Water adsorption on metal surfaces: A general picture from ...everest.iphy.ac.cn/papers/prb69.195404.pdf · Water adsorption on metal surfaces: A general picture from density functional

PHYSICAL REVIEW B 69, 195404 ~2004!

Water adsorption on metal surfaces: A general picture from density functional theory studies

Sheng Meng,1,2 E. G. Wang,1 and Shiwu Gao21Institute of Physics, Chinese Academy of Sciences, P.O. Box 603, Beijing 100080, China

2Department of Applied Physics, Chalmers University of Technology and Go¨teborg University, SE-412 96 Go¨teborg, Sweden~Received 3 October 2003; revised manuscript received 15 December 2003; published 5 May 2004!

We present a density functional theory study of water adsorption on metal surfaces. Prototype water struc-tures including monomers, clusters, one-dimensional chains, and overlayers have been investigated in detail ona model system—a Pt~111! surface. The structure, energetics, and vibrational spectra are all obtained andcompared with available experimental data. This study is further extended to other metal surfaces includingRu~0001!, Rh~111!, Pd~111!, and Au~111!, where adsorption of monomers and bilayers has been investigated.From these studies, a general picture has emerged regarding the water-surface interaction, the interwaterhydrogen bonding, and the wetting order of the metal surfaces. The water-surface interaction is dominated bythe lone pair–d band coupling through the surface states. It is rather localized in the contacting layer. Asimultaneous enhancement of hydrogen bonding is generally observed in many adsorbed structures. Somespecial issues such as the partial dissociation of water on Ru~0001! and in the RT39 bilayer phase, the H-upand H-down conversion, and the quantum-mechanical motions of H atoms are also discussed.

DOI: 10.1103/PhysRevB.69.195404 PACS number~s!: 68.43.Bc, 68.35.2p, 82.30.Rs

leimllsat

ienngo

fam

e

tea

ah

staheteutu

soniagut

l to

dd vi-the

ther-

aces

ddi-

atex-theer-s,

centa

pe

net-antics

uc-d 1Dnsi-ese

thentstoces

very

I. INTRODUCTION

Water interaction with solid surfaces1,2 plays central rolesin a variety of phenomena in nature such as catalysis, etrochemistry, corrosion, and rock efflorescing, and hasportant applications in, e.g., hydrogen production, fuel ceand biological sensors. During the past two decades, wadsorption on single crystalline metal surfaces has beentensively investigated in laboratories by various experimtal techniques1,2 as a prototype system for understandiwater-solid interfaces and their interactions. Dependingthe coverage and experimental conditions, water on a surforms different low-dimensional structures, ranging froisolated monomers and clusters, to one-dimensional~1D!chains, and two-dimensional~2D! ordered overlayers.3 Whilethe ordered 2D structures were accessible in earlier expments by low-energy electron diffractions~LEED!,4,5 recentexperiments using scanning tunneling microscope~STM!have made it possible to locally image and probe isolawater clusters. For instance, water monomers, dimers,hexamers were recently observed by STM on Ag~111!,6

Cu~111!,7 and Pd~111! ~Ref. 8! surfaces. A 1D water chainwas observed on the steps of a Pt~111! surface.3 As the cov-erage increases, water forms hydrogen-bonded~H-bond! net-works of various phases, depending on the substrate,continues to grow into multilayers and bulk ice at higcoverages.9

What determine these adsorbed structures and theirbilities are the two fundamental forces at the water-meinterfaces, namely,~i! the water-surface interaction, whicoccurs predominantly in the water-metal contacting layand ~ii ! the interwater hydrogen bonding, whose characand strength may be modified by the presence of the sstrates. On most metal surfaces, these two interactionsout to be comparable in strength. Their competition resulta rich class of adsorbed structures especially at submlayer coverages. Characterizing these structures, especthose at low coverages, is essential to the understandinthe water-surface interaction at the interfaces. Comp

0163-1829/2004/69~19!/195404~13!/$22.50 69 1954

c--,ern--

nce

ri-

dnd

nd

ta-l

r;rb-rn

ino-llyof

er

simulation based onab initio density functional theory~DFT! has proven to be a useful and supplemental toostudy the water-solid interfaces.10

Water on Pt~111! represents one of the most well studiesystems by experiments, where various adstructures anbrational spectra have been documented. Among allstructures, water bilayer in aA33A3R30° ~RT3! phase wasfirst proposed by earlier experiments on the Pt~111! surface.4

This bilayer phase is most interesting because it marksinitial formation of the H-bonded water networks on the suface. It has generally been observed on other metal surfsuch as Rh~111! ~Ref. 11! and Au~111!,12 and has beenviewed as a model water structure at the interfaces. In ation to the RT3 bilayer, two more bilayer phases, theA393A39R16.1° ~RT39! andA373A37R25.3° ~RT37!,13,9 havealso been observed in recent experiments on Pt~111! at 130–140 K. These bilayers were found to be interconvertiblecertain experimental conditions. Despite the tremendousperimental and theoretical efforts, our understanding onsimplest bilayer, the RT3 bilayer, remains to be controvsial. While two RT3 bilayers, the H-up and H-down onewere proposed in earlier experimental studies and a reDFT calculation,14 Ogasawara and co-workers argued inrecent experiment that only a flat bilayer of H-down tywas observed in the RT3 phase on Pt~111! with a verticalO-O distance as small as 0.25 Å.15 This conclusion has beerecently questioned by Feibelman, who claimed that the wting layer of the water/Pt should be the RT39 rather thRT3, based on the comparison of the adsorption energebetween the two phases.16

Besides the controversy in the bilayer, other nanostrtures of water at surfaces such as monomers, clusters, anchains, remain to be poorly understood due to their insetivity to experimental probes and due to the fact that thnanostructures are computationally more demanding than2D periodic systems. Although recent STM experimewere able to image individual water clusters, it is difficultdetermine the structure and bonding properties at surfabecause water molecules and clusters are usually

©2004 The American Physical Society04-1

0erosnDd

wae

sorhe

o

laop

o-

cuco

arsrat

uray

ibbothhchtcthss

h

rresaitydd.ruw

heafontSeio

ry

h

thattal

bedeter-he

helose

in

re

1Donstoas

tedset0.01

acetherchereesinguc-of

eas

ityn-didnal

edes a

SHENG MENG, E. G. WANG, AND SHIWU GAO PHYSICAL REVIEW B69, 195404 ~2004!

mobile on the surfaces, even at temperatures as low as 18

A particularly interesting type of water is the 1D watchains3 on stepped surfaces, which resembles the 1D cfined water in biomembranes. The latter has been invegated intensively in model confined geometries ananotubes17 by computer simulations. In contrast, the 1chain observed on the Pt surface has been neither studieunderstood.

Another important and fascinating issue of adsorbedter is its dissociation and proton transfer at surfaces, whour understanding is far from conclusive. While water disciation on oxide surfaces has been widely observed, watemetal surfaces is usually believed to be intact except wcoadsorbed with other molecules or atoms.1,2 However, arecent DFT calculation suggested that water bilayerRu~0001! is half dissociated with one OH broken.10 Thisconclusion contradicts the conventional picture of molecuwater on metal surfaces. Detailed vibrational spectroscusing sum frequency generation has recently been carriedfor water bilayers on Ru~0001!.18 The measured data compared well with the calculated vibrational spectra for molelar bilayers, suggesting that the water bilayer is undissoated. This issue remains unresolved and deservesattention in future studies.

This paper presents a computational study of watersorption on transition- and noble-metal surfaces using fiprinciples DFT calculations, with the goal to gain a geneunderstanding of the water-metal interfaces and some ofissues mentioned above. First, various adsorption structincluding water monomers, small clusters, 1D chains, bilers, and multilayers, are investigated on the Pt~111! surface.The energetics of the adsorbed states, geometries, and vtional spectra are determined and compared with availaexperiments. These results demonstrate the role of electrstructure in the water-metal interactions, as revealed byinterface charge transfer and H-bond enhancement, whicturn can be recognized vibrationally via the OH stretmode. Some of the specific issues such as lattice mismahydrogen disorder, partial dissociation, and the nature ofhydrogen bonding at surfaces, are investigated and discuin detail. Second, the understanding gained on the Pt~111!surface is extended to other close-packed surfaces sucRh~111!, Ru~0001!, Pd~111!, and Au~111!, adsorbed withtwo prototype structures: the monomers and bilayers. Colation between thed-electron occupancy of the substratand the structure and energetics of the adsorbed wmolecules are illustrated. A simple picture of hydrophobicand hydrophilicity, which was proposed in a previous stuof water on Pt and Au,19 is further examined and discusseBesides, the vibrational spectra for the representative sttures are given and provide a database for comparisonexperiments.

