ArticlePDF Available

Indium Incorporation into InGaN Quantum Wells Grown on GaN Narrow Stripes

Authors:
  • Atende Industries Sp. z o. o.

Abstract and Figures

InGaN quantum wells were grown using metalorganic chemical vapor phase epitaxy (vertical and horizontal types of reactors) on stripes made on GaN substrate. The stripe width was 5, 10, 20, 50, and 100 µm and their height was 4 and 1 µm. InGaN wells grown on stripes made in the direction perpendicular to the off-cut had a rough morphology and, therefore, this azimuth of stripes was not further explored. InGaN wells grown on the stripes made in the direction parallel to the GaN substrate off-cut had a step-flow-like morphology. For these samples (grown at low temperatures), we found out that the InGaN growth rate was higher for the narrower stripes. The higher growth rate induces a higher indium incorporation and a longer wavelength emission in photoluminescence measurements. This phenomenon is very clear for the 4 µm high stripes and less pronounced for the shallower 1 µm high stripes. The dependence of the emission wavelength on the stripe width paves a way to multicolor emitters.
Content may be subject to copyright.
materials
Article
Indium Incorporation into InGaN Quantum Wells
Grown on GaN Narrow Stripes
Marcin Sarzy ´nski 1, 2, * , Ewa Grzanka 1,2, Szymon Grzanka 1,2, Grzegorz Targowski 1,2,
Robert Czernecki 1,2, Anna Reszka 3, Vaclav Holy 4, Shugo Nitta 5, Zhibin Liu 6,
Hiroshi Amano 5,7,8 and Mike Leszczy ´nski 1,2
1Institute of High Pressure Physics PAS, Sokołowska 29/37, 01-142 Warsaw, Poland
2TopGaN Ltd., Sokołowska 29/37, 01-142 Warsaw, Poland
3Institute of Physics PAS, Al. Lotnikow 32/46, 02-668 Warsaw, Poland
4Faculty of Mathematics and Physics, Charles University, Ke Karlovu 5, 121 16 Praha 2, Czech Republic
5Institute of Materials and Systems for Sustainability, Nagoya University, Furo-cho, Chikusa-ku,
Nagoya 464-8603, Japan
6Department of Electrical Engineering and Computer Science, Nagoya University, Furo-cho, Chikusa-ku,
Nagoya 464-8603, Japan
7Akasaki Research Center, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
8Venture Business Laboratory, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
*Correspondence: sarzyn@unipress.waw.pl; Tel.: +48-501-717-528
Received: 3 July 2019; Accepted: 6 August 2019; Published: 14 August 2019


Abstract:
InGaN quantum wells were grown using metalorganic chemical vapor phase epitaxy
(vertical and horizontal types of reactors) on stripes made on GaN substrate. The stripe width was 5,
10, 20, 50, and 100
µ
m and their height was 4 and 1
µ
m. InGaN wells grown on stripes made in the
direction perpendicular to the o-cut had a rough morphology and, therefore, this azimuth of stripes
was not further explored. InGaN wells grown on the stripes made in the direction parallel to the GaN
substrate o-cut had a step-flow-like morphology. For these samples (grown at low temperatures),
we found out that the InGaN growth rate was higher for the narrower stripes. The higher growth
rate induces a higher indium incorporation and a longer wavelength emission in photoluminescence
measurements. This phenomenon is very clear for the 4
µ
m high stripes and less pronounced for the
shallower 1
µ
m high stripes. The dependence of the emission wavelength on the stripe width paves a
way to multicolor emitters.
Keywords: InGaN; vapor phase epitaxy; patterned substrate; quantum wells; multicolor emitters
1. Introduction
Most electronic and optoelectronic devices are fabricated on laterally homogeneous epitaxial
structures, however, having structures of properties varying in lateral directions would oer new
possibilities, for example, monolithic integration of dierent devices. Such epitaxial structures are
prepared by lateral patterning (lithography and masking) and overgrowth.
The first epitaxial lateral overgrowth (ELOG) of silicon and GaAs over SiOx masks was
demonstrated almost 40 years ago [
1
4
]. This technology was then used for GaN growth on highly
mismatched substrates—sapphire [
5
7
], SiC [
8
], or silicon [
9
]. In all cases, laterally grown GaN over the
mask had substantially lower threading dislocation density than in areas of mask openings. The next
step of this lateral overgrowth was pendeo epitaxy on stripes made on the GaN layer on foreign
substrate [
10
]. Also, in that kind of epitaxy, the suspended wings contained much lower dislocations.
Pendeo epitaxy of GaN paved the way to growth of AlGaInN epitaxial structures on laterally
pattered substrates. In the case of highly mismatched materials, lateral patterning may not only lead
Materials 2019,12, 2583; doi:10.3390/ma12162583 www.mdpi.com/journal/materials
Materials 2019,12, 2583 2 of 15
to lateral chemical composition variation, but also to elastic strain relaxation. Having submicron
patterning, one may obtain quantum dots and wires [11,12].
In the present paper we focus on the growth of the ternary alloy, InGaN, on patterned GaN
substrates. InGaN is the key material in III-Nitride optoelectronic devices such as light emitting diodes
(LEDs) and laser diodes. operating in the blue and green ranges of the electromagnetic spectrum. Single
and multiple quantum wells (SQWs, MQWs or just QWs for short) made of InGaN usually form the
active region of such devices, which are typically grown on dierent sorts of GaN substrates. However,
InGaN is relatively dicult to be grown high-quality (compared to the base binary compound, GaN)
because of its large lattice mismatch to GaN 11% [
13
,
14
] and low growth temperature—700–800
C for
InGaN [
15
17
] versus 950–1050
C for GaN. These two factors induce a number of crystallographic
defects, such as indium concentration fluctuations [
18
21
] and a large concentration of point defects [
22
].
Such defects, in turn, adversely aect optical properties of devices.
Growth of InGaN on three-dimensional structures has been studied in a number of papers [
23
,
24
].
These papers dealt mainly with the growth of nanocolumns. Such three-dimensional growth is possible
due to the low atomic incorporation at the side walls, and a high incorporation on the (00.1) surface.
InGaN growth on a substrate with stripes was studied by Fang and coworkers [
25
] who studied the
mass transport mechanism in InGaN epitaxy on ridge-shaped selective area growth GaN by metal
organic chemical vapor deposition. In that work, however, emphasis was put on dierent aspects and
the authors did not describe indium composition as a function of the stripe width. They also used
selective mask for the growth, which is not the case in the present paper.
In one of our previous papers [
26
] we reported on InGaN quantum wells grown on stripes with
various o-cuts. As indium incorporation into InGaN depends on the o-cut [
27
], by varying the o-cut
spatially we could produce monolithically integrated multicolor emitters. However, when decreasing
the size of the stripes we found out that not only the substrate o-cut, but also the size of the stripes
influences indium incorporation into InGaN. The present paper reports on this phenomenon—we will
show that the growth rate depends on the geometric dimensions of the stripes (width and height).
Because indium incorporation into InGaN depends on the growth rate [
28
,
29
], consequently. In content
also changes with the width of the stripes.
2. Experimental
2.1. Substrate Preparation
We used freestanding, c-plane (0001) GaN substrates from Saint-Gobain Lumilog company
(Vallauris, France). Dislocation density was about 10
7
/cm
2
and the o-cut angle was 0.6
±
0.05
towards
the m-direction <1–100>. Laser-beam photolithography (Microtech LW405B, Palermo, Italy) and ion
etching with chlorine chemistry (Oxford Plasma Lab 100, Yatton, UK) were used to define the substrate
pattern. The substrate scheme is shown in Figure 1a,b.
All prepared GaN substrates (11
×
14 mm in size) had five patterned regions marked P5, P10, P20,
P50, and P100, each with size 1.5
×
8 mm. Stripe width, w, was 5, 10, 20, 50, and 100
µ
m in regions P5,
P10, P20, P50, and P100, respectively, as shown in Figure 1a,b. In each region, stripe separation (g) was
equal to the respective stripe width (w), as shown in Figure 1a. Between regions P50 and P100 there
was a flat 1000
µ
m wide reference region without any stripes. Stripe direction was chosen along the
m-direction <1–100>as to be sure that during the epitaxial growth atomic steps could flow along the
top of the stripe. However, we also prepared substrate with stripes perpendicular to the m-direction,
as shown in Table 1. To check the morphology of QWs grown this way, this sample will be described in
Section 3.1. Typically, the pattern consisted of a set of 4
µ
m high parallel stripes. However, to check
how luminescence depends on the height of the stripes we also prepared substrate with 1
µ
m high
stripes, as shown in Table 1, and results are described in Section 3.5. Before epitaxy, all substrates were
cleaned in organic solvents and piranha.
Materials 2019,12, 2583 3 of 15
Materials 2019, 12, x FOR PEER REVIEW 2 of 15
to lateral chemical composition variation, but also to elastic strain relaxation. Having submicron
patterning, one may obtain quantum dots and wires [11,12].
In the present paper we focus on the growth of the ternary alloy, InGaN, on patterned GaN
substrates. InGaN is the key material in III-Nitride optoelectronic devices such as light emitting
diodes (LEDs) and laser diodes. operating in the blue and green ranges of the electromagnetic
spectrum. Single and multiple quantum wells (SQWs, MQWs or just QWs for short) made of InGaN
usually form the active region of such devices, which are typically grown on different sorts of GaN
substrates. However, InGaN is relatively difficult to be grown high-quality (compared to the base
binary compound, GaN) because of its large lattice mismatch to GaN 11% [13,14] and low growth
temperature—700–800 °C for InGaN [15–17] versus 950–1050 °C for GaN. These two factors induce a
number of crystallographic defects, such as indium concentration fluctuations [18–21] and a large
concentration of point defects [22]. Such defects, in turn, adversely affect optical properties of devices.
Growth of InGaN on three-dimensional structures has been studied in a number of papers
[23,24]. These papers dealt mainly with the growth of nanocolumns. Such three-dimensional growth
is possible due to the low atomic incorporation at the side walls, and a high incorporation on the
(00.1) surface. InGaN growth on a substrate with stripes was studied by Fang and coworkers [25]
who studied the mass transport mechanism in InGaN epitaxy on ridge-shaped selective area growth
GaN by metal organic chemical vapor deposition. In that work, however, emphasis was put on
different aspects and the authors did not describe indium composition as a function of the stripe
width. They also used selective mask for the growth, which is not the case in the present paper.
In one of our previous papers [26] we reported on InGaN quantum wells grown on stripes with
various off-cuts. As indium incorporation into InGaN depends on the off-cut [27], by varying the off-
cut spatially we could produce monolithically integrated multicolor emitters. However, when
decreasing the size of the stripes we found out that not only the substrate off-cut, but also the size of
the stripes influences indium incorporation into InGaN. The present paper reports on this
phenomenon—we will show that the growth rate depends on the geometric dimensions of the stripes
(width and height). Because indium incorporation into InGaN depends on the growth rate [28,29],
consequently. In content also changes with the width of the stripes.
2. Experimental
2.1. Substrate Preparation
We used freestanding, c-plane (0001) GaN substrates from Saint-Gobain Lumilog company
(Vallauris, France). Dislocation density was about 10
7
/cm
2
and the off-cut angle was 0.6 ± 0.05°
towards the m-direction <1–100>. Laser-beam photolithography (Microtech LW405B, Palermo, Italy)
and ion etching with chlorine chemistry (Oxford Plasma Lab 100, Yatton, UK) were used to define
the substrate pattern. The substrate scheme is shown in Figure 1a,b.
(a) (b)
Figure. 1. Stripes fabricated in GaN substrates. (a) Schematic perspective view and definition of stripe
dimensions. (b) Scheme of a 11 × 14 mm c-plane (0001) GaN substrate with five patterned regions
marked P5, P10, P20, P50, and P100. The stripe width (w) is 5, 10, 20, 50, and 100 μm in regions P5,
P10, P20, P50, and P100, respectively. In all regions the gap between stripes (g) was equal to the stripe
Figure 1.
