Through-silicon via (TSV)-based 3D Integrated Circuits (ICs) offer various advantages including higher power efficiency by shorter interconnection, better performance, smaller form factor, and wider bandwidth while enabling heterogeneous integration compared to conventional ICs. Nevertheless, during the fabrication process, testing, and operation, TSVs are exposed to multiple thermal cycles, which will lead to considerable thermal stresses in and around the TSVs. Thermal stress in TSVs raises multiple thermo-mechanical reliability issues in the 3D ICs, among which "via extrusion" is an important one. Via extrusion is a crucial reliability concern as the protrusion of Cu vias in the axial direction can deform and crack the interconnect layers and result in device failure.
Via extrusion has been studied and proven difficult to be thoroughly controlled in most investigations. The stochastic nature of via extrusion may be traced to the different mechanisms of extrusion, the dominance of each is dependent upon the stress and microstructure state of the via. Hence, obtaining a better understanding of the mechanisms of extrusion in a large number of Cu TSVs is the key to the development of practical solutions for this important reliability issue in 3D ICs. Statistical variation has been also observed in the morphology of extrusion, suggesting a correlation between the extrusion mechanism and morphology. Therefore, statistical investigation of via extrusion morphology assists with addressing the extrusion variation concerns in 3D-ICs. However, studying the extrusion morphology in TSVs through conventional methods is extremely time-taking which results in low throughput. Thus, in this project, we will leverage deep learning paradigms to recognize the TSV extrusion morphology patterns and categorize them in different classes. Deep learning approaches enable fast and accurate analyses of a large number of vias, which can help to realize the statistical variation of extrusion morphology in TSV-based 3D-ICs.
Dr. Tengfei Jiang, Department of Materials Science & Engineering and Advanced Materials Processing and Analysis Center at University of Central Florida (UCF).
Dr. Golareh Jalilvand, Department of Chemical Engineering at the University of South Carolina (UofSC).