Potassium-promoted iron oxide model catalyst films for the dehydrogenation of ethylbenzene: An example for complex model systems

Article abstract of DOI:10.1006/jcat.2002.3785

The stability ranges of potassium-promoted iron oxide model catalyst films were studied in dependence on the annealing temperature and oxygen partial pressure using scanning tunneling microscopy (STM) and AES. In agreement with the stability ranges of potassium-promoted iron oxide model catalyst films prepared on a Pt(111) substrate, the same KFeO₂ phase formed at 700 K, while at higher temperatures single-crystalline K_xFe₂₂O₃₄(0001) became more stable. The Ru(0001) substrate enabled the growth of thinner closed films than on Pt(1 H), leading to sufficient conductivity of the insulating KFeO₂ film image this surface with STM. The catalytic dehydrogenation of ethylbenzene was carried out at 870 K, and a depletion of potassium at this temperature was observed, leading to the single crystalline K_xFe₂₂O₃₄(0001) phase.

Full text of DOI:10.1006/jcat.2002.3785

Journal of Catalysis 212, 104–111 (2002)
doi:10.1006/jcat.2002.3785
Potassium-Promoted Iron Oxide Model Catalyst Films
for the Dehydrogenation of Ethylbenzene:
An Example for Complex Model Systems
∗,
∗
Guido Ketteler, † Wolfgang Ranke,∗,1 and Robert Schlo¨gl
∗
Department Inorganic Chemistry, Fritz-Haber-Institute of the MPG, Faradayweg 4-6, 14195 Berlin, Germany;
and †FU Berlin, Fachbereich Biologie, Chemie, Pharmazie, Takustr. 3, 14195 Berlin, Germany
Received June 10, 2002; revised August 12, 2002; accepted August 12, 2002
interaction of potassium with well-defined oxide surfaces
has been subject to several investigations (10–18). In com-
bination with catalytic studies, e.g., with a recently devel-
oped single-crystal flow reactor (19), one should be able
to determine the atomic details and mechnism of cata-
lytic reactions. Potassium-promoted iron oxides are used in
the catalytic dehydrogenation of ethylbenzene to styrene
in the presence of steam at temperatures of about 870 K
(20–22). Recently, we were able to prepare epitaxial
potassium-promoted iron oxide model catalyst films by an-
nealing multilayers of potassium deposited on Fe3O4(111)
at 970 K (10–12). As evidenced by STM and low-energy
electron diffraction (LEED) (11), these films have ex-
cellent single-crystalline quality, and in combination with
X-ray photoelectron spectroscopy (XPS) and thermody-
namic considerations (12), we concluded that they con-
sist of single-phase Kx Fe22O34(0001). These films can be
prepared with varying potassium content giving rise to a
(4 × 4), (2 × 2), and (1 × 1) LEED pattern with increasing
potassium content (11). XPS measurements showed that
around 700 K another ternary potassium iron oxide phase
can be prepared which corresponds to KFeO2 (12). Anneal-
ing this phase beyond 870 K leads to a depletion of potas-
sium, establishingaKFeO2 layerontopofKx Fe22O34(0001)
which agrees excellently with investigations of technical
catalyst samples (23). Nevertheless, the detailed atomic ar-
rangement of this phase is unknown. All UHV measure-
ments showed that it is difficult to control the potassium
content reproducibly, as the residual water gas in an ultra-
high vacuum (UHV) environment was found to be highly
reactive because it converts the surface of the film to KOH,
which desorbs at elevated temperatures. The aim of this
study is to get more insight in the preparation, stability,
and surface structure of the ternary potassium iron oxide
phases forming at 700 K and to reveal details of the surface
structure at the technical dehydrogenation temperature of
870 K. The Kx Fe22O34(0001) phase, which forms at 970 K,
has been previously characterized by STM, LEED, and
Details of the surface structure and composition of potassium-
promoted iron oxide model catalyst films prepared on Ru(0001)
are investigated by scanning tunneling microscopy (STM), low-
energy electron diffraction (LEED), and Auger electron spec-
troscopy (AES). At 700 K, polycrystalline KFeO2 films are formed.
On annealing at the temperature of the technical dehydrogena-
tion reaction (870 K), the potassium content in KFeO2 films de-
creases and a mixed phase which consists of polycrystalline KFeO2
on top of partially uncovered KxFe22O34(0001) forms. Both phases
are exposed at the surface. At 970 K, the whole film transforms to
KxFe22O34(0001). Similarities with technical catalyst samples sug-
gest that the pressure gap is bridged, and for the first time, atomic
details of the potentially catalytically active surface phase are pre-
c
sented. ꢀ 2002 Elsevier Science (USA)
Key Words: potassium; iron oxide; scanning tunneling micro-
scopy; ethylbenzene; dehydrogenation; surface structure; surface
phases.
1. INTRODUCTION
Thin oxide films grown on metal substrates have gained
much interest as well-defined model catalysts, since they
can be prepared with defined stoichiometry and since the
conducting substrate enables a detailed characterization by
common surface science techniques such as electron spec-
troscopy and scanning tunnelling microscopy (STM) (1–3).
Such films have been prepared for many binary oxides as,
for example, Al, Cr, Co, Fe, Ni (1–4). For iron oxides, it
has been shown that preparation conditions and electronic
structure are comparable to the thermodynamically formed
bulk counterparts (5, 6). Recently, even more complex cata-
lyst model systems have been prepared. These include
metal particles deposited on oxide supports (7–9), as well
as alkali metal-promoted model catalyst films (10–12). The
1 To whom correspondence should be addressed. Fax: +49 30 8413 4401.
E-mail: ranke@fhi-berlin.mpg.de.
104
0021-9517/02 $35.00
c
ꢀ 2002 Elsevier Science (USA)
All rights reserved.

POTASSIUM IRON OXIDE MODEL CATALYST FILMS
105
Auger spectroscopy (11) and thus is not addressed in this
paper.
3. RESULTS
3.1. Characterization of the Fe3O4(111) Precursor Films
Fe3O4 crystallizes in the inverse spinel structure. Along
the (111) axis, oxygen layers alternate with two kinds of iron
layers, the so-called kagome´ layer and three mix-trigonal
layers (Figs. 1a and 1b). Figures 1c and 1d show a LEED
2. EXPERIMENTAL
The experiments were performed in an ultrahigh vacuum
chamber (UHV) with a base pressure of 1 × 10−10 mbar
which is equipped with a scanning tunnelling micro-
scope (STM, Burleigh), a combination of low-energy elec-
tron diffraction (LEED) and Auger electron spectroscopy
(AES) optics, and oxygen and argon gas inlet valves and
a sputter gun. Details are described elsewhere (24). The
Ru(0001) substrate single crystal was cleaned by repeated
cycles of Ar+ sputtering (1 keV, ∼3–5 µA) and annealing at
1450 K. Iron was evaporated by resistively heating an iron
wire wrapped around a tungsten filament. On Ru(0001),
closed FeO(111) films with thicknesses of up to 4 ML can
be obtained by deposition of the corresponding amount
of Fe in one run and subsequent oxidation in 10−6 mbar
oxygen at 870 K with a final anneal to 970 K (25). On pro-
longed oxidation, films ≥4 ML transform completely into a
closed Fe3O4(111) film. This offers the opportunity to pre-
pare thin potassium-promoted iron oxide films which allow
STM and LEED measurements. Thick films of the K–Fe–O
phase forming around 700 K display poor conductivity and
the tip approach does not abort, as no tunneling currents
can be measured.
˚
pattern and an atomically resolved STM image of a ∼20-A
˚
thick Fe3O4(111) film grown on Ru(0001) with the 6 × 6 A
periodicity of the iron layers in this oxide. Additionally,
bright features can be seen. Since AES shows only sig-
nals from Fe, O, and Ru, we believe that they represent
Fe clusters which form because of mass transport during
the stoichiometric conversion of closed FeO(111) films to
Fe3O4(111). LEED and STM images of the films grown on
Ru(0001) and Pt(111) look very similar, suggesting that the
same surface structure forms on both substrates, which was
found by a dynamical LEED analysis to be strongly relaxed
and terminated by 1/4 ML of iron cations in the outermost
layer (28).
˚
Figure 2a shows the surface morphology of 20- to 40-A
thick, closed Fe3O4(111) films. The mesoscopic surface
˚
roughness is 30 A on a length scale of 1 µm, which is in
the same range as the roughness of the clean Ru(0001) sub-
strate. Occasionally, hexagonal crystallites with diameters
˚
of ∼1000 A are formed (marked by the arrows). The step
˚
density is low, with terraces several hundred A wide. Most
˚
steps and crystallites are multiples of 4.8 A high, which cor-
Alternatively, hexagonal Fe3O4(111) islands can be ob-
tained by a sequence of several iron deposition (submono-
layers) and oxidation cycles (26). In this case, the growth is
closer to equilibrium conditions, leading to the thermody-
namically favored (5) Volmer–Weber growth of Fe3O4(111)
islands. On prolonged annealing, the Fe3O4(111) islands
grow at the expense of the surrounding FeO(111) film,
which is removed in this way. This is different from Pt(111),
where 1- to 2-ML FeO(111) persists, leading to an overall
Stranski–Krastanov-like growth mode of iron oxides (27).
