A first principles analysis of CO oxidation over Pt and Pt66.7%Ru33.3% (111) surfaces
-
Upload
sanket-desai -
Category
Documents
-
view
212 -
download
0
Transcript of A first principles analysis of CO oxidation over Pt and Pt66.7%Ru33.3% (111) surfaces
A first principles analysis of CO oxidation over Pt and Pt66.7%Ru33.3%
(111) surfaces
Sanket Desai, Matthew Neurock *
Department of Chemical Engineering, University of Virginia, Charlottesville, VA 22904-4741, USA
Received 2 February 2003; received in revised form 27 May 2003; accepted 29 May 2003
Electrochimica Acta 48 (2003) 3759�/3773
www.elsevier.com/locate/electacta
Abstract
Nonlocal gradient corrected DFT slab calculations were carried out to determine the overall reaction energies along with the
barriers for the activation of water and the oxidation of CO over well-defined Pt(111) and PtRu(111) surfaces. We attempt to
elucidate the features that control the bifunctional mechanism proposed for the oxidation of CO in solution. The addition of Ru to
Pt along with the presence of solution helps to enhance the elementary steps that comprise the bifunctional mechanism. The
activation of water over Pt(111) in the vapor phase is energetically unfavorable with an activation barrier of �/142 kJ/mol and
overall heat of reaction of �/53 kJ/mol. Water will desorb before it will react on Pt(111). The presence of solution reduces the barrier
to about 75 kJ/mol. The addition of Ru lowers the barrier for water activation in the vapor phase. Alloying Pt with Ru lowers the
barriers for the homolytic as well as the heterolytic activation of water in solution. The presence of solution, however markedly
favors the heterolytic activation which leads to the formation of an adsorbed OH* surface intermediate, a proton which migrates
into solution and an electron. The energetics for this path over the Pt66.7%Ru33.3%(111) surface were calculated to be quite favorable
with an activation barrier of �/27 kJ/mol and an overall energy of reaction of �/26 kJ/mol. Ab initio molecular dynamics results
indicate that water can be activated over PtRu at 300 K. The resulting surface Ru-OH group that forms induces the adsorption and
subsequent activation of water at a neighboring Pt site. The surface hydroxyl intermediate continues to diffuse across the surface via
a sequence of proton transfer steps. The barrier for the subsequent oxidation of adsorbed CO by surface OH groups over Pt(111) in
the vapor phase is 86 kJ/mol with an overall reaction energy of 21 kJ/mol. The barrier is reduced to 71 kJ/mol when carried out over
the Pt66.7%Ru33.3% alloy in the vapor phase. The barrier for the homolytic disproportionation of CO and OH in solution (over
Pt66.7%Ru33.3%) is 90 kJ/mol with an overall reaction energy of �/24 kJ/mol. The heterolytic path which involves the oxidation of CO
by OH to form CO2�/H�(aq)�/e�/ is more favorable than the homolytic path. The activation barrier for the heterolytic path over
Pt66.7%Ru33.3% is �/60 kJ/mol with an overall reaction of �/6 kJ/mol (exothermic). These results indicate that at potentials which are
less than or equal to the potential of zero total charge, water activation over Pt(111) may be difficult. CO oxidation, however is more
likely limiting over the Pt66.7%Ru33.3% alloy at these potentials. These results only hold for the ideal conditions studied herein
including ideal single crystal surfaces, an ideal liquid water solution phase and the absence of an applied potential.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: First principles analysis; CO oxidation; DFT slab calculations; Bifunctional mechanism; Solvent effects; PtRu
1. Introduction
Site blocking or poisoning of Pt electrodes by CO at
concentrations of less than 1% presents a critical
obstacle for meeting the desired catalytic activity or
overpotentials at the anode of the direct methanol and
the reformate PEM fuel cells [1�/3]. For direct methanol,
it is difficult to avoid CO, since it is produced either as a
direct intermediate or as a side reaction product [1,4].
Surface poisoning significantly degrades fuel cell per-
formance and is a primary contributor to low power
densities. The oxidative stripping of CO from Pt occurs
at high overpotentials which range from 0.7 to 0.8 V
[5,6]. It is well established that alloying Pt with Ru [2,7�/
9], Mo [7,8,10], Sn [7,8,10,11], Fe [12] and other non-
precious metals including Co, Fe, and Ni [13�/17] can
significantly enhance the tolerance of the anode to CO
poisoning. The addition of Ru lowers the overpotential
* Corresponding author. Tel.: �/1-434-924-6248; fax: �/1-434-982-
2658.
E-mail address: [email protected] (M. Neurock).
0013-4686/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0013-4686(03)00509-7
by anywhere between 0.1�/0.3 V [2,3,5] depending on the
composition and atomic structure of the alloy that
forms. The optimal PtRu alloy is thought to be between
Pt50%Ru50% and Pt80%Ru20% [2�/4,7,8,18] depending onhow the catalyst is synthesized. Pt tends to segregate to
the surface, thus obscuring surface characterization
results. Smotkin recently reported that Ru can phase
segregate out leaving the alloy enriched in Pt at the
surface [19]. They estimate that for a high surface area
catalyst, Pt50%Ru50% may really be closer to a
Pt80%Ru20% alloy at the surface [19]. Two mechanisms
have been proposed to explain the significant shift in theoverpotential as Ru is alloyed into Pt. The first is known
as the ‘‘ligand’’ effect. Ru induces changes in the
electronic structure of Pt which ultimately weakens the
Pt�/CO bond [20�/22]. The second is referred to as the
‘‘bifunctional’’ mechanism [2,3,7,8,23,24]. In the bifunc-
tional mechanism, Ru activates water to form the active
oxidant, i.e. a surface hydroxyl intermediate (Ru�/OH),
which subsequently oxidizes CO bound to a neighboringPt site. The bifunctional mechanism is commonly
written as:
H2O�Ru 0 OH�Ru�H�(aq)�e� (1)
CO�Pt�OH�Ru 0 CO2�H�(aq)�e��Pt�Ru
(2)
These steps are considered overall reaction paths. Thetrue identity of the elementary steps and the active
oxidant are still debated. For example, the active
oxidant may be OH, water or some other form of
adsorbed oxygen. In addition, the oxidation step shown
in Eq. (2) could also occur via the formation of the
COOH* intermediate which would then react to form
CO2, H�/(aq) and e�/. For simplicity, we focus herein
primarily on steps 1 and 2 above.There is theoretical as well as experimental evidence
that indicates that both the ligand and bifunctional
mechanisms can occur under different conditions.
