Nanostructured potential well/barrier engineering for realizing unprecedentedly large thermoelectric power factors

Abstract This work describes through semiclassical Boltzmann transport theory and simulation a novel nanostructured material design that can lead to unprecedentedly high thermoelectric power factors, with improvements of more than an order of magnitude compared to optimal bulk material power factors. The design is based on a specific grain/grain-boundary (potential well/barrier) engineering such that: i) carrier energy filtering is achieved using potential barriers, combined with ii) higher than usual doping operating conditions such that high carrier velocities and mean-free-paths are utilized, iii) minimal carrier energy relaxation after passing over the barriers to propagate the high Seebeck coefficient of the barriers into the potential wells, and importantly, iv) the formation of an intermediate dopant-free (depleted) region. The design consists thus of a ‘three-region geometry’, in which the high doping resides in the center/core of the potential well, with a dopant-depleted region separating the doped region from the potential barriers. It is shown that the filtering barriers are optimal when they mitigate the reduction in conductivity they introduce, and this can be done primarily when they are ‘clean’ from dopants during the process of filtering. The potential wells, on the other hand, are optimal when they mitigate the reduced Seebeck they introduce by: i) not allowing carrier energy relaxation, and importantly ii) by mitigating the reduction in mobility that the high concentration of dopant impurities cause. It is shown that dopant segregation, with ‘clean’ dopant-depletion regions around the potential barriers, serves this key purpose of improved mobility towards the phonon-limited mobility levels in the wells. Using quantum transport simulations based on the non-equilibrium Green’s function method (NEGF) as well as semi-classical Monte Carlo simulations we also verify the important ingredients and validate this ‘clean-filtering’ design.