Turbulence and self organization

Chapter 3
Closed System of Hydrodynamic Equations
to Describe Turbulent Motions
of Multicomponent Media
The growing interest in investigating developed turbulent flows of compressible
gases and liquids in recent years (see, e.g., van Mieghem 1973; Ievlev 1975, 1990;
Kompaniets et al. 1979; Bruyatsky 1986; Kolesnichenko and Marov 1999) has been
triggered by the necessity of solving numerous problems of rocket, space, and
chemical technologies and problems related to environmental protection. Concur-
rently, the methods for theoretical modeling of natural media, including the previ-
ously inaccessible regions of near-Earth space and the atmospheres of other planets
in the Solar system, are improved. In particular, it has become obvious that
modeling the upper planetary atmosphere requires developing an appropriate
model of turbulent motion that would take into account the multicomponent
structure and compressibility of the medium, the heat and mass transfer processes,
and chemical reactions (Marov and Kolesnichenko 1987).
We begin this chapter with the derivation of a closed system of averaged
hydrodynamic equations designed to describe a wide class of turbulent flows and
physical–chemical processes in multicomponent media. We analyze the physical
meaning of the individual terms in these equations, including the energy transition
rates between various energy balance components. Here, we systematically use the
weighted-mean Favre (1969) averaging, which allows the form and analysis of the
averaged equations of motion for chemically active gases with variable densities
and thermophysical properties to be simplified considerably, along with the tradi-
tional probability-theoretic averaging. Special attention is paid to the derivation of
closing relations for the turbulent diffusion and heat fluxes and the Reynolds
turbulent stress tensor by thermodynamic methods. For the reader’s convenience,
all calculations are performed comprehensively and can be traced in all details.
Progress in developing and applying semiempirical turbulence models of the
first closure order (the so-called gradient models) for a single-fluid medium allows
some of them to be generalized to the case of turbulent flows of reacting gas
mixtures that is important in astrophysics and geophysics (see, e.g., Libby and
Williams 1994). At the same time, assessing the status of the first-order closure
problem on the whole, it should be recognized that at present there is actually no
general phenomenological theory of turbulent heat conduction and turbulent
M.Y. Marov and A.V. Kolesnichenko, Turbulence and Self-Organization:
Modeling Astrophysical Objects, Astrophysics and Space Science Library 389,
DOI 10.1007/978-1-4614-5155-6_3, # Springer Science+Business Media New York 2013
189

diffusion for multicomponent mixtures. The gradient relations used in the literature
(see, e.g., Hinze 1963; Monin and Yaglom 1992) are not general enough and were
derived mainly for turbulent flows with a well-defined dominant direction under
strong and not always justified assumptions, such as, for example, the conservatism
of the flow characteristics transferred by turbulent fluctuations or the equality of the
mixing lengths for various turbulent transport processes. This necessitates consid-
ering other approaches to the closure of averaged hydrodynamic equations for a
mixture at the level of first-order turbulence models, in particular, using the
methods of extended irreversible thermodynamics. In this case, the Onsager for-
malism allows the most general structure of the closing gradient relations to be
obtained both for the Reynolds stress tensor and for the turbulent heat and diffusion
fluxes in a multicomponent mixture, including those in the form of generalized
Stefan–Maxwell relations for multicomponent turbulent diffusion. At the closure
level under consideration, such relations describe most comprehensively the turbu-
lent heat and mass transport in a multicomponent medium. Both classical models
dating back to Prandtl, Taylor, and Karman (see, e.g., Problems of Turbulence
2006) and more recent second-order closure models based, in particular, on the
differential balance equations for the turbulent energy and integral turbulence scale,
can be used to determine the turbulent exchange coefficients.
3.1 Basic Concepts and Equations of Mechanics of Turbulence
for a Mixture of Reacting Gases
One of the main tasks of theoretical geophysics is to numerically calculate the
spatial distributions and temporal variations of the density, velocity, temperature,
and concentrations of chemical components as well as some other thermohy-
drodynamic characteristics of a gas mixture in a turbulized planetary atmosphere
at large Reynolds numbers Re ¼ UL=n (here, U is the characteristic flow velocity in
the atmosphere, L is the scale of the main energy-carrying vortices, and n is the
molecular kinematic viscosity). Below, we assume that the system of differential
equations for a reacting gas mixture given in the Chap. 2 also describes all details of
the true (instantaneous, pulsating) state of the fields of these quantities under
specified initial and boundary conditions in the case of developed turbulence in
the atmosphere. However, it is essentially useless without a certain averaging-
related modification, because it cannot be solved with present-day computing
facilities. The application of numerical computation methods in this case would
entail the approximation of an enormous spatiotemporal flow field by a finite
number of grid points that should be used when the differential equations are
replaced with their finite-difference analogs. At present, there is only one economi-
cally justified way out of this situation: to solve the stochastic hydrodynamic
equations of a mixture only for large spatiotemporal scales of motion that determine
the averaged structural parameters of a turbulized atmosphere and to model small-
scale motions (the so-called subgrid turbulence) phenomenologically.
190 3 Closed System of Hydrodynamic Equations to Describe Turbulent Motions. . .

In this case, stochasticity implies the existence of an ensemble of possible
realizations of the turbulent flow field for which the concept of a statistical
(mathematically expected) average is defined for all fluctuating thermohy-
drodynamic characteristics. Any flow parameter can then be averaged either over
a set of realizations at various times at a given point of coordinate space or over a
set of values at various spatial points of some volume at a fixed time. As has already
been mentioned in Chap. 1, to eliminate the obvious inconsistency in the averaged
hydrodynamic equations (when the flow parameters are defined as time-averaged
ones, although they are represented in these equations by time derivatives), the time
interval T over which this averaging is performed should be sufficiently long
compared to the time scale of individual turbulent fluctuations but, at the same
time, short compared to the time scale of a noticeable change in averaged quantities
if the averaged motion is nonstationary. Accordingly, the spatial averaging scale
should satisfy conditions similar to those imposed on the time interval T .
In particular, in atmospheric dynamics it is customary to distinguish the mean
zonal motions (with horizontal sizes ~104
km) and the deviations from these
mean motions (called pulsations, fluctuations, vortices). These fluctuations can
have various spatial scales, from several meters to thousands of kilometers. Thus,
by the “turbulent fluctuations” we often mean simply the deviations from the mean
irrespective of their scales (Brasseur and Solomon 1984).
Thus, the separation of the real stochastic motion of a turbulized medium into
slowly varying mean and turbulent (irregular, fluctuating near the means) motions
depends entirely on the choice of the spatiotemporal region for which the means are
defined. The size of this region fixes the scale of averaged motion. All larger
vortices contribute to the averaged motion determined by the mean values of
the state parameters r; u; T; Za ða ¼ 1; 2; . . . ; NÞ. All smaller vortices filtered out
in the averaging process contribute to the turbulent motion determined by the
corresponding fluctuations of the same parameters. To obtain representative
means and the corresponding fluctuations of physical quantities, the spatiotemporal
averaging region must include a very large number of vortices with sizes smaller
than the averaging region and a very small fraction of vortices with sizes larger than
the averaging region (see van Mieghem 1973).
3.1.1 Choosing the Averaging Operator
Averaging is a central problem in the mechanics of continuous media and, in the
case of such a complex system as a turbulized fluid, the construction of its
macroscopic model itself often depends precisely on the averaging method.
In liquid and gas turbulence theories, various methods of averaging physical
quantities Aðr; tÞ are used. For example, these include the temporal averaging
3.1 Basic Concepts and Equations of Mechanics of Turbulence for a Mixture. . . 191

Aðr; tÞ ¼ 1=Tð Þ
ðT
0
Aðr; t þ tÞdt; (3.1)
where the averaging interval T is assumed to be sufficiently long compared to the
characteristic period of the corresponding fluctuation field but much shorter than
the period of variation in the averaged field; the spatial averaging through integra-
tion over a spatial volume W ; and the probability-theoretic averaging over a
statistical ensemble of possible realizations of random hydrodynamic turbulent
flow fields. The latter approach is most fundamental. It uses the concept of an
ensemble, i.e., an infinite set of hydrodynamic systems of the same physical nature
that differ from one another by the state of the field of velocities and/or other
thermohydrodynamic parameters at a given time. According to the well-known
ergodicity hypothesis (see Monin and Yaglom 1992), the time and ensemble
averages are identical for a stationary stochastic process. Without discussing here
the advantages and disadvantages of various averaging methods in more detail, we
only note that “the practice of constructing phenomenological models to study
turbulent motions shows that the techniques for introducing the averaged
characteristics of motion are, in general, unimportant for setting up the complete
system of averaged hydrodynamic equations if one requires the fulfillment of the
following Reynolds postulates in the process of any averaging” (Sedov 1980):
A þ B ¼ A þ B; aA ¼ aA; AB ¼ A B: ð1
Þ (3.2)
Here, Aðr; tÞ and Bðr; tÞ—are some fluctuating characteristics of the turbulent field,
Aðr; tÞ and Bðr; tÞ are their mean values, and a is a constant (without any
fluctuations). Next, we assume that any averaging operator used in (3.2(1
))
commutes with the differentiation and integration operators both in time and in
space:
@Aðr; tÞ=@t ¼ @Aðr; tÞ=@t;
ð
Aðr; tÞdt ¼
ð
Aðr; tÞdt; ð2
Þ
@Aðr; tÞ=@r ¼ @Aðr; tÞ=@r;
ð
Aðr; tÞ dr ¼
ð
Aðr; tÞdr: ð3
Þ (3.2*)
Note that in the case of temporal (and/or spatial) averaging, some of relations
(3.2), in general, hold only approximately, although the smaller the change of
Aðr; tÞ in time and space in the domain of integration under consideration, the more
accurate they are. At the same time, for the probability-theoretic averaging of the
hydrodynamic equations (over the corresponding statistical ensemble of
realizations), the Reynolds postulates (3.2) hold exactly, because they simply
follow from ordinary properties of the mathematical expectation in the probability
theory. That is why we use them below without any restrictions.

In the classical theories of turbulence for homogeneous incompressible fluids
that have been developed by now fairly thoroughly (see, e.g., Townsend 1956;
Monin and Yaglom 1992), the averagings are introduced in a similar way and, as a
rule, without any weight factors for all thermohydrodynamic parameters without
exception. In the case of averaging over time (space) or over an ensemble of
possible realizations,
Aðr; tÞ ¼ lim
M!1
1
M
XM
p¼1
AðpÞ
; (3.3)
(where the summation is over the set of realizations ðp ¼ 1; 2; . . . ; MÞ, while the
corresponding average field Aðr; tÞ is defined as the expected value of A for an
ensemble of identical hydrodynamic systems), the instantaneous value of the param-
eter A is represented as the sum of the averaged, A, and fluctuation, A0
, components:
A ¼ A þ A0
; ðA
0
¼ 0Þ: (3.4)
However, when applied to a multicomponent continuum with a varying density
rðr,tÞ, such averaging, which is the same for all physical parameters of the medium,
not only leads to cumbersome hydrodynamic equations for the scale of mean motion
(because it is necessary to retain correlation moments liker0u0; r0u0u0; r0Z0
a, etc. in
the equations), but also makes it difficult to physically interpret each individual term
of these averaged equations. Bearing in mind the various applications of the phe-
nomenological turbulence model for a reacting mixture being developed in this
book, in particular, to some astrophysical phenomena in which the ratio of the
characteristic fluid velocity to the averaged speed of sound (a measure of signifi-
cance of the density fluctuations) is much greater than unity, below we assume the
mass density r to be variable.
As is well known (see, e.g., Kolesnichenko and Marov 1999), when constructing
a model of developed turbulence in a compressible multicomponent medium, apart
from the “ordinary” means of physical quantities (such as the density, pressure,
molecular mass, momentum, and energy transfer fluxes), it is convenient to use the
so-called weighted means (or Favre means (see Favre 1969)) for some other
parameters (e.g., the temperature, internal energy, entropy, hydrodynamic velocity,
etc.) specified by the relation
hAi r A=r ¼ lim
M!1
1
M
XM
p¼1
rðpÞ
AðpÞ
!
= lim
M!1
1
M
XM
p¼1
rðpÞ
!
; (3.5)
in this case,
A ¼ hAi þ A00
; ðA00
6¼ 0Þ; (3.6)

where A00
is the corresponding turbulent fluctuation of the field Aðr; tÞ. Thus, two
symbols are used below in the book to denote the means of physical quantities: the
overbar designates averaging over an ensemble of realizations (time and/or space),
while the angle brackets designate weighted-mean averaging. The double prime is
used below to denote the fluctuations of the same Favre-averaged quantities.
If r ffi r À const (e.g., in a fluid with Boussinesq properties (Boussinesq 1977)),
then both averaging procedures coincide. At the same time, using averaging (3.5)
for a number of fluctuating physical quantities that characterize a multicomponent
continuum simplifies considerably the form and analysis of the averaged hydrody-
namic equations (Kolesnichenko and Marov 1999). In addition, it is also conve-
nient, because precisely these means are probably measured in experimental studies
of turbulent flows by conventional methods (see, e.g., Kompaniets et al. 1979).
3.1.1.1 Weighted Means
Some properties of the weighted-mean averaging of physical quantities widely used
below can be easily derived from definition (3.5) and the Reynolds postulates (3.2)
(see van Mieghem 1973; Kolesnichenko and Marov 1979):
hAi ¼ hAi; hAi ¼ A; hAhBii ¼ hAihBi; r0A0
¼ r0A00
;
rA00
¼ 0; A00
¼ Àr0A00
=r; rAB ¼ rhAihBi þ rA00
B00
;
ðABÞ00
¼ hAiB00
þ hBiA00
þ A00
B00
À rA00
B00
=r; ðrAÞ0
¼ rA00
þ r00
hAi;
@hAi
@r
¼
@hAi
@r
; rA
@B
@r
¼ rhAi
@hBi
@r
þ rA
@B00
@r
;
r
dA
dt
¼ r
DhAi
Dt
þ
@
@r
Á rA00
u00

; (3.7)
whereDhAi=Dtis the substantial time derivative for averaged motiondefined by (3.11).
3.1.1.2 Averaged Continuity Equation
It is easy to verify that the average density r and weighted-mean hydrodynamic
velocity of a mixture hui r u=r satisfy the continuity equation for mean motion
@r
@t
þ
@
@r
Á rhui

¼ 0 (3.8)
This equation can be obtained by applying the Reynolds averaging operation
(3.2) to the continuity equation (2.2), which is assumed to be valid for the instanta-
neous (true) density and hydrodynamic velocity. Since the turbulent mass flux
r u00 ¼ 0 ðu00 6¼ 0Þ, there is no mass transport through turbulence on average (in

the case of Favre averaging). Given the well-known difficulties of modeling the
correlations r0 u0 that appear in the case of “ordinary” averaging (without any
weight) of (2.2), the retention of the standard from of the continuity equation (when
formally replacing the true density and velocity by the averaged ones) is a strong
argument for using the weighted-mean averaging hui for the hydrodynamic flow
velocity (see van Mieghem 1973). Below, when developing the model of multi-
component turbulence, we use the stochastic averaging operator (3.3) unless any
other averaging method is specified specially.
3.1.1.3 Averaged Operator Relation
Averaging the operator relation (2.4) when using (3.7) and (3.8) leads to the identity
r
dA
dt
¼
@
@t
rhAið Þ þ
@
@r
Á rhAihuið Þ

þ
@
@r
Á rA00
u00

¼ r
@hAi
@t
þ rhui
@hAi
@r
þ
@
@r
Á rA00
u00

:
(3.9)
Let us define the turbulent flux of the attribute Aðr; tÞ, which is the second
statistical moment (a one-time one-point pair correlation function) representing the
transport of some fluctuating characteristic A00
of a turbulent medium by turbulent
velocity fluctuations u00
, by the formula
Jturb
ðAÞ rA00
u00 ¼ rhA00
u00
i (3.10)
and denote the substantial time derivative for an averaged continuum by
D
Dt
Á Á Áð Þ
@
@t
Á Á Áð Þ þ hui
@
@r
Á Á Áð Þ: (3.11)
Identity (3.9) then takes the form
r
dA
dt
¼ r
DhAi
Dt
þ
@
@r
Jturb
ðAÞ

: (3.12)
In addition, in view of (3.8), the operator relation
r
DA
Dt

@
@t
rAð Þ þ
@
@r
rAhuið Þ

(3.13)
between the substantial and local changes in Aðr; tÞ in an averaged flow is valid.
It should be emphasized that the quantity A in the latter relation can be both the
instantaneous value of some specific flow field characteristic (a scalar, a vector, or a
tensor) and its averaged value hAi or fluctuation component A00
.

