Waves in Continuous Media PDF

Lecture Notes in Geosystems Mathematics
and Computing
S.L. Gavrilyuk
N.I. Makarenko
S.V. Sukhinin
Waves in
Continuous
Media
Lecture Notes in Geosystems Mathematics
and Computing
Series editors
W. Freeden, Kaiserslautern
Z. Nashed, Orlando
O. Scherzer, Vienna
More information about this series at http://www.springer.com/series/15481
S.L. Gavrilyuk N.I. Makarenko
S.V. Sukhinin
Waves in Continuous Media

S.L. Gavrilyuk N.I. Makarenko
Aix-Marseille University Russian Academy of Sciences
Marseille, France Lavrentyev Institute of Hydrodynamics
Novosibirsk State University
Novosibirsk, Russia
S.V. Sukhinin
Russian Academy of Sciences
Lavrentyev Institute of Hydrodynamics
Novosibirsk State University
Novosibirsk, Russia
Lecture Notes in Geosystems Mathematics and Computing

ISBN 978-3-319-49276-6 ISBN 978-3-319-49277-3 (eBook)
DOI 10.1007/978-3-319-49277-3
Library of Congress Control Number: 2017930812
Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
Printed on acid-free paper
This book is published under the trade name Birkhuser, www.birkhauser-science.com

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Wave phenomena occur everywhere in nature and therefore are studied in many
areas of science for a long time.
The mathematical wave theory emerged as an independent discipline in the mid-
1970s due to numerous applications in natural science and engineering stimulating
the further development of mathematical methods.
The lecture course Waves in Continuous Media is one of the disciplines on
continuum mechanics and mathematical modeling included into the education
program at the Department of Mechanics and Mathematics, Novosibirsk State
University. This course was first given by Professor L. V. Ovsyannikov,1 a distin-
guished scientist who obtained a number of fundamental results in the field of wave
hydrodynamics. Based on Ovsyannikovs principles of selecting the material, the
authors developed new variants of the course adapted to groups of masters students
specialized in applied mathematics, mechanics, and geophysics.
The textbook contains a rich collection of exercises and problems which have
been carefully selected and tested at practical works and seminars of courses
given by the authors at Novosibirsk State University (Russia) and Aix-Marseille
University (France) for many years. Most of the problems and exercises are supplied
with answers and hints. Solutions of some typical problems are explained in detail,
and some theoretical background material is included in order to make the book
self-contained and give students the necessary tools for self-education. More than
200 problems formulated in the book allowed us to propose to each masters student
an individual semester mini-project consisting in solving up to six problems. Most
of them are solved by applying the theoretical approaches from the course, but the
other ones demand a deeper understanding of the methods discussed in the course.
During the semester, the students have also been working in research laboratories,
so a set of problems specific to the research activity of the students was usually
proposed.
1
Ovsyannikov, L. V.: Wave Motions of Continuous Media. Novosibirsk State University, Novosi-
birsk (1985) [in Russian].
v
vi Preface
The textbook consists of three chapters. Chapters 1 and 2 present the basic
notions and facts of the mathematical theory of waves illustrated by numerous
examples and methods of solving typical problems. The reader learns how to
recognize the hyperbolicity property; find characteristics, Riemann invariants, and
conservation laws for quasilinear systems of equations; construct and analyze solu-
tions with weak or strong discontinuities; and investigate equations with dispersion:
analysis of dispersion relations, the study of large time asymptotic behavior of
solutions, the construction of traveling wave solutions for models reducible to
nonlinear evolution equations, etc. The majority of problems are formulated within
the framework of wave models arising in gas dynamics, magnetohydrodynamics,
elasticity and plasticity, linear and nonlinear acoustics, chemical adsorption, and
other applications.
Chapter 3 deals with surface and internal waves in an incompressible fluid. The
efficiency of mathematical methods is demonstrated on a hierarchy of approximate
submodels generated from the Euler equations of homogeneous and inhomogeneous
fluids. Some problems illustrate the influence of viscosity and vorticity on the wave
processes.
The list of references consists mainly of monographs and textbooks recom-
mended for further reading. Three of them are generic [13], while others [433]
are more specific for each chapter. These have been selected to allow readers to
understand better mathematical statements whose proofs were skipped, and find
solutions of relatively hard exercises. A separate bibliography for each chapter is
maintained. The reference list for Chap. 3 also contains five research articles on the
theory of water waves [16, 17, 21, 26, 31] we explicitly refer to. The books for
further reading are not cited in the text.
The authors thank their colleagues at the Chair of Hydrodynamics, Novosibirsk
State University, for their help in the preparation of the manuscript.
The authors would like to express their special gratitude to Professor
V. M. Teshukov, who recently passed away, for his numerous useful advices and
discussions.
Marseille, France S.L. Gavrilyuk

Novosibirsk, Russia N.I. Makarenko
Novosibirsk, Russia S.V. Sukhinin
October 2016
Contents
1 Hyperbolic Waves .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.1 Hyperbolic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 Propagation of Weak Discontinuities . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4
1.3 Motion with Strong Discontinuities . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
1.4 Kinematic Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10
1.5 Multi-dimensional Wave Fronts. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13
1.6 Symmetrization of Hyperbolic Systems
of Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 16
1.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
2 Dispersive Waves .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.1 Dispersion Relation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.2 Multi-dimensional Wave Packets . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 46
2.3 Group Velocity .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 49
2.4 Stationary Phase Method .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 51
2.5 Nonlinear Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 56
2.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 60
3 Water Waves .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 77
3.1 Equations of Motion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 77
3.2 Linear Theory of Surface Waves . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 82
3.3 Shallow Water Theory .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.4 Shear Flows of Shallow Water . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 88
3.5 Nonlinear Dispersive Equations . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 90
3.6 Stationary Surface Waves . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 94
3.7 Waves in Two-Layer Fluids . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 97
3.8 Waves in Stratified Fluids . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 103
3.9 Stability of Stratified Flows . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 107
vii
viii Contents
3.10 Stationary Internal Waves . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109

3.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 112
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 137
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
Chapter 1
Hyperbolic Waves
1.1 Hyperbolic Systems
We consider the quasilinear system of first order equations
ut C A.u; x; t/ux C b.u; x; t/ D 0; (1.1)
where the n n-matrix A and vector b depend on x, t and u D .u1 ; : : : ; un /T . A

direction dx=dt D c is called characteristic if there exists a linear combination of
equations of the form (1.1) such that each unknown function ui is differentiable
along this direction. The quantity c in the definition of a characteristic direction is
an eigenvalue of the matrix A, i.e.,
det .A cI/ D 0: (1.2)
For any eigenvalue c and the corresponding left eigenvector l D .l1 ; : : : ; ln / of the
matrix A (i.e., lA D cl) the system (1.1) implies the following condition on the
characteristic (the curve corresponding to a characteristic direction):
l .dt u C b/ D 0; (1.3)
where dt D @t C c@x is the operator of differentiation along the characteristic.

The system (1.1) is hyperbolic if all eigenvalues ci of the matrix A are real (in this
case, they can be ordered: c1 6 c2 6 : : : 6 cn ) and there exist n linearly independent
real left eigenvectors of the matrix A.
Figure 1.1 shows the location of characteristics emanating from a given point
M in the .x; t/-plane. A hyperbolic system of equations is equivalent to a system
of n relations on characteristics. A system is hyperbolic if and only if the normal
Jordan form of the matrix A is diagonal. We indicate the sufficient hyperbolicity
Springer International Publishing AG 2017 1

S.L. Gavrilyuk et al., Waves in Continuous Media, Lecture Notes in Geosystems
Mathematics and Computing, DOI 10.1007/978-3-319-49277-3_1
2 1 Hyperbolic Waves
Fig. 1.1 Location of

characteristics emanating
from a given point M in the
.x; t/-plane
conditions:
(a) the matrix A is symmetric,
(b) all roots of Eq. (1.2) are real and distinct.
In case (b), where the matrix A has no multiple eigenvalues, the system (1.1) is
called strictly hyperbolic.
Example 1.1 The process of chemical adsorption used for separating substances in
a liquid or gas mixture by the chromatography method is described by the equations
@t .u C f.u// C v@x u D 0; (1.4)
where u D .u1 ; : : : ; un /T are the concentrations of the separated substances passing

through the adsorption column, f.u/ D . f1 .u/; : : : ; fn .u//T are the concentrations
of the substances adsorbed by the adsorbent, and v D const > 0 is the mixture
velocity. Let the vector-valued function f.u/, called the adsorption isotherm, be such
that all eigenvalues of the Jacobi matrix
@. f1 ; : : : ; fn /
f0 .u/ D
@.u1 ; : : : ; un /
are real, positive, and distinct:
0 < 1 < : : : < n :
Then Eqs. (1.4) can be transformed to the form (1.1) with the matrix A.u/ D v.I C
f0 .u//1 and vector b D 0. Since
A cI D ..v c/I cf0 .u//.I C f0 .u//1 ;

1.1 Hyperbolic Systems 3
the eigenvalues of the matrix A are connected with the eigenvalues of the matrix
f0 .u/ by the identity
v
cj D . j D 1; 2; : : : ; n/:
1 C j
Consequently, the system (1.4) is strictly hyperbolic; moreover, all its characteristic
velocities are positive and do not exceed the mixture velocity. The noncoincidence
of the velocities ci cj (i j) is the basis of the chromatography method.
If there exist scalar functions r.u/ and .u; x; t/ such that
@r
D li .i D 1; : : : ; n/;
@ui
then the relation (1.3) is equivalent to the equation
dt r.u/ D l b;
where r.u/ is called a Riemann invariant. The motivation of this definition becomes
clear in the case l b D 0, where the Riemann invariant r is constant along
characteristics. Riemann invariants always exist for the system (1.1) consisting of
one or two equations and for the system (1.1) with constant matrix A of arbitrary
order n (in the second case, r D l u). In the general case n > 3, Riemann invariants
do not necessarily exist. In the case n D 3, the identity l curl l D 0 is necessary and
sufficient for the existence of a Riemann invariant for the characteristic dx=dt D c
corresponding to a simple eigenvalue c of the matrix A with eigenvector l.u/ D
.l1 ; l2 ; l3 /.
Problem 1.1 Find characteristics and Riemann invariants for the system describing
shallow water flows over the flat bottom
ht C .uh/x D 0;
(1.5)
ut C uux C ghx D 0;
where h.x; t/ is the layer depth, u.x; t/ is the horizontal fluid velocity, and g is the
acceleration of gravity.
Solution We compose the matrix of coefficients of the original system of equations

uc h
A cI D :
g uc
p
Then we find the characteristic velocities c D u gh. The system is hyperbolic
in the domain h > 0. For the characteristic dx=dt D cC the left eigenvector, defined
p p
up to an arbitrary scalar factor, has the form l D . g; h/. Consequently, to find
the Riemann invariant r.h; u/, we should find a solution to the system of equations
@r p
D g;
@h
@r p
D h;
@u
where .h; u/ is an unknown integrating factor. Excluding this factor, we obtain the
linear first order partial differential equation for r
s
@r h @r
D 0:
@u g @h
p
p equation of characteristics du D g=hdh we find the first integral r D
From the
u C 2 gh. Since there is a certain functional arbitrariness in the definition of a
Riemann invariant, the obtained first integral can be taken for the sought invariant.
The characteristic dx=dt D c is studied in a similar way.
Answer:
dx p p
D u C gh W u C 2 gh D const ;
dt
dx p p
D u gh W u 2 gh D const :
dt
1.2 Propagation of Weak Discontinuities
The Cauchy problem for the system (1.1) is to find a solution for t > t0 provided
that ui .x; t0 / D ui0 .x/ is given at t D t0 .
Theorem 1.1 (uniqueness) Assume that the system (1.1) is hyperbolic and the
coefficient matrix A and vector b are continuously differentiable. Let a continuously
differentiable solution u.x; t/ be defined in the characteristic triangle X1 MXn
(cf. Fig. 1.2). If u is another continuously differentiable solution to the system (1.1)
in X1 MXn and u D u on the segment Xn X1 , then u D u in the entire characteristic
triangle X1 MXn .
This theorem implies the existence of wave fronts defined by the characteristics
X1 M and Xn M and defining the domain of determinacy of the solution to the Cauchy
problem by the initial data solely at the domain of dependence Xn X1 of the point M.
Let a domain D be divided by a smooth curve W x D .t/ into two subdomains
D and DC (cf. Fig. 1.3).
1.2 Propagation of Weak Discontinuities 5
Fig. 1.2 The solution to the

Cauchy problem for the
system (1.1) is uniquely
determined by the initial data
on the interval Xn X1 inside
the curvilinear characteristic
triangle X1 MXn
Fig. 1.3 The weak

discontinuities always
propagate along the
characteristic curves
We assume that the solution to the hyperbolic system is continuous in the closed
domain D and continuously differentiable in the closures D and DC . Moreover,
the derivative @x u D v of the solution on can have a discontinuity of the first
kind with jump v D vC v . By the continuity of the solution u, the jump of its
tangent derivative dt u D @t u C 0 .t/@x u on vanishes, which implies the following
expressions for the jumps of derivatives:
@x u D v;
@t u D 0 v:
Consequently, from the system (1.1) we have
.A 0 I/v D 0:
Thus, the derivative of the solution can be discontinuous only on the characteristic;
moreover, the jump is a right eigenvector of the matrix A. In the case of a simple
eigenvalue of the matrix A, the amplitude of the weak discontinuity is characterized
by a scalar such that v D r, where r D .r1 ; : : : ; rn /T is the corresponding
right eigenvector. The quantity satisfies the ordinary differential equation along
the characteristic
d
.l r/ C P C Q 2 D 0; (1.6)
dt
where l is the corresponding left eigenvector, whereas P and Q are known functions.
In particular,
X
n @a
ij
QD li rk rj ;
i;j;kD1
@uk
where aij are components of the matrix A. Without loss of generality we can assume
that .l r/ D 1. The relation (1.6) is the Riccati equation. It is called the transport
equation for the amplitude of the weak discontinuity.
Problem 1.2 Consider the system of equations describing the isentropic motion of
a polytropic gas, written in terms of the Riemann invariants r and l,
(
rt C .u C c/rx D 0; 2 2
r DuC c; lDu c;
lt C .u c/lx D 0; 1 1
with the initial conditions

(
0; x > a;
u.x; 0/ D c.x; 0/ D c0 ;
c0 .x a/=.l0 C a x/; x < a;
where a D const, c0 D const, and l0 D const (c0 > 0, l0 > 0). Compute the jump
ux of the derivative on the characteristic x D c0 t C a at time t.
Solution By the uniqueness theorem for the Cauchy problem, u.x; t/ 0 and
c.x; t/ c0 for x > c0 t C a. Further, for the Riemann invariant l along the
characteristic dx=dt D u C c we have
lt C .u c/lx D 0;
lt C .u C c/lx D 0;
where first relation immediately follows from the equation of motion, whereas the
second one is obtained from the continuity of the tangent derivative of l on the weak
discontinuity line. Therefore, along the characteristic under consideration, we have
lt D 0, lx D 0, and, as a consequence, ux D rx =2. Differentiating the first
equation of the original system with respect to x, considering the jump, and taking
into account the above properties lx .x 0; t/ D lx .x C 0; t/ D 0 for x D c0 t C a, we
1.3 Motion with Strong Discontinuities 7
obtain the transport equation
drx C 1 2
rx D 0
dt 4
with the condition rx D c0 =l0 at t D 0 which follows from the initial data.
Integrating the equation, we obtain an expression for the jump ux .
Answer:
c0
ux D C1
:
2l0 C 2
c0 t
1.3 Motion with Strong Discontinuities
The construction of solutions describing the shock wave propagation is based on the
conservation laws
@t '.x; t; u/ C @x .x; t; u/ D f.x; t; u/; (1.7)
where ' D .'1 ; : : : ; 'n /T , D . 1 ; : : : ; n /T , and f D . f1 ; : : : ; fn /T are n-

dimensional vectors interpreted as densities, fluxes, and sources of the sought
quantities. Not every system of differential equations (1.1) can be written in the
conservative form (1.7). For models of continuous media this property automatically
follows from the main equations written as a system of integral conservation laws
on an arbitrary interval x1 ; x2 :
Zx2 xDx2 Zx2

d
'.x; t; u/dx C .x; t; u/ D f.x; t; u/dx: (1.8)
dt xDx1
x1 x1
Systems of the form (1.1) representable as systems of independent conservation

laws with nonzero Jacobian j@'=@uj 0 are called conservative. In many cases,
the number of conservation laws is larger than the number of unknowns. In such a
situation, a right choice of a system of conservation laws for describing motions
with strong discontinuities should be done with taking into account additional
properties (stability of solutions, existence of a discontinuous solution as the limit of
smooth solutions, physical interpretation of the limit of smooth solutions, physical
interpretation of conservation laws, and so on).
Problem 1.3 Find all scalar conservation laws @t '.u; v/ C @x .u; v/ D 0, where
the density ' is a polynomial in v of degree at most 2, for the system of nonlinear
elasticity
ut D vx ;
0 vt D x ;
D .u/;
0 D const :
Solution According to the definition of a conservation law, we have
'u ut C 'v vt C u ux C v vx D 0:
Excluding the derivatives ut and vt by using the original system and then collecting
and equating to zero the coefficients at ux and vx , we obtain the system of equations
for ' and
0 u C 0 .u/'v D 0;
v C 'u D 0:
Applying the cross differentiation, we exclude the function and obtain the linear
second order partial differential equation for the density '
0 'uu D 0 .u/'vv :
In this problem, it is required to find all solutions of the form
' D .u/v 2 C .u/v C .u/
to this equation. For the coefficients of the above polynomial we have the relations
00 D 0, 00 D 0, and 0 00 D 2 0 . Integrating these relations, we obtain the
solution
' D C1 '1 C : : : C C5 '5 ;
where Cj are arbitrary real constants and 'j are the basic densities specified
below.
Answer:
'1 D u; '2 D 0 v; '3 D uv;

Zu Zu
1 1
'4 D 0 v 2 C .
/d
; '5 D 0 uv 2 C .2
u/.
/d
:
2 2
0 0
1.3 Motion with Strong Discontinuities 9
Fig. 1.4 The solutions can

have a discontinuity of the
first kind on the line x D X.t/
Assume that a solution u has a discontinuity of the first kind on the line x D
X.t/ and is smooth on both sides of this line. We fix integration limits x1 and x2
in the integral conservation law (1.8) such that x1 < X.t/ < x2 at a given time t
(cf. Fig. 1.4). Dividing the integration interval by the point x D X.t/ into two parts,
differentiating the obtained integrals with respect to the time, and passing to the
limit as x1 ! x2 , we obtain the relations on the strong discontinuity, called the
RankineHugoniot conditions,
D' D ; (1.9)
where D D dX.t/=dt is the velocity of the strong discontinuity. We fix a state on one
side of the discontinuity given by the point u D u0 in Rn . Then, on the other side of
the wave, the locus of admissible states is the curve in Rn given by (1.9). This curve,
called the shock adiabat, can consist of several smooth branches passing through the
center u0 .
Example 1.2 Consider the system (1.7) of conservation laws in nonlinear elasticity
with ' D .u; v/, D .v; .u/= 0 /, and f D 0. We assume that the stress function
.u/ is such that 0 .u/ > 0, 00 .u/ < 0, .0/ D 0, and 0 .0/ D C 2 (here,
> 0 and > 0 are the Lam constants). We write the relation (1.9) on the strong
discontinuity connecting the states .u0 ; v0 / and .u; v/ and exclude the wave velocity
D. As a result, we obtain the shock adiabat equation
0 .v v0 /2 D f.u/ .u0 /g.u u0 /: (1.10)
This curve in the .u; v/-plane (u is the strain and v is the velocity of the material)
is the locus of states obtained after the shock wave passes through the state .u0 ; v0 /.
The shock adiabat (1.10) consists of two branches describing the shock propagating
to the left (v < v0 ) and to the right (v > v0 ). Respectively, for the wave velocity we
obtain the expression

s
1 .u/ .u0 /
DD :
0 u u0
We consider an elastic semi-infinite rod x > 0 in the equilibrium state under loading
0 D .u0 /. When the loading is suddenly removed from the end x D 0, the rod
comes back to the unstrained state u D 0, D 0 due p to the unloading shock
wave travelling to the right with the velocity D D .u0 /=. 0 u0 /. For small
initialpstrains u0 the shock wave velocity is approximately equal to the velocity
c0 D . C 2/= 0 of the linear longitudinal elastic wave.
1.4 Kinematic Waves
By kinematic waves we mean the class of one-dimensional motions of a continuous

medium with a given dependence q D Q. / of the flow on the density. The
knowledge of such a dependence allows us to obtain a closed model of motion by
using only the law of conservation of mass
t C qx D 0: (1.11)
For particular media the relation between the flow and density is usually found
from experiments or by integrating other equations of a more general model. In
the kinematic-wave approximation, the description of motion is reduced to finding
solutions to the quasilinear first order partial differential equation
t C c. / x D 0;
where c. / D Q0 . / is the characteristic velocity.

Example 1.3 Consider the model of kinematic waves in a one-way traffic flow. A
twice continuously differentiable function Q on 0 6 6 is characterized by the
properties
(a) Q. / > 0 .0 < < /;

(b) Q00 . / < 0; (1.12)
(c) Q.0/ D Q. / D 0:
A typical graph of Q. / is shown in Fig. 1.5.

