Data-driven approaches to optical patterned defect detection
Mark-Alexander Henn, Hui Zhou, Bryan M. Barnes
Computer vision and classification methods have become increasingly popular in recent years due to ever-increasing computation power. While advances in semiconductor devices are the basis for this growth, few publications have probed the benefits of data-driven methods for improving the defect detection and inspection for such devices. As defects become smaller, threshold-based approaches eventually fail to pick up differences between faulty and non-faulty structures. To overcome these challenges we present machine learning methods including convolutional neural networks (CNN) for image-based defect detection. These images are formed from the simulated scattering off realistic geometries with key defects taking into account line edge roughness, a known and challenging problem in semiconductor manufacturing. A fundamental CNN approach is shown to extend detectability to these smaller defects, even those that are more than 30 times smaller than the inspection wavelength.