YOLOv4-Based Semiconductor Wafer Notch Detection Using Deep Learning and Image Enhancement Algorithms

Abstract

This study designs a system to precisely detect the angle of wafers on an ion implanter's electrostatic chuck (ESC). In specific ion implantation processes, ions may penetrate deeper than intended because of the channeling effect, compromising the device performance. To address this issue, the system adjusts the tilt of the ESC and the twist angles of the wafer to control the ion beam direction. Utilizing a camera-based machine learning system, the system identifies the wafer notch to ensure an accurate alignment of the ESC. However, factors such as insufficient lighting and vibrations affect notch detection, which can degrade image quality. To overcome these issues, this study explored various image-enhancement techniques and evaluated the performance of object detection algorithms on enhanced images.