Which of these mechanisms dominates however is still
actively debated. Wieckowski and Masel [5] used 13C-
NMR, temperature programmed desorption spectro-
scopy and cyclic voltammetry to begin to quantify the
relative magnitudes of the ligand and bifunctionaleffects. Their results indicate that the addition of Ru
leads to a total reduction in the overpotential by 170�/
260 mV [5]. Only 40 mV out of this total, however, was
attributed to an electronic effect induced by alloying.
The remaining 130�/220 mV was the thought to be the
result of bifunctional mechanism [5]. Similar ideas have
also been proposed by Iwasita et al. [25].
The addition of Ru lowers the overpotential for COoxidation by 170�/260 mV. The effect of Ru is still
debated in the literature. A number of earlier studies
suggest that water activation over Pt may be rate
limiting especially at lower potentials since water itself
is not activated until 0.7 V. Recent results by Koper et
al. [26], however, indicate that CO oxidation is more
likely the rate limiting step on Pt(111). The results arebased on the analysis of Tafel slope measurements
which were found to be significantly lower than would
be expected if water activation was rate limiting. The
primary role of Ru then is to help drive the equilibrium
from water to its dissociated state. Gasteiger et al. [2]
and others [27,28] have proposed that over PtRu alloys,
the direct oxidation of CO(Pt) by OH(Ru) (Eq. (2)) is the
rate limiting [25]. This suggests that the contact betweenCO(Pt) and OH(Ru) is critical. Koper et al. [27] indeed
showed by kinetic Monte Carlo simulations that the
resulting current densities were markedly increased as
the rate constants for diffusion were increased. There is
still some debate, however, as to whether all of the
reactivity occurs at the boundary [5,27,29] or whether
some of what is seen is due to OH spillover onto Pt [23]
or CO diffusion onto Ru.Advances in surface science, in-situ spectroscopy and
electrochemical have been used to provide an under-
standing of nature of the active surface and its role in
dictating the electrocatalytic kinetics [3,5]. Ab initio
quantum mechanical methods offer a complementary
tool to help elucidate a more complete understanding of
the how the atomic as well as the electronic structure can
dictate the reactivity of metal catalyzed reactions.Modeling electrocatalytic systems, however, presents a
number of challenges for ab initio quantum mechanical
methods. These challenges go beyond the traditional
difficulties for vapor phase metal catalysis which include
the effect of the size and shape of the particle, the
influence of the support, the effect of alloying, and the
interaction between surface species. Modeling electro-
catalytic systems requires the ability to model the effectsof solution, applied potential, and electrolytes. These are
currently active areas of research for a number of
theoreticians but are quite challenging due to their
complexity. Anderson et al. have performed some of
the first theoretical studies using the semiempirical
ASED-MO method [30,31] and, more recently, ab initio
methods [32,33] in order to model electrochemical
surface reactions. In particular, they examined boththe oxidation and reduction of water as well as the
oxidation of CO over Pt and developed an approach by
which to incorporate electric field effects. These results
are quite promising but may be somewhat limited by the
small size of the models used for both the metal
substrate as well as the solution phase.
Herein we focus on the oxidation of CO over Pt and
PtRu alloys. We explicitly probe the effects of carryingthese reactions out in the vapor phase as well as in the
solution phase. We do not, however, attempt to analyze
the effects of an applied potential, electric fields or the
presence of electrolyte. These features can be very
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733760
important in capturing the true kinetics for fuel cell
systems and ultimately need to be explored in detail. In
addition, we only examine the Pt(111) and the
Pt66.7%Ru33.3% alloy surfaces as models. Based onexperimental findings discussed above concerning the
optimal alloy surface composition, the Pt66.7%Ru33.3%
was chosen as an active surface model. Because of the
very idealized reaction environment used herein, the
results can only loosely related to what happens under
actual fuel cell operating conditions. Despite the differ-
ences, we believe that the trends derived from this study
are informative.
2. Calculation details
Self-consistent field (SCF) electronic energy calcula-
tions were performed using plane wave density func-
tional theory as implemented in the Vienna Ab-initio
Simulation Program [34�/36]. A cut off kinetic energy of
25 Rydberg was used to maintain a finite plane wavebasis set. This cut off was found to be sufficient to attain
convergence in both the adsorption and reaction en-
ergies. Electronic states were computed using a 4�/4�/1
k-point grid in the first Brillouin zone. We have verified
that the calculated energetics converge at the chosen k-
point grid density. For calculations performed in the
vapor phase, there is less than a 2 kJ/mol change in the
energy when we increase the k point sampling to 7�/7�/
1. The accuracy is further improved when we move to
the liquid phase calculations. The Perdew Wang (PW91)
functional [37] was used within the Generalized Gradi-
ent Approximation (GGA) to describe non-local ex-
change and correlation effects. Non-local gradient
corrections were found to be important since the local
density approximation (LDA) [38,39] is known to
underestimate the relatively weak interactions such ashydrogen bonds that form between water molecules
[40,41]. Frozen core, scalar relativistic, Vanderbilt ultra-
soft pseudopotentials were used to describe the core
electrons [42].
In order to begin to understand the effects of alloying
we examine the simple closepacked Pt(111) surface and
the Pt66.7%Ru33.3% alloy (66.7% Pt, 33.3% atomically
mixed Ru) as model surfaces. The alloy favors a facecentered cubic structure with an optimal lattice constant
of 2.78 A. Three parallel layers, each with a Pt:Ru ratio
of 2:1, were chosen to describe the thickness of the metal
slab. The top two layers were allowed to relax during
geometry optimization. The metal atoms in the bottom
layer are not directly involved in the reaction and were
therefore held fixed at their bulk positions. The calcula-
tions were performed within a 3�/3-unit cell, with aspacing of 10 A to separate the metal slab and its
periodic images in the direction normal to the surface
plane. In order to model the solution phase, the unit cell
was packed with 23 water molecules, corresponding to a
solution density of �/1 gm/cc. The inter- and the intra-
molecular geometries of the water molecules were
relaxed during the calculation to allow for the formationof hydrogen bonds. Optimization led to the formation
of a hydrogen bonding network.
Adsorption energies (DEADS) were all calculated using
the following equation:
DEADS�Eadsorbate=metal�Eadsorbate�Emetal slab (3)
where, Eadsorbate/metal, Eadsorbate and Emetal slab are the
electronic energies of the adsorbate bound to the metal
surface, the isolated adsorbate, and the bare metal slab
respectively.
