1: \begin{abstract}

2: In large eddy simulations, the Reynolds averages of nonlinear terms are not

3: directly computable in terms of the resolved variables and require a closure

4: hypothesis or model, known as a subgrid scale term.

5: Inspired by the renormalization group (RNG),

6: we introduce an expansion for the unclosed terms,

7: carried out explicitly to all orders.

8: In leading order,

9: this expansion defines subgrid scale unclosed terms,

10: which we relate to the dynamic subgrid scale closure models.

11: The expansion, which generalizes the Leonard stress for

12: closure analysis, suggests a systematic higher order determination of the

13: model coefficients.

14: 

15: The RNG point of view sheds light on the nonuniqueness of the infinite

16: Reynolds number limit. For the mixing of $N$ species,

17: we see an $N+1$ parameter family of infinite

18: Reynolds number solutions labeled by dimensionless parameters of the

19: limiting Euler equations, in a manner intrinsic to the RNG itself.

20: Large eddy simulations, with their Leonard stress and dynamic subgrid models,

21: break this nonuniqueness and predict unique model coefficients on the

22: basis of theory. In this sense large eddy simulations 

23: go beyond the RNG methodology, which does not in general

24: predict model coefficients.

25: \end{abstract}

26: