微發光二極體顯示器亮度提升及巨量檢測

Abstract

近期，隨著不同光電應用，如擴增實境（AR）、可彎曲螢幕、虛擬實境（VR）等的蓬勃發展，微發光二極體（micro-LED）嶄露頭角，因此被視為未來下一代顯示技術。儘管微發光二極體具有低功耗、長壽命和高對比度等優勢，但仍然存在許多需要改進的問題。 微發光二極體面臨的挑戰之一是隨著亮度尺寸變小，外部量子效率（EQE）急劇降低。由於表面體積與體積的比例增加，感應耦合式電漿蝕刻（ICP-RIE）造成的側壁損傷變得越來越顯著。在我們的研究中，我們引入了數字蝕刻和不同處理時間的N2處理技術，以比較不同發光二極體尺寸大小的效果。在注入25 A/cm2 電流密度的電流時，隨著微發光二極體的尺寸從50x50 µm²減少至8x8 µm²，其量子效率（EQE）減少了33.6 %。同樣地，當尺寸進一步從50x50 µm²減少至4x4 µm²，EQE僅減少56.9 %。 此外，由於發光二極體尺寸的減小，檢測也面臨許多限制以及挑戰。在這篇文章中，我們展示了使用無害的ITO玻璃接觸微發光二極體陣列來檢測發光亮度和輻射概況的電致發光巨量檢測方法。巨量檢測包括兩個階段。在第一階段，我們通過多元線性回歸分析對每個單獨的微發光二極體分析得到不同位置的校準方程式，以校準因電阻而有所不同的發光數據。在第二階段，我們透過構建以及比較不同的2D卷積神經網絡（2D-CNN）模型來判斷有破壞性或發光不均勻的微發光二極體並找出最適合的模型。我們的新解決方案在檢測方面實現了更高效和更準確的方式，發光檢測的誤差僅為8 ％，輻射概況透過模型判斷的準確率達到96.7 ％、精確率達到96.7 ％、召回率達到96.7 ％。

Recently, micro-LEDs (light-emitting diode) have emerged by the tremendous development of different optoelectronic applications such as augmented reality (AR), flexible screens, virtual reality (VR), etc. and is thus considered as the future display technology of next generation. Although micro-LEDs possess advantages of low power consumption, long lifetime, and high contrast, they still have many obstacles to improve. One of the challenges of micro-LEDs is the sharply reduction of external quantum efficiency (EQE) as the luminance size gets smaller. As the ratio of surface volume to volume increases, the sidewall damage caused by Inductively Coupled Plasma - Reactive Ion Etching (ICP-RIE) becomes increasingly significant. In our study, we introduce a digital etching and N2 treatment technology with different treatment times compared to different mesa sizes to repair the sidewall damage. As a result, at 25 A/cm2 of injection current, as the size of micro-LEDs decreases from 50x50 µm² to 8x8 µm², the EQE shows a reduction of 33.6%. Likewise, with a further reduction in size from 50x50 µm² to 4x4 µm², the EQE decreases by 56.9%. Furthermore, the detection is subject to many limitations due to the reduction in the size of micro-LEDs. Here, we demonstrated mass detection methods to examine both luminance and radiation profiles by using harmless ITO-glass contact on micro-LED array. The mass detection consists of two stages. In the first stage, we perform a calibration equation to calibrate the luminance data from resistance variation on every single micro-LED by multi-variable regression analysis. In the second stage, we determine the defective or non-uniform micro-LEDs by constructing and comparing different 2-D Convolutional Neural Network (CNN) models to find the most optimized one for micro-LED detection. Our new solution achieves in more efficiency and accuracy way for detection with error down to only 8 % for luminance detection and 96.7 % accuracy, 96.29 % precious, and 99.45 % recall for radiation profile recognition.