Open Access
Add to BibSonomy
Abstract
A simulation scheme was developed to explore the light distribution of full-color micron-scale light-emitting diode (LED) arrays. The influences of substrate thickness, patterning, and cutting angle of the substrate on several important features, such as light field pattern, light extraction efficiency, and color variation, were evaluated numerically. An experiment was conducted; the results were consistent with simulation results for a 225 × 125 µm2 miniLED and those for an 80 × 80 µm2 microLED. Based on the simulation results, the light extraction efficiency of LED devices with a substrate increases by 67.75% over the extraction efficiency of those without a substrate. The light extraction efficiency of LED devices with a substrate increases by 113.55% when an additional patterned design is used on green and blue chips. The calculated large angle Δu′v′ can be as low as 0.015 for miniLED devices.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Micro- or miniLEDs whose sizes range from several microns to several hundred microns have been considered as the technology for next generation displays due to their many benefits, such as wide color coverage, high brightness, good reliability, small footprint, and significantly high power efficiency [1–5]. These LEDs have better brightness and power efficiency than liquid crystal displays (LCDs) and organic LEDs and thus their deployment in brightness-sensitive or power-sensitive environments such as wearable devices, cell phones, outdoor displays, and augmented reality (AR) and visual reality applications is highly anticipated [6,7].
Currently, several methods have been proposed for large-scale full-color micron-scale LED displays. In addition to the direct bonding of micron-scale red, green, and blue (RGB) LED chips on the same substrate, methods such as color conversion and optical lens assembly are potential candidates for the realization of full-color micro-displays [8]. The optical lens assembly uses specialized optics such as trichroic prisms to combine individual single-color images into a full-color image. The color-conversion method uses phosphors or colloidal quantum dots (CQDs) to absorb photons with high energy (as at wavelengths in the ultraviolet or blue regime) and then to re-emit photons with a lower energy (as at wavelengths in the green or red regime). Different problems are observed when different methods are adapted for use in full-color displays. Although the optical lens assembly method does not require the three colors to be on the same panel, traditional optical components are necessary and the footprint cannot be reduced easily. While mixing photons from different single-color panels, the loss of photons is inevitable. Thus, a certain efficiency reduction can be expected. Similar reduction in efficiency can also occur in the color-conversion method. The quantum yields from the phosphors or CQDs in the absorption and re-emission process are usually not 100%, and the loss of energy due to Stoke shifts in photon energy is also inevitable. Other problems related to the color-conversion materials, such as wide linewidth of phosphors [9] and long-term reliability issue of CQDs [10–13], can undermine their performance and raise their eventual costs. Moreover, the progress in the semiconductor technology and epitaxial growth in the past decade has not only pushed the fabrication accuracy into the nanometer scale but also provided a wide variety of manufacturing capabilities to facilitate large-scale micro-assembly. The direct-bonding method relies considerably on the semiconductor and micro-assembly technologies. The color-purity and high illumination intensity of the direct-bonded chips can be crucial for fabricating high-quality displays. The high efficiency and high reliability of LED chips have also been examined in the past two decades with large success and high market penetration. Thus, the use of direct-bonded micron-scale LED array is particularly suitable for applications such as outdoor signage, head-up displays in cars, long-lasting outdoor wearable devices and AR goggles.
Previous studies on micron-scale LEDs have been focused on the electrical and photonic properties of a single device [14–20]. Moreover, investigations on the photometric properties of an array of such devices has just been initiated [21–23]. Conversely, the angular color-shift problem pertaining to LCDs has been addressed properly, and these problems are seldom mentioned in current micro/mini LED studies [24,25].
We demonstrated a full-color microLED-based display with good characteristics last year [26]. The quality of different types of colors from the individual pixels and the overall performance of an array have been preliminarily evaluated in our recent study [27]. In this study, the photometric properties of a micron scale LED full-color display are thoroughly investigated. A simulation scheme and optical model are developed for these devices by using a commercially available software. An experiment was also conducted to compare the results of the experiment with the simulation results. Moreover, many factors that are not very important when LEDs are used for lighting or when the length and breadth of LEDs are larger than their thickness become quite critical for uniform color mixing. These factors are investigated in the following sections, and we believe that the identification of these influences is very important for the realization of a high-quality full-color display by using micron scale LEDs.
