Apr 25, 2024
2:30pm - 2:45pm
Room 327, Level 3, Summit
Meng An1,Wenyang Ding1,Jiang Guo1,Junichiro Shiomi1
The University of Tokyo1
Meng An1,Wenyang Ding1,Jiang Guo1,Junichiro Shiomi1
The University of Tokyo1
Engineering materials with extremely low thermal conductivity are crucial and desirable for many applications such as thermal management of electronic devices, thermal insulation and thermoelectric conversion. Recently, two-dimensional (2D) materials with heteroatoms from multiple distinct materials blocks show excellent properties and are considered as the promising candidates for constructing integrated circuits and quantum computers. However, Rayleigh's law implies that acoustic phonons, which carry most of the heat, are insensitive to scattering by atomic-scale defects. Therefore, realizing low thermal conductivity of lateral heterostructure with complex composition and diverse arrangement remains challenging due to the high degree of freedom and complex objectives. Herein, we develop a hybrid materials informatics approach which combines the variational autoencoder and Bayesian optimization to design two-dimensional patterned graphene/h-BN lateral heterostructure with extreme low thermal conductivity. With only several hundred training data sets, new structure with extremely low thermal conductivity can be quickly determined in a compressed latent space by calculating far less than 0.0001% of the total candidate structures, which greatly decreases the design period and cost. More interestingly, the thermal conductivity of newly obtained structure is much lower than that of randomly arranged graphene/h-BN lateral heterostructure, where the extracted phonon relaxation time of newly obtained structure from phonon energy spatial density are calculated to demonstrate the validity of Rayleigh's law. Finally, the mode-resolved atomistic Green's function (AGF) method was utilized to calculate the phonon modal transmission of newly obtained structure and analyze the elastic phonon scattering behavior. This study not only can be easily extended to other thermal conduction metamaterials with higher dimensional features but would be beneficial to screening the phonon transport mechanism of two-dimensional heterostructures with multiple thresholds of thermal conductivity.