A computational study of H2 dissociation on silver surfaces: The effect of

oxygen in the added row structure of Ag(110)wzAmjad B. Mohammad,a Kok Hwa Lim,ya Ilya V. Yudanov,ab

Konstantin M. Neymancand Notker Rosch*

a

Received 15th November 2006, Accepted 6th December 2006

First published as an Advance Article on the web 19th January 2007

DOI: 10.1039/b616675j

We studied computationally the activation of H2 on clean planar (111), (110) and stepped (221) as

well as oxygen pre-covered silver surfaces using a density functional slab model approach. In line

with previous data we determined clean silver to be inert towards H2 dissociation, both

thermodynamically and kinetically. The reaction is endothermic by B40 kJ mol�1 and exhibits

high activation energies of B125 kJ mol�1. However, oxygen on the surface, modeled by the

reconstructed surface p(2 � 1)O/Ag(110) that exhibits –O–Ag–O– added rows, renders H2

dissociation clearly exothermic and kinetically feasible. The reaction was calculated to proceed in

two steps: first the H–H bond is broken at an Ag–O pair with an activation barrier Ea B70 kJ

mol�1, then the H atom bound at an Ag center migrates to a neighboring O center with

Ea B12 kJ mol�1.

1. Introduction

Dissociation of molecular hydrogen H2 on metals to form

atomic H species is a crucial step in numerous technologically

important processes; we just mention hydrogen storage.1

Differently from many other metals, H2 dissociation on clean

planar silver surfaces appears to be unlikely according to a

variety of surface science2–6 and theoretical7–9 studies. This

contradicts experimental findings which show silver, well-

known to catalyze oxidation reactions,10 also to be an efficient

hydrogenation catalyst which transforms a,b-unsaturated al-

dehydes to unsaturated alcohols.11 The mechanism of the

latter process is not well understood at the microscopic level

and activation of H2 on pure and correspondingly chemically

modified silver substrates is definitely one of the puzzling key

issues to be clarified. The present study is motivated by this

puzzle, but we addressed the problem of H2 dissociation on

well-defined silver surfaces. With hydrogenation reactions in

mind where oxygen adsorbates may be present (e.g. generated

by aldehyde decomposition12), we also considered model

surfaces with well-defined overlayers of oxygen.

We will briefly summarise relevant results on the interaction

of molecular and atomic hydrogen with silver surfaces. At a

very low temperature, 10 K, only physisorption of H2 was

observed on the Ag(111) surface indicating that H2 dissocia-

tion on silver is an activated process.2 From experimental

results, the binding energy of H2 at that surface has been

estimated at 15 kJ mol�1,4 to be compared with a density

functional (DF) result of 10 kJ mol�1, obtained in the local

density approximation (LDA).7 No H2 adsorption at all was

observed on the Ag(110) surface at 90 K.3 Based on the finding

that H2 does not dissociate on Ag(111) at 100 K the upper

limit for the binding energy of H atoms at this surface was

estimated at 218 kJ mol�1, i.e. half of the H2 dissociation

energy in the gas phase.5 DF calculations at the gradient-

corrected (GGA) level yielded even smaller values of the

binding energy that vary slightly with the exchange–correla-

tion approximation, B200 kJ mol�1 (PW91 functional)13 and

B185 kJ mol�1 (RPBE functional)14 for H adsorption at

3-fold hollow sites of the Ag(111) surface.9 Such an energy

gain by H adsorption is not sufficient for dissociating H2 from

the gas phase onto the Ag(111) surface, which requires an

energy expense of 436 kJ mol�1 (experiment)15 or 439 kJ

mol�1 (DF calculations with PBE16 and PW91 functionals).16

On the Ag(100) surface, a rather high activation barrier of

B105 kJ mol�1 has been determined from DF calculations

with the PW91 functional.8 Also, on the more open Ag(110)

surface dissociative adsorption of H2 is experimentally found

to be a strongly activated process;3 no accurate theoretical

results are yet available for this system.

As H2 cannot be activated under common experimental

conditions on clean low-index surfaces of silver, one has to

discuss alternative activation sites, either sites on higher-index

surfaces, which contain more active low-coordinated metal

centers, or surface sites formed by impurities, e.g. atomic

oxygen.12,17

aDepartment Chemie, Theoretische Chemie, Technische UniversitatMunchen, 85747 Garching, Germany. E-mail: roesch@ch.tum.de;Fax: +49-89-289 13468

b Boreskov Institute of Catalysis, Russian Academy of Sciences,630090 Novosibirsk, Russia

c Institucio Catalana de Recerca i Estudis Avancats (ICREA), 08010Barcelona and Departament de Quımica Fısica i Centre especial deRecerca en Quımica Teorica, Universitat de Barcelona i ParcCientıfic de Barcelona, 08028 Barcelona, Spainw The HTML version of this article has been enhanced with colourimages.z Electronic supplementary information (ESI) available: Calculatedstructural, energetic and vibrational parameters of selected investi-gated systems. See DOI: 10.1039/b616675jy Present address: School of Chemical and Biomedical Engineering,Nanyang Technological University, Singapore 637459.

This journal is �c the Owner Societies 2007 Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 | 1247

PAPER www.rsc.org/pccp | Physical Chemistry Chemical Physics

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online / Journal Homepage / Table of Contents for this issue

The interaction of oxygen with silver has been intensively

studied, due to its important industrial application in the

epoxidation of ethene; yet, the nature of oxygen species under

epoxidation conditions is still debated controversially.10,18–25

Surface science studies under ultra-high vacuum (UHV) con-

ditions provided more detailed structural information. In

particular, an added-row reconstruction of the Ag(110) sur-

face by adsorbed oxygen is experimentally well-estab-

lished:26–30 Ag–O–Ag–O chains develop along the [001]

direction. The interaction between the chains is repulsive and

ordered (n � 1) structures (n = 2–8) have been observed, with

n larger for lower oxygen coverage. This added-row p(n� 1)O/

Ag(110) system appears to be the only surface phase of oxygen

on silver with a well-defined structure. Such Ag–O added rows

are formed on the Ag(110) surface at UHV conditions and

very low oxygen coverages,26,29 thus representing a tempting

model of surface oxygen species that may exist even in oxygen-

deficient atmosphere.