Stripes fabricated in GaN substrates. (
a
) Schematic perspective view and definition of stripe
dimensions. (
b
) Scheme of a 11
×
14 mm c-plane (0001) GaN substrate with five patterned regions
marked P5, P10, P20, P50, and P100. The stripe width (w) is 5, 10, 20, 50, and 100
µ
m in regions P5, P10,
P20, P50, and P100, respectively. In all regions the gap between stripes (g) was equal to the stripe width
(g =w). Stripe height (h) was 4
µ
m in standard samples and 1
µ
m in all regions in the testing sample.
The flat area named “Ref” of 1000
µ
m wide, without stripes, was made as a reference. Each region P
has dimensions 1.5 ×8 mm and typically is fabricated along the m-direction.
Table 1. List of substrates. Symbols kand mean “parallel to” and “perpendicular to”, respectively.
Substrate d (µm) Azimuth 3
1 4 km
2 4 m
3 1 km
2.2. Epitaxy Method
The sidewall angle of patterned substrates before the growth was about 70–80
with respect to
the (0001) plane. Then, InGaN/GaN structures were grown on them, using metalorganic vapor phase
epitaxy (MOVPE). The growth processes were carried out on two MOVPE reactors—one of them was
a home-built, vertical reactor [
30
] located in the Institute of High Pressure Physics, Warsaw, Poland
(IHPP) and the other was a commercial Taiyo Nippon-Sanso horizontal reactor [
31
] located in the
Amano Lab, Institute of Materials and Systems for Sustainability, Nagoya University, Nagoya, Japan
(IMaSS). We used two dierent reactors to confirm that the observed eect was not specific to a
particular epitaxy system. The epitaxial structure grown in the above mentioned reactors consisted of
subsequent layers: 0.5
µ
m GaN:Si (growth temperature T
gr
=1040
C), 170 nm undoped In
0.03
Ga
0.95
N
buer layer (T
gr
=810
C), and a single pair of undoped, 2 nm thick In
0.25
Ga
0.75
N quantum well (QW)/8
nm GaN quantum barrier (QB) (T
gr
=760
C). An identical structure was grown on substrates with
stripes in both directions, i.e., parallel and perpendicular to the m-direction (see Section 3.1). For the
purpose of HR-XRD examination, a similar structure was grown on substrates with stripes parallel
to the m-direction using the vertical reactor. It contained 5 QB/QW pairs (thicknesses 2 and 8 nm,
respectively) instead of one to improve the ability to track QW properties with X-ray methods—results
are described in Section 3.3.
Eorts were made to carefully tune growth recipes for both reactors, to obtain identical structures.
However, in MOVPE, indium composition is always very sensitive to many parameters, mainly growth
temperature. For this reason, average indium composition achieved in the horizontal reactor was
lower than for the vertical one. Nevertheless, the main eect was observed for structures grown in
both reactors.
Materials 2019,12, 2583 4 of 15
2.3. Sample Characterization
Morphology of samples was checked by tapping mode atomic force microscope (AFM, Veeco
Dimension 3100, Plainview, NY, USA).
Optical properties were studied at room temperature using two methods—micro-photoluminescence
(
µ
PL) and cathodoluminescence (CL). In the
µ
PL setup, luminescence was excited by a Kimmon He-Cd,
continuous wave, 325 nm laser with 15 mW output power, and spectra were acquired by a iHR 320
spectrometer (Horiba Scientific, Piscataway, NJ, USA). Excitation spot size in this system was about 5
µ
m.
CL measurements were performed with a SU-70 scanning electron microscope (Hitachi High Technologies,
Tokyo, Japan) equipped with the Gatan Mono CL3 system (Pleasanton, CA, USA). Accelerating voltage
ranged from 5 to 15 kV and the beam current from 2.4 to 14 nA.
Structural properties were evaluated on the sample with 5 QW/QB pairs to increase signal
intensity from QWs/QBs and improve ability to calculate indium content and width of QWs and
QBs. The HR-XRD system (Empyrean-Malvern Panalytical, Almelo, The Netherlands) with a CuK
α1
X-ray source and equipped with hybrid 2-bounce monochromator and a threefold Ge (220) analyzer
was used.
The structure grown in the horizontal reactor was investigated by scanning transmission electron
microscopy (STEM) in high-angle annular dark-field (HAADF) operation mode, using Tecnai G2 F20
S-TWIN microscope (FEI Company, Hillsboro, OR, USA) operated at 200 kV. A cross-sectional STEM
specimen was prepared by mechanical polishing and subsequent Ar+final ion milling until electron
transparency using a Veraion precision ion polishing system (Gatan, Pleasanton, CA, USA).
Additionally, we examined the structure grown on patterned substrate with 4
µ
m high stripes
with a Dektak 150 stylus profiler (Veeco, Plainview, NY, USA).
3. Experimental Results
3.1. Influence of Stripe Azimuth on the InGaN Morphology
All structures were grown in the step-flow mode. The structure morphology on substrates with
stripes parallel to the m-direction were similar to each other, both grown using vertical and horizontal
MOVPE reactors. AFM images for structure grown using a vertical reactor is shown in Figure 2a,b.
The morphology of the same structure grown on substrate with stripes perpendicular to the m-direction,
i.e., along a-direction, was dierent and is shown in Figure 2c.
Materials 2019, 12, x FOR PEER REVIEW 4 of 15
were acquired by a iHR 320 spectrometer (Horiba Scientific, Piscataway, NJ, USA). Excitation spot
size in this system was about 5 µm. CL measurements were performed with a SU-70 scanning electron
microscope (Hitachi High Technologies, Tokyo, Japan) equipped with the Gatan Mono CL3 system
(Pleasanton, CA, USA). Accelerating voltage ranged from 5 to 15 kV and the beam current from 2.4
to 14 nA.
Structural properties were evaluated on the sample with 5 QW/QB pairs to increase signal
intensity from QWs/QBs and improve ability to calculate indium content and width of QWs and QBs.
The HR-XRD system (Empyrean-Malvern Panalytical, Almelo, The Netherlands) with a CuKα1 X-ray
source and equipped with hybrid 2-bounce monochromator and a threefold Ge (220) analyzer was
used.
The structure grown in the horizontal reactor was investigated by scanning transmission
electron microscopy (STEM) in high-angle annular dark-field (HAADF) operation mode, using
Tecnai G2 F20 S-TWIN microscope (FEI Company, Hillsboro, OR, USA) operated at 200 kV. A cross-
sectional STEM specimen was prepared by mechanical polishing and subsequent Ar+ final ion
milling until electron transparency using a Veraion precision ion polishing system (Gatan,
Pleasanton, CA, USA).
Additionally, we examined the structure grown on patterned substrate with 4 μm high stripes
with a Dektak 150 stylus profiler (Veeco, Plainview, NY, USA).
3. Experimental Results
3.1. Influence of Stripe Azimuth on the InGaN Morphology
All structures were grown in the step-flow mode. The structure morphology on substrates with
stripes parallel to the m-direction were similar to each other, both grown using vertical and horizontal
MOVPE reactors. AFM images for structure grown using a vertical reactor is shown in Figure 2a,b.
The morphology of the same structure grown on substrate with stripes perpendicular to the m-
direction, i.e., along a-direction, was different and is shown in Figure 2c.
(a) (b) (c)
Figure 2. Atomic force microscope (AFM) height images of an InGaN/GaN structure grown on two
substrates with stripes. In (a,b), stripe direction was parallel to substrate miscut azimuth and in (c) it
was perpendicular.
The surface roughness (RMS) of QW grown on stripes parallel to the m–direction and measured
by AFM on a 20 × 20 µm area was 0.52 nm while RMS of the QW grown on substrate with stripes
perpendicular to the m-direction was 3.2 nm.
These differences in layer morphology correspond to the directions of the stripe with respect to
the atomic step direction of the substrate. For the substrates with stripes parallel to the m-direction,
atomic steps were perpendicular to the stripe boundary and during growth they could flow along
the stripe top without disturbances, resulting in a smooth surface, as shown in Figure 2a,b. On the
contrary, for the substrate with stripes perpendicular to the m-direction, atomic steps were parallel
to the stripe boundary and during growth they flew across the stripe, as shown in Figure 2c.
Propagation length of an atomic step during growth, l, can be estimated as l = d/tan
δ
where d is layer
thickness and
δ
is the substrate miscut. In our case, d = 0.68 µm and
δ
= 0.6°, so l = 65 µm which is
similar or larger than the stripe width. As a result, a large portion of the layer grew in an island mode
due to the lack of source of atomic steps. The morphology of the InGaN QWs in the case of the stripes
Figure 2.
Atomic force microscope (AFM) height images of an InGaN/GaN structure grown on two
substrates with stripes. In (
a
,
b
), stripe direction was parallel to substrate miscut azimuth and in (
c
) it
was perpendicular.
The surface roughness (RMS) of QW grown on stripes parallel to the m–direction and measured
by AFM on a 20
×
20
µ
m area was 0.52 nm while RMS of the QW grown on substrate with stripes
perpendicular to the m-direction was 3.2 nm.
These dierences in layer morphology correspond to the directions of the stripe with respect to
the atomic step direction of the substrate. For the substrates with stripes parallel to the m-direction,
atomic steps were perpendicular to the stripe boundary and during growth they could flow along
the stripe top without disturbances, resulting in a smooth surface, as shown in Figure 2a,b. On the
Materials 2019,12, 2583 5 of 15
contrary, for the substrate with stripes perpendicular to the m-direction, atomic steps were parallel to
the stripe boundary and during growth they flew across the stripe, as shown in Figure 2c. Propagation
length of an atomic step during growth, l, can be estimated as l=d/tan
δ
where dis layer thickness and
δ
is the substrate miscut. In our case, d=0.68
µ
m and
δ
=0.6
, so l=65
µ
m which is similar or larger
than the stripe width. As a result, a large portion of the layer grew in an island mode due to the lack of
source of atomic steps. The morphology of the InGaN QWs in the case of the stripes perpendicular to
the m-direction was not acceptable due to large and non-uniform roughness; all further results are
only for the QWs grown on substrates with stripes parallel to the m-direction.
3.2. Influence of Stripe width on Luminescence of InGaN QW
The most important result in the present paper is that luminescence wavelength of an InGaN
QW/QB structure strongly increased when it was grown on a narrower stripe. More precisely, for
single QW grown using a vertical reactor,
µ
PL wavelength shifted from 461 up to 495 nm, and for stripe
width variation, from 100 down to 5
µ
m, respectively. For nominally the same structure grown using a
horizontal reactor, a similar eect was found, and
µ
PL wavelength increased from 434 up to 451 nm,
and for stripe width change, from 100 to 5
µ
m, respectively. In both cases
µ
PL was measured on top of
each stripe, in the middle between stripe edges at room temperature. Luminescence wavelength as a
function of stripe width is shown in Figure 3a and µPL spectra are shown in Figure 3b,c.
Materials 2019, 12, x FOR PEER REVIEW 5 of 15
perpendicular to the m-direction was not acceptable due to large and non-uniform roughness; all
further results are only for the QWs grown on substrates with stripes parallel to the m-direction.
3.2. Influence of Stripe width on Luminescence of InGaN QW
The most important result in the present paper is that luminescence wavelength of an InGaN
QW/QB structure strongly increased when it was grown on a narrower stripe. More precisely, for
single QW grown using a vertical reactor, µPL wavelength shifted from 461 up to 495 nm, and for
stripe width variation, from 100 down to 5 µm, respectively. For nominally the same structure grown
using a horizontal reactor, a similar effect was found, and µPL wavelength increased from 434 up to
451 nm, and for stripe width change, from 100 to 5 µm, respectively. In both cases µPL was measured
on top of each stripe, in the middle between stripe edges at room temperature. Luminescence
wavelength as a function of stripe width is shown in Figure 3a and µPL spectra are shown in Figure
3b,c.
(a) (b) (c)
Figure 3. Micro-photoluminescence (µPL) of nominally identical InGaN/GaN structures grown on
substrates with 4 μm high stripes using vertical and horizontal reactors. (a) Central wavelength of
µPL as a function of stripe width, (b) normalized µPL spectra of structure grown using a vertical
reactor, (c) normalized µPL spectra of structure grown using a horizontal reactor.
