Machine learning assisted multiphysics simulation for electroplating copper in high aspect ratio through silicon via

Abstract

During the electroplating copper process of through silicon via (TSV), the process induced defects such as voids and seams, which originate from the inappropriate process parameters, critically compromise the structural integrity and long-term reliability of integrated chips. To solve the present challenges, this study integrates the multiphysics finite element simulation method with the machine learning technology to systematically elucidate the regulatory mechanisms of electroplating additives during the filling process. The results indicate that appropriately increasing the concentration of the suppressor can achieve defect-free filling. In further research, to overcome the limitations of experiment and simulation approaches in terms of material consumption and computational demand, this study employed the data-driven machine learning model for rapid and accurate evaluation of electroplating filling quality, achieving a prediction accuracy of up to 98%. This study provides theoretical support for understanding the defect-free electroplating filling mechanisms of high aspect ratio TSV and its intelligent optimization, offering the valuable reference for the three-dimensional (3D) advanced packaging technologies.