Series Editor
Nicola Bellomo
Politecnico di Torino
Italy
Foundations of Fluid
Mechanics with Applications
Problem Solving Using
Mathematica
ISBN 9781461271987
Mathematica is a registered trademark of Wolfram Research, Inc., 100 Trade Center Drive,
Champaign, IL 618207237, USA.
987654321
Contents
Preface xi
Index 565
Preface
courses of L.D. Landau and E.M. Lifschitz, L.1. Sedov, N.1. Kochin, L.
Prandtl, G. Batchelor, W. Prager, P. Germain, and other authors. New
interesting results have appeared in FM after the publication of these
lecture courses, however, and there is now a need in presenting them in
the educational literature, so that students can be rapidly introduced
into the scope of presentday FM problems and methods.
The development of computer algebra and of a powerful universal
software system Mathematica has led to the fact that the task of the FM
presentation with the use ofthe M athematica system has become topical.
The present authors have undertaken an attempt in this book at solving
this task. A large number of programs are presented that have enabled
us to perform in the process of material presentation both analytical
and numeric computations with the aid of a personal computer. All
M athematica 3.0 programs are stored on the Birkhauser server. The url
address of which is as follows (see Appendix B for further details of our
programs):
http://www.birkhauser.com/book/isbn/O817639950
In addition, we aimed at taking into account the international char
acter of science. Therefore, we considered it necessary to familiarize
Western readers in more detail with the achievements of the Russian
scientists in the field of fluid mechanics.
The present book has been written on the basis of the lecture courses,
which were presented by the authors during the past few years at the
Novosibirsk State Technical University and at the Novosibirsk State Uni
versity. The lectures were intended for graduate and postgraduate stu
dents that have already attended the introductory lecture courses in FM.
At the same time, this book gives all of the necessary FM concepts and
the derivations of all formulas are presented. There are also a large num
ber of problems with the solutions. All of these features enable one to
use this book both for an initial and a deeper study of FM.
Note that, although we present FM as a theoretical discipline, its
development is closely related to the experiment and the practical needs
of industry. The role of experiment in the formation of the basic concepts
and closing relationships of FM is very significant. After these concepts
have been established, however , the deductive methods of mathematics
take on deciding significance.
The book consists of seven chapters. Chapter 1 presents the basic
concepts of continua: the Lagrangian and Eulerian description, tensors
of strains and stresses, equations of continuity, momentum, and energy.
We use throughout the chapter a tensor invariant form, which does not
depend on the choice of coordinate system. Therefore, the presentation
of the basic FM concepts is preceded by a brief introduction in tensor
analysis. The basic definitions of the tensor analysis, which are used
Preface xiii
in the following, are briefly introduced. From the very beginning, the
strains are not assumed to be small; therefore, various tensors of strains
and stresses are introduced that are related to the initial and current
configurations at the Lagrangian description and in a fixed Eulerian
frame.
In Chapter 2, we present the derivation of the differential equations
for continuity, energy, and motion. The concept of local thermodynamic
equilibrium as well as the the thermodynamics laws enable us to indicate
the general form of the closure relationships for the FM governing equa
tions. We formulate the Hamilton Ostrogradsky variational principle,
which enables us, on the one hand, to find the FM motion equations
and, on the other hand, to establish a relation between the integral
conservation laws for energy and momentum and the isotropicity and
homogeneity properties of space and time.
In Chapter 3, the fundamentals of the similarity and dimension the
ory as well as the mathematical methods for studying the weak disconti
nuities (the characteristics) and strong discontinuities in fluid mechanics
are presented.
Chapter 4 is devoted to the fundamentals of the theory of dynamics of
ideal incompressible fluid. For the ideal fluid, we derive the Bernoulli and
Thomson integrals and consider the planar and axisymmetric irrotational
flows. We also study the general propert ies of the vortex flows. We
further consider in detail the methods for the solution of problems on
the ideal fluid flow around planar and axisymmetric bodies.
Chapter 5 deals with the viscous fluid flows. The NavierStokes
equations are derived, and a number of the solutions of these equations
are obtained at small Reynolds numbers. The basics of the Prandtl's
boundary layer theory are presented. Some approaches to the description
of turbulent fluid flows are considered.
Chapter 6 is devoted to the gas dynamics of ideal and viscous com
pressible gases. We present the theory of the Laval nozzle, shock waves,
and Riemann waves. We also give the solution of a problem on shock
wave structure in a viscous gas. The Chaplygin's theory for the transfor
mation of gas dynamics equations to the hodograph plane is presented
in sufficient detail.
Chapter 7 is devoted to a new FM branch, the mechanics of mul
tiphase heterogeneous media. This branch of mechanics has appeared
during the past 20 30 years, and it now enjoys a period of intense devel
opment. One can speak today about the fact that the general approaches
have been developed, which are applicable to the description of an arbi
trary multiphase medium. Various methods of averaging belong to them,
which enable one to go over to an averaged description, as well as the
idea of interpenetrating continua, each of which refers to a corresponding
phase.
xiv Preface
(1.1.1)
d;
(1.1.3)
~i . ~.
e eJ
_

5:i _
Uj 
{O,1 i =I j
. _ . (1.1.4)
, z ],
where 8j is the Kronecker symbol and the dot denotes the scalar product.
Let us consider an arbitrary coordinate transform yi = yi(xl, x 2 , x 3 )
with the Jacobian I~ I =I O. The quantity df' does not depend on
the choice of the coordinate system; therefore, along with (1.1.1) the
equation
df'= dyie~ (1.1.5)
is valid, from which it follows that
~I of' of' ox j ox j _
e =  = j  =  e (1.1.6)
2 oyi ox oyi oyi J.
1.1 Vectors and Tensors 3
~  Ci~.
C e,  C'i~'
ei. (1.1.9)
The quantity T = Tij ... s;,j . .. es , which is invariant under the coordi
nate transform yi = yi(xj), is called a tensor T:
(1.1.10)
In the particular case of two vectors the polyadic products eij are called
the dyads. There exist nine linearly independent dyads el el, el e2, ... ,
e3 e3 in the base of which the secondrank tensor will have the form
The second equation in (1.1.11) follows from the invariance of the tensor
T under the coordinate transformation yi = yi(xj) and the law for the
transformation of the base dyads
4 1 Definitions of Continuum Mechanics
The metric tensor gij plays an important role; its specification deter
mines the coordinate system. Let dr= dXi ~. Then the squared distance
is
ds 2 = dr dr= ei' 0dxidxj = gijdxidx j .
From here, the formulas
(1.1.13)
(1.1.15)
With the aid of gij and gi j , one can lower and raise the indices
(1.1.16)
(1.1.18)
g11 0
II gij 11= g22 (1.1.20)
o g33
By using (1.1.13), we find:
..
g"'J  g"J  5'J
..  .. {O1,
52J  ,
i=/=j

i = j,
(1.1.21)
where rji(xi) are the Christoffel symbols (rji is not a tensor). Substi
tuting (1.1.21) into the foregoing formula, we obtain:
8B
8x

(8b
i = 8x i
k
+ Lirk) 
if ji ek =
't"'7
Vi
bk
ek, (1.1.22)
k 8b k . k
V'i b = 8x i + bJrji (1.1.23)
8e j
~
  r Jki. e k ,. (1.1.24)
ux'
from where the formula follows for the covariant derivative of bk:
(1.1.25)
Interchanging in the second item the indices j ft I, and in the third item
k ft I, we obtain:
(1.1.26)
(1.1.27)
1.1 Vectors and Tensors 7
One can show that the covariant derivative of the metric tensor is equal
to zero: 'l kgij = O. Consider the tensor Tij = 'l i bj, which can be
presented in the form 'libj = gjk'libk . On the other hand , bj = gjkb k ;
therefore, 'li(gjkbk) = gjk'libk, from where it follows that
(1.1.28)
cPr oe,.,
oxjox k ox j '
from where
r sjkeS
 = r skje s
Let us find the expression for r~j in terms of the derivatives of g ij:
(1.1.30)
(1.1.31)
Adding two equations (1.1.31) , subtracting (1.1.30), and taking into ac
count (1.1.29), we obtain:
(1.1.32)
cijk = ei . (e j x ek ),
Cijk = e; . (~ x ek). (1.1.33)
8 1 Definitions of Continuum Mechanics
C )C )
e1 e31
2 e21 e13
1
e21 e22 e32 e1
2
1 e22 e32
e e~ e~ er e~ e33
el . el el . e2 el . e3
e2 . el e2 e2 e2 e3 = det II gij 11= g,
ej . el e3 e2 e3 e3
 x ek
th a t ej  = Cijke i = Cjkie i, from were
h
The formulas (1.1.37) enable us to write the vector product in the form
(1.1.38)
dV eldx1 . (e2 dx 2 x e3 dx 3)
el . (e2 x e3) dx 1dx 2dx 3=
c123dXldx2dx3 = Jgdx dx 2 dx 3.
1 (1.1.39)
1.1 Vectors and Tensors 9
(1.1.40)
T'
dIV = 't"'7
v'
T' = Tjk ~i ~ ~
't"'7
V i e . ej ek = 't"'7
vi
Tik ~
ek;
~
8Ji
(1.1.43)
where ifi = ~/Ieil is a unit vector, which is parallel with the basis vector
~ . For the contravariant components of the tensor, we can obtain in a
similar way the formulas
(1.1.44)
10 1 Dennitions of Continuum Mechanics
(1.1.45)
(1.1.46)
where the summation is not carried out over the repeating indices.
Thus, the tensor f is a linear operator whose action on the vector byields
the vector c = Tkbkei. There exist among the vectors bthe eigenvectors,
which satisfy the condition
(1.1.48)
(1.1.52)
In the case of a symmetric tensor, all eigenvalues are real. Let A and
A* be the complex conjugate values to which the eigenvectors band b*
correspond. Then
1.1 Vectors and Tensors 11
(1.1.53)
(1.1.54)
Let us choose a coordinate system whose basis vectors coincide with the
eigenvectors ei = bi ; therefore, T = Tijb;bj. Multiplying this relation by
bk and bl, we obtain:
(1.1.55)
(1.1.57)
(1.1.59)
(1.1.60)
~V(t+ ~t)
S
v~t
V(t) ~
Figure 1.3: To the derivation of the formula (1.1.66) for the differentia
tion of an integral over the moving volume V(t).
ii and the coordinate axes :fi . One can consider the functions Q1, Q2 , Q3
as the components of the vector 13 , and one can rewrite equation (1.1.60)
is l
in the form
13 iidS= divBdV. (1.1.61 )
The integrands in (1.1.61) are the scalar quantities, which do not de
pend on the choice of the coordinate system. By choosing an arbitrary
coordinate system xi with the basis vectors ei and taking into account
the relationships
123
B ii = b nk, d Xl d X2 d X3= ..;g dx dx dx ,
= bk ek, ii = nji?., 
 k 000
B
(1.1.62)
8b k
= dx dx dx , dV = ..;g dT,
 k .. 1 2 3
div B = V' kb = 8x k + b' r ki , dT
let us write the Ostrogradsky Gauss theorem in an arbitrary coordinate
system:
(1.1.63)
where the expression for the area element dS is obtained below [see
(1.1.76)]. Let us find with the aid of the Gauss Ostrogradsky theorem
the formula for the differentiation of an integral taken over a moving
volume V(t):
dl
d
t V(t)
.
f(x" t) dV
lim
bot +O
1
I....l.t
(rJV(t+bot)
f(xi,t+~t)dV  r
JV(t)
f(Xi,t)dV)
14 1 Definitions of Continuum Mechanics
AV = is iJ iiAtdS.
The last integral in (1.1.64) can be written with regard for this formula
and the Ostrogradsky Gauss theorem (1.1. 63) as
(1.1.65)
dJ
d .
f(x"t)dV= J (8f 8 +\7 i Uvt). )dV (1.1.66)
t V(t) V(t) t
i B . is
df' = ii . rot B dS, (1.1.67)
1.1 Vectors and Tensors 15
where the contour I and the surface B are shown in Fig. 1.4.
(1.1.71)
where Ais is the cofactor of the element gis' On the other hand, the
elements of the inverse matrix are equal to (1.1.17):
gis = 1 A is)
g
f
i
=
1 og
~ =
a
~(lnJ9). (1.1. 73)
ij
2g ux J ux J
where the summation is carried out over all lateral facets i , j, k. Let us
identify two opposite lateral facets with the normals nil and n12. As will
be shown below [see (1.3.3) and (1.3.6)] the area of a lateral prism facet
is determined by formula
(1.1.76)
Since the projections of the normals to the opposite prism facets onto
the xi axis have the opposite signs (see Fig. 1.5), we have with regard
for (1.1.76) that
(1.1. 77)
where
1.1 Vectors and Tensors 17
e x
3
lv'li bidV = L t
(isi
2
bi nidS i l2 + isi
1
binidSil 1 ) = is binidS.
Problem 1.1. Find the metric tensor components gij in the cylin
drical coordinate system. Write the expression for the length element
ds . Determine the components gij and the basis vectors ei.
Xl ecase; x2 = e
sine; x3 =e;
OX1 2 axIl. 2 ox 1
O~l
cos~; O~2 = ~ sm~ ; o~3 = 0;
OX2 2 ox 2 1 2 ox 2
sin~; o~2 = ~ cos~ ; o~3 = 0;
O~l
OX3 ox 3 ox 3
O~l = 0; o~2 = 0; o~3 = 1.
= ox
i ox i 1 2 2 1 2 2
g12 O~l oe = ~ cos~ sm~ +~ cos~ sm~ = O.
It follows from equation (1.1.17) that the matrix gij = (gij )1; therefore,
22 1 gij = 0 at i 1= j.
9 = 2'
r
The basis vectors ei are directed along the tangents to the corresponding
coordinate lines ~i (see Fig. 1.6). The modulus leil is determined by the
formula gii = leil 2 (there is no summation over i) and is equal to
The relation between the ei and the basis vectors of the Cartesian coor
dinate system is determined from equation (1.1.2) and the formulas for
1.1 Vectors and Tensors 19
8Xi.
8f,j
of ax k ~ aXl ~ ax 2 ~ ax 3 ~
ae = ae Ek = ae El + a~l E2 + a~l E3
cos eEl + sin eE2 = cos <PEl + sin <pE2;
of
ae = e sin~ E1 + ~
2~ 1 2~
cos~ E2
The above task can easily be solved with the aid of the software
system Mathematica 3.0 (see our notebook progll.nb) . The syntax
of all Mathematica functions, which we use in our book, is explained in
Appendix A. In what follows, we present the output of the Mathematica
notebook progll.nb.
e(3) = B~3
ax k ax k
gij = a~j a~j ,
where Xi = Xi(~j) determines the relation of arbitrary coordinates ~i to
the Cartesian coordinates Xi. The expressions for the Cartesian coordi
nates X i in terms of the spherical coordinates ~i (see Fig. 1.7) have the
form
ax k ax k ax k a x k
gn a~l a~l = 1, g22 = a~2 a~2 = (e)2,
ax k ax k = (Cl)2 . 2 c2 O L .
g33 a~3 ae <" sm <", gij = at Z r J.
form
gl1 = 1, g22 = r2 , g33 = r2 sin2 0, (1.1.78)
and the length element is equal to
22 1 33 1
9 = r2' 9 = . (1.1.80)
r2 sin2 0
We have made the Mathematica notebook prog12. nb , which solves
the same problem. We present in what follows the results of symbolic
computations by this notebook.
It is easy to see that the results obtained by the above notebook coincide
with the results that we have presented above and have obtained by
hand , i.e., without using a computer.
Problem 1.3. Find the metric tensor gij for the obliqueangled
coordinate system shown in Fig. 1.8. Determine the covariant coordinate
system ~i' i?, gi j .
Solution: Let a Cartesian coordinate system Xl, x 2 be given together
with the basis El , E2, IEll = 1, IE21 = 1. Construct an obliqueangled
coordinate system e, ~2 by a clockwise rotation of the x 2 axis by the
angle ~  VJ (see Fig. 1.8). It follows from the construction that the basis
vectors of the obliqueangled coordinate system will satisfy the condition
Jell = 1, Ie21 = 1. Let M be a point with the radius vector R in the
plane, which can be written as
22 1 Definitions of Continuum Mechanics
x2 e
/M
/
E2 e2
/
/
<p xl
El el e
Figure 1.8: The obliqueangled coordinate system.
we find:
Xl = e + ecos <p, x2 = esin <p,
OXl)2 (OX2)2
911 = g22 =( oe+ = oe 1;
912 =
6 ~2
"
"
tJ. e2 7. M
/ 1 e
0 e1
e1
~1
Taking into account the explicit expressions for gij and gi j , we obtain:
1 1 (_ _ ) 2 1 (_ _)
e = .2 e1  cos<pe2 ; e = .2 e2  cos<pe1 ; (1.1.81)
sm <P sm <P
e1 .e2
 1  (
= .2   cos <P (
e1 . e2 e2 )2) = 0 ,
sm <P
( 1 cos( <p) )
9 = cos(<p) 1
24 1 Definitions of Continuum Mechanics
The Mathematica 3.0 system has presented the expressions for the basis
vectors l and 2 in a very compact form in terms of the trigonometric
functions cot <p and csc <p. By definition, cot <p = cos <p / sin <p and csc <p =
1/ sin <po It is easy to see that the substitution of these definitions into
the expressions for l and 2 yields the familiarformulas (1.1.81).
t>O
: ~2
O~T~~~
1 / //X 2
~ ____ V I /
Xl ____________________ V
Figure 1.10: The relation between the Eulerian and Lagrangian coordi
nates of points of a continuum.
(1.2.2)
 
ei . ej = s;
Uij = {O,1, i=i:j
i = j,
(1.2.3)
(1.2.4)
26 1 Definitions of Continuum Mechanics
(1.2.5)
(1.2.6)
In this case, the coordinate system is frozen in the continuum and it de
forms together with it and is generally curvilinear. Consequently, equa
tion (1.2.6) is local and determines the radius vector of particles located
in a small neighborhood of a given particle whose position is taken as
a coordinate origin ~i = O. While moving from one point to another,
however, one can determine the radius vector of an arbitrary point of a
continuum with respect to a given point with ~i = o.
Thus, there exist two techniques for the description of a continuum.
In the Lagrangian description technique, the Lagrangian coordinates ~i
and the time t are the independent variables. The Lagrangian coordi
nates are generally curvilinear and deform in an accompanying coordi
nate system together with a continuum. In the Lagrangian approach,
a complete description of the motion of a given particle with the coor
dinate ~i is given, which is determined by equation (1.2.1). Let (1.2.1)
be given in a Cartesian basis of an Eulerian coordinate system (1.2.3).
Then the velocity v and the acceleration a of a particle ~i will be equal
to
_
v=ve=
i
at
(axi)_e (1.2.7)
t ~i t,
(1.2.8)
1.2 Strain Tensor 27
where the motion law (1.2.1) is found by integrating the system of equa
tions dtti = vi(xj , t) under the initial conditions xlt=o = ~i. If the
function <pi ... s = <pi ... s (~j , t) is given conversely in the Lagrangian coor
dinates, then, in the Eulerian coordinates, we will have
. ... ,.., i s
<p.... s = <p, .. S(C(xJ, t), t) = <p'" (x J, t), (1.2.9)
. ovi
a'  +vJ
. ovi (1.2.11)
 at oxi'
Let us present a formula determining the substantive derivative in La
grangian coordinates:
O<pi ... S)
( o . .
 = <p'" S(e t) . (1.2.12)
at ~j at '
The stationary (steady) processes are of great importance in mechanics.
For these processes, the motion characteristics in the Eulerian descrip
tion depend only on the coordinates xi and not explicitly on the time
t:
(1.2.13)
Note that the same motion can be stationary in one reference frame and
nonstationary in another reference frame. It is convenient to represent
the flow pattern at a stationary flow with the aid of streamlines, which
are determined in such a way that the tangent to a streamline at each
point is directed along or in a counterdirection of the velocity vector.
28 1 Definitions of Continuum Mechanics
since the time t enters the righthand side of equation (1.2.15) . It is seen
from the comparison of (1.2.14) and (1.2.15) that the streamlines and
the trajectories coincide in the stationary case. If we choose some line l ,
which is no streamline, and if we draw a streamline through each point
of l, then we obtain in the result a stream surface, the tangent to which
is parallel to the velocity vector v. If l is a closed line, then the stream
surface forms a stream pipe.
It follows from the foregoing that the Lagrangian and Eulerian de
scriptions are in the mechanical respect complete and equivalent to each
other. The experience shows, however, that in the problems of the me
chanics of deformable solids (the elasticity and plasticity theory) a pref
erence is given to the Lagrangian description, and in the aerohydrody
namics problems (the flows of fluid and gas) the Eulerian description
technique is usually employed.
e3
df'
0 e2 X2
Xl el
The components of the strain tensor (1.2.18) relate the state in two
configurations; therefore, one can introduce two strain tensors. Let us
determine the strain tensor in the reference frame basis, which is called
in the literature the Green's strain tensor
o 0 . 0 ,
E= Cij e' (!J. (1.2.19)
(1.2.21 )
then one can express Cij in terms of the first derivatives of xi. Consider
the actual configuration. Since ds 2 is a scalar, its magnitude is the
same in the accompanying and Eulerian coordinate systems, gijd~id~j =
gkldxkdxl, from where we have
(1.2.22)
(1.2.23)
from where
o 8x~ 8xb
gij= gkl 8~i 8~j' (1.2.24)
o 1 (A 8x~ 8xb )
Cij =2 9ij  gkl 8~i 8~j . (1.2.25)
Employing the formula for the transformation of the strain tensor com
ponents from the accompanying coordinate system to the Eulerian co
ordinate system
(1.2.26)
1.2 Strain Tensor 31
Let us find the expressions for Eij and tij in terms of the displacement
vector u.
As follows from Fig. 1.11, = i"o + 71, from where r
o ,
Expressing ei = ei  g~ and
(1.2.28)
o 0 ,
Assuming 71 =U i ei= Ui~ and using the definition of a covariant deriva
tive ar.fr =\liU 0 0 k 0
ek, a;;a ,
= \ljU
k'
ek, we find the scalar products:
o k
ei (\ljU
0 0 0
) ek
EJu au
EJ~i . a~j
au EJu
EJ~i . EJ~j
(1.2.29)
(1.2.30)
32 1 Definitions of Continuum Mechanics
o
0 A
(1.2.31)
Let us find the expressions for the strain deformation Cij in the Eule
rian coordinates in terms of the displacement vector il. Since f' = fQ + il,
the variation of an infinitesimal vector df' will be given by the formula
df'  dfQ = dil. Since df' = dxiei, dil = g;
dxi, therefore, we have that
dfQ = dxi (ei  g;). With regard for these relations, we can find the
variation of a squared distance between the points A and B (see Fig.
1.11):
(1.2.33)
It is seen from the comparison of (1.2.30) and (1.2.34) that the Eulerian
strain tensor coincides with the Almansi strain tensor written in the Eu
lerian coordinate system. Expression (1.2.34) can be obtained directly
from (1.2.30) if one chooses in the reference frame such a system of ac
o
companying coordinates gij that it coincides in the actual configuration
with the Eulerian coordinate system gij = gij . In this case, one can
perform in (1.2.30) a substitution 9 i > "Vi, Ui > Ui , uk > Uk, and
then (1.2.30) will coincide with (1.2.34).
1.2 Strain Tensor 33
(1.2.35)
gij
A
= ei . ej = 1:'ei 11:'ej1 cos 'Pij
:.:. A
(1.2.36)
By using the definition of the basis vectors, we can find the relative
variation of their moduli
(1.2.37)
where dS i and dS Oi are the infinitesimal elements of the length along the
coordinate lines ~i and li is the coefficient of a relative lengthening along
~i. Substituting (1.2.35) (1.2.37) into (1.2.18), we obtain:
o 0 0
2Cij = ((1 + li)(1 + lj) cos CPij  cos 'Pij) I ei II ej I (1.2.38)
(1.2.39)
'Pij= ~, and after the deformation between the vectors ei and ~ is equal
e
to CPij = ~  ij . Substituting these relationships into (1.2.38) , we find
that 2Cij = I~ill~jl sineij , from where it follows that
(1.2.40)
where i I j and there is no summation over the repeating indices. For
small deformations with Cij 1, we determine from (1.2.40) that Cij =
()ij/2 . Thus, if we choose an infinitesimal, rectangular parallelepiped in
the reference frame , it will go over after the deformation in an oblique
34 1 Definitions of Continuum Mechanics
dV  dVo ~ J _ i
(1.2.43)
dVo ~ 1  ci'
To prove this fact, let us choose in the reference frame the coordinate
a _
1
(1.2.44)
VI  2Ei'
(1.2.45)
(1.2.46)
(1.2.47)
36 1 Definitions of Continuum Mechanics
As was pointed out in Section 1.1 [see (1.1.21)], the derivative of the
basis vector is expressed in terms of the Christoffel symbols
(1.2.48)
(1.2.50)
(1.2.51 )
Rk orfj orfl
jli = Of;l  of;j + r ijs rksl  r s rk
il sj
0
= , (1.2.52)
(1.2.53)
Rjlik = o. (1.2.54)
Thus, the conditions (1.2.49) and (1.2.54) should be imposed on Eij for
the determination of xi by the given Eij (or gij); i.e., the space determined
by gij should be Euclidean. [The space is called Euclidean if at each
1.2 Strain Jrensor 37
point the curvature tensor (1.2.53) is equal to zero; therefore, one can
introduce for the overall space a Cartesian coordinate system in which
ds 2 = d(x 1)2 + d(x 2 )2 + d(X 3 )2.J It follows from equations (1.2.52) and
(1.2.54) that
(1.2.55)
therefore, there are in (1.2.54) only six independent equations for
At small strains, Cij = ~ (gij  6ij) 1 and the product of the Christoffel
symbols will be a small quantity of the second order with respect to Cij'
Therefore, neglecting their products in (1.2 .52) , we have that
8fij .k _ 8f i lk
(1.2 .56)
a[l 8~j
(1.2.59)
Assuming the velocity field to be continuous, we will have with the ac
curacy up to the terms of the order df'o that
 =  + (8v)
VB 8~i
VA
d ci
A <" (1.2.60)
Since the strains are infinitesimal during the time dt, no difference is
made in (1.2.61) between the bases of the reference frame and the actual
o ,
configuration e(;::;; ei' The Eulerian coordinate system is assumed to be
Cartesian; therefore, gij = 6ij. Let us define the rateofstrain tensor
(1.2.63)
It can be seen from (1.2.62) and (1.2.63) that the tensor iij is symmetric,
iij = iji, and the tensor Wij is antisymmetric, Wij = Wji . Since Wii = 0 ,
the Wij has three different components, for which one can define the
corresponding vector (more exactly, a pseudovector) w = wiei of the
angular velocity in accordance with
1
W = W32/V g,
/0 (1.2.64)
(1.2.65)
. (1.2.66)
Let us prove that the equation (1.2.65) coincides identically with (1.2.61).
Representing the vector product in terms of the Levi Civita tensor, we
obtain with regard for (1.2.66):
(1.2.67)
where
~ijk=~i .(~ X ~k)' ~ ijk =~i. (e j x e k ).
In order to find the ~ijk~ ils, let us define the direct product and the con
volution of two tensors. According to l  4 ,6, the direct product of a tensor
ofrank n A = Ai ... je:; . .. j by a tensor 13 = Bk",sek . .. es ofrank m is de
fined as a tensor of rank n + m of the form 6 = Ci ... jk",sei .. ~ek ... e s .
The components Ci ... jk .. .s are obtained by a simple multiplication of the
components of the tensors (matrices) A and 13
Ci ... jk ... s = Ai ... j B k ... s . (1.2.68)
At a convolution of a tensor 6, for example, over the indices j and k,
it is necessary to make a substitution of the direct product of two basis
vectors ~ and ek by a scalar product, as a result of which the tensor
rank will be reduced by two:
C i ... jk ... s e t
(e ' .
J
ek) e 
s 
g' Ci ... jk ...s e
Jk t ...
e' e  .
J1 k+1 e s ,
Substituting these expansions into the definitions of the Levi Civita ten
sors [the last two equations in (1.2.67)], we obtain:
o
0
eix
0
eiy
0
eiz e xt eoty o
e" z
0
Cijk=
0
ejx
0
ejy
0
ejz
Eijk = o
e xJ
o
e Jy
o
e zJ
0 0 0 Ok ok ok
ekx eky ekz ex ey ez
40 1 Definitions of Continuum Mechanics
While determining the direct product of these tensors, let us make use of
a wellknown theorem of linear algebra 7 that states that the determinant
of the product of two matrices is equal to the product of the determinants
of these matrices:
where the last equality is a consequence of the fact that the determinant
value does not change if the rows are replaced with columns (the trans
position). With regard for this theorem , the components of the tensor
of rank six formed by a direct product of Cijk and cnls will be equal to
) C~ )
exl exS
C"en
eiy eiz
Cijk cnls ejx ejy ejz eny el eSy
ekx eky ekz enz eYz eSz
ei' ei e l ~es 8:'t 8t1 Mt
~ e n  .e
ej I ~. e S 8Jn 81. 8J8 , (1.2 .70)
ek. en  .e
ek I ek. e 8 8kn 81 k 8k8
where we have omitted for brevity of notation the index zero over the
vector components. Calculating the determinant (1.2.70), we obtain the
following compact expression:
Note that formulas (1.2.70) and (1.2.71) for the product of the Levi
Civita tensors are valid in an arbitrary basis.
Substituting (1.2.71) into (1.2.67), we obtain:
_ 1 0 1 0
'0 x d~ = 2(t5Jt5~  t5J8i)V'IVsd~j e k = 2(V'jVk  V'kvj)d~j ek,
from where it follows with regard for (1.2.62) that (1.2.65) coincides
identically with (1.2.61). Thus, the proof of (1.2.65) is completed.
The relationship (1.2.65) expresses the contents of the Cauchy Helm
holtz theorem: the velocity of any point of an infinitesimal particle is
composed of the velocity vA of the translational movement of the particle
center, the rotation velocity 0 x d[ of a particle as an absolutely rigid
o
body with respect to the point A, and the velocity fijd~i e j related to
the particle deformation.
1.2 Strain Tensor 41
uuT = E, (1.2.73)
C=UKV, (1.2.74)
= Wi~. (1.2.75)
In the case of a Cartesian coordinate system, gij = 6ij and the formula
(1.2.75) simplifies to
el e2 e3
w= ox}
0 0
OX2
0
OX3
i
= wei (1.2 .76)
VI V2 V3
42 1 Definitions of Continuum Mechanics
Choosing the line 1 that does not coincide with the vortex line and draw
ing the vortex lines through each point of l, we obtain a vortex surface.
If 1 is closed, a vortex tube emerges.
X3 V
dx
e3 x
X2
0
el e2
Xl
af = 0, a~ = 26, ar = 66(1+t).
aEI OVI
ot +
V OVI _ _ Xl
IOXI  (1 + t)2 + (1 + t)2
Xl 

