In this study, AlGaInP red light emitting diodes with sizes ranging from 5 to 50 micrometers were fabricated and characterized. The atomic layer deposition technology is applied to coat a layer of silicon dioxide for passivation and protection. The top emission area is covered by ITO layer to maximize the optical output. From the optical measurement, the linewidth and emission peaks shift very little among different current levels (from 30 to 150 A/cm2). High current level lifetests are performed and a 15 µm ALD device can last 27 hours of continuous operation at 100 A/cm2 before their diode junction failed. A much shorter lifetime of 5.32 hours was obtained when the driving current is raised to 400 A/cm2. When the same condition was applied to 15 µm PECVD devices, 25 hours and 4.33 hours are registered for 100 A/cm2 and 400 A/cm2 tests, respectively. The cross-sectional SEM reveals the voids, defects, and dark lines developed during the aging tests, and most of them are caused by top contact failure. The surface layers of ITO and SiO2 were melted and the dark lines which were originated from the top surface propagated through the device and led to the eventual failure of the diode. The optical intensity degradation slopes of different sizes of devices indicate a large device can last longer in this accelerated aging test. The efficiencies of the devices are also evaluated by the ABC model and the fitted bimolecular coefficient ranges from 1.35 to 3.40×10−10 cm3/s.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The rise of full-color displays brings the micro and mini LEDs to our attention lately because of their long lifetime, bright color, and good power efficiencies [1–3]. To achieve a high resolution in a display module, the size of the pixels needs to be small. For the resolution of 2000 pixel-per-inch (PPI) or beyond, this number can be as small as 12 µm (for a monochromatic screen and 4 µm if full-color scheme is considered). Methods to implement such a display without monolithic growth include integration of a color conversion layer with monochromatic LED arrays [4–7] and mechanical transfer of individual red, green, and blue chips onto the same substrate . While using a color conversion layer is an attractive method, the need to realize an all-semiconductor and high resolution RGB pixel array cannot be ignored . Nowadays the blue and green devices based on gallium nitride materials have shown great progress in terms of device scaling [10,11]. Meanwhile, the red color device is GaAs-based and its traditional size is around 100 µm or larger , and to fabricate a device with size less than 10 µm in GaAs based materials is not an easy job. Previous researches indicated that the surface recombination velocity of GaAs based material is high [13,14] and thus the degradation in quantum efficiencies can be significant [12,15,16]. Recent advances in fabrication processes have put red micro LEDs in a better position to play an important part in the micro-display [15,17]. Reported earlier this year, M. Wong, et al., applied atomic layer deposition (ALD) technology with plasma surface treatment to enhance the 20 µm device performances, and this shed some light for researchers to continue scaling efforts for these red devices . In this paper, the red micro LEDs with mesa width as small as 5 µm were fabricated and tested for their optical and electrical characteristics. Both plasma enhanced chemical vapor deposition (PECVD) and ALD technologies are used for sidewall passivation and protection. The lifetime of the device is evaluated via accelerated burn-in processes. In a display module, pulse-width-modulation (PWM) is widely applied to drive micro LEDs or other display elements. This modulation method can effectively reduce the average power in the module. So those devices that can pass the continuous high current aging shall be able to handle the PWM signal. The partial quantum efficiencies of devices were measured and fitted by the standard ABC model [17,18]. We hope these results can further extend our knowledge towards these micron-scale devices.