In the next step we studied optical properties of our structure with single QW grown using a
vertical, and also a horizontal reactor, in more detail. For both samples we performed µPL scans along
and across stripes in all regions P5–P100, as shown in Figure 4a. We also made CL measurements in
three spots on top of the 50 µm wide stripe—in the center, 5 µm from the edge, and at the edge, as
shown in Figure 5a. The μPL measurement scheme and results (for 100 µm wide stripe) are shown in
Figure 4. The CL measurement scheme and results (for 50 µm wide stripe) are shown in Figure 4. It
turned out that central µPL wavelength is constant along the stripe. However, central luminescence
wavelength and full-width at half maximum (FWHM) of µPL and CL spectra increased from the
center of each stripe towards its edges.
(a) (b) (c)
Figure 4. μPL of the InGaN/GaN structure. Stripe width 100 µm, stripe height 4 μm. (a) Measurement
scheme with two scan lines across and along the stripe. (b) Wavelength across and along the stripe.
(c) Spectra acquired at stripe center, 10 µm from stripe edge, and at the stripe edge.
Figure 3.
Micro-photoluminescence (
µ
PL) of nominally identical InGaN/GaN structures grown on
substrates with 4
µ
m high stripes using vertical and horizontal reactors. (
a
) Central wavelength of
µ
PL
as a function of stripe width, (
b
) normalized
µ
PL spectra of structure grown using a vertical reactor,
(c) normalized µPL spectra of structure grown using a horizontal reactor.
In the next step we studied optical properties of our structure with single QW grown using a
vertical, and also a horizontal reactor, in more detail. For both samples we performed
µ
PL scans along
and across stripes in all regions P5–P100, as shown in Figure 4a. We also made CL measurements in
three spots on top of the 50
µ
m wide stripe—in the center, 5
µ
m from the edge, and at the edge, as
shown in Figure 5a. The
µ
PL measurement scheme and results (for 100
µ
m wide stripe) are shown in
Figure 4. The CL measurement scheme and results (for 50
µ
m wide stripe) are shown in Figure 4. It
turned out that central
µ
PL wavelength is constant along the stripe. However, central luminescence
wavelength and full-width at half maximum (FWHM) of
µ
PL and CL spectra increased from the center
of each stripe towards its edges.
Materials 2019,12, 2583 6 of 15
Materials 2019, 12, x FOR PEER REVIEW 5 of 15
perpendicular to the m-direction was not acceptable due to large and non-uniform roughness; all
further results are only for the QWs grown on substrates with stripes parallel to the m-direction.
3.2. Influence of Stripe width on Luminescence of InGaN QW
The most important result in the present paper is that luminescence wavelength of an InGaN
QW/QB structure strongly increased when it was grown on a narrower stripe. More precisely, for
single QW grown using a vertical reactor, µPL wavelength shifted from 461 up to 495 nm, and for
stripe width variation, from 100 down to 5 µm, respectively. For nominally the same structure grown
using a horizontal reactor, a similar effect was found, and µPL wavelength increased from 434 up to
451 nm, and for stripe width change, from 100 to 5 µm, respectively. In both cases µPL was measured
on top of each stripe, in the middle between stripe edges at room temperature. Luminescence
wavelength as a function of stripe width is shown in Figure 3a and µPL spectra are shown in Figure
3b,c.
(a) (b) (c)
Figure 3. Micro-photoluminescence (µPL) of nominally identical InGaN/GaN structures grown on
substrates with 4 μm high stripes using vertical and horizontal reactors. (a) Central wavelength of
µPL as a function of stripe width, (b) normalized µPL spectra of structure grown using a vertical
reactor, (c) normalized µPL spectra of structure grown using a horizontal reactor.
In the next step we studied optical properties of our structure with single QW grown using a
vertical, and also a horizontal reactor, in more detail. For both samples we performed µPL scans along
and across stripes in all regions P5–P100, as shown in Figure 4a. We also made CL measurements in
three spots on top of the 50 µm wide stripe—in the center, 5 µm from the edge, and at the edge, as
shown in Figure 5a. The μPL measurement scheme and results (for 100 µm wide stripe) are shown in
Figure 4. The CL measurement scheme and results (for 50 µm wide stripe) are shown in Figure 4. It
turned out that central µPL wavelength is constant along the stripe. However, central luminescence
wavelength and full-width at half maximum (FWHM) of µPL and CL spectra increased from the
center of each stripe towards its edges.
(a) (b) (c)
Figure 4. μPL of the InGaN/GaN structure. Stripe width 100 µm, stripe height 4 μm. (a) Measurement
scheme with two scan lines across and along the stripe. (b) Wavelength across and along the stripe.
(c) Spectra acquired at stripe center, 10 µm from stripe edge, and at the stripe edge.
Figure 4. µ
PL of the InGaN/GaN structure. Stripe width 100
µ
m, stripe height 4
µ
m. (
a
) Measurement
scheme with two scan lines across and along the stripe. (
b
) Wavelength across and along the stripe.
(c) Spectra acquired at stripe center, 10 µm from stripe edge, and at the stripe edge.
Materials 2019, 12, x FOR PEER REVIEW 6 of 15
(a) (b)
Figure 5. Cathodoluminescence (CL) of the InGaN/GaN structure. Stripe width was 50 µm. (a) Scheme
of CL measurement with three characteristic points. (b) CL spectra acquired at stripe center, 5 µm
from the stripe edge, and at the stripe edge.
3.3. Structural Properties
Structure parameters, i.e., QW and QB thicknesses and QW indium content were examined
independently in each of the patterned regions P5–P100 and in the reference region (see Figure 1 for
a region layout explanation). Parameter values were obtained using Panalytical Epitaxy software
with implemented dynamical theory of diffraction, by fitting calculated X-ray scans to measured
ones, as shown in Figure 6a. It was found that QW thickness increased from 2 up to 3 nm when stripe
width was decreased from 100 down to 5 µm, respectively, as shown in Figure 6b. Changes of QW/QB
thickness is clearly seen in Figure 6a by shifts of the Pendelosung fringes towards (0002) reflection of
GaN substrate. Similarly, QB thickness increased slightly from 8.2 up to 9 nm. At the same time,
indium composition in QWs increased from 19% on 100 µm wide stripes up to 23% on 5 µm wide
stripes, as shown in Figure 6b. Of course, due to the composition pulling effect [32], on narrow stripes,
the 1-QW and the 5-QW structures could differ. However, the observed wavelength variations with
respect to the stripe width were similar for both structures.
(a) (b)
Figure 6. HR-XRD measurements of the InGaN/GaN MQW structure on substrate with 4 μm high
stripes; (a) 2theta-omega scans collected at 1000 µm wide reference area, at 10 µm, and 5 µm stripe
regions (other regions not shown for picture clarity); (b) quantum well (QW) width and QW indium
composition for different stripe widths obtained by fitting respective HR-XRD scans. The X-ray beam
covers the whole width of each stripe.
3.4. Influence of TMIn Flow and Temperature on Growth Rate on the Stripes
To get more detail on the growth modes on the stripes, we examined the structure grown on
patterned substrate with 4 μm high stripes with a stylus profiler. We scanned all patterned regions
of the structure in a direction perpendicular to the stripes, i.e., across the stripes, as shown in Figure
Figure 5.
Cathodoluminescence (CL) of the InGaN/GaN structure. Stripe width was 50
µ
m. (
a
) Scheme
of CL measurement with three characteristic points. (
b
) CL spectra acquired at stripe center, 5
µ
m from
the stripe edge, and at the stripe edge.
3.3. Structural Properties
Structure parameters, i.e., QW and QB thicknesses and QW indium content were examined
independently in each of the patterned regions P5–P100 and in the reference region (see Figure 1for
a region layout explanation). Parameter values were obtained using Panalytical Epitaxy software
with implemented dynamical theory of diraction, by fitting calculated X-ray scans to measured ones,
as shown in Figure 6a. It was found that QW thickness increased from 2 up to 3 nm when stripe
width was decreased from 100 down to 5
µ
m, respectively, as shown in Figure 6b. Changes of QW/QB
thickness is clearly seen in Figure 6a by shifts of the Pendelosung fringes towards (0002) reflection
of GaN substrate. Similarly, QB thickness increased slightly from 8.2 up to 9 nm. At the same time,
indium composition in QWs increased from 19% on 100
µ
m wide stripes up to 23% on 5
µ
m wide
stripes, as shown in Figure 6b. Of course, due to the composition pulling eect [
32
], on narrow stripes,
the 1-QW and the 5-QW structures could dier. However, the observed wavelength variations with
respect to the stripe width were similar for both structures.
Materials 2019,12, 2583 7 of 15
Materials 2019, 12, x FOR PEER REVIEW 6 of 15
(a) (b)
Figure 5. Cathodoluminescence (CL) of the InGaN/GaN structure. Stripe width was 50 µm. (a) Scheme
of CL measurement with three characteristic points. (b) CL spectra acquired at stripe center, 5 µm
from the stripe edge, and at the stripe edge.
3.3. Structural Properties
Structure parameters, i.e., QW and QB thicknesses and QW indium content were examined
independently in each of the patterned regions P5–P100 and in the reference region (see Figure 1 for
a region layout explanation). Parameter values were obtained using Panalytical Epitaxy software
with implemented dynamical theory of diffraction, by fitting calculated X-ray scans to measured
ones, as shown in Figure 6a. It was found that QW thickness increased from 2 up to 3 nm when stripe
width was decreased from 100 down to 5 µm, respectively, as shown in Figure 6b. Changes of QW/QB
thickness is clearly seen in Figure 6a by shifts of the Pendelosung fringes towards (0002) reflection of
GaN substrate. Similarly, QB thickness increased slightly from 8.2 up to 9 nm. At the same time,
indium composition in QWs increased from 19% on 100 µm wide stripes up to 23% on 5 µm wide
stripes, as shown in Figure 6b. Of course, due to the composition pulling effect [32], on narrow stripes,
the 1-QW and the 5-QW structures could differ. However, the observed wavelength variations with
respect to the stripe width were similar for both structures.
(a) (b)
Figure 6. HR-XRD measurements of the InGaN/GaN MQW structure on substrate with 4 μm high
stripes; (a) 2theta-omega scans collected at 1000 µm wide reference area, at 10 µm, and 5 µm stripe
regions (other regions not shown for picture clarity); (b) quantum well (QW) width and QW indium
composition for different stripe widths obtained by fitting respective HR-XRD scans. The X-ray beam
covers the whole width of each stripe.
3.4. Influence of TMIn Flow and Temperature on Growth Rate on the Stripes
To get more detail on the growth modes on the stripes, we examined the structure grown on
patterned substrate with 4 μm high stripes with a stylus profiler. We scanned all patterned regions
of the structure in a direction perpendicular to the stripes, i.e., across the stripes, as shown in Figure
Figure 6.
HR-XRD measurements of the InGaN/GaN MQW structure on substrate with 4
µ
m high
stripes; (
a
) 2theta-omega scans collected at 1000
µ
m wide reference area, at 10
µ
m, and 5
µ
m stripe
regions (other regions not shown for picture clarity); (
b
) quantum well (QW) width and QW indium
composition for dierent stripe widths obtained by fitting respective HR-XRD scans. The X-ray beam
covers the whole width of each stripe.
3.4. Influence of TMIn Flow and Temperature on Growth Rate on the Stripes
To get more detail on the growth modes on the stripes, we examined the structure grown on
patterned substrate with 4
µ
m high stripes with a stylus profiler. We scanned all patterned regions of
the structure in a direction perpendicular to the stripes, i.e., across the stripes, as shown in Figure 7a. It
turned out that the structure height in the vicinity of the stripe edges is 100–150 nm larger than in the
stripe center.
Materials 2019, 12, x FOR PEER REVIEW 7 of 15
7a. It turned out that the structure height in the vicinity of the stripe edges is 100–150 nm larger than
in the stripe center.
