Thermoelectric generators convert heat energy into electrical energy. Their high reliability and low weight make them ideal not only for remote and space power generation applications, but also for more common applications like waste heat recovery. Thermoelectric generators are often realized using semiconductor devices where the rapid progress during the last decades has opened numerous possibilities, both in the choice of the material system and in the design of the device geometry and doping profiles.
Currently the influence of graded alloys is being investigated. Additional driving forces due to the effective mass gradient are taken into account. The single material properties are used at certain areas in the device to optimize the electrical conductivity and the generation rate, as well as to minimize the thermal conductivity in order to minimize the internal lattice heat flux and, therefore, maximize the global figure of merit. To intensify these effects, the geometrical structure is optimized by an appropriate local doping profile using Minimos-NT in connection with the optimization framework SIESTA.
In order to achieve predictive simulation, an accurate model set has to be established and verified, and thermodynamically rigorous coupling mechanisms between the electrical and the thermo-mechanical subsystem are required. Since areas of high temperature are expected, state-of-the-art models and material parameters have to be re-examined and extended, if necessary, to extend their validity to high temperature ranges. In addition to the already available models for mobility, thermal conductivity, and heat capacitance, new models for the thermoelectric forces and temperature dependent recombination have to be introduced. Special attention will be paid to the contact models. Physically correct modeling of the boundary conditions must not disturb local energy balance at higher temperatures and has to be consistent with the models for isothermal conditions. The single parameter values will be extracted from Full Band Monte Carlo simulations and from data available in the literature as well as from measurement data.