With the advancement of nanofabrication techniques, the sizes of semiconductor electronic and optoelectronic devices keep decreasing while the operating speeds keep increasing. High-speed operation leads to more heat generation and puts more thermal stress on the devices. Since the heat conduction in semiconductors is dominated by the lattice (i.e., phonons), understanding phonon transport in nanostructures is essential to addressing and alleviating the thermal-stress problem in these modern devices.

In addition to the increased thermal stress, the advanced techniques that have allowed for the shrinking of the devices routinely rely on heterostructuring, doping, alloying, and the growth of intentionally strained layers to achieve the desired electronic and optical properties. These introduce impediments to phonon transport such as boundaries, interfaces, point defects (alloy atoms or dopants), and strain. Phonon transport is strongly affected by this nanoscale disorder. This dissertation examines how different types of disorder interact with phonons and degrade phonon transport.

First, we study thermal transport in graphene nanoribbons (GNRs). GNRs are quasi-one-dimensional (quasi-1D) systems where the edges (boundaries) play an important role in reducing thermal conductivity. Additionally, the thermal transport in GNRs is anisotropic and depend on the GNR's chirality (GNR orientation and edge termination). We use phonon Monte Carlo (PMC) with full phonon dispersions to describe two highly-symmetric types of GNRs: the armchair GNR (AGNR) and the zigzag GNR (ZGNR). PMC tracks phonon in real space and we can explicitly include non-trivial edge structures. Moreover, the relatively low computational burden of PMC allows us to simulate samples up to 100 microm in length and predict an upper limit for thermal conductivity in graphene.

We then investigate the thermal conductivity in III-V superlattices (SLs). SLs consist of alternating thin layers of different materials and III-V SLs are widely used in nanoscale thermoelectric and optoelectronic devices. The key feature in SLs is that it contains many interfaces, which dictates thermal transport. As III-V SLs are often fabricated using well-controlled techniques and have high-quality interfaces, we develop a model with only one free parameter---the effective rms roughness of the interfaces---to describe its twofold influence: reducing the in-plane layer thermal conductivity and introducing thermal boundary resistance (TBR) in the cross-plane direction. Both the calculated in-plane and cross-plane thermal conductivity of SLs agree with a number of different experiments.

Finally, we study thermal conductivity of ternary III-V alloys. In modern optoelectronic devices, ternary III-V alloys are used more often than binary compounds because one can use composition engineering to achieve different effective masses, electron/hole barrier heights, and strain levels. Ternary alloys are usually treated under the virtual crystal approximation (VCA) where cation atoms are assumed to be randomly distributed and possess an averaged mass. This assumption is challenged by a discrepancy between different experiments, as well as the discrepancy between experiments and calculations. We use molecular dynamics (MD) to study the ternary alloy system as both atom masses and atom locations are explicitly tracked in MD. We discover that the thermal conductivity is determined by a competition between mass-difference scattering and the short-range ordering of the cations.