You are on page 1of 13

Article

pubs.acs.org/JPCC

Adsorption of CO2 at ZnO: A Surface Structure Effect from DFT+U


Calculations
Qian-Lin Tang* and Qing-Hong Luo
Department of Applied Chemistry, School of Advanced Materials and Nanotechnology, Xidian University, No. 2 South Taibai Road,
Xi’an 710071, People’s Republic of China

ABSTRACT: A complete understanding of the carbon dioxide (CO2)


interaction with zinc oxide (ZnO) is a basis for the development of
new ZnO-containing materials for catalytic fixation of CO2 into useful
chemicals such as methanol. In this work, the density functional theory
plus U (DFT+U) method coupled with periodic boundary conditions
has been employed to investigate the adsorption of CO2 molecules on
five relevant exposed surfaces of ZnO in an effort to capture the
surface chemistry of the adsorbed molecules. Through our calculations, we demonstrated that the CO2 adsorption is sensitive to
the oxide surface structure. The activation of CO2 by ZnO via the carbonate-ion formation requires the presence of empty
surface Zn−O pairs. It was found that the binding strength of CO2 follows the order of ZnO(0001̅) ≤ ZnO(0001) < ZnO(112̅1)
< ZnO(112̅0) ≤ ZnO(101̅0), which is, surprisingly, different from the decreasing stability sequence of these surfaces
(ZnO(101̅0) > ZnO(112̅0) > ZnO(0001̅)/ZnO(0001) > ZnO(112̅1)). The counterintuitive difference was discussed in detail
on the basis of the adsorption energy decomposition into two antagonistic effects: adsorbate−substrate interaction and their
deformation energies. The bonding nature of CO2 to the oxide substrates was analyzed, and the 2p states of CO2 were verified to
be able to mix both with the 4s orbital of surface zinc and with the 2p orbitals of surface oxygen. Our calculations indicate that
different from the previous viewpoint CO2, as an electronegative adsorbate, may also be positively charged by certain ZnO
surfaces, like the stepped (1121̅ ) facet presented here. Interestingly, the stronger the electron transfer between the two
interacting moieties, the stronger their interaction, and the stronger the CO bond activation. Also, it has been observed that
the change in the surface work function upon the adsorption of CO2 does not completely depend on the sign of charge
transferred.

1. INTRODUCTION the CH3OH formation process over Cu/ZnO-based catalysts at


Chemical fixation of carbon dioxide (CO2) through catalytic the atomic level needs detailed knowledge of how the CO2
hydrogenation has received considerable attention over the reactant interacts with the ZnO surface. Such a fundamental
recent decades1 because the reaction can not only produce raw question is also of paramount practical importance when CO2 is
materials for fuels and valuable chemicals2−8 but also may have utilized as a gas-phase probe molecule to characterize the
great potential as a useful technology to mitigate the emission surface structure and reactivity of freshly prepared ZnO
of the greenhouse gas into the atmosphere.1,8 Efficient catalysts nanoparticles.23,24
based on metal and metal oxide composites have been As a component of the industrial methanol synthesis
proposed for CO2 hydrogenation.7,9−12 For example, it was catalysts, ZnO adopts the stable hexagonal wurtzite struc-
demonstrated that the copper (Cu)/zincite (ZnO) based ture,25−27 in which each O2− ion is tetrahedrally coordinated
catalysts give high conversion and selectivity in hydrogenation with four Zn2+ ions along the crystallographic c-axis, and vice
of CO2 to methanol (CH3OH), the third-most large-scale versa.28 In a recent work of Diebold et al.29 using high-
chemical product, at pressures less than 100 bar with the resolution scanning tunneling microscopy (STM), it was
reaction temperatures in the range of 473−573 K.10,13−16 revealed that the surface morphology of wurtzite ZnO is
Although hitherto some uncertainties and controversy still strongly dominated by five low Miller index planes: the
remain in regard to the nature of the active centers on the nonpolar prism (101̅0) and (112̅0) faces as well as the polar
conventional Cu/ZnO catalysts,10,11,17,18 mostly agreement has (0001̅), (0001), and (112̅1) faces. The mixed-terminated
been reached about a pivotal role played by the formed Cu− (101̅0) and (112̅0) facets can be conceived as containing
ZnO interface in promoting CH3OH synthesis.10,19 As a matter exclusively neutral Zn2+−O2‑ ion pairs (hereinafter called Zn−
of fact, even without any metal phase or further additive, this O dimers), thus leading to the nonpolarity of the two flat
oxide itself is also an active catalyst for the synthesis reaction, in surfaces. The two basal planes O-terminated (0001̅)/Zn-
pronounced contrast with SiO2 and Al2O3.14,20 These indicate terminated (0001) are always created together by cutting the
that ZnO is capable of doing the CO2 activation process in the
Cu/ZnO systems. In the proposed hydrogenation reaction Received: August 9, 2013
mechanisms,14,21,22 the first and foremost step is the activation Revised: October 4, 2013
of the relatively inert CO2 molecule. Thereby, understanding

© XXXX American Chemical Society A dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

hexagonal crystal perpendicular to the c-axis, wherein the first the ones determined from TPD data by Burghaus et al.,36−39,55
surface is located on one side of the as-cleaved stoichiometric −0.38 to −0.32 eV.
sheet and, concurrently, the second surface is located on the Different from the two polar basal facets, the ZnO(101̅0)
other, opposite side. The stepped (112̅1) surface is constituted surface was unanimously believed to be capable of activating
by terraces of four atomic rows and monatomic steps, which are CO 2 through the generation of a surface carbo-
(1120̅ ) and (0001) orientations, respectively. All step edges are nate.24,31,33,40−48,53 However, there is a hot debate about the
terminated by 2-fold coordinated oxygen sites. More recently, bonding mode of CO2 with the nonpolar plane. By means of
Muhler and co-workers30,31 thoroughly elaborated the surface the near-edge X-ray adsorption fine structure (NEXAFS)
structure of polycrystalline ZnO nanoparticles with thermody- spectra, Davis and co-workers44 stated that CO2 is anchored
namic and kinetic methods, ultrahigh vacuum Fourier trans- to the oxide substrate through a bidentate linkage. However,
form infrared spectroscopy (UHV-FTIRS), and CO2 adsorp- recent theoretical researches have countered this argument. For
tion measurements. They concluded that these particles exhibit the adsorption system, ONIOM calculations53 obtained a
a quite significant surface population of structural defects such monodentate adsorption structure, while DFT slab computa-
as steps, edges, kinks, and vacancies. tions45 found a tridentate form as the most stable binding
In the past, extensive experimental16,24,31−48 and a few geometry. The previously published chemisorption energies of
theoretical efforts45,46,49−54 relevant to the adsorption of CO2 CO2 on the prism surface are divergent as well. According to
on ZnO have been focused on the (0001)̅ ,16,24,31,32,35,49−51,54 the Freundlich-type adsorption isotherms, Hotan et al.47 found
(0001),24,31,33,35−39,52,54,55 and (101̅0) facets.24,31,33,40−48,53 the adsorption energy for CO2 over ZnO(101̅0) at various
Nevertheless, only limited, often conflicting, information is coverages to lie between −1.45 and −0.67 eV. In this reference,
available for the adsorption geometry and energy.56 In what they further narrowed down the range to −1.04 to −0.83 eV
follows, a brief comment about the literature is made, which is using the TPD method.47 Afterward, the TPD experiment of
of direct importance for the discussion of our present results. Waugh and co-workers,35,43 however, yielded much stronger
The interplay of CO2 and a perfect (defect-free, pristine) adsorption energies (−1.59 to −1.13 eV). Again, the
corresponding theoretical values estimated by Taft et al.53
ZnO(0001̅) surface was carefully examined by Hirschwald et
and Wang et al.,45 −0.70 to −-0.47 eV, are surprisingly weaker
al.,16 who employed a combination of X-ray photoelectron
than those experimental data.43,47
spectroscopy (XPS) and ultraviolet photoelectron spectroscopy
The reasons for the existing discrepancies described above
(UPS). In their study, exposure to carbon dioxide was identified
are unclear at present30,34 but likely reflect the complex
to result in the formation of surface carbonate. Furthermore, interaction of CO2 with ZnO surfaces. It is important to note
FTIRS measurements from Lavalley et al.33 and first-principles that almost all of the prior theoretical studies45,46,49−54 of this
embedded cluster calculations from Koßmann et al.49 both interaction were based either on cluster models that suffer from
show that in the formed carbonate complex the CO2 moiety the size-dependent effect,57,58 or on the simple standard DFT
prefers a monodentate bent binding mode on the surface. By (without Hubbard U correction) that fails to properly address
contrast, Sokol and co-workers50 argued against the existence of the behavior of ZnO due to the self-interaction error.59,60
the carbonate species in the CO2/ZnO(0001̅) system based on Therefore, these calculation results might not be accurate. To
quantum mechanical/molecular mechanical embedded cluster the best of our knowledge, no detailed theoretical work has
results. The authors suggested that the surface CO2 is in a been dedicated to CO2 adsorption on the different ZnO
physisorbed linear state instead of a chemisorbed bent state. surfaces. In view of these situations, in this article, we presented
Their theoretical findings could be corroborated by STM,32 for the first time a systematic investigation of the CO2 binding
XPS,32 and UHV-FTIRS.24,31 These experiments additionally to various ZnO surfaces by the use of ab initio DFT+U
revealed that the carbonate species detected was ascribed to the calculations61,62 in the slab-supercell approximation, aimed to
activation of CO2 at surface defects of ZnO(0001̅), especially O pinpoint the influence of the oxide crystallographic orientation
vacancy sites. on the adsorption process. The post-DFT approach proves to
In the above-mentioned study by Lavalley’s group,33 they be able to study a large variety of strongly correlated
also used the FTIRS technique to evaluate CO2 adsorption on compounds with substantial improvement with respect to the
ZnO(0001) and demonstrated that at the Zn-terminate surface conventional DFT results.63 With several main low-index facets
the resulting surface species are only carboxylates and being considered as substrates, the surface structure sensitivity
physisorbed linear CO2 molecules. Analogously, on the basis in CO2 activation has been identified and rationalized. Such
of UHV-FTIRS spectra, Noei et al.31 ruled out the high substrate effects could provide a basis for further insight into
possibility that the perfect surface exhibits reactivity with CO2 the catalytic mechanism of CO2 fixation over Cu/ZnO-based
to generate carbonate. Consistently, temperature-programmed catalysts.
desorption (TPD) spectroscopy36,37 provided convincing
evidence for the absence of the carbonates species on 2. COMPUTATIONAL DETAILS
ZnO(0001). On the contrary, according to another UHV- 2.1. First-Principles Periodic Calculations. The calcu-
FTIRS analysis, Fischer and co-workers24 deduced that the lations were performed at the DFT+U level of theory using the
carbonate species is readily formed on the surface via the Vienna ab initio simulation package (VASP),64−68 in which the
binding of CO2 to the surface Zn atoms. Their result is in line valence electronic states were expanded with a plane-wave basis
with an earlier Austin model 1 (AM1) study with large cluster set. The exchange-correlation potential was treated within the
models.52 On the other hand, at the AM1 level the CO2 generalized gradient approximation (GGA) with the Perdew−
adsorption energy on ZnO(0001) is −8.46 eV,52 which is Wang 1991 (PW91) functional,69 known as GGA-PW91. Such
significantly larger in magnitude than that (−0.07 eV) obtained a gradient-dependent functional has proven to be reliable and
with the density functional theory (DFT) slab calculations.54 accurate for a wide variety of molecule−surface systems.70,71
Obviously, both theoretical values are not consistent at all with The plane-wave cutoff was 400 eV, which guaranteed a good
B dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Figure 1. Side views of the optimized bulk and surface structures of wurtzite ZnO considered in this study: (a) bulk ZnO crystal, (b) O-terminated
(0001)̅ with pseudohydrogen atoms of nuclear charge Z = 1/2, (c) Zn-terminated (0001) with pseudohydrogen atoms of Z = 3/2, (d) (1010̅ ), (e)
(112̅0), and (f) (112̅1) with pseudohydrogen atoms of Z = 1/2. All artificial atoms (small white balls) were attached to the bottom of the slab for
each polar surface. The medium red spheres stand for O atoms and the large blue spheres Zn atoms.

