Physics-Based α-IGZO TFTs Compact Modeling and Neural Network Application with 2T0C DRAM Cell

Abstract

Advanced amorphous oxide devices such as amorphous InGaZnO (α-IGZO) operate based on mechanisms that differ significantly from those of conventional Si-based devices, primarily due to structural differences. While both types of devices utilize field-effect mobility as the primary mode of charge transport, there is no consensus on the additional complexities involved in charge movement within α-IGZO devices, which arise from their unique material properties. The BSIM series model commonly used for silicon devices cannot fully explain the charge transport mechanism of α-IGZO devices. Unfortunately, physics-based compact models for α-IGZO, which reflect the intrinsic nature of charge transport involved in electrical conduction have not been completely proposed with a standard formula. This paper presents a compact model for α-IGZO TFTs that incorporates charge transport mechanisms such as percolation, Variable-Range Hopping (VRH), and Trap-Limited Conduction (TLC), along with a methodology for calculating surface potential using the Lambert W function. The model is implemented in Verilog-A for circuit-level simulation and provides high accuracy with fabricated devices measurement. The model’s performance is further evaluated using the MNIST dataset by comparing the classification accuracy across various shallow-layer neural network architectures, demonstrating the model’s potential in neuromorphic system applications.