In the present computational study, we compared the

activation of molecular hydrogen on clean Ag surfaces and

on added-row modified Ag(110) surfaces. To examine the

effect of the surface morphology on the activation of hydro-

gen, we first considered the adsorption of H2 on the most

stable surface Ag(111), on the more open surface Ag(110), and

on the stepped surface Ag(221). Then, we went on to char-

acterize the reaction path of H2 dissociation on the Ag–O

chains of the p(n � 1)O/Ag(110) (n = 2, 3) surfaces.

2. Models and computational details

The calculations were performed with the plane-wave based

Vienna ab initio simulation package (VASP)31–33 using the

PW91 functional. The interaction between atomic cores and

electrons was described by the projector augmented wave

(PAW) method.34,35 For integrations over the Brillouin zone,

we combined (5 � 5 � 1) Monkhorst-Pack grids36 with first-

order Methfessel-Paxton smearing technique (smearing value

0.15 eV).37 An energy cut-off of 400 eV was used throughout.

To model (111) and (110) surfaces, we employed unit cells

with four Ag atoms per layer, enabling a surface coverage of

1/4 or higher. For the stepped Ag(221) [4(111) � (111)] surface,

we used unit cells of eight (or twelve) Ag atoms per layer. In

Fig. 1 we show the various adsorption sites studied. A vacuum

spacing of about 1 nm was adopted to separate the periodi-

cally repeated slabs. The adsorbed moieties were placed on one

side of a slab of five Ag layers. Three layers at the ‘‘bottom’’ of

the metallic substrate were kept fixed at the optimized geo-

metry of the bulk material (Ag–Ag = 293 pm), while the

atomic positions of two ‘‘top’’ Ag layers were optimized for

clean substrates and then kept fixed during the subsequent

adsorption studies. We have justified this type of models for

studies of this kind;38,39 see also below. Upon relaxation of the

clean Ag(111) surface, the interlayer distances changed by at

most 2 pm. At variance, the interlayer distance between the

top two layers of the Ag(110) surface decreased by 12 pm and

the distance between the second and third layers increased by

7 pm. For the Ag(221) surface, the second-layer atoms changed

their positions by less than 2 pm. The first-layer atoms showed

larger displacements; the edge atoms moved 6 pm toward the

‘‘bulk’’ while the terrace atoms moved 1–7 pm in the opposite

direction. With these changes, the distances between the first-

layer Ag atoms shortened to 289 pm. Further relaxation of the

top two Ag layers under the influence of adsorbed atomic

hydrogen at various sites of the Ag(110) surface (see Fig. 1)

was found to be insignificant. For instance, the distance H–Ag

changed by 1 pm compared to our standard model where the

substrate was kept at the structure optimized for a free surface.

At a coverage of 1/4, adsorbate-induced relaxation affects the

Ag–H binding energy by less than 2 kJ mol�1.

As already indicated, besides pure Ag substrates, we also

examined the reactivity of silver with respect to H2 dissocia-

tion in the presence of surface oxygen. For this purpose, we

chose the added-row p(2 � 1)O/Ag(110) surface, where the

AgO row structure is formed by formally depositing two Ag

and two O atoms per surface unit cell. Here, too, we used a

five-layer slab to model the Ag(110) substrate (Fig. 2). For

adsorption and reaction studies, the positions of the H ad-

sorbates and those of the AgO added row were allowed to

relax while keeping the Ag(110) substrate fixed at the geometry

optimized for the bare slab model.

Fig. 1 Sketches of the top view of the adsorption sites studied at the

(111), (110) and (221) surfaces of silver. For clarity, only two Ag layers

are shown (Ag1, Ag2). Light shading—top-layer Ag atoms (Ag1);

dark shading—second-layer Ag atoms (Ag2). Ag(111): 1—fcc, 2—hcp,

3—bridge, 4—top; Ag(110): 5—short bridge, 6—3-fold hollow,

7—long bridge, 8—4-fold hollow, 9—top; Ag(221): 10—edge 3-fold

hollow-1, 11—edge bridge, 12—edge 3-fold hollow-2 and 13—terrace

3-fold hollow. The arrow at the bottom of the Ag(221) sketch points

along a step edge.

1248 | Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 This journal is �c the Owner Societies 2007

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

The energy of adsorption, Ead = Etot(ad_sub) – [Etot(ad) +

Etot(sub)], was calculated by subtracting the sum of the total

energies of the adsorbate (H atoms in the gas phase) and of the

clean substrate from the total energy Etot(ad_sub) of the slab

covered by the adsorbate in the optimized geometry. With this

definition, a negative value of Ead implies favorable adsorp-

tion, associated with a release of energy.

The potential energy surface for H2 adsorption is rather flat;

therefore, we first located approximate configurations for the

transition states (TS) of H2 dissociation on the clean Ag(110)

surface with the help of a constrained search, and subsequently

applied the nudged elastic band method.40 We distinguished

different dissociation pathways by the structure of the final

state (FS) which we labeled according to the sites where the

dissociated atoms H ended up (Table 2). For example, the FS

short-bridge/short-bridge 5/5 refers to dissociated H atoms at

two short bridge sites 5 (Fig. 1). The TS search of H2

dissociation on the p(2 � 1)O/Ag(110) surface was directly

carried out with the nudged elastic band method.40 We

checked all TS structures with a normal mode analysis invol-

ving all degrees of freedom of the geometry optimization to

ensure that the adsorption complex exhibited exactly one

imaginary frequency.