The rest of this paper is organized as follows. With tintroduction in Sec. I, computational methods and detailsgiven in Sec. II. The main results are presented in Sec. IIIwater on Pt~111! and for monomers and bilayers on differemetal surfaces. Vibrational spectra are also presented.tion IV focuses on a few specific issues such as the naturthe H bonding at surfaces, the H-up and H-down convers

19540

K.

n-ti-d

nor

-re-onn

n

ryut

-i-ur

d-t-l

hees-

ra-lenicein

h,eed

as

e-

ter

y

c-ith

rer

ec-ofn,

and partial dissociation of water bilayers. A short summaand conclusions are given in Sec. V.

II. COMPUTATIONAL METHOD

The calculations were carried out with the Viennaab ini-tio simulation program, VASP,20 which enables us to do botstructure optimization and molecular dynamics~MD! simu-lations. The metal surfaces were modeled by a supercellcontains a slab of typically four to seven layers of meatoms and a vacuum region of;13 Å. Water moleculeswere put on one side of the slab to simulate the adsorsystems. The lattice constants of the surfaces were dmined from bulk calculation and usually agree well with texperimental values~Table I!. Different supercells, 333 and2A332A3R30° for small water clusters and aA33A3R30° cell for the RT3 overlayer, were calculated. Tsizes of these unit cells are large enough to yield results cto convergence, with a typical accuracy around 5–10 %energetics. Monkhorst-Pack scheme21 with 33331 and 53531 k-point sampling in the surface Brillouin zone weused for the two sizes of supercells, respectively. A singleGpoint sampling was adopted for large supercells includingwater chains, the RT39, and the RT37 overlayersPt~111!. A plane-wave cutoff at 300 eV was used in mocalculations, while a higher cutoff of 400 eV was alsperformed to check convergence. The Fermi level wsmeared by the Methfessel and Paxton22 approach with aGaussian width of 0.2 eV. The free energy was extrapolato zero kelvin to yield total energies of the systems. Thisof parameters assures a total energy convergence ofeV/atom.

In structural search, the water molecules and the surflayer of the slabs were relaxed simultaneously, whilebottom layers were fixed at their bulk positions. The seawas stopped when the forces on all relaxed atoms wsmaller than 0.05 eV/Å. In MD simulations, the moleculand the surface layer atoms were allowed to move accordto the forces calculated from the converged electronic strture. A 300 eV cutoff in plane-wave basis and a time step0.5 fs were utilized in all MD simulations. To obtain thvibrational spectra, a 2 ps production run at 90–140 K wperformed after equilibrating the system for;1 ps. The vi-brational spectrum was obtained from the velocity-velocautocorrelation function in the MD simulation. Higher eergy cutoff at 400 eV and a shorter time step of 0.25 fsnot change the peak positions or the shape of the vibratiospectra.

Reaction barriers were calculated by the nudgelastic band method,23 available in VASP. For searching thminimum energy reaction pathway, this method employ

TABLE I. The calculated and experimental lattice constants~Å!for several hcp~Ru! and fcc~Rh, Pd, Pt, Au! metals.

Ru Rh Pd Pt Au

Theor. 2.72 3.83 3.96 3.99 4.18Expt. 2.71 3.81 3.89 3.92 4.08

4-2

alleiz

urm

s-

l-s

ni-r-

th

la

doIeoraernd

tes,a

ethtiri

ith

tialastics

leisan

acendtry.isno-yer

o-the

the

re

an

ndpec-

WATER ADSORPTION ON METAL SURFACES: A . . . PHYSICAL REVIEW B69, 195404 ~2004!

constrained minimization of the total free energy oftrial path by relaxing a number of ‘‘images’’ in the path. Athe water molecules and the surface-layer atoms wrelaxed under the constraint, as in the ground-state optimtions.

The adsorption energy for an adsorbed water structEa, has been defined as the mean adsorption energy perecule of the adstructure,

Ea5~Emetal1n3EH2O2E(H2O)n /metal!/n. ~1!

Here E(H2O)n /metal is the total energy of the adsorption sy

tem,Emetal andEH2O are those for the surface and free moecules, respectively, andn is the number of water moleculein the cell.

In our calculations, the Vanderbilt ultrasoft pseudopotetials ~USPP! ~Ref. 24! and the generalized gradient approxmation~GGA! for the exchange-correlation potential by Pedew and Wang~PW91! ~Ref. 25! were used. The GGAextension is crucial for the accurate treatment ofhydrogen bonds and water structures.26 The PW91 form hasbeen tested extensively for a variety of intermolecuinteractions including H bonding.27 To illustrate the feasibil-ity of the USPP1PW91 approach for describing water anhydrogen bond, the calculated geometries and energeticsfree water molecule and dimer were tabulated in TableThe OH bond length and dipole moment of the monomand the geometry and formation energy of the dimer, shexcellent agreement with experiments. Moreover, the vibtional spectrum obtained from MD for a free water dim~Fig. 1! also agrees well with other calculations aexperiments.28,29

III. RESULTS

The first part of this section presents results of waadsorption on Pt~111! in various phases including monomerclusters, and overlayers. The structures, energetics,the interaction between water molecules and the substratstudied in detail. The second part extends this study to ometal surfaces, where general features of water adsorpand the effect of different substrates are investigated. Vibtional spectra obtained from the simulations are given

TABLE II. The calculated geometries and energies of a fwater monomer and a dimer. The bond angles (a, b) are as de-picted in Fig. 1. The experimental data are taken from Refs. 2829.

Theor. Expt.

Free H2O OH bond length~Å! 0.973 0.957HOH bond angle~deg! 104.85 104.52

Dipole moment~D! 1.856 1.855Free H2O dimer OO distance~Å! 2.86 2.98

a ~deg! 2.79 2166b ~deg! 126.35 12366

Formation energy~kJ/mol! 24 23

19540

rea-

e,ol-

-

e

r

f aI.r,w-

r

ndareerona-n

the last part, providing a database for comparison wexperiments.

A. Water adsorption on Pt„111… surface

1. Water monomers on Pt(111)

The adsorption of water monomer contains the esseninformation regarding the water-metal interaction, and hbeen investigated first. Structure optimization and energeindicate that adsorption on top site@see Fig. 2~a!# is moststable compared to bridge and hollow sites~Table III!. Thisis further supported by a shorter H2O-metal ~O-M! bondlength on the top site (dOM52.43 Å). Water lies almostflatly on the surface with its polar axis making a small angu513° –14° with the surface plane. The OH bondstretched slightly, while the HOH angle is more open ththe free water molecule~0.973 Å and 104.85°, Table II!.These results indicate electron transfer from O to surfatoms. Although the top site adsorption has been fouin most recent studies, bridge site with an upright geomewas also reported for monomer in an earlier study30

Our calculation shows that the upright configuration40 meV unfavorable compared to the flatly adsorbed momer on the bridge site. More accurate results with a six-laPt slab and a higher-energy cutoff~400 eV! are also given inTable III, which shows minor variations in the structure(;1%) and adsorption energy (;3%) for the adsorbedmonomer.

To gain insight into the dynamics of the adsorbed monmer, the distance- and angular-dependent energies of

FIG. 2. The water monomer and small clusters adsorbed onPt~111! surface.

e

d

FIG. 1. Vibrational spectrum for a free water dimer. Solid adashed lines correspond to the proton donor and acceptor, restively. The inset shows the optimized geometry of the dimer.

4-3

e inshown

34


TABLE III. Water monomer adsorption on the~111! surface of Pt. Energies, distances, and angles arunits of meV, Å, and deg, respectively. Results for different Pt layers in the slab and energy cutoff arefor comparison.

