Developing silicon-based thermoelectrics for cooling applications

Abstract

Research on silicon thermoelectric coolers lies at the intersection of semiconductor physics, nanoscale heat transfer, industrial manufacturing, and device engineering. The electronic properties of doped silicon (σ≈50,000 S/m and S≈200 µV/K at 1020 cm-3) are highly desirable, but the intrinsic thermal conductivity is at least two orders of magnitude too high for thermoelectric applications. The phononic contributions to the thermal conductivity dominate in silicon and have mean free paths that span a wide range of length scales at room temperature. Conversely, electronic contributions to the thermal conductivity span a much narrower mean free path spectrum at smaller length scales. The thermoelectric potential of bulk silicon may be realized in nanoporous silicon (np-Si) that selectively impedes phonons. The task of minimizing thermal conduction, without significantly affecting the electronic transport, represents an opportunity to use recent scientific understanding of thermal transport in silicon for the important engineering application of cooling. Furthermore, the development of np-Si creates an opportunity for experimental measurements that may further the scientific understanding of nanoscale physics. This dissertation includes (i) a scalable fabrication process used to produce np-Si from degenerately-doped silicon powders, (ii) experimental measurements of the thermoelectric properties of the np-Si samples, (iii) microstructural and compositional characterization of the np-Si samples, (iv) a numerical model that applies the characterization results to predict the effective thermoelectric properties of np-Si, and (v) an augmentation of frequency-domain thermoreflectance to measure the thermal conductivity of anisotropic samples.