1. INTRODUCTION

To date, light-emitting diodes (LEDs) have been used in a wide variety of applications in our daily lives [1–3]. In recent years, micro-LEDs have received considerable attention for applications in next-generation displays and visible light communication (VLC) due to their fast response, light weight, low power consumption, high brightness, and high efficiency [4–9]. The use of blue micro-LEDs as pumping sources for different color converters to achieve full-color micro-displays has been extensively studied [6,10–12]. However, color converters such as quantum dots (QDs) or phosphors encounter several challenges, including luminous uniformity, color conversion efficiency, and stability issues [12–14]. Therefore, assembling red/green/blue (RGB) micro-LEDs to realize full-color micro-displays is a viable method that uses ultra-small LEDs and allows each pixel’s brightness to be regulated in a dynamic range. As the LED chip size shrinks to below 10 μm, blue and green InGaN-based micro-LEDs are able to retain remarkable performance in terms of external quantum efficiency (EQE) [15]. However, most red LEDs are made of aluminum indium gallium phosphide (AlInGaP) material systems, which are not stable at high temperatures and suffer from significant size-dependent efficiency droop due to serious surface recombination [16–22]. Moreover, combining different semiconductors makes the construction of RGB micro-LEDs difficult and expensive. Therefore, InGaN-based red micro-LEDs are of interest for realizing monolithic RGB micro-LED technology [23–27]. Recently, much effort has been dedicated to developing InGaN-based red LEDs. Ohkawa et al. realized 633 nm InGaN red LEDs grown on thick underlying GaN layers with reduced in-plane residual stress, exhibiting an EQE of 1.6% at 20 mA [28]. InGaN micro-LEDs have exhibited stable temperature dependence and higher light output power (LOP) density compared to InGaP [29,30]. In addition, Wuu et al. have investigated the size effect on InGaN red micro-LEDs, and the output power density is very similar to small micro-LED chips at the same injection current density, which is different from AlGaInP micro-LEDs [31]. A previous paper showed that InGaN RGB micro-LEDs cover 84% of Rec.2020, which is very promising for achieving full-color micro-displays [32].

RGB micro-LEDs hold a lot of promise for VLC applications since they can be utilized for both displays and optical communication simultaneously. In this regard, having high-efficiency blue, green, and red micro-LEDs monolithically integrated on a single platform is desired. In terms of efficiency and modulation bandwidth, InGaN-based blue and green micro-LEDs are particularly promising for such device applications [7,9,33–35]. In our recent study, we demonstrated a $2\times 2$ green micro-LED array with emission aperture size of 50 μm that achieved the world’s fastest record for visible light wireless data communication of LEDs [36]. However, InGaN-based red LEDs still have the serious problem of efficiency reduction since red light emission requires high indium content in quantum wells (QWs) [24,37]. The large lattice mismatch between InN and GaN would cause a significant number of defects to occur during the growth of high indium concentration QWs, limiting InGaN red LED device performance [38]. Furthermore, the InGaN red LED with high indium concentration QWs would suffer the quantum-confined Stark effect (QCSE), which separates the overlap of electron–hole wave functions and hinders radiative recombination. However, the modulation bandwidth is limited by QCSE induced by the polarization related electric field in $c$-plane micro-LEDs, and there is yet room for improvement. Semipolar micro-LEDs with reduced QCSE have been quite successful in achieving high-modulation-bandwidth systems when it comes to blue and green light emission [39]. When it comes to red micro-LEDs, the existing QCSE induced by a high indium content severely limits its potential for VLC applications. Many other approaches to address the issue of high-indium-content InGaN growth have been reported, including growth on semipolar or polar surfaces, nanowire or nanocolumn structures, InGaN QDs, and so on [40–43]. However, InGaN red micro-LEDs grown on $c$-plane substrates are currently the best choice owing to their compatibility with current production processes and the ability of the matured InGaN-based blue/green LEDs to be integrated with minimal modification. InGaN red micro-LEDs for VLC applications have become a research focus, and high-bandwidth red micro-LEDs are desired to realize a high-performance RGB micro-LED VLC system.

