Neural Network-Based and Modeling With High Accuracy and Potential Model Speed<i/> <i/>
Chien-Ting Tung, Ming-Yen Kao, Chenming Hu
Abstract
In this brief, we demonstrate a neural network (NN)-based device modeling framework. This NN model is built to model advanced field-effect transistors (FETs). Specific transfer functions and loss functions are chosen to achieve high accuracy and smoothness in the output of this NN model. Both <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${I}$ </tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}$ </tex-math></inline-formula> (current–voltage) and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${C}$ </tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}$ </tex-math></inline-formula> (capacitance–voltage) characteristics are studied in this work. Speed comparison between the NN-based model and Berkeley short-channel IGFET model (BSIM) has been done to show that NN has a great potential to accelerate circuit simulation speed. We also present that this NN modeling framework is not only useful for more Moore technologies [e.g., gate-all-around FET (GAAFET)] but also beyond Moore transistors [e.g., negative capacitance FET (NCFET)].