2. Device fabrication and measurements
The wafer was purchased from Epileds Technologies Inc. in Taiwan. The active layer is made of epitaxial layers of III-V compound semiconductors. The window layer is a Mg-doped GaP with about 2 µm in thickness, and the active layer is made of AlGaInP multiple quantum well with 0.2 µm in thickness. The n-type layer is composed of AlInP and AlGaInP layers, both of which are doped by Tellurium, and the total thickness is around 4 µm. The bottom n-type contact layer is silicon-doped GaAs. The wafers went through standard semiconductor processes to make micro-LED devices. A square-shaped mesa was created to isolate and regulate the current flow, and the mesa-width design is 5, 15, 20, and 50 µm, respectively. The dry etch by inductive coupled plasma (ICP) was conducted to define the size of the mesa. The reactive gases used in the ICP etching process are Cl2, BCl3, and Ar. Once the ICP power reached steady state, the computerized program can control the etching time to obtain suitable depth. The etch rate of the ICP is tuned to be 43.5 nm/sec. On the top of the mesa, an ITO layer of 100 nm was deposited for the purpose of current conduction and also preservation of light-emission window. The p-type and n-type metal contacts are located outside of mesa area to maximize the light output, as shown in Fig. 1(a). The sidewall passivation was accomplished by either PECVD or ALD technologies. In the PECVD sample, a 200 nm layer of SiO2 was deposited. For ALD-coated samples, the procedure is the same except the thickness of SiO2 is reduced to 20 nm, and it was finished by a Picosun system. The choice of the SiO2 layer thickness mostly follows the general recipe, and part of the reason is due to the huge deposition rate difference between PECVD and ALD system (about 460:1 in our case). It will be difficult to find a thickness design that can be suitable both systems. To remove ALD-deposited SiO2, another ICP machine was used and SF6 / C4F8 gas mixture was the main etchants. Regular Cr-Au metal was evaporated by an E-gun system for pad contact and metal connector network if necessary.
After fabrication, the individual devices are tested on the wafer by electrical probe stations as shown in Fig. 1(b). The I-V characteristics can be measured by Keithley 2400 source meters, and the optical results are collected by an integrating sphere (Pegasus Instrument Inc., PG2000). When the accelerated aging tests were carried out, a calibrated photodetector (Newport 818 series) was used to collect the relative optical power intensity. After the accelerated aging tests, the device was examined in a focus ion beam (FIB) system made by TESCAN (model GAIA3). The system uses focused Ga ion to reveal the interior layer of the device. The beam current was controlled at 6.76 nA and the beam size is 50 nm.
3. Results and discussions
3.1 Electrical characteristics
Figure 2 shows the electrical characteristics of these devices with different passivation layers. There are no cases for no-passivation condition due to the small physical sizes. From these plots, the series resistances of the diodes can be extracted and they are listed in Table 1. Small devices, like 5 or 15 µm ones, tend to have high resistance. The comparison between PECVD and ALD conditions show that the PECVD-coated diodes are usually less conductive than ALD ones.
Meanwhile, on the reverse bias side, as shown in Figs. 2(c) and (d), results from two groups are distinguished from each other. The ALD group has much lower leakage currents in average. A 7.8 times of reduction is recorded in the 50 µm device at −4 V. This distinctive separation of leakage current indicates that the effect of passivation should be quite different between PECVD and ALD coatings. The obtained reverse leakage density at −4 V is comparable or smaller than the devices reported earlier this year  (in the range of mid 10−6 A/cm2 at −3 V) and an increase from 4.67×10−7 A/cm2 to 6.22×10−6 A/cm2 was observed between 50 µm and 5 µm ALD devices, which might be caused by increased surface-to-volume ratio of the small devices and thus extra leakage path under reverse bias .
By using ITO as the contact layer, the top emission area can be enlarged. However, the shortcoming is the higher resistance across various devices. This situation can lead to extra heating generated during current injection which can cause the early breakdown and the performance degradation of the devices in turn. A better electrical conduction method needs to be found to facilitate the fabrication and extend the lifetime of the red LEDs. Generally speaking, except the epitaxial layer problem, the resistance in the diode can rise from the ITO/metal structure due to several reasons: (a) the thickness of the ITO; (b) the quality (or inherent conductivity) of ITO; (c) Post-processing conditions. In our cases, because we have all wafers deposited their ITO at the same time (in the same run), so we believe (a) and (b) are less likely. Meanwhile, the subsequent PECVD and ALD depositions could make these ITO layers seeing different temperatures: 300°C for PECVD and 250°C for ALD. Both deposition took a similar time period to finish. Another possible factor is the hydrogen plasma of PECVD damaging the ITO’s conductivity [19–21]. The environment of hydrogen plasma in PECVD chamber leads to elemental tin to yield on the surface and later gets oxidized when exposed. These process can effectively reduce the conductivity of the PECVD-treated ITO film and cause the subsequent high series resistance in the device. Another concern about ITO is from the processing side. Researchers have shown the ITO layer etched in the subsequent ICP etch . From the SEM and microscope pictures we took, no significant ICP etch during the window opening step can be observed. However, images of pits formed on ITO surface can be seen under the microscope, and these could be the potential culprit of reduced ITO conductivity in the device.