(a) (b)
Figure 7. Surface profiles of InGaN/GaN structures grown on substrates with 20 µm wide stripes. All
profiles have been shifted in the z-direction to match each other’s height in the center of the stripe. In
(a) profile SUBS is for GaN substrate, S1—after growth of HT-GaN, S2—after HT-GaN and InGaN,
S3 after HT-GaN, InGaN, and 5QWs. In (b) profile SUBS is for GaN substrate and S4 after growth of
HT-GaN + LT-GaN. See text and Table 2 for more details.
Therefore, we decided to split our examined InGaN/GaN structure into parts, grow each of them
separately on new patterned substrates, as shown in Table 2, and then check surface profile for each
growth. Firstly, we measured the surface of patterned substrate in an identical manner (sample
SUBS). Another profile was measured after growth of 0.5 µm high-temperature GaN:Si (HT-GaN, Tgr
= 1040 °C, sample S1, see Table 2). Another profile was measured on epitaxial structure with 0.5 µm
HT-GaN GaN, followed by 170 nm InGaN, Tgr = 810 °C, sample S2—see Table 2. The next profile was
measured on epitaxial structure with five QW/QB pairs, sample S3—see Table 2. The results are
shown in Figure 7a. Finally, we compared measurement of patterned substrate (sample SUBS), as
shown in Table 2, to the sample which consisted of 0.5 µm HT-GaN:Si followed by 170 nm thick LT-
GaN (low temperature GaN, Tgr = 810 °C), sample S4, Table 2. The results are shown in Figure 7b.
Table 2. Summary of samples used for surface profile characterization. QB: quantum barrier.
Sample Name Epitaxial Structure Growth Temperature (°C)
SUBS Patterned substrate
S1 HT-GaN 1040
S2 HT-GaN + InGaN buffer 1040 + 810
S3 GaN + InGaN buffer + 5 × (QW + QB) 1040 + 810 + 760
S4 HT-GaN + LT-GaN 1040 + 810
It turned out that the initial angle of the stripe sidewalls (70–80° to the c-plane) changed after
growth of HT-GaN, because the exact-oriented, (11–22) facet has formed. An identical effect was
observed in both reactors. That is why surface profiles of SUBS and S1 differ. Taking this fact into
account, it can be seen that structure height at the stripe edges was larger only for GaN and InGaN
layers grown at relatively low temperature, i.e., 810 °C or lower. For HT-GaN, structure height on
top of the stripe was the same as for the patterned substrate without any grown structure.
From HR-XRD results we learnt that the thickness of InGaN buffer, QWs, and QBs grown on
top of the stripe were not lower but rather greater than the thickness of the same layers grown on the
flat reference region. Nominally, the InGaN buffer thickness was 170 nm. However, the height of the
stripe measured near the stripe’s edge is 100 nm larger than the height measured near the stripe’s
center. This means that the growth rate of the InGaN buffer close to the stripe edge must have been
60% larger than in the stripe center. Nominal thickness of five pairs of QW/QB was 50 nm. Their
Figure 7.
Surface profiles of InGaN/GaN structures grown on substrates with 20
µ
m wide stripes. All
profiles have been shifted in the z-direction to match each other’s height in the center of the stripe.
In (
a
) profile SUBS is for GaN substrate, S1—after growth of HT-GaN, S2—after HT-GaN and InGaN,
S3 after HT-GaN, InGaN, and 5QWs. In (
b
) profile SUBS is for GaN substrate and S4 after growth of
HT-GaN +LT-GaN. See text and Table 2for more details.
Table 2. Summary of samples used for surface profile characterization. QB: quantum barrier.
Sample Name Epitaxial Structure Growth Temperature (C)
SUBS Patterned substrate
S1 HT-GaN 1040
S2 HT-GaN +InGaN buer 1040 +810
S3
GaN +InGaN buer +5
×
(QW +QB)
1040 +810 +760
S4 HT-GaN +LT-GaN 1040 +810
Materials 2019,12, 2583 8 of 15
Therefore, we decided to split our examined InGaN/GaN structure into parts, grow each of
them separately on new patterned substrates, as shown in Table 2, and then check surface profile for
each growth. Firstly, we measured the surface of patterned substrate in an identical manner (sample
SUBS). Another profile was measured after growth of 0.5
µ
m high-temperature GaN:Si (HT-GaN,
T
gr
=1040
C, sample S1, see Table 2). Another profile was measured on epitaxial structure with 0.5
µ
m
HT-GaN GaN, followed by 170 nm InGaN, T
gr
=810
C, sample S2—see Table 2. The next profile was
measured on epitaxial structure with five QW/QB pairs, sample S3—see Table 2. The results are shown
in Figure 7a. Finally, we compared measurement of patterned substrate (sample SUBS), as shown in
Table 2, to the sample which consisted of 0.5
µ
m HT-GaN:Si followed by 170 nm thick LT-GaN (low
temperature GaN, Tgr =810 C), sample S4, Table 2. The results are shown in Figure 7b.
It turned out that the initial angle of the stripe sidewalls (70–80
to the c-plane) changed after
growth of HT-GaN, because the exact-oriented, (11–22) facet has formed. An identical eect was
observed in both reactors. That is why surface profiles of SUBS and S1 dier. Taking this fact into
account, it can be seen that structure height at the stripe edges was larger only for GaN and InGaN
layers grown at relatively low temperature, i.e., 810
C or lower. For HT-GaN, structure height on top
of the stripe was the same as for the patterned substrate without any grown structure.
From HR-XRD results we learnt that the thickness of InGaN buer, QWs, and QBs grown on top
of the stripe were not lower but rather greater than the thickness of the same layers grown on the flat
reference region. Nominally, the InGaN buer thickness was 170 nm. However, the height of the stripe
measured near the stripe’s edge is 100 nm larger than the height measured near the stripe’s center.
This means that the growth rate of the InGaN buer close to the stripe edge must have been 60% larger
than in the stripe center. Nominal thickness of five pairs of QW/QB was 50 nm. Their surface profile
close to the stripe edge extends by ~50 nm over the InGaN buer. It means that MQWs must have
grown 100% faster close to the stripe edge compared to the center.
3.5. Influence of the Stripe Height on Luminescence of InGaN QWs.
To investigate the influence of stripe height on morphology and optical properties of InGaN/GaN
structures, we prepared another patterned substrate. It was nearly identical to the substrate described
in Figure 1, but the stripe height was only 1
µ
m instead of 4
µ
m. On this substrate we grew an
InGaN/GaN epitaxial structure, identical to the one described in Section 2.2. It turned out that the
luminescence wavelength shift in the case of 1
µ
m high stripes was only 10 nm between the reference
region and 5
µ
m wide stripes, as shown in Figure 8a. Moreover, the eect of faster growth near the
stripe edges was not observed for 1 µm high stripes, as shown in Figure 8b.
Materials 2019, 12, x FOR PEER REVIEW 8 of 15
surface profile close to the stripe edge extends by ~50 nm over the InGaN buffer. It means that MQWs
must have grown 100% faster close to the stripe edge compared to the center.
3.5. Influence of the Stripe Height on Luminescence of InGaN QWs.
To investigate the influence of stripe height on morphology and optical properties of
InGaN/GaN structures, we prepared another patterned substrate. It was nearly identical to the
substrate described in Figure 1, but the stripe height was only 1 µm instead of 4 µm. On this substrate
we grew an InGaN/GaN epitaxial structure, identical to the one described in Section 2.2. It turned out
that the luminescence wavelength shift in the case of 1 µm high stripes was only 10 nm between the
reference region and 5 µm wide stripes, as shown in Figure 8a. Moreover, the effect of faster growth
near the stripe edges was not observed for 1 µm high stripes, as shown in Figure 8b.
(a) (b)
Figure 8. Luminescence (a) and surface profiles (b) of QW grown on substrates with stripes of
different height.
4. Discussion
The observed luminescence wavelength variations on different stripe widths may depend on
the following factors and their combinations: (i) variations in quantum-confined Stark effect (QCSE)
due to strain, (ii) variations in QW thickness, and (iii) variations in QW indium composition.
In order to assess the influence of elastic strains we performed numerical simulations of elastic
energy per one InGaN molecule at the free surfaces of the stripe and the trench between the stripes,
and the results are presented in Figure 9. For the simulations we used a standard finite-element
method, and elastic constants of GaN and InGaN were taken from [13,33]. We performed strain
simulations for the 20 µm wide stripe and it turned out that in our design of patterned substrate, the
influence of strain was limited to 2–3 µm from the stripe edge, as shown in Figure 9. However, from
µPL data we see that luminescence wavelength shifts were observed not only near the stripe edges
but on the whole 20 µm wide stripe (wavelength shifts more than 20 nm at the stripe center, compared
to the flat region of the substrate). Therefore, we conclude that mechanical strain could have an
influence on wavelength shift only near the stripe edge and is probably not the main cause for the
observed phenomena.
Figure 8.
Luminescence (
a
) and surface profiles (
b
) of QW grown on substrates with stripes of
dierent height.
Materials 2019,12, 2583 9 of 15
4. Discussion
The observed luminescence wavelength variations on dierent stripe widths may depend on the
following factors and their combinations: (i) variations in quantum-confined Stark eect (QCSE) due
to strain, (ii) variations in QW thickness, and (iii) variations in QW indium composition.
In order to assess the influence of elastic strains we performed numerical simulations of elastic
energy per one InGaN molecule at the free surfaces of the stripe and the trench between the stripes, and
the results are presented in Figure 9. For the simulations we used a standard finite-element method,
and elastic constants of GaN and InGaN were taken from [
13
,
33
]. We performed strain simulations for
the 20
µ
m wide stripe and it turned out that in our design of patterned substrate, the influence of strain
was limited to 2–3
µ
m from the stripe edge, as shown in Figure 9. However, from
µ
PL data we see that
luminescence wavelength shifts were observed not only near the stripe edges but on the whole 20
µ
m
wide stripe (wavelength shifts more than 20 nm at the stripe center, compared to the flat region of the
substrate). Therefore, we conclude that mechanical strain could have an influence on wavelength shift
only near the stripe edge and is probably not the main cause for the observed phenomena.
Materials 2019, 12, x FOR PEER REVIEW 9 of 15
Figure 9. (a) Structure model used for the calculation of elastic energy. (b) Lateral profiles of the elastic
energy per one InGaN molecule at the top and bottom InGaN surfaces.
Next, we tried to correlate QW thickness, growth rate, and indium composition on different
stripes. According to HR-XRD measurements in the present work (Section 3.3), made on the 5QW
structure, thicknesses of QWs and QBs on 5 µm wide stripes were 3 and 9 nm, respectively. However,
on 100 µm wide stripes the thicknesses were only 2 and 8 nm, respectively. Indium composition in
QWs was 23% for the 5 µm stripe and 19% for the 100 µm one, as shown in Figure 6. Thus, the growth
rate of the QW was 0.33 Å/s on 100 µm stripes and 0.5 Å/s for 5 µm ones.
The influence of the growth rate on indium composition has been investigated by Leszczyński
et al. [29]. In that work, it was explained that incorporation of indium is larger at the higher growth
rate because it is necessary to overbuild the indium atoms by gallium, otherwise it desorbs from the
surface. Increased indium composition on narrow stripes could have been solely the result of faster
growth. Hence, the increased µPL wavelength observed on narrow stripes can be explained by the
joint effect of increased indium content and thicker QWs (thicker wells emit in longer wavelengths
due to stronger QCSE).
Finally, we would like to know why the growth rate is higher on narrow stripes. As the observed
effects occurred during MOVPE growth on patterned substrates, they should be discussed taking
into account the following phenomena: (i) gas phase transport [34–36], (ii) gas phase reactions [37–
39], (iii) gas phase diffusion [40–45], and (iv) surface diffusion [46–49].
Gas phase transport depends on the reactor design details. However, since we observed the
effect in two very different reactors, we conclude that gas phase transport is not the most important
factor here and we neglect it.
