Display technology has developed rapidly in recent years, with III–V system-based micro-light-emitting diodes (μLEDs) 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 μLED displays to be successfully commercialized, their technical shortcomings need to be addressed. This review comprehensively discusses important issues associated with μLEDs, 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 μLEDs, advanced transfer techniques for production, and the detection and repair of defects. Finally, industrial efforts to commercialize μLED displays are summarized. This review thus provides important insights into the potential realization of next-generation display systems based on μLEDs.

Recent advanced display technologies such as liquid crystal displays (LCDs) and organic light-emitting diodes (OLEDs) have revolutionized 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 information 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 attention as the light source for micro-LED (μLED) 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–V μLEDs can be fabricated in various sizes down to 1 μm2 [∼20 000 pixels per inch (PPI)] using a typical top-down strategy 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 μLEDs 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 efficient μLED displays.

In principle, alloying GaN and InN in InGaN μLEDs can produce an emission wavelength that covers the entire visible range, including the RGB colors, thus enabling the fabrication of full-color μLED displays. Since the emergence of GaN-based blue LEDs 30 years ago, many researchers have sought to optimize the external quantum efficiency (EQE) of blue LEDs.13–15 InGaN alloys with a high indium content can reduce the bandgap energy, leading to the emission of longer wavelengths (i.e., green and red). However, when controlling the number 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 μLEDs. However, these suffer from a more dramatic loss of efficiency with smaller chip sizes than that faced by InGaN-based red μLEDs 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 μLEDs are best suited for use in μLED displays.

To produce high-performance display systems, the light source for the μLEDs (referred to as the frontplane) should operate with a backplane such as thin film transistors (TFTs),19,20 high electron mobility transistors,21,22 or complementary metal-oxide semiconductors (CMOS)23,24 that contains a circuit with appropriate capacitance for charging. A key factor for realizing a μLED-based display system is to develop a precise transfer technique with a high transfer rate, reliability, 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 methods with the aim of commercializing μLED displays.

In this review, we comprehensively discuss the major issues associated with the fabrication of μLEDs and their use in display systems. First, the traditional ABC model for LEDs, which is still used to interpret μLED properties, is introduced. We then review in detail the mechanisms involved in the loss of efficiency suffered by μLEDs 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 μLED 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 μLED displays. We also describe methods for detecting and repairing defective pixels to advance the commercialization of μLED displays. Finally, we present industry efforts on μLED displays to provide insight into what is needed to commercialize μLED displays.

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 μLEDs. The maximum efficiency of an LED is determined by its internal quantum efficiency (IQE), which refers to the number of photons produced per recombination in the active region as a function of the carrier density. 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 proportional to the probability of two carriers overlapping (∼n2), and (3) Auger recombination, which is proportional to the three-carrier process (∼n3). The IQE can thus be calculated using
IQE = Photons Recombinations = B n 2 A n + B n 2 + C n 3 ,
(1)
with n the carrier density, A (s−1) is the non-radiative recombination rate, B (cm3⸱s−1) is the radiative recombination rate, and C (cm6⸱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 B and both A and C .27 The use of ABC constants that depend on the carrier density complicates 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 = Photons a t detector carriers injected = L I × e 0 h v = η e η inj B n 2 A n + B n 2 + C n 3 = η e η inj I Q E ,
(2)
where L is the optical light output, e0 is the elementary charge, h is Planck's constant, v is the frequency, I is the current, η e is the LEE, and η inj is the carrier injection efficiency. In this formulation, the most limiting factor is η e. 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 particular, 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 μLEDs, shrinking the size increases the LEE28–30 because light is not only extracted from the top surface but also from the sidewalls in increasingly larger proportions [Fig. 1(a)]. When η e is reduced by 20%, the EQE decreases by twice the non-radiative recombination rate ( A). Accordingly, LED companies have sought to improve the LEE, typically by roughening the surface, using thin-film flip chips, shaping the packages, and/or other often closely guarded methods.8 
FIG. 1.

(a) Calculated EQE for a typical set of a good blue LED with increasing non-radiative recombination rate A using typical values B 10 12 cm3⸱s−1, and C 10 31 cm6⸱s−1. (b) Numerical calculated dependence of J peak as a function of the non-radiative recombination rate using a fixed set of B and C using equation ( J peak). Fitting the slope gives J peak A 1.3 for a typical set of ABC coefficients. (c) Calculated impact sidewall recombination, i.e., l λ W for square LEDs and assuming λ = 0.1, 1, and 10 μm.

FIG. 1.

(a) Calculated EQE for a typical set of a good blue LED with increasing non-radiative recombination rate A using typical values B 10 12 cm3⸱s−1, and C 10 31 cm6⸱s−1. (b) Numerical calculated dependence of J peak as a function of the non-radiative recombination rate using a fixed set of B and C using equation ( J peak). Fitting the slope gives J peak A 1.3 for a typical set of ABC coefficients. (c) Calculated impact sidewall recombination, i.e., l λ W for square LEDs and assuming λ = 0.1, 1, and 10 μm.

Close modal
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 η e is 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 n2, so the EQE can be plotted against L = η e η inj B n 2 and fit with
EQE ( L ) = η e η inj P + 2 P + L L peak + L peak L ,
(3)
which yields η e η inj, Lpeak, 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
n peak = A C = L peak η e B .
(4)
The parameter P gives then the peak IQE
IQE peak = P P + 2 = B B + 2 A C .
(5)
This was first introduced by Dai et al.31 In principle, the IQE for different 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 P = B A C, only the relationship between A, B, and C can be calculated using P. However, absolute values cannot be obtained from EQE measurements without another independent measurement of the carrier lifetime or the use of modulated currents. However, additional indications can be obtained from other sources. For example, because B and C are mostly given by geometry and the materials used, A is a particular focus, especially because it may also vary depending on the processing and size of the μLEDs. Thus, the current density J peak at peak EQE can be determined. Calculating J peak from Eq. (2) at n peak gives the following:
J peak = η inj e 0 d A B C + 2 A C ,
(6)
where d is the total thickness of the active area. In extreme cases, if A is very large, then the second term in the parentheses ( 2 A / C) is larger and J peak η inj e 0 d 2 A 3 / C. Conversely, if A is very small, then the first term will be dominant, giving J peak η inj e 0 d A B / C. In between, changes in the non-radiative recombination rate A will cause the current density to change slightly more than linearly at the peak EQE,
J peak A q with q = 1 1.5 .
(7)
As shown in Fig. 1(b), for a typical set of B and C, the exponent q in 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 B = 10−11 cm3⸱s−1 and C = 10−29 cm6⸱s−1, yields similar exponents. The reduced size means that a greater number of carriers are located near the sidewalls. At the surface, they undergo non-radiative recombination through surface states (i.e., sidewall recombination). 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 μLEDs, allowing carriers (and especially holes) to efficiently diffuse laterally to the sidewalls and recombine non-radiatively with surface recombination rate As.33 The sidewalls therefore make an additional contribution to the non-radiative recombination rate. Assuming that all carriers within a certain distance λ from the edge are swept to the sidewall via band bending and then recombine at the sidewall with non-radiative surface recombination rate As, the total non-radiative recombination rate A used in Eq. (1) for the IQE is given by the sum of the bulk non-radiative recombination rate Ao and the sidewall surface term as follows:
A = A 0 + A s λ l W ,
(8)
where l is the peripheral length of the LED and W is the area of the LED. Figure 1(c) describes the impact of sidewall surface recombination as a function of device width variation λ. It is understood that the effect of sidewall recombination becomes stronger as the LED width decreases. Additionally, because A is proportional to λ according to Eq. (8), increasing λ consequently increases the sidewall recombination. Thus, if the size of the μLEDs is sufficiently small, the sidewall recombination rate A s λ l W dominates all non-radiative recombination, unless As or λ is significantly reduced by smoothing or passivation, as discussed later.
When comparing differently sized μLEDs processed on the same wafer, it can be assumed that B and C remain the same. The QW can relax laterally only under 2 μm, releasing strain and changing B directly. Thus, above a width of 2 μm, J peak should behave as follows:
J peak A 0 + A s λ l W 1.3 ,
(9)
which allows the effect of sidewall recombination to be assessed. The exact contribution of surface recombination and carrier catchment distance λ is difficult to quantify. However, if the lateral movement of the carriers can be reduced (e.g., by increasing background doping) then λ can be reduced (Fig. 11). Chemical passivation would change both A s and λ because the surface states can shift their energy position or experience a decrease in density, both of which will affect surface band bending and thus λ. Therefore, A s λ is also called the surface recombination velocity and is discussed together. In conclusion, the optimization 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.

