Molecular Dynamics Simulation Of Heat Transport Properties Of Carbon Based Low-dimensional Nano Materials

Thermal conductivity is an important property of materials and has a wide range of applications.It is a key parameter to determine the energy conversion efficiency of thermoelectric materials.To explore the heat transfer mechanism of micro and nano-scales and design new controllable heat transfer nano-devices have become the focus and frontier direction of heat transfer research in recent years.To solve the heat transport problems such as miniaturization,high integration and high heat flux density of chips and devices,it is necessary to face a series of complex processes such as cross-scale and multi-level heat generation,heat transfer,and heat dissipation.Due to the difficulty of experimental measurement of nanoscale heat conduction,there is an urgent need for calculation methods to assist experimental research,which also promotes the progress and improvement of simulation calculation methods.Carbon-based low-dimensional nanomaterials have attracted extensive attention and research due to their attractive structures and properties.In this paper,molecular dynamics simulations are used as a means,starting from the research of simulation methods,to representative interfaces,superlattices and complex structures with interlayer van der Waals interactions,and then to machine learning potential functions.The theoretical simulations of heat transport,the mechanism of heat transport and the application of machine learning potentials in the field of heat transport simulations are studied comprehensively and systematically.Firstly,in terms of theoretical simulation method,the latest application of equilibrium molecular dynamics(EMD)simulations for thermal conductivity calculation in finite system is proposed,and a properly physical interpretation is given.For a finite length system with fixed or open boundary conditions in the transport direction,a maximum thermal conductivity can be obtained using the EMD simulations due to the scattering of the boundary.By comparing the results from the nonequilibrium molecular dynamics(NEMD)simulations,the equivalence relationship between the two methods in finite systems is established.The successful implementation of the EMD method in finite systems can compensate for situations where the NEMD method is not directly applicable,such as characterizing the thermal conductivity of finite nanoparticles.Secondly,the heat transport properties of interfaces,superlattices,and complex structures with interlayer Van der Waals interactions are investigated.For the in-plane heterostructure Graphene/h-BN interface,the interfaces with different tilt angles under lattice matching and lattice mismatch conditions are constructed based on the phase-field crystal(PFC)model.The interfacial thermal conductance varying with tilt angles is obtained by combining NEMD simulations and spectral decomposition method.At the same time,the classical spectral thermal conductance is quantum corrected.Due to the small lattice difference between Graphene and h-BN,the lattice match and lattice mismatch at the interface of Graphene/h-BN heterostructures have similar hindering effects on phonon transport.For homogeneous Graphene grain boundary superlattices,the intrinsic thermal conductivities of Graphene grain boundary superlattices with a series of tilt angles in the diffusion region are systematically studied by using an efficient homogeneous nonequilibrium molecular dynamics(HNEMD)simulation.It is found that there is a minimum thermal conductivity induced by two competing transport mechanisms of phonon coherence and grain boundary scattering in the Graphene grain boundary superlattice.It is noted that there is also a minimum thermal conductivity along the grain boundary due to phonon coherence.For the heterogeneous structure of fullerene(C60)encapsulated Single-walled carbon nanotubes(SWCNTs)with interlayer Van der Waals interaction outside the plane,EMD,NEMD and HNEMD simulation methods consistently predicted that the thermal conductivity of SWCNTs will decrease by 20-30%after C60 encapsulation,which is consistent with the experimental results of carbon nano-bundles.It has successfully resolved the dispute on theoretical simulation and experimental measurement of C60 encapsulated SWCNTs for a long time.It reproduces the previously theoretically calculated artefact of the increased thermal conductivity of C60 encapsulated SWCNTs,and explains that C60 introduces additional phonon scattering to reduce the phonon mean free path of SWCNTs,which will lead to the decrease of the thermal conductivity.Finally,the application of machine learning potential in the field of heat transport.Taking two-dimensional C-N material as the research object,from the preparation of the DFT training set to the NEP machine learning potential training,then to potential function testing and evaluation,and finally the NEP potential is used to specifically study its thermal transport related performance.A set of feasible schemes for studying heat transport performance based on NEP machine learning potential are provided.In additions,it is found that phonon scattering due to large holes in the C2N structure leads to the expansion of the length range of periodic unit with the minimum thermal conductivity of coherent transport in the superlattice.This phenomenon is very conducive to the more efficient and controllable application of superlattices in thermoelectric conversion.