to this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.electacta.2013.11.027.
References
[1] J.OM. Bockris, S.U.M. Khan, Surface Electrochemistry: Molecular level
approach, Plenum Press, New York and London, 1993.
[2] A. Wieckowski, Interfacial Electrochemistry: Theory, Experiment, and Applica-
tions, Marcel Dekker, Inc, New York, Basel, 1999.
[3] T. Iwasita, ed., Infrared Spectroscopy in Electrochemistry, Electrochim. Acta 41
(1996) Special Issue 41.
[4] A. Wieckowski, M. Neurock, Contrast and Synergy between Electrocatalysis and
Heterogeneous Catalysis, Adv. Phys. Chem. (2011) 907129.
[5] M. Salmeron, R. Schlogl, Ambient pressure photoelectron spectroscopy: A new
tool for surface science and nanotechnology, Surf. Sci. Rep. 63 (2008) 169199.
[6] M. Neurock, Engineering Molecular Transformation for Sustainable Energy
Conversion, Ind. Eng. Chem. Res. 49 (2010) 10183.
[7] J.K. Nrskov, F. Abild-Pedersen, F. Studt, T. Bligaard, Density functional theory
in surface chemistry and catalysis, Proc. Natl. Acad. Sci. USA 108 (2011) 939.
466 K.S. Exner et al. / Electrochimica Acta 120 (2014) 460466
[8] R.A. van Santen, M. Neurock, Molecular Heterogeneous Catalysis: A conceptual
and Computational Approach, Wile-VCH, Weinheim, 2006.
[9] K. Reuter, M. Schefer, Composition, structure, and stability of RuO2(110) as a
function of oxygen pressure, Phys. Rev. B 65 (2001) 035406.
[10] K. Reuter, M. Schefer, First principles kinetic Monte-Carlo simulations for het-
erogeneous catalysis: Application to the CO oxidation at RuO2(110), Phys. Rev.
B 73 (2006) 045433.
[11] A. Farkas, F. Hess, H. Over, Experiment-Based Kinetic Monte Carlo Sim-
ulations: CO Oxidation over RuO2(110), J. Phys. Chem. C 114 (2012)
581591.
[12] F. Hess, A. Farkas, H. Over, First-Principles Kinetic Monte Carlo Simula-
tions Revisited: CO Oxidation over RuO2(110), J. Comp. Chem. 33 (2012)
757766.
[13] F. Studt, F. Abild-Petersen, H.A. Hansen, I.C. Man, J. Rossmeisl, T. Bligard, Volcano
Relations for the Deacon Process over Transition-Metal Oxides, ChemCatChem
2 (2010) 98102.
[14] J.A. Keith, T. Jacob, Theoretical Studies of Potential-Dependent and Compet-
ing Mechanisms of the Electrocatalytic Oxygen Reduction Reaction on Pt(111),
Angew. Chem. Int. Ed. 49 (2010) 95219525.
[15] W. Gao, J.A. Keith, J. Anton, T. Jacob, Theoretical Elucidation of the Competitive
Electro-oxidation Mechanisms of Formic Acid on Pt(111), J. Am. Chem. Soc 132
(51) (2010) 1837718385.
[16] J.K. Nrskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J.R. Kitching, T. Bligaard,
H. Jonsson, J. Phys. Chem. B 108 (2004) 17886.
[17] T. Jacob, Ab-initio Atomistic Thermodynamics for Electrocatalysis, in: M.T.M.
Koper, A. Wieckowski (Eds.), Fuel Cell Catalysis: A Surface Science Approach,
John Wiley&Sons Inc, New Jersey, USA, 2009.
[18] T. Jacob, Theoretical Investigations on the Potential-Induced Formation of Pt-
Oxide Surfaces, J. Electroanal. Chem. 607 (12) (2007) 158166.
[19] P. Kazghazchi, K.A. Soliman, F.C. Simeone, L.A. Kibler, T. Jacob, Bridging the
Gap between Nanoparticles and Single-Crystal Surfaces, Faraday Discuss. 140
(2008) 6980.
[20] M. Pourbaix (Ed.), Atlas of Electrochemical Equilibria in Aqueous Solutions,
NACE, Houston, TX, 1974.
[21] H. Over, Atomic scale insights into electrochemical versus gas phase oxidation
of HCl over RuO2-based catalyst: A comparative review, Electrochim. Acta 93
(2013) 314333.
