Ensemble-Dependent Diffusivity Selectivity in Amorphous Polyethylene: A Statistically Validated Molecular Dynamics Study of H ₂ /CH ₄ Transport

The integration of hydrogen into natural gas networks (H2NG) demands a rigorous understanding of relative gas transport properties in polymeric infrastructure to mitigate risks such as leakage and Rapid Gas Decompression (RGD). Polyethylene (PE) is a critical material for this transition due to its resistance to hydrogen embrittlement, yet its susceptibility to gas permeation remains a safety concern. In this work, we present a fully open-source, statistically validated Molecular Dynamics (MD) framework to quantify the diffusive transport of H2 and CH4 in amorphous polyethylene. Using the TraPPE-UA force field and a multistage high-pressure annealing protocol, we generated dense, well-equilibrated matrices (pure-polymer reference ρpure = 0.814 ± 0.001 g cm–3 and mix systems at ρMix,NPT = 0.803 ± 0.001 and ρMix,NVT = 0.819 ± 0.003 g cm–3) and characterized them using radial distribution functions, the full pore size distribution, and the standard porous-materials descriptors LCD (Largest Cavity Diameter) and PLD (Pore Limiting Diameter). Diffusion coefficients of H2 and CH4 were extracted from 48 independent production simulations (3 seeds × 2 systems × 2 ensembles × 4 temperatures) over the range 298–373 K under both isobaric (NPT, P = 1 atm) and isochoric (NVT) ensembles. Our central finding is an ensemble-dependent diffusivity-selectivity trade-off: thermal expansion in NPT conditions opens the free-volume network and reduces the H2/CH4 diffusivity selectivity αD from ∼4.8 at 298 K to ∼1.9 at 373 K, whereas volumetric confinement (NVT) suppresses the free-volume dilation and yields a far more gradual decrease (αD ≈ 4.6 → 2.5). When the eight D(T) data points for each gas are pooled into a Cohen–Turnbull plot of ln D versus 1/FV0 (matrix void fraction), the NPT and NVT data collapse onto a single master line per gas with slopes whose ratio matches the squared kinetic-diameter ratio (σCH4/σH2)2 ≈ 1.73. This collapse provides direct quantitative evidence that the apparent ensemble dependence of selectivity is governed by free-volume modulation alone and that the activation parameter γv* scales with the cross-sectional area of the penetrant. Trajectory analysis further confirms an activated-hopping mechanism for H2 with a broad displacement distribution, while CH4 dynamics are dominated by cage rattling. We explicitly distinguish the diffusivity selectivity αD reported here from the experimental permeability selectivity αP = αD × αS, noting that for H2/CH4 in PE the solubility selectivity αS < 1 and therefore αP < αD. The combination of statistically validated diffusivities, density-resolved free-volume metrics, and the Cohen–Turnbull master curve provides a transferable framework for assessing how thermomechanical boundary conditions modulate the kinetic sieving behavior of polymer infrastructure exposed to hydrogen–methane mixtures.