PHYSICAL REVIEW B 81, 075430 共2010兲

Density functional study of oxygen on Cu(100) and Cu(110) surfaces

X. Duan,1,* O. Warschkow,1,† A. Soon,1 B. Delley,2 and C. Stampfl1

1School of Physics, The University of Sydney, Sydney, New South Wales 2006, Australia
2Paul-Scherrer-Institut, WHGA/123, Villigen CH-5232, Switzerland

共Received 20 July 2009; revised manuscript received 11 November 2009; published 23 February 2010兲
Using density-functional theory within the generalized gradient approximation, we investigate the interac-
tion between atomic oxygen and Cu共100兲 and Cu共110兲 surfaces. We consider the adsorption of oxygen at
various on-surface and subsurface sites of Cu共100兲 for coverages of 1/8 to 1 monolayers 共ML兲. We find that
oxygen at a coverage of 1/2 ML preferably binds to Cu共100兲 in a missing-row surface reconstruction, while
oxygen adsorption on the nonreconstructed surface is preferred at 1/4 ML coverage consistent with experi-
mental results. For Cu共110兲, we consider oxygen binding to both nonreconstructed and added-row reconstruc-
tions at various coverages. For coverages up to 1/2 ML coverage, the most stable configuration is predicted to
be a p共2 ⫻ 1兲 missing-row structure. At higher oxygen exposures, a surface transition to a c共6 ⫻ 2兲 added strand
configuration with 2/3 ML oxygen coverage occurs. Through surface Gibbs free energies, taking into account
temperature and oxygen partial pressure, we construct 共p , T兲 surface phase diagrams for O/Cu共100兲 and
O/Cu共110兲. On both crystal faces, oxygenated surface structures are stable prior to bulk oxidation. We combine
our results with equivalent 共p , T兲 surface free energy data for the O/Cu共111兲 surface to predict the morphology
of copper nanoparticles in an oxygen environment.

DOI: 10.1103/PhysRevB.81.075430 PACS number共s兲: 68.43.Bc, 81.65.Mq, 68.47.Gh

I. INTRODUCTION co-workers,33 explaining the observed 共2冑2 ⫻ 冑2兲R45° and

c共2 ⫻ 2兲 phases as two sides of an entropy-driven order-
Copper-based catalysts are of importance to a number of disorder transition.
industrial processes including the synthesis of methanol,1 the On Cu共110兲, oxygen adsorption results in a major recon-
reduction and decomposition of nitrogen oxides,2–4 the oxy- struction of the surface.20 Two well-ordered superstructures
dehydrogenation of ammonia,5,6 electroless plating,7 fuel cell are known: a 共2 ⫻ 1兲 phase with an oxygen coverage of 1/2
electrodes,8 and the treatment of waste water.9 In copper ML and a c共6 ⫻ 2兲 phase with a coverage of 2/3 ML. The
catalysis—as for many other metal catalysts—surface oxida- oxygen-induced 共2 ⫻ 1兲 reconstruction has been the subject
tion and oxidic overlayers are widely believed to play a cru- of several experimental and theoretical studies,15,20,22,37–49
cial role.10,11 and the added row structure41 is now generally accepted. The
Oxygen adsorption on Cu共100兲 has been the subject of oxygen-induced c共6 ⫻ 2兲 reconstruction is receiving increas-
numerous studies12–36 and a number of distinct oxide recon- ing attention50–54 for its hypothesized role as a precursor
structions are known. Low energy electron diffraction phase of bulk oxidation. The transition between the 共2 ⫻ 1兲
共LEED兲 experiments14,19 identified a 共2冑2 ⫻ 冑2兲R45° recon- and c共6 ⫻ 2兲 surface phases was directly examined by Sun et
struction that appears at oxygen coverages above 0.3 ML. al.53 using reflectance difference spectroscopy which deter-
This phase is also observed in scanning tunneling micros- mined a transition temperature of 660 K.
copy 共STM兲 and x-ray diffraction 共XRD兲 experiments and In light of the importance of temperature and the avail-
was characterized as a missing-row 共MR兲 reconstruction ability of oxygen on determining surface reconstruction, we
with an oxygen content of 1/2 ML.16,17,20,23 At low coverage use the theoretical framework of ab initio atomistic
共⬍0.3 ML兲, STM images show islands with a c共2 ⫻ 2兲 struc- thermodynamics11,55,56 to describe the relative stability of
ture on the clean Cu共100兲 surface,23 supporting earlier LEED various O/Cu共100兲 and O/Cu共110兲 reconstructions in contact
evidence for the existence of such a reconstruction.19 Early with a realistic gaseous environment. This leads us to a com-
theoretical work has rationalized the structural properties of plete thermodynamic model of the stable surface phase be-
the O/Cu共100兲 system in terms of orbital hybridization15 and tween the oxygen-free surface to the bulk oxide Cu2O. Im-
charge transfer processes.18 More recent ab initio studies25–27 portantly, the thermodynamical analysis shows that bulk
have focused on the two observed phases and discuss their Cu2O formation imposes an upper limit on the oxygen cov-
formation in terms of the electronic structure. Kangas and erage of the surface, which, interestingly, is very nearly the
co-workers.29,35 contrast the stability of oxygen at surface same for the three principal low-index surfaces 共100兲, 共110兲,
and subsurface sites for coverages up to 2 ML, finding stable and 共111兲. The results presented here follow on, and comple-
subsurface binding at coverages above approximately 3/4 ment, our earlier results for the O/Cu共111兲 surface.57 This
ML. Using density-functional theory and density-functional completes our free energy treatment of the three principal
perturbation theory, Bonini et al.31 compare the 共2冑2 low-index surfaces of the oxygen-on-copper system at a con-
⫻ 冑2兲R45° missing-row reconstructions of O/Cu共100兲 and sistent level of theory allowing free-energy comparisons be-
O/Ag共100兲 and find similar structural and vibrational prop- tween the three crystal phases. We use the combined free
erties. An important insight was recently provided by the energy data to predict the morphology of copper nanopar-
joint experimental and theoretical work of Iddir and ticles in an oxygen environment.

1098-0121/2010/81共7兲/075430共15兲 075430-1 ©2010 The American Physical Society

DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

Our discussion is organized as follows: In Sec. II, our 0.002 eV/ Å2, respectively. For Cu共110兲, coverages of be-
computational methodology is detailed, including the defini- tween 1/8 to 1 ML are achieved using 共2 ⫻ 1兲, 共3 ⫻ 1兲, 共4
tions of several calculated quantities and a brief description ⫻ 1兲, 共2 ⫻ 2兲, 共3 ⫻ 2兲, 共8 ⫻ 1兲, 共2 ⫻ 4兲, and c共6 ⫻ 2兲 cells.
of the ab initio atomistic thermodynamics approach. Section Brillouin-zone integrations are performed using 共12⫻ 12
III A revisits the clean Cu共100兲 and Cu共110兲 surfaces, which ⫻ 1兲 and 共6 ⫻ 12⫻ 1兲 Monkhorst-Pack grids for the 共1 ⫻ 1兲
serves to validate our methodology against previous results surface unit cells of Cu共100兲 and Cu共110兲, respectively. For
in the literature. In Sec. III B, we present our calculations for the larger surface unit cells, correspondingly smaller grids
the O/Cu共100兲 system discussing energetic, structural, and are used to ensure an equivalent sampling of reciprocal
electronic properties. The corresponding results for space. The convergence with respect to k-point density of
O/Cu共110兲 are presented in Sec. III C. In Sec. IV we com- binding energies and free surface energies to below 0.01 eV
bine our data into thermodynamic 共p , T兲-phase diagrams to and 0.001 eV/ Å2, respectively, was earlier established by
facilitate comparison with experiment. Soon et al.61 using the same computational model of the Cu
surface.
Average binding energies per oxygen atom, EO b , are given
II. COMPUTATIONAL METHOD
relative to the clean surface and an isolated oxygen mol-
All calculations are performed using density-functional ecule. We calculate EO b as
theory 共DFT兲 in the generalized gradient approximation
共GGA; Perdew, Burke, and
exchange-correlation兲58 as implemented in the DMol3
Ernzerhof EO
b =−
1
NO
冉slab
EO/Cu slab
− ECu − ⌬NCuECu −
NO
2
E O2 , 冊 共1兲

