Large Eddy Simulation (LES) of a mixing layer with evaporating drops has been performed with various Sub Grid Scale (SGS) models. The objectives of this work are to identify the statistics that are essential for validation of LES, to quantify the accuracy of their prediction by the SGS model and to clarify the grid-spacing influence on the results. Four SGS models namely, the dynamic Smagorinsky (SM) model, the dynamic Gradient (GR) model, the dynamic mixed model which combines SM with a

more »

... imilarity term and the dynamic Clark model which combines the SM and GR gradients are tested. A set of first order, second order and drop conditioned statistics from LES predictions are compared to the Direct Numerical Simulation (DNS) results. First order statistics such as Favre averaged streamwise velocity, Favre averaged vapor mass fraction, and the drop streamwise velocity are predicted accurately by all the models tested. Second order statistics such as the streamwise Reynolds stress, the cross-stream Reynolds stress, the spanwise Reynolds stress and the vapor mass fraction fluctuation variance are predicted with a maximum of 15 percent error by the SGS models tested. The dynamic Smagorinsky model predicts the drop conditioned statistics, which describe the segregation of the drops in high strain rate region, more accurately compared to the other models tested. Essentially, only the mean values are grid-spacing independent.