When determining the performance of μLEDs, it is important to monitor the defects. Across the entire μLED fabrication process, from epitaxial growth to processing, two types of defects are considered particularly deleterious for the performance of μLEDs: 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 processing for several reasons. For example, the lattice mismatch between layers can create dislocations. It is well known that the dislocation density for GaN (∼108 cm−2) arises from the difference between the lattice constant of GaN (3.19 Å) and that of sapphire (2.75 Å) and/or Si (111) (3.84 Å). 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 compared the performance of devices with dislocation densities of 5.3 × 108, 1.2 × 109, and 5.7 × 109 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 cathodoluminescence (CL) near a threading dislocation at room temperature. Figure 2(b) presents intensity profiles of the CL across the threading dislocation. 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 degradation. 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 μLEDs.

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 Lähnemann 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 

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 Lähnemann 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 

Close modal

Another cause of bulk defects is that an InGaN layer with a high indium content promotes the emission of longer wavelengths because it has a narrower bandgap than GaN and AlN alloys. The growth conditions for InGaN LEDs with a high-In InGaN layer can produce longer 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 wavelengths over 600 nm, 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 approximately 470 nm and increasing the total number of defects.

Trench defects can also be generated when the growth temperature 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 quantum barriers (QBs) at temperatures of 730, 830, and 880 °C with the 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 additional defects such as threading dislocations and trench defects is required to achieve the fabrication of high-performance μLEDs.

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 μm2) prevented damage during the ICP-RIE process. However, in recent years, LED technology has shifted to smaller LEDs (i.e., μLEDs), leading to the size-dependent performance of LEDs related to sidewall defects.39 The chip size contributing to the display resolution should be small because the human eye can distinguish display 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 controlled to below 10 μm, which is desirable for use in μLED displays. However, ICP-RIE, which is conducted with plasma, can unintentionally induce surface damage at the sidewalls of μLEDs and thus degrade their efficiency.42 Therefore, understanding the mechanisms responsible for sidewall recombination is a key factor for the production of highly efficient μLEDs, particularly in terms of reducing the power consumption.

In line with this, Park et al.43 recently observed the width of sidewall damage and the atomic structure near the sidewall. Figure 3(a) presents an HAADF-STEM image of μLEDs, 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 distortion. 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 particular, 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 centers, and these centers have a stronger effect for carriers with a longer diffusion length, which determines how many carriers reach the sidewall.45 

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 dislocation 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 observing carrier diffusion at nominal current densities of 25 (left) and 9500 (right) A/cm2 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 number of defect densities and a condition of sidewall surface.

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 dislocation 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 observing carrier diffusion at nominal current densities of 25 (left) and 9500 (right) A/cm2 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 number of defect densities and a condition of sidewall surface.

Close modal

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 function 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 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 μLEDs and AlGaInP-based red μLEDs.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 recombination centers and longer diffusion lengths for carriers allow a large number of carriers to recombine non-radiatively at the sidewall surface, 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 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 μLEDs.51 Instead, carriers predominantly 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 efficiency because sidewall surface recombination can be avoided.52 However, as their size decreases, the degradation in the efficiency becomes a serious concern49 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 passivation layers produced with atomic layer deposition (ALD).18,53 Though these methods can improve the performance of μLEDs, efficiency degradation at smaller sizes still occurs. For example, Park et al.51 investigated size-dependent InGaN-based blue μLEDs 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 recombination, both sidewall treatment and the reduction of the diffusion length of carriers should be pursued.

The performance of μLEDs 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 degradation of μLEDs. Typically, size-dependent degradation occurs during the ICP-RIE of InGaN μLEDs with a low indium content (e.g., blue μLEDs). Smith et al.62 experimentally observed the degradation of the EQE of InGaN-based μLEDs with chip sizes ranging from 30 μm to 1 μm. Figure 4(a) shows the degradation of blue μLEDs with a reduction in their size. Similar to Eq. (9), the J peak of each size dramatically increased but EQE peak decreased, which suggested that carrier loss at the sidewall can strongly affect the performance of μLEDs. However, the size-dependent behavior of J peak and EQE peak for green μLEDs exhibited almost no change, suggesting that the carrier loss at the sidewall was not serious, unlike that for the InGaN-based blue μLEDs [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 surface recombination velocity of 5 × 103 and 3 × 102 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 μLEDs with more non-radiative carrier recombination because many of the carriers can reach the sidewall.33,64,65

FIG. 4.

Changes of EQEs with various chip sizes as a function of injection current density (a) InGaN-based blue μLEDs and (b) InGaN-based green μLEDs. 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 μLEDs. 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 μLEDs. Reprinted with permission from Oh et al., Opt. Express 26, 11194 (2018). Copyright 2018 The Optical Society.68 

FIG. 4.

Changes of EQEs with various chip sizes as a function of injection current density (a) InGaN-based blue μLEDs and (b) InGaN-based green μLEDs. 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 μLEDs. 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 μLEDs. Reprinted with permission from Oh et al., Opt. Express 26, 11194 (2018). Copyright 2018 The Optical Society.68 

Close modal

In terms of red emissions, both InGaN-based and AlGaInP-based red μLEDs 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 μm; 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 μLEDs are compatible with AlGaInP-based μLEDs. Li et al.66 compared the EQEs of InGaN and AlGaInP-based red μLEDs that varied in size from 20 μm to 100 μm 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 density (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 μLEDs 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 μLEDs increased with decreasing chip size, which was in contrast to the typical EQE of InGaN-based blue μLEDs [Fig. 4(e)]. This behavior could be attributed to suppressed surface recombination and increased LEE via the sidewall. Similarly, Lim et al.67 investigated InGaN-based red μLEDs using thermal decomposition to relax the strain and found that the EQE peak of red μLEDs increased as their size decreased from 100 × 100 to 5 × 5 μm2. These results indicate that InGaN-based red μLEDs have a strong advantage in terms of overcoming the size-dependent degradation of μLEDs.

On the other hand, much research groups have demonstrated that the degradation of AlGaInP-based red μLEDs when their size is reduced is dramatic. For example, Oh et al.68 investigated the degradation of AlGaInP-based red μLEDs in terms of the EQE, ideality factor, and leakage current. They found that EQE peak was lower by a factor of two and that J peak was considerably higher when the size decreased from 350 × 350 to 15 × 15 μm2 [Fig. 4(f)]. These results suggest that the diffusion length of the carriers should be taken into account when seeking to fabricate highly efficient μLEDs with smaller sizes. Unlike InGaN μLED systems, although many attempts have been made to overcome this efficiency degradation of AlGaInP-based red μLEDs, 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 μLEDs for use as a red source in future display systems needs to be developed.

