MORPHOLOGY OF SELF-ORGANIZED InAs ISLANDS
ON InP
As already pointed out for the case of InAs on AlAs, one
of the attractive features of MBE growth is that it is possible to engineer
the bandgap by choice of material. For example, for device applications,
it is of interest to investigate InAs 3-D island formation on different
starting substrates. InP is particularly interesting since it would
permit the shifting of the emission wavelength to the technologically
important 1.3 to 1.5µm range when using the binary InAs as the low bandgap
material. In addition to the different emission range, coherent 3-D
InAs islands on InP will result in a significantly smaller carrier effective
mass and a correspondingly larger intersubband energy level spacing.
In addition to bandgap engineering, by choosing an InP
substrate, the self-assembled InAs/InP system provides us control over
the quantum nanostructures, i.e. either quantum wire or quantum dot.
Direct deposition of several MLs of InAs on the substrate induces self-assembled
nanowires. The nanowires can also be produced by exchange of substrate
phosphorous with incoming arsenic. The nanowires, either by direct deposition
or by exchange, transform into dots when they are annealed under no
flux. The most significant difference between the InAs/InP and the
InAs/AlGaAs system is the observations of stable InAs quantum wires
and their evolution into quantum dots. This morphological change enables
us to select either quantum wire or quantum dot in the InAs/InP system.
Growth
InAs coherently strained islands on InP were grown in
the same MBE chamber described in 1.2.1. We used n-doped InP wafers
previously degassed at 250°C in a separate chamber. After loading a
commercial n-type, planar InP (001) wafer into the MBE chamber, the
oxide layer on the wafer is removed by annealing the wafer above 480°C
under a cracked (950°C) phosphorous (P2) beam equivalent pressure (BEP)
of 10–5 Torr. The resultant surface yields 2´4 RHEED patterns, similar
to those of 2´4 patterns of a GaAs (001) surface as in previous sections.
Once the oxide is removed, a 0.5mm thick buffer layer
of InP is grown on the substrate at 470°C. To have a smooth InP (001)-2´4
surface, the substrate is annealed at 500 °C under a P2 pressure of
1´10–5 Torr for 15 minutes and then cooled to 480°C. At 480°C, the
valve for the P2 flux is closed and the residual phosphorous is pumped
out for about 3 minutes. During this period the RHEED continuously indicates
a stable 2x4 surface reconstruction that is independent of the P2 overpressure.
After preparing an InP surface, the sample is rapidly cooled down below
250°C and transferred to the STM. STM images are taken for filled states
(-3V on the sample) with tunneling current around 100 pA and confirm
a flat and smooth surface.
On this smooth surface of InP, we deposit a desired amount
of InAs with growth rate ~0.3ML/s and As4 flux around 1´10–5 Torr. The
sample is, either immediately or after some delay for annealing, cooled
down below 250°C and transferred to the STM chamber through the ultra
high vacuum modutrack. STM images are taken for filled states (-3V
on the sample) with tunneling current around 100 pA. Observation of
RHEED patterns during the growth indicates that the surface reconstruction
remains the same 2x4. Upon closing the arsenic valve, however, the RHEED
pattern immediately changed to 4x2, indicating an In rich surface. Nanowires
were also produced by simple application of high As4 flux around 1´10–5Torr
on InP surface without indium deposition, utilizing the intermixing
or exchange of V-elements.
InP Starting Surface and Nanowires
The STM images of the starting surface of InP (001), just
before deposition of InAs, are shown in Figure 29-(a) and (b). The
images
show terraces, vacancy islands, monolayer (ML) high steps and 2x4 surface
reconstruction with dimer rows, the inset in Figure 29-(b), which are
along the [
10]
direction. The STM images indicate a beautiful InP (001) surface even
after annealing without phosphorous for 15 minutes although there are
vacancy islands. The shape of the vacancy island is anisotropic, elongated
more along the dimer row direction of the 2x4 surface reconstruction.
No 4x2 surface reconstruction on the InP (001) surface is observed,
even after 15 minutes of annealing without a phosphorous flux at 480°C,
either by RHEED or by STM.