In 2020, Carreira et al. demonstrated AlGaInP red micro-LED devices printed on diamond and glass with maximum bandwidths of 170 MHz and 85 MHz, respectively [44]. Later in the same year, Haggar et al. reported a semipolar (11–22) InGaN/GaN LED with a peak emission wavelength of 610 nm, exhibiting a modulation bandwidth of 140 MHz [45]. In 2021, Haggar et al. reported a semipolar red LED with an emission wavelength of 596 nm, achieving a maximum bandwidth of 150 MHz [46]. However, the reported values of modulation bandwidth are still poorer than those of blue and green micro-LEDs, whose values are limited for VLC applications. In this study, we report a high-efficiency InGaN red micro-LED device with a chip diameter of 25 μm, exhibiting a maximum bandwidth of 271 MHz at an injection current density of $2000{\mathrm{A}/\mathrm{cm}}^2$. High-efficiency InGaN red micro-LEDs hold great promise for realizing high-speed VLC as well as micro-display technologies (see Fig. 1).

 

Fig. 1. Schematic diagram of $c$-plane InGaN red micro-LED epitaxial structure; inset: optical image and illumination image of the red micro-LEDs.

Download Full Size | PDF

2. EXPERIMENTS

A schematic diagram of the $c$-plane InGaN red micro-LED epitaxial structure is shown in Fig. 1. The epitaxial structure is described below from the bottom layer to the top layer. Metal organic vapor-phase epitaxy (MOVPE) was used to grow the InGaN red micro-LED structure on a $c$-plane patterned sapphire substrate (PSS). First, the epitaxial process started with a 2 μm undoped-GaN layer to reduce the residual stress, followed by an 8 μm thick n-GaN:Si ($n=2\times {10}^{18}{\mathrm{cm}}^{-3}$) and 0.8 μm n-AlGaN:Si [$n=(2–3)\times {10}^{19}{\mathrm{cm}}^{-3}$]. Then, 15 pairs of $\mathrm{GaN}/{\mathrm{In}}_{0.08}{\mathrm{Ga}}_{0.92}\mathrm{N}$ superlattice (SL) layers with a thickness ratio of 6 nm/2 nm were grown for strain engineering. The SL layers improved the quality of the following InGaN QW, thereby mitigating the effect of QCSE. Then, a 15 nm n-GaN:Si, 2 nm blue InGaN single-QW (SQW), and red InGaN double-QW (DQW) with a high indium content were grown as active regions. The QW structures consist of an ${\mathrm{In}}_{0.2}{\mathrm{Ga}}_{0.8}$$\mathrm{N}$ blue SQW (2 nm) with $\mathrm{GaN}(2\mathrm{n}\mathrm{m})/{\text{Al}}_{0.13}{\mathrm{Ga}}_{0.87}$N (18 nm)/GaN (3 nm) barrier layers, and ${\mathrm{In}}_{0.34}{\mathrm{Ga}}_{0.66}$$\mathrm{N}$ red DQWs(2.5 nm) with $\mathrm{AlN}(1.2\mathrm{nm})/\text{GaN}(2\mathrm{nm}){\text{/Al}}_{0.13}{\mathrm{Ga}}_{0.87}$$\mathrm{N}$ (18 nm)/GaN (3 nm) barrier layers. GaN was used to replace the ${\mathrm{Al}}_{0.13}{\mathrm{Ga}}_{0.87}$$\mathrm{N}$ part in the upper barrier of the second red QW. Finally, a 15 nm undoped-GaN layer, 100 nm $\mathrm{p}$-GaN:Mg layer, and 10 nm $\mathrm{p}+$-GaN:Mg contact layer were grown [47].