3.1.2 Mass and Momentum Conservation Laws
for Averaged Motion
Below, we consider a turbulized multicomponent gas mixture as a continuous
medium whose true (instantaneous) states of motion can be described by the system
of hydrodynamic equations (2.2), (2.7), (2.9), (2.29), and (2.31) for a random sample
of initial and boundary conditions. This is possible for spatiotemporal scales between
the scales of molecular motions and the minimum turbulence scales (determined by
the linear sizes and lifetimes of the smallest vortices). The latter generally exceed the
scales of molecular motions, i.e., the separation between molecules, let alone the
molecular sizes, by several (at least three) orders of magnitude. Highly rarefied gases,
which are not considered here, constitute an exception.
3.1.2.1 General Averaged Balance Equation
Using identity (3.12) for the probability-theoretic averaging of the balance equation
(2.1), we obtain a general differential form of the substantial balance equation for
some structural parameter Aðr; tÞ for an averaged continuum:
r
DhAi
Dt
r
@hAi
@t
þ rhui
@hAi
@r
¼ À
@
@r
JS
ðAÞ

þ sðAÞ: (3.14)
Here,
JS
ðAÞ JðAÞ þ Jturb
ðAÞ (3.15)
is the substantial total flux density including the averaged molecular, JðAÞ , and
turbulent, Jturb
ðAÞ, fluxes of the attribute A; sðAÞ is the averaged volume density of the
internal source of A . Note that the main problem of the phenomenological
turbulence theory, the so-called closure problem, is related precisely to finding
the unknown turbulent fluxes Jturb
ðAÞ via the medium’s averaged state parameters.
Finally, if we transform the left-hand side of (3.14) using relation (3.10), then we
obtain a local form of the differential balance equation for the Favre-averaged field
quantity Aðr; tÞ:
@
@t
rhAið Þ þ
@
@r
JS0
ðAÞ ¼ sðAÞ: (3.16)
Here,
JS0
ðAÞ rhAihui þ JS
ðAÞ ¼ rhAihui þ JðAÞ þ Jturb
ðAÞ

is the local total flux density of the characteristic hAi in an averaged turbulized
continuum including the convective term rhAihui. The flux density JS0
ðAÞ is the
amount of hAi passing per unit time through a unit surface area @W (the position of
the surface area is specified by a unit vector n lying on the outer side of the surface
@W bounding the turbulized fluid volume W).
Let us now turn to the derivation of averaged multicomponent hydrodynamic
equations by successively considering the cases of different defining parameters A
that describe the instantaneous thermohydrodynamic state of a turbulized medium
in (3.14). In contrast to the ordinary hydrodynamic equations for a mixture that are
assumed to describe random fluctuations of all physical parameters, these equations
contain only smoothly varying averaged quantities; it is this circumstance that
allows the powerful mathematical apparatus of continuous functions and efficient
numerical methods to be used for their solution.
3.1.2.2 Specific Volume Balance Equation for Averaged Motion
Let us assume that A 1=r in (3.14) and use (2.6) for the quantities Jð1=rÞ Àu
and sð1=rÞ 0. We then obtain
r
D
Dt
ð1=rÞ r
@
@t
ð1=rÞ þ ru Á
@
@r
ð1=rÞ ¼ Àdiv JS
ð1=rÞ; (3.17)
where
JS
ð1=rÞ ðr,tÞ Jð1=rÞ þ Jturb
ð1=rÞ (3.18)
is the substantial total flux density of the specific volume in a turbulized continuum;
the averaged molecular and turbulent fluxes of ð1=rÞ are defined, respectively, by
the relations [see (3.7)]
Jð1=rÞ ¼ Àu ¼ Àhui À u00; (3.19)
Jturb
ð1=rÞðr; tÞ rð1=rÞ00
u00 ¼ u00 ¼ Àr0u00=r: (3.20)
Therefore, for the total flux of the specific volume J S
ð1=rÞ we have
JS
ð1=rÞ ðr,tÞ ¼ Àhui À u00 þ u00 ¼ Àhui: (3.18*)

Thus, the substantial averaged specific volume balance equation takes the
following ultimate form:
r
D
Dt
ð1=rÞ ¼ div hui: (3.21)
Finally, below we widely use the relation
r ð1=rÞ00
¼ Àr0
= r; (3.22)
between the fluctuations in density r0
and specific volume ð1=rÞ00
. This relation
follows directly from the definition of ð1=rÞ00
:
ð1=rÞ00
¼ 1=r À 1=r ¼ 1=r À 1=r ¼ r À rð Þ=r r ¼ Àr0
= r r:
3.1.2.3 Chemical Component Balance Equations for Averaged Motion
To derive the averaged diffusion equations, we assume in (3.14) thatA Za ¼ na=r.
The quantities JðZaÞ Ja and sðZaÞ sa ¼
Pr
s¼1
na sxs are then, respectively, the
diffusion fluxes of components a and the generation rates of particles of type a in
chemical reactions [see Sect. 2.1]. As a result, the sought-for balance equation
takes the form
r
DhZai
Dt

@ na
@t
þ div nahuið Þ ¼ ÀdivJS
a þ
Xr
s¼1
na sxs; ða ¼ 1; 2; . . . ; NÞ (3.23)
where
JS
a ðr,tÞ Ja þ Jturb
a (3.24)
is the total diffusion flux of component a in an averaged turbulized medium;
Jturb
a ðr,tÞ rZ00
a u00 ¼ rhZ00
a u00
i ¼ nau00 (3.25)
is the turbulent diffusion flux of a substance of type a; hZai na=r.
Using the weighted-mean averaging properties (3.7), it is easy to obtain a
different (more traditional) form for the turbulent diffusion flux: Jturb
a ¼ n0
au0 À
na=rð Þ r0u0. The cumbersomeness of this expression compared to (3.25) once again
suggests that using the weighted Favre averaging for a turbulized mixture with a
variable density is efficient.

Applying the averaging operator (3.3) to equalities (2.8) and (2.9) yields their
equivalents for averaged motion:
XN
b¼1
mbhZbi ¼ 1 ð1
Þ;
XN
b¼1
mbJb ¼ 0 ð2
Þ;
XN
b¼1
mbsb ¼ 0 ð3
Þ: (3.26)
In addition, the identity
XN
b¼1
mbJturb
b ¼
XN
b¼1
mbr Zbu00 ¼ r
XN
b¼1
mbZb
!
u00 ¼ r u00 ¼ 0
is valid for the turbulent diffusion ﬂuxes Jturb
a and, hence,
XN
b¼1
mbJS
b ðr,tÞ ¼ 0: (3.27)
Thus, the averaged diffusion equations (3.23) for a multicomponent turbulized
continuum, just like their regular analogs (2.7), are linearly dependent; for this
reason, one of them can be replaced by the algebraic integral (3.27).
3.1.2.4 Averaged Momentum Equation
The equation of averaged motion for a mixture (called the Reynolds equation in
the literature) can be derived from (3.14) by assuming that A u. In this case, the
quantity JðuÞ ÀP (viscous stress tensor) corresponds to the surface forces [see
(2.11)], as in ordinary hydrodynamics, while the source density
sðuÞ À
@p
@r
þ 2r u Â O þ r
XN
a¼1
ZaFa
is related to the volume forces acting on a unit volume of a multicomponent mixture
(below, we neglect the ﬂuctuations in O and Fa). As a result, the averaged equation
of motion can be written in vector form as
r
Dhui
Dt
¼ À
@ p
@r
þ
@
@r
PS

þ 2rhui Â O þ r
XN
a¼1
hZaiFa: (3.28)
Here,
PS
ðr,tÞ ÀJS
ðuÞ ¼ ÀJðuÞ À Jturb
ðuÞ ¼ P þ R (3.29)

is the total stress tensor in a turbulized flow acting as the viscous stress tensor with
respect to the averaged motion; P is the averaged viscous stress tensor describing
the momentum exchange between fluid particles due to the forces of molecular
viscosity; and
Rðr,tÞ ÀJturb
ðuÞ ¼ Àru00u00 ¼ Àrhu00
u00
i (3.30)
is the so-called Reynolds tensor having the meaning of additional (turbulent)
stresses. The appearance of tensor R in (3.28) is a direct consequence of the
nonlinearity of the original (instantaneous) equations of motion (2.9). The Reynolds
tensor written in Cartesian coordinates is
Rijðr; tÞ Àru00
i u00
j ¼
Àru00
1
2 Àr u00
1 u00
2 Àr u00
1u00
3
Àr u00
2u00
1 Àr u00
2
2 Àr u00
2u00
3
Àr u00
3 u00
1 Àr u00
3u00
2 Àr u00
32
0
B
B
@
1
C
C
A; (3.31)
whereu00
1 ; u00
2, andu00
3 are the velocity fluctuation components relative to thex1; x2, andx3
axes, respectively. It is a symmetric second-rank tensor and describes the turbu-
lent stresses attributable to the interaction of moving turbulent vortices. The turbulent
stresses, like the molecular ones, are actually the result of momentum transfer from
some fluid volumes to others but through turbulent mixing produced by turbulized fluid
velocity fluctuations. When turbulent mixing dominates in a flow (e.g., in the case of
developed turbulence emerging at very large Reynolds numbers), the averaged viscous
stress tensorP can generally be neglected compared to the Reynolds stressesR (except
the viscous-sublayer regions bordering the solid surface). The turbulent stress tensor
components Rijðr,tÞ are, thus, new unknown quantities. Note once again that the
construction of various shear turbulence models is actually associated with the pro-
posed methods of finding the closing relations for these quantities [see Chap. 4].
As has been pointed out above, the choice of JðuÞ and sðuÞ is not unique and, in
general, can be different. For example, for geophysical applications the total pres-
sure of a mixture is commonly represented as the sum of two terms p ¼ pd
þ p0 ,
where pd
is the so-called dynamic pressure and p0 is the part of the pressure that
satisfies the hydrostatic equation
@p0=@xj ¼ r0gj ¼ Àr0gd3j; ðj ¼ 1; 2; 3Þ: (3.32)
Here, r0 is some constant mass density typical of the atmosphere (e.g., at the sea
level) and g ¼ 0,0, À gf g is the gravity vector, g ¼ gj j. In this case, when the
quantities JðuÞ and sðuÞ are defined by the relations

JðuÞ pd
U À P;
sðuÞ Dr U Á gð Þ þ 2ru Â O þ r
XN
a¼1
ZaFÃ
a; ðwhere Dr ¼ r À r0Þ;
we can obtain the averaged equation of motion for a mixture in a different form:
r
Dhui
Dt
¼ U Á gð ÞDr À
@p d
@r
þ
@
@r
Á PS

þ 2rhui Â O þ r
XN
a¼1
hZaiFÃ
a: (3.33)
For flows in a free stratified atmosphere, where the buoyancy forces (the first term
on the right-hand side of (3.33)) are important, all terms in (3.33) generally have the
order gDr or smaller. Since the total pressure gradient is the sum of the dynamic and
hydrostatic pressure gradients, the following approximate equality holds:
@ p=@xj ¼ @ pd
=@xj þ @p0=@xj % Àdj3gDr À dj3r0g ¼ Àdj3r0gð1 þ Dr=r0Þ:
Hence it follows that the total pressure gradient in the cases where the estimate
Dr=r0 ( 1 is valid can be represented by the approximate relation
@ p=@xj % Àdj3r0g: (3.34)
This relation is used in Chap. 4.
3.1.3 The Energetics of a Turbulent Flow
In the averaged flow of a turbulized mixture, in contrast to its laminar analog, there
are a large number of all possible exchange mechanisms (transition rates) between
various forms of energy of the moving elementary fluid volumes that contribute to
the conserved total energy of the total material continuum. For the most comprehen-
sive physical interpretation of the individual energy balance components, we ana-
lyze here various energy equations for the averaged motion of a multicomponent
mixture, including the kinetic energy balance equation for turbulent fluctuations.
3.1.3.1 Balance Equation for the Averaged Potential Energy of a Mixture
In view of identity (3.12), the Reynolds averaging of (2.14) leads to the following
balance equation for the averaged specific potential energy of a multicomponent
mixture:
r
DhCi
Dt
¼ ÀdivJS
ðCÞ þ sðCÞ; (3.35)

where
JS
ðCÞðr,tÞ JðCÞ þ Jturb
ðCÞ ¼
XN
a¼1
CaJS
a (3.36)
is the total substantial potential energy flux in a turbulized continuum;
JðCÞ ¼
XN
a¼1
CaJa (3.37)
is the averaged molecular potential energy flux in the mixture; and
Jturb
ðCÞðr,tÞ rhC00
u00
i ¼
XN
a¼1
CarZau00 ¼
XN
a¼1
CaJturb
a (3.38)
is the turbulent potential energy flux in the mixture.
The averaged potential energy source for a multicomponent mixture is specified
by the relation
sðCÞ ¼ À
XN
a¼1
JS
a Á FÃ
a
!
À r hui Á
XN
a¼1
hZaiFa
!
: (3.39)
Here, the quantity
PN
a¼1
JS
a Á FÃ
a

is the total transformation rate (per unit mixture
volume) of the potential energy of mean motion into other forms of energy, which
follows from the comparison of (3.39) and (3.54); the quantity r hui Á
PN
a¼1
hZaiFa

is related to the transformation rate of the averaged potential energy into the kinetic
energy of mean motion [see (3.40)], with this process being reversible (adiabatic).
3.1.3.2 Balance Equation for the Kinetic Energy of Mean Motion
Scalar multiplication of (3.28) by the velocity vector hui yields an equation for the
averaged motion of a multicomponent mixture (the work-kinetic energy theorem)
in the following substantial form:
r
D
Dt
huij j2
=2

¼ pdivhui þ div Àphui þ PS
Á hui
À Á
À PS
:
@
@r
hui

þ hui Á
XN
a¼1
naFa
!
;
(3.40)

where huij j2
=2 is the specific kinetic energy of the mean motion. This equation
describes the transformation law of the kinetic energy of the mean motion into the
work of external mass and surface forces and into the work of internal forces (and
back) without allowance for the irreversible transformation of mechanical energy
into thermal one or other forms of energy.
Let us explain the physical meaning of the individual terms in (3.40):
div PS
Á hui
À Á
represents the rate at which the total surface stress PS
does the
work per unit volume of the averaged moving system; the quantity pdivhui is
related to the reversible (adiabatic) transformation rate of the averaged internal
energy (heat) into mechanical one [see (3.54)] and represents the work done by the
moving mixture flow against the averaged pressure p per unit time in a unit volume;
the sign of pdivhui depends on whether the mixture flow expands ð0divhuiÞ or
compresses ð0divhuiÞ; the quantity PS
: ð@=@rÞhui
À Á
represents the total irre-
versible transformation rate of the kinetic energy of the mean motion into other
forms of energy per unit volume [see (3.54) and (3.68)], with the energy of the mean
motion dissipating under the influence of both molecular viscosity at a rate
P : ð@=@rÞhui
À Á
0 and turbulent viscosity at a rate R : ð@=@rÞhuið Þ (generally,
this quantity can be different in sign).
Adding (3.35) and (3.40) yields the balance equation for the mechanical energy
hEmi huij j2
=2 þ h Ci for the averaged motion of a turbulized multicomponent
continuum:
r
D
Dt
huij j2
=2 þ hCi

¼ Àdiv phui À PS
Á hui þ
XN
a¼1
CaJS
a
!
þ pdivhui À PS
:
@
@r
hui

À
XN
a¼1
JS
a Á Fa
!
:
(3.41)
3.1.3.3 Heat Influx Equation for the Averaged Motion of a Mixture
We derive this equation from the general balance equation (3.14) by assuming that
A H and using the expressions
JðHÞ q; sðHÞ
dp
dt
þ P :
@
@r
u

þ
XN
a¼1
Ja Á FÃ
a
!
for the mixture enthalpy flux and source, respectively [see (2.26)]. As a result, we
have

r
DhHi
Dt
¼ ÀdivqS
þ
dp
dt
þ P :
@
@r
u

þ
XN
a¼1
Ja Á FÃ
a
!
; (3.42)
where
qS
ðr,tÞ q þ qturb
(3.43)
is the total heat flux in an averaged turbulized multicomponent continuum;
qturb
ðr,tÞ rH00u00 ffi hcpirT00u00 þ
XN
a¼1
hhaiJturb
a (3.44)
is the turbulent heat (explicit—the first term and latent—the second term) flux that
results from the correlation between the specific enthalpy fluctuations H00
and the
hydrodynamic mixture flow velocity fluctuations u00
. The approximate equality
(3.44) is written here to within terms containing triple correlations. It can be easily
obtained using the expression
H00
¼
XN
a¼1
hZaih00
a þ hhaiZ00
a þ ðZ00
a h00
aÞ00À Á
ffi hcpi T00
þ
XN
a¼1
hhaiZ00
a (3.45)
for the specific mixture enthalpy fluctuations and the properties of weighted-mean
Favre averaging suitable for this case [see (3.7)]. Here, the formulas
h00
a ¼ cpaT00
; ð1
Þ
hcpi ¼
XN
a¼1
cpahZai ð2
Þ
(3.46)
define, respectively, the fluctuations in the partial enthalpies of individual
components and the averaged specific isobaric heat capacity of a turbulized mix-
ture. Below, we assume the following relation to be valid for the averaged total
enthalpy in (3.42):
hHi ffi hcpihTi þ
XN
a¼1
h0
ahZai ¼
XN
a¼1
hhaihZai: (3.47)
This relation can be derived from (2.25) through its Favre averaging and
by neglecting the small fluctuations of the heat capacity cp in a turbulized medium
ðc00
p ffi 0Þ.
It is convenient to transform the substantial derivative of the total mixture
pressure in the expression for the source sðhÞ to

dp
dt
¼
D p
D t
þ u00 @p
@r

¼
D p
D t
þ u00 @p
@r

þ u00 @p0
@r

¼
D p
D t
þ u00 @p
@r

þ divðp0
u00
Þ À p0
divu00
:
Hence it follows that
dp
dt
¼
D p
D t
þ Jturb
ð1=rÞ Á
@p
@r

þ div p0u00
À Á
À p0divu00: (3.48)
In addition, below we use the transformation
P :
@u
@r
¼ P :
@hui
@r

þ P :
@u00
@r
¼ P :
@hui
@r

þ rhebi; (3.49)
where the formula
rhebi P :
@u00
@r

(3.50)
defines the so-called (specific) dissipation rate of turbulent energy into heat under
the influence of molecular viscosity. We note at once that the quantity hebi is among
the key statistical characteristics of a turbulized medium.
Substituting now (3.43), (3.48), and (3.49) into (3.42) yields an averaged heat
influx equation for a turbulized mixture in the following substantial form [cf. (2.24)]:
r
DhHi
Dt
¼ Àdiv q þ qturb
À p0u00
À Á
þ
Dp
Dt
þ P :
@hui
@r

À p0divu00 þ Jturb
ð1=rÞ Á
@p
@r

þ
XN
a¼1
Ja Á FÃ
a
!
þ rhebi:
(3.51)
For the subsequent analysis, we need (3.51) written via the averaged internal
energy hEi. The quantity hEi is defined by the expression
hEi ¼ hHi À
p
r
ffi hcVihTi þ
XN
a¼1
h0
ahZa i; (3.52)
which is the result of the Favre averaging of (2.32). Using the transformation
r
DhEi
Dt
þ pdivhui ¼ r
DhHi
Dt
À
Dp
Dt
; (3.53)

which is a corollary of (3.52) and (3.8), we then ultimately obtain
r
DhEi
Dt
¼ Àdiv q þ qturb
À p0u00
À Á
À pdivhui þ P :
@hui
@r

þ
XN
a¼1
Ja Á FÃ
a
!
ð1=rÞ Á
@p
@r

þ rhebi:
(3.54)
The quantity p0divu00 in (3.54) is related to the transformation rate of the kinetic
energy of turbulent vortices into the averaged internal energy [see (3.69)] and
represents the work done by the environment on the vortices per unit time in
a unit volume as a consequence of the pressure fluctuations p0
and the expansion
ðdivu00
0Þ or compression ðdivu00
0Þ of vortices. Comparison of (3.54) and (3.35)
shows that the quantity
PN
a¼1
Ja Á FÃ
a

defines the transition rate between the
averaged internal and averaged potential energies as a result of the work done by
nongravitational external forces. Similarly, comparison of (3.54) and (3.40) shows
that the quantities pdivhui and P : ð@=@rÞhui
À Á
are related to the transition
rate between the internal and kinetic energies of the mean motion. The correlation
rhebi P : ð@=@rÞu00ð Þ ffi P0
: ð@=@rÞu0ð Þ

in a developed turbulent flow
[see Chap. 4] can be identified with the mean work (per unit time per unit volume)
done by the viscous stress fluctuations on turbulent vortices with a velocity shear
ðð@=@rÞu00
6¼ 0Þ. This work is always positive, because hebi represents the dissipa-
tion rate of turbulent kinetic energy into heat under the influence of molecular
viscosity.
Let us now analyze the transformation rate Jturb
ð1=rÞ Á @p=@r

. Under the action of
buoyancy forces, it is convenient to extrapolate this quantity using (3.34) by the
expression
Jturb
ð1=rÞ Á @p=@r

% gðr0=rÞr0u00
3: (3.55)
The following two cases are known (see, e.g., van Mieghem 1973) to be
generally admissible in a turbulized fluid flow:
• For large vortices, the quantity gr0u3
00 is negative. This is because the large-scale
density fluctuation r0
(of a thermal origin) determines the sign of the vertical
vortex displacement under the effect of buoyancy. Indeed, since lightðr0
0Þand
heavy ðr0
0Þ vortices are, respectively, warm and cold ones, for example, for
warm vortices ðr0
0Þ rising ðu00
30Þ in a gravitational field gr0u00
30. Thus, large
vortices transform the thermal (internal) energy of the flow into the kinetic
energy of turbulent motion.

