Thermoelectric power devices provide an attractive possibility for directly converting heat energy to electricity. Their lack of moving parts results in a long lifetime and practically no need of maintenance. Due to the relatively low efficiency and power density of today's commercially available devices, they are generally only used in environments where their solid-state nature outweighs their poor efficiency. A prominent example is their use in satellites and spacecraft. Further increase in efficiency and proper customization of given thermal and geometrical conditions will open the wide range of combustion engine applications. Predictive simulation is an essential tool for thermoelectric device development and efficiency improvement.
An accurate model set for non-isothermal simulation incorporating large temperature gradients has been assessed and implemented in Minimos-NT. Therefore, the theoretical background has been elaborated between the two approaches of phenomenological irreversible thermodynamics and systematic derivation from Boltzmann's equation using the method of moments.
Every thermoelectrically relevant material defines its optimum figure of merit in a certain temperature range. While silicon and Si/Ge are good performers for relatively high temperatures, the lower temperature range can be covered by lead chalcogenides and other selected compound semiconductors. Relevant data of the important materials PbTe and PbSnTe has been collected from literature. Mobility data has been extracted from Monte Carlo simulations in addition to available measurement data. Proper models have been formulated and implemented in Minimos-NT. In order to estimate the accuracy of the simulation results carried out with Minimos-NT, a sensitivity analysis for several model parameters has been carried out. The band gap model's accuracy has turned out to be most important because of its big influence on carrier generation. A simulation study of conventional and large-area pn-junction thermoelectric generators has been carried out. The novel device structures benefit from carrier generation in the heated zones, and thus higher electrical currents can be obtained.
Several software tools working in conjunction with Minimos-NT have been developed and extended. The structure generation tool, initially limited to rectangular structures and Manhattan grids, was extended to non-rectangular device structures and proper meshes. The optimization framework SIESTA has been partially revised and extended by a user-friendly graphical interface using state-of-the-art graphics libraries.