Hough Transform for Indirect Estimation of Wafer Placement Errors in Photoresist Spin Coating Processes

This paper discusses a low-cost, computer vision based approach to indirectly estimate on-chuck substrate (wafer) placement errors for photoresist spin coating processes in a semiconductor manufacturing environment. Placement errors are estimated by calculating the relative displacement vector between circles bounding the wafer and the photoresist region post edge bead removal (EBR) processing. On-chuck wafer placement is critical in maintaining concentric EBR performances and without a method of detection it is challenging to contain mechanical tool failures, incorrectly performed preventive maintenance (PM) or other human errors. The study revisits the Hough transform (HT) for circle detections from accuracy and computational viewpoints using synthetically generated images. The detection accuracy of HT is proven outstanding. However, processing times dramatically increase (hours) in case of high resolution, real wafer images despite adequate preprocessing. This drawback is compensated by processing only subsets of images relying on mechanical wafer position controls during the wafer scan although, this potentially undermines the overall accuracy of this classical approach.