3. Results and discussion

3.1 Hydrogen adsorption on clean Ag(111), Ag(110) and

Ag(221) surfaces

To establish a consistent data set for comparison, we started

with calculations of H2 adsorption on the clean surfaces

Ag(111), Ag(110) and Ag(221). We calculated H2 molecules

to be essentially unbound at all these surfaces; the adsorption

energy Ead was at most �2 kJ mol�1. This agrees with the

experimental observations already mentioned (section 1): H2

does not adsorb on Ag(110) at 90 K3 and adsorbs only very

weakly at Ag(111) with Ead = �15 kJ mol�1.4 Common

GGA exchange–correlation functionals are not accurate en-

ough for such weak interactions to allow a more quantitative

comparison.41

In Table 1 we collected our results for pertinent structure

parameters of various adsorption complexes of atomic hydro-

gen on the clean surfaces Ag(111), Ag(110), and Ag(221) as

well as the corresponding energies of adsorption (see also

Fig. 1).

Adsorption of H on the Ag(111) surface. On Ag(111), we

computed adsorption complexes at 3-fold hollow sites to be

most stable and confirmed by normal mode analysis that they

are local minima; we were not able to determine a preference

among fcc 1 or hcp 2 sites. The calculated adsorption energy,

�198 kJ mol�1, is essentially the same as the value obtained in

earlier slab model DF PW91 calculations, about �200 kJ

mol�1,8 and lies (per absolute values) below the experimental

estimate which sets an upper limit of �218 kJ mol�1.5

HREELS (high-resolution electron-energy-loss spectroscopy)

data are most consistent with adsorption complexes at 3-fold

hollow sites.6 Adsorption complexes of H at bridge positions 3

are slightly less stable, with a calculated adsorption energy of

�187 kJ mol�1. Computed vibrations of these complexes

feature one imaginary frequency, indicating that the bridge

sites 3 represent a TS of the diffusion of H between 3-fold

hollows 1 and 2. Top sites 4 are least favorable for atomic H

on Ag(111), with Ead = �150 kJ mol�1; a vibrational analysis

identified this structure as a saddle point with two imaginary

frequencies. The calculated binding energies of H among the

sites 1 to 4 vary by only B50 kJ mol�1 and the estimated

barrier for H diffusion between the local minima 1 and 2 via

the bridge site 3 is merely 11 kJ mol�1. One expects from these

results that atomic hydrogen exhibits a notable mobility on the

Ag(111) surface already at rather low temperatures. The

calculated H–Ag distances of the adsorption complexes H/

Ag(111), ranging from 166 to 192 pm (Table 1), show an anti-

correlation with the coordination number of the adsorbate.

Adsorption of H on the Ag(110) surface. It is somewhat

surprising that we calculated similarly strong binding of atom-

ic H at the short bridge 5 (�196 kJ mol�1) and at the long

bridge 7 (�193 kJ mol�1) of the Ag(110) surface. Both

adsorption complexes are local minima according to vibra-

tional analysis. The corresponding adsorption energies are

very close to the value computed for the 3-fold hollow sites

of Ag(111) (Table 1, Fig. 1). At the long bridge site 7, H is

located essentially at the height of the top layer Ag1 of the

substrate. The distances, H–Ag2 = 200 pm, of adsorbed H to

Fig. 2 Top view (a) and side view (b) of the added-row p(2 � 1)O/

Ag(110) structure. Dark-shaded spheres—O atoms, medium-shaded

spheres—Ag atoms of the added row, light-shaded sphere—Ag sub-

strate atoms. The rectangular area selected in the top view represents a

unit cell. Selected interlayer distances in relaxed structures: d0—height

of the oxygen atoms above the Ag atoms of the added row;

d12—height of Ag atoms of the added row above the top crystal plane

of the Ag substrate; d23—height of the top Ag plane of the substrate

above the subsurface Ag plane. The rectangle indicates the smallest

surface elementary cell considered.

This journal is �c the Owner Societies 2007 Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 | 1249

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

the Ag atoms of the second substrate layer are shorter than

those to the top-layer Ag atoms, H–Ag1 = 208 pm. This

finding may be interpreted as an adsorption complex of H

involving the four Ag atoms that form a long bridge site 7: two

atoms belong to the first layer Ag1 and the other two to the

second layer Ag2 (Fig. 1).

As suggested by experimental findings,3 we also studied

adsorption at the 3-fold hollow site 6 where H is displaced

from the short bridge 5 toward a 4-fold hollow position 8 (Fig.

1), but we were unable to find a stable structure with 1/4

surface coverage of hydrogen. At higher coverage, H atoms

may occupy 3-fold hollow sites 6 to minimize H–H repulsion

as observed experimentally at coverage values of 1/3 or

larger.3 Indeed, according to our calculations this is the case

at a hydrogen coverage of 1/2: the zig-zag ordering of hydro-

gen atoms along the [1�10] direction (corresponding to a

confirmed local minimum) is more stable than hydrogen

at the short bridge structure by 12 kJ mol�1 per H atom

(Table 1).