convergence of total energies. The 2s22p2, 2s22p4, and 3d104s2 × 1), (6 × 7 × 1), (7 × 7 × 1), and (6 × 7 × 1) were used for
electrons were included explicitly as the valence for the C, O, these surfaces, respectively.
and Zn atoms, respectively, whereas the remaining electrons Since this work is interested in adsorption at the same
were kept fixed as core states. To describe adequately the coverage on the five different surfaces, in all the calculations
strongly localized zinc 3d orbitals, the GGA+U scheme of there is a single CO2 molecule within the unit cell of each of
Dudarev et al.72 was applied, where the Coulomb U and them, corresponding to a coverage of 1/4. The coverage is
exchange J parameters were combined into a single effective defined as the ratio of the number of adsorbate molecules to
Hubbard U-parameter Ueff = U − J. The Ueff value for the zinc the total number of Zn and O atoms exposed per surface unit
centers was chosen according to a previous optimization59 of cell. The adsorbates were placed only on the top surface of the
the electronic band structure of bulk ZnO to be 7.5 eV. We slab, where a dipole correction was made to get rid of the
found that this value works well for calculating the surface work spurious interaction owing to the presence of nonequivalent
functions of ZnO (see below). The valence−core interaction surfaces.83,84 The top two ZnO layers, together with the
adsorbates, were relaxed up to the maximal force of less than
was represented by the full-potential projector augmented wave
0.03 eV/Å, while the remaining substrate atoms were frozen in
(PAW) formalism.73,74 The convergence tolerance for the self-
their ideal bulk positions. However, for the rather open (112̅1)
consistent electronic minimization was set to 10−5 eV/cycle. To
surface, which undergoes a notable relaxation, two more top
improve convergence, the tetrahedral smearing with Blöchl layers of the slab were allowed to move freely in three
corrections75 was employed to determine how the partial dimensions. All the optimized geometrical structures were
occupancies were set for the wave functions. The first Brillouin confirmed by subsequent vibrational analysis to have no
zone (BZ) was sampled using different Monkhorst−Pack (MP) imaginary vibrational frequencies. The numerical calculation
meshes76 containing the Γ point in all the calculations. of the second derivatives of the potential energy surface within
With a (12 × 12 × 7) k-point grid, the lattice parameters a, the harmonic approximation gave the vibrational frequencies
a/c, and u of ZnO bulk (Figure 1a) were computed to be 3.159 and corresponding normal modes of the adsorbed species. The
Å, 1.608, and 0.381, respectively, which agree well with previous central finite difference method with a geometrical displace-
GGA (Perdew, Burke, and Ernzerhof (PBE) functional77)+U ment of 0.015 Å was utilized for all the vibrational calculations.
(3.160 Å, 1.607, and 0.380)78 and experimental values (3.258 Å, 2.2. Calculation of Surface Energies. The stability of a
1.602, and 0.382).79,80 To simulate the common ZnO surfaces, particular crystal facet is characterized by the surface energy γ
we adopted periodically repeated supercells containing 5, 5, 6, which measures the work required to create the facet from the
5, and 10 formula unit layers of ZnO for the (0001̅), (0001), crystal.85,86 According to a traditional “bond-cutting” model
(101̅0), (112̅0), and (112̅1) surfaces, respectively. Their (BCM),87,88 the surface energy was quickly estimated as follows
structures are shown in Figure 1b−f. The surface dangling
bonds on the lowermost atomic layer of each polar surface were γ = [1 − (Zsurf /Z bulk )1/2 ]E bulk /A (1)
saturated by hydrogen-like atoms to eliminate an unphysical Here Zsurf and Zbulk are the average coordination number (CN)
transfer of charge between the top and bottom sides of the slab. of surface atoms and bulk, respectively; Ebulk denotes the bulk
Explicitly, the artificial atoms with a fractional nuclear charge of energy; and A is the surface area per atom. The computed Ebulk
1/2 and 3/2 were added for O- and Zn-terminated surfaces, value of ZnO is −7.15 eV, which is in agreement with the
respectively.81,82 Consecutive slabs were separated by 21 and 16 experimental result of −7.52 eV.80
Å of vacuum for polar surfaces and for nonpolar surfaces, A rigorous procedure for calculating the surface energy is
respectively, to reduce the interactions of surface states through based on the difference between the total energy of an N-layer
the vacuum region. To accommodate the adsorbates with small slab, ENslab, and that of atoms in bulk, i.e.89
lateral interactions between periodic images, the ZnO(0001̅), N
γ = (Eslab − NE bulk )/A (2)
ZnO (0001), ZnO(101̅0), ZnO(112̅0), and ZnO(112̅1) planes
were modeled by (2 × 2), (2 × 2), (2 × 1), (1 × 1), and (1 × Note that the symmetry of a polar slab makes it not possible to
1) unit cells, respectively. The MP grids of (6 × 6 × 1), (6 × 6 evaluate an absolute surface energy of either side of the slab.
C dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Thus, A in eq 2 was taken as the surface area of the surface unit 4−6, the adsorption energy can be resolved into the interaction
cell (not twice the area). This means that the computed γ value energy and two deformation energies via
is actually the sum of two complementary surface terminations. CO2 sub
It is desirable to rewrite eq 2 as Ead = E int + Edef + Edef (7)

N From this equation, it is obvious that the adsorption energy is


Eslab = NE bulk + Aγ (3) an adequate balance between the three competing terms.
which implies that ENslab manifests a linear dependence on the Because of the negative impact that the interaction contribution
thickness N of the slab if it is sufficiently thick. In practice, our usually has on the adsorption energy, the adsorbate−substrate
test calculations show that in the systems under study the interplay constitutes the driving force for any adsorption
relationship is essentially linear already for very thin slabs (with process. One should notice that the energy produced by the
correlation coefficients of larger than 0.9999). Therefore we can surface to adsorbate bond formation may be counterbalanced,
extract the surface energy by fitting a straight line to all of the to a very large extent, by the electronic and structural
calculated ENslab data versus N and then taking its intercept with relaxations of the whole system. In this situation, it is possible
the ordinate.90,91 Obviously, the incremental energy approach that the value of the final measured chemisorption energy
used in the present DFT+U simulations converges quite rapidly closely resembles typical values of physisorption energies
because no separately calculated bulk energy enters.89,91 All (−0.43 to −0.09 eV).94
ENslab values of each surface were acquired using a (1 × 1) unit
3. RESULTS AND DISCUSSION
cell. The MP grids adopted are (12 × 12 × 1), (12 × 7 × 1), (7
× 7 × 1), and (6 × 7 × 1) for the (0001̅)/(0001), (101̅0), 3.1. Relative Stability for ZnO Surfaces. Before the
(1120̅ ), and (1121̅ ) ZnO surfaces, respectively. The surface reactivity of ZnO is investigated, it is necessary to address the
energies reported correspond to energy fits for the ranges N = stability of its exposed crystal facets based on clear, consistent
5, 7, 9, and 11 for ZnO(0001̅)/ZnO(0001), 6, 8, 10, and 12 for information about the surface energy. In Table 1, we present
ZnO(101̅0), 5, 7, 9, and 11 for ZnO(112̅0), and 10, 14, 18, and
22 for ZnO(1121̅ ). Surface layers on both sides of each slab Table 1. Surface Energy γ (in J·m−2) Calculated for the
were allowed to relax. The central one, two, one, and two ZnO Different ZnO Surfacesa
unit layers were held fixed for the four surface systems, surface reference method γ
respectively.
(0001̅), (0001) this work PW91+U 3.71, 2.87b
2.3. Decomposition of Adsorption Energy. The energy
this work PW91 3.00
of adsorption/binding Ead for CO2 is defined by
this work BCM 1.77
Ead = Eads/sub − Eads − Esub (4) 29 PBE 3.50
95 PBE 3.40
where Eads/sub is the total energy of the ZnO surface covered 95 LDA 4.30
with the adsorbate in the optimized geometry, and Eads and Esub 96 B3LYP 4.00
are the total energies of the adsorbate in the gas phase and of (101̅0) this work PW91+U 2.15
the clean substrate, respectively. The free, isolated adsorbate this work PW91 1.88
was optimized in a large cubic vacuum box of 16 Å side, so that this work BCM 1.92
the Γ-point alone was used to sample the BZ. With this rule, a 29 PBE 1.80
negative value of the adsorption energy, Ead, indicates a release 95 PBE 1.60
of energy or a favorable adsorption. 95 LDA 2.30
Note that three alterations take place during the adsorption 97 B3LYP 2.13
of CO2 on the underlying oxide surface, namely, the structural (1120̅ ) this work PW91+U 2.21
rearrangements of the two partners and their interplay. this work PW91 1.94
Accordingly, to get a deeper insight into the adsorption this work BCM 2.21
process, we applied the following scheme, analogous to that 29 PBE 1.90
first proposed by Raybaud et al.,92,93 to decompose the 95 PBE 1.70
adsorption energy into deformation and interaction contribu- 95 LDA 2.50
CO
tions. The deformation energy for the adorabate, Edef 2, was 98 B3LYP 4.10
evaluated by (112̅1) this work PW91+U 4.47
this work PW91 3.72
CO2
Edef = Eads ′ − Eads (5) this work BCM 2.69
a
Other theoretical data are tabulated where available. bObtained from
where Eads′ stands for the energy of the adsorbate fragment with the fully hydroxylated slab calculations in which the ZnO(0001̅) and
the deformed geometrical structure adopted in the adsorption ZnO(0001) sides of each slab were occupied by the H and OH
complex. Apparently, the ECO 2
def parameter can be considered as a adsorbates at one coverage, respectively.
measure of the activation degree of the surface CO2 molecule.
The substrate deformation energy, Esub def , was calculated in a calculated surface energies in terms of GGA (PW91)+U, GGA-
similar way. The interaction energy Eint between the two
PW91, and BCM for the five low-index ZnO surfaces without
partners is defined as
any point defects. The outcomes of previous studies29,95−98 are
E int = Eads/sub − Eads ′ − Esub ′ (6) also listed for comparison. From this table, it can be seen that
different computational approaches predicted a wide range of
where Esub′ is the energy of the surface moiety and was the absolute values of the surface energy for a given surface. As
determined in an analogous fashion to Eads′. By combining eqs an example, both a large value of 4.30 J·m−2 within the local
D dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Figure 2. Possible (a−d) monodentate, (e,f) bidentate, and (g,h) tridentate adsorption modes for CO2 on ZnO surfaces as represented by a
ZnO(0001̅) surface. Only the topmost ZnO double layer of the plane is displayed for the sake of simplicity. Atomic spheres: dark gray, C; yellow,
adsorbate O; red, lattice O; blue, Zn.