Application of a high arsenic flux at 480°C on the InP
(001)-2x4 surface produces a high density of nanowires as shown in Figure
30-(a). The substrate temperature and the As BEP were 480°C and 10-5
Torr respectively. The dimer rows on the top surface of nanowires are
clearly visible as seen in Figure 30-(b), indicating the arsenic-terminated
2x4 surface reconstruction. This STM picture was taken after the substrate
was annealed under the growth conditions for 8 minutes and then quenched
under enough arsenic flux to support the 2x4 surface reconstruction
as monitored by RHEED. The necessary time period of arsenic application
is about 3 minutes to form nanowires under the high arsenic BEP, indicating
that the nanowires are stable when the arsenic flux is high enough to
maintain a 2x4
surface
reconstruction. The typical wire length is over 1 mm and the density
of the nanowires is about 50/mm. The average height of the nanowires,
along with the standard deviation, is 1.9±0.5nm. The width and average
separation between nanowires are 14.3±1.9nm and 3.7±1.6nm respectively
while the shape is a trapezoid as shown in the line profile along the
[110] direction.
InAs growth on the InP (001)-2x4 surface at 480°C produces
a high density of nanowires, similar to nanowires produced by the exchange
process described above. When the nanowires are cooled down under no
arsenic flux, the surface reconstruction changes to 4x2 from 2x4 still
preserving the nanowire morphology as shown in Figure 31-(a). The dimer
rows on the top surface of nanowires are clearly visible as seen in
Figure 31-(b), indicating the indium-terminated 4x2 surface reconstruction.
Imaging of InAs 3-D Islands
While the nanowires were stable under high As flux, this
was not the case for a low flux. For example, when the needle valve
of the arsenic cell was closed, the As-flux dropped below 10-9 Torr
in several seconds. This resulted in an In-rich 4x2 surface reconstruction
for the nanowires as in Figure 31-(a) and (b). As expected, the RHEED
pattern changed immediately from 2x4 to 4x2 upon closing the arsenic
valve. Continuing this condition of no flux at the same temperature
transforms the nanowires into dots as shown in Figure 31-(c) and (d)
as well as in Figure 30-(c) and (d). A residual of the nanowires, now
2D islands (only 1ML high), are also visible in the Figure 30-(c) and
31-(c). Interestingly, the dots are expanded more along the [110] direction,
the dimer row direction of the 4x2 surface reconstruction. The 4x2 surface
reconstruction on the rectangular flat-top of the dot is clearly visible
in Figure 30-(d) and 31-(d). The same surface reconstruction is also
observed on the wetting layer between dots as well as on the residual
of nanowires.
The density of the transformed dots in Figure 30 along
with the standard deviation is 3.6±1.3x109/cm2. The average height of
the dots is 4.7±0.7nm. The dots have a larger width along the dimer
row direction of the 4x2 surface reconstruction as shown in Figure 30-(d)
and the line profile below. The base and top widths are 58.1±8.3nm and
32.1±5.1nm respectively along the [
10]
direction while 75.5±8.0nm and 38.4±12.4nm along the [110] direction.
The ratio of length/width has changed from ~70 (nanowire) to 0.77 (dot),
where the width is measured along the [110] direction, resulting in
a rectangular based, truncated pyramid elongated along the dimer row
direction. Also the height has increased by 2.5 times over the nanowires.
The morphology change from wire to dot is correlated to
the change in surface reconstruction from 2x4 to 4x2. The change from
wires to dots, only and always occurred, with a change in surface reconstruction.
One plausible explanation for this behavior is as follows: Initially
the morphological change from film to wires is driven by the gain in
elastic relaxation energy of the coherent islands which overcompensates
the energy cost due to the increase of surface area. This would be reasonable
if the energy cost due to the increase of surface area for wires is
less than that for dots and if a 4x2 surface is a higher energy cost
than a 2x4 surface [117]. In this case, four possible stable morphologies
would be 2x4 wires oriented [
10],
2x4 dots, 4x2 wires along [110], and 4x2 dots. However, in our experiment,
the change in surface reconstruction from 2x4 to 4x2 produces 4x2 wires
along [
10].
This could be a very high unstable energy configuration that falls to
the next lowest stable morphology, that of 4x2 dots.
[1][2][3][4][5]