In this model of a continuous medium, the velocity of particles (individual
vehicles) is equal to u D Q. /= . The value > 0 yields the limit vehicle density
on the highway when the vehicles stand bumper to bumper, so that no motion is
1.4 Kinematic Waves 11
Fig. 1.5 Flow-density

relationship
possible in view of the second identity in (1.12c). On the other hand, according to
the first identity in (1.12c), lim Q. /= > 0 exists and is equal to the velocity of
!0
the motion over the free highway. From (1.12b) it follows that c0 . / < 0, so that the
characteristic velocity c. / is a monotonically decreasing function of the density
. Perturbations with abrupt fronts can propagate through moving vehicles when a
sudden deceleration occurs somewhere in the traffic flow.
In the case of the conservation law (1.11), the relation on the strong discontinuity
has the form
D D q
or, in detail,
Q. 2 / Q. 1 /
DD : (1.13)
2 1
The Riemann problem for the kinematic wave equation is the Cauchy problem with
piecewise constant initial data
(
1 ; x < 0;
.x; 0/ D
2 ; x > 0;
where i D const, 1 2 . This problem has a solution in the class of self-similar

motions D .x=t/ with strong and weak discontinuities. For 1 < 2 the solution
is piecewise constant and has a strong discontinuity on the line x D Dt, where D is
given by (1.13). According to this formula, we have
Z 2
1
DD c. /d ;
2 1
1
which implies c1 > D > c2 , ci D Q0 . i /, provided that c. / is monotonically

decreasing. The strong discontinuities satisfying the above condition are stable.
If 1 > 2 , then the velocities of shock wave propagation satisfy the opposite
inequalities c1 < D < c2 which means that the strong discontinuities are unstable
under small perturbations of the initial data. In this case, there exists a continuously
stable solution representable as a centered wave with the following distribution of
the characteristic velocity:
8

<c1 ;
x < c1 t;
c.x; t/ D x=t; c1 t < x < c2 t;

:c ; x > c2 t:
2
Thus, the choice of a stable solution to the Riemann problem for the kinematic wave
equation is determined by the sign of the difference D 1 2 .
Problem 1.4 The traffic flow moves with velocity u0 > 0 and density 0 along a
street where the traffic light, located at the point x D 0, turns red at time t D 0. In
the kinematic-wave approximation, describe the initial stage of the traffic flow for
t > 0 in a neighborhood of the traffic light.
Solution In the motion for t > 0 at the left from the traffic light, the following two
states should be connected: the incoming traffic flow with velocity u0 far from the
traffic light and the convoy of immobile vehicles with density in close vicinity
of the traffic light. This situation is simulated by the problem with discontinuous
initial-boundary conditions at the point x D 0, t D 0
.x; 0/ D 0 .x < 0/; .0; t/ D :
Since 0 < , a shock wave appears in motion (the wave caused by a sudden halt
of vehicles) which propagates in the direction opposite to the traffic flow direction.
By formula (1.13) and the property (1.12c), the wave velocity is equal to D0 D
Q. 0 /=. 0 / < 0. This quantity is the slope of the chord joining the points
. 0 ; Q. 0 // and . ; 0/ (cf. Fig. 1.5). Further, on a road part at the right from the
traffic light, there are no vehicles for t > 0. This state should be agreed with the
outgoing traffic flow of vehicles passed by the traffic light before turning on the red
light. This situation is described by the problem with discontinuous data .x; 0/ D
0 .x > 0/, .0; t/ D 0. Here, a shock wave appears that travels to the right with
velocity D D .Q. 0 / Q.0//= 0 . Since Q.0/ D 0, the wave front corresponds
to the location of the tail of the convoy of vehicles outgoing from the cross-road
with constant velocity u0 D Q. 0 /= 0 . The quantity u0 is the slope of the chord
joining the points .0; 0/ and . 0 ; Q. 0 // in Fig. 1.5. The above-described stage of
the traffic motion near the traffic light corresponds to a part of .x; t/diagram in
Fig. 1.6 observed on the time interval 0 < t < Tk of the red light.
1.5 Multi-dimensional Wave Fronts 13
Fig. 1.6 .x; t/ - diagram

showing the traffic motion
near traffic lights
Answer:
u D u0 ; D 0 .x < D0 t/; u D 0; D .D0 t < x < 0/;

u D 0; D 0 .0 < x < u0 t/; u D u0 ; D 0 .x > u0 t/:
1.5 Multi-dimensional Wave Fronts
Consider the linear system of equations
3
X
@t u C Ai @xi u C Bu D 0 (1.14)
iD1
for an n-dimensional vector u D .u1 ; : : : ; un /T , where Ai .x; t/ and B.x; t/ are given
n n-matrices depending on t and x D .x1 ; x2 ; x3 /T . We consider the hyperplane

1 x1 C
2 x2 C
3 x3 C t D const
in R4 with normal vector D .

1 ;
2 ;
3 ; /T . A direction is called characteristic
if
3
!
X
det I C
i A D 0:
i
(1.15)
iD1
The system (1.14) is said to be hyperbolic at a point .x; t/ if for any D .

1 ;
2 ;
3 /T
Eq. (1.15) has n real roots k D Hk .I x; t/ (k D 1; : : : ; n) and the characteristic
matrix
3
X
A./ D I C
i Ai
iD1
has n linearly independent vectors l 2 Rn such that l A./ D 0.

A smooth hypersurface in R4 such that the tangent hyperplane at each point has
the characteristic direction is called a characteristic of the system (1.14). Let the
characteristic corresponding to the root D H.I x; t/ be given by the equation
'.x1 ; x2 ; x3 ; t/ D 0:
Then its normal D .'x1 ; 'x2 ; 'x3 ; 't /T has the characteristic direction. Therefore,
' satisfies the nonlinear first order partial differential equation, called the Hamilton
Jacobi equation,
't D H.rx 'I x; t/: (1.16)
Characteristics of this equation are called bicharacteristics (rays in wave theory) of

the system (1.14). The differential equations for bicharacteristics have the form of
the Hamilton system of equations
dxi @H
D ;
dt @pi
dpi @H
D
dt @xi
where i D 1; 2; 3, pi D @'=@xi . For the linear hyperbolic system (1.14) with

constant matrices Ai the function H is independent of x and t, and, consequently,
dpi =dt D 0, so that the right-hand sides Hpi . p1 ; p2 ; p3 / of the differential equations
for xi are constant along the rays. Consequently, the rays themselves are rectilinear:
.0/
xi D xi Hp.0/
i
t:
Problem 1.5 The front of a two-dimensional acoustic wave propagating through a

gas at rest with the sound speed c0 has at t D 0 the shape of a parabola y D x2 . Find
the time T at which the sound reaches the observer at the point A.3; 0/.
1.5 Multi-dimensional Wave Fronts 15
Solution The acoustic equations for a gas at rest are written in the form
0 ut C px D 0;
0 vt C py D 0;
pt C 0 c20 .ux C vy / D 0;
where u and v are components of the velocity vector and p is the pressure
perturbation. We write this system in the matrix form (1.14) and compose the
characteristic determinant with the normal vector D .
; ; /T :

0
= 0

det .I C
Ax C Ay / D 0 = 0 D . 2 c20 .
2 C 2 // D 0:
c2
c2
0 0 0 0
p
Hence for the sound characteristics we have H. p; q/ D c0 p2 C q2 , where p D
'x , q D 'y ('.x; y; t/ D 0 is the location of the front at time t). Since the function H
is independent of x and y, the bicharacteristics are rectilinear. Integrating, we obtain
the equations of acoustic rays in the form
p 0 c0 t
x D x0 q ;
p20 C q20
q 0 c0 t
y D y0 q :
p20 C q20
Consequently, at each time moment, the bicharacteristics are directed along the
normal to the front; moreover, perturbations propagate along the rays with constant
velocity c0 . This means that the first perturbation reaching the observer is that
outgoing from the point B.x; x2 / lying on the initial location of the front and is
the nearest to A. The minimum of distance jABj is attained at a point B with the
x-coordinate satisfying the extremum condition 2x3 C x 3 p D 0. Such a point is
unique and is B.1; 1/. The corresponding distance is equal to 5.
p
Answer: T D 5=c0 .
If the Hamiltonian H is independent of t, then the equation of characteristic
surfaces can be looked for in the form t D .x/. By Eq. (1.16) with ' D .x/ t,
the function satisfies the equation
H.rx ; x/ D 1;
called the eikonal equation. In this case, the system for bicharacteristics is the
autonomous system
dxi @H
D ;
ds @pi
dpi @H
D ;
ds @xi
where i D 1; 2; 3, pi D @ =@xi , and s is the parameter along the bicharacteristics.
1.6 Symmetrization of Hyperbolic Systems of Conservation

Laws
Consider the system of quasilinear equations
3
X
A.u/@t u C Bi .u/@xi u D 0 (1.17)
iD1
for a vector-valued function u D .u1 ; : : : ; un /T depending on t and x D .x1 ; x2 ; x3 /T .

It is assumed that the square n n-matrices A.u/ and Bi .u/ are symmetric and the
matrix A D .aij /ni;jD1 is positive definite:
X
n
aij pi pj > 0
i;jD1
for any vector p D . p1 ; : : : ; pn /T , p 0. The system (1.17) with such matrices

is called a symmetric t-hyperbolic in the sense of Friedrichs system (or simply t-
hyperbolic Friedrichs system). In particular, the system (1.1) with two independent
variables t, x and a symmetric matrix A is a symmetric t-hyperbolic Friedrichs
system.
The t-hyperbolicity and conservation properties play an important role in the
analysis of qualitative properties of solutions and numerical solving of such systems.
Therefore, it is useful to know how to reduce a given hyperbolic system of
quasilinear equations to the symmetric form (1.17). Such a reduction is possible
in the case described by the following assertion.
Theorem 1.2 (Godunov, Friedrichs, and Lax) Let the system of conservation
laws
3
X
@t u C @xi i
.u/ D 0 (1.18)
iD1
1.6 Symmetrization of Hyperbolic Systems of Conservation Laws 17
admits the additional conservation law

3
X
@t e.u/ C @xi f i .u/ D 0; (1.19)
iD1
where the function e.u/ is convex in the variables u D .u1 ; : : : ; un /T , i.e., the
Hessian matrix e00 .u/ D .@ui @uj e.u//ni;jD1 is positive definite. Then the system (1.18)
is reduced to the form (1.17).
Proof The simultaneous validity of Eqs. (1.18) and (1.19) implies the following
compatibility condition:
ru f i .u/ D ru e.u/. i
.u//0 .i D 1; : : : ; n/ (1.20)
i
for scalar-valued functions f i and vector-valued functions . We introduce the
Legendre transform e .v/ of a function e.u/ by the formula
e .v/ D v u e.u/;
where the vector u is implicitly defined by the equation v D ru e.u/. It is possible

to invert the dependence of v on u since the Jacobian
@.v ; : : : ; v /
1 n
D det e00 .u/
@.u1 ; : : : ; un /
differs from zero for convex functions e.u/. It is obvious that u D rv e .v/.
Furthermore, e .v/ is a convex function of v. We introduce the functions
f i .v/ D v i
.u/ f i .u/ .i D 1; : : : ; n/:
By (1.20), we have i
.u/ D rv f i .v/. Consequently, the system (1.18) can be written
as
3
X
@t rv e .v/ C @xi rv f i .v/ D 0;
iD1
i.e., in the form (1.17) with the Hessian matrices A.v/ D .@vk @vj e .v//nk;jD1 and
Bi .v/ D .@vk @vj f i .v//nk;jD1 . t
u
The proof of this theorem contains a constructive symmetrization method for
systems of conservation laws.
Example 1.4 Consider the system of equations with two independent variables t
and x
ht C .uh/x D 0;

2 1 2 (1.21)
.hu/t C hu C gh D 0:
2 x
This system is the conservative form of the shallow water equations (1.5) for
the laws of conservation of mass and momentum. For the additional conservation
law (1.19) we take the law of conservation of energy
@t e.h; u/ C @x f .h; u/ D 0
with the functions

1 2 1
eD u h C gh2 ;
2 2
1
f D u3 h C guh2 :
2
We first write Eqs. (1.21) in the original form (1.18) with the vector of conservative
quantities u D .u1 ; u2 /, where u1 D h and u2 D uh. In this notation, the functions
e, f and the components of vector D . 1 ; 2 / are expressed as
u22 1
e.u/ D C gu21 ;
2u1 2
u32
f .u/ D C gu21 u2 ;
2u21
u22 1
1 .u/ D u2 ; 2 .u/ D C gu21 :
u1 2
Since the function e.u/ is convex in u, we have
u22
v1 D eu1 D gu1 ;
u21
u2
v2 D eu2 D :
u1
Inverting the dependence v D ru e.u/, we find

1 1 2
u1 D v1 C v2 ;
g 2

1 1
u2 D v2 v1 C v22 :
g 2
1.7 Problems 19
Consequently, the functions e D v u e and f D v f take the form

1 1 2 2
e .v/ D v1 C v2 ;
2g 2

1 1 2 2
f .v/ D v2 v1 C v2 :
2g 2
Computing the Hessian matrices for these functions, we find the symmetric form
of the system (1.21)
A.v/vt C B.v/vx D 0
with matrices

1 v2 v2 v1 C 32 v22
AD ; BD :
v2 v1 C 32 v22 v1 C 32 v22 3v1 C 52 v22
There are other symmetric forms of hyperbolic systems of equations. For

example, for systems with two independent variables one can obtain such a form by
transformation of the original equations to equations written in Riemann invariants
(if they exist). Thus, for the shallow water equations (1.5) we have
p
rt C .u C gh/rx D 0;
p
lt C .u gh/lx D 0;
p p
where r D u C 2 gh and l D u 2 gh. However, this approach is less general
since it requires the existence of Riemann invariants.
1.7 Problems
1. Find a velocity field u.x; t/ for the one-dimensional motion of a continuous

medium, where all particles are moving by inertia provided that, at the initial
time t D 0, the medium occupies the half-space x > 0 and the velocity
distribution has the form
x2 ,
a. u.x; 0/ D p
b. u.x; 0/ D x.
Answer:
2x2
a. u.x; t/ D p ,
2xt
p C 1 C 4xt C 1
4x C t2 t
b. u.x; t/ D :
2
2. Integrate the equation of characteristics for solutions to the Cauchy problem
ut C uux C u D 0;
u.x; 0/ D ax C b;
where a and b are constants.

Answer: x D x0 C .ax0 C b/.1 et /:
3. Construct a solution to the Cauchy problem
ut C c.u/ux D 0;
u.x; 0/ D c1 .ax C b/;
where a and b are constants and c.u/ is a monotone smooth function. For what
values of a and b does the gradient catastrophe occur?
Answer:

1 ax C b
u.x; t/ D c :
1 C at
4. Find characteristics and Riemann invariants of the system of equations
ut C 2 cos vux C sin uvx D 0;

vt C cos vux C .sin u C cos v/vx D 0:
Answer:
dx
D cos v W u v D const
dt
dx 1 C sin v u
D sin u C 2 cos v W tan D const :
dt cos v 2
5. Show that the Jacobian j@.r1 ; : : : ; rn /=@.u1 ; : : : ; un /j of the transformations r D

r.u/ reducing the hyperbolic system of equations ut C A.u/ux D 0 to a system
in Riemann invariants
rt C C.r/rx D 0; (1.22)
where C.r/ D diag.c1 ; : : : ; cn / is a diagonal matrix, is different from zero.

6. Consider the strictly hyperbolic system (1.22) reduced to Riemann invariants
(i.e., ci cj for i j). Prove that for a nondegenerate simple wave r D
r..x; t//, r0 ./ 0, all except one invariants ri are identically constant,
whereas the level lines of the simple wave .x; t/ D const form a family
1.7 Problems 21
of rectilinear characteristics corresponding to a Riemann invariant that is not

identically constant.
7. For the hyperbolic system of equations
ut C c.u; v/ux D 0;
vt C c.u; v/vx D 0
with multiple characteristic dx=dt D c.u; v/ prove that ux > vx > 0 along this
characteristic provided that these inequalities are valid at t D 0.
8. The one-dimensional motion of a barotropic continuous medium is governed
by the system of equations
t C u x C ux D 0;
c2 . /
ut C uux C x D 0;

where c. / is a smooth function such that c0 . / > 0 and c.0/ D 0. Find a

function c D c. / for which all characteristics of this system for any motion
are straight lines.
Answer: c D A , where A D const > 0.
9. Consider the system of equations of one-dimensional motions of a mixture of
two barotropic fluids
t C u x C ux D 0;
1
ut C uux C px D 0; (1.23)

y1t C uy1x D 0;
where u is the mixture velocity, D 1 1 C2 2 is the mixture density (here, i

are the densities and i are the volume fractions of the fluids; 0 6 i 6 1, 1 C
2 D 1), p D p1 . 1 / D p2 . 2 / is the pressure (dp1 =d 1 > 0, dp2 =d 2 > 0 for
1 > 0, 2 > 0), and y1 is the mass fraction of the first fluid (yi D i i = ). Show
that the system of equations (1.23) for , u, y1 with an implicitly given pressure
p D p. ; y1 / is hyperbolic and its characteristics have the form dx=dt D u,
dx=dt D u c, where the sound speed c in the mixture (the Wood speed) is
given by
1 1 2
D C ;
c2 1 c21 2 c22
dpi
c2i D .i D 1; 2/:
d i
Hint. Use the relation D y1 1 C y2 2 for the specific volumes D 1= ,

1 D 1= 1 , and 2 D 1= 2 .
10. Under the conditions of the previous problem, we consider an air-water mixture
with parameters 1 D 1 kg/m3, 2 D 1000 kg/m3 , c1 D 340 m/sec, and c2 D
1500 m/sec. Show that the sound speed c D c.1 / in the mixture has a unique
minimum cmin over the interval 0 6 1 6 1. Find this minimum.
Answer: cmin 21:5 m/sec.
11. Show that any hyperbolic system (1.1) with two independent variables t and x
can be reduced to the following form by the left multiplication by a suitable
matrix:
But C Cux C Db D 0;
where B, C, and D are symmetric matrices depending on u, x, and t; moreover,

the matrix B is positive definite (recall that such a system is called a symmetric
t-hyperbolic Friedrichs system).
12. Let u D .u1 ; : : : ; un /T 2 Rn , and let e W Rn ! R be a given smooth mapping
such that the Hessian matrix e00 .u/ D k@ui @uj e00 .u/kni;jD1 is positive definite, i.e.,
e.u/ is a convex function. We consider the Legendre transform e of a function
e defined by
e .v/ D v u e.u/;
where v D ru e.u/ and u D u.v/ is the preimage of an element v under the

action of the locally invertible mapping v D ru e.u/. Show that e .v/ is a
convex function of v. Show that the Legendre transform is an involution, i.e.,
double application of this transform yields the same function e.
Hint. Show that u D rv e .v/ and .e .v//00 D .e00 .u//1 .
13. Consider the equations of one-dimensional motion of an ideal gas with zero
pressure
t C . u/x D 0;
ut C uux D 0:
(This approximation appears in astrophysics.) Show that this system is not

hyperbolic. Find all conservation laws @t P. ; u/ C @x Q. ; u/ D 0 admitted by
this system. Find out whether there are laws with a convex function P among
these conservation laws.
Answer: P. ; u/ D a.u/ C b.u/; where a and b are arbitrary smooth functions;
P is not convex.
1.7 Problems 23
14. Consider the system of conservation laws of gas dynamics
t C . u/x D 0;
. u/t C . p C u2 /x D 0;
. s/t C . us/x D 0;
p D p. ; s/;
where is the density, u is the velocity, p is the pressure, and s is the entropy.
The thermodynamical state of the medium is characterized by the internal gas
energy ". ; s/ and temperature T. ; s/ connected by the Gibbs identity

1
Tds D d" C pd :

Show that the system of gas dynamics admits the additional conservation law
et C fx D 0;
where

1
e D " C u2 ;
2

1 2
f D u " C u C pu:
2
15. Under the conditions of the previous problem, compute the Hessian matrix
e00 .u/ of the function e D ." C u2 =2/ in the variables u D .u1 ; u2 ; u3 /T , where
u1 D , u2 D u, and u3 D s.
Answer:
0 2 1
u C K u " s s"ss
1
e00 .u/ D @ u 1 0 A;

" s s"ss 0 "ss
where K D 2 " 2 s" s C s2 "ss C 2 " .

16. Using the result of the previous problem, prove that e D ." C u2 =2/ is convex
in the variables u D . ; u; s/T if and only if E.; s/ D ".1=; s/ is convex in
the variables .; s/.
17. Using the GodunovFriedrichsLax theorem, write the system of conservation

laws of gas dynamics
t C . u/x D 0;
. u/t C . p C u2 /x D 0;
. s/t C . us/x D 0;
p D p. ; s/
as the symmetric t-hyperbolic Friedrichs system
A.v/vt C B.v/vx D 0
where A D AT > 0 and B D BT , by using the law of conservation of energy

with the function e D ." C 12 u2 / (here, " is the internal energy).
Answer:
A D .e .v//00 D .e.u//00 /1 ; B D .ue .v//00 ;
where
p 1 T
e .v/ D p; v D " C Ts u2 ; u; T :
2
18. Show that for any conservation law @t P C @x Q D 0 of the strictly hyperbolic
system (1.22) written in Riemann invariants r D .r1 ; : : : ; rn /T the function
P.r1 ; : : : ; rn / satisfies the system of linear equations

@2 P 1 @cj @P @ci @P
D .i; j D 1; : : : ; nI i j/:
@ri @rj ci cj @ri @rj @rj @ri
19. Find all conservation laws @t P.r; l/ C @x Q.r; l/ D 0 of the system in Riemann
invariants
rt C lrx D 0;
lt C rlx D 0
(the Chaplygin isentropic gas with polytropic exponent D 1).

Answer: P.r; l/ D f .r/g.l/
rl
; where f and g are arbitrary smooth functions.
20. Find out whether there are conservation laws
@t P.r; l; s/ C @x Q.r; l; s/ D 0
1.7 Problems 25
for the system of equations
rt C lrx D 0;
lt C slx D 0;
st C rsx D 0:
Answer: Do not exist.

