Content uploaded by Jeong-Hwan Park
Author content
All content in this area was uploaded by Jeong-Hwan Park on May 16, 2024
Content may be subject to copyright.
Recent advances in micro-pixel light emitting
diode technology
Cite as: Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550
Submitted: 22 September 2023 .Accepted: 19 April 2024 .
Published Online: 16 May 2024
Jeong-Hwan Park,
1,a)
Markus Pristovsek,
2
Hiroshi Amano,
1,2
and Tae-Yeon Seong
2,3,a)
AFFILIATIONS
1
Deep Tech Serial Innovation Center Laboratory, Nagoya University, Furo-Cho, Chikusa-ku, 464-8603 Nagoya, Japan
2
Institute of Materials and Systems for Sustainability, Nagoya University, Furo-Cho, Chikusa-ku, 464-8601 Nagoya, Japan
3
Department of Materials Science and Engineering, Korea University, 02841 Seoul, Korea
a)
Authors to whom correspondence should be addressed: jh1490.park@gmail.com and tyseong@korea.ac.kr
ABSTRACT
Display technology has developed rapidly in recent years, with III–V system-based micro-light-emitting diodes (lLEDs) attracting attention as
a means to overcome the physical limitations of current display systems related to their lifetime, brightness, contrast ratio, response time, and
pixel size. However, for lLED displays to be successfully commercialized, their technical shortcomings need to be addressed. This review com-
prehensively discusses important issues associated with lLEDs, including the use of the ABC model for interpreting their behavior, size-
dependent degradation mechanisms, methods for improving their efficiency, novel epitaxial structures, the development of red lLEDs,
advanced transfer techniques for production, and the detection and repair of defects. Finally, industrial efforts to commercialize lLED displays
are summarized. This review thus provides important insights into the potential realization of next-generation display systems based on lLEDs.
V
C2024 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (https://
creativecommons.org/licenses/by/4.0/).https://doi.org/10.1063/5.0177550
TABLE OF CONTENTS
I. INTRODUCTION. ................................. 1
II. INTERPRETATION OF lLED CHARACTERISTICS
USING TRADITIONAL ABC MODEL . . ............. 2
III. INTRODUCING EFFICIENCY DEGRADATION
OF lLEDS . . ..................................... 4
A. Non-radiative recombination of lLEDs ......... 4
B. Size-dependent EQE behaviors of InGaN-based
blue, green, red and AlGaInP-based red lLEDs . . 7
C. Quantitative analysis for sidewall damage of
lLEDs . ..................................... 8
IV. EFFECTIVE PASSIVATION OF SIDEWALL
SURFACE . . ..................................... 9
V. EFFICIENCT LIGHT EXTRACTION FROM lLEDS. . . 11
VI. NOVEL EPITAXIAL STRUCTURES FOR
III-NITRIDE lLEDS . ............................. 13
VII. CHALLENGING RED lLEDS ..................... 14
A. Diffusion length of III–V...................... 14
B. Sidewall surface recombination of AlGaInP
red lLEDs. . ................................. 15
C. Importance of growth conditions of InGaN red
lLEDs . ..................................... 17
D. State-of-the-art AlGaInP and InGaN red LEDs. . . 20
VIII. ADVANCED TRANSFER TECHNOLOGIES
FOR PRODUCTION . ........................... 21
IX. DETECTION AND REPAIR PROCESSES. ........... 24
X. INDUSTRY EFFORTS . . ........................... 24
XI. CONCLUSION AND OUTLOOK................... 25
I. INTRODUCTION
Recent advanced display technologies such as liquid crystal dis-
plays (LCDs) and organic light-emitting diodes (OLEDs) have revolu-
tionized the display industry, but limitations associated with their
lifetime, brightness, response time, and contrast ratio have hindered
the realization of next-generation displays.
1–3
These next-generation
displays include augmented reality (AR), virtual reality (VR), mixed
reality (MR), and head-up displays (HUDs), which allow for informa-
tion to be accessed rapidly anywhere but must meet high specifications
to do so. For example, they should be able to operate for a long time
with high efficiency and under bright conditions (e.g., direct sunlight),
employ ultra-small pixels for high resolution, and emit full RGB
color.
4–6
These requirements thus require more advanced light sources.
In this context, III–V light-emitting diodes (LEDs) have attracted
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-1
V
CAuthor(s) 2024
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
16 May 2024 21:33:15
attentionasthelightsourceformicro-LED(lLED) displays because
they have already been employed as solid-state light sources for
reduced energy consumption compared with fluorescent and halogen
lamps while exhibiting superior physical performance to LCDs and
OLEDs.
7–9
Inorganic III–VlLEDs can be fabricated in various sizes down to
1lm
2
[20 000 pixels per inch (PPI)] using a typical top-down strat-
egy with inductive-coupled plasma-reactive ion etching (ICP-RIE).
The easily controllable size of these LEDs allows for the production of
ultrahigh-resolution displays (i.e., with a high PPI). Because the human
eye can perceive high-resolution images when the display is close to
the eye, an ultrahigh-resolution display system is required for AR, VR,
and MR applications.
10
However, the performance of lLEDs degrades
as their size decreases due to sidewall damage and carrier diffusion.
11,12
Solving this is the main priority when seeking to fabricate highly effi-
cient lLED displays.
In principle, alloying GaN and InN in InGaN lLEDs can produce
an emission wavelength that covers the entire visible range, including
the RGB colors, thus enabling the fabrication of full-color lLED dis-
plays. Since the emergence of GaN-based blue LEDs 30 years ago,
many researchers have sought to optimize the external quantum effi-
ciency (EQE) of blue LEDs.
13–15
InGaN alloys with a high indium con-
tent can reduce the bandgap energy, leading to the emission of longer
wavelengths (i.e., green and red). However, when controlling the num-
ber of defects in InGaN systems using a high indium content, a low-
efficiency issue known as the green gap arises.
16,17
This green gap has
motivated the search for alternative red sources such as AlGaInP-
based red lLEDs. However, these suffer from a more dramatic loss of
efficiency with smaller chip sizes than that faced by InGaN-based red
lLEDs due to the longer diffusion length of the carriers.
18
Due to these
issues associated with the chip size, it remains unclear whether
AlGaInP or InGaN-based red lLEDs are best suited for use in lLED
displays.
To produce high-performance display systems, the light source
for the lLEDs (referred to as the frontplane) should operate with a
backplanesuchasthinfilmtransistors(TFTs),
19,20
high electron
mobility transistors,
21,22
or complementary metal-oxide semiconduc-
tors (CMOS)
23,24
that contains a circuit with appropriate capacitance
for charging. A key factor for realizing a lLED-based display system is
to develop a precise transfer technique with a high transfer rate, reli-
ability, and scalability in the backplane.
25,26
This goal is currently being
pursued by major display companies using novel technologies such as
electrostatic, elastomer, magnetic, fluidic, laser, and roll-to-roll meth-
ods with the aim of commercializing lLED displays.
In this review, we comprehensively discuss the major issues asso-
ciated with the fabrication of lLEDs and their use in display systems.
First, the traditional ABC model for LEDs, which is still used to inter-
pret lLED properties, is introduced. We then review in detail the
mechanisms involved in the loss of efficiency suffered by lLEDs in
terms of carrier behavior. Subsequently, we outline the solutions that
have been proposed to overcome this degradation in efficiency, with a
focus on the sidewall conditions, light extraction efficiency (LEE), and
epitaxial structure. Next, we introduce red emission source candidates
for lLED displays, such as AlGaInP and InGaN systems, and compare
their advantages and disadvantages. In particular, we summarize the
variables that determine their performance, such as the diffusion
length of the carrier, the number of defects, strain, and quantum well
(QW) structures. Next, we review recently reported mass transfer
methods for lLED displays. We also describe methods for detecting
and repairing defective pixels to advance the commercialization of
lLED displays. Finally, we present industry efforts on lLED displays
to provide insight into what is needed to commercialize lLED
displays.
II. INTERPRETATION OF lLED CHARACTERISTICS
USING TRADITIONAL ABC MODEL
The ABC model has been reviewed several times, with the most
recent complete treatment produced by David et al.
27
Thus, this brief
section only summarizes the general principles as applied to lLEDs.
The maximum efficiency of an LED is determined by its internal quan-
tum efficiency (IQE), which refers to the number of photons produced
per recombination in the active region as a function of the carrier den-
sity. Thus, the upper limit is one photon per injected carrier pair, i.e.,
IQE 1. To calculate the IQE, three main processes are assumed to
occur in the active region: (1) non-radiative Shockley–Read–Hall
(SRH) recombination, whose probability is proportional to the density
of the non-radiative recombination sites (point and extended defects)
and the carrier density n, (2) radiative recombination, which is propor-
tional to the probability of two carriers overlapping (n
2
), and (3)
Auger recombination, which is proportional to the three-carrier pro-
cess (n
3
).TheIQEcanthusbecalculatedusing
IQE ¼Photons
Recombinations ¼Bn2
An þBn2þCn3;(1)
with nthe carrier density, A(s
1
) is the non-radiative recombination
rate, B(cm
3
⸱s
1
) is the radiative recombination rate, and
C(cm
6
⸱s
1
] is the Auger recombination rate. These constants are
not strictly constants in the technical sense of the word, especially the
radiative recombination constant, which is affected by the screening of
charges at higher carrier densities and band-filling effects in InGaN
LEDs,
28
while a correlation can occur between Band both Aand C:
27
The use of ABC constants that depend on the carrier density compli-
cates practical analysis, so many papers restrict their analysis to small
carrier density windows, and assume that the change in the carrier
density is much larger than these proportional constants. The resulting
experimentally accessible value is the EQE
EQE ¼Photons at detector
carriers injected
¼L
Ie0
hv ¼geginj
Bn2
An þBn2þCn3¼geginjIQE;(2)
where Lis the optical light output, e
0
is the elementary charge, his
Planck’s constant, vis the frequency, Iis the current, geis the LEE, and
ginj is the carrier injection efficiency. In this formulation, the most lim-
iting factor is ge. Due to the high refractive index of GaN (2.4), it can
be very low. This means that, because the critical angle is only 24,a
significant proportion of the light will be reflected back and ultimately
lost if no measures are employed to improve light extraction. In partic-
ular, the LEE for LEDs measured on-wafer, which is when the contact
is deposited but not packaged, can be reduced to less than 20% for
smooth layers even with large-area detectors. However, for lLEDs,
shrinking the size increases the LEE
28–30
because light is not only
extracted from the top surface but also from the sidewalls in increas-
ingly larger proportions [Fig. 1(a)]. When geis reduced by 20%, the
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-2
V
CAuthor(s) 2024
16 May 2024 21:33:15
EQE decreases by twice the non-radiative recombination rate (A).
Accordingly, LED companies have sought to improve the LEE, typi-
cally by roughening the surface, using thin-film flip chips, shaping the
packages, and/or other often closely guarded methods.
8
When seeking to optimize LEDs, the IQE is generally the initial
target because it ultimately limits the number of photons. Thus, the
accurate measurement of the IQE of an LED is vital, but this is not
straightforward because geis unknown and the direct fitting of Eq. (2)
is not possible. Only the current (I), which is equal to the number of
recombinations, and the photon flux (L), which is proportional to the
square of the carrier density, can be measured. The carrier density,
which is not equal to the current, is also required. The number of
detected photons is proportional to n
2
, so the EQE can be plotted
against L¼geginjBn2and fit with
EQE L
ðÞ
¼geginj
Pþ2
Pþffiffiffiffiffiffiffiffiffi
L
Lpeak
sþffiffiffiffiffiffiffiffiffi
Lpeak
L
r;(3)
which yields geginj ,L
peak
, and the parameter P. At the maximum IQE,
the first derivative of both IQE and EQE is zero, which gives the carrier
density at peak efficiency assuming constant ABC parameters
npeak ¼ffiffiffiffi
A
C
r¼ffiffiffiffiffiffiffiffiffi
Lpeak
geB
s:(4)
The parameter PgivesthenthepeakIQE
IQEpeak ¼P
Pþ2¼B
Bþ2ffiffiffiffiffiffiffi
AC
p:(5)
This was first introduced by Dai et al.
31
In principle, the IQE for differ-
ent light outputs can also be obtained using special plots.
32
In any case,
it is important to know which of the three ABC factors is responsible
for the change in the IQE. Because P¼B
ffiffiffiffiffi
AC
p, only the relationship
between A,B,andCcan be calculated using P. However, absolute val-
ues cannot be obtained from EQE measurements without another
independent measurement of the carrier lifetime or the use of modu-
lated currents. However, additional indications can be obtained from
other sources. For example, because Band Care mostly given by
geometry and the materials used, Ais a particular focus, especially
because it may also vary depending on the processing and size of the
lLEDs. Thus, the current density Jpeak at peak EQE can be determined.
Calculating Jpeak from Eq. (2) at npeak gives the following:
Jpeak ¼ginje0dA B
Cþ2ffiffiffiffi
A
C
r
!
;(6)
where dis the total thickness of the active area. In extreme cases, if Ais
very large, then the second term in the parentheses (2 ffiffiffiffiffiffiffiffiffi
A=C
p)islarger
and Jpeak ginje0d2ffiffiffiffiffiffiffiffiffiffiffi
A3=C
p.Conversely,ifAis very small, then the
first term will be dominant, giving Jpeak ginje0dAB=C. In between,
changes in the non-radiative recombination rate Awill cause the cur-
rent density to change slightly more than linearly at the peak EQE,
Jpeak Aqwith q¼11:5:(7)
As shown in Fig. 1(b), for a typical set of Band C, the exponent qin Eq.
(7) should be close to 1.3, that is, both halves of the parentheses in Eq.
(7) contribute. Additionally, using other values for ABC reported in the
literature, such as B¼10
11
cm
3
⸱s
1
and C¼10
29
cm
6
⸱s
1
,yields
similar exponents. The reduced size means that a greater number of car-
riers are located near the sidewalls. At the surface, they undergo
non-radiative recombination through surface states (i.e., sidewall recom-
bination). Although the surface states are limited to the first few atomic
layers near the surface, these surface states typically pin the Fermi level
inside the bandgap. Hence, the surface state causes the surface band
bending to extend much further into the lLEDs, allowing carriers (and
especially holes) to efficiently diffuse laterally to the sidewalls and recom-
bine non-radiatively with surface recombination rate A
s
.
33
The sidewalls
therefore make an additional contribution to the non-radiative recombi-
nation rate. Assuming that all carriers within a certain distance kfrom
the edge are swept to the sidewall via band bending and then recombine
at the sidewall with non-radiative surface recombination rate A
s
,the
total non-radiative recombination rate Aused in Eq. (1) for the IQE is
given by the sum of the bulk non-radiative recombination rate A
o
and
the sidewall surface term as follows:
A¼A0þAskl
W;(8)
where lis the peripheral length of the LED and Wis the area of the
LED. Figure 1(c) describes the impact of sidewall surface
FIG. 1. (a) Calculated EQE for a typical set of a good blue LED with increasing non-radiative recombination rate Ausing typical values B1012 cm
3
⸱s
1
, and
C1031 cm
6
⸱s
1
. (b) Numerical calculated dependence of Jpeak as a function of the non-radiative recombination rate using a fixed set of B and C using equation (Jpeak ).
Fitting the slope gives Jpeak A1:3for a typical set of ABC coefficients. (c) Calculated impact sidewall recombination, i.e., lk
Wfor square LEDs and assuming k¼0.1, 1, and 10 lm.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-3
V
CAuthor(s) 2024
16 May 2024 21:33:15
recombination as a function of device width variation k.Itisunder-
stood that the effect of sidewall recombination becomes stronger as the
LED width decreases. Additionally, because A is proportional to k
according to Eq. (8),increasingkconsequently increases the sidewall
recombination. Thus, if the size of the lLEDs is sufficiently small, the
sidewall recombination rate Askl
Wdominates all non-radiative recombi-
nation, unless A
s
or kis significantly reduced by smoothing or passiv-
ation, as discussed later.
When comparing differently sized lLEDs processed on the same
wafer, it can be assumed that Band Cremain the same. The QW can
relax laterally only under 2 lm, releasing strain and changing B
directly. Thus, above a width of 2 lm, Jpeak should behave as follows:
Jpeak A0þAskl
W
1:3
;(9)
which allows the effect of sidewall recombination to be assessed. The
exact contribution of surface recombination and carrier catchment dis-
tance kis difficult to quantify. However, if the lateral movement of the
carriers can be reduced (e.g., by increasing background doping) then k
can be reduced (Fig. 11). Chemical passivation would change both As
and kbecause the surface states can shift their energy position or expe-
rience a decrease in density, both of which will affect surface band
bending and thus k. Therefore, Askis also called the surface recombi-
nation velocity and is discussed together. In conclusion, the optimiza-
tion of EQE strongly focuses on two topics: increasing the LEE and
reducing losses due to the sidewalls. For this purpose, fitting the EQE
to obtain the IQE and analyzing the peak current density is a suitable
methodology.
III. INTRODUCING EFFICIENCY DEGRADATION
OF lLEDS
A. Non-radiative recombination of lLEDs
When determining the performance of lLEDs, it is important to
monitor the defects. Across the entire lLED fabrication process, from
epitaxial growth to processing, two types of defects are considered par-
ticularly deleterious for the performance of lLEDs: bulk defects from
the epitaxial structure and surface defects from etching damage at the
sidewall. Bulk defects, which broadly include dislocation and trench
defects in III-nitride, can increase non-radiative recombination. This
type of defect typically occurs during epitaxial growth and device proc-
essing for several reasons. For example, the lattice mismatch between
layers can create dislocations. It is well known that the dislocation den-
sity for GaN (10
8
cm
2
) arises from the difference between the lattice
constant of GaN (3.19 ˚
A) and that of sapphire (2.75 ˚
A) and/or Si (111)
(3.84 ˚
A). These dislocations can act as non-radiative recombination
centers, resulting in carrier and efficiency loss. Dai et al.
34
investigated
the impact of dislocation density on the IQE and systemically com-
pared the performance of devices with dislocation densities of
5.3 10
8
,1.210
9
, and 5.7 10
9
cm
2
as a function of the carrier
concentration [Fig. 2(a)]. A lower dislocation density guarantees a
higher IQE and a lower non-radiative recombination rate. Lahnemann
et al.
35
also systemically investigated the intensity of cathodolumines-
cence (CL) near a threading dislocation at room temperature. Figure
2(b) presents intensity profiles of the CL across the threading disloca-
tion. The intensity of CL closer to the threading dislocation decreased
remarkably, with the intensity difference between the location with
the threading dislocation and the location without almost 40%.
This indicates that the threading dislocation can act as a non-radiative
recombination center and consequently promote efficiency degrada-
tion. This supports the finding that dislocations can aggravate the IQE
presented in Fig. 2(a). These previous studies indicate that defects
induced by the dislocation density need to be overcome to produce
highly efficient lLEDs.
AnothercauseofbulkdefectsisthatanInGaNlayerwithahigh
indium content promotes the emission of longer wavelengths because
it has a narrower bandgap than GaN and AlN alloys. The growth con-
ditions for InGaN LEDs with a high-In InGaN layer can produce lon-
ger wavelengths, leading to the generation of In platelets and/or voids
and thus trench defects [Fig. 2(c)]. Massabuau et al.
36
experimentally
observed thermal degradation in the active region due to the growth of
a p-type layer that produced indium platelets/voids and trench defects
at those locations. These trench defects can significantly decrease the
performance of LEDs (via carrier loss), associated with an increase in
non-radiative recombination in the bulk. In particular, as shown in
Fig. 2(d), trench defects in InGaN/GaN LEDs can greatly reduce the
efficiency, which is consistent with the green gap.
This degradation due to trench defects (LED 2) was more severe
in decreasing the drive current. In addition, the efficiency of LED 2
was lower than without trench defects (LED 1) across the whole drive
current. These results provide evidence that trench defects can act as
non-radiative recombination centers and degrade efficiency.
Similarly, Kirilenko et al.
37
experimentally observed that a partial
area of an InGaN QW with a high indium content emitted wave-
lengths over 600nm, with the relatively low indium content of this
area originating from the decomposition of the InGaN QW potentially
acting as a defect. This partially disrupted InGaN QW originated from
the decomposition itself while the capping and barrier layers were
grown. Therefore, it had a lower indium content than the rest of the
InGaN QW, resulting in the emission of an additional peak at approxi-
mately 470 nmand increasing the total number of defects.
Trench defects can also be generated when the growth tempera-
ture of multi quantum wells (MQWs) is sufficiently low. Smalc-
Koziorowska et al.