2. Device fabrication and measurement
Micron scale LED chips are different from the traditional LED chips in terms of their sizes. The most distinguished feature is their miniature size when compared with the size of traditional ones. Two categories of devices are defined—miniLEDs and microLEDs. Usually the devices with a size of more than 100 µm can be defined as miniLEDs, whereas the smaller devices are known as microLEDs. From our perspective, the difference between micro- and miniLED is the existence of a substrate at the end of manufacturing processes. The device with a substrate (usually sapphire) is known as miniLED, and the device with no substrate is known as a microLED. These two definitions usually coincide with each other because the capability of the current process to remove the substrate from the microchip is limited to large devices. In our simulation model, we investigated both types of LEDs in terms of the substrate thickness, which can greatly affect the optical results.
The display sample contains an 80 × 80 pixel2 array on a 6 × 6 cm2 PCB. Each pixel contains RGB chips with dimension of 225 × 125 µm2 or 80 × 80 µm2 , as shown in Fig. 1(a). The manufacturing process of LED displays can be divided into two parts—LED components manufacturing and PCB preparation. The LED chips with sapphire substrate can be bought from vendor companies and shall follow the standard LED process procedures. As for chips without a substrate, the LED wafers were grown using the metal organic chemical vapor deposition system (MOCVD). The light generation layer contains InGaN quantum wells for blue and green colors and AlGaInP quantum wells for the red color. The epitaxial wafer can then be processed in regular cleanroom. The individual device was patterned by standard photolithography and isolated by ICP etch with Cl2/BCl3/Ar mixing gases. The contact layer is ITO formed by E-beam coater. The silicon nitride layer was deposited by PECVD to use for further protection of the sidewall. Finally, a metal layer (Ti/Ni/Au/Sn) was formed by e-beam coater for bonding purpose. After the devices were fabricated, laser dicing was performed using a pulsed solid state laser for chips with a substrate. For those without a substrate, the wafer was prepared for mass-transfer procedure which is similar to the standard pick-and-place method. Simultaneously, PCBs were prepared by conducting stencil printing on a diameter of 50–60 µm with solder paste, and solder paste inspection was conducted to check the amount of solder paste on each pixel. Then, RGB LED chips were periodically assembled on the PCB by using the regular hot-press method with a temperature between 200°C to 300°C. It is recommended to perform a standard reflow process under a nitrogen atmosphere to form a good connection between the LED die and the PCB backplane. Finally, the PCB package was finished using the injection molding process to protect the LED die. Figure 1(b) displays the full-color panel fabricated using such LED arrays, and Fig. 1(c) illustrates the display when it is switched on.
Fig. 1. (a) SEM images of a display comprising miniLEDs (225 × 125 µm2). (b) Full-color miniLED display with 80 × 80 pixel2 arrays on a 6 × 6 cm2 PCB board. (c) Display comprising miniLEDs when switched on.
Figure 2 displays the measurement setup that is in line with the SID Information Display Measurement Standard (IDMS). The miniLED display was fixed on a multiaxis positioning stage and could be rotated in the horizontal direction. In the measurement, a fixture was used to clamp the PCB board. The height of this fixture is approximately 1 cm, whereas the center of the display is 3 cm away from the edge. Our calculations indicate that a larger error could possibly exist at approximately 18.5° above the plane, which corresponds to 71.5° from the normal direction. Thus, we inferred that a measurement value larger than 70° is susceptible to this interference from the fixture. Spectroradiometer spectrascan PR670 with a color accuracy of ± 0.0015 was used to measure the photometric and colorimetric performance. Figure 3 shows the L–I–V characteristics of LED chips when operated from 100 mA to 1 A.
Figure 4 displays the measured color gamut and optical spectra from our full-color panel. The coverage is 84.3% when compared to the Rec. 2020 standard. From the measured spectra, the linewidth of generic RGB chips ranges from 18 to 30 nm, which is much less than a lot of color conversion materials. The narrow linewidth spectra of RGB chips substantiate this performance, and the narrow linewidth is one of the great strengths of this direct-bonding method.
Fig. 4. (a) The color gamut coverage of a direct-bonded LED display compared to the Rec. 2020 specification. The CIE plot was prepared by ColorCalculator software of OSRAM Sylvania, Inc.. (b) The bonded LED generated three narrow-linewidth color emission.