21. Find jumps of the derivative x through the lines of weak discontinuity of the
solution to the Cauchy problem for the system of equations
t C u x C ux D 0;
ut C uux C x D 0
with the initial data

(
0; x 6 0;
.x; 0/ D 0 ; u.x; 0/ D k; 0 D const > 0:
kx; x > 0;
Answer: x D 0 t: x D 2.1Ckt/
k
.
22. Show that discontinuities of the second order derivatives of a solution to a
hyperbolic system that, together with the first order derivatives, is continuous
can propagate only along characteristics.
23. Prove that the velocity of propagation of weak kinematic shock waves in a
continuous medium with density is expressed as
1
DD .c1 C c2 / C O. 2 /;
2
where ci is the limit value of the characteristic velocity on the discontinuity line.
Verify that this equality is exact for kinematic waves with a quadratic function
q D Q. / of .
24. Show that D1 < D < D2 for the velocity D of the kinematic shock wave caused
by merging two shock waves with velocities D1 < D2 if the interaction of waves
is described by the law of conservation @t C @x Q. / D 0, where Q is a convex
function (Q00 . / > 0/.
25. At point x D 0 at time t D 0, a kinematic shock wave travelling with constant
velocity D1 catches up a shock wave travelling with velocity D0 over a constant
state 0 > 0. It is known that, after the wave interaction, the density doubles
and the process is described by the conservation law @t C c0 @x . 3 = 02 / D 0,
c0 D const > 0. Find the density .x; t/ for t > 0. What are the velocities of all
shock waves participated in the motion?
Answer: .x; t/ D 0 .x > Dt/, .x; t/ D 4 0 .x < Dt/, D D 3D0 , D1 D 4D0 ,
D0 D 7c0 .
26. A shock wave described by the conservation law @t uC@x .u2 =2/ D 0 propagates
with constant velocity D D 2u0 > 0 in a fluid over the state u D u0 and, at time
t D 1, catches up the trailing edge of the centered wave u D x=t (u0 t < x <
2u0 t) travelling to the right in a fluid over the state u D 2u0 . Find the trajectory
of the shock wave in the .x; t/-plane before and after the interaction with the
centered wave. Find out whether the shock wave catches the leading edge of
the centered wave.
Answer:
8

<u0 .2t 1/;

p
t 6 1;
x D u0 .3t 2 t/; 1 6 t 6 4;

:u 5 t 2 ; t > 4:
0 2
It catches up.
27. Consider kinematic waves in a traffic flow with the quadratic dependence q D
Q. / of the flow on the vehicle density 2 0; :

Q. / D 4qm 1 ; qm D const > 0:

Verify the condition (1.12) for Q. /. What is the maximal value um D

max u. / of the vehicle velocity u. / D Q. /= ? Find the dependence of
06 6
the flow q D q.u/ on the velocity u 2 0; um . For what values of u is the
maximum of the flow qm D max q.u/ attained? What are the extremal values
06u6um
of the characteristic velocity c. / D Q0 . / on the interval 0; ? Derive an
expression in the form c D c.q/ for q D Q. / on the interval 0; =2.
Answer:

4qm u
um D ; q.u/ D u 1 ; qm D q.u/uDu =2 ;
um m
max c. / D c. / D0 D um ; min c. / D c. / D D um ;

06 6 06 6
r
q
c.q/ D um 1 :
qm
28. Prove that, under the conditions (1.12), all continuous perturbations propagate
only upstream in the traffic flow: u. / > c. / for 0 < 6 , where u. / D
Q. /= is the vehicle velocity and c. / D Q0 . / is the characteristic velocity.
29. Prove that the propagation velocity of the kinematic shock wave D. / D
.Q. / Q. 0 //=. 0 /, regarded as a function of state with a fixed state 0
1.7 Problems 27
on the other side of the front, satisfies the relation
Z
1
D0 . / D .% 0 /Q00 .%/d%:
. 0 /2
0
Show that D. / is a monotone decreasing function if Q00 . / < 0.

30. The traffic flow moves with velocity u0 and density 0 along a street where the
traffic light turns red at time t D 0 and is kept during the time Tk . Using the
kinematic wave equations with flow

Q. / D 4qm ; 1 .0 < 0 < ; u0 D Q. 0 /= 0 /;

describe the traffic motion in a neighborhood of the traffic light for t > 0:
a. Find the trajectory OAB of the deceleration shock wave propagating
upstream in the traffic flow (cf. Fig. 1.6).
b. Determine the time tB when the wave OAB comes to a point B (if Tgreen >
tB Tk , then no traffic bottleneck occurs at the traffic light working in the
green light mode during the time Tgreen ).
c. Indicate the time tC when the vehicles which have been halted because of a
red traffic light catch up the flow of the vehicles which have passed by the
traffic light before the time t D 0.
Answer:
a.

1 0 4qm
OA W x D .c0 um /t c0 D Q . 0 /; um D ;
2
q
AB W x D c0 .t Tk / .u2m c20 /Tk .t Tk /:
b.
qm
tB D Tk .q0 D 0 u0 /:
qm q0
c.
um
tC D Tk :
um u0
31. Because of a road accident, the traffic flow q0 moved with density 0 decreases
up to q1 < q0 at the accident place during the time interval T. Using the
kinematic wave equation with the same function Q. / as in the previous
problem and assuming that 0 < q0 D Q. 0 / < qm and 0 < 0 < =2,
find the maximal distance l from the accidence place, where the road accident
affects the transport delay.
Answer:
p
qm .q0 q1 /
lD p T:
qm q0
32. Show that a simple wave for the equations of one-dimensional isentropic flows
of a polytropic gas propagating through the state at rest with density 0 and
sound speed c0 is a kinematic wave. Find the dependence q D Q. / of the flow
q D u on the density .
Answer:
1 !
2c0 2
Q. / D 1 :
1 0
33. The process of filtration of a fluid in a porous medium is governed by the

BuckleyLeverett equation
@t s C @x Q.s/ D 0; Q.s/ D 3s2 2s3 ;
where 0 6 s.x; t/ 6 1 is the saturation of pores by the fluid. Show that the
Riemann problem with the initial data s.x; 0/ D 0 for x > 0 and s.x; 0/ D
1 for x < 0 has at least two self-similar solutions. One of the solutions is a
combination of a shock and the adjacent centered wave (a stable solution):

x 9
s.x; t/ D 1 .x < 0/; s.x; t/ D 0 > ;
t 8
r
1 1 2x x 9
s.x; t/ D C 1 0< < ;
2 2 3t t 8
whereas the other is a piecewise constant (an unstable solution):

x x
s.x; t/ D 1 <1 ; s.x; t/ D 0 >1 :
t t
Is the function Q.s/ convex on the interval 0 6 s.x; t/ 6 1?

34. Construct a travelling wave type solution u.x; t/ D U.x Dt/ for the Burgers
equation
ut C uux D uxx . D const > 0/
such that u ! u1 as x ! C1 and u ! u2 as x ! 1, where u1 and u2 are

constants .u1 u2 /. Find asymptotics of the solution as ! 0.
1.7 Problems 29
Answer:
u2 u1 1
u.x; t/ D u1 C u1 ; DD .u1 C u2 /;
1 C exp u22 .x Dt/ 2
(
u1 ; x > Dt;
lim u.x; t/ D
!0 u2 ; x < Dt:
35. Show that the function u D 2vx =v is a solution to the Burgers equation if
the function v satisfies the heat equation
vt D vxx
(the ColeHopf transform). What solution v .x; t/ to the heat equation yields
a self-similar solution to the Burgers equation u .x; t/ D x=t under this
transform?
Answer:
1 2
v .x; t/ D p ex =.4t/ :
t
36. Verify that the solution to the Burgers equation in Problem 34 is the ColeHopf
transform of the sum v D v1 C v2 of two travelling wave type solutions to the
heat equation of the form
n u uj o
j
vj .x; t/ D exp x t . j D 1; 2/: (1.24)
2 2
37. Consider the solution u.x; t/ to the Burgers equation which is the ColeHopf
transform of the sum v.x; t/ D v1 .x; t/ C v2 .x; t/ C v3 .x; t/ of three solutions
to the heat equation of the form (1.24) with parameters u3 > u2 > u1 . Find the
asymptotics of u as ! 0.
Answer:
8

<u1 ; x > D1 t;

lim u.x; t/ D u2 ; D2 t < x < D1 t; for t < 0;
!0
:u ; x < D t;
3 3
(
u1 ; x > D3 t;
lim u.x; t/ D for t > 0;
!0 u3 ; x < D3 t;
where
1 1 1
D1 D .u1 C u2 /; D2 D .u2 C u3 /; D3 D .u1 C u3 /:
2 2 2
38. Prove that for the system of equations
ut C A.u/ux D 0;
u D .u1 ; u2 ; u3 /T ;
there exists a Riemann invariant constant along the characteristic dx=dt D c if

and only if l curl l D 0, where l.u/ is a left eigenvector of the 3 3-matrix A
corresponding to the eigenvalue c.
39. For what characteristics do Riemann invariants exist for the equations of one-
dimensional motion of a polytropic gas
t C u x C ux D 0;
1
ut C uux C px D 0;

pt C upx C pux D 0:
Answer:
dx
D u.x; t/ W p D const
dt
(the entropy is constant on the contact characteristic).
40. For what dependence of the sound speed c D c. ; p/ on the density and
pressure p do Riemann invariants r exist and are conserved on the sound
characteristics dx=dt D u c of the system of equations of one-dimensional
gas dynamics
t C u x C ux D 0;
1
ut C uux C px D 0;

pt C upx C c2 ux D 0:
Answer:
Z
a. p/ dp
c. ; p/ D ; r D u ;
a. p/
where a. p/ is an arbitrary function.

1.7 Problems 31
41. Consider the hyperbolic system of conservation laws
@t u C @x .u/ D 0: (1.25)
Prove that for any branch of the shock adiabat the tangent vector to this branch
at the adiabat center u D u0 is a right eigenvector of the matrix A.u0 / D 0 .u0 /.
42. We look for a self-similar solution u D u.
/,
D x=t, to the strictly hyperbolic
system of equations
@t u C A.u/@x u D 0:
Show that a necessary conditions for the existence of such a solution is the
condition r.u/ ru c.u/ 0 for at least one eigenvalue c.u/ of the matrix A.u/
(here, r is a right eigenvector of the matrix A corresponding to the eigenvalue
c). Show that if the characteristic field c.u/ satisfies this condition, called the
genuine nonlinearity in the sense of Lax, then the solution u.
/ is obtained by
integrating the system of ordinary differential equations
du r.u/
D :
d
r.u/ ru c.u/
43. The characteristic field c.u/ is said to be linearly degenerate in the sense of Lax
if r.u/ ru c.u/ 0 for a right edigenvector r corresponding to the eigenvalue
c. Show that if c.u/ is a simple eigenvalue of the matrix A.u/ D 0 .u/ of
the hyperbolic conservation law system (1.25), then the shock adiabat branch
D.u u0 / D .u/ .u0 / tangent to the vector r.u0 / at the point u D u0
coincides with the integral curve u D u.s/ of the system of ordinary differential
equations
du
D r.u/;
ds
ujsD0 D u0 :
Moreover, D D c.u/ for the shock wave velocity.

44. Show that for the system of gas dynamics
t C u x C ux D 0;
1
ut C uux C px D 0;

st C usx D 0;
p D p. ; s/
p
the sound characteristics dx=dt D u c (here, c D p . ; s/ is the sound
speed) satisfy the condition of genuine nonlinearity in the sense of Lax if the
equation of state p D g.; s/, D 1= , is such that g .; s/ > 0. Show that the
contact characteristics dx=dt D u are linearly degenerate in the sense of Lax.
45. Consider the system of conservation laws of mass, momentum, and energy of
the one-dimensional motion of an polytropic gas
t C . u/x D 0;
. u/t C . u2 C p/x D 0;

1 1 1 3
p C u2 C u C pu D 0:
1 2 t 2 1 x
Write the shock adiabat equation for a polytropic gas in the space . ; u; p/
with the origin . 0 ; u0 ; p0 / in the parametric form with the Mach number
M D pju0 Dj=c0 for a parameter, where D is the shock wave velocity and
c0 D p0 = 0 .
Answer:

2.M 2 1/ 2jM 2 1j
D 0 1 C ; u D u 0 c0 ;
. 1/M 2 C 2 . C 1/M

2 2
p D p0 1 C .M 1/ :
C1
46. Find characteristics of the system of linear elasticity
@ui @vi
D ;
@t @x
@vi @ @
0 D .i D 1; 2; 3/
@t @x @ui
with potential
1 1
.u1 ; u2 ; u3 / D .u21 C u22 / C . C 2/u23 ;
2 2
where 0 < 0 D const is the density of the material in the nondeformed state,
ui are the strains, vi are the velocities, and are the Lam constants ( >
0; > 0). Is this system hyperbolic?
Answer:
dx
D cj . j D 1; : : : ; 6/;
dt
1.7 Problems 33
where
r r

c1;2 D ; c3;4 D (transverse waves);
0 0
s
C 2
c5;6 D (longitudinal waves):
0
The system is hyperbolic.

47. Under the conditions of the previous problem, show that the linear equations of
elasticity have a solution with a strain vector u D .u1 ; u2 ; u3 /T of the form
u.x; t/ D f .x c1 t/a C g.x c5 t/b

p p
where c1 D = 0 , c5 D . C 2/= 0 , a; b 2 R3 are constant vectors such
that a b D 0, and the functions f and g are not identically constants.
48. Find the characteristic velocities for the system of quasitransverse wave
equations
@t ui D @x vi ;
0 @t vi D @x ui ;
where i D 1; 2, in an isotropic nonlinear medium with elastic potential
1 1
.u1 ; u2 / D .u21 C u22 / C 2 .u21 C u22 /2 . D const /:
2 4
Answer:
s
C 2 .u21 C u22 /
c1;2 D ;
0
s
C 3 2 .u21 C u22 /
c3;4 D :
0
49. Find characteristics and Riemann invariants of the system of equations of

longitudinal nonlinear elastic waves in a rod
ut D vx ;
0 vt D x ;
where u is the strain and D .u/ is the stress ( 0 .u/ > 0).
Answer:
s
dx 0 .u/
D ;
dt 0
Zu s 0
.
/
r .u; v/ D v d
:
0
0
50. Show that the shock adiabat (1.10) of the system of equations of nonlinear
elastic waves in a rod has the second order tangency at the center .u0 ; v0 / with
the level lines of the Riemann invariants r .u; v/ D r .u0 ; v0 /.
51. Find characteristics and Riemann invariants of the equilibrium equations of an
ideal rigid-plastic body for plane deformations

@ @ @
2k cos 2 C sin 2 D 0;
@x @x @y

@ @ @
2k sin 2 cos 2 D 0;
@y @x @y
where .x; y/ D .11 C 22 /=2 is the mean stress, k D const is the yield limit
under shear (the maximal tangent stress), .x; y/ is the angle of slope of the line
of maximal tangent stress at the point .x; y/.
Answer:
dy
D tan W 2k D const
dx
dy
D cotan W C 2k D const :
dx
52. Consider the equation of plane stress state of a rigid-plastic material under the
von Mises yield criterion of the form
p @! p @! @'
3 sin ! cos 2' cos ! C 3 sin ! sin 2' 2 sin ! D 0;
@x @y @y
p @! p @! @'
3 sin ! sin 2' 3 sin ! cos 2' C cos ! C 2 sin ! D 0;
@x @y @x
where the function !.x; y/ is connected with the principal stresses 1 and 2 by
the equalities

1 D 2k cos ! ;
6

2 D 2k cos ! C ;
6
1.7 Problems 35
where k D const is the yield limit, and '.x; y/ is the angle between the first
principal direction of the stress tensor and the Ox-axis. Find characteristics and
Riemann invariants in the domain where the system is hyperbolic.
Answer:
p p
dy 3 sin ! sin 2' 3 4 cos2 !
D p W
dx 3 sin ! cos 2' cos !
Z! p
3 4 cos2 s
' ds D const
2 sin s
=6

5 7 11
<!< ; <!< :
6 6 6 6
53. Derive the HamiltonJacobi equation for the characteristic surfaces of the
system of two-dimensional equations of linear elasticity
@v1 @11 @12

0 D C ;
@t @x @y
@v2 @21 @22

0 D C ;
@t @x @y
@11 @v1 @v2
D . C 2/ C ;
@t @x @y
@22 @v1 @v2
D C . C 2/ ;
@t @x @y

@12 @v1 @v2
D C .12 D 21 /;
@t @y @x
where v D .v1 ; v2 /T is the velocity vector and ij are components of the stress
tensor.
Answer:
r q s
C 2 q 2
't D 0; 't D 'x2 C 'y2 ; 't D 'x C 'y2 :
0 0
54. We consider the chemical adsorption equations (1.4) for a two-component

mixture with the Langmuir isotherm
1
f.u/ D .1 u1 ; 2 u2 /T ; p.u1 ; u2 / D 1 C 1 u1 C 2 u2 ;
p
where k are the Henry coefficients (0 < 1 < 2 ). Show that for u1 > 0 and
u2 > 0 the following assertions hold.
a. 2 R is an eigenvalue of the matrix f0 .u/ if and only if is a root of the
equation
12 u1 22 u2
C D p: (1.26)
1 p 2 p
b. The eigenvalues j of the matrix f0 .u/ are real and distinct: 0 < 1 < 1 =p <
2 < 2 =p (the strict hyperbolicity of the Langmuir adsorption equations).
55. Verify that the chemical adsorption equations (1.4) can be transformed to the
form
@
u C @ f.u/ D 0
by making the change of independent variables D vtx,

D x. Show that for
every eigenvalue of the matrix f0 .u/ given by Eq. (1.26) the function r D p
is a Riemann invariant .@
C @ /r D 0 for the characteristic d=d
D .
56. Consider the system of equations written in Riemann invariants
r12 r2
@
r1 C @ r1 D 0;
1 2
r1 r22
@
r2 C @ r2 D 0:
1 2
Show that the functions

2 =1 1
1
u1 D 1
1 ;
1 2 r1
r2

1 =2 2 2
u2 D 1 1
2 1 r1 r2
with constant 1 2 satisfy the system of the Langmuir adsorption equations

for a two-component mixture

1 u 1
@
u1 C @ D 0;
1 C 1 u 1 C 2 u 2

2 u 2
@
u2 C @ D 0:
1 C 1 u 1 C 2 u 2
1.7 Problems 37
Remark The Langmuir adsorption equations are also reduced to a system in

Riemann invariants in the general case of an n-component mixture.
57. Show that the laws of conservation of mass and momentum in gas dynamics
t C div. u/ D 0;
. u/t C div. u u C pI/ D 0
imply the equation for the density
X 3
@2 Tij
tt c20 4 D ;
i;jD1
@xi @xj
where c0 is the sound speed in a gas at rest, Tij D ui uj C . p c20 /ij .i; j D
1; 2; 3/ are components of the acoustic stress tensor, u D .u1 ; u2 ; u3 /T is the
velocity vector, p is the pressure, and ij is the Kronecker symbol.
58. Show that the potential ' of the velocity field u D r' of a vortex-free
isentropic gas flow satisfies the equation

1
dt 't C jr'j2 D c2 4'; .dt D @t C r' r/;
2
where c is the sound speed.

59. Show that for the system of linear acoustic equations in a gas flow of density
0 D const moving with a given constant speed u0 ,
0 dt u C rp D 0;
dt p C 0 c20 div u D 0;
where dt D @t C u0 r, the following integral law of conservation of energy

holds:

1d 2 p2 1 2 p2
0 juj C d C 0 juj C u0 n dS
2 dt 0 c20 2 0 c20
S

C pu n dS D 0;
S
where R3 is an arbitrary domain with piecewise smooth boundary S and

n is the outward unit normal to S.
60. For the acoustic equations in a gas moving with the constant speed u0 D
.u0 ; 0; 0/T find sound characteristics and corresponding bicharacteristics pro-
vided that each of these characteristic surfaces at t D 0 is the sphere jxj D R.
Answer: .x C u0 t/2 C y2 C z2 D .R c0 t/2 are
characteristics,
x D 1 cR0 t x0 u0 t, y D 1 cR0 t y0 , and z D 1 cR0 t z0 are
bicharacteristics.
61. Let the dependence of p.x; t/ on the variables x D .x; y; z/T and t be implicitly
given by
p D f .x k. p/ C c0 tjk. p/j/;
where f ; k 2 C2 . Show that p is a solution to the wave equation
ptt D c20 4p
(the SmirnovSobolev functional-invariant solution).

62. Find a general form of a spherical wave type solution p D p.jxj; t/ for the wave
equation.
1
Answer: p D jxj . f .jxj c0t/ C g.jxj C c0 t//, where f and g are arbitrary smooth
functions.
63. The acoustic radiation by a ball of radius r0 that pulsates according to
the harmonic law with frequency ! is described by the boundary condition
ur .x; t/ D U cos !t (U D const 0) for jxj D r0 for the radial component
ur of the gas velocity u. Looking for a solution to the acoustic equation in the
form of a spherical wave outgoing from the sound source (i.e., depending on
jxj c0 t), find the pressure field p in the domain jxj > r0 .
Answer:

a i!.tjxj=c0 / i 0 c0 r0 mU im !r0
p.x; t/ D Re e ; aD e mD :
jxj 1 C im c0
64. Find the acoustic energy flux I through the sphere SR W jxj D R,
Z
ID pu n dS;
SR
for the sound source of frequency ! in a gas at rest that creates the pressure
distribution p.x; t/ D ajxj1 cos !.t jxj=c0/.
Answer:
4a2 2 c0
ID cos !.t R=c0 / C sin2!.t R=c0 / :
0 c0 2!R
1.7 Problems 39
65. The Maxwell equations governing the propagation of electromagnetic waves in

a medium with dielectric permeability " and magnetic permeability have the
form
bt C curl e D 0;
"et curl b D 0;
div e D 0;
div b D 0:
Show that the electric-field vector e and the magnetic induction vector b satisfy
the wave equation. Determine the wave propagation velocity.
66. Show that for the Maxwell equations the energy integral identity holds:

1d 2 2
."jej C jbj /d C s n d D 0;
2 dt

where s D eb is the UmovPoynting vector (the energy flux vector), R3

is an arbitrary domain with piecewise smooth boundary D @, and n is the
outward unit normal vector.
67. Find characteristics of the equations of magnetic gas dynamics
t C u x C ux D 0;
.ut C uux / C px D jb;
pt C upx C pux D 1 . 1/j2 ;
bt C ex D 0;
"et C bx C j D 0;
where j D .e ub/ is the electrical current in the gas and D const is the
electrical conductivity.
Answer:
r
p 1
c1 D u; c2;3 D u ; c4;5 D p :
"
68. Find the characteristics of equations of one-dimensional motion of an ideal gas

with infinite conductivity
t C u x C ux D 0;
bt C ubx C bux D 0;
1
.ut C uux / C px C bbx D 0;

pt C upx C pux D 0:
Answer:
s
p b2
c1;2 D u; c3;4 D u C :

69. Construct a simple centered wave type solution with density playing the
role of a parameter of a simple wave of the equation of a gas with polytropic
exponent D 2 having infinite conductivity.
Answer:
p 2 x 2 16
k A2
b D A ; u D k C B; p D 2 ; D B ;
2 32 2 t 9k2
where A, B and k are constants.

