Principal investigators: Lifang He and Zheng Yao
University: Lehigh University
Industry partner: Broadcom Inc.
The semiconductor manufacturing process is becoming increasingly complicated, with more production tools and processes heightening the importance of recipe control. The tools’ recipes are maintained and controlled by a process engineer. If an engineer unintentionally changes the production recipes and/or tool parameters, wafer production could be scrapped or wafers’ yield degraded. This can result from the recipe parameter modification without being set back, or the wrong recipes are chosen for tools’ operation. If these inline mistakes are not detected and corrected early enough, the potential damage will be significant and extremely detrimental to manufacturing quality, revenue, and reliability. This project proposes a machine learning-based image processing technique to enable manufacturers to preventively resolve dynamic recipe selection and parameter values verification for recipe control and management, enhancing the overall operational efficiency for manufacturers.