Reduction of FeO/Pt(1 1 1) thin films by exposure to atomic hydrogen

Article abstract of DOI:10.1016/j.susc.2009.10.008

Using scanning tunneling microscopy (STM), X-ray photoemission spectroscopy (XPS) and density functional theory (DFT) calculations we have studied the reduction of ultra-thin films of FeO(1 1 1) grown on Pt(1 1 1) after exposure to atomic hydrogen at room temperature. A number of new ordered, partly-reduced FeO_x structures are identified and as a general trend we reveal that all the reduced FeO_x structures incorporate 2-fold coordinated Fe atoms as opposed to the original 3-fold coordinated Fe atoms in the FeO film. We find that when all the Fe atoms are 2-fold O-coordinated the FeO_x surface structure is resistant to further reduction at room temperature. We observe that water easily dissociates on the most heavily reduced FeO_x, structure in contrast to the initially inert FeO film, and reveal that it is possible to partially re-oxidize the FeO_x film by heating the surface slightly in the presence of water.

Full text of DOI:10.1016/j.susc.2009.10.008

Surface Science 604 (2010) 11–20
Contents lists available at ScienceDirect
Surface Science
journal homepage: www.elsevier.com/locate/susc
Reduction of FeO/Pt(1 1 1) thin films by exposure to atomic hydrogen
Jan Knudsen a, Lindsay R. Merte a, Lars C. Grabow b, Falk M. Eichhorn b, Soeren Porsgaard a, Helene Zeuthen a,
Ronnie T. Vang a, Erik Lægsgaard a, Manos Mavrikakis b, Flemming Besenbacher a,
*
a Interdisciplinary Nanoscience Center (iNANO), Department of Physics and Astronomy, Aarhus University, DK-8000 Aarhus C, Denmark
b Department of Chemical and Biological Engineering, University of Wisconsin-Madison, Madison, WI 53706, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Using scanning tunneling microscopy (STM), X-ray photoemission spectroscopy (XPS) and density func-
tional theory (DFT) calculations we have studied the reduction of ultra-thin films of FeO(1 1 1) grown on
Pt(1 1 1) after exposure to atomic hydrogen at room temperature. A number of new ordered, partly-
reduced FeOx structures are identified and as a general trend we reveal that all the reduced FeOx struc-
tures incorporate 2-fold coordinated Fe atoms as opposed to the original 3-fold coordinated Fe atoms in
the FeO film. We find that when all the Fe atoms are 2-fold O-coordinated the FeOx surface structure is
resistant to further reduction at room temperature. We observe that water easily dissociates on the most
heavily reduced FeOx, structure in contrast to the initially inert FeO film, and reveal that it is possible to
partially re-oxidize the FeOx film by heating the surface slightly in the presence of water.
Ó 2009 Elsevier B.V. All rights reserved.
Received 22 September 2009
Accepted for publication 8 October 2009
Available online 14 October 2009
Keywords:
Iron-oxide film
Reduction
Scanning tunneling microscopy
Water
Metal-oxide
X-ray photoelectron spectroscopy
1. Introduction
experiments appear to confirm this. For example, electrochemical
measurements by Wang et al. [11] have shown that a charge trans-
Numerous studies have demonstrated that ultra-thin oxides ex-
hibit novel physical and chemical properties compared to their
bulk counterparts. In particular, reducible metal oxides grown on
noble metal surfaces have been shown to form a variety of phases
having structures and stoichiometries with no known bulk ana-
logues, resulting from both two-dimensional confinement and
electronic interactions with the substrate [1,2]. Netzer has noted
the general tendency of transition metals to stabilize ultra-thin
oxides in lower oxidation states compared to bulk phases [3], an
effect which has significant implications for heterogeneous cataly-
sis and the reactivity of transition metal oxides.
Such effects may be important for FeOx/Pt catalysts, which have
been found to be active for selective CO oxidation in hydrogen
[4,5], the oxygen reduction reaction [6], and the water–gas shift
reaction [7,8]. Qin et al. [9] have shown that platinum nanoparti-
cles on Fe3O4 become encapsulated by a thin layer of FeO when ex-
posed to reducing conditions, a phenomenon that has been
implicated in the strong metal-support interaction. Huang and
Ranke [10] studied the reduction of a 1 ML FeO film on Pt(1 1 1)
by atomic hydrogen and identified a substoichiometric FeOx phase
that wets the surface and is resistant to further reduction. Whereas
metallic platinum is observed to stabilize iron oxides in low oxida-
tion states one would expect Pt to destabilize oxides in higher oxi-
dation states, effectively making them more reactive oxidants, and
fer from Fe3O4 to Pt enhances the reducibility of the oxide and al-
ters the chemical properties of the metal. Apparently for similar
reasons, an ultra-thin, oxygen-rich FeOx film on Pt(1 1 1) was found
to be a highly effective catalyst for CO oxidation [12]. Clearly, to
attain a complete understanding of the interactions between iron
oxides and platinum, atomic-scale characterization of the struc-
tures and chemical properties of FeOx/Pt interfaces is highly
desirable.
Here, we make use of high-resolution scanning tunneling
microscopy (STM) and X-ray photoelectron spectroscopy (XPS) in
combination with density functional theory (DFT) calculations to
study the reduction of the 1 ML FeO film on Pt(1 1 1) by atomic
hydrogen. The pristine film has been well-characterized in the past
using STM [13,14], low-energy electron diffraction (LEED) and
X-ray photoelectron diffraction [15], and its atomic structure is de-
picted in Fig. 1a. It consists of a hexagonal Fe-lattice sandwiched in
between the Pt(1 1 1) substrate and a hexagonal surface layer of O
atoms. The lattice mismatch between the hexagonal Fe/O-lattice
(3.1 Å) and the Pt(1 1 1) substrate (2.78 Å) as well as a small rota-
tion of the Fe/O-lattice relative to the substrate (ꢀ0.6°) leads to a
characteristic moiré structure with ꢀ25 Å periodicity, exhibiting
a floreted pattern in LEED (Fig. 1b), and clearly discernible super-
cell in STM measurements (Fig. 1c). When the FeO surface struc-
ture is exposed to atomic hydrogen, oxygen is removed,
desorbing as H2O [10]. The O:Fe ratio of the film decreases steadily
from 1 to ꢀ0.65, and correspondingly we observe formation of sev-
eral new FeOx structures, beginning with vacancy dislocation
* Corresponding author. Tel.: +45 8942 3604/3711; fax: +45 8612 0740.
E-mail address: fbe@inano.dk (F. Besenbacher).
