An adaptive algorithm for n-body field expansions
Weinberg, Martin D.
1998-05-28
Date
2007-01-24
Description
Comment: 31 pages
The adsorption of CO on unreconstructed and reconstructed Ir{100} has been
studied, using a combination of density functional theory and thermodynamics,
to determine the relative stability of the two phases as a function of CO
coverage, temperature and pressure. We obtain good agreement with
experimentaldata. At zero temperature, the (1X5) reconstruction becomes less
stable than the unreconstructed (1X1) surface when the CO coverage exceeds a
critical value of 0.09 ML. The interaction between CO molecules is found to be
repulsive on the reconstructed surface, but attractive on the unreconstructed,
explaining the experimental observation of high CO coverage on growing (1X1)
islands. At all temperatures and pressures, we find only two possible stable
states: 0.05 ML CO c(2X2) overlayer on the (1X1) substrate, and the clean
(1$\times$5) reconstructed surface.
studied, using a combination of density functional theory and thermodynamics,
to determine the relative stability of the two phases as a function of CO
coverage, temperature and pressure. We obtain good agreement with
experimentaldata. At zero temperature, the (1X5) reconstruction becomes less
stable than the unreconstructed (1X1) surface when the CO coverage exceeds a
critical value of 0.09 ML. The interaction between CO molecules is found to be
repulsive on the reconstructed surface, but attractive on the unreconstructed,
explaining the experimental observation of high CO coverage on growing (1X1)
islands. At all temperatures and pressures, we find only two possible stable
states: 0.05 ML CO c(2X2) overlayer on the (1X1) substrate, and the clean
(1$\times$5) reconstructed surface.
Type
Database
Link to record
Show preview
Hide preview
ar
X
iv :c
on d-
m at
/0 70
15 97
v1 [
co nd
-m at.
mt rl-
sc i]
24 Ja
n 2 00
7 Lifting of Ir{100} reconstruction by CO adsorption:
an ab initio study
Prasenjit Ghosh and Shobhana Narasimhan
Theoretical Sciences Unit,
Jawaharlal Nehru Centre for Advanced Scientific Research,
Jakkur, Bangalore 560 064, India
Stephen J. Jenkins and David A. King
Department of Chemistry, University of Cambridge,
Lensfield Road, Cambridge CB2 1EW, U.K.
(Dated: February 6, 2008)
The adsorption of CO on unreconstructed and reconstructed Ir{100} has been studied, using a
combination of density functional theory and thermodynamics, to determine the relative stability of
the two phases as a function of CO coverage, temperature and pressure. We obtain good agreement
with experimental data. At zero temperature, the (1×5) reconstruction becomes less stable than
the unreconstructed (1×1) surface when the CO coverage exceeds a critical value of 0.09 ML.
The interaction between CO molecules is found to be repulsive on the reconstructed surface, but
attractive on the unreconstructed, explaining the experimental observation of high CO coverage on
growing (1× 1) islands. At all temperatures and pressures, we find only two possible stable states:
0.05 ML CO c(2× 2) overlayer on the (1×1) substrate, and the clean (1×5) reconstructed surface.
2 I. INTRODUCTION
Automotive exhausts contain several noxious gases, such as CO and NO, which need to be converted to less harmful
products (e.g., CO oxidized to CO2, and NO reduced to N2) before they are released into the atmosphere. The surfaces
of metals like Pt, Ir and Rh are good catalysts for such reactions. A better understanding of the catalytic activity
of these surfaces could lead to the development OF cheaper and better catalysts, and these reactions have therefore
been the subject of extensive experimental and theoretical studies1.
In response to reduced coordination at the surface, the {100} surfaces of the face centered cubic (fcc) metals Au,
Pt and Ir reconstruct to form a corrugated quasi-hexagonal overlayer (hex) on top of the square fcc substrate. While
Pt{100} and Au{100} display complex periodic patterns with large unit cells2, Ir{100} has a similar structure, but
with a comparatively small (1×5) unit cell. Low energy electron diffraction (LEED) experiments3,4,5,6 and density
functional theory (DFT) calculations7 on Ir{100} have shown that the stable overlayer registry with respect to the
substrate is in accordance with the “two-bridge model”. The lack of registry between overlayer and substrate atoms
results in a significant buckling in the two topmost layers6, as well as a lateral shift in the second layer. Moreover, the
reconstruction appears to go down deep, into at least the fourth layer8. Though most previous studies have focused
on adsorption and reaction mechanisms on Pt{100} rather than Ir{100}, in this work we choose to focus on Ir{100},
since its smaller unit cell makes it more accessible to ab initio calculations. However, the structural similarity of the
reconstructions on Pt{100} and Ir{100} suggests that the chemical reactions and physical processes taking place on
the two surfaces might have similar mechanisms.
Surface adsorbates such as CO, NO and O2 are known to lift the reconstruction on Pt{100} and Ir{100}. For
Pt{100}, the adsorbate-induced lifting of the reconstruction has been widely investigated9,10,11,12,13. There have also
been a few studies on the lifting of reconstruction on Ir{100} by molecular adsorbates14,15,16,17,18. Measurements of
heats of adsorption, on both unreconstructed and reconstructed Pt{100} and Ir{100}, by Hopkinson et al.11, Yeo et
al.17 and Ali et al.14, have shown that the binding energy of CO is greater on the unreconstructed surface than the
reconstructed one; this provides the driving force for the lifting of the reconstruction upon adsorption of CO.
The catalytic oxidation of CO on metal surfaces is certainly one of the most important catalytic processes studied in
surface science. When the catalyst is Pt{100} or Ir{100}, the CO-induced lifting of the reconstruction forms a crucial
component of the catalytic cycle, since the surface alternates rapidly between being CO-rich and (after the CO has
combined with oxygen to form CO2) being clean. Thus, the surface alternates also between being unreconstructed
and reconstructed, and the thermodynamics and kinetics governing this process are of great interest if one wishes to
3 gain a better understanding of the catalytic cycle.
The parameters that characterize the process by which the reconstruction is lifted are: (i) the adsorbate coverage
on the metal surface, (ii) the nucleation of unreconstructed (1× 1) islands, and (iii) the growth rate of these islands.
In the following paragraphs, we summarize the present state of knowledge regarding these three parameters:
(i) Coverage: What is the coverage of CO on the (1×1) islands, and at what critical coverage is the lifting initiated?
Is the critical parameter the global CO coverage, or is it a local CO coverage (that differs from the average value)
that is important? In experimental investigations of the restructuring process on Ir{100} and Pt{100}, it has been
reported that the local CO coverage on the growing (1×1) islands is 0.5 ML. However, the critical value of the total
CO coverage, for the onset of the transition, has been found to be much lower than this, on both Ir{100} and Pt{100}.
For example, thermal energy atomic scattering measurements on Ir{100} and Pt{100} suggest that the restructuring
begins somewhere between 0.05 and 0.13 ML on Ir{100}14, and between 0.01 and 0.03 ML on Pt{100}11; while electron
energy loss spectroscopy measurements by Behm et al.19 and Rutherford back scattering measurements by Jackman
et al.20 on Pt{100} have indicated that the lifting of the reconstruction is initiated at a critical CO coverage of 0.05
ML and 0.08 ± 0.05 ML respectively. When the total CO coverage is 0.5 ML, the entire surface appears to be in the
(1×1) phase.
(ii) Nucleation: It is not clear from the literature whether the nucleation of (1×1) islands is homogeneous or
heterogeneous. On the one hand, the finding of similar CO adsorption energies on steps and terraces reported by
Hopster et al.21, indicates that the nucleation may be homogeneous. Using scanning tunneling microscopy (STM),
Ritter et al.12 have proposed that homogeneous nucleation of the (1×1) islands takes place due to fluctuations in
the density of CO molecules. According to them, when the islands grow beyond a critical size, they become stable
and act as nucleation centers. The spatial progress of the transformation occurs when the rate of growth of the
islands becomes large compared to their rate of nucleation. On the other hand, STM studies by Borg et al.22 suggest
that the restructuring is initiated by heterogeneous nucleation, the nucleation centers being step edges and structural
irregularities disrupting the hexagonal structure along a direction close to the [1 5] direction of their [N 1 ; 1 5]
reconstruction. In agreement with this, molecular dynamics (MD) simulations by van Beurden et al.13,23 of the lifting
of the reconstruction on Pt{100}, at CO coverages between 0.4 and 0.5 ML, indicate that the transformation is
heterogeneously nucleated at step edges aligned along the [011] direction.
(iii) Growth rate: From their molecular beam experiments, Hopkinson et al. deduced how r1×1, the growth rate of
the (1×1) islands, depends on the local CO coverage on the hex-surface (ΘhexCO ). They obtained a non-linear variation
4 of the growth rate: r1×1 = k(Θ hex CO )
nΘhex, where k is independent of Θ hex CO , Θhex is the fraction of surface remaining
in the reconstructed form, and the reaction order n was found to be 4.5± 0.4. From this, they concluded that 4 to 5
CO molecules must be involved cooperatively in the growth of the (1×1) phase. Similar studies by Ali et al.14 of the
CO-induced lifting of reconstruction on Ir{100} also showed a power law behavior for the growth of (1×1) islands,
with 3.9 ≤ n ≤ 5.8. Recent time-resolved reflection absorption infrared spectroscopy (RAIRS) results indicate a
similar power law relationship by an alternative, and very direct, technique24.
For the Pt surface, the presence of a non-linear term in the growth law appears to be crucial to the appearance of
oscillations in many catalytic processes, e.g., CO oxidation with NO25,26, CO oxidation with O2 27,28 and NO reduction
with H2 29,30. Under reaction conditions, the catalyst surfaces undergo transformation; these are non-equilibrium
processes. Moreover the reaction rate may not remain constant but changes periodically or exhibits chaotic behavior.
There can also be the formation of spatial patterns on the catalyst surface31, and these have been modelled recently
within a sophisticated reaction-diffusion scheme32. Although we are not aware of any literature concerning oscillatory
reactions on Ir{100}, the experimental findings by Ali et al. suggest that such oscillatory behavior may be present at
temperatures above 900 K.
What is the underlying mechanism that is responsible for the non-linear growth law? Hopkinson et al.10 and Ali
et al.14 have proposed a simple mechanism for the restructuring process which involves a cooperative phenomenon
among CO molecules: due to statistical fluctuations of the local CO coverage on the hex phase, 4-5 CO molecules
come together at the boundary of the growing (1×1) domain, or at a step, and convert 8-10 Pt atoms from a hex
to square arrangement; however it is not clear why 4-5 CO molecules are needed or precisely how they cooperate.
One explanation for the ‘magic’ value has been suggested by Passerone et al.33, who performed MD simulations on
Au{100}, and found that when islands/craters are formed by adsorbing/desorbing Au atoms, they do not remain
stable unless they exceed a critical size of 8-10 Au atoms. However, it is not clear that this number will translate
to Pt or Ir surfaces, especially in the presence of CO. The restructuring of Pt{100} has also been studied in more
recent MD simulations by van Beurden and coworkers13, who have found that the CO molecules initiate surface
relaxations and lead to a shear tension between adjacent [011] rows due to the preference for a square rather than
a hexagonal coordination sphere, and the restructuring proceeds through the ejection of chains of Pt atoms, and a
rearrangement of the remaining surface atoms. Their simulations showed no evidence of a cooperative phenomenon
between 4-5 CO molecules, and the source of the non-linear growth law remains a mystery. However, it is important
to note that their MD simulations were done at much higher CO coverages (0.4-0.5 ML) than the critical coverage at
5 which the experiments showing the fourth-order power law dependence of the rate of (1×1) formation were conducted.
The mechanism for the lifting of the reconstruction is clearly not the same for low and high CO coverage. At low
coverage, lifting of the reconstruction occurs when random statistical fluctuations bring about a sufficiently high local
coverage to instigate nucleation of (1×1) islands; at the high total coverages studied in the MD simulations, such
high local coverage would be found across the whole surface at all times. The experimentally-observed power law
dependence stems from the probability of bringing together 4-5 CO molecules within a sufficiently small area at low
overall coverage.
As a first step towards quantifying these arguments theoretically, it is desirable first to understand the thermody-
namics, i.e., at what coverage the phase transition becomes thermodynamically favorable. In this work, we determine
this quantity through the study of the thermodynamics of the reconstructed and unreconstructed surfaces using ab
initio DFT. We then extend these results to finite temperatures and pressures, by making use of the chemical potential.
This is useful because it brings the results into regimes where it becomes possible to compare with experiment.
Our work is similar in spirit to a recent DFT study by Deskins et al.16, who showed that the adsorption of atomic
oxygen on Pt{100} makes the unreconstructed surface more thermodynamically stable than the reconstructed one.
In their work, the reconstructed structure of Pt{100} was approximated by a (1 × 5) unit cell. Note however that:
(i) our focussing on Ir{100} enables us to use a relatively small unit cell that corresponds to the true reconstructed
structure; (ii) we have studied the adsorption of CO; (iii) details of our analysis are also different.
The rest of this paper is structured as follows: Section II provides some details about our first-principles calculations.
