Executive Summary : | The demand for low-power consumption and high speed in electronic systems has driven the search for novel devices based on two-dimensional materials (2DMs), such as MoS₂, WS₂, WSe₂, InSe, black phosphorus, etc. However, introducing 2DMs in industrial process integration for technology development can be expensive and time-consuming. Additionally, selecting 2DMs with optimized geometry and the interface is a challenging task from the family of more than 1800 exfoliate materials. To address these challenges, a modeling framework that enables integrated device performance estimation from the crystallographic information of constituent transistor materials is needed to guide experiments and efforts in the right direction. Further, the transport properties of 2DM-based devices are found to be primarily affected by a vast number of non-ideal effects, such as surface roughness, defects and impurity scattering, interface strain, and processing-related defects. Thus, a modeling framework that can provide a more realistic description of material-to-device level non-idealities is the need of the hour. This project focuses on developing a computationally efficient dissipative quantum transport simulation for 2DM-based devices that can be used to design and optimize at the material, device, and circuit levels. Using first-principles DFT simulation, the project will start by understanding the role of material-level non-idealities, such as surface roughness, defects, impurity, and interfacial states. The DFT-based simulation will be used to extract the tight-binding Hamiltonian matrix and material attributes. The material attributes will be used to develop numerical simulation tools based on iterative solutions of 3-D Poisson's equation and non-equilibrium Green's function method (NEGF). To include the more realistic phonon scattering effects, NEGF equations are often solved within the self-consistent Born approximation (SCBA) through an iterative resolution of the Dyson equation. However, the complexity and computational burden of SCBA limits its application to realistic device geometries. To reduce the computational complexity associated with scattering processes, the lowest order approximation (LOA) approach of the interactions will be optimized and rescaled for 2DM-based devices. Finally, a deep neural network (DNN) will be designed by combining network algorithms and device models to develop a computationally efficient solution to the self-consistent NEGF and Poisson's loop. The Verilog-AMS interface using a look-up table approach will be developed to provide circuit and system-level descriptions of 2DM-based devices. The open-source platform will be made available to encourage further efforts in the 2DM-based device and circuit design. |