0039-6028/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2009.10.008

12
J. Knudsen et al. / Surface Science 604 (2010) 11–20
deposition of submonolayer quantities at room temperature and
subsequent oxidation at 870 K in 1.3 ꢁ 10ꢂ6 mbar O2 for 2 min. It
is straightforward to distinguish the FeO film from the Pt(1 1 1)
substrate in the STM images, and because a layer-by-layer growth
mode is observed up to 2.1 monolayers [14], the thickness of the
FeO film can easily be controlled. One monolayer is here and in
the remainder of this article defined as the iron and oxygen cover-
age in a perfect bilayer FeO film grown on Pt(1 1 1) (i.e. 0.8 ML with
respect to the atom density of Pt(1 1 1)). The structure and clean-
liness of the film were always checked with STM before
experiments.
a
HCP
b
c
FCC
TOP
As molecular hydrogen does not react readily with the FeO film
under UHV conditions [10,20], we used atomic hydrogen to study
the reduction of the ultra-thin film. Atomic hydrogen was dosed
with a commercial gas cracker (Oxford Applied Research TC-50)
located approximately 10 cm from the sample. Before the cracker
was turned on the H2 pressure in the chamber was adjusted to
the desired value. Fixed power settings of the cracker were used
in all experiments (54 W) and the exposure time and pressure
were used to vary the exposure. The cracker was always preheated
for 30 s. before the sample was turned towards the H-beam to en-
sure a constant flux of atomic hydrogen during the entire exposure
time. Exposures are calculated from the H2 background pressure
measured before the cracker was turned on. The actual exposure
of Hat is unknown, but it is proportional to the dosing time with
fixed H2 pressure and cracker power settings. The flux of atomic
hydrogen at the sample surface, for a given background H2 pres-
sure, is determined by the effective pumping speed of the chamber.
The latter may vary for STM and XPS set-ups, but we estimate the
flux to be approximately equal in both chambers.
d
p(3x3)
p(2x2)
Oxygen vacancy
dislocation loops
To obtain more ordered structures and to desorb excess hydro-
gen we flashed the samples to ꢀ500 K directly after exposure.
Deionized, triply distilled water was filled in a small glass bulb
and degassed by several freeze–pump-thaw cycles prior to use,
and water exposures were performed by backfilling the UHV
chamber.
Fe-mode
Fig. 1. (a) Atomic structure of the FeO(1 1 1) film with the three high-symmetry
domains highlighted. Pt (white), Fe (black), and O (red). (b) 53.9 eV LEED image of
the 1 ML FeO film, showing the characteristic floret pattern due to the film’s moiré
structure. (c) STM image (60 ꢁ 60 Å2, 3 nA, 65 mV) of the 1 ML FeO film, with the
unit cell from (a) marked. (d) STM image (150 ꢁ 120 Å2, 1.76 nA, 55 mV) obtained
after exposure of 9 ꢁ 10ꢂ6 mbar s atomic hydrogen and subsequent flashing to
450 K. Three different reduced structures are observed to coexist on the surface.
Distinct protrusions correspond to 3-fold coordinated Fe atoms. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
The XPS measurements were performed at MAX-Lab, beam line
I311, which is described in detail elsewhere [21]. All spectra were
collected in normal emission with photon energies of 153 eV for Fe
3p, 380 eV for C 1s, 700 eV for O 1s, and 860 eV for Fe 2p. The spec-
tra are normalized to the background and the binding energies are
calibrated to the Fermi edge. The background due to the neighbor-
ing Pt 5p line was subtracted from all the Fe 3p spectra and a poly-
nomial background was used for the O 1s spectra. Coverages of
FeOx and coverages of OH groups are calculated by fitting the O
1s spectra with two peaks with shape similar to the shape of pris-
tine FeO. The coverage of the FeO film in the XPS experiments was
calibrated by saturating the surface with CO at room temperature.
Because CO only binds to the exposed Pt surface [22] the intensity
of the C 1s signal can be used directly to follow the growth of the
film. By cycles of Fe deposition (submonolayer amounts) at room
temperature and subsequent oxidation, we tuned the coverage of
FeO to the point where the C 1s signal disappeared. Furthermore
we checked the status of the FeO film using LEED, which showed
the expected primary 6-fold spots from the FeO film and a rosette
pattern arising from the moiré structure [15].
loops, which we have previously described briefly [16], followed by
an ordered p(3 ꢁ 3) structure of stoichiometry Fe4O3 and finally a
p(2 ꢁ 2) structure of stoichiometry Fe3O2, which is resistant to fur-
ther reduction. In Fig. 1d, an STM image is depicted showing the
surface in an intermediate state of reduction, where these three
structures are observed to coexist. Additionally, we have studied
the interaction of water with the dislocation loop structure and
with the Fe3O2 structure. The dislocation loops are found to be in-
ert toward water dissociation under the present experimental con-
ditions, similar to the case reported for the pristine FeO film
[17,18], whereas water is found to dissociate reversibly on the
Fe3O2 structure at room temperature and irreversibly at higher
temperatures, partially re-oxidizing the FeOx film.
Spin polarized density functional theory calculations using the
DFT + U approach by Dudarev [23] with U = 4 eV and J = 1 eV [24]
were performed using the VASP total energy code [25,26]. Similar
to a previous study on this system, we used the projector aug-
mented wave (PAW) method with a kinetic energy cutoff of
400 eV combined with the PW91 exchange–correlation functional
[27] and chose anti-ferromagnetic initial guesses for the magnetic
moment of Fe atoms [24]. The unit cells used in our calculations
2. Experimental and computational methods
STM measurements were carried out at room temperature with
a home-built Aarhus-STM [19] mounted in an ultra-high vacuum
(UHV) chamber with base pressure ꢀ1 ꢁ 10ꢂ10 mbar. The Pt(1 1 1)
crystal was cleaned by cycles of Ar+ sputtering and subsequent
annealing at 1000 K until it was clean as judged from atomically-
resolved STM images. The FeO films were grown by cycles of Fe
are consistent with the experimentally observed systems:
a
pffiffiffiffiffiffi pffiffiffiffiffiffi
91 ꢁ 91R5:2ꢃ unit cell with three layers of Pt atoms (Fe–Fe dis-

J. Knudsen et al. / Surface Science 604 (2010) 11–20
13
tance = 3.1 Å) was employed to estimate the energy gain for the
formation of the small triangular dislocation loops and p(2 ꢁ 2)
and p(3 ꢁ 3) unit cells with five layers of Pt atoms were used to
apparent height. Tip changes are most often spontaneous, but
can be induced by applying voltage pulses or by dipping the tip
gently into the surface. The presence of dislocation loops on the
surface allows these modes to be distinguished. STM images of
two similar dislocation loops, which correspond to the smaller tri-
angle depicted in Fig. 2b in different tip imaging modes are shown
as Fig. 2c and d. Comparison of these images with the model re-
veals that Fig. 2c shows 3-fold coordinated Fe atoms as bright pro-
trusions, while the 2-fold coordinated Fe atoms along the edge of
the dislocation loops are invisible. The observation of a registry
shift in Fig. 2d indicates that O atoms are imaged in this mode.
Inspection of Fig. 2b reveals that each of the small dislocation
study the p(2 ꢁ 2) and p(3 ꢁ 3) structures. For the
pffiffiffiffiffiffi pffiffiffiffiffiffi
91 ꢁ 91R5:2ꢃ unit cell, the FeO film and the topmost layer of
Pt were fully relaxed. For the p(2 ꢁ 2) and p(3 ꢁ 3) unit cells, the
FeO film and top two Pt layers were fully relaxed. The Brillouin
zone was sampled using the Gamma-point only for the
pffiffiffiffiffiffi pffiffiffiffiffiffi
91 ꢁ 91R5:2ꢃ cells, a 6 ꢁ 6 ꢁ 1 Monkhorst–Pack k-point grid
for the p(2 ꢁ 2) unit cells and a 4 ꢁ 4 ꢁ 1 Monkhorst–Pack k-point
grid for the p(3 ꢁ 3) unit cells. All simulated STM images were gen-
erated in the Tersoff–Hamann approach with a sample bias of
65 mV and an electron isodensity of 1 ꢁ 10ꢂ5 e/Å3.
