Angular color shift of micro-LED displays

Sidewall emission of a micro-scale light emitting diode (micro-LED) improves the light extraction efficiency, but it causes mismatched angular distributions between AlGaInPbased red micro-LED and InGaN-based blue/green counterparts due to material difference. As a result, color shift of RGB micro-LED displays may become visually noticeable. To address this issue, we first analyze the angular distributions of RGB micro-LEDs and obtain good agreement between simulation and experiment. Next, we propose a device structure with top black matrix and taper angle in micro-LEDs, which greatly suppresses the color shift while keeping a reasonably high light extraction efficiency. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement


Simulation model and experiment
In our analysis, we examine the emission patterns of RGB micro-LEDs at different viewing angles with ray-tracing software LightTools. Table 1 lists the commonly used major structure layers and their thicknesses of flip-chip RGB chips. For AlGaInP-based red micro-LED, it has metal contact layer, n-cladding AlGaInP, n-type AlInP diffusion barrier, GaInP/AlGaInP MQWs, p-type AlInP diffusion barrier, p-cladding AlGaInP, and p-GaP window layer [16]. For InGaN-based green and blue micro-LEDs, their structures consist of metal contact layer, p-type GaN, AlGaN electron block layer, InGaN/GaN MQWs and n-type GaN. Although only the refractive indices at central wavelength are included here, during simulations the wavelength dispersion of refractive index for each material [17,18] is also considered. The die sizes are all 35 × 60 μm in order to be comparable to some commercial products. For simplicity, the metal pad is set to be the same size as chip size. Please note that the layout and materials for blue and green micro-LED are similar, but they are quite different from those of red chip. Light radiation from the multi-quantum wells (MQWs) with uniform angular distribution travels through each layer and across interfaces according to Snell's law. Table 1

. Optical parameters of commonly used AlGaInP-based red micro-LED and
InGaN-based blue and green micro-LEDs adopted in simulations.  For the emission towards sidewall, the escape cone may not be completed, depending on the position of point-like sources located at the MQW layer. For example, the point source at a short distance from sidewall can get a completed escape cone. However, as the distance increases, the escape cone will become uncompleted because of the small thickness of micro-LED chip (< 4 μm). Figure 5(c) shows the side-view of light emissions from the point source with angle θ i satisfying 90°−θ c < θ i < 90°. Light within this angle range will escape from the sidewall but experience different absorption or reflection losses. For example, when the light emission from the point source reaches the top edge of the sidewall without experiencing reflection [gray line in Fig. 5(c)], we can obtain a light escape cone θ 1 = tan −1 (h 1 /x) without absorption and reflection loss. The emission ratio can be calculated using Eqs. (2) and (3). If the emission angle θ i satisfies 90°−θ c < θ i < 90°−θ 1 , although the light can still escape from sidewall, it will experience multiple reflections between top and bottom surfaces as well as absorption of MQW, which are illustrated by green and orange lines in Fig. 5(c). For the light reflected by the bottom surface n times, the sidewall emission ratio can be estimated by modifying Eqs. (2) and (3)

Red Layers
In Eq. (4), α and d represent the absorption coefficient and thickness of MQW layer, and R s is the reflectance of bottom metal pad. For simplicity, we have neglected the Fresnel loss at semiconductor/air interface and the refraction between different layers inside micro-LED. Thus, the total sidewall emission ratio from four side surfaces can be calculated by adding all the ratios together. The calculated results by MATLAB and the simulated data using raytracing in LightTools for micro-LEDs with different chip size are listed in Table 2. The agreement between these two methods is very good. As the chip size shrinks from 50 × 100 µm 2 to 15 × 30 µm 2 , the sidewall emissions from RGB micro-LEDs increase because of lower absorption and reflection losses inside the chip. For red micro-LEDs, the sidewall emission ratio is much smaller than that from green and blue ones due to stronger absorption of red MQW, resulting in a mismatched angular distribution among RGB micro-LEDs.