Layers Ecut ~eV! Top Bridge Hollow dOH /HOH udOM Ea dOM Ea dOM Ea

4 300 2.43 291 3.11 123 3.12 121 0.978 105.36 16 400 2.40 304 2.89 117 3.02 102 0.980 105.62 1

Pt-3

ile

y

erd

y

i

cacinm

usin

snatctro

ra-dentsers

arera-ble

ook

heyaceton

han

elso

ite

e

allin

0

8

9

H2O/Pt~111! are plotted in Fig. 3, as functions of the O-distancedOPt and the bending angleu. The distance dependence~left panel! shows an equilibrium bond length at 2.4Å for the O-Pt bond. In the angular dependence,u50 cor-responds to the molecule lying down on the surface, whu590°(290°) corresponds to the upright position with thO atom pointing toward~away from! the surface. The energprofile was obtained by rotating H2O molecule while keep-ing O fixed at its equilibrium position. The rotational barriat u590° is 140 meV, lower than what Michaelides anco-workers31 reported recently,;190 meV. The discrepancmight result from the smaller supercell, 232, used in theircalculation.~Our calculation used a 333 supercell.! In ad-dition, the rotational barrier along the azimuthal anglefound to be very small~less than 2 meV!, which suggeststhat water molecule can rotate freely on the surface. Weconclude from these results that the adsorbed monomerrotate freely in two dimensions on the surface. The bendmotion could also be quite active near the equilibriuangles.

2. Water clusters on Pt(111)

The adsorption of water clusters is interesting becaboth the H bonding and water-surface interactions arevolved in the adsorbed clusters. Studying these clustersurfaces may help us understand the competition betweetwo interactions at low coverages. On metal surfaces, wclusters were observed by a number of experimental teniques such as high-resolution electron-energy-loss speccopy ~HREELS!,32 infrared adsorption spectroscopy,33 Heatom scattering~HAS!,34 and STM.6,8 On Pt~111!, water

FIG. 3. The variation of the total energy for a water monomon Pt~111! as a function of the H2O-Pt distancedOPt ~left panel! andthe tilt angleu ~right panel!.

19540

e

s

nang

e-at

theerh-s-

dimers, trimers, and other clusters were identified by vibtional spectroscopies.34,35 Yet their detailed structures anbonding properties have not been determined by experimso far. We calculated water dimers, trimers, and hexamadsorbed on Pt~111!. A supercell of 333 was employed fordimer and trimer adsorption, while a larger cell of 2A332A3 was used for hexamers. The obtained structuresdepicted in Fig. 2. The energetics and geometric configutions for each molecule in the clusters are specified in TaIV.

Generally speaking, the geometries of these clusters lquite similar to their gas-phase counterparts.36 Water mol-ecules prefer atop site adsorption, whenever possible. Ttend to lie down onto the surface, due to the cluster-surfinteraction. In the dimer case, for example, both the prodonor and acceptor take an atop site as shown in Fig. 2~b!,although the donor couples more strongly to the surface tthe acceptor, forming two O-Pt bonds~with dOPt52.26 and3.05 Å, respectively! plus an internal H bond. Besides thdifference in O-Pt bond length, the donor and acceptor adiffer in other details. The geometry of the donor is qusimilar to that of the adsorbed monomer withu525.1°,while the acceptor lands onto the surface withu5241.8°.The donor and acceptor make thus an angleb of around120°, as in the free dimer. The O-O distance,dOO52.70 Å, is shortened~2.86 Å for the free dimer, Table II!;and the OH bond is stretched slightly~1.012 Å, compared to0.985 Å for the free dimer!. One can thus infer that the Hbond in the adsorbed dimer is enhanced. A wider/HOH in

r

TABLE IV. The adsorption energies and geometries for smwater clusters on Pt~111!. Energies, distances, and angles areunits of meV, Å, and deg, respectively.

Cluster Ea dOPt u dOH1 dOH2 /HOH dOO

Monomer 304 2.40 13.8 0.980 0.980 105.62Dimer 433 2.26 25.1 0.978 1.012 106.72 2.7

3.05 41.8 0.981 0.982 103.52Trimer 359 2.76 3.5 0.975 0.985 107.75 2.7

2.76 3.5 0.975 0.985 107.86 2.802.76 3.1 0.974 0.985 107.71 2.79

Hexamer 520 2.32 31.1 0.997 1.001 106.22 2.93.38 32.9 0.974 0.991 104.49 2.802.77 1.8 0.978 0.990 107.25 2.893.35 0.3 0.975 0.988 106.88 3.012.77 3.7 0.979 0.987 107.14 2.803.39 32.3 0.974 0.991 104.83 2.88

4-4

otu

aces

indu

3le

veren

eVay

iti

eVe

riad

erT

re

reoteden

thTP

tu

aterOa

oa

a

ve-i

ison

ind

ebehe431g tozig-

the


the donor and a narrower/HOH in the acceptor are alsobserved, which may be caused by electron transfer fromdonor to the substrate and the back donation from the sstrate to the acceptor, due to the interactions with the surf

The trimer and hexamer retain their ringlike structurEach molecule in the trimer lies very flatly (u;3.5°) on thesurface, with one OH forming an H bond and the other befree, as shown in Fig. 2~c!. Cyclic hexamer forms a puckerehexagonal ring with three molecules lying closer to the sface (dOPt52.32, 2.77, 2.77 Å!. The other three are a littlehigher (dOPt;3.4 Å). The adsorbed hexamer thus formO-Pt bonds and 6 H bonds. One water molecule is a doubproton donor, and lies much close to the surface. The aaged O-O distance is slightly larger than that of the fdimer and trimer. In the gas phase, there are two additiohexamer structures, the cage and prism hexamers.36 The ad-sorption energy for the prism structure on Pt is 321 mwhich is 200 meV lower than the cyclic hexamer. So it mnot exist on the Pt surface.

Among these clusters, cyclic hexamer is most stable wadsorption energy of 520 meV per molecule. The trimerleast stable with an adsorption energy of only 359 mThe adsorption energy of the dimer, 433 meV, lies in btween. This energy difference reflects dominantly the vation in the number of water-metal bonds and the H-bonformed in the adsorbed clusters. Compared with expments, cyclic water hexamer was already observed by Sexperiments on Ag~111!,6 Cu~111!,7 and Pd~111! ~Ref. 8! sur-faces, although no experiment has been available on Pt~111!.Small clusters including dimers and trimers were alsoported on Pd~Ref. 8! by STM, formed via diffusion ofmonomers.

3. One-dimensional water chains on Pt steps

The 1D water chain is an interesting type of structubecause it is believed to exist in the water pores acrbiomembranes. It also provides an ideal 1D model sysbased on water molecules. Modeling the structure andnamical properties of 1D water has been carried out intsively for water confined in carbon nanotubes17 and modelconfining potentials. Such 1D model structures have neibeen observed nor realized by any experiments so far.1D water chain found experimentally on the steps of thesurface3 is therefore extremely interesting and has been sied in our calculation. To model the110&/$100% step foundin the experiment,3 a slab with 15 layers of Pt in a~322!surface is used in the calculation. The unit cell is schemcally shown in Fig. 4. The water at the step can form diffent chain structures depending on the H bonding andorientations at the step. The two simplest water structuresthe ones shown in Fig. 4, where one OH of each water cnects the chain and the other OH bond points either inw@H-in, molecule 1 in Fig. 4~b!# or outward~H-out, molecule2!. A zigzag chain, with one H-in molecule coupled alterntively to an H-out molecule, has also been [email protected]~b!#. The results for the three calculated structures are giin Table V. For comparison, monomers adsorbed in the H

19540

heb-e..

g

r-

r-eal

,

hs.--si-M

-

,ssmy--

erhet

d-

i--Hre

n-rd

-

nn

and H-out configurations are also obtained. The last rowfor the isolated monomer and the zigzag chain adsorbedthe terrace.

The isolated monomers at the two configurations bstrongly to the Pt step, with adsorption energiesEa5449 meV for H-in, and 426 meV for H-out. Among ththree chains studied, only the zigzag chain is found tostable with a binding energy of 480 meV per molecule. TH-in and H-out chains, whose adsorption energies areand 385 meV, respectively, are unstable, when comparinthe corresponding monomers. The H-bond energy of the

FIG. 4. The 1D water chains at a110&/$100% step on the Ptsurface, as shown by the side view~a! and the top view~b!. Theunit cell contains 15 layers of Pt atoms in a~322! surface.

