Presentation Information
[24p-12C-10]Dimension Reduction and Mode Analysis of Electromagnetic Field on the Top Surface of 3D Through Silicon Via Array by Using Singular Value Decomposition
〇(D)CHIH-CHUNG WANG1, SONG-EN CHEN1, HUNG-WEI HSU1, JIA-HAN LI1 (1.National Taiwan Univ.)
Keywords:
through silicon via (TSV),finite-difference time-domain (FDTD),singular value decomposition (SVD)
As the semiconductor industry moves toward ad-vanced packaging, high aspect ratio through silicon via (TSV) structures with smaller critical dimensions and high-er aspect ratios are being designed that can be applied to 3D IC stacking. The critical dimensions of the TSV hole array structure are important, such as depth and top dimen-sions, as well as side roughness and bottom detail. To explore the issue, we used the finite-difference time-domain (FDTD) method to analyze the simulation results of electromagnetic field data from different TSV arrays. However, the dimension of near-field data is very large, and it is necessary to compress the huge data set by finding dominant singular value terms to facilitate subse-quent analysis. The singular value decomposition (SVD) can be used for noise reduction in the field of electromag-netic waves, and can also be used for modal calculation and order and dimensionality reduction. After SVD processing, the amount of data can be compressed. This approach has the potential to be used on data analysis and processing inverse problems. We successfully used the SVD method to compress the TSV near-field data, which is of great help in building a database of TSV critical dimensions. Subsequently, we can find the relationship between the geometric parameters of the TSV and the near-field data to solve inverse problem. This method can not only achieve effective dimension re-duction calculations, but also further explore the analysis of electromagnetic field modes in TSV.