The role of vicinal silicon surfaces in the formation of epitaxial twins
during the growth of III-V thin films
G. A. Devenyi, S. Y. Woo, S. Ghanad-Tavakoli, R. A. Hughes, R. N. Kleiman et al.

Citation: J. Appl. Phys. 110, 124316 (2011); doi: 10.1063/1.3671022
View online: http://dx.doi.org/10.1063/1.3671022
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v110/i12
Published by the American Institute of Physics.

Related Articles
The observation of unoccupied quantum-well states in Bi thin film grown on Si(111) by two-photon photoemission
spectroscopy
Appl. Phys. Lett. 99, 243101 (2011)
On the theoretical description of nucleation in confined space
AIP Advances 1, 042160 (2011)
Threading-dislocation blocking by stacking faults formed in an undoped GaN layer on a patterned sapphire
substrate
Appl. Phys. Lett. 99, 211901 (2011)
Effect of anisotropic strain on the charge ordering behavior in Bi0.4Ca0.6MnO3 films
Appl. Phys. Lett. 99, 191914 (2011)
Negative ions: The overlooked species in thin film growth by pulsed laser deposition
Appl. Phys. Lett. 99, 191501 (2011)

Additional information on J. Appl. Phys.
Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors

The role of vicinal silicon surfaces in the formation of epitaxial twins during
the growth of III-V thin films
G. A. Devenyi,1,2,3,a) S. Y. Woo,3,4 S. Ghanad-Tavakoli,2 R. A. Hughes,3,b)
R. N. Kleiman,1,2,3 G. A. Botton,3,4 and J. S. Preston1,2,3
1Department of Engineering Physics, McMaster University, Hamilton, Ontario L8S 4L7, Canada
2Centre for Emerging Device Technologies, McMaster University, Hamilton, Ontario L8S 4L7, Canada
3Brockhouse Institute for Material Research, McMaster University, Hamilton, Ontario L8S 4M1, Canada
4Department of Materials Science and Engineering, McMaster University, Hamilton, Ontario L8S 4L7,
Canada and Canadian Centre for Electron Microscopy, McMaster University, Hamilton,
Ontario L8S 4M1, Canada
(Received 30 September 2011; accepted 15 November 2011; published online 22 December 2011)
We examine the role of vicinal surface steps in the formation and propagation of twins during the
growth of epitaxial III-V thin films (GaAs, InP, GaSb, AlSb) on silicon substrates. This is achieved
through the combined use of two-dimensional X-ray diffraction and conventional transmission
electron microscopy techniques, which allow for both a macro and nano/micro characterization of
the material systems. Observed is a systematic suppression of twins formed opposite to the tilt
direction of vicinal substrates through a process of step-flow overgrowth of nucleated twins, and an
enhancement of twins toward the tilt direction when the fastest growth planes are aligned with the
step-flow. These results indicate a probable path to the enhancement of the electronic mobility of
lateral devices based on III-V semiconductors on silicon. VC 2011 American Institute of Physics.
[doi:10.1063/1.3671022]
INTRODUCTION
Anti-phase boundaries (APBs) have long been identified
as highly detrimental to the electro-optic properties of epi-
taxial thin films.1 Extensive research has been performed in
the use of vicinal substrates and other means2 to control the
formation and propagation of APBs during the growth of
epitaxial thin films. A far more common planar defect during
growth is epitaxial twins.3 Twins are regions of mirrored
stacking sequence in the closest packed plane of {111}
within a crystal with zinc-blende structure, where each con-
secutive layer is shifted after the first bounding stacking
fault.4 These twin defects, often observed through transmis-
sion electron microscopy (TEM), have been attributed to a
variety of causes, including APBs, threading dislocations,
and simply stacking faults.5–13 X-ray diffraction single peak
analysis, which only examines X-ray peaks perpendicular to
the substrate surface, can completely disregard the additional
peaks due to the presence of twins, leading to their misattri-
bution. Similarly, the sole use of conventional TEM has the
disadvantage of being a local measurement, and not com-
pletely representative of the bulk. Further work has defini-
tively noted the key features of twins and provided guides to
their identification.3
The twin has been considered a low-impact defect, due
to its coherent interface and lack of dangling bonds, causing
only weak scattering due to the minor disruption in the crys-
tal symmetry. However, this view of twins considers them
isolated in the material, ignoring the fact that in epitaxial
films, twins are likely to intersect during growth, as the inter-
section of twins of different habit planes will lead to unsatis-
fied bonds. For example, first-order twins formed on
opposing habit planes can terminate from mutual intersection
along a {114} plane, which is highly incoherent, resulting in
unsatisfied bonds and trapped charge.14 Furthermore, twins
that intersect from adjacent habit planes will also form inco-
herent interfaces on high-order planes. Finally, second- and
higher-order twinning interfaces are expected to create dan-
gling bonds and will result in further unsatisfied bonds
between the twinned regions and epitaxial film.
