Establishing a data-based scattering kernel model for gas–solid interaction by molecular dynamics simulation

Abstract

Scattering kernel models for gas–solid interaction are crucial for rarefied gas flows and microscale flows. However, most existing models depend on certain accommodation coefficients (ACs). We propose here to construct a data-based model using molecular dynamics (MD) simulation and machine learning. The gas–solid interaction is first modelled by 100 000 MD simulations of a single gas molecule reflecting on the wall surface, which is fulfilled by GPU parallel technology. The results showed a correlation of the reflection velocity with the incidence velocity in the same direction, and also revealed correlations that may exist in different directions, which are neglected by the traditional gas–solid interaction model. Inspired by the sophisticated Cercignani–Lampis–Lord (CLL) model, two improved scattering kernels were constructed to better reproduce the probability density of velocity determined from MD simulation. The first one adopts variable ACs which depend on the incidence velocity and the second one combines three CLL-like kernels. All the parameters in the improved kernels are automatically chosen by the machine learning method. Compared with the numerical experiments of a molecular beam, the reconstructed scattering kernels are basically consistent with the MD results.

Gas-Solid Interface Interactions Based on Molecular Dynamics Simulations

“…A limit should be placed on the simulation time to make the simulation much more efficient. In some scholars' studies, simulation time limits ranging from 20-50 ps [14,19,20] were used, and it was considered that if the gas molecules could not escape from the wall in such a long time, they were considered to be absorbed by the wall and could not escape. In this paper, the simulation time limit is set to 30 ps.…”

A hybrid Gaussian Mixture/DSMC approach to study the Fourier thermal problem

“…Nevertheless, in all these scattering kernels, the gas-gas interactions that can affect the reflected gas molecules properties in the early transition regime (0.1¡Kn¡1) are ignored, and these wall models cannot deal with adsorption-related problems. Machine learning is another promising technique that can be used to establish a gas scattering kernel directly based on the collisional data obtained from MD simulations [28][29][30][31][32]. As an example, in our previous works [29,31], the Gaussian mixture (GM) approach, an unsupervised machine learning approach, was employed to construct a scattering kernel for monoatomic gases (Ar, He) interacting with the Au surface and diatomic gases (H 2 , N 2 ) interacting with the Ni surface, respectively.…”

Boundary conditions for the Boltzmann equation from gas-surface interaction kinetic models

“…A main disadvantage of these boundary conditions, however, is that they are mathematical in nature, derived empirically, and not related to physical quantities like atom characteristics, crystal characteristics, or interaction potentials [1]. A more physical approach is to obtain information about the reflected molecules by the use of molecular dynamics (MD) simulations [7][8][9][10]. However, it is not easy in general to construct handy models of gas-solid interaction on the basis of the results of MD simulations.…”