Controlled cluster condensation into preformed nanometer-sized pits
H. Hövela)
IBM Research Division, Zurich Research Laboratory, CH–8803 Rüschlikon, Switzerland
Th. Becker
Omicron Vakuumphysik GmbH, D–65232 Taunusstein, Germany
A. Bettac
Universität Rostock, D–18051 Rostock, Germany
B. Reihl, M. Tschudy, and E. J. Williamsb)
IBM Research Division, Zurich Research Laboratory, CH-8803 Rüschlikon, Switzerland
~Received 23 July 1996; accepted for publication 30 September 1996!
We have performed scanning tunneling microscopy ~STM! in ultrahigh vacuum and transmission
electron microscopy ~TEM! on silver and gold clusters grown in preformed nanometer-sized pits on
the surface of highly oriented pyrolytic graphite. We describe the preparation method and evaluate
the three-dimensional shape of the clusters using a combination of STM and TEM applied to the
same cluster sample. The nanometer-sized pits were essential to fix the clusters in position when
using STM. The influence of the tip shape on the STM imaging of nanometer-sized clusters is
discussed. © 1997 American Institute of Physics. @S0021-8979~97!06301-9#
I. INTRODUCTION
There is a growing interest in the study of mesoscopic
structures having a structural size of a few nanometers. For
example, the further miniaturization of electronic circuits
will lead to this regime. The investigation of clusters, i.e.,
small particles with a size of a few nanometers, is an important step because their properties generally differ from ‘‘expected’’ bulk behavior.
Scanning probe methods offer the unique capability of
directly mapping physical properties with atomic resolution;
therefore, they are well suited for the investigation of single
clusters, grown or deposited on a substrate. Corresponding
experiments, however, present conflicting demands. On the
one hand, the substrate should be smooth and have a weak
interaction with the clusters because on a rough substrate the
clusters are masked by the topology and with a strong
cluster–substrate interaction the structural and electronic
properties of the clusters can be changed drastically compared to those of free clusters,1 which complicates the interpretation of the experiments. On the other hand, the investigation of clusters on a smooth and weakly interacting
substrate with a scanning probe technique often proves impossible because the clusters are displaced by the tip. Only if
a ‘‘gentle’’ probe such as noncontact scanning force microscopy is combined with at least an intermediate cluster–
substrate interaction do the clusters stay fixed during
scanning.2 In order to use probes with a stronger tip–sample
interaction such as scanning tunneling microscopy ~STM! or
scanning force microscopy in contact mode, which generally
offer a higher spatial resolution, the cluster–substrate interaction has to be stronger. Junno et al.,3 for example,
a!
Present address: Universität Dortmund, Experimentelle Physik I, D-44221
Dortmund, Germany. Electronic mail: [email protected]
b!
Also at: Départment de Physique de la Matière Condensée, Université de
Genève, CH-1211 Genève 4, Switzerland.
154
J. Appl. Phys. 81 (1), 1 January 1997
achieved this with a sintering of the sample, which introduced an interfacial chemical interaction.
In this article, we will show that it is possible to use a
topological structuring of the substrate ~preformed
nanometer-sized pits! in order to fix the clusters in position.
This has the advantage that a substrate can be used that has
no chemical interaction with the cluster; hence, the cluster
properties are not influenced strongly by the substrate. In
addition, the use of such prestructured substrates offers the
capability of simple cluster production in ultrahigh vaccum
~UHV! by evaporation of the cluster material onto the substrate. If the mobility of the incoming atoms is high enough,
the material condenses exclusively in the pits, which act as
condensation centers.
The development of STM attracted renewed interest in
the mechanism of oxidation of graphite surfaces at monolayer steps. With the possibility of imaging the surface of
highly oriented pyrolytic graphite ~HOPG! with atomic resolution using STM, it was proved that this mechanism leads to
pits only 1 monolayer deep which form by the preferential
oxidation of the first monolayer originating from existing
surface defects. The existence of large circular pits was
already perceived before the invention of STM with the
technique of etch–decoration transmission electron
microscopy,4,5 but with STM both the initial small-pit stage
of the process and the depth of 1 monolayer could be
measured.6,7 Besides using pit formation for the study of the
chemical reaction of graphite with oxygen and other gases,8
the pits are promising structural elements of mesoscopic size
because their diameter is uniform and is given by the parameters temperature T, oxygen partial pressure p O2 , and duration of the oxidation process t. Thus, diameters in the range
of a few nanometers for the smallest pits and up to several
hundred nanometers ~until the pits overlap! can be achieved
by controlling the oxidation process parameters. These welldefined, 1-monolayer-deep ‘‘corrals’’ were used for example
for the study of nucleation in self-assembled molecular
0021-8979/97/81(1)/154/5/$10.00
© 1997 American Institute of Physics
films.9 The diameter d of the pits depends on the parameters
of the oxidation process corresponding to the formula
d5C exp~ 2E a /RT ! p O2t,
~1!
where R is the gas constant and E a 51.273105 J mol21 is
the activation energy for the monolayer etching with O2 .7
The constant C is of the order of 104 nm mbar21 s21. Yang
and Wong found that the speed of pit growth is dependent on
the pit density.5 Using their data, one can calculate
C523104 nm mbar21 s21 for a pit density of 1 pit mm22.
