Crystal structures and growth mechanism for ultrathin films of ionic compound materials:
FeO„111… on Pt„111…
W. Ranke, M. Ritter, and W. Weiss
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany
共Received 28 December 1998兲
The growth and atomic structures of epitaxial iron-oxide films on Pt共111兲 were studied with scanning
tunneling microscopy and high-resolution low-energy electron diffraction. During the initial layer-by-layer
growth of FeO共111兲 four different structures are formed as the coverage increases to 2.5 monolayers, then a
three-dimensional growth of Fe
3
O
4
共111兲 islands begins. The structural transformations demonstrate that the
relaxations within the FeO共111兲 films and the Stranski-Krastanov growth mode are induced by electrostatic
surface energies, which dominate the energetics of thin film systems made up of ionic compound materials.
关S0163-1829共99兲11827-7兴
The growth of epitaxial thin films is of great importance
for many basic research studies and technological applica-
tions. New materials with novel properties can be obtained in
the form of ultrathin films with modified crystal structures.
Great efforts are put into the preparation of thin metal-oxide
films, which serve as model systems for fundamental surface
science and catalysis studies.
1–3
For all applications the con-
trol of the film structure and morphology is crucial. Whereas
a detailed picture of the atomic scale processes during het-
eroepitaxial growth has evolved for metals
4
and semiconduc-
tor materials,
5
not much is known about ionic compound
materials like metal oxides in this respect.
If heteroepitaxial growth takes place close to equilibrium
conditions, the growth mode depends on the surface free
energies of the substrate
␥
s
and the film
␥
f
, as well as on the
interfacial energy
␥
i,n
.
6
In the case of pseudomorphic
growth
␥
i,n
also contains the elastic energy of the strained
film and therefore increases with the number of layers n. The
surface energy of the strained film
␥
f
does not depend on the
film thickness for metal and semiconductor materials studied
so far. This often results in a Stranski-Krastanov growth
where the initial pseudomorphic layer-by-layer growth is
limited to a critical film thickness by the elastic energy. Then
a three-dimensional growth of unstrained islands with dislo-
cations at the substrate-overlayer interface or of coherently
strained islands with limited sizes becomes favorable.
5
In this work the growth of iron oxides onto Pt共111兲 is
investigated on an atomic and mesoscopic scale. Previous
investigations revealed four different atomic FeO共111兲 struc-
tures to occur during the initial growth,
7
and the lattice pa-
rameters of the last unknown structure could be clarified
here. The entire data give new insight into the energetics of
thin film systems of ionic compound materials, and they re-
veal a novel mechanism that determines the atomic struc-
tures and epitaxial growth mode of such systems. In contrast
to metals and semiconductors, interfacial and elastic strain
energies
␥
i,n
play no significant role any more; instead elec-
trostatic surface energies
␥
f,n
, which now depend on the
number of layers n, dominate the energetics of the system.
The experiments were performed in two ultrahigh vacuum
chambers with base pressures of 10
⫺ 10
mbar in order to com-
bine scanning tunneling microscopy 共STM兲 with high-
resolution low-energy electron diffraction 共LEED兲 measure-
ments, where the latter allowed us to determine lateral lattice
constants with an accuracy of 0.01 Å.
3,7
Onto a clean Pt共111兲
surface prepared by numerous sputter-annealing cycles iron-
oxide films were grown by repeated cycles of iron deposition
at room temperature and subsequent oxidation at T⫽ 870 K
in 10
⫺ 6
mbar oxygen.
7
All STM measurements were per-
formed at room temperature in the constant current mode.
Figure 1 shows large scale 共left兲 and atomic resolution
STM images 共right兲 of consecutive iron-oxide growth stages.
The large scale images reveal a layer-by-layer growth up to
2.1 monolayers 共ML兲 of coverage. At submonolayer cover-
age 共a兲 two monoatomic substrate steps indicated by the ar-
rows are visible, and bare platinum substrate areas coexist
with iron-oxide covered areas. In 共b兲 the first iron-oxide layer
is completed and hexagonally shaped second layer islands
with diameters around 200 Å have formed. The second layer
growth has proceeded in 共c兲 and is completed in 共d兲, where
some third layer islands have formed. At this growth stage
three-dimensional Fe
3
O
4
共111兲 islands start to grow with the
bulk lattice constant of this oxide.
7
All oxide overlayer steps
in Fig. 1 run along the
关
011
¯
兴
and
关
1
¯
10
兴
directions of the
substrate surface and exhibit height differences around 2.1 Å.