,
OV2 OV2 2X2 4X2
& + V2 a X2 =  (1 + t)2 + (1 + t)2
aV3 aV3 6X3
&+V3 aX3 = (l+t)2
3) It follows from Fig. 1.12 that the vector tangent to the streamline
(SL) dx is directed along the velocity v;
therefore, x dx = 0, or v
VI V2 V3 = O.
dXI dX2 dX3
Expanding the determinant into the first row, we obtain:
el (V2dx3  V3dx3)  e2 (VI dX3  V3dxI)  e3( VI dX2  V2dxd = 0;
V2dx3  V3dx2 = 0, Vldx3  V3dxI = 0, Vldx2  V2dxI = 0,
44 1 Definitions of Continuum Mechanics
log6 + log~;1/2 = C, 6
C = log172;
~2
therefore,
Xl (X2)1/2
log 6 = log 6 '
We can find in a similar way from the second and third equations that
The above task can easily be solved with the aid of the software sys
tem Mathematica 3.0. We present below the output of the corresponding
program prog14. nb.
6
2(1+ t)6
3(1 + t)26
x 36
{{X3[X]t ~r }}
46 1 Definitions of Continuum Mechanics
It is easy to see that the symbolic computation results by the above Note
book progl4 . nb coincide with the results presented above by us, which
we have obtained by hand. The powerful builtin computer graphics en
able the user of Mathematica 3.0 a rapid visualization of the results. We
show in Fig. 1.13 the trajectories of 27 particles of a continuum, which
we have chosen in our Notebook in a sufficiently arbitrary way.
o
Problem
_
1.5. The displacement field is given in the Cartesian basis
ei = ei = Ei in the form
OUl
O,
OUl OUl = 1' OU2 = O. OU2 = l'
06 06 =1 ; 06 ' 06 ' 06 '
OU2
l', OU3 = 1' OU3 = O. OU3 =0
06 06 ' 06 ' 06 .
1.2 Strain Jrensor 47
Substituting the values of ~ into the Green 's formula (1.2.29) , we ob
tain:
Since the Eulerian coordinate system coincides with the coordinate sys
tem of the reference frame , we determine from the equation Xi = ~i + Ui
that
xl=6+~26 , x2=26+6, x3=66
Using the elimination method, we find:
6 = X2  2Xl  3X3 , 6 = Xl + X3, 6 = X2  2XI  2X3
Ell
Problem 1.6. Find the tensor of small strains Cij, the rotation
tensor Wij, the principal strains Ci , and the principal axes ni of the
strain tensor for the displacement field
OUl OU2
4;    1 ', =7'
OX2  OX2 '
O. OU3  3'
, OXl  ,
1.2 Strain Tensor 49
The strain tensor Eij and the rotation tensor are determined by formulas
(1.2.62) and (1.2.63)
OUl 1 (OUl
Ell  = 3; El2 =   + OU2) 1
 = (1 + 1) = 0;
OXl 2 OX2 OXl 2
0; E22 = 7; E23 = 2; E33 = 4;
The principal values Ci and the vectors iii are determined from a
system of three equations (Ekl  Ebkl)nl = 0 (the summation is carried
out over the index l, but not over the k). The system of homogeneous
linear equations has a nonzero solution provided that det (EklEbkL) = O.
Substituting the values of Eij into this equation, we obtain:
(4  E) o o
o (7  E) 2 =0, (4E)((7E)(4E)4) =0.
o 2 (4  E)
The cubic equation has three roots El = 8, E2 = 4, and E3 = 3. Let us
write the system of equations (Ekl  Eibkl)ni = 0 in the matrix form :
=0, i = 1,2,3.
50 1 Definitions of Continuum Mechanics
Since the determinant of this system is equal to zero, there are two
equations for the determination of three components n~. Therefore, it
is necessary to augment them by the normalization condition (n;)2 +
(n;)2 + (ny)2 = 1. Substituting Cl = 8 into the system of equations, we
obtain two equations 4ni = 0, 2nr  ni = 0, from where ni = 0, ni =
2nr , and from the normalization condition 4(nn2 + (nn2 = I , we have
that 5(nr)2 = 1. We finally obtain for Cl = 8: ni = (0, + ./s, Js).
We can find in a similar way for C2 = 4 that n~ = (1,0,0) and at
C3 = 3, n~ = (0,"*, "75)' One can verify by a direct substitution that
ni . nj = 6ij; therefore, the vectors ni can be taken as a new orthonormal
basis in which the strain tensor will have the diagonal form
It follows from the definition of the strain tensor Ci = b..xi/ Xi and the
form of C~j that along the direction nl the substance has extended by
a factor of eight, along the direction n2
it has extended by a factor of
four, and along the direction n3 by a factor of three.
In what follows, we present the output of the Mathematica Notebook
progl6. nb, which solves this problem.
E = ( 40 07 0)
2
o 2 4
Calculation of the Rotation Tensor Wij
1
W = ( ~
3
o
2 o
3 )
2
{3,4,8}
{{O, I, 2}, {I, 0, O}, {O, 2, I}}
1.2 Strain Tensor 51
BX2 + AX3
BXl  CX3
AXl + CX2
52 1 Definitions of Continuum Mechanics
AXl
x_2 = e[l] + C
BXl
x_3= C[2] +C
b)  CX2  CX1_
a) V = 2
Xl
2 el
X2+ + Xl2 + X 2 e2 ,
2
1
el e2 e3 C_ C
w=2 fJ
fJXI
fJ
fJX2
fJ
fJ X 3
=e3,
2 W3 = 2"' Wl = W2 = O.
CX2 0 0
D'
where ki = const; i.e., the streamlines are the lines parallel with the Xl
axis.
We have in a similar way for the vortex lines (1.2.77) that
from where it follows that the vortex lines are the lines parallel with the
X3 axis (Xl = k 4' X2 = k5, k i = const).
In case b) , where
CX2
Vl =  2 + X 22' V3 = 0,
Xl
el e2 e3
a a a
w=21 ax, aX2 aX3
~ ~
xi+x~ xi + x~
= _ e3 (
2 xi+x~
C _ 2Cxi
(xi+x)2
+ C _ 2Cx~ ) =
xi+x (xi+x)2 '
from where it follows that WI = W2 = W3 = 0, Xi =j:. 0.
Equations for the streamlines (1.2.14) have the form
Thus, the streamlines are the circles (Fig. 1.14). A paradox arises here:
the liquid particles rotate along a circle, however, the angular velocity
w= O. If it were a solid body, then by virtue of v<p = wr the quantity
w =J 0 if v<p =J O. There arises in the liquid an additional rotation of a
liquid particle at the expense of the velocity gradient (see Fig. 1.14).
We have for the given velocity field that
C
v<p = .
r
Thus, the farther from the coordinate origin the particles are located,
the slower their velocities. Let us choose at the moment of time t = 0
a liquid particle ABeD (see Fig. 1.14). By virtue of the dependence
v<p = C /r it will revert at a later moment of time t > 0 into a particle
A' B' C' D', which has undergone a rotation as a solid by the angle e and
the reverse rotation by the same angle because of the velocity gradient.
As a result, the total rotation angle of a liquid particle will be equal
to zero. Note that w = 0 everywhere with the exception of the point
x = 0, where a vortex filament is located, which is perpendicular to the
plane Xl X2. The dependence W3(Xl, X2) in the case of a vortex filament
is determined by the Dirac delta function:
w_1 = 0
w_2 = 0
C
w_3 =  
2
x_2 = k2
x_3 = k3
1.3 Stress Tensor 55
The volume dV ::::; !dedede , and the area dSi ::::; d~jd~k; therefore,
taking d~i > 0, we can see that the righthand side is a quantity of third
order smallness and the lefthand side is the quantity of the secondorder
smallness. Therefore, we can neglect the righthand side and obtain the
equation:
l idSi + f:dS n = 0. (1.3.1)
With regard for the third Newton's law li  p, we can rewrite
56 1 Definitions of Continuum Mechanics
 1
in = dSn j'dSi , (1.3.2)
where dSi is the ith facet area, and dSn is the area of a facet with
the normal ii. We have taken into account the fact that the vectors of
external normal to the areas dSi are directed into an opposite side with
respect to the vectors ~. It follows from Fig. 1.15 that
therefore,
1  1"
dSn . ii = '2(A 1 A2 X A 1 A3) = "2((e2 x e3)d~ d~ + (e3 x el)d~ d~
2 3" 13
, , 1 2 1 '1 2 3
+(el x e2)d~ d~ ) = "2(c231 e d~ d~ + c312e'2 d~ 1 d~ 3 + c123e' 3 1
d~ d~ ).
2
It follows from here that the area of a triangle constructed on the vectors
~d~j and i!kd~k is equal to
(1.3.4)
1.3 Stress Tensor 57
 dS1 ~l
dS 'n=e dS2 ~2 dS3 ~3
n ygrr
+e +e .
ygn H3 (1.3.5)
Expanding ii into the basis ~i, ii = ni~i , we rewrite (1.3.5) in the form
(1.3.6)
(1.3.7)
(1.3.8)
(1.3 .9)
Employing the expansion into the basis in = f~~i' [; j = aij~i ' let us
rewrite (1.3.7) as
(1.3.10)
The obtained relation (1.3.10) is called the Cauchy formula , which en
ables one to find the force acting on an area with a normal ii, if the
Cauchy contravariant stress tensor a ji is known in the basis ~i' The
force acting along a perpendicular to the area with the normal ii is equal
to
(1.3.11)
(1.3.12)
is ln dS = 0, is (r x In)dS = 0. (1.3.13)
"V/Jj = 0, (1.3.14)
'J. _ __
off j
+ CfJr'ji
'..
'r7 .
v JCf  o~j (1.3.15)
Expanding ffj into the basis ffj = (jiji, let us rewrite (1.3.14) with
regard for (1.1.27) in the form
(1.3.16)
where
(1.3 .17)
(1.3.18)
(1.3.19)
't""7
':., ckis  Uj
v j (Cka'ij)c s:ka'ijc
ckis + 'C:., k v j a'iJcc k is
't""7
 cck is a'ik  0
.
Assuming
, 'ik
Ckis a
'"""',
= ~ ckisa
'ik
+ '"""' ,
~ ckis(J
'ik
k<i k>i
'"""', 'ki) 
('ik  (J
~ C kis (J
0,
k<i
dBi r;::;;;;
dB o = V gg". (1.3.21 )
t
According to (1.3.6), the equations dBi = Uni dBn , dB? = n? dB~ are
valid. Substituting them into (1.3.21), we obtain:
(1.3.22)
Let us determine the force vector fin related to an area dB~ in the refernce
frame with the aid of the equation
60 1 Definitions of Continuum Mechanics
Figure 1.16: To the derivation of the formula for the Piola Kirchhoff
tensor.
(1.3.23)
o
Applying formula (1.3.7) to the reference frame fin =(ji n?, we find from
(1.3.23) the contravariant stress vector in the reference frame basis 5 :
(1.3.24)
o 0
Expanding (ji into the basis ei , we can determine the Piola Kirchhoff
o. 0 ji 0
tensor in the reference frame basis (j'=a ej. Using (1.3.24), we de
oij ..
termine the relation between a and at). Let the law of continuum
o
deformation Xi = xi(~j, t) be given at which the vector d f' goes over to
df' (the time t is a parameter). The expression for df' is the same in the
accompanying Lagrangian and Eulerian Cartesian coordinate systems,
i.e., de ~ =
. '" . 0
dx' ei , from where we have that
o ox i 0
(1.3.25)
~= o~j ei
Substituting (1.3.25) into the relation l;i = (yii~ , we find with regard
o 0
for equation and equation (1.3.24) the expression for the
(ji=/} j i ej
Piola Kirchhoff tensor in terms of the Cauchy tensor
It can be seen from (1.3.26) that the Piola Kirchhoff tensor is no sym
metric tensor. The Cauchy formula in the reference frame basis has the
form p~ =/] ijn~.
Let us find the stress tensor in Eulerian coordinates. We will as
sume the Eulerian coordinate system to be for simplicity a Cartesian
system. Let us at first determine a passage from the accompanying La
grangian coordinate system to the Eulerian one for a given motion law
xi = Xi(~j, t). In accordance with the rule for the transformation of the
tensor contravariant components, we obtain:
ij _ ax i ax j Akl
(1.3.27)
(J  a~k a~l (J .
It follows from the relationships (1.3.27) that the stress tensor (Jij is
symmetric. Along with (1.3.27), we present the expression for a ij in
terms of (Jij
(1.3.28)
The stress tensor a ij can be expanded into the spherical and deviator
components
(1.3.29)
SI = 0,
where Sij is called the tensor of the stress deviator and P is the pressure.
In the Cartesian coordinate system, equation (1.3.29) simplifies to
(1.3.30)
1
P  (J
 3 tl,
where Sij is the tensor of the stress deviator and P is the pressure.
(1.3.31 )
62 1 Definitions of Continuum Mechanics
(1.3.32)
Let us determine the Cauchy stress surface from the condition N r2 =
const, and then we obtain the equation for the Cauchy stress surface by
substituting (1.3.32) into (1.3.31):
(1.3.33)
from where it follows that the normal k to the Cauchy stress surface is
parallel with a vector of the force acting on the area with the normal
ii = ';B;, where r = Xiei is the vector r emanating from the coordinate
origin 0 into a given point of the Cauchy stress surface (see Fig. 1.17).
It is known from the analytic geometry that the secondorder surfaces
have at least three principal directions for which the vector k is parallel
with r. (If the tensor surface is spherical, then there are infinitely many
such directions, and the corresponding tensor a ij = P8 ij is called
the spherical tensor.) With regard for (1.3.34), the condition for the
r
parallelism of k and may be written as
(1.3.35)
where a is a proportionality coefficient. It follows from (1.3.35) that
(1.3.36)
1.3 Stress Tensor 63
(1.3.37)
where h ,12 , and h are the first, second, and third invariants of the
stress tensor, respectively:
h = det II a{ I . (1.3.38)
(1.3.39)
(1.3.40)
(1.3.41 )
does not determine the rate of change of the stress tensor. Let us consider
some parallelepiped, which rotates as an absolutely rigid body together
with the forces applied to its boundaries. In this case, the components
of the stress tensor in the Eulerian coordinate system a ij = OikOjlakl
64 1 Definitions of Continuum Mechanics
will change with time: d~:j =I 0, where Oik is the rotation matrix. This
change, however, is not related to the physical processes occurring in a
chosen material volume because, in the accompanying coordinate system
of the parallelepiped, d~:j = O. It is seen from here that the quantity
d~;t is no tensor, since, by virtue of the tensor definition (1.1.11), the
equality Tik = OikOjd'kl should be valid, and if jkl = 0, then Tij = O.
Let us determine the derivative of the tensor in such a way that we
again obtain a tensor as a result of the differentiation of a tensor with
respect to time. This can be done in different ways5,6,1O13, however, we
will follow 5, since it uses substantially the invariance property of a tensor
Cauchy surface. Let the stress tensor (jkl(~k, t) be determined in some
neighborhood of the point ~k = const. According to formula (1.3 .31), we
can find the magnitude of a normal force acting on a unit area with the
normal n, both in the Eulerian Cartesian coordinate system for which
gij = 6ij and in the accompanying curvilinear coordinate system
(1.3.43)
Specifying some infinitesimal vector dr, let us rewrite (1.3.32) in the form
ni = dxijdr. Then, by multiplying (1.3.43) with (dr)2 = 6ij dx i dx j =
gijd~id~j , we find the equation for the Cauchy stress surface
(1.3.44)
(1.3.45)
ax i ax j
aij a~k a~1 = A kl = gkiglj
"Aij 'Aj
= glj ' Ai .
k. = gki .1' (1.3.46)
1.3 Stress Tensor 65
Differentiating (1.3.45) with respect to time and equalling the first and
third terms, we obtain:
(1.3.47)
A scalar quantity is on the lefthand side of (1.3.47), which does not chan
ge under the coordinate transformations. The quantity d~j d~k trans
forms as the contravariant tensor components; therefore, the
(1.3.49)
if.() =daijd
'J!t id J' d(dxi)d J'
  x x +a '   x +a" x
d(dxj)d i (1.3.50)
dt tJ dt tJ dt '
. (da ij s
+ asj ov oV S
q,(t) = ;It ox i + ais ox j (1.3.51 )
)"
dxtdx J ,
Comparing (1.3.49) and (1.3.51) , we now find the derivative of the tensor
in the Eulerian coordinates (aij)' , which corresponds to the derivative
with respect to time in the accompanying coordinates of A~j
I oaij oaij ov s ov s
aij =~
ut
+ Vk~k
uX
+ asj~
ux t
+ ais~
ux J
(1.3.52)
acY tJ k acY tJ av s av s
at + V axk + cYis axj + cYsj axi '
4>" (f>ij)'d~id~j = cY~jdXidxj ,
, ax i ax j
cYij a~k a~l
'
( cYkl )' _ gkiglj
' , (' cY i j )' _'glj (cY k. ' (cY.1i.), .
'j)' _ gki
(1.3.54)
<I>
ax i ax j
Eij a~k ae
we obtain:
(9'ik)' = g'il'J'k(')'
9 glj =  2 E
~i k . (1.3.55)
Using the formulas (1.3.53) (1.3.55) , one can show that the (aij)', (a ij ).,
(aij)', (a/J form in the accompanying coordinate system different ten
sors for the rate of change of the stresses t , to which different tensors of
1.3 Stress Tensor 67
aaij k aaij av j av i .
at + V axk  aik axk  akj axk '
(1.3.57)
(1.3.58)
The above tensors of the stress rates (1.3.53) and (1.3.56) (1.3.58) are
linearly dependent between each other:
and as a linearly independent tensors, one can take (aij)', (aij)tl, aD,
where the tensor (aij)V' = ~((aij)' + (aij)V') is called the Jaumann
derivative and has the components
(aij ) V' = at
aaij k aaij
+ V ax k  aikWjk  ajkWik, (1.3 .59)
where
1 (aVi aV j )
Wij ="2 ax j  axi
Figure 1.18: The octahedron area with the forces given on it.
It follows from here that IJn is equal to the pressure p with the opposite
sign: IJ n = ilJii = po We now find the modulus of the shear force iJt
1.3 Stress Tensor 69
(1.3.60)
_ 1/2
aoct  y'3V 8 1 + S22 + S32 (1.3.61)
Problem 1.11. Find the extremal values of the shear stresses Ti.
0"3
Choosing some area with the normal ii, we can determine the compo
nents Ii, O"n, O"t (see Problem 1.9).
The Lagrange function <I> = 0";  Anini satisfies the following extremum
equations:
n1 1; n2 = 0; n3 = 0; A = O"r; 71 = 0;
n1 O, n2 = 1; n3 = 0; A  _0"2.
2' 72 = 0;
n1 O, n2 = 0; n3 = 1; A = O"~; 73 = 0,
1 1 0"2  0"3
n1 0, n2 = J2' n3 = J2' A= 0"20"3, 71 =
2
1 1 0"1  0"3
= 0, J2 ' A= = (1.3.62)
n1
J2' n2
n3 = 0"10"3, 72
2
1 1 0"1  0"2
n1 J2' n2 = J2' n3 = 0, A = 0"10"2, 73 =
2
It follows from here that the extremal values of 72 lie on the areas, which
halve the right angles (see Fig. 1.20), where we show two such areas.
Problem 1.12. Find the rate of change of the contravariant Aij and
mixed Aij components of the stress tensor A in the accompanying and
Eulerian Cartesian coordinate systems:
(1.3.63)
(1.3.64)
72 1 Definitions of Continuum Mechanics
.. ax n av s .. ax n ax k (av S a~j ) av s
AtJ a~i a~j = AtJ a~i a~j a~j axk = ank axk '
.. ax s av n .. ax s ax k (aV n a~i ) av n
AtJ a~j a~i = AtJ a~j a~i a~i axk = aks axk '
for the components of the rate of change of stresses in the Eulerian coor
dinate system. Expressing the quantity (a ns )6 in (1.3.66) , we obtain:
(1.3.68)
Using (1.3.46) again , let us write the expression for the mixed com
ponents:
ax i ax j .
aij a~l a~k = .9liA'k
ax s a~k i.
a sn = a~i axn A. k (1.3.69)
OX S (O~k ) . Ai.
o~' ox n k
ov s _
_ oc"_A"
k .
O~i oxn k
References
=
consisting of the same particles at any time. Let us choose at the time
t in a reference frame the volume in the form of a parallelepiped
0 0 0
constructed on the vectors de el, de e2, de e3 (see Fig. 2.1). After
the deformation, it will go over to a parallelepiped spanned on the vectors
del, de2, de3, where the coordinates d~i remain constant and
the basis vectors change in time, which leads to the alteration of the
parallelepiped volume dV. Since the coordinates d~i refer at all times
to the same particles (they are frozen in the continuum), the planes
d~ii x d~j~ also consist at any time of the same particles. This means
that the particles do not penetrate the walls of the parallelepiped and
do not leave it during the deformation process. Consequently, the given
parallelepiped is an individual volume in which the number of particles
is preserved: dN = n dV as well as the particles mass: dm = p dV. Since
the mass of a substance comprised within the parallelepiped is conserved,
we obtain:
PodVo = pdV, (2.1.1)
where the subscript zero corresponds to the initial moment of time. For
the volumes of the parallelepiped dVo and dV, the formulas
are valid, which upon substitution into (2.1.1) yield, with regard for
(1.1.33), the formula
(2.1.2)
Since (1.1.36) holds, we have that E~23 = J[ji, 123 = ,;g. Therefore,
the continuity equation in Lagrangian coordinates will have the form
J[ji
p=p0,;g , (2.1.3)
dm
dt =0, m= r
iV(t)
pdV. (2.1.4)
78 2 Fundamental Principles and Laws of Continuum Mechanics
~ __ S(t)
The individual volume surface S(t) consists of the same particles; there
fore , it changes in time, and the points of the volume V(t) boundary are
described by the trajectories equations
dx i i
dt=v , (2.1.5)
r
JV(t)
(c:;
ut
+ 'ViPV i ) dV = o. (2 .1.6)
Since (2.1.6) is valid for any individual volume V(t), we obtain from
(2.1.6) the continuity equation in the Eulerian coordinates:
ap
at + 'ViPv i _
 0, (2.1.7)
where
i _ apv i jri
+ Pv
D.
v.pv  axi ji
Using the Weyl formula (1.1.73), we can rewrite the continuity equation
in the form
ap 1 a i
~ + rn~(y/gpv) = o. (2.1.8)
ut yg ux'
(2.1.9)
2.1 Equations of Continuity, Motion, and Energy 79
where ais the acceleration, F is the body force, and In is a surface force
applied to the surface S bounding the volume V. Equation (2.1.9) points
J
to the fact that the variation of the momentum pa dV of a continuum
located within the volume V is equal to the sum of the body forces
J pF dV and the surface forces IndS. J
Consider at first equation (2 .1.9) in the accompanying Lagrangian
coordinates ~i, with the basis vectors i and the metric tensor gij = i'0.
Using equations (1.1.63) and (1.3.7), let us rewrite the expression for the
surface force as follows:
where
(2.1.12)
upon substitution of which into (2.1.12) , we will have the motion equa
tion
(2.1.13)
where
+ fJ"ki r jt J + fJ"ijr t)'
Eykj
"V .fJ"kj
J
= ~
8~j
k,
where the second derivative a;t=;i is taken for a fixed particle ~j = const
and the basis vectors are related by equations (1.3.25) ~ = $ej. Mul
tiplying equation (2.1.14) by ~k, we obtain with regard for relations
(2.1.15)
pk _ .. 'kl ax j Fi (2.1.16)
 9'J9 a~l '
.. 'kl
9'J9
ax
a~l
j
(a x _ Fi)
2 i
at2
_

