Density functional study of water-gas shift reaction on M3O(3x)/Cu(111)

16626 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 This journal is c the Owner Societies 2012

Cite this: Phys. Chem. Chem. Phys., 2012, 14, 16626–16632

Density functional study of water-gas shift reaction on M3O3x/Cu(111)w

Alba B. Vidalab

and Ping Liu*a

Received 21st June 2012, Accepted 10th August 2012

DOI: 10.1039/c2cp42091k

Density functional theory (DFT) was employed to study the water dissociation and water-gas

shift (WGS) reaction on a series of inverse model catalysts, M3O3x/Cu(111) (M = Mg, Ti, Zr,

Mo, W; x = 1, 2, 3). It has been found that the WGS reaction on Cu can be facilitated by

introducing various oxides to lower the barrier of water dissociation. Accordingly, the calculated

reaction energy for water dissociation was used as a scaling descriptor to screen the WGS activity

of oxide–Cu model catalysts. Our calculations show that the activity towards water dissociation

decreases in a sequence: Mg3O3/Cu(111) > Zr3O6/Cu(111) > Ti3O6/Cu(111) > W3O9/Cu(111),

Mo3O9/Cu(111). It seems that Mg3O3/Cu(111) is the best WGS catalyst among the systems

studied here, being able to dissociate water with no barrier. During the process, both Cu and

oxides participate in the reaction directly. The strong M3O3x–Cu interaction is able to tune the

electronic structure of M3O3x and therefore the activity towards water dissociation. Further

studies of the overall WGS reaction on Mg3O3/Cu(111) show that water dissociation may not be

the key step to control the WGS reaction on Mg3O3/Cu(111) and the removal of H from Mg3O3

can be problematic. The strong interaction between H and O from Mg3O3 blocks the O sites for

further water dissociation and therefore the WGS reaction. Our study observes a very different

behavior of oxide clusters in such small size from the bigger ones supported on Cu(111) and

provides new insight into the rational design of the WGS catalysts.

I. Introduction

The water-gas-shift (WGS) reaction (CO + H2O - CO2 + H2)

has attracted considerable interest, due to its potential applications

in fuel cells, Fischer–Tropsch processes and hydrogen produc-

tion.1–4 Common industrial catalysts for the WGS (mixtures

of Fe–Cr or Zn–Al–Cu oxides) are pyrophoric and normally

require lengthy and complex activation steps before usage.5

Recently, inverse model catalysts of CeOx nanoparticles

supported over Au(111) or Cu(111) surfaces have been found

to display a superior catalytic activity in the WGS reaction.

In particular, the recent study showed that CeOx/Cu(111) is

more active than the traditional catalysts, Cu/CeO2(111) and

Cu/ZnO(0001), towards the WGS reaction,6 though the corner

and edge atoms present in the Cu nanoparticles facilitate the

dissociation of water.7,8 On one hand, the oxide 2 metal

interactions can alter the electronic states of the supported

oxide and lead to new chemical properties.6,9 On the other

hand, an inverse oxide/metal catalyst allows the reactants to

interact with not only metal sites and the metal–oxide interface

as in the case of a traditional metal/oxide catalyst,10 but also

defect sites of oxide nanoparticles, which can be very important

for the overall conversion.6,11 By adopting the inverse model,

the catalyst can gain activity due to the active participation of

oxide in the catalytic reaction.6,11–13

In our previous study,14 the oxide chain (b-MOx, M = Zn,

Zr, Ti, Mo) deposited on the Cu(111) surface was used as a

simplified model to simulate the interface between Cu and

relatively big oxide particles observed experimentally under

the WGS conditions, which display bulk-like structures.6

It was found that the calculated reaction energy for water

dissociation, the rate-limiting step for the WGS reaction on

pure Cu surfaces and nanoparticles,7,15,16 correlates well

with the experimentally measured activity. The high WGS

activity of oxide/Cu(111) relies heavily on the direct participation

of both oxide and metal sites, where the oxide–Cu interaction

plays an important role. The reducible oxides (e.g. ZrO2, TiO2

and MoO3) that are fully oxidized can be reduced due to the

interaction with Cu, which help in releasing the bottleneck water

dissociation and therefore facilitating the WGS reaction on Cu.

In the present study, we move from the previous MOx/

Cu(111) system to M3O3x/Cu(111) (M = Mg, Ti, Zr, Mo, W;

x = 1, 2, 3), where the oxide trimer is used to model the

relatively small oxide particles. Our goal is to gain more

insight into the WGS reaction and screening good catalysts

using density functional theory (DFT). Extensive studies have

shown that for conventional metal/oxide model catalysts, the

a Chemistry Department, Brookhaven National Laboratory, Upton,New York 11973, USA. E-mail: [email protected]

bCentro de Quımica, Instituto Venezolano de InvestigacionesCientıficas (IVIC), Apartado 21827, Caracas 1020-A, Venezuela

w This article was submitted as part of a collection on ComputationalCatalysis andMaterials for Energy Production, Storage and Utilization.

