Femtosecond time-domain thermoreflectance measurements of cross-plane thermal conductivity of GaAs/AlAs superlattices

Nanostructuring is a proven method to reduce the thermal conductivity and increase the thermoelectric figure of merit of semiconductor materials for applications such as solid-state refrigeration. Accurate measurement of the thermal properties of nanostructured materials is necessary to utilize them properly. One technique to measure the cross-plane thermal conductivity of these materials is time-domain thermoreflectance (TDTR). In TDTR, the sample is coated with a thin metal layer which is heated by a laser pulse. As the metal layer cools by heat conduction into the sample, the change in optical reflectivity is detected by a probe laser. For temperature changes of a few kelvin, the change in probe reflectivity is directly proportional to the change in temperature of the metal. By comparing the experimental measurements to a simulated one-dimensional conduction heat flow analysis, the thermal conductivity of the sample and the thermal boundary resistance between the metal and the sample can be determined. In this thesis, we use TDTR to measure the cross-plane thermal conductivity of seven GaAs/AlAs superlattices with period thickness ranging from 2 nm to 41 nm. We compare the results from two TDTR setups, one uses a Ti:sapphire oscillator system with a repetition rate of 76 MHz and the other uses an amplifier system with a repetition of 5 kHz. Most of the thermal conductivity values from the amplifier measurements are consistent with the literature values. However, most of the thermal conductivity values from the oscillator measurements are larger than the literature values. The discrepancy is likely due to pulse accumulation. A sensitivity analysis of the thermal model shows a time delay of at least 2.5 ns should be used to minimize the uncertainty of the thermal boundary resistance measurement. Generally, increasing the time delay will further reduce the uncertainty of the thermal conductivity, but the diffusion length within the superlattice at the maximum time delay also has to be considered. The minimum superlattice thickness that can be measured within the 2.5 ns time delay without substrate interaction is about 120 nm for material with a diffusivity of 6x10⁻⁶ m²/s.Mechanical Engineerin