3.2 Optical characteristics
The spectra of the ALD coated devices are shown in Figs. 3(a) and (b). The peak wavelengths of the spectra are fixed around 625 nm, and the shift due to current injection is little. The measured values are around 0.35 nm in 30–150 A/cm2 range. It was discussed before that micro-LED has better heat dissipation because of large surface to volume ratio [23,24], and thus the junction temperatures have smaller current-dependence than traditional ones. The linewidth was also extracted from these measurements. The full-width-half-maximum (FWHM) as wide as 9.66 nm was detected, and also varies little for various current levels. The summarized results is charted in Table 1. Some of the peak changes are recorded as zero because the wavelength shift due to current injection cannot be resolved by the equipment. The narrow linewidth spectra indicate a good red color on the CIE color space and will be helpful for wide coverage of color in display applications.
The absolute optical power of the individual device is difficult to be precisely measured because of the absorbing GaAs substrate and the on-wafer probing setup we choose. The relative power can be obtained via an integrating sphere and we can deduce specific trend on device sizes and their partial efficiencies.
Further analysis on L-I characteristics, we could find out that the photonic power density (power/area) reduces when the size of the device reduces. The GaN-based micro-LEDs can sustain their output power density at the same current density among various sizes of devices before the roll-over due to high current self-heating or leakage . From the comparison between devices with different coatings and different current levels, a dramatic drop in terms of integrated optical power can be observed in Fig. 3(c). Between the 50 µm and 5 µm devices, the difference can be as large as 570 times for ALD coating samples and 850 times for PECVD coating ones, much larger than the ratio of emission area (100:1), and this phenomenon could be brought up mainly due to comparatively high surface recombination in the small devices.
When compared to each other, we saw a very similar performance in terms of best device between PECVD and ALD coated devices. The only different case is at 5 µm ones, when ALD sample outperform the PECVD one. In other report , the PECVD coated device can deliver 8% EQE in 15 µm devices, which is also a good example for PECVD passivation. From the result, we believe that PECVD can provide the same passivation as ALD for medium sized microLEDs initially. Meanwhile, in the later contents of this study, including accelerated aging and ABC model fitting, devices with two different coating can be examined under extreme current stress and the surface recombination difference.
Previously, it was reported that the sidewall passivation layer can also enhance the light extraction efficiency as well , due to the intermediate index of refraction provided by the coated dielectric material. As high as 40% of increase was demonstrated by device comparison . In our cases, because we did not fabricate the device without any coating, we can evaluate the difference between 200 nm and 20 nm SiO2 conditions. From the simulation point of view, they are close in terms of Fresnel reflection and total reflection angles, thus, we believe their light extraction efficiencies for these two coatings should be close. Our L-I measurements also have similar results from these two types of devices.
3.3 Accelerated aging tests
One of the important characteristics of these devices is their reliability. With proper passivation, we expect they can extend their lifetime against continuous and elevated current injection. The ICP dry-etch can introduce damages and defects on the sidewalls. When the device size shrinks, the effect brought by sidewalls can potentially rise in proportion and lead to early degradation of the performances. Because the semiconductor devices usually last a long time (months, if not years) when they are under mild (or operating) current level, a method called accelerated aging can be used to speed up any degradation in the device . In this test, the driving current is raised up many times compared to operating level to accelerate the happening of any defects in the device. To examine the durability of devices, two different current levels were set up: 100 A/cm2, and 400 A/cm2. Both of them are far higher than the regular operational point which is around several A/cm2, and similar range of current level was used previously [26,27]. Because the power from 5 µm device is too weak for our calibrated photodetector which was placed close to the device, we did not do any 5 µm device aging. The 100 A/cm2 condition was applied to 15, 20, and 50 µm devices, and only 15 µm devices were tested under 400 A/cm2. When performing these tests, we check the power degradation and record the time when the power fell under 50% of its initial values, and it is often called half-life of the device (LT50).