This decreases for higher pit densities and becomes constant
(C543103 nm mbar21 s21! for pit densities above 15 pit
mm22. The reason for this is the migration of oxygen from
defect-free parts of the graphite surface to the pits, which
causes a higher growth speed in the case of low pit densities.
The idea behind our work was to use small pits ~diameters of about 5 nm! and a suitable number of pits per area
for the production of separated metal clusters by condensation of evaporated metal atoms into these pits. For the STM
investigation of the clusters we expected that the pits would
also help fix the clusters on the HOPG substrate, which otherwise interacts only very weakly with the metals used ~silver and gold!.
II. RESULTS AND DISCUSSION
In preliminary experiments we found that the natural defect density on the HOPG surface, which was freshly prepared by tape cleaving before oxidation, was too low for the
use of nanometer-sized pits. With a density of below 1 pit
mm22, small pits could rarely be found with STM and, during the evaporation of metal atoms, the condensation centers
given by the pits would compete with a surplus of other
condensation centers formed by surface steps or other irregular surface structures. Hence we increased the defect density
on the HOPG surface by sputtering with Ne ions of 1 keV
kinetic energy. A sputter current of 231029 A was applied
for 20 s to the sample surface of 636 mm2. Given a sputter
efficiency of about 0.1,10 this corresponds to a defect density
of about 73102 defects per mm2. Alternative means for defect production are described in Ref. 11 ~a microwave
plasma discharge! and Ref. 12 ~accelerated carbon clusters!.
For the oxidation process we used T5530 °C, p O25213
mbar ~heating in air! and t520 min. This should lead to a pit
diameter of approximately 6 nm according to Eq. ~1! using
the constant C543103 nm mbar21 s21, which should be
valid for this high defect density.5
After this two-step preparation, we imaged the HOPG
surface before the metal deposition using an UHV–STM
~Fig. 1!. The experimental values 53102 pits mm22 and a
mean pit diameter of 5 nm are in agreement with the values
expected from the preparation parameters used. The standard
deviation of the diameters is 62 nm. Figure 2 shows a typical pit with atomic resolution. The oxidized area in the surface layer is not as perfectly circular as it is for larger pits
with a diameter of several tens of nanometers. At the edges
of the pit, the lattice structure of the surface layer is distorted. Near the edge of the pit superstructures similar to
those discussed in Ref. 13 can be observed. It should be
J. Appl. Phys., Vol. 81, No. 1, 1 January 1997
FIG. 1. Surface of the HOPG after sputtering and oxidation. This UHV–
STM image shows the nanometer-sized pits in the first monolayer.
mentioned that this picture was measured at a temperature of
5 K with a newly built low-temperature UHV–STM, which
might have been advantageous with respect to stability while
scanning the edges of the pit. All other STM images were
taken in UHV at room temperature.
On this preformed surface with its nanometer-sized pits,
we first deposited silver with an effective layer thickness of
about 0.1 nm. The substrate temperature was 350 °C. We
transferred the sample into the UHV–STM without removing it from the vacuum chamber. Vacuum conditions are not
critical during the preparation of the pits on the inert HOPG,
but during this step it is important to keep the surface of the
silver clusters clean. Cluster production by evaporation of
the cluster material onto the HOPG surface with preformed
pits is compatible with UHV conditions. This should allow
the use of a broad range of cluster materials, including reac-
FIG. 2. Typical pit produced in the HOPG surface imaged with lowtemperature UHV–STM in atomic resolution (10.4310.4 nm2 ).
Hövel et al.
155
FIG. 3. Topology of the sample after deposition of silver onto the HOPG
surface with preformed nanometer-sized pits. The UHV–STM image indicates that nanometer-sized metal clusters with a uniform size have grown
exclusively into the pits.
tive ones. We have hitherto been restricted to the noble metals silver and gold only because we evaporated the metal in
the load-lock chamber of our system, which has a base pressure of about 1026 mbar. Owing to the fixation in the pits it
was possible to image the clusters with STM ~Fig. 3! using
the constant-current mode. However, a low tunneling voltage
and a low tunnel current were necessary in order not to
knock the clusters out of the pits (U T 50.23 V, I T 50.18 nA
are the parameters for Figs. 3 and 4!. The number of clusters
per area is the same as the number of pits per area before
silver deposition. This indicates that the silver has condensed
exclusively at the preformed pits. During prolonged scanning
of a given area, a cluster was sometimes pushed out of its pit.