As mentioned above, a major challenge for small μLEDs is overcoming carrier loss at the sidewall, with a higher surface recombination 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 damage would clarify the performance degradation of μLEDs, few studies have investigated this to date. In this section, we introduce a quantitative interpretation of sidewall damage using various methods 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 μLEDs using time-resolved PL (TRPL) and found that the carrier lifetime decreased closer to the sidewall. They also investigated luminescence intensity profiles for a size of 5.61 × 5.61 μm2 using CL under different temperatures. The ratio of CL intensity of 296 K to 30 K demonstrated that a 5.61 μm mesa with a width of approximately 3.4 μm had an efficiency lower than 80% of the maximum value at the center [Fig. 5(a)]. For III-nitride blue μLEDs, Finot et al.70 systemically estimated the change in the lifetime of carriers under different sidewall conditions using photon-correlation CL. Similar to the results of Boussadi et al.,69 they found that the carrier lifetime across the mesa decreased considerably closer to the sidewall. For example, processed KOH+ALD-Al2O3 exhibited an increase in the carrier lifetime from 6 to 8 ns compared to an unprocessed sample near the sidewall, suggesting that the better the sidewall conditions, the less carrier loss at the sidewall due to surface recombination [Fig. 5(b)].

FIG. 5.

(a) CL intensity ratio of at 296 K to 30 K showing efficiency degradation near the sidewall for 5.61 × 5.61 μm2 AlGaInP-based red μLEDs. 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) KOH+ALD-Al2O3 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 μm2 μLEDs 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 μm diameter LED demonstrating the level of traps center hole traps in QWs at Ev + 0.7 eV and Ev + 0.75 eV electron traps in –he QWs at Ec − 1eV. (e) DLTS spectra of 30 (green), 50 (yellow), and 100 (red) μm 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 

FIG. 5.

(a) CL intensity ratio of at 296 K to 30 K showing efficiency degradation near the sidewall for 5.61 × 5.61 μm2 AlGaInP-based red μLEDs. 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) KOH+ALD-Al2O3 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 μm2 μLEDs 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 μm diameter LED demonstrating the level of traps center hole traps in QWs at Ev + 0.7 eV and Ev + 0.75 eV electron traps in –he QWs at Ec − 1eV. (e) DLTS spectra of 30 (green), 50 (yellow), and 100 (red) μm 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 

Close modal

Yu et al.71 quantitatively investigated the difference in the lifetime of blue and green InGaN for sizes of 10 × 10 and 20 × 20 μm2 fabricated using a typical ICP-RIE process [Fig. 5(c)]. It was found that the lifetime of blue InGaN μLEDs near the sidewall at sizes of 10 × 10 and 20 × 20 μm2 was 3.1 and 3.0 ns, respectively, compared to 9.6 and 9.8 ns for the green InGaN μLEDs. 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 performance degradation of a device due to non-radiative recombination to be understood, including the analysis of sidewall damage in μLEDs. Recently, DLTS has been used to confirm the change in the energy level of hole and electron traps in III-nitride μLEDs. Lee et al.72 reported that deep hole traps and electron traps can form at Ev + 0.75 eV and Ec − 1.0 eV, respectively, at the sidewall of InGaN-based blue μLEDs. As shown in Fig. 5(d), the DLTS signal as a function of temperature under a reverse bias of −0.5 V with a forward bias pulse of +3 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–V curves of p–n junction devices, indirectly indicates the recombination mechanisms,
n ideality q K T ( InI V ) 1 .
(10)
According to the theory outlined by Shockley et al.,73 an ideality factor of 1.0 is associated with band-to-band radiative recombination, an ideality factor of 2.0 represents SRH recombination, and an ideality 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 μLEDs and found differences in the ideality factor for different sidewall conditions [Fig. 5(f)]. Optimized sidewall treatment 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 SiO2 compared with conventional PECVD SiO2 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 μLEDs.

The optical and electrical properties of optoelectronic devices are negatively affected by native defects, crystallographic defects, and contaminants present on the surface of the semiconductor layer.53 Thus, KOH, buffer oxide etching, and HNO3:HCl have been employed to eliminate surface oxides and contaminants. Aqueous solutions and dielectric layers, including TMAH, (NH4)2Sx, CH3CSNH2, SiO2, Al2O3, and Ta2O5, have also been adopted to passivate semiconductor surfaces.53 For μLEDs, a combination of solution treatment and dielectric 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 μLEDs by removing non-radiative recombination centers.53 

In particular, plasma-enhanced chemical vapor deposition (PECVD) SiO2 has been widely used for this purpose.76–78 For example, in blue μLEDs, it was found that the recombination lifetime of carriers increased when samples were passivated with a PECVD SiO2 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 μLEDs.77 ICP-RIE has been found to create plasma-induced lattice disorders (about 2 nm wide) on the surface of the mesa sidewall, which can subsequently be removed using TMAH treatment.78,79 TMAH-treated blue μLEDs (chip size: 2 μm) exhibited a considerable increase in their luminous intensity when passivated with a SiO2 layer.

ALD is widely used to deposit highly compact and dense dielectric layers, meaning that ALD is also suitable for the deposition of passivation layers for μLEDs.74,80–82 For AlGaInP-based red μLEDs, passivation with ALD SiO2 resulted in a 22.8% longer lifetime than that with PECVD SiO2 under the same aging conditions (e.g., continuous operation at 400 A/cm2).80 Furthermore, ALD SiO2 passivation combined with KOH treatment was found to reduce the ideality factor of blue μLEDs from 3.4 to 2.5, indicating that the combined process was effective in reducing SRH non-radiative recombination and surface recombination.74 Additionally, Al2O3 deposited using ALD is considered an effective dielectric material for passivation.83–92, Figure 6(a) presents the EQEs of 20 × 20 μm2 blue μLEDs fabricated using different passivation methods.83 For LED-4 (ALD passivation and HF etching), 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 μLEDs (10 × 10 μm2) with ALD-Al2O3 sidewall passivation exhibited a 66.7% higher EQE and a 46.2% lower surface recombination velocity than those without ALD-Al2O3 treatment. For AlGaInP/GaInP red μLEDs (15 × 15 μm2), a combined process of sulfur treatment and ALD-Al2O3 passivation yielded a 20% higher EQE and a 14% lower surface recombination velocity compared to untreated samples.86 For GaN-based green μLEDs, ALD-Al2O3 passivation and KOH treatment produced ideality factors lower than 1.5 for all samples (varying from 3 to 100 μm).88 Treated μLEDs (6 × 6 μm2) produced a peak EQE of 16.59% at 20 A⸱cm2 and over 600 k cd cm2 at 1 A cm2. On the other hand, compared to ALD Al2O3, ALD AlN passivation has been found to offer a stronger ability to remove sidewall defects in μLEDs as a result of the uniform passivation interface.90 Blue μLEDs (25 × 25 μm2) with AlN passivation exhibited an 18.3% higher EQE than Al2O3-passivated samples in a luminescence application.

FIG. 6.

(a) EQEs of 20 × 20 μm2 μLEDs 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 μLEDs (40 × 40 μm2) with and without ALD-Ta2O3 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 μLEDs 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.

FIG. 6.

(a) EQEs of 20 × 20 μm2 μLEDs 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 μLEDs (40 × 40 μm2) with and without ALD-Ta2O3 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 μLEDs 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.