• For small-scale turbulence, the quantity gr0u00
3 is always positive. Indeed, in this
case, the approximate relationðr0
% Àð@ r=@x3Þx3, wherex3 is the vertical vortex
displacement or mixing length, holds for the Eulerian density fluctuation r0
. The
mean density distribution in a gravitational field is stable ½Àð@ r=@x3Þ0Š. As a
result, the turbulent vortices coming to a given level from below ½u00
30; x30Š
produce positive density fluctuations ðr0
0Þ, while those coming from above
½u00
30; x30Šproduce negative onesðr0
0Þ; whencegr0u00
30. Thus, in this case,
the buoyancy force is a restoring one, i.e., turbulence expends its energy on the
work against the buoyancy forces. The quantity gr0u00
3 represents the transforma-
tion rate of turbulent energy into averaged internal energy per unit volume of the
medium or, in other words, small-scale vortices transform the turbulence energy
into heat [см. (3.69)].
Finally, let us write the averaged heat influx equation for a multicomponent
turbulized mixture via the temperature. Using (3.46) and (3.47) for the quantities
hcpi and hHi, respectively, and the diffusion equations for mean motion (3.23), it is
easy to obtain the expression [cf. (2.27)]
r
DhHi
Dt
¼ rhcpi
DhTi
Dt
À div
XN
a¼1
hhaiJS
a
!
þ
@hTi
@r
Á
XN
a¼1
cpaJS
a
!
þ
Xr
s¼1
hqsixs;
(3.56)
where
hqsi ¼
XN
a¼1
nashhai; ðs ¼ 1; 2; . . . ; rÞ (3.57)
is the averaged heat of the sth reaction. Using this expression, the averaged heat
influx equation (3.52) takes the following final form [cf. (2.29)]:
rhcpi
D h Ti
Dt
¼ Àdiv qS
À p0u00 À
XN
a¼1
hhaiJS
a
!
þ
D p
D t
þ P :
@hui
@r

À
Xr
s¼1
hqsixs þ
XN
a¼1
Ja Á FÃ
a
!
À p0divu00
þ Jturb
ð1=rÞ Á
@p
@r

þ rhebi À
@hTi
@r
Á
XN
a¼1
cpaJS
a
!
(3.58)
(the last term is usually discarded [see Chap. 2]). This is the most general form of
the energy equation that can be used in reacting turbulence models of various
complexities, in particular, those based on simple gradient closure schemes. It is

important to emphasize that the heat influx equation (3.58) written via the averaged
temperature hTi allows the contribution from the heats of individual chemical
reactions to the energetics of a turbulized reacting gaseous medium to be separated
out in explicit form, with the chemical source being an averaged quantity in the case
of a turbulent flow. The nonlinearity of the algebraic dependence of the reaction
rate xsðT; naÞ on the mixture temperature and composition implies that, generally,
the quantities xs cannot be calculated only from the averaged mixture temperature
and composition (i.e., xs 6¼ xsðhTi; naÞ), because they depend significantly on the
intensity of the turbulent fluctuations in these parameters. We postpone a detailed
consideration of this question to the next chapter.
3.1.3.4 Total Energy Conservation Law for the Averaged Motion
of a Mixture
Let us now write out the averaged total energy conservation law for a turbulized
multicomponent mixture in substantial form. This equation allows us to obtain, in
particular, the transfer equation for turbulent energy (the averaged kinetic energy of
the turbulent velocity fluctuations), which is fundamental in the turbulence theory.
Applying the averaging operator (3.3) to (2.16) and using relations (2.17) and (2.18)
for the quantities Eðr; tÞ and JðEÞ, we have
r
D Utoth i
Dt
þ div JUtot
þ Jturb
Utot

¼ 0; (3.59)
where
Utoth i ¼ uj j2
=2
D E
þ C þ hEi; (3.60)
is the total specific energy of the averaged continuum;
Jturb
Utot
ðr,tÞ rhU00
totu00
i ¼ rð uj j2
=2 þ C þ EÞu00 (3.61)
is the turbulent total energy flux in the mixture; and
JUtot
q þ p u À P Á u þ
XN
a¼1
CaJa
¼ q þ p hui À P Á hui þ pu00 À P Á u00 þ
XN
a¼1
CaJa
(3.62)
is the averaged molecular total energy flux in the multicomponent medium.

For the subsequent analysis, it is convenient to transform the kinetic energy of
the instantaneous motion of the medium as
r uj j2
=2 r hui þ u00
ð Þ Á hui þ u00
ð Þ=2 ¼ r uh ij j2
=2 þ r uh i Á u00
ð Þ þ r u00
j j
2
=2:
Performing the (Reynolds) averaging of this expression yields
r uj j2
=2 r uh ij j2
=2 þ r u00j j2
=2;
or
h uj j2
=2i huij j2
=2 þ hbi; (3.63)
where the formula
bh iðr,tÞ rb=r ¼ r u00j j2
=2r (3.64)
defines yet another key statistical characteristic of turbulent motion—the turbulent
energy; the quantity bðr,tÞ u00
j j2
=2 represents the specific fluctuation kinetic
energy of the flow. As a result, (3.60) and (3.61) can be rewritten as
hUtoti ¼ huij j2
=2 þ hCi þ hEi þ hbi; (3.65)
Jturb
Utot
ðr,tÞ rhU00
totu00
i ¼ rð uj j2
=2 þ C þ EÞu00
¼ rhbu00
i À R Á hui þ Jturb
ðCÞ þ Jturb
ðEÞ ; (3.66)
where the correlation function
Jturb
E ðr,tÞ rhE00
u00
i ¼ rðH À p=rÞu00 ¼ qturb
À pu00 (3.67)
defines the turbulent specific internal energy flux in the mixture.
Finally, combining (3.38), (3.62), (3.66), and (3.67), we rewrite the balance
equation (3.59) for the total energy of the mean motion of a turbulized mixture as
r
D
Dt
huij j2
2
þ hCi þ hEi þ hbi
!
þ div qS
À p0u00 þ r b þ
p0
r

u00
À P Á u00 þ phui À PS
Á hui þ
XN
a¼1
caJS
a
!
¼ 0: (3.68)
Here, qS
ðr,tÞ q þ qturb
is the total heat flux due to the averaged molecular and
turbulent transport; p hui is the mechanical energy flux; PS
Á hui is the total energy

flux due to the work done by the viscous and turbulent stresses; ðr b u00 À P Á u00Þ is
the turbulent vortex energy flux as a result of turbulent diffusion; and
PN
a¼1
CaJS
a is
the total potential energy flux due to the averaged molecular and turbulent diffusion.
It should be emphasized that the term p0u00 in (3.68) does not act as the energy
flux, because, as is easy to see, it drops out of the complete energy equation and is
introduced here and below for convenience.
3.1.3.5 Turbulent Energy Balance Equation
The turbulent energy balance equation (or some of its modifications), which is
fundamental in the turbulence theory, is known to underlie many present-day
semiempirical turbulence models (see, e.g., Monin and Yaglom 1992). It can be
derived by various methods, one of which is presented in Chap. 4. Here, we
consider its derivation for a multicomponent mixture based on the above averaged
energy equations.
Subtracting (3.41) and (3.54) from (3.59) we obtain the sought-for balance
equation for the specific turbulent energy hbi r u00j j2
=2r in the following
general form:
r
Dhbi
Dt
¼ ÀdivJturb
hbi þ shbi; ð1
Þ
Jturb
hbi rð u00j j2
=2 þ p0=rÞu00 À P Á u00; ð2
Þ
shbi R :
@hui
@r

þ p0divu00 þ
XN
a¼1
Jturb
a Á FÃ
!
À Jturb
ð1=rÞ Á
@p
@r

À rhebi; ð3
Þ
(3.69)
where Jturb
hbi ðr; tÞ and shbiðr; tÞ are, respectively, the turbulent diffusion flux and the
local source (sink) of the averaged kinetic energy of turbulent fluctuations (turbu-
lent energy). The left-hand part of this equation characterizes the change in
turbulent energy hbi with time and the convective transport of hbi by the averaged
motion; the second term in the right-hand part of (3.69(3)
) represents the work done
by the pressure forces in the fluctuation motion; the third and fourth terms represent
the turbulence energy generation rate under the action of nongravitational forces
and buoyancy; finally, the fifth term represents the dissipation rate of turbulent
kinetic energy into thermal internal energy due to molecular viscosity. The quantity
R : ð@hui=@rÞ on the right-hand sides of (3.40) and (3.69(3)
) has opposite signs and,
hence, it can be interpreted as the transition rate of the kinetic energy of the mean
motion into the kinetic energy of turbulent fluctuations. It is important to emphasize
once again that this energy transition is a purely kinematic process dependent only
on the choice of the turbulent field averaging procedure. Since it is well known that

R : ð@hui=@rÞ0 in the case of small-scale turbulence, the latter always transforms
the kinetic energy of the mean motion into the kinetic energy of turbulent
fluctuations. This is the so-called dissipative effect of small-scale turbulence.
However, large-scale turbulence can transform the turbulence kinetic energy into
the energy of the mean motion (see van Mieghem 1973).
3.1.3.6 Heat Influx Equation for Quasi-stationary Turbulence
In many practical applications, the heat influx equation (3.54) for a turbulized
mixture in its general form defies solution. However, it can be simplified consider-
ably in some special cases. If a stationary-nonequilibrium state in which the
turbulent energy hbi r u00j j2
=2r is conserved both in time and in space is
established in the structure of the fluctuation field in the case of developed turbu-
lence, then shbi ffi 0. In this case, an important relation follows from (3.69(3)
):
R :
@hui
@r

¼ Àp0divu00 À
XN
a¼1
Jturb
a Á FÃ
!
þ Jturb
ð1=rÞ Á
@p
@r

þ rhebi =E;b:
Using this relation, the heat influx equation for a turbulized mixture (3.54) can be
written in an almost “classical” form [cf. (2.22)]:
r
DhEi
Dt
¼ Àdiv qS
À p0u00
À Á
À pdivhui þ PS
:
@hui
@r

þ
XN
a¼1
JS
a Á FÃ
a
!
:
(3.54*)
Accordingly, the averaged heat influx equation (3.58) written via the tempera-
ture takes the form
rhcpi
DhTi
Dt
¼ Àdiv qS
À p0u00 À
XN
a¼1
hhaiJS
a
!
þ
Dp
Dt
þ PS
:
@hui
@r

À
Xr
s¼1
hqsixs:
(3.58*)
3.1.4 Equation of State for a Turbulized Mixture as a Whole
The averaged equations of motion for a turbulized reacting mixture should be
supplemented with the averaged equation of state for pressure. Throughout this
book, the multicomponent gas mixture is considered as a compressible baroclinic
medium for which the equation of state for pressure is the equation of state for a

mixture of perfect gases (2.31). Applying the statistical averaging operator (3.3) to
the equation of state (2.31), we obtain the following exact expression for the
averaged pressure:
p ¼
XN
a¼1
pa ¼ rkBh T i
XN
a¼1
hZai þ rkB
XN
a¼1
h T00
Z00
a i
¼ rkBhTi
XN
a¼1
hZai 1 þ
hT00
Z00
a i
h T ihZa i

: (3.70)
Generally, it contains a large number of correlation functions that relate the
fluctuations in the temperatures and concentrations of individual components. In
the cases where the correlation terms hT00
Z00
a i are small compared to the first-order
terms h T ihZa i (e.g., when ma ffi m; in this case, we have Za ffi na=n m ¼ xa=m,P
a hT00
Z00
a i ¼
P
a hT00
Zai % hT00
i=m ¼ 0), the equation of state for pressure relates
the averaged density, temperature, and pressure in a turbulent flow in the same way
as in a regular flow:
p ¼ rkBhTi
XN
a¼1
hZai ¼ rhRÃ
ihTi; (3.71)
where
hRÃ
i ¼ kB
XN
a¼1
hZai ¼ kB n=r (3.72)
is the Favre-averaged “gas constant” of the mixture. The thermal equation of state
(3.71) is usually applied in simple models of multicomponent turbulence based on
gradient closure hypotheses.
3.1.5 The Closure Problem of the Averaged Equations
for a Mixture
Thus, we derived the basic hydrodynamic partial differential equations designed to
describe turbulent flows (on the scale of mean motion) of gas-phase reacting
mixtures within the continuum model of a multicomponent medium based on the
general mass, momentum, and energy conservation laws using weighted-mean
Favre averaging1
. These equations are the same in general form as the
1
Note that Favre averaging allowed us to obtain exact balance equations for various quantities
conserved in a flow, because when deriving them we made no simplifying assumptions as a result
of which it would be possible to discard a priori some indefinite terms in the averaged equations.

hydrodynamic equations of a reacting mixture for laminar motion presented in
Chap. 2. However, the system of averaged turbulent equations (3.21), (3.23), (3.28),
(3.54), (3.69), and (3.72) is not closed, because it contains new indefinite fluxes that
emerged when averaging the original nonlinear hydrodynamic equations for a
mixture, along with the mean thermohydrodynamic state parameters r; hui; p; hTi,
hZai and their derivatives. It can be seen from this system that, apart from the
averaged molecular fluxes q; P; Ja, and xs, the averaged motion is also described by
the unknown mixed second-order (one-point and one-time) correlation moments.
This raises the central problem of the turbulence theory (known as the closure
problem) associated with the construction of defining relations for all of the indefi-
nite quantities that appear in the turbulent averaged hydrodynamic equations. This
problem for a chemically active multicomponent mixture is also coupled with
additional difficulties. The first difficulty is related to the necessity of allowance
for the compressibility of the total continuum corresponding to the fluid motion
under consideration. The existence of density gradients is one of the most important
properties of reacting flows that was barely considered by the classical models of
nonreacting turbulence. In particular, turbulent convective flows were considered in
meteorology exclusively in the Boussinesq approximation. In this approximation,
the density change is known to be taken into account only in the terms describing the
influence of the acceleration due to gravity. However, this approach is absolutely
inapplicable, for example, to slow (deflagration) turbulent burning, when multiple
density changes emerge in the flow. The second difficulty (to be considered in more
detail in Chap. 4) is revealed when modeling a large number of additional pair
correlations of temperature and concentration fluctuations. These appear (as shown
below) when averaging the source terms of substance production sa in the diffusion
equations (3.23) describing the change in the composition of a reacting mixture. The
evolutionary transfer equations for such correlations in the case of turbulized motion
of a compressible reacting mixture are complicated enormously.
Regarding the averaged molecular fluxes, it is important to note the following:
since the Favre averaging does not allow their regular analogs given, for example,
in the Chap. 2 of this book (in particular, as is easy to verify, the Reynolds
averaging of the Navier–Stokes relation (2.64) for the viscous stress tensor P
complicates considerably its form when using the weighted mean value hui for
the velocity) to be easily averaged, from the viewpoint of consistently constructing
a phenomenological model of compressible turbulence, it seems more appropriate
to directly derive the defining relations for these fluxes in terms of an averaged
turbulized continuum, for example, by the methods of nonequilibrium thermody-
namics, as was done in Sect. 2.3 for their regular analogs.
It is also appropriate to perform this procedure, because the linear algebraic
relationships (turbulence models) between the turbulent fluxes appearing in the
averaged hydrodynamic equations and the averaged state parameters of the medium
(or their derivatives), which are assumed to be known or can be easily calculated,
can be obtained simultaneously and by exactly the same thermodynamic method
(see Kolesnichenko 1980). We are talking primarily about the turbulent heat flux

qturb
(3.44), the turbulent diffusion fluxes Jturb
a ða ¼ 1; 2; . . . ; NÞ (3.25), the
turbulent Reynolds stresses R (3.30), and the large number of pair correlations
hZ00
a T00
i and hZ00
a Z00
bi ða; b ¼ 1; 2; . . . ; NÞ that enter explicitly into the averaged
equation of state for pressure (3.70) or appear (for a chemically active mixture)
when averaging the source terms in the diffusion equations (3.95). In addition, it
is required to also model the turbulent specific volume flux Jturb
ð1=rÞ (3.20) related
to the density fluctuations, the averaged source terms of mass production sa in
chemical reactions, and a number of unknown correlation terms including the
pressure fluctuations.
Recall that the simplest closure schemes based on the Boussinesq gradient
hypothesis (Boussinesq 1977) initially gained the widest acceptance in the simplest
turbulence models for an incompressible single-component fluid (including those
with a passive admixture that does not affect the dynamic regime of turbulence).
This approach allows the unknown turbulent mass, momentum, and energy fluxes
to be related linearly to the gradients of the medium’s averaged state parameters via
some local proportionality coefficients, the so-called turbulent transport (or
exchange) coefficients. For a compressible multicomponent mixture, such relations
were first derived in the most general form by the methods of nonequilibrium
thermodynamics (Kolesnichenko and Marov 1984) and are given in the next
section. Using the gradient closing relations for turbulent flows, we can write the
turbulent averaged hydrodynamic equations for a reacting mixture in exactly the
same form as that for a regular motion. In particular, this allows the hydrodynamic
problems for which the transitions of a laminar reacting gas mixture flow to a
turbulent one are very important to be solved numerically. At the same time, it
should be noted that the gradient hypothesis by no means solves the closure
problem unless some additional assumptions about the turbulent exchange
coefficients are made and the methods of their calculation are specified. Moreover,
this approach is completely inapplicable when the influence of the turbulization
prehistory on the local flow characteristics is significant; in these cases, adequate
turbulent exchange coefficients cannot be determined at all (see Ievlev 1990).
An objective assessment of the status of the first-order closure problem shows
that, in fact, no general phenomenological theory of turbulent heat conduction and
turbulent diffusion for multicomponent mixtures has been developed as yet. As has
already been pointed out, the gradient relations widely used in the literature (see,
e.g., Monin and Yaglom 1992; van Mieghem 1973; Lapin and Strelets 1989) are not
general enough and were derived mainly for a single-fluid medium with a passive
admixture. This necessitates considering more general approaches to the closure of
the turbulent equations for a mixture at the level of first-order models, for example,
through thermodynamic modeling of the turbulence of a compressible continuum.
In this case, the Onsager formalism of nonequilibrium thermodynamics allows the
most general structure of the defining (rheological) relations to be obtained for
turbulent flows, including those in the form of generalized Stefan–Maxwell
relations for turbulent multicomponent diffusion and the corresponding expression
for the total heat flux. At the closure level under consideration, these defining