Concerning gas phase reactions and gas phase diffusion, the presence of stripes (uneven surface)
influences decomposition of the active species, their diffusion, and incorporation. The observed faster
growth at the stripe edge would suggest that locally (edge area) the density of the active species
participating in the growth is higher than in the stripe center or on the flat reference region. Because
in our experiments the III-element sources were trimethylgallium and trimethylindium, the gas phase
was dominated by their respective trimethyl (TM) and dimethyl (DM) species—DM resulting from
TM decomposition. Although the DM lifetime is extremely short, it decomposes into monomethyl
(MM), which diffuses on the surface with a given diffusion length, and in turn, is decomposed in
adatoms, which also diffuse on the surface with their own diffusion lengths. All these species are
affected by the presence of the stripes. The gas phase was preferentially decomposed at the stripe
edges (possibly because of a higher temperature gradient [50]), thus inducing higher density of
species (MM and adatoms) diffusing on the surface. As the stripe is higher, this effect becomes more
pronounced.
Surface diffusion of MM species and adatoms will affect observed phenomena in a different
way. The MM influences the distribution of adatoms on the surface, which consequently influences
the local growth rate. Diffusion lengths of MM and adatoms can be estimated from the present
experimental results. In this view, different indium compositions on different stripes, as shown in
Figure 9.
(
a
) Structure model used for the calculation of elastic energy. (
b
) Lateral profiles of the elastic
energy per one InGaN molecule at the top and bottom InGaN surfaces.
Next, we tried to correlate QW thickness, growth rate, and indium composition on dierent
stripes. According to HR-XRD measurements in the present work (Section 3.3), made on the 5QW
structure, thicknesses of QWs and QBs on 5
µ
m wide stripes were 3 and 9 nm, respectively. However,
on 100
µ
m wide stripes the thicknesses were only 2 and 8 nm, respectively. Indium composition in
QWs was 23% for the 5
µ
m stripe and 19% for the 100
µ
m one, as shown in Figure 6. Thus, the growth
rate of the QW was 0.33 Å/s on 100 µm stripes and 0.5 Å/s for 5 µm ones.
The influence of the growth rate on indium composition has been investigated by
Leszczy´nski et al. [
29
]. In that work, it was explained that incorporation of indium is larger at
the higher growth rate because it is necessary to overbuild the indium atoms by gallium, otherwise it
desorbs from the surface. Increased indium composition on narrow stripes could have been solely
the result of faster growth. Hence, the increased
µ
PL wavelength observed on narrow stripes can be
explained by the joint eect of increased indium content and thicker QWs (thicker wells emit in longer
wavelengths due to stronger QCSE).
Finally, we would like to know why the growth rate is higher on narrow stripes. As the observed
eects occurred during MOVPE growth on patterned substrates, they should be discussed taking into
account the following phenomena: (i) gas phase transport [
34
36
], (ii) gas phase reactions [
37
39
],
(iii) gas phase diusion [4045], and (iv) surface diusion [4649].
Gas phase transport depends on the reactor design details. However, since we observed the eect
in two very dierent reactors, we conclude that gas phase transport is not the most important factor
here and we neglect it.
Materials 2019,12, 2583 10 of 15
Concerning gas phase reactions and gas phase diusion, the presence of stripes (uneven surface)
influences decomposition of the active species, their diusion, and incorporation. The observed faster
growth at the stripe edge would suggest that locally (edge area) the density of the active species
participating in the growth is higher than in the stripe center or on the flat reference region. Because in
our experiments the III-element sources were trimethylgallium and trimethylindium, the gas phase
was dominated by their respective trimethyl (TM) and dimethyl (DM) species—DM resulting from TM
decomposition. Although the DM lifetime is extremely short, it decomposes into monomethyl (MM),
which diuses on the surface with a given diusion length, and in turn, is decomposed in adatoms,
which also diuse on the surface with their own diusion lengths. All these species are aected by the
presence of the stripes. The gas phase was preferentially decomposed at the stripe edges (possibly
because of a higher temperature gradient [
50
]), thus inducing higher density of species (MM and
adatoms) diusing on the surface. As the stripe is higher, this eect becomes more pronounced.
Surface diusion of MM species and adatoms will aect observed phenomena in a dierent way.
The MM influences the distribution of adatoms on the surface, which consequently influences the local
growth rate. Diusion lengths of MM and adatoms can be estimated from the present experimental
results. In this view, dierent indium compositions on dierent stripes, as shown in Figure 6b, would
be an eect of MMIn diusion (InGaN growth being In-limited), which is probably between 5 and
10
µ
m, since the composition was nearly constant for stripe widths larger than 20
µ
m. Similarly,
diusion of In adatoms is probably around 3–3.5
µ
m and it would explain faster growth observed
up to 6–7
µ
m from the stripes’ edge, as shown in Figure 7. Thus, the adatom’s diusion explains the
overgrowth at the stripe edge, and the diusion of MM species explains why the concentration of
indium is higher on small stripes.
On the other hand, such explanation is based on experiments where selective masks were used,
and this is not the case of the present work. Therefore, we would like to present other arguments too.
We take into account the following eects: (i) gas-phase diusion, (ii) alloy-pulling eect of
indium due to mechanical strain, and (iii) kinetic processes on the surface. It ’is important to note that
the faster growth at the edges is present only when growth temperature is relatively low (820
C rather
than 1000
C) and when the stripe height is relatively large (4
µ
m rather than 1
µ
m). Moreover, the
eect is two times stronger for LT InGaN than for LT-GaN grown at the same conditions. There is no
eect for structures grown at 1000
C, even on 4
µ
m high stripes, and there is no eect for structures
grown on 1 µm high stripes even at 820 C.
Certainly, gas phase diusion would promote faster growth at the stripe edge. In the general case,
based on a numerical solution of the diusion equation, the distance which adatoms have to travel
through the gas before they can adsorb at the crystal surface is the smallest at the edges, as in [
40
,
51
], as
shown in Figure 8. Our own results obtained by a simple 2D, Monte Carlo simulation of the diusion
process confirms this eect.
Another point is that the faster growth of LT-InGaN compared to LT-GaN could be explained by
the reduced alloy-pulling eect for indium at the 2–3
µ
m wide area close to the stripe edge [
32
]. Partial
elastic relaxation of the structure near the edges would account for the low energy barrier for indium
adatoms to incorporate into the crystal, as shown in Figure 9.
However, the most important here are kinetic processes at the growth surface. First, based on
AFM data, we point out that the structure grows in the step-flow mode and thus the growth could be
described in the framework of the Burton-Cabrera-Frank model [
52
], as shown in Figure 10. In this
model adatoms adsorb on the growth surface from the gas phase and then diuse, before they are
caught by atomic steps and incorporate into the crystal. The second possibility is, however, that they
desorb and return to the gas phase. The latter can happen mainly in high temperatures and low miscut
conditions, i.e., when the atomic terrace width is comparable or larger than the adatom’s surface
diusion length.
Materials 2019,12, 2583 11 of 15
Materials 2019, 12, x FOR PEER REVIEW 10 of 15
Figure 6b, would be an effect of MMIn diffusion (InGaN growth being In-limited), which is probably
between 5 and 10 µm, since the composition was nearly constant for stripe widths larger than 20 µm.
Similarly, diffusion of In adatoms is probably around 3–3.5 µm and it would explain faster growth
observed up to 6–7 µm from the stripes’ edge, as shown in Figure 7. Thus, the adatom’s diffusion
explains the overgrowth at the stripe edge, and the diffusion of MM species explains why the
concentration of indium is higher on small stripes.
On the other hand, such explanation is based on experiments where selective masks were used,
and this is not the case of the present work. Therefore, we would like to present other arguments too.
We take into account the following effects: (i) gas-phase diffusion, (ii) alloy-pulling effect of
indium due to mechanical strain, and (iii) kinetic processes on the surface. It ’is important to note that
the faster growth at the edges is present only when growth temperature is relatively low (820 °C
rather than 1000 °C) and when the stripe height is relatively large (4 µm rather than 1 µm). Moreover,
the effect is two times stronger for LT InGaN than for LT-GaN grown at the same conditions. There
is no effect for structures grown at 1000 °C, even on 4 µm high stripes, and there is no effect for
structures grown on 1 µm high stripes even at 820 °C.
Certainly, gas phase diffusion would promote faster growth at the stripe edge. In the general
case, based on a numerical solution of the diffusion equation, the distance which adatoms have to
travel through the gas before they can adsorb at the crystal surface is the smallest at the edges, as in
[40] and [51], as shown in Figure 8. Our own results obtained by a simple 2D, Monte Carlo simulation
of the diffusion process confirms this effect.
Another point is that the faster growth of LT-InGaN compared to LT-GaN could be explained
by the reduced alloy-pulling effect for indium at the 2–3 µm wide area close to the stripe edge [32].
Partial elastic relaxation of the structure near the edges would account for the low energy barrier for
indium adatoms to incorporate into the crystal, as shown in Figure 9.
However, the most important here are kinetic processes at the growth surface. First, based on
AFM data, we point out that the structure grows in the step-flow mode and thus the growth could
be described in the framework of the Burton-Cabrera-Frank model [52], as shown in Figure 10. In this
model adatoms adsorb on the growth surface from the gas phase and then diffuse, before they are
caught by atomic steps and incorporate into the crystal. The second possibility is, however, that they
desorb and return to the gas phase. The latter can happen mainly in high temperatures and low
miscut conditions, i.e., when the atomic terrace width is comparable or larger than the adatom’s
surface diffusion length.
Figure 10. Different processes taken into account in the BCF model of crystal growth on vicinal
surfaces.
Second, it is important to look at the structure morphology after growth. Judging from AFM,
SEM, and STEM data, the top of the stripe is formed of (0001) plane which preserves the initial miscut
of the substrate (0.6° towards the <1–100> direction), i.e., it contains regular atomic steps, with the
average direction perpendicular to the stripe edge, separated by 20–30 nm wide terraces, as shown
in Figure 2b and inset to Figure 11a. On the contrary, the stripes’ direction was precisely aligned 90 ±
0.1° to the easy-cleavage direction of GaN, <11–20>, and the sidewalls of the stripe (which are about
6 µm wide for a 4 µm stripe height) after growth become exact-oriented, low-index (11–22) planes,
without any steps, as shown in Figure 11b and 11c. Hence, the density of sites where adatoms can
attach is relatively high on the top of the stripe, compared to the sidewalls, where there are almost
Figure 10.
Dierent processes taken into account in the BCF model of crystal growth on vicinal surfaces.
Second, it is important to look at the structure morphology after growth. Judging from AFM,
SEM, and STEM data, the top of the stripe is formed of (0001) plane which preserves the initial miscut
of the substrate (0.6
towards the <1–100>direction), i.e., it contains regular atomic steps, with the
average direction perpendicular to the stripe edge, separated by 20–30 nm wide terraces, as shown in
Figure 2b and inset to Figure 11a. On the contrary, the stripes’ direction was precisely aligned 90
±
0.1
to the easy-cleavage direction of GaN, <11–20>, and the sidewalls of the stripe (which are about 6
µ
m
wide for a 4
µ
m stripe height) after growth become exact-oriented, low-index (11–22) planes, without
any steps, as shown in Figure 11b,c. Hence, the density of sites where adatoms can attach is relatively
high on the top of the stripe, compared to the sidewalls, where there are almost no such sites. It is
important to note that on the (11–22) facet there was almost no growth, as shown in Figure 11c.
Materials 2019, 12, x FOR PEER REVIEW 11 of 15
no such sites. It is important to note that on the (11–22) facet there was almost no growth, as shown
in Figure 11c.
(a) (b) (c)
Figure 11. (a) Top-view SEM image of the structure grown on substrate with stripes. Image size 40 ×
30 µm. Inset: 2 × 2 µm AFM image acquired on the top of a 10 µm wide stripe. (b) Schematic
perspective view of the structure. The top of the stripe is formed of the vicinal (0001) plane and the
sidewalls are exact-oriented (11–22) planes. (c) Cross-sectional scanning transmission electron
microscopy (STEM) image.
