"The performance of thermoelectric generation has long since been limited by the fact that it depends on hardly tunable intrinsic materials properties. At the heart of this problem lies a trade-off between sufficient Seebeck coefficient, good electrical properties and suitably low thermal conductivity. The two last being closely related by the ambivalent role of electrons in the conduction of both electrical and thermal currents. Current research focuses on materials composition and structural properties in order to improve this trade-off also known as the figure of merit (zT). Recently, evidences aroused that nanoscale structuration (nanowires, quantum dots, thin-films) can improve zT by means of electron and/or phonon confinement. The aim of this project is to tackle the intrinsic reasons for this low efficiency and bring TE conversion to efficiencies above 10% by exploring two unconventional and complementary approaches:Phononic Engineering Conversion consists of modulating thermal properties by means of a periodic, precisely designed, arrangement of inclusions on a length scale that compares to phonon means free path. This process is unlocked by state of the art lithography techniques. In its principles, phononic engineering offers an opportunity to tailor the phonon density of states as well as to artificially introduce thermal anisotropy in a semiconductor membrane. Suitable converter architecture is proposed that takes advantage of conductivity reduction and anisotropy to guide and converter heat flow. This approach is fully compatible with standard silicon technologies and is potentially applicable to conformable converters.The Micro Thermionic Conversion relies on low work function materials and micron scale vacuum gaps to collect a thermally activated current across a virtually zero heat conduction device. This approach, though more risky, envisions devices with equivalent zT around 10 which is far above what can be expected from solid state conversion."