Close modal

Dielectric materials such as Ta2O5, Ga2O3, HfO2, and a TiO2/SiO2 omnidirectional reflector (ODR) have also been shown to act as efficient passivation layers.50,75,91,92 ALD Ta2O5 (with a dielectric constant of 26) effectively served as a passivation oxide for GaN μLEDs (a peak wavelength of ∼425 nm).91 The use of Ta2O5 suppressed current diffusion to the plasma-etched mesa edges. As shown in Fig. 6(b), passivated μLEDs (40 × 40 μm2) yielded a 34.6% larger EQE than unpassivated samples and exhibited an on/off current ratio of 108. In addition, for InGaN-based blue μLEDs (15 × 15 μm2), photoelectrochemically oxidized Ga2O3 passivation resulted in a 22% higher light output and a lower reverse leakage current by over two orders of magnitude at −5 V compared to reference samples.92 It has also been confirmed that HfO2 passivation can effectively improve the EQE of 580 nm blue nano-LEDs by approximately 18% compared to conventional SiO2 passivation.50 In addition, a low-temperature sol-gel process 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 SiO2 passivation was compared with that of PECVD SiO2 passivation.75 EQEs were obtained from 60 pixels, with each pixel consisting of 6–9 nanorods (diameter: 580 nm) fabricated with different forms of SiO2 passivation. 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 process significantly improved LED performance by blocking current injection into the mesa-etch-induced sidewall defects. For H-passivated InGaN-based green μLEDs (20 × 20 μm2), the reverse leakage current was reduced more than 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 sidewall surface defects.96 The PL intensity of InGaN-based green μLEDs 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 μLED chips were about 33% higher compared to the EQE of the reference. Moreover, ion implantation with an As+ source has been used to replace the plasma-etching mesa process in the fabrication of GaN μLEDs.97 The optimal As+ implantation process at 40 keV demonstrated excellent I–V characteristics, including a forward voltage of 3.1 V at 1 mA and a leakage current of 10−9 A at −5 V for InGaN-based blue μLEDs.

The dry etching conditions for the mesa of μLEDs are also a critical factor in determining their performance. Typically, ICP-RIE, which is a plasma-based process, unintentionally damages the sidewall of μLEDs, 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 recombination at the sidewall surface. The authors compared the EQE and efficiency degradation of blue μLEDs fabricated using NBE (3.5 × 3.5 μm2) and using a conventional ICP-RIE process with KOH treatment (3 × 3 μm2). It was reported that the Jpeak of the 3.5 × 3.5 μm2 chips etched with NBE was approximately 3 A/cm2, whereas the Jpeak of the 3 × 3 μm2 chips was 9 A/cm2. According to Eq. (9), this change in Jpeak indicates that NBE reduces non-radiative recombination at the sidewall surface. Furthermore, the efficiency droop, which is defined at 0.01 A/cm2 and can quantitatively confirm non-radiative recombination at the sidewall surface, was 26% and 60% for the 3.5 × 3.5 and 3 × 3 μm2 LEDs, respectively. The lower efficiency droop at low current densities suggested that non-radiative recombination was relatively low. Additionally, the size independence of Jpeak for LEDs with a size of 3.5 × 3.5, 6.5 × 6.5, 10.5 × 10.5, and 20.5 × 20.5 μm2 confirmed that NBE led to negligible non-radiative recombination at the sidewall surface.

Various methods have been proposed to increase the LEE of μLEDs, including the shaping of the sidewalls, diffracted Bragg reflectors, 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 μLEDs.99–103 For example, it has been shown that deep ultraviolet (DUV) μLEDs (λ = 280 nm; chip size = 20 μm) with a sidewall angle of 33° yielded a 19% higher EQE than those with an inclination angle of 75°.99 This improvement was attributed to the fact that more photons 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 20 μm μLEDs 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 20 μm diameter DUV μLEDs, a sidewall angle of 25°–35° was optimal for light extraction from the bottom.99 It was also found that the LEE of red, green, and blue flip chip (FC) μLEDs (< 70 × 112 μm2) increased when raising the sidewall inclination angle from 90° to 120°, although this was less effective for the red μLEDs.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 SiO2/Ag ODR microstructure array was used to improve top emission light extraction and eliminate the color shift in red, green, and blue μLEDs.102 For truncated-pyramid-shaped blue FC μLEDs, the strong dependence of the LEE on the inclination angle has also been reported using the SimuLED package, considerably enhancing the overall emission efficiency.103 The texturing of the sidewall surface has also been utilized to increase light extraction. For example, the sidewall surface of circular blue μLEDs (chip size: 50 μm) was patterned with concave–convex circular composite structures.104,105 A circular pattern radius of 2 μm was 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.

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 μLEDs 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 μLEDs (size: 20 μm) 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 μLEDs. 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 González-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 μm-sized LEDs with different sidewall angles simulated by FDTD. Reprinted with permission from Vögl et al. Opt. Express 31, 22997 (2023). Copyright 2023 The Optical Society.122 

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 μLEDs 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 μLEDs (size: 20 μm) 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 μLEDs. 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 González-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 μm-sized LEDs with different sidewall angles simulated by FDTD. Reprinted with permission from Vögl et al. Opt. Express 31, 22997 (2023). Copyright 2023 The Optical Society.122 

Close modal

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 μLEDs 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 a higher number of AlGaN/GaN SL DBR pairs.106 Furthermore, the use of a TiO2/HfO2 (35 nm/50 nm) conductive DBRs combined with a Cr/Ni/Au p-type electrode was found to effectively increase the light output power (LOP) of near-UV μLEDs (λ = 385 nm; 100 × 100 μm2) by 5%.107 Based on FDTD results, the LOP improvement was associated with a higher LEE (6%). Figures 7(c) and 7(d) show SEM images of a 3.6 μm circular μLED (500 nm) with epitaxially embedded DBRs (11 pairs of nanoporous GaN/undoped GaN) and the PL spectra of μLEDs with and without the DBR, respectively.108 The PL intensity of the μLEDs 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 TiO2/SiO2-based ODRs109 and Ta2O5/SiO2 multilayer high reflectors110 has been found to improve the LEEs of AlGaInP FC μLEDs (100 × 100 μm2) and GaN μLEDs, respectively.

Surface roughening and texturing have been widely used to increase the light extraction of μLEDs.112–116 For example, Gong et al.112 investigated the effect of the nano-texturing of p-GaN on the performance of blue μLEDs. The textured surface consisted of hexagonally arranged nano-cone arrays (height of cones: ∼75 nm). It was found that nano-textured μLEDs with rhomboidal geometries produced 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 μLEDs (chip size: 25 × 17 μm2), surface roughening of p-GaP (etched for 24 s) resulted in a 42.3% higher LOP at 20 μA 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 μLEDs using the finite element method and reported that the LEE of trapezoidal-shaped FC μLEDs improved from ∼53.0% to 64.5% as the mesa angle increased from 0° to 12°. The use of surface grating on n-GaN was also observed to increase the LEE of the μLEDs; 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 investigate the effects of substrate and sidewall texturing on the LEE of blue FC μLEDs.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 μLEDs. 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 μLEDs. Simulation results demonstrated that the sidewall texturing of blue FC μLEDs significantly increased the total LEE by increasing the sidewall LEE. The patterned sapphire substrate technology was thus shown to increase the LEE of the encapsulated μLEDs. Ryu et al.116 examined the effects of chip shape, surface roughness, and p-GaN thickness on the LEE of GaN-based vertical μLEDs using three-dimensional FDTD simulations. It was observed that the cone-shaped n-GaN surface effectively increased the LEE of the FC μLEDs. However, surface roughness had little effect on the LEE of blue FC μLEDs smaller than 5 μm, while square μLEDs produced a higher LEE than circular μLEDs. The lower LEE was attributed to the coupling of whispering gallery modes. For the square μLEDs, 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 μLEDs. For example, suspended GaN-based blue μLEDs exhibited a 150% higher light emission compared to conventional μLEDs.117 The suspended structure effectively increased the LEE of μLEDs by providing 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 μm) on the LEE of top-emission blue vertical μLEDs on an Al backside mirror (chip size: 5 × 5 μm2). Their simulation results revealed a considerable improvement (∼150%) in the LEE after reducing the optical cavity length of the μLED to 1.5 μm. This improvement in the LEE was associated with the constructive interference of topside emitted photons and reflected photons from the Al backside mirror. Surrounding the μLEDs 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 μLEDs (chip size: 20 × 20 μm2). 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 μLEDs (40 × 40 μm2) with a double-sided polished sapphire substrate as a transparent submount increased the LEE.120 

As μLEDs are reduced in size (especially below 10 μm), the LEE part of the EQE becomes dominant, regardless of the emission wavelength. González-Izquierdo et al.121 investigated the performance of μLEDs with different chip sizes between 2.5and 10 μm and experimentally 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 μLEDs 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 μLED produced a larger LEE. Vögl et al.122 numerically investigated the changes in the LEE, focusing on sizes from 1 μm to 5 μm and sidewall angles from 0° to 60°. Figure 7(h) shows the far-field distribution of blue μLEDs depending on the sidewall angle. For 5 μm μLEDs, the emission pattern was less dependent on the sidewall angle, while the effect of the sidewall angle became stronger when smaller than 5 μm. Specifically, the far field of 1 μm μLEDs with a sidewall angle of 40° was mostly strong. This means that the optimization of the sidewall angle is critical to realizing highly efficient ultra-small-pixel μLEDs. This finding suggests that the enhancement of the LEE is a promising approach to improve the overall EQE when μLEDs are smaller than 10 μm.