PCCP Dynamic Article Links

www.rsc.org/pccp PAPER

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

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

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 16627

size of the supported metal particle can affect the catalytic

activity significantly.17–19 In contrast, less effort has been made

to understand the size effect of supported oxide particles, which is

very attractive for both fundamental research and technological

applications of metal oxides. It has been reported that very small

oxide clusters supported on a stoichiometric oxide with different

compositions or metal surface offer an opportunity to control the

atomic and electronic structure of oxides and therefore the

catalytic performances.20–24 Here, we focus our interest on the

interaction of oxide trimers (Mg3O3, Ti3O6, Zr3O6, Mo3O9 and

W3O9) with the Cu(111) surface and their WGS activities. These

oxides may vary in reducibility, oxidation states and structures,

which allows us to gain better understanding of: what is the

behavior of such small oxide clusters during the WGS reaction?

Does the small oxide cluster behave similar to the big ones

supported on Cu(111) studied previously?7 What is the role of

Cu? What is the key to achieve higher WGS activity for

oxide–Cu systems?

II. Computational method

Theoretical calculations were performed using the plane-wave

density functional theory (DFT) approach within the projector

augmented wave method (PAW)25 using the GGA exchange–

correlation functional proposed by Perdew et al.26 as imple-

mented in the VASP 5.2 code.27–29 A plane-wave cutoff energy

of 400 eV was used. We treated the Mg (2p, 3s), Ti (3s, 3p, 3d,

4s), Zr (4s, 4p, 4d, 5s), Mo (4p, 4d, 5s) and W (4f, 5d, 6s), O

(2s, 2p), C(2s, 2p) and H (1s) electrons as valence states, while

the remaining electrons were kept frozen as core states. To

obtain faster convergence, thermal smearing of one-electron

states (kBT = 0.05 eV) was allowed using the Gaussian

smearing method to define the partial occupancies. Previous

DFT calculations using similar setups have been successful in

modeling these oxide systems, being able to well describe the

experimental observations.24

To model the Cu(111) surface, we used four-layer slabs

repeated in a (5 � 5) supercell geometry with a 20 A vacuum

between the slabs. In the calculations, the two top layers of

Cu(111) were allowed to relax in all dimensions together with

the oxides, while the bottom two layers were kept fixed at the

calculated bulk positions. To model the oxide nanoparticles

supported on Cu(111), an oxide trimer was deposited on

Cu(111). The oxide clusters considered here are Mg3O3,

Zr3O6, Ti3O6, Mo3O9 and W3O9. The charge transfer between

the Mo3O3x oxide cluster and Cu(111) was calculated using the

Bader analysis program,30 based on Bader’s theory.31,32

The calculated water adsorption energy was expressed as

DEads = E(H2O/surface) � E(H2O) � E(surface), where E is

the total energy. The reaction energy for water dissociation was

expressed as DE = E[(H + OH)/surface] � E(H2O/surface).

III. Results and discussion

III.1 Geometries of M3O3x/Cu(111)

Our calculations show that all M3O3x clusters strongly interact

with Cu(111). Fig. 1 displays the most stable adsorption

geometries of Mo3O3x clusters supported on Cu(111). In all

cases the Cu(111) surface greatly stabilizes the oxide cluster.

The Mg3O3 cluster with a six-member ring geometry has been

found as the most stable structure for Mg3O3 previously in

both theory and experiment.33–37 Accordingly, such ring

structure was considered here for Mg3O3. As shown in

Fig. 1a, the ring is adsorbed strongly (DEads = �3.42 eV)

and in parallel to the Cu(111) surface. Both Mg and O ions are

bound to Cu, which is accompanied by a significant structural

change in the surface. One can see that the Cu(111) surface is

rippled (Fig. 1a). The Cu atoms interacting with O are pulled

outward the surface by 0.35 A and those bound to Mg are

suppressed 0.17 A below the surface. For Zr3O6 and Ti3O6

clusters, it has been shown that a zig-zag ring is formed,

consisting of three – (Ti(Zr)–O) – units, a triple-bridge O at

the center, and two terminal O.38–41 When depositing on

Cu(111), Zr3O6 and Ti3O6 clusters are bound to Cu(111) via

a central O, a terminal O and a metal ion with DEads of�3.32 eVfor Zr3O6/Cu(111) and �4.08 eV for Ti3O6/Cu(111) (Fig. 1b

and c). Mo3O9 and W3O9 clusters prefer a coplanar cyclic

conformation where each metal ion is tetrahedrally coordinated

with two terminal and two bridging O.42–44 Similar to the cases

of Zr3O6 and Ti3O6, our calculations show that Mo3O9 and

W3O9 are bound to Cu(111) via three terminal O (Fig. 1d and e)

and the corresponding DEads is �1.71 eV for Mo3O9/Cu(111)

and �1.78 eV for W3O3/Cu(111).