The Fe3O4(111) islands on Ru(0001) have a characteristic
hexagonal or triangular shape, and morphology changes of
this shape (e.g., facet angles), which might occur on pro-
motion with potassium, may reveal structural information
even if no atomic resolution can be achieved.
responds to the repeat unit along (111) and indicates the
dominance of one surface termination. Occasionally, lower
Potassium was evaporated at room temperature onto
closed Fe3O4(111) substrate films or on Fe3O4(111) islands
using a commercial getter source (SAES). The sample was
placed about 3 cm from the K source and the pressure
did not exceed 5 × 10−9 mbar during deposition. Subse-
quently, the K/Fe3O4(111) films were annealed at 700, 870,
and 970 K in vacuum or in 10−6 mbar O2. Auger measure-
ments were carried out at a 3-kV primary beam energy
with a modulation voltage of Vpp = 7 V. STM measurements
were performed in the constant current mode using electro-
chemically etched tungsten tips cleaned in situ by electron
bombardment.
FIG. 1. (a) Slightly tilted side view of Fe3O4(111). (b) Top view of the
two different iron sublayers of Fe3O4(111). (c) LEED pattern (60 eV).
2
˚
(d) 150 × 150 A (It = 0.9 nA, UB = +0.7 V) atomic resolution STM image
˚
of a ∼20-A thin Fe3O4(111) film grown epitaxially on Ru(0001). The unit
cell is indicated in the LEED pattern and the STM image.

¨
KETTELER, RANKE, AND SCHLOGL
106
stable ternary compounds in the system K2O–Fe2O3, and
therefore it is very likely that potassium iron oxides adapt
one of these stoichiometries, assuming that the film thick-
ness is sufficient to form the bulk composition. Potassium-
poor Kx Fe22O34 phases (also called K–β-Fe2O3 (x = 2) and
K–βꢁꢁ-Fe2O3 (x = 4)) exist with varying potassium content
(even x < 2hasbeenproposedforisostructuralcompounds;
see references in (12)) and are layered structures consist-
ing of spinel blocks which are very similar to those in
Fe3O4(111). The spinel blocks are separated by KO or K2O
layers. Differences in the stacking of the spinel blocks leads
˚
to different lattice constants c (23.8 vs 35.9 A), while all
˚
structures are hexagonal with a = b = 5.9 A. The potassium-
˚
richKFeO2 isorthorhombicwithlatticeconstantsa = 5.6A,
˚
˚
b = 11.25 A, and c = 15.9 A. The potassium distribution in
this phase is much more homogeneous.
FIG. 2. (a) 1 × 1 µm2 (It = 0.5 nA, UB = +1.2 V) large-scale STM im-
Figure 4 shows the evolution of the Auger peak inten-
sity ratios of K/O and Fe/O with time when annealing thin,
˚
age of a ∼40-A thin Fe3O4(111) film grown on Ru(0001). The arrows
indicate hexagonal crystallites. (b) 1 × 1 µm2 (It = 1.0 nA, UB = +0.1 V)
large-scale STM image of Fe3O4(111) islands grown on Ru(0001).
step heights are observed. These are due to the formation
of “biphase-ordered” FeO(111)/Fe3O4(111) domains (29,
30), as also indicated by their characteristic contribution
to the LEED pattern (not shown here). As described in
Section 2, alsoFe3O4(111)islandscanbeprepared(Fig. 2b).
These crystallites have hexagonal or triangular shape. They
˚
are less than 50 A high, with diameters between 100 and
˚
3000 A.
3.2. Identification of Potassium Iron Oxide Phases
Figure 3 shows the crystal structures of the ternary potas-
sium iron oxides believed to be involved in the dehydro-
genation, as deduced from investigations with technical
catalyst samples (23). Formation and stability of ternary
potassium iron oxides are reviewed in Ref. (12). KFeO2
and Kx Fe22O34 (x = 2–4) are the only thermodynamically
FIG. 4. (a) Evolution of the K/O (squares) and Fe/O (circles) Auger
peak intensity ratios with annealing time when annealing K/Fe3O4(111) to
700, 870, and 970 K in a vacuum. Annealing in 10−6 mbar O2 is necessary to
establish a distinct stoichiometry at 700 K (shaded area). (Diamonds) K/O
Auger intensity ratios of a K/Fe3O4(111) film which was annealed to 700,
870, and 970 K in a 10−6-mbar oxygen atmosphere. The arrow indicates
the intensity ratio of the model catalyst film shown in the STM images in
Fig. 8. The different annealing procedures are indicated by the symbols I,
IIa–c, IIIa,b, and IVa,b and are explained in the text. (b) Magnification of
the Fe/O intensity ratios. Finally, constant Fe/O ratios are attained for the
different annealing procedures (circled data points). (Dashed lines) Antic-
ipated Fe/O AES ratios for stoichiometric KFeO2 (0.213) and Kx Fe22O34
(0.276).
