Summary: | 博士 === 臺灣大學 === 機械工程學研究所 === 98 === Thermoelectrics receives great attention in the past decades because its efficiency barrier has been broken by the recently advanced nanoengineering. Nanostructures such as nanowire and superlattice are shown to have remarkably low thermal conductivity than their bulk counterparts, and this property is particularly favorable to the design of highly efficient thermoelectric (TE) devices. Encouraged by the TE advances since 1990, much experimental and theoretical research toward novel nanostructures of low thermal conductivity is currently undertaken.
This thesis explores the structure-dependent thermal conductivity of nanosized silicon porous materials and curved silicon nanowires. First, an approximate model is proposed to predict the thermal conductivity of porous silicon with aligned pores. This model, by taking both diffusive and ballistic transport mechanisms into consideration, is suitable for bulk as well as nanoscale materials. The effects of geometry are accounted for by introducing the geometry-dependent porosity functions and view factors. We also compared the model predictions with Monte Carlo simulation data under two pore characteristic sizes (100nm and 500nm) and various porosities (0.05~0.31). The agreement between the model predictions and the simulation results is excellent if the phonon mean free path is properly modeled. Next, a new factor, the wire curvature, is introduced for the nanowire case. In addition to the wire thickness and surface roughness, curvature is shown to decrease the effective thermal conductivity in nanoscale. For instance, a rough nanowire of curvature 30 nm has 80% thermal conductivity as compared with its straight counterpart.
|