Endpoint Detection Methods in Implementing AI-assisted Polishing Process

Abstract

Sample preparation plays a critical role in the failure analysis process of modern IC chips. The common problem in sample preparation is over-polishing. To reduce this problem, an AI-assisted monitoring system can be deployed, which can evaluate the progress of sample polishing process and suggest the steps following up. However, to build such an AI system, a tremendous number of images with proper classification are needed. To prepare these images, a reliable endpoint detection method for image analysis is necessary. In this paper, two endpoint detection methods are studied, and grayscale line profile analysis is discussed in detail. The current results are very promising for further development.