A nondestructive automated defect detection system for silicon carbide wafers

Abstract.

Manual methods of defect detection and classification on silicon carbide (SiC) wafers are tedious, time consuming, and prone to error. We have developed a nondestructive optical stress technique (photoelasticity) to isolate structural defects on SiC wafers. The technique is rapid, nondestructive, and inexpensive. In this paper we present an image processing system that exploits the optical system to detect structural defects on SiC wafers automatically. We are specifically interested in detecting defects known as micropipes. The philosophy of our approach is to reduce the dependency of the environment factors, image acquisition factors, and user parameters while maintaining the performance and computational speed. The goal is achieved by careful study of patterns that are invariant to the contrast and shift of pixel intensities and a combination of simple image processing techniques that are locally adaptive.