Before the micro-LED process flow, the epi-wafer was cleaned with acetone, isopropanol, and deionized water to remove unwanted materials on the surface. The first process was a 200 nm thick indium tin oxide (ITO) layer deposited on the top of the $\mathrm{p}$-GaN layer as the current spreading layer. Micro-LEDs with circular mesas ranging from 100 μm to 25 μm were fabricated on the same wafer by a photolithographic process. The ITO layer and mesa were etched by wet and dry etching through HCl solution and an inductively coupled plasma reactive ion etching (ICP-RIE) system, respectively. Then, the samples were annealed at 450°C for 5 min to form a $\mathrm{p}$-type ohmic contact. Subsequently, Ti/Al/Ti/Au metals with a thickness ratio of 25/125/45/75 nm were deposited as the electrodes. A 30 nm thick ${\mathrm{Al}}_2{\mathrm{O}}_3$ layer was deposited on the wafer via atomic layer deposition (ALD), followed by a 200 nm thick ${\mathrm{SiO}}_2$ layer deposition through plasma-enhanced chemical vapor deposition (PECVD) and the via hole process by ICP-RIE. The aluminum oxide (${\mathrm{Al}}_2{\mathrm{O}}_3$) passivation layer was grown at 300°C in an argon (Ar) environment using the trimethylaluminum (TMA) and ${\mathrm{H}}_2\mathrm{O}$ cycle with Ar purging. The passivation layer helps to repair surface damage caused by dry etching and isolates the metal electrodes. Next, a pad metal comprising Ti/Al/Au metals having a thickness of 20/50/300 nm was evaporated. Finally, a distributed Bragg reflector (DBR) consisting of 5.5 pairs of ${\mathrm{SiO}}_2/{\mathrm{TiO}}_2$ layers was introduced to enhance the light extraction efficiency of packaged devices.

To achieve high performance in VLC applications, micro-LED device arrays were fabricated by a parallel connection and optimized from different aspects. A round active area and circular-shaped electrodes were designed to improve the current spreading and light extraction efficiency [48]. Furthermore, the surface defects caused by the mesa etching process cannot be effectively repaired using traditional passivation methods such as PECVD, and hence the ALD technique to passivate etched mesas for decreasing leakage currents of the devices [49,50]. Finally, the piezoelectric polarization induced by an electric field in the $c$-plane causes QCSE, resulting in low oscillator strength and weak light emission [51]. The QCSE can be mitigated by growing SL layers before the growth of QW active layers for stress engineering so that the quality of QWs can be improved, thereby enhancing device performance.

3. RESULTS AND DISCUSSION

The optical and electrical properties of 25-μm-sized packaged micro-LEDs are summarized in Fig. 2. The voltage–current density-output power curve of the device is shown in Fig. 2(a). The relation of current density and LOP is linear before the output power reaches 0.18 mW. The forward voltage is 3.21 V, corresponding to an injected current density at $20{\mathrm{A}/\mathrm{cm}}^2$. The characteristics of the device are reasonably consistent with red micro-LEDs in a previous study [28]. Figure 2(b) shows EQE as a function of injection current density. The maximum value of EQE is 5.02% when the injection current density is $112{\mathrm{A}/\mathrm{cm}}^2$. Then, the EQE droops from 5.02% to 4.56% when the current density is up to $400{\mathrm{A}/\mathrm{cm}}^2$ due to the existing QCSE in the polar $c$-plane GaN. The electroluminescence (EL) emission spectra of the device are shown in Fig. 2(c). The peak wavelength and FWHM at different current densities are summarized in Fig. 2(d). The peak wavelength of the red micro-LED is shifted from 652 nm to 614 nm when current density increases from $32{\mathrm{A}/\mathrm{cm}}^2$ to $400{\mathrm{A}/\mathrm{cm}}^2$. This blueshift phenomenon is well known in InGaN-based blue and green micro-LEDs, and is caused by the screening effect of the polarization field and band filling effect of the localized state. It also occurs significantly in InGaN-based red micro-LEDs with high indium content. Moreover, the full width at half maximum (FWHM) of the micro-LED increases by 16.1 nm when the current density changes from $32{\mathrm{A}/\mathrm{cm}}^2$ to $400{\mathrm{A}/\mathrm{cm}}^2$. The results show that an InGaN red micro-LED with high EQE can be achieved by proper design and packaging, which reveals the great potential for replacing AlGaInP-based ones. Therefore, InGaN-based RGB micro-LEDs for full-color display applications can be implemented on $c$-plane sapphire substrates. It can solve the cost issue due to heterogeneous integration and make the monolithic full-color LED system a reality.

