垂直式刷輪於畫素顯示基板異物瑕疵改善研究

Abstract

本論文主要探討起因於TFT-LCD(Thin Film Transistor - Liquid Crystal Display)量產品客訴CP (Cell Particle)過高(8~20% Yield Loss) ，造成整批產品遭客戶驗退，對公司信譽及營收造成嚴重損失。由客戶端退貨的產品中剖片分析CP種類，依據分析結果將Q&A Map (Question & Answer Map )與魚骨要因圖結合並給予直接相關(5分)、強相關(4分)..等相關分數，分數越高代表影響比例越高，並針對各種Defect做改善。經由分析CP種類歸納出兩種Defect Type: UPI (Under Polyimide) Type: 於PI層之下的Particle Defect 及 OPI (Over Polyimide) Type: 於PI層之上的Particle Defect. 並計算良率損失百分比為OPI Type 17.3%，UPI Type 82.7%;其中以UPI Type內的(PI製程前基板表面微粒髒污)佔整體68.56%為大宗，此Particle Defect主要成因為玻璃基板(TFT or CF)於PI液Coating基板表面前的清洗機無法有效的將表面髒污去除乾淨，以致於在做PI液Coating時PI液無法均勻的覆蓋於基板表面造成的配向不良，且UPI Type Defect位於PI層之下所以後製程清洗段是無法清除，rework也不符合成本效益。本研究重點於提升PI製程前清洗能力，由原先的水平式(臥式)刷輪改造為立式(垂直式)刷輪，並利用田口式(Taguchi method)進行實驗找出製程參數(順逆刷法、刷輪轉速、刷輪下壓量、Convey傳送速度)之最佳水準，經由改善前之Yield Loss為8~20% ，利用最佳製程參數得到Yield Loss大幅下降至為2~6%。

This study explored the improvement of substrate cleaning process before Polyimide film coating in the thin-film transistor liquid crystal display (TFT-LCD) production line. An entire batch of TFT-LCD panel was returned by customer caused by the excessive number of 8%–20% cell particles (CP) defects. Profile analysis process of CP types was conducted on the defected products by question and answer (Q&A) map and fishbone diagram to assign scores for various defect types. Through the analysis of CP, the defects were categorized into two types: (1) under polyimide (UPI), which refers to particle defects under the polyimide (PI) layer; or (2) over polyimide (OPI), which refers to particle defects above the PI layer. Additionally, the yield loss statistic showed that the OPI type accounted for 17.3% of yield loss, whereas the UPI type accounted for 82.7% of yield loss, among which dirt particles on the substrate surface before PI processing accounted for 68.56% of the overall yield loss of the UPI type. The main cause of these particle defects was due to the failure of cleaning the peripheral machines to prevent dirt from getting on the substrate surface before coating the PI fluid onto the glass substrate surface (TFT or CF), which causing an uneven coverage of PI fluid on the substrate surface during PI film coating, and resulting in poor alignment. Furthermore, because the UPI-type defects were located below the PI layer, the dirt could not be deleted during the post-process cleaning, and reworking of panel was not cost-effective. To improve the pre-PI process cleaning ability, this study replaced the original horizontal brush wheels with standing (vertical) brush wheels, and used the Taguchi method to optimize the process parameters (forward–reverse brushing method, brush wheel speed, brush wheel pressure, and convey speed). The optimized parameters were applied to the substrate cleaning process, and the yield loss was reduced from 8%~20% to 2%~6%.