3. Modeling and color variation calculation
3.1 Simulation model
A schematic of an LED array panel is shown in Fig. 5, and the side view is shown in Fig. 6. Two types of structures are simulated: structures with and without a substrate. In the structure without a substrate, as shown in Figs. 5(c) and 6(b), a layer of insulation is coated on top of the chips for protection. The simulation structure contains a 5 × 5 pixel2 array of RGB LEDs bonding on a printed circuit board (PCB). The proposed simulation scales are 225 × 125 µm2 (9 × 5 mils2) for miniLEDs and 80 × 80 µm2 for microLEDs. The refractive indices of all layers are listed in Table 1. The light source or the emitting layer was placed within a multiple quantum well region, and the omnidirectional emission was assumed. The photon emitted from the quantum well region can be assumed to be uniform. However, due to the diffraction inside the semiconductor chip, a planar light source in the semiconductor can have the following light intensity I based on Snell’s law and photon energy conservation [28]:
where P is the source; r is the distance from the source; nair and ns are the refractive indices of air and semiconductor, respectively; and θ is the angle with respect to the surface normal direction. The light pattern in the air then becomes Lambertian [28]. In the simulation, a detector was set 1 m away from the panel as the major data collection point. By simulating a set of RGB micron scale LEDs, the light field pattern of the RGB individual profile was determined. In this study, the thickness and cutting angles of the substrates are treated as important variables that could affect the outcome of light field distribution. Other parameters such as illumination efficiency and angular color variations will also be discussed once the model is built. The GaN-based chips usually come with patterned sapphire substrates (PSSs); however, the ordinary red chips do not contain such a feature. To adapt to this situation, a simulated PSS interface was designed, as shown in Fig. 6(a). The purpose of a PSS is usually to improve the photon extraction percentage [29,30]. In Fig. 6, the cross-sectional views of LED with and without a substrate are shown. The corresponding SEM of the chips in one full pixel can be seen in Figs. 6(c) and 6(d), with the same viewing angle in Figs. 5 (b) and 6(c).Fig. 5. Schematic of the simulation model. (a) Simulated microLED display with 5 × 5 pixel2 arrays. (b) Zoom-in view of a single pixel contains RGB LEDs with a substrate and (c) RGB LED chips with an insulation layer coating with no substrate.
Fig. 6. Schematic cross-section diagrams of individual LED chip on the substrate (a) a chip with substrate (b) a chip without substrate. (c) The close-up SEM of the chips with substrate and (d) the ones without substrate. The solder mask can be seen in both pictures as the isolation wall-like structure.
Table 1. Material parameters of the simulation
The simulation was conducted using ASAP software. The calculation was based on the ray-tracing technology with the Monte Carlo method. For each run, more than 107 rays in each color were calculated in random directions to achieve full color mixing.
3.2 Color variation calculation
The color variation calculation follows the Commission International de l’Eclairage (CIE) standard and IDMS. After the ray-tracing calculation, the RGB light distribution could be obtained in different view angles. Then, the tristimulus X, Y, and Z of the color stimulus S(λ) that represents the luminance of the colors was calculated using the following equations, where $\bar{x}(\lambda ),\;\ \bar{y}(\lambda )$, and $\bar{z}(\lambda )$ are the color-matching functions [31]:
The chromaticity coordinates x, y, and z could be obtained by calculating as follows [31]: The results from (5) to (7) can be marked in the CIE 1931 color space or can also be transferred to CIE 1976. The relation between the chromaticity coordinates (u′, v′) of CIE 1976 and (x, y) is expressed as follows [31]: Then, the color variation (or chromaticity difference) Δu′v′ could be calculated [31]:4. Results and discussion
4.1 Substrate effect
Once the optical structures are developed and the parameters listed in Table 1 are input, the calculation of the light field distribution can be executed. Figure 7 presents a simulation using the 80 × 80 µm2 and 225 × 125 µm2 LED chips without a substrate and with a 90-µm sapphire substrate. Although the case involving no substrate shows a flatter angular distribution than the case involving a 90 µm substrate, it is noticeable that the differences among RGB colors are significant in the calculations. Moreover, the full width at half maximum of the highest peak luminance intensity can be as wide as 130° for the 225 × 125 µm2 chips and 135° for the 80 × 80 µm2 chips.
Fig. 7. (a) Comparison between the light distribution of the 80 × 80 µm2 chips with and without substrate and (b) for the 225 × 125 µm2 chips for the same conditions.
The pronounced twin-peak distribution can be observed in Fig. 7. From previous studies, the side-wall emission from the substrate can contribute greatly in this situation. The areal size of the proposed device is comparable to the added sapphire substrate thickness (90 µm) such that the emitted photons can come out from the side wall and cause a large-angle light distribution [25]. The case with no substrate can be treated more as a planar light source that leads to a Lambertian-like distribution.
To better understand the outcome of the simulation, a separate calculation was conducted for testing the wavelength-dependent extraction. In the Fig. 8(a), two plots are displayed for cases with 100 and 500 beams that were generated in the program. Surface reflection and diffraction can be seen in the plots. The reflection plane above the devices is the boundary between the molding layer and air. In the real calculation, more than one million beams are required to provide credible distribution. In such a case, the plot will be very crowded and viewing any information would be difficult.
