ABSTRACT
Process variation control and root causes elimination are critical to the issue of yield enhancement in semiconductor manufacturing industry. To detect the existence of process variation, one of the most effective ways is to analyze the spatial defect patterns exhibiting on the wafers. Many research works have been proposed to help recognize the spatial patterns. These works can basically be classified into statistical approach and training based approach. In statistical approach, the defect data are statistically analyzed to find out the clustering phenomena, and it usually cannot conduct further analysis to identify the specific defect patterns. Comparing to the statistical approach, the training based approach has the capability of classifying different spatial defect patterns. But it requires the collection of enough training samples, which is usually very time-consuming. Moreover, when the product or process changes, the training process needs to be executed again. For these reasons, this research proposes a feature-based pattern recognition algorithm to automatically identify and classify different defect patterns such that human intervention can be replaced. The experimental results show that this approach can effectively achieve the intended purpose.