III.2 Water adsorption and dissociation on M3O3x/Cu(111)

III.2.1 Geometrics and energetics.We examined the activities of

M3O3x/Cu(111) systems towards water adsorption and dissociation

(Fig. 2), which has been identified as the key step for the WGS

reaction on Cu, Au and Au, Cu–oxide systems.14

For Mg3O3/Cu(111), the water prefers to adsorb at the

Mg3O3–Cu interface (DEads = �1.14 eV). In the initial state,

the molecule binds to Cu via O and one of the H is tilted

Fig. 1 Side view of the optimized geometry of the clusters Mg3O3 (a),

Zr3O6 (b), Ti3O6 (c), Mo3O9 (d) and W3O9 (e) supported on a Cu(111)

surface. (Light brown: Cu; light green: Mg; turquoise: Zr; gray: Ti;

green: Mo; blue: W; red: O.)

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online

16628 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 This journal is c the Owner Societies 2012

towards one of the O in Mg3O3. However, this structure is not

stable and a spontaneous O–H bond cleavage is observed

during the geometry optimization (Fig. 2a). The dissociated

state corresponds to the formation of OH species on both

metal surface and Mg3O3 cluster. In contrast, the water

adsorption on Mg3O3 is slightly less favorable (DEads =

�1.06 eV, Fig. 2b). More importantly, no spontaneous water

dissociation occurs. The corresponding DE for water dissocia-

tion is �0.84 eV and a barrier is expected to be overcome for

O–H bond cleavage. The dissociated OH fragment adsorbed

at the Mg site shifts to the Mg–Cu bridging site and the

dissociated H interacts with a nearby O of the oxide cluster to

form OH species (Fig. 2c). Although the dissociated states on

Mg3O3 (DEads = �1.90 eV, Fig. 2c) are more stable than that

at the Mg3O3/Cu interface (DEads = �1.14 eV, Fig. 1a), the

latter should be preferred as a barrierless process. It is known

that MgO is a very stable oxide and water only physisorbs on

the MgO(100) surface.45–47 Similarly to MgO, using the chain

model, our previous studies observed weak water adsorption

and thermoneutral dissociation on b-ZnO/Cu(111).14 Our

present calculations show that Cu(111) is able to promote

the activity of small MgO clusters and the synergy between Cu

and MgO leads to a barrierless water dissociation.

Different from Mg3O3/Cu(111), water prefers the oxide site

of Zr3O6/Cu(111) and Ti3O6/Cu(111). Two possible sites have

been considered for the water adsorption. As shown in Fig. 2d

and h, on the M (Zr or Ti) ion that is bound to two bridging O

and one terminal O at the M3O6–Cu(111) interface (MO),

DEads is �1.05 eV for Zr3O6/Cu and �1.09 eV for Ti3O6/Cu.

In comparison, a slightly stronger adsorption is observed for

the water molecule adsorbed on the M ion that is bound to

two bridging O and one Cu(MCu) (DEads, �1.16 eV for

Zr3O6/Cu and �1.17 eV for Ti3O6/Cu, Fig. 2e and i). Both

configurations are more stable than those at the oxide–Cu

interface, where the molecule binds to Cu via O and one H is

tilted towards the oxide (DEads, �0.54 eV for Zr3O6/Cu and

�0.91 eV for Ti3O6/Cu). In addition, although the molecular

adsorptions of water on Zr3O6/Cu(111) and Ti3O6/Cu(111) are

as strong as that of Mg3O3/Cu(111), no spontaneous dissocia-

tion is obtained and a barrier has to be overcome to form two

OH species. Similar behavior was also observed previously for

the bigger oxide clusters, b-TiO2 and b-ZrO2/Cu(111).14 For

water dissociation, there is a clear preference for water at the

MO site, where the dissociated H interacts with the terminal O

and the OH fragment stays on top of MO (DE, �1.03 eV for

Zr3O6/Cu and �0.74 eV for Ti3O6/Cu, Fig. 2f and j). In

contrast, the dissociation of water at MCu sites with the

dissociated H bound to a bridging O is less favorable

(DE, �0.21 eV for Zr3O6/Cu and �0.19 eV for Ti3O6/Cu,

Fig. 2g and k). The dissociations for relatively weakly adsorbed

water at the oxide–Cu interface are not considered here.

Mo3O9/Cu(111) and W3O9/Cu(111) behave similarly to

Mg3O3/Cu(111), where the oxide–Cu interface is preferred,

though DEads is smaller (DEads, �0.58 eV for Mo3O9/Cu and

�0.85 eV for W3O9). The water molecule prefers to bind to

Cu(111) via O and one of H tilts toward O of the oxide cluster

(Fig. 2l and n). When water is attached to the metal sites of the

supported oxide clusters, the molecule desorbs during the

geometry optimization. Due to the relatively weak interaction,

there is no spontaneous water dissociation onMo3O9/Cu(111) and

W3O9/Cu(111). This is different from our previous calculations on

b-MoO3/Cu(111), where a spontaneous water dissociation is

observed.14 In addition, the dissociation in both cases is

endothermic (DE, 0.65 eV for Mo3O9/Cu and 0.74 eV for

W3O9/Cu, Fig. 2m and o) and therefore a high barrier is

expected according to our previous calculations.14

III.2.2 Understanding from electronic structure. Our calcu-

lations show that the ability to dissociate water decreases on

going from Mg3O3/Cu(111), Zr3O6/Cu(111) and Ti3O6/

Cu(111) to Mo3O9/Cu(111) and W3O9/Cu(111). This is different

from that observed previously for b-MOx/Cu(111), with b-MoO3/

Cu(111) > b-ZrO2/Cu(111), b-TiO2/Cu(111) > b-ZnO/Cu(111)

