Multi-layered stacking structures and atomic mixing interfaces were constructed. The effects of various factors on the thermal conductivity of different lattice structures were studied by non-equilibrium molecular dynamics simulations, including the number of atomic mixing layers, temperature, total length of the system, and period length. The results showed that the mixing of two and four layers of atoms can improve the thermal conductivities of the multi-layer structure with a small total length due to a phonon “bridge” mechanism. When the total length of the system is large, the thermal conductivity of the multi-layer structure with atomic mixing interfaces decreases significantly compared with that of the perfect interfaces. The interfacial atom mixing destroys the phonon coherent transport in the multi-layer structure and decreases the thermal conductivity to some extent. The thermal conductivity of the multi-layer structure with perfect interfaces is significantly affected by temperature, whereas the thermal conductivity of the multi-layer structures with atomic mixing is less sensitive to temperature.

The non-equilibrium molecular dynamics (NEMD) method is used to study the thermal conductivities of Si/Ge superlattices with tilted interface under different period lengths, different sample lengths, and different temperatures. The simulation results are as follows. The thermal conductivity of Si/Ge superlattices varies nonmonotonically with the increase of interface angle: when the period length is 4–8 atomic layers, the thermal conductivity for the interface angle of 45° is one order of magnitude larger than those for other interface angles, and the thermal conductivity increases linearly with the sample length increasing and decreases with the temperature increasing. However, when the period length is 20 atomic layers, the thermal conductivity is weakly dependent on sample length and temperature due to the existence of phonon localization.