A Novel Neural-Network Device Modeling Based on Physics-Informed Machine Learning

2023/10/03 by Bokyeom Kim, Mincheol Shin · 37 citations
Engineering · Mathematics · Physics and Astronomy · #Advancements in Semiconductor Devices and Circuit Design #Artificial intelligence #Artificial neural network #Computer science #Electrostatic Discharge in Electronics #Enhanced Data Rates for GSM Evolution #Extrapolation #Interpolation (computer graphics) #Machine learning #Mathematics #Model Reduction and Neural Networks #Statistics

paper · doi:10.1109/ted.2023.3316635

published in IEEE Transactions on Electron Devices 70(11), 6021-6025 (Institute of Electrical and Electronics Engineers)

openalex publication_date 2023/10/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/29

Abstract

In this work, we present a novel physics-informed machine learning (PIML)-based neural-network device modeling that predicts both device performance and spatial physical quantities in real-time. Using cutting-edge technologies such as physics-informed neural network (NN) and physics-informed deep operator networks, our approach suggests interpolation and extrapolation strategies in device physics modeling. Despite being trained with a small number of bias voltages, our model demonstrates remarkable accuracy, with a mean absolute percentage error (MAPE) of 0.12% for predicting potential for interpolation and 0.19% for extrapolation. Our approach can be used for data-efficient NN modeling for TCAD and real-time physics analysis in the spatial domain.

Cited by

• DDNet: A Unified Physics-Informed Deep Learning Framework for Semiconductor Device Modeling
• Frequency transfer and inverse design for metasurface under multi-physics coupling by Euler latent dynamic and data-analytical regularizations. [europepmc]
• Combining quasi-one-dimensional modeling with region-wise structure analysis for rapid technology computer-aided design simulations of gate-all-around MOSFETs. [europepmc]
• A Universal Deep Learning Model for Predicting Detection Performance and Single-Event Effects of SPAD Devices. [europepmc]

Related

• Physics-Informed Neural Network for Predicting Out-of-Training-Range TCAD Solution with Minimized Domain Expertise
• Physics-informed Neural Networks for Encoding Dynamics in Real Physical Systems
• Deep Learning Assisted Compact Modeling of Nanoscale Transistor
• Physics-Informed Neural Networks for Electrical Circuit Analysis: Applications in Dielectric Material Modeling
• AutoPINN: When AutoML Meets Physics-Informed Neural Networks