High-Throughput Screening of Aperiodic Superlattices Based on Atomistic Simulation-Informed Effective Medium Theory and Genetic Algorithm

•A modified effective medium theory (m-EMT) is established to predict thermal conductivity of random Si/Ge superlattices.•Combination of m-EMT and genetic algorithm achieves high-throughput screening of superlattices for minimum conductivity.•Thermal conductivities of aperiodic superlattices can be constantly reduced to 1.5 W/(m·K) regardless of the total thickness.•The intense Anderson localization occurs at interfaces sandwiched between the highly asymmetric thick and thin layers. Superlattices (SLs) have received great attention as thermoelectric materials because phonon transport in them can be tailored independently without changing the electronic properties to enhance the thermoelectric figure of merit and conversion efficiency. Heterogeneous aperiodic SLs (ap-SLs) have lower thermal conductivity than their periodic counterparts (p-SL) due to Anderson localization of coherent phonons. Due to enormous amounts of layered structure combinations in ap-SLs, it is challenging to efficiently screen through all the combinations for the structure with the minimum thermal conductivity. In this work, a modified effective medium theory (m-EMT) is established to predict the thermal conductivity of Si/Ge p-SLs and ap-SLs. The effects of Anderson localization and layer thickness distribution are considered using a correction function for the degree of randomization (DOR) in layer thickness. The combination of the m-EMT and the genetic algorithm enables high-throughput screening of enormous amounts (∼ 2n, where n is the number of atomic layers) of SL structures with micron-scale total SL thickness. The thermal conductivities of ap-SLs can be constantly reduced to 1.5 W/(m·K) at average periodic layer thicknesses of 2.0 nm and a DOR of 0.8, regardless of the total SL thicknesses. Phonon spatial localization energy distribution further reflects the intense Anderson localization at interfaces, especially for those sandwiched between the highly asymmetric thick and thin layers. This work provides a high-throughput screening tool to theoretically design micron-scale and heterogeneous SLs, with detailed structural guidance for experimentally growing ap-SL with minimum thermal conductivities. [Display omitted]