70. Show that for each root D H.I x; t/ of the characteristic equation
3
!
X
det I C
i A i
D0
iD1
of the system (1.14) the Hamiltonian H satisfies the Euler identity
3
X
HD
i @
i H:
iD1
71. Find a solution to the Cauchy problem for the HamiltonJacobi equation with
initial data of a plane wave type
't D H.rx '/; '.x; 0/ D k x;
where k 2 R3 is a constant vector, under the condition that the Hamiltonian H is

a positivehomogeneous function of the first degree: H.p/ D H.p/ . > 0/.
Answer: '.x; t/ D k x C H.k/t.
1.7 Problems 41
72. It is known that the wave front .t/ is a characteristic of the system of equations
ut C xvy D 0;
vt C yux D 0
and is given by the equation 0 W '0 .x; y/ D 0 at t D 0. Find the trajectory

of the ray outgoing from the point .x0 ; y0 / 2 0 in the .x; y/-plane. Find the
location of the front at time t > 0 provided that, at t D 0, it has the shape of a
hyperbola xy D 1 and perturbations propagate to the domain xy < 1.
Answer:
.x=x0 /x0 p0 D .y=y0 /y0 q0 ; p0 D @x '0 .x0 ; y0 /;

q0 D @y '0 .x0 ; y0 /; xy D et :
Chapter 2
Dispersive Waves
2.1 Dispersion Relation
We consider wave processes governed by the system of linear partial differential

equations with constant coefficients
X
n X
m
bsp @st @px u.x; t/ D 0; (2.1)
sD0 pD0
where t is the time, x 2 R is the spatial variable, and the coefficients bsp and solution
u can be complex-valued. A wave described by a complex-valued solution
u.x; t/ D aei.kx!t/ ; (2.2)
is called an elementary wave packet. Here, a is the wave amplitude, k is the

wavenumber, ! is the frequency, and D kx !t is the wave phase. In the case
of real parameters (Im k D 0 and Im ! D 0), the wavenumber is equal to the
number of waves at a segment of length 2 in the x-axis and the frequency is equal
to the number of wave crests or troughs passing by a fixed observer during the time
2. For such parameters we can introduce the wavelength L D 2=k and the time
period T D 2=!. Each constant value of the phase is carried with the velocity
cp D !=k, called the phase velocity. Taking into account this fact, we can regard
a wave packet as a travelling wave u.x; t/ D a exp ik.x cp t/ propagating with the
phase velocity.
Differentiating the function u defined by (2.2), we get
@t u D i!u;
@x u D iku:

44 2 Dispersive Waves
Therefore, the system (2.1) have solutions in the form of wave packets with an
amplitude a 0 if and only if
D.!; k/ D 0; (2.3)
where
X
n X
m
D.!; k/ D bsp .i!/s .ik/p :
sD0 pD0
The equality (2.3) relating the frequency and the wavenumber is called the
dispersion relation. Since D.!; k/ is a polynomial of degree n in !, for a given
k Eq. (2.3) has, in general, n complex roots !j D !j .k/ ( j D 1; : : : ; n). The family
of wave packets
u.x; t/ D a exp i.kx !.k/t/
with an arbitrary wavenumber k and the frequency !.k/ generated by some fixed
root of the dispersion relation is called a wave mode. The number of wave modes for
the system (2.1) coincides with the order of this system with respect to the variable
t. The dependence cp .k/ D !.k/=k of the phase velocity on the wavenumber means
that the profile of a wave consisting of several wave packets of a given mode with
different k is deformed due to spreading of these packets running with different
velocities. This phenomenon is called dispersion of waves. Respectively, a wave is
dispersive if ! 00 .k/ 0.
Example 2.1 Consider the telegraph equation
utt c20 uxx C 2 u D 0; (2.4)
where c0 0 and are real constants (this equation governs oscillations of elec-
tromagnetic waves along a conductor of large length, referred to as a transmission
line). This equation yields the dispersion relation (2.3) with
D.!; k/ D ! 2 C c20 k2 C 2
which generates two wave modes with the real frequencies

q
! .k/ D c20 k2 C 2 ; k 2 R:
For k > 0 the mode with frequency !C .k/ describes waves travelling to the
right along the Ox-axis, whereas the mode with frequency ! .k/ describes waves
00
propagating to the left. For 0 we have ! .k/ 0, i.e., the waves are dispersive.
2.1 Dispersion Relation 45
In the case D 0, Eq. (2.4) becomes the one-dimensional wave equation
utt D c20 uxx :
By the DAlembert formula, the general solution of this equation is the sum of two
travelling waves
u.x; t/ D f .x c0 t/ C g.x C c0 t/;
and the dispersion relation yields wave modes with frequencies ! .k/ D c0 k and
phase velocities cp D c0 . In this case, dispersion does not take place. We note that
Eq. (2.4) is a second order hyperbolic differential equation written as
u
D . 2 =4c20 /u
in the characteristic variables

D x c0 t and D x C c0 t. In the case
0, dispersion causes distortion of a signal consisting of harmonics with several
frequencies. It is less seen for large kq(i.e., in the domain of high frequencies !)
since the phase velocities c .k/ D c20 C 2 k2 are asymptotically constant as
k ! 1, i.e., c
p .k/ ! c0 .
If a frequency defined by the dispersion relation (2.3) is complex, i.e., ! D !r C

i!i , then the wave packet has the form
u.x; t/ D a.t/ei.kx!r t/ ;
where a.t/ D a exp.!i t/ is the amplitude factor depending on time. If the imaginary
part of the frequency ! is negative (!i < 0), then the wave amplitude exponentially
decays as t ! C1, i.e., dissipation takes place.
Example 2.2 Consider the linearized Korteweg-de VriesBurgers equation
ut C u0 ux C uxxx D uxx ;
where u0 and > 0 are constants. A unique wave mode for this equation is given
by the frequency
!.k/ D u0 k k3 ik2 :
Since Im ! D k2 < 0, the wave packets with k 0 exponentially decay, which
is caused by the fact that the equation involves the second order derivative uxx with
coefficient , interpreted as the viscosity of a continuous medium. Long waves (the
limit as k ! 0/ are less susceptible to dissipation, the greater the viscosity, the
greater the decay is.
In the case of complex frequencies with !i > 0, the amplitude of the solution
increases infinitely with time. In this case, instability of the wave process is
observed. It should be taken into account that the system (2.1) appearing as a
result of linearization of more general nonlinear equations was first designated for
describing propagation of small perturbations in a continuous medium. Therefore, a
wave packet with !i > 0 can simulate the instability process only at the initial stage
since the original linear approximation loses sense with growth of perturbations.
From this point of view, a neutral-stable elementary wave packet with real frequency
(!i D 0) and constant amplitude a describes regular propagation of waves in the
case where the influence of dissipation on instability is negligible.
2.2 Multi-dimensional Wave Packets
In the case of systems of l differential equations of the form (2.1) for u D

.u1 ; : : : ; ul /T , elementary wave packets are described by the solutions u.x; t/ D
a exp i.kx !t/, where a D .a1 ; : : : ; al /T is the amplitude vector.
Example 2.3 Consider the system of quasilinear first order equations
dt u C A@x u C B@2x u D 0 (2.5)
for an n-dimensional vector u D .v; u1 ; u2 ; : : : ; un1 /T , n > 2; with the differenti-

ation operator dt D @t C v@x . The distinguished component v of the vector-valued
function u is interpreted as the displacement velocity of particles of a continuous
medium along the trajectories dx=dt D v.x; t/. We assume that constant n n-
matrices A and B are real; moreover, the matrix A is symmetric and the matrix B is
antisymmetric: A D AT and B D BT . The linearization of the system (2.5) at the
rest state u D 0 yields the system of equations with constant coefficients
@t u C A@x u C B@2x u D 0:
We look for a solution in the form of a wave packet u.x; t/ D a exp i.kx !t/ with
real wavenumber k. Then we have the equation
.ikB C A cI/a D 0
for the amplitude vector a, where c D !=k is the phase velocity. Solutions with an
amplitude a 0 can exist only if the following equality holds:
det.ikB C A cI/ D 0:
This equality is the dispersion relation. All its roots cj D cj .k/ . j D 1; : : : ; n/ are
real in view of properties of the matrices A and B. Consequently, the frequencies
2.2 Multi-dimensional Wave Packets 47
!j .k/ D cj .k/k of all wave modes are also real. We note that the system (2.5) is
invariant under the Galilean transform
e
t D t; e
x D x u0 t; e
v D v u0 ; e
ui D ui .i D 1; : : : ; n 1/;
realizing the passage to the coordinate system moving with velocity u0 D const .
The linearization of the system (2.5) at a constant solution u D .u0 ; 0; : : : ; 0/ yields
the system of equations
@t u C e
A@x u C B@2x u D 0
with the matrix eA D A C u0 I. The phase velocity e

c of wave packets for this system
is given by e
cDe ! =k, where the wave frequencies e! and ! in the moving and fixed
coordinate systems are connected by the relation e! D ! u0 k. In wave theory, the
change in frequency is called the Doppler effect and the change in frequency u0 k is
referred to as the Doppler shift.
In the case x 2 R3 , by an elementary wave packet we mean a function of the
form
u.x; t/ D aei.kx!t/ ;
where k D .k1 ; k2 ; k3 /T is the wave vector. The surface of constant phase is the
plane k1 x1 C k2 x2 C k3 x3 !t D const which is travelling in the space R3 along the
direction of the vector k with the normal phase velocity cp D !=jkj (cf. Fig. 2.1).
Problem 2.1 Find the normal phase velocities for the three-dimensional wave
packets
w.x; t/ D a cos.k x !t/; a 2 R3
Fig. 2.1 The wave vector k

is always orthogonal to
surfaces of the constant phase
satisfying the system of linear elasticity (the Lam equations)
0 wtt D . C /r div w C 4w;
where w D .w1 ; w2 ; w3 /T is the displacement vector, 4 D @2x1 C @2x2 C @2x3 , div w D

w1x1 C w2x2 C w3x3 , r D .@x1 ; @x2 ; @x3 /, 0 D const is the medium density, > 0
and > 0 are the Lam constants.
Solution We first find the result of action of the main differential operators on a
vector-valued function w.x; t/ in the form of a cosinusoidal wave packet:
4w D jkj2 a cos.k x !t/;

r div w D k.a k/ cos.k x !t/:
From the Lam equations it follows that the amplitude vector a, wave vector k, and
frequency ! satisfy the identity
. 0 ! 2 jkj2 /a D . C /k.a k/: (2.6)
For a given vector k 0 we represent the amplitude vector in the form
a D jkj2 .a k/k C b;
where the vector b is orthogonal to the wave vector: b k D 0. The projections of the
vector identity (2.6) onto the direction of k and onto the perpendicular plane yield
the system of equations
. 0 ! 2 . C 2/jkj2 /a k D 0;
. 0 ! 2 jkj2 /b D 0:
If a k 0, then ! 2 D . C 2/jkj2 = 0 and, consequently, b D 0. In this case,

the direction of the displacement vector w coincides with the direction of k, i.e., the
wave packet is a longitudinal wave propagating with the normal phase velocity
p
cp D . C 2/= 0 :
If a k D 0, but a 0, then b 0. Therefore, ! 2 D jkj2 = 0 . In this case,

we have a transverse wave such that the displacement vector w is orthogonal to the
direction
pof the wave propagation, whereas the normal phase velocity is equal to
cp D = 0 .
2.3 Group Velocity 49
Answer:
c2p D . C 2/= 0 .a k 0/
c2p D = 0 .a k D 0/:
2.3 Group Velocity
The dispersion property is highlighted in the interaction of the wave packets

u.x; t/ D a cos.kx !t/ of a fixed mode with real frequency ! D !.k/ and the
same amplitude, but different wavenumbers k. For the sum of such two packets we
have
a cos.kx !t/ C a cos.k1 x !1 t/

k1 k !1 ! k1 C k !1 C !
D 2a cos x t cos x t ;
2 2 2 2
where !1 D !.k1 /. The wave motion described by this sum is represented as a

periodic sequence of groups of waves propagating with the velocity .!1 !/=.k1
k/. As k1 ! k, the velocity of the envelope coincides with the derivative
d!
cg .k/ D ;
dk
called the group velocity. The maximal amplitude of crests of the carrying wave
for each group is approximately equal to the double amplitude of the original wave
packets, whereas their displacement velocity is equal to the phase velocity cp .k/
(cf. Fig. 2.2). The group and phase velocities are connected by the relation
dcp
cg D cp C k :
dk
Hence for dispersive waves the group and phase velocities are different.
Problem 2.2 Consider the linear Boussinesq equation describing long shallow
water waves of small amplitude
1 2
utt c20 uxx D h uxxtt ;
3 0
Fig. 2.2 For dispersive

waves the group and phase 2a
velocities are different
2a
where
p h0 is the depth of the fluid at rest, g is the acceleration of gravity, and c0 D
gh0 . Find the phase velocity cp and the group velocity cg . What are the maximal
values jcp j and jcg j?
Solution Looking for a solution in the form of an elementary wave packet
u.x; t/ D a exp.i.kx !t//;
we obtain the dispersion relation

1 2 2
! 1 C h0 k D c20 k2
2
3
which determines two modes corresponding to the waves travelling to the left and
to the right respectively. Hence for each mode we can find the phase velocity

1 2 2 1=2
cp .k/ D c0 1 C h0 k
3
and the group velocity

3=2
1
cg .k/ D c0 1 C h20 k2 :
3
Since the velocities do not coincide for an arbitrary k 0, the waves are dispersive.
Furthermore, the following estimates holdW
jcp .k/j 6 c0 ; jcg .k/j 6 c0 ;
where equality is attained at the long-wave limit k D 0.

Thus, the absolute
p values of the phase and group velocities do not exceed the
critical velocity gh0 .
2.4 Stationary Phase Method 51
In the multi-dimensional case, the group velocity is defined by the equality
cg .k/ D r!.k/:
In the case of spatial dispersive waves, not only absolute values, but also directions
of the phase velocity vector cp .k/ D .!.k/=jkj/k and the group velocity vector
cg .k/ can be different.
Example 2.4 Let a frequency ! be a homogeneous function of degree 0 in k D
.k; l; m/T , i.e.,
!.k; l; m/ D !.k; l; m/
for all > 0. Differentiating this identity with respect to the parameter at the
point D 1, we obtain the relation
@! @! @!
k Cl Cm D0 (2.7)
@k @l @m
which means that the vectors cg .k/ and k are perpendicular at each point k 2 R3 .
The homogeneity of the frequency is also a necessary condition for the orthogonality
of the group velocity and the wave vector. Indeed, (2.7) is a linear first order partial
differential equation for !.k; l; m/. The equations of characteristics have the form
dk=k D dl=l D dm=m, which implies that the general solution ! D !.k=l; k=m/ is
a homogeneous function of degree 0 in the variables k, l, m.
2.4 Stationary Phase Method
A wave process in a dispersive medium usually becomes regular with time even
if the wave propagation is caused by an initial perturbation of a general form. For
a given mode ! D !.k/ we consider the linear superposition of elementary wave
packets
Z
C1
u.x; t/ D a.k/ei.kx!.k/t/ dk: (2.8)

1
The amplitude factor a.k/ is uniquely determined by the initial function u.x; 0/ via
the Fourier integral
Z
C1
1
a.k/ D u.x; 0/eikx dx:
2
1
It is convenient to study the behavior of the solution (2.8) for large t for a fixed ratio
x=t D U, which corresponds to the motion of the observer with constant velocity U.
We introduce the phase function .k/ D kx=t !.k/. Then for v.t/ D u.Ut; t/ we
have
Z
C1
.k/
v.t/ D a.k/eit dk:
1
We assume that a.z/ and .z/ are analytic functions of the complex variable z D
k C il in the strip j Im zj < l0 , where l0 > 0. One can show that the contribution of
the integration intervals where 0 .k/ 0 to the solution is exponentially small as
t ! C1. In a neighborhood of a point k0 where 0 .k0 / D 0, but 00 .k0 / 0, we
have the expansion
1
.z/ D .k0 / C .z k0 /2 00
.k0 / C O.jz k0 j3 /:
2
Therefore, the structure of level lines Im .z/ D C in a neighborhood of such a

stationary point is similar to the structure of the level lines of the saddle surface
.k k0 /l D C= 00 .k0 /. The level lines of the function Im in a neighborhood of
the point .k0 ; 0/, where 00 .k0 / < 0, are shown in Fig. 2.3.
For solutions of the form (2.8) the phase of the wave packet under the integral
sign is stationary at the point k0 where ! 0 .k0 / D U D x=t, which corresponds to
the point x D x.t/ moving with the group velocity cg .k0 /. Deforming the integration
contour along the real axis in (2.8) to a contour in the complex domain, as shown in
Fig. 2.3, and computing the integral corresponding to the first two nonzero terms in
Fig. 2.3 The integration

contour along the real axis in
.k; l/-plane is deformed into
the other passing through the
saddle point .k0 ; 0/ (bold
solid lines)
the Taylor expansion of the phase function .z/, we obtain the asymptotic expansion
s
2 i.k0 x!.k0 /t 4 sgn ! 00 .k0 // 1
u.x; t/ D a.k 0 /e C O : (2.9)
j! 00 .k0 /jt t
Consequently, the initial perturbation

p is divided into groups of wave packets with
an amplitude decreasing like 1= t as t ! C1. In the case of dispersive media,
the carried energy density is, as a rule, proportional to the squared amplitude of
the solution ju.x; t/j2 . By the above asymptotics of the solution, the part of energy
contained between the straight lines xj D ! 0 .kj /t (j D 1; 2), is preserved in the first
approximation:
Zx2 Zk2
2
ju.x; t/j dx D 2 ja.k/j2 dk:
x1 k1
This means that the energy is carried with the group velocity; moreover, dispersion
of a wave implies scattering of the energy in the space.
Problem 2.3 Consider the Cauchy problem for the EulerBernoulli beam equation
utt C 2 uxxxx D 0 .1 < x < C1/;

u.x; 0/ D u0 .x/; ut .x; 0/ D 0;
where the constant > 0 characterizes elastic properties of the material and u0 .x/ 2
C1 is an absolutely integrable and even function of x. Using the stationary phase
method, find an asymptotics of the solution as t ! C1.
Solution The direct and inverse Fourier transforms of a function u.x; t/ with respect
to x have the form
Z
C1
1
b
u.k; t/ D u.x; t/eikx dx;
2
1
Z
C1
u.x; t/ D b
u.k; t/eikx dk:
1
Since @c
x u.k; t/ D .ik/ b
n n
u.k; t/, the Fourier-image b
u.k; t/ is a solution to the Cauchy
problem for the second order ordinary differential equation in t:
utt C 2 k4b
b u D 0;
u.k; 0/ D ub0 .k/;
b b
ut .k; 0/ D 0:
Integrating, we find
1 2 2
u.k; t/ D ub0 .k/ cos k2 t D
b ub0 .k/fei k t C ei k t g:
2
Applying the inverse Fourier transform, we can represent the solution as
1
u.x; t/ D fIC .x; t/ C I .x; t/g;
2
where
Z
C1
I .x; t/ D ub0 .k/ei.kx! .k/t/ dk; ! .k/ D k2 :

1
Each integral I can be represented in the form (2.8), and it is possible to use the
asymptotic formula (2.9) with the wavenumber k0 D k0 .x; t/ defined implicitly by
the relation x=t D ! 0 .k0 /. For the wave modes ! .k/ D k2 this relation implies
00
k0 .x; t/ D x=.2 t/. Furthermore, for these modes ! .k0 / D 2 0. Therefore,
by formula (2.9),
r 2
x i x 1
I .x; t/ D ub0 e 4 t 4 C O .t ! C1/:
t 2 t t
It remains to note that if the initial function u0 .x/ is even in x, then its Fourier
transform is even in k:
Z
C1
def 1
a.k/ D ub0 .k/ D u0 .x/ cos kx dx D a.k/:
2
1
Taking into account this evenness property, we obtain the desired asymptotic
expansion of the solution.
Answer:
r 2
x x 1
u.x; t/ D a cos CO ;
t 2 t 4 t 4 t
where
Z
C1
1
a.k/ D u0 .x/ cos kx dx:

0
The solution behaves according to formula (2.9) as t ! C1 if the condition

! 00 .k0 / 0 holds. This condition fails at extremum points of the group velocity,
where ! 00 .k0 / D c0g .k0 / D 0. In neighborhoods of such stationary points, the
solution has another asymptotics. To construct the asymptotics, we introduce the
phase function in some other way, namely: .k/ D k! 0 .k0 / !.k/ so that the
wave phase D kx !t is connected with by the equality
D k.x cg .k0 /t/ C .k/
0 00
(it is assumed that x=t is constant). Since .k0 / D 0 and .k0 / D 0, we have the
following decomposition of .z/:
1
.z/ D .k0 / C .z k0 /3 000
.k0 / C O.jz k0 j4 /:
6
Analyzing the structure of level lines of the function Im and taking into account
the above decomposition and suitable deformation of the integration contour in (2.8)
in the complex plane z D k C il, we can characterize the behavior of the solution
u.x; t/ as t ! C1. Namely, in a neighborhood of the point x D cg .k0 /t such that
c0g .k0 / D 0, c00g .k0 / 0, we have
!
2a.k0 / x cg .k0 /t 2
u.x; t/ D q Ai q ei.k0 x!.k0 /t/ CO.t 3 /: (2.10)
3
.1=2/jc00g .k0 /jt 3 00
.1=2/jcg .k0 /jt
This formula involves the Airy function
Z
C1
1 1 3
Ai.x/ D cos kx C k dk
3
0
with the following behavior as jxj ! 1:

