Thermoelectric and structural characterizations of individual electrodeposited bismuth telluride nanowires

The thermoelectric properties and crystal structure of individual electrodeposited bismuth telluride nanowires (NWs) were characterized using a microfabricated measurement device and transmission electron microscopy. Annealing in hydrogen was used to obtain electrical contact between the NW and the supporting Pt electrodes. By fitting the measured Seebeck coefficient with a two-band model, the NW samples were determined to be highly $n$-type doped. Higher thermal conductivity and electrical conductivity were observed in a 52 nm diameter monocrystalline NW than a 55 nm diameter polycrystalline NW. The electron mobility of the monocrystalline NW was found to be about 19% lower than that of bulk crystal at a similar carrier concentration and about 2.5 times higher than that of the polycrystalline NW. The specularity parameter for electron scattering by the NW surface was determined to be about 0.7 and partially specular and partially diffuse, leading to a reduction in the electron mean-free path from 61 nm in the bulk to about 40 nm in the 52 nm NW. Because of the already short phonon mean-free path of about 3 nm in bulk bismuth telluride, diffuse phonon-surface scattering is expected to reduce the lattice thermal conductivity of the 52–55 nm diameter NWs by only about 20%, which is smaller than the uncertainty in the extracted lattice thermal conductivity based on the measured total thermal conductivity and calculated electron thermal conductivity. Although the lattice thermal conductivity of the polycrystalline NW is likely lower than the bulk values, the lower thermal conductivity observed in this polycrystalline sample is mainly caused by the lower electron concentration and mobility. For both samples, the thermoelectric figure of merit $(ZT)$ increases with temperature and is about 0.1 at a temperature of 400 K. The low $ZT$ compared to that of bulk crystals is mainly caused by a high doping level, suggesting the need for better control of the chemical composition in order to improve the $ZT$ of the electrodeposited NWs. Moreover, bismuth telluride NWs with diameter less than 10 nm would be required for substantial suppression of the lattice thermal conductivity as well as experimental verification of theoretical predictions of power factor enhancement in quantum wires. Such stringent diameter requirement can be relaxed in other NW systems with longer bulk phonon mean-free path or smaller effective mass and thus longer electron wavelength than those in bulk bismuth telluride.