TABLE V. The water monomer and 1D chains adsorbed at^110&/$100% step on the Pt~111! surface, modeled by an unit cell inthe ~322! surface.

Monomer 1D chaindOPt ~Å! Ea ~meV! dOPt ~Å! Ea ~meV!

H-in 2.22 449 2.42 431H-out 2.25 426 2.48 385Mixed 2.45 480On terrace 2.43 291 2.62, 2.72 246

4-5

ablep

uacninp

rorsva

s

eo

r

a

-

e

ini-

ton

f

3tis

r isndi-eme

. 5.T3.olet

n

irer,

ns

l (ofO

onsatedigna-nite,ar-bi-l

asx-in

as

she


zag chain can be deduced as 85 meV. This chain is stbecause it favors intermolecular H bonding and the dipodipole interactions.17 In addition, water monomers at the steare much more stable than those on the Pt~111! terrace (Ea5291 meV). The zigzag chain at the step is abo;230 meV more stable than the same chain on the terrSuch comparison suggests that water-Pt interaction is geally stronger at the steps. It explains why the water chawere only observed at the steps in experiments. The oelectronic structure of the steps is responsible for the stger interaction with water for both chains and monomeThis conclusion is in agreement with experimental obsertions.

4. 2D overlayers on Pt(111): the RT3, RT37, and RT39 phase

Water forms bilayers and multilayers at higher coveragOn Pt~111!, different overlayers have been observed. Onethe well-known forms is theA33A3R30° ~RT3! bilayer, inwhich water molecules form a puckered hexagonal netwoas Doering and Madey proposed.5 However, this RT3 phasewas only observed in finite domains by LEED experiment85 K.32 In addition, two complex phases,A393A39R16.1°~RT39! andA373A37R25.3° ~RT37! bilayers, were also observed at temperatures above 135 K.13,9 The RT39 structurewas found to transform into RT3 at ca. five bilayers.9 Con-troversy still exists in the literature regarding the bilaystructure of water on Pt~111!.

Figure 5 shows the structure of the RT3 bilayer in aA33A3R30° surface unit cell with two water moleculeseach cell. The H2O molecule in the lower plane binds drectly to the surface, while the upper one forms H bondsthe molecules in the lower plane and molecules in neighbing unit cells. Three of the four H atoms form hydrogebonds, while the fourth is either free@H-up case, panel~a!# orbinds to the surface@H-down case, panel~b!#. The verticaldistances between the two oxygens are 0.63 Å and 0.35 Åthe H-up and H-down bilayer, respectively~see Table VIII!.Both structures are contracted compared to the bulk icewith zOO50.97 Å. The adsorption energies are 522 and 5meV, for H-up and H-down cases, respectively. The potenbarrier for H-up flipping to H-down bilayer is 76 meV, a

FIG. 5. ~a! The H-up,~b! H-down bilayers, and~c! the doublebilayers in theA33A3R30° symmetry on Pt~111!. Both the sideand top views are shown.

19540

le-

te.

er-s

enn-.-

s.f

k,

t

r

or-

or

Ih4al

calculated by the nudged elastic band method. This barriefurther discussed in Sec. IV C. Both structures can be cadates for bilayer on the Pt~111! surface. In comparison, thadsorption energy for the half-dissociated bilayer, the sastructure suggested for Ru~0001!,10 is 291 meV, and is muchsmaller than those of the molecular bilayers shown in FigTherefore, dissociation is not favored on Pt surface in RSuch a molecular bilayer was suggested earlier by ultraviphotoemission spectroscopy,37,38 the low-energy electron-diffraction measurement,3,32 and recently by x-ray absorptiospectroscopy.15

The A393A39R16.1° ~RT39! and A373A37R25.3°~RT37! ~Refs. 13 and 9! phases are also investigated. Thegross structures look very similar to that of the RT3 bilayi.e., puckered hexagonal networks~Fig. 6!. However, theRT39 bilayer shows a quite disordered atomic distributiodue to compression~by 3.3%! in the 2D unit cell. Theheight-dependent density profile along the surface normazdirection! for O and H atoms are shown in the right panelsFig. 6. Compared to the RT3 phase, in which the twoatoms are located at two positions, the density distributiof the RT39 show broadened peaks with some atoms locfar away from the surface. These broadened peaks are stures of disorder. Among the 32 water molecules in the ucell, the lowest O atom is only 2.10 Å from the surfacwhile the highest H2O is 4.4 Å above the surface, givingrough bilayer with vertical thickness of 2.3 Å. Such a disodered bilayer is in sharp contrast to the picture of a flatlayer proposed recently,15 which claimed that the verticathickness between the upper and lower O atoms issmall as 0.25 Å. The 2D lattice of the RT37 bilayer is epanded slightly by 4.4%, compared to the bulk ice Ih, asthe RT3 ~7.2%! phase. It is also disordered, but not

FIG. 6. ~a! The A393A39R16.1° ~RT39, H-down! bilayer andits atomic density profiles along thez axis for ~b! O and ~c! Hatoms.~d! The A373A37R25.3° ~RT37, H-down! bilayer and itsdensity profiles for~e! O and~f! H atoms. Solid and dashed circleindicate the H3O and OH sites, respectively. For comparison, tdensity profiles for theA33A3 R30° ~RT3! bilayer ~H-down! arealso shown by the gray lines.

4-6

ns,

in

ah

sealy

asinr

a

tin

anaor4thi

eT3ra

he

nd

cee

then

chndtheisicalur-rac-es,ali-n-ater

has

euris

rallyur-

ofthehe

e

, as

sele

rsr

e

wn


much as the RT39~Fig. 6!. The maximum O-Pt distance ithe RT37 is 3.58 Å, similar to that in the RT3 bilayer3.37 Å ~H-up! and 3.14 Å~H-down!. One key feature of theRT39 structure is the existence of a few H3O1 like ~i.e.,dissociated! molecules, which will be discussed furtherSec. IV D.

To see the coverage dependence of the structureenergetics in the overlayers, water in the RT3 phasebeen studied from two bilayers, Fig. 5~c!, to six bilayers. TheO-O distance between two adjacent bilayers is 2.75–2.83The water-metal bond for the bottom molecule decreagradually. The adsorption height of the bottom water equ2.69, 2.63, 2.56, 2.49, 2.52, and 2.47 Å, respectivewhen the coverage goes from one to six bilayers. In contrthe height of the upper water in the first bilayer remaalmost constant, 3.2560.02 Å. This indicates that in bilayeand multilayers, only molecules in the bottom havedirect interaction with the metal surface~see Fig. 5!,while the upper molecule is almost unaffected, suggesthat the water-surface interaction on Pt~111! is ratherlocalized.

Figure 7 compares the adsorption energy of the RT3RT39 phases for up to three bilayers. It shows that thesorption energy of the RT39 phase, 615 meV, is slightly mfavorable ~by 80 meV! compared to the RT3 phase, 53meV, at one bilayer coverage. The adsorption energy forRT37 phase, 597 meV, lies in between. As the coveragecreases, the RT3 phase becomes more favorable comparother two phases, as found in recent experiment. The Rphase was found to transform into RT3 after a structureorientation9 at about five bilayers, as estimated from tsame experiment.

Information concerning the detailed structures and boing information for all the calculated overlayers on Pt~111!are summarized in Table VI.

B. Water on different metal surfaces

Here below, we extend our study to other metal surfasuch as Au, Pd, Rh, and Ru, which have been under exp

FIG. 7. The adsorption energy of various overlayer phaon Pt~111! at coverages from 1 to 3 bilayers. The square, circand triangle represent theA33A3R30° ~RT3!, the A393A39R16.1° ~RT39!, and theA373A37R25.3° ~RT37! structures,respectively.