8

2 3
1 < exp x 2 ; x ! C1;
Ai.x/ p 3
2 jxj 4
1 2 3
:2 cos jxj 2 ; x ! 1:
3 4
The graph of the Airy function (cf. Fig. 2.4) has the shape of the envelope of
wave packets travelling with group velocity cg .k0 /. Hence it is clear that the front
of the first-localized perturbation propagates with the velocity equal to the extremal
value of the group velocity; moreover, the distance between neighboring zeros of the
envelope in the domain of its oscillations unboundedly increases with the growth of
t. By the approximate formula (2.10) and the above asymptotic formula for the Airy
function, the wave packets decay in a neighborhood of the front (like t1=3 ) slightly
Fig. 2.4 The graph of the

Airy function is shown. It has
the shape of the envelope of
wave packets travelling with
group velocity
slower than inside the domain of the wave motion (like t1=2 ). In the unperturbed
domain before the front, the wave process exponentially decays with increasing
distance from the front.
2.5 Nonlinear Dispersion
Decomposition of monochromatic wave packets into groups of waves caused by

their interaction is typical for the evolution of a wave field in a medium with
dispersion. Moreover, the behavior of the envelope of wave packets is often
described by wave type differential equations. Thus, to describe modulated wave
packets of the form
u.x; t/ D .x; t/ exp i.k0 x !.k0 /t/
with a complex-valued amplitude smoothly depending on x and t, many models of

continuum mechanics and mathematical physics involve the nonlinear Schrdinger
equation
i t C xx j j2 D 0: (2.11)
The equations in (2.11) with the sign minus or plus at the nonlinear term have
different mathematical properties. The equation itself has solutions in the form of
wave packets .x; t/ D a exp i.kx !t/ with a constant amplitude a independent
of x and t. Substituting into (2.11), we obtain the dispersion relation
D.!; kI a/ D ! k2 a2 D 0
which, in addition to the frequency ! and wavenumber k, contains the wave

amplitude a. This fact expresses one of the main differences between the case of
nonlinear dispersive waves and the linear case.
2.5 Nonlinear Dispersion 57
Problem 2.4 For the nonlinear Schrdinger equation (2.11) with the sign plus
construct an envelope solitary wave .x; t/ D A.x ct/ exp i.kx !t/, where A is a
real amplitude function A such that A; A0 ! 0 as jxj ! 1.
Solution Substituting the function into (2.11), we obtain the complex ordinary
second order differential equation
A00 k2 A C !A C A3 C i.2k c/A0 D 0:
Since A is real, the imaginary part of this equation yields the relation .2kc/A0 D 0.
Consequently, for the existence of a nonconstant solution A const it is necessary
that c D 2k for the velocity c of the envelope solitary wave. Taking into account this
fact, we obtain the nonlinear equation
A00 2 A C A3 D 0;
where 2 D k2 ! > 0. Multiplying the equation for A by A0 , integrating, and taking

into account the decay at infinity, we obtain a first order equation with separated
variables. The solution of this equation is given by the quadrature formula
Z
dA
p D
.
D x ct/; (2.12)
A .1=2/A2
2
p
where the integral is computed by substitution
p A D =. 2 cosh
/. Thus, we p find
the required envelope A.
/ D a= cosh . 2a
/, where the amplitude a D =. 2/
is connected with the frequency ! and wavenumber k of the carrying wave packet
by the dispersion relation ! D k2 2a2 . Moreover, the displacement velocity of the
envelope coincides with the group velocity cg .k/ D 2k of the linearized Schrdinger
equation i t C xx D 0.
Answer:
aei.kx!t/
.x; t/ D p ;
cosh f 2a.x ct/g
where ! D k2 2a2 , c D 2k.

To describe nonlinear long waves with dispersion, one often uses the Korteweg-
de Vries equation
ut C uux C uxxx D 0: (2.13)
The joint role of nonlinearity and dispersion is already apparent while searching a
travelling wave type solution. Substituting u D u.x ct/, c > 0, into (2.13), we
obtain the third order ordinary differential equation
cu0 C uu0 C u000 D 0:
The double integration reduces this equation to the first order equation
1
u02 D .u u1 /.u u2 /.u3 u/; (2.14)
3
where the roots of the cubic polynomial on the right-hand side are connected with
the integration constants and the wave velocity c by the Vieta formula (in particular,
u1 C u2 C u3 D 3c). Solutions in the form of a periodic wave are obtained for simple
real roots u1 < u2 < u3 . Without loss of generality we can assume that u2 D 0 since
Eq. (2.13) is invariant under the Galilean transform et D t, e
x D x u2 t, e
u D u u2 .
Denoting by a D u3 a positive root u3 > 0, we have u1 D 3c a for the negative
root u1 < 0. Under this choice of the reference system in the travelling wave, we
can reduce Eq. (2.14) to the quadrature formula
p Z
a
ds
3 p D x ct:
.s C a 3c/s.a s/
u
In the general case, this integral is not necessarily expressed in terms of

elementary functions. However, substituting s D a cos2 and making similar
change of the sought function u D a cos2 ', we can reduce this dependence to the
relation
Z'
d
p D
.
D x ct/; (2.15)
1 m2 sin2
0
where
u3 u2 a 2a 3c
m2 D D ; 2 D :
u3 u1 2a 3c 12
The function ' D am.

I m/, defined implicitly by (2.15), is called the Jacobi
amplitude, and the superposition cn.
I m/ D cos am.
I m/ is called the elliptic
cosine. Consequently, the required wave profile has the form
u.x; t/ D acn2 ..x ct/I m/:
Such a periodic wave is said to be cnoidal because its definition involves the
function cn. For a given value of m (0 < m < 1) the function cn. I m/ is periodic
2.5 Nonlinear Dispersion 59
with period 2K.m/, where

Z2
d
K.m/ D p
1 m2 sin2
0
is a complete elliptic integral. Hence the parameter a > 0 characterizing the wave
height, i.e., the difference between the elevation of a crest and a neighboring trough
(the double amplitude) is connected with the wave velocity c and period L with
respect to the independent variable x by the relation
p p
4 3 a
LD p K p .3c < a/: (2.16)
2a 3c 2a 3c
If we formally introduce the wavenumber k D 2=L, then this equality can be

regarded as the dispersion relation for cnoidal waves. Elliptic Jacobi functions are
reduced to elementary functions as m ! 0 and m ! 1 (the parameter m is called
the modulus of an elliptic function). The function cn.I m/ uniformly converges to
the function cos as m ! 0, which corresponds to transformation of a cnoidal wave
of small amplitude a ! 0 to a usual wave packet in the linear theory. On each finite
interval of the variable , the function cn.I m/ uniformly converges to 1= cosh as
m ! 1. In this case, the periodic wave degenerates into the travelling solitary wave
2
p
u.x; t/ D 3c cosh f. c=2/.x ct/g:
A generalization of solitary wave type solutions are exact multi-soliton solutions

to the Korteweg-de Vries equation describing nonstationary interaction between
solitary waves. In soliton theory, it is convenient to write the Korteweg-de Vries
equation as
ut C 6uux C uxxx D 0: (2.17)
n-Soliton solutions to this equation have the form
@2
u.x; t/ D 2 ln det fI C A.x; t/g;
@x2
where I is the identity matrix of order n and A is a symmetric matrix with entries
i j .ki Ckj /x 4.ki3 Ckj3 /t
aij .x; t/ D e e .i; j D 1; : : : ; n/;
ki C kj
where 1 ; : : : ; n are arbitrary real constants and k1 ; : : : ; kn (ki C kj 0).

2.6 Problems
1. Find the phase velocity cp .k/ D !.k/=k and the group velocity cg .k/ D d!=dk
for the dispersion relations ! D !.k/ generated by the following equations:
(a) ut C c0 ux C c0 h20 uxxx D 0,
(b) ut C c0 ux h20 uxxt D 0,
where c0 > 0 and h0 > 0 are constants. Show that for all k > 0
! .a/ .k/ 6 ! .b/ .k/;

c.a/ .b/
p .k/ 6 cp .k/;
c.a/ .b/
g .k/ 6 cg .k/;
where equalities hold only for k D 0. What are the limits of the ratios
.a/ .b/
cg .k/ cg .k/
.a/
; .b/
cp .k/ cp .k/
in the long-wave (k ! 0) and short-wave (k ! 1) approximations? Construct

the graphs of the functions !.k/, cp .k/, and cg .k/ for 1 < k < C1 in cases
(a) and (b).
Answer:
.a/ .b/
cg .k/ cg .k/
lim .a/
D lim .b/
D 1;
k!0 cp .k/ k!0 cp .k/
.a/ .b/
cg .k/ cg .k/
lim .a/
D 3; lim .b/
D 1:
k!1 cp .k/ k!1 cp .k/
2. It is known that the group velocity of waves described by a differential equation

with two independent variables x, t and complex constant coefficients is real
and coincides with the double phase velocity. Restore the form of the equation
provided that it has (a) the first order in t, (b) the second order in t.
Answer:
(a) iut C uxx D 0,
(b) utt i.1 C 2 /uxxt 1 2 uxxxx D 0
where ; 1 ; 2 2 R are constants.
3. Show that for the equation
utt c20 uxx C u D 0;

2.6 Problems 61
where x1 ; x2 2 R, the energy balance equation holds:
Zx2 x2
d
e.x; t/dx C f .x; t/ D 0;
dt x1
x1
where e D .u2t C c20 u2x C u2 /=2 is the energy density and f D c20 ux ut is the
energy flux.
4. The velocity of energy transfer by the wave packet u D a sin.kx !t/ is equal
to the ratio U D F=E of the mean energy flux over the time period T D 2=!
to the mean energy density E over the space period L D 2=k,
tZ
1 CT xZ1 CL
1 1
FD f .x; t/dt; ED e.x; t/dx:
T L
t1 x1
Under the conditions of the previous problem, show that U coincides with the
group velocity cg .k/.
5. Show that the local frequency !, local wave number k, and phase of the group
of waves defined implicitly as functions of x and t by the relations
! D W.k/; x D W 0 .k/t; D kx !t;
satisfy the differential equations
kt C W 0 .k/kx D 0;
x D k;
t D !
provided that W 00 .k/ 0.

6. It is known that for a mode ! D !.k/ with two-dimensional wave vector k D
.k; l/T the group velocity vector cg D .@!=@k; @!=@l/T for all k 0 forms a
constant angle (0 6 6 ) with the wave vector. For what dependence !.k/
is it possible?
Answer: !.k/ D f jkje' tan , where ' D arc tan.k=l/ and the function f 2
C1 satisfies the conditions f 0 .
/ > 0 .
2 R/ for 0 6 < =2, f 0 .
/ 6 0 .
2
R/ for =2 < 6 , and !.k/ D f .k=l/ for D =2.
7. Show that for the dispersion curve ! D !.k/ that satisfies the condition !.0/ D
0 and is convex for k > 0 the group velocity cg .k/ D ! 0 .k/ and the phase
velocity cp .k/ D !.k/=k possess the following properties:
a. cg .k/ > cp .k/ if ! 00 .k/ > 0 for all k > 0 (the dispersion curve is convex).
b. cg .k/ < cp .k/ if ! 00 .k/ < 0 for all k > 0 (the dispersion curve is concave).
8. Show that the resonance of the phase and group velocities cp .k0 / D cg .k0 / at a
point k0 0 is equivalent to any of the following properties.
a. At the point k D k0 , the dispersion curve ! D !.k/ touches the line passing
through the origin in the .k; !/-plane.
b. The value k D k0 of the wave number is a stationary point of the phase
velocity, i.e., c0p .k0 / D 0.
9. For a real mode ! D !.k/ consider the wave packets
uj .x; t/ D aj ei.kj x!.kj /t/ .kj 2 RI aj 2 C; aj 0I j D 1; 2; 3/:
Let m and n be given integers. Find the wavenumbers kj for which there is a
constant 2 C such that .u1 .x; t//m .u2 .x; t//n D u3 .x; t/ identically with
respect to x; t 2 R.
Answer: k3 D mk1 C nk2 , !.k3 / D m!.k1 / C n!.k2 /.
10. Show that for isotropic waves with the dependence ! D !.jkj/, !.0/ D 0, if
! 0 .k/ > 0, ! 00 .k/ > 0, then there exist vectors k1 ; k2 ; k3 satisfying the three-
wave resonance condition
k3 D k1 C k2 ; !.jk3 j/ D !.jk1 j/ C !.jk2 j/;
but this condition cannot be satisfied if ! 0 .k/ > 0, ! 00 .k/ < 0.

11. For what values of real parameters a and b are all wave modes of the following
system stable
utt a2 uxx C vxx D 0;

vtt b2 vxx C uxx D 0
Answer: a2 b2 > 1:
12. For the system of equations
ut C Aux D Cu
with symmetric real constant n n-matrices A and C we look for a solution in

the form of a wave packet u.x; t/ D U exp ik.x ct/ with a wavenumber k 2 R
(k 0) and an amplitude vector U 2 Cn . Show that the imaginary part cim of
the complex phase velocity c D cre C icim is equal to
.CU; U/
cim D ;
kkUk2
2.6 Problems 63
where
X
n
.U; V/ D Uj V j
jD1
is the inner product of vectors U; V 2 Cn (the bar means complex conjugation)

and kUk2 D .U; U/. What is cim if the matrix A is symmetric and the matrix C
is antisymmetric?
13. Consider the system of equations
Aut C Bux C Cuxx D 0
with symmetric real constant n n-matrices A and B (the matrix A is positive

definite) and antisymmetric matrix C. Show that all wave modes of this system
are stable.
14. Consider the wave packet u D a exp i.kx !t/ for the equation
X
n
ut uxx C bs @2sC1
x u D 0;
sD0
where > 0 and bs .s D 1; : : : ; n/ are constants. For what time T is the absolute
value ju.x; t/j halved?
Answer: T D ln 2=.k2 /:
15. For the equation
wtt c20 wxx C c20 wxxt D 0
describing waves in a KelvinVoigt viscoelastic medium (w is the displacement,

c0 is the velocity of the propagation of longitudinal waves, and > 0 is the
relaxation parameter) we look for a solution in the form of the wave packet
w.x; t/ D aei.~x!t/ D aex ei.kx!t/
decaying with the growth of x, where ~ D k C i is the complex wavenumber.

Derive the dispersion relation for ~ and a real frequency !. Find the phase
velocity of the wave cp D !=k and the decay exponent regarded as a function
of frequency !.
Answer:
2c20 .1 C 2 ! 2 /
! 2 D c20 .1 C i!/~ 2 ; pc2p D ;
1 C 1 C 2!2

!2 1 1
2 D 2 p :
2c0 1 C 2!2 1 C 2!2
16. Perform a similar analysis for the equation
1
tt c20 xx C t D 0

describing waves in a Maxwell viscoelastic medium (here, is the stress in the
medium).
Answer:
i! 2c20 !
!2 C D c20 ~ 2 ;
c2p D p ;
1 C 2 ! 2 C !
! 1
2 D 2 p :
2c0 1 C 2 ! 2 C !
17. The initial wave profile u0 has the form of a sequence of triangular impulses
defined on pairwise disjoint intervals Im D .xm 2; xm C 2/ .m D 1; : : : ; n/:
(
2 jx xm j; x 2 Im ;
u0 .x/ D
0; x 62 Im .m D 1; : : : ; n/:
Find the amplitude factor a.k/ for the function u0 .

Answer:
2 sin2 k X ikxm
n
a.k/ D e :
k2 mD1
18. Find the amplitude factor a.k/ if the initial function u0 has the form of a
modulated wave packet:
a. u0 .x/ D ejxj cos k0 x,
2
b. u0 .x/ D ex =2 cos k0 x .k0 D const /.
Answer:
1 C k2 C k02
a. a.k/ D ,
1 C .k C k0 /2 1 C .k k0 /2
1 1 2 2
b. a.k/ D p e 2 .k Ck0 / cosh.kk0 /:
2
19. Prove that for the initial function u0 .x/ such that
Z
C1
def
AD el0 jxj ju0 .x/j dx < 1 .l0 > 0/
1
2.6 Problems 65
the amplitude function a.k/ admits an analytic extension to the strip j Im zj < l0
of the complex variable z D k C il, where the following estimates holdW
A
ja.z/j 6 ;
2
A
ja0 .z/j 6 :
2e.l0 j Im zj/
20. Let the amplitude function a and phase function satisfy the conditions
a.k/ 2 C1 ; .k/ 2 C2 ; 0
.k/ 0 .1 < k1 6 k 6 k2 < C1/:
Show that
Zk2
it .k/ a.k/eit .k/ kDk2 1
a.k/e dk D 0 Co as t ! C1:
it .k/ kDk1 t
k1
Hint. Use the RiemannLebesgue lemma.

21. The wave motion is described by a solution u.x; t/ of the form (2.8) with mode
! D !.k/ 2 C2 and initial data u.x; 0/ D A0 .x/eik0 x , where the function A0 has
the Fourier transform b
A0 with compact support
Zr
b
A0 .k/ D 0 .jkj > r > 0/; MD jb
A0 .k/j dk < 1
r
(the initial perturbation with narrowband spectrum). Prove that for small t > 0
the solution u has the form of a modulated harmonic wave:
u.x; t/ D A0 .x ! 0 .k0 /t/ei.k0 x!.k0 /t/ Ce

u.x; t/;
where the function e

u satisfies the uniform estimate with respect to x
1 2
je
u.x; t/j 6 Mr t max j! 00 .k/j:
2 jkk0 j6r
22. Using the stationary phase method, find the asymptotics of the integral
Z
C1
1 3 2
v.t/ D ei.tk 3 k / ek =t dk
1
as t ! C1.
Answer:
p p
2 2t t 1
v.t/ D p cos CO :
4
e t 3 4 t
23. Using the representation of the Airy function as a contour integral in the plane
of complex variable D k C il,
Z
1 1 3
Ai.x/ D ei.xC 3 / d;
2
C
p
along the contour C D fk C il W l D .1= 3/jkjg, show that the function Ai.z/ is
an entire analytic function of the complex variable z D x C iy and satisfies the
differential equation
Ai00 .z/ D zAi.z/:

p p
24. We look for a self-similar solution u.x; t/ D .1= 3 3t/v.x= 3 3t/ to the linear
Korteweg-de Vries equation
ut C uxxx D 0:
What differential equation does the function v.

/ satisfy? Show that the Airy
function v D Ai.
/ yields one of such solutions.
Answer: v 00 D
v C C .C D const /.
25. Consider the Cauchy problem for the linearized Korteweg-de Vries equation
ut C c0 ux C uxxx D 0;
u.x; 0/ D u0 .x/;
where the initial function u0 satisfies the condition
Z
C1
def
MD u0 .x/dx 0:
1
Using the stationary phase method, find an asymptotic expansion of the solution
in a neighborhood of the point x D c0 t (c0 > 0) as t ! C1.
Answer:
2
M x c0 t
u.x; t/ D p
3
Ai p3
C O t 3 :
3t 3t
2.6 Problems 67
x D ax, e
26. Find a general form of the scaling e t D bt, e
u D cu reducing the
Korteweg-de Vries equation
ut C uux C uxxx D 0
where ; D const, 0, 0, to the form (2.13) in variablese x,e

t, e
u. What
scaling of the variables x, t, u leaves the Korteweg-de Vries equation invariant?
Answer: ex D ax, e t D a3 t, eu D .=/a2 u .a 2 R W a 0/. The equation is
invariant under the transformation e x D ax,et D a3 t, e
u D a2 u.
27. For the Korteweg-de Vries equation written as
ut C 6uux C uxxx D 0 (2.18)
construct a solitary wave solution u D u.x ct/ (u; u0 ; u00 ! 0 as jxj ! 1/

such that the wave crest at time t D 0 is located at a given point x D x0 .
Answer:
c
u.x; t/ D p .c > 0/:
2 c
2 cosh 2
.x x0 ct/
28. Consider the n-soliton solution to the Korteweg-de Vries equation (2.18)
@2
u.x; t/ D 2 ln det fI C A.x; t/g;
@x2
where I is the identity matrix of order n and A is a symmetric matrix with entries
i j .ki Ckj /x 4.ki3 Ckj3 /t
aij .x; t/ D e e .i; j D 1; : : : ; n/;
ki C kj
k1 ; : : : ; kn and 1 ; : : : ; n (ki C kj 0) are real constants.

a. What choice of constants 1 and k1 in the solitary solution (n D 1)
corresponds to a travelling wave type solitary solution (cf. Problem 27)?
b. Write explicitly a two-soliton solution (n D 2) in terms of hyperbolic
functions with parameters k1 D 1 D 2 D 1 and k2 D 2.
Answer:
p pcx0 p
a. 12 D ce , k1 D c=2.
3 C 4 cosh.2x 8t/ C cosh.4x 64t/
b. u.x; t/ D 12 .
f3 cosh.x 28t/ C cosh.3x 36tg2
29. Consider the eigenvalue problem for the second order ordinary differential
operator (the SturmLiouville problem)
d2 '
C .u.x; t/ C /' D 0 .1 < x < 1/
dx2
with a given coefficient u depending on the parameter t (time). Show that if the
function u.x; t/ is a solution to the Korteweg-de Vries equation (2.18), then the
eigenfunction ' and eigenvalue satisfy the relation
d
'2 C .'Qx 'x Q/x D 0;
dt
where Q D 't 'xxx C 3.u /'x. Show that d=dt D 0 if the functions ' and
u, together with their derivatives, converge to zero as jxj ! C1.
30. Show that for solutions u.x; t/ of the Korteweg-de Vries equation (2.13) decay-
ing, together with their derivatives, as jxj ! 1, the following conservation
laws hold:
Z
C1
d
u2 .x; t/dx D 0;
dt
1
C1
Z
d 2 1 3
ux .x; t/ u .x; t/ dx D 0:
dt 3
1
31. We say that a differential equation with two independent variables x and t that
is of the first order with respect to t admits the Hamiltonian formulation if it
can be represented as
ut D Dx u H.u; ux ; : : : ; u.n/
x:::x /:
Here, Dx is the operator of total differentiation with respect to x and u is the

Euler operator (the operator of variational differentiation),
Dx D @x C ux @u C uxx @ux C uxxx @uxx C : : : ;

u D @u Dx @ux C D2x @uxx : : : ;
where x, u, ux , uxx , : : : are regarded as independent variables. Find the Hamilton

function H.u; ux / for the Korteweg-de Vries equation (2.13).
Answer:
1 2 1 3
H.u; ux / D u u :
2 x 6
2.6 Problems 69
32. Show that for any smooth function v.x; t/ 2 C4 the pair of functions v and
u D vx v 2 is connected by the identity (the GardnerMiura transformation)
ut C 6uux C uxxx D .2v C @x /.vt 6v 2 vx C vxxx /:
33. For the modified Korteweg-de Vries equation
vt .6v C 6v 2 /vx C vxxx D 0 (2.19)
find a bounded travelling wave type solution v D v.x t/ such that v; v 0 ; v 00 !

