вход по аккаунту


Atomic-Scale Evidence for an Enhanced Catalytic Reactivity of Stretched Surfaces.

код для вставкиСкачать
NO molecules adsorbed on an Ru surface dissociate preferentially at
edge dislocations. This long-discussed effect was demonstrated with
the aid of scanning tunneling microscope images which show a higher
concentration of nitrogen molecules near such defects. For more
information see the communication by M. Mavrikakis and J. Wintterlin
et al. on the following pages.
Angew. Chem. Int. Ed. 2003, 42, 2849
DOI: 10.1002/anie.200250845
2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Watching NO Activation on a Surface
Atomic-Scale Evidence for an Enhanced Catalytic
Reactivity of Stretched Surfaces**
Joost Wintterlin,* Tomaso Zambelli, J. Trost,
Jeffrey Greeley, and Manos Mavrikakis*
It has recently been predicted that the catalytic activity of
transition metals will be modified by lattice strain.[1] Changes
of the lattice constant alter the d-band width, causing the dband center (with respect to the Fermi level) to move up or
down in energy to keep the d-occupancy constant, which
consequently modifies the chemical activity.[2] The effect
could be of far-reaching importance for heterogeneous
catalysis, because supported catalysts used in industrial
processes probably always exhibit some lattice strain,
caused, for example, by lattice defects or by interactions
with the support material.
Several experimental observations have suggested that
such strain effects actually exist. Older investigations found
that metals treated by cold-working formed dislocations and
displayed an enhanced catalytic activity,[3, 4] and recently a
correlation between activity and strain in small Cu particles of
a Cu/ZnO catalyst was found.[5, 6] In all of these cases it was
unclear if a lattice expansion or compression was responsible
for the reactivity changes or which elementary reaction step
was affected. A further observation was an enhanced
dissociation probability for oxygen on a heteroepitaxial
metal film,[7] although in a bimetallic system the strain
effect from the lattice mismatch is convoluted with electronic
interactions between film and substrate. The dissociative
sticking coefficient of O2 was also enhanced for a mechanically stretched Cu single crystal.[8] However, strain achieved
by this method was very limited because plastic deformations
had to be avoided, and the dissociative sticking coefficient
varied by only approximately 10 %. To play a role in catalysis,
strain must be large, which is probably the case only in the
[*] Prof. Dr. J. Wintterlin,+ Dr. T. Zambelli, Dr. J. Trost
14195 Berlin (Germany)
Prof. M. Mavrikakis, J. Greeley
Department of Chemical Engineering
University of Wisconsin
Madison, WI 53706-1691 (USA)
Fax: (+ 1) 608-262-5434
[þ] Current address:
Department Chemie
Univerisit?t M@nchen
81377 M@nchen (Germany)
Fax: (+ 49) 89-2180-77598
[**] T.Z. thanks the DAAD for financial support. M.M. thanks a NSFCAREER award (CTS-0134561), and NSF-NPACI and DOE-NERSC
for supercomputing time. We thank Lars Grabow for help with the
2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
immediate neighborhood of defects. The only microscopic
observation to date of a strain effect on adsorbates was that O
atoms adsorbed on a Ru surface were attracted to defects at
which the lattice was locally stretched.[9] This confirmed an
important aspect of the predicted strain effect, an influence
on the binding energy of adsorbates, but no chemical reaction
was involved.
Herein we present what we believe is the first microscopic
evidence for a strain effect on the chemical reactivity of a
molecule. By means of scanning tunneling microscopy (STM)
data we show that the dissociation probability of NO
molecules is significantly enhanced near the intersection
points of edge-dislocations with an Ru surface. The effect is
large and could be relevant for real catalytic reactions. The
behavior is in full agreement with results from density
functional theory (DFT) calculations.
