S-ANN: Synchronous TCAD device simulation of FinFET using Artificial Neural Network

Abstract

The Fin Field-Effect Transistor (FinFET) plays a crucial role in integrated circuits due to its superior control capabilities and low leakage current. However, its complex structure often leads to substantial computational demands and significant time consumption during Technology Computer-Aided Design (TCAD) simulations. To address these challenges, we innovatively propose a synchronous TCAD device simulation model based on Artificial Neural Network (ANN) for FinFET at advanced technology nodes, abbreviated as S-ANN. This model systematically extracts FinFET characteristics under various physical sizes and bias conditions from TCAD simulations and further analyzes transient response data contained in TDR files to build a training dataset, thereby enabling effective training of the S-ANN model. Through rigorous testing and evaluation, S-ANN has demonstrated high compatibility with Sentaurus TCAD and the capability to accurately simulate TCAD electrical characteristics. In addition, Nanosheet FET was used to verified the generalization capability of the S-ANN model. Compared to traditional TCAD simulations, the S-ANN model significantly reduces both computational resource usage and simulation time, while effectively overcoming convergence problems. This advancement offers strong support for the rapid design and optimization of advanced semiconductor devices.