Machine Learning Molecular Dynamics Simulations on Diffusion of SF6 and N2 in SF6/N2 Mixtures

Abstract

Sulfur hexafluoride (SF₆) is widely used for an insulating and arc extinguishing medium in high voltage electrical equipment due to its excellent dielectric properties and insulation performances. However, SF₆ is a potent greenhouse gas, leading to a significant influence on the environment by the large use of SF₆. SF₆ mixed with inert gas such as N₂ is a more promising way to reduce the use of SF₆. The diffusion properties of SF₆/N₂ mixtures play a crucial role in the gaseous mixture separation or replenishment, determining the proportion and uniformity of the mixtures. Combining ab initio molecular dynamics (AIMD) and machine-learning molecular dynamics (MLMD) simulations of SF₆ and N₂ dynamics in SF₆/N₂ mixtures, we systematically illustrate the anomalous characteristics of both SF₆ and N₂ motion. At short times, both SF₆ and N₂ performs super-diffusion through the ballistic motion. At intermediate times, a combination of the SF₆ molecular flexibility and cage effect result in the larger plateau value, while N₂ molecular showed no plateau stage. At longer times, both SF₆ and N₂ gradually turns to normal diffusion. A theoretical model is also offered to quantitatively describe this abnormal diffusion. This work gives a more completely physical understanding of SF₆ and N₂ dynamics in SF₆/N₂ mixtures, which provides important references for SF₆ /N₂ mixed electrical equipment leakage supplement.