*
*
loops converts one HCP domain to an FCC (where indicates the
opposite orientation with respect to the substrate, see Fig. 2a) do-
main by shifting the oxygen lattice, whereas each of the large dis-
3. Results and discussion
*
location loops converts three HCP domains to FCC domains and
*
one TOP domain to one TOP domain. As we have previously re-
3.1. Oxygen vacancy dislocation loops
ported, DFT calculations using pseudomorphic P(2 ꢁ 2) unit cells
indicate that FCC-type domains are more stable by 0.1–0.2 eV per
FeO unit compared to HCP-type domains, suggesting that the dis-
location loops form as a result of the lattice mismatch, analogous
to misfit dislocation loops observed in metal-on-metal growth
[31]. Thus, each small dislocation loop stabilizes a single HCP do-
main, while each large dislocation loop stabilizes the three HCP do-
In a previous publication, we have described the atomic struc-
ture of the triangularly shaped features [16], observed at the onset
of the reduction. In summary, they consist of oxygen vacancy dis-
location loops with an edge length of either approximately 5 or 12
Fe atoms (incorporating 7 or 14 O vacancies, respectively). The
smaller dislocation loops are always centered at HCP domains,
while the larger dislocation loops are always centered at TOP do-
mains. A schematic model of the atomic structures of both the
large and small dislocation loops is shown Fig. 2b.
XPS measurements are consistent with this model of the trian-
gular features, and almost idemtical O spectra obtained before and
after mild reduction are shown in Fig. 2e. By comparison with STM
experiments (to be discussed in more detail below, see Fig. 7b), we
estimate that on this surface dislocation loops are found at 30–40%
of the moiré unit cells, corresponding to a vacancy coverage of
3–4%. The O 1s spectrum shows no change in peak position and
only a small decrease in intensity.
As has been reported for a number of different surface struc-
tures (see, for example, Refs. [28–30]), STM imaging of the FeO film
is strongly dependent on the atomic-scale tip termination, leading
to a number of imaging modes, where both the appearance of the
high-symmetry domains of the moiré structure and the chemical
species imaged are different. For example, under similar or identi-
cal current and bias conditions, STM images have been obtained
showing Fe atoms, O atoms, or both simultaneously, and each of
the high-symmetry domains has been imaged with the largest
*
mains at its corners (the TOP and TOP domains are nearly
isoenergetic).
Here, we present additional DFT calculations using the experi-
pffiffiffiffiffiffi pffiffiffiffiffiffi
mentally observed 91 ꢁ 91R5:2ꢃ unit cell to obtain further
atomic-scale insight than was possible with the pseudomorphic
p(2 ꢁ 2) unit cells. In these more elaborate DFT calculations, we
investigate the formation of triangular dislocation loops from se-
ven adjacent oxygen vacancies in both the FCC (converted to
*
*
HCP ) and HCP (converted to FCC ) domains. For the HCP domain,
formation of the dislocation loop structure results in an energy
gain of 4.15 eV, while for the FCC domain, an energy gain of
0.53 eV is obtained. The difference of 3.62 eV between the two
structures clearly explains the experimental observation that the
smaller dislocation loops are always centered at HCP domains.
Interestingly, however, the formation of a dislocation loop at an
FCC domain is still slightly exothermic, despite the fact that the
*
HCP stacking configuration is less stable. This indicates that the
dominant component of the energy gain is related to the structure
of the dislocation loop itself, independent of the local stacking con-
figuration, and suggests that these dislocation loops would form
FeO
Dislocation loops
FCC*
FCC*
FCC
HCP
TOP
HCP*
FCC*
TOP*
c
Fe-mode
O-mode
O 1s
TOP*
FCC*
FCC*
534 530
a
d
e
b
BE [eV]
Fig. 2. (a) Illustration of the change in local stacking of the FeO film upon shifting the O lattice. The six different high-symmetry domains are shown, together with the
nomenclature used in this work. (b) Atomic model of both the big and small dislocation loops. Pt (white), 3-fold Fe (black), 2-fold Fe (grey), and O (red). (c) Fe-mode STM
image (30 ꢁ 30 Å2, 1.04 nA, 19.5 mV) of a small dislocation loop. (d) O-mode STM image (30 ꢁ 30 Å2, 0.52 nA, 6.4 mV) of a small dislocation loop. (e) O 1s XP-spectra of the
pristine FeO film (grey) and the dislocation structure (black) obtained after exposure of 1.2 ꢁ 10ꢂ6 mbar s atomic hydrogen and subsequent flashing to 500 K. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

14
J. Knudsen et al. / Surface Science 604 (2010) 11–20
at 1.0 ꢁ 10ꢂ7 mbar H2 background pressure at room temperature
53.1
54.1
and subsequent flashing to 500 K to desorb any excess hydrogen.
The O 1s spectra obtained before and after reduction are de-
picted in Fig. 3a, and show a reduction in intensity of the oxygen
signal by ꢀ33%, and only a small (+0.1 eV) shift in binding energy.
Corresponding Fe 3p spectra are shown in Fig. 3b, and show a small
increase in intensity following reduction, most likely due to a
reduction in screening by the O layer, and a shift of 1.0 eV to lower
binding energy. The LEED pattern obtained on this surface is shown
in Fig. 3c and displays distinct, albeit somewhat diffuse, p(2 ꢁ 2)
reflexes, indicative of relatively small coherent domains, consistent
with STM and with previous observations [10].
FeO
a
b
Fe O
3 2
Fe 3p
O 1s
The STM image in Fig. 1d, which depicts the partially-reduced
FeO film imaged with an STM tip in the Fe-mode, shows similar
contrast on the 2-fold coordinated Fe atoms along the edges of
the dislocation loops and on the p(2 ꢁ 2) domains. This suggests
that all Fe atoms in the p(2 ꢁ 2) structure are 2-fold coordinated
to O atoms. Based on this observation, together with the results
of XPS measurements, we propose the structural model depicted
in Fig. 3d, which consists of a honeycomb lattice of O atoms super-
imposed on a kagomé lattice of Fe atoms. Each O atom is coordi-
nated to three Fe atoms, while each Fe atom is coordinated to
two O atoms, corresponding to a stoichiometry of Fe3O2, consistent
with the ꢀ1.5:1 Fe:O ratio as estimated from XPS results.