38
systemically investigated the mechanisms
underlying the formation of trench defects in MQWs. They grew quan-
tum barriers (QBs) at temperatures of 730, 830, and 880Cwiththe
QW growth temperature fixed at 730 C for all samples. They observed
a substantial number of trench defects with QB growth at 730 C
[Fig. 2(e)], while QB growth at 830 C had a smaller number of trench
defects (data not shown here). On the other hand, no trench defects
were observed at 880 C[Fig. 2(f)]. The generation of trench defects at
a low temperature was attributed to a basal stacking fault. These results
suggest that the optimization of growth conditions to suppress addi-
tional defects such as threading dislocations and trench defects is
required to achieve the fabrication of high-performance lLEDs.
Another issue is the increase in the number of sidewall defects. In
the past, improving the efficiency of LEDs focused on reducing the
defects in the bulk via growth techniques because the typical LED size
(>300 300 lm
2
) prevented damage during the ICP-RIE process.
However, in recent years, LED technology has shifted to smaller LEDs
(i.e., lLEDs), leading to the size-dependent performance of LEDs
related to sidewall defects.
39
Thechipsizecontributingtothedisplay
resolution should be small because the human eye can distinguish dis-
play quality when it is close.
40,41
A typical dry etching process using
ICP-RIE allows the chip size of InGaN-based LED to be easily
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-4
V
CAuthor(s) 2024
16 May 2024 21:33:15
controlled to below 10 lm, which is desirable for use in lLED displays.
However, ICP-RIE, which is conducted with plasma, can unintention-
ally induce surface damage at the sidewalls of lLEDs and thus degrade
their efficiency.
42
Therefore, understanding the mechanisms responsi-
ble for sidewall recombination is a key factor for the production of
highly efficient lLEDs, particularly in terms of reducing the power
consumption.
In line with this, Park et al.
43
recently observed the width of side-
wall damage and the atomic structure near the sidewall. Figure 3(a)
presents an HAADF-STEM image of lLEDs, revealing plasma-
induced damage at the sidewalls. The width of sidewall damage
marked by the yellow arrow in the figure is a few nanometers in scale,
as also seen in the high-magnified STEM image showing lattice distor-
tion. This lattice distortion indicates that the sidewall damage is more
severe closer to the sidewall. However, after chemical treatment using
tetramethylammonium hydroxide (TMAH), which is based on OH–,
the sidewall damage could be fully removed [Fig. 3(a)]. Yamada et al.
44
also investigated lattice distortion at the sidewall, demonstrating that
lattice distortion induced by ICP-RIE can introduce surface states (i.e.,
surface band bending) that can degrade device performance. In partic-
ular, surface states can induce the movement of carriers to the sidewall,
where they recombine non-radiatively, thus reducing efficiency.
Therefore, surface states can act as non-radiative recombination cen-
ters, and these centers have a stronger effect for carriers with a longer
diffusion length, which determines how many carriers reach the
sidewall.
45
The diffusion length of a carrier depends on not only the material
properties but also the dislocation density and injection current density
(i.e., the carrier density). Karpov et al.
46
computationally confirmed
that the minority carrier diffusion length in GaN could vary as a func-
tion of the dislocation density [Fig. 3(b)]. In other words, the diffusion
length of a carrier becomes shorter with a higher dislocation density,
which could be attributed to the carriers predominantly recombining
at non-radiative centers caused by dislocations. Furthermore, using
FIG. 2. (a) Internal quantum efficiency as a function of injected carrier concentration with various threading dislocation density. Reproduced from Dai et al., Appl. Phys. Lett. 94,
111109 (2009), with permission from AIP Publishing.
31
(b) Normalized CL intensity near the threading dislocation demonstrating that threading dislocation kills efficiency.
Reprinted with permission from L€
ahnemann et al., Phys. Rev. Appl. 17, 024019 (2022). Copyright 2022 American Physics Society.
35
(c) STEM image showing trench defects
generated on the indium platelets/voids caused by growth of the p-type GaN capping layer. (d) Degradation of efficiency with trench defects. LED1 has no trench defect while
LED2 has trench defects. Reproduced from Massabuau et al., Appl. Phys. Lett. 105, 112110 (2014), with permission from AIP Publishing.
36
Growth temperature-dependent
occurrence of trench defects is shown. (e, f) STEM images of InGaN/GaN structure with QBs grown I(e) 730 C and (f) 880 C. Reproduced from Smalc-Koziorowska et al.,
Appl. Phys. Lett. 106, 101905 (2015), with permission from AIP Publishing.
38
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-5
V
CAuthor(s) 2024
16 May 2024 21:33:15
photoluminescence (PL) measurements, David et al.
47
experimentally
demonstrated that the carrier diffusion length of III-nitrides depends
on the excitation density. According to their study, the diffusion length
of a carrier increases with a lower injection current density (similar to
the excitation density) [Fig. 3(c)], which supports the size-dependent
degradation of III-nitride-based blue lLEDs and AlGaInP-based red
lLEDs.
48,49
Cho et al.
50
recently claimed that carrier localization
(which is similar in concept to a short diffusion length) due to indium
aggregation can make it more difficult for carriers to be trapped in the
sidewall. Thus, carriers with a longer diffusion length are one of the
major reasons for efficiency degradation.
Overall, sidewall surface states can act as non-radiative recombi-
nation centers and longer diffusion lengths for carriers allow a large
number of carriers to recombine non-radiatively at the sidewall sur-
face, resulting in efficiency degradation. Figure 3(d) summarizes this
process for three different cases. The first case is with a high defect
density. As presented in Fig. 3(b), a higher dislocation density (which
is the same as a higher defect density) leads to a shorter diffusion
FIG. 3. (a) STEM images showing lattice distortion induced by a typical ICP-RIE process (upper images) and after TMAH treatment for removing lattice distortion (lower
images). Reprinted with permission from Park et al., Adv. Opt. Mater. 11, 2203128 (2023). Copyright 2023 Wiley VCH.
43
(b) Minority carrier diffusion length as a function of dis-
location density. Reproduced from Karpov et al., Appl. Phys. Lett. 81, 4721 (2002), with permission from AIP Publishing.
46
(c) Images of charge-coupled-device camera observ-
ing carrier diffusion at nominal current densities of 25 (left) and 9500 (right) A/cm
2
using photoluminescence. Reprinted with permission from David et al., Phys. Rev. Appl. 15,
054015 (2021). Copyright 2021 American Physics Society.
47
(d) Mechanism of non-radiative recombination and radiative recombination with various conditions including a num-
ber of defect densities and a condition of sidewall surface.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-6
V
CAuthor(s) 2024
16 May 2024 21:33:15
length for the carriers. Furthermore, a higher dislocation density does
not allow a large number of carriers to reach the sidewall, resulting in
lower sidewall surface recombination in lLEDs.
51
Instead, carriers pre-
dominantly recombine at non-radiative recombination centers such as
dislocations and/or defects in the bulk, meaning that sidewall surface
recombination has a minimal effect.
The second case is with a low defect density, which can yield
highly efficient LEDs. Recalling the green gap, the typical size of
InGaN-based blue and AlGaInP-based red LEDs leads to a high effi-
ciency because sidewall surface recombination can be avoided.
52
However, as their size decreases, the degradation in the efficiency
becomes a serious concern
49
because a large number of carriers can
still reach the sidewall and recombine non-radiatively at a low current
density.
The third case is with a low defect density and surface treatment.
To suppress sidewall surface recombination, various strategies have
been developed, including chemical treatment and the use of passiv-
ation layers produced with atomic layer deposition (ALD).
18,53
Though these methods can improve the performance of lLEDs, effi-
ciency degradation at smaller sizes still occurs. For example, Park
et al.
51
investigated size-dependent InGaN-based blue lLEDs and
found that the EQE decreased dramatically at a low current density
even when the sidewall was chemically treated to remove sidewall
damage. This behavior indicates that sidewall surface recombination
still occurs even with the use of advanced passivation techniques. We
speculate that the decrease in EQE is associated with Fermi level
pinning, which can result in surface band bending. Indeed, it has
been discovered that the surface of non-polar GaN is intrinsically
pinned.
54–57
This suggests that some surface states could exist even
though the sidewall surface is recovered using chemical treatment and
passivation. This intrinsic pinning is a natural phenomenon for all
semiconductors, not only those of types III–V but also silicon-based
semiconductors.
58–61
Therefore, to suppress sidewall surface recombi-
nation, both sidewall treatment and the reduction of the diffusion
length of carriers should be pursued.
B. Size-dependent EQE behaviors of InGaN-based blue,
green, red and AlGaInP-based red lLEDs
The performance of lLEDs is determined by their size, with the
sidewall inducing carrier recombination non-radiatively at its surface
as a result of surface recombination, leading to the performance degra-
dation of lLEDs. Typically, size-dependent degradation occurs during
the ICP-RIE of InGaN lLEDs with a low indium content (e.g., blue
lLEDs). Smith et al.
62
experimentally observed the degradation of the
EQE of InGaN-based lLEDs with chip sizes ranging from 30 lmto
1lm. Figure 4(a) shows the degradation of blue lLEDs with a reduc-
tion in their size. Similar to Eq. (9),theJpeak of each size dramatically
increased but EQEpeak decreased, which suggested that carrier loss at
the sidewall can strongly affect the performance of lLEDs. However,
the size-dependent behavior of Jpeak and EQEpeak for green lLEDs
exhibited almost no change, suggesting that the carrier loss at the
FIG. 4. Changes of EQEs with various chip sizes as a function of injection current density (a) InGaN-based blue lLEDs and (b) InGaN-based green lLEDs. Reproduced from
Smith et al., Appl. Phys. Lett. 116, 071102 (2020), with permission from AIP Publishing.
62
(c) Comparison of surface recombination velocity between InGaN-based blue and
green LEDs. Reproduced from Kitagawa et al., Appl. Phys. Lett. 98, 181104 (2011), with permission from AIP Publishing.
63
(d) The green gap. Reprinted with permission from
Maur et al., Phys. Rev. Lett. 116, 027401 (2016). Copyright 2016 American Physics Society.
52
(e) Size-dependent EQEs of InGaN-based red lLEDs. Reprinted with permission
from Park et al., Laser Photonics Rev. 17, 2300199 (2023). Copyright 2023 Wiley VCH.
51
(f) Size-dependent EQEs for six different sizes of AlGaInP-based red lLEDs.
Reprinted with permission from Oh et al., Opt. Express 26, 11194 (2018). Copyright 2018 The Optical Society.
68
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-7
V
CAuthor(s) 2024
16 May 2024 21:33:15
sidewall was not serious, unlike that for the InGaN-based blue
lLEDs [Fig. 4(b)]. These results have been interpreted based on the
surface recombination velocity. Kitagawa et al.
63
estimated that
InGaN-based blue (450 nm) and green (520 nm) LEDs had a sur-
face recombination velocity of 5 10
3
and 3 10
2
cm/s, respectively.
This confirmed that a high-indium-content InGaN LED emitting a
longer wavelength has a lower surface recombination velocity.
Reducing the surface recombination velocity may be consistent with
reducing the diffusion length induced by carrier localization and
indium fluctuations [Fig. 4(c)]. These results suggest that a higher
surface recombination velocity (which is the same as a longer carrier
diffusion length) can lead to lLEDs with more non-radiative carrier
recombination because many of the carriers can reach the
sidewall.
33,64,65
In terms of red emissions, both InGaN-based and AlGaInP-based
red lLEDs are candidates. In accordance with the green gap, the EQE
of phosphide-based AlGaInP red LEDs is larger than that of InGaN-
based red LEDs, with sizes assumed to be large [>300 lm;
Fig. 4(d)].
52
The lower EQE of InGaN-based red LEDs is related to the
growth conditions and the generation of defects. However, InGaN-
based red lLEDs are compatible with AlGaInP-based lLEDs. Li
et al.
66
compared the EQEs of InGaN and AlGaInP-based red lLEDs
that varied in size from 20 lmto100lm and found that the size-
dependent degradation of AlGaInP is more serious than that of InGaN
due to its high surface recombination velocity. Recently, Park et al.
51
hypothesized that InGaN-based red LEDs with a high dislocation den-
sity (i.e., a high defect density) caused by the strain resulting from a
high indium content in the InGaN layer do not experience surface
recombination. They compared panchromatic CL between blue and
red lLEDs and found that a high dislocation density did not allow a
large number of carriers to move to the sidewall. Consequently, it was
shown that the EQE of InGaN-based red lLEDs increased with
decreasing chip size, which was in contrast to the typical EQE of
InGaN-based blue lLEDs [Fig. 4(e)]. This behavior could be attributed
to suppressed surface recombination and increased LEE via the side-
wall.Similarly,Limet al.
67
investigated InGaN-based red lLEDs using
thermal decomposition to relax the strain and found that the EQEpeak
of red lLEDs increased as their size decreased from 100 100 to 5 5
lm
2
. These results indicate that InGaN-based red lLEDs have a strong
advantage in terms of overcoming the size-dependent degradation of
lLEDs.
On the other hand, much research groups have demonstrated
that the degradation of AlGaInP-based red lLEDs when their size is
reduced is dramatic. For example, Oh et al.
68
investigated the degra-
dation of AlGaInP-based red lLEDs in terms of the EQE, ideality
factor, and leakage current. They found that EQEpeak was lower by a
factor of two and that Jpeak was considerably higher when the size
decreased from 350 350 to 15 15 lm
2
[Fig. 4(f)]. These results
suggest that the diffusion length of the carriers should be taken into
account when seeking to fabricate highly efficient lLEDs with
smaller sizes. Unlike InGaN lLED systems, although many attempts
have been made to overcome this efficiency degradation of AlGaInP-
based red lLEDs, the EQE has not yet been notably increased at
smaller sizes, most likely due to the very long carrier diffusion length.
Therefore, a method for suppressing the carrier diffusion length of
AlGaInP-based red lLEDs for use as a red source in future display
systems needs to be developed.
C. Quantitative analysis for sidewall damage of lLEDs
As mentioned above, a major challenge for small lLEDs is
overcoming carrier loss at the sidewall, with a higher surface recom-
bination velocity reducing the efficiency. Carrier loss at the sidewall
also depends on the fabrication processes involved (e.g., ICP-RIE,
chemical treatment, and passivation), the epitaxial quality, and
material properties, making it difficult to quantitatively interpret the
obtained EQE data. Although quantitative analysis of sidewall dam-
age would clarify the performance degradation of lLEDs, few stud-
ies have investigated this to date. In this section, we introduce a
quantitative interpretation of sidewall damage using various meth-
ods such as deep-level transient spectroscopy (DLTS), PL, CL, and
the ideality factor obtained from I–V characteristics reported in the
literature.
Boussadi et al.
69
quantitatively analyzed sidewall damage in
AlGaInP red lLEDs using time-resolved PL (TRPL) and found that
the carrier lifetime decreased closer to the sidewall. They also investi-
gated luminescence intensity profiles for a size of 5.615.61 lm
2
using CL under different temperatures. The ratio of CL intensity of
296 K to 30 K demonstrated that a 5.61 lm mesa with a width of
approximately 3.4 lm had an efficiency lower than 80% of the maxi-
mum value at the center [Fig. 5(a)]. For III-nitride blue lLEDs,
Finot et al.
70
systemically estimated the change in the lifetime of car-
riers under different sidewall conditions using photon-correlation
CL. Similar to the results of Boussadi et al.,
69
they found that the car-
rier lifetime across the mesa decreased considerably closer to the
sidewall. For example, processed KOHþALD-Al
2
O
3
exhibited an
increase in the carrier lifetime from 6 to 8ns compared to an unpro-
cessed sample near the sidewall, suggesting that the better the side-
wall conditions, the less carrier loss at the sidewall due to surface
recombination [Fig. 5(b)].
Yu et al.
71
quantitatively investigated the difference in the lifetime
of blue and green InGaN for sizes of 10 10 and 20 20 lm
2
fabri-
cated using a typical ICP-RIE process [Fig. 5(c)]. It was found that the
lifetime of blue InGaN lLEDs near the sidewall at sizes of 10 10 and
20 20 lm
2
was 3.1 and 3.0 ns, respectively, compared to 9.6 and
9.8 ns for the green InGaN lLEDs. This indicated that the influence of
surface recombination at the sidewall was weaker at higher emission
wavelengths due to the shorter diffusion length and the high-indium-
content InGaN QW.
DLTS allows the physical mechanisms underlying the perfor-
mance degradation of a device due to non-radiative recombination to
be understood, including the analysis of sidewall damage in lLEDs.
Recently, DLTS has been used to confirm the change in the energy
level of hole and electron traps in III-nitride lLEDs. Lee et al.
72
reported that deep hole traps and electron traps can form at E
v
þ0.75 eV and E
c
1.0 eV, respectively, at the sidewall of InGaN-
based blue lLEDs. As shown in Fig. 5(d), the DLTS signal as a func-
tion of temperature under a reverse bias of 0.5 V with a forward bias
pulse of þ3 V confirmed the level of the electron and hole traps.
Furthermore, the chip-size-dependent DLTS results presented in
Fig. 5(e) confirmed similar levels of electron and hole traps regardless
of the chip size, although the change in the signal was stronger with a
decreasing size, indicating a reduction in the carrier lifetime.
The ideality factor, which is defined using Eq. (10) and obtained
from I–Vcurvesofp–n junction devices, indirectly indicates the
recombination mechanisms,
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-8
V
CAuthor(s) 2024
16 May 2024 21:33:15
nideality q
KT
@InI
@V
1
:(10)
According to the theory outlined by Shockley et al.,
73
an ideality fac-
tor of 1.0 is associated with band-to-band radiative recombination,
an ideality factor of 2.0 represents SRH recombination, and an ideal-
ity factor over 2.0 is caused by defect-assisted tunneling. Therefore,
the ideality factor has been used to interpret both bulk defects (e.g.,
dislocation, trench defect), but and surface defects (i.e., sidewall
damage). Wong et al.
74
optimized the sidewall conditions of III-
nitride blue lLEDs and found differences in the ideality factor for
different sidewall conditions [Fig. 5(f)]. Optimized sidewall treat-
ment using KOH and ALD led to enhanced electrical performance,
which resulted in a decrease in the ideality factor. Similar results of
an improved ideality factor for nanorod LEDs fabricated via sol–gel
SiO
2
compared with conventional PECVD SiO
2
have also been
reported.
75
Collectively, previous studies of quantitative sidewall
damage have emphasized that the conditions of the sidewall play a
major role in the performance of lLEDs.
IV. EFFECTIVE PASSIVATION OF SIDEWALL SURFACE
The optical and electrical properties of optoelectronic devices are
negatively affected by native defects, crystallographic defects, and con-
taminants present on the surface of the semiconductor layer.
53
Thus,
KOH, buffer oxide etching, and HNO
3
:HCl have been employed to
eliminate surface oxides and contaminants. Aqueous solutions and
dielectric layers, including TMAH, (NH
4
)
2
S
x
,CH
3
CSNH
2
,SiO
2
,
Al
2
O
3
,andTa
2
O
5
, have also been adopted to passivate semiconductor
surfaces.
53
For lLEDs, a combination of solution treatment and dielec-
tric layer passivation has been found to be effective in alleviating the
size-dependent reduction in efficiency.
18,33
Therefore, there have been
various attempts to use different solutions and dielectric layers to
increase the EQE of lLEDs by removing non-radiative recombination
centers.
53
In particular, plasma-enhanced chemical vapor deposition
(PECVD) SiO
2
has been widely used for this purpose.
76–78
For exam-
ple, in blue lLEDs, it was found that the recombination lifetime of car-
riers increased when samples were passivated with a PECVD SiO
2
FIG. 5. (a) CL intensity ratio of at 296 K to 30 K showing efficiency degradation near the sidewall for 5.615.61 lm
2
AlGaInP-based red lLEDs. Reprinted with permission
from Boussadi et al., J. Lumin. 234, 117937 (2021). Copyright 2021 Elsevier.
69
(b) Investigation of CL lifetime near the sidewall with (lower) or without (upper) KOHþALD-
Al
2
O
3
passivation. Reprinted with permission from Finot et al., ACS Photonics 9, 173 (2022). Copyright 2022 American Chemical Society.
70
(c) Investigation of PL lifetime for
blue and green InGaN-based 10 10 and 20 20 lm
2
lLEDs near the sidewall. Reproduced from Yu et al., Appl. Phys. Lett. 121, 042106 (2022), with permission from AIP
Publishing.
71
(d) DLTS spectrum of 100 lm diameter LED demonstrating the level of traps center hole traps in QWs at E
v
þ0.7 eV and E
v
þ0.75 eV electron traps in –he
QWs at E
c
1eV. (e) DLTS spectra of 30 (green), 50 (yellow), and 100 (red) lm diameter LED showing similar level of traps regardless of size. Reprinted with permission
from Lee et al., J. Alloys Compd. 921, 166072 (2022). Copyright 2022 Elsevier.