FIG. 3. Structural models for (a) K2Fe22O34(0001) and (b) KFeO2.

POTASSIUM IRON OXIDE MODEL CATALYST FILMS
107
TABLE 1
closed potassium iron oxide films at different temperatures.
K, O, and Fe peaks at energies of 252, 503, and 651 eV, re-
spectively, were chosen to calculate the intensity ratios. For
all films, the Ru Auger peaks dominate the spectra, thus
indicating very thin films. Annealing at 700 K in vacuum
leads to a decrease of the potassium signal until a plateau
at K/O = 3.3 is reached (Fig. 4, regime I). This value dif-
fers depending on the thickness of the Fe3O4(111) film and
when different amounts of K are deposited. Nevertheless,
annealing at 700 K in 10−6 mbar O2 always leads to con-
stant K/O and Fe/O intensity ratios of 2.8 and 0.20, respec-
tively, independent of the Fe3O4(111) film thickness and in-
dependent of the potassium amount deposited, given that
it is above the stoichiometric amount needed to form this
phase (Fig. 4, regime IIa). This indicates that the oxygen
reservoir of the thin Fe3O4(111) films is not sufficient and
oxidative conditions are required to form this potassium
iron oxide phase. Once this value is attained, annealing in
vacuum only slightly increases the Fe/O intensity ratio, to
0.21 (Fig. 4, regime IIb).
Auger Peak Intensity Ratios for Fe3O4(111) and the Different
Potassium Iron Oxide Phases
K/Fe3O4(111)
Fe3O4(111) 700 K
870 K
970 K
IK : IO
IFe : IO
—
0.32
2.8
0.21
1.0
0.24
1.0
0.26
Stoichiometry
From XPS (12)
Fe3O4
Fe3O4
“FeO2”
“Fe19O34
”
“Fe21O34”
KFeO2 KFeO2/Kx Fe22O34 Kx Fe22O34
Note. The Fe/O stoichiometries are estimated assuming that IFe/IO
0.32 corresponds to Fe3O4.
=
Kx Fe22O34(0001) (layered with possible K-rich surface ter-
mination) is very different. Fe/O ratios fluctuate strongly
before the equilibrium phases are formed but finally at-
tain constant values (encircled data points in Fig. 4b), in-
dicating stoichiometries of “FeO2” (at 700 K), “Fe19O34”
(at 870 K), and “Fe21O34” (at 970 K). These values are
close to the anticipated values for stoichiometric KFeO2
and Kx Fe22O34(dashed lines in Fig. 4b).
Further annealing at 870 K in 10−6 mbar O2 does not
change the intensity ratios (Fig. 4, regime IIc). Annealing
at the same temperature in vacuum leads to a diminution
of the potassium content, and after 50–60 min, constant
K/O and Fe/O Auger intensity ratios of 1.0 and 0.24, re-
spectively, are observed (Fig. 4, regime IIIa). These val-
ues are close to those observed after annealing very thick
films to 970 K (11). As with those films, we observe a
weak Fe3O4(111)-(2 × 2) pattern with LEED which was at-
tributed to a Kx Fe22O34(0001) phase. The low intensity of
the (2 × 2) LEED pattern may be due to the limited film
thickness, which is in the range of the three-dimensional
3.3. Surface Structure of the Low-Temperature
Phase (KFeO2)
KFeO2 filmspreparedbydepositionofpotassiumonthin,
closedFe3O4(111)filmsandannealingat700Kin10−6 mbar
O2 (regime IIa,b in Fig. 4) display no LEED pattern. STM
images are shown in Fig. 5. On a large scale, the meso-
scopic surface roughness of the precursor Fe3O4(111) film
˚
˚
has increased to about 50 A on a length scale of 6000 A
(Fig. 5a). Terraces several hundred angstroms wide with
˚
¯
unit cell of the Kx Fe22O34 (23.8 A). Further annealing at
hexagonally inclined step edges running along the [101]
970 K leads to a constant decrease of the potassium signal
(Fig. 4, regime IVa). In oxygen atmosphere, this decrease
is significantly slower and a plateau with a ratio of 1.0 is
reached (Fig. 4, regime IIIb) before all potassium desorbs
on prolonged annealing (Fig. 4, regime IVb). As potassium
depletion occurs preferentially via reaction with residual
gas water and desorption of KOH (12), the time axis in
Fig. 4 depends on the residual gas pressure.