Table 2. Optimized Geometric and Energetic Parametersb for CO2 Adsorbed on Clean Low-Index Surfaces of ZnO at 1/4
Coverageca
parameter (0001̅) (0001) (101̅0) (112̅0) (112̅1)
mode linear linear tridentate tridentate bidentate
dC−O1/dC−O2d 1.18/1.18 1.18/1.17 1.26/1.26 1.27/1.26 1.30/1.24
dC−O3 3.13 - 1.39 1.41 1.35
dO1−Zn1/dO2−Zn2 −/− 3.01/− 1.98/1.98 1.96/1.95 1.93/−
ΔdO3 0.02 - 0.20 0.22 0.28
ΔdZn1/ΔdZn2 −/− 0.11/− 0.73/0.73 0.43/0.36 0.44/−
∠O1CO2d 179.6 178.9 128.8 126.4 125.4
α 0.1/0.3 31.1/31.4 89.8 28.6 40.1
CN(O)/CN(Zn)e 3/− −/3 3/3 3/3 2,3/3
Ead −0.06 (−0.19)f −0.09 (−0.09)g −0.83 (−0.83/−0.62)h −0.79 −0.69 (−0.48)i
Ead (exp) −0.39 to −0.26j −0.38 to −0.32,k −0.47 to −1.45 to −0.67,m −1.04 to −0.83,n −1.59 to - -
−0.43l −1.13o
Eint −0.06 −0.09 −7.07 −9.02 −8.20
CO
Eslab2 0.00 0.00 4.50 6.99 6.79
Esub
def 0.00 0.00 1.74 1.24 0.72
a
Also given are experimental binding energies for direct comparison. bMode, preferred adsorption/binding mode; dA−B, bond distance of one pair of
atoms A and B implicated in the adsorbate−substrate interaction; ΔdL, displacement of lattice atom L in close proximity to the adsorbate from its
position in the relaxed bare surface; ∠O1CO2, OCO internal angle of the adsorbate; α, angle formed by the CO bonds in the adsorbate and the
substrate surface for ZnO(0001̅) and ZnO(0001) or angle of the adsorbate molecular plane relative to the substrate surface for other three facets;
CN(D), coordination number (CN) of surface atom D; Ead, adsorption/binding energy per molecule; Eint, adsorbate interaction energy with the
substrate; ECO
slab , molecule deformation energy; Edef , substrate deformation energy. For parameter α on ZnO(112̅1), the substrate surface of the
2 sub

stepped facet is referred to the terrace plane. The atom labels are the same as in Figure 3. Bond lengths and displacements, angles, and energies are in
units of Å, degrees, and eV, respectively. cSee text for coverage definition. dThe calculated bond length and bond angle of CO2 in vacuum are 1.18 Å
and 180.0°, respectively, in excellent agreement with the experimental data of 1.16 Å and 180.0°.125 eCN = 4 for all bulk atoms. fThe value in
parentheses was computed in the presence of coadsorption with atomic H of 1/4 coverage. gThe value in parentheses was obtained by coadsorption
with OH group of 1/4 coverage. hThe two values X/Y in parentheses were estimated at 1/8 coverage on the clean surface without and with intrinsic
point defects (oxygen vacancies with a concentration of 1/8) present,41 respectively. iThe value in parentheses was found from a hydroxylated
surface where each 2-fold coordinated, step-edge oxygen atom was covered with one H atom. jEstimated with the Redhead equation112,115 using the
TPD data.16 kTPD results on pristine sites.36−39,55 lTPD results on intrinsic defect sites.37,38 mAdsorption isotherm results.47 nTPD results from ref
47. oTPD results from refs 35 and 43.

density approximation (LDA)95,99 and a small value of 1.77 J· approximately evaluated with the LDA methodology instead.
m−2 from the BCM model were found in the (0001̅)/(0001) Analyzing Table 1 in greater detail, it is quite clear that the best
system. Besides, the surface energies at the PW91+U and LDA agreement with the LDA values was achieved using the GGA
levels95,99 are notably larger than those at the PW91 and PBE +U technique. It underestimates surface energies by at most
levels.29,95 Therefore, the currently reported results reveal that 0.59 J·m−2 for various ZnO surfaces examined here. In other
the effect of the plus-U is not trivial: an increase in the surface words, the current GGA + U functional offers much
energy by 13.9−23.7% relative to pure GGA-PW91 calcu- improvement over the results produced by the GGA and
lations. Nevertheless, the same ordering of the estimated B3LYP variants.
surface energies for the different surfaces was determined by all Now we focused on the relative stability of the oxide surfaces
the approximations employed, except for the Becke-3- derived with the PW91+U approach that would be adopted for
parameter-Lee−Yang−Parr (B3LYP)100,101 hybrid function- dealing with the adsorption of CO2 on them. As shown by the
al96−98 and BCM model. Despite the fact that the electron surface formation energies at the level, the stability drops in the
probability density of a solid material is not a slowly varying sequence ZnO(101̅ 0 ) > ZnO(112̅ 0 ) > ZnO(0001̅ ) /
function of position, LDA is generally known to do a ZnO(0001) > ZnO(112̅1 ), reflecting that the (101̅ 0 )
surprisingly excellent job of calculating certain surface proper- orientation is the most favored for the possible surfaces. It is
ties, especially the surface energy.102,103 Then, considering that worthy to mention that the energetics of only the first four
so far none of the pertinent experimental data are available in surfaces have been investigated in other calculations,29,95 which
the literature, the accuracy of other modeling methods can be also provided the same stability order for these surfaces as in
E dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

the present work. On the basis of our calculations, the second


most stable surface, ZnO(112̅0), is only slightly higher
energetically by 0.06 J·m−2 as compared to ZnO(101̅0). Such
a small energetic difference reveals that both types of nonpolar
surfaces are exposed and thus most likely to be observed. In
fact, experimental determinations of abundant (101̅0) and
(112̅0) faces on ZnO nanoparticles104 provide direct support
for this conclusion. As expected, the surface energies for the
three polar surfaces were found to be (at least 72% in energy)
much less stable than the nonpolar ones. The worst stability of
the stepped ZnO(112̅1) facet is ascribed to its roughest
morphology, with the edge O atoms having the lowest value of
CN (= 2) among the atoms of all the investigated surfaces.
Taking into account that the polar ZnO surfaces seen
experimentally are inevitably covered by hydrogen atoms,105,106
we have examined the effect of such a hydroxylation on the
stability of the (0001̅)/(0001) system. In the (1 × 1) surface
supercell employed, one hydrogen atom is located on the
terminal oxygen at the O-terminated (0001̅) side, while one
hydroxyl group occupies the 3-fold hollow site at the
complementary Zn-terminated (0001) side. As a consequence,
the surface energy of the hydroxylated system is lowered by
0.84 J·m−2 to 2.87 J·m−2 (Table 1) with respect to that of the
corresponding clean substrate. This points to an enhanced
stabilization of the two opposite facets by their interactions
with respective adsorbates. To reliably work out the depend-
ence of the surface stability on the reactive environment such as
water, discussion should be based on Gibbs free energies
(instead of the total energies at 0 K as used here),107,108 which
is outside the scope of the current paper.
3.2. Adsorption of CO2: Structure and Energies. Next
we turned to the adsorption characteristics of CO2 on the five
different surfaces of ZnO. As sketched in Figure 2, in total eight
initial binding modes were examined for the species at all
nonequivalent adsorption sites on each surface considered,
including various monodentate, bidentate, and tridentate
geometries. All these possible molecular orientations were
proposed by previous theory and/or experiment on many
oxides such as ZnO.34,109−111 Table 2 summarizes our findings
regarding the relevant structural parameters and energy
contributions for CO2 in the energetically most favorable
situation on these ZnO surfaces. These optimized states are
illustrated in Figure 3. Figure 3. Most energetically favorable configurations for adsorbed
3.2.1. CO2 Adsorption on Polar (0001̅), (0001), and (112̅1) CO2 on (a) ZnO(0001̅), (b) ZnO(0001), (c) ZnO(101̅0), (d)
ZnO(112̅0), and (e) ZnO(112̅1). Atoms engaged in the adsorbate−
ZnO Surfaces. From all possible adsorption modes on the substrate interaction were labeled. Upper panels show side views, while
perfect ZnO(0001̅) and ZnO(0001) faces, only somewhat lower panels indicate top views. See Figure 2 for color coding.
stable physisorption states were obtained for the adsorbed CO2.
The adsorbate lies almost flat on the O-terminated facet
(Figure 3a), the angle α between the surface plane and the C Martins et al.,54 who also predicted the binding energies of CO2
O bond axis being 0.1° and 0.3°, whereas it significantly tilts to be about −0.1 eV on both polar surfaces. Such small energies
toward the Zn-terminated facet (Figure 3b) with α = 31.1° and in absolute magnitude imply that the adsorbed CO2 most likely
31.4°. The internal structure of the molecules is barely affected desorbs from these surfaces rather than being activated. Thus,
by the presence of both surfaces. The CO bond lengths of as experiments24,31,32,36,37 and recent theoretical calculations
1.17−1.18 Å and the OCO angles of 179.6° and 178.9° are demonstrate,50 the surface molecule cannot react with the
essentially equal to the corresponding gaseous values of 1.18 Å regular, undefected ZnO(0001̅) and ZnO(0001) faces to
and 180.0°, respectively. In the meantime the distances of the generate carbonate or carboxylate species.
three atoms within CO2 from the surface are at least as far as Redhead analysis of the TPD spectra of CO2 from the two
3.12 Å, disfavoring the interactions of the molecule with the singly terminated facets determined its adsorption energies at
two flat substrates. Not unexpectedly, the linear adsorption various coverages to be −0.39 to −0.26 eV on ZnO(0001̅)16,112
mode was slightly exothermic, with the estimated heats of and −0.38 to −0.32 eV on ZnO(0001).36−39,55 These data are
adsorption of −0.06 and −0.09 eV on ZnO(0001̅) and slightly exothermic compared to our computed results just
ZnO(0001), respectively. The current DFT+U results are mentioned. The discrepancy found could be explained as
comparable to the simple DFT slab model calculations by follows. First, as pointed out in refs 113 and 114, adsorption
F dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

