Simulation of advanced semiconductor devices with 3D Monte Carlo
OA Version
Citation
Abstract
The push to advance the capabilities of semiconductor devices relies on the continual advancement of emerging materials and novel device structures. Due to the expensive costs of semiconductor fabrication in terms of both resources and time, device modeling has always been necessary to estimate device performance and provide design optimizations prior to their fabrication. However, commonly used device simulation methodologies, such as drift-diffusion, require the input of phenomenological parameters, which can be difficult to obtain especially for technologically immature materials. Using the Monte Carlo method, many of these parameters can be extracted with first-principles calculations. Furthermore, the physical models included in Monte Carlo simulations make it better suited for simulating sub-micrometer scale and high-field devices compared to drift-diffusion.
This thesis focuses on the Monte Carlo method for simulating semiconductor materials and devices using the simulation software developed at Boston University. In depth descriptions and visualizations of the Monte Carlo simulation process are provided, including visualizations of the 3D bandstructure and carrier-phonon interactions. We demonstrate three levels of physical accuracy of Monte Carlo simulations, with bandstructures obtained analytically, semi-empirically, or entirely from first-principles. Since higher levels of physical accuracy incur higher computational costs, we also demonstrate appropriate scenarios for the implementation of each type of bandstructure. Results of device simulations and simulations of carrier transport phenomena are presented for a wide variety of materials with bandgaps ranging from 0.3 eV to 6.4 eV. Finally, we also demonstrate how values extracted from Monte Carlo simulations can be used in other simulation methodologies, including drift-diffusion.
In addition to the standard Monte Carlo methodology, we also introduce a 3D quantum correction model for carrier transport at band crossing and anti-crossing points. Quantum corrections are necessary to accurately simulate carrier transport in most wurtzite materials, which includes technologically significant materials such as 4H silicon carbide and aluminum nitride. The methodology behind its implementation and integration within the Monte Carlo infrastructure is discussed, and comparisons of calculations with and without the quantum corrections are presented. Finally, we also present preliminary device simulations incorporating this fully quantum model and speculate about its future within the Monte Carlo paradigm.