Fig. 8. Light tracing analysis for LED arrays. The figures on the left pertain to a device without a substrate and those on the right pertain to a device with a substrate: (a) & (b) random direction (c) & (d) light direction of 16°, and (e) & (f) light direction of 18°
Once standard ray tracing can be conducted, a more specific angle of light emission can be set up as well. This simulation was conducted to explain the differences between a device with a substrate and a device without a substrate and the results are shown in in Figs. 8(c) - (f). The total reflection angle of the red color is 17.8° based on the Snell’s law. Thus, for the 16° case, both devices with and without substrate can have many light rays penetrate through the optical layers and reach the air. However, for the 18° case, a larger difference can be seen between the cases with and without a substrate due to the additional diffractions that are introduced by the 90-µm sapphire substrate. In the case without a substrate (closer to the microLED condition), photons have much less chance to be diffracted at different angles because of a short light pathway in the chip before they can be re-absorbed by the light emitter again. This leads to a more uniform directionality, and thus, photons have a lower probability of getting extracted when the original incident angle is larger than the total internal reflection angle. In the device with a 90 µm substrate (closer to the miniLED condition), the additional non-absorbing sapphire provides a longer optical pathway for photons, and slightly higher extraction percentages can be expected. A similar claim was made in a recent study [25], where the side-wall emission from a micron-scale LED chip was considered and the twin-peak feature can also be deducted. The other important factor is the re-absorption of the emitted photons [25]. The multiple reflection of the photons at the air–semiconductor interface enhances this effect significantly and thus changes the output light distribution.
4.2 Cutting angle of substrates
The sapphire cutting angle also serves an important role in shaping the lighting pattern. The cutting angle concern occurs due to an imperfect dicing process during chip manufacturing. Both traditional and laser cutting methods can produce such side-wall inclinations. Figure 9 shows the SEM image of cross-sections of side-wall with different cutting angles. The cutting angles can range from almost zero to 8 degrees from the surface normal direction. This situation is less important when the area of the chip is large compared to its own thickness like a micro-LED one. However, a mini-LED with a 90µm substrate can suffer from these non-ideal conditions because of significant photon output from the side-wall surfaces.
Fig. 9. The SEM pictures of the mini LED cross-sections with (a) a almost straight sidewall and (b) a slanted sidewall.
To explore the influence of the cutting angle, we simulated different conditions of the sapphire cutting angle. As shown in Fig. 10, the side wall of the chip could be vertical (ideally) or slant (in reality). The inward or outward slant of the two side walls can change the substrate shape from rectangle to parallelogram or trapezium. The diffraction of the emitted photons can thus be modified from the ideal case. Figure 10 shows the effects of the cutting angles of the substrates of the LED devices. The 0° sample, that is, the ideal case, shows the highest peaks at ± 30°. However, the + 10° case has a flatter distribution of light field. The substrate with a parallelogram shape has an asymmetric pattern due to the slant shape of the substrate. The color variation, as shown in Fig. 10(b), can be controlled under 0.012 for a cutting angle between +10° and −10°.
Fig. 10. (a) Comparisons between the light field pattern for a sapphire cutting angle of 10°, 0°, and −10° and a parallel cutting angle of 7°, as shown in the inset images. (b) Comparison between the color variations for the cutting angles of 10°, 0°, and −10° and parallel cutting angles of 7°.
4.3 Patterned substrate
Another common situation that we need to consider is the PSS in the LED chips. The PSS was first designed to improve the quality of growing devices and the light extraction of the devices. Currently, GaN-based blue and green LED designs have become quite popular. To accommodate this feature, we added PSS in our model for comparison. The pattern is designed as repeated circular spots with a diameter of 2.5 µm and pitch of 3 µm, as shown in the inset image in Fig. 11. In the figure, the solid line indicates the chips with a patterned substrate, whereas the dashed line indicates the chips without a patterned substrate. The simulation shows that the PSS actually modifies the light field pattern of the red chip more than it does for the green and blue chips. The higher change in the red color angular distribution can predict a higher difference in the color variation when we put together all different substrates.
Fig. 11. Comparison between RGB light fields of miniLED devices with a 90-µm-thick patterned substrate: solid line (with a patterned substrate) and dashed line (without a patterned substrate). The insert figure schematically shows the designed pattern on a substrate.
