Prediction of Electrical Characteristics of GaAs HEMT Based on Deep Learning

TCAD-based simulation takes a long time and involves a huge amount of data, and it is necessary to establish a physical model that conforms to the actual device parameters and working environment to make the simulation results have reference significance. With the continuous development of artificial intelligence, the current neural network model has stronger and stronger prediction ability, and this paper hopes to apply artificial intelligence to the research of semiconductor devices by using the powerful generalization ability and nonlinear fitting ability of neural network models, which can not only solve the problem of slow simulation solution speed of semiconductor devices, but also effectively reduce the research threshold of non-specialists.