in a decreasing sequence in activity towards water dissociation.14

That is, the particle size can make a big difference in the activity of

metal oxides. To understand the trend, we calculated the partial

density of states (PDOS) of the unsupported (Fig. 3a–c),

supported oxide cluster (Fig. 3d–f) and the Cu atoms

(Fig. 3g–i) on M3O3x/Cu(111) (M = Mg, Zr, W). The PDOS

for Ti3O6/Cu(111) and Mo3O9/Cu(111) are not shown in the

figure, which are very similar to Zr3O6/Cu(111) and W3O9/

Cu(111), respectively. As shown in Fig. 3, the formation of the

M3O3x/Cu(111) introduces significant changes in the electronic

structure of M3O3x. For the free oxide cluster, molecular bands

with strongly mixed contributions from M and O are observed,

indicating a high degree of covalency (Fig. 3a–c). It means that

the isolated cluster is electronically transformed into a molecule.

However, when the cluster is deposited on Cu(111), the electronic

structure of the oxide cluster returns to a solid-like DOS with

wide O 2p and M s or d bands (Fig. 3d–f). The interaction with

Fig. 2 Optimized geometry for water adsorption and water dissocia-

tion on Mg3O3/Cu(111) (a), (b) and (c), Zr3O6/Cu(111) (d), (e), (f) and

(g), Ti3O6/Cu(111) (h), (i), (j) and (k), Mo3O9/Cu(111) (l) and (m), and

W3O9/Cu(111) (n) and (o); respectively. (Light brown: Cu; light green;

Mg; turquoise; Zr; grey: Ti; green: Mo; blue: W; red: O, white: H.)

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 16629

the support is very important since it ‘‘crystallizes’’ the electronic

structure of the oxide clusters from an initial ‘‘molecular’’

distribution. In all cases the major contributions to the valence

band are from O. This is a common feature of several oxides,48

which have been attributed to the unshared oxygen electron

pairs.49 Compared to the free oxide cluster, the O 2p and M s

or d bands are wider in the case of M3O3x/Cu(111). As a

consequence, after deposition, the band-gap is reduced in all cases.

We also examined PDOS of the surface Cu, which includes

Cu 3d, 4s and 4p and the band center is determined by the

position of Cu 3d. The results show that the Cu band center

varies depending on the adsorbed oxide cluster. In the case of

Mg3O3/Cu(111), there are three kinds of Cu atoms on the

surface. Cu–O and Cu–Mg are the Cu atoms interacting with

O and Mg from the cluster, respectively. Cu–neighbors are

those which are not directly interacting with the oxide cluster.

The positions of the band center are �1.64 eV for Cu–O,

�1.67 eV for Cu–neighbors and�1.74 eV for Cu–Mg. Due the

interaction between O and Cu, there is the position change of

the Cu–O band toward higher energy compared to the

Cu–neighbor (shift, 0.03 eV). The upshift suggests that there

is an electron transfer from Cu to O or Cu–O is oxidized.

In contrast, in the case of the Cu–Mg, it shows opposite

behavior, where a downshift (�0.07 eV) is observed. It means

that the electrons are transferred from Mg to Cu. That is,

Cu–Mg is reduced and the reduction of Cu is two times more

significant than the oxidation. This result is in good agreement

with the Bader charge analysis, where a charge transfer of

0.56 electrons from the Mg3O3 to Cu(111) is obtained and

Mg3O3 is oxidized when depositing on Cu(111). As a result,

both valence and conduction bands of Mg3O3 are dominated

by O 2p, which is different from the other oxide clusters we

studied here. For Zr3O6/Cu(111), there are also three kinds of

Cu atoms on the surface, Cu–O and Cu–Zr and Cu–neighbors.

The band centers are �1.62 eV for Cu–O, �1.68 eV for

Cu–neighbors and �1.70 eV for Cu–Zr. Again, Cu–O is

oxidized by transferring electrons to Zr3O6 together with

upshifts of the band center (0.06 eV). Cu–Zr is reduced with

the downshifts of the band center (�0.02 eV). The shift for

Cu–O is three times higher than that for Cu–Zr. In addition,

Cu–O atoms are three times more than Cu–Zr atoms. Given

that, an overall Cu oxidation is likely. This agrees with the

Bader charge analysis, showing that the Zr3O6 cluster receives

about 0.13 e from the Cu(111) surface and the Zr3O6 cluster is

reduced. This is very similar to the case of a bigger ZrO2 cluster

supported on Cu(111).14Ti3O6/Cu(111) behaves similarly, where

the electron transfer from Cu(111) to Ti3O6 is 0.33 e. In the case

of W3O9/Cu(111), there is no W ion interacting with Cu(111).