Using the combined techniques of two-dimensional X-ray
Diffraction (2DXRD) and conventional TEM, this investiga-
tion has demonstrated a strong asymmetric suppression of
twins due to the formation of atomic steps on vicinal substrates,
previously observed for only GaAs.15–17 This combination of
techniques allows macro and nano/micro characterization that
combines to provide a complete analysis of a material system.
Twin suppression has been observed in GaAs, InP, GaSb, and
AlSb semiconductors grown on vicinal silicon substrates under
varying growth conditions. Through the integration of a quanti-
tative reciprocal space mapping from 2DXRD and real-space
imaging techniques from conventional TEM, we can under-
stand the structural defects globally as well as their spatial dis-
tribution and propagation locally. From this, we propose a
comprehensive mechanism for the suppression, which is the
asymmetric step-flow overgrowth of twins on vicinal sub-
strates, applicable to all of the III-V material systems exam-
ined. Given the detrimental nature of intersection of twins to
electrical properties of thin films, this suppression phenomenon
has the possibility of greatly improving the electro-optic prop-
erties of thin film III-V semiconductors grown on Si.
a)Electronic mail: devenyga@mcmaster.ca.
b)Present address: Department of Mechanical Engineering, Temple Univer-
sity, Philadelphia, PA 19122, USA.
0021-8979/2011/110(12)/124316/7/$30.00 VC 2011 American Institute of Physics110, 124316-1
JOURNAL OF APPLIED PHYSICS 110, 124316 (2011)

EXPERIMENTAL
Semiconductor epilayers (GaAs, InP, GaSb, and AlSb)
were deposited on nominal (001)-oriented (60.5) and vici-
nal Si substrates (offcut 4.7 (60.25) toward [110]) using
a SVT Associates molecular beam epitaxy (MBE) system.
As-received epi-ready wafers were cleaned for 1 min in a
4% HF in de-ionized (DI) water dip followed by a 30 s DI
rinse immediately prior to their insertion into the MBE
load-lock. Before film deposition, both the nominal and vic-
inal Si(001) substrates underwent a 15 min degassing proce-
dure at 350 C followed by a thermal treatment at 800 C
for up to 5 min, in order to reconstruct the Si surface into
single domain terraces.18–21 A small number of single steps
are expected to remain on vicinal substrates, a higher num-
ber on nominal substrates because of the larger terrace
length. Growth conditions followed established proto-
cols13,22,23 and yielded comparable rocking curve full-width
half-maximum for the [004] reflection using double crystal
X-ray diffraction. AlSb epilayers were grown to a thickness
of 550 nm with a 20 nm GaSb capping layer to avoid
oxidation. GaAs epilayers were grown to a thickness of
600 nm and GaSb to a thickness of 500 nm. InP samples
were grown at 470 C with a V/III flux ratio of 2 at a
growth rate of 1 lm/h, resulting in a thickness of 600 nm.
Double crystal X-ray and TEM data also revealed that all
films are fully relaxed by a network of interfacial misfit dis-
locations.24 GaSb samples were grown in the presence of a
5 nm AlSb buffer layer, as prescribed by Akahane et al. 22
Stereographic pole figures were generated for each sam-
ple using 2DXRD techniques. A Bruker SMART 6000 CCD
detector on a Bruker 3-circle D8 goniometer (Bruker AXS
Inc., Madison, WI) with a Rigaku RU-200 rotating anode X-
ray generator (Rigaku MSc, The Woodlands, TX) and
parallel-focusing monochromator optics was used for the
data collection. Scans were taken with the detector centered
on the (111) 2h of the material of interest and the sample
rotated through 360 in 0.5 increments about the surface
normal of the sample. A 1D integration of all frames was
used to determine the combined width of the (111) peaks
using MAX3D software (McMaster University).25 The peak
width was then used to integrate (111) reflections from all
frames, including a background and absorption correction
for the corresponding material with GADDS (Bruker-AXS)
software, resulting in a pole figure. Pole intensities were
obtained from pole figures using a circular integration cursor
with a 10 pixel radius, which was centered on the pole so as
to maximize the total intensity captured. All pole intensities
were corrected for structure factor and frame exposure times.