Reaction energies (DER�N) were calculated similarly
as the difference in the electronic energies of theproducts and the reactants:
DER�N�Eproducts�Ereactants (4)
where, Eproducts and Ereactants are the electronic energy ofthe products and the reactants respectively.
The convention used here is that a positive value of
DER�N indicates an endothermic reaction, whereas a
negative value refers to an exothermic and hence
thermodynamically favorable reaction.
The reaction paths are a function of the optimized
reactant and product states. The reactants and products
chosen were those which correspond to the lowestenergy states. Despite this assumption, the optimization
scheme can sometimes overcome local energy minima to
find lower energy states.
Transition states were all first isolated using the
nudged elastic band approach developed by Jonsson et
al. [43]. The nudged elastic band method interpolates a
series of images between the reactant and product states.
Each image is then optimized in parallel to map theminimum energy reaction path. For each image and
every iteration, the forces acting on the nuclei are
decomposed into components parallel and perpendicu-
lar to the reaction path. In order to prevent the images
from rolling over to either the reactant or the product
states (local minima on the potential energy surface)
during optimization, the component of the force parallel
to the reaction path is set to zero. The nuclear positionsare, however, fully optimized along the normal to the
reaction path. This approach helps converge the images
to the minimum energy path. A spring interaction
between the images is added to maintain the spacing
between adjacent images to a constant. The highest
point along the minimum energy path provides an initial
transition state structure for the chosen reaction path.
Quasi-Newton optimization was subsequently per-formed on this initial state to help converge and isolate
the transition state. The system size is too prohibitive to
carry out full numerical second derivative calculations
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3761
to establish the frequencies on the isolated state to help
confirm their identity. Instead, the structures were
slightly perturbed from their transition states. They
fall back to either the reactant or product statesdepending on the initial perturbation.
As discussed earlier, our model of the reaction
environment is overly simplistic whereby we analyze
the effects of solution and a single well-dispersed alloy
structure. We ignore, for the time being the effects on an
applied potential. Our results would be most compar-
able to those found at the potential of total zero charge
in an electrochemical cell. In the following sections wediscuss the results for both vapor phase as well as
solution phase activation of water and CO oxidation.
3. Results and discussion
3.1. Dissociation of water
3.1.1. Vapor phase
3.1.1.1. Pt(111). In the vapor phase, water adsorbs
atop a single Pt atom through its oxygen atom and is
tilted off the axis normal to the surface with an energy of
30 kJ/mol. At higher coverages the adsorption energy
can increase up to 40�/50 kJ/mol due to hydrogen
bonding. Water dissociates over Pt(111) homolytically
leading to formation of metal hydride and metalhydroxyl intermediates. The reactant and product states
were fully optimized. Atomic hydrogen shows a very
slight preference for the three-fold fcc site. The differ-
ence between the atop, bridge and three fold sites,
however, is less than a 1 kcal/mol. This is consistent with
other studies on the adsorption of hydrogen on Pt
[44,45]. The path is essentially the same as that for water
activation over the Pt33.3%Ru66.7% (111) surface which isshown in Fig. 1 and discussed in the next section. The
reaction involves a direct metal atom insertion into the
O�/H bond, thus resulting in a three center (H�/Pt�/O)
transition state. The transition state is quite late along
the reaction channel whereby there is a considerable
stretch of the O�/H bond. The hydride intermediate that
forms adsorbs at the three-fold fcc hollow site whereas
the hydroxyl intermediate adsorbs atop of a Pt atomsite. The barrier for this reaction over Pt(111) is quite
high at �/142 kJ/mol. Under UHV conditions, water
would preferentially desorb before it would ever un-
dergo dissociation. This is consistent with known
experimental results [46,47].
3.1.1.2. PtRu alloy. By alloying the Pt(111) surface with
progressively more Ru we find that the barrier toactivate water in the gas phase is successively decreased.
The barrier over the pure Ru(0001) surface was found to
be �/92 kJ/mol which is 50 kJ/mol lower than pure
Pt(111). It is interesting to note that even if only one out
of every three metal atoms is Ru (Pt66.7%Ru33.3%), the
barrier is still reduced by 37 kJ/mol (�/105 kJ/mol). The
reaction coordinate which is depicted in Fig. 2 is
essentially identical to that described for water activa-
tion over Pt(111) as discussed above. The overall energy
for reaction is endothermic by �/53 kJ/mol. The
Fig. 1. Reaction coordinate for the activation of water over the
Pt66.7%Ru33.3%(111) surface. (A) Adsorbed reactant state (DEads�/�/34
kJ/mol), (B) transition state (Eact�/�/105 kJ/mol), and (C) product
state (DEP�R�/�/53 kJ/mol). The Ru atom inserts into the O�/H bond
of water, forming a three-center complex.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733762
energetics and mechanism of the reaction are affected by
the presence of a solution environment.
3.1.2. Solvent phase dissociation of water
3.1.2.1. Pt(111). In solution, water remains bound atopa Pt atom through a weak van der Waals interaction and
thus remains stable on the surface. Hydrogen bonding,
however, now begins to play a more important role in
solvating the surface layer. The water molecules ad-
sorbed at the surface are now nearly planar with the
surface. In this configuration, they can begin to max-
imize hydrogen bonding between the hydrogen atoms on
the adsorbed water molecules and the oxygen atomsfrom the water molecules in solution. Water bound to
the metal begins to take on a bilayer like structure.
Water is bound flat to the surface through its oxygen
whereas neighboring solution phase water molecules
hydrogen bond to these species but also interact with the
surface through one of their hydrogen atoms which is
directed toward the surface. This forms a water over-
layer which subsequently forms hydrogen bonds withadditional water molecules in solution as shown in Fig.
2. The structural arrangement of water close to the
surface is similar to that proposed by Doering and
Madey for multilayer adsorption of water over
Ru(0001) [48]. The network of hydrogen bonds between
the water molecules extends within the unit cell as wellas across adjacent unit cells, thereby simulating the
effect of an aqueous environment.
In the vapor phase, water was found to dissociate
homolytically thus forming surface hydride and hydrox-
yl intermediates (Eq. (5)). The presence of solution,
however, opens up a second possible reaction pathway
which involves the heterolytic activation of water to
form a proton (which migrates into the solution phase)an OH surface intermediate and an electron (Eq. (6)).
H2O�0 OH��H� (5)
H2O�0 OH��H� (aq)�e� (6)
The heterolytic activation of water is possible because of
the stabilization of protons by water.