relations appear to describe adequately the turbulent heat and mass transport in a
multicomponent medium. However, since the experimental data on turbulent
exchange coefficients are limited, simpler models still have to be often used in practice.
Thus, our subsequent objective is to derive explicit gradient expressions for the
averaged molecular and turbulent heat, momentum, and mass transfer fluxes, i.e.,
to obtain the so-called defining relations for turbulence in a purely phenomenologi-
cal way using the methods of extended nonequilibrium thermodynamics.
3.2 Rheological Relations for the Turbulent Diffusion
and Heat Fluxes and the Reynolds Stress Tensor
This section of the monograph is devoted to developing a thermodynamic model of
multicomponent turbulence that describes the relationships between the correlation
moments in the averaged hydrodynamic equations for a mixture and the averaged
thermohydrodynamic variables that are known or can be easily calculated. Here,
within the framework of nonequilibrium thermodynamics, we develop a method of
deriving the closing gradient relations for the turbulent diffusion, Jturb
a ðr; tÞ, and
heat, qturb
ðr; tÞ, fluxes and for the Reynolds stress tensor Rðr; tÞ that generalize the
corresponding results of regular hydrodynamics presented in Chap. 2 to the turbu-
lent motion of a multicomponent mixture. The phenomenological turbulence model
developed here is based on the representation of the mixture fluctuation motion by a
thermodynamic continuum that consists of two interacting open subsystems
(continua): the subsystem of averaged motion obtained by the probability-theoretic
averaging of the hydrodynamic equations for an instantaneous mixture flow and the
subsystem of turbulent chaos (the so-called turbulent superstructure) related to the
fluctuation motion of the medium (Kolesnichenko 1998). We emphasize at once
that the proposed “two-fluid turbulence model,” just like the model of two fluids in
the theory of helium superfluidity, is only a convenient way of phenomenologically
describing such a complex phenomenon as hydrodynamic turbulence and does not
purport to explain completely the physics of the process. Nevertheless, it allows, in
particular, not only the “classical” gradient relations for a single-component
turbulized fluid but also the most general structure of such relations for a turbulized
multicomponent medium to be obtained using the Onsager formalism of nonequi-
librium thermodynamics.
Here, by averaging the fundamental Gibbs identity, which is assumed to be valid
for the system’s micromotions, we derive the balance equation for the averaged
entropy hSi of a turbulized medium and find an explicit form for the flux JS
hSiðr; tÞ of
entropy hSi and its local production shSiðr; tÞ due to irreversible physical processes
both within the subsystem of averaged motion and during the interaction with
the subsystem of turbulent chaos. Such characteristics as the turbulization entropy
Sturbðr; tÞand temperatureTturbðr; tÞas well as the pulsation pressure pturbðr; tÞcan be
introduced by postulating the Gibbs thermodynamic identity for the subsystem of
3.2 Rheological Relations for the Turbulent Diffusion and Heat Fluxes. . . 215

turbulent chaos (see Nevzglyadov 1945a, b). These generalized parameters are
related to the turbulent fluctuations and dynamical changes in a quasi-stationary
state of chaos in exactly the same way as, for example, the local equilibrium
entropy Sðr; tÞ is related to the molecular fluctuations and dynamical changes in
a quasi-equilibrium state. Using the balance equation for the total entropy SS
hSi þ Sturb of a turbulized mixture, we obtain linear gradient relations for the
turbulent diffusion and heat fluxes and the Reynolds stress tensor. We give a
detailed derivation of these relations for isotropic turbulence, when the statistical
properties of the turbulent field do not depend on the direction. We derive
generalized Stefan–Maxwell relations for turbulent multicomponent diffusion
and an expression for the turbulent heat flux that describe most comprehensively
the heat and mass exchange in a turbulent mixture flow.
3.2.1 Balance Equation for the Weighted-Mean Entropy
of a Mixture
In this chapter, we perform a thermodynamic analysis of the motion of a turbulized
multicomponent medium by assuming that the one-point correlations hA00
B00
i for
any (not equal to the hydrodynamic flow velocity u) fluctuating thermodynamic
parameters A and B are small compared to the first-order terms hAihBi and can be
omitted, i.e., we assume below that
hA00
B00
i
hAihBi
( 1; ðA 6¼ u; B 6¼ uÞ: (3.73)
We obtain the balance equation for the weighted-mean specific entropy hSiðr; tÞ
rS=r of a turbulent mixture by the statistical averaging (3.5) of the evolutionary
equation (2.36) for the fluctuating entropy S:
r
DhSi
Dt

@
@t
rhSið Þ þ div rhSihuið Þ ¼ Àdiv JðSÞ þ Jturb
hSi

þ shSi: (3.74)
Here, shSiðr; tÞ sðSÞ is the local production of the averaged mixture entropy, i.e.,
the production of hSiðr; tÞ per unit time per unit volume of the medium; JðSÞ and
Jturb
hSi ðr,tÞ rS00 u00 are the averaged instantaneous molecular entropy flux of the
mixture and the turbulent entropy flux of the subsystem of averaged motion,
respectively.
There are two possible ways to obtain (decipher) an explicit form of the
expressions for JðSÞ , Jturb
hSi , and shSi in (3.74): either to average (e.g., over an
ensemble of possible realizations) their respective instantaneous analogs or to
compare the averaged equation (3.74) with the equation derived from the averaged

Gibbs identity (2.37) once the respective substantial derivatives of the averaged
state parameters h1=ri; hZai, and hEi have been eliminated from it. Here, we make
use of the latter way.
3.2.1.1 Averaged Gibbs Identity
Averaging the fundamental Gibbs identity (2.37) (written along the trajectory of the
center of mass of a physical elementary volume), which is valid for mixture
micromotions, leads to the following equation for the weighted-mean specific
entropy hSi and specific internal energy hEi of a mixture (Kolesnichenko 1998)
rhTi
DhSi
Dt
¼ r
DhEi
Dt
þ rp
Dh1=ri
Dt
À r
XN
a¼1
hmai
DhZai
Dt
þ D: (3.75)
Here, we use the following notation:
D ÀT00 rdS=dt À hTidiv rS00u00
À Á
þ div rE00u00
À Á
þ pdivu00
À
XN
a¼1
m00
ardZa=d t À
XN
a¼1
hmaidiv rZ00
a u00
À Á
:
(3.76)
It can be shown that if the same thermodynamic relations are valid for the
averaged thermodynamic parameters as those for their values in the case of
micromotions (and this is true when condition (3.73) is met) and, in particular, if
the basic thermodynamic identities
hGi
XN
a¼1
hmaihZai ¼ hEi þ ph1=ri À hTihSi; ð1
Þ
hSidhTi þ
XN
a¼1
hmaidhZai ¼ dhEi þ pdh1=ri; ð2
Þ
(3.77)
are valid, then D 0 (here, d denotes an increment of any form), i.e., the
fundamental Gibbs identity (3.75) in its substantial form retains its “classical”
form for the subsystem of averaged motion as well (Kolesnichenko 1980).
Indeed, averaging the identity
dðrAeÞ À TdðrASÞ þ pdA À
XN
a¼1
madðrAZaÞ 0;
which holds for any field quantity A, over an ensemble of possible realizations,
we have

0 ¼ d rhAihEið Þ À hTid rhAihSið Þ þ pdhAi
À
XN
a¼1
hmaid rhAihZaið Þ ¼ À d rA00
E00

þ hTid rA00
S

þ T00d r SAð Þ À pdA00
þ
XN
a¼1
hmaid rZ00
a A00

þ
XN
a¼1
m00
a d rZaAð Þ;
(3.78)
in view of assumption (3.77), the left-hand side of this equality is equal to zero for
any A. Setting successively A ¼ 1 and A ¼ u in (3.78), we obtain, respectively, the
following two identities:
T00
@ðrSÞ
@t
þ
XN
a¼1
m00
a
@ðrZaÞ
@t
¼ 0; ð1
Þ
À
XN
a¼1
hmaidivðrZ00
a u00Þ þ divðrE00u00Þ À hTidivðrS00u00Þ
À T00divðrSuÞ þ pdivu00 À
XN
a¼1
m00
adivðrZauÞ ¼ 0; ð2
Þ
(3.79)
from which, as is easy to see, it follows that D 0.
3.2.1.2 Formula for the Production of the Weighted-Mean
Entropy of a Mixture
Let us now eliminate the substantial derivatives of the parameters ð1=rÞ; hZai
ða¼ 1; 2; . . . ; NÞ, and hEi from the right-hand side of the averaged Gibbs relation
(3.75) using the averaged equations (3.21), (3.23), and (3.54). As a result, we
obtain a substantial balance equation for the averaged mixture entropy hSiðr; tÞ
in the following explicit form [cf. (2.39) and (2.40)]
r
DhSi
Dt
þ div
qS
À
PN
a¼1
hmaiJS
a
hTi
0
B
B
@
1
C
C
A ¼ sh Si ¼ s
ðiÞ
h Si þ s
ðeÞ
h Si; (3.80)
where the local production of the averaged entropy is deﬁned by the relation

sh Si
1
hTi
À ~JS
q
@lnhTi
@r

þ

P :
@hui
@r

À
XN
a¼1
JS
a hTi
@
@r
hmai
hTi

þ hhai
@lnhTi
@r
!!
þ
Xr
s¼1
hAsixs
À
XN
a¼1
Ja Á FÃ
a
!
ð1=rÞ Á
@ p
@r

þ rhebi
)
(3.81)
Here, using the relations
hAsiðr; tÞ À
XN
a¼1
nashmai; ðs ¼ 1; 2; . . . ; rÞ (3.82)
we introduced the averaged chemical affinities h As i for reactions s in a turbulized
reacting medium [cf. (2.41)] and use the notation
~JS
q Jq þ ~Jturb
q ; Jq ffi q À
PN
a¼1
hhaiJa ; J
turb
q ~qturb
À
PN
a¼1
hhaiJturb
a ;
~JS
q ~qS
À
PN
a¼1
hhaiJS
a ; ~qS
ðr,tÞ q þ ~qturb
¼ qS
À p0u00;
JS
a Ja þ Jturb
a ; ~qturb
qturb
À p0u00
8

:
(3.83)
for the total diffusion and heat fluxes in a multicomponent turbulent continuum.
Comparing now (3.80) and (3.81) with (3.74), we obtain the following expressions
for the two entropy diffusion fluxes (the averaged molecular, JðSÞ, and turbulent,
Jturb
hSi , ones) and for the entropy production shSi in the subsystem of averaged motion:
JðSÞ
1
hTi
q À
XN
a¼1
hmaiJa
!
¼
1
hTi
Jq þ
XN
a¼1
hSaiJa; (3.84)
Jturb
hSi
1
hTi
~qturb
À
XN
a¼1
hmaiJturb
a
#
¼
1
hTi
~Jturb
q þ
XN
a¼1
hSaiJturb
a (3.85)
s
ðiÞ
hSiðr; tÞ
1
hTi
À ~JS
q Á
@lnhTi
@r

þ P :
@hui
@r

þ
Xr
s¼1
hAsixs
(
À
XN
a¼1
JS
a : hTi
@
@r
hmai
hTi

þ hhai
@lnhTi
@r
À Fa
!)
! 0;
(3.86)

s
ðeÞ
hSiðr; tÞ
1
hTi
À
XN
a¼1
Jturb
a Á Fa
!
ð1=rÞ Á
@p
@r

þ rhebi
( )

=E;b
hTi
(3.87)
Here, hmai ffi hhai À hTihSai is the averaged partial chemical potential; the
positive quantity s
ðiÞ
hSiðr,tÞ defines the local production rate of the averaged mixture
entropy hSi due to irreversible transport processes and chemical reactions within
the subsystem of averaged motion; as will be clear from the subsequent analysis, the
quantity s
ðeÞ
hSiðr,tÞ (the sink or source of entropy) reflects the entropy exchange
between the subsystems of turbulent chaos and averaged motion.
It should be noted that the quantitys
ðeÞ
hSiðr,tÞcan be different in sign, depending on
the specific regime of turbulent flow. Indeed, the dissipation rate of turbulent energy
hebiðr,tÞ is always positive. However, the energy transition rate p0divu00
(representing the work done on turbulent vortices per unit time per unit volume
by the environment due to the pressure fluctuations p0
and the expansion ð divu00
0Þ
or compression ðdivu00
0Þ of vortices) can be different in sign. The quantity Jturb
ð1=rÞÁ
ð@=@rÞp % gr0u3, which represents the turbulence energy generation rate under the
action of buoyancy forces, is positive in the case of small-scale turbulence, but it
can be both positive and negative for large vortices (see van Mieghem 1973). Thus,
it follows from (3.81) that, generally, the entropy hSi for the subsystem of averaged
motion can both increase and decrease, which is a characteristic feature of thermo-
dynamically open systems.
Note also that attributing the individual terms in (3.80) to the turbulent flux or to
the production of averaged entropy is to some extent ambiguous: a number of
alternative formulations using various definitions of the turbulent heat flux different
from (3.80) are possible. Considerations of this kind are expounded in de Groot and
Mazur (1962) and Gyarmati (1970).
3.2.2 Entropy Balance Equations and Entropy Production
for the Subsystem of Turbulent Chaos
Thus, we have made sure that the Favre-averaged entropy hSi alone is not enough
for an adequate description of all features of a turbulized continuum, because it
is not related to any parameters characterizing the internal structure of the
subsystem of turbulent chaos and, in particular, to such a paramount parameter
as the turbulence energy (the averaged fluctuation kinetic energy per unit mass of
the medium)

hbiðr; tÞ r u00j j2
=2r: (3.88)
Therefore, when a phenomenological model of turbulence is constructed, a
thermodynamic consideration of the subsystem of turbulent chaos also seems
necessary. This goal can be achieved by increasing the number of independent
variables in the thermodynamic description of this subsystem, which is in a non-
equilibrium stationary state in the case of strongly developed turbulence. Below, we
characterize the physically elementary volumed rof turbulent chaos (as a rule, when
a continuum model is constructed, infinitely small particles are considered as
thermodynamic systems for which the physical concepts of an internal state are
defined) by the following structural parameters: the extensive state variables Eturb
ðr; tÞ (internal turbulization energy density) and Sturbðr; tÞ (generalized local
turbulization entropy) and the intensive state variables Tturbðr; tÞ (generalized
turbulization temperature characterizing the intensity of turbulent fluctuations)
and pturbðr; tÞ (turbulization pressure) (Blackadar 1955). It is important to note
that such generalized parameters of the state of chaos as the turbulization entropy
Sturb and energy Eturb (considered below as primary concepts) are introduced here a
priori to ensure coherence of the theory and, in general, have no precise physical
interpretation (see Jou et al. 2001). Nevertheless, we assume below that the general
thermodynamic relations holding in a quasi-equilibrium state also remain valid for
a quasi-stationary state of turbulent chaos. In particular, an important point is the
formulation of the second law of thermodynamics that serves exclusively as a
constraint on the form of the corresponding constitutive equations. By admissible
physically real processes (i.e., processes in which a sequence of states can be
realized in the course of time within the framework of the applied model of
turbulent motion) we mean the solution of the balance conservation equations
supplemented by defining relations (obtained in a standard way) when the Clausius
principle holds: the changes in the total entropy SS ¼ Sh i þ Sturb of a turbulized
system caused by internal irreversible processes can be only positive or (in the
extreme case) equal to zero.
Let us now turn to corollaries of this formalism. Following the elegant Gibbs
method (see, e.g., Mu¨nster 2002), we choose the following fundamental
Gibbs equation (in integral form) for the generalized entropy as a local characteris-
tic function (containing all thermodynamic information about the subsystem of
turbulent chaos in a stationary state):
Sturbðr; tÞ ¼ Sturb Eturbðr; tÞ; 1=rðr; tÞð Þ; (3.89)
this functional relation is assumed to be specified a priori. Let us now take, as is
usually done in the formalized construction of classical locally equilibrium ther-
modynamics, the following definitions of the conjugate variables Tturbðr; tÞ and pturb
ðr; tÞ (by assuming all these derivatives to be positive):
1=Tturb @Sturb=@Eturbf g1=r; pturb=Tturb @Sturb=@ð1=rÞf gEturb
:

The meaning of generalized (turbulization) temperature and pressure can then be
assigned to the intensive variables Tturbðr; tÞ and pturbðr; tÞ , respectively. The
corresponding differential form of the fundamental Gibbs equation (3.89) written
along the trajectory of the center of mass of a physically elementary volume is
Tturbðr; tÞ
D
Dt
Sturbðr; tÞð Þ ¼
D
Dt
Eturbðr; tÞð Þ þ pturbðr; tÞ
D
Dt
1
rðr; tÞ

: (3.90)
Obviously, it is admissible to interpret the various kinds of functional relations
between the variablesEturb; Tturb; pturb, andSturb, which can be derived by a standard
(for thermodynamics) method from (3.90), as the “equations of state” for the
subsystem under consideration. Below, we identify the quantity Eturbðr; tÞ with
the turbulence energy
Eturbðr; tÞ hbiðr; tÞ þ const ¼ r u00j j2
=2r þ const (3.91)
and assume that the subsystem of turbulent chaos in the thermodynamic sense is
a perfect classical gas with three degrees of freedom in which the energy is
distributed uniformly (the key hypotheses of the model). In particular, we then have
hbi ¼ cturb
V Tturb ¼ 3
2 RÃ
Tturb ¼ 3
2pturb=r; pturb ¼ RÃ
Tturbr;
Sturb ¼ 3
2 RÃ
ln pturb=r
5
3

þ const:
(3.92)
We derive the corresponding balance equation for the turbulization entropy Sturb
from (3.90) by the above method [see Sect. 2.2] using (3.21) for the speciﬁc volume
ð1=rÞ and the balance equation (3.69) for the turbulent energy hbi; as a result, we
obtain
r
DSturb
Dt
þ divJðSturbÞ ¼ sðSturbÞ s
ðiÞ
ðSturbÞ þ s
ðeÞ
ðSturbÞ; (3.93)
where
JðSturbÞ
1
Tturb
rð u00j j2
=2 þ p0=rÞu00 À P Á u00

¼
1
Tturb
Jturb
hbi ; (3.94)
0 s
ðiÞ
ðSturbÞ ¼
1
Tturb
À Jturb
hbi Á
@lnTturb
@r

þ R :
@hui
@r

þ pturbdivhui
'
; (3.95)
s
ðeÞ
ðSturbÞ
1
Tturb
XN
a¼1
Jturb
a Á FÃ
!
þ p0divu00 À Jturb
ð1=rÞ Á
@ p
@r

À rhebi
( )
À
=E;b
Tturb
(3.96)

Here, JðSturbÞðr; tÞ is the substantial ﬂux of the entropy Sturb for the subsystem of
turbulent chaos; the quantities s
ðiÞ
ðSturbÞ and s
ðeÞ
ðSturbÞ mean the local production and sink
rates of the ﬂuctuation entropy Sturb, respectively.
For the subsequent analysis, it is convenient to decompose the gradient of the
averaged velocity @hui=@r (a second-rank tensor) in (3.86) and (3.95) into symmet-
ric and antisymmetric parts [see (2.43)],
@hui=@r ¼ @hui=@rð Þs
þ @hui=@rð Þa
¼ S þ
1
3
Udivhui þ @hui=@rð Þa
;
(3.97)
and represent the symmetric Reynolds stress tensor R (given the equation of
state (3.92)) as
R
0
R À
1
3
R:Uð ÞU ¼ R þ pturbU ¼ R þ
2
3
rhbiU; (3.98)
where
pturb ¼ À
1
3
ðR : UÞ; D @hui=@rð Þs
;
S D
0
@hui=@rð Þs
0
¼ D À
1
3
Udivhui
(3.99)
are, respectively, the turbulization pressure, the strain rate tensor, and the shear rate
tensor for a turbulized continuum. The scalar product of the Reynolds tensor and the
velocity gradient can then be written as R : ð@=@rÞhuið Þ ¼ R
0
: D
0
Àpturbdivhui and
the balance equation for the turbulization entropy Sturb (3.93) takes the form
r
DSturb
Dt
þ div
1
Tturb
Jturb
hbi
'
¼
1
Tturb
À Jturb
hbi Á
@lnTturb
@r

þ R
0
: D
0
À=E;b
'
:
(3.100)
In writing (3.100), we used the fact that the scalar product of symmetric and
antisymmetric tensors is always equal to zero.
3.2.3 Balance Equation for the Total Entropy of a Turbulized
Continuum
The introduction of two entropies, hSi and Sturb, concretizes our view of the initial
turbulized continuum as a thermodynamic complex that consists of two mutually
open subsystems—the subsystems of averaged motion and turbulent chaos.

