Published online by Cambridge University Press: 21 February 2012
The thermoelectric figure of merit ZT depends on the thermal conductivity inverse. uperlattices with periodic thin layers were studied to obtain ZT > 1 due to phonon confinement between their layers. Unfortunately, their synthesis with ZT higher than 1 is hazardous due to lattice mismatches forming dislocations and cracks. Nanowires with low dimensionalities were also proposed. However, as the superlattices, they decrease the thermal conductivity in only one propagation direction. In experiments, these one-dimensional insulating materials usually fail to beat the lowest limit of amorphous Si (+/- 1 W/m/K). In this theoretical study, three-dimensional Ge quantum dot arrays in Si are proposed to obtain an extreme thermal-conductivity reduction. Two decrease effects are shown from a molecular supercrystal model. First, low phonon group velocities are computed by lattice dynamics. Second, near-field scattering is exalted assuming weak interface bonding. This prediction can lead to a significant ZT increase. Indeed, a thermalconductivity global minimum λ* = 0.009 W/m/K is predicted for a Si/Ge supercrystal with nanodot spacing of +/- 30 nm and Ge concentration of +/- 12.5 Ge at.%. This ultralow λ* is computed at 300 K assuming that all Ge nanodots are weakly bonded and scatter the phonons at the Si-Ge interfaces in the geometrical limit. Thermal conductivity evolution is analyzed with respect to the weakly-bonded Ge nanodot density.