 

Fig. 2. (a) Light–current–voltage (L-I-V) characteristics; (b) EQE as a function of current densities; (c) electroluminescence spectra at different current densities; (d) wavelength shift and FWHM as a function of current densities for the 25-μm-sized packaged micro-LED.

Download Full Size | PDF

This paper also analyzes the size effect through different mesa diameters (25 μm, 50 μm, 75 μm, and 100 μm). The optical output power of micro-LEDs with different mesa sizes as a function of injection current densities is shown in Fig. 3(a). An integrating sphere system (Isuzu Optical SLM-12 system) with ${\mathrm{BaSO}}_4$ surface coating was used for optical spectrum collection. The output powers are 0.137 mW, 0.570 mW, 1.392 mW, and 2.572 mW for diameters of 25 μm, 50 μm, 75 μm, and 100 μm, respectively. The output power ratios between 25 μm and the others are 0.240, 0.098, and 0.053, which are slightly less than the emission area ratios of 0.250, 0.111, and 0.063, respectively. This can be attributed to the lower surface aspect ratio (i.e., area of surface/area of sidewall) for 25 μm-sized micro-LEDs, leading to the poor EQE performance compared to others [32].

 

Fig. 3. (a) Optical output power; (b) wavelength shift as a function of current densities for packaged micro-LED with different chip sizes; (c) EQE as a function of current densities for packaged micro-LED with different chip sizes; (d) benchmark of EQE value for InGaN-based red micro-LED.

Download Full Size | PDF

Figure 3(b) shows that the blueshift effect occurs in micro-LEDs with different sizes as a function of current densities. Due to the screening effect of a piezoelectric field and the band filling effect, the blueshift phenomenon can be observed. In Fig. 3(b), the initial wavelength starts at 652 nm at $32{\mathrm{A}/\mathrm{cm}}^2$ and shifts to 617 nm at $300{\mathrm{A}/\mathrm{cm}}^2$ for a 25-μm-sized micro-LED. For 50 μm/75 μm/100-μm-sized devices, the emission wavelength shifts from 663 nm/667 nm/671 nm to 622 nm/631 nm/637 nm, respectively. The peak emission wavelength tends to be shorter as the chip size decreases. This is according to a small micro-LED with a high threshold voltage, high current densities, and less stress, which bring about the QCSE screening and band filling effects. Hence, the shortest wavelengths of micro-LEDs appear on the smallest mesa size.