19540

ndas

Å.s

ls,t,

s

g

dd-e

en-d to9l

-

sri-

mental investigation. The closely packed surfaces, i.e.,~111! for the fcc and~0001! for the hcp metals, were chosefor comparison. These surfaces differ from Pt~111! in twoaspects:~i! the variation of the surface lattice constant, whimatches differently with the H-bonded water networks; a~ii ! the change in the chemical reactivity associated withvariation of the d-band occupancy. Our basic concernwhether and how these two factors, geometry and chemreactivity, affect the water structures and properties at sfaces. What are the general trends of water-surface intetions. To this end, we consider only two prototype structurthe monomers and bilayers. The former is ideal for the cbration of the metal-surface coupling, while the bilayer eables us to examine both the water-surface and the interwinteractions.

1. Water monomers on Ru, Rh, Pd, Pt, and Au

The structure of the water monomer on these surfacesbeen investigated in detail in a recent study.31 The structureof water monomer on Pt~111! seems to be general to all thessurfaces.31 Results for monomer on these surfaces from ocalculation are given in Table VII. Top site adsorptionmost stable on all these surfaces, withu56° –24°. An in-creased OH bond and a more open HOH angle are genefound. This indicates electron transfer from the O to the sface, as we found on the Pt~111! surface.

Regarding the differences and details, the interactionwater with Ru and Rh is found to be much stronger thanPt and Pd, while it is much weaker on the Au surface. Tadsorption energetics suggests a bonding order as Ru.Rh.Pd.Pt.Au, This is further supported by the trend in thdOM and the bonding angles,u and /HOH. This bondingorder reflects the chemical reactivity of these surfaces

s,

TABLE VI. The structures and energetics for water clusteand thin films on the Pt~111! surface. The unit cell, the numbeof moleculesn, the number of H2O-metal bondsNH2O-M , and thenumber of H bondsNHB in the unit cell are shown together with thadsorption energiesEa, and the H-bond energiesEHB ~in meV!.The two energies for the bilayer correspond to the H-up/H-docases.

Ads. species Unit cell n Ea NH2O-M NHB EHB

Monomer 333 1 304 1 0Dimer 333 2 433 2 1 258Trimer 333 3 359 3 3 55Hexamer 2A332A3 6 520 3 6 368Bilayer A33A3 2 505/527 1 3 235Two bilayers A33A3 4 564 1 7 312Three bilayers A33A3 6 579 1 11 303Four bilayers A33A3 8 588 1 15 307Five bilayers A33A3 10 593 1 19 307Six bilayers A33A3 12 601 1 23 320Bilayer A373A37 26 597 13 39 297Bilayer A393A39 32 615 16 48 309Two bilayers A393A39 64 582 16 112 275Three bilayers A393A39 96 572 16 176 276

4-7


TABLE VII. Geometries and energetics of a water monomer on the Ru~0001!, Rh~111!, Pd~111!, Pt~111!,and Au~111! surfaces. Energies, distances, and angles are in units of meV, Å, and deg, respectively.

Substrate Layer Top Bridge Hollow dOH /HOH udOM Ea dOM Ea dOM Ea

Ru~0001! 5 2.28 409 2.55 92 2.56 67 0.981 105.66 16Rh~111! 4 2.32 408 2.57 126 2.70 121 0.978 105.95 24Pd~111! 4 2.42 304 2.74 146 2.77 130 0.977 105.63 20Pt~111! 4 2.43 291 3.11 123 3.12 121 0.978 105.36 13Au~111! 7 2.67 105 2.80 32 2.80 25 0.977 105.04 6

on

tio

ta

r insoro

ten

cea

ltsn

di-mi-er

ithrd-

ls,

. 8,ctevesest

eron

andin

lly

indicated by the periodic table. Our results are largely csistent with those by Michaelides and co-workers,31 althoughsome details differ. For example, the monomer adsorpenergy was ordered as Rh.Ru.Pt.Pd.Au in their results.However, the energy difference between Ru and Rh anddifference between Pt and Pd, are very small. In fact theywithin the accuracy of the calculations.

2. Water bilayers on different surfaces

To examine the H bonding on different surfaces, watethe A33A3R30° bilayer has been calculated on differesurfaces. Structure parameters and adsorption energiesummarized in Table VIII. The puckered hexagonal netwon these surfaces is very similar to the RT3 bilayerPt~111!. The adsorption height of the bottom waterzOM1increases gradually in the order of Ru.Rh.Pd.Pt.Au, asin the periodic table, while the height of the upper wazOM2 keeps almost constant, namely, 3.40 Å for H-up a3.20 Å for the H-down bilayer. The vertical O-O distanzOO decreases therefore along this order. These resultsmore clearly shown in Fig. 8 for the H-up bilayer. The resufor the H-down bilayer look very similar and are not show

19540

-

n

here

ntarekn

rd

re

here. The universal structure of the upper water layer incates that the water-surface interaction is localized donantly in the bottom layer, while the molecules in the upplayer are almost unaffected.

The bonding order shown above correlates directly wthe d-band filling of these surfaces, which increases accoingly along the periodic table. Thed-band occupation is welknown to affect atomic adsorption for the H and O atomand the general activity of the metal surfaces. From Figwe can also conclude that thed-band occupancy has a direeffect on the water-surface bonding properties. We belithat the effect of the 2D lattice constant, which also increafrom Ru ~2.72 Å! to Au ~2.95 Å!, has a smaller and indireceffect on the interaction with water.

In addition to the RT3 bilayer, half-dissociated bilayproposed by Feibelman has also been calculatedthese surfaces. The two O-M bond lengths are 2.102.20 Å ~except for Au!, and are much shorter than thosethe molecular bilayer. It is a very flat overlayer (zOO;0.05 Å). However this structure is only energeticafavorable on Ru~0001!. On Au~111!, it is completelyrepulsive.

watern

al atom,

TABLE VIII. The geometries and energetics for the H-up, the H-down, and the half-dissociatedbilayers adsorbed on different metal surfaces. HerezOO, zOM1 , andzOM2 are the vertical distances betweethe top and bottom O atoms, the bottom O and the underlying metal atom, and the top O and the metrespectively.

Surface Bilayer zOO ~Å! zOM1 ~Å! zOM2 ~Å! Ea ~meV/molecule!

Ru~0001! H-up 0.86 2.46 3.42 531H-down 0.42 2.69 3.22 533

Half-disso. 0.05 2.09 2.16 766Rh~111! H-up 0.79 2.50 3.40 562

H-down 0.42 2.52 3.12 544Half-disso. 0.04 2.09 2.16 468

Pd~111! H-up 0.60 2.78 3.45 530H-down 0.36 2.66 3.18 546

Half-disso. 0.07 2.09 2.20 89Pt~111! H-up 0.63 2.70 3.37 522

H-down 0.35 2.68 3.14 534Half-disso. 0.06 2.12 2.23 291

Au~111! H-up 0.46 2.90 3.38 437H-down 0.29 2.85 3.25 454

Half-disso. 0.14 2.20 2.43 2472

4-8

ye

wstis

ee

etefth

onwr-ta

asatha

biie

yee

as

d,

inhe

onnot

n intheonn aetab-urle-

terula-ruc-en-are

the

d inreeh

lcu-T3

tosed

ef-For

soon

wel toWe

heseionsry tond

ceere

rfacees.


Recent experiments seem to favor the H-down bilaon Pt~111! ~Ref. 15! and Ru~0001!.18 These experimentsare consistent with our calculations, although the H-dobilayer is only a few tens of meV favorable on Pt. Intereingly, our calculation suggests that the H-up bilayermore favorable than the H-down bilayer on Rh~111!,although no direct comparison with experiment has yet bpossible.