Now let ’us assume that adatoms from the gas phase adsorb to both mentioned surfaces, i.e., the
(0001) and (11–22) at equal rates. After adsorption they start to diffuse randomly on the surfaces. In
the BCF theory, mean diffusion length of adatoms on the growth surface is expressed as
λ=a exp [(W
s
U
s
)/2kT] (1)
where: a—lattice parameter in the growth plane, U
s
—energy barrier for adatom to move to the next
stable position (process “2” from Figure 11), W
s
—energy barrier for adatom desorption (process “3”
from Figure 10), T—absolute temperature.
Assuming that the growth is performed in nitrogen-rich conditions (which is almost always true
in the case of MOVPE), growth kinetics will be governed by Ga and In diffusion on the surface. Mean
diffusion lengths of these adatoms on (0001) and (11–22) facets have been estimated by Ueda and
coworkers [42]. Those authors investigated InGaN grown by MOVPE on patterned substrate with a
SiO
2
mask. The growth temperature was not specified but we assume that it must have been adequate
for InGaN growth, i.e., about 800 °C. On (0001) they were 5.2 and 2.7 µm for Ga and In, respectively,
and on (11–22) they were 3.1 and 1.6 µm for Ga and In, respectively. Hence, adatoms caught on the
stripe top easily find an atomic step and attach (distance between adjacent atomic steps is 20–30 nm
there). On the other hand, adatoms caught on the stripe sidewalls continue to diffuse until they reach
the stripe edge because the surface is atomically flat and there are no atomic steps. Eventually they
desorb back to the gas phase.
The balance between the caught and desorbed number of adatoms will depend on their mean
diffusion length, which drops rapidly with temperature. At typical temperatures used for InGaN
growth, the diffusion length is a few micrometers, i.e., of the order of the sidewall width [42]. Hence
the (11–22) facets will catch adatoms and direct them towards the (0001) top facet where they can
incorporate. The conclusion here is that the larger diffusion length, the faster growth near the stripe
edge, i.e., fast growth will be promoted for low growth temperature. Faster growth at the stripe edges
will also be promoted on taller stripes, because then the sidewall surface is larger. Since the (11–22)
facet angle to the (0001) plane is 58.4 degrees [53], the (11–22) sidewall width = 1.6 times the stripe
height. This was confirmed by our results on 1 and 4µm tall stripes and by surface profiles of GaN
grown at 800 and 1000 °C.
To sum up, faster growth of InGaN and its increased indium composition at the stripe edge
could have been caused by all three factors—gas phase diffusion, the alloy-pulling effect, and kinetic
processes—and the last factor is probably the most important.
5. Conclusions
Figure 11. (a
) Top-view SEM image of the structure grown on substrate with stripes. Image size
40
×
30
µ
m. Inset: 2
×
2
µ
m AFM image acquired on the top of a 10
µ
m wide stripe. (
b
) Schematic
perspective view of the structure. The top of the stripe is formed of the vicinal (0001) plane and
the sidewalls are exact-oriented (11–22) planes. (
c
) Cross-sectional scanning transmission electron
microscopy (STEM) image.
Now let ’us assume that adatoms from the gas phase adsorb to both mentioned surfaces, i.e., the
(0001) and (11–22) at equal rates. After adsorption they start to diuse randomly on the surfaces. In the
BCF theory, mean diusion length of adatoms on the growth surface is expressed as
λ=a exp [(WsUs)/2kT] (1)
where: a—lattice parameter in the growth plane, U
s
—energy barrier for adatom to move to the next
stable position (process “2” from Figure 11), W
s
—energy barrier for adatom desorption (process “3”
from Figure 10), T—absolute temperature.
Assuming that the growth is performed in nitrogen-rich conditions (which is almost always true
in the case of MOVPE), growth kinetics will be governed by Ga and In diusion on the surface. Mean
diusion lengths of these adatoms on (0001) and (11–22) facets have been estimated by Ueda and
coworkers [
42
]. Those authors investigated InGaN grown by MOVPE on patterned substrate with a
Materials 2019,12, 2583 12 of 15
SiO
2
mask. The growth temperature was not specified but we assume that it must have been adequate
for InGaN growth, i.e., about 800
C. On (0001) they were 5.2 and 2.7
µ
m for Ga and In, respectively,
and on (11–22) they were 3.1 and 1.6
µ
m for Ga and In, respectively. Hence, adatoms caught on the
stripe top easily find an atomic step and attach (distance between adjacent atomic steps is 20–30 nm
there). On the other hand, adatoms caught on the stripe sidewalls continue to diuse until they reach
the stripe edge because the surface is atomically flat and there are no atomic steps. Eventually they
desorb back to the gas phase.
The balance between the caught and desorbed number of adatoms will depend on their mean
diusion length, which drops rapidly with temperature. At typical temperatures used for InGaN
growth, the diusion length is a few micrometers, i.e., of the order of the sidewall width [
42
]. Hence the
(11–22) facets will catch adatoms and direct them towards the (0001) top facet where they can incorporate.
The conclusion here is that the larger diusion length, the faster growth near the stripe edge, i.e., fast
growth will be promoted for low growth temperature. Faster growth at the stripe edges will also be
promoted on taller stripes, because then the sidewall surface is larger. Since the (11–22) facet angle to
the (0001) plane is 58.4 degrees [
53
], the (11–22) sidewall width =1.6 times the stripe height. This was
confirmed by our results on 1 and 4
µ
m tall stripes and by surface profiles of GaN grown at 800 and
1000 C.
To sum up, faster growth of InGaN and its increased indium composition at the stripe edge
could have been caused by all three factors—gas phase diusion, the alloy-pulling eect, and kinetic
processes—and the last factor is probably the most important.
5. Conclusions
GaN and InGaN layers, and InGaN/GaN MQW structures were grown on (0001) GaN substrates
with stripes, at dierent growth temperatures, by MOVPE. The substrate stripes were parallel to the
<1–100>direction, they were 5–100
µ
m wide, and typically 4
µ
m high (1
µ
m on the test sample).
There was no selective mask on the substrate. It was found that central wavelength of luminescence
of MQW structures depend on the stripe width and height, and also growth temperature. It can be
up to 40 nm larger on 5
µ
m wide, 4
µ
m tall stripes, compared to the flat area of the substrate (QW
growth temperature 760
C). We attribute this eect to faster growth on tall and narrow stripes. Faster
growth promotes more eective indium incorporation, and of course QWs are thicker, which both
account for the observed eect. Faster growth on tall and narrow stripes is caused mainly by kinetic
processes at the growth surface. Notably, the initial substrate miscut angle and azimuth, and their
relation to the direction of stripes is of primary importance for the observed eects. The reduced
alloy-pulling eect for indium, due to partial elastic strain relaxation at the stripe edges, could also
add to the observed eect.
Author Contributions:
Conceptualization, M.S., E.G. and M.L.; funding acquisition, M.L., V.H. and H.A.;
supervision, M.L., V.H. and H.A., writing—original draft preparation, M.S.; writing—review and editing, E.G.;
investigation, M.S., E.G., S.G., G.T., R.C., V.H., A.R., S.N., and Z.L.; software, V.H. and M.S.
Funding:
This work was supported by the Polish National Center for Research and Development, grant
no V4-JAP/1/2016, Polish National Science Center, grant no 2015/17/B/ST5/02835 and JST (Japan Science and
Technology Agency), Strategic International Collaborative Research Program, SICORP. V.H. acknowledges the
support of the project NanoCent funded by the European Regional Development Fund (ERDF), project No
CZ.02.1.01/0.0/0.0/15_003/0000485.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
Rathman, D.D.; Silversmith, D.J.; Burns, J.A. Lateral epitaxial overgrowth of silicon on SiO
2
.J. Electrochem. Soc.
1982,129, 2303–2306. [CrossRef]
Materials 2019,12, 2583 13 of 15
2.
Tsaur, B.; McClelland, R.W.; Fan, J.C.C.; Gale, R.P.; Salerno, J.P.; Vojak, B.A.; Bozler, C.O.
Low-dislocation-density GaAs epilayers grown on Ge-coated Si substrates by means of lateral epitaxial
overgrowth. Appl. Phys. Lett. 1982,41, 347–349. [CrossRef]
3.
Gale, R.P.; McClelland, R.W.; Fan, J.C.C.; Bozler, C.O. Lateral epitaxial overgrowth of GaAs by organometallic
chemical vapor deposition. Appl. Phys. Lett. 1982,41, 545–547. [CrossRef]
4.
Jastrzebski, L. Growth of electronic quality silicon over SiO
2
by epitaxial lateral overgrowth technique.
J. Electrochem. Soc. 1982,129, 2645. [CrossRef]
5.
Kato, Y.; Kitamura, S.; Hiramatsu, K.; Sawaki, N. Selective growth of wurtzite GaN and AlxGa1
xN on
GaN/sapphire substrates by metalorganic vapor phase epitaxy. J. Cryst. Growth
1994
,144, 133–140. [CrossRef]
6.
Usui, A.; Sunakawa, H.; Sakai, A.; Yamaguchi, A.A. Thick GaN epitaxial growth with low dislocation density
by hydride vapor phase epitaxy. Jpn. J. Appl. Phys. 1997,36, 899–902. [CrossRef]
7. Nakamura, S. First III–V-nitride-based violet laser diodes. J. Cryst. Growth 1997,170, 11–15. [CrossRef]
8.
Zheleva, T.S.; Nam, O.H.; Bremser, M.D.; Davis, R.F. Dislocation density reduction via lateral epitaxy in
selectively grown GaN structures. Appl. Phys. Lett. 1997,71, 2472–2474. [CrossRef]
9.
Dadgar, A.; Bläsing, J.; Diez, A.; Alam, A.; Heuken, M.; Krost, A. Metalorganic chemical vapor phase epitaxy
of Crack-Free GaN on Si (111) exceeding 1
µ
m in thickness. Jpn. J. Appl. Phys.
2000
,39, 1183–1185. [CrossRef]
10.
Gehrke, T.; Linthicum, K.J.; Preble, E.; Rajagopal, P.; Ronning, C.; Zorman, C.; Mehregany, M.; Davis, R.F.
Pendeo-epitaxial growth of gallium nitride on silicon substrates. J. Electron. Mater.
2000
,29, 306–310.
[CrossRef]
11.
Oliver, R.A.; Briggs, G.A.D.; Kappers, M.J.; Humphreys, C.J.; Yasin, S.; Rice, J.H.; Smith, J.D.; Taylor, R.A.
InGaN quantum dots grown by metalorganic vapor phase epitaxy employing a post-growth nitrogen anneal.
Appl. Phys. Lett. 2003,83, 755–757. [CrossRef]
12.
Qian, F.; Li, Y.; Gradeˇcak, S.; Park, H.G.; Dong, Y.; Ding, Y.; Wang, Z.L.; Lieber, C.M. Multi-quantum-well
nanowire heterostructures for wavelength-controlled lasers. Nat. Mater.
2008
,7, 701–706. [CrossRef]
[PubMed]
13.
Morales, F.M.; M
á
nuel, J.M.; Garc
í
a, R.; Reuters, B.; Kalisch, H.; Vescan, A. Evaluation of interpolations of
InN, AlN and GaN lattice and elastic constants for their ternary and quaternary alloys. J. Phys. D Appl. Phys.
2013,46, 245502. [CrossRef]
14.
Leszczynski, M.; Teisseyre, H.; Suski, T.; Grzegory, I.; Bockowski, M.; Jun, J.; Porowski, S.; Pakula, K.;
Baranowski, J.M.; Foxon, C.T.; et al. Lattice parameters of gallium nitride. Appl. Phys. Lett.
1996
,69, 73–75.
[CrossRef]
15.
Yoshimoto, N.; Matsuoka, T.; Sasaki, T.; Katsui, A. Photoluminescence of InGaN films grown at high
temperature by metalorganic vapor phase epitaxy. Appl. Phys. Lett. 1991,59, 2251–2253. [CrossRef]
16.
Keller,S.; Keller, B.P.; Kapolnek, D.; Mishra, U.K.; DenBaars, S.P.; Shmagin, I.K.; Kolbas, R.M.; Krishnankutty, S.
Growth of bulk InGaN films and quantum wells by atmospheric pressure metalorganic chemical vapour
deposition. J. Cryst. Growth 1997,170, 349–352. [CrossRef]
17.