The balance equation for the total entropy SS ¼ ð Sh i þ SturbÞ of a multicomponent
system follows from (3.80) and (3.100):
r
DSS
Dt
þ div
Jturb
hbi
Tturb
þ
qS
À
PN
a¼1
hmaiJS
a

hTi
8

:
9
=
;
¼ sS s
ðiÞ
hSi þ s
ðiÞ
Sturb
þ shSi; Sturb
;
(3.101)
where
0 sS
1
hTi
À ~JS
q Á
@lnhTi
@r

þ pdivhui

þ P
0
: D
0
þ
Xr
s¼1
hAsixsÀ
XN
a¼1
JS
a : hTi
@
@r
hmai
hTi

þ hhai
@ lnhTi
@r
À Fa
!'
þ
1
Tturb
À Jturb
hbi Á
@ ln Tturb
@r

þ R

: D

'
þ =E;b
Tturb À hTi
TturbhTi

;
(3.102)
= À
XN
a¼1
Jturb
a Á Fa
!
ð1=rÞ Á
@p
@r

þ rhebi; (3.103)
shSi; Sturb
s
ðeÞ
hSi þ s
ðeÞ
Sturb
: (3.104)
The local production of the total entropy sS related to irreversible processes
within a turbulized continuum is thus seen to be deﬁned by the set of thermody-
namic ﬂuxes ~JS
q , xs, JS
a, p, P,Jturb
hbi ,pturb, R,=E;b and their conjugate thermodynamic
forces [cf. (2.50), (2.51), (2.52), (2.53) and (2.54)]
YS
q À
1
hTi2
@ hTi
@r
¼
@
@r
1
hTi

; Yhbi À
1
T2
turb
@Tturb
@r
(3.105)
YAs

hAsi
hTi
¼ À
XN
b¼1
hmbi
hTi
nb s; ðs ¼ 1; 2; . . . ; rÞ; (3.106)
YÃ
a À
@
@r
hmai
hTi

þ hhai
@
@r
1
hTi

þ
Fa
hTi
; (3.107)
Yp
divhui
hTi
; (3.108)

YD
1
hTi
D
0
; YR
1
Tturb
D
0
; (3.109)
YE;b
Tturb À hTi
TturbhTi

: (3.110)
Using these definitions, the entropy productionsS can be written in the following
bilinear form:
0 sS ¼
Xr
s¼1
xs YAs
þ~JS
q Á YS
q þ
XN
a¼1
JS
a ÁYÃ
a þ pYp þ P
0
: YD
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
s
ðiÞ
hSi
þ Jturb
hbi Á Yhbi þ R
0
: YR
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
s
ðiÞ
Sturb
þ =E;bYE;b
zfflfflfflffl}|fflfflfflffl{
shSi;Sturb
,
(3.111)
which corresponds to three independent sources of nonequilibrium processes in a
turbulized mixture with a distinctly different physical nature.
According to the main postulate of generalized nonequilibrium thermodynamics
[see Sect. 2.2], when the thermodynamic system is near local equilibrium or near a
stable stationary-nonequilibrium state, the thermodynamic fluxes can be repre-
sented as linear functions of their conjugate macroscopic forces: Jg i ¼
P
d
Lij
gdXdj
ðg; d ¼ 1; 2; . . . fÞ. It is important to note that (3.111) allows the defining relations to
be obtained for three main regimes of a turbulized mixture flow—for an averaged
laminar flow, for developed turbulence when the turbulent fluxes are much more
efficient than the corresponding averaged molecular fluxes (Tturb ) hTi, R ) P,
qturb
) q, etc.), and finally, in the general case where the processes of averaged
molecular and turbulent transport are comparable in efficiency. As can be seen from
(3.111), the spectrum of possible cross effects for a turbulent flow is extended
considerably compared to a laminar one. Thus, for example, the reduced heat flux
~JS
q qS
À p0u00 À
PN
a¼1
hhaiJS
a in a turbulized continuum can emerge not only under
the influence of its conjugate thermodynamic forceYS
q but also through the action of
the force Yhbi conjugate to the flux Jturb
hbi (which describes the “diffusion” transfer of
turbulent energy). However, unfortunately, there are no reliable experimental data
at present that quantitatively describe such cross effects in a turbulized medium. In
addition, the contribution from any cross effects to the total transfer rate is generally
an order of magnitude smaller than that from direct effects (see de Groot and Mazur
1962). Taking these circumstances into account, we use below the requirements
that the production rates of the total entropy s
ðiÞ
hSi; s
ðiÞ
Sturb
shSi; Sturb
be positive

independent of one another, i.e., by assuming that any linear relations referring, for
example, to the subsystem of averaged motion (in particular, between the symmet-
ric part of the averaged viscous stress tensor P

with a zero trace and the tensor
viscous force YD) are not affected noticeably by the subsystem of turbulent chaos
(the tensor force YR ). We also omit a number of cross effects in the linear
constitutive relations without any special stipulations.
To conclude this section, we make two remarks:
• The quantity shSi; Sturb
describing the entropy production within the full system
through irreversible entropy exchange between the subsystems of turbulent
chaos and averaged motion is also always positive in view of the second
law of thermodynamics. Therefore, the “direction” of the thermodynamic flux
=E;bðr; tÞ is specified by the sign of the state function YE;b ð1=hTi À 1=TturbÞ,
which should be considered as the conjugate thermodynamic force (macroscopic
factor) producing this entropy flux. Such entropy exchange between two mutu-
ally open subsystems is known to be an indispensable condition for a structured
collective behavior, i.e., it can be a source of self-organization in one of them
(see Chap. 5).
• Generally, the matrix of phenomenological coefficients Lij
g d for a turbulized
continuum depends not only on averaged state parameters (temperature, density,
etc.) but also on characteristics of the turbulent superstructure itself, for exam-
ple, on the parameters r, hebi, and Tturb (or hbi). Such a situation, in which there is
a functional dependence of the tensor of kinematic coefficients Lij
g d on the
thermodynamic fluxes themselves (e.g., on the turbulent energy dissipation
rate hebi), is known to be typical for self-organizing systems (see Haken 1983,
1988). In general, it can lead to the individual terms in the sum sS being not
positive definite, although the sum itself sS ! 0. In this case, a superposition of
various fluxes, in principle, can lead to negative values of individual diagonal
elements in the matrix Lij
g d . This probably explains the effect of negative
viscosity in some turbulent flows (see Chaps. 5 and 8).
3.2.4 Linear Closing Relations for a Turbulized
Multicomponent Mixture of Gases
To concretize the gradient closing relations (constitutive Onsager laws) relating the
averaged molecular and turbulent thermodynamic fluxes to the corresponding
thermodynamic forces, we now use the formalism of nonequilibrium thermody-
namics presented in Sect. 2.2. We consider here the general case where the
averaged molecular and turbulent transport processes are comparable in signifi-
cance and restrict ourselves to the derivation of such relations for meso- and small-
scale turbulence. For the latter, as is well known, there is a tendency for local

statistical isotropy of its characteristics to be established (the statistical properties of
a turbulent flow in this case do not depend on direction). This approach can be
easily generalized to the case of nonisotropic (large-scale) turbulence.
As is well known from the general theory of tensor functions (see Sedov 1984),
the symmetry properties of isotropic media are completely characterized by a
metric tensor gij
: all tensors will be tensor functions of only the metric tensor, in
particular, Lij
gd ¼ Lgdgij
ðg; d ¼ 1; 2; . . . fÞ, where Lgd are scalar coefficients. In
addition, since there is no interference between the fluxes and thermodynamic
forces of various tensor dimensions in an isotropic system (the Curie principle),
we may consider, for example, phenomena described by polar vectors (heat con-
duction or diffusion) independently of scalar and tensor phenomena (see de Groot
and Mazur 1962). Adopting the additional hypothesis that the system is Markovian
(when the fluxes at a given time depend on the generalized forces taken at the same
time), we then obtain the following phenomenological relations (written in rectan-
gular coordinates, gij
dij) (Kolesnichenko 1998) from (3.111):
~JS
q qS
À p0u00 À
XN
a¼1
hhaiJS
a ¼ LS
qq
@
@r
1
hTi

þ
XN
b¼1
LS
qbYÃ
b; (3.112)
JS
a ¼ LS
a q
@
@r
1
hTi

þ
XN
b¼1
LS
a bYÃ
b; ða ¼ 1; 2; . . . ; NÞ; (3.113)
P
À Á
jk
0
¼ L YDð Þjk ¼ m
@huki
@xj
þ
@huji
@xk

À
2
3
djkdivhui
'
; (3.114)
p ¼
lpp
hTi
divhui þ
Xr
s¼1
lpshAsi
!
ffi m#divhui; (3.115)
xs ¼ Àlsp
divhui
hTi
þ
Xr
m¼1
lsm
hAsi
hTi
; ðs ¼ 1; 2; . . . ; rÞ; (3.116)
Rð Þjk ¼ À
2
3
rhbidjk þ Lturb YRð Þjk
¼ À
2
3
rhbidjk þ mturb @huki
@xj
þ
@huji
@xk

À
2
3
djkdivhui
'
; (3.117)
Jturb
hbi ¼ À
lb
T2
turb
@Tturb
@r
¼ À
mturb
sb
@hbi
@r
; (3.118)
=E;b ¼ lE;b
Tturb À hTi
TturbhTi

: (3.119)

Here, the formulas
m L=2hTi; m# lpp=hTi; mturb
Lturb=2Tturb; nturb
mturb
=r (3.120)
introduce the averaged molecular viscosity, mðr,tÞ, and second viscosity, m#ðr,tÞ,
coefficients needed to define the averaged viscous stress tensor P as well as the
turbulent viscosity, mturb
ðr,tÞ , and kinematic turbulent viscosity, nturb
ðr,tÞ ,
coefficients defining the turbulent stress tensor R. The coefficient sb is the “Prandtl
number” for the turbulent energy, whose value is usually assumed to be constant.
The scalar kinematic coefficients LS
qb and LS
a b , as in the laminar case [see (2.61)
and (2.63)], satisfy the Onsager-Casimir symmetry conditions LS
a b ¼ LS
ba ða,b
¼ 1,2, . . . NÞ and the conditions
XN
a¼1
maLS
qa ¼ 0; ð1
Þ
XN
a¼1
maLS
ab ¼ 0; ðb ¼ 1; 2; . . . NÞ: ð2
Þ (3.121)
It should be kept in mind that, in contrast to the ordinary molecular viscosity
coefficients m and m#, the turbulent viscosity coefficient mturb
characterizes not the
physical properties of a fluid but the statistical properties of its fluctuation motion;
that is why it can take on negative values in some cases. In addition, the well-known
increase in turbulent viscosity compared to its molecular analog once again
suggests that a turbulent motion is more ordered (organized) than a laminar one.
Indeed, the viscosity in a laminar motion is determined by the momentum transfer
at a chaotic molecular level. In contrast, in a turbulent motion, momentum is
transferred from layer to layer by collective degrees of freedom and this is an
indubitable indication of its greater order.
Regarding the defining relation (3.117) for the tensor R, we note the following:
when the turbulent field anisotropy is taken into account, this relation becomes
considerably more complicated, because it requires replacing the scalar turbulent
viscosity coefficient mturb
by a (fourth-rank) tensor [see Chap. 7 and the monograph
by Monin and Yaglom (1992)]. Note also that we managed to derive here the
defining relation (in standard form)
Pjk ¼ m
@huki
@xj
þ
@huji
@xk

À
2
3
djkdivhui
'
þ m#divhui (3.114*)
for the averaged viscous stress tensor directly, i.e., without invoking the
corresponding regular analog [see (2.64)] for a laminar motion and its subsequent
averaging.
As we see, the linear law (3.116) can also be used to obtain the limiting form of
the expressions for the averaged chemical reaction rates near a chemical equilib-
rium state. However, since this result has a limited domain of applicability, here we
not dwell on it, deferring a more detailed consideration to Chap. 4.

3.2.4.1 Heat Conduction and Diffusion in a Turbulized Mixture
Using the formal similarity of the defining relations for the vector turbulent
diffusion and heat processes specified by (3.112) and (3.113) to those for a laminar
flow [see (2.56) and (2.57)], we rewrite (using the approach developed in Sect. 2.3)
(3.112) and (3.113) as
JS
a ¼ ÀnaDS
Ta
@lnhTi
@r
À na
XN
b¼1
DS
a bdturb
b ; ða ¼ 1; 2; . . . ; NÞ; (3.122)
~JS
q ¼ À^l
S @hTi
@r
À p
XN
b¼1
DS
Tbdturb
b ; (3.123)
where
dturb
b
@
@r
nb
n

þ
nb
n
À hCbi

@lnp
@r
À
nb
p
Fb À mb
XN
a¼1
h ZaiFa
!
(3.124)
are the generalized thermodynamic forces for a turbulent mixture motion. These are
similar to the corresponding expressions (2.70) for a regular motion and can be
introduced for a turbulized mixture using the relations
dturb
b À
hTinb
p
YÃ
b À hCbi
@lnp
@r
þ
rb
p
XN
a¼1
hZaiFa; 1
À Á
XN
a¼1
dturb
a ¼ 0; ð2
Þ (3.125)
i.e., in exactly the same way as was done in Sect. 2.3.3 (here, hCbi ¼ mbnb=r is
the Favre-averaged mass concentration of particles of type b).
In relations (3.122) and (3.123), by analogy with the formulas for a laminar
fluid flow, we introduced the symmetric multicomponent turbulent diffusion
coefficients DS
ab ða; b ¼ 1; 2; . . . ; NÞ; turbulent thermal diffusion coefficients DS
Tb
ðb ¼ 1; 2; . . . ; NÞ, and turbulent thermal conductivity coefficients ^l
S
for a multi-
component gas using the definitions
^l
S

LS
qq
hTi2
; DS
Tb ¼
LS
qb
hTinb
; DS
a b ¼ DS
ba ¼
p
hTinanb
LS
a q: (3.125b)
In view of (3.121), the scalar turbulent transport coefficients DS
Tb and DS
a b satisfy
the conditions

XN
a¼1
hCaiDS
Tab ¼ 0;
XN
a¼1
hCaiDS
a b ¼ 0; ða; b ¼ 1; 2; . . . ; NÞ: (3.126)
The coefficients defined by (3.125) are the effective transport coefficients
attributable not only to the molecular mass and heat transfer from some fluid
volumes to other ones but also to the turbulent mixing produced by turbulized
fluid velocity fluctuations; therefore, it can be assumed that DS
a b Da b þ Dtyrb
a b and
^l
S
^l þ ^l
turb
. Since the cross processes related to thermal diffusion and diffusive
heat conduction for turbulized mixtures are completely unstudied at present, below
we neglect them by assuming that DS
Tab ffi 0.
Thus, the defining relations for the turbulent diffusion and heat fluxes can be
written in the following final form:
JS
a ¼ Àna
XN
b¼1
DS
a bdturb
b ; ða ¼ 1; 2; . . . ; NÞ; (3.127)
qS
À p0u00 ¼ À^l
S @hTi
@r
þ
XN
b¼1
hhbiJS
b : (3.128)
These relations describe most completely the heat and mass transfer processes in
a developed isotropic turbulent flow of a multicomponent gas mixture. Unfortu-
nately, since the experimental data on multicomponent turbulent diffusion
coefficients are limited at the current stage of development of the phenomenologi-
cal turbulence theory, more simplified models have to be used in practice. It should
also be added that the turbulent exchange coefficients introduced here, in particular,
the coefficients DS
a b, can be defined in terms of the so-called К-theory of developed
turbulence by invoking additional transfer equations for the pair correlations of
fluctuating thermohydrodynamic mixture parameters [see Chap. 4].
3.2.4.2 Generalized Stefan–Maxwell Relations for a Turbulized Mixture
Just as in the case of laminar mass and heat transfer in a mixture, it is convenient to
reduce the defining relations (3.127) and (3.128) for the turbulent diffusion and heat
fluxes (in particular, when multicomponent flows are simulated numerically) to the
form of generalized Stefan–Maxwell relations including the binary (for a binary
mixture) turbulent diffusion coefficients DS
a b . This is because, in contrast to the
multicomponent diffusion coefficients DS
a b, empirical data are, in general, easier to
use for the coefficients DS
a b.
The procedure for deriving the generalized Stefan–Maxwell relations for multi-
component diffusion in a turbulent flow does not differ in any way from that

performed in Sect. 2.3.4 when deriving these relations for a laminar mixture flow.
Using this analogy, we immediately present the final result (Kolesnichenko 1998):
XN
a ¼ 1
a 6¼ b
nbJS
a À naJS
b
n2
DS
ab
¼ dturb
b ; ðb ¼ 1; 2; . . . ; N À 1Þ;
XN
a¼1
mbJS
a ¼ 0; (3.129)
where
dturb
b
@
@r
nb
n