Using the top-down ICP-RIE process for the fabrication of highly efficient μLEDs can be considered a mature technology due to the use of chemical treatment and the deposition of a passivation layer to suppress the surface recombination rate at the sidewall.11,18,53 However, other drawbacks of III-nitride μLEDs 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 μLEDs for next-generation displays that require various brightness levels, such as those used in AR and VR.5,6,8 To overcome these drawbacks, novel epitaxial structures for III-nitride μLEDs have been proposed as a solution. In this section, we review recent progress in novel epitaxial growth for III-nitride μLEDs.

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 To address this polarization issue for MQWs of InGaN μLEDs, the use of semipolar structures in InGaN μLEDs has recently been reported. For example, Chen et al.124 compared c-plane oriented and semipolar (20–21) InGaN μLEDs and confirmed the advantages of semipolar (20–21) InGaN μLEDs 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) μLEDs. It was shown that the c-plane oriented InGaN μLEDs had a 1.7 times larger peak EQE than semipolar (20–21) InGaN, which would be a disadvantage for semipolar InGaN μLEDs under a low current regime. Despite this, semipolar (20–21) InGaN μLEDs operate well under a high current regime, with a lower efficiency drop and smaller wavelength shift, which is promising for potential use in displays that require a higher brightness than conventional c-plane axis μLEDs.

FIG. 8.

(a) Cross sectional SEM image of (20–21) semi-polar GaN grown on a patterned sapphire substrate for semi-polar μLEDs. Reprinted with permission from Chen et al., Photonics Res. 8, 630 (2020). Copyright 2020 The Optical Society.124 (b) SEM image of μLEDs 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 + thick 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 Al0.11Ga0.89N-based tunnel junction. Reproduced from Wong et al., AIP Advances 13, 015107 (2023), with permission from AIP Publishing.134 

FIG. 8.

(a) Cross sectional SEM image of (20–21) semi-polar GaN grown on a patterned sapphire substrate for semi-polar μLEDs. Reprinted with permission from Chen et al., Photonics Res. 8, 630 (2020). Copyright 2020 The Optical Society.124 (b) SEM image of μLEDs 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 + thick 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 Al0.11Ga0.89N-based tunnel junction. Reproduced from Wong et al., AIP Advances 13, 015107 (2023), with permission from AIP Publishing.134 

Close modal

Conventional InGaN-based blue LEDs grown on a silicon substrate have a very high dislocation density of over 109 cm−2, which can reduce the IQE.34,125 Interestingly, Kamikawa et al.126 demonstrated that μLEDs could be fabricated using the epitaxial lateral overgrowth GaN-on-silicon (EGOS) technique to produce regions that have a low dislocation density of 5 × 106 cm−2. Figure 8(b) presents 23.6 and 45.0 μm μLEDs fabricated using EGOS. Because the bridge, which is approximately 3 μm in length, is the only part connecting the μLED to the EGOS substrate, a polydimethylsiloxane (PDMS) stamp generating stress could detach the μLED 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 μLEDs with sizes of 20 × 20, 40 × 40, and 100 × 100 μm2. The epitaxial layer of the μLEDs was generated using pendeo-epitaxy, which resulted in a dislocation density (3.3 × 108 cm−2) that was approximately 59% lower than that produced using conventional growth (8.0 × 108 cm−2). The Al2O3 layer weakly connected the μLED to the substrate, facilitating mechanical liftoff as an alternative to laser liftoff.

Another strategy for the fabrication of high-brightness μLEDs 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 relatively 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 μm2 InGaN-based blue μLEDs with a thin 3 nm GaN QB produced an efficiency droop at 10 000 A/cm2 of approximately 20%, which suggests that a thin QB leads to relatively high brightness under a high current regime while ameliorating the efficiency droop. It has also been found that a thinner QB can improve the total carrier recombination rate of InGaN μLEDs.130 In addition, Back et al.131 systemically investigated the effect of the thickness of the EBL of p-AlGaN on InGaN-based blue μLEDs with sizes ranging from 10 × 10 to 80 × 80 μm2. When the performance of a device fabricated between 2 × 1018 cm−3 Mg-doped 17 nm-thick Al0.2In0.02Ga0.78N and 2 × 1019 cm−3 Mg-doped 45 nm-thick Al0.12In0.02Ga0.86N (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 μLEDs with a balanced EBL was higher, while their J peak values decreased.

The use of III-nitride tunnel junction (TJ) contacts has also been shown to be beneficial for InGaN μLEDs in terms of promoting their 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 μLEDs over standard μLEDs 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 μLEDs. To address this issue, Wong et al.134 recently designed a novel TJ structure by growing an n-type Al0.11Ga0.89N/GaN layer on top of a conventional TJ structure. They experimentally demonstrated that the WPE of a 60 × 60 μm2 n-type Al0.11Ga0.89N/GaN layer was 10% higher than standard TJ μLEDs, which was computationally supported by band-diagram simulation results showing a higher tunneling rate due to a lower polarization charge at the Al0.11Ga0.89N/GaN interface [Fig. 8(e)]. In terms of multi-color emission, Saito et al.135 reported a 330 PPI monolithic R, G, and B μLED array stacked on the same wafer. To realize this structure, they added a n-In0.2Ga0.8N (25 nm)/p+GaN (10 nm) TJ between the blue and green μLEDs and between the green and red μLEDs. 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.

The degree to which non-radiative recombination centers in the bulk and at the sidewall surface affect μLEDs 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 μLEDs are characterized by a low efficiency, making them unsuitable for use in full-color μLED displays. Therefore, it is necessary to clearly understand the reason why red μLEDs exhibit poor efficiency and to develop strategies to overcome this.

In this section, we discuss the surface recombination velocity of μLEDs. A summary of the surface recombination velocity of various materials is presented in Fig. 9(a). III-nitrides have a lower surface 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
v [ m × s 1 ] = λ m × A s [ s 1 ] .
(11)
FIG. 9.

(a) Summarizing surface recombination velocity of III–V compound materials. 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 EQEpeak for InGaN-based blue, green, red, and AlGaInP-based red μLEDs.

FIG. 9.

(a) Summarizing surface recombination velocity of III–V compound materials. 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 EQEpeak for InGaN-based blue, green, red, and AlGaInP-based red μLEDs.