Then the underlying pit could be seen, as shown in the threedimensional plot in Fig. 4 together with some remaining
clusters. The displaced clusters were immediately pushed out
of the scan area. This is in agreement with experiments when
we deposited silver clusters produced by gas aggregation on
FIG. 4. Three-dimensional display of silver clusters on the HOPG surface.
During prolonged scanning one cluster was pushed out of its underlying pit,
as revealed in the UHV–STM image.
156
J. Appl. Phys., Vol. 81, No. 1, 1 January 1997
FIG. 5. Histogram of the cluster size distribution. Because the lateral diameter is strongly influenced by the shape of the tip, the height of the clusters
was evaluated instead. Together with the data for the silver clusters the
height distribution of the larger gold clusters ~described below! is recorded
for comparison.
an atomically flat HOPG surface. Without the fixation to the
pits it was impossible to image the clusters with STM. In this
image one can also observe that all the clusters show the
same shape. This is caused by the convolution with the tip
shape. We used chemically etched tungsten tips that were
freshly prepared before introduction into the UHV system.
Assuming a typical tip radius of the order of 10 nm and a
three-dimensional cluster size of a few nanometers, the con-
FIG. 6. Together with the clusters located in the preformed pits, clusters
condensed at surface steps could also be observed with the UHV–STM.
Hövel et al.
FIG. 7. Gold clusters grown into preformed nanometer-sized pits on a
HOPG surface. Topology (3003300 nm2 ) and line profile measured with
the UHV–STM. The clusters depicted in profile are of mean height according to the statistics. The horizontal bar inserted into the line profile indicates
the mean lateral diameter as measured with TEM ~Fig. 8! to illustrate the
influence of the tip shape.
volution with the tip shape will lead to a lateral cluster diameter in the STM image which is an image of the tip rather
than of the cluster.14 For this reason we evaluated only the
height distribution of the silver clusters. The histogram in
Fig. 5 shows that this height is very uniform. The mean
height and its standard deviation is 3.760.7 nm. In addition
to the clusters located in the pits, condensation of metal particles along surface steps could be observed in some pictures
~Fig. 6!. The condensation of metal particles at surface steps
was investigated and discussed thoroughly using electron microscopy ~see, e.g., Ref. 15!. The cluster nucleation and
growth in the pits represents the interesting special case of
closed circular surface steps where the diameter of the circles
is of similar size as the metal particles. This is probably one
reason for the very uniform cluster size. We observed decoration of the edges of the pits with chains of clusters if the
pits are too large or only part of the pits filled with clusters if
there are too many pits per area for the deposited metal material. The preparation parameters presented here were optimized to obtain a homogeneous distribution of clusters well
fixed to the nanostructured HOPG surface.
In order to clarify the influence of the tip shape and to
determine the lateral diameter of the clusters as well, we
J. Appl. Phys., Vol. 81, No. 1, 1 January 1997
FIG. 8. TEM image of the same sample with gold clusters as in Fig. 7 taken
after thinning the HOPG from the backside. This image allowed the lateral
diameter of the clusters to be evaluated, which, together with the cluster
height measured with the UHV–STM, yields the three-dimensional shape of
the clusters. As in Fig. 6 ~measured with the UHV–STM! condensation at
surface steps can be observed in addition to the clusters grown into the pits.
prepared a sample for an additional investigation of the lateral cluster shape with transmission electron microscopy
~TEM!. We used the same preparation of the HOPG with the
nanometer-sized pits, but evaporated gold onto the surface
with an effective layer thickness eight times larger than for
the silver clusters discussed above. The substrate was heated
to 600 °C during the gold evaporation. The UHV–STM measurement ~Fig. 7! revealed larger clusters with a height distribution of 6.760.7 nm ~see Fig. 5!. The number of clusters
per area stayed equal to that for the sample with the silver
clusters because we used a HOPG surface with the same
number of pits per area. We used gold and a larger cluster
size in order to achieve a good contrast for TEM, which was
performed on the same sample after thinning the graphite to
a thickness of about 100 nm. This was done by gluing the
HOPG with the gold clusters in front onto cement ~glycol
phthalate! and thinning the sample from the backside first by
cleaving with a knife and then by taping until the graphite
film became optically transparent. After dissolving the cement in an acetone bath, the floating graphite film was captured on copper grids. A TEM image ~Fig. 8! shows the
lateral size of the clusters, which also proves to be rather
uniform. The distribution of the cluster diameter is 10.161.9
nm. Together with the height information from the STM data
Hövel et al.
