Thermal transport properties of defective graphene/graphyne van der Waals heterostructures elucidated via molecular dynamics and machine learning

Abstract

Two-dimensional (2D) all-carbon van der Waals (vdW) heterostructures consisting of graphene and graphyne component layers are reported to have enormous application prospects. Understanding the thermal transport properties of such graphene/graphyne (G/GY) heterostructures is critical to control their performance and stability in prospective applications. In this study, using molecular dynamics simulations and a machine learning (ML) method, we investigate the thermal conductivity of pristine G/GY heterostructures and their defective counterparts. Our simulation results show a significant reduction in the thermal conductivity of G/GY heterostructures due to the presence of vacancies, which become more aggressive as the defect concentration increases. Besides the concentration, the distribution of defects is another important factor affecting the thermal conductivity of defective G/GY heterostructures. Moreover, the defect effect on the thermal conductivity of G/GY heterostructures is majorly determined by the defect characteristics of their graphene layer. Such an impact is found to originate from the changes in both phonon scattering and heat flux. Based on the ML method together with a transfer learning strategy, we also develop a convolutional neural network that can be used to quickly and effectively predict the thermal conductivities of massive possible structures of defective G/GY heterostructures.