Molecular reorientation of water adsorbed on charged Ag(111) surfaces

Molecular reorientation of water adsorbed on chargedAg(1 1 1) surfaces

Cristi�aan G. S�aanchez *

Atomistic Simulation Group, School of Mathematics and Physics, Physics Building, Queen�s University Belfast, University Road,

BT7 1NN Belfast, Northern Ireland, UK

Received 8 November 2002; accepted for publication 22 January 2003

Abstract

In this work we present first principles calculations of water adsorption over charged Ag(1 1 1) surfaces. The ori-

entation of the adsorbed water molecule with respect to the surface changes from oxygen pointing away from the

surface at negative charges to oxygen pointing towards the surface at positive charges. At zero charge the water

molecule is oriented approximately parallel to the surface plane. Complete orientation of the molecule in the direction

of the field is achieved for a critical charge density of 15 lCcm�2 for both positive and negative charges.

� 2003 Elsevier Science B.V. All rights reserved.

Keywords: Density functional calculations; Chemisorption; Water; Silver; Low index single crystal surfaces; Solid–liquid interfaces

1. Introduction

Double layer modelling plays a fundamental

role in theoretical electrochemistry since an accu-

rate knowledge of the double layer is needed in any

attempt to describe other electrochemical phe-

nomena such as charge transfer processes. Manydifferent approaches have been applied to the study

of this subject, ranging from first principles mo-

lecular dynamics to Monte Carlo simulations.

These works have been extensively reviewed re-

cently [1–4]. In spite of the effort dedicated there is

not a general consensus on a model for the elect-

rochemical double layer, and existing models have

difficulties to rationalise all of the abundant body

of experimental data. This limited success can be

ascribed to the inherent complexity of the problem.

The interface is a highly anisotropic environment in

which large changes of composition occur within a

very narrow spatial extent. The abrupt difference in

molecular structure between the phases on either

side of the interface, that makes them so interest-ing, is extremely challenging from a theoretical

point of view. Techniques for the study of both the

charged metal surface [5,6] and water from first

principles ([7] and references therein) have been

made available only recently. No fully ab initio

simulation yet exists of the charged interface, only

for the neutral case [8,9].

The study of water adsorption over metallicsurfaces provides valuable information that may

help the understanding of the electrochemical

interface and there is abundant experimental in-

formation available [10,11]. A number of first

* Tel.: +44-28-90273557; fax: +44-28-90241958.

E-mail address: [email protected] (C.G. S�aanchez).

0039-6028/03/$ - see front matter � 2003 Elsevier Science B.V. All rights reserved.

doi:10.1016/S0039-6028(03)00080-3

Surface Science 527 (2003) 1–11

www.elsevier.com/locate/susc

principles theoretical studies have addressed this

system. Paredes Olivera et al. [14] present some

results on the adsorption energy of water on the

Ag(1 1 1) surface. The aim of this work was the

study of hydronium adsorption. The authors could

not determine the adsorption geometry of the watermolecule due to the small energy difference ob-

tained between the upright and flat configurations.

Izvekov and Voth [9] have performed first princi-

ples molecular dynamics of the water/Ag(1 1 1) in-

terface and present some result on the adsorption

of a single molecule. Kua andGoddard [15] address

the problem of methanol oxidation and show re-

sults for water adsorption onPt(1 1 1). Ohwaki et al.[16] have studied water adsorption on Pt(1 1 1) and

Pt(1 0 0). This authors also studied the effect of the

electric field. Two different fields corresponding to

positive and negative charges of about 10 lCcm�2

were used. Their results indicate that for the posi-

tive field the tilting of the water molecule increases

to 41 degrees and for the negative field the water is

normal to the surface with the hydrogen atomspointing to the surface. Jin and Head [17] studied

water adsorption on Al(1 1 1) and Ignaczak and

Gomes [18] on Cu(1 0 0), Ag(1 0 0) and Au(1 0 0)

surfaces. The results presented in this works for

water orientation and adsorption energy are com-

piled in Table 1 together with the model used to

represent the electronic structure of the metal–

water system. Previous work about these and othersurfaces is described by Nazmutdinov and Shapnik

[19].

All these works (with the exception of [9]) have

modelled the metallic surface by means of finite

clusters. The cluster method is a valuable tool for

the study of adsorption phenomena related to the

electrochemical interface [20]. This approach how-

ever, has some limitations that have been ad-dressed by us elsewhere [21] and are discussed by

the authors in some on these works [14,16]. The

most important ones are the oscillatory depen-

dence of adsorption energies on cluster size and

the cluster dipole moment. Although these prob-

lems can be solved in part by carefully preparing

the cluster�s electronic state and geometry, themethod used in the present work which models theinfinite surface with periodic boundary conditions

is more adequate to describe the electronic struc-

ture of the metal.

