Experimental and theoretical investigation of thermal and thermoelectric transport in nanostructures
This work presents the development and application of analytical, numerical, and experimental methods for the study of thermal and electrical transport in nanoscale systems, with special emphasis on those materials and phenomena which can be important in thermoelectric and semiconductor device applications. Analytical solutions to the Boltzmann transport equation (BTE) using the relaxation time approximation (RTA) are presented and used to study the thermal and electrical transport properties of indium antimonide (InSb), indium arsenide (InAs), bismuth telluride (Bi₂Te₃), and chromium disilicide (CrSi₂) nanowires. Experimental results for the thermal conductivity of single layer graphene supported by SiO₂ were analyzed using an RTA-based model and compared to a full quantum mechanical numerical BTE solution which does not rely on the RTA. The ability of these models to explain the measurement results as well as differences between the two approaches are discussed. Alternatively, numerical solutions to the BTE may be obtained statistically through Monte Carlo simulation for complex geometries which may prove intractable for analytical methods. Following this approach, phonon transport in silicon (Si) sawtooth nanowires was studied, revealing that thermal conductivity suppression below the diffuse surface limit is possible. The experimental investigation of energy transport in nanostructures typically involved the use of microfabricated devices or non-contact optical methods. In this work, two such approaches were analyzed to ascertain their thermal behavior and overall accuracy as well as areas for possible improvement. A Raman spectroscopy-based measurement design for investigating the thermal properties of suspended and supported graphene was examined analytically. The resulting analysis provided a means of determining from measurement results the thermal interface conductance, thermal contact resistance, and thermal conductivity of the suspended and supported graphene regions. Previously, microfabricated devices of several different designs have been used to experimentally measure the thermal transport characteristics of nanostructures such as carbon nanotubes, nanowires, and thin films. To ascertain the accuracy and limitations of various microdevice designs and their associated conduction analyses, finite element models were constructed using ANSYS and measurements of samples of known thermal conductance were simulated. It was found that designs with the sample suspended were generally more accurate than those for which the sample is supported on a bridge whose conductance is measured separately. The effects of radiation loss to the environment of certain device designs were also studied, demonstrating the need for radiation shielding to be at temperatures close to that of the device substrate in order to accurately calibrate the resistance thermometers. Using a suspended microdevice like those analyzed using finite element analysis, the thermal conductivities of individual bismuth (Bi) nanowires were measured. The results were correlated with the crystal structure and growth direction obtained by transmission electron microscopy on the same nanowires. Compared to bulk Bi in the same crystal direction, the thermal conductivity of a single-crystal Bi nanowires of 232 nm diameter was found to be 3 - 6 times smaller than bulk between 100 K and 300 K. For polycrystalline Bi nanowires of 74 nm to 255 nm diameter the thermal conductivity was reduced by a factor of 18 - 78 over the same temperature range. Comparable thermal conductivity values were measured for polycrystalline nanowires of varying diameters, suggesting a grain boundary scattering mean free path for all heat carriers in the range of 15 - 40 nm which is smaller than the nanowire diameters. An RTA-based transport model for both charge carriers and phonons was developed which explains the thermal conductivity suppression in the single-crystal nanowire by considering diffuse phonon-surface scattering, partially diffuse surface scattering of electrons and holes, and scattering of phonons and charge carriers by ionized impurities such as oxygen and carbon of a concentration on the order of 10¹⁹ cm⁻³. Using a similar experimental setup, the thermoelectric properties (Seebeck coefficient, electrical conductivity, and thermal conductivity) of higher manganese silicide (HMS) nanostructures were investigated. Bulk HMS is a passable high temperature thermoelectric material which possesses a complex crystal structure that could lead to very interesting and useful nanoscale transport properties. The thermal conductivities of HMS nanowires and nanoribbons were found to be reduced by 50 - 60 % compared to bulk values in the same crystal direction for both nanoribbons and nanowires. The measured Seebeck coefficient data was comparable or below that of bulk, suggesting unintentional doping of the samples either during growth or sample preparation. Difficulty in determining the amorphous oxide layer thickness for nanoribbons samples necessitated using the total, oxide-included cross section in the thermal and electrical conductivity calculation. This in turn led to the determined electrical conductivity values representing the lower bound on the actual electrical conductivity of the HMS core. From this approach, the measured electrical conductivity values were comparable or slightly below the lower end of bulk electrical conductivity values. This oxide thickness issue affects the determination of the HMS nanostructure thermoelectric figure of merit ZT as well, though the lower bound values obtained here were found to still be comparable to or slightly smaller than the expected bulk values in the same crystal direction. Analytical modeling also indicates higher doping than in bulk. Overall, HMS nanostructures appear to have the potential to demonstrate measurable size-induced ZT enhancement, especially if optimal doping and control over the crystallographic growth direction can be achieved. However, experimental methods to achieve reliable electrical contact to quality four-probe samples needs to be improved in order to fully investigate the thermoelectric potential of HMS nanostructures.