~ V' . 'kj
p JCl . (2.1.17)
( aat2
2
Xi _ Fi) ax
a~s
i = ~V'('
p J 9sk Cl
'k j )
.
(2.1.18)
Equations (2.1.17) and (2.1.18) enable one to determine the motion law
Xi = Xi (~j , t) at a given stress tensor (jkj (~i).
Let us find the motion equations in the Lagrangian coordinates in
o 0
the reference frame basis ;. Let the reference frame basis ei coincide
with the Eulerian basis ei. Then we have the relationships
_ 0
F = Fk ek. (2.1.19)
o
..;ga j =a j =3 kj
0
Employing equation (1.3.24) ek as well as the relations
o
(2.1.19) , let us rewrite the motion equation (2.1.12) in the basis ; of the
reference frame
(2.1.20)
2.1 Equations of Continuity, Motion , and Energy 81
o
In particular, in the Cartesian basis of the reference frame gij= 6ij, the
motion equation will have the form
[Px k a /J kj k
Po at 2 = 7if,J + F . (2.1.21)
avk . k) 1 a k k
P( 
at
+ vH\Jv
]
= (y'ga ])
y'g ax]
+ pF . (2.1.23)
Using the Weyl formula (1.1.73) , we can write the motion equations in
the Eulerian coordinates in a more conventional form
p(aVk
at
+ Vhy ].vk ) = Y' .a kj
]
+ pFk , (2.1.24)
where
aa kj
Y'a kj
]
= __
ax]
. + aklrjl] + aljrkl] .
Besides the conservation law for the momentum (2.1.9), the inde
pendent conservation law for the momentum moment takes place. In
the case of a formless medium, when there is no internal moment of
the momentum and the surface and body forces , the conservation law
for the momentum moment , written for some arbitrary volume V of a
continuum, has the form
Equation (2.1.25) implies that the rate of change of the momentum mo
ment of the volume V is equal to the sum of the moments of the body
force and the surface force acting on this volume V.
82 2 Fundamental Principles and Laws of Continuum Mechanics
M = r
JV(t)
pdV = r
JM
dm = const,
~ r
dt JV(t)
<pp dV = ~
dt J M
r <p dm = r dipdt dm = r
JM JV(t)
dip p dV.
dt
(2.1.26)
from which it follows that (2.1.9) represents the second Newton's law,
written for the individual volume. Applying (2.1.26) to (2.1.25), we
obtain with regard for the relations ~: = v, Tn
= aij ~nj, the equation:
Transforming the last integral to the integral over the volume, we find:
where we have used the definition ~ = V'i;'. Using this expression, let
us rewrite (2.1.27) as
Since the basis vectors l are linearly independent, we have for each
component that Ejilaij = 0, from where the condition for the symmetry
of the Cauchy stress tensor (1.3.20) follows:
(2.1.29)
2.1 Equations of Continuity, Motion, and Energy 83
dv
4
(aV k . k)
+vJ\lv ek
dt at J ,
where ei are the basis vectors of the Eulerian coordinate system Xi.
Similar transformations lead thereafter to the condition for the symmetry
of the Cauchy tensor in the Eulerian coordinates
(2.1.30)
In recent years, the models of micropolar media are widely used in the
investigations of the media with microstructurel 4 , in which the material
point (as a matter of fact, an infinitesimal volume is considered, which
contains a large number of particles) possesses, along with the mean
velocity V, density p, and temperature, the internal momentum moment
~ surface moments mn , and body moments M of the pairs of forces. In
this case, the conservation law for the moment has for the volume V the
form
(2.1.32)
Then one can rewrite (2.1.31) with regard for (2.1.26), (2.1.28), and the
expression
in the form
dl _
= pM + \lm J + (e
A. A A
Pdt J J x e)fJ
t
tJ
.
(2.1.33)
84 2 Fundamental Principles and Laws of Continuum Mechanics
It is easy to see that, in this case, the Cauchy tensor (jij is not symmetric.
Performing similar calculations in the Eulerian coordinate system, we
obtain that the Cauchy tensor (Jij is also not symmetric.
2.1.3 The Energy Conservation Law: The First and
Second Laws of Thermodynamics
As was noted above, the state of an infinitesimal volume dV is char
acterized by the parameters p(r, t), iJ(r, t), and T(r, t). The tempera
ture T(r, t) determines the velocity of the thermal (chaotic) motion of
molecules (particles) located in the volume dV. The temperature T is
defined strictly for the thermodynamically equilibrium systems in which
all thermodynamic processes proceed at an infinitely slow rate. In this
case, at each moment of time t, a thermodynamically equilibrium state
takes place. The thermodynamically equilibrium state is such a state
in which all thermodynamic characteristics remain constant as long as
desired under the invariable external conditions. An actually moving
medium is a nonequilibrium system; therefore, the introduction of the
temperature T as a parameter and the use of the thermodynamic de
scription for finding the dependence T = T(r', t) is some approximation
of the physical processes. If the characteristic linear size of an individual
volume dl satisfies the inequalities). dl L, then such approximation
will be sufficiently good and one can then assume that a thermodynam
ically equilibrium state is realized inside the volume dV, where)' is the
molecularfree path or the distance between the molecules and L is the
characteristic size of the variation of the mean parameters of the medium.
At the motion of an individual volume dV , the equilibrium state within
it will change during the characteristic time !J.t "" L/v, where v is the
velocity of the medium motion. The assumption on the realization of
a state that is close to the equilibrium state will be justified if !J.t is
much larger than the characteristic time for achieving the equilibrium
state (relaxation) T s "" 5../c, where c is the velocity of the thermal mo
tion of molecules. It is easy to see that, if the velocity of the medium
motion v is less or of the order of the velocity of the thermal motion of
molecules c, then, from the condition 5.. L, the inequality Ts !J.t
follows. If this inequality is violated, then it is necessary to use for the
description the kinetic equation. This approach is not discussed, and the
interested reader is referred to the monographs 5 ,6. It will be assumed in
the following that the inequalities 5.. L or Ts !J.t are satisfied.
Along with the temperature T, the state of a thermodynamic system
is characterized by the entropy S and by some set of the parameters
Jti. The independent parameters Jts form the basis of a thermodynamic
space of a system, and all remaining thermodynamic parameters can be
expressed in terms of these parameters: Jtl = f(Jts). It is clear that
some other set of independent parameters Jt~ can be used as the basis
2.1 Equations of Continuity, Motion , and Energy 85
parameters; only their overall number will be the same. The choice of
the independent parameters is determined first of all by the problem
formulation.
The most important characteristic of a thermodynamic system is
the specific internal energy E, which is a function of the independent
parameters S, fJi, and the total energy of an individual volume dV with
the mass dm = p dV is found by formula
(2.1.35)
The variation of the total energy dEM is composed of the kinetic energy
2
variation dKM = d Iv PV2 dV and the internal energy variation dE M =
dIvEpdV:
(2.1.36)
Following7, let us find the kinetic energy variation dKM during the time
dt. Consider the accompanying coordinate system ~i , ~i in which the
motion equation has the form (2.1.13)
dv ' .. ' 
p dt = \l/j"J~ + pF. (2.1.37)
86 2 Fundamental Principles and Laws of Continuum Mechanics
Let us multiply the left and righthand sides of equation (2.1.37) by the
displacement vector dr = v dt = Vkk dt and integrate over the volume
Vet) (see Fig. 2.3). As a result, we obtain the equation:
1 Vet)
dv dV dt =
pv d
t
1 Vet)
\7 ift) VidV dt
A. .
+ 1
Vet)
pF . dr dV. (2.1.38)
Using (2.1.26), let us present the lefthand side of (2.1.38) in the form
r
iV(t)
pVddVdV dt =
t
(r iM t
2
dd v2 dm) dt = dd
t
(r
iV(t)
/22 dV) dt = dK M ,
(2.1.39)
where v 2 = 9ijVi Vj = Vivi is a scalar quantity. Let us transform the
righthand side of (2.1.38) :
r
iV(t)
~jaijVidVdt= r
iV(t)
~j(aijih)dVdt r
iV(t)
aij~jVidVdt.
(2.1.40)
Employing the expansion
r
iV(t)
~j(aijvd dV dt = r
i Set)
aijnjVi dB dt = r an' drdB
i Set)
= dA)~/
(2.1.42)
Substituting (2.1.41) and (2.1.42) into (2.1.40), we find:
dE M = r
iv(t)
a ij iij dV dt + dQ~) + dQ~) . (2.1.45)
2.1 Equations of Continuity, Motion, and Energy 87
Introducing the specific velocity (per unit mass) of the body heat supply
tiv and the surface velocity of the heat supply tin, we will have for the
volume V(t) that
where the minus sign is related to the fact that the positive heat fluxes
tiv > 0, qn > 0 lead to the energy reduction in the volume V(t). Going
over to the integration variable dm = pdV, we can write:
dEM=d r
Jv~)
EpdV=d r Edm= JMr dEdm,
~
r
JV(t)
r ~(jij tij dm dt,
(jij t ij dV dt =
JM p
dQt;:;) =  r qV p dV dt =  r tivdm dt ,
iv(t) J M
dE 1 'J ~ 1 n ~,
dt = pa A
'J 
.
qv  p v ,q . (2.1.47)
If we introduce the notation for the total heat flux to the unit mass per
the unit of time
dqe . 1 ~i
dt = qv  pViq , (2.l.48)
A
(2.l.49)
where we have used the first relationship in (1.3.54) of the form dEij =
tijdt. Equation (2.l.49) [or (2.l.47)] is called the energy equation, or the
88 2 Fundamental Principles and Laws of Continuum Mechanics
dE 1 i. dqe

dt
= (JJ C ""
p tJ
+dt'
(2.1.50)
(2.1.51 )
where
.
Cij = 2"1(~v iVj + ~) _ i _
v jVi , V = V ei
i
= Vie .
Note that in the Eulerian coordinates the Eijdt are not differentials of
the components of the strain tensor Cij. It follows from (2.1.49) and
(2.1.51) that the variation of the specific internal energy occurs at the
expense of the work of the internal stresses iiT
ij dEij and the heat supply
(removal) dqe. The increment of the heat quantity dqe is proportional
to the temperature increment
where C is the specific heat. It follows from the energy equation (2.1.49)
and equation (2.1.52) that one can choose the T and Eij as independent
thermodynamic variables. Then E = E(T, Eij) . The specific heat C is a
scalar quantity, and therefore it can only depend on the temperature T
and the invariants of the strain tensor Ji . The experience shows that the
specific heat depends on temperature and the first invariant J 1 = Ekk;
therefore,
(2.1.53)
The stress (jij depends in the simplest case on the strain tensor Eij and
the temperature T
(2.1.54)
Substituting (2.1.53) and (2.1.54) into equation (2.1.49) , we can find
the internal energy increment under the variation of the thermodynamic
Parameters from T(l) E(l) to T(2) t(2).
, tJ ' tJ
2.1 Equations of Continuity, Motion, and Energy 89
(2.1.56)
and it depends not only on the initial strain value ti~l) and the final
strain value ti~2) , but also on those values of the temperature T, which
were taken during the overall thermodynamic process. Exactly the same
heat quantity
r
T(2)
depends on T (1), T (2) and on the values of tkk, i.e. , on the integration
path, the equation of which is <p( tij, T) = 0, where <p is some given
function.
Equation (2.1.49) is valid for the formless media. In the case of the
availability of the internal structure characterized by the independent
parameters fJs, it is necessary to add to the righthand side of (2.1.49)
the work dA' on the increments of the parameters fJs
(2.1.58)
Aij _ aE aE
a p~,
C = aT' (2.1.59)
UCij
8M = r
iV(t)
8pdV = r 8 dm .
iM
(2.1.63)
For the determination of dq', the Gibbs formula is postulated, which has
the following form for twoparametric gases:
are invariant under the substitution of t for t) ; therefore, one can say
on the basis of a given molecule state what happened to a particular
molecule at any previous moment of time. The answer to this question
is that there are in nature unstable systems, the behavior of which differs
fundamentally from the motion of a molecule, as discussed above. The
infinitesimal disturbances, which always exist in nature, lead unstable
dynamic systems to considerable uncontrollable deviations of their tra
jectories; therefore, the behavior of such systems obeys the probabilistic
laws. It is easy to see that the reversibility property is absent for such
systems. If we try at some moment of time t to return our system to its
initial state by inverting the velocities of all molecules to the opposite
ones (v + v), then an infinitesimal disturbance will always be found,
which will lead to a significant deviation of the trajectory from the orig
inal one, and at the time 2t, the system will be in a state, which differs
considerably from the initial one.
Let us present an example of a dynamically unstable system consist
ing of gas and absolutely elastic balls. Let us assume that the free path of
the balls between their collisions I is much larger than the ball diameter
d. Consider the collision of two balls with the target distance b. Then
it follows from the momentum conservation law that the angle of the
deviation of the balls after the collision is 'P = 7r  2(), where sin () = bj d.
An infinitesimal disturbance ob of the target distance leads to an in
finitesimal disturbance of the deviation angle of the balls o() ~ objd.
Until the next collision, the ball passes the distance I; therefore, the
disturbance of the target distance before the next collision is equal to
ObI = o()l ~ (ljd)ob. After N + 1 collisions, the target distance distur
bance is
(2.1.66)
It follows from formula (2.1.66) that at I d the disturbances will
quickly grow, and after several collisions the state of the system will differ
significantly from the state, which was not subject to an infinitesimal
disturbance.
The abovediscussed questions on the relation between the second
law of thermodynamics and the behavior of thermodynamically unstable
systems are discussed in detail in the literatures ,9 .
(2.1.67)
(2.1.68)
The next important principle of the locality, which is valid for the form
less media, states that the functional Ft depends only on those ~i, which
lie in an infinitesimal neighborhood of the ~i point. This is expressed
practically by the fact that the functional Ft depends only on the func
tions Xi(~j, t  T), T(~j, t  T) and the derivatives g~; (e, t  T) , i.e., on
the strains = Eiji!i~' This conclusion follows in an obvious way from
the fact that the stress state of an individual volume will change only
as a result of its deformation and the variation of its thermodynamic
parameters (the temperature T). The dependence of the coordinates xi
should be absent because of the medium homogeneity. In addition, the
medium should be isotmpic, and its properties should not change under
the rotation of the coordinate system. This means that the functional
Ft may depend, besides the strain tensor, only on the identity tensor
J. With regard for the foregoing, the equation of state of a homoge
neous isotropic formless continuum can be written down in the following
general form:
(2.1.69)
(j(~i,t) = rt ~(((tT),~i),T((tT),~i))dT.
ito
(2.1. 70)
(2.1.71)
where
(2.1. 72)
and
94 2 Fundamental Principles and Laws of Continuum Mechanics
Denoting the principal values of the strain tensor iij by the letters
AI, A2 , A3 , let us write the equation for the Ai (1.1.50) , (1.1.51) in the
form
(2.1. 73)
where
a= LAi ' b= LAiAj, C=A1A2A3.
Since equation (2.1.73) is invariant under the coordinate transforma
tions, it can be rewritten in the arbitrary coordinates as
(2.1.74)
Multiplying this equation by i n  3, we obtain the Hamilton Cayley for
mula:
(2.1. 75)
with regard for which, equation (2.1. 71) may be written as follows:
(2.1.76)
where
f3i = f3i(T(T  t), AI, A2, A3), i = i(T).
For a large class of continuous processes, the strain i is an analytic
function of time, therefore, the following expansion is valid:
where the points denote the terms of the series, which contain the
higher derivatives with respect to time "t , .... Substituting (2.1.78) into
(2.1.70), we find the general form of the equation of state under the
assumption that it is an analytic function:
...2 ..
+ W1i + W2 i 2 + W3 i + w4ii + W5 i + W6 i ,
A
Wi t(1+m(Tt)+n(Tt)2+k(rt)3
lto
+ l(r  t)4)f3i(A1' A2, A3, T(r  t, ~i)) dT. (2.1.80)
2.1 Equations of Continuity, Motion, and Energy 95
(2.1.81)
The irreversible processes are absent in the ideal gas in the case of contin
uous motions; therefore, dq' = O. Substituting this relation into (2.1.62)
and the relation (2.1.81) into (2.1.51), let us rewrite with regard for the
equations
(2.1.83)
(2.1.84)
96 2 Fundamental Principles and Laws of Continuum Mechanics
from where it follows that the specific internal energy is a function of the
density and the entropy E = E(p , S); therefore, the ideal gas is called
the twoparametric medium. Equation (2.1.84) coincides with (2.1.64)
and is called the Gibbs formula in the literature. Since dE is a total
differential, then
(2.1.86)
(2.1.87)
/1 = E  TS + P / p,
2.1 Equations of Continuity, Motion, and Energy 97
from where we obtain with regard for J.l = J.l(P, T) and (2.1.84) that
(2.1.88)
dE = (OE) dP + (OE) dp
OP p op P
oS 1 oE oS
op =
1 (OE
T op  p2 .
P) (2.1.90)
oP ToP'
From the condition of equality of the second derivatives
d8 dE
T
+ TP d ( p1) = C v dT
T
+ Rpd ( p1)
(
d Cv In T
1) = d (Cv In T (l)
+ R In p p (R/Cv))
1)h1l) ,
d (CvlnT (p (2.1.93)
8 = 80 + Cv In To
T (PO)I'1
P , (2.1.94)
where 8 0 , To, Po are constants. Expressing from here the T and substi
tuting into (2.1.92), we find the internal energy E as a function of the
eigenvariables p and 8
dE = P!!...(~)
dt p ,
d8 = 0, (2.1.96)
from where it follows that the entropy remains constant at the adiabatic,
reversible processes: 8 = const. Using (2.1.85) and (2.1.95), let us find
the equation relating the pressure P to the density p at the adiabatic
processes:
P=Apl', (2.1.97)
which is called the Poisson's adiabat, where A = A(8) is a constant.
Summarizing the above, let us write a complete system of equations
of the ideal gas in the Eulerian coordinates gij = ei . S:
dp
+p
dt
~
d'IVV= , (2.1.98)
diJ 1 ~
="VP+F
dt p ,
2.1 Equations of Continuity, Motion, and Energy 99
that
dq' 1 i . P .k
dt = UJci
p J
+ ck'
P
(2. l. 104)
*
of (2.1.106) will have the second order of smallness"" EijEij, and one
can assume with good accuracy these processes to be reversible (in the
absence of heat conduction) ~ 0, ~~ ~ O.
Let us determine the entropy variation %
due to the heat supply
1Jt. We neglect for a while the body heat supply qV = 0 and rewrite
formula (2.1.48) in the Eulerian coordinates
Substituting (2.1.108) into (2.1.107) and then into the second law of
thermodynamics, we can find the entropy variation ~ due to the heat
conduction
 = 1 d'Ivq~ =  
p dqe

T dt T
1 (~"T)
q. v
T2
 d'IV (if)

T
(2.1.111)
dS (dST dS,,)
p dt P (it + dt
~(((diVV)2 + 2/leijeij ) + ;2 (V'T)2  div (~). (2.1.112)
Using (2.1.112), one can find the entropy variation in some fixed volume
V
It follows from here that the first integral over the volume yields the
entropy production in the volume V at the expense of the irreversible
and nonequilibrium processes related to the energy dissipation and the
temperature equalization at the expense of the heat conduction. Since
there should be ~~ > 0, the heat conduction coefficient should also be
positive: Ii: > O.
Using formulas (2.1.51), (2.1.102), (2.1.105), and (2.1.109) let us
write a complete system of equations for a viscous, heat conducting gas
in the Eulerian coordinates:
dp d' ~ 0 d 8 
dt +p IV V = , dt = 8t + v J V' j ,
dv 1
=V'P+ ((
+ .~
V) V'dlvv+v~v
~ A
(2.1.114)
dt p p 3 '
dE = _P~ (~) + f(divv)2 + 2vei e ij + ~~T,
dt dt p p J P
T= ( 8E)
8S p' P=p 2 (8E)
8p s' E=E(p,S).
102 2 Fundamental Principles and Laws of Continuum Mechanics
(2.1.115)
(2.1.116)
where
. k 1 .
Jk = q , Xk =  T2 \lk T , Jkl = T kl , (2.1.117)
and Tkl = gkigljTij , and the Tij are determined by formula (2.1.100). As
follows from the heat conduction law (2.1.108) and the expression for
the stress tensor (2.1.100) , the fluxes ja are proportional to the thermo
dynamic forces Xc>' On the other hand, this result is a consequence of
the fact that u is a homogeneous quadratic function of Xc> and can be
generalized for a wide class of thermodynamic systems, which are near
their equilibrium state. Consider a thermodynamic system whose state
is characterized by the independent parameters JLi. Let these parame
ters have in the equilibrium state the values JL?' Let us introduce the
notation Xi = JLi  JL?, so that the Xi will determine the deviation of the
system from the equilibrium state Xi = O. In the equilibrium state, the
entropy 50 reaches its maximum value; therefore, for the states different
from the equilibrium one, but close to it, we will have that
o 1
5 5 = f3'kX'Xk
2' , , (2.1.118)
where the matrix f3ik is symmetric, i.e., f3ik = f3k i (the substitution of
the indices i ...... k should not change the scalar quantity 5). The system
deviation from the equilibrium state causes in the system the fluxes
Xi = !i(Xk), which tend to return the system to the equilibrium state
Xk = 0, Xk = O. If the deviations are small, one can expand Ii (Xk) into
a Taylor series in the neighborhood of the equilibrium point Xk = 0 and
retaining only the linear terms in the expansion, we obtain:
(2.1.119)
2.1 Equations of Continuity, Motion, and Energy 103
(2.1.121 )
where the symbols {3ii.,1 denote the matrix, which is inverse with respect
to {3ik. Substituting the second equation from (2.1.121) into equation
(2.1.119), we find:
(2.1.122)
where 'Yik = Aij{3jk1 . There exists in the theory of nonequilibrium pro
cesses the Onsager's principle 11 , according to which
Differentiating (2.1.118) with respect to time, we can find the rate of the
entropy production
(2.1.125)
(2.1.126)
(2.1.127)
" "
X3
" " e3 x2
I e2
leI
Xl
I y
o "
,,1'
rp
X
"" 'JI
Figure 2.4: The cylindrical coordinate system.
from where we obtain, with regard for the orthogonality of the coordinate
system gii = 1/gii, the expressions for the components of the covariant
and contravariant metric tensors
where we have used, for the determinant 9 = det " gij II, the expression
9 = g11g22g33 Substituting into the continuity equation the explicit
form of the metric coefficients gii (the summation over i is absent) and
going over to the notations r = Xl, 'P = x 2 , Z = x 3 , u r = U I, Ucp =
u 2 , U z = u 3 , let us rewrite the continuity equation in the form
(2.1.132)
Taking into account the expressions for the Christoffel symbols and going
over to the physical velocities u i and the accelerations a i by the formulas
106 2 Fundamental Principles and Laws of Continuum Mechanics
Let us determine the physical components C'V j~ij) of the vector \l jaij ~
by the formula (\lj~ij)Ei = \ljaij~, where the unit vectors Ei =
~/I~I = ei/y?iii. For the ideal gas aij = _Pgij , therefore, with re
gard for Ei = E i , gii = II gii , we have that
Going over to the notations r , <p, and z, let us rewrite this formula in a
componentwise form
(2.1.136)
(2.1.137)
2.1 Equations of Continuity, Motion, and Energy 107
where
. dv i
pi = Fi y!gii, a' = Yty!gii.
Substituting (2.1.134) and (2.1.136) into (2.1.137), we obtain the motion
equations in the cylindrical coordinates:
oU z oU z u<p oU z oU z _ _ ~ oP F
ot + Ur or + r orp + Uz oz  p oz + z
Z
Xl
"
X3
e3
r
X2
0
Y
X " I
" ~
(2.1.142)
let us write the squared distance between two infinitesimally close points
from where the formulas follow for the metric tensor components of the
orthogonal spherical coordinate system
g33 = ( X 1 smx
2)2
, (2.1.143)
10
g contravariant = ( 0 ~
o 0
L22 ~ 1 = r
L33~1 = rsin 2 (8)
L21 ' 2 = ~
r
f_33'2= cos(8)sin(8)
f_13'3= ~
r
L23'3 = cot(8)
Thus, we have obtained the following expressions for the nonzero Chris
toffel symbols:
110 2 Fundamental Principles and Laws of Continuum Mechanics
. ap . 1 ap . 1 ap
(VaIJ )
J
= 
ar
, (Va2J)
J
= ra()
 , (\7a
J
3J) =   ..
r sm () ac.p
(2.1.146)
With regard for (2.1.146), the motion equations in the spherical coordi
nates will have the form
1 ap 1 f)P 1 ap
ar = p f)r +Fr, ag = pr f)() +F(}, acp = pr sin () f)c.p +Fcp,
(2.1.147)
where the an a(}, and acp are determined by formulas (2.1.145). If there
is no heat exchange (the adiabatic motion) , then the motion equation
will coincide with (2.1.139) , which with regard for (2.1.143) will have the
form
as + Ur as + U(} as + ~ as = 0. (2.1.148)
f)t ar r f)() r sin () ac.p
(2.1.149)
3) In the case of the Cartesian coordinates (see Fig. 2.6), the squared
element of the length between two infinitesimally close points is equal to
It follows from here that the metric tensor is a unit matrix, gIl = g22 =
g33 = gIl = g22 = g33 = 1, and the remaining gij are equal to zero. For
this reason, the coordinate and physical components coincide:
and the equations of continuity and motion will have the form
112 2 Fundamental Principles and Laws of Continuum Mechanics
(2.1. 151)
(2.1.152)
Problem 2.2. Assuming the body forces and the body heat sources
to be equal to zero (Fi = 0, qv = 0), write the equations of continuity
(2.1.7), motion (2.1.24) , and energy (2.1.50) in divergence form (the
Eulerian coordinates). (Under the divergence form of equations, the
2.1 Equations of Continuity, Motion, and Energy 113
following form of equations is meant: aV;;t" s + \7 j'ljJi ... s j = 0, where cpi ... s
and 'ljJi ... s j are some arbitrary tensor functions . Equations in divergence
form are sometimes called the differential conservation laws.)
Solution: Let us multiply the continuity equation (2 .1.7), which is al
ready written in divergence form
ap
at + \7 J.pvj = 0
,
(2.1.154)
(2.1.155)
(2.1.157)
Let us multiply the motion equation by the velocity Vi and sum over i .
As a result, we obtain a single equation:
(2 .1.159)
2
Multiplying the continuity equation by v2 and adding to (2.1.159), we
obtain the equation:
V2) +V'j (. v 2) ..
ata (P2 pvJ 2 =ViV'j(jtJ. (2.1.160)
(2.1.161)
dqe = TdS.
Substituting this equation into the definition of the specific heats, we
obtain:
(2.1.162)
(2.1.164)
To determine (g~ )T, let us make use of the condition for the total dif
ferential of the free energy F
By virtue of
dF = (~~)v dT + (~~)TdV,
it follows from (2.1.165) that
(2.1.166)
From the equality of the mixed derivatives aEf;.tv = aa::T and equations
(2.1.166), we obtain:
(2.1.167)
Since (gC)r < 0, then C p > C v ; i.e., the specific heat under a constant
pressure is larger than the specific heat under a constant volume. For
the perfect gas PV = RT, the Meyer's relation Cp  Cv = R follows
from (2.1.168).
to
Figure 2.7: The true trajectory Xi = Xi(~j,t) and its variations Jxi in
the action functional.
S= itll
to v
LdV dt+ itl
to
Asdt (2.2.1)
oS = it'Jvr
to
JL dV dt = O. (2.2.2)
In the second case, the particle trajectory is varied both inside the vol
ume and on the boundary. The condition for the minimum of functional
(2.2.1) under the trajectory variation oxi leads to the equation
JS = i t'l
to V
JL dV dt + it'
to
JAs dt = o. (2.2.3)
(2.2.4)
dedede:
itotlivo{ aXi
a~ bxidV dt = itl ( a~ a(bxi) dV dt
to ivo aXi at
itotlivo{ ut~ (~~uX bXi)dV dt ihto ivo{ut~ (~~uX )bxidV dt
10 (:~ I::bxi) dV 1: 10 1
%t (:~ )bxidV dt.
Since the particle trajectory is fixed at the moments of time to , tl, we
to tl
have bx 1 = bx i 1 = 0; therefore, the first integral vanishes, and we
ito l
have
itotllVo ~bxtdV
aL
uX
dt = 
tJ a (~)bxtdV
!'l
Vo ut uX
aL . dt. (2.2.8)
(2.2.10)
2.2 The Hamilton Ostrogradsky's Variational Principle 119
Since the variations 8xi as well as Va and S~ are arbitrary, the Euler
Lagrange equations
(2.2.11)
:t (:~)  :~ = o.
It follows from a comparison of this formula with (2 .2.11) and (2.2.12)
that the Euler Lagrange equations for a continuum contain an addi
tional item at: j
related to the action of internal stresses /} ij. Equation
(2.2.11) is the partial differential equation for the integration of which
it is necessary to specify the boundary conditions on the surface S~
bounding the given material volume. One can specify on the surface
o
S~ either the boundary positions (displacements) xn = x~ ei  the first
boundaryvalue
_ . 0
problem  or the surface forces applied to the boundary
Pn = p; ei  the second boundaryvalue problem. A mixed problem is
possible when the displacement xn
is specified on a part of the boundary
120 2 Fundamental Principles and Laws of Continuum Mechanics
S~l and the forces are specified on the remaining part S~2 (see Fig. 2.8).
The given displacements u i = xi  ~i are the boundary conditions for
equations (2.2.11). If a vector of forces fin is given on the boundary,
then the boundary conditions for the Euler Lagrange equations (2.2.11)
are determined from the first equation (2.2.12) .
Note that these boundary conditions refer to a reference frame in
Lagrangian coordinates. The second equation in (2.2.12) is the equation
of state and will be discussed in detail in the following.
As an example, let us derive the equations, which can be obtained
from (2.2.11) for the Lagrange function L of a formless medium of the
form
L
aX )2
= Po (( at
i . .)
E(x :j , S)  <I>(x') , (2.2.13)
and substituting into (2.2.11) and (2.2.12), we obtain the motion equa
tions:
a ..
.. i a (J 'J Fi
(2.2.14)
pox = a~j + Po ,
boundary conditions, and expression for the Piola Kirchhoff tensor (equa
tion of state) in terms of the derivatives of the internal energy
(2.2.15)
It can be seen from here that equations (2.2.14) coincide with motion
equations (2.1.21) obtained from the second Newton's law written for a
formless continuum.
Let us discuss in more detail the second equation in (2.2.15), which
is the equation of state of a formless continuum (see Section 2.1.4). The
internal energy E is usually specified as a function of the strain tensor
Eij , which in the case under consideration gij = bij, g?j = bij has the
form (1.2.23)
1 (axk ax k )
Eij ="2 a~i a~j  bij . (2.2.16)
o .. ox i oE
a tJ = Po o~k OEkj (2.2.17)
a
A Ii 1
= vg ox k a
oe 0 ki
. (2.2.18)
P= po/J9 (2.2.19)
and the relationship (2 .2.17) , we write the equation of state in the actual
configuration:
a ik = P oE . (2.2.20)
OEik
The equation of state (2.2.20) can be obtained independently from
the energy equation. Let us define some process in the space of states
Eij, S with the aid ofinfinitesimal increments (variations) of independent
parameters tStij = tijlSt, tSqe = T tSs, which lead in accordance with
(2.1.49) to the variation of the specific internal energy:
1 ..
tSE = _a tJ Mij + T tSS. (2.2.21 )
P
On the other hand, the quantity E = E(Eij, S) is a full differential of its
variables
oE oE
+ aS tSS,
A
/f ( OUi) dV dt = 
'TriO at /f O'Tri . dt,
at0u"dV (2.2.29)
II :: O(~~;)dV
J
dt =  II O~j (::JouidV dt.
J
(2.2.30)
1J(( O'Tri oH
+
 at  ou i
0 (OH)) i
o~j oui . c5u
,J
+ (OU
i OH) )
at  O'Tri c5'Tri dV dt = O.
(2.2.31 )
(2.2.32)
iI = r H dV.
ivo
For the Lagrange function density (2.2.13), the quantity iI is equal to
the total continuum energy in the volume Vo. The total energy variation
is determined by the expression
dH
dt
r (OH'
= iVa oui U
i oH. 8H)
+ 07ri 7ri + 7ft dV (2.2.35)
dH
 8H1
dV
dt  Va 8t '
from where an important conclusion follows that the total energy of a
continuum is conserved if the Lagrange function density L or the Hamil
ton's density H of a closed system does not depend explicitly on time.
Since the law of energy conservation is confirmed by all experiments
without any exception, the time cannot enter explicitly the set of the
independent variables of the Lagrange function. This points to the fact
that the time is homogeneous, and the motion equations are invariant
with respect to the translations in time.
Let us generalize the abovedeveloped method and find the conditions
for the function a = a( u i , U~j' 7ri, t, S), at which it is the motion integral
A= r adV
iVa
= const.
dA
dt
= r (8a8t + !!::....ui
iVa ou i
+ !!::""ir)dV = r (8a + !!::.... oH _ !!::.... OH)dV
07ri' 8tiVaoui 07ri ou
07ri i .
(2.2.36)
By introducing the Poisson brackets
(2 .2.37)
dA = 8A + [AH]. (2.2.38)
dt 8t
If A does not depend explicitly on the time t, then it follows from (2.2.38)
that
dA = [AH]
dt '
2.2 The Hamilton Ostrogradsky's Variational Principle 125
[AH] =0
that A is an integral A = const.
2.2.3 EulerLagrange Equations in Eulerian Coordinates
and Murnaghan's Formula
Let us obtain the Euler Lagrange equations from the variational prin
ciple (2.2.2) in Eulerian coordinates. The main purpose of the present
section is the construction of the technique for the action variation in
Eulerian coordinates and the derivation of the motion equations and
equation of state of a formless medium in the case of nonlinear defor
mations. This equation of state is called the Murnaghan's formula in
the literature and has a very nontrivial form. It can also be obtained
by other techniques, which are discussed below. In the case of using the
variational principle, however, the Murnaghan's equation is obtained in
a natural way as a result of the action variation.
We will assume the Eulerian coordinate system to be a Cartesian
system, i.e., gij = bij , and the EulerLagrange equations will follow
from the variational principle
(2.2.39)
(2.2.42)
Let us express all functions in (2.2.40) in terms of the new variables. For
this purpose, we define the direct b; and inverse aj distortion tensors
(2.2.43)
(2.2.45)
(2.2.46)
(2.2.47)
2.2 The Hamilton Ostrogradsky's Variational Principle 127
aL a (aL) k a( aL) k
aui  axl auk,I 8i  at auk 8i
auk aL auk a (aL) auk a (aL)
axi auk + axi axl auk,1 + axi at aui = O. (2.2.48)
Using the expression for the total derivative of L with respect to the
coordinate Xi
(2.2.49)
:~ lu k = P:!, aL
aui = 0, (2.2.50)
where we have taken into account the fact that p is a function of U~j
(2.2.45). Using (2.2.44), let us find the derivative with respect to time:

ap
=
ap aaj ap alai I
auk.
,J
aakauk
 = aak
 = Po
J
aak .
,J J J
(2.2.52)
ur
vve now determme
. t he d .
envatIves ifi,
Obi ov
7)k.
2 L et us dJ:r .
1uerentIate equa
U ,k U,j
tions in (2.2.43):
ab~ = bi b8 (2.2.55)
au1 I n
,8
It is easy to find with the aid of (2.2.44) and (2.2.55) the second deriva
tive:
k av 1 k 8 ab~
ai Vl~ =ai vlU ~
uUJ
uU
J
k b1 d 8 ,I j j
VI ai k US U = VI ui V = Vi V . (2.2.56)
Using (2.2.40) , (2.2.54), and (2.2.56) , we obtain
k a
a+J
, auk
I
,
k ap
a (1
 v 2 Eip ) + pk
a v a 2
,1
'auk,1 2 'auk,1 2
pa k
aE(u1)
, + p (1_v 2  E  ip) Jl
'auk,1 2 '
pa1b~L + PVivl  ai + pLJi = PVivl  ai, (2.2.57)
where we have introduced the following notation for the stress tensor:
I k aE k k) aE
a i = a i P J'l k = p(Ji  u,i J'l k (2.2.58)
uU ,1 uU,1
Substituting (2.2.50), (2.2.51), and (2.2.57) into (2.2.49), we will find the
motion equations
(2.2.59)
2.2 The HamiltonOstrogradsky's Variational Principle 129
8p 8 I
8t + 8x! fYV =0
by vi and subtracting from (2.2.59), we obtain:
(2.2.60)
( 8vi I 8vi ) 8a il i
p {it +V 8x! = 8x! +G , (2.2.61 )
.. = P(8E
0")
8E) .
   2cik (2.2.63)
Beij Bekj
aE aE
6E = 8c ij bCij + as 6S, 6xl = vi 6t, (2.2.65)
where we have used the last equation in (2.2.65). Multiplying the left
and righthand sides of equation (2.2.66) by iaij , we find with regard
for the symmetry a ij = a ji :
1 .. 1 .. 8 .
a') ii' M = a')  . (6x' ). (2.2.67)
p ) pax)
2.2 The Hamilton Ostrogradsky's Variational Principle 131
Let us express the variation of the strain tensor bij in terms of the
displacement variation bxl. According to (1.2.26), we have in Eulerian
coordinates that
1 O~k O~k
Cij = "2 (bij  oxi oxj )'
(2.2.68)
bc tJ"
1(
= 2 (bJ'l  O(bxl)
2cJ'I )oxi
 + (blt 0(8x l ))
 2c t'l )oxj
 . (2.2.70)
Let us substitute (2.2.67) into (2.2.64) and (2.2.70) into (2.2.65). Equal
ling then (2.2.64) and (2.2.65), we find that
I' oE oE
(J J = p(bil  2cil)~' T= oS' (2.2.71)
UCij
Solution: Let us write the Eulerian strain tensor in' the form
(2.2.72)
where
(2.2.73)
Taking into account (2.2.72) and (2.2.73), we can rewrite formula (2.2.71)
as follows:
.. aE
(jt) = 2Pt;ik~' (2.2.74)
ugkj
It follows from the mass conservation law written for an infinitesimal
volume dV that
a~j IdVo,
lax i
pdV = Po dVo, dV = (2.2.75)
p = Pol 1axj ae 1
a~i 1 = Po 1axj = Povg,
 (2.2.76)
where the last equality follows from a linear algebra theorem, which
states that the determinant of the product of matrices is equal to the
product of the determinants of these matrices:
where we have used formula (2.2.53), which has the following form in
the given notations:
ag _1  1
~
ugkj
=ggk)" gik gkj = Uij'
J;
/Ti)' _ 2 aE dj
v  p ap U (2.2.78)
2.2 The Hamilton Ostrogradsky's Variational Principle 133
(2.2.79)
oS =
(2.2.80)
oAs = J.~ r f~ ox dS
~
n = l'J1
~
ox k ~ . ek dSn =
A l J1'
~
ax k
a tj ox k dSn
~
(2.2.82)
With regard for (2.2.81), the variation of the first item in (2.2.80) is
equal to
, l.tll (
oS =
to v
Po >l
uX
aL OX k aL OX. k aL,)
k + >l ' k + >l,Ogij dTdt.
uX ug,)
(2.2.83)
(2.2.84)
Since
134 2 Fundamental Principles and Laws of Continuum Mechanics
8 ( 8E 8x k k) 8 ( 8E 8Xk) k (2.2.86)
= 8f,i 2 8g ij 8f,j 8x + 8f,i 2 8gij 8f,j 8x.
(2.2.87)
Substituting (2.2.84) (2.2.87) into (2.2.83), we find with regard for the
equalities 8x i lh = 8Xilt2 = 0 that
8S' =
The last item in (2.2.88) has been obtained with the aid of the Ost
rogradskyGauss theorem in the following way. Let us introduce for
convenience the quantity
Hi = 8E 8x k 8xk
8iij 8f,j .
Since the divergence ~~,i does not depend on the coordinate system,
we at first go over to the surface integral in a coordinate system of the
reference frame, and by using the formula (1.3.22) ni dSn = v'9n? dSn ,
we then go over to the accompanying coordinate system:
Substituting (2.2.82) and (2.2.88) into (2.2.80) and equalling to zero the
surface integrals and the volume integrals, we obtain:
82 x k 8 (8E 8Xk) 8<fJ
(2.2.89)
8t 2 = 8f,i 8i ij 8f,j  8x k '
2.3 Conservation Laws for Energy and Momentum 135
(2.2.90)
Substituting (2.2.90) into (2.2.89), we obtain with regard for the conti
nuity equation the motion equations
(2.2.91 )
which by virtue of the symmetry {fij = (fji coincide with (1.3.20). Equa
tion (2.2.90) enables one to find the stresses in a medium if the specific
internal energy is given as a function of t ij , S, and it also coincides with
the corresponding equation (2.2.20).
With regard for this remark, let us obtain the conservation laws for
a formless continuum. The expression for the action in Lagrangian co
ordinates in a Cartesian basis of the reference frame has the form
(2.3.1)
(2.3.2)
(2.3.4)
(2.3.7)
2.3 Conservation Laws for Energy and Momentum 137
(2 .3.8)
(2.3.11)
(2.3.12)
+
138 2 Fundamental Principles and Laws of Continuum Mechanics
(2.3.15)
If the real motions are considered, for which the Euler Lagrange equa
tions (2.3.15) are valid, then equation (2.3.14) takes the form
Jr {
r
}
a ( aL ")
at L 5t + ai1 i 5u l
a (
+ a~j " aL
L 5e + aU~j
 ")}
5u l dV dt = 0,
(2.3.16)
(2.3.17)
Formulas (2.3.16) and (2.3.17) can be extended for a more general class
of transformations. Let us introduce for convenience the following nota
tions:
(2.3.18)
in which the Greek letters denote the components of the fourdimensional
vector aC< (a = 0, 1, 2,3) and the Latin letters denote the components of
the threedimensional vector bi (i = 1,2,3). In this notation, equations
(2.3.16) and (2.3.17) can be rewritten in the form
a ( aL ")
axc< L5 x C< + a(aui/axc<) 5u' = 0, (2.3.19)
 " " au i f3
5u' = 5u l  ax f3 5x . (2.3.20)
= y i (I
j
u Ii x , uj , au
ax li ' 10C<) (2.3.21 )
2.3 Conservation Laws for Energy and Momentum 139
(2.3.22)
where
X/3 = aX/31 yi = ayi I (2.3.23)
a aca "=0' a ac a "=0
(2.3.24)
where
. . au i /3
y~ = y~  ax/3 Xa (2.3.25)
Substituting (2.3.24) into (2.3.19) and taking into account the indepen
dence of ca , we obtain n relationships (0: = 1, ... , n):
(2.3.26)
Thus, we have proved the Noether's theorem, which states that, if the
action (2.3.1) is invariant under the infinitesimal transformations (2.3.22)
depending on n small parameters, then n differential conservation laws
(2.3.26) are valid in Cartesian coordinates, in which u i are the solutions
of the Euler Lagrange equations (2.3.15), and xg and y~ are determined
from equations (2.3.23) and (2.3.25), respectively.
For each relationship (2.3.26), one can obtain an integral conservation
law in an Euclidean space. The equations
follow from (2.3.18) and (2.3.26), where the Qa are the Noether's char
ges and J~ are the Noether's currents. Multiplying both sides of the
first equation in (2.3.27) by dV and integrating with the use of the
Ostrogradsky Gauss theorem, we obtain the equation:
(2.3.28)
where the integral over an infinitely distant surface stands on the right
hand side. Since L ; 0 at infinity, the integral on the righthand side of
(2.3.28) is equal to zero. This implies that the integrals
J Qa dV = const ; (2.3.29)
140 2 Fundamental Principles and Laws of Continuum Mechanics
thus, they do not depend on time and are the integral conservation laws.
Strictly speaking, only the integral conservation laws have a direct mean
ing and determine the conservation of the Noether's total charge Qo: in
a finite (and infinite) volume, if the Noether's total current in (2.3.28)
across the surface is equal to zero: fS n Jtnj dSn = O. Note that the
availability of the differential conservation laws does not lead in a gen
eral case to the existence of the integral conservation laws. As will be
shown in the following, the existence of the solutions of Killing's equa
tions is required. In the case of an Euclidean space such solutions exist;
therefore, one can find from the differential conservation laws (2 .3.26)
and (2.3.27) the integral conservation laws (2.3.29).
Consider a particular case of the transformations (2.3.21) , a transla
tion in space and time
(2 .3.30)
Substituting (2 .3.30) into (2.3.22) and (2.3.24), let us write the variatons
t5x/3 = c;/3, t5u i = 0, from which we obtain the formulas:
Y~ =0, (2.3.31 )
where we have with regard for (2.1.12) , (2.1.23) , and (2.1.24) that
r,0
j _ 8L 8u i _ a ij i
. (1 U,
8u~j 8t
110 (2.3.34)
E = J H dV = const (2.3.36)
Pk = JaL au i
au i a~k dV = const , (2.3.37)
(2.3.38)
dMij
dt
J (aT? aT? )
~i7it  ~j7it dV = 
J (OTjk aT;k )
~i a~k  ~j a~k dV
dM ij
dt =
J (TotJ  T)dV
Jt , (2.3.39)
where the integration is performed over the total space. It follows from
here that the conservation law for the momentum moment Mij = const
is valid under the condition of the symmetry of the tensor Iij = T ji . In
the case under consideration (2.3.34) , the tensor Tij is not symmetric,
which formally points to the presence of an internal moment. This is not
so, however, since equation (2.3.38) does not determine the momentum
moment in a reference frame 18 . According to 18 ,20, the shoulder for the
momentum moment in formula (2.3.38) should serve ~i + u i rather than
~i; therefore, one should make in (2.3.38) a substitution ~i + ~i + u i .
In this connection, we recall that the condition for the symmetry of
the stress tensor (j i j = (jji was obtained from the conservation law for
the momentum moment in an accompanying coordinate system, where
the shoulder is equal to ~i (the shoulder vector is equal to ';i~) and
the moment is determined by the formula (2.3.38). By a choice of the
coordinate system in the reference frame, one can achieve the coincidence
of the Eulerian and accompanying coordinate systems; therefore, in the
Eulerian coordinate system, the equation uij = u ji also follows from the
condition of the moment conservation.
Note that the Lagrange function in (2.2.1) and (2.2.5) and, corre
spondingly, the tensor of energy and momentum (2.3.32) are not deter
mined uniquely with the accuracy up to an arbitrary function. Employ
ing this arbitrariness, one can always ensure the symmetry of the tensor
of energy and momentum 16 ,17.
Let us show that the motion equations (2.2.14) will not change if the
total derivative of some vector function c.pi is added to the Lagrangian:
form
where we have taken into account the fact that 8u k (h) = 8u k (t2) = O.
Substituting the above expression into (2.3.41) , we obtain the formula:
l1 h 'L i
U
Ti8u do' +
lit2 (;::)T +
J'lL
U J'l' .hi
U,/
J'l
.r
i +
..
o'J
(J )
i 0
8u nj dSn dt
= 0,
v; to uU s~ tl uU ,J uU
from where we find that the motion equations do not change [compare
with (2.2.11)]:
oij aL acpi
(J =.   (2.3.43)
au'.
,J
au i '
It follows from the comparison of (2.2.12) and (2.3.43) that there exists
a functional arbitrariness in the determination of the stress tensor. Since
cpi3 = cpi3(u i ), we can rewrite (2.3.40) as follows:
, acpi3 au i
L = L + aui ax i3 ' (2.3.44)
(2.3.46)
By the choice of !.po. and correspondingly of tj;~'Y, one can attain the
symmetry of the tensor T = T;J; therefore, the tensor of energy and
momentum for a formless continuum is assumed to be symmetric in the
following .
2.3.2 Conservation Laws in an Arbitrary
Coordinate System
As was shown above, the conservation laws (2.3.29) follow in a Cartesian
coordinate system of the Euclidean space from the differential continuity
equations (2.3.27). The majority of the authors usually believe that this
is valid also in the general case for an arbitrary metric tensor gij(~).
For a fourdimensional Riemann space, this question has been discussed
in detail, for example, in Logunov's work 17 . The results obtained by
Logunov can be transferred to the threedimensional Euclidean space of
a classical mechanics.
Let a curvilinear coordinate system be defined at each point of space,
which is specified by the tensor gij(X) . In this case, equations (2.3.27)
and (2.3.32) can be written in the form
(2.3.47)
where
Qk=Tg , J~=T1, j,k=1,2,3
and \7j is the covariant derivative (1.1.25), (1.1.27). The tensor Tjk is
assumed to be symmetric below; therefore, the conditions Tjk = Tkj are
satisfied. By using the Weyl formula (1.1.73), we can rewrite the first
equation in (2.3.47) as follows:
8H 1 8 j _
7ft + V 8~j (VTo)  O.
(2.3.48)
2.3 Conservation Laws for Energy and Momentum 145
J
conservation law
H dV = const , (2.3.49)
where the integration is performed over the total space. Thus, the space
properties gij(X) do not affect the energy conservation law. This is re
lated to the fact that the energy conservation law is related to the action
invariance with respect to the time shifts (2.3.30) X'O = xO + co , and the
time in the classical mechanics does not depend on the spatial coordi
nates.
Multiplying the second equation in (2.3.47) by gik with regard for
8g;k = \7 jgik = 0, we rewrite this equation for the contravariant com
ponents:
(2.3.50)
It follows from this formula that the conservation laws will take place
provided that
J ygrjJlj de de de = o.
Since r)l r)1(9kl) , 9 = g(gkl), and Jlj is an arbitrary function of
xi, it is indeed required to find the conditions for gkl under which the
identities
(2.3.51 )
are satisfied.
146 2 Fundamental Principles and Laws of Continuum Mechanics
Let us find the conditions for gkl at which the conditions (2 .3.51) are
satisfied. It turns out that the metric tensor gij should satisfy for this
purpose the forminvariance condition
(2.3.52)
(2.3.53)
(2.3.54)
where
(2.3.55)
The quantity g'i j (e'k) is determined in accordance with the rules for the
transformation of contravariant components under the transformations
(2.3.53) :
g'ij (ek) = gij (ek) _ ~~~ 1Jk + gil ~~; + gkj ; ; : = gij (~k) + ~iryi + ~j1Ji.
(2.3.57)
Substituting (2.3.57) into (2.3.54), we obtain:
It follows from here that the metric tensor gij will be forminvariant un
der the infinitesimal transformations (2.3.59) if 1Ji will satisfy equations
2.3 Conservation Laws for Energy and Momentum 147
(2.3.60). The solutions of equations (2.3.60) are called the Killing's vec
tors TJi. In the Cartesian coordinates xi, the equations (2.3.60) simplify
to
(2.3.61 )
(2.3.62)
OTJiQi ..
~ + \lj(TJiPJ) = O. (2.3.65)
148 2 Fundamental Principles and Laws of Continuum Mechanics
J
%t 1]iQi dV =  J y'g'lj(1]i Jij ) dT
Let us find the conservation laws for the momentum and momen
tum moment. Let the relation between the Cartesian and curvilinear
coordinates be determined by the formula Xi = yi (~j). Then in accor
dance with (2.3.61) and (2.3.62) the Killing's vectors in the curvilinear
coordinate system ~i will have the form
(2.3.67)
J.
into (2.3.66) , we find with regard for (2.3.47) and (2.3.50):
ik 0
~i 9 Tk Cj dV = const,
J~iykQiwikdV ~ J(~iyk
=  y~yj) QiWjk dV = const,
2.3 Conservation Laws for Energy and Momentum 149
from where we obtain by virtue of Wjk = const the conservation law for
the moment of momentum Mjk in an arbitrary curvilinear coordinate
system:
(2.3.71)
[ef. (2.3.38)] .
Thus, we have proved that the conservation laws for momentum and
energy follow from the homogeneity of space and time and the conserva
tion law for the momentum moment follows from the space isotropicity.
The relativity principle of Galilei plays an important role in classical
mechanics. According to this principle, all phenomena in different in
ertial reference frames proceed in the same way. The Galilei's principle
postulates the existence of many inertial reference frames possessing the
property that the isolated material point is either at rest with respect
to these frames or moves at a constant speed. The relation between the
Cartesian coordinates in two reference frames, one of which x'i moves
with respect to the other one, xi, at a constant speed vb, is given by the
Galilean transformations
t' = t. (2.3.72)
S should
It follows from the Galilei's relativity principle that the action
be invariant under the Galilean transformations. As a result, the con
servation law for the momentum of the center of mass of a moving con
tinuum emerges.
Let us derive the conservation law for the momentum of the center
of mass. Differentiating (2 .3.72) with respect to time:
and multiplying the left and righthand sides of this equation by PodV ,
we integrate it over the total space in a reference frame:
a 1 .
a Pov' dV
t Va
from where the conservation law for the momentum of the center of mass
follows:
Po = r Po vi dV = const.
iVa
(2.3.73)
a(
ax) Lrtl
.+ aui.  .) = 0,
aL bu' (2.3.74)
,1
_ . . au i .
bu' = bu'  a ryl.
Xl
2.3 Conservation Laws for Energy and Momentum 151
(2.3.76)
(2.3.77)
from where
(2.3.79)
152 2 Fundamental Principles and Laws of Continuum Mechanics
8 k 8 k 82  k
oxk Ti = oxk Ii + oxkoxj Si J = 0,
and is symmetric with respect to i and k:
References
(3.1.1)
(3.1.2)
Let the first k quantities aI, a2, . .. , ak (k ::::; n) have the independent
dimensions [all = AI , [a2l = A2, .. , [akl = Ak (For the case k = 3
these may be any three quantities having the independent dimensions of
the form U" Mi3T'Y. ) Then, the dimensions of the remaining quantities
may be expressed in terms of the first k quantities:
[al Af' A~2 ... A~k, [ak+ll = Af' A~2 ... A~k , ... ,
[anl AI' A~2 ... Ak'k. (3 .1.3)
In the new system of measurement units, the formula (3.1.2) takes the
form
(3.1.4)
(3.1.5)
where
a
II
a~" a~2 ... a~k '
an
(3.1.6)
........ a
Po, Vo a
To,,,! 
.
E = cvT, P = (T  l)pE, Cij ="21 (OVi OVj)
OXj + OXi '
(3.1.9)
C
2 P
= "( = I (T  1) cvT. (3.1.10)
p
160 3 Features of the Solutions of Continuum Mechanics Problems
2 2 b
Ii = POvOa Ai(a, ,,/,  , M, Re, Pe, Fr), (3.1.11)
a
0,
(3.1.13)
where
v _ p  T
v=,
Vo
p=,
Po
T
To'
a
E : 1 (OVi OVj)
E F=pT, Cij = 2" OXi + OXi .
cvTo'
The formula (3.1.12) follows from the form of equations (3.1.13) and the
boundary conditions fj = on the ellipsoid surface.
As was noted above, two flows will be similar if their similarity cri
teria, i.e., the arguments of the function (3.1.12), coincide:
b1 b2 VI v2 alvl a2 v2
al a2, "/1 = "/2, =,
al a2 cl C2 WI W2
alvl a2 v 2 VI V2
, ,fo2. (3.1.14)
III lI2 fol
3.1 Similarity and Dimension Theory 161
in the region 0 < r < 00 should satisfy at each moment of time the
condition
q=47f LX) Tr 2 dr. (3.1.17)
It follows from the physical formulation of the problem that the temper
ature T will depend on the following parameters:
T = f(w, q, t, r). (3.1.18)
Let us choose the first three parameters as the quantities with inde
pendent dimensions [w] = m 2 Is, [q] = KO 1m3, [t] = s, where s is the
second, m is the meter, and KO is the temperature in Kelvin. Using the
Pi theorem (3.1.5) , (3.1.6), let us rewrite (3.1.18) in a nondimensional
form
T(wt)~ = f(r2),
q wt
from where it follows that the solution of the problem (3.1.15) (3.1.17)
can be written as
T = (w~)~ f(~), (3.1.19)
is such an example. In this case, the total energy of gas in the considered
flow region is not constant, which does not enable one to find the value
of a. The parameter a in this problem is found from the condition of
a passage of the solution through a singular point and is an irrational
number. A feature of this solution is that, at the initial moment of
time, it is not selfsimilar (it depends on the character of the motion of
a piston creating the shock wave) and becomes selfsimilar as the shock
wave approaches the coordinate origin (the notion of shock wave will be
introduced below in this section). The dimensional constant A is deter
mined from the condition of matching of nonselfsimilar and selfsimilar
solutions.
Let us return to our problem and find the function f(~) in (3.1.19).
Substituting the solution (3.1.19) into equations (3.1.15) and (3.1.16),
we obtain an ordinary differential equation:
where f' = ~. From the condition f + 0, l' + at r + 00, ~ + 00,
we obtain from the above equation the equation 4f' + f 0, whose
solution has the form
(3.1.20)
Substituting (3.1.20) into (3.1.19) and then into (3.1.17), we can easily
find the constant C = 1/(2J1f)3. As a result, we obtain the solution of
the given problem in the form
q r2
T = e 4wt . (3.1.21)
(41TWt)3/2
the dependences Y(Xi) are taken at t = 0. If the initial conditions are
specified at t = in the overall space 00 < Xi < +00, then one says that
the Cauchy problem is defined in the hyperplane t = 0. In a boundary
value problem, it is necessary to specify the boundary conditions on
the surfaces bounding the region D at t > 0. As will be shown below,
the formulation of a boundaryvalue problem depends on the type of
equation or system of equations. The type of system of equations is
determined by its characteristics. There are sufficiently many books in
164 3 Features of the Solutions of Continuum Mechanics Problems
a1n bl1
u ,B=
ann bn1
E au au
at + C ax = f ' (3.2.2)
E au C au = f
at + ax '
au au
Edt at + E dx ax = duo (3.2.4)
3.2 The Characteristics of Partial Differential Equations 165
Figure 3.2: To the formulation of the Cauchy problem for system (3.2 .2).
(3.2.5)
dx
dt = Ai, (3.2.7)
If all of the zeroes Ai in (3.2.8) are real and different, then the system
(3.2.2) is called hyperbolic.
An eigenvector Si corresponds to each eigenvalue Ai. It is found from
the system of equations
(3.2.9)
o
o
A = SlCS, (3.2.10)
o
As a result of this transformation, the vector u goes over to v by the
formula u = Sv. Substituting this formula into (3.2.2), we find:
av av
at + A ax + Gv = g, (3.2.11)
where
G=S 1 (aat
~S+CS a)
ax '
One can say that, if the system of equations (3.2.2) can be reduced to
the form (3.2.11) , then it is called hyperbolic (note that the requirement
of the absence of multiple roots Ai is not necessary in this definition).
Let us formulate an initial and boundaryvalue problem for the hy
perbolic system (3.2.11) in the strip 0 :::; x :::; I, t > O. We at first
elucidate the basic features using a simple example of the system of
equations
3.2 The Characteristics of Partial Differential Equations 167
t
dx \
dt = /\1
"
T
x
o
Figure 3.3: To the formulation of a boundaryvalue problem for system
(3.2.12).
(3.2 .12)
where Al > 0, A2 > O. The system of equations (3.2.12) has the charac
teristics
which are shown in Fig. 3.3 by straight lines. To determine the solution
in the strip 0 ::; x ::; I , t > 0, one must specify the initial conditions at
t = 0:
where >'i(X, t) > 0 in the region 0::; x ::; l, 0 ::; t ::; T, with the boundary
conditions on the left boundary
(i = 1, .. . , no),
(i = no + 1, .. . , n),
[aij(t), f3ij(t) are smooth functions], with the initial condition
Vi I = ipi (X) ,
t=O
which are matched with the boundary conditions, will have a unique
solution if the boundary conditions satisfy the dissipativity conditions
expressed in the form of the inequalities
no n
 L >'iV; + L >'iV;::; >'0 on the right boundary,
i=l i=no+l
n no
 L >'iV; + L >'iV; ::; >'0 on the left boundary,
i=no + l i=l
where >'0 > 0 is a constant. If aij , f3ij are constant, one can assume
>'0 = O.
The above definition of the characteristics (3.2.5) (3 .2.8) and the
principle of the formulation of the boundary conditions (the number of
boundary conditions should be equal to the number of characteristics
emanating from the boundary) are also valid for quasilinear equations
in which the matrix C and the vector f in (3.2.2) [or the matrices A, B
and the vector 9 in (3.2.1)] depend on the solution itself:
au au
at + C(t, x, u) ax = f(t , x, u). (3.2.13)
o
Figure 3.4: The picture of the C+ characteristics in the Riemann wave.
The line x = xp(t) is the piston trajectory.
each (t, x) point above the line Se . The line Se is called the caustic and
is an envelope of a family of the characteristics. There arises indeed a
discontinuity of the first kind at point A , which is called a shock wave
in mechanics. As will be shown below, the characteristics (3.2.5) (3.2.8)
of the continuum mechanics are the lines along which the disturbances
propagate. Just this fact conditions their big role in continuum mechan
ics problems.
In the case of a larger number of the variables, the characteristics are
defined in a similar way and are the hypersurfaces in the corresponding
space. For example, for the system of equations3 ,7
(3.2.14)
15
au av_o av_au=o
at + ax  , (3.2.16)
at ax '
a wellposed problem is the Dirichlet problem of determining the function
u(t, x) [and v(t, x) with the accuracy up to a constant] in a region D(x, t)
bounded by a closed curve T x = x (~) , t = t(~) on the basis of a function
ul'Y = <1>(0 given on the boundary "f.
(3.3.1)
8] + 8] di i = 0
8t 8i i dt '
we obtain with regard for the definition (3.3.2) and the relationship
where
it=ifD . (3.3.4)
Figure 3.6: The coordinate system and the integration region in a refer
ence frame in which the velocity of point 0 is equal to zero.
The body forces and the body heat sources are not taken into account
in equations (3.3.5). As was noted above, these equations cannot be
used on a discontinuity; therefore, let us go over from the differential
equations to the integral equations.
For this purpose, let us consider a divergence equation of the general
form
(3.3.6)
where 'ljin = 'ljijnj. In this equation, the <p and 'ljij cannot only be
continuous, but also discontinuous functions of the coordinates xi. In
the particular case of the continuous functions, equation (3.3.6) follows
from it. In a general case, however, it describes a wider class of the
solutions. We recall that equations (3.3.5) were obtained in Section 2.1
from the integral equations.
Let us integrate equation (3.3.7) on a discontinuity in our case. As a
volume V, we choose a cylinder whose generatrix has the length band
is parallel with the vector of normal n at point 0 (see Fig. 3.6). A
part of the cylinder b/2 in length along the normal n
lies on the one
side of the discontinuity surface f(t, Xi) = 0, and the remaining part
lies on the other side. We will mark our functions in the first region
by the subscript "minus," i.e., <p_, 'lji!.., and we will mark the functions
174 3 Features of the Solutions of Continuum Mechanics Problems
in the second region by the plus sign, i.e., '1'+, 7/;~. The point 0 will
be always at rest ; however, the other points of the surface f(t, xi) = 0
have different velocities D = D(t, xi). Therefore, the surface f = 0 will
shift with time. If we choose a sufficiently small radius r of the cylinder
base and consider the surface f = 0 only in a small neighborhood of the
point 0, however, then we can neglect this shift. After these remarks,
let us calculate the integral in (3.3.7) over the volume and the cylinder
surface:
r
iS n
7/;n dSn = r 7/; j nj dSn = (7/;~  7/;i) nj1fr2 + is:.r 7/;n dSn,
iS n
(3.3.9)
where we have taken into account the fact that the normals to the op
posite cylinder bases have the opposite signs. Let us write the integral
over the lateral cylinder surface in the form
(3.3.10)
(3.3.11)
It is assumed that the functions 'I' and 7/;j are bounded and continuous,
with respect to t and Xi, together with their derivatives on both sides
of the discontinuity surface f(t, Xi) = O. In this case, the functions ('1'),
8b~)' and (7/;n) are also bounded and continuous functions of the time
t. Therefore, as the length of the cylinder generatrix tends to zero, i.e.,
J + 0, we obtain from (3.3.11) the relation on the discontinuity:
[7/;nJ = 0, (3.3.12)
[PUnJ = o. (3.3.14)
Un = Vn  D. (3.3.15)
For the momentum equation (3.3.5), one must consider three compo
nents:
f ni = aijnJ' (3.3.16)
(3.3.18)
(3.3.20)
lows from (3.3.15) that Vn = D; i.e., the particles of a continuum do not
intersect the tangential discontinuity surface. Assuming Un = in the
obtained relations, we can find the relations on the tangential disconti
nuities:
(3.3.21 )
It follows from (3.3.14) and (3.3.18) that the density jumps [p] and the
velocity component tangential to the discontinuity surface rUT] may be
arbitrary. At the shock waves, the quantities [un] f:. 0 and the jump
of the normal velocity component [un] f:. O. In the literature, it is con
ventional to supply the parameters ahead of the shock wave with the
176 3 Features of the Solutions of Continuum Mechanics Problems
subscript 1, and the parameters behind the shock wave are supplied
with the subscript 2 instead of the subscripts +,  . With regard for this
remark, let us write the continuity equation (3.3.14) in the form
(3.3.22)
Here, two substantially different cases are possible depending on the sign
of [un]. The first case UnI > Un2, PI < P2 corresponds to the compression
shock wave, and the second case UnI < Un2, PI > P2 corresponds to the
rarefaction shock wave.
The choice of the needed solution is determined with the aid of the
second law of thermodynamics (2.1.62) , which in the Eulerian coordi
nates xi in the reference frame K' has the form
as + U j\l j S
&t _ 1(.
 T qe + q") , i/ ;::: o. (3.3.23)
(3.3.24)
Integrating this equation over the volume V (see Fig. 3.6), we obtain
similarly to (3.3.11):
[PUnS] = ;.
7rr
rP
Jv T (qe + q') dV.
(3.3.25)
In contrast with (3.3.12), the integral over the volume has emerged on
the righthand side.
Let us show that this integral is different from zero and that it always has
a positive sign. Before proceeding to this proof, we note that the notion
of the discontinuity surface is a mathematical idealization. The flow
parameters at a shock wave rapidly change within narrow transitional
layers, the width of which h reduces with increasing the shock wave
strength down to the order of smallness of several free paths A. To
estimate the integral in (3.3.24), we shall assume the shock wave to be
weak, so that h A. In this case, continuum mechanics hypotheses will
be valid in the transitional layer. A specific heat flux is determined by
formulas (2.1.107) and (2.1.108):
. 1 i
qe = \liq, qi = ",\liT, (3.3.26)
P
and with their regard , we obtain:
3.3 Discontinuity Surfaces in Continuum Mechanics 177
Since the gradients outside a narrow transitional layer are small, the
integrals taken over the cylinder bases Sn1 and Sn2 will be small. In
addition, we note that the gradients of all quantities in the transitional
layer are large along the normal fi to the surface f(t , xi) = 0, and they
are small along all other directions lying in a plane tangent to the surface
f = o. A normaltothelateral cylinder surface fi' is perpendicular to
the normal fi (see Fig. 3.7); therefore, the integral along the lateral
surface Is' ~qn' dSn will also be small. One can consequently neglect in
(3.3.26) the integral over the cylinder surface Sn and assume that
(3.3.28)
Substituting (3.3.27) and (3.3.28) into (3.3.24) and taking into account
the relations j = P1Un1 = P2Un2 > 0 [see formula (3.3.14)], we obtain a
criterion for the solution choice:
Thus, the entropy S2 behind the shock wave should always be larger
than the entropy Sl ahead of the shock wave: S2 > Sl. Let us estimate
178 3 Features of the Solutions of Continuum Mechanics Problems
A
B
v
o
Figure 3.8: The adiabat of the van der Waals' gas.
the entropy increment behind the shock wave. The magnitude of a non
compensated heat is determined by the work of viscous forces . Using
formula (2.1.106) for II' and formula (3.3.27), we can find the entropy
change for a viscous, heatconducting gas:
Assuming in this formula that \1 iT '" [T]j h, eij rv div it '" , j '" Pu,
we obtain:
J [TF [uF
[S] rv puh 2 ('" T2 + /1 r ) (3.3.31 )
Since h ::; J, [T] =1= 0, and [u] =1= 0, the entropy jump in (3.3.30) will be a
finite quantity. One can find from the conditions at a shock wave (3.3.14),
(3.3.18) , (3.3.20) the entropy jump [S], which does not depend on the
transitional layer structure and is determined by shock wave strength
specified, for example, by the velocity jump [un]. As will be shown below
in Problem 3.1, it follows from (3.3.29) that, for the "normal" media for
which the adiabat is convex downwards, the compression shock waves are
stable. For the media having the intervals that are convex upwards on
the adiabat (the interval AB in Fig. 3.8), the existence of the rarefaction
shock waves is possible.
Problem 3.1. Determine with the aid of the Pi theorem the drag
force of a sphere as it enters at a constant speed v, a halfspace filled
with a viscous, incompressible fluid (see Fig. 3.9). The sphere radius is
equal to a. Consider the limiting cases v t 0 and v t 00.
Solution: The drag force F depends on the following parameters:
o
a
where p and f.1 are the fluid density and viscosity and t is the time.
Choosing p, v, and t as the independent dimensional quantities, let us
rewrite equation (3.3.32) in the nondimensional variables:
(3.3.33)
where
IT  ~ IT  _f.1_ IT2 = a .
F  pkv1tn' I  [pvi3t'Y' psvmth
k  1 = 0, l  3k  1 = 0, 2+ n  l = 0.
k = 1, l = 4, n =2;
therefore,
F
IT F = :t2 .
pv t
We have in a similar way for the ITI that
1
180 3 Features of the Solutions of Continuum Mechanics Problems
0:  1 = 0, 1  30: + f3 = 0 , 1 f3 + 'Y = O.
0: = 1, f3 = 2, 'Y = 1;
therefore,
III = fJ/(pv 2t).
We obtain in a similar way that II2 = a/(vt). Substituting the expres
sions for II F , III, and II2 into equation (3.3.33) , we obtain the relation:
Denoting the unknown function IIF by <p, we can rewrite equation (3.3.34)
in the form
F = pv 4 t 2<p (pv2 t  .
  , vt) (3.3.35)
fJ a
The nondimensional variable III = 7 2
is an analog of the Reynolds
number, and II2 = ~ is an analog of the Strouhal number.
In the limit of small velocities v + 0, the force F is determined by the
viscosity fJ and does not depend on the inertial properties p; therefore,
F = F(v, fJ , t, a).
IIF = IIF(II l ),
2
IIF = F/(fJv t), III = vt/a,
(3.3.36)
In the limit of large velocities v + 00, the force F does not depend
on fJ; therefore, we find from equation (3.3.35) that
F = pv4t21jJ(~). (3.3.37)
II1_ pva. _ vt
II 2 
, ,
fJ a
3.3 Discontinuity Surfaces in Continuum Mechanics 181
F pva , vat) .
= pv 2 a 2 t{i2 ( ;;: (3.3.38)
It follows from here that the formulas (3.3.35) and (3.3.38) coincide at
(3.3.39)
AU au
j=j(t, x , u, ax' at)'
where
While moving along the coordinate line a = const, the (3 changes, and
conversely. Let us formulate a Cauchy problem on the coordinate line
a = const for the secondorder equation (3.3.40). For this purpose, we
specify on the line a = const the u((3) and ()~<!). Differentiating these
functions with respect to (3, we can find that , in addition to them, the
va1ues {){3'
{)u
8iJ'i
{)2u
' ()a{){3
{)2u
are speCIfied on t h e gIven
coordmate 1me. We
determine the characteristics from the condition that the given Cauchy
problem for equation (3.3.39) is unsolvable [although the solution u(a, (3)
itself exists]. It follows from the form of (3.3.40) that the coefficient L
affecting the second derivative ~ should vanish for this purpose. Thus,
the lines a(t, x) = const on which
(3.3.41 )
(3.3.42)
( 8a)2
8t
_c2(8a)2
8x
= (8a _c(8a)) (8a +c(8a)) =0,
8t 8x 8t 8x
3.3 Discontinuity Surfaces in Continuum Mechanics 183
which determine two families of lines, along which the disturbances prop
agate at the speed c to the right and to the left.
For the Laplace equation, we set A = 1, B = 0, and C = 1, and the
characteristics obey the equation
it  x = const, it + x = const,
i.e., the Laplace equation has the elliptic type.
For the heat equation, we have A = B = 0, C = 1, and the equation
for the characteristics
(~~)2 = 0
has the solution a = a(t). It follows from here that the characteristics
are represented by the straight lines a(t)
= const parallel with the x
axis. It follows from here that the Cauchy problem for the heat equation
on the line t = 0 is illposed. Only one function is specified as an initial
condition for the heat equation, although the equation itself has the
second order. Such a problem is no Cauchy problem.
Un 11 ii = uJnj , qn = qjnj = O.
Substituting these formulas into (3.3.14), (3.3.18), and (3.3.20), we ob
tain:
184 3 Features of the Solutions of Continuum Mechanics Problems
(3.3.45)
where [<pj = <P2  <Pl , UT is the velocity component parallel with a tangent
plane to the discontinuity surface at point 0, and H = E + P/ p is the
enthalpy (2.1.87).
Let us find the entropy jump in a weak shock wave. For this purpose,
we combine the relations (3.3.45) with each other and rewrite them in a
slightly different form
where j = PUn is the mass flux, and V = 1/ P is the specific volume. The
last equation in (3.3.46) is called the Hugoniot adiabat in the literature.
Since H = H(P, S), the difference H2  Hl will depend on S2  Sl and
(P2  Pd Expanding H2  Hl into a Taylor series in (S2  Sd up to
the firstorder terms, and in (P2  P l ) up to the thirdorder terms, we
obtain the formula:
All derivatives are taken in (3.3.47) at point Sl , Pl. Using (2.1.87), let
us rewrite (3.3.47) in the form
H2  Hl (8V)
T l (S2  Sl) + Vl (P2  Pl ) +"21 8Pl s(P2  Pl )2
+ ~(82V)
6 8Pr s
(P2 P
l )3. (3.3.48)
Since the product (Vl + V2)(P2  P l ) in the Hugoniot adiabat [the last
equation in (3.3.46)] already has the first order of smallness with respect
to (P2  Pd, the specific volume V2 in this expression should be expanded
up to the secondorder terms with respect to (P2  Pl ):
(3.3.49)
3.3 Discontinuity Surfaces in Continuum Mechanics 185
Substituting (3.3.48) and (3.3.49) into the equation for the Hugoniot
adiabat, we obtain:
(3.3.50)
For the normal gases (~)s > 0; therefore, the condition S2 > Sl
will be satisfied if P2 > P 1 , which corresponds to a compression shock
wave. At the phase transitions, the regions can arise, where (~) S < 0;
therefore, the entropy will increase (S2 > Sd in the rarefaction shock
waves, where P 2 < Pl .
References
This chapter deals with incompressible ideal fluid flows. While choosing
the material, the authors aimed at presenting those reults that have now
become the classical results and are widely used in the current research
work of the aerohydrodynamicists l  9 . In particular, the Bernoulli and
Lagrange integrals are derived in Section 4.1. They enable one to find
the pressure distribution in the fluid from a given velocity field.
In Section 4.2, we consider planar irrotational fluid motions. The
method of the theory of functions of a complex variable is presented in
detail. This has enabled one to solve a variety of problems on the fluid
flow around sufficiently arbitrary planar profiles. The computation of a
lift force of a wing, which is known in the literature as the Joukowskii
force, should be considered as the most important result of this method.
In sections 4.3 and 4.4, we study the threedimensional fluid flows.
We demonstrate here with the aid of Mathematica programs the effi
ciency of the method for computation of threedimensional flows around
the bodies, which is based on a substitution of the body surface by the
distributed sources and sinks or dipoles.
In Section 4.5, we consider the general properties of vortex flows and
derive the BiotSavart formula, which enables one to find the fluid flow
velocity from a given distribution of vortices.
C2
2
V' + P + II) = v x rot v,
J
where
P= d; .
Let us multiply the left and righthand sides of this equation by vas in
a scalar product:
C2
2
V' + P + II) . v = (v x rot v) . 71.
Since (v x rot v) . 71= 0, we have:
:l C; + P+ II) = 0,
where v V' = 1ft is a derivative along the streamline. Integrating the
last equation along a streamline, we obtain:
v2
"2 + P + II = C(l). (4.1.2)
4.1 Integrals of Motion Equations of Ideal Fluid and Gas 189
at + \7 (V22 ) + \7
aV' (/ dP) \7 _
dp + . II  0.
\
\ \ \
\ \ \
\ \ \
\ \ \
\ \ \
\ \
"" \ \
\
""""" \
\
"~~
~~Z:Z:::2Z:Z:Z::Z::::2Z:Z:~2Zl B PA
+II+
v
2
2
JdP
=C.
P
The constant C is here the same for the overall flow in contrast with
the Bernoulli integral (4.1.2), in which the constant C(l) is different on
different streamlines.
The Bernoulli and Lagrange integrals play an important role in fluid
and gas mechanics. They enable one to find the pressure distribution on
the basis of the known velocity modulus. The velocity v can be found
from the continuity and momentum equations. In a particular case of
a potential flow v = Vcp, and upon substitution into the continuity
equation at P = Po we obtain the equation t::.cp = O. This circumstance
is illustrated in the following on a number of simple problems.
us write the Bernoulli integral along the streamline AB, where the point
A is located on the surface So and B is located on the surface S1; that
is,
V~
+ PA +goH= v~ +, PB
2 Po 2 Po
where Po is the fluid density. Since the pressure at points A and B is
equal to the atmospheric pressure P A = PB = Pa , the Bernoulli integral
simplifies greatly:
V2 v2
=i + goH = J!...
2 2
We find from the equation for the constancy of flow rates VA SO = VB S1
and the last relationship that
V~
 + 'Y P=
A
V~ PB
+  'Y   .
2 'Y  1 PA 2 'Y  1 PB
Since the vessel is large and the orifice is small, VA VB, and one can
assume that VA = 0, PA = Po, and PA = Po. This enables us to find
from the Bernoulli integral:
v~ _ _ 'Y_(PO _ PB)
2  'Y  1 Po PB
Since the flow is adiabatic, we have:
= Po ( Po '
PB)~
PB
192 4 Ideal Fluid
Po Po To
A.
B
Figure 4.2: The adiabatic gas outflow from a vessel through a small
orifice.
Consider the physical meaning of this formula. Introduce the flow rate
per unit area Q = PBVB. We find from formula (4.1.6):
Q = V . PapaCy (1 
2"(
"(1
2
~ ::t=!
"( ),
where { = PBI Po, and P B is the pressure of a medium into which the gas
escapes. At ~ = I, i.e., PB = Po, we have an equilibrium; consequently,
the gas does not escape from the vessel. At the diminution of PB, i.e.,
at the diminution of ~, the flow rate increases and reaches its maximum
at certain ~ = ~*. At a further diminution of ~, the value of Q reduces
and vanishes at ~ = O. It should be noted, however, that the experiment
confirms the validity of the dependence Q(O only for ~ > ~*' At ~ < ~*
Q indeed remains constant and is equal to its maximum value, and the
outflow velocity is equal to the sound velocity. A further diminution of
PB already does not affect the outflow from the orifice, since the distur
bances from the ambient medium do not penetrate the vessel interior.
At ~ < ~*' when the jet becomes supersonic, the assumption on the one
dimensional character of the flow proves to be invalid, and it is necessary
to take into account the multidimensional character of the flow.
4.2 Planar Irrotational Steady Motions of Ideal Fluid 193
(4.2.1)
where II is the potential of body forces. Here the first equation is the con
tinuity equation, and instead of the vector equation for the momentum,
we have written the EulerBernoulli integral and the condition of the
v
curl absence rot = 0 for a planar motion. As was shown in Section 4.1,
194 4 Ideal Fluid
such a form of equations is possible, since the passage from one system
to another one was carried out with the aid of identical transformations.
The energy equation for an incompressible fluid in the absence of
heat supply yields
dE =0.
dt '
that is, the energy is conserved in a particle for an incompressible fluid.
The first two equations (4.2.1) contain only the functions V x and v y ,
and the last equation can be used for finding the pressure on the basis
of the known values of V x and v y . Thus, the system (4.2.1) completely
describes the flow.
The condition for the curl absence enables us to introduce the velocity
potential cp by the formulas
dy
'lj!B
dx
Comparing this relation with the definition d'lj! = vxdy  vydx, we can
see that d'lj! = 0 along the streamline; that is, 'lj! = const along the
streamline. It is conventional to call the function 'lj!( x, y) the stream
function and the equation 'lj!(x, y) = const the equation of a streamline.
The different values of a constant correspond to different streamlines.
One can compute in terms of the stream function the flow rate across a
curve intersecting different streamlines (see Fig. 4.3). Let us calculate
the flow rate per unit time across the curve AB (a piece of a cylindrical
surface with the height .6.z = 1 with the generatrix AB) by the formula
Q = LA Vn ds = LA (vxnx + Vyny) ds ,
av y _ avx = 0.
ax ay
Substituting into this relation the expressions for the velocity compo
nents Vx = ~ and Vy = ~, we obtain:
(4.2.4)
When solving the problem on a flow around the bodies in terms of the
stream function, it is necessary to have the boundary conditions written
also for this function. Consider the fluid flow around an impermeable
body surface S. It is known that ifn = for an ideal fluid on this surface.
Let us write this condition in terms of the stream function '1f;, i.e.,
w = cp(x, y) + i'1f;(x, y)
4.2 Planar Irrotational Steady Motions of Ideal Fluid 197
Problem 4.3. Prove that, if the functions <p and 'ljJ satsfy the rela
tions (4.2.5), then the function w(x, y) = w(z).
Solution: Let w(x, y) = <p(x, y) + i'ljJ(x, y) and introduce z = x+
iy, Z = x  iy. By definition,
z+Z zz _
w(x, y) = w(2' 2i) = w(z, z).
We have that
OW a<p dx a<p dy .a'ljJ dx .a'ljJ dy
 = .  +  .  +z  +z.
az ax dz ay dz ax dz ay dz
Since
1
Y = 2i (z  z),
we have that
dx 1 dy i
dz 2' dz 2
Substituting these values into the expression for ~~, we find:
'IjJ = const
y
~
that is, the real part determines the velocity circulation, and the imag
inary part determines the volume flow rate of fluid per second. Thus,
the introduction of a complex potential enables one to apply a well
developed apparatus of the theory of the functions of complex variable
for finding the solutions of many boundaryvalue problems on the steady
planar potential flows of an ideal incompressible fluid. As already stated
4.2 Planar Irrotational Steady Motions of Ideal Fluid 199
w = .!L lnz,
27r
where z = re i8 , r = Izl, and 0 = argz. We find that i.p + i'ljJ = (In r +
iO) . .!L
211"
or
'l/; = qO.
i.p = .!L In r ,
27r 27r
The streamlines will be the beams emanating from the coordinate origin.
The equipotential lines are the circles r = const (see Fig. 4.5 (a) and
200 4 Ideal Fluid
y
'IjJ = const
<p = const
(b)). For the complex potential w(z) = ~lnz the pojections of the
velocities on the axes of polar coordinates will be Vr = ~2 1r r Ve = O.
.!,
It can be seen from here that the velocity has a constant magnitude on
each circle with a fixed radius and with a center at the coordinate origin;
the velocity vector is directed along the radius and reduces as r 1 when
r increases. At q > 0 the velocity is directed away from the coordinate
origin [see Fig. 4.5 (b)], and at q < 0 the velocity vector is directed to
the coordinate origin and Vr < 0 [see Fig. 4.5 (a)].
Let us compute the flow rate of a fluid across a contour, enclosing
the coordinate origin by using the formula Q = 1m f V dz = f d'IjJ = q.
Thus, q is the source strength. At q > 0, we have a source, and at
q < 0, we have a sink. If the source is located at point z = a rather than
in the coordinate origin, then the complex potential will have the form
w(z) = ~ln(z  a).
Let a source of strength q be located at point A of the (x, y) plane,
let a sink of strength q be located at point B, and let the complex
coordinates of points A and B be ZA = 4eiO:,
ZB = 4eiO:.
The complex
potential of the flow from these two sources has the form
WA ()
Z
q In (z  2
= 271" l eio:)
, WB Z
()
= q I ( l io:)
271" n z + 2e ,
\ <p = const
Let l ; 0 and the intensity q ; 00 , so that the product l q = M remains
constant. Then, the complex potential will have, for such a limiting flow ,
the form
M eio:
w(z) =   .  . (4.2.7)
27r z
Let us study the flow pattern determined by the complex potential
(4.2.7). For the sake of simplicity, we assume a = 0; i.e., the dipole
is located at the coordinate origin and its axis coincides with the xaxis.
In this case, the functions <p and 'IjJ are presented by formulas
M x
<p =   . ,
27r x2 + y2
The streamlines 'IjJ = const are the lines on which
y 1 'IjJ'7r
c=
2c' M
and they represent the circles passing through the coordinate origin and
the centers are on the yaxis (see Fig. 4.6) , where the equipotential lines
<p = const are shown by the dashed lines. It is conventional to call the
flow described by such a complex potential the dipole.
Consider a complex potential of the form
r
w( z) = 2.lnz. (4.2.8)
7rZ
The streamlines 1jJ = const are the circles with a c enter at the coordinate
origin, and the lines t.p = const are the beams e = const (the dashed lines
in Fig. 4.7). The coordinate origin r = is a singular point, since
at.p 1 at.p r 1
Vr = or = 0, Ve =   = '.
ae
r 27r r
The velocity Ve > at r > 0; that is, the motion along a circle in a
counterclockwise direction corresponds to a positive circulation value.
Let us take a contour l, comprising the coordinate origin, and compute
the velocity circulation over this contour by the formula
Re i i V dz = dt.p = r.
Thus, r is the velocity circulation over a closed contour enclosing the
coordinate origin.
points of the (x, y) plane besides the circle with the radius R. It should
be singlevalued, bounded, and take a given value at the infinity. It is
known from the theory of the functions of complex variable that such a
function can be presented in the form of the Laurent series in negative
powers of z:
 C CI C2
V(z) = Co + z +z2 +z2 +"', (4.2.9)
dw  .