Figure 3(c) shows EQE as a function of injection current densities for all micro-LEDs. The maximum EQE is 5.02% at $112{\mathrm{A}/\mathrm{cm}}^2$, 5.25% at $72{\mathrm{A}/\mathrm{cm}}^2$, 6.37% at $50{\mathrm{A}/\mathrm{cm}}^2$, and 6.71% at $35{\mathrm{A}/\mathrm{cm}}^2$ corresponding to diameters of 25 μm, 50 μm, 75 μm, and 100 μm, respectively, which are higher than previously reported values [23,28,31]. The improvement comes mainly from the inclusion of suppressed QCSE by SL, low leakage current by ALD passivation, great current spreading by circular electrodes, as well as enhanced light extraction efficiency by the bottom DBR with proper packaging. EQE will gradually decrease with current density when the current density rises beyond the operating point of maximum EQE. Even though the sidewall has been passivated, the EQE drop is still unavoidable in a high current density state. The decline of EQE will be exacerbated by current crowding. The current crowding effect becomes more significant as micro-LED chip size increases, leading to higher local junction temperature [52]. In addition, the uncertainties also affect the knowledge of numerous key parameters that can cause droop, such as Auger coefficients, Auger recombination, electron overflow, carrier delocalization, and polarization charges at AlGaN/GaN and InGaN/GaN interfaces [53]. The carrier concentrations in LED active regions reveal that current crowding can increase local non-radiative recombination, heating, and electron overflow, resulting in decreased efficiency in larger LEDs at high injection current densities [54]. Therefore, smaller micro-LEDs (i.e., 50 μm and 25 μm) exhibit a lower efficiency droop as compared to the rest. The droop efficiencies of 25 μm and 50 μm are 7.52% and 7.42%, which are less than 16.01% and 17.02% for 75 μm and 100 μm, respectively.

The first InGaN-based red micro-LED was created in 2012 by Ohkawa et al. [55]. A $500\mathrm{\mu m}\times 500\mathrm{\mu m}$ emission area emitted a peak wavelength at 740 nm. This was the first time to overcome the growth limitation between high indium content and high growth temperature. Nevertheless, the EQE is much lower than that of a commercial blue micro-LED due to the lattice mismatch that induces a piezoelectric field in QW. Fortunately, strain engineering is beneficial to improve the quality of QW, lightening the QCSE and numerous defects in the active region. Nunoue et al. first demonstrated a nitride-based red LED with LOP exceeding 1 mW at 20 mA [24]. The AlGaN interlayer embedded on the QW could compensate for the strain effect of a high-indium-content QW. Hence, the EQE was 2.9%, achieved on a $460\mathrm{\mu m}\times 460\mathrm{\mu m}$ injection area. EQE was further improved to 3.2% through the SL layer on a $40\mathrm{\mu m}\times 40\mathrm{\mu m}$ InGaN amber micro-LED [56]. Based on this paper, Li et al. employed a tunnel junction contact to realize a high peak EQE of 4.5% at $1{\mathrm{A}/\mathrm{cm}}^2$ on a $60\mathrm{\mu m}\times 60\mathrm{\mu m}$ emission area [57]. EQE values in recent papers versus the emission area of a red micro-LED chip are summarized in Fig. 3(d) to compare with the results of this experiment [23,28,30,56–65]. To our best knowledge, this study shows the highest EQE values of 5.02%, 5.25%, 6.37%, and 6.71% for diameters of 25 μm, 50 μm, 75 μm, and 100 μm, respectively.

To analyze the VLC performance of $c$-plane InGaN red micro-LEDs, we measured the frequency response of the micro-LED array through a vector network analyzer (VNA; HP 8720ES). The alternating current signal generated by VNA was coupled with a direct current via a bias tee. Then, the coupled signal was fed into micro-LEDs by a microprobe (ACP40-GS-250), resulting in an optical signal with a small signal. The optical signals of the micro-LEDs were collected by a plastic optical fiber (POF500) and transmitted to a photodetector (Graviton, SPA-3). The optical signals received by the photodetector would be transferred into electrical signals, analyzing the frequency responses via VNA. Figure 4 illustrates the frequency response of the red micro-LEDs with different sizes. The micro-LEDs exhibited a maximum bandwidth of 94 MHz at $350{\mathrm{A}/\mathrm{cm}}^2$, 163 MHz at $800{\mathrm{A}/\mathrm{cm}}^2$, 201 MHz at $1500{\mathrm{A}/\mathrm{cm}}^2$, and 271 MHz at $2000{\mathrm{A}/\mathrm{cm}}^2$ corresponding to diameters of 100 μm, 75 μm, 50 μm, and 25 μm, respectively. Because the SL structure relieves the stress in red micro-LEDs, the influence of QCSE is mitigated. It can shorten the carrier lifetime, resulting in a high modulation bandwidth. Moreover, the frequency bandwidth can be effectively increased by the density of the injection current, which is attributed to a reduction in carrier lifetime at higher current densities and built-in electric field screening. The ABC rate equation model of device current can further explain this phenomenon. The rate Eq. (1) can be described by three factors, namely, Shockley-Read-Hall (SRH) recombination current, radiative recombination current, and Auger recombination current. Hence, the total current can be expressed as