0 as x ! 1.
Answer:
1
vD (a front type wave)
1 C e.xt/
34. Construct a travelling solitary wave type solution v D v.x ct/ to Eq. (2.19)
with a velocity 0 < c < 1 such that v; v 0 ; v 00 ! 0 as x ! 1.
Answer:
c
vD p p :
1 C 1 c cosh c.x ct/
35. Derive the dispersion relation for wave packets
.x; t/ D a exp.ik x !t/
satisfying the Schrdinger equation with constant potential V D const :
2
i t C 4 V D 0;
2m
where 4 D @2x C @2y C @2z is the Laplace operator, is the Planck constant (the
de Broglie wave in quantum mechanics describing the behavior of a particle of
mass m in a field with potential V). Show that the particle velocity U D p=m,
defined as the ratio of its momentum p D k (k is the wave vector) to the mass
m, coincides with the group velocity cg .k/.
36. Consider the nonlinear Schrdinger equation
2
i t C 4 V.j j2 / D0
2m
with nonlinear potential V D V.j j2 /. Show that, under the change of the
sought complex-valued functions
p
.x; t/ D .x; t/ei.x;t/= ; v D r;
where the functions > 0 and are real-valued (the Madelung transformation),
this equation goes to the following system of equations for and v:
t C div . v/ D 0;
p
2 . / (2.20)
m.vt C .v r/v/ C r V. / p D 0:
2m
37. Show that the system (2.20) in the one-dimensional case takes the form
t C . v/x D 0;

2 xx 2
m.vt C vvx / C V. / x2 D 0:
4m 2 x
Linearize this system at the constant solution D 0 , v D 0 and derive the

dispersion relation. Show that the wave modes are real for all wavenumbers if
and only if V 0 . / > 0.
38. Consider the nonlinear Schrdinger equation
i t C xx C j j2 D0 .; 2 R W 0; 0/: (2.21)
Verify that this equation is invariant under the transformation t ! t, !

(the bar means complex conjugation). Find all homothetic transformations
x D ax, e
e t D bt, e D c , where a, b, and c are real parameters, that leave
Eq. (2.21) invariant. Show that, under a suitable homothetic transformation of
the variables x, t, , Eq. (2.21) takes one of the following two canonical forms:
i t C xx j j2 D 0:
To which form does the equation reduce? What is the most general form of the
reducing transformation?
Answer:pThe sign plus if > 0 and minus if < 0; e t D a2 t,
x D ax, e
e D j= ja1 .
39. Show that for solutions to the nonlinear Schrdinger equation (2.11) decaying
as jxj ! 1 the following conservation laws hold:
Z
C1
d
j .x; t/j2 dx D 0;
dt
1
C1
Z
d 2 1 4
j x .x; t/j j .x; t/j dx D 0:
dt 2
1
2.6 Problems 71
40. For the nonlinear Schrdinger equation (2.11) with the sign minus
i t C xx j j2 D0
construct a solution of the form .x; t/ D A.x ct/ exp i.kx !t/ with the real
amplitude A such that jAj ! a D const > 0, A0 ! 0 as jxj ! 1.
Answer:

a.x ct/ i.kx!t/

.x; t/ D a tanh p e ;
2
where ! D a2 C k2 , c D 2k. p p
41. Find the general form of the self-similar solution u.x; t/ D v.x= t/= t for the
one-dimensional Schrdinger equation
iut C uxx D 0:
Answer:
0 p 1
Z t
x=
1 2 B 2 =4 C
u.x; t/ D p eix =4t @C1 C C2 ei
d
A .C1 ; C2 2 C/:
t
0
42. Using the Fourier transform, construct a solution to the Cauchy problem
iut C uxx D 0;
sin x
u.x; 0/ D :
x
Answer:

Zz
1 ix2 2t x 2t C x 2
u.x; t/ D p e 4t f p Cf p ; f .z/ D eis ds:
2 t 2 t 2 t
0
43. Let a complex-valued function u.x; t/ 2 C4 be a solution to the Cauchy problem

for the linear Schrdinger equation
iut C uxx D 0;
u.x; 0/ D u0 .x/;
where is a constant and the initial function u0 takes only real values,
Im u0 .x/ 0. Prove that the function v.x; t/ D Re u.x; t/ is a solution the
Cauchy problem for the equation of flexural waves in an elastic rod
vtt C 2 vxxxx D 0;
v.x; 0/ D u0 .x/; vt .x; 0/ D 0:
44. Using the Fourier transform, construct explicitly a solution to the Euler
Bernoulli beam equation
utt C 2 uxxxx D 0 .1 < x < C1/;
with the initial data

2 =4
u.x; 0/ D aex ; ut .x; 0/ D 0 .a D const /
Hint: To take the inverse Fourier transform, use the formula
Z
C1 r
2 x42
ek eikx dk D e . 2 C W Re > 0/:

1
p
Choose the branch for which 1 D 1.
Answer:

a x2 tx2 1
u.x; t/ D p e 4.1C 2 t2 / cos arc tan t :
4
1 C 2 t2 4.1 C 2 t2 / 2
45. Derive the dispersion relation, find the normal phase and group velocities for
the equation of bending oscillations of a plate
utt C 2 42 u D 0;
where 4 D @2x1 C @2x2 is the two-dimensional Laplace operator, 2 D

Eh2 =.12 0 .1 2 //, h is the plate thickness, E is the Young modulus, and
is the Poisson coefficient.
Answer: !.k/ D jkj2 , cp .k/ D jkj, cg .k/ D 2 k.
46. For the displacement vector w D .w1 ; w2 ; w3 /T solving the three-dimensional
Lam system of equations
0 wtt D . C /r div w C 4w
consider the Helmholtz representation
w D r' C curl v .div v D 0/;

2.6 Problems 73
where ' and v are defined for all x D .x1 ; x2 ; x3 /T 2 R3 and have bounded
second order derivatives with respect to x and t. Show that the scalar potential
' and vector potential v satisfy the wave equations
'tt D c21 4';

vtt D c22 4v:
What are the propagation velocities c1 and c2 of the corresponding waves?

Answer:
s
C 2
c1 D is the velocity of longitudinal waves
0
r

c2 D is the velocity of transverse waves.
0
47. The equality (2.6) connecting the amplitude vector a, wave vector k, and
frequency ! of the wave packet for the three-dimensional Lam system of
equations can be written as the linear system of equations
A.!; k/a D 0
which is homogeneous with respect to a, where
A.!; k/ D . 0 ! 2 jkj2 /I . C /k k;
the symbol means the tensor product of vectors. Find the dispersive function
D.!; k/ D det A.!; k/ for the Lam equations.
Answer:
D.!; k/ D . 0 ! 2 jkj2 /2 0 ! 2 . C 2/jkj2 :
48. Using the two-dimensional Lam system of equations with displacement vector
w D .w1 ; w2 /T , find the phase velocity cp of the Rayleigh surface waves
described by the wave packets
w1 D A1 .x2 / cos.kx1 !t/;

w2 D A2 .x2 / sin.kx1 !t/
with the stress-free condition
@x2 w1 C @x1 w2 D 0;
@x1 w1 C . C 2/@x2 w2 D 0
on the boundary of the half-spaces x2 6 0 and the decay condition
w!0 as x2 ! 1:
Show that the phase velocity cp of the Rayleigh waves satisfies the inequalities
c1 > c2 > cp , where c1 is the velocity of longitudinal waves and c2 is the
velocity of transverse waves (cf. Problem 46).
p
Answer: cp D s0 c2 , where s0 is the least real root of the equation
s3 8s2 C 16.3=2 .c2 =c1 /2 /s 16.1 .c2 =c1 /2 / D 0:
49. Small oscillations of identical masses linked by springs of rigidity 2 are

described by the system of ordinary differential equations
mwR n D 2 .wnC1 2wn C wn1 /; n 2 Z:
Find the phase and group velocities of the signal propagation along the chain in
the form of the wave packet wn .t/ D a expfi.kn !t/g (k is the dimensionless
wavenumber).
Answer:
sin.k=2/
cp .k/ D p ; cg .k/ D p cos.k=2/:
m .k=2/ m
50. Under the conditions of the previous problem, show that for those motions of
masses that satisfy the decay condition wn ; wP n ! 0 (n ! 1) the total energy
E of the system is preserved with time:
C1
1 X
P 2n C 2 .wnC1 wn /2 g D const :
def
E.t/ D fmw
2 nD1
51. Write an equation of small oscillations and deduce the dispersion relation for a
chain with alternating masses m1 m2 linked by identical springs of rigidity
2 . What is the dispersion relation in the limit of identical masses m1 D m2 D
m?
Answer:
8 s 9
< 1 2 2 k=
1 1 1 4 cos
! 2 .k/ D 2 C C :
: m1 m2 m1 m2 m1 m2 ;
2.6 Problems 75
52. Consider a nonlinear chain of elastic beads of the same mass and radius R that
interact according to the Hertz law:
p 2 n
2R E xnC1 xn o3=2
Fn D 2 ;
3.1 2 / R
where E is the Young modulus, is the Poisson coefficient, xnC1 > xn are
the centers of neighboring beads, and Fn is the force of their elastic interaction.
The equations describing displacements of the bead centers wn in dimensionless
variables is written as
wR n D . wn C wn1 /3=2 . wnC1 C wn /3=2 ; n 2 Z;
where the parameter > 0 characterizes the initial displacement of the bead
centers. Linearize the equation of chain in the case 0 and deduce the
dispersion relation.
53. Consider the equation
utt D .u3=2 C u1=4 .u5=4 /xx /xx
governing the propagation of one-dimensional nonlinear waves in weakly

compressible granular materials (here, u D wx > 0 is the strain and w is
the displacement). This equation corresponds to the long-wave approximation
for a chain of elastic beads interacting in accordance with the Hertz law (cf.
the previous problem) without initial displacement ( D 0). Verify that this
equation has an exact travelling wave type solution
25 4 x ct
uc .x; t/ D c cos4 :
16 5
Show that the function .x; t/ defined by the conditions D uc for jx ctj <
.5=2/ and D 0 for jx ctj > .5=2/ is also a solution (a compacton, i.e., a
solitary wave with compact support).
54. Derive the dispersion relation for the linearized Whitham integro-differential
equation
Z
C1

ut C e 2 jxyj uy .y; t/dy D 0:
4
1
Answer:
2k
!.k/ D :
2 C 4k2
55. The Hilbert integral transform H acts on a function u by the formula

0 x" xCA 1
Z
C1 Z Z
1 u.y/dy 1@ A u.y/dy :
Hu.x/ D v:p: D lim C
xy A!C1 xy
1 "!0C xA xC"
c of Hu.x/ and the Fourier transform of u.x/

How are the Fourier transform Hu.k/
connected? Find the Hilbert transform Hv.x/ of the function v.x/ D b=.x2 C
b2 / .b > 0/.
c
Answer: Hu.k/ u.k/, Hv.x/ D x=.x2 C b2 /.
D i sgn kb
56. Looking for a solution in the form of an elementary wave packet u.x; t/ D
aei.kx!t/ , deduce the dispersion relation for the linearized BenjaminOno
equation
ut C c0 ux C Huxx D 0;
where H is the Hilbert transform.

Answer: !.k/ D .c0 C jkj/k:
57. Construct a solution to the BenjaminOno equation
ut C uux C Huxx D 0
in the form of a travelling solitary wave u D u.x ct/, where the fractional-
rational function u.
/ is a linear combination of the functions v1 .
/ D b=.
2 C
b2 / and v2 .
/ D
=.
2 C b2 / (use the result of Problem 55 and the fact that the
Hilbert transform commutes with the operator of differentiation).
Answer:
4c
u.x; t/ D :
c2 .x ct/2 C 1
Chapter 3
Water Waves
3.1 Equations of Motion
We consider the motion of an ideal incompressible inhomogeneous fluid in the

gravity field g D .0; 0; g/T . The unknowns are the velocity vector u D .u; v; w/T ,
the density , and the pressure p depending on x D .x; y; z/T 2 R3 and the time t.
The integral laws of conservation of volume

u n dS D 0;
S
mass

d
d C u n dS D 0;
dt
S
momentum

d
u d C . u.u n/ C pn/dS D g d;
dt
S
and energy

d 1 2 1 2
juj C gz d C juj C p C gz .u n/ dS D 0
dt 2 2
S
hold in any fixed domain with piecewise smooth boundary S (n is the unit outward
normal to S). In the domain, where the motion is described by smooth functions u,

78 3 Water Waves
, and p, the above-mentioned set of conservation laws is equivalent to the system

of differential equations
div u D 0;
t C u r D 0;
(3.1)
1
ut C .u r/u C rp D g:

Moreover, the vorticity ! D curl u satisfies the Helmholtz equation
1
!t C .u r/! D .! r/u rp r :
2
Hence vorticity in a nonviscous inhomogeneous fluid changes under the influence

of two factors: transfer of the initial vorticity distribution and formation of new
vorticity caused by the noncoincidence of level surfaces for the pressure and density
in the flow (the isobaric surfaces p.x; t/ D const and the isochoric surfaces .x; t/ D
const ).
In the case of a constant density D 0 , the solutions to the system (3.1) describe
homogeneous fluid flows. In this case, the system is reduced to
div u D 0;
1 (3.2)
ut C .u r/u C rp D g:
0
Respectively, the Helmholtz equation takes the form
@u
dt ! D h!i; dt D @t C .u r/: (3.3)
@x
Due to the special structure, the system (3.3) is integrated by passing from the
Eulerian coordinates .x; t/ to the Lagrangian coordinates .; t/. The dependence
x D x.; t/ between the Eulerian and Lagrangian coordinates is determined by the
solution to the Cauchy problem for differential equations for trajectories of fluid
particles
dx
D u.x; t/;
dt (3.4)
xjtD0 D :
In the Lagrangian coordinates, for the sought function !

e.; t/ D !.x.; t/; t/ we
have @t !
e D dt !. Therefore, for each trajectory with a given initial location of a
particle the system (3.3) is a system of ordinary differential equations for the
3.1 Equations of Motion 79
vector-valued function !
e.; t/. On the other hand, differentiating (3.4) with respect
to the parameter , we obtain the equation in variations
@u
@t M D M; MjtD0 D I
@x
with the Jacobi matrix
@.x; y; z/
M D x0
.; t/ D
@.
; ; /
and the identity matrix I. By the OstrogradskiiLiouville formula, for the determi-
nant jMj D det M we have
jMjt D jMjtru0x D jMj div u D 0:
Consequently, the matrix M is nonsingular: jM.; t/j D 1. Hence M is the

fundamental matrix of solutions to the linear system (3.3) of ordinary differential
equations. Thus, the variation of vorticity along trajectories of fluid particles is
described by the Cauchy formula
! D M!0 ;
where !0 is the initial vorticity field. As a consequence, we arrive at the Lagrange

theorem asserting that ! 0 in a volume of a homogeneous fluid .t/ for all t > 0
if the vorticity vanishes in .0/. For a potential flow the velocity vector field u has a
potential ' such that u D r'. The function ' is harmonic with respect to the spatial
variables x, y, z (' D 0). In this case, the momentum equation in the system (3.2)
is reduced to the CauchyLagrange integral
1 1
't C jr'j2 C p C gz D b.t/;
2 0
where b is an arbitrary function.

To determine uniquely the unsteady motion in the whole domain occupied by a
fluid, we set u D u0 at t D 0 and impose boundary conditions at the boundary of
the domain for t > 0, in particular, the impermeability condition u n D 0 at the
fixed boundary part with normal vector n and the kinematic condition
. ft C u rf /jf D0 D 0
together with the dynamic condition
p De
p;
80 3 Water Waves
where e p is a given function, at the free boundary f .x; y; z; t/ D 0. For example, for a
fluid in contact with the atmosphere the pressure at the free surface is assumed to be
constant if the air motion is not taken into account: p D p0 . In this case, without loss
of generality we can assume that p0 D 0. The kinematic condition at the interface
of two immiscible fluids has the form
. ft C uj rf /jf D0 D 0 .j D 1; 2/;
where uj are the limiting velocity values from both sides of the contact surface. If
the surface tension is not taken into account, then the dynamic condition implies the
pressure continuity
p D p2 p1 D 0:
Stationary flows are described by solutions to the system (3.1) such that t D 0 and
ut D 0. In this case, for any stream line
dx dy dz
LW D D
u v w
the Bernoulli integral holds:
1 2 1
juj C p C gz D b.L/
2
(the Bernoulli constant b.L/ depends on the stream line).

Example 3.1 Consider the plane flow of a two-layer fluid simulating the process
of penetrating the bottom layer of a heavy fluid into a light fluid (cf. Fig. 3.1). We
assume that the fluid densities are constant and are equal to 1 and 2 respectively
( 2 < 1 ). The flow is stationary in the reference frame moving with the point
of contact of the interface and the bottom. It is additionally assumed that, in this
moving reference frame, the heavy fluid is at rest (this flow scheme is due to [17].
Fig. 3.1 A gravity current head of a heavy fluid penetrating a light fluid
3.1 Equations of Motion 81
In Fig. 3.1, A and B denote the limiting flow states with constant velocities as x !
1 and x ! C1 respectively. The law of conservation of mass for the upper layer
is written as
UH D u2 .H h1 /:
The law of conservation of total horizontal momentum of the fluid has the form
ZH ZH ZH

pA dy C 2 u22 dy D pB C 2 U 2 dy: (3.5)
0 h1 0
By the vertical momentum equation, the pressure at the state B is distributed

hydrostatically
pB .y/ D pB jyD0 g 2 y:
Using the Bernoulli integral on the bottom part OB, we get
1
pB jyD0 D pO 2 U 2 :
2
Similarly, applying the Bernoulli integral on the bottom part AO, we see that the
pressure is constant: p D pO . From the vertical momentum equation for the state A
we obtain the hydrostatic distribution of pressure
(
pO g 1 y; 0 < y < h1
pA .y/ D
pO g 1 h1 g 2 .y h1 /; h1 < y < H:
Substituting the obtained expressions for pressure distributions into (3.5) and
removing u2 by using the law of conservation of mass, we find
U2 1 2 .2H h1 /.H h1 /h1

D :
gH 2 H 2 .H C h1 /
On the other hand, applying the Bernoulli integral to the upper fluid along the
interface, we have
1
pA jyDh1 C 2 u22 C g 2 h1 D pO :
2
Using again the law of conservation of mass and the known expression for pA .y/,
we find
U2 2. 1 2 / .H h1 /2 h1
D :
gH 2 H3
82 3 Water Waves
Comparing these two different expressions for U 2 =gH, we conclude that h1 D H=2.
Thus, the laws of conservation of mass, momentum, and energy simultaneously hold
for the flow configuration under consideration only if the depth of the lower layer is
equal to half the total depth of the channel.
3.2 Linear Theory of Surface Waves
We consider the potential flow of a homogeneous fluid in the layer .t/ D f.x; y/ 2
R2 W 0 < z < h.x; y; t/} bounded by the free surface z D h.x; y; t/ and the flat
bottom z D 0 (cf. Fig. 3.2).
The equations for the velocity potential ' and function h have the form
' 'xx C 'yy C 'zz D 0; x 2 .t/;

'z D 0; z D 0;
9
ht C 'x hx C 'y hy 'z D 0; =
1 z D h.x; y; t/:
't C .'x2 C 'y2 C 'z2 / C gh D 0;;
2
The system of these equations with the initial condition
h D h.0/ .x; y/; ' D ' .0/ .x; y; z/ .' .0/ D 0; x 2 .0//
is called the CauchyPoisson problem.

The equations under consideration have an exact solution h D h0 D const, ' D
'0 D u0 x 12 u20 t gh0 t describing the uniform motion of a fluid layer of constant
depth h0 with constant velocity u0 directed along the Ox-axis. Small perturbations
' D '0 C , h D h0 C of a given state are approximately described by the linear
Fig. 3.2 A sketch for flow with free surface over a flat bottom
3.2 Linear Theory of Surface Waves 83
system of equations
xx C yy C zz D 0 .0 < z < h0 /;
z D 0; z D 0;
)
t C u0 x z D 0
z D h0 :
t C u0 x C g D 0
We consider solutions to the Laplace equation in the form of wave packets satisfying
the impermeability condition at the bottom
p
D aei.kxCly!t/ ; D b ch mzei.kxCly!t/ ; m D k2 C l2 :
From the boundary conditions at z D h0 it follows that such solutions with nonzero
amplitudes a and b exist if and only if the frequency ! and wave vector k D .k; l/T
satisfy the dispersion relation
.! u0 k/2 D gm tanh mh0 : (3.6)
For plane waves we have l D 0 and m D jkj. In this case, the phase and group
velocities are expressed by
s
! p tanh kh0
cp D D u0 gh0 ;
k kh0
d! p
cg D D u0 gh0 f .kh0 /;
dk
where
s s !
d p 1 tanh
1
f .
/ D
tanh
D C :
d
2
cosh2
tanh
We emphasize the following particular cases.

