**Turbulent dissipation and other rates of change.**

When I was working for my PhD with Sam Edwards in the late 1960s, my second supervisor was David Leslie. We would meet up every so often to discuss progress, and I recall that David was invariably exasperated by our concentration on asymptotic behaviour at high wavenumbers. He was strongly motivated towards applications, and felt that the production process at low wavenumbers was more important. To him the dissipation was uninteresting. He used to say to us: `you are messing about down in the drains, when the interesting stuff is all in the production region.’

We had many good humoured arguments but none of us changed our positions. Yet with the passage of time, I increasingly feel that David had a point; even when we restrict our attention to isotropic turbulence. In fact, I would go further and argue that much of the confusion over the Kolmogorov (1941) picture, arises from a failure to see that the dissipation is not the primary quantity. And even when one arrives legitimately at the dissipation (having first considered the production and then the inertial transfer rates), there is often confusion between the instantaneous dissipation rate and the mean dissipation rate. I have myself contributed to that confusion, and this is an opportunity to set matters straight. But first let us consider what the usual practices are.

Kolmogorov used for the instantaneous dissipation and for the mean. Then, in 1953, Batchelor used for the mean dissipation (see equation (6.3.2) in the second edition of his book). A few years later, in 1959, Hinze favoured for the mean dissipation, and this has tended to prevail ever since, particularly in theoretical physics, where is used as an expansion parameter: e.g. the famous -expansion!

In my 1990 book [1], I used for instantaneous dissipation in equation (1.17) and for the mean dissipation in equation (A18). Unfortunately, where I discuss the Kolmogorov variables, in Chapter Two and elsewhere, it is clear that I intend to be the mean dissipation rate. In fact this is the most prevalent usage throughout the literature, at least in theoretical work. When one thinks about it; well it makes sense. One is only ever really interested in mean quantities and a hat notation can be used for instantaneous values where they are required. In my later book [2], I tried to sort this out, as follows:

.

So how does this help us with Kolmogorov, back in 1941? Well, in fact it helps us with Obukhov who, unlike Kolmogorov, worked in wavenumber space where there actually is a turbulence cascade. Obukhov realised that as the Reynolds number increased, there would be a limit where the inertial transfer rate became equal to the dissipation. As the Reynolds number continued to increase, this region of maximum energy transfer would increase in extent, to ever higher wavenumbers. This behaviour has been amply confirmed and is an example of scale invariance. It was recognized by both Obukhov and Onsager that in this range of wavenumbers the spectrum would take the form

If you wish, you can replace the rate of inertial transfer with the dissipation rate. If you want to derive Kolmogorov’s law, then just Fourier transform the Obukhov result for the spectrum. It is the form that has been derived by a properly formulated physical argument. It would be difficult to see how anyone could drag in the so-called intermittency corrections!

[1] W. D. McComb. The Physics of Fluid Turbulence. Oxford University Press, 1990.

[2] W. David McComb. Homogeneous, Isotropic Turbulence: Phenomenology, Renormalization and Statistical Closures. Oxford University Press, 2014.

If one introduces the intermittency correction to the power of the wavenumber k (or to the separation in the structure function as the usual practice), i.e., -5/3+mu, then the dimension is no longer consistent. Then one has to modify the power of the dissipation to restore the dimensional consistency. Even though it can be easily done, seems (as far as I can recall) no one has ever talked about this …

I’m interested in your opinion, sir.

Hello, Sijie.

Thank you for your email and original question. I have read it again and it is perfectly clear. You feel that changing the power of k from -5/3 to -5/3 + mu upsets the dimensional analysis and that there must be a corresponding change to the power of the dissipation which can no longer be 2/3.

In fact you cannot change the power of the dissipation as that determines the consistency of the time dimension: see my blog of 4 March 2020.

What you have to do is introduce a factor L^{\mu} to keep the length dimensions consistent and leave the time dimension unaffected: see my blog of 26 March 2020. This is what was done by Kolmogorov in 1962 and the length L is only very vaguely defined as being of the size of the system.

Does that clear it up?

Best wishes,

David McComb.