Schenk, H.P.D.; de Mierry, P.; Laügt, M.; Omn
è
s, F.; Leroux, M.; Beaumont, B.; Gibart, P. Indium incorporation
above 800
C during metalorganic vapor phase epitaxy of InGaN. Appl. Phys. Lett.
1999
,75, 2587–2589.
[CrossRef]
18.
Kaneta, A.; Funato, M.; Narukawa, Y.; Mukai, T.; Kawakami, Y. Direct correlation between nonradiative
recombination centers and threading dislocations in InGaN quantum wells by near-field photoluminescence
spectroscopy. Phys. Status Solidi Curr. Top. Solid State Phys. 2006,3, 1897–1901. [CrossRef]
19.
Nakamura, S. The Roles of Structural Imperfections in InGaN-Based Blue Light-Emitting Diodes and Laser
Diodes. Science 1998,281, 956–961. [CrossRef]
20.
Jouvet, N.; Kappers, M.J.; Humphreys, C.J.; Oliver, R.A. The impact of substrate miscut on the morphology
of InGaN epitaxial layers subjected to a growth interruption. J. Appl. Phys. 2013,113, 063503. [CrossRef]
21.
Lermer, T.; Pietzonka, I.; Avramescu, A.; Brüderl, G.; Müller, J.; Lutgen, S.; Strauss, U. Interdependency of
surface morphology and wavelength fluctuations of indium-rich InGaN/GaN quantum wells. Phys. Status
Solidi Appl. Mater. Sci. 2011,208, 1199–1202. [CrossRef]
22.
Dreyer, C.E.; Alkauskas, A.; Lyons, J.L.; Speck, J.S.; Van de Walle, C.G. Gallium vacancy complexes as a
cause of Shockley-Read-Hall recombination in III-nitride light emitters. Appl. Phys. Lett.
2016
,108, 141101.
[CrossRef]
Materials 2019,12, 2583 14 of 15
23.
Risti´c, J.; Calleja, E.; Fern
á
ndez-Garrido, S.; Cerutti, L.; Trampert, A.; Jahn, U.; Ploog, K.H. On the mechanisms
of spontaneous growth of III-nitride nanocolumns by plasma-assisted molecular beam epitaxy. J. Cryst. Growth
2008,310, 4035–4045. [CrossRef]
24.
Bengoechea-Encabo, A.; Albert, S.; Müller, M.; Xie, M.Y.; Veit, P.; Bertram, F.; Sanchez-Garcia, M.A.;
Z
ú
ñiga-P
é
rez, J.; de Mierry, P.; Christen, J.; et al. Selective area growth of AlN/GaN nanocolumns on
(0001) and (11–22) GaN/sapphire for semi-polar and non-polar AlN pseudo-templates. Nanotechnology
2017,28, 365704. [CrossRef] [PubMed]
25.
Fang, H.; Yang, Z.J.; Wang, Y.; Dai, T.; Sang, L.W.; Zhao, L.B.; Yu, T.J.; Zhang, G.Y. Analysis of mass transport
mechanism in InGaN epitaxy on ridge shaped selective area growth GaN by metal organic chemical vapor
deposition. J. Appl. Phys. 2008,103, 014908. [CrossRef]
26.
Dr
ó
˙zd˙z, P.A.; Sarzy ´nski, M.; Domagała, J.Z.; Grzanka, E.; Grzanka, S.; Czernecki, R.; Marona, Ł.; Korona, K.P.;
Suski, T. Monolithic cyan—violet InGaN/GaN LED array. Phys. Status Solidi 2017,214, 1600815. [CrossRef]
27.
Franssen, G.; Suski, T.; Kry´sko, M.; Łucznik, B.; Grzegory, I.; Krukowski, S.; Khachapuridze, A.; Czernecki, R.;
Grzanka, S.; Mensz, P.; et al. Influence of substrate misorientation on properties of InGaN layers grown on
freestanding GaN. Phys. Status Solidi Curr. Top. Solid State Phys. 2008,5, 1485–1487. [CrossRef]
28.
Keller, S.; Keller, B.P.; Kapolnek, D.; Abare, A.C.; Masui, H.; Coldren, L.A.; Mishra, U.K.; Den Baars, S.P.
Growth and characterization of bulk InGaN films and quantum wells. Appl. Phys. Lett.
1996
,68, 3147–3149.
[CrossRef]
29.
Leszczynski, M.; Czernecki, R.; Krukowski, S.; Krysko, M.; Targowski, G.; Prystawko, P.; Plesiewicz, J.;
Perlin, P.; Suski, T. Indium incorporation into InGaN and InAlN layers grown by metalorganic vapor phase
epitaxy. J. Cryst. Growth 2011,318, 496–499. [CrossRef]
30.
Czernecki, R.; Kret, S.; Kempisty, P.; Grzanka, E.; Plesiewicz, J.; Targowski, G.; Grzanka, S.; Bilska, M.;
Smalc-Koziorowska, J.; Krukowski, S.; et al. Influence of hydrogen and TMIn on indium incorporation in
MOVPE growth of InGaN layers. J. Cryst. Growth 2014,402, 330–336. [CrossRef]
31.
Nagamatsu, K.; Nitta, S.; Ye, Z.; Nagao, H.; Miki, S.; Honda, Y.; Amano, H. Decomposition of trimethylgallium
and adduct formation in a metalorganic vapor phase epitaxy reactor analyzed by high-resolution gas
monitoring system. Phys. Status Solidi 2017,254, 1600737. [CrossRef]
32.
Hiramatsu, K.; Kawaguchi, Y.; Shimizu, M.; Sawaki, N.; Zheleva, T.; Davis, R.F.; Tsuda, H.; Taki, W.;
Kuwano, N.; Oki, K. The Composition Pulling Eect in MOVPE Grown InGaN on GaN and AlGaN and its
TEM Characterization. MRS Internet J. Nitride Semicond. Res. 1997,2, e6. [CrossRef]
33.
Wright, A.F. Elastic properties of zinc-blende and wurtzite AIN, GaN, and InN. J. Appl. Phys.
1997
,82, 2833–2839.
[CrossRef]
34.
Theodoropoulos, C.; Mountziaris, T.J.; Moat, H.K.; Han, J. Design of gas inlets for the growth of gallium
nitride by metalorganic vapor phase epitaxy. J. Cryst. Growth 2000,217, 65–81. [CrossRef]
35.
Harafuji, K.; Hasegawa, Y.; Ishibashi, A.; Tsujimura, A.; Kidoguchi, I.; Ban, Y.; Ohnaka, K. Complex flow and
gas-phase reactions in a horizontal reactor for GaN metalorganic vapor phase epitaxy. Jpn. J. Appl. Phys. Part
1 Regul. Pap. Short Notes Rev. Pap. 2000,39, 6180–6190. [CrossRef]
36.
Momose, T.; Kamiya, T.; Suzuki, Y.; Ravasio, S.; Cavallotti, C.; Sugiyama, M.; Shimogaki, Y. Kinetic analysis
of GaN-MOVPE via thickness profiles in the gas flow direction with systematically varied growth conditions.
ECS J. Solid State Sci. Technol. 2016,5, P164–P171. [CrossRef]
37.
Sugiyama, M.; Yasukochi, S.; Shioda, T.; Shimogaki, Y.; Nakano, Y. Reaction kinetics of GaN Metal-Organic
Vapor-Phase epitaxy analyzed by multi-scale profiles of growth rate. ECS 2009,25, 507–512.
38.
Sekiguchi, K.; Shirakawa, H.; Chokawa, K.; Araidai, M.; Kangawa, Y.; Kakimoto, K.; Shiraishi, K.
Thermodynamic analysis of trimethylgallium decomposition during GaN metal organic vapor phase
epitaxy. Jpn. J. Appl. Phys. 2018,57, 04FJ03. [CrossRef]
39.
Zhang, Z.; Fang, H.; Yao, Q.; Yan, H.; Gan, Z. Species transport and chemical reaction in a MOCVD reactor
and their influence on the GaN growth uniformity. J. Cryst. Growth 2016,454, 87–95. [CrossRef]
40.
Tanikawa, T.; Honda, Y.; Yamaguchi, M.; Amano, H. Eect of lateral vapor phase diusion during the
selective growth of InGaN/GaN MQW on semipolar and nonpolar GaN stripes. Phys. Status Solidi Appl.
Mater. Sci. 2011,208, 1175–1178. [CrossRef]
41.
Shioda, T.; Tomita, Y.; Sugiyama, M.; Shimogaki, Y.; Nakano, Y. Selective area metal–organic vapor
phase epitaxy of nitride semiconductors for multicolor emission. IEEE J. Sel. Top. Quantum Electron.
2009,15, 1053–1065. [CrossRef]
Materials 2019,12, 2583 15 of 15
42.
Ueda, M.; Hayashi, K.; Kondou, T.; Funato, M.; Kawakami, Y.; Narukawa, Y.; Mukai, T. Mechanisms of
metalorganic vapor phase epitaxy of InGaN quantum wells on GaN microfacet structures. Phys. Status Solidi
Curr. Top. Solid State Phys. 2007,4, 2826–2829.
43.
Tanaka, A.; Chen, R.; Jungjohann, K.L.; Dayeh, S.A. Strong geometrical eects in submillimeter selective area
growth and light extraction of GaN light emitting diodes on sapphire. Sci. Rep.
2015
,5, 1–12. [CrossRef]
[PubMed]
44.
Coltrin, M.E.; Willan, C.C.; Bartram, M.E.; Han, J.; Missert, N.; Crawford, M.H.; Baca, A.G. Transport,
Growth Mechanisms, and Material Quality in GaN Epitaxial Lateral Overgrowth. MRS Proc.
1998
,537, G6.9.
[CrossRef]
45.
Mitchell, C.C.; Coltrin, M.E.; Han, J. Mass transport in the epitaxial lateral overgrowth of gallium nitride.
J. Cryst. Growth 2001,222, 144–153. [CrossRef]
46.
Gotschke, T.; Schumann, T.; Limbach, F.; Stoica, T.; Calarco, R. Influence of the adatom diusion on selective
growth of GaN nanowire regular arrays. Appl. Phys. Lett. 2011,98, 103102. [CrossRef]
47.
Matsuda, Y.; Funato, M.; Kawakami, Y. Growth mechanism of Polar-Plane-Free Faceted InGaN quantum
wells. IEICE Trans. Electron. 2018,101, 532–536. [CrossRef]
48.
So, B.M.; Youn, S.B.; Im, I.T. Growth rate and composition of InGaN during InGaN/GaN quantum wells selective
area metal-organic vapor phase epitaxy considering surface diffusion. J. Ind. Eng. Chem.
2013
,19, 1747–1751.
[CrossRef]
49.
Wang, X.; Hartmann, J.; Mandl, M.; Mohajerani, M.S.; Wehmann, H.H.; Strassburg, M.; Waag, A. Growth
kinetics and mass transport mechanisms of GaN columns by selective area metal organic vapor phase epitaxy.
J. Appl. Phys. 2014,115, 163104. [CrossRef]
50.
Li, Z.; Jiang, Y.; Yu, T.; Yang, Z.; Tao, Y.; Jia, C.; Chen, Z.; Yang, Z.; Zhang, G. Analyses of surface temperatures
on patterned sapphire substrate for the growth of GaN with metal organic chemical vapor deposition.
Appl. Surf. Sci. 2011,257, 8062–8066. [CrossRef]
51.
Xiao, R.F.; Alexander, J.I.D.; Rosenberger, F. Microscopic-growth morphologies in binary systems. Phys. Rev. A
1992,45, 571–574. [CrossRef]
52.
Myers-Beaghton, A.K.; Vvedensky, D.D. Generalized Burton-Cabrera-Frank theory for growth and
equilibration on stepped surfaces. Phys. Rev. A 1991,44, 2457–2468. [CrossRef]
53.
Scholz, F.; Meisch, T.; Caliebe, M.; Schörner, S.; Thonke, K.; Kirste, L.; Bauer, S.; Lazarev, S.; Baumbach, T. Growth
and doping of semipolar GaN grown on patterned sapphire substrates. J. Cryst. Growth
2014
,405, 97–101.