The band center positions are �1.63 eV for Cu–O and �1.68 eV

for Cu–neighbor atoms, with the upshift of 0.05 eV. That is, the

Fig. 3 Partial density of states for Mg3O3 (a), (d) and (g), Zr3O6 (b), (e) and (h) and W3O9 (c), (f) and (i) for M3O3x clusters in gas-phase, Cu(111)

supported M3O3x clusters and the Cu (sum of Cu 3d, 4s and 4p) atoms on the surface of M3O3x/Cu(111).

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online

16630 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 This journal is c the Owner Societies 2012

electrons are transferred from the Cu(111) surface to the W3O9

cluster, and a charge of �1.37 e for the cluster is obtained by

Bader charge analysis. This result is in agreement with a

previous study where W3O9 gas phase clusters were generated

via vacuum sublimation and deposited under ultrahigh vacuum

and low temperature conditions on a Cu(110) surface.20 The

same charge is also obtained for the Mo3O9 cluster supported on

Cu(111). This is different from the case of a bigger MoO3 cluster

supported on Cu(111), where the electron transfer is much small

(�0.47 e).14

Now the question is how such electronic structure leads to

the spontaneous water dissociation. Extensive studies indicate

that the lone pair coupling with the surface band near the

Fermi level is crucial for the formation of the water–surface

bond.50 It is rather localized at the interfaces, which is always

accompanied by electron transfer from O to surface.50,51

For water dissociation on metal oxide and O-covered metal

surfaces, an OH ion is produced by the charge transfer from

metal or metal cations to water. For H, however, the charge

transfer is reversed from H to O anions, leading to another OH

group on the surface. That is, the portion of metal or metal

cations in the valence band as well as that of surface O 2p in

the conduction band can be important for O–H breaking.

Indeed, it has been found that the most dramatic effect to

trigger a fast water dissociation is associated with the surface

O 2p level.52 If the surface O 2p level is close to the Fermi level,

water in dissociated form can attach a H to a surface O,

forming an O–H bond. The formation of the O–H bond

greatly reduces the energy of the O 2p level and drives the

dissociation of water on the surfaces.

On Mg3O3/Cu(111), as shown above, the molecular water is

not stable and the fast dissociation is preferred at the interface.

The positively charged Mg is not capable of stabilizing OH by

itself, due to the lack of electrons near the Fermi level.

External help from Cu is necessary and OH is adsorbed on

the Mg–Cu bridge site (Fig. 2c and 3d). In contrast, Cu atoms

on the surface that are rich in electrons near the Fermi level

(Fig. 3g) are able to provide electrons to form OH anions;

while the O anions of Mg3O3 that dominate the conduction

bands readily accept the electron transfer from H to form H+.

In this way, a spontaneous O–H bond breaking is observed.

As a consequence of electron transfer from Cu(111) to Zr3O6/

Cu(111), O 2p and Zr 4d dominate the valence band and

conduction band, respectively (Fig. 3e). The Zr cations readily

accept electrons from water molecules via O, and the mole-

cular adsorption is stabilized. Different from Mg3O3/Cu(111),

O 2p does not dominate the conduction band, which hinders

the fast transfer of H to H+ during water dissociation. There-

fore, water does not dissociate spontaneously and a barrier is

expected though the reaction is thermodynamically favored.

This is also observed for Ti3O6/Cu(111) as well as b-TiO2 and

b-ZrO2/Cu(111).14 When depositing W3O9 on Cu(111), the

bigger electron transfer from Cu to W3O9, or reduction, than

the case of Zr3O6/Cu(111) leads to the decrease in DOS at the

conduction band (Fig. 3f). As a result, W ions are not capable

of stabilizing water molecules and therefore water molecules

prefer to adsorb on Cu(111). For water dissociation, the

decreased portion of O anions in the conduction band

(Fig. 3f) obstructs a fast transfer from H to H+ by forming

OH with O from W3O9. Thus, the dissociation is endothermic

and therefore higher barriers than Zr3O6/Cu(111) and Ti3O6/

Cu(111) are expected.Mo3O9/Cu(111) displays the same behavior

as W3O9/Cu(111), which greatly differs from b-MoO3/Cu(111).14

In the case of b-MoO3/Cu(111), a barrierless water dissociation

is observed. By adopting different structures and sizes, the

O 2p in the conduction band of b-MoO3 shifts closer to the

Fermi level than that of Mo3O9, which allows a fast O–H bond

breaking.

Overall, one can see that the features near the Fermi level

are very important for water dissociation on M3O3x/Cu(111).

The presence of conduction band dominated by O 2p from

M3O3x allows a fast transfer of H from water to the O from

M3O3x. In addition, adopting the valence band dominated by

the contributions from Cu or metal ions in the oxide is able

to stabilize the dissociated OH from water. Mg3O3/Cu(111)

displays both features, being able to break the O–H bond

spontaneously at the interface. Other systems studied here

miss the first feature and water dissociation is hindered

kinetically and/or thermodynamically. Our previous studies

show that for b-MoO3/Cu(111), the reduction of oxide promotes

the water dissociation and therefore the WGS reaction. According

to the present calculations, the oxide trimers behave differently. In

the case of Mo3O9/Cu(111), Mo3O9 is heavily reduced and yet

water dissociation is endothermic. For Mg3O3/Cu(111), the stable

MgO in such small size and unique conformation can be further

oxidized, which facilitates the O–H bond cleavage. The PDOS of

metal ions and O 2p near the Fermi level are essential in this case.