72
(f) Size and sidewall condition dependent ideality factor extracted from I-V curves. Reprinted
with permission from Wong et al., Appl. Phys. Express 12, 097004 (2019). Copyright 2019 IOP Publishing.
74
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-9
V
CAuthor(s) 2024
16 May 2024 21:33:15
layer,
76
reducing the number of non-radiative recombination centers
and improving the IQE. This has also been shown to be the case for
InGaN-based red lLEDs.
77
ICP-RIE has been found to create plasma-
induced lattice disorders (about 2nm wide) on the surface of the mesa
sidewall, which can subsequently be removed using TMAH treat-
ment.
78,79
TMAH-treated blue lLEDs (chip size: 2 lm) exhibited a
considerable increase in their luminous intensity when passivated with
aSiO
2
layer.
ALD is widely used to deposit highly compact and dense dielec-
tric layers, meaning that ALD is also suitable for the deposition of pas-
sivation layers for lLEDs.
74,80–82
For AlGaInP-based red lLEDs,
passivation with ALD SiO
2
resulted in a 22.8% longer lifetime than
that with PECVD SiO
2
under the same aging conditions (e.g., continu-
ous operation at 400 A/cm
2
).
80
Furthermore, ALD SiO
2
passivation
combined with KOH treatment was found to reduce the ideality factor
of blue lLEDs from 3.4 to 2.5, indicating that the combined process
was effective in reducing SRH non-radiative recombination and sur-
face recombination.
74
Additionally, Al
2
O
3
deposited using ALD is con-
sidered an effective dielectric material for passivation.
83–92
Figure 6(a)
presents the EQEs of 20 20 lm
2
blue lLEDs fabricated using differ-
ent passivation methods.
83
For LED-4 (ALD passivation and HF etch-
ing), the maximum EQE was 32% higher than LED-1 (without
passivation). This increase in EQE was associated with an increase in
the light extraction and a reduction in the leakage current. Similarly,
Lee et al.
85
reported that InGaN-based blue lLEDs (10 10 lm
2
)
with ALD-Al
2
O
3
sidewall passivation exhibited a 66.7% higher EQE
and a 46.2% lower surface recombination velocity than those without
ALD-Al
2
O
3
treatment. For AlGaInP/GaInP red lLEDs (15 15 lm
2
),
a combined process of sulfur treatment and ALD-Al
2
O
3
passivation
yielded a 20% higher EQE and a 14% lower surface recombination
velocity compared to untreated samples.
86
For GaN-based green
lLEDs, ALD-Al
2
O
3
passivation and KOH treatment produced ideality
factorslowerthan1.5forallsamples(varyingfrom3to100lm).
88
Treated lLEDs (6 6lm
2
) produced a peak EQE of 16.59% at
20 A⸱cm
2
and over 600 k cd cm
2
at 1 A cm
2
.Ontheotherhand,
compared to ALD Al
2
O
3
, ALD AlN passivation has been found to
offer a stronger ability to remove sidewall defects in lLEDs as a result
of the uniform passivation interface.
90
Blue lLEDs (25 25 lm
2
)with
AlN passivation exhibited an 18.3% higher EQE than Al
2
O
3
-passivated
samples in a luminescence application.
DielectricmaterialssuchasTa
2
O
5
,Ga
2
O
3
,HfO
2
,andaTiO
2
/
SiO
2
omnidirectional reflector (ODR) have also been shown to act as
efficient passivation layers.
50,75,91,92
ALD Ta
2
O
5
(with a dielectric con-
stant of 26) effectively served as a passivation oxide for GaN lLEDs (a
peak wavelength of 425 nm).
91
The use of Ta
2
O
5
suppressed current
diffusion to the plasma-etched mesa edges. As shown in Fig. 6(b),pas-
sivated lLEDs (40 40 lm
2
) yielded a 34.6% larger EQE than unpas-
sivated samples and exhibited an on/off current ratio of 10
8
.In
addition, for InGaN-based blue lLEDs (15 15 lm
2
), photoelectro-
chemically oxidized Ga
2
O
3
passivation resulted in a 22% higher light
output and a lower reverse leakage current by over two orders of mag-
nitude at 5 V compared to reference samples.
92
It has also been con-
firmed that HfO
2
passivation can effectively improve the EQE of
580 nm blue nano-LEDs by approximately 18% compared to conven-
tional SiO
2
passivation.
50
In addition, a low-temperature sol-gel pro-
cess has been reported to be able to minimize the generation of defects
during the sidewall passivation process.
75,93
For InGaN/GaN-based
blue nano-LED arrays, device performance with sol–gel SiO
2
passiv-
ation was compared with that of PECVD SiO
2
passivation.
75
EQEs
were obtained from 60 pixels, with each pixel consisting of 6–9nano-
rods (diameter: 580 nm) fabricated with different forms of SiO
2
passiv-
ation. The sol-gel-passivated nano-LED arrays produced an EQE of
27.7%, which was 14% higher than that of the PECVD-passivated
devices. The improvement was ascribed to a reduction in the number
of defects generated during the passivation process.
75
Hydrogen passivation can reduce non-radiative recombination at
the sidewall by preventing the injection current from flowing into the
sidewall region.
94,95
In a previous study, p-GaN was passivated with
hydrogen by intentionally exposing the sidewall to hydrogen. This pro-
cess significantly improved LED performance by blocking current
injection into the mesa-etch-induced sidewall defects. For H-
passivated InGaN-based green lLEDs (20 20 lm
2
), the reverse leak-
age current was reduced morethan 10-fold and the EQE was enhanced
by 140% compared to the reference sample [Fig. 6(c)].
94
Furthermore,
a nitrogen ion implantation process has been adopted to modify side-
wall surface defects.
96
The PL intensity of InGaN-based green lLEDs
FIG. 6. (a) EQEs of 20 20 lm
2
lLEDs with different sidewall passivation methods. Reprinted with permission from Wong et al., Opt. Express 26, 21324 (2018). Copyright
2018 The Optical Society.
83
(b) Measured EQEs and optical power of lLEDs (40 40 lm
2
) with and without ALD-Ta
2
O
3
passivation. Reprinted with permission from Zhang
et al., IEEE Trans. Electron Devices 69, 3213 (2022). Copyright 2022 IEEE Publishing.
91
(c) On-wafer EQEs of lLEDs for before and after hydrogen passivation as a function
of the injection current density. Reprinted with permission from Kirilenko et al., Appl. Phys. Express 15, 084003 (2022).
94
Copyright 2022 IOP Publishing.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-10
V
CAuthor(s) 2024
16 May 2024 21:33:15
was enhanced sevenfold after passivation. Auger electron spectroscopy
results showed that this improvement was associated with a significant
reduction in sidewall defects including Ga-N or Ga-OH bonds.
Subsequently, the EQEs of packaged N-ion-implanted lLED chips
were about 33% higher compared to the EQE of the reference.
Moreover, ion implantation with an As
þ
source has been used to
replace the plasma-etching mesa process in the fabrication of GaN
lLEDs.
97
The optimal As
þ
implantation proce ss at 40 keV demon-
strated excellent I–V characteristics, including a forward voltage of
3.1 V at 1 mA and a leakage current of 10
9
Aat5 V for InGaN-
based blue lLEDs.
The dry etching conditions for the mesa of lLEDs are also a criti-
cal factor in determining their performance. Typically, ICP-RIE, which
is a plasma-based process, unintentionally damages the sidewall of
lLEDs, resulting in a decrease in efficiency. In this regard, Wang
et al.
98
demonstrated that a novel etching technique called neural
beam etching (NBE) results in negligible non-radioactive recombina-
tion at the sidewall surface. The authors compared the EQE and effi-
ciency degradation of blue lLEDs fabricated using NBE (3.5 3.5
lm
2
) and using a conventional ICP-RIE process with KOH treatment
(3 3lm
2
). It was reported that the J
peak
of the 3.5 3.5 lm
2
chips
etched with NBE was approximately 3 A/cm
2
,whereastheJ
peak
of the
33lm
2
chips was 9 A/cm
2
. According to Eq. (9),thischangeinJ
peak
indicates that NBE reduces non-radiative recombination at the side-
wall surface. Furthermore, the efficiency droop, which is defined at
0.01 A/cm
2
and can quantitatively confirm non-radiative recombina-
tion at the sidewall surface, was 26% and 60% for the 3.5 3.5 and
33lm
2
LEDs, respectively. The lower efficiency droop at low cur-
rent densities suggested that non-radiative recombination was rela-
tively low. Additionally, the size independence of J
peak
forLEDswitha
size of 3.5 3.5, 6.5 6.5, 10.5 10.5, and 20.5 20.5 lm
2
confirmed
that NBE led to negligible non-radiative recombination at the sidewall
surface.
V. EFFICIENCT LIGHT EXTRACTION FROM lLEDS
Various methods have been proposed to increase the LEE of
lLEDs, including the shaping of the sidewalls, diffracted Bragg reflec-
tors, ODRs, and surface treatment. An inclined sidewall has often been
adopted to redirect the in-plane transverse magnetic (TM) photons
toward the substrate for extraction, thus enhancing the LEE of
lLEDs.
99–103
For example, it has been shown that deep ultraviolet
(DUV) lLEDs (k¼280 nm; chip size ¼20 lm) with a sidewall angle
of 33yielded a 19% higher EQE than those with an inclination angle
of 75.
99
This improvement was attributed to the fact that more pho-
tons underwent total internal reflection (TIR) with the 33-inclined
sidewall, resulting in a higher LEE [Figs. 7(a) and 7(b)]. Normalized
electric field distributions inside the epi-layers of 20lmlLEDs were
used to analyze the propagation paths of the light originating from the
source to the sidewall. With the 33-inclined sidewall [Fig. 7(a)], more
light was reflected downward to the sapphire, thus increasing the LEE
from the bottom. Two-dimensional finite-difference time-domain
(FDTD) simulation results showed that, for 20lm diameter DUV
lLEDs, a sidewall angle of 25–35was optimal for light extraction
from the bottom.
99
It was also found that the LEE of red, green, and
blue flip chip (FC) lLEDs (<70 112 lm
2
) increased when raising
the sidewall inclination angle from 90to 120, although this was less
effective for the red lLEDs.
102
This was consistent with the fact that
the effect of the inclination angle on the LEE decreased with the higher
absorption of the materials. Additionally, a conically patterned SiO
2
/Ag
ODR microstructure array was used to improve top emission light
extraction and eliminate the color shift in red, green, and blue lLEDs.
102
For truncated-pyramid-shaped blue FC lLEDs, the strong dependence
of the LEE on the inclination angle has also been reported using the
SimuLED package, considerably enhancing the overall emission effi-
ciency.
103
The texturing of the sidewall surface has also been utilized to
increase light extraction. For example, the sidewall surface of circular
blue lLEDs (chip size: 50 lm) was patterned with concave–convex cir-
cular composite structures.
104,105
A circular pattern radius of 2 lmwas
found to be optimal, leading to the highest LEE. This improvement in
the LEE was associated with a reduction in the TIR at the sidewall
caused by the concave–convex circular composite structure.
Distributed Bragg reflectors (DBRs), ODRs, and reflective mirrors
have also been adopted to increase the LEE.
102,106–113
For example,
APSYS simulation results showed that, for GaN-based blue lLEDs
fabricated with a superlattice (SL) DBR as a p-type electron-blocking
layer (EBL), the reflectance of the p-region and the LEE increased with
ahighernumberofAlGaN/GaNSLDBRpairs.
106
Furthermore, the
use of a TiO
2
/HfO
2
(35 nm/50 nm) conduc tive DBRs combined with a
Cr/Ni/Au p-type electrode was found to effectively increase the light
output power (LOP) of near-UV lLEDs (k¼385 nm; 100 100 lm
2
)
by 5%.
107
Based on FDTD results, the LOP improvement was associ-
ated with a higher LEE (6%). Figures 7(c) and 7(d) show SEM images
of a 3.6 lm circular lLED (500 nm) with epitaxially embedded DBRs
(11 pairs of nanoporous GaN/undoped GaN) and the PL spectra of
lLEDs with and without the DBR, respectively.
108
The PL intensity of
the lLEDs with the DBRs was 150% higher than without the DBR,
indicating a markedly increased LEE due to the DBRs.
108
In addition,
increasing the reflectance of mirrors such as TiO
2
/SiO
2
-based
ODRs
109
and Ta
2
O
5
/SiO
2
multilayer high reflectors
110
hasbeenfound
to improve the LEEs of AlGaInP FC lLEDs (100 100 lm
2
) and GaN
lLEDs, respectively.
Surface roughening and texturing have been widely used to
increase the light extraction of lLEDs.
112–116
For example, Gong
et al.
112
investigated the effect of the nano-texturing of p-GaN on the
performance of blue lLEDs. The textured surface consisted of hexago-
nally arranged nano-cone arrays (height of cones: 75 nm). It was
found that nano-textured lLEDs with rhomboidal geometries pro-
duced a 57% higher output power than conventional square LEDs and
that the surface-textured samples enhanced the output power by 32%,
indicating that surface texturing can effectively increase the LEE. Lee
et al.
113
also found that, for lateral AlGaInP-based red lLEDs (chip
size: 25 17 lm
2
), surface roughening of p-GaP (etched for 24 s)
resulted in a 42.3% higher LOP at 20 lA than unetched p-GaP because
of the more effective light extraction due to surface scattering. Wang
et al.
114
investigated the effects of the surface microstructure and shape
on the LEE of GaN lLEDs using the finite element method and
reported that the LEE of trapezoidal-shaped FC lLEDs improved
from 53.0% to 64.5% as the mesa angle increased from 0to 12.
The use of surface grating on n-GaN was also observed to increase the
LEE of the lLEDs; grating with a period of 300 nm, a height of
147 nm, and a width of 243 nm yielded a maximum LEE of 72%.
Random surface roughening of n-GaN was also found to be effective
in increasing light extraction, resulting in an LEE of 56%.
The Monte Carlo ray tracing method has also been used to inves-
tigate the effects of substrate and sidewall texturing on the LEE of blue
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-11
V
CAuthor(s) 2024
16 May 2024 21:33:15
FC lLEDs.
115
The patterning of the substrate surface with circular
truncated cones, circular cones, pyramids, or a hemisphere structure
was found to be effective in enhancing the luminous intensity of these
lLEDs. For the circular cone-patterned substrate, the highest LEE was
attained at a cone inclination angle of 38. However, because of the
small difference in the refractive index between the sapphire substrate
and encapsulation, substrate surface texturing had little effect on the
LEE of encapsulated blue FC lLEDs. Simulation results demonstrated
that the sidewall texturing of blue FC lLEDs significantly increased
the total LEE by increasing the sidewall LEE. The patterned sapphire
substrate technology was thus shown to increase the LEE of the encap-
sulated lLEDs. Ryu et al.
116
examined the effects of chip shape, surface
roughness, and p-GaN thickness on the LEE of GaN-based vertical
lLEDs using three-dimensional FDTD simulations. It was observed
that the cone-shaped n-GaN surface effectively increased the LEE of
the FC lLEDs. However, surface roughness had little effect on the LEE
of blue FC lLEDs smaller than 5 lm, while square lLEDs produced a
higher LEE than circular lLEDs. The lower LEE was attributed to the
coupling of whispering gallery modes. For the square lLEDs, the LEE
was observed to strongly depend on the p-GaN thickness due to the
interference effect [Fig. 7(e)]. The highest total LEE was 77%, attained
at a thickness of 90–100 nm.
Other methods have also been employed to enhance the LEE of
lLEDs. For example, suspended GaN-based blue lLEDs exhibited a
150% higher light emission compared to conventional lLEDs.
117
The
suspended structure effectively increased the LEE of lLEDs by provid-
ing a much larger light-escaping area and eliminating light trapping by
the sapphire substrate [Fig. 7(f)], resulting in a dramatic increase in the
PL intensity. Asad et al.
118
also employed the FDTD method to study
the effects of various backside etch depths (from 0 to 5 lm) on the
FIG. 7. FDTD calculation of normalized electric field distributions for TM-polarized light inside the epi-layers from a 20 m diameter LED with (a) a vertical sidewall and (b) an
inclined sidewall. Reprinted with permission from Tian et al., Opt. Lett., 46, 4809 (2021). Copyright 2021 The Optical Society. (c) An SEM image of lLEDs with DBR and (d)
the PL spectra of samples with and without DBR reflector, respectively. Reprinted with permission from Bai et al., ACS Nano 14, 6906 (2020). Copyright 2020 American
Chemical Society. (e) LEE of square lLEDs (size: 20 lm) as a function of p-GaN thickness. Reprinted with permission from Ryu et al., IEEE Photonics J. 12, 1600110 (2020).
Copyright 2020 IEEE Publishing. (f) PL spectra of normal and suspended lLEDs. Inset denotes different light extraction behaviors. Reprinted with permission from Mei et al.,
ACS Photonics 9, 3967 (2022). Copyright 2020 American Chemical Society. (g) The intensity of CL at 10 K as a function of chip size, showing the change in LEE. Reprinted
with permission from Gonz
alez-Izquierdo et al., ACS Photonics 10, 4031 (2023).
121
Copyright 2023 American Chemical Society. (h) Far field characteristics of 1.0, 2.0, and
5.0 lm-sized LEDs with different sidewall angles simulated by FDTD. Reprinted with permission from V€
ogl et al. Opt. Express 31, 22997 (2023). Copyright 2023 The Optical
Society.
122
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-12
V
CAuthor(s) 2024
16 May 2024 21:33:15
LEE of top-emission blue vertical lLEDs on an Al backside mirror
(chip size: 5 5lm
2
). Their simulation results revealed a considerable
improvement (150%) in the LEE after reducing the optical cavity
length of the lLE D to 1.5 lm. This improvement in the LEE was asso-
ciated with the constructive interference of topside emitted photons
and reflected photons from the Al backside mirror. Surrounding the
lLEDs with a reflective Al layer increased the LEE by an additional
75%. Feng et al.
119
introduced a hybrid scheme to increase the LEE of
269 nm DUV lLEDs (chip size: 20 20 lm
2
). The hybrid scheme
included multiple cycles of inductively coupled plasma (ICP) and
TMAH processes with a 60 s ICP and 10 min TMAH treatment in
each cycle. Devices fabricated using the hybrid process demonstrated
the highest LEE at 3.88%. This improvement was attributed to the fact
that the hybrid process resulted in a hierarchical nanoscale structure
on the sidewall, allowing for the extraction of more light rays.
Furthermore, the use of a transparent and vertical package of InGaN-
based blue lLEDs (40 40 lm
2
) with a double-sided polished sap-
phire substrate as a transparent submount increased the LEE.
120
As lLEDs are reduced in size (especially below 10 lm), the LEE
part of the EQE becomes dominant, regardless of the emission wave-
length. Gonz
alez-Izquierdo et al.
121
investigated the performance of
lLEDs with different chip sizes between 2.5and 10 lm and experimen-
tally demonstrated through low-temperature CL that the LEE increases
with lower chip sizes. The size-dependent LEE can be quantitatively
examined at 10 K because SRH is not activated at low temperatures
and so an IQE of 1.0 is generally assumed for all sizes. Figure 7(g)
shows the CL intensity at 10 K for circular and square-shaped lLEDs
as a function of the chip size. The intensity of the circular and square
samples decreased with an increasing chip size, indicating that a
smaller lLED produced a larger LEE. V€
ogl et al.
122
numerically inves-
tigated the changes in the LEE, focusing on sizes from 1lmto5lm
and sidewall angles from 0to 60.Figure 7(h) shows the far-field dis-
tribution of blue lLEDs depending on the sidewall angle. For 5 lm
lLEDs, the emission pattern was less dependent on the sidewall angle,
while the effect of the sidewall angle became stronger when smaller
than 5 lm. Specifically, the far field of 1 lmlLEDs with a sidewall
angle of 40was mostly strong. This means that the optimization of
the sidewall angle is critical to realizing highly efficient ultra-small-
pixel lLEDs. This finding suggests that the enhancement of the LEE is
a promising approach to improve the overall EQE when lLEDs are
smaller than 10 lm.
VI. NOVEL EPITAXIAL STRUCTURES FOR
III-NITRIDE lLEDS
Using the top-down ICP-RIE process for the fabrication of highly
efficient lLEDs can be considered a mature technology due to the use
of chemical treatment and the deposition of a passivation layer to sup-
press the surface recombination rate at the sidewall.
11,18,53
However,
other drawbacks of III-nitride lLEDs systems, such as an efficiency
droop in a high current regime, emission wavelength shifts due to
QCSE, and a high dislocation density, have hindered the realization of
highly efficient lLEDs for next-generation displays that require vari-
ous brightness levels, such as those used in AR and VR.