The presence of water on the electrochemical

interface is often neglected, but it is known to

profoundly affect processes occurring at electrodes

[22]. Since the early flip-flop models [23] the pic-

ture of water molecules close to the electrodechanging their orientation from oxygen-up to oxy-

gen-down in response to the electric field has

dominated our understanding of the electrochem-

ical double layer. These ideas have been confirmed

by experimental data such as X-ray diffraction [24]

and IR spectroscopy [25–27] (see also [10] and

references therein). But, to the best of our knowl-

edge, up to now no first principles results existto support this ideas. In this work we present re-

sults on the relative orientation with respect to the

Table 1

Compilation of first principles results existing in the literature for water adsorption on different metallic surfaces

Surface Method Angle (deg) Eads (eV) Ref.

Ag(1 1 1) Cluster, 28 atoms, MP2 undetermined 0.35 [14]

Ag(1 1 1) Super-cell, PW, BLYP 30 0.53 [9]

Pt(1 1 1) Cluster, 8 atoms, B3LYP 20 0.70 [15]

Pt(1 1 1) Cluster, 7 atoms, B3LYP 16.5 0.56 [16]

Al(1 1 1) Cluster, 10 atoms, ROHF 90 0.56–0.87 [17]

Cu(1 0 0) Cluster, 12 atoms, B3LYP 35 0.31 [18]

Ag(1 0 0) Cluster, 12 atoms, B3LYP 40 0.27 [18]

Au(1 0 0) Cluster, 12 atoms, B3LYP 25 0.30 [18]

Ag(1 1 1) Super-cell, PBE )17 0.2 This work

Positive angles indicate that the oxygen atom is pointing towards the surface and negative ones that the oxygen atom is pointing away

from the surface. The acronyms used in the method column are as follows: MP2 ¼ Moller–Plesset second-order perturbation theory,PW ¼ plane wave basis set, BLYP ¼ density functional theory using BLYP GGA functional, B3LYP ¼ Becke 3 parameter hybrid

Hartree–Fock DFT functional, PBE ¼ DFT using PBE GGA functional.

2 C.G. S�aanchez / Surface Science 527 (2003) 1–11

surface of a water molecule adsorbed on a charged

Ag(1 1 1) surface as a function of charge. The

critical charge necessary to achieve full orientation

of the water molecule in the direction of the field is

obtained. Fully ab initio calculations are compu-

tationally very demanding and hence at present theproblem of a changed electrode in contact with

water containing ions at a non-zero temperature

cannot be calculated exactly. The work described

in this paper represents a first step towards the

fully ab initio description of the charged electro-

chemical interface.

2. The model

The electronic structure is treated quantum me-

chanically using standard state-of-the-art methods

[21,29]. In this section we give the technical details

necessary to reproduce the work. In order to rep-

resent the surface we use a slab geometry and

periodic boundary conditions. The metal is repre-sented by a four layer slab with the lattice param-

eter obtained from the bulk calculation. Water is

adsorbed on both sides of the slab. Although we are

interested in studying an isolated water molecule

adsorbed on the surface our model requires the use

of a periodic system. We require a super-cell of the

lowest possible coverage so as to avoid water–water

interactions. We use a ð2� 2Þ surface unit cell,which is equivalent to a water monolayer with

coverage degree of 1/4. Test calculations made for a

ð3� 3Þ unit cell (coverage degree of 1/9) showed novariation on the adsorption geometries obtained

and hence we do not expect that the presence of

water molecules in neighbouring cells would affect

the results significantly. A calculation of the ð2� 2Þwater monolayer isolated in the vacuum has anenergy per molecule 0.01 eV lower than the isolated

water molecule, and this correction has been taken

into account for the calculation of the adsorption

energy at zero charge. A vacuum space equivalent

to six (1 1 1) layers separates periodic images of the

metal slab on the direction parallel to the surface.

The size of the super-cell in the direction normal to

the surface is 24.3�AA, this gives a separation betweenwater molecules in neighbouring cells in the direc-

tion normal to the surface of 11.6 �AA. The super-cell

used is shown in Fig. 1. The outermost layer of the

substrate was allowed to relax in all optimisations.

To model the electronic structure of the system

we use the generalized gradient approximation

(GGA) to density functional theory (DFT) in the

Perdew–Burke–Ernzerhof (PBE) [28] form. Cal-

culations were carried out using the programSIESTA [30,31] a code designed for DFT calcu-

lations in systems with a large number of atoms.