3. The wetting order of the metal surfaces

With the results of the adsorbed monomers and bilayavailable, we now concentrate on a specific issue of wasurface interaction at surfaces, namely, the wettability osurface. This question is general and important to bothfundamental understanding of the water-solid interactiand to technological applications such as biosensor andterproof materials. Experimentally, the wettability of a suface has been characterized macroscopically by the conangle at the interfaces. In a recent study,19 we have proposeda molecular picture of wettability, which is simply definedthe ratio between the H-bond energy of the adsorbed wstructures and the monomer adsorption energy. Such a cacterization has been justified on three surfaces: Pt, Au,graphite.19

The upper panel of Fig. 8 shows the so-defined wettaity, w5EHB /Ea, of these surfaces with the H-bond energdeduced from the bilayer. More explicitly, we used

EHB5~Ea@bilayer#322Ea@monomer# !/3, ~2!

which characterizes the mean H-bond energy in the bilaThe smaller thew is, the stronger is the wettability of thsurface. The upper panel of Fig. 8 shows an order ofw aswRu<wRh,wPd<wPt,wAu , giving a wetting order of Ru.Rh.Pd.Pt.Au. The w51 line has been suggestedthe approximate border dividing the hydrophilic (w<1) andhydrophobic (w@1) surfaces. According to this, Ru, Rh, P

FIG. 8. The structure parameters (zOO, zOM1 , zOM2) and thewettability, defined asw5EHB /Ea, for an H-up bilayer on theRu~0001!, Rh~111!, Pd~111!, Pt~111!, and Au~111! surfaces. Thecase for the H-down bilayer is very similar.

19540

r

n-

n

rsr-aesa-

ct

erar-nd

l-s

r.

and Pt lie in the hydrophilic region. On the contrary, Au isthe hydrophobic region. This division is consistent with texperimental understanding.39,40 The wetting order resultsessentially from the variation of the water-metal interaction these surfaces, because the H-bond energy doeschange appreciably on different surfaces. The trend showFig. 8 also implies a general relationship betweenhydrophilicity-hydrophobicity and the monomer adsorptienergy. Such a relationship was independently found irecent model study,41 where a linear correlation between thcontact angle and the monomer binding energy was eslished from the Monte Carlo simulations. It justifies omodel of wettability, based on the parameters of molecusurface interactions.

C. Vibrational spectra

To provide a database for vibrational recognition of wastructures at the surfaces, we have carried out MD simtions to extract the vibrational spectra for the adsorbed sttures. Vibrational spectra has been quite useful for the idtification of surface and interface structures, because theymeasurable by experiments.

For comparison with experiments, we have calculatedspectra for water monomers, dimers, and bilayers on Pt~111!;bilayers on Pd~111!, Rh~111!, and Au~111!; and H-up,H-down, and half-dissociated bilayers on Ru~0001!. Theeigenfrequencies for the adsorbed structures are listeTable IX. These spectra are generally characterized by thregions:~A! the low-energy modes below 120 meV, whiccorrespond to the translational and librational motions;~B!The HOH bending modes at;200 meV; and~C! the OHstretch modes between 300 and 470 meV. On Pt~111! sur-face, an excellent agreement was found between the calated spectra and the EELS and HAS data for the Rbilayer.14 On Ru~0001!, the vibrational spectra seemsmatch better with that of the H-up bilayer, an issue discusby a recent experiment.18

The vibrational spectra also enable us to estimate thefect of zero point energy on the adsorption energetics.instance, the zero-point energy (Ezp5S i\wi /2) is ;90 meVper molecule for the first bilayer on Pt~111!. It stabilizes thebilayer by 30 meV compared to ice Ih, whoseEzp is 120meV. Such estimation can be applied to other cases, asas the zero-point energy is of concern.

IV. DISCUSSION ON A FEW TOPICS

With the results presented in the preceding section,now turn to discuss a few specific topics, which are centrathe water-metal interactions and the interface properties.bring up this discussion because the understanding on tissues are so far not yet conclusive. Although the discusshere are made on specific structures and systems, we tpoint out their possible implications on other systems aprocesses.

A. The nature of the water-surface bond

First of all, we discuss the nature of the water-surfabond. This issue is relevant and important because thare general concerns about the character of the water-subond, due to the dipole moment of water molecul

4-9

se

2


TABLE IX. The calculated and experimental vibrational energies for the water bilayers on the Pt~111!,Pd~111!, Rh~111!, Au~111!, and Ru~0001! surfaces~in meV!. See Refs. 12,14,32 for the assignment of themodes.

Substrate Translations and librations dHOH nO-HB nO-H

Ru~0001! H-up 34 40 50 67 87 119 200 378, 424 462H-down 20 48 61 73 89 111, 129 196 347, 440 440

Half-disso. 20 32 53 77 117, 129 186, 196 300–380, 428Expt.a 48 68 87 114 189 364, 422, 442 457Expt.b 384, 427 457

Rh~111! H-up 18 44 61 89 111, 129 198 349, 422 466H-down 20 44 75 89 133 200 347, 420 440

Pd~111! H-up 14 40 53 67 89 109, 117 198 374, 424 466H-down 20 42 57 71 89 111, 123 202 380, 426 444

Pt~111! monomer 16 40 61 89 113, 121 190 440dimer 20 32 44 65 85 105, 133 198 347 432, 45H-up 18 32 53 69 87 107, 119 198 388, 432 467

H-down 16 34 57 69 91 111, 119 196, 202 384, 424 438Expt.c 16.5 33 54 65 84 115, 129 201 424 455

Au~111! H-up 17 36 108 201 400, 444 466H-down 18 36 77 105 202 402, 436 468Expt.d 31 104 205 409 ~452!e

aRef. 42.bMultiple by an isotope factor 1.35 from D2O/Ru~0001! ~Ref. 18!.cRef. 32 and 43.dRef. 12.eTaken from water/Ag~111! ~Ref. 44!.

acciao

owser,

thetiesd

s,ega-ec-er–le-

isgs.s

lu-ithtes.es,

nd

cesork

fo

s

ies

The dipole-dipole interaction of van der Waals~vdW!type has been believed to play a role in the water-surfinteractions. Indeed, vdW has been found to be crufor the weakly interacting surfaces, such as watergraphite.45

FIG. 9. Isodensity contours of the difference electron density~a! the water monomer,~b! the dimer,~c! the H-up bilayer, and~d!the H-down bilayer on Pt~111!. The difference density is defined aDr5r@(H2O)n /Pt#2r@(H2O)n#2r@Pt#. Heren is the number ofH2O molecules in the unit cell. The contours have densitDr560.00532ke/ Å 3, for k50, 1, 2, 3, 4. Solid and dashed linecorrespond tonr.0 andnr,0, respectively.

19540

el

n

To see the nature of the water-surface bond, Fig. 9 shthe difference electron density for a water monomer, a dimthe H-up and H-down bilayers on the Pt~111! upon adsorp-tion. The horizontal axis is in the@110# direction, and alsogoes approximately along one of the OH bonds, whilevertical axis is normal to the surface. The induced densiin Fig. 9 exhibit adxz anddz2 character for all the calculatestructures on the Pt~111!. It indicates that thed bands,especially the surface states ofdxz and dz2 characters, ofPt~111!, are generally involved in the water-Pt interactionwhich leads to;0.02 electron transfer from O to Pt. ThH-O-H binding is weakened, as observed by the OH elontion and HOH widening, due to the reduced bonding eltrons in water. This picture is consistent with earlistudies46,47of water on other surfaces, where the lone pairdband coupling was found to be crucial for the molecusurface interaction.

Figure 9 also shows that the water-surface bondingrather localized in the bottom layer, as shown clearly in Fi9~c! and 9~d!. The upper molecule of the RT3 bilayer showvery little coupling to the surface. This led us to the concsion that the water at surfaces forms chemical bond wmetal electrons, especially with those of the surface staThis water-surface bond is rather localized at the interfacand mostly in the bottom layer of molecules. Similar conclu-sion has recently been drawn by Michaelides aco-workers.47

The chemical bonding between water and the surfaoften induces electron transfer. Figure 10 shows the w

r

s

4-10

du-o

fotha

pei-,w

ele

ip

lear

ac

teade43

eVwto

neE

thepg

Wt

nds

tes, as

tratesatters

ndgies,

ar-n ase Hrees.methethe

isat-

ms

tates

i-

al.ing

781eisveur-wt of


function change of the Pt~111! surface adsorbed with H2Omonomers, dimers, and bilayers~averaged over the H-up anH-down cases!. The size of the unit cells used in the calclations provides an approximate calibration of adsorbate cerages, which are 1/9, 1/3, and 2/3 monolayers~ML !, respec-tively. The mean adsorption energy is also plottedcomparison. Water adsorption results in a reduction ofwork function from 5.8 eV to 5.0 eV. In experiments,monotonic decrease of work function was measured for uthe formation of a bilayer,48 with 0.7–0.8 eV decrease at onbilayer. This reduction of the work function is a clear indcation of the electron transfer from water to the surfacepicture consistent with the induced electron densities shoin Fig. 9.

