This work involves large eddy simulations (LES) of passive scalar mixing in a round, turbulent jet. We study the effects of resolution on the resulting LES velocity and scalar concentration fields, and focus on the interaction between the modeled, subgrid, and resolved-scale quantities. The simulations are performed in spherical coordinates for Reynolds number Re _d = 5,000 at three levels of grid resolution, and for Re_d = 10,000 at a single resolution level. Throughout this work, the LES-filtered momentum equations are closed using a dynamic Smagorinsky model for the SGS stress.

In the first part of this study, the LES-filtered scalar transport equation is closed using both the dynamic eddy diffusivity and dynamic mixed models for the SGS scalar flux. The mean velocity and scalar concentration fields and jet scaling parameters are accurately reproduced by the LES on all grid systems used. However, there are some variations observed in the mean quantities at different levels of resolution that suggest a non-monotonic influence of subgrid parameters on the resolved LES fields.

The contribution from the SGS model is evaluated by comparing distributions of the modeled subgrid stress and scalar flux components, as well as the SGS and molecular dissipation of resolved-scale fluctuations, at different levels of resolution. Overall, the contribution from the SGS model is observed to increase as the grid is coarsened, indicating as expected that more of the smaller-scale turbulent motions are represented by models at lower levels of resolution.

In the second part of this study, we investigate the effects of different modeling approaches for the SGS scalar flux. The scalar flux models evaluated in this work include, in addition to the eddy diffusivity and mixed models, the recently proposed dynamic structure and multi-fractal models. Resulting mean and fluctuating resolved concentration fields predicted by the multi-fractal model are in close agreement with those predicted by the eddy diffusivity and mixed models, which also show good agreement with available experimental results. The concentration fields predicted by the dynamic structure model differ noticeably from those for the other models in this study, particularly for fluctuating quantities. Analyses of the SGS scalar flux components show that the SGS scalar flux is overpredicted by the dynamic structure model compared to the other models, which may explain some of these trends observed in the resolved and fluctuating concentration fields. The results suggest that model performance, as assessed through the resolved-scale simulation results, is significantly more sensitive to scalar energy transfer across the resolved scale than to the structural details of model formulation.