Post Bonding Defect Analysis using Deep Learning

In the semiconductor industry, defect detection is very important as it affects final device performance. Hybrid bonding yield is especially sensitive to defect, and therefore, validation of defects post bonding is critical to ensure robust downstream device performance. This paper presents a novel approach to identifying and analyzing defects in post hybrid bonding processes using Computer Vision and Deep Learning- based classification methods with limited availability of data. With this approach, both defect identification throughput and accuracy are improved, thereby improving device yield and performance.