[22] S. Trasatti, Progress in understanding of the mechanism of chlorine evolution
at oxide electrodes, Electrochim. Acta 32 (1987) 369.
[23] L.A. Kibler, Hydrogen electrocatalysis, Chem. Phys. Chem. 7 (2006) 985.
[24] H.A. Hansen, I.C. Man, F. Studt, F. Abild-Pedersen, T. Bligaard, J. Rossmeisl,
Electrochemical chlorine evolution at rutile oxide (110) surfaces, Phys. Chem.
Chem. Phys. 12 (2010) 283.
[25] SeqQuest website: http://dft.sandia.gov/Quest/SeqQ Home.html
[26] J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made
simple, Phys. Rev. Lett. 77 (1996) 38653868.
[27] J.P. Perdew, K. Burke, M. Ernzerhof, Errata: Generalized gradient approximation
made simple, Phys. Rev. Lett. 78 (1997) 1396.
[28] G. Kresse, J. Furthmller, Vienna ab initio simulation package (VASP), Comput.
Mater. Sci. 6 (1996) 1550.
[29] J.K. Nrskov, T. Bligaard, A. Logadottir, J.R. Kitching, J.G. Chen, S. Pandelov, U.
Stimming, Trends in the Exchange Current for Hydrogen Evolution, Journal of
The Electrochemical Society 152 (3) (2005) J23J26.
[30] Jaguar, version 7. 9, Schrdinger LLC, New York, NY, 2012.
[31] Jaguar User Manual, 2009, http://theochem.chem.columbia.edu/group/
computing
[32] P.W. Atkins, J. de Paula, Physical Chemistry, Fifth completely revised edition
German edition, Wiley-VCH, 2006.
[33] T.E. Madey, W. Chen, H. Wang, P. Kazghazchi, T. Jacob, Nanoscale Surface Chem-
istry over Faceted Substrates: Structure, Reactivity and Nanotemplates, Chem.
Soc. Rev. 37 (10) (2008) 23102327.
[34] L.I. Krishtalik, Kinetics and mechanism of anodic chlorine and oxygen evolution
on transition-metal oxide electrodes, Electrochim. Acta 26 (1981) 329.
[35] D. Crihan, M. Knapp, S. Zweidinger, E. Lundgren, C.J. Weststrate, J.N. Andersen,
A.P. Seitsonen, H. Over, Stable Deacon process for HCl Oxidation over RuO2,
Angew. Chem. 120 (2008) 21612164.
[36] J.P. Hofmann, S. Zweidinger, A.P. Seitsonen, A. Farkas, M. Knapp, O. Balmes,
E. Lundgren, J.N. Andersen, H. Over, Dynamic response of chlorine atoms
on a RuO2(110) model catalyst surface, Phys. Chem. Chem. Phys. 12 (2010)
1535815366.
[37] J.P. Hofmann, S. Zweidinger, M. Knapp, A.P. Seitsonen, K. Schulte, J.N. Andersen,
E. Lundgren, H. Over, Hydrogen-promoted chlorination of RuO2(110), J. Phys.
Chem. C. C 114 24 (2010) 1090110909.
[38] H. Wang, W.F. Schneider, D. Schmidt, Intermediates and Spectators in O2 Dis-
sociation at the RuO2(110) Surface, J. Phys. Chem. C. 113 (2009) 1526615273.
[39] K. Exner, J. Anton, T. Jacob, H. Over, Importance of solvent effects for ab initio
thermodynamic studies, not published yet.
[40] J. Rossmeisl, private communication.
[41] S. Trasatti, W.E. OGrady, in: H. Gericher, C.W. Tobias (Eds.), Advances in Elec-
trochemical Science and Engineering, vol. 12, Wiley, New York, 1981, p. 117.
[42] L.J.J. Jansen, L.M.C. Starmans, J.G. Visser, E. Barendrecht, Mechanism of chlorine
evolution on a ruthenium oxide-titanium oxide electrode and on a ruthenium
electrode, Electrochim. Acta 22 (1977) 1093.
[43] L. Petkov, Electrooxidation of chloride systems, A review, Oxidation Commu-
nications 32 (2009) 654.
[44] M. Thomassen, C. Karlsen, B. Borresen, R. Tunold, Kinetic investigation of the
chlorine reduction reaction on electrochemically oxidised ruthenium, Elec-
trochim. Acta 51 (2006) 2909.