In summary, the Auger measurements show that two dis-
tinct stoichiometries can be obtained. The first one is a
potassium-rich phase at 700 K which requires oxidation of
the K/Fe3O4(111) film. The second phase transition to a
compound with lower potassium content depends on the
oxygen partial pressure and occurs at 870 K in vacuum
or 970 K in 10−6 mbar O2. The stability ranges agree per-
fectly with the results from XPS annealing cycles in vac-
uum and water atmosphere (12), where these equilibrium
phases have been attributed to KFeO2 (at 700 K) and
Kx Fe22O34 (above 870 K). Table 1 summarizes the obtained
Auger ratios for the identified phases. Only the “Fex Oy”
stoichiometries are calculated from the Auger intensity ra-
2
˚
A
FIG. 5. (a) 5000 × 5000
(It = 1.0 nA, UB = +1.2 V) and
2
˚
(b) 1500 × 1500 A (It = 1.0 nA, UB = +1.3 V) STM images of the KFeO2
phase obtained after annealing potassium deposited on Fe3O4(111) to
700 K in a vacuum. Crystallographic directions of the Fe3O4(111) precur-
tios, as the K distribution in KFeO2 (homogenous) and sor film are indicated.

¨
KETTELER, RANKE, AND SCHLOGL
108
mains which exist beside the sintered K–Fe–O phase. An-
nealing at 870 K in 10−6 mbar O2 does not affect the island
morphology but the morphology of the regions in between
changes (Fig. 7d). Flat terraces are covered by a higher dis-
˚
persed polycrystalline film. This film is 6–10 A thick and has
similarities to the covering film shown in Fig. 8 (see next sec-
tion), which can be identified as KFeO2. Edges of this phase
are in most cases aligned at the Ru(0001) substrate, but
no crystalline order is visible. Also here, the K/Ru(0001)-
(3 × 3) LEED pattern is obtained. Experiments with al-
kali metals on 1-ML FeO(111)/Pt(111) showed that pro-
longed annealing leads to desorption of the alkali metal
and as a result the iron oxide spreads back over the
2
˚
FIG. 6. (a) 500 × 500 A (It = 0.7 nA, UB = +2.7 V) and (b) 600 ×
2
˚
600 A (It = 0.7 nA, UB = +2.7 V) STM images of the phase obtained
after annealing potassium deposited on Fe3O4(111) to 870 K in a Pt(111) (18).
10−6-mbar oxygen atmosphere. Crystallographic directions of the
Fe3O4(111) precursor film are indicated. These films have the same K/O
3.4. Surface Structure of Potassium-Promoted
and Fe/O Auger intensity ratios as the one shown in Fig. 5.
Iron Oxides at 870 K
As shown in Section 3.2, annealing a KFeO2 film at 870 K
in a vacuum leads to the formation of a phase with lower
potassium concentration. Based on XPS intensity evalua-
tions, this depletion was suggested as proceeding from the
bulk under formation of Kx Fe22O34 while a potassium-rich
¯
and [110] directions of the precursor Fe3O4(111) film are
observed. The surface morphology thus still is governed by
the precursor Fe3O4(111) film or the Ru(0001) substrate.
Closer inspection reveals that the terraces are very rough
(Fig. 5b). The film has cracked and is disrupted by several
trenches. The terraces are too rough to determine precise
step heights.
STM images of a film prepared by oxidation in 10−6 mbar
oxygen at 870 K (regime IIc in Fig. 4) are shown in
Figs. 6a and 6b. Higher bias voltages are now required
to resolve this surface. The morphology has changed and
apart from 120◦ or 60◦ angles, also rectangular step edges
are occasionally formed. Also this film displays no LEED
pattern.
Experiments with K-promoted Fe3O4(111) islands re-
confirm that the macroscopic morphology of potassium-
promoted iron oxides is governed by the precursor
Fe3O4(111) morphology. After promotion with potassium
and prolonged annealing at 700 K, triangular or hexago-
nal islands identical to unpromoted islands are observed
(compare Fig. 7a with Fig. 2c). Between the islands, where
originally a flat, well-ordered FeO(111) film existed, new
features have formed (Fig. 7b). Flat terraces are now par-
tially covered by a film with a dendritic appearence. Step
edges are decorated by this phase as well. The dendritic
shape shows that the high dispersion of the FeO(111) film
has vanished, and a sintered phase with an unfavorable epi-
taxial relationship with respect to the substrate has formed.