dz
= V(z) = Vx  tv
y
+ Cz + CI
Z2
+....
Integrating with respect to z, we obtain:
00
~CP I =
ur r=R
Ux cos B + u y sin B.
This relation serves for finding c, CI, ... , Cn. Going over to the polar
coordinates in w(z), subdividing into cp and 'lj; and differentiating cp with
respect to r, we substitute the obtained expression into the boundary
condition and determine C and Cn. This yields
+ 2: (An Bn) ,
cosne+ sinne
00
rn rn
n=2
204 4 Ideal Fluid
that is
This is the general form of a complex potential for the flow past the
cylinder, where there are the following complex potentials:
1) ifooZ coresponds to the translational flow;
2) (VOO  U)~2 corresponds to the dipole;
3) 2~i lnz corresponds to a point vortex.
Let us write the complex potential in the following particular cases:
1) The cylinder is at rest, i.e. , u == o. Then
_ R2 r
w(z) = Vooz + Voo  + .ln z. (4.2.11)
z 2~z
3) The fluid at infinity is at rest and the cylinder is also at rest. Then
r
w(z) = lnz
2~i '
this is the circulatory flow past the cylinder.
Let us study in detail the flow pattern for a stationary flow of a
cylinder, which is at rest, that is, Case 1 when u = 0 and the flow at
infinity is directed along the xaxis. The complex potential at Vx =
v oo , Vy = 0 takes the form
R2 r
w(z) = voo(z + ~) + 2~i lnz.
4.2 Planar Irrotational Steady Motions of Ideal Fluid 205
Ol    ....:....:..j ..o;.....j''+='1 X
2
4 2 o 2 4 6
or
iy )
cp + 2'ljJ x + 2y + R .
.
= V(X) ( . 2 X 
2 2'
X +y
From here, we find:
R2 R2
cp = v(x)x (1 + x 2+ y2) , 'ljJ =v(X) y(l 2 2)'
x +y
(4.2.12)
which are the thirdorder curves symmetric with respect to the yaxis.
The lines 'ljJ = Cl and 'ljJ = Cl are symmetric with respect to the xaxis.
At 'ljJ = 0, the equation for the streamline is split into two equations:
y = 0 is the xaxis and x 2 + y2 = R2 is the circle. The flow pattern past
the cylinder is presented in Fig. 4.9. We have obtained Fig. 4.9 with the
aid of the Mathematica Notebook prog41.nb (see also Appendix B).
Consider the field of the velocity vector. Let us go over in (4.2.12)
to the polar coordinates (r, e) by the formulas x = rcose, y = rsine.
Then
206 4 Ideal Fluid
or
o<.p
= Voo cos e(R2)
1 ~ , (4.2.13)
;;:10<.p . (
oe = Voo sm R2) .
e 1+ ~
The formulas (4.2.13) yield the velocity components at any point of the
flow. Assuming r = R in (4.2.13) , we obtain the magnitude of the
velocity on the surface:
Vr = 0, Ve = 2voo sine.