The method used to include the effects of the

electric field has been described in detail elsewhere

[5,6] in the context of plane wave calculations, and

has been adapted by us to the SIESTA code. The

method relies on the inclusion of a charged plane

at the boundary of the cell parallel to the surface;bearing a charge equal in magnitude but opposite

sign to the charge on the slab. Since overall the

system (metallic slab plus charged plane) is neu-

tral, only dipole–dipole and higher multi-pole

moment interactions can occur between periodic

images as in any slab calculation. The dipole–

dipole interaction is avoided by using a symmetric

slab with no net dipole moment. This is achievedby adsorbing water molecules on both sides of a

symmetric slab placed at the centre of the cell.

Albeit no symmetry restriction is imposed during

the geometry optimisation the symmetry is main-

tained within 0.001 �AA.

Fig. 1. The surface super-cell used in the calculations corre-

sponds to the area enclosed by the black polygon. This is a

ð2� 2Þ surface unit cell with an equivalent water coverage of1/4.

C.G. S�aanchez / Surface Science 527 (2003) 1–11 3

Norm conserving Troullier–Martins [32] pseudo-

potentials were used to represent the atomic cores.

We have used a basis set composed of pseudo-

atomic orbitals for the expansion of the Kohn–

Sham orbitals. This set is constructed using a

generalisation of the method proposed by Sankeyand Niklewski [33], which consists of solving the

problem of the pseudo-atom with the boundary

condition that the electronic valence wave func-

tions vanish beyond a certain cutoff radius. In this

way, the number of non-zero Hamiltonian matrix

elements is dramatically reduced. To give more

flexibility to the basis set, a second group of va-

lence orbitals and a set of polarisation orbitals areadded to the valence set. This corresponds to a

double-f plus polarisation (DZP) basis set in theusual quantum chemistry terminology, with po-

larisation functions for both heavy elements and

hydrogen. The procedure used to construct this set

is described in detail elsewhere [34,35]. The local-

isation radius of the basis set (corresponding to an

energy shift of 0.025 eV [33–35]) was chosen as acompromise between accuracy and computational

efficiency. Further increase of the localisation ra-

dius changed adsorption energies in less than 0.05

eV, the change in forces being smaller than the

tolerance for geometry optimisation (0.005 eV/�AA).In order to check for the accuracy provided by the

basis and pseudo-potential, calculations for bulk

silver and an isolated water molecule were per-formed. For bulk silver we obtained a lattice

constant of 4.21 �AA and a bulk modulus of 0.83Mbar, which can be compared to the experimen-

tal values of 4.09 �AA and 1.04 Mbar respectively.The relatively expanded structure and small bulk

modulus are a result of the application of the

GGA. A plane wave calculation performed with

the Carr–Parrinello moleclar dynamics (CPMD)[36] code with an essentially complete basis set

(using a cutoff of 70 Ry) gives values of 4.20 �AA and0.82 Mbar for the lattice constant and bulk mod-

ulus respectively. The same plane wave calculation

using the local density approximation (LDA) gives

4.05 �AA and 1.29 Mbar, much improving the latticeconstant but not so the bulk modulus, which is off

by a similar magnitude but opposite sign. It is awell know fact that GGA works better for mo-

lecular systems and LDA does a better job for

some condensed matter systems due to a fortuitous

cancellation of errors [37]. However, in order to

describe water properly, which is the main objec-

tive of the present study, a GGA approximation

was required. For the isolated water molecule we

obtain an OH distance of 0.979 �AA and an HOHangle of 103.9 degrees. The experimental values

are 0.957 �AA and 104.5 respectively and the planewave results obtained for a 130 Ry cutoff are 0.969�AA and 104.3 degrees.

3. Results and discussion

The geometry obtained for water adsorbed over

the neutral surface is shown in Fig. 2a and b. No

attempt was made to optimise the structure at

adsorption sites other than top since from our own

calculations [38] and the results published in the

literature [9,11] this is the most stable site. The

geometry reported by Izvekov and Voth [9] was

used as a starting point for the optimisation. In theoptimised structure the oxygen atom is slightly

displaced from the top site and the hydrogen

atoms point slightly towards the surface. The mole-

cular plane forms an angle of 17 degrees with re-

spect to the plane of the substrate. The relaxation

of the metal layer is very small. The silver atom

closest to the water molecule shows an inward

relaxation of about 0.01 �AA. This result is in con-trast with the geometry normally assumed for ex-

perimental evidence [11] for adsorption at UHV,

where the tilt is interpreted to be in the opposite

direction. Similarly, for the electrochemical inter-

face the temperature coefficient of the potential of

zero charge seems to indicate a natural orientation

of water molecules with the oxygen atom pointing

towards the surface [12]. Experimental UHV data,however, is affected by the formation of water

clusters and should be taken with care in order to

predict the structure of water adsorbed at low cov-

erage. With respect to the thermal coefficient of the

potential of zero charge it has been suggested that it

can be explained from the metal electronic contri-

bution [13].