This agrees well with the alkali-metal-induced sintering ob-
served previously for 1-ML FeO(111) on Pt(111) (18). The
FIG. 7. (a) 1 × 1 µm2 (It = 0.7 nA, UB = +1.5 V) large-scale STM im-
age of K deposited on Fe3O4(111) islands after annealing in a 10−6-mbar
2
˚
film produces the Fe3O4(111) LEED pattern superimposed oxygen at 700 K. (b) 3000 × 3000 A (It = 0.7 nA, UB = +1.5 V) large-scale
STM image of the flat regions inbetween the Fe3O4(111) islands after de-
with a Ru(0001)-(3×3) pattern (Fig. 7c). An intense (3 × 3)
position of K and annealing in 10−6 mbar oxygen at 700 K. (c) LEED image
LEED pattern is also obtained when potassium is deposited
2
˚
(60 eV) of the surface shown in (a) and (b). (d) 2000 × 2000 A (It = 0.8 nA,
on a clean Ru(0001) substrate without iron oxide present;
UB = +0.8 V) large-scale STM image after oxidation in 10−6 mbar oxygen
therefore it does not arise from a new ternary potassium
iron oxide phase but results from K/Ru(0001)-(3 × 3) do- cated.
at 870 K. Crystallographic directions of the Ru(0001) substrate are indi-

POTASSIUM IRON OXIDE MODEL CATALYST FILMS
109
odicity is also displayed in the LEED pattern (Fig. 8d, inset)
and is well known from Kx Fe22O34(0001) films on Pt(111)
(10, 11). The upper terraces cover about 85% of the sur-
face and have a morphology similar to the KFeO2 phase
(compare Figs. 5 and 6). This phase is in most cases aligned
to the crystallographic directions of the Kx Fe22O34(0001)
phase but no crystalline order is visible. Further anneal-
ing causes this KFeO2 phase to disappear. The remain-
ing well-ordered Kx Fe22O34(0001) surface was investigated
in (11).
4. DISCUSSION
The Auger measurements indicate the formation of two
different distinct stoichiometric compounds when potas-
sium is deposited on Fe3O4(111) and annealed to different
temperatures. The stoichiometries deduced from AES for
both phases (Table 1) as well as the characteristic tempera-
tures of formation agree perfectly with previous XPS inves-
tigations and thermodynamic considerations (12). The first
phase corresponds to polycrystalline or amorphous KFeO2.
These KFeO2 films on Ru(0001) could not be prepared in
single-crystalline quality since annealing at temperatures
beyond 870 K leads to formation of other ternary phases
(Kx Fe22O34(0001)) or desorption of potassium. Even after
prolonged annealing for several hours at 700 or at 870 K in
oxygen atmosphere, we could not observe a defined LEED
pattern of this phase. STM images obtained under these
two conditions are quite similar. In general, films annealed
at 870 K in oxygen are less corrugated in large-scale images
and require higher bias voltages for tunneling. This indi-
cates that these films have fewer defects, which according
to the differences in the preparation are very likely oxy-
gen vacancies. This would explain the higher bias voltages
necessary to tunnel oxygen-annealed films. Nevertheless,
the K/O and Fe/O Auger intensity ratio is not significantly
different for both conditions; thus differences in the oxide
stoichiometry are quite small. On a small scale, both films
show rough terraces without long-range order. No atomic
2
˚
A
FIG. 8. (a) 5000 × 5000
(It = 0.8 nA, UB = +1.2 V) and
2
˚
(b) 1500 × 1500 A (It = 0.8 nA, UB = +1.5 V) large-scale STM images
of K deposited on Fe3O4(111) after annealing for ∼35min at 870 K.
The K/O and Fe/O intensity ratios are indicated by an arrow in Fig. 4.
2
2
˚
˚
(c) 500×450 A (It =0.2 nA, UB = +1.5 V) and (d) 300×265 A (It =
0.2 nA, UB = +1.5 V) higher resolution STM images with line scans. (Dia-
2
˚
monds) The two-dimensional 12 × 12 A unit cell of the Kx Fe22O34(0001)
phase. (Inset) The weak (2 × 2) LEED pattern (13 eV) of the
˚
Kx Fe22O34(0001) phase with a lattice constant of ∼12 A.
KFeO2-like surface region is maintained. In order to pre- resolution was obtained with STM, which may be due to
pare such a mixed phase, we annealed a KFeO2 film at the high bias voltages and to the roughness of the film.
870 K in a vacuum until the constant Auger ratios of the In some cases, hexagonal as well as right angles are ob-
Kx Fe22O34 phase were nearly attained (indicated by the ar- served at step edges. The latter may indicate that the mate-
row in Fig. 4). The characteristic morphology of this film is rial is of a space group with right angles (e.g., orthorhombic
shown in Figs. 8a and b. The mesoscopic surface roughness KFeO2).