software.59 The electronic eigenfunctions are expanded in where NO is the number of adsorbed oxygen atoms, EO/Cu slab
is
terms of a localized atomic orbital basis set of double nu- the total energy of the adsorbate-substrate system, EO2 is the
merical plus polarization 共DNP兲 quality and a real-space cut- energy of an isolated oxygen molecule 共see below兲, and ECu slab
off of 9 Bohr. Our calculations include scalar-relativistic cor- is the energy of the clean surface. ⌬NCu is the difference in
rections. A thermal broadening of 0.1 eV is applied to the number of Cu atoms between the O/Cu system and the
improve the convergence of the self-consistent procedure. clean surface slab, and ECu is the energy of a Cu atom in bulk
For bulk copper, we use a 12⫻ 12⫻ 12 Monkhorst-Pack Cu. The term ⌬NCuECu accounts for differences in the num-
grid to integrate over the Brillouin zone of the elemental ber of Cu atoms, when we compare oxygen adsorption be-
fcc-unit cell and obtain a calculated lattice constant of tween clean, missing-row, and added-row reconstructions.
3.64 Å, in good agreement with experiment 共3.61 Å兲.60 The Average binding energies are defined such that a positive
calculated bulk modulus and cohesive energy are 136 GPa value indicates a thermodynamically favorable adsorption
and 3.45 eV, respectively; also in good agreement with the process.
experimental values of 137 GPa and 3.49 eV.60 The energies of an oxygen molecule 共EO2兲 and an isolated
Periodic slabs are used to model the surfaces of copper.
oxygen atom are obtained from spin-polarized calculations
The Cu共100兲 slab is composed of 9 atomic layers with an
where we use a real-space basis set cutoff of 20 Bohr. Using
approximate thickness of 14.6 Å. The Cu共110兲 surface is
these two energies, we calculate the molecular binding en-
represented using a slab of 11 atomic layers which is ap-
ergy of an oxygen molecule to be 6.08 eV 共or 3.04 eV per
proximately 13.0 Å thick. Successive slabs in the three-
atom兲, somewhat larger than the experimental value of 5.12
dimensionally repeated unit cell are separated by a large
eV 共or 2.56 eV per atom兲. Our overestimation of the O2
vacuum region of ⬇30 Å. In all our calculations, oxygen
binding energy is inline with previous studies that use DFT
atoms are adsorbed on both surfaces of the slab in order to
in the GGA approximation.47,62 For the molecular bond
preserve inversion symmetry. In our geometry optimizations,
length and the vibrational frequency we obtain 1.22 Å and
three atomic layers at the center of the slab are held fixed at
1544 cm−1, respectively, in good agreement with the experi-
the calculated bulk positions. All other atoms are fully re-
mental values of 1.21 Å and 1580 cm−1 共Ref. 63兲.
laxed until all force components acting on the atoms are
The effect of the gaseous environment on the relative sta-
below 0.015 eV/ Å. Test calculations using larger slabs have
bility of the considered surface structures is captured by the
confirmed that the resultant error to the average oxygen bind-
method of ab initio atomistic thermodynamics.11,55,56 The
ing energy and the Gibbs free surface energy 共see definitions
availability of oxygen from the environment is represented
below兲 are smaller than 0.01 eV and 0.001 eV/ Å2, respec-
by the oxygen chemical potential, ␮O, which is a function of
tively.
the gas-phase temperature and the oxygen partial pressure,
Several surface supercells are used to study the effects of
pO2. For a given ␮O共pO2 , T兲, the thermodynamically pre-
oxygen coverage, ⌰, on the electronic and geometric struc-
ture of the surface. For the Cu共100兲 surface we consider ferred surface is the one with the lowest Gibbs free energy ␥
coverages of between 1/8 and 1 ML, using 共1 ⫻ 1兲, 共2 ⫻ 2兲, which we calculate as
共4 ⫻ 4兲, 共2冑2 ⫻ 冑2兲R45°, 共2冑2 ⫻ 2冑2兲R45°, and c共4 ⫻ 6兲 sur- 1
face unit cells. For the large c共4 ⫻ 6兲 unit cell, computational ⌬␥ = 共Gslab − GCu
slab
− ⌬NCu␮Cu − NO␮O兲. 共2兲
2A O/Cu
restrictions necessitated a thinner, five-layer slab, as opposed
to the regular nine-layer slab. This increases for this structure In this equation, ⌬␥ is the surface Gibbs free energy rela-
the error in the average oxygen binding energy and the Gibbs tive to the oxygen-free, nonreconstructed surface. The calcu-
free surface energy to smaller than 0.03 eV and lated Gibbs free energies of the oxygenated and the oxygen-

075430-2
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

free slab model used to represent the surface are denoted, both surfaces and compare our calculated interlayer spacings
slab
GO/Cu slab
and GCu , respectively. The term ⌬NCu␮Cu represents with previously reported values. In the following, interlayer
the cost of exchanging Cu atoms with a reservoir with separations near the surface are given as percent changes ⌬ij
chemical potential ␮Cu. In our case, this reservoir is bulk relative to the respective bulk separations of 1.82 and
copper with which the surface is assumed to be in equilib- 1.30 Å in the 具100典 and 具110典 directions. For the Cu共100兲
rium; thus, ␮Cu equals the free energy of a copper atom in surface, the change between the first and second atomic layer
bulk. In Eq. 共2兲, the factor 2A 1
is the reciprocal of the total 共this is denoted ⌬12兲 is calculated to be −2.3%, correspond-
共in-plane兲 surface area of the slab model 共i.e., the sum of ing to a small contraction relative to bulk. Between the sec-
both sides兲. The factor serves to normalize to a common unit ond and third layer the change is ⌬23 = +1.0%. These values
area the surface free energies of slabs of varying unit cells. are in good agreement with previous DFT-GGA
In our calculations, we neglect in GO/Cuslab
and GCuslab
the free calculations65 that report ⌬12 = −2.6% and ⌬23 = +0.9% and
energy contributions due to vibrational motion and the experimental LEED data68,69 with ⌬12 = −2.1% and ⌬23 =
pressure-volume term.56 This allows us to replace these +0.5%. For the Cu共110兲 surface, we obtain ⌬12 = −10.0% and
terms with the respective DFT total energies, EO/Cu slab
and ECuslab
. ⌬23 = +3.0%, consistent with the experimental values of ⌬12
It is convenient to combine Eqs. 共1兲 and 共2兲 to obtain the = −10.0⫾ 2.5% and ⌬23 = 0.0⫾ 2.5% 共Ref. 66兲, and theoreti-
surface free energy in terms of the average oxygen binding cal results using the linearized augmented plane wave
energy, i.e., 共FLAPW兲 method,67 yielding ⌬12 = −9.7% and ⌬23 = +3.6%.
Our calculated work functions for clean Cu共100兲 and
1 Cu共110兲 surfaces are 4.39 and 4.18 eV, respectively, which
⌬␥ ⬇ b − NO⌬␮O兲,
共− NOEO 共3兲
2A are in reasonable agreement with the experimental values of
4.59 and 4.48 eV.64 Earlier theoretical results include those
In this equation, ⌬␮O denotes the chemical potential of Skriver and Rosengaard71 who calculated values of 5.26
given relative to the dissociation energy EO2 of an isolated and 4.48 eV using linear muffin-tin orbitals and the local
O2 molecule, i.e., we define density approximation 共LDA兲. Tao et al.65 report a work
1 function of 4.49 eV for Cu共100兲 using DFT-GGA in good
⌬ ␮ O = ␮ O − E O2 . 共4兲 agreement with our value of 4.39 eV. We note that the work
2
functions of metal surfaces are usually underestimated by
In our electronic structure analysis, we consider electron DFT-GGA calculations.67
density differences Our calculated surface energies for Cu共100兲 and Cu共110兲
are 0.64 and 0.97 eV per surface unit cell, respectively.
n⌬共r兲 = nO/Cu共r兲 − nCu共r兲 − nO共r兲, 共5兲 These results correlate well with the FLAPW calculations of
where n O/Cu
共r兲 is the total electron density of the substrate- Da Silva et al.72,73 reporting surface energies of 0.60 eV for
adsorbate system, and nCu共r兲 and nO共r兲 are the electron den- Cu共100兲 and 0.90 eV for Cu共110兲. The more approximate
sities of the clean substrate and the isolated O ad-layer, re- LDA muffin tin calculations of Skriver and Roosengard71
spectively. The density difference is evaluated and plotted for yield 0.85 and 1.33 eV, respectively.
the relaxed adsorbate system, which highlights those regions
near the surface where oxygen adsorption induces a deple-
B. Oxygen adsorption on Cu(100)
tion or accumulation of the electron density.
The surface dipole moment 共in Debye兲 is evaluated using Cu sublattices and O binding sites. We examine oxygen
the Helmholtz equation adsorption on three types of Cu共100兲 surface structures 共Fig.
1兲. The first structure is the simple, nonreconstructed
A⌬⌽ Cu共100兲 surface shown in Fig. 1共a兲 which defines the
␮= , 共6兲
12␲⌰ 共1 ⫻ 1兲 surface unit cell of this crystal face. On this surface
there are three distinct high-symmetry adsorption sites: The
where A is the area 共in Å2兲 of a 共1 ⫻ 1兲 surface unit, and ⌬⌽
top site 共T兲 is located directly on top of a surface Cu atom,
is the work-function change 共in eV兲 relative to the clean
the bridge site 共B兲 is a twofold coordinated site between two
surface. The work function is defined as the difference be-
Cu atoms, and the hollow site 共H兲 is coordinated by four Cu
tween the electrostatic potential 共U兲 in the middle of the
atoms. The second surface structure 关Fig. 1共b兲兴 is a missing-
vacuum region and the Fermi energy 共EF兲, so that,
row reconstruction with a 共2冑2 ⫻ 冑2兲R45° unit
⌬⌽ = U − EF − ⌽clean , 共7兲 cell,14,16,19,20,23 which features a chain of Cu atom vacancies
in the 关011兴 direction. On this reconstruction the most favor-
where ⌽clean is the work function of the clean surface. able adsorption site is the pseudohollow site 共pH兲, which is
similar to the hollow site on the clean surface, but only three-
III. RESULTS fold coordinated due to the missing row of Cu atoms. The
third structure prototype is the distributed vacancy recon-
A. Clean Cu(100) and Cu(110) surfaces
struction 共not shown in Fig. 1兲, which is closely related to the
The clean Cu共100兲 and Cu共110兲 surfaces are well charac- missing-row reconstruction, but is characterized by a nonor-
terized experimentally and theoretically in the literature.64–70 dered, random distribution of Cu vacancies in the surface.
As a point of reference, we relax the atomic positions for The importance of the vacancy order/disorder transition was