Results obtained from cluster calculations forother surfaces in general agree that the water

molecule is tilted with the hydrogen atoms point-

4 C.G. S�aanchez / Surface Science 527 (2003) 1–11

ing away from the surface [15,16,18] with the ex-

ception of the Al(1 1 1) surface [17] for which theresults indicate orientation normal to the surface.

Previous first principles results for the Ag(1 1 1)

surface are inconclusive with respect to the ad-

sorption geometry. In Ref. [14], as mentioned in

Section 1, the tilting angle could not be obtained

because of the small energy difference between

different orientations, and this was attributed to a

limitation of the cluster model. Most cluster cal-culations avoid full geometry optimisation; nor-

mally the internal geometry of the adsorbate and

the adsorption site are constrained in order to save

computational time and no attempt to relax the

substrate is made. Furthermore, the experimental

lattice constant of the metal is used, which can be

different from the optimum lattice constant for the

model used. These approximations may be im-portant to determine the final water configuration.

The result that is more comparable to ours in

terms of the model used is that of Ref. [9]. Thegeometry obtained by these authors differs from

ours in that their results show the water molecule

tilted outwards at an angle of 30 degrees with re-

spect to the surface plane. This discrepancy might

be attributed to the lower (less accurate) plane

wave cutoff energy used in Ref. [9] (60 Ry). In our

experience cutoffs of over 90 Ry are needed in

order to ensure proper convergence of the forceswhen norm conserving Troullier–Martins pseudo-

potentials are used for oxygen. We have performed

plane wave calculations using the CPMD code

with a plane wave energy cutoff of 70 Ry and ob-

tained an adsorption geometry very similar to the

one presented here. All of this can be rationalised

from the fact that the potential energy surface for

water adsorption over Ag is a relatively shallowfunction of the tilt angle and molecular position

Fig. 2. Water orientation with respect to the substrate. For the sake of clarity only a ð1� 1Þ surface unit cell of substrate is shown(note that the unit cell used for the calculation is a ð2� 2Þ cell, 4 times larger, as shown in Fig. 1). (a) Zero charge, top view; (b) zerocharge, side view; (c) charge density of 15.7 lCcm�2, top view; (d) charge density of 15.7 lCcm�2, side view; (e) charge density of )15.7lCcm�2, top view; (f) charge density of )15.7 lCcm�2, side view.

C.G. S�aanchez / Surface Science 527 (2003) 1–11 5

and hence it is difficult to obtain a conclusive op-

timum geometry. The high surface diffusion rate

experimentally observed is an evidence of this

characteristic of the potential energy surface. With

respect to the adsorption energy, we can estimate a

value of )0.2 eV, taking into account the basis setsuperposition error by means of the counterpoise

method. The result obtained is comparable to

previous theoretical [9,14,18] and experimental

[10,11] results, the later being somewhat larger

owing to water–water interaction as already men-

tioned.

The most important results in this work concern

the adsorption over charged surfaces. A total of 10different surface charges where studied between

)15.7 and +15.7 lCcm�2 corresponding to an

excess of between 0.0075 and )0.0075 electrons persurface atom. For a Ag(1 1 1) surface in contact

with a 50 mM KClO4 solution the highest negative

and positive charges used correspond to potentials

of approximately )1.5 and )0.5 V vs. SCE re-

spectively [40]. The convergence criterion used forthe optimisation of the charged systems was 0.03

eV/�AA tolerance in the residual forces (for the

neutral surface a tolerance of 0.005 eV/�AA was

used). Once again due to the smooth variation in

energy with orientation the use of the looser con-

vergence criterion gives an uncertainty in the angle

formed with respect to the surface. This uncer-

tainty is less than �1 degree for positive and about�6 degrees for negative charges. For the sake ofconsistency, the results for the neutral surface

shown in Figs. 3, 6 and 7 are those obtained with

the looser criterion.

In Fig. 3 the angle formed between the water

molecule plane and the surface plane is shown.

The curve shows a linear behaviour for charges

between � 5 lCcm� 2, while for larger charges astrong deviation from linearity occurs which is

more pronounced for negative surface charges.

Complete orientation in the direction of the field is

attained for both positive and negative charge

densities of around 15 lCcm�2 or above (see Fig.

