Multiscale insights into the diffusion of SF6/N2 mixtures via machine-learning interatomic potentials

Abstract

Sulfur hexafluoride (SF 6 ) is widely used as an insulating and arc-extinguishing medium in high-voltage electrical equipment due to its excellent dielectric properties and insulation performance. However, SF 6 is also a potent greenhouse gas, so mixing SF 6 with an inert gas such as N 2 is a promising way to reduce its usage. The diffusion properties of SF 6 /N 2 mixtures play a crucial role in gas-mixture separation or replenishment, because they determine the proportion and uniformity of the mixtures. By combining ab initio molecular dynamics (AIMD) and machine-learning molecular dynamics (MLMD) simulations of SF 6 and N 2 in SF 6 /N 2 mixtures, we systematically characterize their multiscale diffusion dynamics. At short times, both SF 6 and N 2 exhibit the expected ballistic regime with super-diffusive scaling. At intermediate times, SF 6 shows a more pronounced plateau-like crossover than N 2 , which is consistent with the molecular flexibility of SF6 and the associated rotational and vibrational dynamics. At longer times, both SF 6 and N 2 gradually approach normal diffusion. A theoretical model is also proposed to quantitatively describe the non-Fickian features of the MSD curves. This work provides a more complete physical picture of SF 6 and N 2 dynamics in SF 6 /N 2 mixtures and offers a useful reference for gas replenishment in SF 6 /N 2 mixed electrical equipment.