For each sample two {110} TEM cross-sections were
prepared, one parallel to the [110] miscut direction and the
other perpendicular. The specimens were prepared by the
standard procedure of mechanical polishing, dimpling, and
ion-milling (4 keV Ar-ions at an incident angle of64 using
a liquid nitrogen cold stage for InP) until perforation. Crys-
tallographic information of the epitaxial layer was obtained
using diffraction contrast imaging with a Philips CM12 con-
ventional transmission electron microscope (TEM) operated
at 120 kV and equipped with a LaB6 filament. In addition,
electron diffraction analysis was performed using selected
area electron diffraction (SAD).
RESULTS
Figure 1 shows the {111} pole figures generated from
2DXRD data. The pole figures presented are a stereographic
projection of the (111) X-ray reflections of all orientations of
the material of interest, oriented with the top of the figure
corresponding to the [110] direction of the substrate. The
pole positions indicate that each of the III-V materials have
the same symmetry and have variations of one orientation
relationship with the Si substrate. For each of the III-V films,
the pole figures indicate a dominant [100]-orientation, as
well as four weaker orientations associated with twins hav-
ing (111), (111) (111), and (111) habit planes, as indicated
by a single shared pole (and hence a shared plane) between
the bulk film and each twin variant. A simulated pole figure,
indicating the origin of each pole is labeled and given in
Fig. 2. The experimental pole figures differ from the simula-
tion due to instrumental broadening of poles, and in that the
AlSb and InP pole figures exhibit second-order twinning,
which is not included in the figure in an effort to maintain
clarity. The outermost poles are partially visible in the exper-
imental pole figures due to limitations of the measurement
range of samples in reflection, combined with substrate tilt
tolerance. The background of Figs. 1(a) and 1(b) is higher
FIG. 1. (Color online) Stereographic
{111} pole figures generated from
2DXRD show a bulk (100) phase plus
four twinned variants as identified in
Fig. 2. Twin variant intensity is asym-
metric for vicinal substrates.
124316-2 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)

due to the relatively weaker X-ray scattering from GaAs. For
films grown on vicinal substrates (Figs. 1(b), 1(d), 1(f), 1(h))
the center of symmetry of the pole figure is shifted away
from the tilt direction, i.e., towards the step edge. Figure 3
shows the corrected measured and normalized intensity of
the unique twin poles, as labeled in Fig. 2. Measured inten-
sities are normalized to the maximum value of pole intensity
for each pole figure. On nominal substrates, pole intensities
are equal within experimental uncertainty and hence the twin
volume fraction is equal for all four h111i twin plane direc-
tions. Films grown on vicinal substrates, show a 50-75%
reduction in the volume fraction of the twin forming opposite
to the step direction of the (111) habit plane (Pole 3). The
overall intensity of poles does not increase on vicinal sub-
strates, as seen in the corrected intensity plot of Fig. 3(a),
indicating no overall increase in twinning due to vicinal sub-
strates. Conventional cross-sectional TEM measurements
indicate that microtwins are visible for all four of the III-V
material systems studied. Selected area electron diffraction
(SAD) patterns of areas containing microtwins give rise to
additional twin reflections in the form of streaks or spots in
positions mirrored across the twin plane direction. Analysis
of the SAD patterns indicates that the twin variants in the
GaAs epilayers are highlighted in strong contrast with dark-
field (DF) images formed by isolating only the twin reflec-
tion of the dominant variant in the diffraction pattern
(Figs. 4(a) and 4(d)). In comparison to their complementary
DF images, where no particular twin variant is highlighted
(Figs. 4(b) and 4(e), respectively), the preferential twinning
direction is clearly present in GaAs grown on vicinal
substrates. This is also true for GaSb grown on vicinal sub-
strates, as shown in Fig. 4(f). The asymmetry in twinning
directions for vicinal substrates is evident when examining
their spatial distribution, as well as propagation into the epi-
layer with conventional TEM. The preferential twinning
direction along (111) planes away from the step edge (Pole 1)
induces microtwins that propagate deep into much of the epi-
layer thickness, as indicated by microtwins with bright con-
trast in the DF image of GaSb and GaAs shown in Figs. 4(f)
and 4(d), respectively. The intrinsic asymmetry to a vicinal
substrate becomes very obvious when viewing perpendicular
to the tilt direction of [110]. The two edge-on {111} twin
habit planes, namely (111) and (111) now lie 59.4
(54.7 þ offcut angle) and 50 (54.7  offcut angle) from
the interface, respectively. The reduction in peak intensity of
FIG. 2. Simulated {111} pole figure for a cubic III-V semiconductor film
deposited on a nominal silicon substrate. The pole figure containing poles
from the dominant [100] film orientation and from the four primary twins
along their associated habit plane. The poles associated with each orientation
are labeled with unique markers for use in the intensity measurements
(Fig. 3), and labels on the edge of the pole figure indicate absolute crystal
directions of the silicon substrate. All poles from Fig. 1 are accounted for,
although second-order twin poles are omitted for clarity.