Sections III A and III B contain the results of our calculations on the clean Ir{100} surface (both unreconstructed
and reconstructed), while our results for CO adsorbed on the unreconstructed and reconstructed Ir{100} surfaces
are contained in Sections III C and III D respectively. We emphasize that it is particularly important to choose k-
point meshes (used for Brillouin zone integrations) very carefully; this issue is discussed in Section II, while some
illustrative examples are presented in III C. The ab initio density functional theory results feed into an analysis using
the entropy and the chemical potential, presented in Section III E. Finally, we discuss the implications of our results
and summarize in Section IV.
II. DETAILS OF AB INITIO CALCULATIONS
It is known that the lifting of the reconstruction in the systems under study is governed by very small differences in
the energies of the competing structures; it is therefore vital to do as accurate a calculation as possible. For this reason,
6 we have chosen to perform ab initio calculations within the framework of density functional theory34 (DFT), since
this is perhaps the most reliable method currently available for obtaining accurate values of ground state properties
such as structures, surface energies and adsorption energies.
Our calculations have been performed using the CASTEP package, wherein the Kohn-Sham35 equations are solved
iteratively by conjugate gradient minimization.36 We have used a plane wave basis set, with a cut-off of 25 Ry (340)
eV, and ultrasoft pseudopotentials. The pseudopotentials for Ir, C and O are Ir OO.usp, C OO.usp and O OO.usp
respectively which are provided along with the CASTEP distribution (Version 4.2). For the exchange-correlation
interactions, we have used the Perdew-Wang form of the generalized gradient approximation (GGA)37; note that
earlier calculations7 showed that gradient corrections are essential in properly describing this surface, and that use
of the local density approximation (LDA) instead of the GGA would incorrectly predict clean Ir{100} to be stable
against reconstruction. Integrations over the Brillouin zone have been evaluated with a Monkhorst-Pack (MP)38 mesh
(further details of which are given below), along with a Gaussian smearing function of width 0.1 eV.
In order to test the reliability of the pseudopotentials used in our calculations, we first performed calculations on
bulk Ir and a CO molecule in the gas phase. For the former we obtain a lattice constant of 3.86 A˚, which is in excellent
agreement with the experimental value of 3.84 A˚, while for the latter, we obtained a C-O bond length of 1.14 A˚which
also agrees very well with the experimental value of 1.13 A˚39.
For surface calculations, we use a supercell consisting of a slab of six layers of Ir atoms separated by a vacuum
thickness of about 10 A˚. The top four Ir layers are allowed to relax their positions, whereas the bottom two are
fixed at the bulk separation (‘asymmetric’ slab). Further, only the top surface is allowed to reconstruct; note that
the density of atoms is different on a reconstructed and unreconstructed surface. Thus, all such asymmetric slabs
will contain one surface that is of interest to us, while the other side consists of a bulk-truncated (unrelaxed and
unreconstructed) Ir{100} surface. In order to determine the surface energy of the latter, we also perform a calulation
on a ‘symmetric’ slab, comprised of eight layers, of which the middle two are fixed at the bulk spacing, while the top
three and bottom three layers are allowed to relax. A comparison of the symmetric and asymmetric slabs for the
unreconstructed structure enables one to determine separately the surface energies of a relaxed and bulk-truncated
(1× 1) surface.
The size of the surface unit cell used depends upon whether whether we are looking at an unreconstructed surface or
a reconstructed one, and what CO coverage we are considering. Calculations for the clean unreconstructed surface were
carried out using both (1× 1) and (1× 5) cells; the results obtained with the two were almost identical. For the clean
7 reconstructed surface, we use a (1×5) cell. To study CO adsorption on the unreconstructed and reconstructed surface
we adsorb CO on only one side (the side which we are allowing to relax) of the ’asymmetric slab’. For adsorption of
CO on the unreconstructed surface, we consider coverages of 0.11, 0.125, 0.2, 0.25 and 0.5 ML. The unit cells used for
these calculations are shown in Fig. 1; note that (i) all the unit cells are square, i.e., the distance between adjacent
CO molecules is the same in both directions, and (ii) the cells for different coverages are not necessarily commensurate
with one another. Finally, for studying adsorption of CO on the reconstructed surface, we use a (2 × 5) cell, within
which we consider CO coverages of 0.1, 0.2 and 0.6 ML. (Note that all CO coverages in this paper are given with
respect to the density of atoms in the topmost layer of the unreconstructed surface.)
An important consideration is the choice of k-point meshes for Brillouin zone (BZ) sampling. For the surface cells,
we have used Monkhorst-Pack meshes, of the form (n1×n2×1), where n1 and n2 determine the fineness of the mesh.
This issue becomes particularly crucial in the present problem, since we are interested in computing energy differences
that are comparable to the errors introduced by incorrect (unconverged) BZ sampling. Such errors can be reduced
by using a k-point mesh that is as fine as possible, while keeping computational feasibility in mind. We have found
that, as expected, the convergence is faster when one uses a mesh that does not include high symmetry points [the
Brillouin zone center (BZC) and the k-points on the edges of the BZ]; this corresponds to choosing n1 and n2 to be
even numbers. Examples of this are presented below in Section IIID.
The specific choices made for n1 and n2 for the various surface cells used in the present paper are given in Table I;
we emphasize that convergence with respect to k-point sampling has been carefully established for all the cases. Note
also that computing surface energies and adsorption energies requires comparing the total energies of two systems
(bulk and clean surface, or clean and covered surface); in such cases, we are careful to make sure that the unit cells
and k-point meshes used for the two systems are either identical or related by folding, thereby further reducing the
errors introduced by incomplete k-point sampling.
III. RESULTS
A. Unreconstructed Clean Surface
Using the asymmetric six-layer slab described above, we find that the first interlayer distance d12 is contracted
(with respect to the bulk interlayer separation) by 6.14%. This result compares well with that obtained from ab initio
calculations by Ge et al.7(6.5%), but is larger than the reported experimental value (∼ 3.6%)40. Upon comparing the
8 total energy of this asymmetric slab with that of the bulk structure, we find that the sum of the surface energies of the
two surfaces (one relaxed and the other unrelaxed) is 3.01 eV per (1× 1) area. Next, using the symmetric eight-layer
slab, which possesses two relaxed surfaces, we determine the energy of the relaxed unreconstructed surface, Γrel1×1, to
be 1.50 eV per (1 × 1) area, while the energy of the unrelaxed (bulk-truncated) unreconstructed surface, Γunrel1×1 , is
1.51 eV per (1 × 1) area. This latter quantity is subtracted out when determining surface energies for slabs that are
reconstructed and/or have CO adsorbed on only one side.
B. Reconstructed Clean Surface
In the reconstructed surface, the topmost layer forms a quasi-hexagonal layer on top of the square substrate, as can
be seen in Fig. 2(a). The surface unit cell is (1× 5). The reconstruction results in buckling within the layers, as well
as lateral shifts of the atoms with respect to their bulk-truncated positions. The parameters used to specify these
structural rearrangements are indicated in Fig. 2(b), and the values we obtain for them are presented in Table II. It
can be seen that our results are in excellent agreement with those obtained from LEED and STM by Schmidt et al.8.
In accordance with their observations, we find that in addition to the reconstruction of the topmost layer, there are
significant lateral shifts and buckling in the three layers below. The only (minor) difference between our results and
theirs is that our calculations yield a very small lateral shift of the third atom in the second layer (p32) in a direction
opposite to that determined by them.
The value we obtain for the surface energy of the reconstructed surface is Γrel1×5 = 1.45 eV per (1 × 1) area, and
we thus correctly obtain the result that the clean surface would prefer to reconstruct. Note also that the energy of
reconstruction for Ir{100} is very small, viz. 0.05 eV/(1×1) area. This is in reasonably good agreement with the
value of 0.07 eV/(1×1) area obtained by Ge et al.. The slight difference between our results and theirs presumably
arise from the use of slabs of different thicknesses and different k-point meshes. (Note that the energy difference here
is somewhat smaller than the value of 0.21 eV/(1×1) area found experimentally for the Pt{100} surface17).
C. CO on unreconstructed Ir{100}
As discussed in Sections I and II, the choice of a proper k-point mesh is very crucial for our calculations. Even
values of n1 and n2 lead to better sampling of the BZ and quicker convergence of the adsorption energy. This fact
becomes evident on inspecting Fig. 3, where we have shown how the adsorption energy, for CO at 0.5 ML and using
9 a ( √ 2×√2) unit cell, converges as a function of k-point sampling. Accordingly, we use a (12× 12× 1) mesh for this
particular unit cell; similar checks were performed for other coverages and cells.
Experiments and previous theoretical calculations41 indicate that the atop site is the most probable site for CO
adsorption. To verify this, and compare our results with previous DFT calculations and LEED measurements,41, we
calculate the geometry and adsorption energies of CO molecules occupying hollow, bridge and atop sites within a
(2×2) surface unit cell, and at 0.5 and 0.25 ML CO coverages. The adsorption energy per CO molecule, Eads, is given
by:
Eads = Eslab+CO − Eslab − nCOECO
nCO , (1)
where Eslab+CO, Eslab and ECO are the the total energies of the slab with CO adsorbed on it, the clean slab, and a
CO molecule in the gas phase, respectively, while nCO is the number of CO molecules adsorbed per surface unit cell.
The results obtained by us for Eads, for different sites at the two coverages considered, are summarized in Table III.
Although there are slight numerical differences between the adsorption energies obtained from our calculations and
those of Titmuss et al.41, both sets of calculations predict that CO molecules adsorb on the atop site. The differences
in the numerical values of the adsorption energy obtained from the two calculations presumably arise from the use of
slabs of different sizes and different k-point mesh. The structural parameters are also in good agreement with both
the LEED measurements and theoretical calculations.
We go on to study the variation of the adsorption energy of CO on the atop site as a function of CO coverage. The
results are summarized in the second column of Table IV. The difference in adsorption energy between 0.5 ML and
0.25 ML CO coverage is 0.04 eV. For lower coverage, (Θ ≤ 0.25 ML) the adsorption energy is more or less constant
(∼ 2.51 eV). The slight variation in the numerical values of Eads is most likely due to numerical errors that arise
because the k-point meshes used for different surface unit cells are not exactly commensurate. At lower coverage,
the distance between CO molecules increases, decreasing the interaction between them. At sufficiently low coverage,
the CO molecules are so far apart that they do not interact with one another, resulting in a constant value of the
adsorption energy. The increase in Eads with increase in CO coverage indicates the presence of very weak attractive
interactions between nearest-neighbor CO molecules on the unreconstructed surface.
In addition to the CO adsorption energies, we also calculate the “surface energies” (Γs) at different CO coverages.
These “surface energies” will be used below as a measure of the stability of the reconstructed and unreconstructed
surfaces. We define the “surface energy” of the CO covered surface (Γs) as:
10
Γs = Eslab+CO − nIrEbulk − nCOECO
Ns − Γunrels , (2)
where Ns is the ratio of the area of the surface unit cell to that of the (1×1) cell, and the superscript unrel represents
the unrelaxed and unreconstructed lower surface of the slab.
D. CO on reconstructed Ir{100}
CO adsorption on the reconstructed surface has been studied for coverages of 0.1, 0.2 and 0.6 ML. In all the cases,
we use a (2×5) supercell, so as to minimize the interaction between periodic images of CO molecules in adjacent
supercells. Since the reconstructed surface is quasi-hexagonal in nature, there are many possible adsorption sites.
The different adsorption sites which we consider for CO adsorption at 0.1 ML coverage are shown in Fig. 5. Our
result for the adsorption energies for these different sites are given in table V. Of all the different possibilities, we
find that the “atop3” site (T3) is the most probable one. While it is not surprising that CO prefers an atop site, it
is somewhat unexpected that the most favoured site is atop the Ir atom that is lies lowest within the buckled surface
layer, rather than the Ir atom that protrudes. We intend to address the origin of this very surprising result in a future
publication.
For a coverage of 0.2 ML, we have to adsorb two CO molecules in the (2×5) cell. Since CO clearly prefers to adsorb
on the “atop3” (T3) sites, we choose a combination of two T3 sites. There are three inequivalent combinations of two
T3 sites in the (2×5) cell, namely, A and B, A and C, and A and D (see Fig. 6). Our results for the CO adsorption
energies at these different sites are listed in Table V. Out of these three possible combination of sites, we find similar
adsorption energies for the AB and AC combinations. The fact that the AD combination is disfavored suggests that
CO molecules on this surface would prefer not to sit too close to each other. Comparing the adsorption energies
for coverages of 0.1 and 0.2 ML, we find that the interaction between CO molecules on the reconstructed surface is
repulsive in nature, in contrast to the very weak attractive interaction on the unreconstructed surface. For a coverage
of 0.6 ML, we need to adsorb six CO molecules per (2×5) cell. Since the CO molecules interact repulsively on the
unreconstructed surface, they will tend to spread out uniformly at low temperatures. Hence we assume that at a
coverage of 0.6 ML, CO will adsorb at every atop site, as shown in Fig. 6.