The experiments were performed in an ultrahigh vacuum
chamber, in which also the surface of the (0001) oriented Ru
sample was cleaned.[10] In addition to monoatomic steps, the
sample surface displayed another type of structural defect
(Figure 1). The hexagonal pattern in Figure 1 is the (2 : 2)O
structure that was prepared by dosing the Ru surface with
oxygen. The atomic resolution of this ordered structure allows
one to construct a Burgers loop around the central defect. A
nonvanishing Burgers vector is found, which indicates that at
the image center an edge dislocation from the bulk intersects
the surface.[11]
The important effect for the present purpose is that
around each of these defects the lattice is deformed. On the
one side of each dislocation, where the additional half plane
is, the lattice is compressed (upper right in Figure 1), and on
the opposite side it is stretched (lower left in Figure 1). This
deformation primarily represents horizontal strain, parallel to
the surface. There is also a small vertical lattice deformation
that, because of the extreme vertical resolution power of
STM, becomes visible on higher magnification scans
(Figure 2). On these scales it becomes apparent that each of
the defects is surrounded by two extended regions, one of
which appears brighter than the surrounding surface, and the
other of which appears darker. These areas extend about
100 < into the undisturbed surface. The brightness variations
represent vertical deformations, where, at the bright side, the
surface protrudes approximately 0.2 < above the level of the
undisturbed lattice, and on the dark side it is approximately
0.2 < lower. These vertical deformations reflect a tendency of
the surface to try to relax some of the horizontal strain. On
the (horizontally) compressed side of each dislocation, the top
lattice layers bulge up a little bit (appearing brighter), and on
the stretched side they bend inward (appearing darker). This
relaxation is much smaller than the horizontal strain (which is
of the order of the Burgers vector b), but it can be seen as a
marker of the range of the horizontal deformation field
around each dislocation.
Dosing of NO on Ru(0001) (without the (2 : 2)O phase)
at 300 K leads to dissociation into N and O atoms.[12] Upon
adsorption, the NO molecules can diffuse across the surface
until they dissociate, mainly at defects (for example, at atomic
steps),[13] which indicates that the dissociation is an activated
process. The resulting N atoms are easily resolved by the STM
DOI: 10.1002/anie.200250845
Angew. Chem. Int. Ed. 2003, 42, 2850 – 2853
Figure 2. STM image of two edge dislocations, after exposure of 0.3 L
of NO on the clean Ru(0001) surface at 300 K
(1 L = 1.33 F 106 mbar s). The small dark dots are N atoms.
(440 E F 320 E). Inset: Edge dislocation from a different experiment,
after exposure of 0.1 L of NO (70 E F 50 E). At the dark, stretched parts
near the dislocations concentrations of N atoms are enhanced by factors of 7 in the main image and 18 in the inset.
Figure 1. STM image of the Ru(0001) surface (180 E F 240 E), covered
with a (2 F 2)O structure (hexagonal pattern). The Burgers loop around
the central defect (white dots, the length unit is the lattice constant of
the (2 F 2) structure) has an open end, which indicates that an edge
dislocation intersects the surface at the position of the defect. The Burgers vector b is half a (2 F 2) lattice constant long, that is, one lattice
constant of the substrate (inset lower right). The Burgers loop crosses
two domain boundaries in the (2 F 2) phase (broken lines), where the
dot markers switch positions relative to the O atoms of the overlayer
(see insets). This observation results from the fact that for an edge
dislocation in the (1 F 1) bulk the (2 F 2) overlayer has to develop
domain boundaries. The blurred area above the dislocation is caused
by fluctuations in the oxygen layer, which has a lower binding energy
on the compressed side of the dislocation..
as almost stationary dark dots, whereas O atoms are more
difficult to see at low coverages because of their greater
mobility.[14, 15] STM images of the edge dislocations, recorded
after exposure of small amounts of NO on the Ru surface
(Figure 2), reveal much larger concentrations of N atoms at
the dark, stretched sides than on the bright, compressed sides
or on defect-free terraces. The N and O atoms are easily
differentiated from each other based on their different
imaging by STM,[10] which allows the atoms on the dark
areas to be clearly identified as N atoms. Because of the low
mobility of the N atoms,[14] their distribution marks the
reaction area for the NO molecules. We conclude that the NO
dissociation probability is significantly enhanced on the
stretched part of the strain field of edge dislocations.
The magnitude of the horizontal lattice stretch responsible for the observed effect can be estimated from the
diameter of the vertically deformed area. A horizontal
distortion of the size b (2.70 <) has to be distributed over
the diameter of the strained core region (half width 100 <),
Angew. Chem. Int. Ed. 2003, 42, 2850 – 2853
which gives an average lattice strain of approximately 3 %
with respect to the unstrained Ru. This value is a lower limit,
because the NO molecules may in fact have dissociated closer
to the edge dislocation, where the strain is larger. Because of
the time delay between NO adsorption and the recording of
the STM images, the finite mobility of the N atoms may have
already changed the N distribution somewhat. Theoretical
models of edge dislocations in the bulk reveal core half widths
of a few lattice constants, which correspond to distortions of
10 %.[16, 17] The lattice distortions mentioned above, which
are caused by subsurface “bubbles” of Ar[9] did not exhibit an
enhanced activity for NO dissociation. The lattice stretch of
2–3 % associated with such distortions is thus below the limit
for an enhanced reactivity.