075430-3
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

FIG. 1. 共Color online兲 Oxygen binding sites on the Cu共100兲

surface showing in 共a兲 the nonreconstructed 共n兲 surface, and in 共b兲
the missing-row 共MR兲 reconstructed surface. Top-layer Cu atoms
are white and second layer Cu atoms are gray. The in-plane position
of surface and subsurface binding sites are indicated by red squares FIG. 2. 共Color online兲 Calculated average binding energy per
and labeled below the diagram. oxygen atom on Cu共100兲 in the on-surface and subsurface sites, for
various coverages. The solid lines connecting the calculated binding
highlighted by the work of Iddir et al.33 and we will be using energies are used to guide the eye. The dashed line indicates, as a
their 共2冑2 ⫻ 2冑2兲R45° quasi-disordered model of this sur- guide to the eye, the convex hull connection between the 1/4 ML
face to study oxygen binding at the pseudohollow sites 共pH兲. nonreconstructured surface and the 1/2 ML missing-row
Additionally, for nonreconstructed and missing-row surfaces, reconstruction.
we will be exploring subsurface binding sites which are in-
dicated in Fig. 1. In the following discussion, we will use N-c共2 ⫻ 2兲 1/2 ML hollow-site structures. The average oxy-
prefixes N-, MR-, and DV- with our unit cell notation to gen binding energies is 1.81 eV for the c共4 ⫻ 6兲, to be com-
indicate nonreconstructed, missing row, and distributed va- pared with 1.83 eV for the linear combination of N-共2 ⫻ 2兲
cancy reconstructions, respectively. 1/4 and N-c共2 ⫻ 2兲 1/2 ML. This is just within the error of
Oxygen on the nonreconstructed surface. We first con- our calculations due to slab thickness effects.
sider the adsorption of oxygen on the three high-symmetry For a coverage of 1/2 ML we also consider subsurface
sites 共top, bridge, and hollow兲 of the nonreconstructed sur- binding of oxygen on the nonreconstructed surface. Two sites
face 共N兲 for a range of coverages. For each structure, we were tested: the tetrahedral subsurface site 关see Fig. 1共a兲兴,
optimize the geometry and calculate the average oxygen which is located directly under the surface bridge site, be-
binding energy, the work function change, and the surface tween the first and second atomic layers, and the octahedral
dipole moment. The calculated binding energies are plotted site directly under the surface top site in the second layer.
in Fig. 2 and are listed for selected structures in Table I. For Our calculations show that oxygen atoms in these subsurface
coverages of 0.5 ML and above, the binding energy de- sites are much less stable than the on-surface hollow and
creases almost linearly with coverage indicating a gradual bridge sites as shown in Fig. 2. Overall, our binding energy
build-up of repulsive interactions between the adsorbed oxy- results are in qualitative and quantitative agreement with the
gen atoms.28,76 The highly coordinated hollow site 共red data previous GGA calculations of Kangas et al.29
points兲 is energetically preferred at all coverages, and top site We note in passing that the binding of oxygen to Cu共100兲
adsorption is always endothermic 共i.e., EO b ⬍ 0兲. The bridge is significantly stronger than to Ag共100兲. For 1/2 ML hollow-
site is less stable than the hollow site by approximately 0.8 to site adsorption on Ag共100兲, Cipriani et al.74 and Gajdoŝ
0.9 eV. For the energetically preferred hollow site, we have et al.75 report GGA binding energies of 0.74 and 0.71 eV,
additionally investigated a lower coverage of 1/8 ML but find respectively. This is less than half of the 1.71 eV binding
only a marginal change in the binding energy relative to the energy that we find for Cu共100兲.
1/4 ML result. This suggests that at these low coverages the Oxygen on missing row and disordered vacancy recon-
interaction between oxygen adatoms is effectively screened. structions. Evidence from STM 共Ref. 16兲 and diffraction17
Included in our nonreconstructed surfaces is a c共4 ⫻ 6兲 experiments suggest the formation of a 共2冑2 ⫻ 冑2兲R45° re-
structure with a hollow-site oxygen coverage of 1/3 ML. construction when the oxygen coverage reaches 1/2 ML.
This structure is inspired by earlier work,77 which shows this Shown in Fig. 3共b兲, this reconstruction is characterized by a
structure to be thermodynamically stable on the Ag共100兲 sur- missing row of copper atoms with 1/2 ML oxygen atoms
face. For the equivalent Cu共100兲 structure the stability is not adsorbed at the pseudohollow sites 共pH兲 alongside the miss-
as clear—as shown in Fig. 2, the average oxygen binding ing row. In order to evaluate the stability of these missing-
energy of the c共4 ⫻ 6兲 is very slightly smaller 共i.e., less row reconstructions, we consider a number of variants of the
stable兲 than the linear combination of our N-共2 ⫻ 2兲 1/4 and 共2冑2 ⫻ 冑2兲R45° structure. One variant reconstruction with a

075430-4
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

TABLE I. Calculated structural parameters for different O coverage 共from 1/8 to 1 ML兲 on Cu共100兲 in the
hollow site. The values of oxygen adsorbed on the missing-row 共MR兲 reconstructed surface for 1/4 and 1/2
ML are included for comparison. The binding energy, EO b , is calculated with respect to the isolated oxygen
molecule. dCu-O indicates the average bond length between oxygen and the nearest copper atom. dO1 is the
minimum vertical height of O with respect to the surface Cu layer, ⌬ij 共i = 1 , 3 , j = 2 , 4兲 represents percent
change relative to the bulk interlayer spacing, between the ith and jth atomic interlayer, where the center-
of-mass of the layer is used. Also given are the workfunction change, ⌬⌽, and the corresponding surface
dipole moment, ␮.

N-共4 ⫻ 4兲 N-共2 ⫻ 2兲 MR-共4冑2 ⫻ 2冑2兲R45° N-共2 ⫻ 2兲 MR-共2冑2 ⫻ 2冑2兲R45° N-共2 ⫻ 2兲 N-共1 ⫻ 1兲

Structure 1/8 ML 1/4 ML 1/4 ML 1/2 ML 1/2 ML 3/4 ML 1 ML

dCu-O 共Å兲 2.01 2.01 1.86 1.96 1.86 1.95 1.89

dO1 共Å兲 0.81 0.82 0.13 0.71 0.18 0.68 0.51
⌬12 共%兲 +1.1 +0.6 +6.0 +5.0 +4.4 +14.3 +18.7
⌬23 共%兲 +0.5 +0.4 +2.2 +1.1 +2.2 +6.0 −1.7
⌬34 共%兲 +0.6 +0.6 +0.6 +0.6 +1.1 +0.6 +1.1
b 共eV兲
EO 1.85 1.89 1.53 1.71 1.81 1.06 0.60
⌬⌽ 共eV兲 0.37 0.74 0.05 0.92 0.32 1.31 1.51
␮ 共Debye兲 0.52 0.52 0.03 0.32 0.11 0.31 0.27

reduced oxygen coverage 共1/4 ML兲 is created using a same coverage 共see Fig. 2兲. For 1/4 ML coverage, the aver-
doubled unit cell, MR-共4冑2 ⫻ 2冑2兲R45°, in which only half age oxygen binding energy of the missing-row 共4冑2
of the pseudohollow sites are occupied by oxygen. Missing- ⫻ 2冑2兲R45° structure is 0.36 eV less stable, suggesting that
row structures with increased coverage 共3/4 and 1 ML兲 are a missing-row reconstruction is not preferred at this low cov-
created in the 共2冑2 ⫻ 冑2兲R45° unit cell by placing additional erage. However, this preference reverses when we consider
oxygen atoms into subsurface tetrahedral or octahedral sites. higher coverages. At 1/2 ML, we find for the MR-共2冑2
The in-plane position of the subsurface sites are labeled as ⫻ 冑2兲R45° structure an oxygen binding energy that is 0.11
shown in Fig. 1共b兲 where sT1 and sT2 are tetrahedral sites, eV more favorable than that of the N-c共2 ⫻ 2兲 structure. For
and sO1 and sO2 are octahedral sites. For 3/4 ML coverage, 3/4 and 1 ML oxygen coverages, the missing-row reconstruc-
two reconstructions were considered: The tetrahedral recon- tions are also more stable than the nonreconstructed Cu共100兲
struction contains oxygen at the sT1 subsurface site, and the surface; however, the average oxygen binding energies are
octahedral reconstruction has one oxygen at the sO1 site. We somewhat smaller relative to the 1/2 ML case, probably due
also construct tetrahedral and octahedral subsurface recon- to the less favorable subsurface oxygen binding sites in-
structions for 1 ML coverage. The tetrahedral reconstruction volved. Overall, our results are consistent with the STM ex-
has two oxygen atoms at two sT2 subsurface sites, the octa- periments of Fujita et al.23 who report a 共2冑2 ⫻ 冑2兲R45°
hedral reconstructions has one oxygen atom each at sO1 and missing-row phase only for higher coverages.
sO2 sites. Iddir et al.33 recently made the convincing argument that
We now compare the energetics of the missing-row recon- the experimentally observed 共2冑2 ⫻ 冑2兲R45° to c共2 ⫻ 2兲
structions to those of the nonreconstructed surface at the transition at 475 K is due to an order-disorder phase transi-
tion involving the Cu vacancies that make up the missing
row. They propose that above 475 K, the vacancies in the Cu
sublattice become disordered, such that the apparent symme-
try in the diffraction is determined by the 1/2 ML oxygen
sublattice with a c共2 ⫻ 2兲 unit cell.
In our calculation, we use the same 1/2 ML disordered
vacancy 共DV兲 structure 关Fig. 3共c兲兴 that Iddir et al.33 employ
to represent this surface 共referred to as a “spaced vacancy
structure” in Ref. 33兲. This structure has a 共2冑2
FIG. 3. 共Color online兲 Top views of the favorable oxygen ad-
⫻ 2冑2兲R45° unit cell and arises by shifting every other Cu
sorption structures on Cu共100兲: 共a兲 1/4 ML oxygen coverage on a atom from a complete row into the missing row, thereby
nonreconstructed surface with a 共2 ⫻ 2兲 unit cell, 共b兲 the missing effectively creating a surface with two half-missing rows.
row 共2冑2 ⫻ 冑2兲R45° reconstruction with 1/2 ML coverage, and 共c兲 The calculated average oxygen binding energy for this struc-
the “disordered vacancy” 共2冑2 ⫻ 2冑2兲R45° structure that becomes ture is 1.78 eV, which is 0.02 eV less than for the missing-
stable at high temperature due to configurational entropy. Large row structure with ordered Cu vacancies. The energy differ-
white and gray spheres represent top layer Cu atoms and underlying ence per Cu vacancy is 0.058 eV, from which we can
substrate Cu atoms, respectively. Black 共red兲 spheres represent O estimate, via the configurational entropy 共see Ref. 33 for de-
atoms. The rectangles 共blue兲 indicate the supercells used in the tails兲, a transition temperature of 300 K between the ordered
calculations. MR-共2冑2 ⫻ 冑2兲R45° and the quasi-disordered 共2冑2

