Unsupervised Pattern Classification for Dual-Layer EPE Characterization in Advanced Logic Devices

Abstract

This work presents a simulation-driven pattern classification methodology tailored for multi-layer logic device applications. Leveraging aerial image data of both line-space and contact hole patterns, we demonstrate the approach on a two-layer stack. The methodology integrates data science techniques to extract and cluster similar patterns, enabling the identification of EPE-critical pattern groups. These critical pattern groups provide valuable insights for lithography process setup, inline monitoring, and advanced process control strategies.