The heterolytic activation of water over Pt(111) was
found to be the more favorable of the two paths. The
activation baffler for the heterolytic path is �/75 kJ/mol.
The overall reaction energy was also found to be �/75
kJ/mol thus reflecting that the potential energy surfacein the region of the transition and product states is quite
flat. Charge transfer appears to occur at the transition
state. Once the charge transfers the proton can readily
Fig. 2. Optimized structure of water adsorbed on the Pt66.7%Ru33.3%(111) surface in the presence of an aqueous environment. The arrows represent
the path along which the adsorbed water prefers to dissociate. For the sake of clarity, only the water molecules involved in the chemistry are
highlighted. Hydrogen bonds are shown by dotted lines.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3763
move from the transition state structure into solution.
The potential energy surface therefore increases up to
the transition state but then remains relatively flat
whereby the proton is stabilized by the surroundingsolution. This is consistent with the literature on proton
migration through solution where the energy barriers
are only on the order of 2.7 kcal/mol. Ab initio
molecular dynamics studies show the formation and
breakup of various different water clusters. This is
attributed to the relatively shallow potential energy
surface that forms due to the hydrogen bonding net-
work. This is similar to recent findings by Marx et al.[49,50] who indicate that proton transfer in pure water
can be quite rapid. There is a low activation barrier for
hydrogen transfer as the result of the shallow potential
due to the local hydrogen bonding.
While the presence of solution helps to markedly
lower the activation barrier and the overall endothermi-
city, this step is still thought to be rate-limiting under the
conditions examined here. This is qualitatively consis-tent with experimental results which suggest that at
much lower potentials the activation of water is rate
limiting. In addition, these results agree with those
recently reported by Anderson [33] for the activation
of water over a 2 atom Pt cluster.
3.1.2.2. PtRu alloy. Both reactions (5) and (6) were
found to be thermodynamically more favorable having
lower activation barriers than the dissociation of waterin the vapor phase. The reaction leading to the forma-
tion of adsorbed surface hydride and hydroxyl inter-
mediate was found to be endothermic by �/43 kJ/mol,
while that leading to the formation of protons in the
aqueous phase and an adsorbed OH- intermediate
(which subsequently transfers its charge to the metal)
was significantly less endothermic at �/26 kJ/mol.
Transition state calculations indicate that the activationbarrier for the homolytic pathway involving the forma-
tion of a surface hydride (Eq. (5)) is �/90 kJ/mol while
that for the heterolytic path involving the formation of
an aqueous proton is only �/27 kJ/mol. In the presence
of solution, water is more likely to dissociate to form an
OH- surface intermediate and a proton rather than
surface hydroxyl intermediate and hydrogen adatom
thus favoring the heterolytic path. The results describedhere are for the potential of zero total charge. Qualita-
tively they can be extended to more positive potentials.
At potentials which are more negative, the hydride form
becomes more favorable.
The hydroxyl intermediate that forms sits atop of a
Ru atom site. It is partly stabilized by the solvent and
partly stabilized by the metal surface. The electron that
is produced upon the heterolytic dissociation is deloca-lized over the oxygen as well as the metal surface. The
proton, on the other hand, desorbs from the surface into
solution whereby it can be stabilized by a shell of solvent
molecules. Static calculation indicate that the proton is
most stable as the H5O2� Zundel intermediate in
solution which resides near the metal surface but is
clearly stabilized by hydrogen bonding with solution(Fig. 3). Only one water molecule separates the proton
from the surface hydroxyl group. Subsequent dynamics
simulations, however, show that proton is quite mobile
and can take on the H5O2� and the H4O9
� forms as well
as additional hydrogen bonding states. This is consistent
with recent results reported by Marks et al. [50] who
indicate that the proton defect is delocalized over
different states in the hydrogen bonding network andfluctuates between H5O2
�, H4O9� as well as other
intermediates.
The transition state structure for the dissociation of
water in the presence of solution is shown in Fig. 4. In
the reactant state, water adsorbs atop of Ru and forms
hydrogen bonds with neighboring water molecules. The
reaction coordinate involves the stretch of the OH bond.
The presence of solution here helps to stabilize theheterolytic splitting water. The aqueous solution pro-
vides a much better medium than the metal surface to
stabilize the formation of an Hd� intermediate. The
proton therefore preferentially transfers directly into
solution. At the transition state, the proton that is
formed as the result of the activation of water is
transferred to a neighboring water molecule in solution.
The solution molecule in turn transfers one of itshydrogen atoms to the next adjacent water molecule.
The dissociation of water into hydroxyl intermediates
and protons thus involves a concerted migration of
protons from the adsorbed water molecule into solution
along the hydrogen-bonded water network. The water
molecules which make up the hydrogen bonding net-
work provide a direct route for proton conduction away
from the adsorbed water. This is analogous to theGrotthus mechanism for proton transfer through solu-
tion and has also been suggested by Koper [51] for the
activation of adsorbed water over Rh.
In the vapor phase, the dissociation of water was
found to be mediated by the Ru surface, whereby
ruthenium activates the O�/H bond. Both, the hydroxyl
and hydrogen intermediates that form are stabilized by
the surface. In the presence of the solution, theanalogous metal-mediated route, which leads to the
formation of surface bound hydroxyl and hydride
species (Eq. (5)), was found to be enhanced by the
presence of water. Solution molecules interact with each
other only via hydrogen bonds. The solution structure
can readily rearrange during the course of the reaction
to help stabilize the hydrogen as it dissociates from
adsorbed water to the Ru surface. The solution plays anindirect role in that it helps to stabilize the partial
charges on the transition state complex. This effect
however is rather weak as the barrier is lowered by only
15 kJ/mol from that in the vapor phase. The preferred
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733764
route in solution is the heterolytic path, which involves
the direct participation of the solution molecules and
leads to the formation of surface hydroxyl group aproton and an electron (Eq. (6)). The metal is involved
in the initial activation (weakening) of the O�/H bond of
water. Dissociation, however, involves the cooperative
coordination between the metal and water solution
whereby an electron is transferred to the metal along
with the OH surface intermediate, whereas the proton
stabilizes into the water solution.
3.1.3. Dynamics of water on PtRu
In order to capture the influence of the fluctuating
nature of the solution at temperatures greater than 0 K
and associated entropic effects, we carried out ab initiomolecular dynamics simulations for liquid water over
the Pt66.7%Ru33.3% alloy. The dynamics were performed
within the canonical NVT ensemble for up to 2 ps in
time with a time step of 1 fs. The temperature was
maintained near constant at 300 K using a Nose-Hoover
thermostat [52,53]. The other computational parameters
were described earlier in Section 2.