5,6,8
To over-
come these drawbacks, novel epitaxial structures for III-nitride lLEDs
have been proposed as a solution. In this section, we review recent pro-
gress in novel epitaxial growth for III-nitride lLEDs.
Although c-plane oriented LEDs are already mature, spontaneous
and piezoelectric polarization, which can suppress electron–hole
overlap and wavelength shifts with increasing voltage, remains an
issue.
123
ToaddressthispolarizationissueforMQWsofInGaN
lLEDs, the use of semipolar structures in InGaN lLEDs has recently
been reported. For example, Chen et al.
124
compared c-plane oriented
and semipolar (20–21) InGaN lLEDs and confirmed the advantages
of semipolar (20–21) InGaN lLEDs under a high current regime.
Figure 8(a) presents a cross-sectional SEM image of semipolar (20–21)
GaN on a patterned sapphire substrate used to fabricate semipolar
(20–21) lLEDs. It was shown that the c-plane oriented InGaN lLEDs
had a 1.7 times larger peak EQE than semipolar (20–21) InGaN, which
would be a disadvantage for semipolar InGaN lLEDs under a low cur-
rent regime. Despite this, semipolar (20–21) InGaN lLEDs operate
well under a high current regime, with a lower efficiency drop and
smaller wavelength shift, which is promising for potential use in dis-
plays that require a higher brightness than conventional c-plane axis
lLEDs.
Conventional InGaN-based blue LEDs grown on a silicon sub-
strate have a very high dislocation density of over 10
9
cm
2
, which can
reduce the IQE.
34,125
Interestingly, Kamikawa et al.
126
demonstrated
that lLEDs could be fabricated using the epitaxial lateral overgrowth
GaN-on-silicon (EGOS) technique to produce regions that have a low
dislocation density of 5 10
6
cm
2
.Figure 8(b) presents 23.6 and
45.0 lmlLEDs fabricated using EGOS. Because the bridge, which is
approximately 3 lm in length, is the only part connecting the lLED to
the EGOS substrate, a polydimethylsiloxane (PDMS) stamp generating
stress could detach the lLED without breaking the chip. Similarly, Oh
et al.
127
investigated multiple-sapphire nanomembranes (MSNMs) at
the interface [Fig. 8(c)] and fabricated InGaN-based blue lLEDs with
sizes of 20 20, 40 40, and 100 100 lm
2
. The epitaxial layer of the
lLEDs was generated using pendeo-epitaxy, which resulted in a dislo-
cation density (3.3 10
8
cm
2
) that was approximately 59% lower
than that produced using conventional growth (8.0 10
8
cm
2
). The
Al
2
O
3
layer weakly connected the lLED to the substrate, facilitating
mechanical liftoff as an alternative to laser liftoff.
Another strategy for the fabrication of high-brightness lLEDs is
to design a band diagram by changing the thickness of the QB. Lin
et al.
128
computationally and experimentally studied the effect of the
QB thickness on the EQE, demonstrating that a thin QB leads to a rel-
atively high EQE at a high current, with less of a blue shift due to the
suppression of the electrical field in the QW. Similarly, Park et al.
129
reported that 10 10 lm
2
InGaN-based blue lLEDs with a thin 3 nm
GaN QB produced an efficiency droop at 10 000 A/cm
2
of approxi-
mately 20%, which suggests that a thin QB leads to relatively high
brightness under a high current regime while ameliorating the effi-
ciency droop. It has also been found that a thinner QB can improve
the total carrier recombination rate of InGaN lLEDs.
130
In addition,
Back et al.
131
systemically investigated the effect of the thickness of the
EBL of p-AlGaN on InGaN-based blue lLEDs with sizes ranging from
10 10 to 80 80 lm
2
. When the performance of a device fabricated
between 2 10
18
cm
3
Mg-doped 17 nm-thick Al
0.2
In
0.02
Ga
0.78
Nand
210
19
cm
3
Mg-doped 45 nm-thick Al
0.12
In
0.02
Ga
0.86
N (referred to
as a balanced EBL) was assessed [Fig. 8(d)], it was found that the
45 nm thick balanced EBL reduced the leakage current at the QW
underlying the EBL. As a result, the EQE for all lLEDs with a balanced
EBL was higher, while their Jpeak values decreased.
The use of III-nitride tunnel junction (TJ) contacts has also been
shown to be beneficial for InGaN lLEDs in terms of promoting their
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-13
V
CAuthor(s) 2024
16 May 2024 21:33:15
LEE and preventing vertical hydrogen diffusion, while also offering
multi-color emission with independent junction control.
132–135
Hwang
et al.
132
reported that TJs can improve the performance of InGaN
lLEDs over standard lLEDs in terms of a uniform current spread and
greater optical transparency. Although TJs can improve the optical
properties of LEDs (e.g., the EQE), they still suffer from a relatively
high turn-on voltage due to high resistance, indicating that there is a
trade-off between the electrical and optical properties of lLEDs. To
address this issue, Wong et al.
134
recently designed a novel TJ structure
by growing an n-type Al
0.11
Ga
0.89
N/GaN layer on top of a conven-
tional TJ structure. They experimentally demonstrated that the WPE
of a 60 60 lm
2
n-type Al
0.11
Ga
0.89
N/GaN layer was 10% higher than
standard TJ lLEDs, which was computationally supported by band-
diagram simulation results showing a higher tunneling rate due to a
lower polarization charge at the Al
0.11
Ga
0.89
N/GaN interface
[Fig. 8(e)]. In terms of multi-color emission, Saito et al.
135
reported a
330 PPI monolithic R, G, and B lLED array stacked on the same
wafer. To realize this structure, they added a n-In
0.2
Ga
0.8
N(25nm)/
pþGaN (10nm) TJ between the blue and green lLEDs and between
the green and red lLEDs. Considering light extraction through the
substrate, blue QWs were positioned at the bottom, green QWs in the
center, and red QWs at the top. This was because the bandgap energy
of red QWs is lower than that of blue and green QWs.
VII. CHALLENGING RED lLEDS
A. Diffusion length of III–V
The degree to which non-radiative recombination centers in the
bulk and at the sidewall surface affect lLEDs differs depending on the
type of material that is used (e.g., InGaN or AlGaInP). Because of non-
radiative recombination centers, both InGaN and AlGaInP red lLEDs
are characterized by a low efficiency, making them unsuitable for use
in full-color lLED displays. Therefore, it is necessary to clearly under-
stand the reason why red lLEDs exhibit poor efficiency and to develop
strategies to overcome this.
In this section, we discuss the surface recombination velocity of
lLEDs. A summary of the surface recombination velocity of various
materials is presented in Fig. 9(a).III-nitrideshavealowersurface
recombination velocity than other members of the III–V family.
64
This
is true because the surface recombination velocity is proportional to
not only the surface recombination rate but also the diffusion length of
the carriers
v½ms1¼km
½
As½s1:(11)
Although it is unclear whether the surface recombination rate (As)of
AlGaInP or InGaN is larger, it is known that the carrier diffusion
length (kÞof GaP or GaAs in AlGaInP-based red lLEDs is longer than
that of GaN. This difference in the diffusion length consequently
affects the surface recombination density at the sidewall. The carrier
diffusion lengths for III–V materials reported by previous studies are
summarized in Table I.
46,136–150
The carrier diffusion length for GaP
and GaAs is on a scale of a few micrometers. However, for GaN, it is
around a few tens to a few hundreds of nanometers. An increase in the
carrier diffusion length leads to efficiency loss for lLEDs via carrier
loss at the sidewall, with the degradation becoming more serious as the
size of the LED reduces. Figure 9(b) presents the EQE
peak
from recently
published papers for InGaN-based blue, green, and red lLEDs and
FIG. 8. (a) Cross sectional SEM image of (20–21) semi-polar GaN grown on a patterned sapphire substrate for semi-polar lLEDs. Reprinted with permission from Chen et al.,
Photonics Res. 8, 630 (2020). Copyright 2020 The Optical Society.
124
(b) SEM image of lLEDs fabricated on ELOS substrate. Reprinted with permission from Kamikawa et al.,
Cryst. Growth Des. 23, 4855 (2023). Copyright 2023 American Chemical Society.
126
(c) Micro-LED on MSNM. Reprinted with permission from Oh et al., ACS Appl. Mater.
Interfaces 14, 25781 (2022). Copyright 2022 American Chemical Society.
127
(d) Cross sectional TEM images having 10.5 nm QB and 10.5 nm QB þthick EBL. Reprinted with
permission from Baek et al., Nat. Commun. 14, 1386 (2023). Copyright 2023 Springer Nature.
131
(e) Simulation on band diagrams and electron tunneling rate at reverse bias of
3 V for GaN and Al
0.11
Ga
0.89
N-based tunnel junction. Reproduced from Wong et al., AIP Advances 13, 015107 (2023), with permission from AIP Publishing.
134
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-14
V
CAuthor(s) 2024
16 May 2024 21:33:15
AlGaInP-based red lLEDs as a function of the ratio of the peripheral
length of the device to the area.
77,80,85,88
For InGaN lLEDs, the slope
of EQE
peak
increases as the wavelength decreases from red to blue.
This is in good agreement with experimental results demonstrating
that vvaries with the indium content.
63
In other words, k,whichisa
component of v, determines the degradation of EQE
peak
. Furthermore,
as summarized in Table I, the diffusion lengths of GaP and GaAs,
which are components of AlGaInP-based red lLEDs, are 10 times lon-
ger than that of GaN. In contrast, InGaN-based red LEDs suffer less
sidewall surface recombination due to the relatively short diffusion
length. Thus, the difference in the diffusion length between AlGaInP
and InGaN is beneficial for InGaN LEDs when they are smaller than a
certain size, resulting in a relatively high EQE. Ultimately, the differ-
ence in the slope between InGaN- and AlGaInP-based red lLEDs
means that their potential device applications need to be considered
based on their PPI values. In other words, InGaN-based red lLEDs
are suitable for displays that require ultrahigh PPI, such as VR and AR,
whereas AlGaInP-based red lLEDs are more suitable for other display
applications.
B. Sidewall surface recombination of AlGaInP
red lLEDs
In AlGaInP red LEDs, the fact that AlGaInP is based on a lattice-
matched system guarantees a high EQE at relatively large sizes due to a
low bulk defect density (i.e., misfit dislocations, point defects, and 2D/
3D defects). Both p-GaP and n-GaAs produce holes and electrons with
a diffusion length on a scale of approximately a few micrometers
(Table I), which is over 10 times longer than that of GaN. Therefore,
AlGaInP-based red lLEDs have surface recombination centers at the
sidewall, causing the carriers to move to those centers and resulting in
their loss. Because surface recombination is a form of non-radiative
recombination, it is the primary determiner of the performance of
lLEDs at a low current. In particular, Jpeak , which is used to numeri-
cally express surface recombination, increases dramatically at smaller
sizes. For example, Fan et al.
151
reported that the Jpeak of a 160 lm
AlGaInP LED was approximately 10 A/cm
2
. However, it dramatically
increased to over 80 A/cm
2
at a size of 10 lm. In addition, the EQE
dramatically decreased from 20% to 2.5% at 20 A/cm
2
, indicating that
surface recombination at the sidewall became stronger as the chip size
decreased. Thus, the IQE of AlGaInP lLEDs was seriously degraded.
Several solutions have been proposed based on chemical treat-
ment and/or the formation of a passivation layer using ALD.
80,152–157
Jung et al.
154
found that HF treatment can effectively remove the side-
wall defects in AlGaInP red lLEDs induced by the conventional ICP-
RIE process. Before HF etching, an atomic arrangement indicating
crystal damage was observed approximately 2 nm from the sidewall
[Fig. 10(a)]. After HF, this type of surface defect disappeared
[Fig. 10(b)], consequently improving the EQE [Fig. 10(c)]. However,
the peak EQE did not occur at a low current, indicating that sidewall
surface recombination was still dominant. Furthermore, it has been
reported that chemical treatment prior to ALD sidewall passivation
has a dominant effect in recovering sidewall damage.
156
Size- and
sidewall-dependent EQE is presented in Fig. 10(d). The 100 lm refer-
ence sample exhibited characteristics that were superior to the other
samples, indicating that the performance of larger lLEDs did not criti-
cally depend on the sidewall conditions. However, the EQE of 20 lm
LEDs under different sidewall conditions (reference: without passiv-
ation; chemical treatment: ALD passivation using Al
2
O
3
;ALDþS:
(NH
4
)
2
SO
4
treatment and passivation using Al
2
O
3
)differedremark-
ably. Although methods to remove sidewall damage could improve the
performance of AlGaInP lLEDs, their performance at a low current
was still significantly lower than that of larger sized LEDs, with Jpeak in
particular very diffe rent between 100 and 20 lm. This indicates that
the minority carriers can still diffuse to the sidewall and recombine
non-radiatively under a low current.
Recently, Mun et al.
157
demonstrated that treating AlGaInP-
based red lLEDs with dual dielectric passivation effectively improves
their electrical characteristics (e.g., ideality factor and leakage current),
thus increasing the EQE. Comparing the ideality factors for various
FIG. 9. (a) Summarizing surface recombi-
nation velocity of III–V compound materi-
als. The surface recombination velocity of
III-nitride is over ten times lower than
others. Reprinted with permission from
Bulashevich et al., Phys. Status Solidi
RRL 10, 480 (2016).
64
Copyright 2016
Wiley VCH. (b) Description of the change
in EQE
peak
for InGaN-based blue, green,
red, and AlGaInP-based red lLEDs.
TABLE I. Summary of diffusion length of minority carriers for various III–V compound
materials (the diffusion length of GaN is typically over ten times lower than that of
GaP or GaAs).
Materials
Hole diffusion
length in n-type
layer (lm)
Electron diffusion
length in p-type
layer (lm) Ref.
GaP 1–6 0.06–3136–139
GaAs 2–35 0.6–20 140–145
GaN 0.05–0.7 0.093–0.95 46,146–150
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-15
V
CAuthor(s) 2024
16 May 2024 21:33:15
chip sizes from 100 100 to 10 10 lm
2
fabricated with different
passivation conditions, the samples with the dual dielectric passivation
of 10 nm Al
2
O
3
(ALD)/300 nm SiN
x
(PECVD) exhibited the lowest
ideality factors regardless of the chip size. Furthermore, electrolumi-
nescence (EL) imaging revealed that the samples passivated with the
Al
2
O
3
/SiN
x
layer produced the brightest images, though their Fresnel
power transmittance (77.7%) was lower than that of Al
2
O
3
(ALD)/
SiO
2
(PECVD) (86.9%). These results suggest that it is important to
design a sidewall passivation layer that considers the electrical and
optical characteristics as a whole.
Although chemical treatment and/or ALD passivation can
enhance the EQE of AlGaInP-based red lLEDs, the EQE sharply
decreased at smaller sizes (Fig. 10). We believe this size-dependent
degradation even with improved sidewall conditions is caused by the
intrinsic surface state. Because the all-semiconductor surface band
diagram is intrinsically pinned, carriers have the possibility of reaching
the surface due to the difference in energy level between bulk and sur-
face. Therefore, the intrinsic surface state and the longer diffusion
length of the carriers are the reasons why AlGaInP-based red lLEDs
suffer from sidewall surface recombination even when the sidewall
conditions are improved, as reflected by case 3 in Fig. 3(d). Thus, the
intrinsic surface state and the long diffusion length remain limitations
that need to be overcome to produce highly efficient AlGaInP lLEDs
as a red source with smaller chip sizes.
68,151
One potential solution for
reducing surface recombination at the sidewall is controlling the diffu-
sion length of the carrier. The carrier diffusion length is defined as a
function of the carrier concentration.
136,143
It is expected that the diffu-
sion length of the carrier in GaP and GaAs could be shortened by
increasing the carrier concentration as shown in Figs. 11(a) and 11(b),
respectively. Thus, optimized sidewall conditions using chemical
FIG. 10. Atomic arrangement at the side-
wall of AlGaInP lLEDs (a) before and (b)
after HF treatment. (c) Relative EQEs for
12, 15, and 19 lm square before and after
HF treatment. Reprinted with permission
from Jung et al., Sci. Rep. 11, 4535
(2021).
154
Copyright 2021 Springer Nature.
(d) Investigating degradation for EQE of
AlGaInP red lLEDs with various sidewall
conditions. Reprinted with permission from
Wong et al., Appl. Physic. Express 16,
066503 (2023).
156
Copyright 2023 IOP
Publishing.
FIG. 11. (a) Hole diffusion length as a func-
tion of electron concentration in GaP.
Reprinted with permission from Smith et al.,
Solid State Electron. 15, 361 (1972).
136
Copyright 1972 Elsevier. (b) Electron diffu-
sion length in p-type GaAs as a function of
hole concentration. Reproduced from Casey
et al.,J.Appl.Phys.44, 1281 (1973), with
permission from AIP Publishing.
143
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-16
V
CAuthor(s) 2024
16 May 2024 21:33:15
treatment and passivation and a shorter carrier diffusion length could
yield highly efficient AlGaInP-based red lLEDs.
C. Importance of growth conditions of InGaN red
lLEDs
Due to the lower surface recombination velocity of III-nitrides
than other III-V materials, the main concerns for InGaN red LEDs are
how to suppress the number of defects in the bulk, relax the strain, and
optimize the growth conditions to improve the QW quality. These
challenges need to be overcome in order to address size-independent
efficiency degradation. Fundamentally, these issues originate from the
growth of an InGaN QW layer with a high indium content. In particu-
lar, the growth of an InGaN QW layer with a high indium content
requires lower temperatures, which leads to poor crystal qual-
ity.
16,17,158,159
Furthermore, this InGaN layer increases the strain due
to lattice mismatch, resulting in defects such as threading dislocations
and trench defects.
To improve the performance of InGaN-based red LEDs, various
strategies have been reported in recent years. In 2020, Iida et al.
160
fab-
ricated 633 nm wavelength red LEDs with a size of 400 400 lm
2
at
20 mA by varying the unde rlying thickness of the n-GaN layer from 2
to 8 lm. They found that a thick underlying n-GaN layer could help
reduce the in-plane compressive stress, thus decreasing the number of
surface defects, including trench defects [Fig. 12(a)]. Furthermore, it
was found that a thick underlying n-GaN layer could lead to a decrease
in the residual in-plane stress, resulting in enhanced indium incorpo-
ration in InGaN QWs and a red shift in the wavelength [Fig. 12(b)]. In
2020, Dussaigne et al.
161
investigated the fabrication of InGaN red
LEDs grown on InGaN and sapphire (InGaNOS, fabricated by Soitec)
and found that the number of threading dislocations decreased using a
substrate with a large alattice parameter. For example, they compared
threading dislocation densities between 3.2069 and 3.2056 ˚
Ain
InGaNOS substrates after full InGaN red LED growth and found that
a large lattice parameter could effectively reduce the number of defects
generated in the InGaN QWs [Fig. 12(c)]. This decrease in the number
of defects consequently increased the EQE of red LEDs over twofold
[Fig. 12(d)]. In 2023, Wu et al.
162
investigated V-defect formation in
InGaN-based red LEDs and claimed that small V-defects forming in
MQWs not from dislocations would be deleterious to the production
of highly efficient red LEDs because these could act as SRH and trap-
assisted Auger recombination centers. By identifying the importance
of V-defects in InGaN-based red LEDs, they were able to achieve a
peak EQE of 6.5% at 600 nm.
163
Another route for improving the crystal quality of InGaN-based
red LEDs is relaxing the strain. In 2023, Lim et al.
67
reported that the
FIG. 12. (a) Surface defect density depending on the underlying n-GaN thickness. (b) EL peak wavelength with various underlying n-GaN thickness demonstrating suppressed
in-plane residual stress with increasing n-GaN thickness. Reproduced from Iida et al., Appl. Phys. Lett. 11 6, 162101 (2020), with permission from AIP Publishing.
160
(c)
Comparison analysis of threading dislocation density between 3.2069 and 3.2056 ˚
A of InGaNOS substrates. (d) EQEs of InGaN-based red LEDs grown on InGaNOS with a lat-
tice parameter of 3.2069 ˚
A (LED B) and InGaNOS with a lattice parameter of 3.2056 ˚
A (LED A). Reproduced from Dussaigne et al., J. Appl. Phys. 128, 135704 (2020), with per-
mission from AIP Publishing.
