A Deep Learning Based Fast Device Simulation Model for Nanosheet FETs

This research introduces a deep learning-powered rapid simulation model for Nanosheet FETs. (DL-Model). By employing an Artificial Neural Network (ANN), the DL-Model accurately captures the device structure and electrical characteristics of Nanosheet FETs, thereby significantly reducing the computational resources and time required by traditional TCAD simulations. We trained the ANN model using simulation data generated by Sentaurus TCAD and validated its accuracy and generalization capability across various bias conditions and geometric dimensions. Compared with conventional TCAD simulations, the DL-Model not only enhances simulation efficiency but also effectively addresses convergence issues associated with solving complex partial differential equations. Experimental results demonstrate that the DL-Model achieves high precision and rapid response in predicting the device structure and key electrical characteristics of Nanosheet FETs, offering strong backing for the swift development and refinement of cutting-edge semiconductor devices.