Figure 4(a) shows the normalized optical intensity (to the initial value) against aging time for 15 µm devices with two different coatings under continuous operation of 400 A/cm2. After the initial drop (within 10 minutes), the devices can maintain a relatively stable operation before another sudden plunge in output power, and the LT50 hours are 4.33 for the PECVD device and 5.32 for the ALD device, respectively. This phenomenon was accompanied by junction breakdown and bring the device into short-circuit condition. All the 100 A/cm2 tests also showed this feature. Table 2 summarize all the failed hour (LT50) in this study. There is no obvious trend for size-dependent failure time. But ALD devices are in general comparable or last longer than their PECVD counterparts.
In Fig. 4(b), a generic power vs. time degradation plot of a LED is shown, and we illustrate different stages of the device during the accelerated aging test. The initial drop (named “pre-aging”) during the accelerated aging tests was also reported previously [28,29]. In some cases, it was attributed to the early failure of the package components, like silicone yellowing or solder degradation [29,30]. However, it is also possibly due to an increase in the resistive leak current or the leak current path outside of active region . In the latter case, the saturation (or termination) of these pathways can stabilize the output power of the light emitting devices such as lasers or LEDs. This situation was discovered early in the laser/LED reliability study, and a specific step often called pre-aging, purge, or screening procedure was developed to enhance the chance to obtain good light emitting devices , and if the device output can stabilize after the initial drop, it is a promising sign from the material standpoint, and extended lifetime can be expected.
The second stage (named “stable” in the figure) in our device aging can be also known as a long-term “wear-out” stage , where the device enjoyed a stable and slow decay in output power. In normal case, if nothing else happens, it will degrade until reaching 50% of its initial value and the LT50, which is the lifetime of the device with 50% of its initial power or better, can be determined. However, in our case, a sudden death of the device can happen due to the top contact failure, as we will see in the next paragraph. This sudden decay can happen randomly and will need to be improved in the process in the future. Since it is random, we cannot see too much size-dependence in Table 2 when judged by the LT50 or the failure time. However, if we extract the magnitude of the slope of degradation in the “stable” region, which is still with some uncertainty due to the early stage of the wear-out phase in our study, we can see an uptick slope in 15 µm devices, while 50 µm devices have a flatter and more gradual decaying curve (i.e. the slope is smaller). The ratio among the average slope values of the 15, 20, and 50 µm devices under 100 A/cm2 condition is 3.93: 2: 1, meaning that 50 µm device can have slower decay in power and thus a longer lifetime if normal wear-out dominates the aging process.
In addition to the device’s optical power, we also record a spectral evolution for a 15 µm device at 100 and 400 A/cm2 condition. Fig. 5(a) shows the development of the spectra of a 15 µm ALD coated device at 400 A/cm2 aging, and the slight red-shifted peak and broadened spectral width can be detected. From the figure, we know that the peak was shifted 3.27 nm, and the linewidth increases from 7.63 to 11.30 nm before the device plunged into short-circuit status. In terms of spectral profile change, while the variation in general is small, the PECVD devices have less amount of peak red-shift and linewidth broadening compared to the ALD ones in both current settings. We believe the main cause behind this might be due to the local material quality variation. The parameters were extracted from measured spectra and summarized in Figs. 5(b) and (c), and the data was obtained by comparing the initial spectrum.
The forward bias voltage of the devices are also monitored during the aging test. As shown in Fig. 6(a), the deviation of the diode voltage was graphed against aging time. An abrupt increase for both ALD and PECVD devices marked the initial “pre-aging” period and the start of “stable” period in the aging process. Then the voltage increase will take a more gradual slope until the breakdown point where the device plunged into short-circuit condition quickly. Although the PECVD and ALD devices share the same slope of voltage increase, the failure point is different, and ALD can sustain a higher forward voltage increase compared to the PECVD one. This kind of forward voltage evolution during the burn-in process was also reported before in the InP-based LEDs .
The reverse biased electrical characteristics of 50 µm devices was also measured and both the series resistance and reverse leakage current increase as the device runs towards its final breakdown. As shown in Fig. 6(b), the reverse leakage current at −4 V of the 50 µm ALD device was increased 12% during the stable period, and almost doubled (1.85 times) when the time approached “sudden degradation” point. On the other hand, the reverse bias current of a 50 µm PECVD device can become 2.97 times in the stable period, and increase dramatically, often more than 100 times at the failure point. In the meantime, the forward bias current at 2.5 V dropped to 17.6% of the initial value for ALD devices, indicating increased resistance of the device and similar degree of reduction (14.1%) can be also observed in the PECVD devices. After the breakdown, the device becomes short-circuited and the observation from microscope indicates serious burnt surface of the device happened.