The variation, with coverage, of the “surface energies” of the reconstructed and unreconstructed surfaces (computed
using Equation. (2)), is plotted in Fig. 7. At zero CO coverage, the reconstructed (1×5) surface is energetically more
11
stable than the unreconstructed (1×1). However, at 0.1 ML coverage the unreconstructed surface is more stable than
the reconstructed one by about 0.03 eV per (1×1) area. The crossover between the stability of the two surfaces takes
place at around 0.09 ML CO coverage. Thus adsorption of CO switches the stability of the Ir{100} surfaces at a very
low total coverage, that appears to be in very good agreement with the experimentally reported values of Hopkinson et
al.. However, we note that this analysis has been done at conditions corresponding to zero temperature and pressure.
In the next section, we extend this results to finite temperatures and pressures, by performing a thermodynamic
analysis.
E. Thermodynamic Analysis
In order to account for the effect of varying gas-phase temperature and pressure upon the surface configuration, we
apply concepts from classical thermodynamics. The free energy (F ) of a system gives a measure of the stability of
the system. For a multi-species system F is given by:
F = E + PV − TS − ∑
i
niµi, (3)
where P , V , T and S are the pressure, volume, temperature and entropy, respectively. ni and µi are, respectively,
the number of molecules and the chemical potential of species i at the surface. E is the internal energy of the system
and is obtained from the DFT calculations.
In order to apply these concepts we need to have a clear demarcation between that which we consider to be a part
of the “system” and that which we take to be its “surroundings”. For our case this can be achieved in two ways:
(i) entropy viewpoint, in which the “system” is considered to include the surface, the adsorbed molecules on the
surface and the gas above the surface; the “surroundings” include only the world beyond some external gas container.
Taking this view, the number of particles in the system is now fixed but the entropy is considerably large, since it
includes large translational and rotational contributions from the gas molecules which are not adsorbed on the surface.
Hence Equation (3) is modified as:
FS = E + PV − TS. (4)
From this viewpoint, the changes in the configuration of the surface are driven by massive changes in entropy between
molecules on the surface and in the gas phase; net adsorption (desorption) occurs when the enthalpy of adsorption is
12
greater (less) than this entropy difference.
(ii) chemical potential viewpoint, in which the “system” consists of the surface and any molecules adsorbed on the
surface. The rest of the world beyond then constitute the “surroundings”. Now to a fair approximation the entropy
of the system is more or less negligible, since it consists only of vibrational contributions. On the other hand, the
number of particles in the system is variable, so the chemical potential of the gas molecules must be included. The
thermodynamic potential F now becomes:
Fµ = E + PV − ∑
i
niµi. (5)
The changes in surface configuration are driven by the need to maintain the equality of surface and gas-phase chemical
potentials; net adsorption (desorption) occurs when the gas-phase chemical potential is instantaneously greater (less)
than the surface chemical potential. Note that from both the approaches one should get the same results.
1. Entropy Viewpoint
Taking the viewpoint that our thermodynamic system includes the surface, adsorbed molecules and the gas above
it, we calculate the necessary free energy that must be minimized. This is achieved by defining free energies relative
to that of the clean unreconstructed surface:
δFS = FSslab+CO − FSclean(1×1) −∆nIrFSbulk − nCOFSCO
Ns , (6)
where FS clean(1×1) and F
S slab+CO have been evaluated for the same unit cell, thus mitigating any systematic errors that
may arise due to different k-point sampling. To evaluate the free energy of the solid phases, slab with CO molecules
adsorbed on it (FSslab+CO) and for the clean one (F S clean(1×1)) we ignore the effects of TS because typically the entropy
of the solid phase is negligible compared to that of the gas phase. For the solid phases we ignore the effects of PV and
approximate it for CO in the gas phase, by assuming that the gas phase obeys the ideal gas equations. With these
approximations the free energy of the solid phases equals the internal energy obtained from DFT calculations. The
third term in Equation (6) contains the free energy per atom in the Ir bulk (FSbulk) and the difference in the number
of Ir atoms (∆nIr) between the DFT calculations used to obtain the first two terms. In the last term, F S CO is the
free energy of the CO molecules in the gas phase. The last term accounts for the decrease in entropy of the system
due to the adsorption of CO molecules from the gas phase. Ns represents the number of (1×1) cells over which the
13
value of FS has been calculated. By construction, therefore, δFS for the clean reconstructed surface is the negative
of the reconstruction energy, while for the clean unreconstructed surface δFS = 0. The lowest value of δFS under
any temperature and pressure conditions indicates the most thermodynamically stable surface configuration, δFSmin.
In order to evaluate FSCO we need to know the value of S for the gas phase at different temperatures and pressures.
The values of S at a pressure of 1 bar and for a temperature range of 298.15 to 1500 K can be obtained from the
CRC handbook39. We then derive the variation of S with pressure and temperature from the following formula:
S(T, P ) = S(T, P 0)−Rln(P/P0). (7)
P 0(=1 bar) is the reference pressure and R is the universal gas constant. In Fig. 8, we plot a series of lines
calculated at a variety of pressures, each plotting the value of δFS as a function of temperature. It is apparent
that for all temperatures and pressures considered, there are only two possible configurations of the surface that
are thermodynamically stable, the clean unreconstructed (1×5) surface and the 0.5 ML c(2×2) CO overlayer on the
unreconstructed substrate. The critical temperature at which the crossover in the stability occurs varies as a function
of pressure and is depicted in Fig. 9.
2. Chemical Potential Viewpoint
In order to evaluate the relative stability of different surface configurations from the chemical potential viewpoint,
we evaluate the free energy as given in by Equation (5). The chemical potential can be obtained by making use of its
relationship to the enthalpy (H) and entropy:
µi = 1
ni (H − TS). (8)
The values of entropy at standard pressure can be extracted from the CRC Handbook39. The CRC Handbook lists only
the change in H at P 0 in going from absolute zero temperature to a variety of finite temperatures (∆H(0→ T, P 0).
Therefore H(T, P 0) can be extracted using the following relation:
H(T, P 0) = H(0, P 0) + ∆H(0→ T, P 0) (9)
where H(0, P 0) can be obtained from DFT calculations. The variation of µi with pressure is computed as:
14
µi(T, P ) = µi(T, P 0) + kBT ln(P/P
0), (10)
where kB is the Boltzmann constant. Again we evaluate δF µ the same way as in Equation (6) and plot the same
quantities as evaluated in the Section III E 1. The results obtained via the chemical potential approach are exactly
the same as those obtained via the entropy approach, plotted in Fig. 8 and 9.
IV. DISCUSSION AND SUMMARY
As in previous calculations, we find that Ir{100} is reconstructed; the stabilization energy being very small, viz.
0.05 eV per (1×1) area. Note that this is comparable to the errors typically introduced by incorrect k-point sampling,
emphasizing the need for careful calculations. The structure of the reconstructed surface obtained by us is in excellent
agreement with experiment. While the increase in density is confined to the topmost layer, the three layers below
also display marked buckling and lateral shifts.
However, as more and more CO is adsorbed on the surface, there is a reversal in phase stability. Our calculations
show that at zero temperature and pressure, when CO is adsorbed on Ir{100}, the relative stability of the reconstructed
and unreconstructed phases is reversed at a very low total CO coverage of 0.09 ML. This shows that the lifting of the
reconstruction is thermodynamically favoured at a low local coverage. The need for statistical fluctuations to bring
4-5 molecules together to trigger the lifting of the reconstruction is kinetic in origin, rather than thermodynamic.
On the unreconstructed Ir{100} surface, CO adsorbs at atop sites, and there is a very weak attractive interaction
between CO molecules. Note that in this respect Ir{100} appears to behave differently from Pt{100}, where there
is a repulsive interaction between CO molecules19,42. The attractive nature of the CO-CO interactions on the unre-
constructed (1×1) surface suggests that at sufficiently low temperature, CO molecules would tend to cluster together
on being adsorbed on the fully unreconstructed surface in order to maximize their coverage, until a saturation cov-
erage of 0.5 ML is achieved. The weakness of the interaction, however, means that such clustering may not easily
be observed in experiments performed at moderate temperature. In contrast, the interaction between CO molecules
adsorbed on the reconstructed (1×5) surface is repulsive, and they tend to spread out uniformly over the surface.
Including the cost of lifting the reconstruction, the adsorption energy of CO molecules at any total coverage to form
an island of 0.5 ML local coverage on a corresponding patch of (1×1) substrate (2.45 eV per CO molecule) is still
higher than the adsorption energy of CO molecules at any coverage on the unreconstructed surface. This suggests
15
that at any total coverage, the molecules will prefer to cluster into islands of local 0.5 ML coverage on patches of the
unreconstructed (1×1) substrate, rather than remain spread out on the reconstructed (1×5) substrate. This bolsters
the interpretation of 0.5 ML local CO coverage on the unreconstructed (1×1) islands by Ali et al. even at low total
CO coverage. The higher heat of adsorption and the attractive nature of the interaction between CO molecules on the
(1×1) unreconstructed surface provide a driving force to lift the reconstruction. For both Ir{100} (where the interac-
tion between CO molecules on the unreconstructed surface is weakly attractive) and Pt{100} (where this interaction
is weakly repulsive), the main effect is that the heat of adsorption on the unreconstructed surface far exceeds that
on the reconstructed surface. As a result, the reconstruction lifting phenomenon remains similar on both Ir{100}
and Pt{100}. However we donot understand the driving force behind the shift of the weakly attractive interaction
between CO molecules on the unreconstructed surface to a repulsive nature on the reconstructed one.
Our thermodynamic analysis extends the results obtained from DFT calculations to a range of pressures and
temperatures. For the entire temperature and pressure range considered by us, there are only two thermodynamically
stable configurations, namely 0.5 ML of CO on the unreconstructed substrate in a c(2×2) structure, and the clean
reconstructed substrate. The phase diagram shows that a small change in temperature and/or pressure can result
in a very large change in coverage. From experiments, we know that the lifting of the reconstruction sets in at 490
K at very low pressures (10−10-10−07 mbar). From our phase diagram we find that at similar pressure ranges, the
transition temperature lies between 550 and 600 K. Thus our results from thermodynamic analysis are reasonably
consistent with experimental observations. The slight discrepancies may have arisen from our use of the ideal gas
equation of state to describe the gas phase. In practical applications such as in catalytic converters in vehicles, the
reactions take place at temperatures above about 750 K and at ∼ 10 mbar pressure. According to our phase diagram,
such temperatures and pressures lie very close to the transition line. Hence slight changes in the reaction conditions
may lead to drastic changes in the surface, which in turn will affect the rate of chemical reactions.
To conclude, we have investigated the thermodynamic stability of unreconstructed and reconstructed phases of
the Ir{100} surface, in the presence of CO. Though our results for the critical coverage, etc., are in good agreement
with experiment, it still remains intriguing to speculate about the role played by kinetic factors in the restructuring
process. In particular, the origin of the nonlinear growth law observed in experiments is still a puzzle; we hope that
future work on the kinetics of the restructuring mechanism will shed light on this.
One of us (P.G.) acknowledges CSIR, India for a research scholarship. The Royal Society is also thanked for a
16
University Research Fellowship (S.J.J.). We are grateful to Dr. Stephanie Pratt for helpful discussions.
1 Z-P. Liu, Pure Appl. Chem. 76, 2069 (2004).
2 M. A. Van Hove, R. J. Koestner, P. C. Stair, J. P. Bibie´rian, L. L. Kesmodel, I. Bartoˇs, and G. A. Somorjai, Surf. Sci. 103,
189 (1981).
3 M. A. Van Hove, R. J. Koestner, P. C. Bibe´rian, L. L. Kesmodel, I. Bartoˇs, and G. A. Somorjai, Surf. Sci. 103, 189 (1981).