FIG. 3. (a) Corrected (via structure factor and exposure time) and (b) normalized intensity of twin poles (as labeled in Fig. 2 above). The normalized X-ray in-
tensity clearly shows a strong reduction in Pole 3 for all vicinal substrates.
124316-3 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)

Pole 3 from 2DXRD can now be explained by the TEM
image of the representative variant, shown in Fig. 5, which
propagates 25 nm from the interface and is small in width
as visible in cross-section. In addition, the in situ RHEED
pattern showed a transition from a spotty to a streaky pattern
after nearly the same film thickness had been grown, indicat-
ing a change in surface morphology.
DISCUSSION
Island nucleation followed by coalescence (i.e., the
Volmer-Weber growth mode) is frequently observed for
III-V thin films3,22,27,28 and is, in fact, anticipated for the
growth parameters used in our film growths. Both substrate
strain and interfacial energy play a strong role in determining
the size of the initial islands during nucleation. After their
initial formation, subsequent island growth occurs on the
{111} planes, creating pyramidal structures, which grow
both laterally and in height. For III-V systems, stacking
along the h111i directions consists of alternating layers of
group III- and group V-terminated surfaces, where two of
the exposed surfaces are (111)A with Ga triply bonded, and
the other two are (111)B with As triply bonded. These two
types of surfaces assemble at different rates. Such a growth
mode can continue indefinitely, but often transforms into a
layer-by-layer growth once some critical thickness is
achieved.29 Epitaxial twins occur when atoms stacked on a
{111} plane shift from their designated positions during
growth, resulting in a stacking fault, characterized by a 180
rotation in bond directionality about the plane normal.
Atoms bonded to the next layer are likely to follow the stack-
ing order defined by their nearest and next-nearest neighbor
and continue the stacking sequence set out after the stacking
fault, creating a rotation twin of the first region. A subse-
quent stacking fault will then rotate the next layer back to its
original orientation, bounding the twinned region. The stack-
ing fault energy (SFE) for a system is the measure of how
costly it is for a single crystal plane to be misordered from
its expected stacking sequence.
The total volume (V) of all twins in a film is given by
Eq. (1) and schematically shown in Fig. 6:
FIG. 4. Conventional TEM images of
the [110] cross-section, (a) GaAs Nomi-
nal DF image of variants with (111) twin
habit plane; (b) GaAs Nominal DF
image of the same area as (a); (c) GaSb
Nominal BF image; (d) GaAs Vicinal
DF image of variants with (111) twin
habit plane; (e) GaAs Vicinal DF image
of the same area as (d); (f) GaSb Vicinal
with the preferential twinning direction
of (111) away from step edge (Pole 1),
as indicated by microtwins with bright
contrast in this DF image; step edge
direction is toward the right for all vici-
nal images.
FIG. 5. Conventional TEM DF image of the [110] cross-section of GaSb
Vicinal, depicting the nanotwins close to the interface, of the twinning direc-
tion of (111) toward the step edges (Pole 3). The black “n”-shaped loops in
this DF image formed with a superlattice reflection are APBs, as described
by Woo et al. (Ref. 26). Step edge direction is toward the right of the image.
