Hydrodynamic and magnetohydrodynamic turbulent flows. Modelling and statistical theory (Q1303753)

Turbulence modelling arises within a stochastic description of turbulent flows where new quantities appear (correlations) but the number of equations remains the same, so that new equations (coming from the model) have to be added leading to solutions. The procedure leads often to finding constants which do not vary from one experiment to the other. This discipline has been relegated to some physicists or engineering circles by some purists, but it nevertheless constitutes an attempt to describe and understand in a predictable way complicated natural phenomena, confirming thus its scientific value. The book is a treatise on this topic. An introductory section motivates turbulence modelling; it allows to handle situations where the structure of the large scale energy input to the flow is taken into account (like in shear flows). Finally, an outline of the book is given. In chapter two the fundamental equations of fluid mechanics are derived (conservation of mass, the Navier-Stokes equation obtained via a model of stress; they are used to get an energy equation). A vorticity equation in a non-inertial frame is derived next. Then, the author defines helicity (the fluid rotates while moving) and gets an evolution equation for it. Finally, the turbulence equations arising through a decomposition of the quantities in average and fluctuating parts are provided. The details and justification of the adopted ``ensemble averaging'' are omitted. In addition to the mean evolution equations, the author obtains evolution equations for the Reynolds stress, the energy, the heat flux and variance, and finally for the turbulent part of the helicity. Chapter three explains homogeneous turbulence (correlation functions are translation-invariant). The author simplifies the situation to the isotropic case without mean velocity and justifies why he does that (simplification of the equations). The Fourier representation is used to express quantities in a frequency domain, so that wavenumbers represent different ranges and physical intuition is enhanced (the nonlinear terms are packaged in an integral expression). This leads to Kolmogorov's picture of a cascade of energy transfer from the energy-production range to the dissipation range via the inertial range where the form of the energy spectrum is given based on dimensional analysis. The attempts of Kraichnan to recover Kolmogorov's law analytically are exposed next. Kraichnan uses a perturbation method for associated Green functions representing the nonlinear interaction of fluctuating velocities. This leads to a substitution procedure after modified expansions are plugged in the expression for the velocity correlations (renormalization), but only low-order terms are retained. Then, Kraichnan assumes some analytic forms for the correlation and the Green function in the isotropic case, and uses them in the evolution equation for the correlation function but Kraichnan fails to obtain Kolmogorov's description. This is due to the effect of large eddies on small ones, which appears in the developed formalism. Some attempts to repair this failure are presented. Chapter four presents conventional turbulence modelling. First, the Reynolds stress occurring in the mean velocity evolution equation takes the following empirical form \[ R_{ij}= \textstyle{{2\over 3}} K\delta_{ij}- \nu_TS_{ij},\tag{1} \] where \(K\) is the turbulent kinetic energy, \(\delta_{ij}\) is the Kronecker symbol, \(S_{ij}\) is the mean velocity strain tensor and \(\nu_T\), the turbulent viscosity, is assumed to be proportional to \(K^2/\varepsilon\). \(\varepsilon\) is the energy dissipation rate. This form implies that the evolution of \(K\) and \(\varepsilon\) have to be taken into account, and this is done next. Several terms in the obtained equations are estimated based on physical considerations. Finally, the model is reduced to determining five constants. Some practical configurations (grid turbulence, channel turbulence) allow to provide numerical values for these constants. Some shortfalls are pointed out especially: the model does not capture anisotropy effects, and this is illustrated on some physical configurations. Quadratic nonlinearities are added in an attempt to remedy these drawbacks, and again constraints on the coefficients of the model are derived from specific physical configurations. For a rotating channel, third-order modelling (in \(S_{ij}\) and mean vorticity) is seen to be necessary. The other section of this chapter is devoted to ``second-order modelling''. Now terms appearing in the evolution equation for the Reynolds stress are modelled in terms of introduced quantities. From the evolution equation for the deviatoric part of Reynolds stress in a second-order modelling framework, one obtains an expression where the first term corresponds to algebraic linear turbulent viscosity model. When this linear model is substituted in the full nonlinear equation, one gets an algebraic second-order nonlinear model. Another term involving second-order derivatives did not appear previously in the algebraic modelling. Truncating expressions (i.e. neglecting higher-order terms) might be dispfutable when the neglected quantities have large values which is the case here, and approaches in work of researchers to handle this issue are mentioned. (This suggests the use of a renormalization procedure ``à la Kraichnan''.) Finally, a similar study for the modelling of scalar diffusion is presented. Chapter five considers less elaborate models in the context of large eddy simulation. This is an intermediate situation between full simulation and modelling via ensemble averaging. Several filtering procedures which eliminate small scale components are presented. But the effect of these components has to be modelled; this is called subgrid-scale modeling. So physical quantities can be decomposed into filtered quantities and the remaining ones. But the algebraic properties of filtered quantities differ from the ensemble averaged ones. This leads to the appearance of new terms in the filtered Navier-Stokes equation. In a first approach, the additional terms are neglected, and the situation is formally analogous to the ensemble averaging configuration. A first algebraic modelling mimics the one developed in the previous chapter using an analog to (1). An evolution equation for the energy is derived in this context. When two terms in this equation cancel (this corresponds to the equilibrium of the energy cascade), using also the analog to (1) and related quantities, one obtains an algebraic