Research

Thermal Reliability Modeling & Characterization of Advanced Packaging

Advanced packaging has emerged as a key solution to the rapid growth of microelectronics. The demand for higher performance and further miniaturization inevitably increases power density and generates extreme heat flux. These conditions pose critical challenges in thermal management. To overcome these limitations, thermal-aware design of advanced packaging is urgently required for next-generation electronic systems. We aim to investigate the thermal phenomena in advanced packaging, ranging from the standard-cell level to the architecture level, and to establish guidelines for the thermal-aware design of advanced packaging. Through this research, we seek to provide a fundamental understanding of heat transfer mechanisms and deliver design strategies that enhance both performance and reliability in future semiconductor systems. 

AI-based Design of Electronic Cooling Technologies

To cope with the extreme power density limits of high-performance systems and high-power devices, conventional microchannel cooling solutions must significantly improve their cooling efficiency. In heterogeneous integration, each chiplet generates different levels of heat, leading to temperature nonuniformity. Such nonuniformity induces thermo-mechanical stress and raises serious reliability concerns. To address these challenges, we propose a novel approach to designing microchannels for liquid cooling using artificial intelligence (AI). AI is employed because the design space of microchannel geometries is vast and highly nonlinear, making traditional trial-and-error or simulation-only methods inefficient. By leveraging AI, we can efficiently explore complex design parameters, optimize cooling performance, and achieve reliable thermal management for advanced packaging systems.

Optimization of Panel-Level Packaging Fabrication Using AI

With the rapid growth of GPU performance, the form factor of advanced devices continues to increase, often exceeding the wafer reticle size limit. This trend highlights the need for panel-level packaging to accommodate larger system integration. In this context, redistribution layer (RDL) plating and polymer insulating dielectric (PID) processes become critical, yet conventional approaches struggle to achieve uniformity and scalability over large areas. To address these challenges, we investigate large-area printing processes as an alternative solution. Furthermore, artificial intelligence (AI) is employed to optimize printing parameters and process conditions, enabling cost-effective, high-throughput, and reliable fabrication of RDL and PID structures for advanced packaging.