|
MORPHOLOGICAL CHANGES TO
InAs ISLANDS BY OVERGROWTH
As described above the growth of highly compressed epilayers in some
heteroepitaxial systems, nanometer scale 3D islands can nucleate on the
surface as a strain relief mechanism at the growth front. If growth is
terminated at this point and the islands are then covered with a larger
bandgap material, the nanometer scale 3D geometrical confinement results
in quasi-zero dimensional electronic levels inside the island. InAs islands
grown in a GaAs matrix are the dominant material system for such quantum
dots (QD’s). These films provided a rich testing ground for the study
of zero dimensional confinement in semiconductors, and hold the potential
for several interesting device applications.
Because the electronic levels of the QD’s depend on the size and shape
of the island, a large body of research has focused on better understanding
and controlling the InAs island morphology. The bulk of structural characterization
studies have examined the surface morphology of the uncapped 3D-islands.
However, research that probes the structure of the QD’s embedded inside
the GaAs matrix indicates that the 3D-island morphology is altered during
GaAs overgrowth. In this section we discuss the use of in-situ scanning
tunneling microscopy (STM) to study how the InAs island morphology changes
as a function of GaAs overgrowth.
Growth and Imaging of InAs Islands as a Function of Overgrowth
The sample preparation and InAs deposition on GaAs is described earlier.
After InAs deposition, the sample is maintained at 495ºC under an As BEP
of 1x10-6 Torr for 50s to allow island ripening which yields larger and
more uniform dots. The GaAs overgrowth was performed at 0.7ML/s (Ga BEP
of 4x10-7 Torr, As BEP of 4x10-6 Torr) in all experiments. The sample
heater was turned off immediately after GaAs deposition for each experiment,
and the sample cooled under a gradually reduced As flux until its temperature
fell below 350ºC.
It is useful to first consider what changes to the surface might be caused
by GaAs deposition. Figure 13 illustrates how the surface profile
of the 3D-island morphology might be changed by GaAs overgrowth. Figures
13-a and 13-b illustrate possible outcomes in the case that the islands
are completely stable. That is, in these cases it is assumed no In-Ga
intermixing occurs, so that the 3D-island morphology is preserved. Figures
13-c and 13-d illustrate the more realistic situation in which cation
intermixing leads to a redistribution of the island material. In this
latter scenario, as overgrowth is initiated the InAs layer becomes an
alloyed InGaAs layer, effectively reducing the lattice mismatch between
epilayer and substrate. Such an alloyed layer would likely possess a 2D
lowest energy surface. In this case, the surface morphology observed after
overgrowth, would depend on how much of the 3D-island volume had redistributed
itself into the 2D region. If the sample were maintained at high temperature
for sufficient time, after a few monolayers of Ga were deposited, a 2D
surface would be expected. But if the sample were quenched during this
change, an unstable, and intermediate 2D-3D surface would be observed.
Such a surface would exhibit a reduced island height, combined with an
enlarged basal area as illustrated in Figures 13-c and 13-d.
InAs Morphology
Figure 14 shows two images of the surface morphology after 2.1ML of InAs
deposition. This coverage corresponds to 0.4ML beyond
the 2D to 3D critical thickness. The growth conditions we have employed
produce relatively large islands which are 9.5 ± 1.5nm tall and have a
base diameter of approximately 33nm. The number density of the islands
is 3 x 1010 cm-2. This morphology of relatively large islands that are
separated by 2D regions is well suited for studying the changes to the
islands after GaAs overgrowth. Figure 14-b shows a high resolution scan
of an individual island. We observe a structure for the uncapped island
that is different from the disc shaped structure commonly reported by
AFM, and supports the structure reported by Hasegawa et al.. This image
shows four facetted planes, which form a pyramidal structure that is elongated
along the (-110). This morphology is strikingly similar to that predicted
by Lee et al., who derived a similar structure from quantitative analysis
of the reflective high-energy electron diffraction (RHEED) pattern.
