An RDL Modeling and Thermo-Mechanical Simulation Method of 2.5D/3D Advanced Package Considering the Layout Impact...

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Advanced packaging designs face reliability challenges due to decreasing-width, increasing-aspect-ratio routing بسیار (RDL).
• Traditional simulation methods struggle to accurately capture the impact of complex RDL layouts on package reliability.

Purpose of the Study:

• To propose and validate a machine learning (ML)-based method for RDL modeling and simulation.
• To accurately predict the impact of RDL layout on the design for reliability (DFR) of advanced packages.

Main Methods:

• Digitalizing RDL structures into tensors by dividing them into blocks and pixels of metal percentage.
• Building and training an artificial neural network (ANN)-based surrogate model to predict equivalent material properties.
• Transforming blocks into finite elements for simulation and validating with line bending tests.

Main Results:

• Neglecting layout impact led to critical errors in simulations, especially with thinned substrates.
• The ML-based method achieved a 2.81% reaction force error, accurately considering layout impact with 200 × 200 elements.
• Thermal cycling test (TCT) simulations showed maximum stress in RDL, with both RDL and bumps critically affected by layouts.

Conclusions:

• The proposed ML method precisely accounts for layout impacts in RDL simulations with minimal resources.
• This approach offers an efficient way to improve the reliability of advanced package designs.
• Maximum stress in advanced packages is more likely in RDL than bumps, influenced by layout and vias/bumps.