075430-5
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

⫻ 2冑2兲R45° 关i.e., the observed c共2 ⫻ 2兲 pattern兴. Within the

errors of our methodology, this compares reasonably well
with the results of Iddir et al.33 who report an experimental
transition temperature of 475 K. Their DFT-GGA calcula-
tions give an energy difference of 0.13 eV 共per Cu vacancy兲
and a transition temperature of 650 K. While the temperature
variation is considerable due to the small energy differences
involved, it is comforting to see that the experimental tem-
perature lies in between the theoretical results of Iddir et al.33
and the present work.
Surface Relaxations. We now consider the atomic struc-
ture of various surfaces which are listed in Table I. For the
nonreconstructed 共N兲 surface, we see that an increase in the
oxygen coverage from 1/4 to 1 ML results in a decrease in
the vertical position, dO1, of oxygen above the surface plane
from 0.81 to 0.51 Å. Similarly, the Cu–O bond length de-
creases from 2.01 to 1.89 Å. Our interlayer separations are
in excellent agreement with the experimentally observed val- FIG. 4. 共Color online兲 Calculated work function change, ⌬⌽,
ues which range from 0.4 to 0.8 Å depending on and surface dipole moment, ␮, as a function of coverage for O on
coverage.21,24 Below 1/4 ML coverage, we find no appre- Cu共100兲. The solid lines connecting the calculated values are to
ciable change in these properties. Overall, the coverage- guide the eye.
dependent changes in the structural parameters match the
trends seen in the binding energy. These trends are consistent in the work function and surface dipole moment as a func-
with a reduction in the oxygen-oxygen repulsion with re- tion of the oxygen coverage. For hollow-site adsorption and
duced coverage. The generally lower binding energies found the missing-row reconstructions we can see that the work
for high coverages can thus be understood as due to the function change ⌬⌽ increases monotonically with coverage.
increased O-O repulsion. We believe the hollow site prefer- This can be understood in terms of the large surface dipole
ence arises, because O-O repulsion is partly screened by the moment arising due to partial electron transfer from the sub-
short Cu–O bond lengths. At low coverage, the electrostatic strate to the adsorbate; the difference between the electrone-
interaction between O adatoms is less effective, so that the gativity of oxygen 共3.44兲 and copper 共1.90兲 is quite large
distance between the adatoms and the metal surface becomes 共1.54兲. With increasing coverage, a repulsion builds up
larger. among the partially negatively charged O atoms. To reduce
Oxygen adsorption results in significant relaxations in the this repulsion, there will be partial electron transfer back to
first and second atomic layers with a concomitant change in the substrate, giving rise to a decrease in the surface dipole
the calculated interlayer spacing. For the clean surface, ⌬12 is moment, resulting in a depolarization. The missing-row re-
negative 共−2.3%兲 corresponding to a small contraction be- constructed surfaces exhibits similar trend with coverage. We
tween the first and second layers relative to the bulk value. note at this point that the work function change for
With 1 ML of oxygen at the hollow site, ⌬12 becomes O/Cu共100兲 behaves differently to what has been reported for
strongly positive 共+18.7%兲 corresponding to a significant ex- the O/Ag共100兲 system.75 For Ag共100兲, the work function in-
pansion. With decreasing oxygen coverage, the interlayer creases up to 1/2 ML and then decreases due to negative
spacing gradually decreases toward the clean surface value. adsorption heights at high coverage. Negative adsorption
The oxygen-induced expansion can be understood as a heights do not occur in our Cu共100兲 structures. This may be
charge transfer from copper to oxygen atoms resulting a due to the 12% smaller lattice constant of Cu compared to
depletion of the electron density between the first and second Ag.
Cu layer and a weakening of the bonding. This is broadly Further insights into the electronic structure are gained
analogous to what has been reported for oxygen on palla- from the electron density difference distributions and pro-
dium and other metallic surfaces.74,78 jected density of states 共PDOS兲. We show in Fig. 5 the elec-
The obtained structural parameters for the 1/4 and 1/2 ML tron density difference 共relative to the clean surface兲 for a
missing-row reconstructions are listed in Table I. At 1/2 ML plane perpendicular to the surface for the N-p共2 ⫻ 2兲 1/4 ML
coverage, the oxygen atoms are located at a height of 0.18 Å and the MR-共2冑2 ⫻ 冑2兲R45° structures. We see that the per-
above the outmost Cu layer, which is comparable to the turbation caused by O adsorption is mostly localized at the
value of 0.28 Å from other DFT-GGA calculations using the oxygen atom and the topmost layer of Cu atoms. The elec-
pseudopotential method and a plane wave basis.31 The ex- tron density around the oxygen adatom is enhanced and is
perimental value is 0.17⫾ 0.1 Å from a photoelectron dif- depleted around the Cu atoms. The electron transfer from Cu
fraction study,24 and 0.1 Å 共Ref. 22兲 by LEED. The calcu- to the O adatom is reflective of the relative electronegativi-
lated bond lengths to the four coordinating Cu atoms are ties of the two elements.
1.88, 1.86, 1.86 and 2.24 Å, close to the experimental values The projected densities of states presented in Fig. 6 for
1.80, 1.83, 1.83, and 2.12 Å, respectively.22 the N-c共2 ⫻ 2兲 1/2 ML and the MR-共2冑2 ⫻ 冑2兲R45° 1/2 ML
Electronic structure. We now turn to the electronic prop- structures, where the O-2p and Cu-3d orbitals are shown. It
erties of the O/Cu共100兲 system. Figure 4 shows the changes can be seen that there is a hybridization between the O-2p

075430-6
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

FIG. 5. 共Color online兲 Difference electron density distribution of

O/Cu共100兲 structures: 共a兲 The nonreconstructured 共2 ⫻ 2兲 structure
with a 1/4 ML coverage of oxygen at the hollow site. 共b兲 The
missing row 共2冑2 ⫻ 冑2兲R45° reconstruction with a coverage of 1/2
ML. Solid and dashed isolines indicate lines of constant charge
accumulation and depletion, respectively. The lowest positive con-
tour line is at 0.001 electron Bohr−3, while the highest negative
contour line corresponds to a value of −0.001 electron Bohr−3. Suc-
cessive isolines differ by a factor of 101/3.

and Cu-3d in the energy range −7 to −5 eV. The greater

occupation of hybridized O-2p and Cu-3d bonding states FIG. 7. 共Color online兲 Oxygen binding sites on the Cu共110兲
compared to antibonding states, reflects the relatively strong surface showing 共a兲 the nonreconstructed 共n兲 surface, 共b兲 the 关001兴
Cu–O bond. The nature of the states in the PDOS is con- added row 共AR兲 reconstructed surface with a 共2 ⫻ 1兲 unit cell, 共c兲
firmed 共not shown兲 by the single particle wave function at the 关11̄0兴 added row 共AR兲 surface with a 共1 ⫻ 2兲 unit cell, and 共d兲 a
the chosen eigenstates at the ⌫ point. The eigenvalues corre- Cu5Ox added strand 共AS兲 on the nonreconstructed surface. A close
sponding to low energies are due to bonding states, while packing of Cu5O4 added strands gives rise to the 2/3 ML c共6 ⫻ 2兲
those close to the Fermi level are due to antibonding states. reconstruction at high oxygen exposures. Top-layer Cu atoms are
Furthermore, surface Cu-3d states are narrowed compared to colored white and second layer Cu atoms are colored gray. The
subsurface Cu-3d states due to reduced coordination at the in-plane position of surface and subsurface binding sites are indi-
surface. For the MR-共2冑2 ⫻ 冑2兲R45° 1/2 ML structure, the cated by red squares and labeled below the diagram.
bonding states of O-2p extend to 8 eV below the Fermi level,
and the occupancy of the antibonding states is lower than C. Oxygen adsorption on Cu(110)
that of the N-c共2 ⫻ 2兲 1/2 ML system, which indicates the Cu sublattices and O binding sites. Being characterized
stronger interaction between O and surface Cu atoms in the by a rectangular surface unit cell, the Cu共110兲 surface offers,
reconstructed system. relative to the square unit cell Cu共100兲, a larger number of
possibilities for oxygen adsorption. On the nonreconstructed
surface 关Fig. 7共a兲兴, we have five distinct sites. The hollow
site 共H兲 is located above the center of four surface Cu atoms.
The long-bridge 共LB兲 and the short-bridge 共SB兲 sites are
located between pairs of surface Cu atoms along the 关001兴
and 关11̄0兴 directions, respectively. The shifted-hollow 共shH兲
site is pseudo threefold coordinated and located roughly half-
way between hollow and short-bridge sites. Oxygen adsorp-
tion at the top site 共T兲 directly above a surface Cu atom has
previously been shown to be unstable47 and is not further
explored here. We also consider two types of added row
共AR兲 reconstructions 关Figs. 7共b兲 and 7共c兲兴 in which rows of
Cu atoms are added along the 关001兴 and the 关11̄0兴 directions.
We consider these rows at different separations using
共2 ⫻ 1兲, 共4 ⫻ 1兲, and 共8 ⫻ 1兲 unit cells for rows in the 关001兴
direction, and a 共1 ⫻ 2兲 unit cell for the 关11̄0兴 direction.
FIG. 6. 共Color online兲 Projected density of states 共PDOS兲 for the
selected O/Cu共100兲 structures with 1/2 ML oxygen coverage: 共a兲
Added row structures with a 共1 ⫻ 1兲 periodicity are equiva-
The nonreconstructed c共2 ⫻ 2兲 and 共b兲 the missing row 共2冑2 lent to a nonreconstructed surface. For the 关001兴 added row
⫻ 冑2兲R45° structure. Energies are given relative to the Fermi level structures, we consider oxygen binding at the long-bridge
EF. The bulklike Cu-3d PDOS of atoms in the center of the slab is site of the added row as illustrated in Fig. 7共b兲. For the 关11̄0兴
indicated using a dotted line. added row, we investigate oxygen binding at short bridge and

