Complex dynamics and performance of inhomogeneous thermoelectrics
This thesis aims to illuminate complex dynamics and performance of inhomogeneous thermoelectrics by means of a theoretical approach. For our investigations, we develop a versatile network model, which is based on the phenomenological On\-sager-de Groot-Callen theory. This model allows us to study three setups related to the fabrication of a new type of thermoelectric generator. The first setup is related to the production of the generator's raw material. Therefore, the current-activated pressure-assisted densification technique is applied to create nanostructured bulk material. The network model is designed to account for the complex dynamics caused by the particle motion during the densification. In particular, we investigate the influence of elongated pores parallel to the electrodes, which lower the conductivity. Moreover, we investigate a self-organized assembly of particles in binary particle mixtures. With the second setup, we scrutinize the application of the Harman method to inhomogeneous material. The Harman method is often used to determine the thermoelectric transport properties. It turns out that this method employed on inhomogeneous material, systematically overestimates the absolute value of the Seebeck coefficient. It is demonstrated that the error is caused by the temperature distribution, which memorizes the influence of the priorly applied electrical current. Related to this, we show that the electrical power of double segmented generators is usually less than expected from the electrical conductivity and the true open circuit Seebeck coefficient. Nonetheless, we prove that by choosing transport parameters from a small range, a power enhancement can be obtained. Finally, we investigate the usage of a pn junction as a generator, where the temperature gradient is parallel to the pn interface and electrodes are attached on the cold side. The dismissal of hot side contacts facilitates the application in high temperature regimes. In a first step, the diode character of the interface is neglected. We investigate the reduction of the electrical power compared to a conventional device. Thereby, we determine a relation between power and current. Furthermore, geometrical optimizations are discussed. In a second step, diode characteristics are included into the model, which leads to a qualitative agreement to experimental results.