FIG. 6. Two twins on a nominal surface, embedded within a film. Twin
dimensions with widths (w1, w2) and length (L) are labeled.
124316-4 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)

V ffi
XN
1
Lw1w2 sin 54:74
ð Þ: (1)
The first factor impacting the volume of twins in an epilayer
is the number of twins, N. This parameter is driven primarily
by the SFE but can also be influenced by the number of
nucleation sites. Atomic registration errors on {111} growth
planes are expected to occur fairly frequently, as their proba-
bility is inversely proportional to the SFE, and SFE is only a
fraction of the average thermal energy of an atom during
growth.3 The formation energy of stacking faults is critical
to the formation of microtwins, this is to say that when this
energy is lower, the probability of a stacking fault, and hence
twin formation, increases.30 Higher assembly rates on the
fault plane can also lock in stacking faults that could other-
wise be corrected in order to minimize energetics. The sec-
ond factor, L, is the vertical length that the twins propagate
through the epilayer, a value that generally is the full thick-
ness of the epilayer. The last two parameters impacting twin
volume are the width in (w1) and perpendicular to (w2) the
fast-growth direction. The width of the twin in the growth
direction is determined by the interplay between the SFE and
the assembly rate of that plane. The twin width perpendicular
to the growth plane is determined by the size of the island in
that dimension, which is influenced both by the chemistry
and misfit of the epilayer-substrate interaction.
Given the role the above parameters play, it is now possi-
ble to account for the differences between the various III-V
semiconductor films grown on nominal (001)-oriented Si sub-
strates. The most obvious difference is the mean volume of
twins, with InP standing out as having the largest value. This
can mainly be attributed to the low value of 17 meV/atom for
its reduced SFE (the energy per atom in a fault plane),31 as
compared to the larger values of 47 and 53 meV/atom exhib-
ited by GaAs and GaSb, respectively.31 It is noted that the
SFE for AlSb has not been reported but is expected to be simi-
lar to that of GaSb due its similar ionicity.32 The higher mean
value exhibited by AlSb is due to its higher nucleation den-
sity,22 which leads to a greater number of islands where each
has the possibility of forming twins.
In all III-V films grown on vicinal substrates, there is a
50-75% reduction in the volume of twins contributing to
Pole 3. This corresponds to a substantial reduction in twin
formation opposite to the tilt direction. GaAs, however, is
unusual in that it exhibits a strong reduction in twins for
directions perpendicular to the step direction (Pole 2 and 4),
and an increase in twins toward the tilt direction (Pole 1 
Pole 3). The TEM studies of Xie et al.16 on GaAs epilayers
grown on vicinal Si substrates observed with conventional
TEM correlate well with our observations of As-initiated
vicinal GaAs films. Irrespective of the rotation in tilt direc-
tion to the perpendicular h110i between Xie’s work and the
work presented here, the same very low density and equal
distribution of microtwins with {111} habit planes perpen-
dicular to the tilt direction (Pole 2 and 4) is evident. Xie’s
work proposed preferentially oriented island nucleation, and
claimed that the asymmetric distribution of microtwinning is
due to the two (111)B planes that are aligned with the offcut
exhibiting a faster growth rate. The reductions in the Pole 2
and 4 twin variants are not observed for the other III-V sys-
tems (i.e., InP, GaSb, and AlSb), indicating that the fastest-
growing plane cannot be solely responsible for such reduc-
tions. The single domain nucleation achieved by Xie due to
Ga and As equal bonding preference to Si33 cannot be guar-
anteed in other systems due to the relatively low bonding
affinity of Sb34 and P.35 These other systems instead exhibit
a behavior in the distribution of microtwins that is a combi-
nation of the As-initiated and Ga-initiated GaAs, as demon-
strated by Xie et al., indicating mixed domain nucleation
(both group III and V atoms as first atomic layer species).
Despite the effort to utilize a group V-soaking prior to
growth, the high affinity of group III atoms for Si can easily
displace any weakly bonded group V atoms, especially at
higher growth temperatures. For growths of group III-
initiated nucleation, the fastest growing planes remain as
group V planes, but are rotated to the two perpendicular
{111} orientations not under the influence of the substrate
offcut asymmetry, as claimed by Xie et al. Therefore, in the
event of mixed domain nucleation, the asymmetric preferen-
tial growth of microtwins tilted toward the tilt direction in
III-Vs grown on vicinal Si cannot be attributed exclusively
to the faster growing group V planes.
