вход по аккаунту


Partial Dissociation of Water Leads to Stable Superstructures on the Surface of Zinc Oxide.

код для вставкиСкачать
Contrary to what was previously believed, water molecules do not
adsorb intact on the surfaces of zinc oxide, but dissociate by an
autocatalytic process and form a complex superstructure consisting of
intact water molecules and hydroxy species. For more details, see the
Communication by B. Meyer, C. Wll et al. on the following pages.
Angew. Chem. Int. Ed. 2004, 43, 6641
DOI: 10.1002/anie.200461696
2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Water Structure
Partial Dissociation of Water Leads to Stable
Superstructures on the Surface of Zinc Oxide**
Bernd Meyer,* Dominik Marx, Olga Dulub,
Ulrike Diebold, Martin Kunat, Deler Langenberg, and
Christof Wll*
Understanding wet surfaces and water/solid interfaces is of
crucial importance in such diverse fields as corrosion,
catalysis, mineralogy, and geology. For H2O, the delicate
interplay between chemical bonding, van der Waals forces,
and hydrogen bonding gives rise to complex phenomena such
as complete dissociation, partial dissociation at defects,
multilayer formation, and wetting.[1] Recently, an intriguing
intermediate scenario was advanced, in which the interaction
between the water molecules results in a partial dissociation
of water on perfect surfaces, leading to superlattices with
long-range order.[2–5] In view of possible failures of common
theoretical approaches and severe experimental difficulties[5]
this proposal has become the subject of a highly controversial
debate.[5–9] For a given surface a consistent picture can only be
obtained by using a broad array of experimental and
computational tools in concert. Applying such a strategy, we
unambiguously demonstrate that water forms a simple,
unusually stable two-dimensional superstructure on defectfree zinc oxide surfaces. Diffraction (He-atom scattering
(HAS), low-energy electron diffraction (LEED)), real-space
(scanning tunneling microscopy (STM)), and thermodynamic
(He thermal desorption spectroscopy (He-TDS)) data show
that water forms a (2 4 1) superstructure with long-range
order that exists up to temperatures close to the boiling point
of water. Car–Parrinello ab initio simulations[10] reveal a
superlattice in which every second water molecule is dissociated. The unusual thermal stability of this superstructure is a
result of a favorable “key–lock” type structural arrangement
in conjunction with hydrogen bonding. The computed struc-
[*] Dr. B. Meyer, Prof. Dr. D. Marx
Lehrstuhl f1r Theoretische Chemie
Ruhr-Universit5t Bochum
44780 Bochum (Germany)
Fax: (+ 49) 234-32-14045
Dr. M. Kunat, Dipl.-Chem. D. Langenberg, Prof. Dr. C. WCll
Lehrstuhl f1r Physikalische Chemie I
Ruhr-Universit5t Bochum
44780 Bochum (Germany)
Fax: (+ 49) 234-32-14182
Dr. O. Dulub, Prof. U. Diebold
Department of Physics
Tulane University
New Orleans, LA 70118 (USA)
[**] We thank Prof. Volker Staemmler for discussions and for carrying
out the coupled-cluster reference calculations on the water dimers.
This work was supported by the DFG (SFB 558), FCI, NSF
(CHE-010908), and NASA.
2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ture and binding energy are in excellent agreement with those
found in the experiment.
Zinc oxide is a material for which significant progress in
understanding its surface properties was achieved
recently,[11–17] stimulated by its importance for a number of
applications ranging from cosmetics and medicine to paints
and heterogeneous catalysis. The nonpolar ZnO (101̄0)
surface chosen for the present study is, in contrast to its
polar counterparts,[13, 16] electrostatically stable and as such
now well understood.[12] Briefly, this Wurtzite-type surface
consists of layers containing slightly tilted ZnO “dimers”,
which are formed by three-fold coordinated Zn and O ions.