534
532
530
528
60
56
BE [eV]
52
BE [eV]
c
d
As is the case for the pristine, unreduced FeO film, the Fe and O
sublattices of the p(2 ꢁ 2) structure can be resolved separately in
different tip-dependent STM imaging modes, as shown in Fig. 3f
and h, and the protrusions in these STM images directly reveal
the expected honeycomb and kagomé lattices, respectively. That
the former corresponds to the oxygen lattice and the latter to the
iron lattice is supported by STM images obtained on the more
mildly-reduced FeO films, where domains of the p(2 ꢁ 2) structure
coexist with the dislocation loop structure. In Fig. 3e an STM image
is depicted where a small patch of the honeycomb structure is
found to co-exist adjacent to two dislocation loops. The protrusions
observed at these dislocation loops are out of registry with the pro-
trusions for the surrounding FeO film, which is indicative of an
O-specific STM tip imaging mode. Likewise, Fig. 3g shows a domain
of the p(2 ꢁ 2) structure adjacent to the dislocation loops in the Fe-
specific imaging mode. The apparent height of the p(2 ꢁ 2) struc-
ture is similar to that observed along the edges of the dislocation
loop structure, and although the corrugation is very weak, a ka-
gomé lattice of protrusions can be discerned, which is evident from
the higher contrast image inset in Fig. 3g.
e
f
O-mode
g
h
Fe-mode
Since we know from DFT calculations that the FCC and HCP do-
mains are more stable as compared to the TOP domains in the
moiré-structured FeO film [16], we propose that Fe atoms occupy
hollow sites on the Pt(1 1 1) surface, with one half of the oxygen
atoms located at hollow sites and the other half in atop-Pt sites.
For this model of the p(2 ꢁ 2) structure we can derive Fe- and
O-coverages of 0.94 ML and 0.63 ML, respectively, with respect to
the coverages of the FeO film. The slightly reduced coverage of Fe
in the p(2 ꢁ 2) structure with respect to the pristine FeO film is
likely compensated by a number of different defects which are al-
ways observed in the STM images of the p(2 ꢁ 2) structure.
To gain further confirmation of the Fe3O2 surface structure, we
have carried out a detailed DFT study of a variety of structures with
different O and Fe coverages. The most stable structures are subse-
quently compared to the structure revealed from the STM results.
DFT calculations confirm that the structure depicted in Fig. 3d,
with Fe at hollow sites, is indeed the most stable p(2 ꢁ 2) structure,
and in the inset we have shown a simulated STM image obtained
from the DFT calculations of the local density of states (LDOS)
using the Tersoff–Hamann model. Although such simulated STM
images must be treated with caution considering our observation
that the tip state may play a strong role in STM imaging, the sim-
Fig. 3. (a) O 1s and (b) Fe 3p XP-spectra of the pristine FeO film (grey) and the
p(2 ꢁ 2) structure (black). (c) 100 eV LEED image of the p(2 ꢁ 2) structure. The
substrate unit cell is indicated in red, and the overlayer unit cell in blue. (d) Ball
model of the proposed Fe3O2 structure. Inset: Tersoff–Hamann simulated STM
image of the structure. (e) O-mode STM image of the mildly-reduced FeO film,
showing dislocation loops and the oxygen honeycomb structure. (f) STM image of
the p(2 ꢁ 2) structure showing a honeycomb pattern similar to (e). (g) Fe-mode
STM image showing the edge of a dislocation loop (top) and the p(2 ꢁ 2) structure
(middle). Inset: Excerpt with high contrast, showing the kagomé lattice of Fe atoms.
(h) Fe-mode STM image of the p(2 ꢁ 2) structure showing the kagomé lattice of Fe
atoms. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
upon oxygen removal even in the absence of any lattice mismatch-
related variation in overlayer-substrate bond strength.
3.2. The p(2 ꢁ 2) – Fe3O2 structure
Upon extensive exposure to atomic hydrogen at room tempera-
ture the FeO film is transformed completely into the p(2 ꢁ 2) struc-
ture, as revealed both in LEED and STM measurements. Typical
experimental parameters used to prepare a uniform, locally well-
ordered p(2 ꢁ 2) structure were 10 min. atomic hydrogen exposure

J. Knudsen et al. / Surface Science 604 (2010) 11–20
15
ulated image is consistent with much of our experimental data, as
seen by comparison with Fig. 3f.
54.1 53.4
FeO
We have also in the DFT calculations considered structural
models with H or Pt atoms which are incorporated in the ‘‘pocket
sites” of the honeycomb structure, which might also be consistent
with experimental data. However, the DFT calculations indicate
that the most stable adsorption site for hydrogen is atop oxygen
on this structure, forming OH, which we do not observe with
XPS, suggesting that hydrogen fully desorbs during the flash to
500 K, in agreement with previous temperature programmed
desorption (TPD) measurements [10]. Our DFT calculations also
predict that CO should adsorb strongly (ꢀ2 eV binding energy) to
a Pt adatom in the pocket site, but only weakly (ꢀ0.6 eV binding
energy) on the Fe3O2 structure. XPS measurements showed no
signs of CO adsorption upon dosing at room temperature, and we
thus rule out the presence of Pt adatoms in the structure.
a
b
Fe O
4
3
O 1s
Fe 3p
534
532 530 528
BE [eV]
60
56
BE [eV]
52
c
e
d
3.3. The p(3 ꢁ 3) – Fe4O3 structure
p(3x3)
At intermediate exposures of atomic hydrogen a p(3 ꢁ 3) struc-
ture is observed, which is always found, in the STM images, to
coexist with the dislocation loop and p(2 ꢁ 2) structures. A LEED
pattern obtained for a film containing both p(2 ꢁ 2) and p(3 ꢁ 3)
structures is depicted in Fig. 4c. Corresponding O 1s and Fe 3p
spectra are shown in Fig. 4a and b, together with that of the unre-
duced, pristine FeO. The decrease in O 1s intensity is ꢀ25% and the
shift in the Fe 3p peak position is 0.7 eV to lower binding energy.
These values, together with the fact that the p(3 ꢁ 3) structure
can be removed and converted into the p(2 ꢁ 2) structure by fur-
ther exposure to atomic hydrogen, suggest that this structure has
an Fe:O ratio intermediate between that of the unreduced FeO
and the Fe3O2 phase.
p(2x2)
f
Again turning to the STM image depicted in Fig. 1d, we find that
in the Fe imaging mode, the p(3 ꢁ 3) structure is imaged as a hon-
eycomb of bright protrusions, which mark the positions of 3-fold
coordinated Fe atoms. Furthermore, a number of STM images ob-
tained of coexisting p(2 ꢁ 2) and p(3 ꢁ 3) domains show distinct
similarity between the two structures, and both tend to show
6-fold symmetry. An example of this is depicted in Fig. 4d, where
similar bright protrusions are observed in positions marked by
blue (p(2 ꢁ 2)) and green (p(3 ꢁ 3)) hexagons. Taking these obser-
vations into account, the structural model shown in Fig. 4e can be
constructed beginning with placement of the 3-fold coordinated Fe
atoms in a p(3 ꢁ 3) unit cell, with three O atoms over hollow sites
and three atop Pt, connected by 2-fold coordinated Fe atoms. An
alternate model, shown in Fig. 4f, is formed by placing all O atoms
over Pt-hollow sites. Both models have Fe4O3 stoichiometry with
Fe- and O-coverages of 1.11 ML and 0.83 ML, respectively, relative
to the coverages of the FeO film.