V(z) = 
dw
= Voo
(R2)
1  + r 1
. ..
dz z2 27rz Z
Let us find the critical points of the flow, where Vx = 0, Vy = 0; that is,
Voo z
2 r .z 
+ 2 vooR = 0
2
7rZ
or
(4.2.15)
y y
2 
0 x
Ihl+l x
1
3
4 2 0 2 4 4 2 o 2 4
(a) (b)
Figure 4.10: The flow patterns around the cylinder at (a) r I 0 and
r and at (b) r .r. /. 0 and 4v 00
2 R2 > 47i7
2
4v 00 2 R2 = r2 .
47i7
2)  L,22 + 4v~R2 = O. The critical points merge into one point located
on the imaginary axis: IZ121 ,
= R, Zl = Z2 = 1rVoo
r4 . i [see Fig. 4.10
(b)]. In order to obtain Fig. 4.10 (b) with the aid of Mathematica, it is
sufficient to replace the value g=l in the above program for Fig. 4.10 (a)
with the value g = 2.
In the case of the circulatory flow past the cylinder, the stream
lines are symmetric with respect to the yaxis. The pressures at the
cylinder points symmetrical with respect to the yaxis have equal mag
nitudes. There is already here no flow symmetry with respect to the
xaxis. Therefore, a force acting on the cylinder in the direction of the
yaxis arises, and the force in the direction of the xaxis is equal to
zero as in the flow without circulation; that is, the D'alembert's paradox
takes place here also. Thus, in the presence of circulation, different flow
patterns can take place and, therefore, it is necessary for the solution
uniqueness to specify the circulation magnitude or some additional con
ditions for its determination. This fact is substantial while solving many
practical problems, which we will show later. Thus, it follows from the
considered examples that, if the complex potential is given, the physical
flow pattern is determined easily.
208 4 Ideal Fluid
4 2 o 2 4
Figure 4.11: The flow pattern around the cylinder at r i= 0 and 4v~R2 <
r2
~.
df I = dz I =k>0
d( (>00 d( (>00 '
iy
0
@ ill
CD
x
n
~ 2lt~
(a) (b)
Figure 4.12: The (a) physical and (b) auxiliary planes in the problem of
the flow around the contour.
take place at the corresponding points of the zplane and of the (plane.
The function w(z) is a complex potential of the flow past a contour I
at rest in the zplane. Therefore, the function 'lj;(x, y) is constant on
I. The circle I' in the (plane corresponds to the contour I, but since
'lj;(x, y) = W(~, 1]), the function w(~, 1]) will also be constant on I'; that
is, the circle is a streamline of the flow whose complex potential is w(().
Let us find the conditions at infinity for this flow. The complex velocity
is
V(z) = dw = dw . dz = V(z) . dz.
dz dz d( d(
At point z > 00, the value Voo is known, consequently
( dw)
d( 00
=V I<: ~oo
= kV I z~oo

= kVoo .
Thus, the w(() determines in (plane a flow outside the circle with the
velocity at infinity kVoo , which corresponds to formula (4.2.11):
 kVooR2 r
w(() = kVoo( + ( +27riln(.
210 4 Ideal Fluid
ill
CD
iy
0
0
x
0
~
Q
21( 0
(a) (b)
Figure 4.13: A profile with (a) one corner point and (b) its mapped form
in the auxiliary plane (.
Formula (4.2.16) gives the solution of the problem of the potential flow
past an arbitrary contour, if the conformal mapping of the region outside
[ onto the exterior of the circle is known; that is, when the function
( = F(z) is known. The value k is found by the formula
The function z = f( () transforms the angle 1T into the angle 21T 15. This
points to the fact that the transformation conformity is violated, and in
4.2 Planar Irrotational Steady Motions of Ideal Fluid 211
the vicinity of point A, the function z(() should have the expansion of
the form
+ ....
27l"tS
Z  ZA = M((  (A')"
i.e., at ( < 7r at point A' ~( lA' = 0; consequently, if d~2C,) IA' =f:. 0, then
VA > 00, which is physically inadmissible.
The requirement that the velocity VA at the sharp trailing edge be
finite makes the contents of the ChaplyginJoukowskiiKutta postula
te 1 ,4, 8, 9,10,1l. The satisfaction of this condition is possible if ~( lA' = 0;
i.e., A' is a critical point of the flow past a cylinder, and as is already
known it depends on the magnitude of the circulation r. From here,
it follows the second formulation of the Chaplygin Joukowskii Kutta
postulate: the circulation in the flow past the profile with a sharp trailing
edge A is such that the point of a circle onto which A is mapped at the
transformation should be a critical point in the flow past the cylinder.
This postulate enables one to determine the value of r. So we have:
where the complex velocity is expressed by the formula ~( Ic,>oo = kVoo '
Let the flow impinging on the profile have the inclination angle 0: to
the xaxis; that is, Voo = Ivoole ia , Voo = Ivoole ia . Let us compute the
quantity
dWI = kVoo _ kVooR2 + ~. _1_ = 0
d( A' (~, 27ri (A' .
We find from here that
Taking into account that (A' = Rei()o and V00 = IVoo Ieia , where eo is the
angle determining the position of point A' on the circle, we obtain:
or
r = 47rkRlv oo l sin(eo  0:) . (4.2.17)
It is conventional to call the angle (0:  eo) the angle of attack. It is seen
that, if eo  0: = 0 then r = 0 automatically. This enables one to ensure
212 4 Ideal Fluid
F=i p ndl ,
where I is the body contour and n is the vector of the normal to it. Let
us rewrite this formula in the projections onto the coordinate axes:
Along the contour I, the Bernoulli integral is valid, which may be written
in the absence of the body forces as
v2
P=poCp02" (4.2.19)
R = 2
.Po
2
iI
V 2d
z. (4.2.20)
Consider the element of the body contour dl and denote by () the angle
between the tangent to the contour l and the xaxis. Then
R =
.Po
22' i
IV
2  2i8 d
e z.
4.2 Planar Irrotational Steady Motions of Ideal Fluid 213
At the contour l points, the velocity is directed along the tangent; that
is, ve iO = v cos e  iv sin e = Vx  ivy = V, which enables us to write
that
1
R = i Po V2 dz.
2 ft
Taking into account the fact that V = ~~ , we have:

R
. .Po
= Fx  zFy = z2 ft dz
1(dw)2 dz. (4.2.21 )
Taking into account the fact that ve iO = V and dE = e 2iO dz, we can
rewrite the second formula of ChaplyginBlasius in the form
L
Po
= Re (  2 ft1(dW)2 )
dz z dz . (4.2.23)
214 4 Ideal Fluid
R  F x 'F  ipo
Z Y 
2
i( I
dW)
d 2 dz.
Z
From the theory of functions of complex variable, we have for the region
outside the profile I the expansion
 dw Al A2
V (z) = dz = Ao + ; + ~ + ....
Let us find the coefficients Ao and Al from the condition that V =V 00
Jdw
Jz dz dz = 27rA I
The integration of both sides of this equation and the use of the residue
f (~:f
theorem yields
dz = 2Voor.
(4.2.24)
Using the expansion into the Laurent series of ~~ in a region outside the
body, we find:
(dW)
z dz 2 r + ( 2A2 V;  r 2 ) 1 + ...
= z V002 + 2V;00 27fi 00 47f2 z .
216 4 Ideal Fluid
Substituting this expression into (4.2.24) and using the residue theorem,
we get:
L = Re [ ~o 21Ti ( 2A2 V00  4~2) ]
or
Thus, if the complex velocity expansion into the Laurent series is known,
i.e., A2 is known, then the magnitude of the moment can easily be com
puted. It is often convenient to use the expansion of the mapping func
tion z = f(() in the neighborhood of a point at infinity
+ ko + <: + (2 + ....
kl k2
z = k(
Let us calculate
2 2
k V00 +2kV00 
 r 1
21Ti' (
r2 + 2k 2Voo VooR
(41T2  2) (21 + "' ,
d( 1 ko 2kl 1
z z ~( =(+T+T''(+''';
dz
Z(()(dW)2. d(
d( dz
Cl( + Co + [2koVoo 2~i + 2k kl vc!
r2 + 2k2V VooR2)] '(1 + (2C2 + ....
(41T2 00
Substituting the obtained formulas into (4.2.25) and applying the residue
theorem, we find:
(4.2.26)
iy
o i'T/
A x
a o +a
(a) (b)
Figure 4.14: The (a) physical and (b) auxiliary planes in the problem of
the flow around the plate.
Figure 4.15: The streamlines in the incompressible fluid flow around the
plate at 0: = 7r / 4.
and
ko = 0, Ro = 1.
Therefore,
+ ~ln(z+Jz2a2),
27rz a
(4.2.29)
Thus, having the complex potential (4.2.29) and (4.2.30), we can find
the complex velocity V and its components Vx and Vy in the region,
including also the plate points. We have made the Mathematica program
prog44 . nb, which solves the problem of the incompressible fluid flow
around a plate. We show in Fig. 4.15 a picture of the streamlines ob
tained with the aid of this program.
Let us determine the main vector of the pressure forces acting on the
plate with the aid of the Joukowskii theorem:
 ~ X
1 2a ~
(4.2.31)
Cp = 1 (1
"2PO Voo S
2 = 27rsina,
since sin a ~ a. Taking into account the expression for the moment L
(4.2.26), one can write that
L= ~cosaF.
2
The experimental data show that the obtained results can be used as
the solution of a problem of the attached flow around slender profiles at
small angles of attack. The profile is assumed to be slender if the ratio of
the profile thickness to its chord length is much smaller than unity, and
the need in an attached flow requires a smallness of the angle between
the tangent direction at any point of the profile and the chord as well as
a small angle of attack.
IF1(x)
~
I,1 I F2(X)
~
I,1 I dFl(X)
~
I ,1 I dF2(X)
~
I.1 (4.2.32)
It should be noted that the limitations (4.2.31) for the subsonic profiles
may not be satisfied in the vicinity of a rounded leading edge. In this
region, one must use with extreme care the solutions that we construct
below.
Introduce a coordinate system x*Oy* fixed in a profile by directing
the x* axis along the profile chord (a, a). The angle between the
velocity Voo direction of the Oxaxis and the chord (the Ox* axis) is the
angle of attack a. Let y; = Ft(x*), Yb = Fb(X*) be the profile equations
in this coordinate system. The relation between the coordinates (x, y)
and (x*, y*) is given by the formulas
x* =xcosexysinex, y* =xsinex+ycosex.
x* = x, y* = xex + y, (4.2.33)
xex + Yt = Ft(x),
or
(4.2.34)
The problem of the flow around the profile will be solved if we find a
function w(z) satisfying the conditions at infinity, the conditions for the
flow around the profile, and the ChaplyginJoukowskii postulate. Let us
present the complex potential w(z) in the form
dWI
dz = Voo ,
00
dW'
dz
1 _ o.
00
dW'
d(
1 _0
00 
(4.2.38)
222 4 Ideal Fluid
and corresponding to the noslip condition on the unit circle. Let us find
this condition. Set ( = pe iO , and introduce the functions 11>' (p, B) and
w' (p, B) such that
w' (() = 11>' (p, B) + iW' (p, B).
Taking into account the formula 1jJ' (x, y) = w' (p, B), we have on the circle
p = 1 that
We will search for the function w' (() given in the exterior of the circle
p = 1 and satisfying the condition (4.2.38) in the form of the series
r
L rn '
00 Cn
w'(() = ~ln(+ (4.2.39)
7r~ n=O"
where Cn = an +ibn are the unknown constants to be determined. Deter
mining the real and imaginary parts in (4.2.39), we find the expressions:
r 00 1
11>' (p, B) B+ ""' (ancosnB+bnsinn61),
27r ~ pn
n=O
r
W'(p, B) = lnp +
27r
00
L
1
n=O pn
(an sin nB + bn cosnB).
We now expand the known function f(B) into the Fourier series:
00
f(61) = L(ancosnB+,Bnsinn61)
n=O
4.3 ThreeDimensional Potential Ideal Fluid Flows 223
L(bn cos nO  an sin nO) = aVoo [a cos 0  L(an cos nO + f3n sin nO)].
n=O n=O
Comparing the coefficients at equal sin nO and cos nO, we find:
an Vooaf3n , n::::: 1,
bo aVooao , b1 = aVoo(a  al), bn = aVooa n , n ::::: 2.
r = 27r L nb n
n=O
Thus, we have determined in the complex potential w' (() all coefficients
of the series, which enables us to perform a complete construction of the
problem solution by the presented algorithm.
1 ocp
v.x=,
rsine OA
The continuity equation for an incompressible fluid in the spherical co
ordinates follows from formula (2.1.144) if we assume = 0: Wi
Let us make use of the condition that cp = cp(r). Then we find from the
continuity equation:
q
q = 41fr 2 Vr = 41fC or C = 41f'
This enables one to write down the velocity potential in the case of a
flow from a source placed at the coordinate origin in the form
4.3 ThreeDimensional Potential Ideal Fluid Flows 225
V
where r = x 2 + y2 + Z2. If the source is placed not in the coordinate
origin but in a fixed point with the coordinates x = a, y = b, and z = c,
then
q
r.p= (4.3.1)
47rV(x  a)2 + (y  b)2 + (z  C)2
Consider a flow from a source and sink located at a distance l from one
another and having the same strengths q but having the opposite signs.
Since the equations governing the flows are linear, the sum of the two
solutions is also the solution r.p = r.pl + r.p2, where
q 1
r.pl =
47r Jx
,:.======
2 + y2 + (z  ~)2
is a potential corresponding to the source and
If the dipole axis l does not coincide with the coordinate axis, then the
potential of a flow from the dipole will in a general case have the form
r r
Figure 4.17: The flow around a sphere moving at a velocity illl voo.
o<P1 = 0, o<p I = 0,
ax 00 oy 00
a ( 2 . o<p ) a (. o<p )
or r smB or + oB smB oB = 0. (4.3.5)
~<PI
un r=Ro
= ucosB, Vrl r>oo = ~<PI
ur r>oo
= voocosB,
a<PI
~ I = voocosB, _1 . a<PI
!:IB I .
= vooslnB,
uT r+oo r U r+oo
or
<P = vooz +"21 (R)3
:;: (voo  u) z.
The first item is the potential of a planeparallel flow at a velocity
v oo , and the second item is the potential of a dipole with the moment
M = 27rR3 (u  v oo ).
Vrl r=Ro = 0,
The maximum value of the velocity magnitude on the sphere surface is
equal to ~voo and is reached at the points B = ~ . It should be noted
that for the case of the flow past a cylinder the velocity maximum on
the surface is equal to 2voo .
We have from the Bernoulli integral on the sphere surface that
where the magnitude of a constant was found from the condition at in
finity. It follows from the symmetry of the pressure distribution that
the main vector of all pressure forces is equal to zero; that is, the
D'Alembert's paradox takes place.
Following 1,9, we now consider the axially symmetric flows in the
cylindrical coordinates, where the zaxis is taken as the symmetry axis.
All hydrodynamic quantities do not depend in this case on 'P. Therefore,
the continuity equation (2.1.139) in the case Po = const can be written
in the form
o 0
or (rvr ) + oz (rvz ) = O.
If we introduce the function 'Ij; by the formulas
1o'lj;
  ror
v z ' (4.3.10)
and by the definition of the streamline, the function 'Ij; will be constant on
a streamline. Thus, we will call the function 'Ij;, introduced by formulas
(4.3.10) , the stream function.
4.3 ThreeDimensional Potential Ideal Fluid Flows 229
If the flow is potential, then a velocity potential exists, which for the
flow with the axial symmetry in the cylindrical coordinates is related to
the velocity components via the formulas
ocp
Vz = oz
or (4.3.11)
cp = cp(ro, ZO) + 1
I
1 (O'ljJ O'ljJ)
~ oz dr  or dz .
Let us construct the stream functions for some specific simplest flows.
1) The translational flow cp = VooZ. We find by formula (4.3.11) that
2
If the flow axis r = 0 is the streamline 'ljJ = 0, then c = 0 and 'ljJ = V oo r2 .
ocp q r ocp
or = 41T . (Vr2 + Z2)3 ' OZ
Mz M 8 ( 1 )
<p =  41Tr3 = 41T . 8z Jr2 + Z2 .
7j; = M r8 (
4 1 ) + f(r).
1T 8 r v'r2+z2
Computing the quantity ~ from this formula and comparing it with the
expression for the ~ found from the condition ~ = we obtain r%;,
1r = 0; that is, f = const . Consequently, the stream function describing
the flow from a dipole has the form
M r2
7j; =   . + c.
41T (Jr2 + z2) 3
In the general case of the potential axially symmetric flows of an ideal
incompressible fluid, the formulation of problems on the flows around
the bodies in terms of <p reduces to the solution of the equation
8<p I = 0, 8<P1
8 = 0 and 8<P1 _ V00,

8n S r r~oo 8z z ...... oo
z
B
A
The obtained equation differs from the Laplace equation by the item
 ~ ~; therefore, the welldeveloped methods of the theory of functions
of complex variable for the given class of problems will already be inap
plicable.
Jar
B
1 ( z ( ) (4.3.13)
"pI =  4n JJ(() 1  Jr 2 + (z _ ()2 d(.
232 4 Ideal Fluid
Let us present the overall flow around the body of revolution in the form
of a sum of two flows : the translational flow 'ljJ2 = r2lf and the flow
determined by 'ljJl; that is
'IjJ = r 2 Voo   1
2 41l'
1B (
A
J.L( () 1  Z  (
Jr 2 + (z  ()2
)
d(. (4.3.14)
that is, the total strength of the sources (sinks) located inside the body
should be equal to zero. Under satisfaction of this condition, equation
(4.3.14) has the form
We shall assume that r = r(z) is the equation of the body contour and,
consequently, 'IjJ = 0 on the contour. Taking the boundary condition into
account , we find :
(4.3.15)
where the J.L((i) at points (i (i = 1,2, ... , n) are the unknown quantities.
Replace the integral in (4.3.15) with this sum, and require that the
obtained equation be satisfied at points Zk belonging to the interval 6k,
where k = 1,2, . .. , n . In this way, we obtain a system of linear algebraic
equations for J.L((i) in the form:
4.3 ThreeDimensional Potential Ideal Fluid Flows 233
'ljJ( z, x) =
2
+ ~ (_ 0.00136163(0.958333 + z)
47r Jx 2 + (0.958333 + z)2
0.00299313( 0.875 + z) 0.00431155( 0.791667 + z)
Jx + (0.875 + Z)2 Jx + (0.791667 + Z)2
2 2
0.4
0.2
o
0.2
0 . 4
1 0.5 o 0.5 1 1.5 2
=z
A~
B
the choice of the singularities replacing the body. In a general case, other
methods for the solution of equation (4.3.15) also exist.
All dipoles located in the interval AB form a flow with the velocity
potential
1 r
B fL(()xd(
'P1 =  47r JA (Jx + y2 + (z  ()2)3
2
or in the cylindrical coordinates
rcosB r
B fL(() d(
'P1 = ~ JA (Jr 2 + (z  ()2)3
Let us present the flow near the body in the form of the sum
where x = r cos O. For the determination of fL( (), we use the boundary
condition on the body surface, since the condition at infinity is satisfied
automatically at such a choice of the velocity potential. Let us write
the equations for the streamlines in cylindrical coordinates (to avoid
confusion with the potential 'P, we denote here the polar angle by the
letter 0) :
dr dz rdO
Vr Vz Vo
and find the expressions for Vr , Vz , and Vo by using (4.3.16) in the form
Vr = a'P =
ar
Voo cos 0 _ cos 0 ~ [r
47l' ar
jB (Jr + (z 
A 2
fL( () d( 1
()2)3 '
Vo
~ a'P
r aO
= Voo sin 0 + sin 0
47l'
jBA (y'r2 + (z 
fL( () d(
()2)3
dr
dz = f(r, z). (4.3.17)
Since the equation of the body is given, r = <I>(z), then ~: = <I>'(z) will
be a known function. Taking this condition, as well as (4.3.17), into
account, we find:
(4.3.18)
Thus, we have reduced the problem of the flow past a body to the solution
of an integral equation. This equation can in practice be easily reduced
to a system of linear algebraic equations, as this has been done in the
foregoing case.
Let us now describe the computer implementation of the above me
thod of singularities. Since the computational results in the problems
of threedimensional flow around a body are usually presented in the
Cartesian coordinates x, y, z, it is desirable to go over to these coordi
nates from the cylindrical coordinates r, 0, z. For this purpose, we must
find the Cartesian velocity components u x , u y, U z of the velocity vector v
from the components V r , Vo, V z in cylindrical coordinates. Let us denote
4.3 ThreeDimensional Potential Ideal Fluid Flows 237
the basis vectors of the cylindrical coordinate system by el , e2, e3' Then
if = vj~, where vI = VTl v 2 = Vii, and v 3 = Vz . While solving Problem
1.2 (see Section 1.1), we have found that
+ v 2 [sin()(EIEd+cos()(E
...,  
2 E I )]
VI cos ()  v 2 sin () = cos () Vr  sin () Vii.
Since the streamlines at infinity coincide with the free stream, it is clear
that Idxjdzl > 00 as Ixl > 00. Therefore, the above ODE cannot be
used directly for numerical integration. In this connection, we use a para
metric representation for each streamline: x = x(t) , y = y(t), z =
z(t) , where t is a parameter, which changes along the streamline. Then
we can write the following ODEs for the streamline:
dx
ux(x(t), y(t), z(t));
dt
dy
uy(x(t), y(t) , z(t)) ; (4.3.19)
dt
dz
uAx(t), y(t), z(t)) .
dt
The integral equation (4.3.18) for fL can easily be rewritten in the form
3<p(z)<p'(z ) {B (z  ()fL(() d(
47r JA (J<p(z)2+(z()2)3
238 4 Ideal Fluid
(4.3.20)
Zk = kh , k = 1, ... , n
i = 1, .. . ,n; k = 1, .. . ,n.
We have used the builtin Mathematica function LinearSolve [A, b] for
the numerical solution of this system of equations. This enables us to
write the approximations of the velocity components V r , Vo, and V z by
the method of singularities. For example,
Since the righthand sides of the ODEs (4.3.19) are very complex,
these equations cannot be integrated in analytic form . Therefore, one
must apply some numerical method to solve these equations. We have
used the classical fourthorder RungeKutta method 13 for the numerical
integration of the system (4.3.19). The integration step tlt = 0.02 along
the taxis proved to be sufficient to obtain a good accuracy of the nu
merical results. For each streamline we have specified the Cauchy data,
i.e., the initial points (xo , Yo , zo) of a streamline:
x (O) = Xo, y(O) = Yo , z(O) = zoo
These initial points were chosen for each streamline at a sufficiently
large distance from the body, where the flow differs little from the free
stream. We have implemented the abovepresented variant ofthe method
of sources and sinks for threedimensional problems in the Mathematica
Notebook prog46. nb. In what follows, we present the following output
of this program:
1) the numerical values of the source strengths p((i);
2) the approximate analytic expression for the velocity component V z
obtained by the method of sources and sinks.
~( ( 0.000970804(0.975 + z)
uz(x, y, z) =
4 X 5/2
Jr (x2 + y2 + (0.975 + z )2)
0.00387169( 0.925 + z)
(x2 + y2 + (0.925 + z)2)5 /2
0.00164998( 0.875 + z)
(x2 + y2 + (0.875 + z)2) 5/2
0.00207601( 0.825 + z)
(x2 +y2 + (0.825 + z)2)5 /2
0.00241959( 0. 775 + z)
(x2 + y2 + (0.775 + z)2)5 /2
0.003267 43( 0.725 + z)
(x2 +y2 + (0.725 + Z)2)5 /2
240 4 Ideal Fluid
0.00397453( 0.675+ z)
(x2 + y2 + (0.675 + Z)2)5 /2
0.00495214( 0.625 + z)
(x2 +y2 + (_0 .625+z)2)5 /2
0.0057672( 0.575+ z)
(x2 + y2 + (0.575 + z)2)5 /2
0.00672378( 0.525 + z)
(x2 +y2 + (0.525 + Z)2)5 /2
0.007 46622( 0.475 + z)
(x2 + y2 + (0.475 + z)2)5 /2
0.00823441( 0.425+ z)
(x2 + y2 + (0.425 + Z)2)5/2
0.00869796( 0.375 + z)
(x2 + y2 + (0.375 + Z)2)5/ 2
0.00905145( 0.325 + z)
(x2 + y2 + (0.325 + z)2)5 /2
0.00895613( 0.275 + z)
(x2 +y2 + (0.275 + z)2)5/2
0.00857577( 0.225 + z)
(x2 + y2 + (0.225 + z)2)5 /2
0.00753456( 0.175 + z)
(x2 +y2 + (0.175 + z )2)5/2
0.00600065( 0.125 + z)
(x2 + y2 + (0.125 + Z)2)5 /2
0.0035396( 0.075 + z)
(x2 + y2 + (0.075 + z)2)5/2
_ 0.000519049( 0.025 + z) ))
(x2 + y2 + (0.025 + z)2)5/2
n = 20. We show in Fig. 4.21 three pairs of the streamlines. In each pair
of the streamlines, the initial points (xo , Yo, zo) were taken symmetrically
with respect to the plane y = O. The interested reader can easily obtain
other streamlines by specifying his own values of (xo, Yo , zo) (see the
lists zOl, y01, x01 in the main program ThreeDimF10w [ ... ] of our
Mathematica Notebook prog46 .nb).
In Fig. 4.22, we show the distribution of the fluid velocity vectors
in the threedimensional flow around the same body of revolution as in
Fig. 4.21. To obtain this figure, we have used the builtin Mathematica
function P1otVectorFie1d3D [J of the software system Mathematica 3.0
(see a description of this function in Appendix A).
As compared with the finite difference or finite element methods
for the numerical solution of threedimensional problems, the above
presented method of sources and sinks is much more efficient because
it does not need any spatial computing mesh in the threedimensional
space around the body of revolution.
The approach we have used above is similar to the boundary ele
ment method. In contrast with the latter method, however, the above
presented method has the advantage, that it enables one to obtain the
analytic formulas, with the aid of which it is possible to study the solu
tion behavior at infinity.
242 4 Ideal Fluid
Figure 4.22: The distribution of the fluid velocity vectors in the flow
around the body of revolution.
y
Zo
x
Xo 00 Yo
Vn = ( o<p
an s
) =
~
(Uo + U~) ~
w . n,
The condition at infinity also retains its form since the relations
are equivalent; i.e., during the time 6.t, the body will pass only a finite
interval. The condition on a body surface will have in this case the
following representation:
244 4 Ideal Fluid
It follows from this formula that the potential <P should linearly depend
on the velocities that are variable in time and will have the following
form 7 :
(4.4.3)
where the functions <Pi (i = 1, ... ,6) are the functions of the coordinates
(x, y , z). Such a form for the representation of the velocity potential was
proposed for the first time by Kirchhoff. Thus, if the body form and the
law of its motion are given, then the determination of <P reduces to the
solution of the external Neumann problem for the Laplace equation.
By virtue of the linearity of problem (4.4.1), all functions <Pi (x, y, z)
must satisfy the Laplace equation
where i is the radius vector of a surface point with respect to the coor
dinate origin (see Fig. 4.23). Let us write the Lagrange integral in the
coordinate system (xo , Yo , zo):
fHp + v2 + P = f(t) ,
at 2 Po
which has the following form at Voo = 0:
 J. r n(acpat + 2:V2)
F = Po }s dB, L= Po JIs, (i x n) . (~~ + V;) dB.
(4.4.7)
Similar results can be obtained if we make use of the conservation laws
for the momentum and for the momentum moment written in integral
form. Let us take an arbitrary surface I:, which is fixed in space and
encloses the body B. By definition, the momentum K available within
a volume T between the surfaces B and I: is equal to
K= Po J~ cp . n dB  Po JIs cp . n dB.
Applying the theorem on the momentum change to the fluid mass within
the volume V , we have:
dK
 = F F
,  (4.4.8)
dt '
246 4 Ideal Fluid
where F' is the main vector of the forces acting on the surface ~ from
the fluid located outside the volume V. With regard for (4.4.7), we have
the following expression for P':
The total variation of the momentum during the time dt within a volume
is equal to
where the last item corresponds to the momentum variation at the ex
pense of a fluid that has flown into the volume V or has left it during
the time dt; i.e.,
Po J'iEr fi (O'P
ot
+ V2) dS 
2
.:idt J'iErP'P' fi dS
+ ! Jis Po 'P . fi dS  J1 Po if Vn dS.
therefore,
lim
R+oo J
iEr(ii. v
2
2
 iJ v n ) dS = O.
Let us compute in a similar way the main moments of the pressure forces
that act on the surfaces S and ~ within the volume V. Then the law of
the variation of moments is equal to
dl = if _ i.
dt
By the definition,
I = Po JL Jis
*
$(f x ii) dS  Po $(r xii) dS.
The expressions for i' and will be similar to the aboveobtained for
mulas (4.4.9) and (4.4.10); i.e.,
+ JL Po (r x v) . Vn dS.
Taking into account the fact that the surface ~ is fixed in space and
turning R to infinity, we obtain:
dO
dt =

F+R,
 dH
dt 
= L+Q.

248 4 Ideal Fluid
where
B4 Po Jis <p. (Y/  z(3) dB = Po Jis <P . a;:4 dB,
B5 Po JIs 'P. (za  xI) dB = Po JIs 8::
'P. dB,
or
6
Bi = L AikUk, (4.4.15)
k=l
4.4 Nonstationary Motion of a Solid in the Fluid 249
where
It follows from these formulas that all Bi are expressed in terms of Uk,
that is, in terms of the components of the solid body velocity and the uo
angular velocity w. The coefficients Aik having the dimension of mass
are essentially determined by the body geometry, and it is conventional
to call them the virtual masses. There are 36 such coefficients Aik, and
if the values of 'Pk are known, then their computation reduces to the
numerical quadratures. One can show that the tensor Aik is symmetric:
Aik = Aki' Therefore, the number of different coefficients Aik is no more
than 2l.
As an example, let us consider the abovesolved problem of the flow
past a sphere of radius Ro moving in a fluid at a velocity V00 under the
action of the force R applied to the sphere center. The velocity potential
for the flow past a sphere moving at a unit velocity along the Ozaxis
has the form
R~ cos()
'P3 =  2r2 '
where the (), r, and A are the spherical coordinates with the origin at
the sphere center. We find that
= o'P31 = cos () Ro
o'P31 'P3 I =   cos ()
on s or r=Ro ' r=Ro 2
and, consequently,
=
P R3
O2 0
rr
io io
27r 2
cos () sin () d() dA = "3Po7rR~ .
2 3
"3P07rROUi, i = 1,2,3,
0, i = 4,5, 6,
or in the vector form:
 2 3 _ 
B="3Po7rRou, 1=0,
250 4 Ideal Fluid
 2 3 dil
F = 3P07rRo dt'
It follows from the obtained formulas that the forces are reduced to a
single resultant force applied to the sphere center, and the main moment
will be equal to zero. If the sphere mass is equal to m and the force
Ii applied to its center acts on the sphere, then the motion equations
(4.4.14) for the sphere may be written in the form
dil
m dt
2 3dil
+ 3P07rR dt = R