During this layer-by-layer growth four different atomic
film structures are formed, which are displayed in atomic
resolution STM images on the right side of Fig. 1. They are
denoted 1–4 according to the sequence of their appearance,
and their formation areas on the first three layer surfaces are
indicated by the numbers in the large scale images on the left
side. All structures exhibit protrusions forming hexagonal
surface lattices with the unit cells indicated by the diamonds,
which are about3Åinsize. The STM image of structure 2
was observed previously by Galloway, Benitez, and
Salmeron.
8
An image simulation based on quantum-
chemistry electron-scattering theory revealed the short peri-
odicity protrusions to occur at oxygen atom positions located
in the topmost layer of this film, and the model shown in Fig.
2 was proposed.
9
It consists of a hexagonal close-packed
iron-oxygen bilayer with oxygen located on top, which is
laterally expanded to a lattice constant of a
FeO
⫽ 3.09 Å 共the
PHYSICAL REVIEW B 15 JULY 1999-IVOLUME 60, NUMBER 3
PRB 60
0163-1829/99/60共3兲/1527共4兲/$15.00 1527 ©1999 The American Physical Society
value in bulk FeO is 3.04 Å兲 and rotated by
␣
⫽ 0.6° against
the Pt共111兲 surface lattice. This creates the moire
´
superstruc-
ture cell observed in the atomic resolution STM image,
which can be defined by the coincidence site 2 chosen arbi-
trarily on top of an underlying platinum atom. The lattice
constant of a
FeO
⫽ 3.09 Å indeed is revealed by the FeO共10兲
spot position in the high-resolution LEED pattern,
7
and the
model was further confirmed by photoelectron diffraction
experiments.
10
Analogous to structure 2, the protrusions in the atomic
resolution images of the other structures also are attributed to
topmost layer oxygen atom positions. Thus all FeO共111兲
films form purely oxygen-terminated surface structures as
depicted in Fig. 2. In this rigid model structures 1, 2, and 4
with their corresponding coincidence sites are obtained for
FeO共111兲 bilayers with the lateral lattice constants a
FeO
and
misfit angles
␣
listed in Table I, which were determined
experimentally by high-resolution LEED measurements and
which almost exactly coincide with the values obtained theo-
retically from the rigid models in Fig. 2.
7
The corresponding
moire
´
angles of these models also agree well with those ob-
served by STM.
Structure 3 is not characterized by a simple moire
´
super-
structure but by triangular features which appear with a pe-
riodicity of 38.4 Å along directions rotated by ⫾30° against
the atom rows on Pt共111兲. This corresponds to a (8)
⫻ 8)) R30° superstructure with respect to Pt共111兲 as indi-
cated by coincidence site 3 in Fig. 2. However, structure 3 is
observed in the second layer and is assumed to form on top
of structure 4 located in the first layer underneath, because
structure 4 forms a simple superstructure both with respect to
Pt共111兲 and structure 3. Structure 3 exhibits considerable
short-range disorder and not all oxygen atoms occur at
equivalent positions within the different triangles. The best
defined oxygen atom positions are aligned along the bright
rows in the STM image and form the triangle edges. They
are represented as black circles in Fig. 3. The triangles were
arranged so that all corners occupy equivalent positions with
respect to the (8⫻ 8) unit cell of structure 4 underneath 共in-
FIG. 1. Large scale 共left兲 and atomic resolution 共right兲 STM
images of epitaxial FeO共111兲 filmsonPt共111兲. Four different coin-
cidence structures 1–4 are formed sequentially as the coverage in-
creases. They exhibit different contrasts in the large scale images as
indicated by the numbers.
FIG. 2. Rigid model of the four FeO共111兲 structures formed on
Pt共111兲. FeO共111兲 bilayers with different lattice constants a
FeO
and
rotated against the platinum surface lattice by different angles
␣
lead to coincidence sites 1–4. Here coincidence structure 2 with its
superstructure cell is shown.
TABLE I. Experimentally determined lateral lattice constants
a
FeO
, rotation misfit angles against the Pt共111兲 surface lattice
␣
,
and Fe-O共111兲 interlayer distances of the four FeO共111兲 structures
formed in the first three layers. The Fe-O bond lengths were calcu-
lated from a
FeO
and the interlayer distances.
Structure,
layer
a
FeO
共Å兲
Misfit
angle
␣
Fe-O
interlayer
distance
Fe-O
bond
length
共Å兲
S1, 1st layer 3.11 1.3°
S2, 1st layer 3.09 0.6° 0.68 1.91
S2, 2nd layer 3.09 0.6° 1.05 2.1
S3, 2nd layer 3.40
a
⬍1.05
S4, 3rd layer 3.15 0° 1.05 2.1
FeO bulk 3.04 1.25 2.16
a
Average value.