Figure 12 presents that the color variation is worsened when a patterned substrate is used. Based on our calculation, Δu′v′ is approximately 0.01 over the entire 180° range for the flat case. However, once we added a PSS structure in the blue and green chips, which is the case in our samples, the maximum deviation increases to 0.015, which is 1.5 times of the value attained without using a PSS structure. If the red chips also have a PSS, which is usually rare, the maximum deviation can be as high as 0.02 in the worst scenario.
Fig. 12. Comparison between the color variation in 225 × 125 µm2 LED devices with a 90-µm-thick patterned substrate: yellow line (with no PSS), cyan line (with a PSS on green and blue chips), and blue line (with a PSS on RGB chips).
4.4 Substrate thickness dependence
The light extraction efficiency is another important parameter that can be obtained from our simulation. Extraction efficiency is defined as the ratio between the photons emitted from the active layer and the photons detected in the air. Figure 13 shows the relation between the light extraction efficiency and substrate thickness. Under a substrate thickness of 20 µm, the light extraction efficiency increases with the substrate thickness. However, the efficiency remains constant when the thickness is larger than 20 µm, and the maximum light extraction efficiency is approximately 13.37% for a substrate thickness of 90 µm. When compared to the case that no substrate is used, the extraction efficiency enhancement can be as high as 67.75%. This graph shows that the benefit provided by a thicker substrate (or additional optical structure for diffraction) can be saturated after the substrate thickness reaches 20 µm. Nevertheless, it is more practical to have a substrate that has a thickness of 90 µm or higher due to the readiness of using the lapping process during the chip fabrication.
Fig. 13. Substrate thickness versus emitting efficiency. Blue line indicates the influence of substrate thickness on the emitting efficiency without using a patterned substrate. The orange line shows a comparison between the devices with patterned substrates on blue and green chips. The green line shows a comparison between patterned substrates on RGB chips.
The pattern on a substrate surface also serves an important role in improving the light extraction efficiency. The light extraction efficiency of miniLED devices with a patterned substrate placed on green and blue chips increases by 27.30% than the efficiency of devices without a patterned substrate. From another viewpoint, the efficiency increases by 113.55% than the efficiency of devices without a substrate. We also simulated the condition of pattern on RGB chips. The light extraction efficiency was 29.2% when a 90-µm-thick substrate was used. This implies that the light extraction efficiency of the devices with a 90-µm-thick substrate increases by 118.62% than the efficiency of devices without a patterned substrate and by 266.75% than the efficiency of devices without a substrate.
The substrate thickness dependence can also be seen in the color variation (Δu′v′). As shown in Fig. 14, Δu′v′ versus thickness values are calculated at various viewing angles. A stronger (or larger) color difference can be observed at large angles, whereas a very slight response is observed when we stand in front of the display, such as in the 10° or 30° case. As the substrate becomes thicker, the Δu′v′ dependence becomes weaker due to the limited escape cone on the side wall. The noticeable dip for a 15-µm-thick substrate can be probably attributed to the difference in the critical angle of total reflection (θc) (and thus the escape cone) between the red chip and GaN-based chip. The former has very narrow escape angle of 17.5°, and the latter has a value that is twice of this value for both green and blue color. A wider escape cone scales blue and green photons with an increase in the substrate thickness, whereas a narrow escape cone of red photons causes early saturation of the side-wall emission. This phenomenon can be observed from the results of different thickness. The thin substrate can generate a Lambertian-like distribution for all the three colors, but the blue and green profiles evolve into a twin-peak shape (like in Fig. 7), which is a clear indicator of side-wall emission, for the cases with thicker substrates. Moreover, the angular distribution of the red color does not change much as the thickness of the substrate increases. The resultant chromaticity difference (Δu′v′) is low for a thin substrate but increases as the thickness of the substrate increases.
Fig. 14. Color variation versus substrate thickness for viewing angles from 10° to 90°. The simulation was performed on the devices with a flat sapphire layer on a red chip and patterned sapphire layers on green and blue chips.
4.5 Evaluation of cross-talk issue
When the size of pixel is reduced, the cross-talk between the individual pixels becomes more important. We can analyze the issue in two categories: the intensity and color-mixing cross-talks. First in the intensity cross-talk, we can simply deal the gray-scaled intensity to understand the problem. Figure 15(a) shows the generic layout of our RGB pixel array with a pitch of 750µm (d1) and a separation of 30µm (d2). In Fig. 15(b), the inset shows the calculated 3D field intensity profile of one pixel. From this inset plot, the intensity cross-talk can happen due to the overflow of the light field intensity from the adjacent pixel. To better quantify this issue, the linear superposition of the fields from the two adjacent LED chips in different pixels can be a good start. This method can provide the first order estimation about the cross-talk from the chip in the neighborhood. When the pitch is too small, it is possible that the combined light field distribution can only be viewed as one peak, as shown in Fig. 15(b). In this case, we can certainly conclude that the cross-talk is too strong, and the display won’t have clear image due to the mixture of the photons from the adjacent pixels. On the other hand, if the two-peak pattern can be resolved after superposition, we believe that the cross-talk is still manageable, and thus the image formed by the pixel array can be clearly defined. Based on this idea, the distance between the centers of the pixel is used as the variable and the superimposed light field is the outcome. The criterion for resolving two peaks can be defined as the 10% of the amplitude as the depth of the dip between the two peaks. From this criterion, our device with raised substrate can have tolerable cross-talk beyond the pitch of 250µm.