These ZnO dimers, in turn, assemble to form characteristic
rows separated by trenches (Figure 1). For the adsorption of
Figure 1. Side and top view of the atomic structure of the molecular
adsorbed water monolayer with (1 1) symmetry (left) and the optimized half-dissociated H2O/ZnO(101̄0) adsorption geometry (right).
The surface unit cells are shown by dashed lines. Zn, O, and H atoms
are represented by grey, red, and white spheres, respectively, the O
atoms of the water molecules are in blue. In the top views, the atoms
in the second surface layer are indicated by lighter colors.
water on this surface one would expect that the oxygen atoms
of the water molecules bind to the coordinatively unsaturated
Zn ions on the surface. In a previous photoelectron spectroscopy study[18] it was indeed concluded that H2O chemisorbs at
the Zn sites of the ZnO(101̄0) surface. Since the distance
between the Zn sites is fairly large (3.25 B versus 2.78 B
between H2O molecules in ice Ih), one would predict the
formation of a simple water (1 4 1) monolayer.
In contrast to this prediction, we find that water on such a
ZnO surface forms an ordered superstructure with a (2 4 1)
periodicity. The He atom angular distributions (Figure 2) as
well as LEED data clearly show sharp and well-defined halforder diffraction peaks. Also high-resolution STM measurements revealed a (2 4 1) periodicity of the adlayer. These
findings effectively rule out that water is adsorbed on the
ZnO(101̄0) surface as a full monolayer of equivalent H2O
molecules as assumed previously. A (2 4 1) structure would
come about if water would only occupy every other Zn site,
with a total coverage of half a monolayer. If the water
molecules were separated by two lattice constants, no hydro-
DOI: 10.1002/anie.200461696
Angew. Chem. Int. Ed. 2004, 43, 6642 –6645
Figure 2. He-atom specular reflectivity of a ZnO(101̄0) surface saturated with water at a temperature of 280 K (solid line) recorded for a
heating rate of 1 K s 1. The dashed line shows the corresponding derivative, revealing a desorption maximum at 367 K. Inset: He-atom diffraction patterns for the clean (top) and water-saturated (bottom)
ZnO(101̄0) surface recorded along the [011̄0] azimuth. The (2 1)
periodicity of the water adlayer is clearly revealed by the prominent
half-order diffraction peaks.
gen bonds between the water molecules would be present.
However, the STM measurements show that for low water
coverages the adsorbates form ordered two-dimensional
islands (Figure 3), which implies a strong lateral attraction
between adsorbed species. In addition, a quantitative determination of the H2O coverage using X-ray photoelectron
spectroscopy (XPS) shows that, at full coverage, the surface
Figure 3. Left: STM image of a ZnO(101̄0) surface after dosing 1 Langmuir of water at room temperature (tunneling conditions: + 1.6 V
sample bias and 4.7 nA tunneling current). The water molecules form
well-ordered two-dimensional islands with a (2 1) periodicity (bright
spots) on the ZnO substrate (faint rows). The solid line shows the
position where the line profile was taken. Right: Simulated STM
images for the (hypothetical) molecular water monolayer with (1 1)
symmetry (top) and the half-dissociated H2O/ZnO(101̄0) structure
(bottom). The height of the tip above the surface increases from blue
through green to red. Bands in the energy range of 1 eV above the
band gap are taken into account, mimicking positive sample bias and
tunneling into the conduction band.
Angew. Chem. Int. Ed. 2004, 43, 6642 –6645
contains one H2O molecule per surface Zn ion (see Experimental Section), so that a (2 4 1) structure caused by only
every second Zn site being occupied can be ruled out.