Close modal

Although it is unclear whether the surface recombination rate ( A s) of AlGaInP or InGaN is larger, it is known that the carrier diffusion length ( λ ) of GaP or GaAs in AlGaInP-based red μLEDs 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 μLEDs via carrier loss at the sidewall, with the degradation becoming more serious as the size of the LED reduces. Figure 9(b) presents the EQEpeak from recently published papers for InGaN-based blue, green, and red μLEDs and AlGaInP-based red μLEDs as a function of the ratio of the peripheral length of the device to the area.77,80,85,88 For InGaN μLEDs, the slope of EQEpeak increases as the wavelength decreases from red to blue. This is in good agreement with experimental results demonstrating that v varies with the indium content.63 In other words, λ, which is a component of v, determines the degradation of EQEpeak. Furthermore, as summarized in Table I, the diffusion lengths of GaP and GaAs, which are components of AlGaInP-based red μLEDs, are 10 times longer 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 difference in the slope between InGaN- and AlGaInP-based red μLEDs means that their potential device applications need to be considered based on their PPI values. In other words, InGaN-based red μLEDs are suitable for displays that require ultrahigh PPI, such as VR and AR, whereas AlGaInP-based red μLEDs are more suitable for other display applications.

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 (μm) Electron diffusion length in p-type layer (μm) Ref.
GaP  1–6  0.06–3  136–139  
GaAs  2–35  0.6–20  140–145  
GaN  0.05–0.7  0.093–0.95  46, 146–150  
Materials Hole diffusion length in n-type layer (μm) Electron diffusion length in p-type layer (μm) Ref.
GaP  1–6  0.06–3  136–139  
GaAs  2–35  0.6–20  140–145  
GaN  0.05–0.7  0.093–0.95  46, 146–150  

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 μLEDs 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 μLEDs at a low current. In particular, J peak, which is used to numerically express surface recombination, increases dramatically at smaller sizes. For example, Fan et al.151 reported that the J peak of a 160 μm AlGaInP LED was approximately 10 A/cm2. However, it dramatically increased to over 80 A/cm2 at a size of 10 μm. In addition, the EQE dramatically decreased from 20% to 2.5% at 20 A/cm2, indicating that surface recombination at the sidewall became stronger as the chip size decreased. Thus, the IQE of AlGaInP μLEDs was seriously degraded.

Several solutions have been proposed based on chemical treatment 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 sidewall defects in AlGaInP red μLEDs 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 μm reference sample exhibited characteristics that were superior to the other samples, indicating that the performance of larger μLEDs did not critically depend on the sidewall conditions. However, the EQE of 20 μm LEDs under different sidewall conditions (reference: without passivation; chemical treatment: ALD passivation using Al2O3; ALD + S: (NH4)2SO4 treatment and passivation using Al2O3) differed remarkably. Although methods to remove sidewall damage could improve the performance of AlGaInP μLEDs, their performance at a low current was still significantly lower than that of larger sized LEDs, with J peak in particular very different between 100 and 20 μm. This indicates that the minority carriers can still diffuse to the sidewall and recombine non-radiatively under a low current.

FIG. 10.

Atomic arrangement at the sidewall of AlGaInP μLEDs (a) before and (b) after HF treatment. (c) Relative EQEs for 12, 15, and 19 μm 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 μLEDs with various sidewall conditions. Reprinted with permission from Wong et al., Appl. Physic. Express 16, 066503 (2023).156 Copyright 2023 IOP Publishing.

FIG. 10.

Atomic arrangement at the sidewall of AlGaInP μLEDs (a) before and (b) after HF treatment. (c) Relative EQEs for 12, 15, and 19 μm 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 μLEDs with various sidewall conditions. Reprinted with permission from Wong et al., Appl. Physic. Express 16, 066503 (2023).156 Copyright 2023 IOP Publishing.

Close modal

Recently, Mun et al.157 demonstrated that treating AlGaInP-based red μLEDs 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 chip sizes from 100 × 100 to 10 × 10 μm2 fabricated with different passivation conditions, the samples with the dual dielectric passivation of 10 nm Al2O3 (ALD)/300 nm SiNx (PECVD) exhibited the lowest ideality factors regardless of the chip size. Furthermore, electroluminescence (EL) imaging revealed that the samples passivated with the Al2O3/SiNx layer produced the brightest images, though their Fresnel power transmittance (77.7%) was lower than that of Al2O3 (ALD)/SiO2 (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 μLEDs, 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 surface. Therefore, the intrinsic surface state and the longer diffusion length of the carriers are the reasons why AlGaInP-based red μLEDs 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 μLEDs as a red source with smaller chip sizes.68,151 One potential solution for reducing surface recombination at the sidewall is controlling the diffusion length of the carrier. The carrier diffusion length is defined as a function of the carrier concentration.136,143 It is expected that the diffusion 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 treatment and passivation and a shorter carrier diffusion length could yield highly efficient AlGaInP-based red μLEDs.

FIG. 11.

(a) Hole diffusion length as a function of electron concentration in GaP. Reprinted with permission from Smith et al., Solid State Electron. 15, 361 (1972).136 Copyright 1972 Elsevier. (b) Electron diffusion 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 

FIG. 11.

(a) Hole diffusion length as a function of electron concentration in GaP. Reprinted with permission from Smith et al., Solid State Electron. 15, 361 (1972).136 Copyright 1972 Elsevier. (b) Electron diffusion 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 

Close modal

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 particular, the growth of an InGaN QW layer with a high indium content requires lower temperatures, which leads to poor crystal quality.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 fabricated 633 nm wavelength red LEDs with a size of 400 × 400 μm2 at 20 mA by varying the underlying thickness of the n-GaN layer from 2 to 8 μm. 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 incorporation 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 a lattice parameter. For example, they compared threading dislocation densities between 3.2069 and 3.2056 Å in 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 

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. 116, 162101 (2020), with permission from AIP Publishing.160 (c) Comparison analysis of threading dislocation density between 3.2069 and 3.2056 Å of InGaNOS substrates. (d) EQEs of InGaN-based red LEDs grown on InGaNOS with a lattice parameter of 3.2069 Å (LED B) and InGaNOS with a lattice parameter of 3.2056 Å (LED A). Reproduced from Dussaigne et al., J. Appl. Phys. 128, 135704 (2020), with permission from AIP Publishing.161 

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. 116, 162101 (2020), with permission from AIP Publishing.160 (c) Comparison analysis of threading dislocation density between 3.2069 and 3.2056 Å of InGaNOS substrates. (d) EQEs of InGaN-based red LEDs grown on InGaNOS with a lattice parameter of 3.2069 Å (LED B) and InGaNOS with a lattice parameter of 3.2056 Å (LED A). Reproduced from Dussaigne et al., J. Appl. Phys. 128, 135704 (2020), with permission from AIP Publishing.161 

Close modal

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 growth of an In0.3Ga0.7N decomposition layer could effectively relax the strain. This high-indium-content In0.3Ga0.7N layer was decomposed 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%, respectively. 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 × 5 μm2 μLEDs. The peak wavelength was estimated to be 660 nm at 1 A/cm2 and 597 nm at 200 A/cm2.

FIG. 13.

(a) XRD RSM of (1124) reflections revealing a strain relaxation using In0.3Ga0.7N 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 content in MQWs and peak wavelength of EL spectra at 1 A/cm2. (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 In0.08Ga0.92N stress-release layer and an AlN dislocation confinement layer. (f) EL spectra of an LED at different current densities, 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.

FIG. 13.

(a) XRD RSM of (1124) reflections revealing a strain relaxation using In0.3Ga0.7N 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 content in MQWs and peak wavelength of EL spectra at 1 A/cm2. (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 In0.08Ga0.92N stress-release layer and an AlN dislocation confinement layer. (f) EL spectra of an LED at different current densities, 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.