 

Fig. 4. Frequency responses for different diameters of (a) 100 μm, (b) 75 μm, (c) 50 μm, and (d) 25 μm micro-LEDs. (e) Frequency versus current density for different chip sized micro-LEDs; (f) time resolved photoluminescence (TRPL) for the InGaN micro-LED.

Download Full Size | PDF

Hence, when the injection current increases, the number of injected carriers (i.e., electrons and holes) in the active region will increase, resulting in a higher carrier density and short differential carrier lifetime. According to Eq. (2), the recombination coefficient is inversely related to the differential carrier lifetime. Recombination efficiency will be enhanced with the number of injected carriers, resulting in a low carrier lifetime. Therefore, a larger injection current will lead to a short differential carrier lifetime at a high carrier density in the active region, which improves the modulation bandwidth of the device. This phenomenon can be observed in Fig. 4(e), which summarizes the $-3\mathrm{dB}$ bandwidth for different mesa sizes at different current densities. The 100-μm-sized device showed a lower bandwidth than 25-μm-, 50-μm-, and 75-μm-sized ones at low current densities due to the current crowding effect and the limitation of the resistance capacitance (RC) time constant, but all of them are almost proportional to the current density. In addition, a small emitter diameter is also promoted for a uniform current-spreading effect, resulting in a high injection current and low junction temperature [52]. Hence, a smaller emitter diameter can be driven to a higher current density than a larger emitter diameter to obtain a higher $-3\mathrm{dB}$ bandwidth. Furthermore, the carrier lifetime can be rewritten from a carrier density relationship to a current density relationship as follows [66]:

If EQE does not change with current density significantly, we can fit the equation by a power law of $f_{-3\mathrm{dB}}\sim J^m$, where $m$ is a power factor changing from $0$ to $1$. If the radiative recombination dominates the modulation bandwidth, the value of $m$ will be close to 0.5 [68]. This can be explained by the bimolecular recombination mechanism as follows [69]:

At this time, the coefficient of SRH and Auger recombination is neglected. However, both should be considered in the modulation bandwidth because the injection current of LED is related to the rate equation. Therefore, the modulation bandwidth is determined by the ratio of the polynomial coefficients of Eq. (5). If the first term “A” of the polynomial is larger than the second term “$2B\alpha$” and the third term “$3C{\alpha }^2$,” the modulation bandwidth is a constant value. The factor of the power law is zero. If the third term “$3C{\alpha }^2$ of the polynomial is larger than the second term “$2B\alpha$” and the first term “A,” the modulation bandwidth will be a linear function. The factor of the power law is one. Generally, the modulation bandwidth depends on bimolecular recombination mechanism as shown in Eq. (6). Hence, the value of $m$ equal to 0.5 can be used as the watershed. When $m$ is smaller than 0.5, the modulation bandwidth is gradually dominated by SRH. Otherwise, the bandwidth will be taken over by the Auger recombination gently. The power factor of the modulation bandwidth depends on the ratio of $A$, $B$, and $C$. If the ratio of $A$, $B$, and $C$ is changed, the trend of the modulation bandwidth of micro-LEDs versus current will be different. According to a previous paper [70], the ratio between radiative recombination and Auger recombination was enlarged from $1.21\times {10}^{19}$ to $2.12\times {10}^{19}$ when the junction temperature was increased from 300 K to 500 K, and the linearity of the modulation bandwidth changed significantly at low current densities. We used the characteristics to identify the dominant carrier recombination mechanism. According to the fitting result, the value of $m$ obtained from Fig. 4(e) was 0.4834. The radiative recombination still dominated the recombination mechanism of the red micro-LED but slightly inclined towards SRH. This result can correspond to the size-effect analysis in Fig. 3. Figure 4(f) presents the time-resolved photoluminescence (TRPL) measurement of the InGaN red micro-LEDs with excited optical power of 8 mW. According to the double-exponential decay fitting, the average lifetime (${\tau }_{\mathrm{avg}}$) is about 1.01 ns, which is correlated with the $-3\mathrm{dB}$ bandwidth of the InGaN red micro-LEDs.