1528 PRB 60
BRIEF REPORTS
dicated by the dotted lines兲, and the large black oxygen at-
oms occupy hollow site positions with respect to the first
FeO bilayer which correctly continue the NaCl type lattice.
The grey oxygen atoms are on ‘‘wrong,’’ mostly near-bridge
positions. This causes a more open packing with interatomic
distances between that of structure 4 共3.17 Å兲 and values as
large as 3.8 Å. The LEED spot positions of structure 3 cor-
respond to a lattice constant of 3.4 Å, which can be consid-
ered as an average interatomic distance. Within the areas in
Fig. 3 where no second layer atoms are plotted the atom
positions could not be determined due to the lack of clear
intensity maxima in the STM images.
All FeO共111兲 lattice constants a
FeO
listed in Table I are
larger than the FeO bulk value, and for coverages above 1.5
ML they increase in the second layer when compared to the
first layer. This is an unexpected behavior because the
Pt共111兲 surface lattice constant 共2.77 Å兲 is considerably
smaller than the FeO bulk value 共3.04 Å兲. Instead of adopt-
ing the FeO bulk structure with increasing film thickness the
lateral expansion increases, and the increasing elastic strain
energy must be overcompensated by another stabilizing con-
tribution. The lateral expansions are accompanied by re-
duced interlayer distances as listed in Table I. For structure 2
formed in the first layer an interlayer distance of 0.68 Å was
deduced from photoelectron diffraction measurements.
10
An
interlayer distance of 1.05 Å is deduced for structures 2 and
4 in the second and third layer, respectively, as deduced from
the height differences between consecutive layers with equal
surface structures measured by STM. The interlayer distance
of the strongly expanded structure 3 is assumed to be even
smaller.
Structures 1, 2, and 4 do occur in the first FeO共111兲 layers
and represent slightly different substrate-overlayer interface
geometries. This indicates that the interfacial energy does not
depend critically on the exact geometry formed at the inter-
face to the substrate. Although an exact description of the
film energetics would require ab initio total energy calcula-
tions, the observed relaxations can be understood qualita-
tively when considering the ionic character of the oxide ma-
terial. All FeO共111兲 films form polar unreconstructed
surfaces terminated by close-packed oxygen layers. In a
purely ionic model as formulated by Tasker
11
each iron-
oxygen bilayer produces an electric field which increases the
surface potential of the oxide film V(z) 共z is the distance
from the substrate surface兲. This results in a surface energy
␥
f,n
that increases with the number of layers n as illustrated
schematically in Fig. 4 for one 共a兲–共c兲 and two 共d兲–共f兲 bi-
layer thick films. Since the potential variation V(z) within
each bilayer is proportional to the interlayer distance indi-
cated by the solid lines, the surface energy of the films can
be decreased by relaxations that reduce the interlayer dis-
tances. This situation is depicted in Figs. 4共c兲 and 4共f兲 and by
the broken lines in Figs. 4共b兲 and 4共e兲.
The electrostatic surface energy is the driving force for
the reduction of the Fe-O interlayer distances in the
FeO共111兲 films and the structural transformations observed.
Since reduced interlayer spacings would reduce the Fe-O
bond lengths, the films respond by increasing their lateral
lattice constants. The resulting Fe-O bond lengths in the sec-
ond and third FeO layers are 2.1 Å, slightly smaller than the
FeO bulk value of 2.16 Å 共see Table I兲. The interlayer dis-
tance for structure 2 in the first layer 共0.68 Å兲 results in an
Fe-O bond length of 1.91 Å, which corresponds to a value
expected for adsorbed oxygen as observed on rhodium.
12
This indicates a reduced ionic charge in the first bilayer,
which must be induced by a charge redistribution within the
overlayer as indicated by the
␦
⫺
in Fig. 4共f兲. Such a charge
redistribution further decreases the surface energy
13
as indi-
cated by the dotted line in Fig. 4共e兲.
That a thickness-dependent electrostatic surface energy
␥
f,n
dominates the total energy of the system also is reflected
by the sequence of structural transformations visible in the
large scale STM images in Fig. 1. At submonolayer cover-
ages structure 1 is formed. Because a second bilayer consid-
erably increases the surface energy, the system tries to avoid
its formation and forms the slightly compressed structure 2
just before the first layer gets completed. Then the second
layer starts to grow with structure 2, and the surface energy
of the film strongly increases with the second layer coverage.