4 E. Lang, Mu¨ller, K. Heinz, M. A. Van Hove, R. J. Koestner, and G. A. Somorjai, Surf. Sci. 127, 347 (1983).
5 N. Bickel and K. Heinz, Surf. Sci. 163, 435 (1985).
6 K. Johnson, Q. Ge, S. Titmuss, and D. A. King, J. Chem. Phys. 112, 10460 (2000).
7 Q. Ge, D. A. King, N. Marzari, and M. C. Payne, Surf. Sci. 418, 529 (1998).
8 A. Schmidt, W. Meier, L. Hammer, and K. Heinz, J. Phys.: Condens. Matter 14, 12353 (2002).
9 P. A. Thiel, R. J. Behm, P. R. Norton, and G. Ertl, Surf. Sci. 121, L553 (1982).
10 A. Hopkinson, J. M. Bradley, X. C. Guo, and D. A. King, Phys. Rev. Lett. 71, 1597 (1993).
11 A. Hopkinson, X. C. Guo, J. M. Bradley, and D. A. King, J. Chem. Phys. 99, 8262 (1993).
12 E. Ritter, R. J. Behm, G. Po¨tschke, and J. Wintterlin, Surf. Sci. 181, 403 (1987).
13 P. van Beurden, B. S. Bunnik, G. J. Kramer, and A. Borg, Phys. Rev. Lett. 90, 066106 (2003).
14 T. Ali, B. Krotzer, A. V. Walker, and D. A. King, J. Chem. Phys. 109, 10996 (1998).
15 T. Ali, B. Krotzer, A. V. Walker, Q. Ge, and D. A. King, J. Chem. Phys. 109, 9967 (1998).
16 A. Deskins, J. Lauterbach, and K. T. Thomson, J. Chem. Phys. 122, 184709 (2005).
17 Y.Y. Yeo, C.E. Wartnaby and D.A. King, Science 268, 1731 (1995).
18 S. Khatua, Z.-P. Liu and D. A. King, Surf. Sc. 584, 214 (2005).
19 R. J. Behm, P. A. Thiel, P. R. Norton, and G. Ertl, J. Chem. Phys. 78, 7437 (1983).
20 T. E. Jackman, K. Griffiths, J. A. Davies, and P. R. Norton, J. Chem. Phys. 79, 3529 (1983).
21 H. Hopster, H. Ibach, and G. Comsa, J. Catalysis 46, 37 (1977).
22 A. Borg, A.-M. Hilmen, and E. Bergene, Surf. Sci. 306, 10 (1994).
23 P. van Beurden and G. J. Kramer, J. Chem. Phys. 121, 2317 (2004).
24 S. Khatua, S.J. Pratt and D.A. King, to be submitted.
25 A. Hopkinson and D. A. King, Chem. Phys. 177, 433 (1993).
26 A. Hopkinson and D. A. King, Faraday Discuss 6, 255 (1993).
27 M. Gruyters, T. Ali, and D. A. King, Chem. Phys. Lett. 232, 1 (1995).
28 M. Gruyters, T. Ali, and D. A. King, J. Phys. Chem. 100, 14417 (1996).
17
29 M. Gruyters, A. T. Pasteur, and D. A. King, J. Chem. Soc., Faraday Trans. 92, 2941 (1996).
30 A. V. Walker, M. Gruyters, and D. A. King, Surf. Sci. 384, L791 (1997).
31 G. Nicolis and F. Baras, Chemical Instabilities (Reidel, Dordrecht, 1984).
32 A.T. Anghel, R.B. Hoyle, I.M. Irurzun, M.R.E. Proctor and D.A. King, to be published.
33 D. Passerone, F. Ercolessi, and E. Tosatti, Surf. Sci. 454-456, 947 (2000).
34 P. Hohenberg and W. Kohn, Phys. Rev. 136, 864B (1964).
35 W. Kohn and L. J. Sham, Phys. Rev. 140, 1133A (1965).
36 M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, and J.D. Joannopoulos, Rev. Mod. Phys. 64, 1045 (1992).
37 Y. Wang and J. P. Perdew, Phys. Rev. B 43, 8911 (1991).
38 Hendrik J. Monkhorst and James D. Pack, Phys. Rev. B 13, 5188 (1976).
39 CRC Handbook of Chemistry and Physics, editor-in-chief David R. Lide, 84th edition (2003-2004).
40 K. Heinz and G. Besold, Surf. Sci. 125, 515 (1983).
41 S. Titmuss, K. Johnson, Q. Ge, and D. A. King, J. Chem. Phys. 116, 8097 (2002).
42 Y. Y. Yeo, L. Vattuone, and D. A. King, J. Chem. Phys. 104, 3810 (1996).
18
TABLE I: Parameters for the Monkhorst-Pack k-point meshes used for different surface cells.
Surface unit cell k-point mesh
1×1 10×10 √ 2×√2 12×12
2×2 8×8 √ 5×√5 8×8
2 √ 2×2√2 6×6
3×3 6×6
1×5 10×2
2×5 5×2
19
TABLE II: Structure of Ir{100}-1×5 described by the parameters defined in Fig. 2. dik’s denote average interlayer spacings
between the centre of mass planes of the layers, while dik’s give the smallest spacing between their subplanes. d b
ik describes the
percent change in dik with respect to the bulk interlayer spacing (db). Note that our calculations are in excellent agreement
with experiment, except for a slight difference in the lateral shift for the third atom in layer two (p32), which we find to be
negligible and in the opposite direction compared to the experimental findings.
Parameters Our Calculations Expt.8 Ge et al.7
db (A˚) 1.93 1.92 1.92
d12 (A˚) 1.96 1.94 1.97
d12 (A˚) 2.24 2.25 -
d b
12 (%) 16.09 16.67 -
d23 (A˚) 1.82 1.79 -
d23 (A˚) 1.88 1.88 -
d b
23 (%) -2.56 -2.08 -
d34 (A˚) 1.88 1.83 -
d34 (A˚) 1.94 1.93 -
d b
34 (%) 0.54 0.52 -
d45 (A˚) 1.91 1.89 -
d45 (A˚) 1.923 1.91 -
d b
45 (%) -0.34 -1.01 -
b131 (A˚) 0.22 0.25 -
b231 (A˚) 0.53 0.55 0.47
b341 (A˚) 0.2 0.20 0.20
p21 (A˚) 0.03 0.05 0.05
p31 (A˚) 0.07 0.07 0.02
b132 (A˚) 0.04 0.07 -
b232 (A˚) 0.08 0.10 -
p22 (A˚) 0.01 0.01 -
p32 (A˚) 0.0004 -0.02 -
b133 (A˚) 0.06 0.10 -
b233 (A˚) 0.03 0.05 -
b134 (A˚) 0.04 0.06 -
b234 (A˚) 0.01 0.03 -
20
TABLE III: Summary of our calculations for the c(2×2) and p(2×2) phases and comparison with experiments and previous
theoretical calculations (Prev)41. Eads denotes the adsorption energy of CO on Ir{100}; rIr−C and dC−O represent the Ir-C
and C-O bond lengths respectively. ∆Z gives the vertical distance between two intraplanar Ir atoms. d12 denotes the distance
between the top two surface layers. The different parameters are also shown in Fig. 4.
c(2×2) phase, 0.5 ML CO coverage
Site Eads rIr−C dC−O ∆Z d12
eV/ CO molecule (A˚) (A˚) (A˚) (A˚)
Ours Prev Expt Ours Prev Expt Ours Prev Expt Ours Prev Expt Ours Prev Expt
Atop -2.69 -2.65 - 1.86 1.86 1.81 ± 0.05 1.16 1.16 1.16 ± 0.05 0.14 0.19 0.13 ± 0.05 1.82 1.82 1.82 ± 0.04
Bridge -2.51 -2.29 - 2.06 2.04 - 1.17 1.18 - 0.0 0.0 - 1.89 1.90 -
Hollow -2.1 -1.5 - 2.29 2.28 - 1.2 1.2 - 0.0 0.0 - 1.90 1.92 -
p(2×2) phase, 0.25 ML CO coverage
Atop -2.83 -2.61 - 1.86 1.86 - 1.16 1.16 - 0.15 0.20 - 1.82 1.83 -
Bridge -2.71 - - 2.05 - - 1.17 - - 0.06 - - 1.84 - -
Hollow -2.34 - - 2.46 - - 1.2 - - 0.0 - - 1.84 - -
21
TABLE IV: Adsorption energies (Eads) of CO at atop sites on unreconstructed and reconstructed Ir{100} surfaces at different
CO coverages. All values of Eads are expressed in units of eV/CO molecule.
CO coverage Eads Eads
(ML) unreconstructed surface reconstructed surface
1.0 -1.93 -
0.6 - -1.90
0.5 -2.55 -
0.25 -2.51 -
0.2 -2.51 -2.27
0.125 -2.52 -
0.11 -2.53 -
0.10 - -2.31
22
TABLE V: Adsorption energies (Eads) at different sites of the reconstructed surface at 0.1 CO coverage. See Figs. 5 and 6 for
the convention used in labelling sites.
coverage site Eads
(ML) (eV/ CO molecule)
T1 -2.07
T2 -2.09
T3 -2.31
T4 -1.92
H1 -1.99
H2 -1.68
H3 -1.70
H4 -2.01
0.1 H5 -2.21
H6 -1.89
B1 -2.00
B2 -1.62
B3 -2.19
B4 -1.63
D1 -1.76
D2 -1.73
D3 -2.00
A and B -2.27
0.2 A and C -2.27
A and D -2.15
0.6 A, B, E, F, G, H -1.90
23
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
(a) 0.5 ML (b) 0.25 ML
(c) 0.2 ML (d) 0.125 ML (e) 0.11 ML
FIG. 1: Schematic top views of the surface unit cells for different CO coverages. The large shaded circles and the small white
ones denote Ir atoms on the topmost layer and CO molecules respectively. The unit cells are indicated by solid black lines:
(a) √ 2×√2, (b) 2×2, (c)√5×√5, (d) 2√2×2√2 and (e) 3×3.
24
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
b
b
b
b
b
b
b
b
b
1
1
1 43
23
13
2 2
3 3
4 4
13
13
13
23
23
23
p p p p1 1 1 1 2 23 3
p p p 2 2 2 2 p
2 233
(b)(a)
1 2 3
4 3
2 1
11
1 1
1 1
2 2
2 2
2 2
3 3
3
3 3
FIG. 2: (a) Exact top view and (b) schematic side view of the reconstructed Ir{100} surface. In (a) the grey spheres are atoms
in the topmost layer, whereas the black ones are atoms in the second layer. The surface unit cell is demarcated by the solid
black rectangle. The horizontal and dashed lines denote the planes of reflection of the unit cell. In (b), bi3n is the distance along
the z-direction between atoms i and 3 in the nth layer and p denotes the lateral shift of the atoms from their bulk truncated
position. The arrows denote the direction of the shift. This is a schematic diagram. The actual values of different parameters
are listed in Table II.
25
0 0.01 0.02 0.03 0.04 −2.7
−2.65
−2.6
−2.55
−2.5
−2.45
−2.4
(⊄2x⊄2) unit cell with 0.5 ML CO coverage
with BZC without BZC
E a ds
(eV pe
r C O
mo lec
ule )
1/nk
12X12
FIG. 3: Convergence of adsorption energy (Eads) of CO at 0.5 ML on the atop site of the unreconstructed surface, with
respect to the k-points used for √ 2×√2 unit cell. Monkhorst-Pack meshes of the form (n1 × n2 × 1) were used, meshes that
include/donot include the BZC have odd/even values of n1 = n2. For n1 = n2 = 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, and
18 the number of points “nk” in the whole BZ are 25, 36, 49, 64, 81, 100, 121, 144, 169, 196, 225, 256, 289 and 324 respectively.
Note that meshes that do not include the BZC (solid line) converge faster than those that include it (dashed line).
26
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
rIr−C ∆ Z
dC−O
d12 �������� ����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
FIG. 4: Schematic side view of the structure of CO adsorbed on unreconstructed Ir{100}-(2×2) at 0.5 ML coverage41. The
big, medium and small circles represent Ir, O and C atoms.
27
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
T1 H2 B2 H3 T3 H6 B4 D1 D2 D3
B1 H1 H4 B3 T4H5T2
FIG. 5: Schematic diagram showing different possible CO adsorption sites on the reconstructed Ir {100} surface (top view).
Shaded circles are atoms in the topmost Ir layer. B, D, H and T denote cross bridge, diagonal bridge, hex and atop sites
respectively. The sites are labelled according to the nomenclature used by Deskins et al.16. Note that T1, T2, T3 and T4
correspond to sites atop the atoms labelled 1, 2, 3 and 4 in Fig. 2.
28
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
A B
CD
E
F G H
FIG. 6: Schematic diagram showing different possible CO adsorption sites, within a (2× 5) supercell, on the reconstructed Ir
{100} surface (top view) at 0.2 ML and 0.6 ML CO coverage. Shaded circles are atoms in the topmost Ir layer. For 0.2 ML
coverage the adsorption sites are A and B, A and C, and A and D. Sites E, F, A, G, B and H are the sites for CO adsorption
at 0.6 ML CO coverage. Note that the positioning of the lateral boundaries for the supercell follows the same convention as
used in Fig. 6; i.e., E, F and A correspond to T1, T2 and T3 sites, etc.
29
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 CO coverage (ML)
0
0.5
1
1.5
2
Su rfa
ce e
ne rg
y (eV
/(1 x1
) s urf
ac e a
rea )
FIG. 7: Relative stability, at T = 0 K, of the reconstructed and the unreconstructed surfaces as a function of CO coverage.
The dashed line and the solid line denote the surface energy for the reconstructed and unreconstructed surface respectively.
The two lines cross in the neighborhood of 0.09 ML.
30
300 600 900 1200 1500 −1
0
1
2
3
300 600 900 1200 1500 −1
0
1
2
300 600 900 1200 1500 −1
0
1
2
300 600 900 1200 1500 −1
−0.5
0
0.5
1
δF (e
V/ (1
x1 ) s
urf ac
e a rea
)
Temperature (K)
(a) (b)
(c) (d)
FIG. 8: Free energies relative to that of the clean unreconstructed surface (δF ) as a function of temperature at different
pressures: (a) 10−8 mbar, (b) 10−3 mbar, (c) 1 mbar and (d) 102 mbar. The lines without any symbols on them denote the
unreconstructed surface and those with open circles denote the reconstructed one. The solid line, the short-dashed line, and
the long-dashed line represent surfaces at a coverage of 0, 0.1 and 0.2 ML respectively. The dot-dashed line and dot-dashed line
with open circles represent the unreconstructed surface at 0.5 ML CO coverage and the reconstructed surface at 0.6 ML CO
coverage respectively. Note that we have dropped the superscripts S and µ from F since both the approaches (as described in
the text) produce the same results. It can be seen that in all the cases, there are only two stable phases, viz., the dot-dashed
line (0.5 ML CO on unreconstructed surface) and the solid black line with circles (clean reconstructed surface).