Close modal

The use of a porous GaN layer can also suppress strain. In 2020, Pasayat et al.164 fabricated InGaN-based μLEDs on porous GaN and compared the performance of μLEDs fabricated on porous and non-porous regions with various tile sizes. It was found that the emission wavelength of μLEDs differed depending on whether a porous underlayer was present. In particular, μLEDs 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 relaxation. Subsequently, they demonstrated that 6 × 6 μm2 μLEDs with an on-wafer EQE of 0.202% at 10 A/cm2 emitted a wavelength of 632 nm under optimized growth conditions for n-In0.04Ga0.96N on a porous GaN substrate and a tile size of 11 × 11 μm2.165 RSM data [Fig. 13(b)] also revealed that the 440 nm-thick underlying n-In0.04Ga0.96N layer was 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 a chip size of 100 × 200 μm2. Figure 13(c) presents the average indium content in the MQWs and the peak wavelength in the EL spectra at 1 A/cm2 as a function of the thickness of the low-temperature buffer layer. It was seen that the average indium content and the peak wavelength 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 7 nm, 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, respectively. This indicates that achieving a high EQE in current InGaN red μLEDs with a longer wavelength is challenging.

Strain relaxation by an In0.08Ga0.92N layer on top of u-GaN, which allows for relatively longer wavelengths, was recently reported by Xing et al.167 The introduction of the In0.08Ga0.92N 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 g=[0002], confirming 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 calculated to be approximately 1.56 × 108 and 1.08 × 108 cm−2, respectively. This indicates that the SRL and DCL suppress the increase in the number of defects. In addition, Raman scattering spectral results revealed that the SRL and DCL alleviated the compressive strain. The EL spectra from LEDs with and without the SRL and DCL at various current densities [Fig. 13(f)] also illustrated effective strain release. For example, the peak wavelengths in the EL spectra at 1 A/cm2 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 factors in achieving longer emissions.

Conventional InGaN-based blue LEDs, which are grown on the sapphire, have a relatively simple MQW structure consisting of 2–3 nm 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 621 nm by optimizing the QW structure [Fig. 14(a)]. This suggests that a single QW structure in the InGaN layer for red emissions can suppress additional defect generation. A hybrid MQW structure 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 injection into the blue QW. Subsequently, the AlN capping layer between the InGaN-based red single QW and u-GaN prevented the decomposition 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 significantly increased from 3.3% to 6.0% with a ∼602 nm emission wavelength by decreasing the InGaN QW thickness from 3 to 1.9 nm. This effect of the QW thickness could be attributed to electron–hole wavefunction 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 Al0.8Ga0.2N interlayer and a 1.5 nm GaN cap layer on changes in the wavefunction overlap and peak wavelength. 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 631 nm. Collectively, these results suggest that it is necessary to carefully consider the tradeoffs between crystal quality (non-radiative recombination) and electron–hole wavefunction overlap (radiative recombination) in order to produce highly efficient InGaN-based red LEDs.

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., AIP Advances 12, 065125 (2022), with permission from AIP Publishing.168 (b) Computational investigation for the overlap of electron–hole wavefunctions 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 simulation 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 

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., AIP Advances 12, 065125 (2022), with permission from AIP Publishing.168 (b) Computational investigation for the overlap of electron–hole wavefunctions 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 simulation 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 

Close modal
Most of the recent results for InGaN-based μLEDs have reported a peak wavelength between 600 and 630 nm at the peak EQE, with very few emitting a wavelength over 650 nm. Thus, emitting pure red using InGaN-based red μLEDs 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
E Q W x , h = E g x + e 1 x , h + h h 1 x , h R y x , h e 0 F x h ,
(12)
where E g is the InGaN QW bandgap, e 1 and h h 1 are the first electron and hole energy levels, respectively, R y is the exciton binding energy, e 0 is the elementary charge, F is 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/cm2 [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 μLEDs 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 electrons and holes becomes worse as the QW thickness increases. To ensure a high efficiency and a longer emission wavelength for InGaN-based red μLEDs, the QW thickness, indium content, and energy band diagram for the QW should be optimized.
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/cm2. (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 

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/cm2. (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 

Close modal

Other novel approaches for emitting red from InGaN-based μLEDs 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 epitaxy growth on a microhole-patterned template and reported an emission wavelength of 642 nm at 10 A/cm2. Cai et 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/cm2. Europium doping is also a candidate for achieving the emission of longer wavelengths.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.

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 625 nm (green symbol).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 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 bulk due 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 average 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 efficiency, it remains an open question as to whether AlGaInP- or InGaN-based red μLEDs are suitable for the typical resolution regime for μLED displays (e.g., 1000–10 000 PPI).

FIG. 16.

Absolute EQEs of red LEDs and red μLEDs 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.

FIG. 16.

Absolute EQEs of red LEDs and red μLEDs 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.

Close modal

The integration of μLEDs and electronics drivers is an essential process for μLED 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-interconnection-free integration circuits. The characteristics of these integration techniques are summarized in Table II.

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 heterojunction field effect transistor for GaN-based LED200   Augmented reality (AR)/virtual reality (VR), head-up display (HUD) 
Monolithic hybrid integration (heterogeneous integration)  Flip-chip or wafer bonding for hybrid integration  A vertically stacked passive matrix μLEDs array,201 RGB LEDs on Si complementary metal oxide semiconductor AM drivers202–207   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 backplanes203,208 Smartwatch, smartphone 
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 heterojunction field effect transistor for GaN-based LED200   Augmented reality (AR)/virtual reality (VR), head-up display (HUD) 
Monolithic hybrid integration (heterogeneous integration)  Flip-chip or wafer bonding for hybrid integration  A vertically stacked passive matrix μLEDs array,201 RGB LEDs on Si complementary metal oxide semiconductor AM drivers202–207   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 backplanes203,208 Smartwatch, smartphone 

For large-screen display applications, a variety of methods have also been proposed for the mass transfer of μLEDs, including electrostatic,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.

TABLE III.

Summary of mass transfer technologies.

Technology Transfer rate Throughput Reliability Scalability Description
Electrostatic209   Moderate  Moderate  Low  Pick up μLEDs from donor wafer with intermediate substrate and place them on TFT or IC backplane 
Elastomer (transfer printing)210,211  0.98–19.7 ×106/h  Moderate  High  High 
Magnetic212,213  Moderate  Moderate  Low 
Fluidic assembly214–217   50 × 106/h  High  High  High  Liquid used to cause LEDs to fill in grooves in a target substrate 
Laser218   100 × 106/h  High  Low  High  Laser irradiation discharges LED chips from donor and place them on the target substrate. 
Roll-to-roll219,220  10 000 × 106/s  High  Moderate  Moderate  Roll transfer steps to an array of interconnected devices onto the target substrate. 
Technology Transfer rate Throughput Reliability Scalability Description
Electrostatic209   Moderate  Moderate  Low  Pick up μLEDs from donor wafer with intermediate substrate and place them on TFT or IC backplane 
Elastomer (transfer printing)210,211  0.98–19.7 ×106/h  Moderate  High  High 
Magnetic212,213  Moderate  Moderate  Low 
Fluidic assembly214–217   50 × 106/h  High  High  High  Liquid used to cause LEDs to fill in grooves in a target substrate 
Laser218   100 × 106/h  High  Low  High  Laser irradiation discharges LED chips from donor and place them on the target substrate. 
Roll-to-roll219,220  10 000 × 106/s  High  Moderate  Moderate  Roll transfer steps to an array of interconnected devices onto the target substrate. 

It is known that electrostatic transfer methods make it possible to transfer μLEDs 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 optoelectronics and electronics.210,211 In this process, soft elastomeric stamps are utilized to remove μLEDs or ICs from the donor substrate and place them on the target substrate to produce 2D and 3D arrays of μLED chips.207,211 In the case of magnetic transfer,212,213 LED chips are deposited with a magnetic layer and fixed to the (electro-)magnetization transfer stamp for chip transfer to the host substrate via magnetic force. In addition, fluidic self-assembly methods use different driving forces, such as gravity214 and directional surface tension,215 and shape-conforming structures, adhesives, liquid solder, or two different liquids are required.216 In particular, automated reel-to-reel fluidic self-assembly217 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 substrate 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 μLED 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 collection and placement of an array of electronic components (e.g., TFTs) on a temporary substrate; (ii) the separation of the μLED 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.