Previous work by Wei and Aindow15 on GaAs epilayers
on vicinal Si concludes that under a balanced Ga- and As-
initiated flux, it is expected to achieve layer-by-layer growth.
They proposed that the high density of twins with habit
planes toward the tilt direction in combination with few
twins on all other habit planes is due to the deformation
resulting from the residual misfit associated with the heteroe-
pitaxy. The critical resolved shear stress (i.e., the threshold
value of stress necessary to cause atomic planes to slip) as
calculated by Wei and Aindow15 along the h112i slip direc-
tion in the {111} planes toward the tilt direction is over 4%
higher than all other slip systems. This means the likelihood
of slip in this direction is lower, so deformation twins would
more likely form. However, this slip direction is not one of
the two contributing partial dislocations (from the dissocia-
tion of a perfect dislocation along an interfacial h110i direc-
tion) responsible for the formation of a stacking fault
common in such epilayers. Therefore, the twins observed
cannot be exclusively or even primarily be due to deforma-
tion, as their asymmetric distribution cannot be explained by
the anisotropy in resolved shear stresses alone.
Comparing nominal and vicinal samples for all systems,
there is an increase in the mean intensity of twins, which we
attribute partially to an increased number of preferential
nucleation sites during initial growth, thus increasing the
number of twins, N, which can form. Next, there is an
enhancement in the fraction of twins with (111) habit planes
oriented toward the tilt direction in GaAs, which is not
observed for GaSb, AlSb and InP. This phenomenon is evi-
dent in both the 2DXRD intensity of Pole 1 (see Fig. 3(a))
and the DF TEM image (see Fig. 4(d)). We can surmise from
Xie et al. that the As-initiated GaAs growth induces an
exclusively As-prelayer, which leads to a domain in which
(111)B surfaces are on (111) and (111) aligned with the step
direction. The high assembly rates of those planes increase the
probability of errors in the form of stacking faults, which, in
124316-5 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)

turn, increases the total number of twins nucleated, N. While
in the case of mixed domain nucleation with GaSb, the
enhancement in the (111) fast-assembly surface is diluted by
the even distribution to the adjacent (111) and (111) planes of
equal assembly rates. Because of this, fewer stacking errors
occur in the (111) and (111) planes, and the number of twins
nucleated along those planes are decreased. The effects of this
enhancement in the fraction of twins in III-Vs (including
GaAs) is expected to be less pronounced with increasing
growth temperature, where the group V-prelayer atoms are
more likely to desorb, causing mixed nucleation and results,
which are more comparable to those of GaSb and AlSb. The
presence of single steps on vicinal substrates can also affect
this enhancement phenomenon, as they cause a rotation of
domains, which result in APBs. The strong asymmetry in
GaAs samples indicates that the Si substrate preparation
results in mostly double-stepped vicinal substrates.
The sum of the four reflection intensities for GaSb
grown on vicinal substrates is close to two times higher than
for GaAs grown on vicinal substrates. This is evident when
comparing the width of twin domains with (111) and (111)
habit planes in GaSb grown on nominal and vicinal Si sub-
strates, which are about two times as wide as those in GaAs,
as is shown in Figs. 4(c) and 4(f), demonstrating an overall
increase in the volume fraction of those twin variants. The
width of twins in GaSb grown on nominal and vicinal Si sub-
strates with two (111) and (111) habit planes are also about
three times as wide as those in GaAs (the TEM images of the
perpendicular [110] cross-section are not presented here).
Our proposed mechanism for the reduction of twins is
the transition to a step-flow growth mode induced by vicinal
substrates. On nominal oriented substrates a transition to
layer-by-layer growth simply allows for the continued propa-
gation of nucleated epitaxial twins throughout the film as the
growth front propagates normal to the substrate. In some sys-
tems on nominal substrates, layer-by-layer growth is never
achieved, as is observed for InP and AlSb. On vicinal sub-
strates, growth does not transition to a pure layer-by-layer
growth mode but instead to a step-flow growth mode, due to
the steps present on the underlying substrate. From the tran-
sition point onwards, the growth front is not normal to the
substrate, but instead flows across the surface at an enhanced
rate toward the tilt direction. This enhancement can also
increase the fraction of twins forming in the step-flow direc-
tion as is the case for GaAs grown on vicinal substrates.
