In this thesis, a simulation pipeline for efficient and accurate atomistic calculations of electron transport in nanoscale devices is developed. This method is based on the non-equilibrium Green's function (NEGF) formalism with tight-binding parameters of the considered materials determined from electronic structures by density-functional theory (DFT) calculations. DFT simulation is a robust technique to model nanostructures, but cannot be scaled to a realistic device sizes due to heavy computational cost. This limitation is circumvented by transforming the delocalized plane-wave states into maximally localized Wannier functions (MLWFs) that serve as the localized basis for the quantum transport solver. This allows accurate modeling of device structures on a micron scale, but with atomic level accuracy. The effectiveness of this approach is demonstrated through the investigation of nanostructures and the comparison with experimental results. Firstly, in order to validate our approach, we compared the transport results obtained by our method with that by full DFT simulation. The two methods agree very well but our method uses three orders of magnitude less time. Then we tested our transport calculations by applying it to the telescopic double wall carbon nanotube, where two nanotube of different radius overlap with each other. The obtained results are similar to the ones in literature. We then applied our simulation pipeline to the important problem of metal-semiconductor contact. Metal-semiconductor contact is a major factor limiting the shrinking of transistor dimension to further increase device performance. Two-dimensional (2D) materials such as graphene and transition metal dichalcogenides (TMDCs) are pushing the forefront of complementary metal-oxide semiconductor (CMOS) technology beyond the Moore's law, and show great promises for realizing atomically thin circuitry. A fundamental challenge to their effective use remains the large resistance of electrical contacts to 2D materials for probing and harnessing their novel electronic properties. There are generally two types of contact geometries, namely top contacts and edge contacts, both of which are examined in in this thesis. Conventional 3D metallic top contacts can achieve low contact resistance with monolayer 2D materials, but cannot avoid the intrinsic problem of large electrode volume. 2D top contacts, including graphene and recently demonstrated atomically flat metal thin films, can achieve both small volumes and low contact resistances of metal-semiconductor interfaces. The analysis of graphene-\ch{MoS2} top contacts reveals that they suffer from weak van der Waals coupling to TMDCs \cite{kang2014computational} so their transfer efficiency depends largely on the contact area and is compromised dramatically below a transfer length which is typically tens of nm scale \cite{ling2016parallel, schulman2018contact}. On the other hand, in-plane edge contacts have the potential to achieve lower contact resistance due to stronger orbital hybridization compared to conventional top contacts. We then present full-band atomistic quantum transport simulations of the graphene/\ch{MoS2} edge contact. We find that the potential barrier created by trapped charges decays fast with distance away from the interface, and is thus thin enough to enable efficient injection of electrons. This results in Ohmic behavior in its I-V characteristics, which agrees with experiments. Our results demonstrate the role played by trapped charges in the formation of a Schottky barrier, and how one can reduce the Schottky barrier height (SBH) by adjusting the relevant parameters of the edge contact system. The thesis provides full details on the application of the MLWF technique to self-consistent quantum transport simulations, as implemented in our open-source software swan. Our framework can be extended conveniently to incorporate more general nanostructure geometries as well as electron-phonon interactions. Such approaches are important for understanding electron flow beyond the quantum limit and have started to draw increasing attention from the device modeling community.




Downloads Statistics

Download Full History