1.4.3 Initial Stages of Overgrowth
Figure 15 shows four 200 x 200nm STM images. Figure 15-a shows the InAs
surface prior to overgrowth. Figure 15 shows images
taken after 0.7ML, 1.4ML and 2.7ML of GaAs overgrowth. After only 0.7ML
of overgrowth we observe small terraces forming at the base of the island.
This effect is more pronounced after 1.4ML of overgrowth as the island
base has become significantly enlarged, and the taller 3D region of the
island has been reduced in size. By 2.4ML of GaAs coverage, the mean island
height has been reduced from 100Å to 20Å. From this series of images it
can be concluded that extensive cation intermixing is taking place during
GaAs overgrowth. As a result the lowest energy surface morphology has
become 2D. We know this to be true because we observe a complete dissolution
of the 3D morphology after 2ML of deposition if the sample is maintained
at 495º C for 30s. Figure 15 shows that we have cooled the sample fast
enough to allow only partial intermixing. Because these surfaces are not
in equilibrium, these images can not be viewed as snap shots of the growth
front.
Figure 16 shows the surface profile data of representative islands taken
for each surface shown in Figure 15 for both the (110) and
(-110) directions. The height reduction of the 3D islands at these low
coverages, compared to the uncapped island profile, demonstrate that material
is leaving the 3D islands. From the (110) surface profiles, it is clear
that no broadening of the island base accompanies this height reduction.
In contrast, after 2.7ML of GaAs overgrowth, the island base has increased
by nearly a factor of three along the (-110). This anisotropic volume
redistribution points to the specific mechanism in which the In is transferred
away from the 3D islands and into the GaAs capping layer. Cation surface
diffusion coefficients are reported to be 100 times larger along the (-110)
than along the (110) due to the corrugated surface of the 2x4 surface
reconstruction. Because the zincblende lattice has non-symmorphic symmetry,
the bulk diffusion rate is expected to be the same in these two directions.
To summarize our results from these initial stages of GaAs overgrowth,
we have confirmed that extensive In-Ga intermixing occurs, and that the
specific mechanism is In surface diffusion on the 2x4 surface.
1.4.4 More Substantial Overgrowth
Figure 17 shows more 200 x 200nm images taken after additional stages
of GaAs overgrowth; the 2.7ML coverage image is shown
again for comparative purposes. At 5ML the islands continue to be reduced
in height while their basal area increases. At 11 ML, the islands become
taller, and even taller at 21ML. In addition, depressions are observed
in the center of the islands at 10 and 20 ML. Traces of this feature are
also apparent at lower coverage. The depression may be explained by preferential
nucleation of the capping layer in the 2D region.
This has been proposed to be the result of the 3D island locally possessing
a larger lattice constant near to that of InAs, so that nucleation of
the smaller GaAs bond in this region would be energetically unfavorable.
The surface diffusion of In away from this region during and/or after
overgrowth could also account for this depression.
These features are revealed more clearly in the line profile data shown
in Figure 18. These line profiles are first normalized in the 2D region
to 0, and then offset by the amount of GaAs overgrowth. This allows for
meaningful comparison to the uncapped island profiles shown with these
data. A very interesting observation may be reached regarding the change
in height shown by these line scans as the coverage in increased from
5 to 10ML. Because, in the center of the line scan, this height increase
is not due to GaAs nucleation, we conclude that a greater portion of the
3D island has been preserved after 10ML of coverage. That is, thicker
or more rapid overgrowth better preserves the 3D-island morphology. Because
we have observed that a small amount of overgrowth can lead to a complete
loss of the 3D islands, this idea is confirmed by the simple fact that
cross sectional studies of capped islands show that the 3D islands are
indeed preserved inside a GaAs matrix. That is, were this observation
untrue it would be unlikely that anyone had ever observed 3D islands buried
inside the GaAs capping layer. Therefore, the volume redistribution process
is not so rapid as to be independent of the GaAs deposition rate. Faster
overgrowth buries more In beneath the growth front. This In would then
have to undergo bulk diffusion through the zincblende lattice to achieve
volume redistribution. This later intermixing process would occur at a
rate insignificant compared to surface diffusion.
[1][2][3][4][5]

|