SELF-ORGANIZED InAs ISLANDS ON AlAs(001)
Because quantum dots are obtained by strain induced self-organized growth, the choice of material is restricted by the need of both strain and confinement. Despite this material restriction, however, the use of alloys for the dot or barrier material, allows some flexibility and makes it possible to tune the lattice mismatch and achieve different size dots with different barrier heights, enabling band structure engineering, and a corresponding wider range of emission wavelengths. For example, for the (In,Ga)As/GaAs system island formation has been achieved with an In composition varying from 100 to as low as 25%.More recently, several groups have added additional flexibility by using alloys for the barrier layers. For example, in one study the room temperature photoluminescence emission was tuned from 0.8 to 1.15 mm using InAs/(Al,Ga)As quantum dots. Composite quantum dots have also been fabricated in the (In,Ga,Al)As/(Ga,Al)As system.
While the use of alloys conceptually makes it possible to tune the lattice mismatch, it is also true that alloying changes the chemical nature of the interacting surfaces. This raises the question about the role of a changing interface on dot formation. For example, is alloying so simple that changing the chemical nature plays little or no role, leaving tuning of the lattice mismatch as the feature that governs the dot formation?
In this section we discuss on the influence of the chemical composition of the (Al,Ga)As surface on the formation of strain induced three dimensional (3D) InAs islands. Results demonstrate that there are major differences between the InAs/GaAs and the InAs/AlAs systems despite the nearly identical lattice mismatch. These differences have been determined by gradually adding aluminum to the starting GaAs surface. We find that increasing the Al fraction enhances the control of the size and density of the 3D islands. Moreover, the observed difference in island density, size, and growth mode is explained by considering the difference in surface mobility attributable to the change in the chemical nature of the interacting surfaces.
Growth and Imaging of InAs Islands on (Al,Ga) As
To investigate the formation of InAs coherently strained islands on (Al, Ga)As, we prepared GaAs buffer layer first as described earlier. The (Al, Ga)As layer is then deposited by adjusting Al and Ga fluxes to produce a 1.0 monolayer per second (ML/s) growth rate. After the deposition of 50 MLs (~140Å) of this alloy, the substrate temperature is cooled to 500°C while reducing the As beam equivalent pressure (BEP) to keep a 2x4 As-rich surface reconstruction. InAs is then deposited with a growth rate of 0.10 ML/s by successive steps of 3s followed by 17s interruptions. The As BEP used is 6.0x10-6 Torr as measured by an ion gauge, and is kept at this high value during the growth interruptions. Once the desired amount of InAs is deposited, the sample is quenched to room temperature. A low As flux is maintained until the temperature falls below 350°C in order to prevent the formation of a cation-rich surface. The 2D to 3D growth mode transition, as well as, the surface reconstruction are observed by watching the evolution of the RHEED pattern in the [110], [-110] and [100] directions. Even though the information obtained from RHEED is also present in the STM data, RHEED patterns are needed to study self-assembled island formation as they provide real-time information about surface morphology changes, such as the 2D-3D transformation and alloying induced modification of the surface reconstruction. After completion of the growth, the sample is transferred under vacuum to the STM where plane view imaging of the surface on an atomic scale is achieved. Constant current mode STM images presented in this work are filled states images and have been acquired at room temperature with -3V, -4V bias at the sample.
InAs/(AlGa)As Islands
Figure 19 displays three STM images of typical starting surface consisting
of (a) GaAs, (b) Al0.5Ga0.5As and (c) AlAs surfaces. It
is quite clear that the aluminum composition impacts the surface roughness.
The strong reactivity of Al atoms drastically reduces the surface mobility
resulting in a rather rough surface morphology. 
As
can be seen from Figure 20, an analysis of the surface roughness as a
function of aluminum composition shows an exponential behavior. The roughness
represents the mean deviation from the average height and has been evaluated
from several 200 x 200 nm scans for each aluminum composition. However,
large scan STM images show that these surfaces, although rough, exhibit
large terraces even in the case of pure AlAs. All (Al,Ga)As surfaces prepared
show a streaky RHEED pattern indicating a 2x4 surface reconstruction.
Nevertheless, STM images reveal that this 2x4 surface is not very well
ordered when both Al and Ga atoms are present on the surface. In this
case, the surface presents a large density of kinks and missing As dimers
that are attributable to the alloying effect.