or
(
m
2 3) dildt = R.
+ 3P07rRo
Thus, the sphere motion occurs in such a way as if it had occurred in a
vacuum, and the sphere mass is increased by the amount ~P07rR~, equal
to a half mass of a fluid displaced by the sphere.
r = [iJ. df'
B
z
Bo
A
Ao y
o
x
r(t) = r, v. ~i ds
JAoBo s
(4.5.1)
Since
v av ds = d(V2) a  d
 aid s=a r
as 2 ' as '
252 4 Ideal Fluid
the final expression for the derivative of circulation will have the form
df = fa. dr + v~ _ v~.
dt AVB 2 2
If the curve AB is closed, then A = Band
~~ = f ~: .dr. (4.5.2)
Thus, the time derivative of the velocity circulation over a closed contour
is equal to the circulation of acceleration over the same contour. For the
ideal fluid and gas, we have the following equation for the momentum
change in the form (2.1.98):
dv 1 
 = V7P+ F. (4.5.3)
dt p
Since the fluid is barotropic [p = p(P)]' a function P(P) exists, such that
~ V7 P = V7P. The body forces are conservative by the definition; there
fore, F =  V7U. Taking into account the Thomson's theorem conditions,
we can rewrite formula (4.5.3) as
dv
dt = V7(P + U).
df
dt =  f d(P + U) = 0,
from where it follows that f(t) = const.
Thus, the velocity circulation over any closed contour moving with a
fluid remains constant for this contour at any time in the motion. The
proof of the Thomson's theorem is then completed.
Now, let us prove the following Lagrange theorem:
If the conditions of the Thomson's theorem are satisfied and there are no
vortices in a fixed fluid mass at some moment of time t = to, then there
will also be no vortices at the subsequent moments of time.
Proof. Assume that at some moment of time t = to there are no vortices
in the fluid mass under consideration lying within a volume V; that is,
n = o. Consequently, the fluid flow will be potential and v = V7cp, where
cp is the velocity potential. The velocity circulation fo over an arbitrary
closed contour lo will be equal to zero by the definition:
(4.5.6)
In this particular case, where the vortex surface has a tube shape, it is
called the vortex tube. Let the liquid particles form a vortex surface lo at
the moment of time to. Let us choose on this surface an arbitrary closed
contour lo bounding a piece of the surface ao. From the Stokes formula,
we have:
r = J v dr =
ho
J.Jaor n ii dB = 0.
254 4 Ideal Fluid
At the moment of time t, the fluid particles, which were located at time
to on la , will pass to the contour I bounding the area (7 of the surface S.
Since the motion of liquid particles obeys the conditions of Thomson's
theorem, we have:
r = iv, dr= 0
or, by virtue of the Stokes formula,
(4.5.7)
Thus, we have proved that the vortex filament remains a vortex filament
in the process of its motion.
Let us now prove the second part of the theorem, which states that
the vortex tube strength will be constant along its length and will not
change with time. Consider a vortex tube with a curvilinear axis (see
4.5 Vortical Motions of Ideal Fluid 255
(4.5.8)
Since On = 0 on ~, then
(4.5.9)
n
where is an outer normal to the surface bounding the volume T (see
Fig. 4.25). Taking into account the fact that iiI = nl and using the
Stokes formula, we obtain:
J.lS2r fJ n dS = ih
v df' = r 2, J'ls,r fJ n dS = i
l,
v df' = r~ = r 1,
(4.5.10)
where fl and f2 are the velocity circulations calculated by a passage
along the contours hand l2 in the same direction. It follows from for
mulas (4.5.9) and (4.5.10) that
Since the conditions of Thomson's theorem are satisfied and the contours
hand l2 have been chosen arbitrarily, the circulation along any liquid
contour does not depend on time and, consequently, the vortex tube
strength does not change with time. The Helmholtz theorem has been
proved.
If the vortex tube ends in the fluid, then S2 + 0, and according to
(4.5.7) the angular velocity ~n in this section tends to infinity, which is
physically impossible. By virtue of the Helmholtz theorem, the vortex
256 4 Ideal Fluid
as
(4.5.11)
Consider the case of a barotropic fluid; i.e., p = p(P), but the body forces
are nonconservative. Equation (4.5.11) can be simplified with regard for
this condition to the form
it
The work F df' for a nonconservative force F is not equal to zero at a
passage along a contour I; therefore, ~~ i= 0 and the Thomson's theorem
is invalid; that is the vortices may arise and vanish. Let us consider the
case in which the body forces are conservative, i.e., F = \lU, but the
fluid is baroclinic. In this case, p depends not only on pressure, but also
on temperature, humidity (of air), or salinity (water). Equation (4.5.11)
takes the form
dr = 
dt
jB WdP  l
ABC
C
WdP  lD WdP  lA
D
WdP = Wo (wo +1) = l.
4.5 Vortical Motions of Ideal Fluid 257
w = Wo +1
W=Wo
c P=Po
/
B PI = Po +1
0:: + rot (\i' C2 ))  rot (iJ x rot iJ) = rot F  rot (~\i' P ) .
 ::'2
Since
rot ( \i' . ~) 0,
rotG\i' p) 1
2" \i' P x \i' P,
p
rot iJ = n ,
we have:
an + (iJ \i')D 
;:;  
(D . \i')iJ  D diviJ = rot F
+ 2"1 \i' p x \i' P,
vt P
258 4 Ideal Fluid
or
dn ~ ~ ~ 1
d = (S1. V') V + S1divv + rotF + 2V'P x V'p. (4.5.12)
t P
It is conventional to call equation (4.5.12) the Friedmann's equation.
If we assume that the field of the body forces is conservative that is
F = V'U, and the fluid is barotropic, then equation (4.5.12) simplifies
to the form
dn ~ ~
  (S1. V')V  S1 . divv = o. (4 .5.13)
dt
If we assume that the fluid is incompressible, then we obtain the equa
tion:
dn
dt
= (0 . V')v. (4.5.14)
divn = O.
We will search for the solution of equations (4.5.15) in the form of the
sum
where the functions VI and V2 satisfy the following equation and the
boundary conditions, respectively:
(4.5.16)
4.5 Vortical Motions of Ideal Fluid 259
Since the original problem is linear, the sum of these two solutions will
be the solution of problem (4.5.15).
Let us construct the solution of problem (4.5.16). Introduce the
function c.p by the formula
Vl = \1c.p.
In this case, the second equation of system (4.5.16) is satisfied automat
ically and the substitution into the first equation yields
c.p = ~
47l'
11100 B(~,
00
r
1], () d~ d1]d(, (4.5.19)
(4.5.20)
One can assume without loss of generality that div A = 0. Let us prove
the validity of this assertion. If we indeed assume that div A = f =I 0,
then, assuming Al = A + \7 cp, we get:
Each of these equations is the Poisson equation, and the solution of these
equations has the form
A = .!...
4n
Jl1O ~ d~
00 r
d",d(,
'!"'rot
4n
Jl1 ~ d~ 00
00 r
d",d(. (4.5.23)
V=.!...\7.JJJ~d~d"'d(+.!...rotJJJoo
4n r 4n
fld~d",d(.
r  00
(4 .5.24)
(4.5.25)
Denote the tube section by (J, and denote the mean tube line by I. Let [
be a unit vector of the tangent to the mean line. Assuming the velocity
vortex n to be constant in each section, we can write for the length
element dl of a vortex tube that
v = 4
1 rot
7r
J 11 0
I
dl
<r r
1
 d(J::::::: rot
47r
Jn
I
(J[
r
 dl.
V= ~.
47r
rotJ!
I r
dl (4.5.26)
~ (i.ltz i.l
47r oy oz
tyI r
dl
I r
dl) ,
z A
j y
i 0
x
(4.5.27)
We have in (4.5.27) under the integral sign a vector product of the two
vectors [and ii = f; that is,
 f1
v = 4
7f I
_dl = 
(t x n)"2
r
f1
47f I
dl
(t x T)3"'
r
(4.5.28)
It follows from the obtained formula that the vortex filament element
dl engenders at point M(T) the velocity f:l.v, which is computed by the
formula _ f _ dl
f:l.v = 47f (t X T) r3 . (4.5.29)
where a is the angle between the vectors f and r. The formulas (4.5.28)
or (4.5.29) are similar to the Biot Savart formulas in electrodynamics.
Consider a particular case in which the vortex filament is rectilinear
and infinite (see Fig. 4.27). Let a rectilinear vortex filament AB (Fig.
4.5 Vortical Motions of Ideal Fluid 263
4.27) pass through point (~, TJ) in parallel with the zaxis. Then fx
fy = 0, fz = k,dl = d( , and the formulas (4.5.28) simplify to
iJ = ~ (')Q k x r de.
47r Loo r3
Mapping this equality onto the coordinate axes and calculating the in
tegrals, we obtain:
rYTJ r x~
vx = 27r 7 ' Vy = 27r 7' Vz = 0, (4.5.30)
where
00 d( _ ~
/
r3  p2
00
The formulas (4.5.30) describe a planar fluid flow, where at each point
perpendicular to the vortex the particles move along a circle at the cen
ter of which the vortex is located. The velocity magnitude is equal to
v = 2r7r . 1.
p
The counterclockwise motion along a circle, p in radius,
corresponds to the positive values of r, and the clockwise motion along
this circle corresponds to the negative values of r . As a consequence of
the symmetry of the fluid flow around a point vortex it is obvious that
the vortex will be at rest . Using the functions of a complex variable
z = x + iy, let us write the formulas (4.5.30) in the form
r 1
Vx ivy =  ..   ,
2m z  Zo
where Zo = ~ +iTJ, 2 = xiy, and 20 = ~ iTJ. We recall that the complex
potential of the vortex w = 2~i In(zzo), and we have by definition that
Vx  ivy = ~~, where r is the vortex intensity, or nothing more or less
than the velocity circulation along any closed contour enclosing the point
zoo
For a qualitative explanation of a variety of the phenomena occurring
in nature one can introduce the concept of the discontinuity surface,
which is the surface on which some hydrodynamic parameters undergo a
discontinuity. The velocity is usually chosen as such a quantity. Such is,
for example, the discontinuity surface in a cyclone, along which the cold
and warm air get in touch, and where a jump in the wind velocity takes
place. Let us show that the surface of a discontinuity in the tangential
velocity component may be considered as a limiting case of a vortex layer,
that is, a space between two close surfaces that is filled by vortices, and
in this case, a continuous although rapid velocity variation takes place.
We assume for simplicity that the discontinuity surface is the plane S
264 4 Ideal Fluid
y
a+c
S a
parallel with the Oxy plane so that its equation is y = a (see Fig. 4.28).
Introduce the plane Sl with equation y = a + c, which lies at a distance
c from the plane S. Let the fluid move at a velocity v on the one side
of S and at a velocity VI on the other side Sl. Both velocities are
constant and parallel with the xaxis. Assume that the components of
the velocities V and VI along the Oxaxis are u and U1. Thus, we have the
discontinuity only in the tangential velocity component. Assume that,
in a layer between Sand Sl, the velocity components are determined by
the formulas
Vy = Vz = o.
Then we will have in the plane S, for which y  a = 0, that Vx = u, and
in the plane S1, where y  a = c, that Vx = u1 .Consequently, at such a
specification of the velocity, its magnitude will change continuously from
U to U1 while passing to the plane Sl from the plane S. The curl vector
in the layer SSl has a direction perpendicular to the Oxy plane, and its
component along the Ozaxis is equal to
o _ avy avx _ U  U1
z  ax  ay  c'
r = Jis Oz dx dy = + 11 l
dx
E
U ~ U1 dy,
which yields
4.5 Vortical Motions of Ideal Fluid 265
It follows from the last formula that the vortex tube strength r = r2 z . C
does not depend on the layer thickness c. In the limit, as c + 0, r2z + 00,
and the intensity r = U  Ul remains constant, we will have a flow with
a surface of the discontinuity in the tangential velocity component. Such
a flow with a tangential discontinuity may be interpreted as a flow en
gendered by a vortex layer in which the vortices of a sufficiently large
intensity are located. The introduction of the vortex layer concept gives
a possibility to explain the origin of the vortices in a fluid . By virtue of
the Lagrange theorem, if there are no vortices in the ideal fluid at the
initial moment of time, then there will be no vortices at all times of mo
tion. In reality, we have that , under conditions close to the conditions of
the Lagrange theorem (the constancy of density, small viscosity of fluid,
and the availability of a potential of the acting forces) , the vortices in
the fluid arise. If we assume that a vortex layer on the surface of a body
flowed past emerges, then it is not difficult to imagine that, in the case
of instability of this layer, the vortices can separate from it , as this often
takes place in reality at the motion of a body in the fluid.
References
This chapter is devoted to the viscous fluid flows, which are described
by the Navier Stokes equations. We derive the Navier Stokes equations
in the Cartesian, cylindrical, and spherical coordinate systems and con
sider their exact solutions at small Reynolds numbers. We present the
Prandtl's theory of boundary layer, which is valid at large Reynolds
numbers. This theory enables one to calculate the drag force acting
on a plate in the viscous fluid flow. We also outline the theory for the
transition from laminar viscous fluid flow to turbulent flow and discuss
a number of the semiempirical theories of turbulence.
As was noted above, in the ideal fluid , the surface forces applied to
the surface elements of any fluid volume represent the normal pressures
directed inside the volume. Any actual fluid possesses, however , the
viscous property, which gives rise to shear stresses. It is this property
that is one of the reasons causing the drag of the fluid flow in pipes and
channels or the drag of bodies moving in the fluid. In this case, there
are, between the layers in the viscous fluid flow, the forces tangent to
the direction of motion of these layers.
The classical viscous fluid is an isotropic medium whose shear drag
is different from zero and linearly depends on the shear strain rate [see
(2.1.101)]. The equations governing the viscous fluid flows were derived
for the first time by Navier (1822), who applied a simplified molecular
model for the gases. This has led to the introduction of a positive viscos
ity Il > 0, which in the opinion of Navier desribes the molecular diffusion
of the momentum. It has been generally recognized at present, however ,
that simple laws for the molecular forces describe inadequately the ac
tual fluids. Therefore, a continual approach proposed by Stokes (1845)
is considered to be more preferable, and we will follow this approach.
divv= 0,
8v
8 + ('<"7) 1 '<"7p
v v V =   v + Vu v,
A _
(5.1.1)
t Po
dT
cVPOdj = K~T + <P,
where <P = 2/L6ij6ij, 6ij = ~(~ + ~~n is the rateofstrain tensor. The
function <P entering the system of equations (5.1.1) is nonnegative and
v =
With the use of the covariant derivative definition (1.1.25), this formula
can be rewritten with regard for the symmetry of the Christoffel symbols
r;k = r~j' as follows:
~_ 1 (OVk OVj) ~
rot V  ,;g oxj  ox k ei.
(rotiJ)i = ~(OOj
y9j9kx
(..j9kuk)  oOk (y9jUj)) ,
x
(5.1.5)
270 5 Viscous Fluid
1 i 09i 1 o,;g;
9=
2 oxj ,;g; ox j ,
1 i 09j 1 09j
9  , =     , . (5.1.9)
2 OX' 29i ox'
Substituting (5.1.9) in (5.1.8), we obtain at i = j:
(5.1.10)
5.1 General Equations of Viscous Incompressible Fluid 271
and at i =1= j:
(5.1.12)
Knowing the (Jij, one can find the physical components of the stress
tensor aij by formulas (1.1.45):
(5.1.13)
 P + 2It ( 1 aUi
+i
 Ui
 av1fi
j + Uk 
av1fi)

v1fi ax vgigj ax Vgigk axk '
Note that the coefficients v1fi == Hi are sometimes called the Lame
coefficients in the literature 1 .
In the Cartesian coordinate system, we set gl = g2 = g3 = 1, Xi =
X, x j = y, xk = Z, Ui = VX , Uj = v Y' and Uk = Vz . Substituting these
values in (5.1.6), (5.1.7), (5.1.15) and (5.1.1), respectively, we obtain the
Navier Stokes equations in the Cartesian coordinates:
OVx OVx OV y )
O"xx = P+ 2f.t ox' O"xy = O"yx = f.t ( oy + ox ; (5.1.16)
oVy oVx OVz)
O"yy = P+2f.t oy' O"xz=O"zx=f.t ( OZ + ox ;
oV z OVy
O"zz = P + 2f.t 8z , O"yz = O"zy = f.t ( 8z + OVz)
By .
The energy equation is not considered here for the reason noted above.
In the cylindrical coordinate system xi = r, x j = lP, Xk = z the
metric tensor components are [see (2.1.130)J
. _ 2
gi = 1, g)  r , gk = 1. (5.1.17)
Introduce the notations
Ui = Vn Uj = v"" Uk = vz, O"ii = O"rr ,
aij = O"r"" aik = O"rz, ajj = 0"",,,,, ajk = O"",z, akk = O"zz
With regard for (2.1.134), we obtain the system of the Navier Stokes
equations in the cylindrical coordinates:
oVr 1 OV'" oVz Vr 0
Tr + ~ OlP + 8z + ~ = ,
Ui Vr , Uj = VB , Uk = Vcp ,
O'ii arr , O'ij = arO , O'ik = a r cp ,
iJ jj aoo, iJjk =aocp, iJkk = acpcp.
1
 (2vl[r,
r
e, i.p] + cot(O)v2[r, e, i.p]
+ csc( O)v~O,O , l) (r , e, i.p) + rv~l , O ,O)(r, e, i.p)) = 0
274 5 Viscous Fluid
With regard for the above expressions for components of !::..V, we obtain
the system of Navier Stokes equations in the spherical coordinates:
divv = 0,
(v\l)v = ~ \l P + v ~v, (5.1.21)
Po
in an infinite region, which satisfies the boundary conditions
vi s = ' vI 00 
= V oo ,
where S is the body surface. If one searches for the solution of (5.1.1)
in the class of irrotational flows, then v = \l <p, and by virtue of the
continuity equation in (5.1.1) , we have ~<p = O. It follows from here that
~v = ~(\l<p) = \l(~<p) = O. In the presence of this relationship, the
Navier Stokes equations will coincide with the Euler equations; that is,
v
the solutions of the Euler equations under the condition = \l <p satisfy
equations (5.1.2) and are the solutions of the Navier Stokes equations.
As shown above, however, the solution of equation ~<p = 0 for the
276 5 Viscous Fluid
Vnl s = ocp
an Is = 0' 'Vcploo = voo
The tangent velocity component vr on the body surface S will be differ
ent from zero, that is Vr Is = ~ lsi O. This means that the potential
flow in the case of a viscous fluid does not satisfy the noslip condition
vi s = 0 at points of a solid body. Thus, the class of the potential flows
cannot be used for solving the problems of the viscous incompressible
fluid flow around the bodies. The viscous fluid flows should always be
rotational in this case.
An inherent nonlinearity of the system of NavierStokes equations
makes impossible the application of the superposition of the solutions,
which enabled us to construct the solutions for a variety of practical
ideal fluid flows, as shown in the foregoing sections. As is known, any
streamline in the inviscid fluid can be replaced with a rigid wall and
thereby the fluid flow will not be disturbed. This is related to the fact
that the boundary condition for the normal velocity component on a
streamline and on a rigid wall is the same: Vn = O. In the case of
a viscous fluid , one more boundary condition Vr = 0 appears on the
wall, which is absent on a streamline. It follows from here that in the
viscous fluid case a substitution of an arbitrary streamline for a rigid
wall is invalid. This should be taken into account while constructing the
solutions.
OVx = 0 oP = oP = 0 Po = const,
ox ' oy oz '
Taking this into account, we can present the last equation of system
(5.2.1) as
OV x _v(02Vx + 02Vx ) = _~ op. (5.2.2)
ot oy2 OZ2 Po ox
oP
ox = f(t) , P = f(t) . x + ft(t).
Thus, in a onedimensional flow the pressure is a linear function of x and
the functions f(t) and ft (t) can be determined if the pressure P is given
in two sections Xl and X2, namely,
Then
oP = F2(t)  FI(t) = !:::.P (t)
oX X2  Xl !:::.x'
and the velocity is found from equation (5.2.2):
The problem is simplified if the flow is steady. In this case, the pressure
jump is constant and equation (5.2.3) reduces to the solution of the
Poisson equation
(5.2.4)
1 b:.P) y2
Vx =
( ;;, b:.x 2 + C 1 y + C2 ,
where C 1 and C 2 are the integration constants. Determining the C 1 and
C2 from the boundary conditions, we obtain:
_ ~ b:.P( 2 _ h2 ) V1 V 2 VI +V2
Vx  2fl b:.x Y + 2h Y+ 2 .
Figure 5.1: The velocity profile vz(y) in the steady flow between two
parallel planes.
Problem 5.1. Find a steady viscous fluid flow between two parallel
plates y = h when one of the plates moves at a velocity Vo and the
second plate is at rest.
Consider a viscous fluid flow in a circular tube of radius R. The tube
is at rest, and the zaxis coincides with the tube axis. Assuming that
only one velocity component V z = V z (r) is different from zero, we obtain
from system (5.1.18) the equation for V z :
d2 v z 1 dv z 1 6.P
+=
dr2 r dr /l 6.z '
where /l = POV . Let us present the solution of this equation as follows:
Taking into account the boundary noslip condition on the tube contour
VZ IR= 0 and the finiteness of the velocity magnitude along the axis , we
find the PO'iseuille formula
1 6.P 2 2
Vz =  . (R  r ). (5.2.7)
4/l 6.z
Let us calculate the flow rate across the tube crosssection by the formula
Q= looR lo27f
0
27r 6.P
vzrdcpdr=
4/l 6. z
loR r(R
0
2 2 7r 6.P 4
r)dr=R.
8/l 6. z
The obtained solution does not always agree well with experimental data
and depends substantially on the Reynolds number Re = ~. If Re ::;
1000  1100, then there is a good agreement with the experiment. At
280 5 Viscous Fluid
Re > 1100, a drastical change in the flow pattern occurs. At small Re,
each fluid particle moves closely to a straight line: the flow is stratified
and quiet. It is conventional to call such a flow the laminar flow. At
Re > 103 , each fluid particle performs a chaotic motion and it is no longer
a onedimensional and steady flow. Such a flow is called turbulent.
The formulas (5.2.6) and (5.2.7) are valid only for laminar flows. The
number Re, at which a flow passage from a laminar to turbulent regime
takes place, is called the critical Reynolds number. The numerical value
of the critical Reynolds number depends substantially on the quality of
the polishing of the material of tubes, the inlet section, and many other
parameters.
Problem 5.2. Extend the solution (5.2.7) obtained for the axisym
metric flows in a circular tube for the case of axisymmetric steady fluid
flows inside an annular tube Ri :::; y2 + Z2 :::; R, where Vlr=R 1 = Vl and
Vl r =R2 = V2 are the velocities at which the tubes move in parallel with
their axes.
Figure 5.2: The viscous fluid flow between two rotating coaxial cylinders.
P = PI + Po i T
Tl
v2

l'
. dr.
Let us calculate the friction force that acts on the elements of the internal
and external cylinders by the formula (5.1.18)
ar !!.)
2 J1 (WI  W2) rir~
= J1(OV  =
1'2 (r~  ri)
O"r<p
l'
r~
v=wr
rr r~ W
1'22  1'2I 1'22  1'21 l'
3) Let 1'2 + 00, W2 = 0. We then obtain the fluid motion outside the
cylinder at a velocity
Wirr
V=.
l'
where r2 = x 2+y2 +z2. Since the solution is sought for at small Reynolds
numbers, it is necessary that the following two conditions be satisfied:
the freestream velocity Voo is sufficiently small or the sphere radius a is
small. Introducing the nondimensional variables by the formulas
x y z _ a_ p =;;
~ = ~' rt = ~' (= ~, w = ;; v,
(a)2 . poP ' (5.2.9)
P I
00
 (va)2 Poo
Po
(5.2.11)
It follows from here that the nondimensional velocity will have the order
Iwl "'" Re in the flow region. Since the Reynolds number is small, Re 1,
the following inequalities will be valid:
r
() Voo
Thus, the convective derivatives Wi~ in the system (5 .2.10) are the
quantities of the secondorder smallness. Neglecting in (5.2.10) the terms
of the secondorder smallness in comparison with the terms of firstorder
smallness and returning to the dimensional variables by formulas (5.2.9),
we obtain the system of the Stokes equations:
Thus, the system of the Stokes equations (5.2.12) governs the incom
pressible fluid flow at small Reynolds numbers Re 1 and is obtained
from the complete system of the Navier Stokes equations (5.1.16) by
neglecting the convective derivatives.
The Stokes equations are linear, which facilitates greatly the con
struction of analytic solutions. It is in particular possible to solve with
its aid the aboveformulated problem of the flow past a sphere. For
this purpose, we introduce a spherical coordinate system whose origin is
placed at the sphere center (see Fig. 5.3). By virtue of the flow symme
try with respect to the Oxaxis, the desired functions will depend only
on rand ():
+
284 5 Viscous Fluid
where the prime denotes the derivative with respect to r. The boundary
conditions for system (5.2.18) follow from (5.2.15) and (5.2.16):
We search for the solution of the first equation in (5 .2.20) in the form
j = rn. As a result of substitution, we obtain an equation for n: n(n
2)(n + l)(n + 3) = 0, which has the solution n = 1, n = 3, n = 0,
and n = 2. Thus, the general solution of the first equation in (5.2.20)
has the form
C 1 C2 2
j= 3++C3+C4r, (5.2.21 )
r r
where C 1 , C2 , C3, and C4 are the integration constants. Substituting
(5.2.21) in the second and third equations of system (5.2.20), we obtain:
C1 C 2 C2
9= 3
2r
+ 2r2 + C3 + 2C4r, b = 2"
r
+ lOC4r. (5.2.22)
(5.2.23)
(5.2.27)
which are also linear. One can compute with the aid of these equations
the velocity field and find a refined formula for the drag force of a sphere l
avoo
Re =  1. (5.2.28)
v
Note that, while solving the problem of the viscous incompressible fluid
flow moving perpendicularly to the cylinder, one must immediately use
the Oseen equations. [The Stokes equations (5.2.12) in this case have
no solution satisfying the boundary conditions on the cylinder and at
infinity2.] The force acting on the unit cylinder length was calculated
for the first time by Lamb and is equal to l
&vx &vy _ 0
&x + &y  ,
&vx &vx &vx 1 &P
~ +VX & +vy &
U~ X Y
= a
Po x
+I/~X , (5.3.1)
av x ~ O(~) (5.3.3)
ay J '
Using the aboveintroduced hypothesis J /L 1 and comparing the
orders of the terms a;:2~ and ~~l we find that a;;l a;:l.
Estimating
the remaining terms entering the second equation of system (5.3.1), we
find that
rYavy vJ
Vy = Jo ay dy ~ 0(1:);
consequently,
v av x ~ O( vJ . ~)
Yay L J.
Thus, the terms of the lefthand side of the second equation of system
2
(5 .3.1) will have the order of magnitude O(VL).
Prandtl assumed that the inertia forces and the viscous friction forces
in the boundary layer have the same order of magnitude; that is,
or
where Re = v!;. It follows from the obtained estimate that the relative
thickness of the boundary layer is inversely proportional to VRe; that
is, the larger the number Re, the thinner the boundary layer.
We will estimate the terms involving the pressure by using the Ber
noulli integral, which will be valid at the outer edge of the boundary
layer; that is,
v2 P
 +  = const.
2 Po
290 5 Viscous Fluid
Consequently, one can neglect the term vo;;f in this equation, and it
takes the form
The continuity equation remains unchanged since the order of its terms
is the same.
Consider the third equation of system (5.3.1). Estimating the orders
of terms, we have:
vYoy
(Ovy) >::::, 0 (~ . V2)
LL'
02v y >::::, o(~)
oy2 L6 .
Thus, taking the above into account, we can simplify the system of
the NavierStokes equations (5.3.1) and write it as
av x av x av x 1 ap a2v x
at +vxa
X
+vY  a
Y
=   a +I/a
POX Y
2 '
avx avy _ 0
ax + ay  , (5.3.4)
P = P(x, t).
auG 1 ap
uo+=o,
ax Po ax
where uo is the flow velocity at the outer edge of the boundary layer.
With regard for this relation, we can simplify the system of equations
(5.3.4) and write it in the form
(5.3.5)
Since at y = 0 the quantity :0~~ =  (uo ~ )y=o, one can take as the
function uo the solution of ideal fluid motions at y = O. The function
uo = U x depends only on x. The functions Vx and Vy are determined as
a solution of the system of equations (5 .3.5). To solve these equations,
it is necessary to specify the initial and boundary conditions. The first
equation in (5.3.5) has the parabolic type (see Problem 3.2 in Chapter
3); therefore, it is necessary to specify for its solution the initial condition
at x = 0 and the boundary condition at y = 0 and y = 8(x) as follows:
1) the initial condition at x = 0: Vx = uo(O);
2) the noslip conditions are satisfied on the body wall: vxly=o = 0 and
vyly=o = 0;
3) at the outer edge of the boundary layer,
Having the velocity distribution in the boundary layer, one can find the
outer edge of the boundary layer b(x) by using the equality vx(x, b) =
(1  e)uo(X), where e is a given small quantity.
Finally, in the case of an unsteady motion, one must also specify
an initial condition at t = 0: Vx = v~(x, y), where v~(x, y) is a given
function of its arguments.
The obtained equations (5.3.5) describe a laminar flow; therefore,
they are often termed the equations of a laminar boundary layer in the
literature. The use of the Cartesian coordinates at the derivation of
(5.3.5) is of no fundamental importance. A similar derivation can also
be performed for an arbitrary curvilinear coordinate system 1 ,7.
Note that the initial condition vxlx=o = uo(O) does not significantly
affect the solution in the boundary layer. It affects the solution in a thin
layer .6.x rv b, and at .6.x b, its influence rapidly decays.
The boundary layer theory enables one to explain the phenomenon
of the flow separation from a smooth surface of the body (see Fig. 5.5).
Let the derivative ~~ < on the body surface (Fig. 5.5) from point
o to point B. At point B, the pressure achieves its minimum ~~ IE =
5.3 Viscous Fluid Flows at Large Reynolds Numbers 293
av x av x a 2v x
Vx ax + Vy ay =v ay2 ' (5.3.8)
Vxl
y=O,x>O
= 0, vyl
y=O,x>O
= 0' Vxl y=8(x) ,x20 = voo' (5.3.9)
Vxl
x=O,y>O
= voo , Vxl y=oo,x>O = Voo' (5.3.10)
aay2
v I
2 x
y=O,x>O
=0
.
Vxl = 0, aay2
v I
2 x
y=O,x>O
=0
'
y=O,x>O
Vxl = voo , vxl = Voo (5.3.13)
x=O,y>O x>O,y"""oo
5.3 Viscous Fluid Flows at Large Reynolds Numbers 295
Since the problem under consideration does not involve any reference
length, the solution will be selfsimilar. The functions to be determined,
for example, the velocity v x , will depend on the variables x and y and
the viscosity v. Using the Pi theorem, we will search for the solution
of equation (5 .3.12) in the form Vx = F(~) , where ~  is avb . rx
selfsimilar variable. We can calculate that
oV
x _ rr' [ 1 ( 1) yx _,,]
ox .r 
ffv
 
