| We study the mechanism of tunable heat transfer through interfaces between solids using a combination of non-equilibrium molecular dynamics simulation (NEMD), vibrational mode analysis and wave packet simulation. We investigate how heat transfer through interfaces is affected by factors including pressure, interfacial modulus, contact area and interfacial layer thickness, with an overreaching goal of developing fundamental knowledge that will allow one to tailor thermal properties of interfacial materials. The role of pressure and interfacial stiffness is unraveled by our studies on an epitaxial interface between two Lennard-Jones (LJ) crystals. The interfacial stiffness is varied by two different methods: (i) indirectly by applying pressure which due to anharmonic nature of bonding, increases interfacial stiffness, and (ii) directly by changing the interfacial bonding strength by varying the depth of the potential well of the LJ potential. When the interfacial bonding strength is low, quantitatively similar behavior to pressure tuning is observed when the interfacial thermal conductance is increased by directly varying the potential-well depth parameter of the LJ potential. By contrast, when the interfacial bonding strength is high, thermal conductance is almost pressure independent, and even slightly decreases with increasing pressure. This decrease can be explained by the change in overlap between the vibrational densities of states of the two crystalline materials. The role of contact area is studied by modeling structures comprised of Van der Waals junctions between single-walled nanotubes (SWCNT). Interfacial thermal conductance between SWCNTs is obtained from NEMD simulation as a function of crossing angle. In this case the junction conductance per unit area is essentially a constant. By contrast, interfacial thermal conductance between multiwalled carbon nanotubes (MWCNTs) is shown to increase with diameter of the nanotubes by recent experimental studies [1]. To elucidate this behavior we studied a simplified model comprised of an interface between two stacks of graphene ribbons to mimic the contact between multiwalled nanotubes. Our results, in agreement with experiment, show that the interfacial thermal conductance indeed increases with the number of graphene layers, corresponding to larger diameter and larger number of walls in MWCNT. The role of interfacial layer thickness is investigated by modeling a system of a few layers of graphene sandwiched between two silicon slabs. We show, by wave packet simulation and by theoretical calculation of a spring-mass model, that the transmission coefficient of individual vibrational modes is strongly dependent on the frequency and the number of graphene layers due to coherent interference effects; by contrast, the interfacial thermal conductance obtained in NEMD simulation, which represents an integral over all phonons, is essentially independent of the number of graphene layers, in agreement with recent experiments. Furthermore, when we heat one atomic layer of graphene directly, the effective interfacial conductance associated with heat dissipation to the silicon substrate is very small. We attribute this to the resistance associated with heat transfer between high and low frequency phonon modes within graphene. Finally, we also replaced graphene layers by a few WSe2 sheets and observed that interfacial thermal resistance of a Si/n-WSe2/Si structure increases linearly with interface thickness at least for 1 < n <= 20, indicating diffusive heat transfer mechanism, in contrast to ballistic behavior of a few graphene layers. The corresponding thermal conductivity (0.048 W m-1 K-1) of a few WSe2 layers is rather small. By comparing phonon dispersion of graphene layers and WSe2 sheets, we attribute the diffusive behavior of a few WSe2 sheets to abundant optical phonons at low and medium frequencies leading to very short mean free path. Our computational studies of effects of pressure and structural properties on interfacial thermal conductance provide fundamental insights for tunable heat transfer in nanostructures. [1] Professor D. Y. Li from University of Vanderbilt, private communication (Nov. 14, 2011). |