Important additional information about the adsorption of
water can be gained from the binding energy of the molecules
on the surface. The reliable determination of this quantity on
metal oxide surfaces is, however, generally hampered by the
presence of defects. In fact, most controversies on the
question whether water adsorbs in a molecular form or
dissociatively on a perfect surface stem from the problem that
water may dissociate at defect sites, in particular at oxygen
vacancies, which contributes to the overall signal.[1] To rule
out contributions of defects, a technique tailored to studying
adsorbates on well-defined solid substrates was employed,
namely the scattering of thermal energy He atoms. When
monitoring the reflectivity of the ZnO(101̄0) surface for
thermal energy He atoms as a function of surface temperature
(He-TDS) the desorption from perfect areas can be studied
exclusively, allowing a precise determination of the adsorbate
binding energies.[13] He-TDS has been used previously to
detect the desorption of water from oxide surfaces such as
MgO[19, 20] or of CO and H2 from ZnO surfaces.[13, 21] The HeTDS data (Figure 2) reveal a sharp, well-defined desorption
maximum close to the boiling point of water at a temperature
of 367 K. This corresponds to a binding energy of 1.02 eV, in
excellent agreement with a value reported previously.[18] This
desorption temperature is astonishingly high. In general, such
high desorption temperatures were only observed for a few
other metal oxide surfaces for which it is known that water
dissociates upon adsorption, for example, the Cr2O3 (0001)[22]
and the (2 4 1) a-Fe2O3(011̄2) surface.[23] It therefore appears
to be likely that also on this ZnO surface H2O dissociates,
hydroxylating a substrate O atom and adding an OH species
on top of a Zn–substrate atom. Such a dissociation may in fact
lead to a (2 4 1) structure. Interestingly, we observed the
formation of the (2 4 1) phase already at temperatures as low
as 200 K, a fact which would imply a very low activation
energy for dissociation.
The precise value for the binding energy of H2O adsorbed
on defect-free regions of the ZnO(101̄0) surface (the coherence length in the He-TDS measurement was of the order of
150 B) opens the pathway for a direct comparison to and
validation of theoretical calculations, which then in turn will
provide the basis to rationalize the different experimental
findings. To this end density functional theory (DFT)
calculations were carried out within a periodic slab setup
using six to eight ZnO layers. Special care was taken to
electrostatically decouple the slabs to suppress artificial
interactions.[24] However, applying DFT to subtle molecule/
surface interactions is often questionable in view of possible
deficiencies for describing hydrogen bonding and van der
Waals interactions. Thus, the gradient-corrected Perdew,
Burke, and Ernzerhof (PBE) functional used here in conjunction with ultrasoft pseudopotentials and a 25 Rydberg
plane wave cutoff (see ref. [12] for further details) was
thoroughly tested for the system of interest. It is well known
that the PBE functional describes very accurately the
equilibrium structure and binding energy of water in isolated
dimers and ice Ih.[4, 26] However, to probe the reliability of PBE
2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
for water separations and orientations that are relevant here,
we have calculated the binding energy of nearest and nextnearest water pairs as they appear in the undissociated (1 4 1)
water monolayer on the ZnO(101̄0) surface (see below). For
these pairs with O O separations of 3.25 and 5.21 B, dimer
binding energies of 0.085 and 0.011 eV, respectively, were
found. This is in astonishing agreement with the results of
0.085 and 0.016 eV of very accurate coupled cluster-type (MCCEPA[25]) calculations. A similar accuracy was also previously
demonstrated to hold for weakly bound molecules and
adatoms on small ZnO clusters.[13, 14, 27] Finally, the structure
of bulk ZnO including its low-index surfaces has been
calculated reliably within the described DFT/PBE setup.[12]
Based on all these gauges, it can be concluded that the PBE
functional should capture the relevant interactions.
The adsorption properties of water on the ZnO(101̄0)
surface was studied for various coverages starting from a full
monolayer and reducing to 1/2, 1/4, and 1/6 of a monolayer. In
a first set of calculations, (1 4 1), (2 4 1), (2 4 2), and (3 4 2)
surface unit cells containing one water molecule were used.
This implies that the water molecules are kept equivalent by
imposing translational symmetry. Under this constraint, only
molecular adsorption of water is found for all coverages with
a very similar orientation of the molecules (Figure 1).