Fig. 15. (a) The schematic diagram of a RGB pixel array (b) The intensity cross-talk evaluation at different horizontal distances. The inset is the 2-D optical field distribution of a pixel.
In the color-mixing related cross-talk, it is more difficult to quantify because it deals with the color-shift due to the interference from the neighborhood pixels of different colors. To simplify the problem, we consider the situation in our panel and handle the closest distance between the two color sub-pixels: the red and blue sub-pixels in the vertical direction as shown in Fig. 16(a). On the real printed circuit board, it can be as close as 30 µm. The color from the adjacent pixel can “contaminate” the original color depends on the distance of the two pixel and the spread of optical field of the chip. By calculating the overlapped part of blue photon in the red pixel area, we can determine the change of the color coordinate on the CIE plot when this color-mixing happens. As shown in Fig. 16(b), the closer the distance between red and blue chip becomes, the larger variation to the original red color will be. On the other hand, when red and blue chips are separated beyond 180µm away, the resulting color-mixing can fall into the MacAdam ellipse region, where the difference in color is indistinguishable. However, we must point out that this calculation is based on the additive color model, which could overestimate the influences sometimes. A more detailed analysis is necessary when the 2-D field distribution and the human color vision model are properly included in the calculation.
Fig. 16. (a) The concept of color-mixing between the adjacent sub-pixels due to the overlapped field distribution. (b) The calculated traces of color coordinates at different separation of the two sub-pixels (red and blue). The d value (defined in the inset) changes from 30µm (the one closest to blue) to 180µm (the one falls in the MacAdam ellipse). The CIE plot was prepared by ColorCalculator software of OSRAM Sylvania, Inc..
To avoid excess cross-talk, narrowing the spread of the optical field is a good way to investigate. To achieve this, one can raise the blocking wall (or black matrix) between the pixels, or another method is to make the optical field distribution more directional by either decorating the surface of the LED chip with focusing micro-lens structure or externally adding this function.
4.6 Measurement and comparison
Figure 17 displays the light field pattern of the measured and calculated results for mini- and microLED panels. The sizes of the chips are as follows: 225 × 125 µm2 for miniLED and 80 × 80 µm2 for microLED. The actual angular distribution of the emitted light is closer to that of the Lambertian shape than the calculated angular distribution. The deviation can be due to an imperfection in the shape of the chip. As aforementioned, the shape of the chip can modify the output light distribution considerably. With one slanted angle, the twin-peak pattern can be suppressed, as shown in Fig. 10. The re-absorption of the photons is included to increase the accuracy of the model [25]. Another possible explanation is random scattering inside the actual PCB panel, which cannot be easily modeled.
Fig. 17. Comparison between simulated light field pattern and experimental measurement (a) for miniLEDs and (b) for microLEDs
Figure 18 shows the color variation comparison between the experiment and simulation results of 225 × 125 µm2 miniLEDs with a 90-µm-thick substrate. The tendency and the variation scale are consistent for the experiment and simulation. The measured data shows larger deviation beyond a viewing angle of 60°. We believe that this situation could be caused by the incomplete capturing of the photons from the microLED panel, which can be blocked easily from the external fixtures, as mentioned in the measurement setup. In Fig. 19, a similar plot for the microLED panel is demonstrated. The measured results are a bit scattered in the plot, but they are closely in line with the calculated profile toward large viewing angles. The nonideal shape of chips and the other optical structures could pose the biggest uncertainty in scattering the photons when we try to fit our model in to the experimental results.
Fig. 18. Comparison between the experiment and simulation results of 225 × 125 µm2 LEDs with a 90-µm-thick patterned substrate on green and blue chips.
Fig. 19. Comparison between the experiment and simulation results of 80 × 80 µm2 LEDs without a substrate layer.