075430-7
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

FIG. 9. 共Color online兲 Top views of the favorable oxygen ad-

sorption structures on Cu共110兲. Shown are 共a兲 the 1/4 ML oxygen
coverage added row 共4 ⫻ 1兲 structure characterized by Cu-O-Cu-
O-Cu chains in the 关100兴 direction, 共b兲 the added row 共2 ⫻ 1兲 struc-
ture with 1/2 ML coverage with a closer spacing of the CuO chains,
and 共c兲 the c共6 ⫻ 2兲 structure with 2/3 ML oxygen composed of
Cu5O4 added strands in the 关100兴 direction. Large white and gray
circles represent top and second layer Cu atoms, respectively. Small
dark 共red兲 circles represent O atoms. The rectangles indicate the
surface unit cells used in the calculations. In the c共6 ⫻ 2兲 structure,
the nonequivalent Cu atom sites Cu1 and Cu2 are indicated 共see
Table III兲.
FIG. 8. 共Color online兲 Calculated average binding energy of
oxygen on Cu共110兲 for various coverages and configurations. The
dashed line connects the binding energies of the most stable struc-
共Fig. 8兲 show that added row reconstructions are preferred
tures and serves to guide the eye. over most of the oxygen coverage range. The most stable
reconstructions are characterized by 关001兴 added rows 共blue
shifted hollow sites 关see Fig. 7共c兲兴. Finally, we consider re- data points in Fig. 8兲 in which oxygen atoms occupy all of
constructions that we describe as added strand structures the long-bridge sites. This creates continuous Cu-O-Cu-
共AS兲 in which Cu adatoms are arranged into strands as illus- O-Cu adatom chains that appear to be particularly stable
trated in Fig. 7共d兲. On this surface oxygen binds to the two which is evident by the fact that a reduced oxygen coverage
long-bridge sites and also to two threefold sites, forming a does not disrupt these chains. Instead the reduced availability
Cu5O4-stoichiometric added strand that is the basic building of oxygen atoms is accommodated by an increased average
block of an experimentally observed c共6 ⫻ 2兲 reconstruction. separation between the Cu-O rows, as illustrated in Figs. 9共a兲
We will examine several variations in this pattern that differ and 9共b兲 for the 1/4 and 1/2 ML structures, respectively.
in the relative alignment of nearby strands or the number of For the 1/2 ML 关001兴 added row structure, our calculated
Cu atoms within the strand. In our structure notation, we use oxygen binding energy is 2.00 eV, which is in good agree-
prefixes N-, AR-, AS- to indicate the Cu-sublattice as nonre- ment with the previously reported values of 2.11 and 2.03 eV
constructed, added row, or added strand, respectively. An by Liem et al.47 and Frechard et al.,46 respectively. Oxygen
overview of our calculated average oxygen binding energies adsorption at the short-bridge site of a 关11̄0兴 added row
for these sublattices and a range of oxygen coverages be- structure 关Fig. 7共c兲兴 is considerably less stable, with a calcu-
tween 1/8 and 1 ML is given in Fig. 8. lated binding energy of only 1.29 eV. The shorter bridge
Oxygen on the nonreconstructed surface. For low cover- causes the oxygen adatoms to be displaced higher above the
age 共1/8 ML兲 on the nonreconstructed surface oxygen is adatom row, which presumably results in less effective
preferentially bonded to the shifted-hollow site 共labeled screening and increased repulsion between oxygen atoms.
N-shH in Fig. 8兲, followed in order of decreasing stability by For the stable 关001兴 AR structures, the average oxygen
the hollow, the long-bridge, and the short-bridge sites. At 1/8 binding energies do not change much when the coverage
ML, our calculated binding energy of the shifted hollow site decreases below 1/2 ML; for example, 1/4 ML and 1/8 ML
is 1.62 eV, which is in good agreement with the earlier result coverages have binding energies of 2.04 and 2.07 eV, to be
of 1.59 eV reported by Liem et al.47 At a coverage of 1/2 compared with 2.00 eV for 1/2 ML. This indicates that the
ML, the shifted-hollow site is still the preferred structure, Cu-O rows are only very weakly repulsive when separated
however the long-bridge site is now the second most stable by more than one lattice unit; however, above 1/2 ML the
structure. This order is reversed above ⬇0.6 ML coverage, binding energies decrease rapidly to 1.71 eV for 2/3 ML and
and the long-bridge site becomes the preferred binding side. 1.27 eV for 1 ML. In this coverage regime, the Cu-O adatom
The general trend on the nonreconstructed surface is that at a rows are separated by less than two lattice units on average,
coverage of 1/4 ML and above, the binding energies begin to meaning that a larger number of rows are brought into single
rapidly decrease, which again reflects the increased repulsion lattice unit separation; the energetics suggests that this an
between an increasing number of oxygen atoms on the unfavorable arrangement. For a full 1 ML coverage, the sur-
surface.28,76 The long-bridge site is an exception to this trend, face is so densely packed with Cu-O adatom rows that the
showing the least variation in the binding energy with oxy- Cu-sublattice is in fact identical to the nonreconstructed sur-
gen coverage. face 关Fig. 7共a兲兴 in which oxygen atoms occupy all of the long
Added row reconstructions. The average binding energies bridge sites.

075430-8
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

It is worthwhile to briefly reflect on the energetic proxim-

ity of the four 2/3 ML structures 关Figs. 10共a兲–10共d兲兴 consid-
ered here and the significance of these results in relation to
experiment. Within the computational errors of our proce-
dure 共⬇0.01 eV in binding energies兲, the AS-c共6 ⫻ 2兲 and
the AR-共3 ⫻ 1兲 are very nearly of the same energy. This is to
say that the presence or absence of the 1/6 ML bridging Cu
atoms has only a very subtle effect on the stability of the
surface. A further increase in the density of bridging Cu at-
oms to 1/3 ML, which leads to the AR-共3 ⫻ 1兲 structure,
produces a significant destabilization of 0.15 eV in the aver-
age oxygen binding energy. Also significant within the mar-
gins of error is the 0.05 eV difference in the oxygen binding
energy between the two added-strand structures c共6 ⫻ 2兲 and
共3 ⫻ 2兲. These last two findings, suggest that the positioning
of bridging Cu atoms is correlated both within and between
adjacent strands. This in turn will prompt any bridging Cu
atoms present to arrange into some local order. While the
amount of bridging Cu atoms between 0 and 1/6 ML is not
significantly explained by our energetics, we may speculate
FIG. 10. 共Color online兲 Variations in added strand reconstruc- that configurational entropy would favor structures that con-
tions on Cu共110兲. 共a兲 The experimental c共6 ⫻ 2兲 reconstruction tain bridging Cu atoms over those that do not. This would
which is composed of Cu5O4 added strands, 共b兲 a 共3 ⫻ 2兲 variant in not be dissimilar to the mechanism that stabilizes the 1/2 ML
which the bridging Cu atoms are not shifted between adjacent disordered vacancy structure on the 共100兲 surface.33
strands, 共c兲 a 共3 ⫻ 1兲 variant of Cu3O2 stoichiometry, containing a Lastly, we also tested the stability of an isolated added
full row of bridging Cu atoms, and 共d兲, for comparison, a 2/3 ML strand as shown in Fig. 10共e兲 with an oxygen coverage 1/3
CuO added row structure. Panel 共e兲 shows a 1/3 ML added strand ML. The calculated average oxygen binding energy is 1.94
structure formed by removing every other strand from the c共6 eV. This is less stable than what we expect 共by linear inter-
⫻ 2兲 structure. Large white and gray circles represent top and sec- polation兲 for a 1/3 ML added-row structure 共EO b = 2.03 eV兲.
ond layer Cu atoms, respectively. Small dark 共red兲 circles represent
Thus, it appears that the close proximity of 共and repulsion
O atoms. The rectangles indicate the surface unit cells used. Note
between兲 the Cu-O adatom rows is an important ingredient in
that the unit cell for the c共6 ⫻ 2兲 structure is shifted away from
making the added strand structures competitive with the
centered to facilitate comparison with the other unit cells.
added row structures.
The added strand reconstructions. Experimental Structural Relaxations. Calculated interplanar separations
evidence52 points to the existence of a c共6 ⫻ 2兲 reconstruc- for several relevant surface structures are listed in Table II.
tion at oxygen coverages above 2/3 ML. This reconstruction The internal coordinates of the two oxygen and two surface
was attributed by Feidenhans’l et al.50 to the 2/3 ML oxy- Cu atoms as indicated in Fig. 9共c兲 are listed in Table III. Our
genated added strand structure shown in Fig. 10共a兲. In our calculated results are in good agreement with the experimen-
calculations, we find that this reconstruction has a binding tal values from low energy ion scattering 共LEIS兲
energy of 1.87 eV, which is marginally more stable than the measurements,52 improving over earlier effective-medium
2/3 ML added row structure 关Fig. 10共d兲, EO b = 1.86 eV兴. The theory 共EMT兲 calculations.38
c共6 ⫻ 2兲 reconstruction is a Cu5O4 added strand structure that Figure 9共b兲 depicts the AR-共2 ⫻ 1兲 structure in which the
differs from the 关001兴 added row structures by having addi- O atoms occupy the long-bridge sites of every other 关001兴
tional Cu atoms placed between pairs of Cu-O added rows. copper row, while the other rows are missing. In this struc-
These Cu atoms cross link between every other pair of oxy- ture, the O atoms are bonded to two surface Cu atoms and
gen atoms within the two rows, thereby creating a strand two subsurface Cu atoms, where the Cu–O bond lengths are
structure that is oriented along the 关001兴 direction and has a 1.83 and 2.05 Å, respectively. These values are in good
width of three lattice units in the 关11̄0兴 direction. In the agreement with the various experimental results which are in
c共6 ⫻ 2兲 reconstruction, these strands are closely packed, and the range of 1.81–1.84 and 2.00– 2.05 Å 共Refs. 22, 43, 50,
the linking Cu atoms are shifted by one 关001兴 lattice unit and 52兲 and with the findings of the first-principles calcula-
from one strand to the next. An alternative 共3 ⫻ 2兲 added tions of Liem et al.47 共1.83 and 2.08 Å兲 and of Frechard et
strand structure 关Fig. 10共b兲兴 is slightly less stable with an al.46 共1.83 and 2.05 Å兲. The vertical height of O, dO1, is
average oxygen binding energy of 1.82 eV. This structure is calculated to be 0.13 Å, which is larger than the LEED val-
similar to the c共6 ⫻ 2兲 reconstruction but the linking Cu at- ues ⬃0.03 Å,22,43 but comparable to the value of 0.10 Å
oms are not shifted from one strand to the next. In a second obtained by Frechard et al.46 where only O atoms and the
variant, shown in Fig. 10共c兲, we tested the stability of a 共3 first Cu layer were optimized.
⫻ 1兲 added strand reconstruction in which all oxygen atom Electronic Structure. To elucidate the binding properties
pairs are cross-linked by Cu atoms. With an average oxygen of oxygen on the Cu共110兲 surface we list the change in the
binding energy of 1.71 eV, this structure is also less stable work function and surface dipole moment for the most favor-
than the c共6 ⫻ 2兲. able oxygen structures in Table II. These values are also

075430-9
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

TABLE II. Calculated structural and electronic parameters for

different O coverages 共from 0.125 to 0.5 ML兲 for the missing-row
reconstructed Cu共110兲 surfaces. The binding energy, EO b is calcu-
lated with respect to the isolated oxygen molecule. dCu-O indicates
the average bond length between oxygen and the nearest copper
atom. dO1 is defined to be the minimum vertical height of O with
respect to the surface Cu layer, ⌬ij 共i = 1 , 3 , j = 2 , 4兲 represents the
percent change in the ith and jth metal interlayer distances, where
the center-of-mass of the layer is used, relative to the bulk interlayer
spacing. ⌬⌽ and ␮ represent the work function change and surface
dipole moment, respectively.