The 4-fold hollow site 8 binds H somewhat weaker, DEad =

�169 kJ mol�1, but it represents a saddle point with two

imaginary vibrational frequencies. In that adsorption com-

plex, the distance of H to the nearest Ag1 atom on the surface

layer, 256 pm, is much longer than the H–Ag2 contact,

178 pm, to a metal atom of the second layer; this finding

indicates that on this site H interacts predominantly with a

single Ag atom of the second layer of the substrate and the

H–Ag2 distance is just 12 pm longer than that with Ag atom of

the top sites 4 and 9. Similarly to the results for Ag(111), H

adsorption on Ag(110) was calculated to be least stable at the

top site 9, with DEad = �147 kJ mol�1 (again a saddle point

with two imaginary frequencies). In fact, the H–Ag distances

of adsorption complexes with the same coordination of H are

quite similar as well for these two surfaces (Table 1).

The similarities between the calculated results for the

Ag(111) and Ag(110) surfaces extend to the variation of the

calculated adsorption energies, which does not exceed

50 kJ mol�1 on Ag(110). To be more precise in quantifying

the propensity of H atoms to diffuse on the Ag(110) surface,

we identified the TS of H diffusion from the short bridge site 5

to the long bridge site 7 using the nudged elastic band method.

The resulting TS is less stable than the IS by only 14 kJ mol�1,

thus even more stable than the adsorption complex on the

4-fold hollow site 8. Therefore, H diffusion on Ag(110) surface

also appears to be very facile.

Adsorption of H on the Ag(221) surface. On the stepped

surface Ag(221) [4(111) � (111)], we focused on the adsorption

sites 10, 11 and 12 which are located at the step edges. When

selecting these sites, we relied on findings that less coordinated

edge atoms commonly form more active sites.38 In addition,

we considered the terrace 3-fold hollow site 13 as a reference to

characterize the mobility of the adsorbates. We confirmed that

all these sites correspond to local minima. Atomic H at the

edge hollow sites 10 and 12 as well as at the edge bridge sites 11

features very similar energy of adsorption, �198 to �202 kJ

mol�1 (Table 1, Fig. 1). These values are close to those

calculated for the most favorable sites of the surfaces

Ag(111) and Ag(110). The terrace reactivity of the Ag(221)

surface is similar to that of the (111) surface. This is mani-

fested by the energy of H adsorption on the terrace 3-fold

hollows 13, �193 kJ mol�1 that essentially matches the energy

of atomic H at the 3-fold hollow sites 1 and 2 of the Ag(111)

surface. Note that the interaction of H with the terrace site 13

is only slightly weaker than with the edge sites 10 to 12.

Consequently, we expect facile mobility of atomic H also on

this stepped Ag(221) surface, just as on the Ag(111) and

Ag(110) surfaces. Inspection of Table 1 reveals that H–Ag

distances are very similar to those of sites with the same

coordination on the other surfaces studied. Reducing the

surface coverage to 1/12 of a monolayer does not reveal any

noticeable change in the adsorption geometries and binding

energies of hydrogen atoms (Table 1).

Activation of H2 on the Ag(110) surface. Before addressing

the activation of H2 molecules on clean silver surfaces, it is

instructive to note that the present computational approach

reproduces pertinent characteristics of H2 in the gas phase,

Table 1 Parameters of adsorption complexes of atomic H on high-symmetry sites of various Ag surfacesa

Substrate Siteb H–Ag/pm Ead/kJ mol�1

Ag(111) 1 fcc 192 �1982 hcp 192 �1983 bridge 183 �1874 top 166 �150Exp.c �218

Ag(110) 5 short bridge 181 (180d) �196 (�192d)6 3-fold hollowd 185, 201e �2047 long bridge 208, 200e �1938 4-fold hollow 178 �1699 top 166 �147

Ag(221) 10 edge 3-fold hollow-2f 191, 200e (192, 200)h �202 (�198)h11 edge bridge 182 (183)h �199 (�195)h12 edge 3-fold hollow-1g 194, 194e (195, 195)h �198 (�192)h13 terrace 3-fold hollow 190, 196e (190, 196)h �193 (�186)h

a Energies of adsorption Ead (a negative sign implies favorable adsorption, see section 2) and lengths of the shortest H–Ag bonds. Unless otherwise

stated, the coverage of H was 1/4 of a monolayer, except for Ag(221) where the standard coverage was 1/8. b See Fig. 1. c Ref. 5. d Surface

coverage 1/2. e The adsorption site comprised two non-equivalent Ag atoms. f Formed by two edge and one terrace Ag atoms. g Formed by one

edge and two terrace Ag atoms. h Surface coverage 1/12.

1250 | Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 This journal is �c the Owner Societies 2007

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

namely the binding energy, 439 kJ mol�1 vs. 436 kJ mol�1

(exp.) and the H–H bond length, 74 pm (calc. and exp.).2,15 As

already mentioned, dissociative adsorption of H2 at the three

Ag surfaces studied is an endothermic process, requiring about

40 kJ mol�1. The two more open surfaces Ag(110) and

Ag(221) do not exhibit any propensity for a more favorable

thermodynamics of H2 dissociation than the close-packed

surface Ag(111). Therefore, we explored in detail this process

at clean silver surfaces only for the example of Ag(110).

Table 2 shows pertinent calculated results for selected

reaction pathways of H2 dissociation on Ag(110): reaction

energies Er = Etot(2H_sub) � [Etot(H2) + Etot(sub)] of dis-

sociative H2 adsorption, the corresponding activation energies

Ea, and geometric parameters of the transition and final states.

