A novel machine learning solution to maximize defective die capture and yield prediction accuracy in inline process control with e-beam inspection

This paper introduces a novel machine learning solution, Computational Guided Inspection (CGI), designed to enhance defective die capture and yield prediction accuracy in semiconductor manufacturing. Traditional inspection methods face limitations in resolution and coverage, necessitating an intelligent approach to die sampling for e-beam inspection. CGI leverages historical fab data and machine learning techniques to guide die sampling, resulting in a substantial 1.5 times improvement in defective die capture rates compared to conventional methods. The study showcases CGI's potential to transform inline defect inspection processes, offering significant advancements in yield management and process optimization in high-volume manufacturing environments.