B. The enhancement of hydrogen bonding at surfaces

As a related issue, the interwater interaction, namthe H bonding in the interface structures is another issugeneral interest. In principle, H bonding at surfacesstrongly entangled with the water-surface interactions, escially in small nanostructures and clusters, where a cseparation of the two interactions is difficult. However,qualitative picture of the H bonding at surfaces is still impotant and relevant for a number of issues related to interfwater.

Water adsorption strengthens the H bonding. This can beseen from the adsorbed dimer, the simplest H-bonded sysat surfaces. From Table IV, the H-bond energy in thesorbed dimer on Pt can be estimated, by subtracting ontwo monomer adsorption energy, to be between 258 (32 –30432) and 562 (43332230431) meV, which islarger than the bond strength of the free dimer, 250 mThis enhancement is unusual because it does not agreethe Pauling’s principle for chemical bonding, accordingwhich the H bond of water molecules should be weakewhen more bonds, here the water-surface bonds, form.hancement of H bonding can generally be seen in ostructures, as shown by the last column of Table VI, excfor the adsorbed trimers. In the latter case, the seeminweakened H-bonding is an artifact of our bond counting.counted three water-Pt bonds plus three H bonds in the

FIG. 10. The work function change for water/Pt~111! upon ad-sorption.

19540

v-

re

to

an

y,ofse-ar

-e

m-or3

.ith

dn-ertlyeri-

mers, where in reality the H bonds and water-surface boare closely entangled.

The enhancement of H bonding by the metal substracan be directly seen by the valence charge redistributionshown for water/Pt~111! in early publication.14 This H-bondenhancement has generally been observed on other subsincluding Pd, Rh, Ru, and Au. It is worth mentioning thH-bonding enhancement was also found in water clusand bulk water.49

C. The H-up and H-down conversion in the RT3 bilayer

The two bilayers in the RT3 phase on Pt, the H-up aH-down cases, are nearly degenerate in adsorption ener522 and 534 meV, respectively~see Table VII!. The 2Dstructures of both bilayers have nearly the same oxygenrangement. They are therefore indistinguishable, as soothe measurements are not sensitive to the positions of thatoms, which is true for most experiments. It is therefointriguing to ask what is the barrier between the two statAre the two states distinguishable or are they in fact the sastate? The answer to this question depends critically onenergy barrier between the two states and the nature ofH-atom motion around the barrier. A full account of thissue requires a quantum-mechanical treatment of the Homs along the minimum energy path~MEP!. Here, our dis-cussion relies fully on the classical treatment of the H atoand the DFT calculations of the electrons.

The calculated MEP and the schematic transition s~saddle point! are shown in Fig. 11. The MEP involvemainly the rotation of the upper H2O molecule in the HOHplane. The potential barrier for H-up flipping to H-down blayer is found to be 76 meV at the reaction angle.33°,corresponding to OH angle relative to the surface normThis barrier is substantially lower than the correspondbarrier, 300 meV, on Ru~0001!.47 But it agrees well with thebarrier for similar conformational change in a free dimer,meV ~Ref. 50! ~the energy difference between structuresand 9 in Fig. 2 there!. In this sense, the presence of thPt~111! surface has little influence on this barrier. Thisconsistent with the fact that the barrier is located well abothe bottom water plane, where the interaction with the sface is predominant. It would be interesting to find out hothis barrier changes in a quantum-mechanical treatmenthe H atoms.51

FIG. 11. The minimum energy pathway~MEP! and the transi-tion state between the H-up and H-down bilayers on Pt~111!.

4-11

lf-

ns

tioi

toisfo

n

l-re

r,

n

nth

uro,

a

eswn

tio

o-in

vehe

son. Aylebe

enlay-terstoitha

ainarts.es,thechly

truc-e-

wa-

as

icaltheub-

b-andpec-da-by

FTtalles.onpre-

haseral-warein-

oret-ple,a-py,rces

serve

da-

ore


D. Partial dissociation of water bilayers

A recent experiment18 questioned the existence of a hadissociated water adlayer on Ru~0001!, proposed by Feibel-man. It is therefore interesting to look at the kinetic costraints for the dissociation and other competing processuch as molecular desorption. The barrier for dissociafrom the H-down bilayer to the half-dissociated structure0.62 eV, as found in another calculation.31 However the ad-sorption energy of the RT3 bilayer is 0.53 eV accordingour calculation, which implies that water in the bilayereasier to desorb than to dissociate. The higher barrierdissociation may thus completely prohibit the existenceany dissociated water on Ru, as found in experiments;140 K.

Partial dissociation of H2O molecules have also beefound in the RT39 phase,16 forming H3O1 and OH2 likegroups on Pt~111!, although the fraction of dissociated moecules is very small. From our calculation, we found thdissociated molecules out of 32~9%! in the first RT39 bi-layer, with all H3O1 lying in the upper layer and the OH2

lying in the bottom layer. When the water film grows thickeless dissociation is found: only 2 out of 64~3%! molecules intwo bilayers and 1 out of 96~1%! in three bilayers. Contraryto the RT39 phase, no dissociation is found in the RT3 aRT37 bilayers.

The partial dissociation of the RT39 bilayer on Pt~111!results from both lateral compression of the water film aits interaction with the substrate. This is evidenced byfact that if we remove the Pt substrate, one and only one H2Odissociates, compared to three in the first bilayer on Pt. Fther evidence for surface induced dissociation comes fthe fact that only the bottom H2O molecules donate protonsindicating the influence of the substrate. In contrast, nonethe upper H2O ~H-down! molecules is found to donateproton to other H2O. Therefore all the H3O1s lie in theupper layer of the first bilayer while the OH2s bind to the Ptsurface with a Pt-O bond length of 2.1 Å. Lateral comprsion also contributes to partial dissociation, because oneter molecule is dissociated in the RT39 even in the abseof the surface, and because the RT37 or RT3 phases doexhibit any dissociation. Compression induced dissociais a well-known phenomenon in bulk water52 and thinfilms.53 The effect of compression in the RT39 bilayer prvides a surface example of this general phenomenontwo-dimensional system.

V. SUMMARY AND CONCLUSIONS

The adsorption of water on metal surfaces has been intigated byab initio DFT calculations. From these studies, tfollowing conclusions can be drawn.

~i! The interaction between water and metal surfacedominated by a chemical bonding formed between the lpair of water molecules and the surface electronic statesa result, the water-surface bond is rather localized, mostlthe contacting regions and bottom layers like, for exampthe donor of the dimer, and the bottom molecule in thelayer. Long-range polarization does exist, however, its effis relatively smaller.

19540

-esns

orfat

e

d

de

r-m

of

-a-cenotn

a

s-

ises

in,

i-ct

~ii ! An enhancement of the interwater H bonding has begenerally observed in both nanometer clusters and overers. This enhancement is especially obvious in small clussuch as the dimer and in the first bilayers. We would likepoint out that such an enhancement is not consistent wPauling’s principle for chemical bonds, but seems to beunique feature of the H-bonded water molecules.

~iii ! The structure of water in the adsorbed states remlargely the same as the gas-phase or bulk ice counterpAlthough water molecules do adjust their bonding featurlike the bond lengths and angles, upon adsorption. Onsurfaces with strong water-metal coupling, like Ru, Rh, suadjustment is more significant. It is minor on the weakinteracting systems such as Au. In the latter case, water stures remain largely rigid. This reflects the competition btween the two fundamental interactions in the adsorbedter molecules. In this sense, water could also be hard.

~iv! The wetting order of the studied surfaces is foundRu.Rh.Pd.Pt.Au, the same order as thed-band occu-pancy of the metals. This ordering results from the chemreactivity of the substrates, and is a direct indication oflocalized electronic coupling between water and the sstrates.

~v! Vibrational spectra of various water structures are otained and are generally consistent with the structuresinteractions present upon adsorption. These vibrational stra, in particular the OH stretch modes, provide a usefultabase for vibrational recognition of interface structuresexperiments.

