Machine-Learning Accelerated Discovery of High-Performance Thermal Switch in two-dimensional Materials Considering High-order Anharmonicity

Heat dissipation and thermal switches are vital for adaptive cooling and extending the lifespan of electronic devices and batteries. In this work, we conducted high-throughput investigations on the thermal transport of 24 experimentally realized 2D materials and their potential as thermal switches, leveraging machine-learning-assisted strain engineering and phonon transport simulations. We identified several high-performance thermal switches with ratios exceeding 2, with germanene (Ge) achieving an ultrahigh ratio up to 9.64 within the reversible deformation range. The underlying mechanism is the strain-induced bond softening, which sensitively affects anharmonicity represented by three- and four-phonon scattering. The widespread four-phonon scattering was confirmed in the thermal transport of 2D materials. Opposite switching trends were discovered, with 2D TMD materials showing negative responses to tensile strain and buckled 2D elemental materials showing positive responses. We further proposed a screening descriptor based on strain-induced changes in the Grüneisen parameter for efficiently identifying new high-performance thermal switch materials. This work establishes a paradigm for thermal energy control in 2D materials through strain engineering, which may be experimentally realized in the future via bending, substrate mismatch, and related approaches, thereby laying a robust foundation for further developments and applications.