Convolutional Neural Network (CNN) Based Process Window Analysising for Lithography

Efficient and highly accurate methods and tools for process window analysis are urgently needed in integrated circuit manufacturing. We have developed a convolutional neural network-based automated process window analysis tool, which can analyze process windows from raw data efficiently. Our tool utilizes convolutional neural network and well-trained models for classifying critical dimension scanning electron microscope images to enhance efficiency and reduce errors for process window analysis. Furthermore, we propose an image segmentation-based classification method to further improve accuracy. The proposed method is potential for expedited and highly accurate process window analysis.