Growth fronts originating from the steps flow across the sur-
face and overgrow islands that formed during the initial
island growth mode. Since step-flow growth occurs down
steps, toward the tilt direction, twins that were initially
nucleated on habit planes away from the tilt direction during
island growth are halted since they need to propagate oppo-
site to the growth front. The transition from island to step-
flow growth is evident due to the presence of nanotwins
which propagate 25 nm from the interface before termina-
tion, as is shown in the GaSb TEM DF image (Fig. 5). This
propagation distance into the film thickness is consistent
with the transition from an island to step-flow growth
mode as is apparent from the observed transition from a
spotty to streaky RHEED pattern during the GaSb growth.
The transition to step-flow growth ultimately decreases the
propagation length, L, of the twins with habit planes opposite
to the tilt direction.
Greater vicinal angles are expected to initiate step-flow
growth from nucleation, and hence completely eliminate
twins opposite to the tilt direction. An offcut toward the
[100] direction can induce surface steps in the two orthogo-
nal h110i directions simultaneously,27 which is expected to
decrease the twin density in two of four {111} habit planes,
reducing the overall intersections of microtwins that do so
along incoherent interfaces.
CONCLUSION
The results presented here demonstrate a newly
expanded understanding of the role that vicinal substrates
play in the formation of epitaxial twins, along with a mecha-
nism by which twins are preferentially suppressed away
from the tilt direction. Twins are reduced by overgrowth
from a step-flow growth process. Furthermore, the first layer
species induced by specific prelayer conditions have also
been shown to be important in twin formation in III-V semi-
conductor films. This work also demonstrates the importance
of combining 2DXRD and TEM analysis techniques to
obtain both a macro and nano/micro understanding of the
material systems. In addition, these results offer a route for
improving lateral III-V semiconductor device performance
where 1D transport is of prime importance. It is predicted
that further increases to the vicinal offcut angle would reduce
or eliminate twins due to a pure step-flow growth, and off-
cuts toward the [100] direction would form 2D steps for the
reduction of microtwins in two directions simultaneously.
ACKNOWLEDGMENTS
The authors thank Dr. B. J. Robinson for the experimen-
tal assistance with growth of the samples. G.A.B. thanks M.
Aindow for helpful discussions regarding this paper. This
work was supported by the Ontario Centres of Excellence
and ARISE Technologies. Electron Microscopy work was
carried out at the Canadian Centre for Electron Microscopy
(CCEM), a facility supported by NSERC and McMaster Uni-
versity. X-ray work was carried out at the McMaster Analyti-
cal X-Ray Diffraction Facility (MAX), a facility supported
by NSERC and McMaster University.
1D. B. Holt, Phys. Chemistry Solids 30, 1297 (1969).
2H. Kroemer, J. Cryst. Growth 81, 193 (1987).
3F. Ernst and P. Pirouz, J. Mater. Res. 4, 834 (1989).
4C. N. J. Wagner, in Local Atomic Arrangements Studied by X-Ray Diffrac-
tion, edited by J. B. Cohen and J. E. Hilliard (Gordon and Breach,
New York, 1966), pp. 224-225.
5H. Toyota, T. Sasaki, Y. Jinbo, and N. Uchitomi, J. Cryst. Growth 310, 78
(2008).
6Y. H. Kim, J. Y. Lee, Y. G. Noh, M. D. Kim, S. M. Cho, Y. J. Kwon, and
J. E. Oh, Appl. Phys. Lett. 88, 241907 (2006).
7H. Y. Xu, Y. N. Guo, Y. Wang, J. Zou, J. H. Kang, Q. Gao, H. H. Tan, and
C. Jagadish, J. Appl. Phys. 106, 083514 (2009).
8A. Proessdorf, F. Grosse, O. Romanyuk, W. Braun, B. Jenichen, A. Tram-
pert, and H. Riechert, J. Cryst. Growth 323, 401 (2011).
9R. Fischer, D. Neuman, H. Zabel, H. Morkoc¸, C. Choi, and N. Otsuka,
Appl. Phys. Lett. 48, 1223 (1986).