161
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-17
V
CAuthor(s) 2024
16 May 2024 21:33:15
growth of an In
0.3
Ga
0.7
N decomposition layer could effectively relax
the strain. This high-indium-content In
0.3
Ga
0.7
Nlayerwasdecom-
posed during n-InGaN/n-GaN decomposition stop-layer (DSL)
growth at 950 C, which generated tensile strain in the DSL layer. The
tensile stress of an n-InGaN/n-GaN buffer layer grown on the DSL
layer and DSL layers was calculated to be 0.16% and 0.28%, respec-
tively. Both were in the line of fully relaxed in an XRD reciprocal space
map (RSM), which indicated that the DSL could generate tensile strain
[Fig. 13(a)]. Subsequently, they demonstrated that a decomposed
InGaN underlayer resulted in a red emission of 643 nm and peak EQE
of 0.44% for 5 5lm
2
lLEDs. The peak wavelength was estimated to
be 660 nm at 1 A/cm
2
and 597 nm at 200 A/cm
2
.
The use of a porous GaN layer can also suppress strain. In 2020,
Pasayat et al.
164
fabricated InGaN-based lLEDs on porous GaN and
compared the performance of lLEDs fabricated on porous and non-
porous regions with various tile sizes. It was found that the emission
wavelength of lLEDs differed depending on whether a porous under-
layer was present. In particular, lLEDs fabricated on the non-porous
region emitted the shortest wavelength, indicating that a relatively
large strain can lead to the lowest indium content in an MQW. As a
result, porous GaN allowed a higher indium uptake via strain relaxa-
tion. Subsequently, they demonstrated that 66lm
2
lLEDs with an
on-wafer EQE of 0.202% at 10 A/cm
2
emitted a wavelength of 632 nm
under optimized growth conditions for n-In
0.04
Ga
0.96
N on a porous
GaN substrate and a tile size of 11 11 lm
2.165
RSM data [Fig. 13(b)]
also rev ealed that the 440 nm-thick und erlying n-In
0.04
Ga
0.96
Nlayerwas
relaxed by about 56%, resulting in MQWs above that layer emitting a
longer wavelength with a relatively high indium content (>26%).
In another study on strain modulation, Chen et al.
166
optimized
the thickness of a low-temperature GaN buffer layer. They found that
a thickness of 7 nm produced the highest indium content in the QW,
resulting in a peak EQE of about 7.4% at a wavelength of 629 nm, with
achipsizeof100200 lm
2
.Figure 13(c) presents the average indium
content in the MQWs and the peak wavelength in the EL spectra at
1A/cm
2
as a function of the thickness of the low-temperature buffer
layer. It was seen that the average indium content and the peak wave-
length both increased as the buffer layer thickness increased. However,
from a thickness of 9 nm, both measures started to decrease. Based on
an optimized low-temperature buffer layer thickness of 7nm, Chen
et al. then studied changes in the EQEs for MQWs with various
indium levels. Figure 13(d) presents the EQE as a function of the peak
wavelength in the EL spectra under various current injections and
indium levels in the MQWs. The estimated indium content in the
MQWs was 5.8%, 6.0%, 6.2%, and 6.5% for S1, S2, S3, and S4, respec-
tively. This indicates that achieving a high EQE in current InGaN red
lLEDs with a longer wavelength is challenging.
Strain relaxation by an In
0.08
Ga
0.92
N layer on top of u-GaN,
which allows for relatively longer wavelengths, was recently reported
FIG. 13. (a) XRD RSM of (1124) reflec-
tions revealing a strain relaxation using
In
0.3
Ga
0.7
N decomposition layer. Reprinted
with permission from Lim et al., Adv.
Photon. Res. 4, 2200286 (2023). Copyright
2023 Wiley VCH.
67
(b) High-resolution XRD
RSM recorded around the GaN (1124)
reflection showing strain relaxation using
porous GaN. Reprinted with permission
from Pasayat et al.,Appl.Phys.Express
14, 011004 (2021). Copyright 2021 IOP
Publishing.
164
(c) Thickness of LT-GaN
buffer layer dependent average indium con-
tent in MQWs and peak wavelength of EL
spectra at 1 A/cm
2
. (d) Degradation of EQE
with increasing indium content in MQWs
and increasing emission wavelength.
Reprinted with permission from Chen et al.,
Adv. Funct. Mater. 33, 2300042 (2023).
Copyright 2023 Wiley VCH.
166
(e) Cross-
sectional TEM images showing an
In
0.08
Ga
0.92
N stress-release layer and an
AlN dislocation confinement layer. (f) EL
spectra of an LED at different current densi-
ties, confirming a relatively longer emission
wavelength via the stress-release layer.
Reprinted with permission from Xing et al.,
Opt. Express 32, 11377 (2024). Copyright
2024 The Optical Society.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-18
V
CAuthor(s) 2024
16 May 2024 21:33:15
by Xing et al.
167
The introduction of the In
0.08
Ga
0.92
N stress-release
layer (SRL) and an AlN dislocation confinement layer (DCL) was
found to effectively release the compressive stress and to reduce the
penetration of threading dislocations, respectively. Figure 13(e)
presents a cross-sectional TEM image with reflection g¼[0002], con-
firming that, for the samples with SRL and DCL layers, the number of
screw dislocations is negligibly small. Moreover, for the same sample,
defect densities above and below the SRL and DCL layers were calcu-
lated to be approximately 1.56 10
8
and 1.08 10
8
cm
2
,respectively.
This indicates that the SRL and DCL suppress the increase in the num-
ber of defects. In addition, Raman scattering spectral results revealed
that the SRL and DCL alleviated the compressive strain. The EL spec-
tra from LEDs with and without the SRL and DCL at various current
densities [Fig. 13(f)] also illustrated effective strain release. For exam-
ple, the peak wavelengths in the EL spectra at 1 A/cm
2
were estimated
to be 651 and 634 nm for samples with and without the SRL and DCL,
respectively. A difference in the peak wavelength of about 17 nm again
suggests that control of the strain and defect density are important fac-
tors in achieving longer emissions.
Conventional InGaN-based blue LEDs, which are grown on the
sapphire, have a relatively simple MQW structure consisting of 2–
3nm InGaN QWs and 12–15 nm GaN QBs. To fabricate high-
performance red InGaN LEDs, specific growth conditions for the
MQWs are required, thus groups and KAUST and UCSB have sought
to optimize these growth conditions. Iida et al.
168
achieved a peak EQE
of 4.3% at 621nm by optimizing the QW structure [Fig. 14(a)]. This
suggests that a single QW structure in the InGaN layer for red emis-
sions can suppress additional defect generation. A hybrid MQW struc-
ture consisting of a lower-indium-content InGaN layer (i.e., a blue
single QW) was also shown to release the strain. An n-AlGaN barrier
layer between the blue and red QWs effectively suppressed hole injec-
tion into the blue QW. Subsequently, the AlN capping layer between
FIG. 14. (a) Optimized epitaxial structure for
red emissions including single red InGaN
QW, blue InGaN QW, hole-blocking
n-AlGaN barrier, and AlN capping layer.
Reproduced from Iida et al.,AIPAdvances
12, 065125 (2022), with permission from AIP
Publishing.
168
(b) Computational investiga-
tion for the overlap of electron–hole wave-
functions in red InGaN QW (left, 3 nm in
thickness; right, 1.9 nm in thickness).
Reprinted with permission from Li et al.,
ACS photonics 10, 1899 (2023). Copyright
2023 American Chemical Society.
169
(c)
Band structure and wavefunction simulatio n
results of samples with/without the AlGaN
interlayer and the GaN cap layer.
Reproduced from Lee et al., Appl. Phys.
Lett. 124, 121109 (2024), with permission
from AIP Publishing.
171
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-19
V
CAuthor(s) 2024
16 May 2024 21:33:15
the InGaN-based red single QW and u-GaN prevented the decomposi-
tion of the QW during high-temperature processing after it had grown.
Li et al.
169
also investigated the effect of the thickness of InGaN QWs
(1.9 and 3 nm) for red emissions. They found that the peak EQE signif-
icantly increased from 3.3% to 6.0% with a 602 nm emission wave-
length by decreasing the InGaN QW t hickness from 3 to 1.9 nm. This
effect of the QW thickness could be attributed to electron–hole wave-
function overlap [Fig. 14(b)]. A thicker QW increased the piezoelectric
field, which led to IQE degradation. However, a thinner QW increased
the electron–hole overlap area in the QW, which increased radiative
recombination and the IQE. Recently, Vichi et al.
170
and Lee et al.
171
reported the effect of a 3 nm Al
0.8
Ga
0.2
N interlayer and a 1.5 nm GaN
cap layer on changes in the wavefunction overlap and peak wave-
length. The AlGaN interlayer and GaN cap layer were shown to induce
band bending [Fig. 14(c)]. As a result, a redshift in the peak wavelength
occurred with a decrease in the electron–hole wavefunction overlap.
However, in terms of the crystal quality of the grown layers, the
AlGaN interlayer led to strain compensation, and the GaN cap layer
prevented indium desorption from the InGaN QW. As a result, based
on their calculations, they achieved an EQE of 15% at 631nm.
Collectively, these results suggest that it is necessary to carefully con-
sider the tradeoffs between crystal quality (non-radiative recombina-
tion) and electron–hole wavefunction overlap (radiative
recombination) in order to produce highly efficient InGaN-based red
LEDs.
Most of the recent results for InGaN-based lLEDs have reported
a peak wavelength between 600 and 630nm at the peak EQE, with
very few emitting a wavelength over 650 nm. Thus, emitting pure red
using InGaN-based red lLEDs remains a challenge. According to
Damilano et al.,
172
the emission color of InGaN-based LEDs is defined
as a function of the indium composition and the thickness of the QWs
[Fig. 15(a)]. This definition is related to the internal electric field.
Therefore, the fundamental energy transition in a QW is given by
EQW x;h
ðÞ
¼Egx
ðÞ
þe1x;h
ðÞ
þhh1x;h
ðÞ
Ry x;h
ðÞ
e0Fx
ðÞ
h;
(12)
where Egis the InGaN QW bandgap, e1and hh1are the first electron
and hole energy levels, respectively, Ry is the exciton binding energy,
e0is the elementary charge, Fis the electric field across the QW, and h
is the thickness of the QW. Park et al.
51
demonstrated the possibility of
using an 8.3 nm thick InGaN QW with 13% indium content
[Fig. 15(b)] to red at a wavelength of 690 nm at 1 A/cm
2
[Fig. 15(c)].
This produced a piezoelectric field in the thick QW, reducing the
energy transition. However, the peak EQE of their InGaN-based red
lLEDs was approximately 0.25%, indicating that there was a trade-off
between the emission wavelength and efficiency. Although a thick QW
emits longer wavelengths, the wavefunction overlap between the elec-
trons and holes becomes worse as the QW thickness increases. To
ensure a high efficiency and a longer emission wavelength for InGaN-
based red lLEDs, the QW thickness, indium content, and energy band
diagram for the QW should be optimized.
Other novel approaches for emitting red from InGaN-based
lLEDs have been reported in recent years by many groups.
173–179
Bi
et al.
173
realized EL red emission (1.98 eV) at a sub-micrometer scale
from InGaN platelets using selective area metal–organic chemical
vapor deposition (MOCVD). Feng et al.
175
also proposed selective epi-
taxy growth on a microhole-patterned template and reported an emis-
sion wavelength of 642 nm at 10A/cm
2
.Caiet al.
177
investigated a
sub-micrometer platelet array for long wavelengths and confirmed
that the optimization of the GaN seed layer and InGaN platelets, with
the layers relaxing the strain and allowing the MQW to reach a
wavelength emission of over 630 nm using PL and CL. Subsequently,
Yu et al.
178
reported that high-indium-content InGaN/GaN quantum
dots inserted between p-AlGaN and an n-doped superlattice layer
could yield a peak wavelength of 630 nm at 10 A/cm
2
. Europium dop-
ing is also a candidate for achieving the emission of longer wave-
lengths.
179
For example, Mitchell et al.
179
confirmed the possibility of
red emission using an Eu-doped GaN layer, with a maximum EQE of
9.2% and a peak wavelength of approximately 630 nm.
D. State-of-the-art AlGaInP and InGaN red LEDs
Figure 16 summarizes the state-of-the-art red LEDs that have
been reported in the last 6 years. The EQE is divided into three groups:
AlGaInP LEDs (sky blue symbol),
68,80,86,151,153,180
InGaN LEDs emitting
wavelengths over 625 nm (red symbol),
51,67,77,160,161,165,166,171,181–189
and
InGaN LEDs emitting wavelengths less 625nm (green sym-
bol).
66,163,168,169,188,190–199
The EQE of the AlGaInP LEDs tends to
decrease with a smaller size. This is typically attributed to the significant
FIG. 15. (a) Definition of emission wavelength of InGaN-based LEDs as a function of Indium content and QW thickness. Reprinted with permission from Damilano et al.,J.
Phys. D: Appl. Phys. 48, 403001 (2015). Copyright 2015 IOP Publishing.
172
(b) STEM image showing a relatively thick InGaN QW that can emit 690 nm peak wavelength at
1 A/cm
2
. (c) Normalized EL spectra with various current injections showing blue shift induced by QCSE. Reprinted with permission from Park et al., Laser Photonics Rev. 17,
2300199 (2023). Copyright 2023 Wiley VCH.
51
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-20
V
CAuthor(s) 2024
16 May 2024 21:33:15
carrier loss at the sidewall due to a relatively long diffusion length.
Unlike AlGaInP LEDs, the growth conditions for InGaN MQWs when
aiming to emit longer wavelengths lead to defects in the bulkdue to the
relatively low growth temperature, which results in a higher number of
defect numbers, which affects the performance of InGaN red LEDs
more strongly than the size. The plotted data show that the EQE of
InGaN LEDs is less dependent on the device area than that of AlGaInP
LEDs. In addition, the EQEs for the green symbols are higher on aver-
age than those for the red symbols, which suggest that LEDs emitting
longer wavelengths contain more defects due to the growth of MQWs
at a relatively low temperature. Nonetheless, the low EQE of InGaN red
LEDs can be overcome through optimized growth conditions and
device processing, including sidewall passivation and reflectors to
improve light extraction efficiency.
171,185,199
However, in terms of effi-
ciency, it remains an open question as to whether AlGaInP- or InGaN-
based red lLEDs are suitable for the typical resolution regime for
lLED displays (e.g., 1000–10 000 PPI).
VIII. ADVANCED TRANSFER TECHNOLOGIES
FOR PRODUCTION
The integration of lLEDs and electronics drivers is an essential
process for lLED applications. For example, for micro- and mini-
displays, several integration methods, including selective area growth,
selective epitaxial removal, and 3D integration using FC bonding or
wafer bonding methods, have been employed to produce different
types of integrated optoelectronics circuit,
18,41,200–208
including
vertically connected, serially interconnected, and metal-interconnec-
tion-free integration circuits. The characteristics of these integration
techniques are summarized in Table II.
For large-screen display applications, a variety of methods have
also been proposed for the mass transfer of lLEDs, including electro-
static,
209
elastomer,
210,211
magnetic,
212,213
fluidic self-assembly,
204–217
laser,
218
and roll-to-roll techniques.
219,220
Mass transfer technologies
and their characteristics, such as the transfer rate, throughput,
reliability, scalability, and transfer characteristics, are summarized in
Table III.
It is known that electrostatic transfer methods make it possible to
transfer lLEDs from a host substrate to a receiving substrate using a
transfer stamp or a target substrate with electrostatically charged
areas.
209
Transfer printing using elastomer stamps integrates optoelec-
tronics and electronics.
210,211
In this process, soft elastomeric stamps
are utilized to remove lLEDs or ICs from the donor substrate and
place them on the target substrate to produce 2D and 3D arrays of
lLED chips.
207,211
In the case of magnetic transfer,
212,213
LED chips
are deposited with a magnetic layer and fixed to the (electro-)magneti-
zation transfer stamp for chip transfer to the host substrate via mag-
netic force. In addition, fluidic self-assembly methods use different
driving forces, such as gravity
214
and directional surface tension,
215
and shape-conforming structures, adhesives, liquid solder, or two dif-
ferent liquids are required.
216
In particular, automated reel-to-reel flu-
idic self-assembly
217
with two main units (an assembly and
component recycling and dispensing units) has been found to be
promising in terms of the assembly rate and yield. Moreover, a laser-
induced mass transfer method in which a laser beam is used to detach
LEDs from the host substrate and then transfer them to a receiving
substrate has been introduced.
218
Similar to laser liftoff technology,
laser irradiation causes ablation at the board/LED interface, pushing
the chips toward the receiving substrate and separating them from
the host substrate. Additionally, a temporary polymer adhesive sub-
strate can be used as the interfacial layer. Roll-to-roll and roll-to-
plate transfer methods are also promising processes for the mass
transfer of lLED chips.
219,220
These techniques have been found to
be compatible with either rigid or flexible substrates for flexible and
stretchable displays. They consist of three transfer steps: (i) the col-
lection and placement of an array of electronic components (e.g.,
TFTs) on a temporary substrate; (ii) the separation of the lLED
chips from the host substrate and their connection to the TFTs; and
(iii) the roll-transfer of the interconnected LED/TFT array to a
receiving substrate.
FIG. 16. Absolute EQEs of red LEDs and
red lLEDs as a function of chip sizes
reported in recent 6 years. Sky dots are
AlGaInP. Red dots are InGaN emitting
over 625 nm wavelength at peak EQE.
Green dots are InGaN emitting less
625 nm wavelength at peak EQE.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-21
V
CAuthor(s) 2024
16 May 2024 21:33:15
Despite these advances in transfer technology, for lLEDs to enter
the mainstream product market, practical transfer challenges such as
the throughput, yield, and production scalability must be addressed.
To address these issues, modified fluidic assembly methods have
been developed.
213,217
For example, Chang et al.
213
developed a mag-
netic-force-assisted dielectrophoretic self-assembly technology
(MDSAT) that combines magnetic and dielectrophoresis (DEP) forces
[Figs. 17(a)–17(e)]. In the fluidic system, an assembly substrate and
lLEDs are placed in a bath chamber and a cluster of lLEDs is formed
via the axial rotational motion of magnets underneath the assembly
substrate (see the inset of Fig. 17). By embedding ferromagnetic nickel
into the n-type electrode of lLEDs, their movement is controlled by
magnets. By applying a localized DEP force centered around the recep-
tor holes, the lLEDs are captured and assembled at the receptor sites.
As shown in Fig. 17(a), COMSOL simulation results revealed that the
DEP force increased with a lower angle, indicating that the lLEDs
were pulled into the receptor hole as the angle decreased. Assembly
occurred when the angle was below 10and when the DEP force was
dominant over the magnetic force. Figures 17(b)–17(d) show lLEDs
arranged around the receptor holes in accordance with the movement
of the magnetic force, vibrating at the edge of the hole in response to
the rotational motion of the magnet. They are assembled when the
angle between the lLEDs and the receptor hole falls below a certain
angle. As presented in Fig. 17(e), the transfer yield initially increased
with an increasing peak-to-peak voltage related to DEP (V
pp
)before
starting to decrease. The proposed MDSAT approach has been shown
to achieve a 99.99% RGB LED simultaneous transfer yield within
15 min, positioning it as an excellent transfer technology candidate for
the mass production of commercial products.
Furthermore, Hwang et al.
221
developed a method for quickly
aligning lLEDs chips at the wafer scale by controlling the van der
Waals (vdW) force between the chip and the interposer, which they
TABLE II. Summary of monolithic and monolithic hybrid integration (flip-chip/wafer bonding). RGB: Red Green Blue (full color); IC: Integrated Circuit.
Display (pixel
density) Main technology Description Main systems Applications
Micro-display
(high pixels per
inch (PPI))
Monolithic integration
(homogeneous integration)
Selective epitaxial removal
and selective
area growth used for
monolithic integration
GaN-based active matrix
(AM), GaN-based hetero-
junction field effect tran-
sistor for GaN-based
LED
200
Augmented reality (AR)/
virtual reality (VR), head-
up display (HUD)
Monolithic hybrid integration
(heterogeneous integration)
Flip-chip or wafer bond-
ing for hybrid integration
A vertically stacked pas-
sive matrix lLEDs
array,
201
RGB LEDs on Si
complementary metal
oxide semiconductor AM
drivers
202–207
AR/VR, HUD
Mini-display (mid
PPI)
Heterogeneous integration Flip-chip/wafer bonding
used to integrate
micro-LED array with Si
IC backplanes
AM micro-LEDs on oxide
thin film transistor (TFT)
or low temperature
polysilicon-TFT
backplanes
203,208
.
Smartwatch, smartphone
TABLE III. Summary of mass transfer technologies.
Technology Transfer rate Throughput Reliability Scalability Description
Electrostatic
209
- Moderate Moderate Low Pick up lLEDs from donor wafer
with intermediate substrate and
place them on TFT or IC
backplane
Elastomer (transfer printing)
210,211
0.98–19.7 10
6
/h Moderate High High
Magnetic
212,213
- Moderate Moderate Low
Fluidic assembly
214–217
50 10
6
/h High High High Liquid used to cause LEDs to fill
in grooves in a target substrate
Laser
218
100 10
6
/h High Low High Laser irradiation discharges LED
chips from donor and place
them on the target substrate.
