Advanced numerical modeling of semiconductor material properties and their device performances

Abstract

With the renewed concept of "Materials by Design" attracting particular attentions from the engineering communities in recent years, numerical methods that can reliably predict the optical and electrical properties of materials is highly preferable. Since the growth or the synthesis of a "designed" material and the ensuing devices is usually prohibitively expensive and time-consuming, numerical simulation tools that predict the properties of a proposed material together with its device performance before production is especially important and cost-effective. Furthermore, as the technology advances, semiconductor devices have been pushed to operate at their material limits, which requires a thorough understanding of the materials' microscopic processes under different conditions. Therefore, developing numerical models that are capable of investigating the semiconductor properties from material level to device level is highly desirable.

This dissertation develops a suite of numerical models in which optical absorption and Auger recombination in semiconductor materials are studied and simulated together with their device performances. In particular, Green's function theory with full band structures is employed to investigate the material properties by evaluating the broadening of the electronic bands under the perturbation of phonons. As a result, both direct and phonon-assisted indirect processes are computed and compared among different materials. Drift-diffusion model and a 3D Monte-Carlo model are subsequently used to simulate the device characteristics with the obtained material parameters. This work first determines the full band structures for Si, Ge, α-Sn, HgCdTe, InAsSb and InGaAs alloys from EPM model, and then investigated the materials' minority carrier lifetime for IR detector applications. Finally device level simulations using drift-diffusion and 3D Monte-Carlo models are demonstrated. In particular, two issues of developing 3D Monte-Carlo device simulation models, namely the use of unstructured spatial meshes and elimination of particle-mesh forces, are discussed, which are crucial in simulating modern semiconductor devices having complex geometry and doping profiles.