Dissociative adsorption, which may have been expected
from the experimentally observed high desorption temperature, is not stable within this constraint. At a coverage of 1=4
and 1=6 monolayer the H2O molecules basically do not
interact, and the binding energy is 0.94 eV per molecule. It
increases to 0.97 and 1.03 eV at half and full coverage,
respectively. Despite the molecular adsorption, the binding
energies are large and in good general agreement with the
experimental value. This unexpectedly strong binding is due
to a “key–lock” type interaction between the water molecules
and the ZnO surface. The water molecules are stabilized by
three different types of attractive interactions to both the
substrate and neighboring adsorbate molecules. 1) The O
atoms of the water molecules occupy the O sites of a
hypothetical next ZnO layer on the ZnO(101̄0) surface so that
the surface Zn ions regain their four-fold tetrahedral coordination as in the underlying bulk, which leads to a strong Zn
O and thus ZnO/H2O bonding. 2) One of the H atoms forms a
hydrogen bond across the ZnO trench of the surface to a
neighboring substrate O atom. 3) In the case of a monolayer
coverage, a water–water hydrogen bond to a neighboring
adwater is formed by the second H atom. Thus, the full
monolayer water film is significantly more stable than the half
monolayer, in agreement with the XPS and STM data.
Although the key–lock configuration of a full monolayer
of molecular water is a low-energy structure, it does not
exhibit the (2 4 1) symmetry observed experimentally. Thus,
in a second set of calculations, we explored the possibility that
the (2 4 1) structure appears because of differently oriented
water molecules. Two water molecules were randomly
displaced in a (2 4 1) surface unit cell and a structural
relaxation was performed. In all cases the adsorbed water
molecules moved back toward the (1 4 1) monolayer structure
and none of the modifications yielded an energy per water
molecule lower than that obtained for the (1 4 1) coverage.
2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
This picture changed when we explicitly considered the
dynamics of the water molecules in full ab initio Car–
Parrinello molecular dynamics (CPMD) simulations.[10] In this
approach, the system can evolve to a more stable structure, if
there is any, without much bias. We set up slabs with (3 4 2)
and (4 4 2) surface unit cells containing six and eight water
molecules, respectively. Starting from the fully relaxed water
monolayer structure with (1 4 1) symmetry, spontaneous
dissociation of a subset of water molecules was observed at
a simulation temperature of 300 K. After the initial dissociation the structures were stable and no further proton
transfers occurred. Since the dissociation starts immediately
after the beginning of the CPMD simulations, the activation
energy has to be very low, which explains why the (2 4 1)
phase already appears at temperatures below 200 K in the
Upon further structural optimization, the most stable
structure turned out to be a configuration in which every
second water molecule is dissociated, leading to a (2 4 1)
superstructure (Figure 1). Interestingly, the undissociated
water molecules remain in the stabilizing key–lock position
akin to the molecular (1 4 1) structure described above. The
molecules tilt a bit more toward the surface and thereby
strengthen the hydrogen bond to their neighboring molecules.
In every second water molecule one OH bond is cleaved and
the H atom is transferred to a substrate O atom. Thus, two
hydroxy groups are formed, one built from an adwater O
atom and the second one originating from a substrate O atom.
For this half-dissociated (2 4 1) structure, the overall binding
energy per H2O molecule is found to be slightly, but
significantly, increased with regard to that of the molecular
(1 4 1) structure (from 1.03 to 1.13 eV, respectively). This is
one of the highest binding energies for water on structurally
well-defined solid substrates.[1] It is noteworthy that the
interaction between the water molecules themselves “catalyzes” the dissociation reaction. As pointed out above,
isolated water molecules do not dissociate on the
ZnO(101̄0) surface. However, calculations show that putting
a second water molecule at a neighboring lattice site is
already sufficient to trigger the dissociation. Between the two
water molecules a hydrogen bond is formed that weakens the
already activated OH bond across the ZnO trences in the
acceptor molecule. The molecule becomes unstable and the
proton is transferred to a substrate O atom without noticeable
activation energy.
