Large eddy simulation of the variable density mixing layer

Abstract In this paper we perform large eddy simulations of variable density mixing layers, which originate from initially laminar conditions. The aim of this work is to capture the salient flow physics present in the laboratory flow. This is achieved through varying the nature of the inflow condition, and assessing the vortex structure present in the flow. Two distinct inflow condition types are studied; the first is an idealised case obtained from a mean inflow velocity profile with superimposed pseudo-white-noise, and the second is obtained from an inflow generation technique. The inflow conditions generated have matching mean and root mean squared statistics. Validation of the simulations is achieved through grid dependency and subgrid-scale model testing. Regardless of the inflow condition type used, the change in growth rate of the mixing layer caused by the density ratio is captured. It is found that the spacing of the large-scale spanwise structure is a function of the density ratio of the flow. Detailed interrogation of the simulations shows that the streamwise vortex structure present in the mixing layer depends on the nature of the imposed inflow condition. Where white-noise fluctuations provide the inflow disturbances, a spatially-stationary streamwise structure is absent. Where the inflow generator is used, a spatially stationary streamwise structure is present, which appears as streaks in plan-view visualisations. The stationary streamwise structure evolves such that the ratio of streamwise structure wavelength to local vorticity thickness asymptotes to unity, independent of the density ratio. This value is in agreement with previous experimental studies. Recommendations are made on the requirements of inflow condition modelling for accurate mixing layer simulations.