A Comprehensive Review of Machine Learning Approaches for Semiconductor Device Modeling and Simulation

Application of machine learning (ML) approaches has recently expanded within a wide range of domains, including semiconductor devices. Their strong potential resulted in increasing number of research in semiconductor device modeling and simulation. This article provides a comprehensive review of the recent research on ML-based semiconductor device modeling and simulations. We organize the literature based on the device category, such as Planar Metal–Oxide–Semiconductor (MOS) device, Field-Effect Transistors (FET) devices, High Electron Mobility Transistor (HEMTs) devices and other semiconductor devices. We discuss the implementation of ML methodologies on each device along with the primary findings and summarize them in a Table. This review finds the use of a wide variety of ML approaches in device modeling and simulation, including traditional and deep learning (DL) models. Furthermore, this study shows the ability ofMLmodels to capture the complex, nonlinear patterns in device simulation. Finally, we identify key challenges associated with implementing ML approaches in device simulations and suggest future directions for implementing ML models efficiently. Overall, this review provides useful resources for researchers, engineers, and industry experts interested in leveraging machine learning algorithms to efficiently model semiconductor devices.