5. Summary
In this study, the chromaticity of full-color micron-scale-LEDs for use in displays was calculated and evaluated in terms of the color angular variation. A good coverage of color gamut and narrow linewidth of emission spectra were observed from the fabricated panel. The simulated results revealed that substrate thickness is an important issue. The optimal thickness of a substrate may affect the emission pattern and improve the angular color variation and image quality. Moreover, a patterned substrate could increase the light extraction efficiency. However, such a substrate also increases color variation due to the disparity of RGB chips. A carefully created optical design is required to make the light distribution uniform.
Although the potential commercial interest on LED without a substrate is high, both the simulation and experimental results suggest that devices with a sapphire substrate can provide better color uniformity and good light extraction efficiency. The measured angular Δu′v′ confirms our simulation for the case using a 90 µm substrate. The steps that can improve the optical quality of such devices are crucial, and thus, it is important to have a numerical model to predict the possible light field distribution. Moreover, a balance between the size, thickness, and arrangement of LED chips can be designed beforehand. The direct-bonding method can bring together the best features of the micron scale LEDs like high brightness, narrow linewidth and stability. With the advances in micro-assembly, the direct-bonding technique can be ready and mature quickly for commercialization of these devices. We believe that this study can be the first step to realize a full-color, high-definition micro-display in the near future.
Funding
Ministry of Science and Technology, Taiwan (MOST) (MOST107-2221-E-009-114-MY3); Ministry of Economic Affairs (MOEA) (107-EC-17-D-11-1554).
References
1. H. X. Jiang, S. X. Jin, J. Li, J. Shakya, and J. Y. Lin, “III-nitride blue microdisplays,” Appl. Phys. Lett. 78(9), 1303–1305 (2001). [CrossRef]
2. C.-W. Jeon, K.-S. Kim, and M. D. Dawson, “Fabrication of Two-Dimensional InGaN-Based Micro-LED Arrays,” Phys. Status Solidi A 192(2), 325–328 (2002). [CrossRef]
3. H. X. Jiang and J. Y. Lin, “Nitride micro-LEDs and beyond - a decade progress review,” Opt. Express 21(S3), A475 (2013). [CrossRef]
4. E. H. Virey and N. Baron, “Status and Prospects of microLED Displays,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 49(1), 593–596 (2018). [CrossRef]
5. T. Wu, C. W. Sher, Y. Lin, C. F. Lee, S. Liang, Y. Lu, S. W. Huang Chen, W. Guo, H. C. Kuo, and Z. Chen, “Mini-LED and Micro-LED: Promising Candidates for the Next Generation Display Technology,” Appl. Sci. 8(9), 1557 (2018). [CrossRef]
6. Z. Y. Fan, J. Y. Lin, and H. X. Jiang, “III-nitride micro-emitter arrays: development and applications,” J. Phys. D: Appl. Phys. 41(9), 094001 (2008). [CrossRef]
7. Z. Liu, W. C. Chong, K. M. Wong, and K. M. Lau, “GaN-based LED micro-displays for wearable applications,” Microelectron. Eng. 148, 98–103 (2015). [CrossRef]
8. T. Wu, C. W. Sher, Y. Lin, C. F. Lee, S. Liang, Y. Lu, S. W. Huang Chen, W. Guo, H. C. Kuo, and Z. Chen, “Mini-LED and Micro-LED: Promising Candidates for the Next Generation Display Technology,” Appl. Sci. 8(9), 1557 (2018). [CrossRef]
9. J. Silver, P. Harris, G. Fern, J. Bonar, G. Valentine, and S. Gorton, “A Novel Approach to the Manufacture of Micro-LED Colour Conversion Structures,” Proceeding Int. Disp. Work. 23, 1183–1186 (2016).
