Revealing the scattering behaviour of gas molecules on porous surfaces is essential to develop accurate boundary conditions for kinetic transport models that describe the gas dynamics in shale reservoirs. Here, we use high-fidelity molecular dynamics simulations to resolve the gas–surface interactions between methane molecules and realistic organic kerogen surfaces, and to assess the applicability of the widely used scattering kernels. Our results show that the tight matrix porosities have a negligible effect on the timescale and lengthscale of the scattering process, which can be considered instantaneous in time and local in space. Although reflected velocity distributions reveal that the common Maxwell, Cercignani–Lampis and Yamamoto scattering models fail to fully capture the scattering details of methane on kerogen, especially when the incident molecular speeds are high, the Maxwell model predicts best the reflected angular beam pattern and the overall reflected velocity distribution for rough kerogen surfaces. However, for low-speed impingement, more characteristic of shale applications, all scattering models give similar velocity distributions, which are driven by the high degree of gas–surface accommodation observed. We find that a Maxwell model with a calibrated tangential momentum accommodation coefficient, which approaches unity as the surface roughness increases to ∼2 nm, is enough to reproduce comparable velocity profiles and mass flow rates inside moderately confined kerogen mesopores. Deviations between the Maxwell model and our molecular simulations are only observed for highly rarefied transport problems, but this rarefaction lies beyond the realm of shale reservoir applications. This paper, therefore, reports the first scattering study on porous and rough kerogen surfaces, and demonstrates the applicability of the Maxwell model, which can be readily incorporated into gas kinetic solvers to predict the apparent permeability of shale with mesopore and macropore networks.
As the scattering physics for kerogen/mineral surfaces was poorly understood, the surface accommodation coefficient of the molecular reflection process (e.g., TMAC) used in the previous kinetic simulations and the empirical permeability models is often treated as a convenient tuneable parameter to fit the experiment data, with used values ranging from 0.01 to 1.0. Given the fact that TMAC approaches unit at atomic-scale roughness of about 2 nm, as also reported on the study of other surfaces, the fully-diffuse Maxwell scattering kernel may be appropriate for modelling gas transport near kerogen or other mineral surfaces. Further investigations will be required when calibration studies with the experiments reveal TMACs much lower than 1.0, for which to be true, there has to be a high concentration of very smooth non-organic surfaces, which we believe is unlikely in shale rock.