Title: AI-Driven Strategies for Accurate and Efficient Transistor Parameter Extraction in Next-Generation Device Modeling
Location: Meeting Room 2, 3rd Floor, Crowne Plaza Kunming City Centre
Time: 15:15-16:45, Oct. 21, 2025, Tuesday
Speaker: Dr. Ningmu Zou，Nanjing University , China

 

Abstract: Accurate transistor parameter extraction is a cornerstone of compact models such as BSIM, directly impacting circuit simulation accuracy and design reliability. However, traditional extraction techniques are often slow, iterative, and highly dependent on expert intervention, making them increasingly impractical for today’s complex process nodes and diverse device architectures. This tutorial provides a systematic introduction to how artificial intelligence—particularly reinforcement learning—can reshape the parameter extraction workflow by efficiently navigating high-dimensional parameter spaces and achieving convergence with far fewer simulations. We will begin by reviewing the fundamentals of compact modeling and the challenges inherent in conventional extraction methods, including their limitations in scalability and portability. We then present an improved dueling DQN framework as a case study, illustrating its ability to accelerate convergence, improve accuracy, and adapt seamlessly across different devices and technology nodes. Step-by-step, the tutorial will cover algorithm design choices, state/action/reward engineering, integration with device simulators, and strategies for balancing exploration with convergence speed. Finally, we will discuss best practices for embedding AI-driven extraction into modern EDA flows, highlighting how such integration can enable faster model calibration, more robust parameter portability, and streamlined deployment for next-generation semiconductor technologies. Attendees will gain both conceptual understanding and practical implementation insights.

 

Bio: 

Dr. Ningmu Zou is an Associate Professor at the School of Integrated Circuits, Nanjing University. Prior to joining academia, he spent six years as a Staff Engineer at AMD Inc., where he contributed to advanced semiconductor manufacturing technologies. He earned his Ph.D. in Optical Engineering from Cornell University in 2017, with a focus on advanced integrated circuit manufacturing and artificial intelligence (AI) applications. Dr. Zou has published over 40 peer-reviewed papers in leading international journals and conferences and holds more than 30 granted patents. His research centers on the development and application of AI algorithms in advanced semiconductor manufacturing, including machine learning models and big data analytics for photolithography process optimization, near-field optical proximity correction, device yield and performance enhancement, and chip defect diagnosis and classification.