As initial state, we chose a configuration where the H2

molecule practically does not interact with the substrate,

namely 300 pm above the top crystalline plane. Many combi-

nations of stable H adsorption sites seem possible as final state

of H2 dissociation because the potential energy surface of

atomic H adsorption on Ag is quite flat. We have chosen the

FS (with hydrogen coverage of 1/2) as combinations of various

stable adsorption complexes (Table 2; Fig. 1), namely the short

bridge site 5, the long bridge site 7, and 3-fold hollow site 6. All

selected FS have been shown to be local minima with the help

of a normal mode analysis. Based on these structures for

coverage 1/2, we also carried out single-point calculations at

coverage 1/3. The zig-zag FS configuration of hydrogen atoms

at 3-fold hollow sites, 6/6, is the least endothermic among the

configurations studied, with Er = 33 and 49 kJ mol�1 for

coverage of 1/2 and 1/3, respectively. Note that reducing the

surface coverage from 1/2 to 1/3 generally makes H2 dissocia-

tion slightly more endothermic except for the long-bridge/

long-bridge configuration 7/7 along one trough where the

reaction energy is essentially the same for both coverage values

(Table 2). This implies that more energy (B10 kJ mol�1) is

required for initial H2 dissociation when the surface is free of

adsorbed H, but the differences are small and within the

accuracy of the current computational method employed.

These differences can be rationalized by the resulting band

structure and nearest-neighbors interaction among the adsor-

bates. In the final states 5/5, 6/6, 7/7, and 5/6 at 1/2 coverage,

the nearest-neighbor distances within a unit cell and between

adjacent unit cells are identical; in other words, the adsorbates

form chain-like structures over the crystal plane, which results

in an additional stability gain. When the coverage is reduced,

nearest-neighbor distances among adsorbates in adjacent unit

cells increase, resulting in a reduced stability of the final state,

hence a more endothermic reaction. This effect is substantial

for the complexes 6/6 and 5/6, but hardly noticeable for the

complexes 5/5 and 7/7.

The lowest activation energies calculated for H2 dissociation

on Ag(110) with surface coverage at 1/2 is about 125 kJ mol�1.

This value is comparable to the reported lowest activation

energies of 106 kJ mol�1 on the Ag(100) surface,8 but notably

higher than on other transitional metal surface, e.g. Pt where a

vanishing activation barrier was calculated.42 Reducing the

surface coverage from 1/2 to 1/3 does not change the activa-

tion energies by more than 5 kJ mol�1 for all the configura-

tions studied. The changes of the activation energies with

coverage can be taken as uncertainty estimates of the results

of the current model strategy. Hence, on the basis of the

present results, one should consider the activation energies of

the final states 5/5, 5/6, 6/6, and 5/7 as very similar, an average

at B125 kJ mol�1 (see above). The present results are con-

sistent with the low dissociation probability of H2 determined

experimentally at silver surfaces (see section 1).5

In summary, we calculated H2 dissociation to be kinetically

and thermodynamically unfavorable on clean surfaces of

silver. Thus, other sites are likely responsible for the activation

of hydrogen on silver, e.g. reaction centers induced by the

presence of surface oxygen species.

3.2 Hydrogen adsorption and activation on p(n� 1)O/Ag(110)

surfaces

We begin our discussion with the surface p(2 � 1)O/Ag(110),

which is characterized by the smallest unit cell among the

(n � 1) reconstructed added-row structures. The surface for

n = 2 is sketched in Fig. 2: added-row Ag atoms are

positioned at 4-fold hollow sites 8 of the non-reconstructed

Ag(110) surface and O atoms occupy short bridge sites 5 on it.

In Table 3, we compare calculated geometric parameters of

the reconstructed p(2 � 1)O/Ag(110) surface with experimen-

tal data26–28 and results of previously reported calcula-

tions.43,44 The Ag and O atoms of the added row apparently

form almost linear chains; we computed the O atoms to be

Table 2 Calculated activation energies Ea and reaction energies Er (kJ mol�1) of H2 dissociation for various final states on the Ag(110) surface atcoverage 1/2 (see also Fig. 1).a Selected distances (pm) of the transition state (TS) and the final state (FS) are also shownb

TS FS

H–H z(H) H–Ag Ea H–H H–Ag Er

Short bridge 5/short bridge 5 155 149 168 132 (134) 293 180 57 (61)3-fold 6/3-fold 6

c 151 155 172 133 (134) 361 185 33 (49)Long bridge 7/long bridge 7d 171 100 183 162 (160) 293 190 68 (67)Short bridge 5/long bridge 7 136 107, 50 178, 183 122 (127) 277 180, 200 54 (63)Short bridge 5/3-fold 6 152 121, 158 170, 168 129 (131) 318 180, 185 44 (55)Long bridge 7/3-fold 6 152 137, 130 167, 169 138 (143) 347 190, 185 52 (58)

a Values in parentheses are obtained in single-point calculations at coverage 1/3 for structures as determined at coverage 1/2. b H–H and H–Ag are

interatomic distances and z(H) is the height of H atom above the Ag(110) surface. c Zig–zag ordering of H atoms. d Complex oriented along the

troughs.

This journal is �c the Owner Societies 2007 Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 | 1251

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

displaced only 7 pm (d0) from the axis of the added Ag atoms

in the direction away from the substrate, in agreement

with the geometry of the photoelectron diffraction (PED)

study (Table 3), which is particularly sensitive to the para-

meter d0.26 The only notable difference with respect to the

PED experiment is the computed distance d23, 144 pm, which

is B10 pm (longer) out of the experimentally determined

range. Probably, for a more quantitative description of the

d23 values one should consider a thicker slab and allow

relaxation of more surface layers. Our results agree quite well

with all recently communicated experimentally derived struc-

tures26–28 and with the structure of a DF PW91 slab model

study.44

However, the structural characteristics obtained in an ear-

lier DF LDA model investigation43 (using the Perdew-Zunger

exchange–correlation functional45) differ considerably, but

they do so also from the values of other studies listed in Table

3. Most noticeable is the deviation of the d0 value, which is

B50 pm longer in the previous LDA work43 than in this work

and a former DF study,44 the latter two carried out with the

PW91 functional. At the same time the d12 value of ref. 43 is

B20 pm shorter compared to our PW91 result.

