The effect of phonon-electron (p-e) scattering on lattice thermal conductivity is investigated for Cu, Ag, Au, Al, Pt, and Ni. We evaluate both phonon-phonon (p-p) and p-e scattering rates from first principles and calculate the lattice thermal conductivity (κ_L). It is found that p-e scattering plays an important role in determining the κ_L of Pt and Ni at room temperature, while it has negligible effect on the κ_L of Cu, Ag, Au, and Al. Specifically, the room temperature κ_Ls of Cu, Ag, Au, and Al predicted from density-functional theory calculations with the local density approximation are 16.9, 5.2, 2.6, and 5.8 W/m K, respectively, when only p-p scattering is considered, while it is almost unchanged when p-e scattering is also taken into account. However, the κ_L of Pt and Ni is reduced from 7.1 and 33.2 W/m K to 5.8 and 23.2 W/m K by p-e scattering. Even though Al has quite high electron-phonon coupling constant, a quantity that characterizes the rate of heat transfer from hot electrons to cold phonons in the two-temperature model, p-e scattering is not effective in reducing κ_L owing to the relatively low p-e scattering rates in Al. The difference in the strength of p-e scattering in different metals can be qualitatively understood by checking the amount of electron density of states that is overlapped with the Fermi window. Moreover, κ_L is found to be comparable to the electronic thermal conductivity in Ni.

Nonequilibrium molecular dynamics (NEMD) simulations on conceptual binary Lennard-Jones systems show that the thermal conductivity ($\kappa$) of a superlattice (SL) can be significantly reduced by randomizing the thicknesses of its layers, by which a SL becomes a random multilayer (RML). Such reduction in $\kappa$ is a clear signature of coherent phonon that can be localized in RMLs. We build a two-phonon model that divides the overall heat conduction into coherent and incoherent phonon contributions. In SL both coherent and incoherent phonons contribute to heat conduction, while in RML coherent phonons are localized so only incoherent phonons contribute. This model can fit the length dependence of the thermal conductances predicted in our NEMD simulations very well. The ballistic-limit thermal conductance and the intrinsic mean free path (MFP) of both coherent and incoherent phonons, and the localization length of coherent phonons, are obtained by fitting our model to the NEMD simulation results. The significant increase in $\kappa$ of SL with total length is due to the long MFP of coherent phonons, and the lower $\kappa$ of RML than SL is caused by the localization of coherent phonons.

We have used a two-temperature nonequilibrium molecular dynamics method for predicting interfacial thermal resistance across metal-nonmetal interfaces. This method is an extension of the conventional nonequilibrium molecular dynamics for the dielectric-dielectric interface, where a temperature bias is imposed and the heat current is derived. We have included the electron degree of freedom for the interfacial thermal transport problem by treating the electron-phonon coupling with the two-temperature model. The method is demonstrated on two model systems, that is, silicon-copper interface and carbon-nanotube\char21{}copper interface. Temperature nonequilibrium between electrons and phonons in the metal side is quantitatively predicted, and a temperature drop across the interface is observed. The results agree with experimental data better than those obtained from conventional nonequilibrium molecular dynamics simulations where only phonons are considered. Our approach is capable of taking into account both the electron and lattice degrees of freedom in a single molecular dynamics simulation and is a generally useful tool for modeling interfacial thermal transport across metal-nonmetal interfaces.