˚
˚
is 50 A on a length scale of 5000 A. The step density is very
As we have shown above, it makes a difference whether
˚
low and terraces are 500–2000 A wide. Most step heights are KFeO2 films are annealed in a vacuum or in an oxygen at-
˚
multiples of ∼5 A. The terraces have hexagonally shaped mosphere: At 870 K, depletion of potassium and transfor-
˚
depressions with diameters of 100–400 A. The two height mation into Kx Fe22O34(0001) was only observed in vacuum
levels correspond to two different phases which are sepa- while in oxygen atmosphere KFeO2 remained stable, which
˚
rated by steps which are 2–5 A high. The lower regions are simply indicates that the stability ranges of the involved
very flat and cover about 15% of the surface. Higher reso- oxides depend on temperature and oxygen partial pres-
lution STM images reveal a single-crystalline phase with a sure. The catalytic dehydrogenation of ethylbenzene is per-
˚
˚
12 × 12 A periodicity (Figs. 8c and 8d). The 12 × 12 A peri- formed at 870 K, and we observe a depletion of potassium at

¨
KETTELER, RANKE, AND SCHLOGL
110
this temperature which finally leads to the single-crystalline
Kx Fe22O34(0001) phase. Slightly before the constant Auger
ratio of this phase is reached, most of the film is still cov-
ered with polycrystalline KFeO2. Only small sections of
Kx Fe22O34(0001) can be seen in large-scale STM images
(Fig. 8). Nevertheless, as the Auger ratio is already very
close to the value of the pure Kx Fe22O34(0001) phase, we
can conclude that a thin KFeO2 film has formed on top
of Kx Fe22O34(0001) instead of thick KFeO2 domains. This
confirms the proposition deduced from XPS, that the deple-
tion of the film goes along with a decrease in the potassium
content in the bulk while the surface remains potassium
rich (12). Both phases are exposed side by side at the sur-
face. These films therefore appear to model the catalytically
active phase quite well, which according to investigations
of the technical catalyst consists of a KFeO2 shell forming
around a core of K2Fe22O34.
The technical dehydrogenation of ethylbenzene is per-
formed in an excess of steam. Water may have two different
functions. It may have an oxidizing impact due to the lib-
eration of oxygen by thermal decomposition into hydrogen
and oxygen at high temperatures. According to the disso-
ciation equilibrium, the partial pressures of O2 and H2 for
1 bar H2O at 870 K are 6.4 × 10−6 and 1.3 × 10−5 mbar,
respectively (5). The oxidizing or reducing impact of water
would thus be close to UHV conditions. On the other hand,
water may react with the potassium iron oxides under for-
mation of KOH, which at this temperature has a consid-
erable vapor pressure and may desorb. Previous XPS ex-
periments showed that the phase transition to KFeO2 and
KFe22O34(0001) occurred at the same temperatures in a
vacuum and in a 10−8-mbar H2O water atmosphere (12);
therefore the mixed KFeO2/KFe22O34(0001) surface struc-
ture shown here may also form at 870 K in water atmo-
sphere. However, formation of the K-rich KFeO2 surface
phase then competes with K removal by KOH formation
and desorption, and its existence therefore depends on its
rate of formation. As the technical catalyst was identified
to consist of KFeO2 and KFe22O34 (21, 23), the removal of
potassium by KOH desorption has to be compensated for
in a dynamic process in which potassium segregates from
the underlying KFe22O34 phase to the surface. In this way,
the KFe22O34 phase serves as a potassium reservoir for the
recovery of the KFeO2 surface phase. The required potas-
5. CONCLUSIONS
Using STM and AES, we investigated the stability ranges
of potassium-promoted iron oxide model catalyst films in
dependence on the annealing temperature and oxygen par-
tial pressure. In agreement with the stability ranges of
potassium-promoted iron oxide model catalyst films pre-
pared on a Pt(111) substrate, we found that the same KFeO2
phase forms at 700 K, while at higher temperatures single-
crystalline Kx Fe22O34(0001) becomes more stable. We see
no effect of the substrate on the stoichiometry or crys-
talline quality of the model catalyst film. The Ru(0001)
substrate enables the growth of thinner closed films than
on Pt(111), leading to sufficient conductivity of the insulat-
ing KFeO2 film to image this surface with STM. At condi-
tions which simulate the technical dehydrogenation reac-
tion, Kx Fe22O34(0001) is partially covered by KFeO2. Both
phases are exposed at the surface. As the same KFeO2
shell around a K2Fe22O34 core was identified for the ac-
tive state in technical catalyst samples (23), our results hint
that the pressure and material gap might be bridged for the
experiments shown here, and for the first time, details of
the surface of the potentially catalytically active phase are
presented.
ACKNOWLEDGMENT
The authors thank Thomas Bunke for experimental assistance.
