Thermal conductivity of silicon using reverse non-equilibrium molecular dynamics

Abstract

Simulations are performed using the reverse non-equilibrium molecular dynamics (rNEMD) method and the Stillinger-Weber (SW) potential to determine the input parameters for achieving ±1% convergence of the calculated thermal conductivity of silicon. These parameters are then used to investigate the effects of the interatomic potentials of SW, Tersoff II, Environment Dependent Interatomic Potential (EDIP), Second Nearest Neighbor, Modified Embedded-Atom Method (MEAM), and Highly Optimized Empirical Potential MEAM on determining the bulk thermal conductivity as a function of temperature (400–1000 K). At temperatures > 400 K, data collection and swap periods of 15 ns and 150 fs, system size ≥6 × 6 UC2 and system lengths ≥192 UC are adequate for ±1% convergence with all potentials, regardless of the time step size (0.1–0.5 fs). This is also true at 400 K, except for the SW potential, which requires a data collection period ≥30 ns. The calculated bulk thermal conductivities using the rNEMD method and the EDIP potential are close to, but lower than experimental values. The 10% difference at 400 K increases gradually to 20% at 1000 K.

Machine learning based prediction of lattice thermal conductivity for half-Heusler compounds using atomic information

“…Therefore, it is necessary to know the exact lattice thermal conductivity for thermal management in half-Heusler devices. Theoretical predictions of the lattice thermal conductivity of solids can be made using non-equilibrium molecular dynamics simulations 1 – 7 or the density functional theory (DFT) calculations 8 – 16 . In recent years, with the advances in machine learning (ML) algorithms, several studies on understanding the heat transport properties of functional materials have been reported 17 – 21 .…”

“…Therefore, it is necessary to know the exact lattice thermal conductivity for thermal management in half-Heusler devices. Theoretical predictions of the lattice thermal conductivity of solids can be made using nonequilibrium molecular dynamics simulations [1][2][3][4][5][6][7] or the density functional theory (DFT) calculations [8][9][10][11][12][13][14][15][16]. The prediction of thermal conductivity by nonequilibrium molecular dynamics requires an enormous amount of computational time because time evolution must be calculated for numerous atomic movements.…”

Molecular Dynamics Simulation of Improving the Physical Properties of Polytetrafluoroethylene Cable Insulation Materials by Boron Nitride Nanoparticle under Moisture-Temperature-Electric Fields Conditions

“…In order to study the influence of BN nanoparticles on the thermal conductivity of PTFE insulating materials, the thermal conductivity of the PTFE model, the BN/PTFE model of tetrafluorosilane grafting, and the BN/PTFE model of non-grafting at different temperatures were calculated by the reverse non-equilibrium molecular dynamics (RNEMD) method [37,38,39]. The specific steps for this are as follows: Firstly, the PTFE model was divided into 40 layers along the z -axis direction, which was defined as having a hot layer at both ends and a cold layer in the middle; secondly, the atoms with the lowest kinetic energy in the hot layer will exchange energy periodically with those with the highest kinetic energy in the cold layer, and the energy exchange process is shown in Figure 8; finally, the exchange step was set to 0.1 ps, and the exchange frequency is 1000.…”

Molecular Dynamics Study of Thermal Conductivity of Silver Chalcogenides

“…[ 11 ] It is analogous to the experimental setting, but much larger computational system is needed to reproduce a reasonable temperature gradient. [ 12–14 ] In the EMD simulation using the Green–Kubo (GK) formula, the ionic thermal conductivity is calculated by applying the linear response theory in a homogeneous equilibrium system. [ 15 ] However, as is generally known, long simulation time is required to obtain thermal conductivity with high statistical accuracy.…”