Reliability Study of Critical Structural Redistribution Layers in Advanced Packaging: A Review

The continuous evolution of semiconductor packaging demands highly reliable redistribution layer (RDL) architectures to support next-generation electronic systems. However, ensuring RDL reliability remains a formidable challenge due to multiphysics interactions, including mechanical stress-induced fatigue, thermal expansion mismatches, and high-frequency signal integrity degradation. This article presents a comprehensive review of RDL reliability across mechanical, thermal, and electrical domains, identifying key failure mechanisms and research gaps. To address these challenges, we introduce an AI-driven optimization framework that integrates machine learning–assisted chip layout optimization, adaptive thermal management, and real-time signal integrity enhancement. Utilizing deep reinforcement learning and graph neural networks, this study demonstrates how AI can dynamically optimize RDL routing, enhance power distribution networks, and mitigate localized heating effects. Furthermore, we explore the integration of AI-driven predictive modeling into electronic design automation tools, enabling real-time multiphysics co-optimization of RDL architectures. This study establishes a structured framework for future research, bridging the gap between theoretical modeling and practical fabrication. By incorporating AI-assisted design methodologies, next-generation RDL architectures can achieve superior reliability, enhanced performance, and improved scalability, supporting applications in 5G communications, high-performance computing, and heterogeneous integration technologies.