When InAs is deposited on the surface, the annealing time following each
0.3 ML shot allows cation intermixing and thus leads to the formation
of an alloyed layer with a drastically reduced surface roughness even
in the case of an Al rich starting surface. STM data taken after the deposition
of a fraction of a monolayer of InAs show a complete smoothing of the
surface. The RHEED pattern changes from a 2x4 to a 1x3 which is characteristic
of an In(Ga, Al)As surface. While depositing InAs, one can follow the
evolution of the surface morphology by observing the RHEED patterns. In
particular, at the onset of the 3D-island formation, the RHEED pattern
exhibits characteristic transmission spots. This information is used to
accurately determine the 2D-3D critical thickness, i.e. the amount of
InAs needed to achieve the 3D growth mode. This critical thickness is
highly sensitive to the growth conditions and especially to the temperature
at which InAs is deposited. When the substrate temperature is increased,
the critical thickness increases in a proportion that is roughly independent
of the (Al,Ga)As composition. This increase in critical thickness is explained
by an enhanced cation intermixing leading to a significantly lower density
of elastic energy. Temperature, however, is not the only parameter that
influences the critical thickness. For example, our experiments demonstrate
that the critical thickness, as defined
above, strongly depends on the Aluminum composition of the starting surface
(despite the same lattice mismatch with InAs). Figure 21 displays the
value of the critical thickness as a function of the Al fraction at a
substrate temperature equal to 500ºC. The critical thickness increases
from 1.75 ML for GaAs to 2.2 MLs for AlAs in an exponential way, much
like the surface roughness. Since the mobility of matter on the surface
and the strain energy have proven to be the driving force for 3-D island
formation, these observations raise the interesting question of whether
the critical thickness increases because of surface roughness, or because
of the reduced mobility of In atoms on the Al-rich surface? To answer
this question, we have conducted experiments where we intentionally enhanced
the starting surface roughness, either using GaAs or AlAs. These experiments
have shown that the initial surface roughness does not change the critical
thickness or the dot density. Hence, this difference in critical thickness
must be due to a higher surface mobility of adatoms on a GaAs surface
compared to the AlAs surface. These observations lead to the conclusion
that the roughening of the starting surface and the larger critical thickness
are each observable consequences of a weaker surface mobility on Al-rich
surfaces.
The resolution of the STM has also allowed us to probe the wetting layer
structure, i.e. its surface morphology and reconstruction.
This is shown in Figure 22 displaying three STM pictures acquired after
the deposition of 2.1 ML of InAs on (a) GaAs, (b) Al0.5Ga0.5As and (c)
AlAs. It is clear that none of these images exhibit a sharp 2x4 surface
reconstruction as expected from a pure InAs layer. The rather distorted
aspect of the surface indicates that alloying is taking place. This results
in the presence of Ga and Al atoms on the surface leading to a hybrid
reconstruction of the InAs 2x4 reconstruction and the GaAs or AlAs c(4x4)
reconstruction produced at that temperature when As is supplied. This
hybrid reconstruction shows some level of ordering, in the sense that
the orientation of the As “dimer rows” still follows the [-110] direction
as with the 2x4 reconstruction. It is worth noting that this is more the
case with samples prepared using a strong Al fraction, suggesting reduced
intermixing with higher Al content.
When the gain in relaxing the elastic energy becomes greater than the
cost in surface energy associated with the creation of 3D features, 3D
islands start to form on the surface. By definition, this is achieved
for an InAs
deposition greater than the critical thickness tc. In order to investigate
the morphology of these islands as a function of the Al composition in
the starting surface, we have conducted a series of experiments for which
the InAs deposition above tc is the same. A series of STM images corresponding
to an InAs amount equal to (tc+0.3) ML is displayed in Figure 23 for different
Al fraction. The density of the islands changes exponentially by almost
an order of magnitude by going from InAs/GaAs to InAs/AlAs. This change
is accompanied by a corresponding decrease in the volume of individual
islands. The average height of the islands goes from 9nm in the case of
InAs/GaAs to 4nm in the case of InAs/AlAs. Wang et al have clearly shown
the correlation between critical thickness, island size, and island density,
pointing out that the higher the critical thickness, the larger the island
number density and the smaller the island size. Our results are in strong
agreement with these predictions. However, in their paper, Wang et al
treat the density as an input parameter changing presumably according
to growth conditions and considering only one system (InAs/GaAs). In
our case the growth conditions are the same but the ratio of material
composition, i.e., Al content, is changing. Physically, our observations
are tied again to the loss of mobility for InAs deposited on an Al-rich
surface. The islanding is being blocked or at least retarded by the lack
of material motion on the surface and thus, when islanding finally occurs
the amount of strain energy is greater (because more material has been
deposited) resulting in a larger density of smaller islands.
We also investigated the effect of the total InAs deposition on 3D-island
density as a function of the Al fraction of the starting
surface. For example, in Figure 24 we plot the evolution of the 3D-island
density as a function of the Al fraction, for two values of the InAs deposition:
(tc+0.1) ML and (tc+0.3) ML. The results are extracted from several 1mm-by-1mm
STM scans taken from different locations on the sample. As already reported
in the literature, the InAs/GaAs data exhibits a significant change in
the island density with deposition. This is not the case, however, for
InAs/AlAs for which the density remains fixed at the significantly higher
value of 1.4x1011cm-2. The dashed lines, which are a linear regression
to the data points of the two series of experiments, indicate that the
density dependence on the deposition decreases monotonically when the
starting surface changes from GaAs to AlAs.
