为了您获得更好的浏览体验,请将您的浏览器更新到最新版本!
+86.21.6027.8500
Chinese
2019年3月20-22日
上海新国际博览中心

马铁中

马铁中
昂坤科技,首席执行官

个人简介

Dr. Aris Ma received BE from Tsinghua University in Materials Science in 1992 and PhD from University of Minnesota (Twin Cities) in Electrical Engineering in 2000. He worked for PDF Solutions (San Jose, CA USA) as an IC yield consultant and engagement manager until 2005. In 2010, he founded AKOptics (Beijing China), a metrology and inspection tool maker for IC and LED industry. He is currently the CEO of the company. Dr. Ma has over 10 publications and holds over 50 domestic and international patents.

摘要

Setting suitable PR defect spec is challenging in PSS manufacturing process. Ideally one would like to detect the PR defects which will cause PSS defect later on after the photo resist is applied and developed so that any “bad” wafers will not enter into the PSS etching process. Reworking PR wafers require less work and can save cost. Once Sapphire substrate is etched and its defect count exceeds the spec limit, the wafer will be lost. Therefore detecting bad wafers in PR phase is critical to save manufacturing cost. Currently PR defect spec was established based on engineers’ experience. A new software system based on deep learning algorithms are developed to better set up the PR spec. By analyzing the correlation between PR defect and PSS defect, one can build the PR defect killer pareto and thus establish the link between PR defect image and PSS defect image. The software system can analyze the defect image and the count of defects in PR phase and determine whether the PR wafer will be reworked during PR phase or enter next etching process.

The developed software system is adopted into our Falcon 300 AOI system and proved it can detect the PR defect early and set up the correct PR spec thus saving PSS manufacturing costs.