Roll-to-roll
219,220
10 000 10
6
/s High Moderate Moderate Roll transfer steps to an array of
interconnected devices onto the
target substrate.
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-22
V
CAuthor(s) 2024
16 May 2024 21:33:15
refer to as fluidic-assisted self-alignment transfer (FAST) [Fig. 17(f)].
In this technique, the top (Au face) and bottom (AlN face) of lLED
chips are designed to exhibit different vdW forces, enabling their selec-
tive bonding to the substrate in a specific pixel location under fluid
and dry treatment conditions.
As illustrated in Fig. 17(g), when the align bar is in contact with
the lLED chips (the contacting stage), a random force is applied in all
directions. The lLED chips can easily be flipped over because the
adhesion force between the lLED chips and substrate is low in wet
conditions. As the align bar sweeps over the lLED chips (the sweeping
stage), a dry region is formed under the AlN face because of the appli-
cation of pressure (the drying stage), which significantly increases the
vdW and adhesion forces and maintains the AlN
d
#state when resub-
merged in the solvent. (AlN
d
#,Au
d
#,andAu
w
#refer to the dry and
wet states of AlN and Au facing downward toward the substrate,
respectively.) In contrast, due to the high surface roughness of the Au
FIG. 17. (a) DEP force acting on lLEDs regarding its angle to the receptor hole. (b–d) Camera images of lLEDs during the three stages of the assembly process and sche-
matic of corresponding stages. (e) Transfer yield and DEP force dependence on changes in applied voltage. Reprinted with permission from Chang et al. Nature 617, 287
(2023). Copyright 2023 Springer Nature.
213
(f) Custom-made lLEDs chip dispenser. Chips are randomly oriented in solvent. (g) Formation of AlN
d
#state by pressure and dry-
ing of solvent. Reprinted with permission from Hwang et al., Nature Electron. 6, 216 (2023). Copyright 2023 Springer Nature.
221
(h) Integration of InGaN blue-green dual color
lLEDs and AlGaInP red lLEDs on CMOS. (i) A cross-sectional SEM image confirming successfully integrated lLEDs on CMOS via Au–Sn and Au–In flip-chip bonding.
Reprinted with permission from Qi et al., Light: Sci. Appl. 12, 258 (2023). Copyright 2023 Springer Nature.
222
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-23
V
CAuthor(s) 2024
16 May 2024 21:33:15
face, the solvent readily penetrates the space between Au and the sub-
strate, transitioning from a dry (Au
d
#)toawet(Au
w
#)stateand
reducing the vdW and adhesion forces, which is illustrated in the SEM
images in Fig. 17(g). Additionally, the AlN face is engineered to facili-
tate solvent washing and to accelerate the drying process. Thus, the
transition to AlN
d
#is irreversible due to high vdW and adhesion force,
while other transitions are reversible. This technology has been shown
to achieve the single-faced and irreversible alignment of 259 200 lLED
chips with an accuracy of 100% and a transfer yield of 99.992% over
40 trials.
Backplane devices with a high mobility are desirable for display
operations with a fast response. Commercially available backplane
devices such as amorphous Si TFTs, oxide TFTs, and low-temperature
polysilicon TFTs have mobilities of approximately 1, 10, and
100 cm
2
/VS, respectively. A low mobility can result in slower
response times, making the device unsuitable for AR displays.
Therefore, integration with a CMOS backplane with a mobility greater
than 1000 cm
2
/VS is desirable to fabricate high-performance lLED
displays. Recently, Qi et al.
222
demonstrated a full-color lLED display
on a CMOS backplane via FC bonding technology. As illustrated in
Fig. 17(h), an InGaN-based blue/green dual-color lLED array with a
subpixel size of 15 15 lm
2
was first bonded to the CMOS via Au-Sn
at 220 C, and then the silicon substrate was removed. AlGaInP-based
red lLEDs with a subpixel size of 15 15lm
2
were then bonded to
the CMOS via Au–In at 180 C, which is lower than the Au–Sn bond-
ing process and does not damage the previously bonded blue/green
lLED array. The GaAs substrate was then removed to open the light-
emitting area. Figure 17(i) presents an SEM image of heterogeneous
InGaN-based blue/green lLEDs and AlGaInP lLEDs on a CMOS for
full-color displays via Au–Sn and Au–In, respectively. In addition, the
color gamut in CIE 1931 and turned-on blue, green, and red lLEDs
confirmed the development of a III–V-based inorganic full-color dis-
play operating on a CMOS.
IX. DETECTION AND REPAIR PROCESSES
To implement lLED displays, a mass repair process is vital.
However, this section only briefly describes detection and repair tech-
nologies; for more detailed information, please refer to previous stud-
ies.
223–225
For commercial applications, a mass transfer yield of
99.9999% and the detection and repair of bad pixels after transfer are
required.
223
Bad pixels are known to be caused by open circuits or
short circuits that can occur during the LED fabrication and mass
transfer process.
The function of the lLEDs must not be affected by the detection
process for bad pixels, and the detection cost must be low. For these
reasons, PL and EL detection methods are widely used. PL detection is
promising due to its non-contact nature, and the spot size of the laser
can be tune d to less than 2 lm, allowing it to be used for the accurate
analysis of lLEDs.
226
However, with PL detection, the electrical char-
acteristics of lLEDs cannot be detected. To resolve this, PL detection
has been combined with confocal Raman microscopy,
227
while tuning
the laser excitation wavelength and power
228
has been shown to
improve the accuracy of PL detection. Thus, PL detection can be both
non-contact and highly accurate. Nevertheless, EL detection is a more
direct way to determine the optical and electrical properties of LEDs
and is considered more reliable than PL detection.
229
In conventional
LED detection, chips are inspected one by one. This electrical testing is
not a suitable process for millions of lLEDs. Thus, a method of using
one row or one column of an LED array in each test has been pro-
posed.
230
The main disadvantage of EL detection is the potential for
damage when the probes come into contact with the LED chip. To pre-
vent this damage, flexible probes based on the principle underlying
MEMS switches
231
and a non-electrical contact EL method (referred
to as capacitive current injection functional testing) have been adopted.
When comparing the two methods, PL detection tends to have a
higher detection efficiency than EL detection in terms of the number
of LED chips used for each detection cycle and the time required due
to differences in their detection principles. Recently, optical coherence
tomography (OCT) has also been proposed as a means to derive infor-
mation from inside lLEDs rather than the LED surface region.
232
OCT is promising due to its high depth resolution below 10 lm, high
detection sensitivity, and real-time imaging in non-invasive and non-
contact modes.
223
Redundant design and selective removal and replacement have
been the primary methods employed for repair. In redundant design,
two or more sets of LED chips are added during the fabrication of the
LED display. If some chips in the display fail, then the redundant chips
can take their place without the need for a repair. This design can
markedly increase the production yield. For example, a device with
twin-emitter lLEDs has been demonstrated,
233
where if one is dam-
aged the other can take over. For selective removal and replacement,
bad pixels are generally repaired with a pick-and-place technique. For
example, two parallel circuits can be preset, one of which is set up as a
redundant circuit.
234
A spare device consisting of three LEDs (RGB)
and a polydimethylsiloxane (PDMS) stamp can be placed nearby.
After removing the defective LEDs, the nearby spare devices are trans-
ferred to a redundancy circuit for repair. A hybrid repair method com-
bining UV selective irradiation and pick-and-place technique has also
been developed,
235
while a laser-based mass transfer method has been
employed to detect and repair defective bad LEDs before mass transfer.
For example, a technology with a high placement rate was developed
by shifting between single-beam and multi-beam lasers for bad chip
removal and good chip transfer.
218
This process was reported to be
suitable for detection before mass transfer, which can improve the
transfer yield.
Redundant design with in situ self-repair has been shown to be
more efficient than selective removal and replacement with ex situ
repair.
223
However, the former strategy faces some practical issues,
such as extra steps during mass transfer, the need for extra circuits and
space, and the low probability of the spare chips being used.
Conversely, the latter is more cost-effective and advantageous for the
high integration of chips due to the lack of additional mass transfer
steps and the large space available for high integration. In conclusion,
pick-and-place technology may be more advantageous for repairing
smaller lLEDs.
207
X. INDUSTRY EFFORTS
Collaboration between academia and industry is important to
accelerate the commercialization of lLED displays. Therefore, it is nec-
essary to understand industry concerns, requirements, and approaches.
In this section, we briefly review the results reported by global compa-
nies on lLED displays. For example, Samsung Display reported that
the EQE of nLEDs depends on the material that is deposited on the
sidewall first. As shown in Fig. 18(a),HfO
2
-passivated devices outper-
form SiO
2
-passivated ones. This is due to the fact that the HfO
2
layer
more effectively prevents the additional damage that can occur during
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-24
V
CAuthor(s) 2024
16 May 2024 21:33:15
the deposition of the second passivation layer. In addition, Lumileds
compared the EQEs of InGaN- and AlGaInP-based red lLEDs as a
function of their size and current density [Fig. 18(b)]. Because of the
impact of carrier diffusion, the EQE of AlGaInP-based red lLEDs
decreased as the size was reduced (from 9 to 4lm), whereas that of
InGaN red lLEDs was almost constant regardless of the size (moving
from 6.7 to 2.2 lm) and the current density. As a result, InGaN-based
red lLEDs appear to be more suitable for devices that are much
smaller and operate at lower currents. It is also expected that InGaN-
or AlGaInP-based red lLEDs can be strategically used to fabricate
lLED displays depending on the target resolution, operating current
regime, and environment. The EQE of InGaN red LEDs grown using
MOCVD demonstrated by Samsung Electronics is around 15%, sug-
gesting that InGaN red LEDs can operate with a high EQE for ultra-
high resolution displays with further development [Fig. 18(c)].
LG display also demonstrated a 100 PPI 12-inch active-matrix
lLED stretchable display using polyimide (PI). Figure 18(d) presents
an image of this stretchable display, which has a stretchability of 20%,
a transparency of 39%, and a luminance of 1000 cd/m
2
,allbasedona
lLED light source. This stretchable lLED display could potentially be
applied to a variety of display designs once lLED technology becomes
more mature. In addition, Jade Bird Display developed technology to
achieve over 5000 PPI using wafer-level monolithic integration
between InGaN-based blue/green and GaAs-based red lLEDs and
CMOS IC wafers. This technology is based on eutectic metal bonding
between the LED and CMOS IC wafers, the removal of the substrate
via laser liftoff for the InGaN-based blue/green LEDs, and selective wet
chemical etching for the GaAs-based red lLEDs. Figure 18(e) displays
nine micro-display device dies with a pixel mesa size of 6 lmona4-
inch wafer. This successful wafer-level device demonstrates that eutec-
tic metal bonding can be used to integrate lLED arrays and CMOS IC
drivers for mass production.
XI. CONCLUSION AND OUTLOOK
The present review describes recent advances in lLEDs based on
compound semiconductors, AlGaInP, and InGaN. As the area ratio of
the sidewall surface increases with chip miniaturization, light extrac-
tion and surface recombination at the sidewall need to be considered
seriously. The ABC model is a suitable guide for the design of chip
structures if physical parameters such as the diffusion length, which
reflects the mobility and lifetime of excess carriers in the active layer,
and the recombination rate at the sidewall surface can be accurately
determined. Reducing sidewall damage is an important issue for the
fabrication of high-efficiency lLED chips; controlling the power input
during dry etching, chemical treatment such as TMAH, and surface
passivation with ALD are effective in reducing this damage. However,
it is important to consider whether there are other options available
that can address this damage. The formation of insulating regions on
FIG. 18. (a) Importance of first passivation layer for suppressing the impact of sidewall surface recombination. Reprinted with permission from Cho et al. J. Soc. Inf. Disp. 31,
289 (2023). Copyright 2023 Wiley VCH.
50
(b) Revealing advantages and disadvantages of InGaN- and AlGaInP-based red lLEDs. Reprinted with permission from Moran et al.
Dig. Tech. Pap. - SID Int. Symp. 54, 414 (2023). Copyright 2023 Wiley VCH.
236
(c) Demonstration that InGaN red LEDs can achieve over 15% EQE at 630 nm wavelength.
Reproduced from Lee et al., Appl. Phys. Lett. 124, 121109 (2024), with permission from AIP Publishing.
171
(d) Demonstration of flexible display based on lLEDs technology.
Reprinted with permission from Jung et al., J. Soc. Inf. Disp. 31, 201 (2023). Copyright 2023 Wiley VCH.
237
(e) Demonstration of the mass production of lLEDs displays on
CMOS IC driver through the eutectic metal bonding. Reprinted with permission from Zhang et al., J. Soc. Inf. Disp. 26, 137 (2018). Copyright 2018 Wiley VCH.
204
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-25
V
CAuthor(s) 2024
16 May 2024 21:33:15
sidewalls via ion implantation is also effective but increases the chip
area. This also occurs in truncated chip structures, which increases
LEE at the expense of a larger chip area. When reducing the chip size
to 100 lm
2
or lower, especially for red LEDs, it is still an open question
as to which method should be used. AlGaInP-based red LEDs have the
problem of a long carrier diffusion length and/or a high surface recom-
bination rate, thus the EQE decreases dramatically when the chip size
decreases. Similarly, InGaN-based red LEDs still exhibit a low EQE
due to defects, strain, reduced electron–hole overlap. However,
through collaboration between academia and industry, it is expected
that these difficulties can be overcome and micro LED displays can be
commercialized. It is almost important to note that the recent develop-
ment of several mass transfer technologies has enabled the very rapid
and reliable assembly of lLED array. However, if the chip size is
reduced to 100 lm
2
or lower, other methods, such as wafer-based tech-
nologies, should be considered. Overall, efficiency is a critical issue for
the commercial implementation of lLED displays, and lLED
researchers are expected to play a more active role than ever before.
ACKNOWLEDGMENTS
T.-Y.S gratefully acknowledges financial support from the
National Research Foundation of Korea funded by the Ministry of
Science and ICT (No. NRF-2022R1A2C2006887).
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Jeong-Hwan Park: Conceptualization (lead); Investigation (equal);
Writing –original draft (equal). Markus Pristovsek: Conceptualization
(supporting); Writing –original draft (equal); Writing –review &
editing (equal). Hiroshi Amano: Conceptualization (supporting);
Funding acquisition (equal); Writing –original draft (equal); Writing –
review & editing (equal). Tae-Yeon Seong: Conceptualization (sup-
porting); Funding acquisition (equal); Writing –original draft (equal);
Writing –review & editing (equal).
DATA AVAILABILITY
Data sharing is not applicable to this article as no new data were
created or analyzed in this study.
REFERENCES
1
Z. Chen, S. Yan, and C. Danesh, J. Phys. D. Appl. Phys. 54, 123001 (2021).
2
T. Y. Lee, L. Y. Chen, Y. Y. Lo, S. S. Swayamprabha, A. Kumar, Y. M. Huang,
S. C. Chen, H. W. Zan, F. C. Chen, R. H. Horng, and H. C. Kuo, ACS
Photonics 9, 2905 (2022).
3
J. E. Ryu, S. Park, Y. Park, S. W. Ryu, K. Hwang, and H. W. Jang, Adv. Mater.
35, 2204947 (2023).
4
V. Lee, Proc. SPIE 11310, 113102S (2020).
5
K. Behrman and I. Kymissis, Nat. Electron. 5, 564 (2022).
6
X. He, Micro-LED Displays 2021-2031: Technology, Commercialization,
Opportunity, Market and Players (IDTechEx, 2021).
7
E. F. Schubert and J. K. Kim, Science 308, 1274 (2005).
8
T. Y. Seong, J. Han, H. Amano, and H. Morkoc, III-Nitride Based Light
Emitting Diodes and Applications (Springer, 2013).
9
J. Cho, J. H. Park, J. K. Kim, and E. F. Schubert, Laser Photonics Rev. 11,
1600147 (2017).
10
Y. Takubo, Y. Hisatake, T. Iizuka, and T. Kawamura, SID Symp. Dig. Tech.
Papers. 43, 869 (2012).
11
S. Hang, C.-M. Chuang, Y. Zhang, C. Chu, K. Tian, Q. Zheng, T. Wu, Z. Liu,
Z.-H. Zhang, Q. Li, and H.-C. Kuo, J. Phys. D. Appl. Phys. 54, 153002 (2021).
12
B. Lu, L. Wang, Z. Hao, Y. Luo, C. J. Chen, M. C. Wu, J. Tang, J. K. Kim, and
E. F. Schubert, Laser Photonics Rev. 16, 2100433 (2022).
13
I. Akasaki, Rev. Mod. Phys. 87, 1119 (2015).
14
H. Amano, Rev. Mod. Phys. 87, 1133 (2015).
15
S. Nakamura, Rev. Mod. Phys. 87, 1139 (2015).
16
D. Iida and K. Ohkawa, Semicond. Sci. Technol. 37, 013001 (2022).
17
Z. Zhuang, D. Iida, and K. Ohkawa, Jpn. J. Appl. Phys. 61, SA0809 (2022).
18
P. J. Parbrook, B. Corbett, J. Han, T. Y. Seong, and H. Amano, Laser
Photonics Rev. 15, 2000133 (2021).
19
Y.-Z. Lin, C. Liu, J.-H. Zhang, Y.-K. Yuan, W. Cai, L. Zhou, M. Xu, L. Wang,
W.-J. Wu, and J.-B. Peng, IEEE Trans. Electron Devices 11, 5656 (2021).
20
T. Jin, S. Kim, J. H. Han, D. H. Ahn, S. U. An, T. H. Noh, X. Sun, C. J. Kim, J.
Park, and Y. Kim, Nanoscale Adv. 5, 1316 (2022).
21
D. Chen, Z. Liu, X. Lu, L. Wan, R. Li, Z. Yang, and G. Li, J. Mater. Chem. C 7,
2823 (2019).
22
Y. Cai, C. Zhu, W. Zhong, P. Feng, S. Jiang, and T. Wang, Adv. Mater.
Technol. 6, 2100214 (2021).
23
C. J. Chen, H. C. Chen, J. H. Liao, C. J. Yu, and M. C. Wu, IEEE J. Quantum
Electron. 55, 3300106 (2019).
24
L. Qi, X. Zhang, W. C. Chong, P. Li, and K. M. Lau, Opt. Express 29, 10580
(2021).
25
H. Kawanishi, H. Onuma, M. Maegawa, T. Kurisu, T. Ono, S. Akase, S.
Yamaguchi, N. Momotani, Y. Fujita, Y. Kondo, K. Kubota, T. Yoshida, Y.
Ikawa, T. Ono, H. Higashisaka, Y. Hirono, and S. Anzai, J. Soc. Inf. Disp. 29,
57 (2021).
26
D. Chen, Y.-C. Chen, G. Zeng, D. W. Zhang, and H.-L. Lu, Research 6, 0047 (2023).
27
A. David, N. G. Young, C. Lund, and M. D. Craven, ECS J. Solid State Sci.
Technol. 9, 016021 (2020).
28
S. Karpov, Opt. Quant. Electron 47, 1293 (2015).
29
R. T. Ley, J. M. Smith, M. S. Wong, T. Margalith, S. Nakamura, S. P.
DenBarrs, and M. J. Gordon, Appl. Phys. Lett. 116, 251104 (2020).
30
S. Bornemann, J. Gulink, V. Moro, J. C. Gil, S. Wolter, G. Schottler, D.
Bezshlyakh, J. D. Prades, A. Dieguez, and A. Waag, IEEE Photonics J. 13,
8200209 (2021).
31
Q. Dai, Q. Shan, J. Wang, S. Chhajed, J. Cho, E. F. Schubert, M. H. Crawford,
D. D. Koleske, M.-H. Kim, and Y. Park, Appl. Phys. Lett. 97, 133507 (2010).
32
M. Pristovsek, Phys. Status Solidi RRL 17, 2200331 (2023).
33
F. jiang, B. R. Hyun, Y. Zhang, and Z. Liu, Phys. Status Solidi - Rapid Res.
Lett. 15, 2000487 (2021).
34
Q. Dai, M. F. Schubert, M. H. Kim, J. K. Kim, E. F. Schubert, D. D. Koleske,
M. H. Crawford, S. R. Lee, A. J. Fischer, G. Thaler, and M. A. Banas, Appl.
Phys. Lett. 94, 111109 (2009).
35
J. L€
ahnemann, V. M. Kaganer, K. K. Sabelfeld, A. E. Kireeva, U. Jahn, C.
Ch
eze, R. Calarco, and O. Brandt, Phys. Rev. Appl. 17, 024019 (2022).