To rationalize these discrepancies, we also optimized the

structure of the surface p(2 � 1)O/Ag(110) at the LDA level45

(Table 3). Indeed, our LDA results agree only qualitatively

with those of the previous LDA calculation.43 Also, in our

LDA study the distance d0 is longer and the distance d12shorter compared to our PW91 results. Likely, the differences

between the two LDA structures have to be attributed to

different representations of the Kohn–Sham orbitals, namely

by localized functions in ref. 43 and the PAW technique in the

present study.34,35 The differences between the present LDA

and GGA (PW91) results are somewhat surprising as one

often observes that LDA structures are closer to experiment

than the corresponding PW91 structures. We also explored

whether a relaxation of the substrate is able to affect the

structure of the added row. However, when the top two layers

of the Ag substrate were allowed to relax together with the

added row, we found only very small structural changes: the

interlayer distance d0 changed from 7 to 8 pm, distance d12from 161 to 165 pm, and distance d23 from 144 to 139 pm (see

Table 3 and Fig. 2b).

The optimized structure of the added row surface system

changed only in a minor way when the interval between the

added rows increases, i.e. on going from the p(2 � 1)O/

Ag(110) to the p(4 � 1)O/Ag(110) structure.44 The corre-

sponding change of the calculated vibrational frequency of

O atoms was about 10 cm�1, in line with experimental

estimates of the effect of reduced oxygen coverage.44 This

implies that the adsorption propensity and, thus, the reactivity

of the Ag–O chains should not change significantly at a lower

oxygen coverage. In the following, we will corroborate this

expectation for the case of hydrogen adsorption.

Compared to a clean Ag(110) surface, a reconstructed

added-row p(n � 1)O/Ag(110) surface offers two new types

of positions for the adsorption of atomic hydrogen, on O and

Ag atoms of the added rows. We calculated the energy of H

adsorption on-top of an added-row Ag center of p(2 � 1)O/

Ag(110) at �77 kJ mol�1; in that structure, each Ag center of

the added rows is occupied with one atom H. The significantly

weaker interaction energy indicates a notable destabilization

compared to on-top adsorption complexes of the oxygen-free

surfaces Ag(110) and Ag(111), with Ead of about �150 kJ

mol�1 (Table 1). In contrast, the interaction with O centers

was calculated to be very strong, Ead = �331 kJ mol�1, in the

system with 1H : 1O where each of the O centers forms an H

adsorption complex. For a system with the lower coverage

1H : 2O, we calculated Ead = �317 kJ mol�1. This trend in

binding energies is likely due to the more flexible structure of

the added Ag–O row, when the O atoms are fully hydroge-

nated compared to the more rigid added Ag–O structure where

some O atoms are not hydrogenated. Note that, at 1H : 1O

coverage, all O atoms are bound to two Ag atoms of the added

row, whereas the O centers not hydrogenated at 2H : 3O

coverage are bound to two substrate Ag atoms and two added

row Ag atoms (Fig. 3).

On going from the (2 � 1) to the (3 � 1) and (4� 1) periodic

structures, where the distance between the added rows in-

creases, the adsorption energy at full coverage, 1H : 1O,

changes only negligibly, from �331 kJ mol�1 to �328 and

�329 kJ mol�1, respectively. Indeed, even in the (2 � 1)

structure the distance between neighboring added rows,

585 pm, is large enough to prevent any noticeable adsorbate–

adsorbate interaction between different rows. Thus, at var-

iance with clean silver surfaces, dissociation of H2 molecules

involving oxygen centers of added rows is exothermic by more

than 200 kJ mol�1.

Hydrogen adsorption induces significant changes in the

structure of the added row. Oxygen centers with hydrogen

adsorbed shift away from the substrate (Fig. 3); the Ag–O

distance increases from 207 pm, without adsorbate, to 226 pm

(Table 4).

Table 3 Comparison of calculated and experimental distancesa (pm) characterizing the added-row p(2 � 1)O/Ag(110 surface; for thedesignations, see Fig. 2

PEDb LEISc SEXAFSd LDAe LDAf GGAg GGAh

d0 3 � 5 �3 � 8 �20 � 20 59 15 7 7d12 155 � 6 166 � 3 167 � 40 142 138 161 161d23 133 � 6 132 � 3 128 134 138 144rnn 204.4 � 0.2 204.5 � 0.5 205 � 3 213 208 205 207rnnn 216 � 10 223 � 8 221 � 3 242 220 224 227

a For the interlayer distances d0, d12, and d23, see Fig. 2; rnn—distance between O and the nearest Ag atoms, rnnn—distance between O and the next-

nearest Ag atoms. b Photoelectron diffraction (PED),ref. 26. c Low energy ion-scattering (LEIS), ref. 27. d Surface extended X-ray adsorption

fine structure (SEXAFS), ref. 28. e DF LDA calculation, substrate of 4 layers, ref. 43. f DF LDA calculation, substrate of 5 layers, this

work. g DF GGA calculation, substrate of 4 layers, ref. 44. h DF GGA calculation, substrate of 5 layers, this work.

1252 | Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 This journal is �c the Owner Societies 2007

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

Finally, we will discuss the reaction path for H2 dissociation

on Ag–O added rows of the p(n � 1)O/Ag(110) surfaces. We

begin with the p(2 � 1)O/Ag(110) structure which features the

most compact unit cell. H2 molecules physisorb on this sur-

face; the calculated energy of adsorption, Ead B �2 kJ mol�1,

is negligible. Therefore, we represented the initial state (IS)

of H2 dissociation (and the energy reference) by almost free H2

molecules at 300 pm above the added Ag–O rows (Fig. 3).

