Session: K9-09: NANOSCALE THERMAL TRANSPORT MODELING AND MACHINE LEARNING I

Paper Number: 137120

137120 - Physics-Informed Neural Networks for Transistor Thermal Modeling Using Phonon Boltzmann Transport Equation 

Abstract: 

Over the past few decades, the reduction in the physical size of integrated circuits has contributed to the widespread of lightweight and compact electronic devices. Transistors, being the fundamental building blocks of CPUs, have evolved to facilitate size reduction and improve computational efficiency. However, the shrinking size of transistors leads to an increase in hot spot temperatures and may result in reduced lifespan and potential malfunctions. Therefore, the prediction of hot spot temperatures plays a vital role in the design and optimization of transistors. To accurately calculate temperature distribution, one needs to solve phonon Boltzmann transport equation (BTE) due to the non-diffusive transport behavior in the nanoscale, which Fourier’s heat equation cannot handle. While numerical solvers exist for phonon BTE, achieving converged solution and eliminating the ray effect usually requires substantial computational memory and time. Moreover, obtaining the temperature distribution for a new design entails repeating the entire procedure. To overcome these limitations, we show a physics-informed neural network to concurrently solve phonon BTE for transistor thermal modeling with varying geometric parameters, taking advantage of the flexibility and automatic differentiation of neural networks. This approach enables the prediction of temperature distribution within a few seconds for various geometric parameters, offering both efficiency and high accuracy.

Presenting Author: Jiahang Zhou University of Notre Dame

Presenting Author Biography: Jiahang Zhou is a 3rd-year graduate student in the Department of Aerospace and Mechanical Engineering at the University of Notre Dame, who is advised by Prof. Tengfei Luo and Prof. Jianxun Wang. His research interest lies in the intersection of thermal transport and machine learning, using novel machine learning tools to speed up the thermal modeling and thermal design of micro-nano electronic devices.

Authors:

Jiahang Zhou University of Notre Dame
Wenjie Shang University of Notre Dame
Pan Du University of Notre Dame
Jyoti Panda University of Notre Dame
Bo Zhang University of Notre Dame
Jianxun Wang University of Notre Dame
Tengfei Luo University of Notre Dame