36
F. C. P. Massabuau, M. J. Davies, F. Oehler, S. K. Pamenter, E. J. Thrush, M. J.
Kappers, A. Kov
acs, T. Williams, M. A. Hopkins, C. J. Humphreys, P.
Dawson, R. E. Dunin-Borkowski, J. Etheridge, D. W. E. Allsopp, and R. A.
Oliver, Appl. Phys. Lett. 105, 112110 (2014).
37
P. Kirilenko, Z. Zhuang, D. Iida, M. Velazquez-Rizo, and K. Ohkawa, Crystals
11, 1123 (2021).
38
J. Smalc-Koziorowska, E. Grzanka, R. Czernecki, D. Schiavon, and M.
Leszczy
nski, Appl. Phys. Lett. 106, 101905 (2015).
39
F. Olivier, A. Daami, C. Licitra, and F. Templier, Appl. Phys. Lett. 111,
022104 (2017).
40
J. Ha, S. Kim, D. Song, S. Park, J. Park, J. Choi, and C. Lee, Information
Display 39, 15 (2023).
41
J. J. Wierer and N. Tansu, Laser Photonics Rev. 13, 1900141 (2019).
42
F. Olivier, S. Tirano, L. Dupr
e, B. Aventurier, C. Largeron, and F. Templier,
J. Lumin. 191, 112 (2017).
43
J. H. Park, M. Pristovsek, W. Cai, H. Cheong, A. Tanaka, Y. Furusawa, D. P.
Han, T. Y. Seong, and H. Amano, Adv. Opt. Mater. 11, 2203128 (2023).
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-26
V
CAuthor(s) 2024
16 May 2024 21:33:15
44
S. Yamada, H. Sakurai, Y. Osada, K. Furuta, T. Nakamura, R. Kamimura, T.
Narita, J. Suda, and T. Kachi, Appl. Phys. Lett. 118, 102101 (2021).
45
M. Boroditsky, I. Gontijo, M. Jackson, R. Vrijen, E. Yablonovitch, T. Krauss,
C. C. Cheng, A. Scherer, R. Bhat, and M. Krames, J. Appl. Phys. 87, 3497
(2000).
46
S. Y. Karpov and Y. N. Makarov, Appl. Phys. Lett. 81, 4721 (2003).
47
A. David, Phys. Rev. Appl. 15, 054015 (2021).
48
C. Becht, U. T. Schwarz, M. Binder, and B. Galler, Phys. Status Solidi Basic
Res. 260, 2200565 (2023).
49
Y. Y. Li, F. Z. Lin, K. L. Chi, S. Y. Weng, G. Y. Lee, H. C. Kuo, and C. C. Lin,
IEEE Photonics J. 14, 7007907 (2022).
50
H. Cho, D. Kim, S. Lee, C. Yoo, and Y. Sim, J. Soc. Inf. Disp. 31, 289 (2023).
51
J.-H. Park, M. Pristovsek, W. Cai, H. Cheong, C.-M. Kang, D.-S. Lee, T.-Y.
Seong, and H. Amano, Laser Photonics Rev. 17, 2300199 (2023).
52
M. Auf Der Maur, A. Pecchia, G. Penazzi, W. Rodrigues, and A. Di Carlo,
Phys. Rev. Lett. 116, 027401 (2016).
53
T. Y. Seong and H. Amano, Surf. Interfaces 21, 100765 (2020).
54
S. Chevtchenko, X. Ni, Q. Fan, A. A. Baski, and H. MorkoR,Appl. Phys. Lett.
88, 122104 (2006).
55
L. Lymperakis, P. H. Weidlich, H. Eisele, M. Schnedler, J. P. Nys, B.
Grandidier, D. Sti
evenard, R. E. Dunin-Borkowski, J. Neugebauer, and P.
Ebert, Appl. Phys. Lett. 103, 152101 (2013).
56
P. Reddy, I. Bryan, Z. Bryan, W. Guo, L. Hussey, R. Collazo, and Z. Sitar,
J. Appl. Phys. 116, 123701 (2014).
57
I. Barto
s, O. Romanyuk, J. Houdkova, P. P. Paskov, T. Paskova, and P. Ji
rí
cek,
J. Appl. Phys. 119, 105303 (2016).
58
A. J. Bard, A. B. Bocarsly, F. F. Fan, E. G. Walton, and M. S. Wrighton, J. Am.
Chem. Soc. 102, 3671 (1980).
59
A. B. Bocarsly, D. C. Bookbinder, R. N. Dominey, N. S. Lewis, and M. S.
Wrighton, J. Am. Chem. Soc. 102, 3683 (1980).
60
R. T. Tung, Phys. Rev. Lett. 84, 6078 (2000).
61
A. Dimoulas, P. Tsipas, A. Sotiropoulos, and E. K. Evangelou, Appl. Phys.
Lett. 89, 252110 (2006).
62
J. M. Smith, R. Ley, M. S. Wong, Y. H. Baek, J. H. Kang, C. H. Kim, M. J.
Gordon, S. Nakamura, J. S. Speck, and S. P. Denbaars, Appl. Phys. Lett. 116,
071102 (2020).
63
H. Kitagawa, M. Fujita, T. Suto, T. Asano, and S. Noda, Appl. Phys. Lett. 98,
181104 (2011).
64
K. A. Bulashevich and S. Y. Karpov, Phys. Status Solidi - Rapid Res. Lett. 10,
480 (2016).
65
S. S. Konoplev, K. A. Bulashevich, and S. Y. Karpov, Phys. Status Solidi Appl.
Mater. Sci. 215, 1700508 (2018).
66
P. Li, H. Li, H. Zhang, C. Lynsky, M. Iza, J. S. Speck, S. Nakamura, and S. P.
Denbaars, Appl. Phys. Lett. 119, 081102 (2021).
67
N. Lim, P. Chan, H. Chang, V. Rienzi, M. J. Gordon, and S. Nakamura, Adv.
Photonics Res. 4, 2200286 (2023).
68
J.-T. Oh, S.-Y. Lee, Y.-T. Moon, J. H. Moon, S. Park, K. Y. Hong, K. Y. Song,
C. Oh, J.-I. Shim, H.-H. Jeong, J.-O. Song, H. Amano, and T.-Y. Seong, Opt.
Express 26, 11194 (2018).
69
Y. Boussadi, N. Rochat, J. P. Barnes, B. B. Bakir, P. Ferrandis, B. Masenelli,
and C. Licitra, J. Lumin. 234, 117937 (2021).
70
S. Finot, C. Le Maoult, E. Gheeraert, D. Vaufrey, and G. Jacopin, ACS
Photonics 9, 173 (2022).
71
L. Yu, B. Lu, P. Yu, Y. Wang, G. Ding, Q. Feng, Y. Jiang, H. Chen, K. Huang,
Z. Hao, J. Yu, Y. Luo, C. Sun, B. Xiong, Y. Han, J. Wang, H. Li, and L. Wang,
Appl. Phys. Lett. 121, 042106 (2022).
72
I. H. Lee, T. H. Kim, A. Y. Polyakov, A. V. Chernykh, M. L. Skorikov, E. B.
Yakimov, L. A. Alexanyan, I. V. Shchemerov, A. A. Vasilev, and S. J. Pearton,
J. Alloys Compd. 921, 166072 (2022).
73
C. T. Sah, R. N. Noyce, and W. Shockley, Proc. IRE 45, 1228 (1957).
74
M. S. Wong, C. Lee, D. J. Myers, D. Hwang, J. A. Kearns, T. Li, J. S. Speck, S.
Nakamura, and S. P. Denbaars, Appl. Phys. Express 12, 097004 (2019).
75
M. Sheen, Y. Ko, D. Kim, J. Kim, J. Byun, Y. S. Choi, J. Ha, K. Y. Yeon, D.
Kim, J. Jung, J. Choi, R. Kim, J. Yoo, I. Kim, C. Joo, N. Hong, J. Lee, S. H.
Jeon, S. H. Oh, J. Lee, N. Ahn, and C. Lee, Nature 608, 56 (2022).
76
J. Yu, T. Tao, B. Liu, F. Xu, Y. Zheng, X. Wang, Y. Sang, Y. Yan, Z. Xie, S. Liang,
D. Chen, P. Chen, X. Xiu, Y. Zheng, and R. Zhang, Crystals 11, 403 (2021).
77
R.-H. Horng, C.-X. Ye, P.-W. Chen, D. Iida, K. Ohkawa, Y.-R. Wu, and D.-S.
Wuu, Sci. Rep. 12, 1324 (2022).
78
F. Yang, Y. Xu, L. Li, X. Cai, J. Li, J. Tao, S. Zheng, B. Cao, and K. Xu, J. Phys.
D: Appl. Phys. 55, 435103 (2022).
79
J.-H. Park, M. Pristovsek, H. Cheong, W. Cai, T. Kumabe, D.-S. Lee, T.-Y.
Seong, and H. Amano, Opt. Lett. 47, 2250 (2022).
80
H.-H. Huang, S.-K. Huang, Y.-L. Tsai, S.-W. Wang, Y.-Y. Lee, S.-Y. Weng,
H.-C. Kuo, and C.-C. Lin, Opt. Express 28, 38184 (2020).
81
S. Gandrothula, T. Kamikawa, P. Shapturenka, R. Anderson, 1. M. Wong, H.
Zhang, J. S. Speck, S. Nakamura, and S. P. Denbaars, Appl. Phys. Lett. 119,
142103 (2021).
82
H. Li, M. S. Wong, M. Khoury, B. Bonef, H. Zhang, Y. Chow, P. Li, J. Kearns,
A. A. Taylor, P. De Mierry, Z. Hassan, S. Nakamura, and S. P. DenBaars, Opt.
Express 27, 24154 (2019).
83
M. S. Wong, D. Hwang, A. I. Alhassan, C. Lee, R. Ley, S. Nakamura, and S. P.
DenBaars, Opt. Express 26, 21324 (2018).
84
Y.-W. Yeh, S.-H. Lin, T.-C. Hsu, S. Lai, P.-T. Lee, S.-Y. Lien, D.-S. Wuu, G. Li,
Z. Chen, T. Wu, and H.-C. Kuo, Nanoscale Res. Lett. 16, 164 (2021).
85
T.-Y. Lee, Y.-M. Huang, H. Chiang, C. Lichao, C.-Y. Hung III, W.-H. Kuo, Y.-
H. Fang, M.-T. Chu, C.-I. Wu, C.-C. Lin, and H.-C. Kuo, Opt. Express 30,
18552 (2022).
86
J. Park, W. Baek, D.-M. Geum, and S. Kim, Nanoscale Res. Lett. 17, 29 (2022).
87
M. S. Wong, J. A. Kearns, C. Lee, J. M. Smith, C. Lynsky, G. Lheureux, H.
Choi, J. Kim, C. Kim, S. Nakamura, J. S. Speck, and S. P. DenBaars, Opt.
Express 28, 5787 (2020).
88
Y. Liu, F. Feng, K. Zhang, F. Jiang, K.-W. Chan, H.-S. Kwok, and Z. Liu,
J. Phys. D: Appl. Phys. 55, 315107 (2022).
89
D.-H. Lee, J.-H. Lee, J.-S. Park, T.-Y. Seong, and H. Amano, ECS J. Solid State
Sci. Technol. 9, 055001 (2020).
90
D. Chen, Z. Wang, F.-C. Hu, C. Shen, N. Chi, W. Liu, D. W. Zhang, and H.-L.
Lu, Opt. Express 29, 36559 (2021).
91
M. Zhang, S. Hang, C. Chu, H. Shao, Y. Zhang, Y. Zhang, Y. Zhang, Q. Zheng,
Q. Li, and Z.-H. Zhang, IEEE Trans. Electron Dev. 69, 3213 (2022).
92
C.-F. Lin, C.-M. Lin, and R.-H. Jiang, Jpn. J. Appl. Phys. 51, 01AG03 (2012).
93
W. Huang, X. Miao, and Z. Liu, Micromachines 14, 566 (2023).
94
P. Kirilenko, D. Iida, Z. Zhuang, and K. Ohkawa, Appl. Phys. Express 15,
084003 (2022).
95
Z. Zhuang, D. Iida, M. Velazquez-Rizo, and K. Ohkawa, Opt. Lett. 46, 5092
(2021).
96
S. Liu, S. Han, C. Xu, H. Xu, X. Wang, D. Wang, and Y. Zhu, Opt. Mater. 121,
111579 (2021).
97
Y.-H. Hsu, C.-H. Wang, X.-D. Lin, Y.-H. Lin, D.-S. Wuu, and R.-H. Horng,
Discover Nano 18, 48 (2023).
98
X. Wang, X. Zhao, T. Takahashi, D. Ohori, and S. Samukawa, Nat. Commun.
14, 7569 (2023).
99
M. Tian, H. Yu, M. H. Memon, Z. Xing, C. Huang, H. Jia, H. Zhang, D.
Wang, S. Fang, and H. Sun, Opt. Lett. 46, 4809 (2021).
100
R. Floyd, M. Gaevski, K. Hussain, A. Mamun, M. V. S. Chandrashekhar, G.
Simin, and A. Khan, Appl. Phys. Express 14, 084002 (2021).
101
P. Tian, X. Shan, S. Zhu, E. Xie, J. J. D. McKendry, E. Gu, and M. D. Dawson,
IEEE J. Quantum Electron. 58, 3300214 (2022).
102
Y. Liu, T. Xia, A. Du, T. Liang, Z. Fan, E. Chen, J. Sun, Q. Yan, and T. Guo,
Opt. Lett. 48, 1650 (2023).
103
K. A. Bulashevich, S. S. Konoplev, and S. Y. Karpov, Photonics 5, 41 (2018).
104
L. Tan, Q. Zhou, W. Hu, H. Wang, and R. Yao, Appl. Sci. 9, 3458 (2019).
105
L. Tan, Q. Zhou, H. Wang, and R. Yao, IEEE Photonics Technol. Lett. 31,
1705 (2019).
106
G. Yan, B.-R. Hyun, F. Jiang, H.-C. Kuo, and Z. Liu, Opt. Express 29, 26255
(2021).
107
S. H. Oh, T. H. Lee, K.-R. Son, and T. G. Kim, J. Alloy. Comp. 773,490
(2019).
108
J. Bai, Y. Cai, P. Feng, P. Fletcher, C. Zhu, Y. Tian, and T. Wang, ACS Nano
14, 6906 (2020).
109
K.-P. Chang, P.-C. Lien, C.-C. Yen, P.-W. Chen, R.-H. Horng, and D.-S. Wuu,
IEEE Photonics Technol. Lett. 33, 1375 (2021).
110
H. M. Ku, C. Y. Huang, C. Z. Liao, and S. Chao, Jpn. J. Appl. Phys. 50,
04DG07 (2011).
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-27
V
CAuthor(s) 2024
16 May 2024 21:33:15
111
I. Y. Hong, A. B. M. H. Islam, T. K. Kim, Y.-J. Cha, and J. S. Kwak, Appl. Surf.
Sci. 512, 145698 (2020).
112
Z. Gong, Y. F. Zhang, P. Kelm, I. M. Watson, E. Gu, and M. D. Dawson, Appl.
Phys. A 103, 389 (2011).
113
S.-Y. Lee, E. Lee, J.-H. Moon, B. Choi, J.-T. Oh, H.-H. Jeong, T.-Y. Seong, and
H. Amano, IEEE Photonics Technol. Lett. 32, 1041 (2020).
114
H. Wang, L. Wang, J. Sun, T.-L. Guo, E.-G. Chen, X.-T. Zhou, Y.-A. Zhang,
and Q. Yan, Displays 73, 102172 (2022).
115
S. Lan, H. Wan, J. Zhao, and S. Zhou, Micromachines 10, 860 (2019).
116
H.-Y. Ryu, J. Pyo, and H. Y. Ryu, IEEE Photonics J. 12, 1600110 (2020).
117
Y. Mei, M. Xie, T. Yang, X. Hou, W. Ou, H. Long, L. Ying, Y. Liu, G. Weng, S.
Chen, and B. Zhang, ACS Photonics 9, 3967 (2022).
118
M. Asad, Q. Li, M. Sachdev, and W. S. Wong, ACS Appl. Electron. Mater. 3,
882 (2021).
119
F. Feng, Y. Liu, K. Zhang, M. Zhanghu, K.-W. Chan, K. Xu, H.-S. Kwok, and Z.
Liu, Appl. Phys. Lett. 121, 221104 (2022).
120
M. S. Wong, S. H. Oh, J. Back, C. Lee, J. S. Speck1, S. Nakamura, and S. P.
DenBaars, Jpn. J. Appl. Phys. 60, 020905 (2021).
121
P. Gonz
alez-Izquierdo, N. Rochat, D. Zoccarato, F. Rol, J. Simon, P. Le Maitre,
M. Volpert, M. Charles, M. Lafossas, S. Torrengo, and Ł. Borowik, ACS
Photonics 10, 4031 (2023).
122
F. V€
ogl, A. Avramescu, F. Knorr, A. Lex, A. Waag, M. Hetzl, and N. von
Malm, Opt. Express 31, 22997 (2023).
123
H. Li, H. Zhang, P. Li, M. S. Wong, Y. C. Chow, S. Pinna, J. Klamkin, P.
DeMierry, J. S. Speck, S. Nakamura, and S. P. DenBaars, J. Phys. Photonics 2,
031003 (2020).
124
S.-W. H. Chen, Y.-M. Huang, K. J. Singh, Y.-C. Hsu, F.-J. Liou, J. Song, J.
Choi, P.-T. Lee, C.-C. Lin, Z. Chen, J. Han, T. Wu, and H.-C. Kuo, Photonics
Res. 8, 630 (2020).
125
A. Dadgar, Phys. Status Solidi B 252, 1063 (2015).
126
T. Kamikawa, T. Kobayashi, Y. Aoki, N. Suda, H. Ogura, M. Seida, K.
Takeuchi, K. Mishima, Y. Taniguchi, F. Yamashita, Y. Hayashi, and K. Masaki,
Cryst. Growth Des. 23, 4855 (2023).
127
J. Oh, D. Kim, D. Yang, K. Hwang, J. Hwang, J. Kim, S. Lee, J. Ryu, S. Park, J.
K. Shin, Y. Kim, Y. Park, E. Yoon, and H. W. Jang, ACS Appl. Mater.
Interfaces 14, 25781 (2022).
128
G. B. Lin, D. Y. Kim, Q. Shan, J. Cho, E. F. Schubert, H. Shim, C. Sone, and J.
K. Kim, IEEE Photonics J. 5, 1600207 (2013).
129
J. H. Park, W. Cai, H. Cheong, Y. Ushida, D. H. Lee, Y. Ando, Y. Furusawa, Y.
Honda, D. S. Lee, T. Y. Seong, and H. Amano, J. Appl. Phys. 131, 153104
(2022).
130
Z. Yuan, Y. Li, X. Lu, Z. Wang, P. Qiu, X. Cui, P. Tian, Q. Wang, and G.
Zhang, IEEE Trans. Electron Devices 69, 4298 (2022).
131
W. J. Baek, J. Park, J. Shim, B. H. Kim, S. Park, H. S. Kim, D. M. Geum, and S.
H. Kim, Nat. Commun. 14, 1386 (2023).
132
D. Hwang, A. J. Mughal, M. S. Wong, A. I. Alhassan, S. Nakamura, and S. P.
Denbaars, Appl. Phys. Express 11, 012102 (2018).
133
P. Li, H. Li, Y. Yao, H. Zhang, C. Lynsky, K. S. Qwah, M. Iza, J. S. Speck, S.
Nakamura, and S. P. DenBaars, Opt. Express 29, 22001 (2021).
134
M. S. Wong, A. Raj, H. M. Chang, V. Rienzi, F. Wu, J. J. Ewing, E. S. Trageser,
S. Gee, N. C. Palmquist, P. Chan, J. H. Kang, J. S. Speck, U. K. Mishra, S.
Nakamura, and S. P. Denbaars, AIP Adv. 13, 015107 (2023).
135
T. Saito, N. Hasegawa, K. Imura, Y. Suehiro, T. Takeuchi, S. Kamiyama, D.
Iida, K. Ohkawa, and M. Iwaya, Appl. Phys. Express 16, 084001 (2023).
136
B. L. Smith and M. Abbott, Solid State Electron. 15, 361 (1972).
137
M. L. Young and D. R. Wight, J. Phys. D. Appl. Phys. 7, 1824 (1974).
138
P. L. Gourley, R. M. Biefeld, T. E. Zipperian, and J. J. Wiczer, Appl. Phys. Lett.
44, 983 (1984).
139
T. Meoded, R. Shikler, N. Fried, and Y. Rosenwaks, Appl. Phys. Lett. 75, 2435
(1999).
