Neural Network Assisted Si-on-Nothing MOSFET Current-Voltage Modeling with Incremental Learning

In this work, the current-voltage model of Silicon-on-nothing (SON) MOSFET is developed by extending BSIM-SOI model using neural network (NN) assisted method, which model accuracy reaching 99%. The implementation of incremental learning in device modeling field is also verified by sequential tasks in actual device development. Both two incremental approaches, continual-learning NN and block-modular NN, provide NN-assisted device model with expanding ability. Block-modular NN shows better compatibility in device modeling due to the maintaining of prior knowledge.