Semiconducting large bandgap oxides as potential thermoelectric materials for high-temperature power generation?

Semiconducting large bandgap oxides are considered as interesting candidates for high-temperature thermoelectric power generation (700–1,200 °C) due to their stability, lack of toxicity and low cost, but so far they have not reached sufficient performance for extended application. In this review, we summarize recent progress on thermoelectric oxides, analyze concepts for tuning semiconductor thermoelectric properties with view of their applicability to oxides and determine key drivers and limitations for electrical and thermal transport properties in oxides based on our own experimental work and literature results. For our experimental assessment, we have selected representative multicomponent oxides that range from materials with highly symmetric crystal structure (SrTiO3 perovskite) over oxides with large densities of planar crystallographic defects (Ti n O2n−1 Magneli phases with a single type of shear plane, NbO x block structures with intersecting shear planes and WO3−x with more defective block and channel structures) to layered superstructures (Ca3Co4O9 and double perovskites) and also include a wide range of their composites with a variety of second phases. Crystallographic or microstructural features of these oxides are in 0.3–2 nm size range, so that oxide phonons can efficiently interact with them. We explore in our experiments the effects of doping, grain size, crystallographic defects, superstructures, second phases, texturing and (to a limited extend) processing on electric conductivity, Seebeck coefficient, thermal conductivity and figure of merit. Jonker and lattice-versus-electrical conductivity plots are used to compare specific materials and material families and extract levers for future improvement of oxide thermoelectrics. We show in our work that oxygen vacancy doping (reduction) is a more powerful driver for improving the power factor for SrTiO3, TiO2 and NbO x than heterovalent doping. Based on our Seebeck-conductivity plots, we derived a set of highest achievable power factors. We met these best values in our own experiments for our titanium oxide- and niobium oxide-based materials. For strontium titanate-based materials, the estimated highest power factor was not reached; further material improvement is possible and can be reached for materials with higher carrier densities. Our results show that periodic crystallographic defects and superstructures are most efficient in reducing the lattice thermal conductivity in oxides, followed by hetero- and homovalent doping. Due to the small phonon mean free path in oxides, grain boundary scattering in nanoceramics or materials with nanodispersions is much less efficient. We investigated the impact of texturing in Ca3Co4O9 ceramics on thermoelectric performance; we did not find any improvement in the overall in-plane performance of a textured ceramic compared to the corresponding random ceramic.