energies obtained by TPD experiments depend sensitively on ZnO(1010̅ ) orientation, while it is inclined by only 28.6°
some guessed value for the preexponential factor which is with respect to the ZnO(112̅0) orientation. The two O atoms
typically 1013 s−1.115 However, this guessed prefactor often in the molecule fragment have an almost equivalent bonding
deviates orders of magnitude from the correct one. Second, with their closest Zn atoms, with the newly formed O−Zn
underestimation of absolute adsorption energies with the GGA bond distances of 1.98 Å on the former facet and of 1.95 and
+U method is likely.116,117 From Table 2, it can be seen that the 1.96 Å on the latter one, all of which are only slightly larger
experimental adsorption energies of CO2 on both the polar than the calculated bulk-terminated values (1.92 and 1.93 Å).
surfaces do not result from the surface impurity contamination The internal CO bonds of CO2 were lengthened upon
at all.105,106 For instance, even if coadsorption with atomic adsorption from an original value of 1.18 Å to 1.26−1.27 Å,
hydrogen of 1/4 coverage further stabilizes CO2 by 0.13 eV on which are 0.13−0.15 Å shorter relative to the external C−O
ZnO(0001̅), the resulting adsorption energy of −0.19 eV is bonds. The chemisorbed molecule again possesses a bent O
again weaker than the corresponding TPD results.16,112 It is CO skeleton with an angle of about 125°.
worth mentioning that the present slab results in magnitude are The binding energy for CO2 on ZnO(101̅0) was calculated
always smaller than the corresponding ones predicted by cluster to be −0.83 eV, which is very close to (but slightly more
models (−0.40 eV on ZnO(0001̅)49 and −8.46 eV on H/ exothermic than) the one (−0.79 eV) predicted at ZnO(112̅0).
ZnO(0001)52). The former value does just fall in the range of the
In contrast to the cases of the ZnO(0001̅) and ZnO(0001) chemisorption energies from Hotan et al.,47 namely, −1.45 to
surfaces, a stable bidentate carbonate was found to form on −0.67 eV obtained through adsorption isotherms and −1.04 to
another polar, stepped ZnO(112̅1) surface upon exposure to −0.83 eV estimated by the TPD measurements. Obviously our
CO2 (Figure 3e). This is in full agreement with recent finding is more reasonable than other DFT data reported on
experimental observations24,31 that clearly indicate that the the same (101̅0) surface, −0.70 and −0.47 eV.45,53 Never-
appearance of carbonate species on ZnO(0001̅ ) and theless, the present theoretical value is still much weaker than
ZnO(0001) can be related with activation of CO2 at their the experimental TPD results of Waugh et al.35,43 that vary
defect sites such as steps. In the calculated chemisorbed state, between −1.59 and −1.13 eV. With this disagreement in mind,
the CO2 moiety bridges a low coordinated Zn−O dimer we continued to explore whether adsorbate coverage and
located directly at the step edge via one CO bond, thus surface intrinsic point defects (in the form of oxygen vacancies)
creating a nascent C−O bond and a nascent O−Zn bond. The can further promote the stability of the adsorbed CO2 on
surface molecule is strongly deformed compared to its gas- ZnO(101̅0) by employing surface models with a 1/8 coverage
phase counterpart, with the internal CO bonds elongated to and the k points of the (6 × 4 × 1) MP mesh. On the regular,
an average length of 1.27 Å and an OCO angle shrunk by ca. pristine surface, the computed adsorption energy per CO2
55° to 125.4°. Such a drastic structural deformation points to molecule is exactly equal to that at 1/4 coverage, indicating that
very high activation of the CO bonds by the CO2 interaction a lower fractional coverage does not enhance the adsorption at
with the defective substrate. The preferred bent configuration is all. On the other hand, on the defected surface with a single O
tilted 40.1° from the parallel of the (112̅0) terraces. The vacancy considered in one unit cell,41 we found that even if the
predicted chemisorption energy per molecule amounts to vacant site could adsorb the CO2 molecule the adsorption
−0.69 eV, which is, however, more favored than the TPD- energy would be weakened by 0.21 to −0.62 eV, which deviates
estimated values of −0.47 to −0.43 eV on the ZnO(0001) more from the TPD values of −1.59 to −1.13 eV35,43 than that
surface strongly sputtered by Ar+ ions (i.e., exhibiting a rather on the defect-free counterpart. It reveals that the presence of
large density of defects).37,38 To clarify this difference, on the point defects on ZnO(1010̅ ) will hinder rather than facilitate
ZnO(112̅1) orientation, we also investigated the influence of the adsorption process. Therefore, according to these
atomic hydrogen coadsorbate, inevitably present in the real calculation results, the CO2 adsorption energies published in
surface,30 on the CO2 adsorption. Intriguingly, when each step- refs 35 and 43 appear, we believe, to be questionable.
edge O atom was preoccupied by one H atom, the binding It should be emphasized that on an actual ZnO(1010̅ )
mode of CO2 evolved from bidentate to tridentate (not surface CO2 can react with a surface H impurity to produce
presented here). Moreover, the coadsorption decreased the more strongly held intermediates like H2CO.30,118 Some of the
exothermicity of CO2 adsorption by 0.21 to 0.48 eV, which now surface intermediate species most likely give rise to TPD peaks
agrees excellently with the experimental data acquired from the centered at high temperatures between 390 and 560 K.35,43
defected ZnO(0001) surface.37,38 As a conclusion, our However, Waugh and co-workers35,43 attributed these peaks to
calculation results clearly show that these TPD adsorption the adsorbed CO2 on the surface and thus derived the CO2
energies37,38 could be assigned to chemisorption, rather than adsorption energies from these TPD data. As a support to our
physisorption, of the CO2 molecules at these H-covered defect hypothesis, using the computed binding energy of H2CO
sites. (−1.25 eV) and Redhead analysis,112,115 we found that the
3.2.2. CO2 Adsorption on Nonpolar (101̅0) and (112̅0) predicted temperature of the desorption peak maximum for the
ZnO Surfaces. For the two nonpolar planes, ZnO(101̅0) and adsorbed H2CO on the nonpolar surface, 425 K, indeed lies
ZnO(112̅0), CO2 prefers to bind with two neighboring surface within the CO2 desorption peak temperature range claimed by
Zn atoms via its two O ends, respectively, and with a lattice O those authors (i.e., 390−560 K).
atom via a carbon atom, thus yielding a tridentate carbonate 3.2.3. Comparison between All Investigated ZnO Surfaces.
species (Figure 3c,d). Consistent with the slab calculations of According to results reported in Table 2, the comparison of
Wang et al.,45 the present result also disagrees with an earlier CO2 deformation energies on the five differently oriented
experimental work44 based on NEXAFS that identified a surfaces of this study reveals that CO2 activation by ZnO relies
bidentate adsorbed configuration for the CO2 on ZnO(101̅0). strongly on the nature of the substrate. For the perfect O-
As far as our structural properties are concerned, the molecular terminated (0001̅) and Zn-terminated (0001) facets, the
plane of the adsorbate is exactly perpendicular to the molecule preserves its linear structure upon adsorption and
G dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