FIG. 3. Model of the ‘‘triangle’’ structure 3 formed in the sec-
ond layer on top of structure 4 in the first layer. The (8)⫻8) )
and (8⫻ 8) superstructure cells of structures 3 and 4 are indicated
by the solid and dotted lines, respectively.
FIG. 4. Schematic side view representation of one and two bi-
layer thick films of an ionic material on top of a metal substrate.
The electrostatic potential V(z) for unrelaxed and relaxed overlay-
ers are indicated by the solid and broken lines, respectively 共z is the
distance from the substrate surface兲. The dotted line indicates the
potential after a charge redistribution of
␦
⫺
has occurred within the
overlayer.
PRB 60
1529BRIEF REPORTS
This is the driving force for the transformation of structure 2
into structure 3, which takes place at coverages around 1.5
ML. In this extremely expanded structure with O-O distances
up to 3.8 Å topmost oxygen atoms relax deeply into the
underlying iron layer, leading to a strongly reduced Fe-O
interlayer distance. Before the second layer gets completed
structure 3 transforms into the more compressed structure 4,
because again the system tries to avoid the formation of the
next layer. Then third layer islands with structure 4 start to
grow. As can be seen in the large scale STM image in Fig.
1共d兲 the more compressed structure 2 forms around these
third layer islands. Again the system tries to incorporate as
much material as possible into the first two bilayers in order
to avoid the third layer growth. The maximal FeO coverage
is 2.5 ML, then the growth of Fe
3
O
4
共111兲 islands forming a
more stable surface structure becomes favorable.
14
In summary, iron oxide grows in a Stranski-Krastanov
mode onto Pt共111兲 with two closely related crystal structures
involved, FeO and Fe
3
O
4
. The initial FeO layer-by-layer
growth is limited to a maximal thickness of 2.5 ML, then a
three-dimensional growth of Fe
3
O
4
共111兲 islands becomes fa-
vorable. This growth mode and the large relaxations within
the ultrathin FeO共111兲 films are caused by the electrostatic
FeO共111兲 surface energy which increases with the number of
layers n, whereas interfacial and elastic strain energies play
no significant role. This represents a new mechanism that
dominates the energetics of thin film systems of ionic mate-
rials and that determines their growth mode and crystal struc-
tures. Ultrathin films with polar surface orientations always
reduce their interlayer distance, resulting in laterally ex-
panded structures, which recently also was observed for
␣
-Fe
2
O
3
共0001兲 grown onto Al
2
O
3
共0001兲 substrates.
15
We thank M. Swoboda for technical assistance and the
Deutsche Forschungsgemeinschaft for financial support.
1
H.-J. Freund, Angew. Chem. 109/5, 444 共1997兲.
2
S. C. Street, C. Xu, and D. W. Goodman, Annu. Rev. Phys.
Chem. 48,43共1997兲.
3
W. Weiss, M. Ritter, D. Zscherpel, M. Swoboda, and R. Schlo
¨
gl,
J. Vac. Sci. Technol. A 16,21共1998兲.
4
H. Brune, Surf. Sci. Rep. 31, 121 共1998兲.
5
E. Pehlke, N. Moll, A. Kley, and M. Scheffler, Appl. Phys. A:
Mater. Sci. Process. 65, 525 共1997兲.
6
E. Bauer and J. H. van der Merve, Phys. Rev. B 33, 3657 共1986兲.
7
M. Ritter, W. Ranke, and W. Weiss, Phys. Rev. B 57, 7240
共1998兲.
8
H. C. Galloway, J. J. Benitez, and M. Salmeron, Surf. Sci. 298,
127 共1993兲.
9
H. C. Galloway, P. Sautet, and M. Salmeron, Phys. Rev. B 54,
R11 145 共1996兲.
10
Y. J. Kim, C. Westphal, R. X. Ynzunza, H. C. Galloway, M.
Salmeron, M. A. Van Hove, and C. S. Fadley, Phys. Rev. B 55,
R13 448 共1997兲.
11
P. W. Tasker, J. Phys. C 12, 4977 共1979兲.
12
S. Schwegmann, H. Over, V. De Renzi, and G. Ertl, Surf. Sci.
375,91共1995兲.
13
C. Noguera, Physics and Chemistry at Oxide Surfaces 共Cam-
bridge University Press, Cambridge, England, 1996兲.
14
W. Weiss and M. Ritter, Phys. Rev. B 59, 5201 共1999兲; M. Ritter
and W. Weiss, Surf. Sci. 共to be published兲.
15
S. Chambers 共private communication兲.
1530 PRB 60
BRIEF REPORTS