2). Hence, complete dielectric saturation of the

water sub-monolayer in contact with the metal is

expected to occur at these charge densities. Therelaxation of the substrate for negative and small

positive charges is similar to that observed for the

neutral surface. For larger positive charges how-

ever, the silver atom closest to the water molecule

relax outwards and this relaxation grows with

charge up to 0.08 �AA for the highest positive charge.The fact that both hydrogen atoms approach

the surface simultaneously can be explained from

an electrostatic point of view. The charged system

tends to displace the charge away from the surface

as far as possible in order to minimise the energy.

For a negatively charged surface, this accounts to

moving the oxygen atom away and attract the

hydrogen atoms closer; the optimum way to ac-

complish this is with both hydrogen atoms at thesame distance to the surface.

Some insight into the nature of the water–metal

interaction can be gained from the analysis of the

projected densities of states (PDOS) [39], defined

as:

guð�Þ ¼Xi;j

Z Z1

4p3dð�� �iðkÞÞ

� hwiðk; rÞj/ujðk; rÞid�dk ð1Þ

-15 -10 -5 0 5 10 15

σ / µC cm-2

-80

-60

-40

-20

0

20

40

60

80

100

Ang

le w

rt s

urfa

ce p

lane

/ de

g

Fig. 3. Angle between the water molecule plane and the surface

plane as a function of charge density on the surface. Positive

angles indicate that the oxygen atom is pointing towards the

surface and negative ones that the oxygen atom is pointing

away from the surface.

6 C.G. S�aanchez / Surface Science 527 (2003) 1–11

where /ujðk; rÞ represents the basis set orbital j onatom u, wiðk; rÞ the i Kohn–Sham orbital with ei-genvalue �iðkÞ and the integration on k runs over

the Brillouin zone. In practice, the integration over

the Brillouin zone is replaced by a finite sum, andDirac�s d function is replaced by a Gaussian ofwidth r (0.25 eV was used for the PDOS shownhere):

guð�Þ ¼Xi;j;k

Zwk

1

rffiffiffiffiffiffi2p

p

� exp �� �iðkÞ2r2

� �hwiðk; rÞj/ujðk; rÞid�

ð2Þ

guð�Þ represents the portion of the total density ofstates due to orbitals on atom u. In Fig. 5b thedensity of states of the water molecule is shown,the three highest occupied molecular orbitals can

be seen. Iso-surface plots for these orbitals are

shown in Fig. 4. In Fig. 5b the total DOS is di-

vided into H and O contributions. From this we

can see that orbital 1B2 is formed from almost

equal contributions from oxygen and hydrogen

atomic orbitals. Orbital 3A1 (the r lone pair) has asmall contribution from H orbitals which is evensmaller for 1B1 (the p lone pair). 1B2 is the main rbonding orbital between O and H atoms, orbitals

3A1 and 1B1 are basically oxygen lone pairs. Upon

adsorption on the neutral surface (Fig. 5d) both

1B2 and 3A1 orbitals are stabilised but the most

important changes occur in the 1B1 orbital. This

orbital interacts mainly with the d-band (shown in

Fig. 5a) producing significant stabilisation andwidening. This fact explains that the most stable

orientation of the water molecule over the neutral

surface is almost horizontal. This is the orientation

that allows maximum overlap of the 1B1 orbital

with the surface. Integration of the OþH PDOSfrom )7 eV up to the Fermi level indicates that theoccupation of these orbitals increases, henceforth asmall negative charge (�0.02 electrons) is trans-ferred to the water orbitals. The localisation of this

charge is however, difficult to assess from the

analysis of the PDOS. Upon charging of the sur-

face the interaction with the electric field is even-

tually dominant and a perpendicular orientation is

attained. In this orientations the overlap between

the 1B1 orbital with the surface is smaller and itremains almost unchanged. The interaction is not

equivalent for positive and negative charges (com-

pare 4c and 4e) and some mixing can be appreci-

ated for positive charges when the orbital is closer

to the surface. Other differences exist between

positive and negative charges. While orbital 3A1energy remains almost unchanged the 1B2 orbital

energy is higher with respect to neutral charge andthis effect is more important for positive charges.

The same trend is observed for the 1B1 orbital.

This changes in orbital energies may be a result of

the combination of both orientation change and

charge effects. From the integration of the PDOS

in Fig. 5a change in occupation of around 0.1

electrons can be observed for both 3A1 and 1B1orbitals. This occupation increase is positive fornegative charge and the opposite holds for positive

surface charge.

In Fig. 7 of Ref. [40] Valette shows a plot of the

inner layer capacity of the water/Ag(1 1 1) interface

as a function of charge density. The inner layer

capacitance has a maximum at the potential of

zero charge (pzc) and levels off at about the same

Fig. 4. The three highest occupied molecular orbitals of H2O. The energy order is 1B1 > 3A1 > 1B2.