31
400 600 800 1000 1200 1400 10−10
10−8
10−6
10−4
10−2
100
102
104
Pr es
su re
(m ba
r)
Temperature (K)
Surface remains clean (5x1) reconstructed
Surface saturates 0.5 ML CO c(2x2) overlayer on (1x1) substrate
10
FIG. 9: Temperature-pressure phase diagram for CO on Ir{100}. The black curve dividing the two phases gives the critical
temperature at which the crossover in stability of the two phases occurs, as a function of pressure.
iv :c
on d-
m at
/0 70
15 97
v1 [
co nd
-m at.
mt rl-
sc i]
24 Ja
n 2 00
7 Lifting of Ir{100} reconstruction by CO adsorption:
an ab initio study
Prasenjit Ghosh and Shobhana Narasimhan
Theoretical Sciences Unit,
Jawaharlal Nehru Centre for Advanced Scientific Research,
Jakkur, Bangalore 560 064, India
Stephen J. Jenkins and David A. King
Department of Chemistry, University of Cambridge,
Lensfield Road, Cambridge CB2 1EW, U.K.
(Dated: February 6, 2008)
The adsorption of CO on unreconstructed and reconstructed Ir{100} has been studied, using a
combination of density functional theory and thermodynamics, to determine the relative stability of
the two phases as a function of CO coverage, temperature and pressure. We obtain good agreement
with experimental data. At zero temperature, the (1×5) reconstruction becomes less stable than
the unreconstructed (1×1) surface when the CO coverage exceeds a critical value of 0.09 ML.
The interaction between CO molecules is found to be repulsive on the reconstructed surface, but
attractive on the unreconstructed, explaining the experimental observation of high CO coverage on
growing (1× 1) islands. At all temperatures and pressures, we find only two possible stable states:
0.05 ML CO c(2× 2) overlayer on the (1×1) substrate, and the clean (1×5) reconstructed surface.
2 I. INTRODUCTION
Automotive exhausts contain several noxious gases, such as CO and NO, which need to be converted to less harmful
products (e.g., CO oxidized to CO2, and NO reduced to N2) before they are released into the atmosphere. The surfaces
of metals like Pt, Ir and Rh are good catalysts for such reactions. A better understanding of the catalytic activity
of these surfaces could lead to the development OF cheaper and better catalysts, and these reactions have therefore
been the subject of extensive experimental and theoretical studies1.
In response to reduced coordination at the surface, the {100} surfaces of the face centered cubic (fcc) metals Au,
Pt and Ir reconstruct to form a corrugated quasi-hexagonal overlayer (hex) on top of the square fcc substrate. While
Pt{100} and Au{100} display complex periodic patterns with large unit cells2, Ir{100} has a similar structure, but
with a comparatively small (1×5) unit cell. Low energy electron diffraction (LEED) experiments3,4,5,6 and density
functional theory (DFT) calculations7 on Ir{100} have shown that the stable overlayer registry with respect to the
substrate is in accordance with the “two-bridge model”. The lack of registry between overlayer and substrate atoms
results in a significant buckling in the two topmost layers6, as well as a lateral shift in the second layer. Moreover, the
reconstruction appears to go down deep, into at least the fourth layer8. Though most previous studies have focused
on adsorption and reaction mechanisms on Pt{100} rather than Ir{100}, in this work we choose to focus on Ir{100},
since its smaller unit cell makes it more accessible to ab initio calculations. However, the structural similarity of the
reconstructions on Pt{100} and Ir{100} suggests that the chemical reactions and physical processes taking place on
the two surfaces might have similar mechanisms.
Surface adsorbates such as CO, NO and O2 are known to lift the reconstruction on Pt{100} and Ir{100}. For
Pt{100}, the adsorbate-induced lifting of the reconstruction has been widely investigated9,10,11,12,13. There have also
been a few studies on the lifting of reconstruction on Ir{100} by molecular adsorbates14,15,16,17,18. Measurements of
heats of adsorption, on both unreconstructed and reconstructed Pt{100} and Ir{100}, by Hopkinson et al.11, Yeo et
al.17 and Ali et al.14, have shown that the binding energy of CO is greater on the unreconstructed surface than the
reconstructed one; this provides the driving force for the lifting of the reconstruction upon adsorption of CO.
The catalytic oxidation of CO on metal surfaces is certainly one of the most important catalytic processes studied in
surface science. When the catalyst is Pt{100} or Ir{100}, the CO-induced lifting of the reconstruction forms a crucial
component of the catalytic cycle, since the surface alternates rapidly between being CO-rich and (after the CO has
combined with oxygen to form CO2) being clean. Thus, the surface alternates also between being unreconstructed
and reconstructed, and the thermodynamics and kinetics governing this process are of great interest if one wishes to
3 gain a better understanding of the catalytic cycle.
The parameters that characterize the process by which the reconstruction is lifted are: (i) the adsorbate coverage
on the metal surface, (ii) the nucleation of unreconstructed (1× 1) islands, and (iii) the growth rate of these islands.
In the following paragraphs, we summarize the present state of knowledge regarding these three parameters:
(i) Coverage: What is the coverage of CO on the (1×1) islands, and at what critical coverage is the lifting initiated?
Is the critical parameter the global CO coverage, or is it a local CO coverage (that differs from the average value)
that is important? In experimental investigations of the restructuring process on Ir{100} and Pt{100}, it has been
reported that the local CO coverage on the growing (1×1) islands is 0.5 ML. However, the critical value of the total
CO coverage, for the onset of the transition, has been found to be much lower than this, on both Ir{100} and Pt{100}.
For example, thermal energy atomic scattering measurements on Ir{100} and Pt{100} suggest that the restructuring
begins somewhere between 0.05 and 0.13 ML on Ir{100}14, and between 0.01 and 0.03 ML on Pt{100}11; while electron
energy loss spectroscopy measurements by Behm et al.19 and Rutherford back scattering measurements by Jackman
et al.20 on Pt{100} have indicated that the lifting of the reconstruction is initiated at a critical CO coverage of 0.05
ML and 0.08 ± 0.05 ML respectively. When the total CO coverage is 0.5 ML, the entire surface appears to be in the
(1×1) phase.
(ii) Nucleation: It is not clear from the literature whether the nucleation of (1×1) islands is homogeneous or
heterogeneous. On the one hand, the finding of similar CO adsorption energies on steps and terraces reported by
Hopster et al.21, indicates that the nucleation may be homogeneous. Using scanning tunneling microscopy (STM),
Ritter et al.12 have proposed that homogeneous nucleation of the (1×1) islands takes place due to fluctuations in
the density of CO molecules. According to them, when the islands grow beyond a critical size, they become stable
and act as nucleation centers. The spatial progress of the transformation occurs when the rate of growth of the
islands becomes large compared to their rate of nucleation. On the other hand, STM studies by Borg et al.22 suggest
that the restructuring is initiated by heterogeneous nucleation, the nucleation centers being step edges and structural
irregularities disrupting the hexagonal structure along a direction close to the [1 5] direction of their [N 1 ; 1 5]
reconstruction. In agreement with this, molecular dynamics (MD) simulations by van Beurden et al.13,23 of the lifting
of the reconstruction on Pt{100}, at CO coverages between 0.4 and 0.5 ML, indicate that the transformation is
heterogeneously nucleated at step edges aligned along the [011] direction.
(iii) Growth rate: From their molecular beam experiments, Hopkinson et al. deduced how r1×1, the growth rate of
the (1×1) islands, depends on the local CO coverage on the hex-surface (ΘhexCO ). They obtained a non-linear variation
4 of the growth rate: r1×1 = k(Θ hex CO )
nΘhex, where k is independent of Θ hex CO , Θhex is the fraction of surface remaining
in the reconstructed form, and the reaction order n was found to be 4.5± 0.4. From this, they concluded that 4 to 5
CO molecules must be involved cooperatively in the growth of the (1×1) phase. Similar studies by Ali et al.14 of the
CO-induced lifting of reconstruction on Ir{100} also showed a power law behavior for the growth of (1×1) islands,
with 3.9 ≤ n ≤ 5.8. Recent time-resolved reflection absorption infrared spectroscopy (RAIRS) results indicate a
similar power law relationship by an alternative, and very direct, technique24.
For the Pt surface, the presence of a non-linear term in the growth law appears to be crucial to the appearance of
oscillations in many catalytic processes, e.g., CO oxidation with NO25,26, CO oxidation with O2 27,28 and NO reduction
with H2 29,30. Under reaction conditions, the catalyst surfaces undergo transformation; these are non-equilibrium
processes. Moreover the reaction rate may not remain constant but changes periodically or exhibits chaotic behavior.
There can also be the formation of spatial patterns on the catalyst surface31, and these have been modelled recently
within a sophisticated reaction-diffusion scheme32. Although we are not aware of any literature concerning oscillatory
reactions on Ir{100}, the experimental findings by Ali et al. suggest that such oscillatory behavior may be present at
temperatures above 900 K.
What is the underlying mechanism that is responsible for the non-linear growth law? Hopkinson et al.10 and Ali
et al.14 have proposed a simple mechanism for the restructuring process which involves a cooperative phenomenon
among CO molecules: due to statistical fluctuations of the local CO coverage on the hex phase, 4-5 CO molecules
come together at the boundary of the growing (1×1) domain, or at a step, and convert 8-10 Pt atoms from a hex
to square arrangement; however it is not clear why 4-5 CO molecules are needed or precisely how they cooperate.
One explanation for the ‘magic’ value has been suggested by Passerone et al.33, who performed MD simulations on
Au{100}, and found that when islands/craters are formed by adsorbing/desorbing Au atoms, they do not remain
stable unless they exceed a critical size of 8-10 Au atoms. However, it is not clear that this number will translate
to Pt or Ir surfaces, especially in the presence of CO. The restructuring of Pt{100} has also been studied in more
recent MD simulations by van Beurden and coworkers13, who have found that the CO molecules initiate surface
relaxations and lead to a shear tension between adjacent [011] rows due to the preference for a square rather than
a hexagonal coordination sphere, and the restructuring proceeds through the ejection of chains of Pt atoms, and a
rearrangement of the remaining surface atoms. Their simulations showed no evidence of a cooperative phenomenon
between 4-5 CO molecules, and the source of the non-linear growth law remains a mystery. However, it is important
to note that their MD simulations were done at much higher CO coverages (0.4-0.5 ML) than the critical coverage at
5 which the experiments showing the fourth-order power law dependence of the rate of (1×1) formation were conducted.
The mechanism for the lifting of the reconstruction is clearly not the same for low and high CO coverage. At low
coverage, lifting of the reconstruction occurs when random statistical fluctuations bring about a sufficiently high local
coverage to instigate nucleation of (1×1) islands; at the high total coverages studied in the MD simulations, such
high local coverage would be found across the whole surface at all times. The experimentally-observed power law
dependence stems from the probability of bringing together 4-5 CO molecules within a sufficiently small area at low
overall coverage.
As a first step towards quantifying these arguments theoretically, it is desirable first to understand the thermody-
namics, i.e., at what coverage the phase transition becomes thermodynamically favorable. In this work, we determine
this quantity through the study of the thermodynamics of the reconstructed and unreconstructed surfaces using ab
initio DFT. We then extend these results to finite temperatures and pressures, by making use of the chemical potential.
This is useful because it brings the results into regimes where it becomes possible to compare with experiment.
Our work is similar in spirit to a recent DFT study by Deskins et al.16, who showed that the adsorption of atomic
oxygen on Pt{100} makes the unreconstructed surface more thermodynamically stable than the reconstructed one.
In their work, the reconstructed structure of Pt{100} was approximated by a (1 × 5) unit cell. Note however that:
(i) our focussing on Ir{100} enables us to use a relatively small unit cell that corresponds to the true reconstructed
structure; (ii) we have studied the adsorption of CO; (iii) details of our analysis are also different.
The rest of this paper is structured as follows: Section II provides some details about our first-principles calculations.
Sections III A and III B contain the results of our calculations on the clean Ir{100} surface (both unreconstructed
and reconstructed), while our results for CO adsorbed on the unreconstructed and reconstructed Ir{100} surfaces
are contained in Sections III C and III D respectively. We emphasize that it is particularly important to choose k-
point meshes (used for Brillouin zone integrations) very carefully; this issue is discussed in Section II, while some
illustrative examples are presented in III C. The ab initio density functional theory results feed into an analysis using
the entropy and the chemical potential, presented in Section III E. Finally, we discuss the implications of our results
and summarize in Section IV.
II. DETAILS OF AB INITIO CALCULATIONS
It is known that the lifting of the reconstruction in the systems under study is governed by very small differences in
the energies of the competing structures; it is therefore vital to do as accurate a calculation as possible. For this reason,
6 we have chosen to perform ab initio calculations within the framework of density functional theory34 (DFT), since
this is perhaps the most reliable method currently available for obtaining accurate values of ground state properties
such as structures, surface energies and adsorption energies.