The dynamic simulations were started using theoptimized structures found from the static water/metal
interface calculations. Snapshots of the system at
various time intervals in the simulation are shown in
Fig. 5. Water preferential adsorbs only at the Ru sites.
At about 0.8 ps of the dynamics run, one of the
adsorbed water molecules begins to heterolytically
dissociate, transferring one of its hydrogen atoms
directly into solution. The dissociation leads to the
formation of a solvated proton, and a surface hydroxyl
intermediate centered at the initial Ru adsorption site.
The negative charge is delocalized over the top layer of
the metal surface, particularly near to the Ru adsorption
site, and the adsorbed hydroxyl intermediate. The
reaction products formed and the reaction pathway
are in agreement with the predictions from static
calculations.
The dissociation of water at the Ru site acts to induce
coadsorption of water at a neighboring Pt site. The
OH� intermediate at the Ru site abstracts a proton
from the water that has adsorbed on the Pt site. As a
result, the Ru�/hydroxyl intermediate is converted into
Ru�/water, whereas the Pt�/water molecule is converted
into Pt�/hydroxyl. The hydroxyl species thus diffuses
from Ru to Pt. The hydroxyl intermediate subsequently
diffuses to another neighboring Ru site during the
course of the dynamics via a similar proton transfer
path.
This suggests that the diffusion of the surface
hydroxyl intermediate is not due to Brownian motion
Fig. 3. DFT optimized structure of the reaction products formed from the heterolytic dissociation of water. The hydroxyl intermediate is bound atop
Ru whereas the proton is solvated in water (DEP�R�/26 kJ/mol).
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3765
along the surface, but instead the result of proton
transfer between adsorbed water molecules and the
hydroxyl intermediates as is shown schematically in
Fig. 6. This may be important in describing the kinetics
for CO oxidation. The results suggest that surface
hydroxyl groups can readily migrate from Ru to Pt
provided that there is a free path along the Pt sites.
Water initially adsorbs and activates on Ru. It can
continue to transfer along Pt as long as there are vacant
sites nearby where water can adsorb and provide
additional protons for abstraction. Our calculations
indicate the overall reaction energy for proton transfer
from water adsorbed on an adjacent Pt atom to the Ru�/
OH intermediate (Pt�/OH2�/Ru�/OH0/Pt�/OH�/Ru�/
OH2) is nearly thermoneutral at 5 kJ/mol.
This path of proton transfer as a potential mechanism
of surface diffusion may be of direct relevance to the fuel
cell chemistry where CO oxidation is thought to occur at
the Pt�/Ru boundaries [18,54]; CO forms islands on Pt
whereas water is adsorbed and activated on Ru. As CO
at the periphery is oxidized to CO2, more CO begins to
diffuse from the interior to the Pt�/Ru boundaries.
Recent evidence by Davies et al. [23] suggests that in
addition to the reaction at the Pt�/Ru boundary, there
may be spill-over of the oxidant onto Pt. Our results
indicate that OH groups can migrate onto Pt but they
require vacant Pt sites to provide a diffusion path. In
addition, diffusion must be significantly faster than
adsorption and desorption. Both of these are significant
assumptions and need to be further explored before fullyaccessing the relevance of these dynamic simulations to
fuel cell catalysis.
3.2. CO�/OH disproportionation over Pt�/Ru
The second step of the bifunctional mechanism
involves the disproportionation of CO with the hydroxylintermediates. Again, we examine the chemistry over
Pt(111) and the well-dispersed Pt66.7Ru33.3(111) alloy in
the vapor phase as well as in the solution phase in order
to probe the effects of alloying and solvation.
3.2.1. Vapor phase surface reaction
3.2.1.1. Pt(111). The most favorable adsorption sites
for coadsorbed CO and OH on Pt(111) are on adjacent
bridge and atop sites, respectively. These sites are
directly across from one another and form a 3-fold
surface site. It is well known that DFT tends to predict
that CO will adsorb at the higher fold coordination sites
whereas experiments show that CO binds atop [58]. Thecalculated energy differences between these sites, how-
ever, are actually quite small. CO and OH can subse-
quently react to form CO2 and surface hydrogen. The
Fig. 4. Transition state structure for the heterolytic dissociation of water into adsorbed hydroxyl intermediates and protons (Eact�/27 kJ/mol).
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733766
reaction therefore involves both the formation of an
OC�/OH bond as well as the dissociation of the O�/H
bond. If the reaction only involved the formation of the
OC�/OH bond then one might suspect that Pt (and other
metals to the right in the periodic table) would be more
active than Ru since the weaker metal-adsorbate bonds
should tend to favor coupling reaction paths. Ru,
however, aids the scission of the OH bond just as it
does in the activation of water. The CO�/OH reaction
proceeds by the simultaneous displacement of the
adsorbed CO and OH intermediates toward one another
along with the metal atom insertion into the OH bond.
The barrier for this reaction in the vapor phase over
Pt(111) was found to be �/86 kJ/mol whereas the overall
reaction energy was endothermic by �/21 kJ/mol. This is
in very good agreement with a value of 85 kJ/mol
estimated from the experimental TPD results by Weibel
et al. [55]. They found a peak in the desorption spectra
at 340 K which they attributed to CO2 formation from
the reaction of CO and OD on Pt(111). The barrier for
CO�/OH (or OD) in the vapor phase is significantly
lower than that calculated for the activation of water
over Pt(111).
The CO�/OH disproportionation over Pt(111) in
solution proceeds by a heterolytic path thus forming
CO2*�/H*(aq) and e(�/) as products. The barrier for
the heterolytic path was found to be �/79 kJ/mol. The
overall reaction energy was calculated to be �/22 kJ/mol
Fig. 5. Snapshots taken from an ab-initio molecular dynamics simulation performed at 300 K. (a) shows the initial structure wherein water is
adsorbed over the Ru sites on the surface. (b) shows the Ru-bound hydroxyl intermediate and the solvated proton that form on the dissociation of
water. (c)�/(f) indicate the migration of the hydroxyl intermediates over the surface from Ru (shown as darker atoms on the surface) to Pt (shown as
the lighter atoms) to Ru sites via proton transfer.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3767
endothermic. CO, however, can also react with an
adsorbed hydroxyl to form the COOH surface inter-
mediate. This reaction, shown in Eq. (7), is only 10 kJ/
mol less favorable than the reaction to form CO2�/
H��/1e�.