AR-共8 ⫻ 1兲 AR-共4 ⫻ 1兲 AR-共2 ⫻ 1兲 AS-c共6 ⫻ 2兲

Structure 1/8 ML 1/4 ML 1/2 ML 2/3 ML

dCu-O 共Å兲 1.98 1.97 1.97 2.09

dO1 共Å兲 0.13 0.13 0.13 0.05
⌬12 共%兲 +14.5 +12.5 +10.8 +23.1 FIG. 11. 共Color online兲 Calculated work function change, ⌬⌽,
⌬23 共%兲 −10.0 −6.4 −1.5 −3.9 and surface dipole moment, ␮, as a function of coverage for O on
⌬34 共%兲 +2.7 +0.5 −3.1 −1.5 the Cu共110兲 surface for the various structures considered. Solid
lines connect the low-energy added row 共AR兲 and added strand
b 共eV兲
EO 2.07 2.04 2.00 1.87
共AS兲 structures. Filled and open circles indicate the work functions
⌬⌽ 共eV兲 0.20 0.33 0.56 0.54 of other nonreconstructed and reconstructed phases, respectively.
␮ 共Debye兲 0.40 0.33 0.27 0.20
Figure 13 shows the PDOS of the O-2p, Cu-3d orbitals
plotted in Fig. 11. It can be seen that as more effective for the 1/2 ML oxygenated AR-共2 ⫻ 1兲 and the 2/3 ML
charges 共i.e., oxygen atoms兲 are added to the surface, ⌬⌽ AS-c共6 ⫻ 2兲 phases. In both cases, a renormalization of the
increases with coverage from 1/8 to 1/2 ML, while the sur- O-2p states to lower energies is found. The hybridization of
face dipole moment decreases with coverage. bonding states between the O-2p and Cu-3d orbitals is lo-
We plot the difference electron density for the two stable cated mainly in the energy window from −4 to −8 eV. The
structures, AR-共2 ⫻ 1兲 and AS-c共6 ⫻ 2兲, in Fig. 12. It can be corresponding antibonding states are located at −2 to 2 eV.
seen that the oxygen atoms appear to be almost coplanar Furthermore, the surface Cu-3d states in Figs. 13 are nar-
with adjacent Cu atoms. Due to the large electronegativity of rower than those of the bulk Cu-3d states, and this is again
oxygen, the electron density of the oxygen atoms is en- due to reduced coordination of surface Cu atoms.
hanced, while the electron density of the nearest-neighbor Cu IV. DISCUSSION
atoms is depleted. The slight accumulation of electron den-
sity toward the vacuum side of the surface results in an in- Having surveyed oxygen binding at varying coverages on
ward pointing surface dipole moment and a positive ⌬⌽. Cu共100兲 and Cu共110兲, we are now in a position to assess the

TABLE III. Atomic positions in Å of nonequivalent atoms of

the added-strand c共6 ⫻ 2兲 2/3 ML reconstruction. The results of this
work are compared to the positions obtained by low-energy ion-
scattering 共LEIS兲 共Ref. 52兲 and effective-medium theory 共EMT兲
共Ref. 32兲. In-plane coordinates are given relative to the Cu1 site
关see Fig. 9共c兲兴. Coordinates not listed in the table are zero due to
symmetry. The results of the present work use a rectangular unit
cell of a = 7.2856 Å and c = 15.455 Å.

Atom This work EMTa LEISb

Cu1 z 0.55 1.2 0.45⫾ .1

Cu2 x 2.29 2.38 2.28
FIG. 12. 共Color online兲 Difference electron density distribution
y 1.78 1.805 1.78⫾ 0.07 of O/Cu共110兲 structures: Shown are 共a兲 the added row 共2 ⫻ 1兲 struc-
O1 x 1.85 1.36 1.78 ture with a 1/2 ML oxygen coverage and 共b兲 the added strand re-
z 0.50 0.4 0.4⫾ 0.1 constructed c共6 ⫻ 2兲 structure with a 2/3 ML coverage. Dashed
O2 x 5.21 5.094 5.2⫾ 0.1 isodensity lines represent charge depletion relative to the clean sur-
face and the solid isodensity lines depict charge accumulation. The
z −0.06 −0.2 −0.12⫾ 0.1
lowest positive contour line is at 0.001 electron Bohr−3, while the
aReference 70. highest negative contour line corresponds to a value of −0.001 elec-
bReference 57. tron Bohr−3. Successive isolines differ by a factor of 101/3.

075430-10
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

FIG. 13. 共Color online兲 Projected density of states 共PDOS兲 for

selected O/Cu共110兲 structures showing 共a兲 the added row 共2 ⫻ 1兲
structure with 1/2 ML oxygen coverage and 共b兲 the added strand
c共6 ⫻ 2兲 structure with 2/3 ML coverage. Energies are given relative
to the Fermi level EF. The bulklike Cu-3d PDOS of atoms in the
center of the slab is indicated using a dotted line.

relative stability of these structures when they are in thermo-

dynamic contact with a real gas phase environment. Within
the framework of ab initio atomistic thermodynamics a sur-
face is assumed in thermodynamic equilibrium with both the
gas phase above the surface and the bulk elemental solid
共Cu兲 below. The gas phase is represented by the oxygen
chemical potential ⌬␮O共p , T兲 关given here relative to molecu-
FIG. 14. 共Color online兲 Calculated surface phase stability dia-
lar O2 as per Eq. 共4兲兴, which in turn is given by the gas phase
gram for 共a兲 the O/Cu共100兲 system and 共b兲 the O/Cu共110兲 system as
oxygen partial pressure p and the temperature T. Under these a function of the oxygen chemical potential. A thick 共blue兲 line
conditions, the preferred surface for a given p and T is the highlights the minimum free energy curve for each surface. Vertical
one of lowest surface free energy ␥ and it is this surface that dashed lines indicate the phase transition points on the ⌬␮O axis. A
we should expect to observe experimentally at these condi- shaded background on the left and right hand side denotes the re-
tions. gions of stability of the clean surface and the bulk oxide 共Cu2O兲,
This type of analysis leads to the phase stability diagrams respectively. In between 共without shading兲 are the surface oxide
shown in Fig. 14 in which the free energies of the considered phases. Note that the free surface energy is reported as ⌬␥ relative
surface structures are plotted against the oxygen chemical to the respective clean 共oxygen-free兲 surface.
potential ⌬␮O. In these diagrams, free energies are reported
as differences ⌬␥ relative to the clean surface as defined in oxygenated surface structures are found to be thermodynami-
Eq. 共3兲. Plotted in this way, individual surface structures ap- cally stable. We note in passing that at some oxygen potential
pear as straight lines with a negative slope 共−NO / 2A兲 that is above −1.24 eV a phase transition between bulk Cu2O and
proportional to the oxygen coverage. Critical for this discus- bulk CuO is expected to occur. This transition was not fur-
sion are the curves of minimum free energy 共colored red in ther explored in this work as it occurs outside the ⌬␮O limits
the online version兲 that form when the free energy lines of relevant to surface oxide formation.
the most stable surfaces intersect one another. These points Starting with the clean Cu共100兲 surface 关see Fig. 14共a兲兴,
of intersection on the ⌬␮O axis define the conditions of a an increase in ⌬␮O leads at −1.89 eV to a phase transition to
phase transition from one minimum free energy surface to a N-共2 ⫻ 2兲 structure with an oxygen coverage of 1/4 ML.
another. In this way, the ⌬␮O scales for Cu共100兲 and Cu共110兲 This oxygenated surface corresponds to the nonrecon-
are divided into several distinct segments in which different structed, hollow-site adsorption structure that we found in
clean and oxygenated surface structures are preferred. These Fig. 2 to be preferred for a 1/4 ML coverage. A further in-
segments are indicated in Fig. 14 and are labeled by the type crease in the oxygen exposure results at ⌬␮O = −1.72 eV in
of the surface and the oxygen coverage in ML. An upper another phase transition from the N-共2 ⫻ 2兲 1/4 ML structure
limit for ⌬␮O exists at −1.24 eV. This corresponds to the to the missing row 共2冑2 ⫻ 冑2兲R45° reconstruction with a 1/2
calculated heat of formation of bulk Cu2O which defines the ML coverage. This surface phase remains the preferred re-
point at which a bulk phase transition from Cu to Cu2O will construction until the bulk-oxide limit at ⌬␮O = −1.24 eV is
occur. In the limit of low oxygen exposure 共i.e., large nega- reached. None of the higher coverage structures considered
tive ⌬␮O兲, the clean Cu surface is preferred for both 共100兲 by us are thermodynamically stable anywhere on the ⌬␮O
and 共110兲 crystal faces. In between these two limits, several scale.

