低维热电材料往往可以通过降低声子热导率实现其热电性能的提升。由两种材料交替生长获得的超晶格薄膜较单一材料薄膜具有更低的热导率，通过改变材料的厚度排布，随机排列的非周期性超晶格甚至可以实现更低的热导率。本文基于非平衡分子动力学模拟计算了硅和锗薄膜热导率和非对称界面热阻，构建了随机硅–锗超晶格热导率的数值拟合等效介质模型。引入邻间因子和修正函数后，获得了可以更为准确预测随机排列硅–锗超晶格热导率的修正等效介质模型。将此模型与遗传算法相结合，可以对大量随机超晶格结构进行高通量筛选，实现了热导率的快速优化。结果表明，即使总厚度大的超晶格最低热导率仍能维持在1.4∼1.8 W·m⁻¹·K⁻¹，平均周期厚度稳定在2.0∼2.5 nm。

The thermoelectric properties of low-dimensional materials can be improved by reducing the phonon thermal conductivity (TC). The superlattice films obtained by alternatively growing the two materials have lower TC than the single material films, especially for the randomly arranged aperiodic superlattice with special thickness arrangement. In this paper, TC of Si and Ge thin films and the thermal resistance of asymmetric interfaces are calculated by nonequilibrium molecular dynamics simulation (NEMD), and their numerical fitting models are constructed. After introducing adjacent factor and correction function, an effective medium theory (EMT) is obtained to accurately predict the thermal conductivity of randomly arranged Si-Ge superlattices. Combined with genetic algorithm, EMT can be used for high-throughput screening of aperiodic superlattice structure to realize thermal conductivity optimization. The results show that the minimum TC of even thick superlattices can be maintained in the range of 1.4∼1.8 W·m⁻¹·K⁻¹, and the average periodic thickness is stable in the range of 2.0∼2.5 nm.