[CrossRef]
©
2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
... However, one should be aware of the fact that nitrides are very difficult to grow and process. The popularity of nitride semiconductors is reflected in the Special Issue in papers of R. Czernecki et al. [1], Y.L. Casallas-Moreno et al. [2], M. Sarzynski et al. [3], Akira Kusaba et al. [4], and Takeshi Ohgaki et al. [5]. ...
... In this case, the strain can be accommodated by elastic deformation. This issue is discussed in the paper of Y.L. Casallas-Moreno et al. [2] for InN nanocolumns on AlN/Si substrates and of M. Sarzynski et al. [3] for InGaN quantum wells on narrow stripes made on GaN substrates. ...
Article
Full-text available
In this Special Issue, we have 10 excellent papers on epitaxy. In this editorial preface, I will make comments on the following issues: (1) applications of the materials examined, (2) lattice mismatch, (3) epitaxial growth methods used, (4) characterization methods used, (5) material problems: solved and still to be solved. The “Advances in Epitaxial Materials” has a big advantage of having, in one issue, papers on different materials, but in every paper the reader should find interesting information on epitaxial growth and characterization.
... There is a redshift in the emission wavelength with the increasing number of SSPM-L AlN/GaN pairs: thus, we consider that the most favourable indium (In) composition probably corresponds to the structure with 120 pairs of SSPM-L AlN/GaN, which exhibits the lowest tensile strain, as shown in Fig. 8. It is challenging to incorporate In in growing GaN-based LEDs, because excessive TMI flow during the growth process induces a compressive strain state in the InGaN/GaN layer 50,51 . Our results suggest that growing an a-GaN layer with a large number of SSPM-L AlN/GaN pairs would provide a tensile strain state for In incorporation that facilitates a near-relaxation strain state within the epi layer. ...
Article
Full-text available
We demonstrated high-quality single crystalline a-plane undoped-gallium nitride grown on a nonpatterned r-plane sapphire substrate via metal–organic chemical vapor deposition. The effect of four different numbers of sandwiched strain-periodic AlN/GaN multilayers on the strain state, crystal quality, optical and electrical properties was investigated. Field emission scanning electron microscopy and atomic force microscopy showed that the surface morphology was improved upon insertion of 120 pairs of AlN/GaN thin layers with a root-mean-square roughness of 2.15 nm. On-axis X-ray ω-scan rocking curves showed enhanced crystalline quality: the full width at half maximum decreased from 1224 to 756 arcsec along the [0001] direction and from 2628 to 1360 arcsec along the [1–100] direction for a-GaN grown with 120 pairs of AlN/GaN compared to a-GaN without AlN/GaN pairs. Reciprocal space mapping showed that a-plane GaN with a high number of AlN/GaN pairs exhibits near-relaxation strain states. Room-temperature photoluminescence spectra showed that the sample with the highest number of AlN/GaN pairs exhibited the lowest-intensity yellow and blue luminescence bands, indicating a reduction in defects and dislocations. The a-plane InGaN/GaN LEDs with 120 pairs of SSPM-L AlN/GaN exhibited a significant increase (~ 250%) in light output power compared to that of LEDs without SSPM-L AlN/GaN pairs.
... Growth conditions are determined by individual and mutual dependences among the pressure and temperature distributions in the reactor chamber; the type, ratio, and flow rate of the precursors and carrier gasses; as well as some other factors. The process is even more difficult to control if the whole structure-with different layers, various doping, and quantum wells [8,9]-is required, or if selective area growth is considered [10][11][12][13]. Furthermore, the temperature measurements in the reactor chamber are limited due to the restricted access and the expected range (e.g., 600 • C to 1200 • C in the case of GaN). ...
Article
Full-text available
The present paper focuses on the high-pressure metal-organic vapor phase epitaxy (MOVPE) upside-down vertical reactor (where the inlet of cold gases is below a hot susceptor). This study aims to investigate thermo-kinetic phenomena taking place during the GaN (gallium nitride) growth process using trimethylgallium and ammonia at a pressure of above 2 bar. High pressure accelerates the growth process, but it results in poor thickness and quality in the obtained layers; hence, understanding the factors influencing non-uniformity is crucial. The present investigations have been conducted with the aid of ANSYS Fluent finite volume method commercial software. The obtained results confirm the possibility of increasing the growth rate by more than six times through increasing the pressure from 0.5 bar to 2.5 bar. The analysis shows which zones vortexes form in. Special attention should be paid to the transitional flow within the growth zone as well as the viewport. Furthermore, the normal reactor design cannot be used under the considered conditions, even for the lower pressure value of 0.5 bar, due to high turbulences.
Article
Full-text available
We carried out a kinetic analysis of metallorganic vapor phase epitaxy (MOVPE) of GaN to investigate the dependence of the growth rate on the process conditions as a function of residence time of the precursors in the reactor. The wafer was not rotated during growth, allowing us to analyze the thickness profile of the film in the direction of gas flow, and hence the dependence of the growth rate on the residence time. The growth rate is determined mainly by the concentration of the growth species and mass transfer of the growth species to the wafer surface. The growth rate peaked in the flow direction, and the position of this peak could, in most cases, be explained by considering a combination of the linear gas velocity and the time constant for vertical diffusion of trimethylgallium (TMGa) and/or growth species across the NH3 feed stream to the wafer surface. In some cases this was not possible, indicating that more complex effects were significant. This work is expected to contribute to understanDing of the reaction pathways for GaN-MOVPE, and the growth rate data reported here are expected to provide useful benchmarks for growth simulations that combine computational fluid dynamics and reaction models.
Article
Full-text available
Advanced semiconductor devices often utilize structural and geometrical effects to tailor their characteristics and improve their performance. We report here detailed understanding of such geometrical effects in the epitaxial selective area growth of GaN on sapphire substrates and utilize them to enhance light extraction from GaN light emitting diodes. Systematic size and spacing effects were performed side-by-side on a single 2" sapphire substrate to minimize experimental sampling errors for a set of 144 pattern arrays with circular mask opening windows in SiO2. We show that the mask opening diameter leads to as much as 4 times increase in the thickness of the grown layers for 20 μm spacings and that spacing effects can lead to as much as 3 times increase in thickness for a 350 μm dot diameter. We observed that the facet evolution in comparison with extracted Ga adatom diffusion lengths directly influences the vertical and lateral overgrowth rates and can be controlled with pattern geometry. Such control over the facet development led to 2.5 times stronger electroluminescence characteristics from well-faceted GaN/InGaN multiple quantum well LEDs compared to non-faceted structures.
Article
The growth mechanisms of three-dimensionally (3D) faceted InGaN quantum wells (QWs) on (1122) GaN substrates are discussed. The structure is composed of (1122), {1101}, and {1100} planes, and the cross sectional shape is similar to that of 3D QWs on (0001). However, the 3D QWs on (1122) and (0001) show quite different inter-facet variation of In compositions. To clarify this observation, the local thicknesses of constituent InN and GaN on the 3D GaN are fitted with a formula derived from the diffusion equation. It is suggested that the difference in the In incorporation efficiency of each crystallographic plane strongly affects the surface In adatom migration.
Article
We analyzed the decomposition of Ga(CH3)3 (TMG) during the metal organic vapor phase epitaxy (MOVPE) of GaN on the basis of first-principles calculations and thermodynamic analysis. We performed activation energy calculations of TMG decomposition and determined the main reaction processes of TMG during GaN MOVPE. We found that TMG reacts with the H2 carrier gas and that (CH3)2GaH is generated after the desorption of the methyl group. Next, (CH3)2GaH decomposes into (CH3)GaH2 and this decomposes into GaH3. Finally, GaH3 becomes GaH. In the MOVPE growth of GaN, TMG decomposes into GaH by the successive desorption of its methyl groups. The results presented here concur with recent high-resolution mass spectroscopy results.
Article
In this study, by using time-of-flight (TOF) high-resolution and high-sensitivity mass spectrometry measurements, the decomposition of trimethylgallium into dimethylgallium and monomethylgallium along with its adduct formation with NHx in a commercially available horizontal metalorganic vapor-phase epitaxy (MOVPE) reactor with a resistive heater was investigated and observed in more detail than previously reported. The results confirmed the use of the TOF monitoring system in analyzing the elementary reaction process in actual MOVPE systems.
Article
Despite the strong interest in optoelectronic devices working in the deep ultraviolet range, no suitable low cost, large-area, high-quality AlN substrates are available up to now. The aim of this work is the selective area growth of AlN nanocolumns by plasma assisted molecular beam epitaxy on polar (0001) and semi-polar (11-22) GaN/sapphire templates. The resulting AlN nanocolumns are vertically oriented with semi-polar {1-103} top facets when grown on (0001) GaN/sapphire, or oriented at 58° from the template normal and exposing {1-100} non-polar top facets when growing on (11-22) GaN/sapphire, in both cases reaching filling factors ≥ 80 %. In these kinds of arrays each nanostructure could function as a building block for an individual nano-device or, due to the large filling factor values, the overall array top surfaces could be seen as a quasi (semi-polar or non-polar) AlN pseudo-template.
Article
In the case of InGaN alloys grown by metalorganic vapour phase epitaxy on a c-plane GaN, indium content decreases as the substrate miscut is increased. This phenomenon has been previously used to fabricate laser diodes with variable wavelength on one chip [Appl. Phys. Express 5, 021001 (2012)]. In that work, however, wavelength variation was only 5 nm. In the present work we show independent, electrically driven array of light emitting diodes (LED), covering 40 nm emission wavelength range on one chip. This is achieved by a particular patterning technique, which enables the change in the local miscut of the substrate by introducing large enough slopes for practical devices. This technological approach offers a new degree of freedom for InGaN/GaN bandgap modification and device engineering. It can be applied to freestanding GaN as well as to GaN/sapphire templates used for mass production of LEDs. Once optimized, this approach could eventually lead to truly monolithic RGB LEDs.
Article
Fluid flow, heat transfer, and species transport with chemical reactions have been investigated for gallium nitride (GaN) growth in a commercial metal-organic chemical vapor deposition (MOCVD) reactor. Both the growth rate and the growth uniformity are investigated zone by zone, as the wafers are divided into three zones/groups according to their distances to the susceptor center. The results show that species transport in the reactor is affected by the inlet conditions, i.e., the premixed or non-premixed inlet, the inlet temperature, the total gas flow rate, and the V/III component ratio, and reveal that the premixed inlet condition is preferred for uniform growth. Especially, a large total flow rate or a low V/III ratio results in both increase of the growth rate and improvement of the growth uniformity.
Article
We describe a mechanism by which complexes between gallium vacancies and oxygen and/or hydrogen act as efficient channels for nonradiative recombination in InGaN alloys. Our identification is based on first-principles calculations of defect formation energies, charge-state transition levels, and nonradiative capture coefficients for electrons and holes. The dependence of these quantities on alloy composition is analyzed. We find that modest concentrations of the proposed defect complexes (∼1016 cm−3) can give rise to Shockley-Read-Hall coefficients s−1. The resulting nonradiative recombination would significantly reduce the internal quantum efficiency of optoelectronic devices.A=(107−109)
Article
The microstructure and the lateral epitaxy mechanism of formation of homoepitaxially and selectively grown GaN structures within windows in SiO2 masks have been investigated by transmission electron microscopy (TEM) and scanning electron microscopy. The structures were produced by organometallic vapor phase epitaxy for field emission studies. A GaN layer underlying the SiO2 mask provided the crystallographic template for the initial vertical growth of the GaN hexagonal pyramids or striped pattern, The SiO2 film provided an amorphous stage on which lateral growth of the GaN occurred and possibly very limited compliancy in terms of atomic arrangement during the lateral growth and in the accommodation of the mismatch in the coefficients of thermal expansion during cooling, Observations with TEM show a substantial reduction in the dislocation density in the areas of lateral growth of the GaN deposited on the SiO2 mask. In many of these areas no dislocations were observed. (C) 1997 American Institute of Physics. [S0003-6951(97)00843-7].