These data (Figure 23) also support the density/size behavior predicted by Wang et al, namely that the size increases as the density decreases. This behavior is further confirmed by the variation in density with deposition reported in the InGaAs/GaAs system where even bigger dots and an extremely low density at the onset of the 2D-3D transition are observed. The reason for a coverage-dependent density is likely to be related to a very high ripening rate resulting in the vanishing of the smallest islands present at the onset of 2D-3D transition. Our data thus indicate that the ripening rate is very low for InAs/AlAs islands with respect to InAs/GaAs islands, which again, is consistent with the reduction of the overall surface kinetics when growth occurs in an Al-rich environment.
InAs/AlAs Islands
We have used the STM to study the details of the evolution of the surface
at the onset of 3D InAs island formation on AlAs and also
for further InAs deposition. The results are summarized in figure 25 displaying
four STM images obtained for a surface preparation corresponding to 2.1
ML (Fig.25-a), 2.2 ML (Fig.25-b), 2.85 ML (Fig.25-c) and 3.9 ML (Fig.25-d).
Images (a) and (b) show the surface morphology before islanding and just
after formation of the 3D islands respectively. The difference in morphology
between these two surfaces has been confirmed by RHEED exhibiting a streaky
pattern for 2.1 ML of InAs whereas the pattern turns to a transmission
pattern when 2.2 ML of InAs is deposited. A careful analysis of these
two pictures shows that the high density of two-dimensional islands in
image (a) corresponds roughly to that of the 3D islands in image (b).
This suggests that the 2D islands must reach a given size to give rise
to a 3D island. For further deposition, the density remains equal to that
of the 3D features observed in image (b). This observation agrees quite
well with the model of Priester and Lannoo who predicted the 2D-3D transition
to be initiated by 2D “platelets” with size and density defining those
of the 3D features. In the case of InAs/AlAs it is however clear that
these 2D structures are extremely small and contain just a few atoms leading
to the formation of ultra-small 3D features with a height of a few monolayers
(Fig.25-b). When additional InAs is deposited the density holds and thus
the size of the islands increases. This happens in two steps, first the
height increases faster than the base but when the deposition is around
2.7 ML the island growth begins to scale, i.e. (height and base grow proportionally
maintaining a constant aspect ratio). For coverage greater than 3.0 ML
big islands can be seen. One of them is displayed in Figure 25 (d). Its
height is 21 nm and its base length is 75 nm so that its aspect ratio
is the same as that of the other islands. The density of these “big” islands
stays low (5x108cm-2) even when the coverage is brought to 3.9 ML.
As suggested by Figure 24 and by the STM images of Figure 25, the InAs
deposition does not provide a means to control the island density in the
InAs/AlAs system. The density is plotted in Figure 8 as a function of
the amount of InAs deposited and is found to be constant in the range
investigated. However, it is possible to accurately control the density
of islands by either changing the growth temperature or by annealing the
sample at an elevated temperature for a variable amount of time. Changing
the growth temperature
toward cooler temperatures allows us to significantly increase the number
density. For example, it is found to change by an order-of-magnitude (from
2x1011cm-2 to 2x1012cm-2 ) by reducing the growth temperature to 450ºC.
This phenomenon is accompanied by a reduction in the size of the islands.
At 400ºC, we found the islands to occupy most of the surface area so that
most of them were touching each other. On the other hand, providing the
surface the time needed for matter to move, leads to the well known Ostwald
ripening effect and therefore to a significant decrease in the 3D island
number density. The density again varies by an order-of-magnitude between
no annealing (2x1011cm-2) and 10 minute annealing (2.5x1010cm-2) as can
be seen in Figure 26. The behavior of the InAs/AlAs island density suggests
that this system is very well suited for device application and electronic
engineering. The density of islands can be controlled over two orders-of-magnitude
and it is also possible to independently control the size. While low temperature
deposition leads to a high density of small islands, (important to achieve
large confinement energy and inter-island carrier tunneling), the annealing
time provides a means to reduce this density to a regime in which a single
dot might be isolated.
Keeping the islands at high temperature (500ºC) for an extended period
of time has some additional consequences. Figure 27 
summarizes
the main effects of annealing the samples and shows the evolution of the
height and the base length of the 3D-islands grown on GaAs and AlAs surfaces.
The fact that the average size of the islands changes with time clearly
indicates that those islands do not represent a stable state but that
they are rather involved in an evolutionary process dictated by the combination
of Ostwald ripening and cation intermixing. The physical origin of the
ripening in the case of self-assembled strained 3D-islands is still unclear
as several tentative explanations can be found in the existing literature.