(a) Stationarypwaves. In this case, the phase velocity of the wave vanishes: cp D 0,
p
and F D tanh kh0 =.kh0 / for the Froude number F D ju0 j= gh0 . The quantity
p
gh0 is called the critical velocity. Hence we can assert that linearp stationary
waves can occur only for the subcritical flow with velocity ju0 j < gh0 .
(b) Deep water waves. In the limit as h0 ! 1, the phase and group velocities p of
waves propagating
p in a fluid at rest (u 0 D 0) are expressed as c p D g=k and
cg D 12 g=k. Thus, for a deep water wave the group velocity is half the phase
velocity. p
(c) Long waves. In this limit case, we let k ! 0. Consequently, cp D u0 gh0 D
cg . The coincidence of the phase velocity and the group velocity indicates the
hyperbolicity of the long waves.
84 3 Water Waves
Problem 3.1 Find trajectories of particles of a plane travelling wave with the
velocity potential D ga! 1 eky sin.kx !t/; where ! and k are connected by the
dispersion relation ! 2 D gk in the linear theory of deep water waves (it is assumed
that the parameter D ak is small).
Solution The potential .x; y; t/ is a harmonic function of x and y for y < 0, satisfies
the decay condition r ! 0 as y ! 1 and the boundary conditions t D y ,
t C g D 0 on y D 0, where the function defining the free surface profile y D
.x; t/ has the form of a real wave packet .x; t/ D a cos.kx!t/. The trajectories of
particles with the velocity field u D r are described by the differential equations
dx
D g! 1 eky cos.kx !t/;
dt
(3.7)
dy
D g! 1 eky sin.kx !t/
dt
and the initial data .x.0/; y.0// D .

; /. The smallness of the parameter D ak D
2a=L implies the smallness of the ratio of the wave amplitude a to the length L.
Taking into account this property, we look for a solution x D .x; y/ in the form
x.t/ D x0 .t/ C x1 .t/ C O. 2 /:
Substituting the last expression into Eqs. (3.7) and collecting terms at the same
powers of , we obtain the following equation for x0 .t/ and x1 .t/:
dx0 dy0
D 0; D 0; .x0 ; y0 /jtD0 D .
; /;
dt dt
9
dx1 !
D eky0 cos.kx0 !t/;>>
=
dt k
.x1 ; y1 /jtD0 D .0; 0/:
dy1 ! >
>
D eky0 sin.kx0 !t/; ;
dt k
Hence .x0 .t/; y0 .t// D .

; / and, consequently,
1 k
x1 .t/ D e .sin k
sin.k
!t//;
k
1
y1 .t/ D ek .cos.k
!t/ cos k
/:
k
Thus, the functions x0 .t/ and x1 .t/ are periodic with period T D 2=! equal to
the time period of a progressive harmonic wave. For the period T the particle with
coordinates x.t/ D x0 .t/ C x1 .t/, taken up to a summand of order O. 2 /, runs
the circle jx xc j D r with radius r D aek and center xc D .
C aek sin k
,
aek cos k
/. Hence the oscillation amplitude of particles in the wave is maximal
3.3 Shallow Water Theory 85
on the fluid surface and exponentially decreases with increasing the submersion
depth.
Answer:
x.t/ D
C aek .sin k
sin.k
!t// C O.a2 k2 /;
y.t/ D aek .cos k
cos.k
!t// C O.a2 k2 /:
3.3 Shallow Water Theory
In the theory of nonlinear long waves, it is assumed that the characteristic wave-
length L is much larger than the fluid depth h0 . With these scales, we introduce the
dimensionless variables x0 , t0 , ' 0 , h0 :
L 0 p
.x; y/ D L.x0 ; y0 /; .z; h/ D h0 .z0 ; h0 /; tD p t; ' D L gh0 ' 0 :
gh0
Then the original equations with small parameter " D h0 =L take the form
"2 .'xx C 'yy / C 'zz D 0 .0 < z < h/;

'z D 0 .z D 0/;
9
ht C 'x hx C 'y hy D "2 'z >
=
1 .z D h/
't C .'x2 C 'y2 C "2 'z2 / C h D 0>
;
2
(we omit the primes in the notation of dimensionless quantities). The Lagrange
method in the theory of long waves uses the representation of the potential ' in
terms of its value on the bottom 'jzD0 D A.x; y; t/ as the power series
X1
z2n n
'.x; y; z; t/ D .1/n "2n A.x; y; t/;
nD0
.2n/ 2
where 42 D @2x C @2y is the Laplace operator with respect to the horizontal variables
x and y. By this representation, the boundary conditions on the free surface z D
h.x; y; t/ lead, in the approximation of lower order with respect to ", to the shallow
water equations for the functions h and u D rA which can be written in dimensional
variables as
ht C div .hu/ D 0;
(3.8)
ut C .u r/u C grh D 0;
86 3 Water Waves
where the operations div and r hold with respect to x and y. By the Cauchy
Lagrange integral, the pressure p in the fluid layer is expressed as
p.x; y; z; t/ D 0 g.h.x; y; t/ z/
with accuracy O."2 /, i.e., the pressure distribution is hydrostatic.

There is an analogy between gas dynamics and shallow water theory. Denoting
by D h the density and by P D 12 gh2 the pressure, we can write (3.8) as
t C div . u/ D 0;
1
ut C .u r/u C rP D 0;
(3.9)
1 2
PD g ;
2
which coincides with the equations of the isentropic flow of a polytropic gas with
polytropic exponent D 2.
Problem 3.2 At the time t D 0, a dam holding a storage basin of depth h0 is
suddenly destroyed. In the shallow water approximation, find the free surface shape
y D h.x; t/ for t > 0, the velocity u0 of the water front moving along a dry bottom,
and the water discharge q0 in the dam site.
Solution This problem can be regarded as the Cauchy problem for the one-
dimensional shallow water equations
ht C .uh/x D 0;
ut C uux C ghx D 0
with discontinuous initial data

(
h0 ; x < 0;
h.x; 0/ D u.x; 0/ D 0:
0; x > 0;
A gas-dynamic counterpart of this problem is the problem on the flow into vacuum
of a gas initially at rest. The continuous solution for t > 0 is represented as a simple
centered wave propagating
p to the left through the fluid at rest with the characteristic
velocity c0 D gh0 equal to the critical velocity for a given water storage basin.
The equations for this simple wave have the form
p
u C 2 gh D 2c0 ;
p x
u gh D :
t
3.3 Shallow Water Theory 87
The first relation is the condition that the Riemann invariant is constant, and the
second relation is the equation of a centered family of rectilinear characteristics.
Hence we find the explicit dependence of u and h on x and t in the domain c0 <
x=t < 2c0 :
2
u.x; t/ D .x C c0 t/;
3t
1
h.x; t/ D .x 2c0 t/2 :
9gt2
Thus, the free surface has the shape of a parabola

p with vertex x D 2c0 t travelling to
the right with the velocity u0 D 2c0 D 2 gh0 . The depth and velocity of water in
the dam site remain unchanged during the whole time of motion: h.0; t/ D .4=9/h0
and u.0; t/ D .1=3/c0 . Thus, we find the water discharge q0 D .4=27/h0c0 , i.e., the
water amount flowing out from the dam each second.
Answer:
8

<h0 ;
x < c0 t;
h.x; t/ D .x 2c0 t/ =.9gt /; c0 t < x < 2c0 t;
2 2

:0; x > 2c0 t;
p
u0 D 2c0 ; q0 D .4=27/h0c0 .c0 D gh0 /:
Discontinuous solutions to the hyperbolic shallow water equations describe

propagation of bore type waves where the depth and velocity of the fluid change
abruptly. To describe such motions, the system (3.8) is taken for conservation laws
of mass and depth-averaged horizontal momentum equation which, in the case of
the one-dimensional motion, can be written as
@t h C @x .uh/ D 0;
1 (3.10)
@t .uh/ C @x .u2 h C gh2 / D 0:
2
The system (3.10) implies the relations on the strong discontinuity
1
Dh D uh; Duh D u2 h C gh2 ;
2
where D is the bore velocity.
88 3 Water Waves
3.4 Shear Flows of Shallow Water
In the case of two-dimensional flows, the Euler equations (3.2) are written in
dimensionless variables as
ut C uux C vuy C px D 0;
"2 .vt C uvx C vvy / C py D 1;
ux C vy D 0;
where u.x; y; t/ is the horizontal velocity, v.x; y; t/ is the vertical velocity, and
p.x; y; t/ is the fluid pressure. In the long wave approximation, the vertical momen-
tum equation implies the hydrostatic law
py D 1:
Using the dynamic condition on the free boundary y D h.x; t/ and excluding the
pressure from the horizontal momentum equation, we obtain the following system
of equations written in dimensional variables:
ux C vy D 0;
(3.11)
ut C uux C vuy C ghx D 0:
Integrating the continuity equation with respect to y and using the kinematic
condition on the free boundary, we obtain the integro-differential equations
0 h 1
Z
@ @ 0 0A
ht C u.t; x; y /dy D 0;
@x
0
ut C uux C vuy C ghx D 0; (3.12)

Zy
vD ux .t; x; y0 /dy0 ;
0
called the Benney equations or shallow water equations for shear flows. They
coincide with the classical shallow water equations if the horizontal velocity u is
independent of the vertical variable y.
Methods for studying equations of the form (3.12) use a generalization of the
theory of nonlinear hyperbolic equations. To write these equations in the form
convenient for our purposes, we introduce the new independent variable by setting
y D Y.x; ; t/, where Y.x; ; t/ is a solution to the Cauchy problem
Yt Ce
uYx D e
v;
(3.13)
YjtD0 D h0 .x/ ;
3.4 Shear Flows of Shallow Water 89
where eu.x; ; t/ D u.x; Y.x; ; t/; t/, e

v.x; ; t/ D v.x; Y.x; ; t/; t/, and the function
h0 .x/ determines the initial location of the free boundary: htD0 D h0 .x/. This
change of variables parametrizes the initial domain D0 D f1 < x < C1; 0 <
y < h0 .x/g in such a way that D0 is represented as the union of the one-parameter
family of curves W y D h0 .x/, where the value D 0 corresponds to the bottom
y D 0, and D 1 corresponds to the initial location of the free boundary. Then the
evolution of each initial curve is described by the equation y D Y.x; ; t/. At
each time t > 0, the value D 1 corresponds to the free boundary y D h.x; t/.
By the definition of the functions e u and e
v , we have
e
ut D ut C uy Yt ; e
ux D ux C uy Yx ; e
u D uy Y ;
v D ux Y D .e
e ux Y e
u Yx /:
ut Ce
Consequently, e ueux C ghx D 0. Differentiating (3.13) with respect to , we get
.Y /t C .e
uY /x D 0:
Since
Z1
h.x; t/ D Y .x; ; t/d;
0
we finally obtain the shallow water equations for shear flows
.H/t C .e
uH/x D 0;
Z1 !
(3.14)
e
ut Ce
ueux C g Hd D0
0 x
for H.x; ; t/ D Y .x; ; t/ and e

u.x; ; t/. The system (3.14) can be treated as an
infinite-dimensional system of equations parametrized by the continuous parameter
in the interval 0 6 6 1. Such systems can be represented in an abstract form
[31] as follows. Let B be a Banach space of vector-valued functions u.x; ; t/, and
let A be a linear operator acting on such vector-valued functions only with respect
to : Ahi W B ! B. We consider the equation
ut C A.u/hux i D 0 (3.15)
for u. The characteristic of the system (3.15) is defined by the equation dx=dt D
c.x; t/, where c is an eigenvalue of the problem
.F; .A cI/hfi/ D 0:
90 3 Water Waves
Here, F is a linear continuous functional in the dual B of the space B and f 2 B is

a test vector. The action of the proper functional F on the system (3.15) yields the
relation on the characteristic
.F; ut C cux / D 0:
The system (3.15) is called hyperbolic if all eigenvalues c are real and the
corresponding proper functionals F form a complete system. In this case, the
relations on characteristics are equivalent to the system (3.15).
3.5 Nonlinear Dispersive Equations
Using higher order approximations with respect to the parameter ", we can take into
account the dispersion properties of nonlinear long water waves. We consider the
two-dimensional potential motion in the .x; y/-plane. In this case, it is convenient
to represent the velocity field in terms of the stream function: u D . y ; x /.
The stream function and the velocity potential form a pair of conjugate harmonic
functions connected by the CauchyRiemann equations
'x D y;
'y D x:
We write the original p

equations of fluid motion in terms of the dimensionless stream
function 0 D =.h0 gh0 /
"2 xx C yy D 0 .0 < y < h/; D 0 .y D 0/;

ht C . x C y hx /jyDh D 0;
1 @ 2
. yt "2 hx xt /jyDh C f" 2
x .x; h; t/ C 2
y .x; h; t/g C hx D 0;
2 @x
where we omit the primes to simplify the notation. The last equation is obtained by
differentiating the CauchyLagrange integral with respect to x at points of the free
surfaces (the condition p.x; h.x; t/; t/ D const is taken into account). Further, we
introduce the depth-averaged horizontal fluid velocity
Z
h.x;t/
1
u.x; t/ D 'x .x; y; t/dy
h.x; t/
0
such that jyDh D uh for the stream function . The function admits the
following representation in terms of u and h in .t/, similar to the Lagrange series
3.5 Nonlinear Dispersive Equations 91
for the potential ':
1
D uy C "2 .h2 y y3 /uxx C O."4 /:
6
Substituting this expression into the boundary condition at y D h.x; t/ and leaving
quantities of order up to "2 we obtain the second-order long wave approximation
(called the SerreSuGardner equations in the one-dimensional case) written in
dimensional variables as
ht C .uh/x D 0;
1 3 (3.16)
ut C uux C ghx D .h .uxt C uuxx u2x //x :
3h
The terms involving third order derivatives yield the dispersion correction to the
classical shallow water equations. In the case of small perturbations h D h0 C ,
u D u0 C v of the uniform flow, the linearized equations (3.16) take the form
t C u0 x C h0 vx D 0;
1 2
vt C u0 vx C gx D h .vt C u0 vx /xx :
3 0
Hence for the elementary wave packets
.x; t/ D a expfi.kx !t/g;

v.x; t/ D b expfi.kx !t/g
we obtain the dispersion relation
gh0 k2
.! u0 k/2 D :
1 C 13 h20 k2
Comparing with the exact dispersion relation (3.6), we see that the linearization
of (3.16) is equivalent to the fractional rational approximation of the function
tanh
in the limit of long waves or small depth as
D h0 k ! 0.
The dispersive terms in (3.16) can be written in other form by using the operator
of total differentiation dt D @t C u@x with the average velocity u. The first equation
in (3.16) implies the relation
dt2 h .@t C u@x /2 h D h.uxt C uuxx u2x /

92 3 Water Waves
which in dimensional variables leads to another form of the SerreSuGardner

equations
ht C .uh/x D 0;
1 2 2 (3.17)
ut C uux C ghx C .h dt h/x D 0:
3h
As in the case of the usual shallow water equations, this approximate model can
be interpreted as the gas-dynamic equations (3.9), but with the following more
complicated equation of state
1 2 1 2 2
PD g C dt :
2 3
For the class of motions described by the functions h D 1 C "2 , u D "2 v

in dimensional variables the nonlinear dispersion equations are simpler than (3.16)
and (3.17). Such a modelling means that we deal with weakly nonlinear dispersive
shallow water waves of small amplitude. In this case, the system (3.16) leads to the
approximate equation
ht C .uh/x D 0;
1 (3.18)
ut C uux C ghx h20 uxxt D 0
3
and the SerreSuGardner system leads to the equations
ht C .uh/x D 0;
1 (3.19)
ut C uux C ghx C h0 hxtt D 0;
3
Both systems (3.18) and (3.19) are called the Boussinesq equations.
An approximate description of long waves travelling in only one direction, to the
left or to the right, is obtained in the independent variables D "2 t,
D x c0 t
with c20 D gh0 . This stretching transformation of t means that the long time wave
evolution is observed in the slow time scale. In this case, the system (3.16), up to a
quantity of order O."4 /, is reduced to the Korteweg-de Vries equation

3 1
t C c0 1 C x C c0 h20 xxx D 0
2h0 6
written in the original dimensional variables for the function .x; t/ interpreted as
the elevation of the free surface in a weakly nonlinear long wave on the surface of a
fluid of finite depth h0 .
3.5 Nonlinear Dispersive Equations 93
The multi-dimensional counterpart of the SerreSuGardner equations is often

called the GreenNaghdi equations
ht C div.hu/ D 0;
1
dt u C grh C r.h2 dt2 h/ D 0; (3.20)
3h
dt D @t C u r;
where the operations div and r are taken with respect to the horizontal variable
x D .x; y/ and the vector u, as in the one-dimensional case, is interpreted as the
depth-averaged horizontal fluid velocity
Z
h.x;t/
1
u.x; t/ D rx '.x; z; t/dz:
h.x; t/
0
Since the vector structure of the GreenNaghdi equations (3.20) is similar to the
structure of the main hydrodynamics equations, similar conservation laws and the
first motion integrals hold, in particular, the conservation law of total momentum
.hu/t C div .hu u Ce

pI/ D 0;
1 2 1 2 2 (3.21)
e
pD gh C h dt h:
2 3
From the practical point of view it is not convenient to use the momentum equation
in the form (3.21) since the momentum flux contains second order time derivatives
of the function h. Therefore, instead of the velocity vector u, we introduce the new
sought vector-valued function
1
vDuC r.h2 dt h/
3h
and write the local momentum equation (the second equation in (3.20)) as
1 1 1
dt v C dt hr.h2 dt h/ dt .r.h2 dt h// C grh C r.h2 dt2 h/ D 0;
3h2 3h 3h
where we used the identity dt .rf / D r.dt f / .u0x /T rf for any smooth function
f .x; t/ 2 C2 . After simple transformations, we obtain the relation
T
@u 1
dt v C .v u/ C grh r.dt h/2 D 0
@x 2
94 3 Water Waves
which can be written as

1 2 1 2
vt C curl v u C r v u juj C gh .dt h/ D 0; (3.22)
2 2
where the operation curl and the vector product are applied to the three-dimensional
vectors .v; 0/ and .u; 0/. A consequence of (3.22) is the Helmholtz equation in the
form
t C curl . u/ D 0
for the generalized vorticity vector D curl v. In particular, this implies the
conservation of the generalized circulation
I
D v dx
C.t/
is preserved along any contour C.t/ in the horizontal plane x D .x; y/ consisting of
the same particles moving with the velocity field u. Another consequence of (3.22)
is a counterpart of the CauchyLagrange integral
1 1
't C v u juj2 C gh .dt h/2 D b.t/
2 2
for generalized potential flows in the case v D r'.
3.6 Stationary Surface Waves
In the reference frame moving with a travelling wave, the motion is described
by the stationary solution, where the sought functions are independent of t. The
problem for two-dimensional stationary surface waves is formulated as follows.
Find a stream function .x; y/ and a function h.x/ > 0 such that
xx C yy D0 .0 < y < h.x//;

.x; 0/ D 0; .x; h.x// D Q; (3.23)
2 2
x C y C 2gh D 2b .y D h.x//;
where b is the Bernoulli constant and Q is the flow discharge, constant on each
vertical cross-section of the layer.
The problem (3.23) can be reduced to finding solutions to a single integro-
differential equation for a function determining the free surface shape. The complex
potential w.z/ D '.x; y/ C i .x; y/ realizes a conformal mapping from the flow
3.6 Stationary Surface Waves 95
domain in the plane z D x C iy to the strip 1 < ' < C1, 0 < < Q.
Passing to dimensionless variables in which the strip has unit width, we look for
the inverse mapping z D w C Z.w/. For solitary wave type solutions the analytic
function Z.w/ D X.'; / C iY.'; / in the strip 0 < < 1 should satisfy the
boundary conditions
YD0 . D 0/;
1
C 2F 2 Y D 1 . D 1/;
.1 C X' /2 C Y'2
where F 2 D Q2 =gh30 is the square Froude number (h0 is the unperturbed fluid depth).
Furthermore, the decay condition Z ! 0 as j'j ! C1 should hold. We introduce
the function .'/ D Y.'; 1/ describing the free boundary shape. We define the
auxiliary operator N of normal derivative sending .'/ to the derivative Y .'; 1/,
where Y is the solution to the Dirichlet problem
Y' ' C Y D 0 .0 < < 1/;

Y.'; 0/ D 0; Y.'; 1/ D .'/:
We consider the analytic function f .w/ D 1=.1 C dZ=dw/ 1, where Im f D 0 at

D 0 and f .w/ ! 0 as j'j ! C1. We have
1 C X'
Re f D 1;
.1 C X' /2 C Y'2
Y'
Im f D :
.1 C X' /2 C Y'2
By the CauchyRiemann equations .Re f /' D .Im f / and the definition of N, we

obtain the following relation at the boundary D 1:
!
1 C X' Y'
D N :
.1 C X' /2 C Y'2 .1 C X' /2 C Y'2
'
Making simple transformations and taking into account the boundary conditions for
the function Z, we obtain the required equation [16] for the function
1
F 2 C N C N. 2 / D 0:
2
The nonzero solution corresponding to a solitary surface wave bifurcates from the
zero solution D 0 at the point F D 1 and exists in the interval 1 < F < 1:290 (the
numerical estimate).
96 3 Water Waves
An approximate description of solutions to the problem (3.23) can be obtained

by looking for stationary solutions to the SerreSuGardner equations (3.17). In
this case, from the first equation in (3.17) we obtain the flow discharge integral
uh D Q D const . Excluding u from the second equation with the help of the above
integral and integrating twice, we obtain the first order ordinary differential equation
(the BoussinesqRayleigh equation) for the function h

1 2 dh 2
Q D gh3 C bh2 2ch C Q2 ; (3.24)
3 dx
where b and c are the integration constants connected with the roots hi (i D 1; 2; 3)
of the cubic polynomial on the right-hand side of the Vieta formulas
2b Q2
h1 C h2 C h3 D ; h1 h2 h3 D :
g g
The solutions are nontrivial if all the roots are real and the function h takes the values
in the interval h1 6 h2 < h < h3 . The free surface shape is implicitly given by the
quadrature formula
Zh3
Q ds
xD p p :
3g .s h1 /.s h2 /.h3 s/
h
Consequently, as in the case of the Korteweg-de Vries equation (cf. Sect. 2.6), the
shape of the free surface is given by the equation
h.x/ D h2 C .h3 h2 /cn2 .rxI m/;

p
3g.h3 h1 / h3 h2
rD ; m2 D :
2Q h3 h1
This cnoidal wave has amplitude a D .h3 h2 /=2 and is periodic with period
Zh3
2Q ds
LD p p :
3g .s h1 /.s h2 /.h3 s/
h2
In the limit of waves of small amplitude a ! 0, we obtain the value m D 0, so that

the Jacobi amplitude becomes a linear function D rx and a cnoidal wave takes
the form of an elementary wave packet h.x/ D h0 C a cos kx with the wavenumber
k D 2r and the average fluid depth h0 D .h2 C h3 /=2. Introducing the phase velocity
u0 by the identity u20 C2gh0 D 2b (it is legitimate since the parameter b has the sense
of the Bernoulli constant), we obtain the dispersion relation
1
F2 D ;
1 C 13 h20 k2
3.7 Waves in Two-Layer Fluids 97
p
where F D u0 = gh0 is the Froude number. Thus, in the linear limit, the stationary
wave turns out to be subcritical, which agrees with the conclusions of Sect. 3.2.
In the other limit case, where the root h2 ! h1 becomes double, the elliptic
cosine is transformed to an elementary function: cn.
I m/ ! 1= cosh
as m ! 1.
In the limit, the period L of the cnoidal wave infinitely increases and is transformed
to a solitary wave with the amplitude a D h3 h2 > 0 and the free surface profile
r
3 hh32 1
a
h.x/ D h0 C ; rD q :
cosh 2 rx 2 hh32 h2
The quantity h0 D h2 yields the asymptotics of the layer depth as jxj !p 1.