After the devices failed, they were examined carefully, under optical microscope and also later in FIB. The before and after accelerated aging test pictures of the device demonstrated burnt marks on the surface of mesa. Using FIB technology, we can then cut down the device and reveal its inner region. In Fig. 7, we demonstrate the detailed SEM pictures of the devices after accelerated aging tests. In Fig. 7(a), it is the original SEM picture of a failed ALD device. We enhanced the contrast of this SEM and put it in the Fig. 7(b). The dark lines elongate all the way from the top and penetrate through the active region. Voids, cracks and burnt marks can be clearly seen in the picture. These defects are found in the location near the sidewall of the device, adding the possibility that the defects on the sidewall could assist the propagation of dark lines. The deformed interfaces among layers of materials can be also seen near the top of device mesa. In Fig. 7(c), the area close to contact window opening of a failed PECVD device is shown and an illustrative diagram is on the side for clarity. The details of the surface burnt marks, voids and molten parts can be seen inside the circled area. There are two ITO layers in our devices because the first ITO layer was used for direct p-type mesa contact and the second ITO layer is for interconnect to metal pads. These voids and hillocks which were caused by high temperature generated locally and could lead to the crack in semiconductor [25,31,33]. The local hot spot could rise from the high current injection and high series resistance inherent to this device. Our SEM in Fig. 7 not only showed the disintegrated surface layers of the device due to high heat, but also demonstrated the propagation of dark lines that can go through the device. Eventually, this phenomenon originated in p-contact area can lead to the sudden failure in Fig. 4. The quickly flattened optical intensity of the device during the aging test (like in Fig. 4(a)) also tells us that the semiconductor material quality is good and stable. Therefore, to move forward, it is quite important to solve the electrical conduction issue associated with these small devices.
Combined with the SEM observation and the leakage current measurements, the failure mechanism of a microLED under the high current stress can be mainly due to the p-type contact failure at the device top. This failure can lead to defect propagation (in the form of dark lines) through the device junction, and this propagation is prone to happen at the edge of the mesa where the surface defects and traps are located. The increase in reverse leakage can be viewed as the prelude of the massive failure. The traps on the sidewall can certainly contribute to this situation. As the defects slowly penetrate through the device, the SEM also showed us the possible scenario of the dark line going deep into semiconductor via the sidewall region. Thus devices from the same wafer but with different sidewall coatings can have different lifetime in this accelerated aging test, and ALD coating could be valuable to passivate the sidewall and make the device stronger against such dark line propagation.
3.4 ABC model for devices
Although our GaAs substrate was not removed, the partial quantum efficiency of the devices can be evaluated from the top emission measurement. The external quantum efficiency model can be approximated as [17,18]:34]. The carrier concentration can be obtained via the current density formula : 18,35]. By the measured EQE numbers, the proper numerical values of A, B, and C can be fitted. In order to demonstrate their relative magnitude in terms of quantum efficiencies, the highest efficiency obtained in a 50 µm device was used for normalization factor. The normalized partial quantum efficiency vs. current density is plotted in Fig. 8(a) and (b). Also can be seen in Fig. 8(b), the 5 µm device is very difficult to be fitted due to low optical output.
The measured quantum efficiency can be as high as 9.26% in the 50 µm device, and decrease in efficiency can be found as the device size shrinks down to 15 µm. To extract these A, B and C parameters, Equations (1) and (2) are applied to obtain the carrier concentration n, and the initial values of A, B, C are from the published results. Then we use the following formula to minimize the distances between the measured efficiency and the calculated number :12]. Meanwhile, the extracted bimolecular coefficients, B, shows less variations and their values are between 1.35 to 3.40 ×10−10 cm3/s, which are close to the published values [37–39]. Comparatively speaking, there is a 5-time reduction of peak quantum efficiency between the 50 µm and 15 µm devices. Another eight times of reduction in peak efficiency can be observed between 15 µm and 5 µm devices.