The classical ripening called Ostwald ripening refers to the fact that
the smallest clusters posses lower surface binding energies relative to
the largest clusters due to the greater curvature. It has to be noted
that this interpretation does not involve any strain consideration. More
recent studies have pointed out that (i) the energy gain associated with
strain relaxation scales with the volume of individual islands V, while
the cost associated with surface energy scales with V2/3 rendering favorable
the evolution toward larger islands, (ii) the total surface energy of
a system is decreased when small particles combine to form larger islands
with no loss of mass, (iii) ripening involves strain dependent kinetics
on the surface as 3D-islands act as local stressors. In any case, ripening
results in a decrease in island density and an increase in the average
size of the islands. This agrees very well with the data plotted in Figure
27 for annealing times typically less than 2 minutes. However, it is clear
from the data that classic ripening alone can not account for the evolution
of the average height and base-length, which slowly decreases for annealing
times greater than 2 minutes. Instead, with a decreasing density, ripening
becomes less efficient since the decreasing density leads to a greater
spatial separation between islands and a corresponding weaker exchange
of material or greater isolation. However, this does not explain a decreasing
size after 2 minutes. On the other hand, high temperature annealing results
in intermixing in the element III species such as In and Ga or In and
Al. In this case In atoms go deep into the substrate as a strain relief
mechanism. As a result, elastic energy is no longer concentrated near
the surface of the sample and the driving force for the 3-D islanding
process progressively vanishes, resulting in a decrease in the number
of islands and in the size of each island. In fact, for long annealing
times we observe that 3D-islands begin to dissolve within the wetting
layer. The dissolving process results in the presence of monolayer thick
plateaus surrounding each island. The same behavior is observed for InAs/GaAs
and InAs/AlAs 3D-islands. Due to the presence of the plateaus, the height
decreases much faster than the base-length which remains always larger
than its original value. Because the density is always decreasing, the
total volume of islands after several minutes of annealing is drastically
reduced compared to the initial volume. This agrees very well with the
model of a reduced density of elastic energy close to the surface. It
has to be noted that at the growth temperature investigated (495-500 ºC),
no significant In desorption is expected. Concerning the homogeneity of
the islands, it is unclear if annealing provides a means to achieve a
more homogeneous population of islands especially for InAs/GaAs islands.
In the case of InAs/AlAs islands, long annealing times do provide a slightly
lower standard deviation value. This is another observation that is consistent
with the lower surface mobility for this system since the InAs/GaAs island
population reaches its optimum homogeneity almost instantaneously.
Investigation of the integrated volume of the islands when the InAs deposition
is varied also adds insight into the growth mechanisms.
The results are shown in Figure 28 in which our experimental results obtained
with the InAs/AlAs system are compared with measurements carried out by
Joyce et al on InAs/GaAs 3D self-assembled islands. The dotted line shows
the InAs volume deposited after the critical thickness for InAs/AlAs islands,
while the solid line is for the total InAs volume deposited. Despite evidence
of the presence of facets, we have assumed for simplicity, a conic shape
with an elliptic base, accounting for the shape anisotropy, to compute
the volume of each island. Island sizes (height, base length and width)
have been measured using line profiles on typically a hundred of islands
for each data point. The error bars presented in figure 28 account for
differences found between different 200 x 200 nm STM scans performed at
different locations on the sample surface. As can be seen from the two
set of data, InAs islands formed on AlAs surfaces grow in a very different
manner compared to those formed on GaAs surfaces. The volume of InAs islands
formed on GaAs increases more rapidly than the total amount of InAs deposited
which leads to the suggestion that material is taken from the buried GaAs
layer. This observation accounts for the very strong intermixing observed
in the case of InAs/GaAs strain induced 3D islands. On the other hand,
the behavior of InAs/AlAs islands is markedly different as evidenced by
the experimental data points plotted in Figure 28. The islands form at
first by incorporating not only the material deposited, but also material
from the wetting layer. However, after 2.7 ML, the growth continues by
only consuming the deposited material, maintaining a wetting layer of
1 ML. This change in slope, which occurs around 2.7 ML, corresponds to
the change in slope in the evolution of the aspect ratio. Below this threshold,
the height of the islands grows much faster than the base length resulting
in steeper islands. Once above this threshold, the islands dimensions
scale equally. In this sense, one can conclude that the island growth
reaches an equilibrium state in which the shape is stable, the islands
growing with the additional material, in a proportional way. In regard
to these experimental facts, one has a clear picture of the growth mode
of strain induced InAs/AlAs self-assembled islands which can be classified
to be Stranski-Krastanov as opposed to islands grown on a GaAs surface.