Our calculations have been performed using the CASTEP package, wherein the Kohn-Sham35 equations are solved
iteratively by conjugate gradient minimization.36 We have used a plane wave basis set, with a cut-off of 25 Ry (340)
eV, and ultrasoft pseudopotentials. The pseudopotentials for Ir, C and O are Ir OO.usp, C OO.usp and O OO.usp
respectively which are provided along with the CASTEP distribution (Version 4.2). For the exchange-correlation
interactions, we have used the Perdew-Wang form of the generalized gradient approximation (GGA)37; note that
earlier calculations7 showed that gradient corrections are essential in properly describing this surface, and that use
of the local density approximation (LDA) instead of the GGA would incorrectly predict clean Ir{100} to be stable
against reconstruction. Integrations over the Brillouin zone have been evaluated with a Monkhorst-Pack (MP)38 mesh
(further details of which are given below), along with a Gaussian smearing function of width 0.1 eV.
In order to test the reliability of the pseudopotentials used in our calculations, we first performed calculations on
bulk Ir and a CO molecule in the gas phase. For the former we obtain a lattice constant of 3.86 A˚, which is in excellent
agreement with the experimental value of 3.84 A˚, while for the latter, we obtained a C-O bond length of 1.14 A˚which
also agrees very well with the experimental value of 1.13 A˚39.
For surface calculations, we use a supercell consisting of a slab of six layers of Ir atoms separated by a vacuum
thickness of about 10 A˚. The top four Ir layers are allowed to relax their positions, whereas the bottom two are
fixed at the bulk separation (‘asymmetric’ slab). Further, only the top surface is allowed to reconstruct; note that
the density of atoms is different on a reconstructed and unreconstructed surface. Thus, all such asymmetric slabs
will contain one surface that is of interest to us, while the other side consists of a bulk-truncated (unrelaxed and
unreconstructed) Ir{100} surface. In order to determine the surface energy of the latter, we also perform a calulation
on a ‘symmetric’ slab, comprised of eight layers, of which the middle two are fixed at the bulk spacing, while the top
three and bottom three layers are allowed to relax. A comparison of the symmetric and asymmetric slabs for the
unreconstructed structure enables one to determine separately the surface energies of a relaxed and bulk-truncated
(1× 1) surface.
The size of the surface unit cell used depends upon whether whether we are looking at an unreconstructed surface or
a reconstructed one, and what CO coverage we are considering. Calculations for the clean unreconstructed surface were
carried out using both (1× 1) and (1× 5) cells; the results obtained with the two were almost identical. For the clean
7 reconstructed surface, we use a (1×5) cell. To study CO adsorption on the unreconstructed and reconstructed surface
we adsorb CO on only one side (the side which we are allowing to relax) of the ’asymmetric slab’. For adsorption of
CO on the unreconstructed surface, we consider coverages of 0.11, 0.125, 0.2, 0.25 and 0.5 ML. The unit cells used for
these calculations are shown in Fig. 1; note that (i) all the unit cells are square, i.e., the distance between adjacent
CO molecules is the same in both directions, and (ii) the cells for different coverages are not necessarily commensurate
with one another. Finally, for studying adsorption of CO on the reconstructed surface, we use a (2 × 5) cell, within
which we consider CO coverages of 0.1, 0.2 and 0.6 ML. (Note that all CO coverages in this paper are given with
respect to the density of atoms in the topmost layer of the unreconstructed surface.)
An important consideration is the choice of k-point meshes for Brillouin zone (BZ) sampling. For the surface cells,
we have used Monkhorst-Pack meshes, of the form (n1×n2×1), where n1 and n2 determine the fineness of the mesh.
This issue becomes particularly crucial in the present problem, since we are interested in computing energy differences
that are comparable to the errors introduced by incorrect (unconverged) BZ sampling. Such errors can be reduced
by using a k-point mesh that is as fine as possible, while keeping computational feasibility in mind. We have found
that, as expected, the convergence is faster when one uses a mesh that does not include high symmetry points [the
Brillouin zone center (BZC) and the k-points on the edges of the BZ]; this corresponds to choosing n1 and n2 to be
even numbers. Examples of this are presented below in Section IIID.
The specific choices made for n1 and n2 for the various surface cells used in the present paper are given in Table I;
we emphasize that convergence with respect to k-point sampling has been carefully established for all the cases. Note
also that computing surface energies and adsorption energies requires comparing the total energies of two systems
(bulk and clean surface, or clean and covered surface); in such cases, we are careful to make sure that the unit cells
and k-point meshes used for the two systems are either identical or related by folding, thereby further reducing the
errors introduced by incomplete k-point sampling.
III. RESULTS
A. Unreconstructed Clean Surface
Using the asymmetric six-layer slab described above, we find that the first interlayer distance d12 is contracted
(with respect to the bulk interlayer separation) by 6.14%. This result compares well with that obtained from ab initio
calculations by Ge et al.7(6.5%), but is larger than the reported experimental value (∼ 3.6%)40. Upon comparing the
8 total energy of this asymmetric slab with that of the bulk structure, we find that the sum of the surface energies of the
two surfaces (one relaxed and the other unrelaxed) is 3.01 eV per (1× 1) area. Next, using the symmetric eight-layer
slab, which possesses two relaxed surfaces, we determine the energy of the relaxed unreconstructed surface, Γrel1×1, to
be 1.50 eV per (1 × 1) area, while the energy of the unrelaxed (bulk-truncated) unreconstructed surface, Γunrel1×1 , is
1.51 eV per (1 × 1) area. This latter quantity is subtracted out when determining surface energies for slabs that are
reconstructed and/or have CO adsorbed on only one side.
B. Reconstructed Clean Surface
In the reconstructed surface, the topmost layer forms a quasi-hexagonal layer on top of the square substrate, as can
be seen in Fig. 2(a). The surface unit cell is (1× 5). The reconstruction results in buckling within the layers, as well
as lateral shifts of the atoms with respect to their bulk-truncated positions. The parameters used to specify these
structural rearrangements are indicated in Fig. 2(b), and the values we obtain for them are presented in Table II. It
can be seen that our results are in excellent agreement with those obtained from LEED and STM by Schmidt et al.8.
In accordance with their observations, we find that in addition to the reconstruction of the topmost layer, there are
significant lateral shifts and buckling in the three layers below. The only (minor) difference between our results and
theirs is that our calculations yield a very small lateral shift of the third atom in the second layer (p32) in a direction
opposite to that determined by them.
The value we obtain for the surface energy of the reconstructed surface is Γrel1×5 = 1.45 eV per (1 × 1) area, and
we thus correctly obtain the result that the clean surface would prefer to reconstruct. Note also that the energy of
reconstruction for Ir{100} is very small, viz. 0.05 eV/(1×1) area. This is in reasonably good agreement with the
value of 0.07 eV/(1×1) area obtained by Ge et al.. The slight difference between our results and theirs presumably
arise from the use of slabs of different thicknesses and different k-point meshes. (Note that the energy difference here
is somewhat smaller than the value of 0.21 eV/(1×1) area found experimentally for the Pt{100} surface17).
C. CO on unreconstructed Ir{100}
As discussed in Sections I and II, the choice of a proper k-point mesh is very crucial for our calculations. Even
values of n1 and n2 lead to better sampling of the BZ and quicker convergence of the adsorption energy. This fact
becomes evident on inspecting Fig. 3, where we have shown how the adsorption energy, for CO at 0.5 ML and using
9 a ( √ 2×√2) unit cell, converges as a function of k-point sampling. Accordingly, we use a (12× 12× 1) mesh for this
particular unit cell; similar checks were performed for other coverages and cells.
Experiments and previous theoretical calculations41 indicate that the atop site is the most probable site for CO
adsorption. To verify this, and compare our results with previous DFT calculations and LEED measurements,41, we
calculate the geometry and adsorption energies of CO molecules occupying hollow, bridge and atop sites within a
(2×2) surface unit cell, and at 0.5 and 0.25 ML CO coverages. The adsorption energy per CO molecule, Eads, is given
by:
Eads = Eslab+CO − Eslab − nCOECO
nCO , (1)
where Eslab+CO, Eslab and ECO are the the total energies of the slab with CO adsorbed on it, the clean slab, and a
CO molecule in the gas phase, respectively, while nCO is the number of CO molecules adsorbed per surface unit cell.
The results obtained by us for Eads, for different sites at the two coverages considered, are summarized in Table III.
Although there are slight numerical differences between the adsorption energies obtained from our calculations and
those of Titmuss et al.41, both sets of calculations predict that CO molecules adsorb on the atop site. The differences
in the numerical values of the adsorption energy obtained from the two calculations presumably arise from the use of
slabs of different sizes and different k-point mesh. The structural parameters are also in good agreement with both
the LEED measurements and theoretical calculations.
We go on to study the variation of the adsorption energy of CO on the atop site as a function of CO coverage. The
results are summarized in the second column of Table IV. The difference in adsorption energy between 0.5 ML and
0.25 ML CO coverage is 0.04 eV. For lower coverage, (Θ ≤ 0.25 ML) the adsorption energy is more or less constant
(∼ 2.51 eV). The slight variation in the numerical values of Eads is most likely due to numerical errors that arise
because the k-point meshes used for different surface unit cells are not exactly commensurate. At lower coverage,
the distance between CO molecules increases, decreasing the interaction between them. At sufficiently low coverage,
the CO molecules are so far apart that they do not interact with one another, resulting in a constant value of the
adsorption energy. The increase in Eads with increase in CO coverage indicates the presence of very weak attractive
interactions between nearest-neighbor CO molecules on the unreconstructed surface.
In addition to the CO adsorption energies, we also calculate the “surface energies” (Γs) at different CO coverages.
These “surface energies” will be used below as a measure of the stability of the reconstructed and unreconstructed
surfaces. We define the “surface energy” of the CO covered surface (Γs) as:
10
Γs = Eslab+CO − nIrEbulk − nCOECO
Ns − Γunrels , (2)
where Ns is the ratio of the area of the surface unit cell to that of the (1×1) cell, and the superscript unrel represents
the unrelaxed and unreconstructed lower surface of the slab.
D. CO on reconstructed Ir{100}
CO adsorption on the reconstructed surface has been studied for coverages of 0.1, 0.2 and 0.6 ML. In all the cases,
we use a (2×5) supercell, so as to minimize the interaction between periodic images of CO molecules in adjacent
supercells. Since the reconstructed surface is quasi-hexagonal in nature, there are many possible adsorption sites.
The different adsorption sites which we consider for CO adsorption at 0.1 ML coverage are shown in Fig. 5. Our
result for the adsorption energies for these different sites are given in table V. Of all the different possibilities, we
find that the “atop3” site (T3) is the most probable one. While it is not surprising that CO prefers an atop site, it
is somewhat unexpected that the most favoured site is atop the Ir atom that is lies lowest within the buckled surface
layer, rather than the Ir atom that protrudes. We intend to address the origin of this very surprising result in a future
publication.
For a coverage of 0.2 ML, we have to adsorb two CO molecules in the (2×5) cell. Since CO clearly prefers to adsorb
on the “atop3” (T3) sites, we choose a combination of two T3 sites. There are three inequivalent combinations of two
T3 sites in the (2×5) cell, namely, A and B, A and C, and A and D (see Fig. 6). Our results for the CO adsorption
energies at these different sites are listed in Table V. Out of these three possible combination of sites, we find similar
adsorption energies for the AB and AC combinations. The fact that the AD combination is disfavored suggests that
CO molecules on this surface would prefer not to sit too close to each other. Comparing the adsorption energies
for coverages of 0.1 and 0.2 ML, we find that the interaction between CO molecules on the reconstructed surface is
repulsive in nature, in contrast to the very weak attractive interaction on the unreconstructed surface. For a coverage
of 0.6 ML, we need to adsorb six CO molecules per (2×5) cell. Since the CO molecules interact repulsively on the
unreconstructed surface, they will tend to spread out uniformly at low temperatures. Hence we assume that at a
coverage of 0.6 ML, CO will adsorb at every atop site, as shown in Fig. 6.
The variation, with coverage, of the “surface energies” of the reconstructed and unreconstructed surfaces (computed
using Equation. (2)), is plotted in Fig. 7. At zero CO coverage, the reconstructed (1×5) surface is energetically more
11
stable than the unreconstructed (1×1). However, at 0.1 ML coverage the unreconstructed surface is more stable than
the reconstructed one by about 0.03 eV per (1×1) area. The crossover between the stability of the two surfaces takes
place at around 0.09 ML CO coverage. Thus adsorption of CO switches the stability of the Ir{100} surfaces at a very
low total coverage, that appears to be in very good agreement with the experimentally reported values of Hopkinson et
al.. However, we note that this analysis has been done at conditions corresponding to zero temperature and pressure.
In the next section, we extend this results to finite temperatures and pressures, by performing a thermodynamic
analysis.
E. Thermodynamic Analysis
In order to account for the effect of varying gas-phase temperature and pressure upon the surface configuration, we
apply concepts from classical thermodynamics. The free energy (F ) of a system gives a measure of the stability of
the system. For a multi-species system F is given by:
F = E + PV − TS − ∑
i
niµi, (3)
where P , V , T and S are the pressure, volume, temperature and entropy, respectively. ni and µi are, respectively,
the number of molecules and the chemical potential of species i at the surface. E is the internal energy of the system
and is obtained from the DFT calculations.
