Molecular-Dynamics-Derived Gas–Surface Models for Use in Direct-Simulation Monte Carlo

Nonreactive Scattering of N2from Layered Graphene Using Molecular Beam Experiments and Molecular Dynamics

“…For all incidence angles, more flexible surfaces lead to lower final translational energies. With the experimental validation of the MD derived post-collisional velocities, the DSMC direct velocity sampling method used in our previous work 28 can be used to analyze the mesoscale influence of gas−surface interactions for boundary layer dominated flows.…”

“…Further details regarding the surface generations are available in our previous work. 28 The dependence of collision statistics on the number of surface layers is shown in Figure 8a,b for the lower-and highervelocity trajectories, respectively, at different incidence angles. At lower incidence angles (closer to surface normal), the majority of simulated trajectories were single collision events for both incidence velocities.…”

“…When an incident N 2 molecule approaches the graphite surface, which was observed to have transient wave-like structures as a part of its topology during our simulations, it is exposed to a surface morphology that has a trough. This may cause it to interact with as many as 13 surface carbon atoms 28 and generate a larger repulsion which imparts additional kinetic energy to the gas particle during the collision and produces a scattered trajectory that is closer to the surface normal.…”

“…Finally, in our previous work, we took an ensemble average of our MD trajectory simulations of N 2 scattering from an HOPG surface to calculate the CLL accommodation coefficients. Note that, in this case, we did not have experimental data for comparison and used the −C-N– Lennard-Jones parameters in the ReaxFF potential that were obtained from the work by Kamat et al Similar to the work of Duan et al, we calculated values of α n = 0.62 and α t = 0.05, which are generally consistent with the trends seen in the experiment.…”

Modeling rarefied gas-solid surface interactions for Couette flow with different wall temperatures using an unsupervised machine learning technique

“…However, it is well known that MD simulations are computationally expensive. A new class of scattering kernels, known as nonparametric scattering kernels, were proposed in the literature [29][30][31]. The main idea behind developing such models is to overcome some intrinsic limits of the existing empirical GSI models, which are mainly caused by employing a limited number of fitting parameters (i.e., accommodation coefficients).…”