Using these extracted numbers, we believe the ratio of SRH coefficient (A) between the ALD and PECVD devices can be a good parameter to watch:
In 50 µm devices, this number is 1.09, and it becomes 1.14 when the side length reduces to 15 µm. The larger values of the SRH coefficient in PECVD devices indicate that they have stronger trap-related recombination during operation than the ALD ones do. Also a rising ratio for smaller devices means this difference becomes more serious when device size shrinks.
In conclusion, we demonstrated AlGaInP micro-LED devices with good performance via sidewall passivation. Both PECVD and ALD coating were used, and the leakage current level for ALD samples are 7.8 times lower than those of PECVD ones. High current levels of 100 and 400 A/cm2 were applied to evaluate their reliabilities, and detailed photonic and electrical characterization was carried out. The standard ABC model was applied and the radiative bimolecular coefficients are close to what was reported before. Although the initial L-I measurements showed the larger PECVD devices had comparable optical output to the ALD counterparts, the surface recombination and the accelerated aging test can really distinguish them. From the accelerated aging tests, the 15 µm ALD device can outperform the PECVD one by longer lifetime (7.4% at 100 A/cm2, 18.6% at 400 A/cm2). The reverse leakage current increase in ALD device is also dramatically lower than the PECVD one (1.12 times vs. 2.97 times in stable period). Further FIB and SEM measurements showed top p-type contact deterioration during the aging and dark lines propagated near the edge area and through the device junction. The ABC model also indicates that the PECVD device can have stronger SRH recombination component by 9% in 50 µm devices and by 14% in 15 µm devices. The increased SRH recombination was observed in the small devices, indicating the necessity of sidewall protection. The minimal wavelength shift and narrow linewidth of this AlGaInP red device can be very attractive for the application of the future micro-display technology.
Ministry of Science and Technology, Taiwan (MOST 107-2221-E-009-114-MY3).
The authors declare no conflicts of interest.
1. Y. Huang, E.-L. Hsiang, M.-Y. Deng, and S.-T. Wu, “Mini-LED, Micro-LED and OLED displays: present status and future perspectives,” Light: Sci. Appl. 9(1), 105 (2020). [CrossRef]
2. K. Ding, V. Avrutin, N. Izyumskaya, Ü. Özgür, and H. Morkoç, “Micro-LEDs, a Manufacturability Perspective,” Appl. Sci. 9(6), 1206 (2019). [CrossRef]
3. J. Y. Lin and H. X. Jiang, “Development of microLED,” Appl. Phys. Lett. 116(10), 100502 (2020). [CrossRef]
4. X. Zhou, P. Tian, C.-W. Sher, J. Wu, H. Liu, R. Liu, and H. -C. Kuo, “Growth, transfer printing and colour conversion techniques towards full-colour micro-LED display,” Prog. Quantum Electron. 71, 100263 (2020). [CrossRef]
5. Y. Yin, Z. Hu, M. U. Ali, M. Duan, L. Gao, M. Liu, W. Peng, J. Geng, S. Pan, Y. Wu, J. Hou, J. Fan, D. Li, X. Zhang, and H. Meng, “Full-Color Micro-LED Display with CsPbBr3 Perovskite and CdSe Quantum Dots as Color Conversion Layers,” Adv. Mater. Technol. 5, 2000251 (2020). [CrossRef]
6. 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. Hirano, and H. Kawanishi, “25-5: Late-News Paper: 1,053 ppi Full-Color “Silicon Display” based on Micro-LED Technology,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 50, 353–355 (2019). [CrossRef]
7. F. Templier, “GaN-based emissive microdisplays: A very promising technology for compact, ultra-high brightness display systems,” Jnl Soc Info Display 24(11), 669–675 (2016). [CrossRef]
8. S.-M. Yang, P.-H. Wang, C.-H. Chao, C.-W. Chu, Y.-T. Yeh, Y.-S. Chen, F.-P. Chang, Y.-H. Fang, C.-C. Lin, and C.-I. Wu, “Angular color variation in micron-scale light-emitting diode arrays,” Opt. Express 27(16), A1308–A1323 (2019). [CrossRef]
9. Y.-M. Huang, K. J. Singh, A.-C. Liu, C.-C. Lin, Z. Chen, K. Wang, Y. Lin, Z. Liu, T. Wu, and H.-C. Kuo, “Advances in Quantum-Dot-Based Displays,” Nanomaterials 10(7), 1327 (2020). [CrossRef]
10. D. Hwang, A. Mughal, C. D. Pynn, S. Nakamura, and S. P. DenBaars, “Sustained high external quantum efficiency in ultrasmall blue III–nitride micro-LEDs,” Appl. Phys. Express 10(3), 032101 (2017). [CrossRef]
11. Y. Tsai, Y. Huang, S. Yang, W. Kuo, Y. Fang, S. Hsu, C. Huang, H. Shih, S. Wang, H. Huang, and C. Lin, “High Performance Ultraviolet Micro-LED Arrays for Fine-Pitch Micro Displays,” in 2019 IEEE Photonics Conference (IPC), 2019), 1–2.