DFT calculations indicate that the structure depicted in Fig. 4e,
with Fe at hcp hollow sites and O at fcc and atop sites, is the most
stable one, being lower in energy by >1 eV than the structure de-
picted in Fig. 4f. Simulated STM images of this structure, shown
as insets in Fig. 4e and f show protrusions over O atoms, as was
the case for the p(2 ꢁ 2) structure. The STM image depicted in
Fig. 4g shows an O-mode STM image of the p(3 ꢁ 3), p(2 ꢁ 2) and
dislocation loop structures. The honeycomb lattice of the p(2 ꢁ 2)
and the characteristic triangles of the p(3 ꢁ 3) structure can be
clearly observed, in excellent agreement with DFT simulated STM
images, which strongly support the structural models proposed.
g
O-mode
Fig. 4. (a) O 1s and (b) Fe 3p XP-spectra of the pristine FeO film (grey) and a
partially-reduced FeO film exhibiting both p(2 ꢁ 2) and p(3 ꢁ 3) structures (black).
(c) 100 eV LEED image of the surface corresponding to the XP-spectra. The Pt(1 1 1)
unit cell is indicated in red, the p(2 ꢁ 2) unit cell in blue and the p(3 ꢁ 3) in green.
(d) STM image showing coexisting p(2 ꢁ 2) (blue hexagon) and p(3 ꢁ 3) (green
hexagons) phases. Note the similar bright protrusions and the 6-fold symmetry of
both phases around these sites. (e and f) Structural models incorporating 3-fold
coordinated Fe in the observed honeycomb pattern, with Tersoff–Hamann simu-
lated STM images shown as insets. Structure (e) is more stable by ꢀ1 eV. (g)
Atomically-resolved STM image showing the p(3 ꢁ 3) (green unit cell), p(2 ꢁ 2)
(blue unit cell) and dislocation loop structures in an O-specific STM tip mode. Note
the good correspondence with simulated STM images of both the p(2 ꢁ 2) and
p(3 ꢁ 3) structures. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
Following exposure to atomic hydrogen, we observe by XPS the
formation of hydroxyls on the surface, indicated by the appearance
of an O 1s peak at 531.7 eV, shifted +2.2 eV relative to the main
peak at 529.5 eV assigned to oxygen in FeO. A shift of +2.2 eV
3.4. Overview of the reduction process – XPS results
agrees well with the reported value for OH groups on a-
Fe2O3(0 0 0 1) [32,33]. O 1s spectra of the clean (a1) and hydroxyl-
Having described the structures formed upon reduction of the
FeO film by atomic hydrogen, we now examine the overall reduc-
tion process in further detail.
ated FeO film (a2) are shown in Fig. 5a.
Spectrum (a2) in Fig. 5a clearly shows a peak at 531.7 eV corre-
sponding to OH and a peak at 529.5 eV corresponding to FeOx. The

16
J. Knudsen et al. / Surface Science 604 (2010) 11–20
FeO
x
54.2
Total
before flash
1.0
0.8
0.6
0.4
0.2
0.0
54.0
53.8
OH
FeO
Before flash
a1
a2
Total
after flash
53.6
53.4
+H
at
OH groups
before flash
53.2
53.0
Flash
500K
After flash
1
a3
3 4 56
3 4 56
3 4 56
3 456
2 3 456
2 3 456
534 532 530 528
BE [eV]
1
10
10
a
b
Exposure of atomic hydrogen
[10 mbar s]
c
Exposure of atomic hydrogen
-6
-6
[10 mbar s]
Fig. 5. (a) XP-spectra of the O 1s region of the clean FeO film (a1), the FeO film exposed to 3 ꢁ 10ꢂ6 mbar s hydrogen (a2) and subsequently flashed to 500 K (a3). (b) Total
oxygen coverage and coverage of OH groups calculated by fitting the O 1s spectra. The grey curve shows the oxygen coverage after the H-exposed surface was flashed to
500 K. (c) Fe 3p peak maxima as function of the exposure to atomic hydrogen for the exposed FeO films before (red) and after flashing to 500 K (grey). (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. (a–e) STM images (140 ꢁ 140 Å2) obtained directly after exposure to atomic hydrogen at room temperature. TOP, HCP and FCC domains are indicated with s,
D, and h,
respectively, using the convention from Refs. [16,24]. Exposures are given in the figure. (f–j) Representative STM images (140 ꢁ 140 Å2) obtained after exposure and a
subsequent flash to 500 K.
areas of the both peaks, referenced to the O 1s signal of the pristine
FeO film, are measured as function of atomic hydrogen exposure
and the total oxygen coverage and the coverage of OH groups are
plotted in Fig. 5b. The XPS measurements show that the OH cover-
phase. At exposures above 3 ꢁ 10ꢂ6 mbar s, the total oxygen cover-
age start to decrease, indicating that significant removal of oxygen
begins at this point.
After flashing the hydrogen-exposed FeO film to 500 K the O 1s
peak assigned to OH groups disappears as Fig. 5a3 demonstrates.
The energy of the O 1s peak assigned to FeOx is unaffected, but a
clear reduction of the intensity compared to the pristine FeO film
is observed. Fig. 5b shows the reduction of the FeOx peak, refer-
enced to the O 1s signal of the pristine FeO film, after different
exposures of hydrogen and a subsequent flash. In this case, no
age increases with exposure to
a maximum of 0.23 ML at
3 ꢁ 10ꢂ6 mbar s, after which it decreases to a limiting value of
0.19 ML following long atomic hydrogen exposure. Until the max-
imum OH coverage is reached, the total oxygen coverage is approx-
imately unchanged, indicating that accumulation of hydrogen on
the surface, without detectable oxygen removal, dominates in this

J. Knudsen et al. / Surface Science 604 (2010) 11–20
17
[35] and a metal surface [36]. Below an exposure of 1.6 ꢁ
10ꢂ6 mbar s, we observe that the coverage of hydroxyl groups in-
creases steadily with the atomic hydrogen exposure (Fig. 7a), with
no indications that oxygen is removed from the surface. Above this
point, where the OH coverage deduced from STM images is
0.12 ML, the triangular oxygen vacancy dislocation loops begin to
form, as shown in Fig. 6b. More extensive reduction of the FeO sur-
face is observed above an exposure of ꢀ4 ꢁ 10ꢂ6 mbar s, where
dark pits are observed in STM images, as depicted in Fig. 6c. Atom-
ically-resolved images of these pits have not been obtained, but
they presumably correspond to small oxygen-depleted domains
incorporating elements of the p(2 ꢁ 2) and p(3 ꢁ 3) structures.
Thereafter, STM imaging becomes more difficult, with poor tip sta-
bility indicative of weakly-bound, mobile species on the surface.
Nonetheless, the formation and eventual saturation of the surface
by the p(2 ꢁ 2) phase at higher exposures can be seen in Fig. 6d
and e.
a
0.12
0.08
0.04
p(3x3) +
p(2x2)
formation
Hydrogen
adsorption
Triangle
formation
0.0
0.8
0.6
0.4
0.2
0.0
FLASH TO 500 K
Triangle, p(3x3),
b
Triangle
formation
and p(2x2) formation
0
1
2
3
4
-6
Exposure of atomic hydrogen [10 mbar s]
Fig. 7. (a) Coverage of OH groups calculated from the protrusion observed in STM
images for different exposures of atomic hydrogen. (b) Relative number of HCP
domains converted to FCC domains after hydrogen adsorption and subsequent
flashing to 500 K.
