Thermal conductivity of van der Waals heterostructure of 2D GeS and SnS based on machine learning interatomic potential

Abstract

The van der Waals heterostructure has provided an unprecedent platform to tune many physical properties for two-dimensional materials. In this work, thermal transport properties of van der Waals heterostructures formed by lateral stacking of monolayer GeS and SnS have been investigated systematically based on machine learning interatomic potential. The effect of van der Waals interface on the lattice thermal transport of 2D SnS and GeS can be well clarified by introducing various stacking configurations. Our results indicate that the van der Waals interface can strongly suppress the thermal transport capacity for the considered heterostructures, and either the average thermal conductivity per layer or the 2D thermal sheet conductance for the considered heterostructures is lower than that of corresponding monolayers. The suppressed thermal conductivity with tunable in-plane anisotropy in SnS/GeS heterostructures can be ascribed to the enhanced interface anharmonic scattering, and thus exhibits obvious interface-dependent characteristics. Therefore, this work highlights that the van der Waals interface can be employed to effectively modulate thermal transport for the 2D puckered group-Ⅳ monochalcogenides, and the suppressed lattice thermal conductivity together with interface-dependent phonon transport properties in the SnS/GeS heterostructure imply the great potential for corresponding thermoelectrical applications.

Machine Learning Enabled Fast Optical Identification and Characterization of 2D Materials

“…Once a sufficient set of optical images is labeled, a trained machine learning model can automatically classify the layers. As reviewed by Mao et al [27], previous studies have successfully implemented machine learning algorithms for automatic layer characterization [32] and other material properties [33][34][35] such as mechanical strength [36]. In our work, we systematically tested various approaches to find the best generalized solution for 2D material thickness characterization.…”