12. P. Royo, R. P. Stanley, M. Ilegems, K. Streubel, and K. H. Gulden, “Experimental determination of the internal quantum efficiency of AlGaInP microcavity light-emitting diodes,” J. Appl. Phys. 91(5), 2563–2568 (2002). [CrossRef]
13. O. Demichel, M. Heiss, J. Bleuse, H. Mariette, and A. Fontcuberta i Morral, “Impact of surfaces on the optical properties of GaAs nanowires,” Appl. Phys. Lett. 97(20), 201907 (2010). [CrossRef]
14. K. A. Bulashevich and S. Y. Karpov, “Impact of surface recombination on efficiency of III-nitride light-emitting diodes,” Phys. Status Solidi RRL 10(6), 480–484 (2016). [CrossRef]
15. 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, “Improved performance of AlGaInP red micro-light-emitting diodes with sidewall treatments,” Opt. Express 28(4), 5787–5793 (2020). [CrossRef]
16. M.-C. Tseng, C.-L. Chen, N.-K. Lai, S.-I. Chen, T.-C. Hsu, Y.-R. Peng, and R.-H. Horng, “P-side-up thin-film AlGaInP-based light emitting diodes with direct ohmic contact of an ITO layer with a GaP window layer,” Opt. Express 22(S7), A1862–A1867 (2014). [CrossRef]
17. 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, “Light output performance of red AlGaInP-based light emitting diodes with different chip geometries and structures,” Opt. Express 26(9), 11194–11200 (2018). [CrossRef]
18. A. Daami and F. Olivier, InGaN/GaN µLED SPICE modelling with size-dependent ABC model integration, SPIE OPTO (SPIE, 2019), Vol. 10912.
19. J.-H. Lan and J. Kanicki, “ITO surface ball formation induced by atomic hydrogen in PECVD and HW-CVD tools,” Thin Solid Films 304(1-2), 123–129 (1997). [CrossRef]
20. S. Major, S. Kumar, M. Bhatnagar, and K. L. Chopra, “Effect of hydrogen plasma treatment on transparent conducting oxides,” Appl. Phys. Lett. 49(7), 394–396 (1986). [CrossRef]
21. H. -C. Lin and I. -M. Lu, “Influence of the deposition of PECVD hydrogenated silicon nitride on the transparency of an indium tin oxide underlayer,” in Asia Pacific Symposium on Optoelectronics ‘98 (SPIE, 1998), Vol. 3421.