Fig. 6f–j show STM images of each H-exposed surface after
flashing to 500 K. Surface hydroxyls are removed during the tem-
perature flash and the coverage of reduced structures is observed
to increase, consistent with TPD measurements by Huang and
Ranke showing that hydrogen desorbs from the surface both as
H2 and H2O upon heating from 300 K to 500 K [10]. The structures
observed after flashing show an even better ordering and as a re-
sult of this, as well as the much-improved tip stability following
desorption of hydrogen, characterization of the surface with STM
becomes much easier. As was the case during room-temperature
exposure to H, the initial stage of reduction involves the formation
of dislocation loops, and analysis of a large number of STM images
indicates that their coverage increases approximately linearly with
exposure up to ꢀ1.6 ꢁ 10ꢂ6 mbar s (see Fig. 7b), at which point va-
cancy dislocation loops have converted half of the film’s HCP do-
*
significant decrease in the O 1s signal can be detected for expo-
sures up to 1.2 ꢁ 10ꢂ6 mbar s. Above this point, reduction in the
O 1s signal becomes significant, and continues until an oxygen cov-
erage of ꢀ0.65 ML is reached after long exposure.
The Fe 3p core level in the FeO and the FeOx structures can also
be used to follow the details of the reduction process. Both the Fe
2p and Fe 3p peaks have complicated multiplet structures [34] and
rather than deconvoluting into a large number of peaks we simply
used the position of the peak maximum to follow the reduction of
the FeO film. For the H-exposed and subsequently flashed FeO
films (see Fig. 5c) we observe that the Fe 3p peak position begins
to decrease rapidly above 1.2 ꢁ 10ꢂ6 mbar s from its initial posi-
tion at 54.1 eV. After long exposure the Fe 3p peak position reaches
a value of 53.1 eV, halfway between that of the pristine film and of
metallic Fe deposited on Pt(1 1 1), for which we measure a Fe 3p
peak position of 52.2 eV. Thus, the FeO film does not completely re-
duce to metallic Fe. Comparing the measurements obtained for
reduction at room temperature and upon heating to 500 K, we ob-
serve that induction periods, prior to onset of the oxygen reduc-
tion, exist in both cases, and that the atomic hydrogen threshold
exposure at room temperature (ꢀ3 ꢁ 10ꢂ6 mbar s) is roughly dou-
ble that observed when the sample is flashed to 500 K
(ꢀ1.2 ꢁ 10ꢂ6 mbar s).
*
mains to FCC domains. At higher atomic hydrogen exposures,
p(2 ꢁ 2) and p(3 ꢁ 3) structures are formed, covering increasing
areas of the surface until finally the p(2 ꢁ 2) structure is the only
phase observed, in excellent agreement with XPS and LEED
measurements.
The induction periods observed in XPS measurements are clar-
ified by the STM results, and the different stages of the reduction
process are shown in Fig. 7a and b. In both cases, during exposure
at room temperature and following the 500 K flash, the formation
of defects, in the form of triangular dislocation loops, precedes the
growth of the more reduced p(3 ꢁ 3) and p(2 ꢁ 2) phases. At room
temperature, however, the formation of these defects is slow, and
only occurs when a sufficient OH coverage is reached. We note that
significant removal of oxygen from the surface only occurs during
formation of the p(2 ꢁ 2) and p(3 ꢁ 3) structures, as the vacancy
concentrations present in the form of dislocation loops (at the
threshold density of 0.5 dislocation loops per unit cell) amount
to only a few percent, and are therefore difficult to detect on the
basis of O 1s intensities. The longer induction period observed in
XPS measurements at room temperature thus corresponds directly
to the difference in the onset of formation of the more reduced
structures observed in STM and is limited exclusively by the initial
rate of vacancy formation to produce the dislocation loops.
The combination of XPS and STM in our study adds further in-
sight into the autocatalytic reduction mechanism originally pro-
posed by Huang and Ranke. Vacancies created during the initial
stage of reduction are stabilized in the form of dislocation loops
and the formation of the p(2 ꢁ 2) structure does not start directly
from single oxygen vacancies as a result of destabilized next-near-
est O atoms along the O rows as they proposed [10]. As we have so
far been unable to obtain well-resolved atomic scale STM images of
the surface directly after H exposure, but before flashing to 500 K,
we can not provide a precise mechanism for the reduction, which
must involve restructuring of the Fe lattice in addition to removal
Coincident with the reduction in O 1s intensity and the shift in
Fe 3p binding energy, both p(3 ꢁ 3) and p(2 ꢁ 2) reflexes appear in
the LEED patterns obtained following the XPS measurements
(Figs. 3c and 4c). The ordered LEED patterns, together with the
intermediate Fe 3p binding energy and presence of oxygen on
the surface, indicate that ordered, partially-reduced structures
are formed on the surface, in agreement with our STM results pre-
sented above and with previous results [10].
3.5. Overview of the reduction process – STM results
In order to gain further insight into the reduction of the FeO
film, we have performed a detailed STM study of the evolution of
FeO surface structure following exposure to atomic hydrogen and
subsequent heating. An overview of selected STM images after dif-
ferent exposures is shown in Fig. 6a–e.
The hydroxyls formed initially on the surface upon atomic
hydrogen exposure are clearly visible as bright protrusions in the
STM images shown in Fig. 6a–c. We observe that the individual
OH-related protrusions can be removed locally by applying tip
voltage pulses (U P 2.5 V) during scanning; similar behavior has
been observed previously for hydrogen adsorbed on an oxide

18
J. Knudsen et al. / Surface Science 604 (2010) 11–20
of oxygen. Nucleation and growth of the p(2 ꢁ 2) and p(3 ꢁ 3)
tices. The Fe3O2 phase is unique in that the metal atoms, rather
than the oxygen atoms, form the kagomé lattice. Given the current
interest in highly-frustrated magnetic structures [37], the Fe3O2
film could be an interesting system for surface characterization
by spin-sensitive techniques.
structures appears to begin at the ‘‘pits” observed to form either
*
at TOP domains in the centers of larger dislocation loops or at
TOP domains surrounded by three smaller dislocation loops (see,
for example, Fig. 6c). This supports the notion that the presence
of partially-reduced structures further facilitates the formation of
more reduced structures. The reason for the rapid growth of these
more reduced structures following the initial nucleation is still un-
clear, though we speculate that restructuring of the Fe lattice may
be the limiting factor, and that the presumably greater flexibility of
the more open Fe3O2 honeycomb lattice may further facilitate this
transformation. If this is the case, the autocatalytic reaction ob-
served for the purely two-dimensional FeO film may have little
similarity to those observed on bulk-like iron oxides.
Fe atoms coordinated to two O atoms in a ‘‘linear” (neglecting
coordinates normal to the surface) configuration are common to
all the reduced FeOx structures we have observed by STM. Reduc-
tion of the FeO film by atomic hydrogen can be described as a grad-
ual progression from the stoichiometric FeO film, consisting
entirely of 3-fold coordinated Fe atoms, to the dislocation loop
structure, having ꢀ15% 2-fold Fe atoms at half-saturation, to the
p(3 ꢁ 3) Fe4O3 structure with 75% 2-fold Fe and finally to the
p(2 ꢁ 2) Fe3O2 with 100% 2-fold Fe. The high stability of this
p(2 ꢁ 2) Fe3O2 structure appears to prevent complete reduction
to metallic Fe at room temperature.
