Dhruv Singhal1 2 3 Dimitri Tainoff2 3 Jessy Paterson2 3 Meriam Ben-Khedim2 3 4 Pascal Gentile1 Laurant Cagnon2 3 Daniel Bourgault2 3 Olivier Bourgeois2 3 Denis Buttard1 3 5

1, Université Grenoble Alpes, CEA INAC-Pheliqs- SiNaPS, F-38000, Grenoble, , France
2, Institut Néel, CNRS, Grenoble, , France
3, Université Grenoble Alpes, Grenoble, , France
4, STMicroelectronics, Rousset, , France
5, Université Grenoble Alpes/IUT, Grenoble, , France

Thermal conductivity in semiconductors is dominated by phonons that have a broad spectral distribution at room temperature. The rate of Umklapp processes dominates the phonon scattering at room temperature; it scales with ω2 (ω is the phonon frequency). This implies that low frequency acoustic phonons will have longer mean free path that can be reduced by rational incorporation of phonon scattering elements. For instance, when the dimensions of the material are comparable to the mean free path of the phonons, the thermal conductivity drops down significantly due to surface scattering mechanism.
In this work, we have fabricated an array of nanowires (silicon and various bismuth telluride) in nanoporous alumina templates with very high nanowire density (≈ 109/cm2). The fabrication is performed by pulse electrodeposition for Bi2Te3 alloys and by Chemical Vapour Deposition for silicon nanowires. We have employed 3ω method to extract the thermal conductivities of the nanowires as it is a well-developed technique to measure the thermal conductivity of thin films and bulk materials. The 3ω method is often employed with a simplified approximation and this, in turns, brings several limitations on its applicability especially in the case of thin nanoporous materials. For using it in our case, we have developed a model using an analytical solution of 2D heat conduction on a multilayer system to extract the thermal conductivity of the whole array of nanowires embedded in porous alumina, which is an anisotropic film. Fitting of the experimental data with the model considering all the different layers on the sample is done to extract the correct value of thermal conductivities of the embedded nanowires. The variation in thermal conductivities as a function of temperature (ranging from 80 K to 350 K) has been obtained. The results are correlated with the geometry (roughness) and composition of the nanowires. High-resolution TEM imaging along with the composition analysis with EDX gives an insight of the variation in phonon transport. The difference in thermal conductivities between nanowire and bulk is significant in silicon showing a strong confinement effect on silicon. A thermal conductivity around 10 W/mK is observed in 60 nm diameter nanowires. The thermal conductivity is reduced with the reduction in nanowire diameter, which shows stronger confinement in thinner silicon nanowires. Comparison of the thermal conductivity in between silicon and ternary alloys of Bi2Te3 with reduction in dimensions is clearly interpreted. By analysing the different parameters (Seebeck effect, electrical conductivity etc.), it can be seen that the array of highly-doped thin nanowires of silicon embedded in nanoporous alumina would result in highly-efficient thermoelectric devices.