140
L. W. Aukerman, M. F. Millea, and M. McColl, J. Appl. Phys. 38, 685 (1967).
141
T. S. Rao-Sahib and D. B. Wittry, J. Appl. Phys. 40, 3745 (1969).
142
R. D. Ryan and J. E. Eberhardt, Solid State Electron. 15, 865 (1972).
143
H. C. Casey, B. I. Miller, and E. Pinkas, J. Appl. Phys. 44, 1281 (1973).
144
L. Jastrzebski, J. Lagowski, and H. C. Gatos, Appl. Phys. Lett. 27, 537 (1975).
145
M. Niemeyer, J. Ohlmann, A. W. Walker, P. Kleinschmidt, R. Lang, T.
Hannappel, F. Dimroth, and D. Lackner, J. Appl. Phys. 122, 115702 (2017).
146
T. S. Ugahara, H. S. Ato, M. H. Ao, Y. N. Aoi, S. K. Urai, and S. T. Ottori, Jpn.
J. Appl. Phys. 37, 398 (1998).
147
J. C. Gonzalez, K. L. Bunker, and P. E. Russell, Appl. Phys. Lett. 79, 1567
(2001).
148
K. Kumakura, T. Makimoto, N. Kobayashi, T. Hashizume, T. Fukui, and H.
Hasegawa, Appl. Phys. Lett. 86, 052105 (2005).
149
G. P. Yablonskii, A. L. Gurskii, V. N. Pavlovskii, E. V. Lutsenko, V. Z. Zubialevich,
T. S. Shulga, A. I. Stognij, H. Kalisch, A. Szymakowski, R. H. Jansen, A. Alam, B.
Schineller, and M. Heuken, J. Cryst. Growth 275, e1733 (2005).
150
S. Hafiz, F. Zhang, M. Monavarian, V. Avrutin, H. Morkoc¸,
€
U.
€
Ozg€
ur, S.
Metzner, F. Bertram, J. Christen, and B. Gil, J. Appl. Phys. 117, 013106 (2015).
151
K. Fan, J. Tao, Y. Zhao, P. Li, W. Sun, L. Zhu, J. Lv, Y. Qin, Q. Wang, J. Liang,
and W. Wang, Results Phys. 36, 105449 (2022).
152
S. Han, C. Xu, H. Li, S. Liu, H. Xu, Y. Zhu, A. Fang, and X. Wang, Opt. Mater.
114, 110860 (2021).
153
Y. Zhao, J. Liang, Q. Zeng, Y. Li, P. Li, K. Fan, W. Sun, J. Lv, Y. Qin, Q. Wang,
J. Tao, and W. Wang, Opt. Express 29, 20217 (2021).
154
B. O. Jung, W. Lee, J. Kim, M. Choi, H. Y. Shin, M. Joo, S. Jung, Y. H. Choi,
and M. J. Kim, Sci. Rep. 11, 4535 (2021).
155
C.-K. Yee, J.-M. Lin, M.-J. Wu, H.-T. Cheng, C.-W. Huang, C.-A. Lee, K.-H.
Lin, C.-C. Wu, and C.-H. Wu, Opt. Lett. 48, 2933 (2023).
156
M. S. Wong, R. C. White, S. Gee, T. Tak, S. Gandrothula, H. Choi, S. Nakamura,
J. S. Speck, and S. P. DenBaars, Appl. Phys. Express 16, 066503 (2023).
157
S. Mun, J. Lee, S. Shin, S. R. Jeon, S. Choi, H. Kwak, K. Kim, J. Kim, and C.
Kang, ECS J. Solid State Sci. Technol. 13, 26002 (2024).
158
P. Li, H. Li, M. S. Wong, P. Chan, Y. Yang, H. Zhang, M. Iza, J. S. Speck, S.
Nakamura, and S. P. Denbaars, Crystals 12, 541 (2022).
159
X. Zhao, K. Sun, S. Cui, B. Tang, H. Hu, and S. Zhou, Adv. Photonics Res. 4,
2300061 (2023).
160
D. Iida, Z. Zhuang, P. Kirilenko, M. Velazquez-Rizo, M. A. Najmi, and K.
Ohkawa, Appl. Phys. Lett. 116, 162101 (2020).
161
A. Dussaigne, F. Barbier, B. Damilano, S. Chenot, A. Grenier, A. M. Papon, B.
Samuel, B. Ben Bakir, D. Vaufrey, J. C. Pillet, A. Gasse, O. Ledoux, M.
Rozhavskaya, and D. Sotta, J. Appl. Phys. 128, 135704 (2020).
162
F. Wu, J. Ewing, C. Lynsky, M. Iza, S. Nakamura, S. P. Denbaars, and J. S.
Speck, J. Appl. Phys. 133, 035703 (2023).
163
J. Ewing, C. Lynsky, M. Wong, F. Wu, Y. C. Chow, P. Shapturenka, M. Iza, S.
Nakamura, S. Denbaars, and J. Speck, Opt. Express 31, 41351 (2023).
164
S. S. Pasayat, R. Ley, C. Gupta, M. S. Wong, C. Lynsky, Y. Wang, M. J.
Gordon, S. Nakamura, S. P. Denbaars, S. Keller, and U. K. Mishra, Appl. Phys.
Lett. 117, 061105 (2020).
165
S. S. Pasayat, C. Gupta, M. S. Wong, R. Ley, M. J. Gordon, S. P. Denbaars, S.
Nakamura, S. Keller, and U. K. Mishra, Appl. Phys. Express 14, 011004
(2021).
166
Z. Chen, B. Sheng, F. Liu, S. Liu, D. Li, Z. Yuan, T. Wang, X. Rong, J. Huang, J.
Qiu, W. Liang, C. Zhao, L. Yan, J. Hu, S. Guo, W. Ge, B. Shen, and X. Wang,
Adv. Funct. Mater. 33, 2300042 (2023).
167
K. XIng, H. Junwei, Z. Pan, Z. Xia, Z. Jin, L. Wang, X. Jiang, H. Wang, H.
Zeng, and X. Wang, Opt. Express 32, 11377 (2024).
168
D. Iida, P. Kirilenko, M. Velazquez-Rizo, Z. Zhuang, M. A. Najmi, and K.
Ohkawa, AIP Adv. 12, 065125 (2022).
169
P. Li, H. Li, Y. Yao, N. Lim, M. Wong, M. Iza, M. J. Gordon, J. S. Speck, S.
Nakamura, and S. P. Denbaars, ACS Photonics 10, 1899 (2023).
170
S. Vichi, Y. Robin, S. Sanguinetti, M. Pristovsek, and H. Amano, Phys. Rev.
Applied 14, 024018 (2020).
171
D.-G. Lee, Y. Choi, S. Jung, Y. Kim, S. Y. Park, P. Choi, and S. Yoon, Appl.
Phys. Lett. 124, 121109 (2024).
172
B. Damilano and B. Gil, J. Phys. D. Appl. Phys. 48, 403001 (2015).
173
Z. Bi, F. Lenrick, J. Colvin, A. Gustafsson, O. Hultin, A. Nowzari, T. Lu, R.
Wallenberg, R. Timm, A. Mikkelsen, B. J. Ohlsson, K. Storm, B. Monemar,
and L. Samuelson, Nano Lett. 19, 2832 (2019).
174
A. Gustafsson, A. Persson, P. Persson, V. Darakchieva, Z. Bi, and L.
Samuelson, Nanotechnology 35, 255703 (2024).
175
P. Feng, C. Xu, J. Bai, C. Zhu, I. Farrer, G. M. De Arriba, and T. Wang, ACS
Appl. Electron. Mater. 4, 2787 (2022).
176
W. Cai, Y. Furusawa, J. Wang, J. H. Park, Y. Liao, H. J. Cheong, S. Nitta, Y.
Honda, M. Pristovsek, and H. Amano, Appl. Phys. Lett. 121, 211105 (2022).
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-28
V
CAuthor(s) 2024
16 May 2024 21:33:15
177
W. Cai, J. Wang, J. H. Park, Y. Furusawa, H. Cheong, S. Nitta, Y. Honda, M.
Pristovsek, and H. Amano, Jpn. J. Appl. Phys. 62, 020902 (2023).
178
L. Yu, L. Wang, P. Yang, Z. Hao, J. Yu, Y. Luo, C. Sun, B. Xiong, Y. Han, J.
Wang, H. Li, and L. Wang, Opt. Mater. Express 12, 3225 (2022).
179
B. Mitchell, V. Dierolf, T. Gregorkiewicz, and Y. Fujiwara, J. Appl. Phys. 123,
160901 (2018).
180
Z. Wang, X. Ye, W. Tu, and C. Yang, IEEE Trans. Electron Devices 71, 2253 (2024).
181
D. Iida, Z. Zhuang, P. Kirilenko, M. Velazquez-Rizo, and K. Ohkawa, Appl.
Phys. Express 13, 031001 (2020).
182
P. Chan, V. Rienzi, N. Lim, H. M. Chang, M. Gordon, S. P. DenBaars, and S.
Nakamura, Appl. Phys. Express 14, 101002 (2021).
183
Z. Zhuang, D. Iida, M. Velazquez-Rizo, and K. Ohkawa, Photonics Res. 9,
1796 (2021).
184
Z. Wang, S. Zhu, X. Shan, Z. Yuan, Z. Qian, X. Lu, Y. Fu, K. Tu, H. Guan, X.
Cui, and P. Tian, Opt. Express 30, 36403 (2022).
185
Y.-M. Huang, C.-Y. Peng, W.-C. Miao, H. Chiang, T.-Y. Lee, Y.-H. Chang, K.
J. Singh, Z. D. Iida, R.-H. Horng, C.-W. Chow, C.-C. Lin, K. Ohkawa, S.-C.
Chen, and H.-C. Kuo, Photonics Res. 10, 1978 (2022).
186
X. Lu, Y. Li, Z. Jin, S. Zhu, Z. Wang, Z. Qian, Y. Fu, K. Tu, H. Guan, X. Cui,
and P. Tian, J. Light. Technol. 41, 5394 (2023).
187
X. Pan, J. Song, H. Hong, M. Luo, and R. N€
otzel, Opt. Express 31, 15772
(2023).
188
L. Yu, H. Zhibiao, Y. Luo, C. Sun, B. Xiong, Y. Han, J. Wang, H. Li, L. Gan, Y.
Jiang, H. Chen, and L. Wang, Appl. Phys. Lett. 123, 232106 (2023).
189
X. Zheng, X. Xu, C. Tong, Y. Fu, M. Zhou, T. Huang, Y. Lu, Z. Chen, and W.
Guo, Appl. Phys. Lett. 124, 051103 (2024).
190
P. Li, A. David, H. Li, H. Zhang, C. Lynsky, Y. Yang, M. Iza, J. S. Speck, S.
Nakamura, and S. P. Denbaars, Appl. Phys. Lett. 119, 231101 (2021).
191
Z. Zhuang, D. Iida, M. Velazquez-Rizo, and K. Ohkawa, IEEE Electron Device
Lett. 42, 1029 (2021).
192
Z. Zhuang, D. Iida, and K. Ohkawa, Opt. Lett. 46, 1912 (2021).
193
A. Dussaigne, P. Le Maitre, H. Haas, J. C. Pillet, F. Barbier, A. Grenier, N.
Michit, A. Jannaud, R. Templier, D. Vaufrey, F. Rol, O. Ledoux, and D. Sotta,
Appl. Phys. Express 14, 092011 (2021).
194
P. Li, H. Li, Y. Yang, H. Zhang, P. Shapturenka, M. Wong, C. Lynsky, M. Iza,
M. J. Gordon, J. S. Speck, S. Nakamura, and S. P. Denbaars, Appl. Phys. Lett.
120, 041102 (2022).
195
P.Li,H.Li,H.Zhang,Y.Yang,M.S.Wong,C.Lynsky,M.Iza,M.J.Gordon,J.S.
Speck,S.Nakamura,andS.P.Denbaars,Appl. Phys. Lett. 120, 121102 (2022).
196
F. H. Hsiao, T. Y. Lee, W. C. Miao, Y. H. Pai, D. Iida, C. L. Lin, F. C. Chen, C.
W. Chow, C. C. Lin, R. H. Horng, J. H. He, K. Ohkawa, Y. H. Hong, C. Y.
Chang, and H. C. Kuo, Discov. Nano 18, 95 (2023).
197
K. Xing, Z. Xia, G. Xie, Z. Pan, Z. Zhuang, and J. Hu, IEEE Photonics Technol.
Lett. 35, 1439 (2023).
198
Y. Sang, Z. Zhuang, K. Xing, Z. Jiang, C. Li, and F. Xu, IEEE Electron Device
Lett. 45, 76 (2024).
199
R. Armitage, Z. Ren, M. Holmes, and J. Flemish, Phys. Status Solidi - Rapid
Res. Lett. 18, 2400012 (2024).
200
Z. Li, J. Waldron, T. Detchprohm, C. Wetzel, R. F. Karlicek, Jr., and T. P.
Chow, Appl. Phys. Lett. 102, 192107 (2013).
201
C. M. Kang, D.-J. Kong, J.-P. Shim, S. Kim, S.-B. Choi, J.-Y. Lee, J.-H. Min, D.-
J. Seo, S.-Y. Choi, and D.-S. Lee, Opt. Express 25, 2489 (2017).
202
D. Peng, K. Zhang, V. S.-D. Chao, W. Mo, K. M. Lau, and Z. Liu, J. Disp.
Technol. 12, 742 (2016).
203
J. G. Um, D. Y. Jeong, Y. Jung, J. K. Moon, Y. H. Jung, S. Kim, S. H. Kim, J. S.
Lee, and J. Jang, Adv. Electron. Mater. 5, 1800617 (2019).
204
L. Zhang, F. Ou, W. C. Chong, Y. Chen, and Q. Li, J. Soc. Inf. Disp. 26, 137
(2018).
205
K. Tsuchiyama, K. Yamane, S. Utsunomiya, H. Sekiguchi, H. Okada, and A.
Wakahara, Appl. Phys. Express 9, 104101 (2016).
206
H.-A. Ahn, S.-K. Hong, and O.-K. Kwon, IEEE Trans. Circuits Syst. II 65, 724
(2018).
207
R. S. Cok, M. Meitl, R. Rotzoll, G. Melnik, A. Fecioru, A. J. Trindade, B.
Raymond, S. Bonafede, D. Gomez, T. Moore, C. Prevatte, E. Radauscher, S.
Goodwin, P. Hines, and C. A. Bower, J. Soc. Inf. Disp. 25, 589 (2017).
208
J. Fan, C.-Y. Lee, S-j Chen, L. M. Gang, Z. L. Jun, S. Yang, L. M. Cai, X. H. Fei,
L. Nian, and S. I. D. Intern, Symp. Dig. Tech. Pap. 50, 326 (2019).
209
A. Bibl, J. A. Higginson, H.-H. Hu, and H.-F. S. Law, US patent 9,773,750
(2017).
210
K. Ding, V. Avrutin, N. Izyumskaya,
€
U.
€
Ozg€ur, and H. Morkoc¸, Appl. Sci. 9,
1206 (2019).
211
J. Yoon, S. -M. Lee, D. Kang, M. A. Meitl, C. A. Bower, and J. A. Rogers, Adv.
Opt. Mater. 3, 1313 (2015).
212
E. E. Kuran and M. Tichem, IEEE Trans. Autom. Sci. Eng. 10, 536 (2015).
213
W. Chang, J. Kim, M. Kim, M. W. Lee, C. H. Lim, G. Kim, S. Hwang, J. Chang,
Y. H. Min, K. Jeon, S. Kim, Y.-H. Choi, and J. S. Lee, Nature 617, 287 (2023).
214
J. K. Tu, J. J. Talghader, M. A. Hadley, and J. S. Smith, Electron. Lett 31, 1448
(1995).
215
S.-C. Park, J. Fang, S. Biswas, M. Mozafari, T. Stauden, and H. O. Jacobs, Adv.
Mater. 26, 5942 (2014).
216
S. Biswas, M. Mozafari, T. Stauden, and H. O. Jacobs, Micromachines 7,54
(2016).
217
D. Lee, S. Cho, C. Park, K. R. Park, J. Lee, J. Nam, K. Ahn, C. Park, K. Jeon, H.
Yuh, W. Choi, C. H. Lim, T. Kwon, Y. H. Min, M. Joo, Y. H. Choi, J. S. Lee, C.
Kim, and S. Kown, Nature 619, 755 (2023).
218
V. R. Marinov, SID Int. Symp. Dig. Tech. Papers 49, 692 (2018).
219
S. H. Ahn and L. J. Guo, ACS Nano 3, 2304 (2009).
220
M. Choi, B. Jang, W. Lee, S. Lee, T. W. Kim, H.-J. Lee, J.-H. Kim, and J.-H.
Ahn, Adv. Funct. Mater. 27, 1606005 (2017).
221
J. Hwang, H.-J. Kim-Lee, S. W. Hong, J.-Y. Park, D. K. Kim, D. Kim, S. Song, J.
Jeong, Y. Kim, M. J. Yeom, M-c Yu, J. Kim, Y. Park, D.-C. Shin, S. Kang, J.-K.
Shin, Y. Kim, E. Yoon, H. Lee, G. Yoo, J. Jeong, and K. Hwang, Nat. Electron.
6, 216 (2023).
222
L. Qi, P. Li, X. Zhang, K. M. Wong, and K. M. Lau, Light Sci. Appl. 12,258
(2023).
223
G. Zhu, Y. Liu, R. Ming, F. Shi, and M. Cheng, Sci. China Mater. 65, 2128
(2022).
224
S. Zhang, H. Zheng, L. Zhou, H. Li, Y. Chen, C. Wei, T. Wu, W. Lv, G. Zhang,
S. Zhang, Z. Gong, B. Jia, H. Lin, Z. Gao, W. Xu, and H. Ning, Crystals 13,
1001 (2023).
225
J. Choi, K. Cho, and S.-H. Cho, Appl. Phys. A 130, 140 (2024).
226
K. R. Bagnall, E. A. Moore, S. C. Badescu, L. Zhang, and E. N. Wang, Rev. Sci.
Instrum. 88, 113111 (2017).
227
Raman enhancement detection method and Raman enhancement detection
device for micro-LED chip. China patent 111610177.A (2020).
228
K. Kim, G. Jung, J. Kim, Y. Sung, J. Kang, W.-J. Lee, Y. Moon, T. Jeong, and
J.-H. Song, Opt. Mater 120, 111448 (2021).
229
D. Lewis, R. Dechter, R. Ley, M. Wong, M. Khoury, and S. I. D. Symp, “Digest
Techn,”Papers 52, 328 (2021).
230
B. M. Ahn and S. H. Park, US patent 11069584.B (2021).
231
Z. W. Liang and W. Q. Luo et al., US patent 2021098646.A (2020).
232
D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang,
M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and A. G. Fujimoto, Science
254, 1178 (1991).
233
W. Henry and C. Percival, “55-2: Invited paper: ILED displays: Next genera-
tion display technology,”SID Symp. Dig. Tech. Papers 47, 747 (2016).
234
C. Prevatte, E. Radauscher, M. A. Meitl, D. Gomez, K. Ghosal, S. Bonafede, B.
Raymond, T. Moore, A. J. Trindade, P. Hines, and C. A. Bower, “Miniature
heterogeneous fan-out packages for high-performance, large-format systems,”
in Proceedings of the 2017 IEEE 67th Electronic Components and Technology
Conference (ECTC) (IEEE, 2017), pp. 1098–1106.
235
J. B. Park, K. H. Lee, S. H. Han, T. H. Chung, M. K. Kwak, H. Rho, T. Jeong,
and J.-S. Ha, Sci. Rep. 9, 11551 (2019).
236
B. Moran, M. J. Holmes, J. Flemish, R. Armitage, Z. Ren, H. Lotfi, T. Chung,
H. J. Kim, H. Masui, R. Pathak, J. C. Tan, B. H. Tan, Y. M. Tio, S. Lim, K. B.
Lim, and O. Shchekin, Dig. Tech. Papers. - SID Int. Symp. 54, 414 (2023).
237
H. Jung, C. Il Park, M. B. Gee, J. Choi, Y. R. Jeong, S. J. Min, J. H. Song, M. S.
Lim, M. Kim, T. Kim, S. Ham, H. Lee, H. Kim, I. T. Jeong, G. H. Kim, J. Y.
Yang, and S. Yoon, J. Soc. Inf. Disp. 31, 201 (2023).
Applied Physics Reviews REVIEW pubs.aip.org/aip/are
Appl. Phys. Rev. 11, 021319 (2024); doi: 10.1063/5.0177550 11, 021319-29
V
CAuthor(s) 2024
16 May 2024 21:33:15