þ
nb
n
À hCbi

@lnp
@r
À
nb
p
Fb À mb
XN
a¼1
h ZaiFa
!
:
In the case of a direct numerical solution of these relations for the turbulent
diffusion fluxes JS
a , it is convenient to reduce them, by analogy with a laminar
mixture flow, to the form of a generalized Fick law [see (2.116)]. As a result, we
obtain
JS
b ¼ ÀDS
b ndturb
b À
1
n
XN
a ¼ 1
a 6¼ b
nb
DS
ab
JS
a
0
B
B
B
B
@
1
C
C
C
C
A
¼ Àr DS
b
@
@r
nb
r

þ dJS
b ; (3.130)
where
dJS
b nbDS
b
Â À
@lnM
@r
À 1 À
mb
M
@lnp
@r
þ
n
p
Fb À mb
XN
a¼1
hZaiFa
!
þ
1
n
XN
a ¼ 1
a 6¼ b
JS
a
DS
ab
8

:
9
=
;
;
(3.131)
DS
b
1
n
XN
a ¼ 1
a 6¼ b
na
DS
ab
0
B
B
B
B
@
1
C
C
C
C
A
À1
; M
XN
a¼1
mana=
XN
a¼1
na ¼
r
n
: (3.132)
By introducing the effective diffusion coefficient DS
b, we can simplify consider-
ably the numerical solution of the problem despite the fact that the generalized Fick

law in form (3.130) generally does not allow each diffusion equation (3.23) to be
considered separately from the other ones. However, since the methods of succes-
sive approximations are commonly used for the numerical solution of problems, the
presence of the term d JS
b in (3.130) is often not important.
We see from relations (3.130), (3.131) and (3.132) that the ordinary Fick
diffusion law strictly holds for a turbulized mixture in the following cases: (a) the
thermal diffusion is negligible; (b) the mixture is binary; (c) the mass force per unit
mass is the same for each component (Fa=ma ¼ Fb=mb); and (d) either the pressure
gradients are zero or the molecular weights of both substances are identical (if ma
mb ¼ m, then M ¼ m). These conditions are rather stringent and it is often
difficult to justify them when modeling real turbulent transport processes. Never-
theless, since the generalized Stefan–Maxwell equations for multicomponent diffu-
sion are complex and since the turbulent coefficients DS
a b have been studied
inadequately, for simplicity, the generalized Fick diffusion law (3.130) (without
the second term on the right-hand side) can be used in many analytical applications.
For the integral mass balance condition
P
a mbJS
a ¼ 0 to be retained, all Wilkey
coefficients must be assumed to be equal, Dturb
b Dturb
.
3.2.5 Formulas to Determine the Correlations Including
Density Fluctuations
Let us now consider the derivation of a defining relation for the turbulent specific
volume flux Jturb
ð1=rÞ that so far remains unknown. In contrast to a single-fluid
turbulized continuum, where the compressibility effects are often negligible, the
total mass density rðr; tÞ in a multicomponent chemically active turbulent medium
generally changes significantly from point to point, for example, due to the forma-
tion of new components and local heat release in chemical reactions. As we have
already seen, when the compressibility of the mass density is taken into account (in
the turbulence model), one more unknown correlation function Jturb
ð1=rÞ rð1=rÞ00
u00
¼ u00 ¼ Àr0u00=r, the turbulent specific volume flux, enters into the heat influx
equation for mean motion (3.54) and the turbulent energy balance equation (3.69).
Correlation moments of this type (e.g., r0 Z00
a =r, r0 T00=r, etc.) also appear in other
transfer equations for the second moments of the local turbulent field characteristics
that are invoked below when developing complicated models of multicomponent
turbulence in the second approximation [see Chap. 4].
It should be noted that in the case of so-called developed turbulent flows, where
the turbulence energy production and dissipation rates are approximately equal,
these additional balance equations for the second correlation moments transform
from differential ones into a system of algebraic relations between the sought-for
second-order correlation moments (like rA00
B00
and rA00
B00
) and the gradients of the

averaged velocities, temperatures, and concentrations for the individual mixture
components. It is these relations that can also be used to establish various forms of
complicated algebraic dependences for the turbulent exchange coefficients on the
gradients of the averaged defining parameters for a medium (in particular, for
stratified flows).
To determine the correlations A00
¼ Àr0A00
=r , it is generally necessary to
invoke special differential equations for them. These, in turn, contain a number
of new correlation terms that are poorly amenable to modeling. Nevertheless, this
approach was investigated in the literature (see, e.g., Methods of Turbulent
Flow Calculation 1984; Kolesnichenko and Marov 1999). At the same time, a
simpler way of determining the correlationsr0A00
=ris possible, which allows them
to be related algebraically to the turbulent diffusion and heat fluxes. It is based on
the fact that the relative density fluctuations caused by the pressure fluctuations
are often negligible compared to their variations caused by the temperature and
concentration fluctuations of the individual components in a multicomponent
medium.
To derive such algebraic relations, let us first find the expression for the density
fluctuations r0
in a gas mixture. For this purpose, we rewrite the equations of state
for a multicomponent mixture of perfect gases
p ¼ RÃ
rT; RÃ
¼ kBn=r ¼ kB
XN
a¼1
Za (3.133)
as
p=r ¼ hRÃ
i hTi þ ðRÃ
Þ00
hTi þ hRÃ
iT00
þ ðRÃ
Þ00
T00
¼ hRÃ
ihTi þ kBhTi
XN
a¼1
Z00
a þ hRÃ
iT00
þ kB
XN
a¼1
ðZ00
a T00
Þ:
(3.134)
Here, we write the true values of the quantities T and RÃ
as the sum of averaged
and fluctuation values, T ¼ hTi þ T00
and ðRÃ
¼ hRÃ
i þ ðRÃ
Þ00
Þ, and use the easily
derivable formulas
hRÃ
i ¼ kB
XN
a¼1
hZai ¼ kB n=r; ðRÃ
Þ00
¼ kB
XN
a¼1
Z00
a : (3.135)
Let us now Favre average (3.134). As a result, we have
hRÃ
ihTi ¼ p=r À kB
XN
a¼1
hZ00
a T00
i: (3.136)

If we eliminate hRÃ
ihTi from (3.136) rewritten as
p ¼ p r=r þ hRÃ
irT00
þ kBrhTi
XN
a¼1
Z00
a þ kBr
XN
a¼1
ðZ00
a T00
Þ À kBr
XN
a¼1
hZ00
a T00
i
¼ p þ pr0
=r þ hRÃ
irT00
þ kBrhTi
XN
a¼1
Z00
a þ kBr
XN
a¼1
ðZ00
a T00
Þ00
;
using this expression, then we obtain the sought-for exact relation
r0
r
¼
1
p
p0
À hRÃ
irT00
À kBhTir
XN
a¼1
Z00
a À kBr
XN
a¼1
ðZ00
a T00
Þ00
!
(3.137)
relating the turbulent mass density fluctuations r0
to the pressure, temperature, and
concentration fluctuations of the individual gas mixture components.
When the turbulence models in the second approximation are developed, it is
convenient to express the temperature fluctuations T00
in (3.137) in terms of the
enthalpy fluctuations H00
. This is because the transfer equations of the second
correlation moments for a multicomponent mixture containing temperature
fluctuations are much more complex than the transfer equations of the moments
containing total mixture enthalpy fluctuations (see Chap. 4). In particular, the
balance equation for the mean square of the enthalpy fluctuations hH002
i, in contrast
to that for the variance hT002
i, does not contain a large number of pair correlations
hZ00
a T00
i whose presence in the equation for hT002
i is ultimately determined by the
presence of a chemical source of thermal energy in the instantaneous equation for
the temperature [cf. (2.24) and (2.29)]. To eliminate the fluctuations T00
from
(3.137), we use relation (3.45) written as
hcpiT00
¼ H00
À
X
a
hhaiZ00
a : (3.45*)
Finally, to calculate the sought-for correlation A00
Àr0A00
=r, we multiply
(3.137) by A00
and average the derived expression over an ensemble of possible
realizations. As a result, after discarding the third-order correlation moments,
we obtain
A00
¼ À
p0A00
p
þ
hRÃ
i
phcpi
rH00A00
À
XN
a¼1
hhairZ00
a A00
!
þ
kBhTi
p
XN
a¼1
rZ00
a A00
: (3.138)
Relation (3.138) is used below both in constructing relatively simple models of
multicomponent turbulence based on the gradient closure schemes and in
constructing more complex turbulence models based on the differential transfer
equations for various second correlation moments of fluctuating thermohy-
drodynamic quantities [see Chap. 4].

Let us now successively identify the parameter A in the general formula (3.138)
with the hydrodynamic flow velocity, the specific number density Za , and the
mixture enthalpy H; as a result, we obtain a number of key relations for semiem-
pirical modeling of multicomponent turbulence (when using the weighted Favre
averaging):
Jturb
ð1=rÞ À
r0u00
r
¼ À
p0u00
p
þ
1
rhcpihTi
Jturb
q þ
1
n
XN
a¼1
JT
a ; (3.139)
H00 À
r0H00
r
¼ À
p0H00
p
þ
1
rh cpihTi
rH002 À
XN
a¼1
hhairZ00
a H00
!
þ
1
n
XN
a¼1
rZ00
a H00;
(3.140)
Z00
b À
r0Z00
b
r
¼ À
p0Z00
b
p
þ
1
rhcpihTi
rH00Z00
b À
XN
a¼1
hhairZ00
a Z00
b
!
þ
1
n
XN
a¼1
rZ00
a Z00
b;
(3.141)
It should be kept in mind that in the case of flows with small Mach numbers Ma,
the relative turbulent pressure fluctuations in (3.137) may be neglected compared to
the relative density and/or temperature fluctuations. This also implies that the first
terms in (3.139), (3.140) and (3.141) may be discarded. This principle is known as
Morkovin’s hypothesis, whose validity for flows without chemical reactions was
confirmed in the literature up to Ma ¼ 5 (Morkovin 1961). In particular, for forced
convection in the terrestrial atmosphere, which is pronounced only in jet flows
where the wind velocity gradients reach fairly high values, the ratios of the
fluctuations r0
; p0
, and T00
to the corresponding means r; p, and Th i have the
following order of magnitude (see van Mieghem 1973):
r0
j j=r % T00
j j= Th i % 10À4
; r0
j j=r % 10À5
: (3.142)
At the same time, the scale of turbulent pressure fluctuations for free convection
in the terrestrial atmosphere is generally much larger than that for forced convec-
tion. Therefore, in general, the terms with the pressure fluctuations p0
in (3.137)
cannot be discarded.
3.2.6 Rheological Relations for the Turbulent Diffusion and Heat
Fluxes in the Case of Strongly Developed Turbulence
In Sect. 3.2.4, we derived the defining relations (3.127) and (3.128) for the turbulent
diffusion, Jturb
a , and heat, qturb
, fluxes in the case where the molecular and turbulent
mass and energy transport processes are equivalent in their contribution to the flow

dynamics and energetics. Let us now obtain the analogous relations for strongly
developed turbulence when turbulent mixing is much more efficient than molecular
one, i.e., whenR ) P,qturb
) q, etc. In this case, the turbulization temperatureTturb,
which is known to be a measure of the intensity of turbulent mixing, is much higher
than the averaged mixture temperaturehTi, a characteristic of the mean energy in the
chaotic thermal motion of atoms. Given the inequality Tturb ) hTi, (3.102) for the
production rate of the system’s total entropy sS can be represented as
0 sS ~Jturb
q
@
@r
1
hTi

þ
1
rh cpihTi2
@p
@r
( )
þ Jturb
hbi Á
@
@r
1
Tturb

þ
XN
a¼1
Jturb
a : À
@
@r
hmai
hTi

þ hhai
@
@r
1
hTi

þ
1
hTin
@p
@r
'
þ
1
Tturb
R
0
: D
0
þ
Xr
s¼1
xs
hAsi
hTi
þ
1
hTi
rhebi:
(3.143)
In writing (3.143), we discarded the small terms referring to the averaged
molecular motion and used (3.139) for the turbulent flux Jturb
ð1=rÞ without the small
term that allows for the pressure fluctuations (the approximation of free convection).
Thus, in the case of strongly developed turbulence, the entropy production sS is
determined by the set of thermodynamic fluxes ~Jturb
q , Jturb
hbi , Jturb
a , R
0
, x and the
corresponding conjugate thermodynamic forces
Yturb
q À
1
hTi2
@hTi
@r
À
1
rh cpi
@p
@r

; ð1
Þ Yhbi À
1
T2
turb
@Tturb
@r
; ð2
Þ (3.144)
YÃÃ
a À
@
@r
hmai
hTi

þ hhai
@
@r
1
hTi

þ
1
nhTi
@p
@r
¼ À
p
nahTi
@
@r
na
n

(3.145)
YR
1
Tturb
D
0
; ð1
Þ YAs

hAsi
hTi
¼ À
XN
b¼1
hmbi
hTi
nb s; ðs ¼ 1; 2; . . . ; rÞ; ð2
Þ
(3.146)
which correspond to the scalar, vector, and tensor sources of nonequilibrium in the
system. Using these thermodynamic fluxes and forces, the production of the total
entropy SS ¼ hSi þ Sturb in a turbulized mixture can be written in the following
bilinear form:
0 sS ~Jturb
q Á Yturb
q þ Jturb
hbi Á Yhbi þ
XN
a¼1
Jturb
a Á YÃÃ
a þ R
0
: YR þ
Xr
s¼1
xsYAs
þ
1
hTi
rhebi:
(3.147)

In particular, (3.147) allows us to obtain defining relations corrected for the
case of strongly developed turbulence for the vector thermodynamic fluxes that
differ insignificantly from relations (3.127) and (3.128). For an isotropic medium,
the corresponding relations when the small cross effects are neglected have
the form
~Jturb
q ¼ À
Lturb
qq
hTi2
@hTi
@r
À
1
rh cpi
@p
@r

; (3.148)
Jturb
a ¼ ÀLturb
aa
pr
nanhTi
@
@r
na
r

þ
na
r
@ ln M
@r
'
: (3.149)
If we now define the turbulent thermal conductivity and diffusion coefficients by
the formulas
lturb

1
hTi2
Lturb
qq ; ð1
Þ Da
p
hTinan
Lturb
aa ð2
Þ (3.150)
then the closing relations (3.148) and (3.149) can be written in the following final
form:
qturb
¼ p0u00 À lturb @hTi
@r
À
1
rhcpi
@p
@r

þ
XN
a¼1
hhaiJturb
a ; (3.151)
Jturb
a ffi ÀrDa
@
@r
na
r

: (3.152)
Expression (3.151) generalizes the analogous expression for the heat flux, which
is commonly used in meteorology for atmospheric turbulence with a passive
admixture and in a number of other important cases, for example, when modeling
free turbulent convection (see Monin and Yaglom 1992), to multicomponent
mixtures.
Note that, as a rule, the so-called potential temperature
y hTi
p0
p
gÀ1
g
¼ hTi
p0
p
RÃ
cp
: (3.153)
(here, p0 is some standard pressure), which is related to the entropy S of a gas
mixture by the relation S ¼ cp ln y þ const, is used in meteorology instead of the
“ordinary” temperature hTi. Hence it is clear that, for example, during adiabatic
vertical motions of small elements of a turbulized medium, the potential tempera-
tureydoes not change (i.e., it is a strictly conservative flow characteristic), while the

ordinary temperature hTi changes with altitude. It is easy to see that the following
approximate equality holds:
@y
@r
ffi
@hTi
@r
À
1
rhcpi
@p
@r