According to our calculations, the activation of H2 on the

added row proceeds in two steps (Fig. 3). First, H2 dissociates

and the two atomic H species bind to neighboring O and Ag

atoms of an added row. The calculated activation energy of

this dissociation step is 71 kJ mol�1 relative to the above

mentioned IS where H2 is essentially free (Table 4). This

activation barrier is reduced almost to half of that calculated

for H2 dissociation on the clean Ag(110) surface (Table 2). The

H–H distance in the corresponding transition state TSdiss,

122 pm, elongates to 231 pm in the intermediate state after

complete splitting of the H–H bond, to better match the Ag–O

distances of 207–226 pm (Table 4). That intermediate state is

destabilized by 32 kJ mol�1 with respect to the IS, due to the

relatively weak binding, �77 kJ mol�1 (see above), of an H

atom at an Ag center of the added row. In the second step, this

weakly bound H atom moves from Ag to a neighboring

(vacant) O center. The activation energy of this diffusion

transition state TSdiff is very small, only 12 kJ mol�1. As

already discussed, the resulting FS lies 224 kJ mol�1 lower in

energy than the IS (Table 4).

To examine an eventual effect of adsorbate–adsorbate inter-

action, we also studied a model with lower coverage of

adsorbed hydrogen—one H2 molecule per (2 � 3) unit cell,

where in the end one out of three O adsorption sites remains

unoccupied (Fig. 3). Table 4 shows that the reaction path is

affected only slightly, both structurally and energetically, by

this reduction of the hydrogen coverage along the added row.

In addition, we estimated the effect of a reduced oxygen

coverage, by studying H2 activation on the structure p(3 � 1)

O/Ag(110) with a (3 � 2) unit cell where the distance between

the added rows is increased; here, we returned to a model

where in the final state all oxygen centers are occupied by

hydrogen atoms. As expected, also in this case, we did not find

any essential changes of the characteristics of the reaction path

(Table 4, values in parentheses). Thus, the reaction profile is

almost independent of the hydrogen coverage along the added

rows and the distance between the added rows. Therefore, we

propose to generalize the present mechanism of H2 activation

Fig. 3 H2 dissociation on the added-row p(2 � 1)O/Ag(110) surface

for the coverage H :O = 2 : 3 (Table 4). The various panels (from top

to bottom) show (IS) the initial state with H2 in the gas phase above

the surface, (TSdiss) the transition state for H2 dissociation, (I) the

intermediate state with one atom H adsorbed at O and another one at

the neighboring Ag center, (TSdiff) the transition state for H diffusion

from the center Ag to a neighboring unoccupied O, and (FS) the final

state with both H atoms adsorbed at O sites.

Table 4 Calculated geometric parameters and relative energies of initial states (IS), transition states for H2 dissociation (TSdiss), intermediates (I),transition states for diffusion of one atom H form Ag to O (TSdiff), and final states (FS) for the dissociation of hydrogen at the p(2 � 1)O/Ag(110)surface; see Fig. 3. Energy values in parentheses are for p(3 � 1)O/Ag(110) structures, calculated in a single-point approximation

Interatomic distances/pm

H–H Ag–H O–H Ag–O E,a kJ mol�1

H :O = 1 : 1

IS 75 371 402 207 0TSdiss 122 189 109 220 71 (69)I 231 165 97 211 32 (29)TSdiff 325 166 98 212 44 (43)FS 414 279 98 226 �224 (�218)H :O = 2 : 3

IS 75 372 404 207 0TSdiss 122 189 109 220 76I 233 166 98 213 46TSdiff 324 168 98 212 63FS 414 284 98 227 �205a Energy change relative to the IS.

This journal is �c the Owner Societies 2007 Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 | 1253

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online

also to very low oxygen coverage, as long as local fragments

resembling added-row structures are formed.

4. Conclusions

We studied the activation of H2 on clean silver surfaces and on

a model surface pre-covered by oxygen. Using a density

functional method and a periodic slab models, we investigated

the adsorption of atomic hydrogen on the surfaces Ag(111),

Ag(110), and Ag(221) as well as on added-row structures

p(n � 1)O/Ag(110). According to these computational results,

the dissociation of H2 is thermodynamically (endothermic by

B40 kJ mol�1) and kinetically (activation barrier about 125 kJ

mol�1) unfavorable on clean silver surfaces. In contrast,

dissociation of H2 on the –Ag–O– added-row chains of

reconstructed p(n � 1)O/Ag(110) surfaces is highly exothermic

(by 200–220 kJ mol�1) and proceeds in two steps with low

activation barriers: (i) dissociation of H2 on an Ag–O pair with

an activation energy of B70 kJ mol�1, and (ii) diffusion of

hydrogen atoms from silver to more favorable neighboring

vacant oxygen positions with an activation barrier of B10 kJ

mol�1. On the basis of these results, the presence of atomic

oxygen species on silver surfaces seems essential for their

propensity to activate molecular hydrogen.

Acknowledgements

We thank Dr M. Bron and Prof. P. Claus for stimulating

discussions. Dr Z.-X. Chen provided advice during the initial

stage of the project. K.H.L. is grateful to Deutscher Aka-

demischer Austauschdienst for a fellowship. This study was

supported by Deutsche Forschungsgemeinschaft and Fonds

der Chemischen Industrie (Germany). K.M.N. acknowledges

financial support by the Spanish Ministry of Education and

Science (grants CTQ2005-08459-CO2-01, UNBA05-33-001)

and the Generalitat de Catalunya (grant 2005SGR-00697).