C.G. S�aanchez / Surface Science 527 (2003) 1–11 7

charge densities we obtain for full water orienta-

tion. Our results support the idea that the lowering

of the inner layer capacitance at large charge

densities is due to the dielectric saturation of the

water layers closest to the electrode. According toBockris et al. [23] the dielectric saturation of the

first water layer occurs in the immediate vicinity of

the pzc and hence this water layer has a small

contribution to the capacitance. Our results indi-

cate however that adsorbed water molecules retain

orientational polarisability over a much wider

range of charges. This is due to the interaction with

the metal, which acts as a restoring force orientingthe molecules parallel to the surface, and opposing

to the effects of the field. The dielectric constant of

these water molecules is lower than that of bulk

water molecules, but on the other hand they do not

saturate close to the pzc as would be the case for

free water molecules. The interaction with the

metal plays a similar role to thermal agitation and

interaction with other water molecules.Our calculations agree with recent interfacial IR

spectroscopy results that indicate that for Au(1 1 1)

[25] and Pt(1 1 1) [26] surfaces water orientation

changes from hydrogen-down to hydrogen-up in

going from negative to positive surface charges. It

is interesting to compare our results with those

obtained by Toney and co-workers [24] using

surface X-ray diffraction. These authors presentoxygen distribution functions in the direction per-

pendicular to the Ag(1 1 1) surface for positive and

negative charge surfaces. In Fig. 6 the distances

between oxygen and hydrogen atoms with respect

to the outermost substrate plane are shown. Val-

ues for negative charges are somewhat noisy owing

to the higher uncertainty in the angle we already

mentioned. The 3.4 �AA saturation value we obtainfor the oxygen–silver distance for negative surface

charge agrees with the first peak in the distribution

shown by Toney and co-workers. For positive

charges the comparison is more difficult since the

peak in the experimental distribution is broad, but

our value of 2.6 �AA is within the broad maximum inthe distribution. Some models of the inner layer

have taken the water–surface distance as a varyingparameter and this distance has been considered to

change appreciably (by around 0.7 �AA) [2,41,42].Our calculations do not support this large ampli-

0

20 DS

01.5

34.5

01.5

34.5

01.5

34.5

O+HOH

-12 -10 -8 -6 -4 -2 0 2

E / eV

01.5

34.5

Ag

H2O

σ>0

σ=0

σ<0

1B23A1

1B1

(a)

(b)

(c)

(d)

(e)

Fig. 5. PDOS for the H2O/Ag(1 1 1) system as a function of energy, the energy zero corresponds to the Fermi level. (a) Densities of

states projected over silver D and S orbitals. (b) PDOS for the isolated H2O molecule. The DOS projected over both H and O orbitals,

only O orbitals, and only H orbitals is shown. (c) PDOS for the positively charged H2O/Ag(1 1 1) system (r ¼ 10 lCcm�2). (d) PDOS

for the neutral H2O/Ag(1 1 1) system. (e) PDOS for the negatively charged H2O/Ag(1 1 1) system (r ¼ �10 lCcm�2).

8 C.G. S�aanchez / Surface Science 527 (2003) 1–11

tude, but the tendency obtained is the same. Even

if the O–Ag distance experiences variations of the

order of 0.8 �AA, the centre of charge of the watermolecule remains quite still, varying about 0.2 �AAon going from negative to positive charges. The

molecule is closer to the surface for positive

charges.

In Fig. 7 we have plotted the adsorption energy

of the water molecule as a function of surface

charge density. This adsorption energy has been

calculated as the difference in energy between thecharged surface covered with water at its optimum

orientation and the sum of the energies of the clean

surface bearing the same charge and the isolated

water molecule:

Eads ¼ EH2O=Agð1 1 1ÞðrÞ � EAgð1 1 1ÞðrÞ � EH2O ð3Þ

In order to explain the parabolic shape of thecurve in Fig. 7 we can give the following argument:

For relatively large charges, the main contribution

to the total energy of both the clean silver surface

and the water covered surface is electrostatic [6].

The system is a behaves like a parallel plate ca-

pacitor. The charge distribution over the surface

forms one of the plates and the countercharge

placed at the reference electrode the second one.

The electrostatic contribution is therefore propor-tional to the square of the charge placed on the

plates, and inversely proportional to the (integral)

capacitance. The parabolic shape of the curve

can be understood from the fact that this adsorp-

tion energy represents the difference in energy be-

tween a capacitor with and without a dielectric

between its plates. The system containing water

has a higher capacitance, hence its energy is lower,the difference in energy with respect to the clean

surface being proportional to the square of the

charge. For equivalent charge densities the adsorp-

tion is stronger for the positively charged surface

owing to the favourable metal–oxygen interaction.