;
using which (3.151) takes a habitual form for meteorological studies.
To conclude this section, let us summarize the basic (reference) differential
equations and defining relations that characterize the relatively simple gradient
model of multicomponent reacting turbulence. It corresponds to the case where the
second-order correlation moments hA00
B00
i for the fluctuating thermohydrodynamic
parameters of a mixture A and B (different from the hydrodynamic flow velocity u)
in all of the applied formulas are small compared to the first-order termshAihBiand
can be discarded. The gradient model of multicomponent turbulence is defined
primarily by the hydrodynamic equations for mean motion (3.21), (3.23), (3.28),
(3.58), the averaged equation of state for pressure (3.71), relation (3.139) defining
the turbulent specific volume flux for a mixture Jturb
ð1=rÞ, and the rheological relations
(3.127), (3.128), and (3.117) for the turbulent diffusion, JS
a , and heat, qS
, fluxes and
the Reynolds stress tensor R. Data on the kinetics of reactions in a turbulized flow
and at its boundaries are also needed to completely describe the flow and the heat
exchange between chemically reacting gases. In the simple turbulence model under
consideration, the quantities xs can be assumed to be calculated only from the
averaged temperature and composition of the mixture, xs ¼ xsðhTi; naÞ [cf. (4.1.4)].
In addition, the system of differential equations and final relations should be
supplemented by the set of chemical components with allowance made for their
gasdynamic, thermophysical, and chemical properties; the universal laws of kinet-
ics and thermodynamics including the equations of state and the expressions for
various thermodynamic functions retaining its ordinary form in the approximation
under consideration; the formulas for the molecular and turbulent transport
coefficients; and the initial and boundary conditions. Although the continuum
model of reacting multicomponent turbulence constructed in this way is a
“simplified” one, it can, nevertheless, be used to formulate and plausibly solve
various astrophysical and geophysical problems.
3.3 Modeling the Turbulent Transport Coefficients
3.3.1 The Turbulence Scale
The gradient model or the so-called algebraic model of turbulence in the first
approximation constructed in the preceding section works efficiently for simple
quasi-stationary flows with well-chosen turbulent exchange coefficients. However,

as has been noted above, this model is completely unsuitable for flows in which the
local average conditions of motion change abruptly or when there is a significant
flow region affected by the turbulent flow as a whole. This is because when such
models are derived, it is commonly assumed that some “local equilibrium” state is
established in the structure of developed turbulence in which the turbulence
characteristics at each point of the flow are entirely determined only by the local
characteristics of the averaged flow field (and the field of volume forces) near the
same point, i.e., by the local values of the dissipation rate and the turbulence scale
as well as the local averaged state parameters of the medium itself. If, alternatively,
the convective and diffusive terms are significant in the balance equation for
turbulent energy (i.e., the flow parameters at a point depend on the characteristics
of the turbulent flow as a whole), then these local formulas are, in general, invalid
(see Ievlev 1975).
Thus, the specific application of gradient models to the calculation of turbulent
flows requires a preliminary development of methods for determining the turbulent
exchange coefficients. The formulas for calculating the turbulent diffusion, Dturb
,
turbulent thermal conductivity, lturb
, and turbulent viscosity, nturb
, coefficients can
be obtained by various methods differing from one another by the degree of
complexity. One of the possible methods for a semiempirical determination of
these coefficients based on the differential transfer equations for the second corre-
lation moments (simplified to algebraic relations) is analyzed below in Chap. 4. In
this section, we briefly discuss the traditional approach to modeling turbulence
based on the concept of mixing length and analyze some simple semiempirical
methods for modeling the turbulent exchange coefficients.
3.3.2 Gradient Hypothesis
The concept of mixing length introduced by Prandtl (1925) in the theory of
turbulence allows one not only to simply express the turbulent exchange
coefficients in the defining relations for turbulent fluxes in terms of the mixing
length but also to derive these relations themselves for some special cases. We now
demonstrate this for the turbulent transport of a conservative (passive) admixture as
an example. But first we once again draw the reader’s attention to the fact that the
defining relations for turbulent fluxes derived thermodynamically in Sect. 3.2 are
not related to the key assumption of this approach, namely, the assumption that the
flow characteristics transported by turbulent fluctuations are conservative, and
hence, have a wider domain of applicability.
Thus, we assume that the transport of a particular scalar flow characteristic by
turbulent fluctuations is a diffusion process and that the existence of some mixing
length scale can be admitted—the distance traversed by an elementary volume of
gas in a turbulized flow before it irreversibly mixes with the environment as a result
of velocity fluctuations. Denote the Lagrangian turbulent fluctuation of a field
3.3 Modeling the Turbulent Transport Coefficients 239

quantity Aðr; tÞ corresponding to the Eulerian fluctuation A00
by A00
L and the
effective mixing length of the attribute A traversed by the turbulent vortices in the
flow before they are destroyed through their interaction with other perturbations by
xðAÞðr; tÞ. We then have
A00
L ¼ A00
þ xðAÞ Á ð@=@rÞhAi: (3.154)
3.3.2.1 Turbulent Diffusion Flux
Assuming now the component composition of a turbulized mixture to be conserva-
tive (i.e., assuming that, being displaced by the distance xðaÞ, the vortices retain the
same specific number density ðZa na=rÞ of component a in the Lagrangian
volume that they possessed at the initial level), we obtain
ðZaÞ00
L ¼ 0; Z00
a ¼ ÀxðaÞ Á ð@=@rÞhZai: (3.155)
Hence follows a simple rheological relation for the turbulent flux Jturb
a ða ¼ 1;
2; . . . ; NÞ [cf. (3.130)]:
Jturb
a ðr,tÞ rhu00
Z00
a i ¼ Àrhu00
xðaÞi:
@
@r
na
r

¼ Àr Dturb
a Á
@
@r
na
r

; (3.156)
where the formula Dturb
a ðr; tÞ hu00
xðaÞi defines the nonsymmetric tensor of turbu-
lent diffusion coefficients that, in the general anisotropic case, allows for the
differences in the intensities of the turbulent velocity and composition fluctuations
along different coordinate axes. Obviously, the length xðaÞ here is, in a sense,
analogous to the mean free path in the kinetic theory of gases. Thus, the turbulent
diffusion coefficientDturb
a is expressed very simply in terms of the mixing lengthxðaÞ
(which is a random variable). Relation (3.156) is equivalent to the assertion that the
turbulent flux of a substance of type a is proportional to the gradient of the mean
concentration hZai na=r and has the opposite direction with respect to it. The
coefficients Dturb
a ðr; tÞ are functions of the coordinates and time and usually exceed
their laminar analogs by many times.
Generally, for different mixture components, it is, of course, necessary to
introduce different effective mixing lengths xðaÞ for the transport of a substance.
In addition, the conservatism (i.e., Lagrangian invariance) hypothesis for the
concentrations Za of the individual components in a chemically active flow is not
justified enough, because the averaged Arrhenius chemical kinetics affects signifi-
cantly the turbulent mass transport processes at finite chemical reaction rates (see

Ievlev 1975). Nevertheless, since this question has not been studied adequately in
the literature, below, for simplicity, we assume that xðaÞ x.
Note once again that, in contrast to the molecular diffusion coefficients, the
turbulent diffusion coefficients Dturb
describe not the physical properties of a gas
mixture but the properties of a specific turbulent fluid motion and, therefore,
directly depend both on the turbulent field intensity and on the method (scale) of
averaging the fluctuating characteristics of this motion. For this reason, the way of
introducing the averaged turbulence parameters is the fundamental basis on which
both the approaches to developing the methods of experimental measurements for
the turbulent exchange coefficients themselves and the results of comparing the
theoretical and experimental data depend. Of course, this remark also pertains to all
of the other turbulent exchange coefficients considered below.
3.3.2.2 Turbulent Heat Flux
Let us now derive the rheological relation for the turbulent heat flux in a multicom-
ponent mixture defined by the formulaqturb
ðr; tÞ r H00u00 [see (3.44)] in exactly the
same way. Since the mixture entropy S rather than enthalpy H is a conservative flow
characteristic, to obtain the needed relation we use (3.83) and (3.85) rewritten as
qturb
¼ ~Jturb
q þ p0u00 þ
XN
a¼1
hhaiJturb
a ; (3.157)
~Jturb
q ¼ hTiJturb
hSi À hTi
XN
a¼1
hSaiJturb
a : (3.158)
Using definition (3.10) for the turbulent flux Jturb
ðAÞ rA00
u00 of an attribute A, we
rewrite (3.158) as
~Jturb
q ¼ hTiJturb
hSi À hTi
XN
a¼1
hSai Jturb
a ¼ hTiru00 S00 À
XN
a¼1
hSaiZ00
a
!
and substitute here the expressions
S00
L ¼ S00
þ x:ð@=@rÞhSi ffi 0 and Z00
aL ¼ Z00
a þ xð@=@rÞhZai ffi 0; (3.159)
which relate the Lagrangian and Eulerian turbulent fluctuations for the entropy S
and concentration Za (the characteristics of a multicomponent medium that are
assumed to be conservative). As a result, we obtain

~Jturb
q ¼ hTiru00 S00
L À
XN
a¼1
hSaiZ00
La
!
À hTiru00x:
@hSi
@r
À
XN
a¼1
hSai
@hZai
@r
!
ffi ÀrhTiDturb
:
@hSi
@r
À
XN
a¼1
hSai
@hZai
@r
!
¼ Àlturb
:
@hTi
@r
À
1
rhcpi
@p
@r

:
(3.160)
In writing the last equality in (3.160), we used the transformation
hTidhSi À hTi
XN
a¼1
hSaidhZai hTi
XN
a¼1
hZaidhSai ¼
XN
a¼1
hZai dhhai À d pa=nað Þ
¼ hcpidhTi À dp=r;
(3.161)
which is a corollary of the Gibbs–Duhem identity, and the relation
lturb
ðr; tÞ ¼ rhcpiDturb
; (3.162)
which defines the tensor of turbulent thermal conductivity coefficients. Thus, for
the turbulent heat flux we ultimately have [cf. (3.151)]
qturb
ðr; tÞ ¼ p0u00 À lturb
Á
@hTi
@r
À
1
rhcpi
@p
@r

þ
XN
a¼1
hhaiJturb
a : (3.163)
We make several more remarks regarding the derivation of (3.163). First, it was
obtained by assuming the parameters S and Za to be conservative characteristics of
the medium, i.e., by assuming that the turbulent motion of a Lagrangian vortex
particle in the mixture from the levelrðxj; tÞwhere it separated from the overall flow
to the level rðxj; tÞ þ xj occurs not only isentropically but also with an invariable
spatial distribution of chemical gas components. However, as has been noted
above, these flow characteristics are generally not the Lagrangian invariants
ðS00
L6¼ 0; Z00
aL 6¼ 0Þ of the turbulent field, because the motion of vortices can be
accompanied by various thermal effects (e.g., local heat release through chemical
reactions or small-scale turbulent heating through viscous dissipation) and by
changes in chemical composition, which leads to the inverse effect of heat transfer
on the development of turbulence [see Chap. 7]. Second, when deriving (3.163),
we assumed that the so-called turbulent Lewis number is equal to unity, Leturb

wturb
=Dturb
¼ 1, wherewturb
is the turbulent thermal diffusivity related to the ordinary
thermal conductivity by the relation 9b
wturb
lturb
=rhcpi: (3.162*)

This ordinary assumption in the theory of turbulence (see, e.g., Lapin and
Strelets 1989) is equivalent to the mixing lengths for the mixture substance and
entropy being equal, x ¼ xðSÞ [see (3.159)]. However, of course, it is generally
necessary to distinguish these scales, because the turbulent vortices can be involved
in heat transfer more actively than in mass transfer (and vice versa).
For a stratified atmosphere, the rheological relation (3.163) for the vertical
component of the turbulent heat flux can be written by taking into account the
hydrostatic equation (3.34) as
qturb
z ¼ p0u00
z þ
XN
a¼1
hhaiJturb
az À lturb @hTi
@z
þ
g
hcpi

: (3.164)
Recall that the quantity ga g=cp in meteorological literature is called the
dry adiabatic lapse rate (ga ¼ 0:98
/100 m in the terrestrial troposphere) and the
expression y hTi þ gaz roughly coincides with the potential temperature used in
meteorology [see (3.153)].
Thus, according to relation (3.164), there exist two mechanisms of thermal energy
transfer through a turbulized multicomponent gas: under the action of an averaged
temperature gradient (the potential temperature y in a stratified medium) and by the
turbulent diffusion fluxesJturb
a when each particle of substanceatransfers, on average,
hhaiof thermal energy. It is important to recall once again that the first term in (3.164)
does not act as the energy flux—the quantity p0u00 drops out of the averaged heat
influx equation (3.58) when (3.164) for the heat flux is substituted into it.
3.3.2.3 Reynolds Tensor
The turbulent stresses, along with the molecular ones, are actually the result of
momentum transfer, but through the turbulent velocity fluctuations. In the simplest
case of a plane shear (along the z axis) flow, the horizontal component of the
Reynolds stress (3.185) takes the form
Rxz Àr u00
x u00
z ¼ rnturb
@huxi=@z; (3.165)
where the formula nT
¼ hxz u00
z i introduces the vertical turbulent viscosity coeffi-
cient defining the ratio of the apparent internal stress to the corresponding averaged
strain rate. In writing (3.165), we assumed that u00
x ¼ Àxz@huxi=@z, i.e., the vortices
being displaced vertically by the distance xz retain the momentum at the level z þ xz
that they had at the initial level z (Prandtl’s hypothesis).
However, the Reynolds tensor Rij is generally related to the strain rate tensor by
the more complex linear relation (3.117), which, in projections onto the Cartesian
coordinate axes, takes the form,

Rxx Àrðu00
x Þ2
¼ À
2
3
rhbi þ 2 rnturb @huxi
@x
À
1
3
divhui

;
Ryy Àrðu00
y Þ2
¼ À
2
3
rhbi þ 2 rnturb @huyi
@y
À
1
3
divhui

;
Rzz Àrðu00
z Þ2
¼ À
2
3
rhbi þ 2 rnturb @huzi
@z
À
1
3
divhui

;
Rxy ¼ Ryx Àru00
x u00
y ¼ Àru00
y u00
x ¼ rnturb @huxi
@y
þ
@huyi
@x

;
Ryz ¼ Rzy Àr u00
z u00
y ¼ Àr u00
y u00
z ¼ rnturb @huzi
@y
þ
@huyi
@z

;
Rzx ¼ Rxz Àru00
z u00
x ¼ Àru00
x u00
z ¼ rnturb @huzi
@x
þ
@huxi
@z

: (3.166)
3.3.3 First Approximation Modeling for the Turbulent
Transport Coefficients
Obviously, the rheological relations for turbulent fluxes derived above using the
concept of mixing length do not solve the turbulence modeling problem either:
formulas of the type Dturb
h u00
xi do not allow the turbulent transport coefficients
to be determined experimentally, because the local mixing length x is a quantity too
uncertain to be measured. Thus, the problem of closing the averaged hydrodynamic
equations for a mixture is again reduced to the problem of finding approximating
algebraic formulas for the turbulent exchange coefficients. This approach is called a
semiempirical first-order (approximation) turbulence theory.
Let us initially analyze the simplest method of modeling the turbulent viscosity
coefficient nturb
. But first we note that the assumption commonly used to find the
turbulent thermal conductivity, lturb
, and turbulent diffusion, Dtyrb
, coefficients is
that the turbulent Prandtl and Schmidt numbers defined by the formulas
Prtyrb
rhcpinturb
=lturb
; Sctyrb
nturb
=Dturb
(3.167)
are approximately constant in a fluctuating multicomponent flow. This assumption
stems from the fact that, in contrast to the turbulent transport coefficients them-
selves, these ratios change only slightly both within some turbulized flow and when
passing from flow to flow. It is also commonly assumed that the turbulent thermal
diffusivity wturb
lturb
=rhcpi and the turbulent diffusion coefficient Dturb
coincide,
i.e., that the Lewis number is Leturb
wturb
=Dturb
¼ 1. In this case, the turbulent
Prandtl and Schmidt numbers are equal, Prtyrb
¼ Sctyrb
. According to the currently
available experimental data, Prtyrb
¼ 0:86=0:90 for flows near a solid wall and
Prtyrb
¼ 0:5 for plane jets and in mixing layers.

3.3.3.1 Prandtl Algebraic Model
As an example, let us consider an averaged flow where the gravitational forces
produce a preferential direction in coordinate space. The following principle of
local similarity (which is consistent with the principle of local internal equilibrium
in the structure of a turbulent flow) is commonly used in the theory of turbulent
transport: the turbulent transport coefficients at each point depend only on the
properties of the medium at the same point, on the local turbulence scale, and on
some characteristics of the averaged flow and volume force fields at this point.
Therefore, the expression for the turbulent viscosity coefficient nturb
is then
nturb
¼ nturb
ðr; n; L; fi
; @huxi=@zÞ; (3.168)
where LðrÞ is the external turbulence scale at a given point of the flow, fi
ðrÞ are the
local characteristics of the fields of volume forces (in particular, the forces of inertia
related to the longitudinal fluid acceleration; then, f / @huxi=@x ). When the
influence of the first and second derivatives of the averaged velocity on the local
properties of the fluid flow is taken into account, this dependence can be different.
The scale LðrÞ characterizes the geometric structure of the turbulent field or the
characteristic size (and it is then the integral turbulence scale L) of the large vortices
involved in turbulent transport that carry a major fraction of the flow kinetic energy.
Occasionally, LðrÞ can be treated as the mean mixing length LðzÞ ¼
ffiffiffiffiffi
x2
q
(as was
originally done by Prandtl (1925, 1942)). In this case, it coincides in order of
magnitude with the correlation length of the velocity field. The external turbulence
scale LðrÞ should be determined from additional considerations. Because of this
uncertainty, it remains possible to take into account the integrated properties of the
flow and its prehistory in the purely local formulas for the turbulent exchange
coefficients. In particular, for free shear layers, the parameter L can be assumed to
be equal to a length proportional to the layer thickness over the entire layer.
However, the proportionality coefficient depends on the pattern of the free flow.
For example, for a flow around an infinite flat wall, it has been established that the
mean mixing lengthL is proportional to the distance to the wall:LðzÞ ¼ kz, wherek
is the Karman constant that can be set equal to ~0.4.
Far from a solid wall, turbulence depends weakly on the molecular viscosity of
the mediumnand, hence, the quantityncan be excluded from the set of arguments in
(3.168). For the turbulent viscosity coefficient, we can then establish (at fi
¼ 0)
nturb
¼ nturb
ðr; L; @huxi=@zÞ;
whence follows the famous Prandtl formula, when using the dimensional theory:
nturb
ðzÞ ¼ aL2
@huxi=@zÞj j;

or
nturb
ðzÞ ¼ L2
@huxi=@zÞj j; ðLðzÞ ¼ kzÞ: (3.169)
The constant factor a is determined for each specific type of motion based on
experimental data; in some cases, it is convenient to drop this factor by appropri-
ately redefining the turbulence scale L.
Near a solid wall, where the influence of molecular viscosity n is significant, the
well-known functional dependence
nturb
¼ n’
L2
@huxi=@zÞj j
n

(3.170)
also follows from (3.168) (at f i
¼ 0). This expression can be further improved both
using theoretical (to be more precise, semiempirical) considerations and purely
experimentally (see, e.g., Lapin and Strelets 1989; Monin and Yaglom 1992). If the
forces of inertia in the longitudinal direction are included in the set of arguments,
then relations (3.169) and (3.170) are modified and take the form
nturb
¼ L2 @huxi
@z

@huxi=@x
@huxi=@z

; (3.169*)
nturb
¼ n’
L2
n
@huxi
@z

,
@huxi=@x
@huxi=@z

: (3.170*)
Relation (3.165) by no means always describes correctly the turbulent momen-
tum transport. In particular, there can exist regions in a turbulent flow behind the
grid where the averaged flow velocity is constant and the gradient @huxi=@zÞ ¼ 0,
while the correlator hu00
x u00
z i 6¼ 0 , because turbulence is generated immediately
behind the grid and is subsequently transported downstream by the averaged
flow. However, the mixing length hypothesis (3.169) requires zero nturb
and,
according to the Prandtl model (Prandtl 1942), turbulence is absent. This circum-
stance reveals the main shortcoming of such models: the mixing length hypothesis
suggests local equilibrium of the turbulent field. Fortunately, the displacement of
the points in space at which hu00
x u00
z i 6¼ 0 and @huxi=@zÞ ¼ 0 relative to each other is
often small and, therefore, applying (3.165) does not lead to significant errors in
numerical flow simulations.
Now, using the rheological relations (3.117), (3.156), and (3.164), we give
convenient (for practical purposes) expressions for the Reynolds stresses and the
turbulent diffusion and heat fluxes that describe the vertical momentum, mass, and
thermal energy transfer for a turbulent motion of a multicomponent mixture:

Rxz ¼ rL2 @huxi
@z

@huxi
@z
; ð1
Þ
Jturb
a z ¼ À
1
Prturb
rL2 @huxi
@z

@
@z
na
r

; ð2
Þ
qturb
z ¼ À
1
PrT
rhcpiL2 @huxi
@z

@hTi
@z
þ
g
hcpi

þ
XN
a¼1
hhaiJturb
az : ð3
Þ (3.171)
It is important to keep in mind that in the general case of a temperature-
inhomogeneous medium, some correction to the linear scale L that allows for the
inverse effect of a nonuniform temperature distribution (determining the degree of
flow stability) on the turbulent mixing efficiency should be introduced into these
relations. This correction is necessary, because the hypothesis about Lagrangian
invariance of any transferable substance is invalid for a chemically active gas
mixture stratified in a gravitational field. In a homogeneous stratified medium (for
example, in a well-mixed lower planetary atmosphere), a correction of this kind can
emerge only due to the existing vertical temperature gradients in certain regions of
space; as a result, the buoyancy forces facilitating or preventing the generation of
turbulence energy appear. To take this fact into account, Richardson proposed a
dimensionless criterion—the Richardson gradient number
Ri ¼
g
hTi
@hTi
@z
þ
g
hcpi
0
@huxi
@z
2
: (3.172)
Based on the similarity theory, it is natural to assume that all dimensionless
characteristics of a turbulent flow are certain functions of the dimensionless number
Ri. To take into account the buoyancy forces in relations (3.171), the following
corrections to the scale L can be applied:
• in the case of stable stratification ( Ri0 ) hampering the development of
turbulence: L ¼ LÃ
ð1 À b1RiÞ , 5b110 (typically b1 ffi 7 ) (Monin and
Yaglom 1992);
• in the case of unstable stratification (Ri0) increasing the turbulence energy
through the instability energy: L ¼ LÃ
ð1 À b2RiÞÀ1=4
ðb2 ffi 14Þ (Lumley and
Panofsky 1964); the formula L ¼ LÃ
ð1 À cRiÞ0;25
, where с is an empirical
coefficient (Bradshaw 1969), is also recommended for the same case;
• in the limiting case ( Ri ¼ 0) where the adiabatic height distribution of the
temperature ð@hTi=@z ¼ Àg=hcpi ÀgaÞ does not affect the development of
turbulence: L ¼ LÃ
, the mixing length in the absence of buoyancy forces.
At the same time, the buoyancy forces change the Prandtl–Schmidt number
(Munk and Anderson 1948):
PrturbÃ
¼ Prturb
ð1 þ 3:33RiÞ1:5
=ð1 þ 10RiÞ0:5
: (3.173)

Apart from the Prandtl first-order turbulence model, other mixing length models
based on different concepts of a conserved quantity when passing from one layer of
a turbulent flow to another are also known. These modes include, for example, the
Taylor (1932) turbulence model, which is based on the assumption that the vorticity
rather than the momentum has the property of a conserved transferrable substance
in a turbulent flow, and the Karman model, in which the mixing length L in (3.169)
depends on two derivatives, L ¼ const @huxi=@zð Þ= @2
huxi=@z2
ð Þ . However, the
shortcomings of the Taylor model are the same as those of the Prandtl one—the
impossibility to describe the turbulent transport processes at the points where
@huxi=@zÞ ¼ 0, while the Karman formula is difficult to apply, for example, for
free turbulent flows where the velocity profile has a point of inflection at which
@2
huxi=@z2
¼ 0 and, hence, nturb
becomes infinite, which does not correspond to
reality (see Loitsyanskii 1978).
3.3.4 Differential Kolmogorov–Prandtl Model [b À L Model]
To overcome the limitation of the mixing length hypothesis noted above, it was
necessary to construct turbulence models that would, nevertheless, allow the
absence of a real internal equilibrium between the turbulence field and the field
of averaged flow parameters to be somehow taken into account. In equilibrium
turbulence, where the production of turbulence energy at each point of the flow is
compensated for by its dissipation, there is no need to include the turbulence energy
hbi r u00j j2
=2r in the set of arguments defining the turbulent viscosity coefficient
nturb
, because it is defined by the same set of local parameters. To approximately
take into account the real “nonequilibrium” of the turbulent field, Kolmogorov
(1942) and Prandtl (1945) proposed introducing the parameter hbi into the set of
arguments defining the coefficientnturb
and using also the turbulence energy balance
equation (3.69), along with the hydrodynamic equations for averaged motion, to
solve specific problems.
In the field of gravitational forces, i.e., when
PN
a¼1
Jturb
a Á FÁÃ
a

¼ 0, (3.69) to
determine the quantity hbi takes the form
r
@hbi
@t
þ r hui Á
@hbi
@r

¼ ÀdivJturb
hbi þ R :
@hui
@r

À Jturb
ð1=rÞ Á
@p
@r

À rhebi;
(3.174)
where
Jturb
hbi rð u00j j2
=2 þ p0=rÞu00 À P0
Á u00 (3.175)

is the turbulent energy flux. According to the (second-order) Kolmogorov–Prandtl
closure model, the turbulent viscosity coefficient nturb
and the turbulent energy
dissipation rate heei [see (3.50) and (3.88)] are related to the turbulent flow energy
hbi by the fundamental relations
nturb
¼ cmL
ffiffiffiffiffiffiffi
hbi
p
; (3.176)
hebi ¼ ce
hbi
3
2
L
(3.177)
(where cm and ce are empirical constants). These relations basically follow from the
dimensional theory and are a generalization of the well-known Kolmogorov (1941,
1942) hypothesis that the turbulent energy dissipation rate hebiðr; tÞ at a given point
of a developed turbulent flow is determined only by the local values of the turbulent
energy hbiðr; tÞ and the turbulence scale Lðr; tÞ.
To determine the other indefinite terms in (3.174), we use relations (3.117),
(3.118), and (3.139). The diffusion term
Jturb
hbi rðb þ p0=rÞu00 À P0
Á u00 ¼ À
r nturb
sb
@hbi
@r
(3.178)
describes the total substantial turbulence energy flux related to various transport
mechanisms in space. In particular, the quantity Jtyrb
ðbÞ r bu00 is interpreted as the
kinetic energy flux of the fluctuation (vortex) motion, so that divJtyrb
ðbÞ describes the
mean rate of decrease in the vortex kinetic energy per unit volume due to “turbulent
diffusion.” In turn, the quantity À div P0
Á u00

characterizes the mean rate of
increase in the vertex kinetic energy due to the work done by the viscous stress
tensor fluctuations at the boundary of an elementary volume.
The expression for the rate of turbulence production by a shear flow appearing in
(3.174) can be written as
R :ð@=@rÞhuið Þ ¼ mturb
D
0
: D
0
Àpturbdivhui
¼ mturb @huki
@xj
þ
@huji
@xk
À
2
3
djkdivhui
2
À
2
3
rhbidivhui
Fn À
2
3
rhbidivhui;
(3.179)
where the positive function
Fn ¼ rnturb
D
0
: D
0
¼ rnturb @huki
@xj
þ
@huji
@xk
À
2
3
djkdivhui
2
(3.180)

is the rate at which heat is generated by viscous turbulent friction per unit volume
per unit time and, accordingly, is called the dissipative function. According to
(3.40), the quantity R : ð@=@rÞhuið Þ can be treated as the rate of exchange between
the turbulent energy and the kinetic energy of mean fluid motion (in this case, it
should be kept in mind that this energy exchange is a purely kinematic process
depending significantly on the choice of the averaging operation).
The generation of turbulent energy in the gravity field r G due to a nonuniform
temperature and/or mixture composition distribution present in (3.174) can be
represented as
rG Jturb
ð1=rÞ Á
@ p
@r

¼ À
nturb
Prtyrb
1
hTi
@hTi
@r
À
1
rhcpi
@p
@r

:
@p
@r
þ
nturb
Prtyrb
@lnM
@r
:
@p
@r
(3.181)
using the formulas
Jturb
a ðr; tÞ ¼ À
r nturb
Prturb
@
@r
na
r

; (3.182)
Jturb
q ðr; tÞ qturb
ðr; tÞ À
XN
a¼1
hhaiJturb
a
¼ À
rhcpinturb
Prtyrb
@hTi
@r
À
1
rhcpi
@p
@r

: (3.183)
to determine the turbulent flux Jturb
ð1=rÞ (the term with pressure fluctuations in (3.139)
was discarded).
3.3.4.1 Richardson and Kolmogorov Numbers
As can be seen from (3.181), two additional turbulence generation mechanisms are
possible in turbulized multicomponent mixture flows. Whereas the first mechanism
is thermal in nature, the second turbulence generation mechanism is diffusive in
nature and emerges when there is a nonuniform distribution of individual
components in the region of space under consideration. This eventually stems
from the fact that the spatiotemporal nonuniformity (fluctuations) of the mass
density is attributable to two factors: spatial temperature nonuniformity and con-
centration nonuniformity [see (3.137)]. As is well known, if a local region with a
density lower than that of the environment appears in a liquid, then the so-called
buoyancy force will act on it in the gravity field. Under certain conditions,
the equilibrium stability is lost and this force sets the liquid in convective motion.

The appearance (disappearance) of turbulence of two types, thermal turbulence and
“concentration” turbulence produced by spatial composition nonuniformity, is also
eventually related to the dual nature of the buoyancy force.
It is generally accepted to allow for the influence of thermal stratification of a
medium on the evolution of a turbulent flow using the Richardson number Ri. In the
model being developed, it can be written in the following general form:
Ri ¼ À
@lnhTi
@r À 1
hcpihTi
1
r
@p
@r
n o
: 1
r
@p
@r

D
0
: D
0
ffi
g
hTi
@hTi
@z þ g
hcpi

D
0
: D
0
(3.184)
(the second form of the number Ri was obtained by taking into account the
hydrostatic equation (3.34))
To allow for the influence of a nonuniform chemical composition distribution in
a multicomponent mixture on the generation (disappearance) of turbulence under
the action of buoyancy forces, below we use the Kolmogorov number
Ko ¼
@lnM
@r Á 1
r
@ p
@r
n o
D
0
: D
0
ffi
g @lnM
@z
D
0
: D
0
; M r=n; (3.185)
which was introduced into the turbulence theory by Barenblatt (1978).
Using these definitions, the mixture turbulent energy transfer equation to deter-
mine the parameter hbi can be written in the following general form:
r
@hbi
@t
þ r hui:
@hbi
@r

¼
@
@r
rnturb
sb
:
@hbi
@r

¼ À
2
3
rhbidivhui
þ rnturb
D
0
: D
0
1 À
Ri þ Ko
Prturb
'
À r ce
hbi
3
2
L
:
(3.186)
It can be seen from (3.186) that if Ri þ Ko0 , then the turbulent energy is
generated by both wind shear and buoyancy forces. When Ri þ Ko ! Prturb
, the
corresponding sum of the terms in the turbulent energy balance equation becomes
zero, implying that no turbulent motion is maintained. In those practically important
cases where one of the above two generation mechanisms of buoyancy forces is
inefficient, one can talk about critical Richardson numbersRic (Kolmogorov numbers
Koc). These numbers can be determined from the condition that the turbulent motion
exists only at RiRic (at constant composition) or only at KoKoc (at constant
temperature). When the two generation mechanisms of buoyancy forces are
switched on, they can generally act both in the same and opposite directions. If the
thicknesses of the thermal and diffusive mixing layers differ significantly, then this
different directionality of the action of the turbulent energy source can lead to a
reversal in some (external for the thinner layer) region of the jet flow.
There exists a whole group of semiempirical Kolmogorov–Prandtl turbulence
models in which (3.186) is used to determine the turbulent viscosity coefficient in

free shear flows (plane or axisymmetric ones) by its numerical solution together
with the hydrodynamic equations of mean motion and the differential equation (or
algebraic expression) for the external turbulence scale L (see, e.g., Turbulence:
Principles and Applications 1980).
3.3.5 Equations for the Turbulence Scale: A Model
with Two Transfer Equations
To close (3.186), it is necessary to have some algebraic relations or an additional
differential equation to determine the turbulence scale L. Deriving the differential
equation for the scale L is one of the most complex problems in the semiempirical
theory of developed turbulence. The fact is that the parameter L cannot be defined
only via the one-point moments of the fluctuating characteristics for a moving flow.
Being a measure of the distance between two points r1 and r2 in a turbulized flow at
which the two-point correlation moments hA00
ðr1ÞA00
ðr2Þi still differ noticeably
from zero, the scale L is determined from the complex differential equations for
these moments by their integration over the distance between points r1 and r2 (see,
e.g., Lumley and Panofsky 1964; Lewellen 1977). The differential equations for L
derived in this way, which describe its convection, generation, and dissipation,
contain a large number of proportionality coefficients poorly established from
experiments, i.e., they are generally much less reliable than, for example, the
balance equation for the Reynolds stress tensor where many terms are determined
almost exactly [see Chap. 4].
For this reason, to ensure the efficiency of practical calculations, the turbulence
scaleLis often specified in the form of purely empirically determined functions or is
found using an algebraic formula (and, occasionally, a simplified differential
equation) that takes into account only the flow geometry (the distance to the wall
z, the boundary layer thickness d, the channel shape, etc.) and that does not depend
on the peculiarities of the fluid flow. For example, the empirical formula derived by
Nikuradze when investigating a fluid flow in smooth tubes can be used for L=d.
Replacing the tube radius R by d, we have the following formula to determine L in
the boundary layer:
L=d ¼ 0:14 À 0:08 ð1 À z=RÞ2
À 0:06ð1 À z=RÞ4

; (3.187)
where z is the distance along the normal from the wall. In the case of free
convection in stratified shear layers, the following simple differential equation
(see Laikhtman 1970) can be used to determine the scale L:
L ¼ Àkc
1
4
hbi
nturb
. @
@z
hbi
nturb
!
; (3.188)

which allows Lto be calculated via the mean characteristics of a moving flow. Here,
k ¼ 0; 4 is the Karman constant; с is an empirical constant. In the limiting cases of
free convection and strong stability, this formula transforms into the well-known
asymptotic relations characteristic of these limiting stratification regimes.
At the same time, the above formulas are distinguished by relatively small
universality and, being suitable for one class of flows, should be modified when
passing to the description of another type of flows. In addition, formulas of type
(3.187) can be used only in the case of “local equilibrium” turbulence, whose
characteristics are determined by the local conditions at each point. For “nonequi-
librium” turbulence, where the influence of the flow prehistory on the flow
characteristics at a point is very important, the quantity L should still be determined
using a dynamical equation including all forms of energy transformations in a
turbulent flow and with coefficients independent of the flow geometry.
In the case of invariant modeling of the turbulent transport processes, one seeks
to use a “universal” evolutionary equation for the integral turbulence scale L that to
some extent removes the above shortcomings in its derivation. Lewellen (1977)
provided such an equation for the turbulence macroscale L defined by the formula
L ¼
const
hbi
ððð
w
hu00
k ðrÞu00
k ðr þ lÞi
dw
l2
;
which was obtained by integrating the transfer equation for the two-point velocity
correlations hu00
k ðrÞu00
k ðr þ lÞi over the volume. This equation is
@L
@t
þ hui
@L
@r
¼ 0:3
@
@r
Lhbi
1
2
@L
@r

À 0:35
L
rhbi
R :
@hui
@r
þ 0:6
nL
l2
À
0:375
hbi
1
2
@
@r
ðhbi
1
2LÞ
!2
þ
0:8L
hbi
g
hTi
qturb
z ;
(3.189)
where l ¼ L=ð3 þ 0:125Returb
Þ
1
2 is the so-called Kolmogorov–Taylor microscale.
The term including the scale l to some extent allows for the relationship between
the velocity fluctuations at some distance from the wall and the pressure
fluctuations at the wall as well as the difference between the vortex sizes in the
transverse direction and along the wall. The difficulties of deriving this and similar
equations for L stem from the fact that none of the terms in the original equation for
the correlations hu00
k ðrÞu00
k ðr þ lÞi can be integrated and, hence, they must all be
modeled. On the other hand, the difficult problem of boundary conditions at the free
boundary of the turbulent flow region where the scale L does not approach zero
arises for the differential equation (3.189).
For this reason, for multicomponent turbulence it is often convenient to invoke
some transfer equation for the combination hA002
i
m
Ln
that, when used jointly with
the transfer equations for the moment hA002
i, defines the scale L instead of (3.189)

(see the collective monograph Turbulence: Principles and Applications 1980). One
of such equations is the transfer equation for the dissipation rate of turbulent energy
hebi, which in combination with the empirical relation (3.177) allows the system
of averaged hydrodynamic equations to be completely closed at the level of second-
order moments. The question about the boundary conditions in this case is simplified
considerably, because the quantity hebi tends to zero at the outer boundary.
The evolutionary transfer equation for the dissipation rate of turbulent energy
hebi in the case of a homogeneous incompressible fluid flow was first obtained by
Davydov (1959, 1961). For developed turbulence, this equation is
r
@hebi
@t
þ hui
@hebi
@r

¼ 0:15r
@
@r
:
hbi2
hebi
@hebi
@r
!
þ 1:45
hebi
hbi
R :
@hui
@r
þ 0:48 r G

À 1:92
rhebi2
hbi
;
(3.190)
where the quantities R : ð@hui=@rÞ and rG are defined by (3.179) and (3.181),
respectively. The terms modeled in the simplest way, which remain at large
numbers Returb
, appear on the right-hand side of (3.190): the generation of hebi
under the action of a velocity gradient, the production of turbulent energy dissipa-
tion by buoyancy forces and nongravitational forces, and the molecular destruction
of hebi. This equation considered simultaneously with the empirical relations
L ¼ hbi
3
2=hebi; nturb
¼ 0:09hbi=hebi (3.191)
following from dimensional considerations allows the turbulent energy transfer
equation (3.186) to be completely closed. Thus, the b À e -model proposed by
(Launder 1975) for the calculation of free shear flows is basically a version of the
Kolmogorov–Prandtl model. In the calculations of near-wall flows, when the
calculation should be performed up to the wall (where the local Reynolds numbers
are small), the empirical coefficients in (3.190) obtained at large Reynolds numbers
do not work. Lander and Morse (1979) proposed modifications for this case as well.
To conclude this section, we emphasize once again that the constants in the
b À e model depend both on the flow geometry and on the physical–chemical
nature of the medium being modeled. Another shortcoming of such models is the
assumption about a gradient nature of the turbulent transport processes. Neverthe-
less, these models are widely used for practical purposes, for example, in such
programs as PHOENICS™, FLUENT™, FIRE™, NUMECA™, STAR-CD™, and
KIVA™ developed for numerical simulations of chemically reacting turbulent
flows in engines, turbines, burners, and chemical reactors (see, e.g., Rosten and
Spalding 1987).

Turbulence and self organization

More Related Content

What's hot (15)

Viewers also liked (8)

Similar to Turbulence and self organization (20)

More from Springer (20)

Recently uploaded (20)

Turbulence and self organization