Spectral approach to finite Reynolds number effects on Kolmogorov's 4/5 law in isotropic turbulence

The Kolmogorov's 4/5 law is often considered as the sole exact relationship of inertial range statistics. Its asymptotic character, however, has been evidenced, investigating the finite Reynolds number (FRN) effect for the third-order structure function S 3 ( r ) (e.g., for longitudinal velocity increments with r separation length) using variants of the Kármán-Howarth equation in physical space. Similar semi-empirical fits were proposed for the maximum of the normalized structure function, C 3 =−max r S 3 ( r )/( ɛr ), expressing C 3 −4/5 as a power law of the Taylor-based Reynolds number. One of the most complete studies in this domain is by Antonia and Burratini [J. Fluid Mech. 550 , 175 (2006)]. Considering that these studies are based on a model for the unsteady second-order structure function S 2 ( r,t ), with no explicit model for the third-order structure function itself, we propose to revisit the FRN effect by a spectral approach, in the line of Qian [Phys. Rev. E 55 , 337 (1997), Phys. Rev. E 60 , 3409 (1999)]. The spectral transfer term T ( k,t ), from which S 3 ( r,t ) is derived by an exact quadrature, is directly calculated by solving the Lin equation for the energy spectrum E ( k,t ), closed by a standard triadic (or three-point) theory, here Eddy Damped Quasi Normal Markovian. We show that the best spectral approach to the FRN effect is found by separately investigating the negative (largest scales) and positive (smaller scales) bumps of the transfer term, and not only by looking at the maximum of the spectral flux or max k ∫ k ∞ T ( p , t ) dp → ɛ . In the forced case, previous results are well reproduced, with Reynolds numbers as high as Re λ =5000 to nearly recover the 4/5 value. In the free decay case, the general trend is recovered as well, with an even higher value of Re λ =50 000, but the EDQNM plots are systematically below those in Antonia and Burattini [J. Fluid Mech. 550 , 175 (2006)]. This is explained by the sensitivity to initial data for E ( k ) in solving the Lin equation at moderate Reynolds numbers. Accordingly, an ad hoc initialization yields results consistent with the experimental spectrum measurements of Comte-Bellot and Corrsin [J. Fluid Mech. 48 (2), 273 (1971)], from which S 3 ( r ) are recalculated. Present results show that the dispersion observed in existing data at low Reynolds number may be due to sensitivity to initial spectrum shape, a feature of the flow which is not under control in most of laboratory experi