CO��OH�0 COOH� DER�N��4 kJ=mol (7)
With such a small change in the energy difference, it is
difficult to distinguish which of these paths (Eq. (2) or
Eq. (7)) would be more prevalent under electrocatalytic
conditions.
3.2.1.2. PtRu. On the clean Pt66.7%Ru33.3% alloy sur-
face, OH preferentially adsorbs atop Ru while COprefers the Pt bridge site (Fig. 7a). In the vapor phase,
the O�/H bond is activated by the direct participation of
the metal. The reaction coordinate here also involves the
simultaneous formation of an OC�/OH bond along with
the simultaneous scission of the O�/H bond as is shown
in Fig. 7B. Ru inserts into the O�/H bond, thus
elongating the O�/H bond from 0.99 A in its reactant
state to 1.40 A at the transition state. The products ofthe reaction in the vapor phase are CO2 and a surface
bound hydrogen atom. CO2 weakly interacts with the
surface in a di-s adsorption mode through its carbon
and oxygen atoms. The resulting atomic hydrogen binds
at the three-fold hollow site on the surface (Fig. 7c). The
overall homolytic reaction was found to be endothermic
by�/11 kJ/mol and has an activation barrier of �/71 kJ/
mol. In the vapor phase, the energetics are only slightlymore favorable on the alloy than on pure Pt. The
energetics and the mechanism, however, may change in
the presence of the solution.
3.2.2. Solution phase surface reaction
The favored adsorption sites for the CO and OH
intermediates in solution are found to be the same as
those in the vapor phase*/CO prefers the Pt�/Pt bridge
site whereas OH binds atop Ru (Fig. 8). An important
point to note is that the surface hydroxyl intermediate
can form hydrogen bonds with the water solution thus
stabilizing the reactant state.
The disproportionation between the coadsorbed CO
and OH surface intermediates in solution can either
occur homolytically*/leading to the formation of CO2
Fig. 6. Schematic representation of pathways for the dissociation of
water and the subsequent migration of hydroxyl intermediates over the
surface. The arrows represent the paths for proton transfer.
Fig. 7. CO oxidation over Pt66.7%Ru33.3%(111) in vapor phase (a) CO
and OH coadsorbed on Pt66.7%Ru33.3%(111) surface (b) Transition
state for CO oxidation (Eact�/�/71 kJ/mol) (c) Products CO2 and H
(DER�N�/�/11 kJ/mol).
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733768
and surface hydrogen (Eq. (8)) or heterolytically*/
leading to the formation of CO2, a proton and an
electron (Eq. (9)).
CO��OH�0 CO2��H� (8)
CO��OH�0 CO2��H��e� (9)
The former is analogous to the reaction that occurs inthe vapor phase; the latter is possible in the solution
phase because of the energy gain in the stabilization of
protons in water.
We have examined the energetics associated with both
of these reactions in solution. The homolytic reaction
that leads to the formation of CO2 and a surface hydride
(Eq. (8)) was calculated to be endothermic by �/24 kJ/
mol and has an activation barrier of �/90 kJ/mol. Thisreaction is 13 kJ/mol less favorable than the vapor phase
reaction, likely because there is a loss in the solvation of
the polar hydroxyl group in the reactant state that is not
compensated for by the formation of the surface bound
products when carried out in solution.
In the heterolytic reaction path (Eq. (9)), CO and OH
react at the metal�/solution interface to form weakly
adsorbed CO2, an electron which is delocalized over themetal (in an actual electrocatalytic system, the electron
would instead be drawn out of the metal as current)
whereas the proton would migrate into solution (Eq.
(9)). This heterolytic solution phase coupling reaction is
slightly exothermic at �/6 kJ/mol thus making it more
favorable than the homolytic vapor phase reaction (�/11
kJ/mol) and the homolytic solution phase surface
reaction (�/24 kJ/mol). This is primarily due to the
greater degree of solvation of the protons as compared
to that of chemisorbed hydrogen. The activation barrier
for the heterolytic reaction is �/60 kJ/mol, which is 30
kJ/mol lower than that for the homolytic solution phase
reaction and 11 kJ/mol lower than that for the vapor
phase path. The disproportionation of CO and OH at
the solution�/metal interface is therefore more likely to
form CO2 and a proton rather than CO2 and a surface
hydrogen. The CO2 that forms adsorbs di-s at the
surface through the carbon and oxygen atoms. If we
allow the proton to remain in the local vicinity of the
CO2, it will react with the CO2d� surface intermediate to
form the COOH* surface species. This path is only
favorable when the electron is not pulled out of the
metal fast enough. The proton resides in solution as an
H5O2� where it is solvated by two water molecules (Fig.
9). There is experimental evidence as well as very recent
theoretical results over Pt [56] which indicate that
COOH and formic acid are possible intermediates or
products [18]. Anderson [56] suggests that the COOH
intermediate can subsequently react with a second
Fig. 8. Reactants CO and OH co-adsorbed over well-dispersed Pt�/Ru alloy. CO prefers to adsorb at the Pt�/Pt bridge site. OH adsorbs atop Ru.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3769
surface hydroxyl intermediate to form CO2 and H�/
(aq). This, however, presumes that there are two
adjacent hydroxyl groups next to the bound CO
intermediate or that OH migration is rapid.
The reaction at the solution�/metal interface is similar
to that at the vapor�/metal interface in that it involves
the simultaneous coupling of the carbon of CO with the
oxygen of the OH intermediate, and the activation of the
O�/H bond of the hydroxyl intermediate. In the presence
of the solution, the latter was found to occur by the
direct involvement of the solution phase water mole-
cules. During the reaction, the O�/H bond of the
hydroxyl group begins to stretch. Here, however, it is
aligned toward the solution where it can form hydrogen
bonds rather than toward the surface. The transition
state associated with the reaction is shown in Fig. 10.
The OH group dissociates heterolytically. The oxygen is
transferred to the adsorbed CO to form a surface CO2
intermediate. The proton, on the other hand, transfers
to a neighboring water molecule in solution, which in
turn transfers one of its hydrogen atoms to the next
adjacent water molecule in the form of a proton. The
disproportionation in solution thus occurs by the
concerted movement of protons from the hydroxyl
group into the solution along the hydrogen bonds of
water. The heterolytic reaction appears to be morefavorable than the homolytic reaction by 30 kJ/mol.
This reaction is less affected by the presence of solution
than the water activation step.