Color shift and light efficiency
To analyze the color shift induced by subpixels' angular distribution mismatch, a more representative way is to calculate the color shifts of the mixed colors by RGB color mixing. We have defined 10 reference colors in total in order to evaluate color shift throughout the entire color gamut. These reference colors include three primary colors, white point D65, and six mixed colors (three 100% saturated colors and three 50% saturated colors), which are plotted in CIE 1976 color space [20,21], as shown in Fig. 6.   Figure 7 depicts the color triangle of the RGB micro-LED display and CIE coordinates of 10 reference colors at viewing angles from 0° to 80° with 10° intervals. For primary colors, no color shift is observed at fixed driving current due to weak cavity effect inside micro-LEDs. While for mixed colors and white point, color shift becomes worse as viewing angle increases as expected. The average color shift Δµ′ν′ of all colors at 80° is 0.061 and the maximum value is 0.169 for magenta channel, which exceeds the just-noticeable level (Δµ′ν′ < 0.02). This issue will get worse as the micro-LED size decreases because of increased sidewall emissions from green and blue chips, as listed in Table 2. Therefore, it is necessary to improve the color performance of RGB micro-LED display system, especially for high resolution applications.
In terms of color gamut, our display device can cover 97% of DCI-P3 standard and 78% of Rec. 2020 standard, as indicated by the color triangle in Fig. 7. The relatively narrow color gamut coverage results from the spectral crosstalk between green and red chips, as Fig. 8 shows. Compared to our device, the Osram's green LED has a slightly shorter central wavelength and narrower FWHM. As a result, it has less crosstalk with the red LED and its RGB LED system can achieve wider color gamut. The calculated color gamut coverage is 99% of DCI-P3 standard and 90% of Rec.2020 standard. To achieve a desired color gamut, we can tune the electroluminescent (EL) spectrum of green micro-LED by controlling the well/barrier width and indium composition of MQWs [22].
To reduce patterns, i.e. blue micro-L configuration region. The g refractive ind completely ab micro-LEDs obtained, but ratios of gree Table 2.
Since pow taper angle α surface [23-2 angle and wit 140°, the ligh that from red of black matr narrower angu on the red cou By considerin is an optimal green and blu device with 9 shift. ED display with to ward method i hich means the ed. Figure 9 array and top re filled with r idewall emissi allowing only distributions. T optical efficie to ~50% of th key performanc re (Fig. 9)   The simulated color shifts for RGB micro-LED display with top black matrix and 120° taper angle is plotted in Fig. 11. The average Δµ′v′ of all 10 reference colors at 80° is 0.005 and the maximum value is 0.014 for magenta channel, which is below 0.02 and is acceptable for commercial applications. Besides these 10 reference colors, we also evaluate the color shift using the first 18 colors in Macbeth ColorChecker [26], which is commonly used in color tests and reproductions to mimic the colors of natural objects such as human skin, foliage and flowers. Figure 12 depicts the simulated results. The color shifts of all 18 reference colors within 80° viewing cone are below 0.01 and the maximum average Δµ′ν′ is 0.007, which remains visually unnoticeable by human eye. Fig. 11. Simulated color shifts of 10 reference colors from 0° to 80° viewing angle for RGB micro-LED display with top black matrix and 120° taper angle.   -LED displays φ). It is notewo stay the same a result, the ifts of RGB mi x, the color shi l the sidewall e th RGB color absorb the amb n improves the shift include° p when the orthy that when the e angular icro-LED ift can be emissions rs can be bient light e ambient e making asymmetric subpixel arrangement [29] or employing a patterned scattering film on top of the display panel to achieve matched angular distributions [30]. Each approach has its own merits and demerits. In addition to angular color shift caused by angular distribution mismatch, micro-LED display may also suffer from color shift originating from spectral shift at different driving current densities. Due to the competition between band-filling effects and self-heatinginduced bandgap shrinkage, blueshift at the low current density level and redshift at the high current density level have been reported [31,32]. This color shift issue might be addressed by image rendering, optimization of epitaxial wafer [33] or driving method, etc.

Conclusion
We have analyzed the color shift of RGB micro-LED displays due to angular distribution mismatch. The simulation model is validated by experiment and a good agreement is obtained. In order to mitigate the color shift while keeping a high light extraction efficiency, we proposed a device structure with top black matrix and taper angle in micro-LEDs. After optimization, the color shift Δu′ν′ of the RGB micro-LED display with 120° taper angle is suppressed to below 0.01 within 80° viewing cone for the first 18 reference colors in Macbeth ColorCheker, and the efficiency keeps ~85% of the device without black matrix.