III.3 WGS reaction on M3O3x/Cu(111)

To estimate the WGS activity of M3O3x/Cu(111), the correlation

between DFT-calculated DE for water dissociation and experimen-

tally measured WGS activity on Cu-based systems is employed

here. As shown in Fig. 4, the red line represents the correlation

determined in our previous studies by comparing DFT calculations

and experiment on model systems.14 For the Cu-based systems we

have studied previously,14 the water dissociation is always the slow

Fig. 4 Correlations of the DFT-calculated DE for water dissociation

vs. the experimentally measured WGS activity. The red line was taken

from our previous work14 and the dashed line was extrapolated

accordingly.

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online

This journal is c the Owner Societies 2012 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 16631

step under the reaction conditions. Therefore, the WGS reaction

can be greatly facilitated by the fast water dissociation. Accordingly,

the WGS activity can be estimated by only calculating the

reaction energy for water dissociation. Following this idea, we

add DE calculated in the present study in Fig. 4. One can see

that the activity towards water dissociation and therefore the

WGS reaction decreases in a sequence: Mg3O3/Cu(111) >

Zr3O6/Cu(111) > Ti3O6/Cu(111) > W3O9/Cu(111), Mo3O9/

Cu(111). Mg3O3/Cu(111) seems to be the best WGS catalyst

among the systems studied here, being able to break water

with no barrier. However, we also notice that the data point

for Mg3O3/Cu(111) is not in the range of the previously

determined red line (Fig. 4), but in the extrapolated region

(the dashed line). That is, it is possible that the water dissociation

may not be the slow step anymore and the sequential steps can be

the bottleneck for the WGS on Mg3O3/Cu(111).

To gain better understanding of WGS on Mg3O3/Cu(111),

we calculated the potential energy diagram (Fig. 5). The

reaction starts with the dissociation of water into OH and H

at the interface, followed by the chemisorption of CO on

Cu(111). CO is oxidized by OH adsorbed on the Cu site to

form carboxyl (HOCO) on Cu(111), which decomposes into

chemisorbed CO2 and H on Cu(111). Finally, CO2 desorbs

and H2 is released by recombining the adsorbed H on Cu(111)

and on Mg3O3. One can see in Fig. 5 that all the elementary

steps at or near the interface, which again demonstrates the

important role of the metal/oxide interface in the WGS

reaction. In terms of energetics, the first two steps, water

dissociation and CO adsorption, are highly exothermic, which

releases the energy of more than 2 eV. In contrast, the rest of

the reactions are uphill. In particular, the removal of H from

Mg3O3 in the phase of H2 is the most endothermic due to the

strong O–H interaction. Our calculations show that including

more water molecules leads to the H saturation of the supported

Mg3O3 at three O sites (Mg3O3H3) and the release of H2 is still

highly endothermic with the energy cost of at least 1 eV. Therefore,

the removal of H from Mg3O3 can be the key step, which slows

down the overall WGS reaction on Mg3O3/Cu(111). Similar

strong H interactions are also observed for Ti3O6/Cu(111) and

Zr3O6/Cu(111). In contrast, Mo3O9/Cu(111) and W3O9/Cu(111)

bind the adsorbates more weakly and the removal of H is more

thermodynamically favorable by at least 1 eV. However, the water

dissociation is activated in the meantime, which hinders the WGS

reaction (Fig. 4). DFT calculations on the transition states and

kinetic modeling are under way to obtain a better picture of the

kinetics.

It has been demonstrated that water dissociation is the rate-

limiting step for the WGS reaction on pure Cu systems,7,15 and

the WGS reaction on Cu can be facilitated by introducing

oxides to lower the barrier of water dissociation.14 According to

our present study of M3O3x/Cu(111), this is not necessarily the

case. Indeed, the strong interaction between O of oxides and H

from water leads to a fast water dissociation on oxide–Cu

systems. However, as a consequence, the removal of H from

oxide can be problematic, which blocks the O sites for further

water dissociation and therefore the WGS reaction. To achieve

the high WGS activity, the oxides of Cu–oxide nanocatalysts

have to compromise, being able to dissociate water easily, but

still allowing the facile H removal. The particle size of oxide can

be very important for the overall activity.

IV. Conclusion

We employed DFT to study the water dissociation and the

WGS reaction on a series of inverse model catalysts M3O3x/

Cu(111). The previous studies have identified water dissociation

Fig. 5 Potential energy diagram for the WGS on Mg3O3/Cu(111) (left) and the corresponding geometries of reactants, intermediates and

products (right, light brown line: Cu; light green: Mg; red: O, white: H, gray: C).