3.3. Comparing reaction paths
A comparison of the of the overall reaction energies
for water activation and CO�/OH disproportionation
over the pure Pt(111) in solution indicates that water
activation is 48 kJ/mol more endothermic. The activa-
tion barriers, however, are very similar. The barrier for
water activation is �/75 kJ/mol whereas the CO�/OH
disproportionation step is �/79 kJ/mol. At the potentialswhich are lower than or equal to the potential of zero
total charge, water activation is likely rate controlling.
We can not make any statements, however, with respect
to higher potentials.
In moving to the Pt66.7%Ru33.3% alloy, the activation
of water in solution drops from �/75 kJ/mol over
Pt(111) to �/27 kJ/mol over the Pt66.7%Ru33.3% alloy
[57]. The CO�/OH disproportionation reaction, on theother hand, decreased by only 19 kJ/mol in moving from
Pt(111) (79 kJ/mol) to the Pt66.7%Ru33.3% (111) alloy (60
kJ/mol). Over the Pt66.7%Ru33.3% alloy, the dispropor-
Fig. 9. The products that result from the CO�/OH disproportionation over the Pt�/Ru alloy in solution include the formation of CO2 and proton
and an electron. (EP�R�/�/6 kJ/mol). CO2 adsorbs di-sigma. Proton resides in solution as H5O2�.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733770
tionation step now becomes the more limiting of the two
steps involved in the bifunctional mechanism. This is
consistent with the results reported by Gastegier et al. [2]and others [27]. The results here agree well with those
reported by Wieckowski and Masel [5] who indicate that
addition of Ru primarily enhances the bifunctional
mechanism. The electronic effects of alloying Pt with
Ru which also promote this reaction were found to be
less important. These comparisons, however, are still
preliminary since the relative energetics for the two
reactions may be altered when electric field effects areincluded.
4. Summary/conclusions
Alloying Pt with Ru significantly lowers the activation
barriers as well as the overall reaction energies for both
steps (water activation and CO oxidation) in the
bifunctional mechanism. The analysis of these reactions
in the vapor phase is useful for setting ideas. The
presence of solution, however, is critical since it not
only influences the overall energetics but also changesthe actual mechanism. The activation of water over
Pt(111) is energetically not very favorable when carried
out in the vapor phase. The activation barrier for water
over Pt(111) is �/142 kJ/mol and endothermic by �/53
kJ/mol which indicates that water would preferentially
desorb before dissociating over Pt under UHV condi-
tions. While the presence of solution significantly lowers
the barrier to dissociate water down to 75 kJ/mol, it is
still considered to be a rate limiting step at potentials less
than or equal to the potential of zero total charge.
Without having explored higher potentials or even the
effect of potential, we can not explicitly comment on
what will happen at higher potentials.Alloying Pt with Ru at a ratio of Pt66.7%Ru33.3%,
significantly lowers the barrier for the dissociation of
water in both the vapor and liquid phases. The barrier
for the heterolytic activation of water over the
Pt66.7%Ru33.3%(111) surface in the presence of solution
decreased to �/27 kJ/mol. The significant drop in the
activation barrier is the result of both alloying as well as
solution effects. Water activation over the
Pt66.7%Ru33.3%(111) surface no longer appears to be
rate limiting. In addition, the preliminary results from
ab initio MD simulations suggest that the presence of
surface hydroxyl groups can induce water to adsorb and
activate at neighboring Pt sites. The migration of surface
hydroxyl intermediates appears to occur via a low
energy proton transfer path. The diffusion of surface
hydroxyl intermediates onto Pt requires a path of vacant
Fig. 10. Transition state structure for the CO�/OH disproportionation over well dispersed Pt�/Ru alloy in solution (Eact�/�/60 kJ/mol).
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3771
Pt sites which would allow water to activate at Ru and
diffuse over Pt ensembles.
The second step in the bifunctional mechanism, the
oxidation of CO, occurs via the disproportionation of
CO and OH. This reaction proceeds homolytically in the
vapor phase over the pure Pt(111) with an activation
barrier of �/86 kJ/mol and an overall reaction energy of
�/21 kJ/mol. The barrier is reduced to 71 kJ/mol when
carried out over the Pt66.7%Ru33.3%(111) surface. Pt aids
in the coupling of CO and OH whereas Ru helps to
activate the O�/H bond. The barrier for this same
homolytic activation path is increased to 90 kJ/mol
when carried out in presence of solution. The reactant
state for the homolytic disproportionation of CO and
OH is stabilized more so than the transition state thus
leading to the increased barrier. In solution, the hetero-
lytic path which results in the formation of CO2�/
H�(aq)�/e(�/) is much more favorable than the homo-
lytic path. The barrier is reduced to 60 kJ/mol. The
transition state involves the formation of a much more
charged separated state (Pt�/CO2d��/Hd�) which is
stabilized by solution much more so than the reactant
state. If the negative charge that forms is allowed to
remain localized near the CO2 it will react with the
resulting proton to form COOH. If, however, the charge
is transferred through the metal the resulting
products are CO2(g), an aqueous phase proton and an
electron. Overall the CO oxidation path is less affected
by the presence of solution than the water activation
path.
The end result is that both alloying and the
presence of solution enhance water activation
as well as CO disproportionation steps. These
effects, however, impact the water activation much
more than the CO�/OH disproportionation step.
The barrier for water activation is lower than that
for CO oxidation over the Pt66.7%Ru33.3%(111)
surface when the reaction is carried out in the presence
of solution. This suggests that for the ideal conditions
examined here, that there may be a change in the rate
limiting step from water activation to CO�/OH dis-
proportionation as we begin to alloy the Pt(111) surface
with Ru.
Acknowledgements
We would like to acknowledge the DuPont Chemical
Company for partial support of this work and would
also like to thank the NCSA Supercomputing Center at
the University of Illinois for the computing resources.
M.N. would like to thank Professor Marc Koper for
invaluable discussions.
References
[1] T.D. Jarvi, E.M. Stuve, in: J. Lipkowski, P.N. Ross (Eds.),
Fundamental Aspects of Vacuum and Electrocatalytic Reactions
of Methanol and Formic Acid on Platinum Surfaces, Wiley-VCH,
1998, p. 75.
[2] H.A. Gasteiger, N. Markovic, P.N. Ross, E.J. Cairns, J. Phys.
Chem. 98 (1994) 617.
[3] S. Wasmus, A. Kuver, J. Electroanal. Chem. 461 (1999) 14.
[4] T. Iwasita, Electrochim. Acta 47 (2002) 3663.
[5] P. Waszczuk, G.Q. Lu, A. Wieckowski, C. Lu, C. Rice, R.I.