To explain the observed enhanced reactivity of NO in the
vicinity of edge dislocations, we performed self-consistent
periodic DFT calculations.[18] A three-layer Ru(0001) slab,
periodically repeated in a super cell geometry with 5.25
equivalent layers of vacuum between any two successive
metal slabs, was used. Adsorption of N, O, and NO, and
dissociation of NO were treated using a (4 : 2) unit cell, a
suitable choice to probe the low-coverage behavior of NO on
the Ru(0001) surface (computational details are given in the
Theoretical Section).
We examined the effect of changing the lattice constant
parallel to the surface of the Ru(0001) slab on the thermochemistry and reaction barrier for the dissociation of NO. The
calculated lattice constant (deq) for bulk Ru was found to be
2.74 <, in reasonable agreement with the experimental value
of 2.70 <. The equilibrium value for c/a of 1.582 was used. The
lattice constant (d) was varied between 2.60 and 2.88 <, which
corresponds to a maximum absolute value of relative strain
(Dd/deq) of approximately 5 %. All the metal atoms were kept
fixed in their bulk positions. Allowing for relaxation of the
topmost metal layer or including more than three metal layers
in the slab had a negligible effect on the results. As in previous
2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
studies on similar systems, we followed the reaction paths by
varying the NO bond length with simultaneous relaxation of
all the other molecular degrees of freedom.[1, 19]
Figure 3 shows the calculated effect of strain on the
adsorption strength of all the species involved in this study.
Data in the top and middle panels suggest that atomic oxygen,
atomic nitrogen, and molecular NO all bind more strongly to
Figure 4. Calculated energy along the reaction coordinate of the NO
bond length during NO dissociation. Continuous lines represent the
best fits through the calculated data points. The difference in energy
between the highest point and the initial point on each curve is taken
as the respective dissociation barrier, shown in the bottom panel of
Figure 3. The zero of the energy scale corresponds to the clean surface
plus a gas-phase NO molecule. N atoms: green circles, O atoms: red
circles, Ru atoms: purple circles.
Figure 3. Top: The effect of a relative change in the lattice constant
Dd/deq of an Ru(0001) surface on the binding energy of atomic nitroads
gen (Eads
N ) and oxygen (EO ; referenced to the clean surface plus the
appropriate gas-phase atomic species), middle: the binding energy of
molecular NO (Eads
NO ; referenced to the clean surface plus a gas-phase
NO molecule), and bottom: the NO dissociation barrier (Ediss
NO ; referenced to the initial state of the dissociation).
the stretched regions of the surface. The bottom panel of
Figure 3 shows that stretched surfaces substantially decrease
the barrier for NO dissociation, thereby strongly favoring the
dissociation of NO molecules at the stretched regions of the
Figure 4 provides further details for the NO dissociation
path on Ru(0001) for three different lattice constants. Strain
did not change the preferred adsorption site for any of the
adsorbates: atomic nitrogen and oxygen, as well as NO
molecules, all prefer hcp sites. The transition-state geometry
was found to be of the bridge–bridge type, regardless of the
degree of strain, and in good agreement with previous studies
for the dissociation of small molecules on metal surfaces.[1, 20]
Because of the varying degree of electronic coupling between
the adsorbate and the surface, along the reaction path, the
energy differences arising from lattice strain increase on going
from the initial to the transition state to the final state. In all
cases though, strain stabilizes the corresponding states.
Therefore, on stretched regions of the surface both the
adsorption and the dissociation of NO are favored. Finally,
2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
because of the enhanced adsorption energy of atomic nitrogen and oxygen, it might be expected that both atomic
products of the dissociation should be found on the stretched
regions of the defects. The experiment, however, only showed
N atoms. We suggest that because the O atoms are much more
mobile than the N atoms,[14, 15] they move away from the
reactive site faster. It is therefore expected that the O atoms
have been spread evenly over the entire surface by the time
the STM images were recorded.
The increased reactivity of stretched surfaces has been
attributed to the shift of the d-band center to higher energy
that results from the stretching of the surface.[1] The work
presented here offers the first microscopic experimental
evidence for the effect of strain on reaction kinetics. The
implications for catalysis are clear. Almost all catalytic
reactions are preceded by bond-activation dissociation steps,
which are often rate limiting, and lattice strain is probably
present in most applications involving supported catalysts.