thus remains chemically inactive. This behavior is not surprising was obtained for the Zn atoms of this plane (Table 2). The
because both the basal surfaces lack coordinatively unsaturated value thus approaches nearly twice that found in either CO2/
Zn−O dimers that are required for CO2 to be activated.24,34 On ZnO(112̅0) or CO2/ZnO(112̅1). As a result, the ZnO(101̅0)
the contrary, the chemisorption and resulting activation of CO2 surface shows a higher deformation energy, by at least 0.5 eV,
are favored on the mixed-terminated (1010̅ ), (1120̅ ), and than the other two surfaces. Furthermore, as seen from Table 2,
(112̅1) planes. The deformation energies for these activated the molecule deformation energy is at least 159% more
molecules, although very similar in structure, differ by 0.2−2.49 endothermic than the substrate deformation energy for any of
eV between the three surfaces. Such large energetic differences the three chemisorption systems. It clearly suggests that a
indicate that the activation extent of the CO bonds in the strong CO2−surface interaction is done primarily at the cost of
chemisorbed CO2 is also structure sensitive. a huge molecule deformation energy.
From the analysis of the computed adsorption energies at 1/ Our adsorption energy decomposition results shed light on
4 coverage on all the clean oxide surfaces, it is clear that the the different ability of CO2 adsorption on different ZnO
binding of CO2 to these substrates is strengthened in the order surfaces. The adsorption energy, which reflects a global stability
of ZnO(0001)̅ ≤ ZnO(0001) < ZnO(1121̅ ) < ZnO(1120̅ ) ≤ of the system,119 is a trade-off between the antagonistic
ZnO(101̅0). The substrate stabilizing effect is congruent with interaction and deformation energies.92 The lowest exother-
prior experimental evidence of the crystal plane dependence of micities for adsorption of the CO2 molecules at the ZnO(0001̅)
CO2 adsorption on ZnO.35 Surprisingly, this order is almost and ZnO(0001) sites with respect to all systems examined here
completely reversed to the decreasing order of the DFT+U- can be straightforwardly traced back to negligible adsorbate−
derived stability of the bare ZnO surfaces as presented in the substrate interactions (Eint > −0.1 eV). For CO2/ZnO(112̅0)
previous section. This finding is not trivial since it contradicts relative to CO2/ZnO(112̅1), the deformation energies for the
the widely accepted statement that a solid surface with better adsorbate and substrate moieties are both increased. Never-
(worse) stability normally exhibits lower (higher) reactivity. theless, the total cost (0.72 eV) in these deformation energies is
Also importantly, we noted that the nonpolar ZnO(101̅0) facet completely compensated by the gain (0.82 eV) in the
not only is the most stable facet of ZnO but also shows the adsorbate−substrate interaction energy. Consequently, the
highest activity in CO2 adsorption. Maybe for this reason, ZnO binding of CO2 on the ZnO(1120̅ ) facet is somewhat improved
particles have been identified as a promising material for use in by 0.1 eV. On moving from ZnO(112̅0) or ZnO(112̅1) to
high-performance hydrogenation catalysts for CO2.13,15 ZnO(101̅0), although the CO2−surface interaction and the
It is also tempting to elucidate how the surface structure of substrate deformation are simultaneously weakened, the
ZnO affects the interplay between CO2 and the oxide. With the decrease in the substrate deformation energy is at least 0.54
calculated interaction energies shown in Table 2, it can be seen eV more than the decrease in the absolute interaction energy.
that the adsorbate−substrate interaction follows ZnO(0001̅) ≤ Accordingly, a stronger value of the final measured CO2
ZnO(0001) < ZnO(101̅0) < ZnO(112̅1) < ZnO(112̅0) in an adsorption energy was obtained for the last surface. All in all,
increasing sequence. Some of this order corresponds well with it can be concluded that the thermodynamically most stable
intuition because the lowering of the surface stability of a (101̅0) surface, however, offers the best site for the adsorption
substrate would be expected usually to improve its interaction of CO2 on ZnO particles.
with the adsorbate molecules. Owing to, as already mentioned, 3.3. Electronic Properties. To reveal the physical origin of
the absence of exposed Zn−O dimers, the first two polar the interaction between CO2 and ZnO surfaces, detailed
surfaces cannot enhance the CO2−surface interaction as analyses of electronic structure were made for the adsorbed
compared to the most stable ZnO(1010̅ ) surface. Indeed, the states described previously. First, the electronic charge density
adsorbate−substrate bonds formed after CO2 adsorption are difference, Δρ(r), for CO2 adsorption on each surface of
much weaker for the ZnO(112̅0) surface than for the stepped interest was constructed to gain insight into the charge density
ZnO(112̅1) surface, as evidenced by their longer distances. The redistribution in surface binding according to the following
presence of the extra dangling bonds of the surface step is expression
responsible for the bond differences. However, the adsorbate− Δρ(r) = ρads/sub (r) − ρads (r) − ρsub (r) (8)
substrate interaction is not strengthened but weakened by 0.82
eV when going from the former to the latter surface. The where ρads/sub(r), ρads(r), and ρsub(r), in turn, are the total
reduced interaction is perhaps interpreted as due to the lower charge densities of the adsorbate on the slab in the optimized
coordination of CO2 with the substrate on the stepped surface, configuration, the separated adsorbate, and the separated
two versus three on the less open (112̅0) surface.34 substrate. In computing ρads(r) and ρsub(r), the adsorbate and
A close inspection of the relationship between the interaction substrate atoms were kept fixed at precisely the same positions
energy and deformation energies of the CO2 and ZnO surface as they were in the adsorption complex. Hence, the values of
shows that the stronger the adsorbate−substrate interplay, the Δρ(r) are positive in the regions where the electronic charge is
larger the geometrical distortions of the two partners. An accumulated after molecular adsorption and negative where the
exception to this would be in the CO2/ZnO(101̅0) system charge is depleted. The contour diagrams of the calculated
where the surface relaxation brought by CO2 adsorption is electron density difference in a cross section through the plane
remarkably greater than those of the ZnO(112̅0) and of the adsorbed CO2 molecules are displayed in Figure 4 for
ZnO(112̅1) fragments, both of which more strongly interact ZnO(101̅0), ZnO(112̅0), and ZnO(112̅1) (no bonding was
with the adsorbate. The stronger structural deformation can be found to occur between the adsorbate with ZnO(0001̅) or
attributed to a breaking of one surface Zn−O bond directly ZnO(0001), and thus the charge density difference maps for
below the adsorbed molecule on the ZnO(1010̅ ) plane, which the two physisorbed systems are not shown on the figure). The
was significantly elongated by 0.50 to 2.31 Å with respect to the cutting plane chosen is exactly the common plane of symmetry
clean counterpart. In particular, the maximum value of the of the chemically active frontier orbitals of a free bent CO2
adsorption-induced displacement of substrate atoms (0.73 Å) molecule, which are the highest occupied molecular orbital
H dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Figure 4. Contour plots of electronic charge density difference for


CO2 chemisorption on (a) ZnO(101̅0), (b) ZnO(112̅0), and (c)
ZnO(112̅1). The cutting plane passes through the CO bonds of the
investigated molecule. Positive/negative values imply electron density Figure 5. Partial density of states (PDOS) for (a) the free linear CO2
accumulation/depletion. The atom labels refer to the ones shown in molecule, (b) the bent CO2 molecule in vacuum, (c) the CO2/
Figure 3. Note that upon adsorption of CO2 on the (112̅0) facet the ZnO(101̅0) complex, (d) the CO2/ZnO(112̅0) complex, and (e) the
charge density redistributions of the Zn atoms bonding with the CO2/ZnO(112̅1) complex. The PDOS spectrum of the isolated bent
adsorbate (not displayed here) are similar to that in plot c. CO2 was calculated with the averaged structural parameters of the
adsorbate in the three chemisorbed systems. The CO2 molecular
(HOMO: CO2-4b2) and the lowest unoccupied molecular orbitals of interest are marked. In plots c−e, shown are only the active
orbital (LUMO: CO2-6a1).109,120 bands (involved in the adsorbate to substrate bond formation) of CO2
and the closest surface atoms. The Fermi level (EF) is indicated by a
It can be seen from Figure 4 that the bonding behaviors of
green dashed vertical line at 0 eV.
CO2 molecules on the three active surfaces are very similar in
two ways. First, upon adsorption the charge density in the 4b2
orbital is depleted, while there are electrons accumulated on the
6a1 orbital. Since CO orbitals in 4b2 and 6a1 possess bonding analysis of the free CO2 PDOSs (Figure 5a,b), it is clear that
and antibonding characteristics, respectively,120 these charge the bending of the gas-phase molecule split the bonding
rearrangements would cause the two CO bonds in CO2 to be HOMO 1πg and antibonding LUMO 2πu into 1a2, 4b2 and 6a1,
highly activated and hence to be elongated as described earlier. 2b1, respectively. As a consequence, the 4b2 orbital becomes the
Second, the charge density changes emerge mainly in the 2p HOMO, while the 6a1 orbital becomes the LUMO, as noted in
orbitals of the C and O of CO2 and lattice O linked to this the previous paragraph. After adsorption of CO2 on the three
molecule as well as the 4s orbital of the first neighbor Zn surfaces, the 6a1 and 2b1 states are both shifted to lower
atom(s), implying that these states are involved in the binding energies below the Fermi level EF (Figure 5c−e), indicative of
of CO2 to ZnO substrates. O(CO2) 2p electrons directly the bond formation between CO2 and the substrates. The
adjacent to the corresponding Zn atom(s) are built up, whereas electron transfer to the antibonding LUMO and LUMO+1
the Zn atom(s) loses electronic charge on the s-like orbital(s). orbitals weakened the CO bonds in the adsorbed molecule,
It is apparent that the Zn center(s) donates electrons to the which is what we have seen from our charge density
adsorbate molecule (also see below). In the case of CO2/ calculations. To mix with surface O 2p and Zn 4s orbtials,
ZnO(101̅0), the Zn 4s electrons are in a fairly delocalized the 2p states of the adsorbed CO2 broaden compared to those
shape, which appears to result from the simultaneous of a free CO2, especially in the energy interval −9 to 0 eV and
interaction of the valence orbital with two O atoms in different near 6 eV. The signature is common for molecules interacting
CO2 molecules due to the periodic boundary condition. The with solid surfaces as predicted by the Newns−Anderson
charge density on the relevant surface O atom is highly model.121 Combining Figure 4 and Figure 5c−e, one can
polarized to maximize its electrostatic interplay with the C therefore conclude that there must be a fairly substantial
atom. It should be emphasized that because of the stable 3d10 covalent contribution to the CO2−ZnO interactions.
configuration of the zinc ion the 3d orbtials do not participate To quantify adsorption-induced energy variations of the
in the chemical bonding of CO2 on ZnO surfaces. closest surface O 2p and Zn 4s orbitals, we have utilized eq 9 to
Next, we calculated the partial density of states (PDOS) of compute the energy of each of these valence bands (Eb, with b
CO2/ZnO(101̅0), CO2/ZnO(112̅0), and CO2/ZnO(112̅1) = 2p or 4s) in the presence/absence of CO2 (named Eads b /
projected onto the C and O belonging to CO2 and the nearest Eclean
b ), which is a measure of the electron contribution to the
lattice O and Zn atoms, which are plotted in Figure 5. From the total energy.122
I dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Table 3. Bader Charges Q (in |e|) for the Atoms on Clean Low-Index ZnO Surfaces and Bader Charge Differences ΔQ of the
Atoms (Entities) in Each Investigated System When Going from CO2 in the Gas Phase Plus the Clean Surface to the Adsorbed
Complexa
parameter (0001̅) (0001) (101̅0) (112̅0) (112̅1)
QO3b
−1.03 - −1.15 −1.15 −0.88
QZn1/QZn2b −/− 0.95/− 1.15/1.15 1.15/1.15 1.19/−
ΔQCb 0.03 0.03 0.07 0.02 0.11
ΔQO1/ΔQO2b −0.02/−0.02 −0.08/−0.04 −0.09/−0.12 −0.10/−0.14 0.00/0.09
ΔQO3b −0.01 - −0.03 0.03 −0.17
ΔQZn1/ΔQZn2b −/− 0.15/− 0.11/0.11 0.10/0.10 0.05/−
ΔQCO2 −0.01 −0.09 −0.14 −0.22 0.21
ΔQOc 0.03 0.06 −0.01 −0.06 −0.07
ΔQZnc −0.01 −0.12 0.07 0.06 −0.01
ΔQpseudoHd 0.00 0.00 - - 0.00
a
A negative value of ΔQ means electron gain. See Figure 3 for the atom labeling. bFor the atoms involved in the adsorbate−substrate interaction.
c
Sum of the changes in net charge on all surface X (X = O or Zn) atoms in a unit cell except the atom(s) bonding to the adsorbate. dSum of the
changes in net charge of all pseudohydrogen atoms.

