Addressing Missing Die Data in CD-SEM Analysis: A Direction-Weighted KNN Approch

In semiconductor manufacturing, handling missing values in measurement data is a crucial step to ensure accurate data analysis or feature extraction. Traditional interpolation algorithms, such as K-nearest neighbor (KNN), have been widely used to estimate missing die locations on wafer maps. However, when applied to CD-SEM measured wafer maps in lithography, such interpolation methods may not accurately approximate the true data due to the lack of consideration for lithography process characteristics. This paper proposes a modified direction-weighted KNN interpolation algorithm with considerations of scanner exposure direction. Comparative analysis demonstrates that the modified algorithm more effectively represents the feature data after interpolation, leading to better alignment with the original wafer map.