3.6. Water dissociation
Spectrum a2 in Fig. 8a shows the O 1s photoemission region of
the Fe3O2 film after water exposure at room temperature at
1.0 ꢁ 10ꢂ6 mbar for 10 min. Comparison with the O 1s spectrum
before exposure (spectrum a1) shows that a new peak develops
at 531.6 eV, i.e. coincident with the observed binding energy for
hydroxyl groups formed on the FeO film after exposure to atomic
hydrogen. Simultaneously, we observe a small reduction in the
FeOx peak intensity, indicating that the water donates a proton to
a neighboring O atom upon dissociation. Ideally we expect that
the decrease in the FeOx signal would be equal to half of the ob-
served peak area due to hydroxyl groups. The measured ratio is
0.20, less than expected, suggesting that some of the donated
hydrogen may be bound to Fe or Pt, or that a portion of the ob-
served OH peak is due to molecularly adsorbed H2O. The direct
binding of a hydrogen atom to Fe seems plausible, based on the
work of Parkinson et al. [38], which demonstrated the formation
of HFeOH species upon reaction of Fe atoms with condensed water.
After flashing the hydroxylated Fe3O2 film to 500 K, we observe
an O 1s spectrum (spectrum a3) which is indistinguishable from
that observed before the water exposure, indicating that the hy-
droxyl groups formed upon water dosing leave the surface as
H2O once the surface is flashed to 500 K. In addition, the p(2 ꢁ 2)
Finally, we note the structural similarity between the Fe3O2
phase identified here and the structures of the Ti2O3 [1] and
V2O3 [2] phases grown on Pt(1 1 1) and Pd(1 1 1), respectively, all
of which consist of superpositions of honeycomb and kagomé lat-
FeO
x
a
a
b
FeO
x
OH
+H O
2
Fe O
(400K)
3
2
a1
OH
Fe 3p peak
maximum
+H O
b
2
c
54.0
+H O
(500K)
a2
2
53.4
1.0
0.9
0.8
0.7
0.6
0.5
x3
c
+Flash
OH
FeO
a3
x3
FeO
x
534
532
530
528
534
530
Fe O
400K
450K
3
2
BE [eV]
BE [eV]
348K 425K 500K
FeO
Fig. 8. (a) The
O 1s region of the clean Fe3O2 phase (a1), after exposure to
600 ꢁ 10ꢂ6 mbar s H2O at room temperature (a2), and after subsequently flashing
to 500 K (a3). Magnified difference spectra referring to the Fe3O2 structure are
shown below each spectrum. (b) 100.1 eV LEED image of the Fe3O2 structure after
water exposure. (c) 100.1 eV LEED image of the water exposed and flashed Fe3O2
structure.
Fig. 9. (a) XP-spectra of the water-exposed Fe3O2 film taken after water exposure of
600 ꢁ 10ꢂ6 mbar s at different temperatures. The pristine FeO film is included for
comparison. (b) 100.1 eV LEED image of the Fe3O2 film exposed to 600 ꢁ 10ꢂ6
mbar s of water at 500 K corresponding to the red points in c. (c) Coverage of FeOx,
OH and Fe 3p peak position calculated from the O 1s and Fe 3p spectra.

J. Knudsen et al. / Surface Science 604 (2010) 11–20
19
LEED pattern is weakened following water exposure (Fig. 8b), but
regains its original intensity after heating to 500 K (Fig. 8c). We
thus conclude that the water dissociation observed at room tem-
perature is fully reversible.
loop shrinks as the model depicted in Fig. 10b shows. This effect is
not observed for O2, which does not dissociate at the dislocation
loops at room temperature, even during STM measurements. In-
stead, we observe that the small quantities of water formed from
O2-induced desorption from the inner chamber walls [39] can re-
oxidize the film during measurement. An example is depicted in
Fig. 10c showing selected STM images from an STM movie re-
corded on the partly reduced FeO(1 1 1)/Pt(1 1 1) film in a back-
ground atmosphere of ꢀ1.3 ꢁ 10ꢂ6 mbar O2. Both dislocation
loops (visible as dark lines and one of which is highlighted with
a green triangle), and hydroxyls (imaged as yellow protrusions),
are clearly visible in STM images. In the image taken after 15 s
an additional bright double protrusion appears and the dimension
of the adjacent dislocation loop shrinks by one lattice distance.
After the bright double protrusion is formed, one of the hydroxyls
diffuses to an adjacent lattice site and the two hydroxyls can easily
be distinguished (image after 44 s). The fact that the double hydro-
xyl group appears exactly when the dislocation loop is observed to
shrink by one FeO-lattice distance clearly demonstrates that the
two hydroxyls are formed due to water dissociation in the pocket
site located at the corner of the dislocation loop. A schematic of
the most probable mechanism for water oxidation of the disloca-
tion loops is depicted in Fig. 10d. The fact that the double hydrox-
yls are observed exclusively at the corner sites suggests that the
barrier for dissociation is lower here than at the 2-fold coordinated
Fe atoms located along the edges of the dislocation loops.
Fig. 9a shows the O 1s spectra of the clean and the water-ex-
posed Fe3O2 films for different exposure temperatures. Relative
oxygen coverages deduced from these and similar plots are shown
in Fig. 9c. The O 1s peak assigned to FeOx increases with increasing
temperature above 350 K while the peak characteristic of hydroxyl
groups decreases. These two observations indicate that Fe3O2 film
becomes re-oxidized by water exposure above 350 K. This conclu-
sion is further supported by the increase in Fe 3p binding energy at
higher dosing temperatures and the LEED pattern observed after
water exposure at 500 K (see Fig. 9b). Here, the p(2 ꢁ 2) reflexes
have disappeared and we start to observe a very weak rosette
structure indicating that small FeO(1 1 1) domains are formed on
the surface. Even though the Fe3O2 film clearly becomes re-oxi-
dized when water is dosed above 350 K, it should be stressed that
it was impossible to form the stoichiometric FeO film again under
the experimentally tested temperature and pressure conditions.
Consistent with these XPS results, we observe with STM that a
mildly-reduced FeO film, exhibiting dislocation loops, is unaffected
by water exposure under high vacuum conditions. When water is
dosed during measurement, however, we observe tip-induced dis-
sociation of water at the dislocation loops inside the scanning area,
while those outside are unaffected. The STM images shown in
Fig. 10a are taken from an STM movie recorded in ꢀ2 ꢁ 10ꢂ8 mbar
H2O and they clearly demonstrate that the dislocation loops can be
removed completely by a stepwise reduction of their size. Consis-
tent with water dissociation the stepwise size reduction suggests
that one oxygen atom is added each time the size of the dislocation
4. Conclusions
Using a combination of STM, XPS and DFT, we have studied the
reduction by hydrogen of single-layer FeO films grown on
Pt(1 1 1) by atomic hydrogen. We found that the reduction process
starts with the development of triangularly shaped oxygen vacancy
dislocation loops. More reduced p(3 ꢁ 3) (Fe4O3) and p(2 ꢁ 2)
(Fe3O2) structures nucleate in the TOP domains once the oxygen
lattice is reconstructed by the development of several oxygen
vacancy dislocation loops in adjacent HCP domains. Structural
models of all three reduced structures are proposed and as a general
trend we find that the overall reduction process can be described as
the gradual transformation of 3-fold oxygen-coordinated Fe atoms
into 2-fold coordinated Fe atoms.