References

1 L. Schlapbach and A. Zuttel, Nature, 2001, 414, 353.2 Ph. Avouris, D. Schmeisser and J. E. Demuth, Phys. Rev. Lett.,1982, 48, 199.

3 P. T. Sprunger and E. W. Plummer, Phys. Rev. B, 1993, 48, 14436.4 K. Christmann, Surf. Sci. Rep., 1988, 9, 1.5 X. Zhou, J. M. White and B. E. Koel, Surf. Sci., 1989, 218, 201.6 G. Lee and E. W. Plummer, Phys. Rev. B, 1995, 51, 7250.7 C. Mijoule and V. Russier, Surf. Sci., 1991, 254, 329.8 A. Eichler, G. Kresse and J. Hafner, Surf. Sci., 1998, 397, 116.9 J. Greeley and M. Mavrikakis, J. Phys. Chem. B, 2005, 109, 3460.

10 R. A. Van Santen and H. P. C. E. Kuipers, Adv. Catal., 1987, 35,265.

11 P. Claus, Top. Catal., 1998, 5, 52.12 M. Bron and P. Claus, personal communication.13 J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R.

Pederson, D. J. Singh and C. Fiolhais, Phys. Rev. B, 1992, 46,6671.

14 B. Hammer, L. B. Hansen and J. K. Nørskov, Phys. Rev. B, 1999,59, 7413.

15 CRC Handbook of Chemistry and Physics, ed. D. R. Lide, CRCPress, London, 77th edn, 1996.

16 J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996,77, 3865.

17 P. Claus, Appl. Catal. A, 2005, 291, 222.18 S. Linic and M. A. Barteau, J. Catal., 2003, 214, 200.19 C. Stegelmann, N. C. Schiodt, C. T. Campbell and P. Stoltze, J.

Catal., 2004, 221, 630.20 V. I. Bukhtiyarov, M. Havecker, V. V. Kaichev, A. Knop-Gericke,

R. W. Mayer and R. Schlogl, Phys. Rev. B, 2003, 67, 235422.21 S. A. Tan, R. B. Grant and R. M. Lambert, Appl. Catal., 1987, 31,

159.22 J. Schnadt, A. Michaelides, J. Knudsen, R. T. Vang, K. Reuter, M.

Lægsgaard, M. Scheffler and F. Besenbacher, Phys. Rev. Lett.,2006, 96, 146101.

23 M. Schmid, A. Reicho, A. Stierle, I. Costina, J. Klikovits, P.Kostelnik, O. Dubay, G. Kresse, J. Gustafson, E. Lungren, J. N.Andersen, H. Dosch and P. Varga, Phys. Rev. Lett., 2006, 96,146102.

24 C. T. Campbell and M. T. Paffett, Surf. Sci., 1984, 143, 517.25 C. T. Campbell, Surf. Sci., 1985, 157, 43.26 M. Pascal, C. L. A. Lamont, P. Baumgartel, R. Terborg, J. T.

Hoeft, O. Schaff, M. Polcik, A. M. Bradshaw, R. L. Toomes andD. P. Woodruff, Surf. Sci., 2000, 464, 83.

27 M. Canepa, P. Cantini, F. Fossa, L. Mattera and S. Terreni, Phys.Rev. B, 1993, 47, 15823.

28 L. Becker, S. Aminpirooz, A. Schmalz, B. Hillert, M. Pedio andJ. Haase, Phys. Rev. B, 1991, 44, 13655.

29 K. Berge and A. Goldmann, Surf. Sci., 2003, 540, 343.30 O. Nakagoe, K. Watanabe, N. Takagi and Y. Matsumoto, J. Phys.

Chem. B, 2005, 109, 14536.31 G. Kresse and J. Furthmuller, Phys. Rev. B, 1996, 54, 11169.32 G. Kresse and J. Hafner, Phys. Rev. B, 1993, 47, 558.33 G. Kresse and J. Furthmuller, Comput. Mat. Sci., 1996, 6, 15.34 P. E. Blochl, Phys. Rev. B, 1994, 50, 17953.35 G. Kresse and D. Joubert, Phys. Rev. B, 1999, 59, 1758.36 H. J. Monkhorst and J. D. Pack, Phys. Rev. B, 1976, 13, 5188.37 M. Methfessel and A. T. Paxton, Phys. Rev. B, 1989, 40, 3616.38 Z.-X. Chen, K. H. Lim, K. M. Neyman and N. Rosch, J. Phys.

Chem. B, 2005, 109, 4568.39 K. H. Lim, Z.-X. Chen, K. M. Neyman and N. Rosch, Chem.

Phys. Lett., 2006, 420, 60.40 G. Mills, H. Jonsson and G. K. Schenter, Surf. Sci., 1995, 324, 305.41 W. Koch and M. C. Holthausen, A Chemist’s Guide to Density

Functional Theory, 2nd edn, Wiley-VCH, Weinheim, 2002.42 S. Kandoi, A. A. Goklale, L. C. Grabow, J. A. Dumesic and M.

Mavrikakis, Catal. Lett., 2004, 93, 93.43 T. Schimizu and M. Tsukada, Surf. Sci., 1993, 295, L1017.44 H. Katagiri, T. Uda and K. Terakura, Surf. Sci., 1999, 424, 322.45 J. Perdew and A. Zunger, Phys. Rev. B, 1981, 23, 5048.

1254 | Phys. Chem. Chem. Phys., 2007, 9, 1247–1254 This journal is �c the Owner Societies 2007

Publ

ishe

d on

19

Janu

ary

2007

. Dow

nloa

ded

by U

nive

rsity

of

Prin

ce E

dwar

d Is

land

on

22/1

0/20

14 1

9:28

:51.

View Article Online