In order to apply these concepts we need to have a clear demarcation between that which we consider to be a part
of the “system” and that which we take to be its “surroundings”. For our case this can be achieved in two ways:
(i) entropy viewpoint, in which the “system” is considered to include the surface, the adsorbed molecules on the
surface and the gas above the surface; the “surroundings” include only the world beyond some external gas container.
Taking this view, the number of particles in the system is now fixed but the entropy is considerably large, since it
includes large translational and rotational contributions from the gas molecules which are not adsorbed on the surface.
Hence Equation (3) is modified as:
FS = E + PV − TS. (4)
From this viewpoint, the changes in the configuration of the surface are driven by massive changes in entropy between
molecules on the surface and in the gas phase; net adsorption (desorption) occurs when the enthalpy of adsorption is
12
greater (less) than this entropy difference.
(ii) chemical potential viewpoint, in which the “system” consists of the surface and any molecules adsorbed on the
surface. The rest of the world beyond then constitute the “surroundings”. Now to a fair approximation the entropy
of the system is more or less negligible, since it consists only of vibrational contributions. On the other hand, the
number of particles in the system is variable, so the chemical potential of the gas molecules must be included. The
thermodynamic potential F now becomes:
Fµ = E + PV − ∑
i
niµi. (5)
The changes in surface configuration are driven by the need to maintain the equality of surface and gas-phase chemical
potentials; net adsorption (desorption) occurs when the gas-phase chemical potential is instantaneously greater (less)
than the surface chemical potential. Note that from both the approaches one should get the same results.
1. Entropy Viewpoint
Taking the viewpoint that our thermodynamic system includes the surface, adsorbed molecules and the gas above
it, we calculate the necessary free energy that must be minimized. This is achieved by defining free energies relative
to that of the clean unreconstructed surface:
δFS = FSslab+CO − FSclean(1×1) −∆nIrFSbulk − nCOFSCO
Ns , (6)
where FS clean(1×1) and F
S slab+CO have been evaluated for the same unit cell, thus mitigating any systematic errors that
may arise due to different k-point sampling. To evaluate the free energy of the solid phases, slab with CO molecules
adsorbed on it (FSslab+CO) and for the clean one (F S clean(1×1)) we ignore the effects of TS because typically the entropy
of the solid phase is negligible compared to that of the gas phase. For the solid phases we ignore the effects of PV and
approximate it for CO in the gas phase, by assuming that the gas phase obeys the ideal gas equations. With these
approximations the free energy of the solid phases equals the internal energy obtained from DFT calculations. The
third term in Equation (6) contains the free energy per atom in the Ir bulk (FSbulk) and the difference in the number
of Ir atoms (∆nIr) between the DFT calculations used to obtain the first two terms. In the last term, F S CO is the
free energy of the CO molecules in the gas phase. The last term accounts for the decrease in entropy of the system
due to the adsorption of CO molecules from the gas phase. Ns represents the number of (1×1) cells over which the
13
value of FS has been calculated. By construction, therefore, δFS for the clean reconstructed surface is the negative
of the reconstruction energy, while for the clean unreconstructed surface δFS = 0. The lowest value of δFS under
any temperature and pressure conditions indicates the most thermodynamically stable surface configuration, δFSmin.
In order to evaluate FSCO we need to know the value of S for the gas phase at different temperatures and pressures.
The values of S at a pressure of 1 bar and for a temperature range of 298.15 to 1500 K can be obtained from the
CRC handbook39. We then derive the variation of S with pressure and temperature from the following formula:
S(T, P ) = S(T, P 0)−Rln(P/P0). (7)
P 0(=1 bar) is the reference pressure and R is the universal gas constant. In Fig. 8, we plot a series of lines
calculated at a variety of pressures, each plotting the value of δFS as a function of temperature. It is apparent
that for all temperatures and pressures considered, there are only two possible configurations of the surface that
are thermodynamically stable, the clean unreconstructed (1×5) surface and the 0.5 ML c(2×2) CO overlayer on the
unreconstructed substrate. The critical temperature at which the crossover in the stability occurs varies as a function
of pressure and is depicted in Fig. 9.
2. Chemical Potential Viewpoint
In order to evaluate the relative stability of different surface configurations from the chemical potential viewpoint,
we evaluate the free energy as given in by Equation (5). The chemical potential can be obtained by making use of its
relationship to the enthalpy (H) and entropy:
µi = 1
ni (H − TS). (8)
The values of entropy at standard pressure can be extracted from the CRC Handbook39. The CRC Handbook lists only
the change in H at P 0 in going from absolute zero temperature to a variety of finite temperatures (∆H(0→ T, P 0).
Therefore H(T, P 0) can be extracted using the following relation:
H(T, P 0) = H(0, P 0) + ∆H(0→ T, P 0) (9)
where H(0, P 0) can be obtained from DFT calculations. The variation of µi with pressure is computed as:
14
µi(T, P ) = µi(T, P 0) + kBT ln(P/P
0), (10)
where kB is the Boltzmann constant. Again we evaluate δF µ the same way as in Equation (6) and plot the same
quantities as evaluated in the Section III E 1. The results obtained via the chemical potential approach are exactly
the same as those obtained via the entropy approach, plotted in Fig. 8 and 9.
IV. DISCUSSION AND SUMMARY
As in previous calculations, we find that Ir{100} is reconstructed; the stabilization energy being very small, viz.
0.05 eV per (1×1) area. Note that this is comparable to the errors typically introduced by incorrect k-point sampling,
emphasizing the need for careful calculations. The structure of the reconstructed surface obtained by us is in excellent
agreement with experiment. While the increase in density is confined to the topmost layer, the three layers below
also display marked buckling and lateral shifts.
However, as more and more CO is adsorbed on the surface, there is a reversal in phase stability. Our calculations
show that at zero temperature and pressure, when CO is adsorbed on Ir{100}, the relative stability of the reconstructed
and unreconstructed phases is reversed at a very low total CO coverage of 0.09 ML. This shows that the lifting of the
reconstruction is thermodynamically favoured at a low local coverage. The need for statistical fluctuations to bring
4-5 molecules together to trigger the lifting of the reconstruction is kinetic in origin, rather than thermodynamic.
On the unreconstructed Ir{100} surface, CO adsorbs at atop sites, and there is a very weak attractive interaction
between CO molecules. Note that in this respect Ir{100} appears to behave differently from Pt{100}, where there
is a repulsive interaction between CO molecules19,42. The attractive nature of the CO-CO interactions on the unre-
constructed (1×1) surface suggests that at sufficiently low temperature, CO molecules would tend to cluster together
on being adsorbed on the fully unreconstructed surface in order to maximize their coverage, until a saturation cov-
erage of 0.5 ML is achieved. The weakness of the interaction, however, means that such clustering may not easily
be observed in experiments performed at moderate temperature. In contrast, the interaction between CO molecules
adsorbed on the reconstructed (1×5) surface is repulsive, and they tend to spread out uniformly over the surface.
Including the cost of lifting the reconstruction, the adsorption energy of CO molecules at any total coverage to form
an island of 0.5 ML local coverage on a corresponding patch of (1×1) substrate (2.45 eV per CO molecule) is still
higher than the adsorption energy of CO molecules at any coverage on the unreconstructed surface. This suggests
15
that at any total coverage, the molecules will prefer to cluster into islands of local 0.5 ML coverage on patches of the
unreconstructed (1×1) substrate, rather than remain spread out on the reconstructed (1×5) substrate. This bolsters
the interpretation of 0.5 ML local CO coverage on the unreconstructed (1×1) islands by Ali et al. even at low total
CO coverage. The higher heat of adsorption and the attractive nature of the interaction between CO molecules on the
(1×1) unreconstructed surface provide a driving force to lift the reconstruction. For both Ir{100} (where the interac-
tion between CO molecules on the unreconstructed surface is weakly attractive) and Pt{100} (where this interaction
is weakly repulsive), the main effect is that the heat of adsorption on the unreconstructed surface far exceeds that
on the reconstructed surface. As a result, the reconstruction lifting phenomenon remains similar on both Ir{100}
and Pt{100}. However we donot understand the driving force behind the shift of the weakly attractive interaction
between CO molecules on the unreconstructed surface to a repulsive nature on the reconstructed one.
Our thermodynamic analysis extends the results obtained from DFT calculations to a range of pressures and
temperatures. For the entire temperature and pressure range considered by us, there are only two thermodynamically
stable configurations, namely 0.5 ML of CO on the unreconstructed substrate in a c(2×2) structure, and the clean
reconstructed substrate. The phase diagram shows that a small change in temperature and/or pressure can result
in a very large change in coverage. From experiments, we know that the lifting of the reconstruction sets in at 490
K at very low pressures (10−10-10−07 mbar). From our phase diagram we find that at similar pressure ranges, the
transition temperature lies between 550 and 600 K. Thus our results from thermodynamic analysis are reasonably
consistent with experimental observations. The slight discrepancies may have arisen from our use of the ideal gas
equation of state to describe the gas phase. In practical applications such as in catalytic converters in vehicles, the
reactions take place at temperatures above about 750 K and at ∼ 10 mbar pressure. According to our phase diagram,
such temperatures and pressures lie very close to the transition line. Hence slight changes in the reaction conditions
may lead to drastic changes in the surface, which in turn will affect the rate of chemical reactions.
To conclude, we have investigated the thermodynamic stability of unreconstructed and reconstructed phases of
the Ir{100} surface, in the presence of CO. Though our results for the critical coverage, etc., are in good agreement
with experiment, it still remains intriguing to speculate about the role played by kinetic factors in the restructuring
process. In particular, the origin of the nonlinear growth law observed in experiments is still a puzzle; we hope that
future work on the kinetics of the restructuring mechanism will shed light on this.
One of us (P.G.) acknowledges CSIR, India for a research scholarship. The Royal Society is also thanked for a
16
University Research Fellowship (S.J.J.). We are grateful to Dr. Stephanie Pratt for helpful discussions.
1 Z-P. Liu, Pure Appl. Chem. 76, 2069 (2004).
2 M. A. Van Hove, R. J. Koestner, P. C. Stair, J. P. Bibie´rian, L. L. Kesmodel, I. Bartoˇs, and G. A. Somorjai, Surf. Sci. 103,
189 (1981).
3 M. A. Van Hove, R. J. Koestner, P. C. Bibe´rian, L. L. Kesmodel, I. Bartoˇs, and G. A. Somorjai, Surf. Sci. 103, 189 (1981).