To investigate if the half-dissociated monolayer structure
is consistent with the observations in the STM experiments,
STM images for both, the half-dissociated (2 4 1) and the
molecular monolayer (1 4 1) were calculated by using the
Tersoff–Hamann approximation.[28] The experimental STM
image in Figure 3 was recorded with a positive sample bias
voltage, that is, under conditions where electrons tunnel into
the conduction band (i.e. the LUMO). Under these conditions, the clean ZnO(101̄0) surface is characterized by faint
stripes that are attributed to the ZnO dimers (see Figure 1).
The bright spots in the (2 4 1) superstructure of the H2O
islands are in registry with these stripes. The theoretical STM
images reveal easily distinguishable differences between the
two different adsorbate structures (Figure 3). The pattern
Angew. Chem. Int. Ed. 2004, 43, 6642 –6645
stemming from the half-dissociated superstructure matches
the experimental STM image, which is clearly not the case for
the molecular adlayer. The bright features in the image of the
partially dissociated water layer are due to the OH groups of
the dissociated water molecules, with a contrast that is
predominantly geometric in nature.
Based on this consistent evidence, it can be concluded
convincingly that half of the water molecules on the fully
covered perfect ZnO(101̄0) surface self-dissociate resulting in
a well-defined (2 4 1) superlattice. This superstructure is
characterized by long-range order and structural simplicity
but in particular by an unusually large stability range
extending from 200 K up to the boiling point of liquid
water. Transcending the specific case, these findings imply
that when ionic materials such as minerals come in contact
with water the possibility of water auto-dissociation has
always to be considered both in experiment and theory—
without necessarily invoking defects! This is of particular
relevance to heterogeneous catalysis so that a number of
reactions, such as methanol synthesis on ZnO or Cu/ZnO in
which water is a side product, have to be reanalyzed.
Experimental Section
The measurements were carried out with an ultrahigh vacuum (UHV)
molecular beam system that was described in detail in reference [21].
The mixed terminated ZnO(101̄0) substrates were oriented to within
0.28 and polished mechanically. After installation in the UHV
chamber the samples were first cleaned by cycles of Ar+ sputtering
(800 V, 1 mA, 4–12 h, T = 650 K). After about 20 preparation cycles
the XP spectra revealed contamination levels of less than 0.05 ML for
carbon-containing species, and the surfaces exhibited (1 4 1) diffraction peaks in HAS and LEED. The surface was exposed to water
either by backfilling the UHV chamber using a leak valve or by
dosing water through a capillary that ended about 5 cm in front of the
ZnO substrate.
The STM experiments were carried out in a separate UHV
chamber equipped with low-energy He+-ion scattering, LEED, and
an STM operating at room temperature. The sample was prepared by
sputtering (1 keV Ar+, 30 min) and annealing (1000 K, 15 min) cycles.
Water was dosed by backfilling the UHV chamber.
Coverage calibration was carried out by ratioing the O1s signal
seen for the H2O monolayer at 533.5 eV to that measured for a (2 4 1)
CO2 adlayer (with each (2 4 1) unit cell containing one CO2 molecule)
at the same energy. The intensity ratio O1sOH/O1sCO2 was consistent
with the presence of three OH species and one CO2 molecule per
(2 4 1) unit cell. The activation energy for desorption was determined
by using the Redhead formula with the experimental heating rate of
1 K s 1 and assuming first-order desorption with a preexponential
factor of 1013 4 1 s 1.
[5] D. Menzel, Science 2002, 295, 58 – 59; Correction: D. Menzel,
Science 2002, 296, 264.
[6] S. Meng, L. F. Xu, E. G. Wang, S. W. Gao, Phys. Rev. Lett. 2002,
89, 176 104; comment by P. J. Feibelman, Phys. Rev. Lett. 2003,
91, 059 601; reply: S. Meng, L. F. Xu, E. G. Wang, S. W. Gao,
Phys. Rev. Lett. 2003, 91, 059 602.
[7] H. Ogasawara, B. Brena, D. Nordlund, M. Nyberg, A. Pelmenschikov, L. G. M. Petterson, A. Nilsson, Phys. Rev. Lett. 2002, 89,
276 102.