2
2 ,
oV x
oy
= .1" [_1 .
ffv fi
_1 ]
'
The substitution of these expressions in (5.3.12) yields
.F' . .1''''  (.1''')2 + .1' . (.1")2 = 0, (5.3.14)
F(O = k 2 F1(k~) ,
where k is a constant. It is easy to be sure by a direct substitution
that the function F(~) satisfies equation (5.3.14) if the function F1(~)
satisfies this equation. Let us choose k in such a way that F(~) is the
solution of problem (5.3.14) with conditions (5.3.15). We find from the
third condition in (5.3.15) that k = yfiii. Consequently,
Vx = ~: F1(V~: . vk :}x) , (5.3.16)
where Co and .1'1 are the known quantities as the solution of the auxiliary
Cauchy problem.
296 5 Viscous Fluid
Let us calculate the drag R x of the plate of a finite length l and width
b by formula
Rx = 2
io
dz
io
r r b
(TY XI dx.
y=o
l
The appearance of the coefficient 2 is related to the fact that both sides
of the plate give the same contribution to the drag force. Since (Ty x ly=o
does not depend on z, then
Since
(Ty x Iy=o = It (avx + av y ) I = I av
t I
x
ay ax y=O ay y=O'
we find with regard for (5.3.16) that
1 a
(T xy y=O = It ay
(voo (voo y ))1 y=o = Jt (Voo)3/2
C :F1 Co . v2vx Co
1
V2vx'
O
where we have taken into account that :F~ (0) = 1. Substituting the
obtained expression into formula (5.3.17) and computing the integral,
we obtain:
_ 2V2. I.. 3/2 . r,
Rx  3/2 povvv oo b vl,
Co
where we have assumed that It = PoV . The drag coefficient
Rx
Cx = 1 2 '
2 Povoo So
which at the total area of both plate sides So = 2b l is equal to
1 ~ 1
Cx = 2V2(cJ . y'Re'
where Re = ~ is the Reynolds number.
We have used the Mathematica 3.0 system to solve numerically the
auxiliary problem
(5.3.18)
1. 75 F(S)
...
1.5
2 4 6 8 10 S
With regard for (5.3.19), the initial conditions for system (5.3.20) have
the form
.rl(O) = 0, Y(O) = 1, Z(O) = O. (5.3.21 )
We have used in our Mathematica Notebook prog52. nb the standard
fourthorder RungeKutta method 8 to solve the initialvalue problem
(5.3.20) , (5.3.21). The software system Mathematica 3.0 contains a large
library of the methods for the analytical solution of the ODEs. There
fore, we hoped to solve the nonlinear ODE (5.3.14) with its aid in analytic
form. The methods available, however, in the corresponding library of
Mathematica 3.0 are still insufficient to solve this task in analytic form .
2V2 ( Co
1 )3/2~ 1.32822;
therefore, the drag coefficient is
(5.3.22)
b(X) rv ~

x
v=
,
298 5 Viscous Fluid
vj (v)
2
1 o 1
Figure 5.8: The Poiseuille profile 1 and the mean velocity profile of the
turbulent flow 26 .
from where it follows that 8(x) increases with x and the Prandtl as
sumptions on which the boundary layer theory is based are violated.
The experiments show that, at some distance from the plate leading
edge, the laminar boundary layer passes to a turbulent boundary layer.
A
0.1
0.05 "
"
b
0.03 a'\
"
0.01 \ 1
103 104
profile v(r) (see Fig. 5.8) and to an increase of the skinfriction drag.
The skinfriction drag is characterized by the coefficient of the loading
loss A = ~~ /(p~Vr) . At a laminar flow, the dependence v = vz(r) is
determined by the Poiseuille formula (5.2.7) , with the aid of which, one
can compute the mean fluid velocity in the tube:
Substituting this value into the expression for A, we obtain the depen
dence A = 64/Re, which is shown by a dashed line 1 in Fig. 5.9. In the
turbulent flow regime, the dependence A(Re) within a wide range of the
Re values is given by an empirical formula A = O.316/Re1 / 4 shown in Fig.
5.9 by a dashed curve 2. The actual dependence A(Re) observed in the
experiment is shown by a solid line in Fig. 5.9. The transitional regime
from a laminar flow to a turbulent one corresponds to the interval " ab"
in Fig. 5.9. In the transitional regime at the values of Re close to Re*,
an alternation of laminar and turbulent regimes (the intermittency) can
take place. This is related to the fact that the turbulence at first forms
in some bounded regions, representing the liquid tubes filling a part of
the tube crosssection. The experience shows that at Re > Re* the front
end of the liquid tube moves faster than the rear end. This leads to
the extension of tubes, their merging, and the flow turbulization. At
Re < Re*, the picture is opposite, which leads to the disappearance of
the turbulent tubes and the flow laminarization. Returning to Fig. 5.9,
we note that the quantity A for a turbulent flow regime is larger than the
value of A for the laminar flow (line 1) at the same Reynolds number.
The abovenoted instability of the laminar flow at Re > Re* and the
presence of fluctuations in the turbulent flow are the general properties
300 5 Viscous Fluid
where v' vo, P' Po. Substituting this solution in (5.1.21) and
neglecting the quadratic disturbance terms, we obtain a linear system of
equations:
divv' = 0, ~
av' + (~n)~' (~'n)~
Vo v V + V v Vo =  v
A~' ,
1 np' + lIuV (5.4.1)
ut P
where the stationary solution VO(Xi), PO(Xi) satisfies the stationary equa
tions
divvo = 0, (vo \7)vo = ~ \7 Po + 1I ~vo (5.4.2)
P
In our case vo = (U(y), 0, 0), where U(y) was found above [see (5.2.6)].
a'IjJ' q:!fi.
Assuming in (5.4.1) that v~ = By and v~ =  ax ' one can find the
equation for the disturbances of the stream function 'ljJ' (see also l ):
Re* Re
Figure 5.10: The neutral curve for the planar flow between the plates.
(5.4.7)
w ~ U(0)Re 3 / 11 (5.4.8)
2h '
the wand k are related to each other on both curves (5.4.8) by relation
w ~ 4(kh)3. It follows from Fig. 5.10 that, for any Re > Re*, the
disturbances exist that amplify in the flow. This is valid, in particular,
for large numbers Re also. Thus, a small viscosity has a destabilizing
effect. It follows from here that the term vb.iJ in the Navier Stokes
equations (5.1.21) cannot be neglected for a correct flow description even
at small v. This is related to the fact that the small parameter v affects
the higher (second) derivative and can lead to a qualitative change of
the solution.
302 5 Viscous Fluid
tT/2
a; dt',
t
(5.4.12)
Averaging this equation with regard for the relation ViVj = V(Uj + v~vj
and the formulas (5.4.10), (5.4.11), we obtain the Reynolds equations:
where
(5.4.15)
(5.4.16)
304 5 Viscous Fluid
is called the tensor of the Reynolds stresses. It is easy to see that the
system of equations (5.4.13)(5.4.16) is not closed since the tensor of the
Reynolds stresses (5.4.16) is not determined. Therefore, it is necessary
to invoke the additional hypotheses, which enable one to find the form
of this tensor 6 ,10,15,16 .
Let us present below some most widespread semiempirical hypothe
ses. Using an analogy between the molecular and fluctuating fluid motion
in turbulent flow, Boussinesq has proposed for 7rij the formula
In this formula, A is the turbulent viscosity coefficient for which one can
write by analogy with molecular viscosity:
A = pv'l'. (5.4.18)
Prandtl proposed to consider the quantity I' , an analog of the mean free
path for the molecules, and termed I' the mixing length. According to
Prandtl, one can write for a planar flow along the xaxis, for example,
between the plates or in a boundary layer in which Vx = vx(y), that
v~ rv l'~. Substituting this formula in expression (5.4.18), we obtain:
A= Pl21~: I. (5.4.19)
The obtained expressions are called the von Karman formulas. The
turbulent viscosity A usually exceeds by a factor of several dozens the
molecular viscosity tt; therefore, one can neglect the terms ttEij in the
Reynolds equations. It should be noted that the aboveconsidered hy
potheses enable one to qualitatively describe the nearwall flows and
freestreams in the far field from the bodies. Since there are no refer
ence dimensions in the problem in these cases, the mixing length will be
proportional to the corresponding coordinate: l rv CXi.
5.4 Turbulent Fluid Flows 305
where
=
qj K,
atj ,
ax Q.J _  CvPVj'T' . ( )
5.4.23
(5.4.24)
of'
Qj = CvAQ~' (5.4.25)
ux J
Since the physical nature of the turbulent transport of the momentum,
energy, and admixture is the same (at the expense of the velocity fluc
tuations), we can find by analogy with formula (5.4.19) for a planar flow
with Vx = vx(y) that 6
21
AQ = plQ avayx 1, (5.4.26)
306 5 Viscous Fluid
where the IQ and 1M are the mixing lengths for the processes of turbulent
transport of heat and admixture.
The coefficients of turbulent viscosity A, energy transport A Q , and
diffusion AM are determined by the nature of the velocity fluctuations
taking place in the turbulent flow. As already said above, there is no
consistent theory here. There are a number of important qualitative
results, however, explaining some properties of the velocity fluctuations
v' (since in the results presented below the direction of the velocity vector
is insignificant, we will omit the vector symbol over the velocity notation
and we will mean the velocity module)1O,17,18.
It follows from the experiments that there are in turbulent flow the
fluctuations with different spatial dimensions A. Denote the magnitude
order of the velocity fluctuation on the scale A by v~, and introduce the
corresponding number ReA = v~ A/ 1/. The largescale fluctuations have
rv
the size A L and the fluctuations velocity v rv Llv, where Land Llv
are the reference scales of the body and the mean velocity alteration,
respectively. The least scale is determined from the condition ReA rv
1 and has the order Ao rv 1/ /vb. For the smaller scales A Ao, the
inequality ReA 1 is valid and the pulsations decay at the expense of
viscosity occurs. Since ReL = LlvL/1/ 1 in a turbulent flow, then AO
L, and there is in the interval from Ao to L a wide spectrum of pulsations
scales Ao A L, the inertial interval. According to Richardsonl, in
the flow past a body the generation of turbulent flow energy at the
expense of largescale pulsations occurs. After a disintegration of these
pulsations into smaller ones, this energy passes to the pulsations with
smaller scales. Since ReA 1 therein, there is practically no energy
dissipation. Thus, a constant specific energy flux t from largescale
pulsations to smallscale pulsations takes place. This flux dissipates
into the heat at the smallscale pulsations of the order Ao. Using the
t,
constancy of a specific energy flux one can find the energy distribution
[; and the velocity v~ in the interval Ao A L for the homogeneous
isotropic turbulence. In this case, the turbulence properties on the scales
A L do not depend on the mean velocity v and the pulsations velocity
t
v~ will depend on and on A; that is, VA = f( , A). Since the dimension
of the specific heat flux [t] = [EJI(kg s) = J/(kgs) =m2 /s 3 , then we
obtain with the aid of the Pi theorem the Kolmogorov Obukhov law lO ,17
(5.4.27)
v~ rv(~)t (5.4.28)
Llv L
5.4 Turbulent Fluid Flows 307
y u(y)
Figure 5.11: The velocity profile of turbulent flow near the wall.
Thus, the similarity law (5.4.28) is valid for the homogeneous isotropic
turbulence. Applying (5.4.28) to the least scale vb '" 6V(AO/ L)I/3, we
can find with regard for Reo = Aovb/v '" 1 the order of the least scale
AO and the velocity vb:
3/4 A /R 1/4
AO '" L/Re L , vaI '" uv eL ,
(5.4.29)
It follows from this formula that the main portion of energy is located
in the largescale pulsations with the large A '" Land k '" 1/ L. These
scales play the main role in the transport of energy and momentum;
therefore, the turbulent viscosity A '" p6vL.
whose integral is
Txy + 1l"xy = Co (5.4.30)
308 5 Viscous Fluid
Since the transverse velocity fluctuations on the wall are equal to zero,
v~!y=o = 0, we obtain from equation (5.4.16):
n XY \
y=o
= pv~v~\ y=o =0,
so that Co = Txy(O) = Tw , the skinfriction stress on the wall. As a result,
equation (5.4.30) can be rewritten as
(5.4.31 )
The stresses conditioned by the molecular friction Txy and the Reynolds
stresses n xy are found by formulas (5.4.15), (5.4.17) , and (5.4.18) and in
our case are equal to
du
nxy = A dy' (5.4.32)
du 2 2 (du)2
/i dy + P"" Y dy = Tw (5.4.33)
u= 
,."
1ffw lny+c.
p
(5.4.34)
On the contrary, near the wall the molecular friction plays the predom
inant role, Txy n xy , and we obtain from (5.4.33) the equation:
du
/i=Tw
dy
the solution of which is with regard for the boundary condition u!y=o =0
Tw
U= y, (5.4.35)
pv
5.4 Turbulent Fluid Flows 309
where we have set J1 = pl/. The nearwall region, where formula (5.4.35)
is valid, is called the viscous sublayer. We can find the viscous sublayer
thickness b from the condition Re = v.1i v ~ I, where v* = u(b) = !>.Lb.
pv
Substituting the expression for v* into the condition Re ~ I, we obtain:
(5.4.36)
Note that such a subdivision of the flow into a viscous sublayer and the
turbulent flow (the turbulent kernel of the flow) is sufficiently rough.
There is indeed a transitional region in which the molecular viscosity
effect gradually decreases as the distance from the wall increases. In
many problems, however, such a rough approximation is sufficient.
Assuming u = v*, y = b in formula (5.4.34), we find the constant
c ~ v*  v; lnb. Substituting this expression in (5.4.34), we find the
Prandtl formula:
(5.4.37)
u = v* In yv* . (5.4.38)
Ii 1/
References
j i
i'
from here that P = P(t, x), p = p(t, x). Assume that vx = vx(t, x).
Neglecting the terms Vy~ and vz~ in the second equation (6.1.1) as
compared with vx~, we can rewrite (6.1.1) in a simplified form
ov x OV x 10P
+ v 
x ox  pox'
at
op
+ op
Vx ox
(OVx oV y
+ p ox + oy + oz  ,
OVz) _ 0 (6.1.2)
at
p <1>( P).
J.Jr
F
(OP
at
op
+ Vx ox + P ox dS +
OVx) JJr
F
(OV y
P oy
ovz)
+ 8z dS = 0,
and taking into account the fact that the relation (* + Vx ~ + P~ )
does not depend on y and z, we find:
( OP
at
op
+ Vx ox + P ox
OVx)F Jr (OV y
+ JF P oy + oz
OVz)dS=O
.
OP op OVx) dF
F ( at + Vx ax + P ax + Pdt = O.
Assuming that the tube does not deform, that is cLJ: = Vx ~~, we can
write the above equation as
or
(6.1.3)
In the stationary flow case, we have from equations (6.1.2) and (6.1.3):
dv x 1 dP
vX d
x
+ d
p x
= 0, P = ~(P)
v2
; +
J dP
~(P) = B, P = ~(P) , (6.1.4)
Since the motion is steady, the formulas (6.1.4) give the general solution
of the problem of the gas flow in a variable section duct, where the
constant Q represents the gas flow rate in a fixed duct section. These
314 6 Gas Dynamics
M<lM=l M> 1
Po x
dF
dx
>0
~>o
dx
c2 dp dp Vx
vxdv x +   = 0 and  = "2. dv x .
p p c
dF + dv x _ Vx dv x =0
F Vx c2
or
dv x Vx dF
= (6.1.6)
dx (M2  l)F dx '
2
where M2 = ~ c
and the boundary conditions are x = Xo, Vx = v~, and
P = Po. The relation (6.1.6) enables us to draw a number of important
conclusions. We will assume for definiteness that Vx > O. The sign of
~ in (6.1.6) depends on the flow type.
1) Let M < 1; that is, Vx < c. Then in the case in which the area
F reduces, i.e., ~~ < 0, then ~ > 0, that is the velocity increases. If
6.1 OneDimensional Stationary Gas Flows 315
~; > 0, i.e., the section area increases, then ~ < 0; that is, the velocity
decreases.
2) Let M > 1, that is Vx > c. In this case, ~ < 0 if ~; < 0; that
is, the velocity decreases as the duct area decreases. If ~; > then
~ > 0; i.e., the duct area increase leads to the velocity increase.
3) At M = 1, it is necessary for the existence of a regular flow regime
that ~; = 0; that is, Vx = C in the minimal section. The satisfaction
of this condition is related to the need for surmounting a singularity at
point M = 1 of equation (6.1.6).
These considerations were put by Laval into the basis of the elemen
tary nozzle theory, on the basis of which it is possible to revert a subsonic
flow into a supersonic flow in a continuous way. The duct section should
at first have a convergent part, in which the flow velocity increases up to
the sound velocity in the minimal section, and then the duct should be
divergent; in this duct part, the flow accelerates. In the minimal section
V x = c, that is M = 1 (see Fig. 6.2) .
We now consider the onedimensional stationary isentropic gas mo
tion in the variable section duct. We will assume that the gas flow is
adiabatic and that the gas is perfect. Under these conditions, we have
with (2.1.97) in view that P = Ap', ~ = c2 , and c2 = "(. %= "(RoT.
We find from the Bernoulli integral (6.1.4): v22 + ~p,1 = B or
v2 "( P
+=B.
2 ,,(1 P
We have assumed here v = V x . It follows from the equation of state
P = (cp  cv)pT that %. , ~1 = cvT. This enables us to write the
Bernoulli integral as
v2
2 + cpT =B.
Choosing the constant B from the condition that T = To at v = 0, we
2
have: v2 +cpT = cpTo , where To is the temperature in the adiabatically
decelerated gas. Divide both sides of this equality by CpT. As a result,
we obtain:
To v2
T = 1 + 2cpT
Observing that 2~;T = . ~: , we can rewrite the above formula as
T,o '\11
 = 1+'M2 (6.1.7)
T 2
Recalling that the sound velocity is proportional to the square root of
temperature, we obtain:
Co ( "(  1 2) 1/ 2
= 1+M . (6.1.8)
c 2
316 6 Gas Dynamics
4
M / Mil
I 1
I I
l. I I
'....~ __ MO I
 _ + 1
 M'
123 x
Figure 6.3: The distributions of the pressure :. and the Mach number
M along the Laval nozzle (see Fig. 6.2).
(6.1.9)
. = PIVI
FI pv
= (1 ++ ~M2)
1 Mi ~
1
1'  1 Ml . Cl
M C
or after simple transformations
. = Ml .
FI M
(1 + l+~Mi
~M2) 2(,1)
2!....
(6.1.10)
F
F
.5..
Fo
1 M
M{ 1 M{'
Y
Yo
0.5
0.25
0.5 1.5 2 2.5
(a)
1. 75 7
1.5 6
1. 25 5
1 4
0.75
3
0.5
2
0 . 25
1 ._.  . 
0 0.5 1 1.5 2 2.5
o 0.5 1 1 .5 2 2.5 3
(b) (c)
4.5 I
.' \
4
0.8
3.5
0.6
2.5
0.4 2
1.5
0.2 1 ._.
0 0.5 1 1.5 2.5 3 0 0.5 1 1.5 2 2.5
(d) (e)
Figure 6.6: The quasionedimensional nozzle flows: (a) the Laval noz
zle geometry; (b) the profiles of the Mach number M(x) (solid line) ,
the pressure ratio P(x)/ PI (dashed line), the density ratio p(x)/ PI (dot
ted line), the temperature ratio T(x)/TI (white circles), supersonic flow
regime; (c) the profile of the gas velocity ratio v( x) / V1, supersonic flow
regime; (d) the profiles of M,P/P1,P/Pl,T/T1, subsonic flow regime;
and (e) the profile of V/Vl, subsonic flow regime.
320 6 Gas Dynamics
. Po, Po . P
Figure 6.7: The gas outflow from a reservoir through the Laval nozzle.
erning the nozzle wall geometry. One can obtain different nozzle shapes
by the variation of these parameters.
We show in Fig. 6.6 (a) the actual nozzle shape for which the quasi
onedimensional gas flow was computed with the aid of our program
prog61.nb [see Figs. 6.6 (b) (e)] . The profiles of M, Pj P1, pj P!' VjV1,
and TjT1, shown in Figs. 6.6 (b),(c) , correspond to the supersonic flow
regime in the diverging nozzle part. The program prog61.nb has the
input parameter reg, which enables the user to choose the subsonic or
supersonic flow regime in the diverging part of the Laval nozzle. We
show in Figs. 6.6 (d),(e) the profiles of M, PjP1, pjp!, VjV1, and TjT1
for the same Laval nozzle in the case of a subsonic flow regime.
Problem 6.1. A reservoir is filled with an ideal gas at rest with "( = 1.4
under the pressure Po and the gas density is Poi Vo = O. Find the
dependence of the gas flow rate Q on the pressure P at the nozzle exit
for the gas escaping from the reservoir through a Laval nozzle (see Fig.
6.7). The area of the nozzle exit crosssection is equal to S.
Co ( "(  1 2) 1/2
Po = l+M ,
c 2
and equals
Q
Qrnaxl~
FIFo
l..._ _ _'  _ _ '   _...
0.53 1
Figure 6.8: The dependence of the flow rate Q on the ratio of the pressure
at the nozzle exit to the pressure inside the reservoir.
where S is the area of the nozzle exit. The obtained formula is valid at
Q< Qrnax, after that Q = Qrnax = const (Fig. 6.8) at I = 1.4. The
appearance of this limitation Q ::; Qrnax is related to the flow choking.
In this case, at FIFo < 0.53 b = 1.4), we have M = 1 in the minimal
crosssection and M > 1 to the right of this crosssection, therefore, the
disturbances from the nozzle exit propagating upstream at the velocity
~~ = v  c cannot enter the reservoir and increase the gas flow rate.
[H + U2n ] = 0,
2
[PUn] = 0, [pu;, + F] = 0, rUT] = 0, (6.1.13)
(6.1.14)
(6.1.16)
(6.1.18)
(6.1.19)
At a given j, equation (6.1.19) [or (6.1.18)] together with the last equa
tion in (6.1.14) enables one to find the gas state behind the shock wave.
Assuming in (6.1.19) that P2 = P and V2 = V , let us present the Hugo
niot adiabat (6.1.19) passing through the point PI, VI by the line H in
Fig. 6.9. The last equation in (6.1.14) describes a straight line R (the
Rayleigh line) passing through the point PI, Vi. The intersection of the
Rayleigh line R and the Hugoniot adiabat H determines the state behind
the shock wave P2 , V2 (point 2 in Fig. 6.9).
Let us determine the mutual disposition of the Hugoniot adiabat H
and the Poisson adiabat Ps passing through the same point PI, VI (the
initial gas state before the compression). The equation for the Poisson
6.1 OneDimensional Stationary Gas Flows 323
Figure 6.9: The Hugoniot adiabat H, the Poisson adiabat Ps , and the
Rayleigh line R.
adiabat Ps follows from the ideal gas equation of state V = V(P, 8), in
which one must set 8 = const. Expanding the righthand side of this
equation in the neighborhood of PI, VI, let us rewrite the equation of
the Poisson adiabat in the form
where the derivatives are taken at P = PI and we have deleted the terms
of a higher order of smallness.
At a gas compression in the shock wave, the final gas state P, V lies at
the Hugoniot adiabat (6.1.18) and the gas entropy increases (see Section
3.3), i.e., 8> 8 1 . Consequently, we obtain for a weak shock wave from
the equation of state V = V (P, 8):
It can be seen that the first three items in (6.1.21) coincide with (6.1.20),
and the last item related to the entropy growth is absent in (6.1.20). We
have found previously in Problem 3.3 for the Hugoniot adiabat (6.1.18)
the entropy change (3.3.50) in a weak shock wave. Substituting (3.3.50)
in (6.1.21), we obtain the equation for the Hugoniot adiabat in the neigh
324 6 Gas Dynamics
+ ~(83V) (p_p)3
6 8p3 S 1
1 (8V) (8 2V)
+ 12Tl 8S P 8p2 s(P  Pt)
3
+ .. . , (6.1.22)
where the derivatives are taken at point V = VI. It follows from (6.1.25)
that a weak shock wave propagates through the gas particles at a sound
speed. In the case of a finite amplitude shock wave, the velocity of
its propagation over the gas particles is given by formula (6.1.24) from
which it follows that D2 = vl tg a , where a is the angle between the
Rayleigh line R and the av axis (see Fig. 6.9). The squared sound
velocity (6.1.25) is cI = Vltg;3, where ;3 is the angle between the
tangent to the adiabats at point PI, VI and the aVaxis [see Fig. 6.9)].
It follows from the concavity condition for the adiabats (~~) s > 0 that
tan a > tan;3 (see Fig. 6.9). Therefore, for the finite amplitude shock
cr
wave P 2 > Pi the inequality D2 > is valid. Since D = Ul , it follows
from here that uI > cI; that is, the flow ahead of the shock wave is
supersonic.
6.1 OneDimensional Stationary Gas Flows 325
\~
P
PI
1 V
VI
:r=! 1
'1'+1
Figure 6.10: The adiabats of Hugoniot and Poisson for a perfect gas.
Let us find the gas velocity U2 behind the shock wave front. We
obtain from the second and third equations of system (6.1.14):
2 2(&P) (6.1.27)
c2 = V2 &V 82
cp
E= PV , /'=  , cp  Cv = R. (6.1.28)
/,1 Cv
(!'+l)VI(!'l)V
(6.1.29)
(!'+l)V(!'l)VI
We show in Fig. 6.10 the Hugoniot adiabat (6.1.29) by a solid line, the
Poisson adiabat PV'1' = PI V? for a perfect gas by a dashed line. It can be
seen from Fig. 6.10 that the Hugoniot adiabat has a vertical asymptote
326 6 Gas Dynamics
Xsw = Dt
Figure 6.11: The trajectory of the piston xp(t) and of the shock wave
xsw(t) in the t, x plane.
VI Ul b+ I)Ml
V2 = U2 = b  I)Ml + 2'
P2 2, M2 ,I
PI ,+1 1  ,+1' (6.1.30)
T2 (2,Ml b  1))(b  I)Ml + 2)
=
Tl b + I)2Ml
2 + b  1 )Ml . M2 _ U~
M2
2 2,Ml b  1)' 2  cr
Problem 6.2. A piston begins to move at a velocity u into a tube filled
with an ideal gas at rest. The gas is under the pressure PI and has the
specific volume Vi. Find the shock wave velocity D and the pressure
behind the shock wave front P2
Solution: Let us represent the given process on the t, x diagram. We
show in Fig. 6.11 by a solid line the shock wave trajectory and by a
6.1 OneDimensional Stationary Gas Flows 327
dashed line the trajectory of a gas particle. It can be seen that the gas
velocity behind the shock wave front is equal to the piston velocity in
the laboratory frame. In the coordinate system comoving with the shock
wave front, we can use the jump relations (6.1.14) and the equation of
the Hugoniot adiabat (6.1.16):
PV
E=CvT =   .
,),1
Since the gas velocity jump is equal to u , we have u 2 = (P2Pt) (VI  V2)'
Dividing both sides of this equation by PI Vl and using the relation
we obtain:
( P2 _ 1) (1 _ V2) ci = u2. (6.1.33)
PI VI ')'
Using the shock adiabat (6.1.32), we find:
1 _ V2 = 2(P2  Pt)
(6.1.34)
VI (')'  l)PI + (')' + 1)P2
2 (P
ci 2 1 ) 2 P2)
(')'+l)u 2 (   1 2')'u 2 =0.
')' Pl PI
328 6 Gas Dynamics
M~
 ,
Cl
from where
(6.1.35)
In order to find D, we make use of the fact that the gas velocity before
the shock wave front in the coordinate system of the shock wave front is
equal to D. It follows from here that
(~ 1)P1 V1
(1 ~)
Taking into account equations (6.1.32) and (6.1.35), we obtain:
from where
(')' + 1)
+ 4
D=c1  U' '
6.1 OneDimensional Stationary Gas Flows 329
a a ( 2 av)
(pv)+
at ax pv +P1]ax =0, (6.1.41)
~
at
(P(E+ V2))
2
+~(pv(H+ V2)) _1]V av
ax 2 ax
_",aT)
ax
=0
'
330 6 Gas Dynamics
Po
o
Figure 6.12: The profile of the pressure P(x l ) in the shock wave propa
gating to the left: jj = Dii.
we can go over to the coordinate system K' comoving with the shock
wave front . Assume that in the coordinate system K' all parameters
depend only on the coordinate x' and do not depend on the time t. Since
the system (6.1.41) is invariant under the Galilean transformations, it
will have the same form in the system K'. Omitting in (6.1.41) all partial
derivatives with respect to time and replacing x 7 x', V 7 U, P 7 p',
and E 7 E' , we can write the equations of a viscous, heatconducting
gas in the coordinate system K' comoving with the shock wave front as
d ( , ) d ( '2 , du )
dx' P u = 0, dx' P u + P  TJ dx' = 0,
2
d ( ( H'+
p'u du dT')
U ) TJUK, =0, (6.1.43)
dx' 2 dx' dx'
p' = P'(p', E').
du
p'u = GI , p'u 2 + p'  TJ dx' = G2 ,
2) du dT'
GI ( H' + U  TJu  K, = G3. (6.1.44)
2 dx' dx'
In the coordinate system K' , we have U > 0, therefore, the integration
constants GI > 0, G2 > 0, and G3 > O. The first integral in (6.1.44)
expresses the mass conservation law. The second integral expresses the
6.1 OneDimensional Stationary Gas Flows 331
momentum conservation law The third integral expresses the energy con
servation law. After simple transformations, we can present the system
(6.1.44) in the form
1 0
7]
dV'
dx' 01 (or (V'  0:) +P'(V' , T')) =M~(V' , T') ,
K,dT' 0 1 (E' _ ~or(v' _0 2 )2 _ 0 3 + ~(02)2) (6.1.45)
dx' 2 01 0 1 2 01
M~(V' , T') ,
v

x
o
It follows from these relations that the functions UI, p~, P{ , T{; U2'P~'P~,
and T~ should satisfy the Hugoniot conditions. Resolving them, we find
that the points (VI, T I ) and (V2' T 2 ) exist in the (V, T) plane and lie on
the Hugoniot shock adiabat.
A detailed qualitative analysis ofthe solution of system (6.1.46) may
be found in 9 . The system of equations (6.1.46) with boundary conditions
(6.1.47) has an analytic solution for some particular cases. Consider
a wellknown solution found by Becker. Let the equation of state of
a viscous, heatconducting gas coincide with the ideal gas equation of
state. Then we have in the nondimensional variables:
1