A deep neural network potential model for theoretically predicting thermal transport, mechanical properties of multi-layered graphitic carbon nitride with molecular dynamics

Abstract

Graphitic carbon nitride (

) as a kind of important 2D materials, shows great potential applications as electric and photoelectric devices. However, most reported works mainly focus on the thermal transport characteristics of monolayer

, with bulk

ignored, which also has many important applications. Herein, anisotropic thermal conductivity of bulk

is investigated with the help of molecular dynamics. Considering the unique layer-by-layer structural characteristics, a deep neural network potential (DNNP) is developed, for accurately describing the interlayer vdw interactions among different C-N layers. It has been demonstrated that DNNP can accurately describe the atomic interactions in

, which is the weakness of classical potential fields. With the help of DNNP, phonon density of states, thermal conductivities, mechanical properties of monolayer, Bi-layer and Tri-layer

are fully investigated, and the relationship among atomic structures, structural symmetry and thermal conductivity is discussed, and mechanisms on vdW interactions making difference to thermal conductivity of

are illustrated on. The anisotropic thermal conductivity of bulk

is evaluated, showing that the thermal conductivity perpendicular to the C-N planes is 1.40 Wm⁻¹K⁻¹ which is about 1/3 of that along with C-N plane (4.41 and 4.12 Wm⁻¹K⁻¹ in x and y direction, respectively).