REFERENCES
1. Street, S. C., Xu, C., and Goodman, D. W., Annu. Rev. Phys. Chem. 48,
43 (1997).
2. Freund, H.-J., Angew. Chem. 109(5), 444 (1997).
3. Noguera, C., J. Phys. Condens. Matter 12, R367 (2000), and references
therein.
4. Weiss, W., and Ranke, W., Prog. Surf. Sci. 70, 1 (2002).
5. Ketteler, G., Weiss, W., Ranke, W., and Schlo¨gl, R., Phys. Chem. Chem.
Phys. 3, 1114 (2001).
6. Schedel-Niedrig, Th., Weiss, W., and Schlo¨gl, R., Phys. Rev. B 52, 17449
(1995).
7. Ba¨umer, M., and Freund, H.-J., Prog. Surf. Sci. 61, 127 (1999).
8. Cosandey, F., and Madey, T. E., Surf. Rev. Lett. 8, 73 (2001).
9. Chusuei, C. C., Lai, X., Luo, K., and Goodman, D. W., Top. Catal. 14,
71 (2001).
sium diffusion is a very fast process. This is evident from 10. Shaikhutdinov, Sh. K., Joseph, Y., Kuhrs, C., Ranke, W., and Weiss, W.,
Faraday Discuss. 114, 363 (1999).
11. Shaikhutdinov, Sh. K., Weiss, W., and Schlo¨gl, R., Appl. Surf. Sci. 161,
XPS measurements which showed that at 300 K a solid-
state reaction occurs and KO2 is formed by reaction with
oxygen from the Fe3O4(111) substrate (12). Presently, it is
not clear how the complete removal of potassium from the
497 (2000).
12. Joseph, Y., Ketteler, G., Kuhrs, C., Ranke, W., Weiss, W., and Schlo¨gl,
R., Phys. Chem. Chem. Phys. 3, 4141 (2001).
catalyst in its active state by formation and desorption of 13. Campbell, C. T., Surf. Sci. Rep. 27, 1 (1997).
14. Murray, P. W., Condon, N. G., and Thornton, G., Surf. Sci. Lett. 323,
KOH is avoided under real catalysis conditions. The deac-
tivation of the catalyst by spatial disintegration into Fe3O4
and bulk KOH is a slow process (21, 23), and it is likely that
carbonaceaous deposits serve as a protective layer against
K depletion (22).
L281 (1995).
15. Huang, H. H., Jiang, X., Zou, Z., Chin, W. S., Xu, G. Q., Dai, W. L.,
Fan, K. N., and Deng, J. F., Surf. Sci. 412/413, 555 (1998).
16. Carley, A. F., Jackson, S. D., Roberts, M. W., and O’Shea, J., Surf. Sci.
454, 141 (2000).

POTASSIUM IRON OXIDE MODEL CATALYST FILMS
111
17. Nerlov, J., Hoffmann, S. V., Shimomura, M., and Møller, P., Surf. Sci. 24. Weiss, W., Ritter, M., Zscherpel, D., Swoboda, M., and Schlo¨gl, R.,
401, 56 (1998).
J. Vac. Sci. Technol. A 16, 21 (1998).
18. Vurens, G. H., Strongin, D. R., Salmeron, M., and Somorjai, G. A.,
Surf. Sci. Lett. 199, L387 (1988).
19. Kuhrs, C., Swoboda, M., and Weiss, W., Top. Catal. 15, 13 (2001).
20. Geus, J. W., Appl. Catal. 25, 313 (1986).
21. Muhler, M., Schlo¨gl, R., and Ertl, G., J. Catal. 138, 413 (1992).
22. Kuhrs, C., Arita, Y., Weiss, W., Ranke, W., and Schlo¨gl, R., Top. Catal.
14, 111 (2001).
25. Ketteler, G., and Ranke, W., Phys. Rev. B 66, 033405 (2002).
26. Ketteler, G., and Ranke, W., manuscript in preparation.
27. Weiss, W., and Ritter, M., Phys. Rev. B 59, 5201 (1999).
28. Ritter, M., and Weiss, W., Surf. Sci. 432, 81 (1999).
29. Condon, N. G., Leibsle, F. M., Parker, T., Lennie, A. R.,
Vaughan, D. J., and Thornton, G., Phys. Rev.
B 55, 15885
(1997).
23. Muhler, M., Schu¨tze, J., Wesemann, M., Rayment, T., Dent, A., Schlo¨gl, 30. Shaikhutdinov, Sh. K., Ritter, M., Wang, X.-G., Over, H., and Weiss,
R., and Ertl, G., J. Catal. 126, 339 (1990). W., Phys. Rev. B 60, 11062 (1999).