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online

16632 Phys. Chem. Chem. Phys., 2012, 14, 16626–16632 This journal is c the Owner Societies 2012

as the key step for the WGS reaction on pure Cu and various

Cu–oxide catalysts. The WGS reaction on Cu can be facilitated

by introducing oxides to lower the barrier of water dissociation.

Accordingly, the present calculations predict that the WGS

activity decreases in a sequence: Mg3O3/Cu(111) > Zr3O6/

Cu(111) > Ti3O6/Cu(111) > W3O9/Cu(111), Mo3O9/Cu(111).

Mg3O3/Cu(111) seems to be the best WGS catalyst among the

systems studied here, being able to break water with no barrier.

The oxide trimers behave differently from the big oxide clusters

studied previously.14 For the big oxide clusters, the reduction of

oxide has been found as the key to promote the water dissocia-

tion and therefore the WGS reaction on oxide/Cu(111) model

catalysts. In contrast, for the oxide trimers, it is more complex.

In the case of Mo3O9/Cu(111), Mo3O9 is heavily reduced, yet

water dissociation is endothermic. For Mg3O3/Cu(111), the

stable MgO in such small size and unique conformation can

be further oxidized, which facilitate the O–H bond cleavage.

The DOS of metal ions and O 2p near the Fermi level are

essential in this case. In addition, further studies show that

water dissociation is not the key step to control the WGS

reaction on Mg3O3/Cu(111). With the O 2p dominated

conduction band, a strong interaction between O of Mg3O3

and H from water is observed, which leads to a barrierless

water dissociation at the interface ofMg3O3/Cu(111). However, as

a consequence, the removal of H fromMg3O3 can be problematic,

which blocks the O sites for further water dissociation and

therefore the WGS reaction. Our results imply that to achieve

the high WGS activity, the oxides of Cu–oxide nanocatalysts

have to compromise, being able to dissociate water easily, but

still allowing the facile H removal. The particle size of oxide can

be very important for the overall activity. Our study provides

new insight into the design of the WGS catalysts.

Acknowledgements

This research was carried out at Brookhaven National

Laboratory under contract DE-AC02-98CH10886 with the

US Department of Energy, Division of Chemical Sciences.

The DFT calculations were carried out using the computing

facilities at the Center for Functional Nanomaterials at Brookhaven

National Laboratory and National Energy Research Scientific

Computing (NERSC) Center.

References

1 M. Z. Jacobson, W. G. Colella and D. M. Golden, Science, 2005,308, 1901–1905.

2 V. Subramani and S. K. Gangwal, Energy Fuels, 2008, 22, 814–839.3 W.Wang, S. Wang, X. Ma and J. Gong, Chem. Soc. Rev., 2011, 40,3703–3727.

4 J. J. Spivey, Catal. Today, 2005, 100, 171–180.5 R. Burch, Phys. Chem. Chem. Phys., 2006, 8, 5483–5500.6 J. A. Rodriguez, J. Graciani, J. Evans, J. B. Park, F. Yang,D. Stacchiola, S. D. Senanayake, S. Ma, M. Perez, P. Liu,J. Fernandez-Sanz and J. Hrbek, Angew. Chem., Int. Ed., 2009,48, 8047–8050.

7 P. Liu and J. A. Rodriguez, J. Chem. Phys., 2007, 126,1647051–1647058.

8 J. A. Rodriguez, P. Liu, J. Hrbek, J. Evans and M. Perez, Angew.Chem., Int. Ed., 2007, 46, 1329–1332.

9 H.-J. Freund, Surf. Sci., 2007, 601, 1438.10 C. T. Campbell, Surf. Sci. Rep., 1997, 27, 1.

11 J. A. Rodriguez, S. Ma, P. Liu, J. Hrbek, J. Evans and M. Perez,Science, 2007, 318, 1757–1760.

12 Q. Fu,W. Li, Y. Yao, H. Liu, H. Su, D.Ma, X. Gu, L. Chen, Z.Wang,H. Zhang, B. Wang and X. Bao, Science, 2010, 328, 1141–1144.

13 J. A. Rodriguez and J. Hrbek, Surf. Sci., 2010, 604, 241–244.14 P. Liu, J. Chem. Phys., 2010, 133, 2047051–2047052.15 A. A. Gokhale, J. A. Dumesic and M. Mavrikakis, J. Am. Chem.

Soc., 2008, 130, 1402–1414.16 Q. L. Tang, Z. X. Chen and X. He, Surf. Sci., 2009, 603, 2138–2144.17 W. E. Kaden, T. Wu, W. A. Kunkel and S. L. Anderson, Science,

2009, 326, 826–829.18 M. Valden, X. Lai and D.WGoodman, Science, 1998, 281, 1647–1650.19 S. Vajda, M. J. Pellin, J. P. Greeley, C. L. Marshall, L. A. Curtiss,

G. A. Ballentine, J. W. Elam, S. Catillon-Mucherie, P. C. Redfern,F. Mehmood and P. Zapol, Nat. Mater., 2009, 8, 213–216.

