A New Hybrid Scheme for Simulations of Highly Collisional RF-Driven Plasmas

TL;DR

This paper introduces a hybrid simulation approach combining kinetic and fluid models for atmospheric pressure RF-driven plasmas, validating the drift-diffusion approximation and highlighting its limitations for accurate plasma modeling.

Contribution

A novel 1D hybrid modeling scheme that couples kinetic electron simulations with fluid heavy species, improving accuracy over traditional drift-diffusion models at atmospheric pressure.

Findings

Drift-diffusion approximation can overestimate ion transport.

Full ion fluid models provide more accurate plasma density profiles.

Hybrid scheme offers similar computational cost with improved accuracy.

Abstract

This work describes a new 1D hybrid approach for modeling atmospheric pressure discharges featuring complex chemistry. In this approach electrons are described fully kinetically using Particle-In-Cell/Monte-Carlo (PIC/MCC) scheme, whereas the heavy species are modeled within a fluid description. Validity of the popular drift-diffusion approximation is verified against a "full" fluid model accounting for the ion inertia and a fully kinetic PIC/MCC code for ions as well as electrons. The fluid models require knowledge of the momentum exchange frequency and dependence of the ion mobilities on the electric field when the ions are in equilibrium with the latter. To this end an auxiliary Monte-Carlo scheme is constructed. It is demonstrated that the drift-diffusion approximation can overestimate ion transport in simulations of RF-driven discharges with heavy ion species operated in the $\gamma$ mode at the atmospheric pressure or in all discharge simulations for lower pressures. This can lead to exaggerated plasma densities and incorrect profiles provided by the drift-diffusion models. Therefore, the hybrid code version featuring the full ion fluid model should be favored against the more popular drfit-diffusion model, noting that the suggested numerical scheme for the former model implies only a small additional computational cost.