22. M. S. Wong, D. Hwang, A. I. Alhassan, C. Lee, R. Ley, S. Nakamura, and S. P. DenBaars, “High efficiency of III-nitride micro-light-emitting diodes by sidewall passivation using atomic layer deposition,” Opt. Express 26(16), 21324–21331 (2018). [CrossRef]
23. M. Asad, Q. Li, M. Sachdev, and W. S. Wong, “Size-dependent optoelectrical properties of 365 nm ultraviolet light-emitting diodes,” Nanotechnology 30(50), 504001 (2019). [CrossRef]
24. Z. Gong, S. Jin, Y. Chen, J. McKendry, D. Massoubre, I. M. Watson, E. Gu, and M. D. Dawson, “Size-dependent light output, spectral shift, and self-heating of 400 nm InGaN light-emitting diodes,” J. Appl. Phys. 107(1), 013103 (2010). [CrossRef]
25. M. Fukuda, O. Fujita, and S. Uehara, “Homogeneous degradation of surface emitting type InGaAsP/InP light emitting diodes,” J. Lightwave Technol. 6(12), 1808–1814 (1988). [CrossRef]
26. W. Guo, N. Chen, H. Lu, C. Su, Y. Lin, G. Chen, Y. Lu, L. Zheng, Z. Peng, H. Kuo, C. Lin, T. Wu, and Z. Chen, “The Impact of Luminous Properties of Red, Green, and Blue Mini-LEDs on the Color Gamut,” IEEE Trans. Electron Devices 66(5), 2263–2268 (2019). [CrossRef]
27. P. N. Grillot, M. R. Krames, H. Zhao, and S. H. Teoh, “Sixty Thousand Hour Light Output Reliability of AlGaInP Light Emitting Diodes,” IEEE Trans. Device Mater. Relib. 6(4), 564–574 (2006). [CrossRef]
28. K. Streubel, N. Linder, R. Wirth, and A. Jaeger, “High brightness AlGaInP light-emitting diodes,” IEEE J. Select. Topics Quantum Electron. 8(2), 321–332 (2002). [CrossRef]
29. J. Xiao, Z. Guo, Y. Xiao, Y. Gao, L. Zhu, Y. Lin, Y. Lu, and Z. Chen, “Multichannel Online Lifetime Accelerating and Testing System for Power Light-Emitting Diodes,” IEEE Photonics J. 9(3), 1–11 (2017). [CrossRef]
30. M. Fukuda, Reliability and Degradation of Semiconductor Lasers and LEDs, The Artech House Optoelectroncis Library (Artech House, Inc., Norwood, MA, USA, 1991).
31. M. Fukuda, “Laser and LED reliability update,” J. Lightwave Technol. 6(10), 1488–1495 (1988). [CrossRef]
32. F. R. Nash, W. J. Sundburg, R. L. Hartman, J. R. Pawlik, D. A. Ackerman, N. K. Dutta, and R. W. Dixon, “Implementation of the proposed reliability assurance strategy for an InGaAsp/InP, planar mesa, buried heterostructure laser operating at 1.3 µm for use in a submarine cable,” AT&T Technical Journal 64, 809–860 (1985). [CrossRef]
33. M.-H. Chang, D. Das, P. V. Varde, and M. Pecht, “Light emitting diodes reliability review,” Microelectron. Reliab. 52(5), 762–782 (2012). [CrossRef]
34. L. A. Coldren and S. W. Corzine, Diode Lasers and Photonic Integrated Circuits, Wiley Series in Microwave and Optical Engineering (John Wiley & Sons, Inc., New York, 1995).
35. C. Tian, W. Wang, J. Liang, Z. Liang, Y. Qin, and J. Lv, “Theoretical and experimental analysis of AlGaInP micro-LED array with square-circle anode,” AIP Adv. 5(4), 041309 (2015). [CrossRef]
36. Q. Shan, D. S. Meyaard, Q. Dai, J. Cho, E. F. Schubert, J. K. Son, and C. Sone, “Transport-mechanism analysis of the reverse leakage current in GaInN light-emitting diodes,” Appl. Phys. Lett. 99(25), 253506 (2011). [CrossRef]
37. N. C. Chen, C. M. Lin, C. Shen, W. C. Lien, and T. Y. Lin, “Redshift of edge emission from AlGaInP light-emitting diodes and correlation with electron-hole recombination lifetime,” Opt. Express 16(25), 20759–20773 (2008). [CrossRef]
38. M. Guina, J. Dekker, A. Tukiainen, S. Orsila, M. Saarinen, M. Dumitrescu, P. Sipilä, P. Savolainen, and M. Pessa, “Influence of deep level impurities on modulation response of InGaP light emitting diodes,” J. Appl. Phys. 89(2), 1151–1155 (2001). [CrossRef]
39. Ł. Piskorski, R. P. Sarzała, and W. Nakwaski, “Self-consistent model of 650 nm GaInP/AlGaInP quantum-well vertical-cavity surface-emitting diode lasers,” Semicond. Sci. Technol. 22(6), 593–600 (2007). [CrossRef]