E
∫−∞F nb(ε)εdε calculations further revealed that, as for the ZnO(101̅0) and
Eb = ZnO(112̅0) facets, the negative charges in CO2δ− are primarily
E
∫−∞F nb(ε)dε the consequence of electron injection from the nearest-
(9)
neighbor Zn cations into the two adsorbate oxygen atoms.
However, the above picture does not account for why the CO2
where nb(ε) is the b-PDOS of a specific surface atom at the
on the stepped surface becomes positively instead of negatively
energy level (eigenvalue) ε. Then, the stabilization energy of
charged. We then turned to analyzing the surface atomic
state b in the adsorption system can be given by ΔEb = Eads b −
charges of all the investigated surfaces and their changes upon
b . The smaller (more negative) ΔEb means the more active
Eclean
the adsorption of CO2, which should reflect the potentiality of
b state, and vice versa. The calculated ΔE2p and ΔE4s are −2.15,
these atoms to further capture or lose electrons. It was found
0.07; −2.58, −0.59; and −2.39, −0.14 for the (101̅0), (112̅0),
that the least coordinated O and Zn atoms on the clean
and (112̅1) ZnO surfaces, respectively. It can be seen that the
ZnO(112̅1) surface bear less negative and more positive
stabilization of both the lattice O 2p and Zn 4s states by CO2
charges, respectively, as compared to the corresponding atoms
adsorption strengthens on going from ZnO(101̅0), through
on the other surfaces. The charge differences demonstrate that
ZnO(112̅1), to ZnO(112̅0). The trend agrees nicely with the
for the defective surface the two kinds of surface atoms are
observed differences in adsorbate−substrate interaction energy
characterized by stronger electrophilic and electrophobic
between these surfaces (see above). We found that although
properties, respectively. Hence, on CO2 adsorption, the edge
the larger the energy of a state on a clean surface (closer to the
O atom accepted as much as 0.17 electrons from this molecule,
Fermi level) the higher its activity would generally be, there is a
while the Zn atom hardly donated any electron to the nearest
caution to apply this rule to the investigated systems. On the
adsorbate O atom, resulting in an overall transfer of charge
basis of our calculations, the Eclean
2p values are, respectively,
from the adsorbate to the surface. In this way the present
−2.20, −1.98, and −1.46 eV for ZnO(101̅0), ZnO(112̅0), and
calculations suggest, for the first time, the possible presence of a
ZnO(112̅1), whereas the 4s orbital energies are −6.11, −5.33,
positively charged CO2 species on certain ZnO surfaces, such as
and −5.34 eV for the three bare substrates, respectively.
ZnO(112̅1).
However, the much higher 2p state energy on ZnO(1121̅ ) than
Our results also indicate that not all CO2/ZnO systems obey
that on ZnO(112̅0) does not lead to a stronger CO2−surface
the well-established rule126−128 that when adsorbates increase
interaction (Eint: −8.20 vs −9.02 eV). It should be pointed out
the work function of a surface they tend to withdraw electrons
that the reactivity of ZnO surfaces toward CO2 adsorption is from the surface. The work function Φ was calculated as
not directly associated with the surface electronic structure.
This is because the adsorption energy depends not only on the Φ = Evac − E F (10)
interaction component but also on geometrical deformation.
Now we were in the position to address the charge states of where Evac is the plane-averaged electrostatic potential energy of
CO2 on ZnO surfaces by using the Bader charge analysis123,124 an electron in the vacuum region far from the surface.129 The
as listed in Table 3. Given a high ionization potential of 13.77 surface work function change, ΔΦ, before and after adsorption
eV and a low electron affinity of ∼3.80 eV, CO2 is a was evaluated by subtracting the work function of the clean
comparatively poor electron donor but a good electron surface from that of the covered surface. As originally proposed
acceptor.34,125 Indeed, judging from the total net charges by Brédas et al.,130 the conduction band minimum was selected
ΔQCO2 on CO2 at the five different ZnO surfaces under as the effective EF value for calculating the work function to
scrutiny, the adsorbed molecules gained electrons (−0.22 to make a meaningful comparison with the experimental data on
−0.01 |e|) from almost all the substrates (with the exception of ZnO. The computed results along with prior theoretical and
ZnO(112̅1)) in the course of the adsorption process, giving rise experimental estimates130−133 are given in Table 4. We found a
to the formation of the partially negatively charged CO2δ−. It is good agreement of our results about the work functions of bare
mentioned in passing that the reduction of CO2 by oxygen ZnO surfaces with the corresponding data reported in the
vacancies on ZnO(0001̅) and by ZnO(101̅0) was observed as literature. For example, the work function of ZnO(0001) was
well in the recent theoretical studies.45,49,50,53 Our Bader charge predicted in the current work to be 4.61 eV. This value is only
J dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

Table 4. Work Function Φ (in eV) of the Differently


Oriented ZnO Surfaces and Change ΔΦ in the Surface Work
Function after the Adsorption of CO2ba
parameter (0001̅) (0001) (101̅0) (112̅0) (112̅1)
Φ 5.96 4.61 (4.68)c
4.42 4.43 6.41
Φ (exp) 4.23−4.95d 3.50−4.25d 4.49−4.73e - -
ΔΦ 0.04 0.08 0.85 0.42 0.59
a
Also listed are the available previous theoretical and experimental
values for the purpose of comparison. bA positive ΔΦ implies an
increase of the surface work function. cIn parentheses is given the prior
PBE+U result.130 dReferences 131−133. eReferences 132 and 133.

somewhat higher than the experimental data ranging from 3.50


to 4.25 eV131−133 and is essentially identical to the GGA-PBE
+U result (4.68 eV) of ref 130. As far as we know, there is no
theoretical or experimental study on the work functions of
CO2-adsorbed ZnO surfaces so far. Here the changes in work
function for the five examined ZnO surfaces exposed to CO2
were identified to be positive, varying between 0.04 and 0.85
eV. Such changes would be expected to result in a donation of
charge from the adsorbate to each of these surfaces. However,
we found one exceptionalthough the CO2 caused the work
function of ZnO(112̅1) to increase, the electron acceptance
from the adsorbed molecule to the stepped substrate could still
take place. In essence, as pointed out by King and co- Figure 6. Effect of the transferred charge between CO2 and the
workers,126 the signs of the changes in work function do not underlying ZnO surface on (a) the interaction energy of the two
necessarily correlate with the direction of the charge transfer partners and (b) the molecule deformation energy. R is the correlation
due to the surface overspill of electron density into the coefficient of the linear regression.
adsorbate region.
Finally, we were interested in the relationship of the On the prism ZnO(101̅0) facet, the influence of coverage and
interaction energies of CO2 with ZnO surfaces to the defects on the adsorption of CO2 was discussed, and its heats of
transferred charges on the adsorbed adsorbate. According to adsorption were estimated to be not smaller than −0.83 eV per
Tables 2 and 3, the molecule with a higher absolute net charge molecule. It is therefore proposed that the too tightly bound
interacts more strongly with the underlying substrate. In fact, as states observed experimentally for CO2 at the prism surface,
shown in Figure 6a, there is a linear dependence of their corresponding to binding energies far below our lower energy
interaction energy on the transferred charge with a poor limit, are most likely attributed to adsorption of other species,
correlation coefficient (R = 0.78). Consequently, the ionic instead of CO2, on this surface. On the basis of the present
character of the adsorbate−surface interplay becomes more computed adsorption energies, the ability of CO2 adsorption is
accentuated in the order of ZnO(0001̅) , ZnO(0001), sensitively affected by the substrate change, and it is in the
ZnO(101̅0), ZnO(112̅1), and ZnO(112̅0). As already pointed sequence ZnO(0001̅) ≤ ZnO(0001) < ZnO(112̅ 1 ) <
out, the CO2 interaction with the ZnO substrate is increased ZnO(112̅0) ≤ ZnO(101̅0). This sequence cannot be explained
predominantly at the expense of a larger molecule distortion. by the stabilities of the five bare surfaces. Our detailed analyses
Naturally, we again found that the calculated deformation reveal that the reason behind no regular relationship seen
energy for CO2 varies roughly linearly with the charge between the two properties is that the final measured
transferred between the two partners (Figure 6b), indicating adsorption energy is the result of a delicate balance between
that the stronger the electron transfer, the stronger the the interaction and geometrical distortions of the adsorbate and
activation of the internal CO bonds. surface moieties.
The interactions of the chemisorbed CO2 with ZnO surfaces
4. CONCLUSIONS follow a simple patternthe 2p state of the molecule develops
In summary, we have presented a comprehensive survey of the overlap with the relevant lattice O 2p and Zn 4s bands. We
adsorption of CO2 on all the relevant low Miller index found that the energy changes of the two kinds of surface
orientations of single-crystal ZnO by using the DFT+U scheme. atomic orbitals, induced by the presence of CO2, can be used as
The oxide surface structure has been analyzed, and the a tool to predict the relative strength of the interplay of the
predicted stability ordering of (112̅1) < (0001̅)/(0001) < molecule and ZnO surface. In general, the stronger the electron
(1120̅ ) < (1010̅ ) is adopted. Our results clearly demonstrate transfer between the two fragments, the stronger is their
that the preferred adsorption state of CO2 depends on the interaction, and also the higher is the activation of the
nature of the substrate: the molecule physisorbs on the singly adsorbate molecule. The electron donation from Zn to the
terminated surfaces, while it chemisorbs on the mixed antibonding orbitals of CO2 and back-donation to lattice O
terminated surfaces. For the basal ZnO(0001̅ ) and from the electron of a CO2 bonding orbital lead to elongated
ZnO(0001) facets, the formation of carbonates from adsorbed CO internal bond distances as well as a rather bent OC
CO2 was validated to occur at steps such as ZnO(112̅1) and/or O angle. More importantly, CO2, albeit being an electro-
other defect sits, compatible with recent experimental findings. negative adsorbate, is partially positively charged on the
K dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX
The Journal of Physical Chemistry C Article