075430-11
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

sition is well documented experimentally, there does not ap-

pear to be any evidence for a 1/4 ML structure with 共2 ⫻ 2兲
symmetry 共see e.g., the phase diagram Fig. 1 in Ref. 33兲.
This is not necessarily a conflict. At coverages of around 1/4
ML 共and below兲 our calculated oxygen binding energies on
the nonreconstructed Cu共100兲 surface are very nearly con-
stant with respect to changes in coverage 共see Fig. 2兲. This
implies that the oxygen adatoms in the N-共2 ⫻ 2兲 phase are
weakly interacting, and plausibly, disordered distribution on
the surface. Thus, the N-共2 ⫻ 2兲 1/4 ML structure in our
phase diagram acts as a “placeholder” for the low-coverage
disordered arrangement of oxygen atoms on the nonrecon-
structed surface, which will appear as a 共1 ⫻ 1兲 structure in
experiment. Only when the oxygen exposure increases, and
the surface progresses toward a 1/2 ML coverage, do the
oxygen binding energies increase. This causes oxygen atoms
to lock into an ordered c共2 ⫻ 2兲 1/2 ML pattern. Thermody-
namically, this structure supports the formation of Cu vacan-
cies, which in turn may order into the MR-共2冑2 ⫻ 冑2兲R45°
1/2 ML phase, or remain disordered in the DV-c共2 ⫻ 2兲 1/2
ML phase.
Looking now at the Cu共110兲 surface, the phase stability
diagram in Fig. 14共b兲 shows at ⌬␮O = −2.07 eV a transition
of the oxygen-free surface into the added row 共8 ⫻ 1兲 struc-
ture with a 1/8 ML oxygen coverage. At −1.98 eV, the
added row 共2 ⫻ 1兲 structure with 1/2 ML oxygen becomes
favorable. This structure corresponds to a denser packing of
the oxygenated added rows. Interestingly, the 1/4 ML added
row structure is not stable as an intermediate phase between
the 1/8 and 1/2 ML added rows, which suggests that there
will be a rather abrupt transition between a low coverage of
oxygen and the AR-共2 ⫻ 1兲 reconstruction with 1/2 ML cov-
erage. This could, for instance, become apparent in experi-
ment in the form of AR-共2 ⫻ 1兲 island formation. As the
oxygen exposure further increases to ⌬␮O = −1.46 eV, the
thermodynamic preference changes from the AR-共2 ⫻ 1兲 to
FIG. 15. 共Color online兲 Stability of oxygen-free, surface-oxide, the added strand c共6 ⫻ 2兲 structure with a 2/3 ML oxygen
and bulk-oxide phases in a 共p , T兲 phase diagram for 共a兲 the coverage. The AS-c共6 ⫻ 2兲 reconstruction is the stable sur-
O/Cu共100兲 and 共b兲 the O/Cu共110兲 system, based on the ⌬␮O phase face phase up to the ⌬␮O = −1.24 eV bulk-oxide limit. While
boundaries determined in Fig. 14. The vertical dotted line in panel higher coverage 共1 ML兲 structures were considered by us
共a兲 indicates the experimental 共Ref. 33兲 phase boundary at 475 K 共see Fig. 8兲, these do not appear as stable phases in the sta-
above which the Cu vacancies of the MR-共2冑2 ⫻ 冑2兲R45° structure bility diagram due to the much smaller oxygen binding en-
become disordered, giving rise to a disordered vacancy structure ergies at these coverages. This results in a predicted maxi-
that appears as a c共2 ⫻ 2兲 reconstruction. mum oxygen coverage of 2/3 ML for the Cu共110兲 surface.
In the 共p , T兲 phase diagram 关Fig. 15共a兲兴 the two ⌬␮O seg- Figure 15 translates these ⌬␮O boundaries into a 共p , T兲 phase
ments corresponding to the stable surface oxides 共white diagram.
background兲 form a band of roughly 10 to 20 orders of mag- In Fig. 16, the minimum free energy curves of the 共100兲
nitude width on the pressure scale and are bounded on the and 共110兲 crystal faces are combined with the equivalent data
high and low pressure side by bulk Cu2O and the oxygen- for the 共111兲 face 共taken from Ref. 57兲 to predict the oxygen-
free surface, respectively. In the region corresponding to the dependent morphology of copper nanoparticles using Wulff
MR-共2冑2 ⫻ 冑2兲R45° 1/2 ML, we have additionally indicated construction.79 This procedure determines the nanoparticle
the 473 K phase boundary33 to the 1/2 ML disordered va- shape by balancing within geometrical constraints the rela-
cancy structure that is experimentally observed as a c共2 tive fractions of the three contributing crystal faces so as to
⫻ 2兲 pattern. This results in a diagram with a total of four minimize the particles overall surface free energy. These cal-
distinct surface phases: the clean Cu共100兲 surface at low culations are performed using the program
oxygen pressure, a nonreconstructed 共2 ⫻ 2兲 structure with WINXMORPH.80,81
1/4 ML oxygen atoms at the hollow site at increased pres- The predicted nanoparticle shapes at four different oxygen
sures, and then either the ordered missing row 共for T chemical potentials are shown in Figs. 16共b兲–16共e兲. Under
⬍ 473 K兲 or the disordered vacancy c共2 ⫻ 2兲 structure 共for highly reducing conditions at ⌬␮O = −2.1 eV, all three crys-
T ⬎ 473 K兲. While the MR-共2冑2 ⫻ 冑2兲R45° to c共2 ⫻ 2兲 tran- tal faces are free of oxygen, and therefore the shape is deter-

075430-12
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

the corresponding particle shape shown in Fig. 16共d兲兴.

Lastly, at ⌬␮O = −1.3 eV, just prior to bulk oxidation, the
free energies of the 共111兲 and 共100兲 surfaces are very nearly
degenerate, whereas the 共110兲 is slightly set apart. This re-
sults in a slight reduction in the 共110兲 surface area in pre-
dicted nanoparticle shape 关Fig. 16共e兲兴 relative to the ⌬␮O =
−1.5 eV particle in 关Fig. 16共d兲兴.
Overall, the particle shape in the course of oxidation tran-
sitions from highly facetted configuration 关Fig. 16共b兲兴 to an
almost spherical shape 关Fig. 16共e兲兴. This observation can be
attributed to the fact that the calculated surface free energies
in the absence of oxygen are much more anisotropic than
those at the oxygen-rich limit as shown in Fig. 16共a兲兴. For
the latter case, the free energy curves are almost degenerate
and, moreover, very nearly parallel. The parallel slope indi-
cates that all three faces, despite differing monolayer cover-
ages, share very similar surface densities of oxygen atoms
共between 0.071 and 0.076 atoms/ Å2 using our GGA unit
cell dimensions兲. The near-degeneracy in the energy in turn
suggests that surface free energies at this point are more
determined by the mutual proximity of oxygen atoms than by
the orientation of the surface.

V. SUMMARY AND CONCLUSIONS

In summary, we have investigated the chemisorption of O

on the Cu共100兲 and Cu共110兲 surfaces using density-
functional theory for a wide range of atomic configurations
and coverages, including oxygen adsorbed on ideal and re-
constructed surfaces. For the Cu共100兲 system, oxygen ad-
sorption prefers the hollow site at low oxygen exposures. At
higher exposures, the 共2冑2 ⫻ 冑2兲R45° missing-row recon-
structed structure becomes energetically favored. For the
O/Cu共110兲 system, the 关001兴 Cu-O added row reconstruc-
tions are favored at low oxygen exposures, with the AR-共2
FIG. 16. 共Color online兲 Copper nanoparticle morphology using ⫻ 1兲 reconstruction prominent under these conditions. At
the Wulff construction. 共a兲 Comparison of the calculated absolute higher oxygen exposures, a transition from the AR-共2 ⫻ 1兲
free surface energies ␥ of 共100兲, 共110兲, and 共111兲 crystal faces as a structure to the c共6 ⫻ 2兲 added Cu5O4-strand structure is pre-
function of the oxygen chemical potential. 共b–d兲 Predicted nanopar- dicted to occur. On both crystal faces considered, oxygenated
ticle shapes for selected values of ⌬␮O. surface structures are stable under conditions prior to onset
of bulk oxide Cu2O formation. Overall, the computational
mined by the free energies of the clean Cu surfaces. With
results are broadly consistent with experimental observation,
Cu共111兲 being considerably more stable than Cu共110兲 and
Cu共100兲, the nanoparticle shape 关Fig. 16共b兲兴 maximizes the which highlights the ability of DFT free energy calculations
exposure of the 共111兲 face; only geometric constraints cause to describe the thermodynamics of surface reconstruction for
some of the next-favored 共100兲 face to be formed. Figure this technologically important material. The combined
16共c兲 shows the nanoparticle morphology at the higher oxygen-dependent free energies of the 共100兲, 共110兲, and
chemical potential of ⌬␮O = −1.8 eV at which both the 共100兲 共111兲 surfaces are used to predict the shape of copper nano-
and 共110兲 faces are oxidized, while the 共111兲 remains oxygen particles in contact with an oxygen environment.
free. Oxidation reduces the free energies of the 共100兲 and
共110兲 surfaces, and at ⌬␮O = −1.8 eV, the 共110兲 face is now ACKNOWLEDGMENTS
slightly more stable than the 共100兲. This change in order
manifests itself in the appearance of 共110兲 edges in the nano- This research is supported by the Australian Research
particle shape 关Fig. 16共c兲兴. At ⌬␮O = −1.5 eV, all three sur- Council 共Grant No. DP0770631兲, the Australian Partnership
faces are oxidized and their free energies are very similar. As for Advanced Computing 共APAC兲, and the Australian Centre
a result, all three crystal faces are prominently expressed in for Advanced Computing and Communications 共ac3兲.

075430-13
DUAN et al. PHYSICAL REVIEW B 81, 075430 共2010兲

*Present address: Department of Physics, Ningbo University, Foster, T. Kangas, and K. Laasonen, Catal. Today 100, 403
Ningbo 315211, China. 共2005兲.
† 31
o.warschkow@physics.usyd.edu.au N. Bonini, A. Kokalj, A. Dal Corso, S. de Gironcoli, and S.
1 A. Guerrero-Ruiz, L. Rodriguez-Ramos, and J. L. G. Fierro, Baroni, Surf. Sci. 600, 5074 共2006兲.
Appl. Catal. 72, 119 共1991兲. 32 M. J. Harrison, D. P. Woodruff, J. Robinson, D. Sander, W. Pan,
2
G. Centi and S. Perathoner, Appl. Catal. 132, 179 共1995兲. and J. Kirschner, Phys. Rev. B 74, 165402 共2006兲.
3 33
P. J. Godowski, J. Onsgaard, A. Gagor, M. Kondys, and Z. S. Li, H. Iddir, D. D. Fong, P. Zapol, P. H. Fuoss, L. A. Curtiss, G.-W.
Chem. Phys. Lett. 406, 441 共2005兲. Zhou, and J. A. Eastman, Phys. Rev. B 76, 241404共R兲 共2007兲.
4 M. Brandhorst, J. Zajac, D. J. Jones, J. Roziére, M. Womes, A. 34 S. Jaatinen, M. Rusanen, and P. Salo, Surf. Sci. 601, 1813

Jimenez-Lòpez, and E. Rodriguez-Castellón, Appl. Catal., B 55, 共2007兲.

