This work consists in an experimental study of the main parameters governing the transport of heat across an interface between two solids, when it is dominated by phonons. The main experimental tool used is Time-Domain ThermoReflectance (TDTR), which is an optical technique able to measure the Thermal Boundary Conductance (TBC) of single interfaces between metallic layers about 100 nm thick and mm-sized, flat dielectric substrates. Emphasis is put on diamond as a dielectric, because the diamond/metal TBC is among the important parameters governing the thermal conductivity of diamond-based Metal Matrix Composites (MMC) with potential applications as heat sink materials. In a first part, TDTR as a technique is reviewed and its main advantages are identified. Experimental parameters that influence it the most strongly are the size of the laser spots used and the thickness of the metallic layer investigated. An experimental procedure is presented, that provides a reliable way of measuring these quantities. In a second part, the effect on the TBC of surface treatments of the dielectric substrate prior to the metallic layer deposition is investigated. On an AlN substrate, it is found that the formation of a native oxide on its surface decreases the thermal transport across its interface with an Al layer. Etching away this oxide improves the measured TBC from 43 to 241 MWm−2K−1. On a diamond substrate, the results are more nuanced. The as-received surface is found to be hydrogen-terminated and has traces of organic contaminants. Ridding the surface of these contaminants and terminating it with hydrogen using an Ar:H plasma treatment improves TBC in a way that depends on the diamond surface orientation, from 23 to 32 ([111] orientation) or 23 to 54 ([100] orientation) MWm−2K−1. Doing so using a mixture of H2SO4 and HNO3 (1:1) at 200°C partly terminates the surface with oxygen and yields a TBC of 125 MWm−2K−1. Forcing a graphitic sp2 termination of the surface by sputtering Ar ions on it further improves TBC, up to 150 MWm−2K−1. Finally, forcing oxygen termination using an Ar:O plasma yields a high TBC of 180 MWm−2K−1. Still on the oxygenated surface, the deposition of the Al layer by DC sputtering instead of evaporation results in the highest TBC value measured between Al and diamond, of 230 MWm−2K−1, an order of magnitude higher than the TBC of 23 MWm−2K−1 between Al and as-received diamond, and thrice as large as the highest value of 90 MWm−2K−1 reported in the literature. All the treatments applied except for the Ar:H plasma are found to yield TBCs independent of the diamond surface orientation. In a third part, some attempts are made to explain the origin of the observed increase in TBC. By comparing the temperature dependence of the TBC between Al on oxygenated and sp2-terminated diamond, we find that the latter case leads to a distinctly different temperature-dependent behavior. This suggests that the presence of oxygen atoms, presumably forming an interfacial oxide, is the main factor influencing TBC in this system. To confirm this hypothesis, the thickness of this presumed oxide is increased by depositing Al2O3 using Atomic Layer Deposition (ALD) of 1.6, 4.5, 6.2 and 10 nm on the substrate before the Al layer. It is found that the difference in temperature dependence of the TBC between the case where these layers are present and the case when the diamond surface is only oxygen-terminated can be explained simply by the resistive contribution of the deposited layer. This suggests that interfacial states created by the reaction of Al and O are a crucial parameter explaining the high TBC found. In another experiment, the work of adhesion of Cu and Ni on clean diamond is calculated using Density Functional Theory. Using known values for the work of adhesion between Al and Ti on diamond, the TBC between diamond and Al, Cu, Ni or Ti is found to scale roughly with the work of adhesion of each of these systems. Furthermore, the presence of hydrogen at the diamond surface is found to dramatically decreases both the work of adhesion and the thermal transfer across its interface with these metals, confirming the previous results obtained between Al and diamond. Still in the same approach, the TBC between diamond and Cr, Mo, Nb and W is measured before and after a high vacuum heat treatment at 800°C. After careful characterization of the samples both before and after heat treatment, it is found that carbides have formed in the metal layers after heat treatment. Despite the presence of these carbides, the TBC between the diamond substrate and the layers stays the same, suggesting that even before the heat treatment, an interfacial carbide is present between the metal and the diamond and governs heat transfer through the interface. In a final experiment, very thin layers of Ni (of 1.8, 3.2 and 7 nm) are inserted between Al or Ag and diamond. Once these layers have been carefully characterized, it is found that these interfacial layers increase the measured TBC between the top layer and diamond to values that, within error and accounting for electron-phonon coupling, equal 260 MWm−2K−1 measured for a layer of pure Ni on diamond, up from TBCs of 47 and 170 MWm−2K−1 measured in the same study for Ag/diamond and Al/diamond interfaces respectively. Finally, throughout this whole thesis, values for TBC between several metals (Ag, Al, Au, Cr, Cu, Mo, Nb, Pt, Ti, W) as well as some carbides (Cr, Mo, Nb, W) and diamond are provided. They are measured in clean conditions and thus are expected to set a maximum value to the TBC in non-ideal systems such as diamond reinforced metals with potential applications as heat sinks. Based on the results presented, solutions to improve thermal transport in diamond-based MMCs are proposed.
EPFL2013