equation for the energy. This is the Smagorinsky model. The constant which characterizes it takes several values depending on the types of flows, so that improvements are called for. They are shortly described next (use of the dynamical energy equation without equilibrium assumption, inclusion of the neglected terms in the Navier-Stokes equation\dots) with reference to related publications. A double filtering procedure investigated by Germano allows to get additional constraints to evaluate the constants of the model. Other approaches are presented, too. Chapter six presents mainly the work of the author as a way to give a statistical justification for the conventional turbulence models. The method of Kraichnan does not apply directly to turbulent shear flows (due to boundary effects which affect the large scales), and the ``two-scale direct-interaction approximation'' method of the author is presented. With this time-scale analysis, the Fourier representation and the appropriate Green functions, the turbulent-viscosity model is derived. Expressions for \(K\), \(\nu_T\) are obtained. Interestingly enough, the energy spectrum deviates from the Kolmogorov one, and this is shown analytically. Finally, the formula \(\nu_T= C_\nu K^2/\varepsilon\) gives in this context \(C_\nu= 0.123\), whereas previously one had \(C_\nu= 0.09\). This discrepancy leads to a higher-order analysis which is performed next. Another section is devoted to modelling the change in \(\varepsilon\), which proves to be the key quantity responsible for the deviation from the Kolmogorov energy spectrum. Finally, the turbulent scalar flux, frame rotation effects and buoyancy effects are modelled in this context. Transport equations are included in the treatment. Some aspects concerning subgrid-scale modelling are added at the end of the chapter. Chapter seven shows a simpler way to recover formula (1) and a corresponding turbulent viscosity from the dynamical equations. This is called the ``Markovianized one-point approach''. By neglecting the long range effect of the pressure, it is a convenient alternative to the approach of the previous chapter which becomes too complicated in compressible and magnetohydrodynamic flows. A scaling (where anisotropy arises in first-order terms) and the Fourier representation are introduced. The first-order speed in the scaling is written as \[ u_1= \int^\tau_{-\infty} G_u(\tau, s)I_u(s) ds\approx I_u(\tau) \int^\tau_{-\infty} G_u(\tau, s) ds\approx I_u(\tau) \tau_T.\tag{2} \] \(I_u\) is the right-hand side of the first-order dynamics, and a model for the Green function \(G_u\) in terms of the characteristic time scale \(\tau_T\) is assumed in the second approximation of (2). \(\tau_T\) is estimated from inertial range spectral quantities, i.e. typically from zero-order information (isotropic). The author does not justify why it is valid to inject zero-order data into first-order quantities. High-order corrections in the model are expected to come from high-order terms in the scaling. The rest of the book focuses on challenging and exciting topics. Chapter eight is devoted to compressible turbulence modelling. The equations (density, speed, energy) for two kinds of averaging, i.e. the conventional ensemble averaging and a mass-weighted ensemble averaging, are written down. The methods of chapter six for the first kind of averaging are used first. The forms of models for the correlation functions which appear in the equations are given without justification on p. 284, and a turbulent viscosity expression which includes compressibility effects is obtained. But the associated Green functions are not known. This leads to another approach using the methods of chapter seven. A compressibility term in the Kolmogorov spectrum is added, and an explicit formula for the turbulent viscosity is obtained. It depends, in particular, on a normalized density variance. So a simplified compressibility model is derived; it uses as usual the evolution equations for \(K\) and \(\varepsilon\), and another one for the density variance. The final model depends on ten constants, and it is supposed that it can account for observed turbulence suppression after a shock wave as well as for the sharp decrease of the mixing layer in the configuration of a turbulent free shear layer. Chapter nine deals with magnetohydrodynamic turbulence modelling. The associated phenomena arise in planetary or stellar physics as well as in plasmas. The basic evolution equations are introduced (mass conservation, Lorentz force in the momentum equations, Maxwell's equations, which are coupled to the fluid motion, Ohm's law, energy conservation). Elsasser's variables are introduced: \(\phi= u+b\) and \(\psi= u-b\), where \(u\) is the speed and \(b\) the magnetic field, and the equations then have a more symmetric form. Some quantities are defined (magnetohydrodynamic energy, cross helicity which is the integral over the volume of \((u\cdot b)\)), and they are conserved provided viscosity and magnetic diffusivity are neglected. Cowling's theorem showing that axisymmetric flows and magnetic fields are incompatible is proved. The magnetic and hydrodynamic Reynolds numbers of the Earth and of the Sun are large in the zones where there is most motion, so that turbulence is expected. Next, mean-field equations and evolution equations for fluctuating quantities are derived; here rotation effects are included (Coriolis force\dots). The fluctuating quantities act on the mean-field quantities via modified Reynolds stress \(R_{ij}= \langle u_i' u_j'- b_i' b_j'\rangle\) and an additional turbulent electromotive force \(E_M= \langle u'\times b'\rangle\) (primed quantities are fluctuating quantities); in the last expression, a component along the magnetic field is called helicity dynamo, and another one arises from a non-vanishing cross helicity \(\langle u'\cdot b'\rangle\). Two studies following the methods of chapter six and seven lead to models for the turbulent Reynolds stress and the electromotive force. The last chapter develops the themes of the previous one, concentrating on a qualitative description of the fields. Typically, one investigates how magnetic fields are produced (helicity effect, rotational or toroidal fluid motion and combinations). Some examples are pointed out, like the sunspots. Each chapter ends with its list of references, and an index is included. The English is awkward at times but it can be felt that the author cares about being understood. This is a very good book on turbulence modelling, which starts from the basics an brings the reader to advanced topics, including important contributions of the author.