157
the three-dimensional shape of the clusters can be deduced.
We have indicated the mean cluster diameter measured with
TEM in the line profile of the STM image ~Fig. 7!. This
illustrates the effect of the convolution with the tip shape.
The increased lateral size of the clusters by convolution with
the tip shape also leads to the impression of connected clusters in the UHV–STM image in Fig. 7 while the TEM image
~Fig. 8! shows that most of the clusters are well separated. In
both the STM and the TEM images the number of clusters
per area is the same, which proves that the clusters are not
pushed away by the STM tip. Otherwise one would observe
fewer clusters per area when using the scanning probe microscope than in the TEM images as it is described in Ref. 2.
The aligned chains of clusters in Fig. 8 can be identified with
clusters condensed at surface steps ~see Fig. 6!.
should allow the use of nonlocal measurement methods,
which cannot be applied for single clusters manufactured using the STM tip. Together with the opportunity of using
STM due to the fixation of the clusters to the pits this
will enable a direct comparision of the nonlocal measurement methods with STM. For example it should be possible
to compare scanning tunneling spectroscopy and photoelectron spectroscopy concerning the electronic structure of the
clusters.
ACKNOWLEDGMENTS
H.H. acknowledges the support by the Feodor Lynen
Program of the Alexander von Humboldt Foundation and
E.J.W. that of the Swiss National Science Foundation.
K.-P. Charlé, W. Schulze, and B. Winter, Z. Phys. D 12, 471 ~1989!; B. N.
J. Persson, Surf. Sci. 281, 153 ~1993!; H. Hövel, S. Fritz, A. Hilger, U.
Kreibig, and M. Vollmer, Phys. Rev. B 48, 18 178 ~1993!.
2
W. Mahoney, D. M. Schaefer, A. Patil, R. P. Andres, and R. Reifenberger,
Surf. Sci. 316, 383 ~1994!.
3
T. Junno, S. Anand, K. Deppert, L. Montelius, and L. Samuelson, Appl.
Phys. Lett. 66, 3295 ~1995!.
4
G. R. Hennig, J. Chem. Phys. 40, 2877 ~1964!.
5
R. T. Yang and C. Wong, J. Chem. Phys. 75, 4471 ~1981!.
6
H. Chang and A. J. Bard, J. Am. Chem. Soc. 112, 4598 ~1990!.
7
X. Chu and L. D. Schmidt, Carbon 29, 1251 ~1991!.
8
X. Chu and L. D. Schmidt, Surf. Sci. 268, 325 ~1992!; H. Chang and A. J.
Bard, J. Am. Chem. Soc. 113, 5588 ~1991!.
9
D. L. Patrik, V. J. Cee, and T. P. Beebe, Jr., Science 265, 231 ~1994!.
10
M. Henzler and W. Göpel, Oberflächenphysik des Festkörpers ~Teubner,
Stuttgart, 1991!.
11
A. Tracsz, A. A. Kalachev, G. Wegner, and P. Rabe, Langmuir 11, 2840
~1995!.
12
G. Bräuchle, S. Richard-Schneider, D. Illig, J. Rockenberger, R. D. Beck,
and M. M. Kappes, Appl. Phys. Lett. 67, 52 ~1995!.
13
J. Valenzuelea-Benavides and L. Morales de la Garza, Surf. Sci. 330, 227
~1995!.
14
J. Barbet, A. Garvin, J. Thimonier, J.-P. Chauvin, and J. Rocca-Serra,
Ultramicroscopy 50, 355 ~1993!.
15
J. J. Métois, J. C. Heyraud, and R. Kern, Surf. Sci. 78, 191 ~1978!.
1
III. CONCLUSION
Using a combination of UHV–STM and TEM we have
investigated metal clusters grown on HOPG. We used preformed nanometer-sized pits produced in the first monolayer
of the HOPG surface by oxidation of surface defects, which
were introduced by sputtering. This had several advantages.
Owing to the nanometer-sized pits the cluster growth could
be performed in a controlled way. The pits acted as welldefined condensation centers, which led to a very uniform
cluster size when the cluster material was evaporated onto
the surface, which can be performed in UHV. In addition the
clusters stayed fixed to the pits during STM scanning at
room temperature. Without the pits the clusters would be
displaced by the STM tip. Combining STM and TEM measurements we were able to deduce the cluster shape and
prove that the clusters are three-dimensional with their
height comparable to their diameter. The cluster samples prepared with our method are homogeneous on an area of several square millimeters with about 53108 clusters/mm2. This
158
J. Appl. Phys., Vol. 81, No. 1, 1 January 1997
Hövel et al.