Despite these advances in transfer technology, for μLEDs 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 magnetic-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 μLEDs are placed in a bath chamber and a cluster of μLEDs 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 μLEDs, their movement is controlled by magnets. By applying a localized DEP force centered around the receptor holes, the μLEDs 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 μLEDs were pulled into the receptor hole as the angle decreased. Assembly occurred when the angle was below 10° and when the DEP force was dominant over the magnetic force. Figures 17(b)–17(d) show μLEDs 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 μLEDs 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 (Vpp) 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.

FIG. 17.

(a) DEP force acting on μLEDs regarding its angle to the receptor hole. (b–d) Camera images of μLEDs during the three stages of the assembly process and schematic 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 μLEDs chip dispenser. Chips are randomly oriented in solvent. (g) Formation of AlNd↓ state by pressure and drying 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 μLEDs and AlGaInP red μLEDs on CMOS. (i) A cross-sectional SEM image confirming successfully integrated μLEDs 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 

FIG. 17.

(a) DEP force acting on μLEDs regarding its angle to the receptor hole. (b–d) Camera images of μLEDs during the three stages of the assembly process and schematic 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 μLEDs chip dispenser. Chips are randomly oriented in solvent. (g) Formation of AlNd↓ state by pressure and drying 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 μLEDs and AlGaInP red μLEDs on CMOS. (i) A cross-sectional SEM image confirming successfully integrated μLEDs 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 

Close modal

Furthermore, Hwang et al.221 developed a method for quickly aligning μLEDs chips at the wafer scale by controlling the van der Waals (vdW) force between the chip and the interposer, which they refer to as fluidic-assisted self-alignment transfer (FAST) [Fig. 17(f)]. In this technique, the top (Au face) and bottom (AlN face) of μLED chips are designed to exhibit different vdW forces, enabling their selective 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 μLED chips (the contacting stage), a random force is applied in all directions. The μLED chips can easily be flipped over because the adhesion force between the μLED chips and substrate is low in wet conditions. As the align bar sweeps over the μLED chips (the sweeping stage), a dry region is formed under the AlN face because of the application of pressure (the drying stage), which significantly increases the vdW and adhesion forces and maintains the AlNd↓ state when resubmerged in the solvent. (AlNd↓, Aud↓, and Auw↓ 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 face, the solvent readily penetrates the space between Au and the substrate, transitioning from a dry (Aud↓) to a wet (Auw↓) state and reducing the vdW and adhesion forces, which is illustrated in the SEM images in Fig. 17(g). Additionally, the AlN face is engineered to facilitate solvent washing and to accelerate the drying process. Thus, the transition to AlNd↓ 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 μLED 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 cm2/V·S, 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 cm2/V·S is desirable to fabricate high-performance μLED displays. Recently, Qi et al.222 demonstrated a full-color μLED display on a CMOS backplane via FC bonding technology. As illustrated in Fig. 17(h), an InGaN-based blue/green dual-color μLED array with a subpixel size of 15 × 15 μm2 was first bonded to the CMOS via Au-Sn at 220 °C, and then the silicon substrate was removed. AlGaInP-based red μLEDs with a subpixel size of 15 × 15 μm2 were then bonded to the CMOS via Au–In at 180 °C, which is lower than the Au–Sn bonding process and does not damage the previously bonded blue/green μLED 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 μLEDs and AlGaInP μLEDs 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 μLEDs confirmed the development of a III–V-based inorganic full-color display operating on a CMOS.

To implement μLED displays, a mass repair process is vital. However, this section only briefly describes detection and repair technologies; for more detailed information, please refer to previous studies.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 μLEDs 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 tuned to less than 2 μm, allowing it to be used for the accurate analysis of μLEDs.226 However, with PL detection, the electrical characteristics of μLEDs cannot be detected. To resolve this, PL detection has been combined with confocal Raman microscopy,227 while tuning the laser excitation wavelength and power228 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 μLEDs. Thus, a method of using one row or one column of an LED array in each test has been proposed.230 The main disadvantage of EL detection is the potential for damage when the probes come into contact with the LED chip. To prevent this damage, flexible probes based on the principle underlying MEMS switches231 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 information from inside μLEDs rather than the LED surface region.232 OCT is promising due to its high depth resolution below 10 μm, 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 μLEDs has been demonstrated,233 where if one is damaged 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 transferred to a redundancy circuit for repair. A hybrid repair method combining 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 μLEDs.207 

Collaboration between academia and industry is important to accelerate the commercialization of μLED displays. Therefore, it is necessary to understand industry concerns, requirements, and approaches. In this section, we briefly review the results reported by global companies on μLED 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), HfO2-passivated devices outperform SiO2-passivated ones. This is due to the fact that the HfO2 layer more effectively prevents the additional damage that can occur during the deposition of the second passivation layer. In addition, Lumileds compared the EQEs of InGaN- and AlGaInP-based red μLEDs 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 μLEDs decreased as the size was reduced (from 9 to 4 μm), whereas that of InGaN red μLEDs was almost constant regardless of the size (moving from 6.7 to 2.2 μm) and the current density. As a result, InGaN-based red μLEDs 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 μLEDs can be strategically used to fabricate μLED 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%, suggesting that InGaN red LEDs can operate with a high EQE for ultrahigh resolution displays with further development [Fig. 18(c)].

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 μLEDs. 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 μLEDs 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 μLEDs 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 

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 μLEDs. 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 μLEDs 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 μLEDs 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 

Close modal

LG display also demonstrated a 100 PPI 12-inch active-matrix μLED 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/m2, all based on a μLED light source. This stretchable μLED display could potentially be applied to a variety of display designs once μLED 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 μLEDs 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 μLEDs. Figure 18(e) displays nine micro-display device dies with a pixel mesa size of 6 μm on a 4-inch wafer. This successful wafer-level device demonstrates that eutectic metal bonding can be used to integrate μLED arrays and CMOS IC drivers for mass production.

The present review describes recent advances in μLEDs based on compound semiconductors, AlGaInP, and InGaN. As the area ratio of the sidewall surface increases with chip miniaturization, light extraction 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 μLED 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 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 μm2 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 recombination 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 development of several mass transfer technologies has enabled the very rapid and reliable assembly of μLED array. However, if the chip size is reduced to 100 μm2 or lower, other methods, such as wafer-based technologies, should be considered. Overall, efficiency is a critical issue for the commercial implementation of μLED displays, and μLED researchers are expected to play a more active role than ever before.

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).

The authors have no conflicts to disclose.

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 (supporting); Funding acquisition (equal); Writing – original draft (equal); Writing – review & editing (equal).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

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.
14.
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.
Lähnemann
,
V. M.
Kaganer
,
K. K.
Sabelfeld
,
A. E.
Kireeva
,
U.
Jahn
,
C.
Chèze
,
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.
Kovács
,
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.
Leszczyński
,
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.
Dupré
,
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
).
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
).
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.
Morkǫ
,
Appl. Phys. Lett.
88
,
122104
(
2006
).
55.
L.
Lymperakis
,
P. H.
Weidlich
,
H.
Eisele
,
M.
Schnedler
,
J. P.
Nys
,
B.
Grandidier
,
D.
Stiévenard
,
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.
Bartoš
,
O.
Romanyuk
,
J.
Houdkova
,
P. P.
Paskov
,
T.
Paskova
, and
P.
Jiříček
,
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
).
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
).
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.
González-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.
Vögl
,
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.
Morkoç
,
Ü.
Özgür
,
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
).
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.
Nötzel
,
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
, and
S. 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
,
Ü.
Özgür
, and
H.
Morkoç
,
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
).