10. D. Bozyigit and V. Wood, “Challenges and solutions for high-efficiency quantum dot-based LEDs,” MRS Bull. 38(9), 731–736 (2013). [CrossRef]
11. 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, “Fabrication of a vertically-stacked passive-matrix micro-LED array structure for a dual color display,” Opt. Express 25(3), 2489 (2017). [CrossRef]
12. S. C. Hsu, L. A. Ke, H. C. Lin, T. M. Chen, H. Y. Lin, Y. Z. Chen, Y. L. Chueh, H. C. Kuo, and C. C. Lin, “Fabrication of a Highly Stable White Light-Emitting Diode with Multiple-Layer Colloidal Quantum Dots,” IEEE J. Sel. Top. Quantum Electron. 23(5), 1–9 (2017). [CrossRef]
13. S. C. Hsu, Y. H. Chen, Z. Y. Tu, H. V. Han, S. L. Lin, T. M. Chen, H. C. Kuo, and C. C. Lin, “Highly Stable and Efficient Hybrid Quantum Dot Light-Emitting Diodes,” IEEE Photonics J. 7(4), 1–7 (2015). [CrossRef]
14. Z. Gong, S. Jin, Y. Chen, J. McKendry, and D. Massoubre, “Size-dependent light output, spectral shift, and self-heating of 400 nm InGaN light-emitting diodes,” J. Appl. Phys. 107(1), 013103 (2010). [CrossRef]
15. 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]
16. S. S. Konoplev, K. A. Bulashevich, and S. Y. Karpov, “From Large-Size to Micro-LEDs: Scaling Trends Revealed by Modeling,” Phys. Status Solidi A 215(10), 1700508 (2018). [CrossRef]
17. H. K. Lee, J. S. Yu, and Y. T. Lee, “Thermal analysis and characterization of the effect of substrate thinning on the peformances of GaN-based light emitting diodes,” Phys. Status Solidi 207(6), 1497–1504 (2010). [CrossRef]
18. C. E. Lee, Y. C. Lee, H. C. Kuo, M. R. Tsai, T. C. Lu, and S. C. Wang, “High brightness GaN-based flip-chip light-emitting diodes by adopting geometric sapphire shaping structure,” Semicond. Sci. Technol. 23(2), 025015 (2008). [CrossRef]
19. 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, and J. O. Song, “Light output performance of red AlGaInP-based light emitting diodes with different chip geometries and structures,” Opt. Express 26(9), 11194 (2018). [CrossRef]
20. I. Zhmakin, “Enhancement of light extraction from light emitting diodes,” Phys. Rep. 498(4-5), 189–241 (2011). [CrossRef]
21. T. K. Kim, M. U. Cho, M. M. Lee, Y. J. Cha, S. K. Oh, B. Chatterjee, J. H. Ryou, S. Choi, and J. S. Kwak, “Improved Light Output Power of 16 × 16 Pixelated Micro-LEDs for Headlights by Enhancing the Reflectivity and Coverage of the p -Electrode,” Phys. Status Solidi A 215(10), 1700571 (2018). [CrossRef]
22. X. Ding, Y. Tang, Z. Li, J. Li, Y. Xie, and L. Lin, “Multichip LED Modules with V-Groove Surfaces for Light Extraction Efficiency Enhancements Considering Roughness Scattering,” IEEE Trans. Electron Devices 64(1), 182–188 (2017). [CrossRef]
23. Y. Huang, G. Tan, F. Gou, M. C. Li, S. L. Lee, and S. T. Wu, “Prospects and challenges of mini-LED and micro-LED displays,” J. Soc. Inf. Disp. 27(7), 387–401 (2019). [CrossRef]
24. S. S. Park, I. Sohn, E. Cho, S. Park, and E. Kim, “Color shift reduction of liquid crystal displays by controlling light distribution using a micro-lens array film,” J. Disp. Technol. 8(11), 643–649 (2012). [CrossRef]
25. F. Gou, E. L. Hsiang, G. Tan, P. T. Chou, Y. L. Li, Y. F. Lan, and S. T. Wu, “Angular color shift of micro-LED displays,” Opt. Express 27(12), A746 (2019). [CrossRef]
26. C.-C. Lin, Y. H. Fang, M. J. Kao, P. K. Huang, F. P. Chang, L. C. Yang, and C. I. Wu, “Ultra-Fine Pitch Thin-Film Micro LED Display for Indoor Applications,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 49(1), 782–785 (2018). [CrossRef]
27. S. Yang, “The substrate thickness dependence on micro LED chip arrays,” Dig. Tech. Pap. - Soc. Inf. Disp. Int. Symp. 50(1), 1724–1727 (2019). [CrossRef]
28. E. F. Schubert, Light-Emitting Diodes2nd ed. (Cambridge University Press, 2006).
29. J. Gan, S. Ramakrishnan, and F. Y. Yeoh, “A review on improvement of led light extraction efficiency through a micro repeating structure,” Rev. Adv. Mater. Sci. 42(1), 92–101 (2015).
30. C. H. Chiu, P. M. Tu, C. C. Lin, D. W. Lin, Z. Y. Li, K. L. Chuang, J. R. Change, T. C. Lu, H. W. Zan, C. Y. Chen, and H. C. Kuo, “Highly efficient and bright LEDs overgrown on GaN nanopillar substrates,” IEEE J. Sel. Top. Quantum Electron. 17(4), 971–978 (2011). [CrossRef]
31. International Committee for Display, Information Display Measurements Standard (SID, 2012), pp. 467–469.