4 E. Lang, Mu¨ller, K. Heinz, M. A. Van Hove, R. J. Koestner, and G. A. Somorjai, Surf. Sci. 127, 347 (1983).
5 N. Bickel and K. Heinz, Surf. Sci. 163, 435 (1985).
6 K. Johnson, Q. Ge, S. Titmuss, and D. A. King, J. Chem. Phys. 112, 10460 (2000).
7 Q. Ge, D. A. King, N. Marzari, and M. C. Payne, Surf. Sci. 418, 529 (1998).
8 A. Schmidt, W. Meier, L. Hammer, and K. Heinz, J. Phys.: Condens. Matter 14, 12353 (2002).
9 P. A. Thiel, R. J. Behm, P. R. Norton, and G. Ertl, Surf. Sci. 121, L553 (1982).
10 A. Hopkinson, J. M. Bradley, X. C. Guo, and D. A. King, Phys. Rev. Lett. 71, 1597 (1993).
11 A. Hopkinson, X. C. Guo, J. M. Bradley, and D. A. King, J. Chem. Phys. 99, 8262 (1993).
12 E. Ritter, R. J. Behm, G. Po¨tschke, and J. Wintterlin, Surf. Sci. 181, 403 (1987).
13 P. van Beurden, B. S. Bunnik, G. J. Kramer, and A. Borg, Phys. Rev. Lett. 90, 066106 (2003).
14 T. Ali, B. Krotzer, A. V. Walker, and D. A. King, J. Chem. Phys. 109, 10996 (1998).
15 T. Ali, B. Krotzer, A. V. Walker, Q. Ge, and D. A. King, J. Chem. Phys. 109, 9967 (1998).
16 A. Deskins, J. Lauterbach, and K. T. Thomson, J. Chem. Phys. 122, 184709 (2005).
17 Y.Y. Yeo, C.E. Wartnaby and D.A. King, Science 268, 1731 (1995).
18 S. Khatua, Z.-P. Liu and D. A. King, Surf. Sc. 584, 214 (2005).
19 R. J. Behm, P. A. Thiel, P. R. Norton, and G. Ertl, J. Chem. Phys. 78, 7437 (1983).
20 T. E. Jackman, K. Griffiths, J. A. Davies, and P. R. Norton, J. Chem. Phys. 79, 3529 (1983).
21 H. Hopster, H. Ibach, and G. Comsa, J. Catalysis 46, 37 (1977).
22 A. Borg, A.-M. Hilmen, and E. Bergene, Surf. Sci. 306, 10 (1994).
23 P. van Beurden and G. J. Kramer, J. Chem. Phys. 121, 2317 (2004).
24 S. Khatua, S.J. Pratt and D.A. King, to be submitted.
25 A. Hopkinson and D. A. King, Chem. Phys. 177, 433 (1993).
26 A. Hopkinson and D. A. King, Faraday Discuss 6, 255 (1993).
27 M. Gruyters, T. Ali, and D. A. King, Chem. Phys. Lett. 232, 1 (1995).
28 M. Gruyters, T. Ali, and D. A. King, J. Phys. Chem. 100, 14417 (1996).
17
29 M. Gruyters, A. T. Pasteur, and D. A. King, J. Chem. Soc., Faraday Trans. 92, 2941 (1996).
30 A. V. Walker, M. Gruyters, and D. A. King, Surf. Sci. 384, L791 (1997).
31 G. Nicolis and F. Baras, Chemical Instabilities (Reidel, Dordrecht, 1984).
32 A.T. Anghel, R.B. Hoyle, I.M. Irurzun, M.R.E. Proctor and D.A. King, to be published.
33 D. Passerone, F. Ercolessi, and E. Tosatti, Surf. Sci. 454-456, 947 (2000).
34 P. Hohenberg and W. Kohn, Phys. Rev. 136, 864B (1964).
35 W. Kohn and L. J. Sham, Phys. Rev. 140, 1133A (1965).
36 M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, and J.D. Joannopoulos, Rev. Mod. Phys. 64, 1045 (1992).
37 Y. Wang and J. P. Perdew, Phys. Rev. B 43, 8911 (1991).
38 Hendrik J. Monkhorst and James D. Pack, Phys. Rev. B 13, 5188 (1976).
39 CRC Handbook of Chemistry and Physics, editor-in-chief David R. Lide, 84th edition (2003-2004).
40 K. Heinz and G. Besold, Surf. Sci. 125, 515 (1983).
41 S. Titmuss, K. Johnson, Q. Ge, and D. A. King, J. Chem. Phys. 116, 8097 (2002).
42 Y. Y. Yeo, L. Vattuone, and D. A. King, J. Chem. Phys. 104, 3810 (1996).
18
TABLE I: Parameters for the Monkhorst-Pack k-point meshes used for different surface cells.
Surface unit cell k-point mesh
1×1 10×10 √ 2×√2 12×12
2×2 8×8 √ 5×√5 8×8
2 √ 2×2√2 6×6
3×3 6×6
1×5 10×2
2×5 5×2
19
TABLE II: Structure of Ir{100}-1×5 described by the parameters defined in Fig. 2. dik’s denote average interlayer spacings
between the centre of mass planes of the layers, while dik’s give the smallest spacing between their subplanes. d b
ik describes the
percent change in dik with respect to the bulk interlayer spacing (db). Note that our calculations are in excellent agreement
with experiment, except for a slight difference in the lateral shift for the third atom in layer two (p32), which we find to be
negligible and in the opposite direction compared to the experimental findings.
Parameters Our Calculations Expt.8 Ge et al.7
db (A˚) 1.93 1.92 1.92
d12 (A˚) 1.96 1.94 1.97
d12 (A˚) 2.24 2.25 -
d b
12 (%) 16.09 16.67 -
d23 (A˚) 1.82 1.79 -
d23 (A˚) 1.88 1.88 -
d b
23 (%) -2.56 -2.08 -
d34 (A˚) 1.88 1.83 -
d34 (A˚) 1.94 1.93 -
d b
34 (%) 0.54 0.52 -
d45 (A˚) 1.91 1.89 -
d45 (A˚) 1.923 1.91 -
d b
45 (%) -0.34 -1.01 -
b131 (A˚) 0.22 0.25 -
b231 (A˚) 0.53 0.55 0.47
b341 (A˚) 0.2 0.20 0.20
p21 (A˚) 0.03 0.05 0.05
p31 (A˚) 0.07 0.07 0.02
b132 (A˚) 0.04 0.07 -
b232 (A˚) 0.08 0.10 -
p22 (A˚) 0.01 0.01 -
p32 (A˚) 0.0004 -0.02 -
b133 (A˚) 0.06 0.10 -
b233 (A˚) 0.03 0.05 -
b134 (A˚) 0.04 0.06 -
b234 (A˚) 0.01 0.03 -
20
TABLE III: Summary of our calculations for the c(2×2) and p(2×2) phases and comparison with experiments and previous
theoretical calculations (Prev)41. Eads denotes the adsorption energy of CO on Ir{100}; rIr−C and dC−O represent the Ir-C
and C-O bond lengths respectively. ∆Z gives the vertical distance between two intraplanar Ir atoms. d12 denotes the distance
between the top two surface layers. The different parameters are also shown in Fig. 4.
c(2×2) phase, 0.5 ML CO coverage
Site Eads rIr−C dC−O ∆Z d12
eV/ CO molecule (A˚) (A˚) (A˚) (A˚)
Ours Prev Expt Ours Prev Expt Ours Prev Expt Ours Prev Expt Ours Prev Expt
Atop -2.69 -2.65 - 1.86 1.86 1.81 ± 0.05 1.16 1.16 1.16 ± 0.05 0.14 0.19 0.13 ± 0.05 1.82 1.82 1.82 ± 0.04
Bridge -2.51 -2.29 - 2.06 2.04 - 1.17 1.18 - 0.0 0.0 - 1.89 1.90 -
Hollow -2.1 -1.5 - 2.29 2.28 - 1.2 1.2 - 0.0 0.0 - 1.90 1.92 -
p(2×2) phase, 0.25 ML CO coverage
Atop -2.83 -2.61 - 1.86 1.86 - 1.16 1.16 - 0.15 0.20 - 1.82 1.83 -
Bridge -2.71 - - 2.05 - - 1.17 - - 0.06 - - 1.84 - -
Hollow -2.34 - - 2.46 - - 1.2 - - 0.0 - - 1.84 - -
21
TABLE IV: Adsorption energies (Eads) of CO at atop sites on unreconstructed and reconstructed Ir{100} surfaces at different
CO coverages. All values of Eads are expressed in units of eV/CO molecule.
CO coverage Eads Eads
(ML) unreconstructed surface reconstructed surface
1.0 -1.93 -
0.6 - -1.90
0.5 -2.55 -
0.25 -2.51 -
0.2 -2.51 -2.27
0.125 -2.52 -
0.11 -2.53 -
0.10 - -2.31
22
TABLE V: Adsorption energies (Eads) at different sites of the reconstructed surface at 0.1 CO coverage. See Figs. 5 and 6 for
the convention used in labelling sites.
coverage site Eads
(ML) (eV/ CO molecule)
T1 -2.07
T2 -2.09
T3 -2.31
T4 -1.92
H1 -1.99
H2 -1.68
H3 -1.70
H4 -2.01
0.1 H5 -2.21
H6 -1.89
B1 -2.00
B2 -1.62
B3 -2.19
B4 -1.63
D1 -1.76
D2 -1.73
D3 -2.00
A and B -2.27
0.2 A and C -2.27
A and D -2.15
0.6 A, B, E, F, G, H -1.90
23
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
(a) 0.5 ML (b) 0.25 ML
(c) 0.2 ML (d) 0.125 ML (e) 0.11 ML
FIG. 1: Schematic top views of the surface unit cells for different CO coverages. The large shaded circles and the small white
ones denote Ir atoms on the topmost layer and CO molecules respectively. The unit cells are indicated by solid black lines:
(a) √ 2×√2, (b) 2×2, (c)√5×√5, (d) 2√2×2√2 and (e) 3×3.
24
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
b
b
b
b
b
b
b
b
b
1
1
1 43
23
13
2 2
3 3
4 4
13
13
13
23
23
23
p p p p1 1 1 1 2 23 3
p p p 2 2 2 2 p
2 233
(b)(a)
1 2 3
4 3
2 1
11
1 1
1 1
2 2
2 2
2 2
3 3
3
3 3
FIG. 2: (a) Exact top view and (b) schematic side view of the reconstructed Ir{100} surface. In (a) the grey spheres are atoms
in the topmost layer, whereas the black ones are atoms in the second layer. The surface unit cell is demarcated by the solid
black rectangle. The horizontal and dashed lines denote the planes of reflection of the unit cell. In (b), bi3n is the distance along
the z-direction between atoms i and 3 in the nth layer and p denotes the lateral shift of the atoms from their bulk truncated
position. The arrows denote the direction of the shift. This is a schematic diagram. The actual values of different parameters
are listed in Table II.
25
0 0.01 0.02 0.03 0.04 −2.7
−2.65
−2.6
−2.55
−2.5
−2.45
−2.4
(⊄2x⊄2) unit cell with 0.5 ML CO coverage
with BZC without BZC
E a ds
(eV pe
r C O
mo lec
ule )
1/nk
12X12
FIG. 3: Convergence of adsorption energy (Eads) of CO at 0.5 ML on the atop site of the unreconstructed surface, with
respect to the k-points used for √ 2×√2 unit cell. Monkhorst-Pack meshes of the form (n1 × n2 × 1) were used, meshes that
include/donot include the BZC have odd/even values of n1 = n2. For n1 = n2 = 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, and
18 the number of points “nk” in the whole BZ are 25, 36, 49, 64, 81, 100, 121, 144, 169, 196, 225, 256, 289 and 324 respectively.
Note that meshes that do not include the BZC (solid line) converge faster than those that include it (dashed line).
26
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
rIr−C ∆ Z
dC−O
d12 �������� ����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
����
����
����
����
����
����
����
����
����
����
����
����
���
���
���
���
���
���
���
���
���
���
���
���
FIG. 4: Schematic side view of the structure of CO adsorbed on unreconstructed Ir{100}-(2×2) at 0.5 ML coverage41. The
big, medium and small circles represent Ir, O and C atoms.
27
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
T1 H2 B2 H3 T3 H6 B4 D1 D2 D3
B1 H1 H4 B3 T4H5T2
FIG. 5: Schematic diagram showing different possible CO adsorption sites on the reconstructed Ir {100} surface (top view).
Shaded circles are atoms in the topmost Ir layer. B, D, H and T denote cross bridge, diagonal bridge, hex and atop sites
respectively. The sites are labelled according to the nomenclature used by Deskins et al.16. Note that T1, T2, T3 and T4
correspond to sites atop the atoms labelled 1, 2, 3 and 4 in Fig. 2.
28
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
���
A B
CD
E
F G H
FIG. 6: Schematic diagram showing different possible CO adsorption sites, within a (2× 5) supercell, on the reconstructed Ir
{100} surface (top view) at 0.2 ML and 0.6 ML CO coverage. Shaded circles are atoms in the topmost Ir layer. For 0.2 ML
coverage the adsorption sites are A and B, A and C, and A and D. Sites E, F, A, G, B and H are the sites for CO adsorption
at 0.6 ML CO coverage. Note that the positioning of the lateral boundaries for the supercell follows the same convention as
used in Fig. 6; i.e., E, F and A correspond to T1, T2 and T3 sites, etc.
29
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 CO coverage (ML)
0
0.5
1
1.5
2
Su rfa
ce e
ne rg
y (eV
/(1 x1
) s urf
ac e a
rea )
FIG. 7: Relative stability, at T = 0 K, of the reconstructed and the unreconstructed surfaces as a function of CO coverage.
The dashed line and the solid line denote the surface energy for the reconstructed and unreconstructed surface respectively.
The two lines cross in the neighborhood of 0.09 ML.
30
300 600 900 1200 1500 −1
0
1
2
3
300 600 900 1200 1500 −1
0
1
2
300 600 900 1200 1500 −1
0
1
2
300 600 900 1200 1500 −1
−0.5
0
0.5
1
δF (e
V/ (1
x1 ) s
urf ac
e a rea
)
Temperature (K)
(a) (b)
(c) (d)
FIG. 8: Free energies relative to that of the clean unreconstructed surface (δF ) as a function of temperature at different
pressures: (a) 10−8 mbar, (b) 10−3 mbar, (c) 1 mbar and (d) 102 mbar. The lines without any symbols on them denote the
unreconstructed surface and those with open circles denote the reconstructed one. The solid line, the short-dashed line, and
the long-dashed line represent surfaces at a coverage of 0, 0.1 and 0.2 ML respectively. The dot-dashed line and dot-dashed line
with open circles represent the unreconstructed surface at 0.5 ML CO coverage and the reconstructed surface at 0.6 ML CO
coverage respectively. Note that we have dropped the superscripts S and µ from F since both the approaches (as described in
the text) produce the same results. It can be seen that in all the cases, there are only two stable phases, viz., the dot-dashed
line (0.5 ML CO on unreconstructed surface) and the solid black line with circles (clean reconstructed surface).
31
400 600 800 1000 1200 1400 10−10
10−8
10−6
10−4
10−2
100
102
104
Pr es
su re
(m ba
r)
Temperature (K)
Surface remains clean (5x1) reconstructed
Surface saturates 0.5 ML CO c(2x2) overlayer on (1x1) substrate
10
FIG. 9: Temperature-pressure phase diagram for CO on Ir{100}. The black curve dividing the two phases gives the critical
temperature at which the crossover in stability of the two phases occurs, as a function of pressure.
Comments