267 共2005兲. 35 T. Kangas and L. Laasonen, Surf. Sci. 602, 3239 共2008兲.
5 36
I. Louis-Rose, C. Méthivier, and C.-M. Pradier, Catal. Today 85, K. Lahtonen, M. Hirsimäki, M. Lampimäki, and M. Valden, J.
267 共2003兲. Chem. Phys. 129, 124703 共2008兲.
6 R. W. Mayer, M. Hävecker, A. Knop-Gericke, and R. Schlögl, 37 G. Ertl, Surf. Sci. 6, 208 共1967兲.

Catal. Lett. 74, 115 共2001兲. 38 R. A. DiDio, D. M. Zehner, and E. W. Plummer, J. Vac. Sci.
7
C. Fukuhara, H. Ohkura, Y. Kamata, Y. Murakami, and A. Iga- Technol. A 2, 852 共1984兲.
rashi, Appl. Catal., A 273, 125 共2004兲. 39 F. M. Chua, Y. Kuk, and P. J. Silverman, Phys. Rev. Lett. 63,
8
A. Atkinson, S. Barnett, R. J. Gorte, J. T. S. Irvine, A. J. McE- 386 共1989兲.
40
voy, M. Mogensen, S. C. Singhal, and J. Vohs, Nature Mater. 3, F. Jensen, F. Besenbacher, E. Læsgaard, and I. Stensgaard, Phys.
17 共2004兲. Rev. B 41, 10233 共1990兲.
9 F. Luck, Catal. Today 27, 195 共1996兲. 41 D. J. Coulman, J. Wintterlin, R. J. Behm, and G. Ertl, Phys. Rev.
10 C. Stampfl, M. V. Ganduglia-Pirovano, K. Reuter, and M. Schef-
Lett. 64, 1761 共1990兲.
fler, Surf. Sci. 500, 368 共2002兲. 42 R. Feidenhans’l, F. Grey, R. L. Johnson, S. G. J. Mochrie, J.
11 C. Stampfl, Catal. Today 105, 17 共2005兲.
Bohr, and M. Nielsen, Phys. Rev. B 41, 5420 共1990兲.
12
M. Wuttig, R. Franchy, and H. Ibach, Surf. Sci. 213, 103 共1989兲. 43
S. R. Parkin, H. C. Zeng, M. Y. Zhou, and K. A. R. Mitchell,
13 H. C. Zeng, R. A. MacFarlane, and K. A. R. Mitchell, Surf. Sci.
Phys. Rev. B 41, 5432 共1990兲.
208, L7 共1989兲. 44
L. Ruan, F. Besenbacher, I. Stensgaard, and E. Laegsgaard, Phys.
14 M. Wuttig, R. Franchy, and H. Ibach, Surf. Sci. 224, L979
Rev. Lett. 70, 4079 共1993兲.
共1989兲. 45
J. Buisset, H.-P. Rust, E. K. Schweizer, L. Cramer, and A. M.
15 K. W. Jacobsen and J. K. Nørskov, Phys. Rev. Lett. 65, 1788
Bradshaw, Surf. Sci. 349, L147 共1996兲.
共1990兲. 46 F. Frechard and R. A. van Santen, Surf. Sci. 407, 200 共1998兲.
16 47
F. Jensen, F. Besenbacher, E. Lægsgaard, and I. Stensgaard, Y. S. Liem, G. Kresse, and J. H. R. Clarke, Surf. Sci. 415, 194
Phys. Rev. B 42, 9206 共1990兲. 共1998兲.
17 I. K. Robinson, E. Vlieg, and S. Ferrer, Phys. Rev. B 42, 6954 48 Y. Uehara, T. Matsumoto, and S. Ushioda, Phys. Rev. B 66,

共1990兲. 075413 共2002兲.

18 E. A. Colbourn and J. E. Inglesfield, Phys. Rev. Lett. 66, 2006 49 K. Bobrov and L. Guillemot, Phys. Rev. B 78, 121408共R兲

共1991兲. 共2008兲.
19
M. Sotto, Surf. Sci. 260, 235 共1992兲. 50
R. Feidenhans’l, F. Grey, M. Nielsen, F. Besenbacher, F. Jensen,
20 F. Besenbacher and J. K. Nørskov, Prog. Surf. Sci. 44, 5 共1993兲.
E. Laegsgaard, I. Stensgaard, K. W. Jacobsen, J. K. Nørskov,
21 T. Lederer, D. Arvanitis, G. Comelli, L. Tröger, and K. Baber-
and R. L. Johnson, Phys. Rev. Lett. 65, 2027 共1990兲.
schke, Phys. Rev. B 48, 15390 共1993兲. 51 R. Feidenhans’l, F. Grey, R. L. Johnson, and M. Nielsen, Phys.
22
W. Liu, K. C. Wong, H. C. Zeng, and K. A. R. Mitchell, Prog. Rev. B 44, 1875 共1991兲.
Surf. Sci. 50, 247 共1995兲. 52
G. Dorenbos, M. Breeman, and D. O. Boerma, Phys. Rev. B 47,
23 T. Fujita, Y. Okawa, Y. Matsumoto, and K.-I. Tanaka, Phys. Rev.
1580 共1993兲.
B 54, 2167 共1996兲. 53
L. D. Sun, M. Hohage, and P. Zeppenfeld, Phys. Rev. B 69,
24 M. Kittel, M. Polcik, R. Terborg, J. T. Hoeft, P. Baumgärtel, A.
045407 共2004兲.
54
M. Bradshaw, R. L. Toomes, J.-H. Kang, D. P. Woodruff, M. L. D. Sun, M. Hohage, R. Denk, and P. Zeppenfeld, Phys. Rev. B
Pascal, C. L. A. Lamont, and E. Rotenberg, Surf. Sci. 470, 311 76, 245412 共2007兲.
共2001兲. 55
W. X. Li, C. Stampfl, and M. Scheffler, Phys. Rev. B 68, 165412
25 S. Stolbov and T. S. Rahman, Phys. Rev. Lett. 89, 116101
共2003兲.
共2002兲. 56
K. Reuter, C. Stampfl, and M. Scheffler, Handbook of Materials
26 S. Stolbov, A. Kara, and T. S. Rahman, Phys. Rev. B 66, 245405
Modeling 共Springer, Berlin, 2005兲, Vol. 1.
共2002兲. 57
A. Soon, M. Todorova, B. Delley, and C. Stampfl, Phys. Rev. B
27 I. Merrick, J. E. Inglesfield, and H. Ishida, Surf. Sci. 551, 158
73, 165424 共2006兲.
共2004兲. 58 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77,
28 M. Alatalo, S. Jaatinen, P. Salo, and K. Laasonen, Phys. Rev. B
3865 共1996兲.
70, 245417 共2004兲. 59
B. Delley, J. Chem. Phys. 113, 7756 共2000兲; 92, 508 共1990兲.
29 T. Kangas, K. Laasonen, A. Puisti, H. Pitkänen, and M. Alatalo, 60 C. Kittel, Introduction of Solid State Physics 共Wiley, New York,

Surf. Sci. 584, 62 共2005兲. 1996兲.

30 A. Puisto, H. Pikänen, M. Alatalo, S. Jaatinen, P. Salo, A. S. 61 A. Soon, L. Wong, B. Delley, and C. Stampfl, Phys. Rev. B 77,

075430-14
DENSITY FUNCTIONAL STUDY OF OXYGEN ON Cu共100兲… PHYSICAL REVIEW B 81, 075430 共2010兲

125423 共2008兲. 共1992兲.

62 B. Hammer, L. B. Hansen, and J. K. Nørskov, Phys. Rev. B 59, 72 J. L. F. Da Silva, C. Stampfl, and M. Scheffler, Surf. Sci. 600,
7413 共1999兲. 703 共2006兲.
63 K. Huber and G. Herzberg, Molecular Spectra and Molecular 73
J. L. F. Da Silva, C. Barreteau, K. Schroeder, and S. Blügel,
Structure IV: Constants of Diatomic Molecules 共Van Nostrand Phys. Rev. B 73, 125402 共2006兲.
Rienhold, New York, 1979兲. 74 G. Cipriani, D. Loffreda, A. D. Corso, S. De Gironcoli, and S.
64
P. O. Gartland, S. Berge, and B. J. Slagsvold, Phys. Rev. Lett.
Baroni, Surf. Sci. 501, 182 共2002兲.
28, 738 共1972兲.
65 X. M. Tao, M. Q. Tan, X. X. Zhao, W. B. Chen, X. Chen, and X.
75
M. Gajdoŝ, A. Eichler, and J. Hafner, Surf. Sci. 531, 272 共2003兲.
76
F. Shang, Surf. Sci. 600, 3419 共2006兲. C. Stampfl, S. Schwegmann, H. Over, M. Scheffler, and G. Ertl,
66 H. L. Davis and J. R. Noonan, Surf. Sci. 126, 245 共1983兲. Phys. Rev. Lett. 77, 3371 共1996兲.
77
67
J. L. F. Da Silva, K. Schroeder, and S. Blügel, Phys. Rev. B 69, I. Costina, M. Schmid, H. Schiechl, M. Gajdoŝ, A. Stierle, S.
245411 共2004兲. Kumaragurubaran, J. Hafner, H. Dosch, and P. Varga, Surf. Sci.
68
R. Mayer, C. S. Zhang, K. G. Lynn, W. E. Frieze, F. Jona, and P. 600, 617 共2006兲.
78 M. Methfessel, D. Hennig, and M. Scheffler, Phys. Rev. B 46,
M. Marcus, Phys. Rev. B 35, 3102 共1987兲.
69
J. Wan, Y. L. Fan, D. W. Gong, S. G. Chen, and X. Q. Fan, 4816 共1992兲.
79 G. Wulff, Z. Kristallogr. Mineral. 34, 449 共1901兲.
Modell. Simul. Mater. Sci. Eng. 7, 189 共1999兲.
70 H. P. Bonzel and M. Nowicki, Phys. Rev. B 70, 245430 共2004兲. 80 W. Kaminsky, J. Appl. Crystallogr. 38, 566 共2005兲.
71
H. L. Skriver and N. M. Rosengaard, Phys. Rev. B 46, 7157 81
W. Kaminsky, J. Appl. Crystallogr. 40, 382 共2007兲.

075430-15