Masel, Electrochim. Acta. 47 (2002) 3637.
[6] Z. Jusys, J. Kaiser, R.J. Behm, Electrochim. Acta 47 (2002) 3693.
[7] M. Watanabe, S. Motoo, J. Electroanal. Chem. 60 (1975) 267.
[8] M. Watanabe, S. Motoo, J. Electroanal. Chem. 60 (1975) 275.
[9] H.A. Gasteiger, N.M. Markovic, P.N. Ross, J. Phys. Chem. 99
(1995) 8290.
[10] H. Massong, H. Wang, G. Samjeske, H. Baltruschat, Electro-
chim. Acta 46 (2000) 701.
[11] K. Wang, H.A. Gasteiger, N.M. Markovic, P.N. Ross, J.
Electrochim. Acta 41 (1996) 2587.
[12] H. Uchida, H. Ozuka, M. Watanabe, Electrochim. Acta 47 (2002)
3629.
[13] M. Watanabe, S. Motoo, J. Electroanal. Chem. 110 (1980) 103.
[14] M. Watanabe, S. Motoo, J. Electroanal. Chem. 110 (1980) 261.
[15] M. Watanabe, S. Motoo, J. Electroanal. Chem. 187 (1985) 161.
[16] M. Watanabe, S. Motoo, J. Electroanal. Chem. 194 (1985) 275.
[17] M. Watanabe, S. Motoo, J. Electroanal. Chem. 206 (1986) 197.
[18] A. Kabbabi, R. Faure, R. Durand, B. Beden, F. Hahn, J.M.
Leger, C. Lamy, J. Electroanal. Chem. 444 (1998) 41.
[19] E. Smotkin, Electrochim. Acta, 2002, this issue.
[20] F.B. deMongeot, M. Scherer, B. Gleich, E. Kopatzki, R.J. Behm,
Surf. Sci. 411 (1998) 249.
[21] J.C. Davies, B.E. Hayden, D.J. Pegg, Surf. Sci. 467 (2000) 118.
[22] E. Christoffersen, P. Liu, A. Ruban, H.L. Skriver, J.K. Norskov,
J. Catal. 199 (2001) 123.
[23] J.C. Davies, B.E. Hayden, D.J. Pegg, M.E. Rendall, Surf. Sci. 496
(2002) 110.
[24] H.F. Octjen, V.M. Schmidt, U. Stimming, F. Trila, J. Electro-
chem. Soc. 143 (1996) 3838.
[25] T. Iwasita, H. Hoster, A. John-Anacker, W.F. Lin, W. Vielstich,
Langmuir 16 (2000) 522.
[26] N.P. Lebedeva, M.T.M. Koper, J.M. Feliu, R.A. van Santen, J.
Electroanal. Chem. 524�/525 (2002) 241.
[27] M.T.M. Koper, J.J. Lukkien, A.P.J. Jansen, R.A. van Santen, J.
Phys. Chem. B 103 (1999) 5522.
[28] T.E. Shubina, M.T.M. Koper, Electrochim. Acta 47 (2002) 3621.
[29] K.A. Friedrich, K.P. Geyzers, U. Linke, U. Stimming, J.
Stumper, J. Electroanal. Chem. 402 (1996) 123.
[30] A.B. Anderson, E. Grantscharova, J. Phys. Chem. 99 (1995) 9149.
[31] A.B. Anderson, E. Granscharova, S. Seong, J. Electrochem. Soc.
143 (1996) 2075.
[32] A.B. Anderson, Electrochim. Acta 47 (2002) 3759.
[33] A.B. Anderson, N.M. Neshev, R.A. Sidik, P. Shiller, Electrochim.
Acta 47 (2002) 2999.
[34] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
[35] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.
[36] G. Kresse, J. Furthmuller, Comput. Mater. Sci. 6 (1996) 15.
[37] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R.
Pederson, D.J. Singh, C. Fiolhais, Phys. Rev. B 46 (1992)
6671.
[38] D.M. Ceperley, B.J. Alder, Phys. Rev. Lett. 45 (1980) 566.
[39] J.P. Perdew, A. Zunger, Phys. Rev. B 23 (1981) 5048.
[40] G. Ortiz, P. Ballone, Phys. Rev. B 43 (1991) 6376.
[41] K. Laasonen, F. Csajka, M. Parrinello, Chem. Phys. Lett. 194
(1992) 172.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/37733772
[42] D. Vanderbilt, Phys. Rev. B 42 (1985) 8412.
[43] G. Mills, H. Jonsson, G.K. Schenter, Surf. Sci. 324 (1991) 305.
[44] G. Papoian, J. Norskov, R. Hoffmann, J. Am. Chem. Soc. 122
(2000) 4129.
[45] S.G. Podkolzin, R.W. Watwe, Q.L. Yan, J.J.
dePablo, J.A. Dumesic, Phys. Chem. B 105 (2001)
8550.
[46] P.A. Thiel, T.E. Madey, Surf. Sci. Rep. 7 (1987) 211.
[47] C. Panja, N. Saliba, B.E. Koel, Surf. Sci. 395 (1998) 248.
[48] D. Doering, T.E. Madey, Surf. Sci. 123 (1982) 305.
[49] D. Marx, M. Sprik, M. Parrinello, Chem. Phys. Lett. 273 (1997)
360.
[50] D. Marx, M. Tuckerman, J. Hutter, M. Parrinello, Nature 397
(1999) 601.
[51] P. Vassilev, M.T.M. Koper, R.A. van Santen, Chem. Phys. Lett.
359 (2002) 337.
[52] S. Nose, Mol. Phys. 52 (1984) 255.
[53] S. Nose, J. Chem. Phys. 81 (1984) 511.
[54] H.A. Gasteiger, N. Markovic, P.N. Ross, E.J. Cairns, J. Electro-
chem. Soc. 141 (1994) 1795.
[55] M.A. Weibel, K.M. Backstrand, T.J. Curtiss, Surf. Sci. 444 (2000)
66.
[56] A.B. Anderson, N.M. Neshev, J. Electrochem. Soc 149 (2002)
E383.
[57] S. Desai, M. Neurock, in press.
[58] P.J. Feibelman, B. Hammer, J.K. Norskov, F. Wagner, M.
Scheffler, R. Stumpf, R. Watwe, J. Dumesic, Phys. Chem. B 105
(18) (2001) 4018.
S. Desai, M. Neurock / Electrochimica Acta 48 (2003) 3759�/3773 3773