stepped ZnO(112̅1) facet, due primarily to the 2-fold (21) Tang, Q.-L.; Hong, Q.-J.; Liu, Z.-P. J. Catal. 2009, 263, 114−
coordinated edge O atom with a strong electrophilic effect. 122.
The result counters the commonly held view that the best and (22) Grabow, L. C.; Mavrikakis, M. ACS Catal. 2011, 1, 365−384.
most appropriate description of CO2 on ZnO surfaces should (23) Hikov, T.; Rittermeier, A.; Luedemann, M.-B.; Herrmann, C.;
Muhler, M.; Fischer, R. A. J. Mater. Chem. 2008, 18, 3325−3331.
be a CO2δ− anion. It was also highlighted that the alterations in
(24) Esken, D.; Noei, H.; Wang, Y.-M.; Wiktor, C.; Turner, S.; Van
work function for ZnO surfaces upon the adsorption of CO2 Tendeloo, G.; Fischer, R. A. J. Mater. Chem. 2011, 21, 5907−5915.
are not necessarily decided by the direction of charge transfer. (25) Xu, Y.-N.; Ching, W. Y. Phys. Rev. B 1993, 48, 4335−4351.
We believed that the current picture of CO2 adsorption on the (26) Wróbel, J.; Piechota, J. Solid State Commun. 2008, 146, 324−
different ZnO surfaces would shed some light on the more 329.
complicated catalytic reactions such as methanol synthesis from (27) Azzaz, Y.; Kacimi, S.; Zaoui, A.; Bouhafs, B. Phys. B 2008, 403,
CO2 hydrogenation. 3154−3158.


(28) Kresse, G.; Dulub, O.; Diebold, U. Phys. Rev. B 2003, 68,
AUTHOR INFORMATION 245409.
(29) Diebold, U.; Koplitz, L. V.; Dulub, O. Appl. Surf. Sci. 2004, 237,
Corresponding Author 336−342.
*E-mail: qltang@xidian.edu.cn. Phone: (+86) 29-8189-1149. (30) Xia, X.-Y.; Strunk, J.; Busser, W.; d’Alnoncourt, R. N.; Muhler,
Fax: (+86) 29-8189-1371. M. J. Phys. Chem. C 2008, 112, 10938−10942.
Notes (31) Noei, H.; Wöll, C.; Muhler, M.; Wang, Y.-M. J. Phys. Chem. C
2011, 115, 908−914.
The authors declare no competing financial interest.


(32) Lindsay, R.; Michelangeli, E.; Daniels, B. G.; Ashworth, T. V.;
Limb, A. J.; Thornton, G.; Gutiérrez-Sosa, A.; Baraldi, A.; Larciprete,
ACKNOWLEDGMENTS R.; Lizzit, S. J. Am. Chem. Soc. 2002, 124, 7117−7122.
The authors acknowledge the financial support from the (33) Saussey, J.; Lavalley, J.-C.; Bovet, C. J. Chem. Soc., Faraday
Natural Science Foundation of China Grant No. 21003098 and Trans. 1 1982, 78, 1457−1463.
the Fundamental Research Funds for the Central Universities (34) Busca, G.; Lorenzelli, V. Mater. Chem. 1982, 7, 89−126.
(35) Bowker, M.; Houghton, H.; Waugh, K. C.; Giddings, T.; Green,
of China Grant No. K50510070003. One of the authors (Q.-
M. J. Catal. 1983, 84, 252−255.
L.T.) is indebted to Dr. W.-Q. Li for the fruitful discussions and (36) Wang, J.; Burghaus, U. Surf. Rev. Lett. 2004, 6, 521−529.
suggestions provided during this work.


(37) Wang, J.; Burghaus, U. Chem. Phys. Lett. 2005, 403, 42−46.
(38) Wang, J.; Burghaus, U. J. Chem. Phys. 2005, 122, 044705.
REFERENCES (39) Wang, J.; Hokkanen, B.; Burghaus, U. Surf. Sci. 2005, 577, 158−
(1) Wang, W.; Wang, S.-P.; Ma, X.-B.; Gong, J.-L. Chem. Soc. Rev. 166.
2011, 40, 3703−3727. (40) Au, C. T.; Hirsch, W.; Hirschwald, W. Surf. Sci. 1988, 199, 507−
(2) Jeong, H.; Cho, C. H.; Kim, T. H. React. Kinet. Mech. Catal. 2012, 517.
106, 435−443. (41) Göpel, W.; Bauer, R. S.; Hansson, G. Surf. Sci. 1980, 99, 138−
(3) Yu, K. M. K.; Yeung, C. M. Y.; Tsang, S. C. J. Am. Chem. Soc. 158.
2007, 129, 6360−6361. (42) Esser, P.; Göpel, W. Surf. Sci. 1980, 97, 309−318.
(4) Zha, F.; Ding, J.; Chang, Y.; Ding, J.-F.; Wang, J.-Y.; Ma, J. Ind. (43) Bowker, M.; Houghton, H.; Waugh, K. C. J. Chem. Soc., Faraday
Eng. Chem. Res. 2012, 51, 345−352. Trans. 1 1981, 77, 3023−3036.
(5) Meunier, F. C. Angew. Chem., Int. Ed. 2011, 50, 4053−4054. (44) Davis, R.; Walsh, J. F.; Muryn, C. A.; Thornton, G.; Dhanak, V.
(6) Liao, F.-L.; Huang, Y.-Q.; Ge, J.-W.; Zheng, W.-R.; Tedsree, K.; R.; Prince, K. C. Surf. Sci. Lett. 1993, 298, L196−L202.
Collier, P.; Hong, X.-L.; Tsang, S. C. Angew. Chem., Int. Ed. 2011, 50, (45) Wang, Y.-M.; Kovácǐ k, R.; Meyer, B.; Kotsis, K.; Stodt, D.;
2162−2165. Staemmler, V.; Qiu, H.-S.; Traeger, F.; Langenberg, D.; Muhler, M.;
(7) Ma, L.-Y.; Tran, T.; Wainwright, M. S. Top. Catal. 2003, 22, Wöll, C. Angew. Chem., Int. Ed. 2007, 46, 5624−5627.
287−293. (46) Wang, Y.-M.; Xia, X.-Y.; Urban, A.; Qiu, H.-S.; Strunk, J.; Meyer,
(8) Song, C.-S. Catal. Today 2006, 115, 2−32. B.; Muhler, M.; Wöll, C. Angew. Chem., Int. Ed. 2007, 46, 7315−7318.
(9) Liang, X.-L.; Dong, X.; Lin, G.-D.; Zhang, H.-B. Appl. Catal. B: (47) Hotan, W.; Göpel, W.; Haul, R. Surf. Sci. 1979, 83, 162−180.
Environ. 2009, 88, 315−322. (48) Noei, H.; Wöll, C.; Muhler, M.; Wang, Y.-M. Appl. Catal. A:
(10) Ma, Y.; Sun, Q.; Wu, D.; Fan, W.-H.; Zhang, Y.-L.; Deng, J.-F. Gen. 2011, 391, 31−35.
Appl. Catal. A: Gen. 1998, 171, 45−55. (49) Koßmann, J.; Roßmüller, G.; Hättig, C. J. Chem. Phys. 2012,
(11) Yang, Y.-X.; Evans, J.; Rodriguez, J. A.; White, M. G.; Liu, P. 136, 034706.
Phys. Chem. Chem. Phys. 2010, 12, 9909−9917. (50) Dutta, G.; Sokol, A. A.; Catlow, C. R. A.; Keal, T. W.; Sherwood,
(12) Zhang, R.-G.; Liu, H.-Y.; Wang, B.-J.; Ling, L.-X. Appl. Catal. B: P. ChemPhysChem 2012, 13, 3453−3456.
Environ. 2012, 126, 108−120. (51) Fink, K. Phys. Chem. Chem. Phys. 2006, 8, 1482−1489.
(13) Saito, M.; Murata, K. Catal. Surv. Asia 2004, 8, 285−294. (52) Martins, J. B. L.; Longo, E.; Taft, C. A. Int. J. Quantum Chem.
(14) Wambach, J.; Baiker, A.; Wokaun, A. Phys. Chem. Chem. Phys. 1998, 70, 367−374.
1999, 1, 5071−5080. (53) Martins, J. B. L.; Longo, E.; Salmon, O. D. R.; Espinoza, V. A.
(15) Spencer, M. S. Top. Catal. 1999, 8, 259−266. A.; Taft, C. A. Chem. Phys. Lett. 2004, 400, 481−486.
(16) Au, C. T.; Hirsch, W.; Hirschwald, W. Surf. Sci. 1988, 197, 391− (54) Farias, S. A. S.; Longo, E.; Gargano, R.; Martins, J. B. L. J. Mol.
401. Model. 2013, 19, 2069−2078.
(17) Nakamura, J.; Choi, Y.; Fujitani, T. Top. Catal. 2003, 22, 277− (55) Wang, J.; Funk, S.; Burghaus, U. Catal. Lett. 2005, 103, 219−
285. 223.
(18) Studt, F.; Abild-Pedersen, F.; Varley, J. B.; Nørskov, J. K. Catal. (56) Wöll, C. Prog. Surf. Sci. 2007, 82, 55−120.
Lett. 2013, 143, 71−73. (57) Ignaczak, A.; Gomes, J. A. N. F. J. Electroanal. Chem. 1997, 420,
(19) Nakamura, J.; Nakamura, I.; Uchijima, T.; Kanai, Y.; Watanabe, 209−218.
T.; Saito, M.; Fujitani, T. J. Catal. 1996, 160, 65−75. (58) Ruuska, H.; Pakkanen, T. A.; Rowley, R. L. J. Phys. Chem. B
(20) Strunk, J.; Kähler, K.; Xia, X.-Y.; Muhler, M. Surf. Sci. 2009, 603, 2004, 108, 2614−2619.
1776−1783. (59) Erhart, P.; Albe, K.; Klein, A. Phys. Rev. B 2006, 73, 205203.

L dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX


The Journal of Physical Chemistry C Article

(60) Calzolari, A.; Ruini, A.; Catellani, A. J. Am. Chem. Soc. 2011, 133, (103) Mattsson, A. E.; Schultz, P. A.; Desjarlais, M. P.; Mattsson, T.
5893−5899. R.; Leung, K. Model. Simul. Mater. Sci. Eng. 2005, 13, R1−R31.
(61) Anisimov, V. I.; Zaanen, J.; Andersen, O. K. Phys. Rev. B 1991, (104) Pan, Z.-W.; Dai, Z.-R.; Wang, Z.-L. Science 2001, 291, 1947−
44, 943−954. 1949.
(62) Solovyev, I. V.; Dederichs, P. H.; Anisimov, V. I. Phys. Rev. B (105) Staemmler, V.; Fink, K.; Meyer, B.; Marx, D.; Kunat, M.; Girol,
1994, 50, 16861−16871. S. G.; Burghaus, U.; Wöll, C. Phys. Rev. Lett. 2003, 90, 106102.
(63) Cococcioni, M.; de Gironcoli, S. Phys. Rev. B 2005, 71, 035105. (106) Meyer, B. Phys. Rev. B 2004, 69, 045416.
(64) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, 558−561. (107) Reuter, K.; Scheffler, M. Phys. Rev. B 2001, 65, 035406.
(65) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 48, 13115−13118. (108) Łodziana, Z.; Nørskov, J. K.; Stoltze, P. J. Chem. Phys. 2003,
(66) Kresse, G.; Hafner, J. Phys. Rev. B 1994, 49, 14251−14269. 118, 11179−11188.
(67) Kresse, G.; Furthmüller, J. Comput. Mater. Sci. 1996, 6, 15−50. (109) Pacchioni, G. Surf. Sci. 1993, 281, 207−219.
(68) Kresse, G.; Furthmüller, J. Phys. Rev. B 1996, 54, 11169−11186. (110) Tai, J.-R.; Ge, Q.-F.; Davis, R. J.; Neurock, M. J. Phys. Chem. B
(69) Perdew, J. P.; Wang, Y. Phys. Rev. B 1992, 45, 13244−13249. 2004, 108, 16798−16805.
(70) Wesolowski, T. A.; Parisel, O.; Ellinger, Y.; Weber, J. J. Phys. (111) Baniecki, J. D.; Ishii, M.; Kurihara, K.; Yamanaka, K.; Yano, T.;
Chem. A 1997, 101, 7818−7825. Shinozaki, K.; Imada, T.; Nozaki, K.; Kin, N. Phys. Rev. B 2008, 78,
(71) Grinberg, I.; Yourdshahyan, Y.; Rappe, A. M. J. Chem. Phys. 195415.
2002, 117, 2264−2270. (112) Redhead, P. A. Vacuum 1962, 12, 203−211.
(72) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; (113) Alfè, D.; Gillan, M. J. J. Phys.: Condens. Matter 2006, 18, L451−
Sutton, A. P. Phys. Rev. B 1998, 57, 1505−1509. L457.
(73) Blöchl, P. E. Phys. Rev. B 1994, 50, 17953−17979. (114) Pozzo, M.; Carlini, G.; Rosei, R.; Alfè, D. J. Chem. Phys. 2007,
(74) Kresse, G.; Joubert, D. Phys. Rev. B 1999, 59, 1758−1775. 126, 164706.
(75) Blöchl, P. E.; Jepsen, O.; Andersen, O. K. Phys. Rev. B 1994, 49, (115) Tang, Q.-L. Int. J. Quantum Chem. 2013, 113, 1992−2001.
16223−16233. (116) Rohrbach, A.; Hafner, J. Phys. Rev. B 2005, 71, 045405.
(76) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188−5192. (117) Huang, M.; Fabris, S. J. Phys. Chem. C 2008, 112, 8643−8648.
(77) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, (118) Edwards, J. F.; Schrader, G. L. J. Phys. Chem. 1984, 88, 5620−
3865−3868. 5624.
(78) Palacios, P.; Sánchez, K.; Wahnón, P. Thin Solid Films 2009, (119) Lei, Y.-H.; Chen, Z.-X. J. Phys. Chem. C 2012, 116, 25757−
517, 2448−2451. 25764.
(79) Decremps, F.; Datchi, F.; Saitta, A. M.; Polian, A.; Pascarelli, S.; (120) Wang, S.-G.; Liao, X.-Y.; Cao, D.-B.; Huo, C.-F.; Li, Y.-W.;
Di Cicco, A.; Itié, J. P.; Baudelet, F. Phys. Rev. B 2003, 68, 104101. Wang, J.-G.; Jiao, H.-J. J. Phys. Chem. C 2007, 111, 16934−16940.
(80) Ö zgür, Ü .; Alivov, Y. I.; Liu, C.; Teke, A.; Reshchikov, M. A.; (121) Newns, D. M. Phys. Rev. 1969, 178, 1123−1135.
Doğan, S.; Avrutin, V.; Cho, S.-J.; Morkoc, H. J. Appl. Phys. 2005, 98, (122) Wang, C.-M.; Fan, K.-N.; Liu, Z.-P. J. Am. Chem. Soc. 2007,
041301. 129, 2642−2647.
(81) Meyer, B.; Marx, D. Phys. Rev. B 2004, 69, 235420. (123) Bader, R. F. W. Acc. Chem. Res. 1985, 18, 9−15.
(82) Yang, Z.; Xiong, S.-J. Surf. Sci. 2011, 605, 40−45. (124) Henkelman, G.; A, A.; Jónsson, H. Comput. Mater. Sci. 2006,
(83) Neugebauer, J.; Scheffler, M. Phys. Rev. B 1992, 46, 16067− 36, 354−360.
16080. (125) Lide, D. R. CRC Handbook of Chemistry and Physics, 84th ed.;
(84) Makov, G.; Payne, M. C. Phys. Rev. B 1995, 51, 4014−4022. CRC Press: Boca Raton, FL, 2003−2004.
(85) Rapcewicz, K.; Chen, B.; Yakobson, B.; Bernholc, J. Phys. Rev. B (126) Michaelides, A.; Hu, P.; Lee, M.-H.; Alavi, A.; King, D. A. Phys.
1998, 57, 7281−7291. Rev. Lett. 2003, 90, 246103.
(86) Siegel, D. J.; Hector, L. G.; Adams, J. B. Surf. Sci. 2002, 498, (127) Leung, T. C.; Kao, C. L.; Su, W. S.; Feng, Y. J.; Chan, C. T.
321−336. Phys. Rev. B 2003, 68, 195408.
(87) Zhao, J.-P.; Wang, X.; Chen, Z.-Y.; Yang, S.-Q.; Shi, T.-S.; Liu, (128) Breedon, M.; Spencer, M. J. S.; Yarovsky, I. Surf. Sci. 2009, 603,
X.-H. J. Phys. D: Appl. Phys. 1997, 30, 5−12. 3389−3399.
(88) Hugosson, H. W.; Eriksson, O.; Jansson, U.; Ruban, A. V.; (129) Fall, C. J.; Binggeli, N.; Baldereschi, A. J. Phys.: Condens. Matter
Souvatzis, P.; Abrikosov, I. A. Surf. Sci. 2004, 557, 243−254. 1999, 11, 2689−2696.
(89) Boettger, J. C. Phys. Rev. B 1994, 49, 16798−16800. (130) Wood, C.; Li, H.; Winget, P.; Brédas, J.-L. J. Phys. Chem. C
(90) Gay, J. G.; Smith, J. R.; Richter, R.; Arlinghaus, F. J.; Wagoner, 2012, 116, 19125−19133.
R. H. J. Vac. Sci. Technol. A 1984, 2, 931−932. (131) Kohl, D.; Moorman, H.; Heiland, G. Surf. Sci. 1978, 73, 160−
(91) Fiorentini, V.; Methfessel, M. J. Phys.: Condens. Matter 1996, 8, 162.
6525−6529. (132) Moormann, H.; Kohl, D.; Heiland, G. Surf. Sci. 1979, 80, 261−
(92) Valero, M. C.; Raybaud, P.; Sautet, P. J. Catal. 2007, 247, 339− 264.
(133) Moormann, H.; Kohl, D.; Heiland, G. Surf. Sci. 1980, 100,
355.
302−314.
(93) Morin, C.; Simon, D.; Sautet, P. J. Phys. Chem. B 2003, 107,
2995−3002.
(94) Masel, R. I. Principles of Adsorption and Reactions on Solid
Surfaces; John Wiley & Sons: NY, 1996.
(95) Meyer, B.; Marx, D. Phys. Rev. B 2003, 67, 035403.
(96) Wander, A.; Schedin, F.; Steadman, P.; Norris, A.; McGrath, R.;
Turner, T. S.; Thornton, G.; Harrison, N. M. Phys. Rev. Lett. 2001, 86,
3811−3814.
(97) Wander, A.; Harrison, N. M. Surf. Sci. 2000, 457, L342−L346.
(98) Wander, A.; Harrison, N. M. Surf. Sci. 2000, 468, L851−L855.
(99) Perdew, J. P.; Zunger, A. Phys. Rev. B 1981, 23, 5048−5079.
(100) Becke, A. D. J. Chem. Phys. 1993, 98, 5648−5652.
(101) Lee, C.; Yang, W.-T.; Parr, R. G. Phys. Rev. B 1988, 37, 785−
789.
(102) Kurth, S.; Perdew, J. P.; Blaha, P. Int. J. Quantum Chem. 1999,
75, 889−909.

M dx.doi.org/10.1021/jp407970a | J. Phys. Chem. C XXXX, XXX, XXX−XXX

You might also like