The results presented in this paper, based on the Dcalculations, gained much insight into the fundamenwater-metal interactions at the atomic to electronic scaDetailed characterization of the prototype water structuresseveral metal surfaces has been documented and a comhensive understanding of the water-metal interactionsemerged. We believe that such an understanding is genizable to other surfaces. Nevertheless, one should be aof the fact that some important aspects of water-surfaceteractions have not been tackled and are beyond the theical approach we adopted here. These include, for examthe quantum-mechanical character of H-atom motion in wter; the kinetic and thermodynamical effects such as entrotemperature, and pressure; and the possible dispersive fosuch as the van der Waals interactions. These issues deour attention in future studies.

ACKNOWLEDGMENTS

This work was supported by the Natural Science Fountion of China~Grant Nos. 60021403 and 10134030!, MOST~Grant Nos. G2000067103 and 2002AA311151!, the Swed-ish Research Council~Grant No. VR 621-2001-2614!, andthe ATOMICS consortium, funded by the Foundation fStrategic Research~SSF!. The allocated computer time at thHigh Performance Computing Center at North~HPC2N! andthe National Supercomputer Center~NSC! of Sweden isgratefully acknowledged.

4-12

n

em

n

M

,

e

m.

g,

y,

ci.

.

u-

rf.

hys.

L.

re

ett.


1P.A. Thiel and T.E. Madey, Surf. Sci. Rep.7, 211 ~1987!.2M.A. Henderson, Surf. Sci. Rep.46, 1 ~2002!.3M. Morgenstern, T. Michely, and G. Comsa, Phys. Rev. Lett.77,

703 ~1996!; M. Morgensternet al., Z. Phys. Chem.~Munich!198, 43 ~1997!.

4L.E. Firment and G.A. Somorjai, J. Chem. Phys.63, 1037~1975!;Surf. Sci.84, 275 ~1979!.

5D.L. Doering and T.E. Madey, Surf. Sci.123, 305 ~1982!.6K. Morgenstern and J. Nieminen, Phys. Rev. Lett.88, 066102

~2002!.7K. Morgenstern and K-H. Rieder, J. Chem. Phys.116, 5746

~2002!.8T. Mitsui, M.K. Rose, E. Fomin, D.F. Ogletree, and M. Salmero

Science297, 1850~2002!.9S. Haq, J. Harnett, and A. Hodgson, Surf. Sci.505, 171 ~2002!.

10P.J. Feibelman, Science295, 99 ~2002!; D. Menzel,ibid. 295, 58~2002!.

11K.D. Gibson, M. Viste, and S.J. Sibener, J. Chem. Phys.112,9582 ~2000!.

12G. Pirug and H.P. Bonzel, Surf. Sci.405, 87 ~1998!.13A. Glebov, A.P. Graham, A. Menzel, and J.P. Toennies, J. Ch

Phys.106, 9382~1997!.14S. Meng, L.F. Xu, E.G. Wang, and S.W. Gao, Phys. Rev. Lett.89,

176104~2002!.15H. Ogasawara, B. Brena, D. Nordlund, M. Nyberg, A. Pelme

schikov, L.G.M. Pettersson, and A. Nilsson, Phys. Rev. Lett.89,276102~2003!.

16P.J. Feibelman, Phys. Rev. Lett.91, 059601~2003!.17D.J. Mann and M.D. Halls, Phys. Rev. Lett.90, 195503~2003!.18D.N. Denzler, Ch. Hess, R. Dudek, S. Wagner, Ch. Frischkorn,

Wolf, and G. Ertl, Chem. Phys. Lett.376, 618 ~2003!.19S. Meng, E.G. Wang, and S.W. Gao, J. Chem. Phys.119, 7617

~2003!.20G. Kresse and J. Hafner, Phys. Rev. B47, 558 ~1993!; 49, 14251

~1994!; J. Phys.: Condens. Matter6, 8245~1994!; G. Kresse andJ. Furthmu¨ller, Comput. Mater. Sci.6, 15 ~1996!; Phys. Rev. B54, 11169~1996!.

21H.J. Monkhorst and J.D. Pack, Phys. Rev. B13, 5188~1976!.22M. Methfessel and A.T. Paxton, Phys. Rev. B40, 3616~1989!.23H. Jonsson, G. Mills, and K.W. Jacobsen, inClassical and Quan-

tum Dynamics in Condensed Phase Simulations, edited by B.J.Berne, G. Ciccotti, and D. F. Coker~World Scientific, Singapore1998!.

24D. Vanderbilt, Phys. Rev. B41, 7892~1990!.25J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. P

erson, D.J. Singh, and C. Fiolhais, Phys. Rev. B46, 6671~1992!.

19540

,

.

-

.

d-

26L. Giordano, J. Goniakowski, and J. Suzanne, Phys. Rev. Lett.81,1271 ~1998!.

27S. Tsuzuki and H.P. Lu¨thi, J. Chem. Phys.114, 3949~2001!.28F. Sim, A. St-Amant, I. Papai, and D.R. Salahub, J. Am. Che

Soc.114, 4391~1992!.29R.N. Barnett and U. Landman, Phys. Rev. B48, 2081~1993!.30S. Izvekov and G.A. Voth, J. Chem. Phys.115, 7196~2001!.31A. Michaelides, V.A. Ranea, P.L. de Andres, and D.A. Kin

Phys. Rev. Lett.90, 216102~2003!.32K. Jacobi, K. Bedu¨rftig, Y. Wang, and G. Ertl, Surf. Sci.472, 9

~2001!.33M. Nakamura, Y. Shingaya, and M. Ito, Chem. Phys. Lett.309,

123 ~1999!; M. Nakamura and M. Ito,ibid. 325, 293 ~2000!.34A.L. Glebov, A.P. Graham, A. Menzel, Surf. Sci.427, 22 ~1999!.35H. Ogasawara, J. Yoshinobu, and M. Kawai, J. Chem. Phys.111,

7003 ~1999!; Chem. Phys. Lett.231, 188 ~1994!.36J.K. Gregory, D.C. Clary, K. Liu, M.G. Brown, and R.J. Saykall

Science275, 814 ~1997!.37G.B. Fisher and J.L. Gland, Surf. Sci.94, 446 ~1980!.38F.T. Wagner and T.E. Moylan, Surf. Sci.191, 121 ~1987!.39R.S. Smith, C. Huang, E.K.L. Wong, and B.D. Kay, Surf. S

367, L13 ~1996!.40P. Lofgren, P. Ahlstro¨m, D.V. Chakarov, J. Lausmaa, and B

Kasemo, Surf. Sci.367, L19 ~1996!.41T. Werder, J.H. Walther, R.L. Jaffe, T. Halicioglu, and P. Ko

moutsakos, J. Phys. Chem. B107, 1345~2003!.42P.A. Thiel, R.A. DePaola, and F.M. Hoffmann, J. Chem. Phys.80,

5326 ~1984!.43B.A. Sexton, Surf. Sci.94, 435 ~1980!.44A.F. Carley, P.R. Davies, M.W. Roberts, and K.K. Thomas, Su

Sci. 238, L467 ~1990!.45D. Feller and K.D. Jordan, J. Phys. Chem. A104, 9971~2000!.46H.P. Bonzel, G. Pirug, and J.E. Mu¨ller, Phys. Rev. Lett.58, 2138

~1987!.47A. Michaelides, A. Alavi, and D.A. King, J. Am. Chem. Soc.125,

2746 ~2003!.48I. Villegas and M.J. Weaver, J. Phys. Chem.100, 19 502~1996!.49S.S. Xantheas, C.J. Burnham, and R.J. Harrison, J. Chem. P

116, 1493~2002!.50B.J. Smith, D.J. Swanton, J.A. Pople, H.F. Schaefer III, and

Radom, J. Chem. Phys.92, 1240~1990!.51D. Marx, M.E. Tuckerman, J. Hutter, and M. Parrinello, Natu

~London! 397, 601 ~1999!.52E. Schwegler, G. Galli, F. Gygi, and R.Q. Hood, Phys. Rev. L

87, 265501~2001!.53S. Meng, E. G. Wang, and S. W. Gao~unpublished!.

4-13

Water adsorption on metal surfaces: A general picture from ...everest.iphy.ac.cn/papers/prb69.195404.pdf · Water adsorption on metal surfaces: A general picture from density functional

Documents