Furthermore, we have studied the interaction of water with the
reduced iron-oxide films. In contrast to the inert FeO film, we find
that the reduced structures are able to dissociate water. By heating
the surface during water exposure, we find that the film can be
partly re-oxidized.
Acknowledgments
We thank Karina Schulte, Joachim Schnadt and Jesper Andersen
and the MAX-Lab staff for help with the XPS-experiments. The sup-
port to the iNANO center from the Danish Research Agency, the
Strategic Research Council and the Danish Council for Independent
Research is gratefully acknowledged as is the support from the
Carlsberg and Villum Kahn Rasmussen Foundations, and the Euro-
pean Research Council. Work at UW was supported by DOE-BES,
Chemical Sciences Division. Supercomputing time was utilized at
NERSC, PNNL, and ORNL.
Fig. 10. (a) STM images taken from an STM movie recorded in 2 ꢁ 10ꢂ8 mbar H2O.
(b) Model showing the stepwise size reduction of the dislocation loops. (c) STM
images from an STM movie taken in 6.6 ꢁ 10ꢂ6 mbar O2. Notice that the dislocation
loop shrinks (indicated by green triangles) simultaneously with the development of
double hydroxyl groups. Finally the hydrogen atoms diffuse away from each other.
(d) Model of the water dissociation in the pockets sites formed by the corners of the
dislocation loops. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
References
[1] G. Barcaro, S. Agnoli, F. Sedona, G.A. Rizzi, A. Fortunelli, G. Granozzi, J. Phys.
Chem. C 113 (2009) 5721.
[2] S. Surnev, G. Kresse, M. Sock, M.G. Ramsey, F.P. Netzer, Surf. Sci. 495 (2001) 91.
[3] F.P. Netzer, Surf. Rev. Lett. 9 (2002) 1553.
[4] X.S. Liu, O. Korotkikh, R. Farrauto, Appl. Catal. A 226 (2002) 293.
[5] M. Watanabe, H. Uchida, K. Ohkubo, H. Igarashi, Appl. Catal. B 46 (2003) 595.

20
J. Knudsen et al. / Surface Science 604 (2010) 11–20
[6] T. Toda, H. Igarashi, H. Uchida, M. Watanabe, J. Electrochem. Soc. 146 (1999)
[22] G.H. Vurens, M. Salmeron, G.A. Somorjai, Surf. Sci. 201 (1988) 129.
[23] S.L. Dudarev, G.A. Botton, S.Y. Savrasov, C.J. Humphreys, A.P. Sutton, Phys. Rev.
B 57 (1998) 1505.
[24] L. Giordano, G. Pacchioni, J. Goniakowski, N. Nilius, E.D.L. Rienks, H.J. Freund,
Phys. Rev. B 76 (2007) 075416.
[25] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.
[26] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.
[27] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.
Fiolhais, Phys. Rev. B 46 (1992) 6671.
[28] W.A. Hofer, A.S. Foster, A.L. Shluger, Rev. Mod. Phys. 75 (2003) 1287.
[29] M. Schmid, H. Stadler, P. Varga, Phys. Rev. Lett. 70 (1993) 1441.
[30] L. Ruan, F. Besenbacher, I. Stensgaard, E. Laegsgaard, Phys. Rev. Lett. 70 (1993)
4079.
[31] G.O. Potschke, R.J. Behm, Phys. Rev. B 44 (1991) 1442.
[32] P. Liu, T. Kendelewicz, G.E. Brown, E.J. Nelson, S.A. Chambers, Surf. Sci. 417
(1998) 53.
[33] Y. Joseph, W. Ranke, W. Weiss, J. Phys. Chem. B 104 (2000) 3224.
[34] F. Sirotti, G. Rossi, Phys. Rev. B 49 (1994) 15682.
[35] S. Wendt et al., Surf. Sci. 598 (2005) 226.
[36] T. Mitsui, E. Fomin, D.F. Ogletree, M. Salmeron, A.U. Nilekar, M. Mavrikakis,
Angew. Chem. Int. Ed. Engl. 46 (2007) 5757.
[37] J.E. Greedan, J. Mater. Chem. 11 (2001) 37.
3750.
[7] T. Healey, P.D. Devries, Patent, WO2007019361-A1, US2007183968-A1,
EP1919823-A1, 2007.
[8] A. Basinska, T.P. Maniecki, W.K. Jozwiak, React. Kinet. Catal. Lett. 89 (2006)
319.
[9] Z.H. Qin, M. Lewandowski, Y.N. Sun, S. Shaikhutdinov, H.J. Freund, J. Phys.
Chem. C 112 (2008) 10209.
[10] W.X. Huang, W. Ranke, Surf. Sci. 600 (2006) 793.
[11] C. Wang, H. Daimon, S.H. Sun, Nano Lett. 9 (2009) 1493.
[12] Y.N. Sun, Z.H. Qin, M. Lewandowski, E. Carrasco, M. Sterrer, S. Shaikhutdinov,
H.J. Freund, J. Catal. 266 (2009) 359.
[13] H.C. Galloway, J.J. Benitez, M. Salmeron, Surf. Sci. 298 (1993) 127.
[14] W. Ranke, M. Ritter, W. Weiss, Phys. Rev. B 60 (1999) 1527.
[15] Y.J. Kim, C. Westphal, R.X. Ynzunza, H.C. Galloway, M. Salmeron, M.A.
VanHove, C.S. Fadley, Phys. Rev. B 55 (1997) 13448.
[16] L.R. Merte, J. Knudsen, L.C. Grabow, R.T. Vang, E. Laegsgaard, M. Mavrikakis, F.
Besenbacher, Surf. Sci. 603 (2009) L15.
[17] U. Leist, W. Ranke, K. Al-Shamery, Phys. Chem. Chem. Phys.
2435.
5 (2003)
[18] Y. Joseph, C. Kuhrs, W. Ranke, M. Ritter, W. Weiss, Chem. Phys. Lett. 314 (1999)
195.
[19] E. Laegsgaard, F. Besenbacher, K. Mortensen, I. Stensgaard, J. Microsc.-Oxford
152 (1988) 663.
[38] G.S. Parkinson, Y.K. Kim, Z. Dohnalek, R.S. Smith, B.D. Kay, J. Phys. Chem. C 113
(2009) 4960.
[20] W.X. Huang, W. Ranke, R. Schlogl, J. Phys. Chem. C 111 (2007) 2198.
[21] R. Nyholm, J.N. Andersen, U. Johansson, B.N. Jensen, I. Lindau, Nucl. Instrum.
Methods Phys. Res. A 467 (2001) 520.
[39] J. Matthiesen et al., ACS Nano 3 (2009) 517.