Micro-Vision-Based 3-DOF Full-Stroke Nanoscale Positioning Method for the Wafer Stage

The quality of semiconductor manufacturing and processing technology is determined by the precision and efficiency of the wafer stage positioning measurement. However, traditional measurement sensors such as laser interferometers (LIs) and grating rulers are expensive to calibrate and have poor orthogonality. To address this, this article presents a micro-vision-based wafer stage positioning system that employs a line-feature-based coarse-to-fine (LFCF) method for global planar three-degrees-of-freedom positioning via the wafer pattern. Line features are extracted from the wafer to estimate the angular displacement. An image database is established, and the residual neural network (ResNet) is improved for coarse positioning to eliminate accumulated errors. The fast and robust feature-based positioning (FRFP) algorithm is proposed for feature extraction and matching, and the fine position is obtained through the constraint equation of feature point pairs. This article also builds a micro-vision-based wafer positioning system to verify the precision and efficiency of the approach. The results demonstrate that the system achieves a precision of 131.4 nm, 82.2 nm, and 0.0015 rad in the x-, y-, and θ-directions, respectively, at an image processing frame rate of 5 Hz.