Size-effect induced thermal conductivity reduction in silicon thin films (TFs) has been broadly studied in last two decades [1]. Ultra-pure, crystalline silicon TFs, present in silicon-on-insulator (SOI) technology, have become a standard platform for thermal metrology. Due to the high quality of the SOI device layer, the large variations in existing data for the thickness-dependent thermal conductivity of silicon thin films is unexpected. Early data, using electrothermal measurements, required the silicon to be coated with multiple layers of dielectrics and metals – for mechanical support, electrical insulation, and to serve as a heater/thermometer for the measurements. In these measurements, interfacial effects and parallel heat conduction pathways complicate the heat transfer analysis and may explain the large variation. Recently, optothermal measurements of free standing silicon thin films from Chávez et al [2] and Cuffe et al [3] provide new thermal conductivity data across a large range of thicknesses (from ~10 nm to ~1 micron), without the presence of any additional metal or dielectric films. Thus, the impact of interfaces and parallel heat conduction pathways are avoided in these non-contact measurements. But still the variation in the measured data exceeds the expected uncertainty of the measurement techniques and the impact of interfaces on previous electrothermal measurements is still not well understood.

Here, we reinvestigate thermal conductivity of silicon TFs using a modified electrothermal measurement technique. Specifically, we develop a differential cross-beam measurement structure, similar to the T-type measurements used for fibers and nanotubes, to extract simultaneously the thermal conductivity of silicon films with and without coated metal and dielectric layers. In addition to interrogating the interface effects on thermal transport, thickness-dependent thermal conductivity data extracted by our electrothermal measurements is compared to the recent non-contact measurements to understand variations between measurement techniques. Finally, the thickness-dependent thermal conductivity is used to characterize the spectral dependent phonon transport. The phonon mean free path (MFP) distribution is obtained based on the analytical relation between film thickness and MFP suppression. Comparison of the accumulation functions, reconstructed using suppression function and atomistic simulations, sheds light on the phonon spectral dependence. This method is useful considering the complexity of accurately capturing anharmonic phonon transport. Results of this work will facilitate improved understanding of thermal transport and comparison of data between different measurement techniques.

References: [1] Marconnet, Amy M., Mehdi Asheghi, and Kenneth E. Goodson. "From the Casimir limit to phononic crystals: 20 years of phonon transport studies using silicon-on-insulator technology." Journal of Heat Transfer 135.6 (2013): 061601. [2] Chávez-Ángel, E., et al. "Reduction of the thermal conductivity in free-standing silicon nano-membranes investigated by non-invasive Raman thermometry."APL Materials 2.1 (2014): 012113. [3] Cuffe, John, et al. "Reconstructing phonon mean free path contributions to thermal conductivity using nanoscale membranes." arXiv preprint arXiv:1408.6747 (2014).