Predicting the flow of cold plasma jets in kINPen: a critical evaluation of turbulent models

Abstract

The flow in a cold plasma jet device may be laminar, transitional, turbulent or a combination, depending on the conditions and the geometry. Choosing the appropriate turbulence model constitutes a crucial step with a direct impact on the accuracy of the results. Three different turbulence models with different levels of accuracy and detail, and different computational cost, namely the standard k–ϵ, realizable k–ϵ and large eddy simulation (LES) models, are compared in terms of turbulence kinetic energy and anisotropy, axial velocities, and on-axis species’ mole fractions. The focus is on the flow in the kINPen plasma jet with Ar as the feed gas and on the mixing of Ar with air after the exit of the jet nozzle. Electric field and species consumption are not considered. However, the plasma-induced elevation of the gas temperature is considered and its effect on the concentration of species is investigated. A systematic study of the accuracy of the LES and the k–ϵ models is performed in order to verify that the produced results are not affected by the mesh density, the time step and the simulation time. The accuracy of the LES model is further supported by the high percentage (98%) of the directly resolved turbulence kinetic energy; only 2% is modeled. The k–ϵ models (standard and realizable) produce similar results, which deviate from those of the LES model. In particular, there are remarkable differences in the velocity profiles after the needle and up to the exit of the jet nozzle. Additionally, the k–ϵ models overestimate the on-axis Ar (feed gas) mole fraction, i.e. the LES model predicts stronger mixing after the exit of the jet nozzle. The origin of the deviations is the Boussinesq approximation utilized by the k–ϵ models to calculate the Reynolds stresses and leads to an underestimation of the turbulence anisotropy. The results of the models are compared to Schlieren measurements of the on-axis Ar mole fraction with very good agreement for the LES model.