20 M. Wagner, S. Surnev, M. G. Ramsey, G. Barcaro, L. Sementa,F. R. Negreiros, A. Fortunelli, Z. Dohnalek and F. P. Netzer,J. Phys. Chem. C, 2011, 115, 23480–23487.

21 J. Graciani, J. J. Plata, J. F. Sanz, P. Liu and J. A. Rodriguez,J. Chem. Phys., 2010, 132, 1047031–1047038.

22 C. Popa, M. V. Ganduglia-Pirovano and J. Sauer, J. Phys. Chem. C,2011, 115, 7399–7410.

23 S. P. Price, X. Tong, C. Ridge, V. Shapovalov, Z. Hu, P. Kemper,H. Metiu, M. T. Bowers and S. K. Buratto, Surf. Sci., 2011, 605,972–976.

24 J. Zhu, H. Jin, L. Zang, Y. Li, Y. Zhang, K. Ding, X. Huang,L. Ning and W. Chen, J. Phys. Chem. C, 2011, 115, 15335–15344.

25 G. Kresse and J. Joubert, Phys. Rev. B: Condens. Matter Mater.Phys., 1999, 59, 1758–1775.

26 J. Perdew, J. Chevary, S. Vosko, K. Jackson, M. Pederson,D. Singh and C. Fiolhais, Phys. Rev. B: Condens. Matter Mater.Phys., 1992, 46, 6671–6687.

27 G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater.Phys., 1993, 47, 558–561.

28 G. Kresse and J. Furthmuller, Comput. Mater. Sci., 1996, 6, 15–21.29 G. Kresse and J. Furthmuller, Phys. Rev. B: Condens. Matter

Mater. Phys., 1996, 54, 11169–11186.30 G. Henkelman, A. Arnaldsson and H. Jonsson, Comput. Mater.

Sci., 2006, 36, 354–360.31 R. F.W. Bader and P.M. Beddall, J. Chem. Phys., 1972, 56, 3320–3329.32 R. F. W. Bader, Atoms in Molecules: A Quantum Theory, Clarendon

Press, Oxford, 1994.33 W. A. Saunders, Z. Phys. D: At., Mol. Clusters, 1989, 12, 601–603.34 P. J. Ziemann andA.W. Castleman, J. Chem. Phys., 1991, 94, 718–728.35 E. de la Puente, A. Aguado, A. Ayuela and J. M. Lopez, Phys. Rev. B:

Condens. Matter Mater. Phys., 1997, 56, 7607–7614.36 K. Kwapien, M. Sierka, J. Dobler, J. Sauer, M. Haertelt, A. Fielicke

and G. Meijer, Angew. Chem., Int. Ed., 2011, 50, 1716–1719.37 L. Hong, H. Wanga, J. Cheng, L. Tang and J. Zhao, Comput..

Theor. Chem., 2012, 980, 62–67.38 G. Von Helden, A. Kirilyuk, D. van Heijnsbergen, B. Sartakov,

M. A. Duncan and G. Meijer, Chem. Phys., 2000, 262, 31–39.39 S. Li and D. A. Dixon, J. Phys. Chem. A, 2010, 114, 2665–2683.40 H. Wu and L. S. Wang, J. Chem. Phys., 1997, 107, 8221–8228.41 S. Li and D. A. Dixon, J. Phys. Chem. A, 2008, 112, 6646–6666.42 X. Huang, H.-J. Zhai, B. Kiran and L.-S. Wang, Angew. Chem.,

Int. Ed., 2005, 44, 7251–7254.43 S. Li and D. A. Dixon, J. Phys. Chem. A, 2006, 110, 6231–6244.44 S. Li and D. A. Dixon, J. Phys. Chem. C, 2011, 115, 19190–19196.45 K. Honkala, A. Hellman and H. Gronbeck, J. Phys. Chem. C,

2010, 114, 7070–7075.46 W. Langel and M. Parrinello, J. Chem. Phys., 1995, 103, 3240–3252.47 C. Xu and D. W. Goodman, Chem. Phys. Lett., 1997, 265,

341–346.48 F. Cora, M. G. Stachiotti, C. R. A. Catlow and C. O. Rodriguez,

J. Phys. Chem. B, 1997, 101, 3945–3952.49 F. Cora, A. Patel, N. M. Harrison, C. Roettic and C. R. A.

Catlowa, J. Mater. Chem., 1997, 7, 959–967.50 A. Michaelides, A. Alavi and D. A. King, J. Am. Chem. Soc., 2003,

125, 2746–2755.51 H. P. Bonzel, G. Pirug and J. E. Muller, Phys. Rev. Lett., 1987, 58,

2138–2141.52 H. Xu, R. Q. Zhang, A. M. C. Ng, A. B. Djurisic, H. T. Chan,

W. K. Chan and S. Y. Tong, J. Phys. Chem. C, 2011, 115,19710–19715.

Dow

nloa

ded

by B

NL

Res

earc

h L

ibra

ry o

n 03

Dec

embe

r 20

12Pu

blis

hed

on 1

3 A

ugus

t 201

2 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

2CP4

2091

K

View Article Online