Moreover, high-speed VLC applications based on InGaN-based RGB micro-LEDs require high-efficiency InGaN red micro-LEDs with short carrier lifetimes. The relation between the modulation bandwidth and carrier lifetime can be expressed in the following equation:

The 25-μm-sized micro-LEDs exhibited the highest modulation bandwidth among the fabricated micro-LEDs, which is also the highest reported modulation bandwidth for red micro-LEDs, to our best knowledge. According to these results, we adopted 25-μm-sized micro-LEDs as light sources in the data transmission experiment. The data transmission characteristics of the red micro-LEDs were analyzed by an off–on keying (OOK) system. The test bit sequence was a non-return-to-zero (NRZ) $2^7-1$ pseudorandom binary sequence (PRBS7) generated by a bit pattern generator (Anritsu MP1800A). The eye diagrams were analyzed and recorded by an 86100A oscilloscope. Figures 5(a)–5(c) illustrate the detected NRZ-OOK eye diagrams at 200 Mbit/s, 300 Mbit/s, and 390 Mbit/s, respectively. There is a clear and open eye diagram at 200 Mbit/s. At 300 Mbit/s, the eye is beginning to close, and is virtually closed at 390 Mbit/s. This shows the potential application of these micro-LEDs at data rates of the order of 300 Mbit/s. The maximum achievable data rate of 350 Mbit/s has a bit error rate (BER) of $2.6\times {10}^{-3}$ satisfying the forward error correction (FEC) threshold of $3.8\times {10}^{-3}$. The measured bandwidth and data transmission rates imply that $c$-plane InGaN red micro-LEDs are very promising for VLC applications. The modulation bandwidth and data rates of micro-LEDs can be further improved by optimizing the active region, packaging condition, and efficient heat sinking design so that they can operate at a higher current density and achieve higher modulation bandwidths.

 

Fig. 5. NRZ-OOK eye diagrams for 25-μm-sized InGaN red micro-LEDs at (a) 200 Mbit/s, (b) 300 Mbit/s, and (c) 390 Mbit/s.

Download Full Size | PDF

4. CONCLUSION

In summary, we present an InGaN red micro-LED with high EQE and a high modulation bandwidth through the incorporation of SL structure, ALD passivation, and the bottom DBR. The SL structure is responsible for the improvement of InGaN QWs by mitigating QCSE, thereby increasing radiation recombination. The 25-μm-sized micro-LED exhibits a maximum EQE of 5.02% with a low efficiency droop at an injection current density of $112{\mathrm{A}/\mathrm{cm}}^2$. A maximum modulation bandwidth of 271 MHz is achieved by a $6\times$ 25-μm-sized micro-LED array at a high injection density of $2000{\mathrm{A}/\mathrm{cm}}^2$ with the data transmission rate of the device reaching up to 350 Mbit/s under an NRZ-OOK modulation format. These results indicate that InGaN red micro-LEDs are very promising for VLC and full-color micro-display applications via the monolithic integration of nitride-based semiconductors.