[8] R. Ludwig, Angew. Chem. 2003, 115, 3580 – 3582; Angew. Chem.
Int. Ed. 2003, 42, 3458 – 3460.
[9] C. Clay, S. Haq, A. Hodgson, Phys. Rev. Lett. 2004, 92, 046 102.
[10] D. Marx, J. Hutter in Modern Methods and Algorithms of
Quantum Chemistry (Ed.: J. Grotendorst), pp. 301–449 (NIC, FZ
JOlich), 2000, see:; CPMD
code: J. Hutter et al., see:
[11] A. Wander, F. Schedin, P. Steadman, A. Norris, R. McGrath, T. S.
Turner, G. Thornton, N. M. Harrison, Phys. Rev. Lett. 2001, 86,
3811 – 3814.
[12] B. Meyer, D. Marx, Phys. Rev. B 2003, 67, 035 403; B. Meyer, D.
Marx, Phys. Rev. B 2003, 67, 039 902(E).
[13] V. Staemmler, K. Fink, B. Meyer, D. Marx, M. Kunat, U.
Burghaus, S. Gil Girol, C. WPll, Phys. Rev. Lett. 2003, 90, 106 102.
[14] B. Meyer, D. Marx, J. Phys. Condens. Matter 2003, 15, L89 – L94.
[15] B. Meyer, Phys. Rev. B 2004, 69, 045 416.
[16] O. Dulub, U. Diebold, G. Kresse, Phys. Rev. Lett. 2003, 90,
016 102.
[17] G. Kresse, O. Dulub, U. Diebold, Phys. Rev. B 2003, 68, 245 409.
[18] G. Zwicker, K. Jacobi, Surf. Sci. 1983, 131, 179.
[19] D. Ferry, A. Glebov, V. Senz, J. Suzanne, J. P. Toennies, H. Weiss,
J. Chem. Phys. 1996, 105, 1697 – 1701.
[20] D. Ferry, A. Glebov, V. Senz, J. Suzanne, J. P. Toennies, H. Weiss,
Surf. Sci. 1997, 377–379, 634 – 638.
[21] T. Becker, S. HPvel, M. Kunat, C. Boas, U. Burghaus, C. WPll,
Surf. Sci. 2001, 486, L502 – L506; Erratum: T. Becker, S. HPvel,
M. Kunat, C. Boas, U. Burghaus, C. WPll, Surf. Sci. 2001, 511,
[22] M. A. Henderson, S. A. Chambers, Surf. Sci. 2000, 449, 135 – 150.
[23] M. A. Henderson, S. A. Joyce, J. R. Rustad, Surf. Sci. 1998, 417,
66 – 81.
[24] B. Meyer, D. Vanderbilt, Phys. Rev. B 2001, 63, 205 426.
[25] R. Fink, V. Staemmler, Theor. Chim. Acta 1993, 87, 129.
[26] D. R. Hamann, Phys. Rev. B 1997, 55, R10 157 – R10 160.
[27] B. Meyer, D. Marx, Phys. Rev. B 2004, 69, 235 420.
[28] J. Tersoff, D. R. Hamann, Phys. Rev. B 1985, 31, 805 – 813.
Received: August 18, 2004
Keywords: ab initio calculations · surfaces · water structure ·
zinc oxide
[1] M. A. Henderson, Surf. Sci. Rep. 2002, 46, 1 – 308.
[2] P. J. D. Lindan, N. M. Harrison, M. J. Gillan, Phys. Rev. Lett.
1998, 80, 762 – 765.
[3] L. Giordano, J. Goniakowski, J. Suzanne, Phys. Rev. Lett. 1998,
81, 1271 – 1273.
[4] P. J. Feibelman, Science 2002, 295, 99 – 102.
Angew. Chem. Int. Ed. 2004, 43, 6642 –6645
2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Без категории
Размер файла
307 Кб
water, oxide, surface, partial, dissociation, leads, superstructures, zinc, stable
Пожаловаться на содержимое документа