Theoretical Section
Adsorption was allowed on only one of the two surfaces exposed and
the electrostatic potential was adjusted accordingly.[21] Ionic cores
were described by ultrasoft pseudopotentials,[22] and the Kohn–Sham
one-electron valence states were expanded in a basis of plane waves
with kinetic energies below 25 Ry. The surface Brillouin zone was
sampled at eight special k points. The exchange-correlation energy
and potential were described by the generalized gradient approximation (PW91).[23, 24]
Received: December 27, 2002 [Z50845]
Angew. Chem. Int. Ed. 2003, 42, 2850 – 2853
Keywords: chemisorption · density functional calculations ·
heterogeneous catalysis · scanning probe microscopy · surface
[1] M. Mavrikakis, B. Hammer, J. K. Nørskov, Phys. Rev. Lett. 1998,
81, 2819.
[2] J. Greeley, J. K. Nørskov, M. Mavrikakis, Annu. Rev. Phys.
Chem. 2002, 53, 319.
[3] J. Eckell, Z. Elektrochem. 1933, 39, 433.
[4] I. Uhara, S. Kishimoto, Y. Yoshida, T. Hikino, J. Phys. Chem.
1965, 69, 880.
[5] M. M. GKnter, T. Ressler, B. Bems, C. BKscher, T. Genger, O.
Hinrichsen, M. Muhler, R. SchlLgl, Catal. Lett. 2001, 71, 37.
[6] M. M. GKnter, T. Ressler, R. E. Jentoft, B. Bems, J. Catal. 2001,
203, 133.
[7] S.-K. Shi, H.-I. Lee, J. M. White, Surf. Sci. 1981, 102, 56.
[8] Y. Uesugi-Saitow, M. Yata, Phys. Rev. Lett. 2002, 88, 256 104-1.
[9] M. Gsell, P. Jakob, D. Menzel, Science 1998, 280, 717.
[10] J. Trost, T. Zambelli, J. Wintterlin, G. Ertl, Phys. Rev. B 1996, 54,
17 850.
[11] The length of the Burgers vector is one half of a (2 : 2)O lattice
constant, that is, one lattice constant of the Ru substrate, and it
points into the close-packed direction of the Ru basal plane (b =
=3 h1210i). Such a Burgers vector corresponds to an added half
plane intersecting the surface as indicated by the “T”-shaped
marker. If we assume that the dislocation is perpendicular to the
surface, the corresponding glide plane, spanned by the dislocation and the Burgers vector, would be a (10
10) plane, which is
precisely the known glide plane of Ru: D. W. Rhys, J. LessCommon Met. 1959, 1, 269.
[12] H. Conrad, R. Scala, W. Stenzel, R. Unwin, Surf. Sci. 1984, 145, 1.
[13] T. Zambelli, J. Wintterlin, J. Trost, G. Ertl, Science 1996, 273,
[14] T. Zambelli, J. Trost, J. Wintterlin, G. Ertl, Phys. Rev. Lett. 1996,
76, 795.
[15] J. Wintterlin, J. Trost, S. Renisch, R. Schuster, T. Zambelli, G.
Ertl, Surf. Sci. 1997, 394, 159.
[16] Y. Minonishi, S. Ishioka, M. Koiwa, S. Morozumi, M. Yamaguchi,
Philos. Mag. A 1981, 43, 1017.
[17] Q. F. Fang, R. Wang, Phys. Rev. B 2000, 62, 9317.
[18] B. Hammer, L. B. Hansen, J. K. Nørskov, Phys. Rev. B 1999, 59,
[19] S. Dahl, A. Logadottir, R. C. Egeberg, J. H. Larsen, I. Chorkendorff, E. TLrnqvist, J. K. Nørskov, Phys. Rev. Lett. 1999, 83, 1814.
[20] Y. Xu, M. Mavrikakis, Surf. Sci. 2001, 494, 131.
[21] J. Neugebauer, M. Scheffler, Phys. Rev. B 1992, 46, 16 067.
[22] D. Vanderbilt, Phys. Rev. B 1990, 41, 7892.
[23] J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R.
Pederson, D. J. Singh, C. Fiolhais, Phys. Rev. B 1992, 46, 6671.
[24] J. A. White, D. M. Bird, Phys. Rev. B 1994, 50, 4954.
Angew. Chem. Int. Ed. 2003, 42, 2850 – 2853
2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Без категории
Размер файла
244 Кб
scala, evidence, catalytic, surface, enhance, atomic, reactivity, stretches
Пожаловаться на содержимое документа