Kolmogorov 1941 theory ($S_p\sim {ϵ}^{p/3}{r}^{p/3}$)

Kolmogorov 1941 theory assumes locally universal small-scale velocity statistics set by mean viscous dissipation and kinematic viscosity at large Reynolds number. In its inertial range dimensional form, $S_p(r)\sim {ϵ}^{p/3}{r}^{p/3}$. Large-scale modulation and internal intermittency require qualifying this unconditional scaling.

In the inertial range, velocity increment statistics also become independent of kinematic viscosity. Combined with the Kolmogorov first similarity hypothesis, this gives $S_p(r)={\beta }_p(ϵr{)}^{p/3}$ by dimensional analysis. Signed integer moments and absolute noninteger moments must be distinguished; the signed third-order coefficient is negative in a forward energy cascade.

The second-order longitudinal structure function has the displayed form in the idealized inertial range of Kolmogorov 1941 theory. It predicts that the typical longitudinal velocity increment grows as the cube root of separation. Unlike the Kolmogorov four-fifths law, its exponent and coefficient are not exact consequences of the Navier-Stokes equation.

In the locally isotropic equilibrium range, the small-scale velocity increment statistics depend on the separation, mean viscous dissipation and kinematic viscosity, independently of large-scale forcing details. Dimensional analysis leaves the Kolmogorov microscales as normalization scales and gives the displayed second-order longitudinal structure function. Universality of $F$ is part of the hypothesis, not something proved by dimensional analysis.