The applicability of our model to the interpre-

tation of experimental results for a charged metal

surface in contact with bulk water is limited by theextent to which water–water interactions affect

these results. In other words, how good can we

expect our calculations to model the behaviour of

an adsorbed water molecule in a real electrochem-

ical interface? We can give a qualitative answer to

this question by analysing the relative importance

of the different interactions involved. As an esti-

mate of the water–water interaction energy we maytake the energy of a typical hydrogen bond in bulk

water of around 0.16 eV (15 kJmol�1) [43]. Around

-15 -10 -5 0 5 10 15

σ / µC cm-2

2.6

2.8

3

3.2

3.4

d / Å

d O-Agd H-Ag

Fig. 6. Oxygen and hydrogen atoms distances to the first sub-

strate layer as a function of charge density on the surface.

-20 -10 0 10 20

σ / µC cm-2

-1

-0.8

-0.6

-0.4

-0.2

Ead

s / e

V

Fig. 7. Water adsorption energy as a function of charge density

on the surface. The values plotted are not BSSE corrected.

C.G. S�aanchez / Surface Science 527 (2003) 1–11 9

the point of zero charge, the water–metal interac-

tion energy is comparable to the water–water in-

teraction and hence we expect the interaction with

other adsorbed molecules, and molecules from the

solution side, to be important in determining the

structure. At higher charges however, the interac-tion with the metal and the electric field is much

stronger than the interaction with other water

molecules. Therefore we argue that our results

better represent the behaviour of water at large

surface charge densities. The problem with bulk

water is that it cannot be modelled by a single

distribution of water molecules, but a statistical

sampling is necessary. First principles moleculardynamics or Monte Carlo simulations of the

charged metal–water interface would provide a

definitive answer about the extent to which water–

water and water–metal interactions and thermal

agitation determine the structure of the inner layer.

Work in this direction is underway.

4. Conclusions

We have studied the adsorption of a water

molecule on charged Ag(1 1 1) surfaces. The ori-

entation relative to the surface for different surface

charge densities was obtained. The metal–water

bond is due to the interaction of the 1B1 oxygen

lone pair with the surface, which is optimal whenthe water molecule is parallel to the surface. The

results agree with experimental findings that water

orientation in the electrochemical interface chan-

ges from oxygen-up to oxygen-down on going

from negative to positive charges. The critical

surface charge needed for full orientation with the

field has been found to be around 15 lCcm�2. This

supports the idea that the lowering and levellingof the inner layer capacitance at this charge den-

sities is due to dielectric saturation of the water

layer closest to the electrode. These results pro-

vide new and important information about the

role of the water–metal interaction in determining

the properties of the inner layer. Results concern-

ing the internal geometry of the water molecule,

vibrational frequencies and the nature of the me-tal–water bond will be published in detail else-

where.

Acknowledgements

The author is grateful to Ruth M. Lynden-Bell,

Alexander Y. Lozovoi and Jorge J. Kohanoff for

fruitful discussions and to the authors of the SI-ESTA program who kindly provided their code.

This work was supported by EPSRC through

grant GR/M03931.

References

[1] R. Parsons, Chem. Rev. 90 (1990) 813.

[2] W. Schmickler, Chem. Rev. 96 (1996) 3177.

[3] W. Schmickler, Ann. Rep. Prog. Chem. Sect. C 95 (1999)

117.

[4] R. Guidelli, W. Schmickler, Electrochim. Acta 45 (1999)

2317.

[5] A.Y. Lozovoi, A. Alavi, J. Kohanoff, R.M. Lynden-Bell, J.

Chem. Phys. 115 (2001) 1661.

[6] A.Y. Lozovoi, A. Alavi, submitted for publication.

[7] P. Vassilev, C. Hartnig, M.T.M. Koper, F. Frechard, A.R.

van Santen, J. Chem. Phys. 115 (2001) 9815.

[8] S. Izvekov, A. Mazzolo, K. VanOpdorp, G. Voth, J. Chem.

Phys. 114 (2001) 3248.

[9] S. Izvekov, G.A. Voth, J. Chem. Phys. 115 (2001) 7196.

[10] M.A. Henderson, Surf. Sci. Rep. 46 (2002) 1.

[11] P.A. Thiel, T.E. Madey, Surf. Sci. Rep. 7 (1987) 211.

[12] A. Hamelin, L. Doubova, L. Stoicoviciu, S. Trasatti, J.

Electroanal. Chem. 244 (1988) 133.

[13] R. Guidelli, G. Aloisi, E. Leiva, W. Schmickler, J. Phys.

Chem. 92 (1988) 6671.

[14] P. Paredes Olivera, A. Ferral, E.M. Patrito, J. Phys. Chem.

B 105 (2001) 7227.

[15] J. Kua, W.A. Goddard III, J. Am. Chem. Soc. 121 (1999)

10928.

[16] T. Ohwaki, K. Kamegai, K. Yamashita, Bull. Chem. Soc.

Jpn. 74 (2001) 1021.

[17] S. Jin, J.D. Head, Surf. Sci. 318 (1994) 204.

[18] A. Ignaczak, J.A.N.F. Gomes, J. Electroanal. Chem. 420

(1997) 209.

[19] R.R. Nazmutdinov, M.S. Shapnik, Electrochim. Acta 41

(1996) 2253.

[20] C. S�aanchez, E.P.M. Leiva, in: W. Vielstich, A. Lamm, H.

Gasteiger (Eds.), Handbook of Fuel Cells: Fundamentals,

Technology and Applications, vol. 2, John Wiley & Sons,

Chichester, 2003 (Chapter 12).

[21] C. S�aanchez, E.P.M. Leiva, in: W. Vielstich, A. Lamm, H.

Gasteiger (Eds.), Handbook of Fuel Cells: Fundamentals,

Technology and Applications, vol. 2, John Wiley & Sons,

Chichester, 2003 (Chapter 3).

[22] C. S�aanchez, E.P.M. Leiva, in: W. Vielstich, A. Lamm, H.

Gasteiger (Eds.), Handbook of Fuel Cells: Fundamentals,

Technology and Applications, vol. 2, John Wiley & Sons,

Chichester, 2003 (Chapter 2).

10 C.G. S�aanchez / Surface Science 527 (2003) 1–11

[23] J.O�M. Bockris, A.K.N. Reddy, M. Gaboa-Aldeco, Mod-ern electrochemistry, Fundamentals of Electrodics, vol.

2A, second ed., Kluwer Academic, New York, 2000.

[24] J.G. Gordon, O.R. Melroy, M.F. Toney, Electrochim.

Acta 40 (1995) 3.

[25] K. Ataka, T. Yotsuyanagi, M. Osawa, J. Phys. Chem. 100

(1996) 10664.

[26] T. Iwasita, X. Xia, J. Electroanal. Chem. 411 (1996) 95.

[27] T. Iwasita, X.H. Xia, H.-D. Liess, W. Vielstich, J. Phys.

Chem. B 101 (1997) 7542.

[28] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77

(1996) 3865.

[29] R. Parr, W. Yang, Density Functional Theory of Atoms

and Molecules, Oxford University Press, 1989.

[30] P. Ordej�oon, E. Artacho, R. Cachau, J. Gale, A. Garc�ııa, J.

Junquera, J. Kohanoff, M. Machado, D. S�aanchez-Portal,

J.M. Soler, R. Weht, Mat. Res. Soc. Symp. Proc. 677

(2001) AA9.6.1.

[31] http://www.uam.es/departamentos/ciencias/fismateriac/

siesta/.

[32] N. Troullier, J.L. Martins, Phys. Rev. B 43 (1991) 1993.

[33] O.F. Sankey, D.J. Niklewski, Phys. Rev. B 40 (1998)

3979.

[34] D. S�aanchez-Portal, P. Ordej�oon, E. Artacho, J.M. Soler, Int.

J. Quant. Chem. 65 (1997) 453.

[35] E. Artacho, D. S�aanchez-Portal, P. Ordej�oon, A. Garc�ııa,

J.M. Soler, Phys. Status Solidi 215 (1999) 809.

[36] CPMD, version 3.3, written by J. Hutter, A. Alavi, T.

Deutsch, and the group of M. Parrinello at the MPI in

Stutgart and IBM Research Laboratory in Z€uurich.

[37] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 80

(1998) 891.

[38] C.G. S�aanchez, J.J. Kohanoff, unpublished results, 1999.

[39] R. Hoffmann, Rev. Mod. Phys. 60 (1988) 601.

[40] G. Valette, J. Electroanal. Chem. 269 (1989) 191.

[41] S. Amokrane, J.P. Badiali, Electrochim. Acta 34 (1989)

39.

[42] S. Amokrane, J.P. Badiali, J. Electroanal. Chem. 150

(1989) 315.

[43] A. Rahman, F.H. Stilinger, J. Chem. Phys. 60 (1974) 154.

C.G. S�aanchez / Surface Science 527 (2003) 1–11 11