RNN-Based Proxy Modeling for TCAD Simulation of Trench Mosfet Manufacturing Processes

This paper presents an efficient recurrent neural networks (RNN) based technology computer aided design (TCAD) proxy model that accurately simulates the process of Trench metal-oxide-semiconductor field-effect transistor (MOSFET) devices. The model significantly accelerates the simulation process while maintaining high accuracy in predicting the device's electrical performance. The verification is carried out on a N-type Trench MOSFET with a 150nm gate feature size. The model generates predictions in under 1 second, achieving an accuracy greater than 98% when compared to traditional TCAD simulation results.