124316-6 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)

10T. Nguyen, W. Varhue, E. Adams, M. Lavoie, and S. Mongeon, J. Mater.
Res. 19, 2315 (2004).
11H. Noge, H. Kano, T. Kato, M. Hashimoto, and I. Igarashi, J. Cryst.
Growth 83, 431 (1987).
12A. Vila`, A. Cornet, J. R. Morante, P. Ruterana, M. Loubradou, R. Bonnet,
Y. Gonza´lez, and L. Gonza´lez, Philos. Mag. A 71, 85 (1995).
13R. Fischer, H. Morkoc¸, D. A. Neumann, H. Zabel, C. Choi, N.
Otsuka, M. Longerbone, and L. P. Erickson, J. Appl. Phys. 60, 1640
(1986).
14J. Hornstra, Physica 25, 409 (1959).
15X. L. Wei and M. Aindow, Appl. Phys. Lett. 65, 1903 (1994).
16Q. H. Xie, K. K. Fung, A. J. Ding, L. H. Cai, Y. Huang, and J. M. Zhou,
Appl. Phys. Lett. 57, 2803 (1990).
17K. C. Rajkumar, A. Madhukar, J. K. Liu, and F. J. Grunthaner, Appl. Phys.
Lett. 56, 1160 (1990).
18H. Neergaard Waltenburg and J. T. Yates, Chem. Rev. 95, 1589
(1995).
19E. Schroder-Bergen and W. Ranke, Surf. Sci. 259, 323 (1991).
20T. Sakamoto and G. Hashiguchi, Jpn. J. Appl. Phys. 25, L78 (1986).
21E. Pehlke and J. Tersoff, Phys. Rev. Lett. 67, 1290 (1991).
22K. Akahane, N. Yamamoto, S. Gozu, and N. Ohtani, J. Cryst. Growth 264,
21 (2004).
23G. Balakrishnan, S. H. Huang, A. Khoshakhlagh, A. Jallipalli, P. Rotella,
A. Amtout, S. Krishna, C. P. Haines, L. R. Dawson, and D. L. Huffaker,
Electron. Lett. 42, 1 (2006).
24S. H. Vajargah, M. Couillard, K. Cui, S. G. Tavakoli, B. Robinson, R. N. Klei-
man, J. S. Preston, and G. A. Botton, Appl. Phys. Lett. 98, 082113 (2011).
25J. F. Britten and W. Guan, Commission Crystallograph. Comput. Newslet-
ter 8, 96 (2007).
26S. Y. Woo, S. H. Vajargah, S. Ghanad-Tavakoli, R. N. Kleiman, and G. A.
Botton, “Direct observation of anti-phase boundaries in heteroepitaxy of
GaSb thin films grown on Si(001) by transmission electron microscopy,”
J. Appl. Phys. (to be submitted).
27S. F. Fang, K. Adomi, S. Iyer, H. Morkoc¸, H. Zabel, C. Choi, and N.
Otsuka, J. Appl. Phys. 68, R31 (1990).
28Y. Kim, Y. Noh, M. Kim, J. Oh, and K. Chung, Thin Solid Films 518,
2280 (2010).
29J. Tersoff, A. W. D. van der Gon, and R. Tromp, Phys. Rev. Lett. 72, 266
(1994).
30O. Oda, Compound Semiconductor Bulk Materials and Characterizations
(World Scientific, Singapore, 2007), p. 556.
31H. Gottschalk, G. Patzer, and H. Alexander, Physica Status Solidi A 45,
207 (1978).
32D. B. Holt and B. G. Yacobi, Extended Defects in Semiconductors: Elec-
tronic Properties, Device Effects and Structures (Cambridge University
Press, Cambridge; New York, 2007).
33R. D. Bringans, M. A. Olmstead, F. A. Ponce, D. K. Biegelsen, B. S.
Krusor, and R. D. Yingling, J. Appl. Phys. 64, 3472 (1988).
34R. A. A. Kubiak, J. Vac. Sci. Technol. B 3, 592 (1985).
35Y. Kiyota and T. Inada, J. Vac. Sci. Techno. A 19, 2441 (2001).
124316-7 Devenyi et al. J. Appl. Phys. 110, 124316 (2011)