Therefore, in this case, it is natural to define the Froude number F D u0 = gh0
from the fluid velocity u0 at infinity (i.e., from the condition u20 C 2gh2 D 2b).
Hence we obtain the Boussinesq formula for the parameters of the solitary wave
a
F2 D 1 C :
h0
By this formula, this nonlinear wave is supercritical.

The constructed solution to the GreenNaghdi equation yields waves of an
arbitrarily large amplitude, whereas the exact solution to the problem (3.23) exists
only in a finite interval of amplitudes. The limiting solitary or periodic wave of finite
amplitude has a cusp with the angle 120 at the wave crest.
3.7 Waves in Two-Layer Fluids
Consider the plane potential motion of an ideal fluid consisting of two layers j .t/
(j D 1; 2) of densities 1 2 that are separated by the surface y D h.x; t/. We
assume that the fluid is bounded by the flat bottom y D 0 from below and by an
impermeable lid y D H from above.
The velocity potentials 'j and function h satisfy the equations
'jxx C 'jyy D 0; x D .x; y/T 2 j .t/ .j D 1; 2/;

'1y D 0 .y D 0/; '2y D 0 .y D H/;
9
ht C 'jx hx 'jy D 0; .j D 1; 2/ >
>
> >
>
1 2 1 2 >
=
1 '1t C '1x C '1y C gh
2 2 y D h.x; t/:
>>
>
>
1 2 1 2 >
D 2 '2t C '2x C '2y C gh > ;
2 2
98 3 Water Waves
We recall that the kinematic condition admits a nonzero jump of the tangent fluid
velocity through the interface, whereas the dynamic condition means the continuity
of pressure.
The linear theory deals with small perturbations of the piecewise constant
horizontal flow with rectilinear boundary h.x; t/ D h1 D const and constant
velocities uj (i.e., with the potentials '0j .x; y; t/ D uj x .gh1 C 12 u2j /t). Looking
for solutions to the linearized equations in the form of wave packets
h D h1 C aei.kx!t/ ;
'1 D '01 C A1 cosh kyei.kx!t/ ;
'2 D '02 C A2 cosh k.H y/ei.kx!t/
with constant amplitudes a, A1 , and A2 , we obtain the dispersion relation
1 .! u1 k/2 coth kh1 C 2 .! u2 k/2 coth k.H h1 / D . 1 2 /gk (3.25)
which is the quadratic equation for the frequency !. If this equation has a pair of
complex roots, then the main flow is unstable. Thus, the frequency ! is complex
for all wavenumbers k in the case 1 < 2 , where a heavier fluid occurs in the
upper layer. Such an instability of the layer interface is called the RayleighTaylor
instability. If 1 > 2 , then the roots are real if and only if
g
.u2 u1 /2 6 .tanh k.H h1 / C tanh kh1 / ; (3.26)
k
where D 2 = 1 and D 1 . If u2 u1 (the relative velocity of the motion
of layers differs from zero), then the inequality (3.26) fails for sufficiently large jkj.
Consequently, in the case of sliding of layers, we have the short-wave instability,
called the KelvinHelmholtz instability.
The shallow water approximation for a two-layer fluid is obtained by represent-
ing the potentials 'j .x; y; t/ as the Lagrange series
X1
y2n 2n
'1 D .1/n "2n @ A1 .x; t/;
nD0
.2n/ x
X1
.H y/2n 2n
'2 D .1/n "2n @x A2 .x; t/;
nD0
.2n/
where A1 .x; t/ D '1 jyD0 and A2 .x; t/ D '2 jyDH . The two-layer shallow water
equations for h, u1 D A1x , and u2 D A2x have the form
ht C .u1 h/x D 0;
.H h/t C .u2 .H h//x D 0;
u1t C u1 u1x C ghx D .u2t C u2 u2x /:
Summarizing the first two equations, we get
u1 h C u2 .H h/ D Q.t/;
where Q is an arbitrary function of t. Assuming that the flow is constant at x1 and

using the Galilean invariance property of the equations of motion, we can assume
without loss of generality that Q D 0. Then we obtain an expression for the fluid
velocity in the upper layer u2 D u1 h=.H h/. As a result, the equations of two-
layer shallow water under a rigid lid are reduced to the system of two equations
ht C .u1 h/x D 0;
H h H 2 .2 C /Hh C h2
u1t C u1 u1x
Hh .H h/2 (3.27)

H 2 2
C g u hx D 0:
.H h/3 1
This system of equations is hyperbolic if and only if

u21 h 2 h
< 1 1 : (3.28)
gH H H
It is easy to verify that the inequality (3.28) agrees with the stability criterion (3.26)
in the limit of long waves as k ! 0. Thus, the two-layer shallow water equations
form an evolution system of mixed type unlike the case of one-layer shallow water
equations that are hyperbolic equations. A simpler approximate model is obtained
by the additional simulation
1CA p 2
h!H ; u2 u1 ! gHB; t! p t
2 gH
with small parameter (approximation in the case of fluid flows with close
densities). Considering the terms of leading order with respect to the parameter ,
we obtain the system of equations
At C ..A2 1/B/x D 0;
(3.29)
Bt C ..B2 1/A/x D 0:
100 3 Water Waves
Nonlinear stationary waves in a two-layer fluid under a rigid lid are described by
the following equations for the stream functions j .x; y/ (j D 1; 2) and the function
h.x/ (0 < h.x/ < H):
1xx C 1yy D0 .0 < y < h.x//;

2xx C 2yy D0 .h.x/ < y < H/;
1 .x; 0/ D 0; 1 .x; h.x// D 2 .x; h.x// D Q1 ;
(3.30)
2 .x; H/ D Q1 C Q2 ;
2 2
1 . 1x C 1y C 2gh 2b1 /
2 2
D 2 . 2x C 2y C 2gh 2b2 / .y D h.x//;
where Qj is the water discharge and bj is the Bernoulli constant for the jth layer. In
the absence of waves, the flow with rectilinear interface and constant velocities is
described by the solution
h.x/ D h1 D const .0 < h1 < H/;

1 .x; y/ D u1 y .0 < y < h1 /;
2 .x; y/ D u2 .y h1 / C u1 h1 .h1 < y < H/:
This solution is obtained for the following values of the discharges and Bernoulli
constants:
Q1 D h 1 u 1 ; Q2 D h 2 u 2 ;
(3.31)
b1 D u21 C 2gh1 ; b2 D u22 C 2gh1 ;
where h2 D H h1 is the depth of the unperturbed upper fluid layer. Theses values
of Qj and bj should be involved in Eqs. (3.30) and in searching solitary wave type
solutions with the asymptotic behavior h.x/ ! h1 , r j ! .0; uj / (j D 1; 2) as
x ! 1.
The second-order shallow water approximation yields the following nonlinear
ordinary differential equation for h.x/:
2
1 2 2 dh
. 1 Q1 .H h/ C 2 Q2 h/ D P.h/; (3.32)
3 dx
where P.h/ D h.h H/.. 1 2 /gh2 2. 1 b1 2 b2 /h Cc/C 1 Q21 .H h/C 2 Q22 h
and c is the integration constant. For solitary waves the conditions h.x/ ! h1 and
h0 .x/ ! 0 as x ! 1 imply
c D 1 u21 h1 C 2 u22 h2 g. 1 2 /h21 C 2. 1 b1 2 b2 /h1 :

Fig. 3.3 A sketch for smooth bore type solutions between horizontal top and bottom rigid walls
Equation (3.32), obtained by L. V. Ovsyannikov for an approximate description of

stationary waves in a two-layer fluid, is a counterpart of the BoussinesqRayleigh
equations (3.24) in the theory of surface waves. If the polynomial P.h/ of the 4th
degree has one double root h D h1 and two simple roots different from h, then
Eq. (3.32) has a solution in the form of a symmetric solitary wave of elevation or
depression. In the case of two double roots, we have nonsymmetric smooth bore
type waves (cf. Fig. 3.3).
We write Eq. (3.32) in dimensionless variables by setting h.x/ D h1 .1 C .x//,
x D h1 x and introducing the ratio r D h2 =h1 and the densimetric Froude numbers
1 u201 2 u202
F12 D ; F22 D :
g. 1 2 /h1 g. 1 2 /h2
According to Eq. (3.32), for the function expressing the deviation of the interface
from the equilibrium state we obtain the equation
2
d 3 2 F12 .r / C rF22 .1 C / .1 C /.r /
D : (3.33)
dx F12 .r / C r3 F22 .1 C /
The interval 1 < < r corresponds to the domain of admissible values 0 <
h.x/ < H. Since for such the denominator of the fraction on the right-hand side
of (3.33) is positive, the numerator should be positive in a neighborhood of D 0,
which is valid if and only if F1 and F2 satisfy the inequality
F12 C F22 > 1: (3.34)
This condition has a certain sense from the point of view of the dispersion properties
of nonlinear waves. Namely, in the case of stationary waves (i.e., wave packets
with the frequency ! D 0 and unknown dimensionless wavenumber
D kh0 ),
the dispersion relation (3.25) is written in dimensionless variables as
F12
coth
C F22 r
coth r
D 1:
102 3 Water Waves
Since the even function f .

/ D
coth
is strictly monotonically increasing for
> 0, the real roots

2 R can exist only if F12 C F22 6 1. Thus, linear stationary
waves in a two-layer fluid of the form of sinusoidal wave packets exist only in the
subcritical domain F12 C F22 6 1, whereas Eq. (3.33) describes nonlinear waves
existing in the supercritical domain (3.34). In particular, smooth bore type solutions
are obtained for the Froude numbers
1Ca ra
jF1 j D p ; jF2 j D p ;
1Cr r.1 C r/
where the dimensionless parameter a .1 < ap< r/ is the p bore amplitude. The
locus of points .F1 ; F2 / is the rhombus jF1 j C rjF2 j Dp 1 C r which p ptouches
the inscribed circle F12 C F22 D 1 at the points .1= .1 C r; r= 1 C r/
corresponding to the value a D 0 (cf. Fig. 3.4). For small a the bore profile is given
by the approximate formula

a ax 3
.x/ D 1 C tanh ; 2 D :
2 2 r.1 r C r2 /
Fig. 3.4 The unit circle in the plane of densimetric Froude numbers .F1 ; F2 / separates the
subcritical (inside the circle) and supercritical (outside the circle) domains. The rhombus touches
the circle at four points from which a one-parameter family of smooth bore type solutions
bifurcates.
3.8 Waves in Stratified Fluids 103
3.8 Waves in Stratified Fluids
The system (3.1) at rest u D 0 is reduced to the equation
rp D g
which implies p D p0 .z/, D 0 .z/, where p0 and 0 are connected by the

hydrostatic law
dp0
D g 0 .z/:
dz
Consequently, under the action of the gravity force, an inhomogeneous fluid at rest
splits by the planes z D const into horizontal layers (strata) of constant density.
This phenomenon is called stratification and the fluid is referred to as a stratified
fluid. In this sense, the two-layer fluids considered in the previous section provide
the simplest examples of stratification, where the density is piecewise constant. The
continuous stratification can also occur in the horizontal shear flow described by
the exact solution u D u0 .z/, v D v0 .z/, w D 0, D 0 .z/, p D p0 .z/ with
arbitrary smooth functions u0 , v0 , 0 , and the pressure p0 connected with 0 by the
hydrostatic law.
In the general case, from the second equation in (3.1) it follows that the density
is constant along the trajectories of fluid particles, i.e., the integral curves of the
system of ordinary differential equations
dx
D u.x; t/:
dt
Therefore, each level surface of the density .x; t/ D const consists of the same
fluid particles, which means that it is conserved as a material object. Perturbations
propagating along such interfaces within a fluid are called internal waves.
To describe the wave motion in a fluid layer of finite depth 0 < z < h.x; y; t/
bounded by the flat bottom z D 0 and the free surface z D h.x; y; t/, we impose the
boundary conditions
w D 0 .z D 0/;
ht C uhx C vhy D w; p D p0 D const .z D h/:
To exclude the surface waves from consideration, the rigid impermeable lid z D h0
is taken for the upper layer boundary, where the impermeability condition w D 0 is
imposed.
104 3 Water Waves
Small perturbations u D u0 C u0 , D 0 C 0 , p D p0 C p0 of the shear flow with

the velocity vector u0 D .u0 .z/; v0 .z/; 0/T are described by the linearized system of
equations
u0x C vy0 C w0z D 0;

D0 0 C 0z w0 D 0;
1 0
D0 u 0 C p C u0z w0 D 0;
0 x
1 0
D0 v 0 C p C v0z w0 D 0;
0 y
1 0 g
D0 w0 C pz C 0 D 0;
0 0
where D0 D @t C u0 @x C v0 @y . Successively excluding the sought functions, we

reduce the system to a single equation for the vertical fluid velocity w0
@ @w0
0 .N 2 C D20 /2 w0 C D20 0
@z @z
@w0 @w0
D0 . 0 u0z /z C . 0 v0z /z D 0; (3.35)
@x @y
where 2 D @2x C @2y . The quantity N D N.z/ is defined by the formula
g 0z
N2 D
0
and has the same dimension as the frequency. The quantity N D N.z/ is called the
BruntVisl frequency (or the buoyancy frequency). To describe internal waves
in a weakly stratified fluid, one often uses the Boussinesq approximation (do not
confuse with the Boussinesq model in the shallow water theory!). In accordance
with the Boussinesq approximation, the coefficients 0 and N in (3.35) are constants:
0 .z/ 00 D const > 0 and N.z/ N0 D const 0. If there is no velocity shear,
we have the equation
wtt C N02 2 w D 0;
where D @2x C @2y C @2z .

Example 3.2 Consider the process of generating internal waves in a weakly strat-
ified fluid occupying the whole space R3 . We assume that a point source of
perturbations is concentrated at the origin x D 0, oscillates with a small amplitude
and a given frequency !0 , and perturbs the wave field in the form of elementary
wave packets w D a expfi.kx C ly C mz !0 t/g. In this case, the components of the
3.8 Waves in Stratified Fluids 105
wave vector k D .k; l; m/T are connected with the frequency !0 by the dispersion
relation
k2 C l2
!02 D N02 :
k2 C l2 C m2
It is clear that the frequency of generated waves cannot exceed the BruntVisl
frequency: !0 6 N0 . Furthermore, the angle 0 at which the wave vector is inclined
to the horizontal Oxy-plane is known:
!
0
0 D arc cos :
N0
Since the energy of wave packets is transferred from the perturbation source along
the direction of the group velocity vector cg D r!.k/, it is of interest to understand
the relative orientation of the vectors cg and k. By the dispersion relation, the
frequency ! D !.k/ can be regarded as a homogeneous function of degree 0 in
k. Therefore, the group velocity vector is always orthogonal to the wave vector:
k cg .k/ D 0. Consequently, the wave motion perturbed by a point source of
oscillations is concentrated in a neighborhood of a conical surface with the axis
coinciding with the Oz-axis and the opening angle 20 .
In the case of a stratified fluid layer of finite deep h0 that is unbounded in the
horizontal direction, the internal waves are described by elementary wave packets
w.x; y; z; t/ D W.z/ei.kxCly!t/ ;
where the amplitude W depends on the vertical variable z. We denote
.z/ D ku0 .z/ C lv0 .z/ !; m2 D k2 C l2 :
In view of (3.35), for the amplitude W we obtain the differential equation

d 2 d W W
0 D 0 m2 . 2 N 2 / .0 < z < h0 /:
dz dz
The boundary conditions for a fluid under a rigid lid take the form
W.0/ D W.h0 / D 0
and, in the presence of the free surface,
W D 0 .z D 0/;
d W W
2 D gm2 .z D h0 /:
dz
106 3 Water Waves
In these equations with a given wave vector k D .k; l/T and the unknown function
W, the parameter ! involved in the coefficient and the condition on the free surface
is also unknown. The values of ! for which there is a nonzero solution W form the
spectrum. In the case of internal waves propagating in a fluid at rest (u0 D v0 D 0)
under a rigid lid, the spectral problem takes the form

2 N2
. 0 Wz /z D 0 m 1 2 W .0 < z < h0 /;
! (3.36)
W.0/ D W.h0 / D 0:
If the upper boundary of the fluid layer is free, then the boundary conditions take
the form
m2
W.0/ D 0; W 0 .h0 / D g W.h0 /:
!2
In the general case, the spectrum of the problem (3.36) can be complex. We note
that if ! 2 is complex, then the eigenfunction W is also complex. Multiplying by the
complex-conjugate function W, integrating over the interval .0; h0 /, and taking into
account the boundary conditions, we obtain the relation
Zh0 Zh0
2 2 2 2 2
! 0 .jWz j C m jWj /dz D g 0 .h0 /jW.h0 /j C 0 N 2 jWj2 dz:
0 0
If the density 0 .z/ does not decrease with decreasing z, we have N 2 .z/ > 0 for
all z 2 .0; h0 /. In this case, ! 2 > 0 and the spectrum is real. The following
assertions give more detailed information on properties of the spectrum of the
problem (3.36).
(a) There exists a countable family of wave modes !n2 .m/ (n D 1; 2; : : :) such that
!12 .m/ > !22 .m/ > : : : > !n2 .m/ > : : : and !n2 .m/ ! 0 as n ! 1.
(b) The eigenfunction Wn .z/ has exactly n zeros on the interval 0; h0 if the upper
boundary is free and exactly n C 1 zeros in the case of the wave motion under a
rigid lid.
(c) For each wave mode the phase velocity cn .m/ D !n .m/=m is a strictly
monotonically decreasing function of the parameter m (the modulus of the wave
vector).
Problem 3.3 In the Boussinesq approximation, find the spectrum of phase veloci-
ties and eigenfunctions of the problem on two-dimensional linear internal waves in
the layer of a stratified fluid under a rigid lid. What is the group velocity for each
wave?
Solution We set m2 D k2 in (3.36) for the plane motion, where k is the wavenumber.
We also set 0 .z/ D const and N.z/ D N0 D const in accordance with the
3.9 Stability of Stratified Flows 107
Boussinesq approximation. In the case ! 6 N0 , we denote

N2
2 D m2 1 02 :
!
As a result, we obtain the equations
Wzz C 2 W D 0 .0 < z < h0 /;

W.0/ D W.h0 / D 0;
which imply D n D n=h0 , 'n .z/ D sin n z .n D 1; 2; : : : /. Expressing ! in

terms of , we find the wave modes !n2 .k/ D N02 k2 =.k2 C 2n /. Respectively, the
.n/ .n/
phase cp D !n .k/=k and group cg D d!n .k/=dk velocities have the form
N0
c.n/
p .k/ D p 2 ;
k C 2n
N0 2n
c.n/
g .k/ D :
.k2 C 2n /3=2
.n/ .n/
It is clear that jcg .k/j < jcp .k/j for all k 0, where equalities hold only in the
.n/ .n/
long-wave limit k D 0 which provides the maximal value cp D cg D N0 h0 =.n/
of the phase and group velocities of the nth mode. In the case ! > N0 , there are no
nontrivial solutions to the problem under a lid on the upper boundary of the layer.
3.9 Stability of Stratified Flows
If the spectrum is such that Im ! 6 0, then the basic flow with the velocity vector
u0 D .u0 .z/; v0 .z/; 0/T is stable; otherwise, we have instability. In particular, if the
eigenfunction W.z/ of the problem (3.36) has zero at the point z D h in .0; h0
and N 2 .z/ < 0 for z 2 .0; h /, then for the corresponding eigenvalue we have
! 2 < 0. Consequently, the state at rest or, which is equivalent, the uniform flow with
u0 D const and v0 D const is unstable. For example, a constant flow is unstable if
the fluid density 0 .z/ is a monotonically increasing function on .0; h0 /.
In order to analyze the stability of the plane shear flow u0 D .u0 .z/; 0; 0/T , where
the wave packets w.x; z; t/ D W.z/ei.kx!t/ propagate, it is convenient to take the
phase velocity c D cp .k/ D cr C ici for the spectral parameter. The instability takes
place in the case ci > 0. Replacing the amplitude function W.z/ D .z/ .z/, for
the flow under a rigid lid we obtain the spectral problem
. 0 .u0 c/2 z /z C 0 .N 2 k2 .u0 c/2 / D 0 .0 < z < h0 /;

(3.37)
.0/ D .h0 / D 0:
108 3 Water Waves
Then the dimensionless Richardson number
N 2 .z/
Ri.z/ D
.du0 .z/=dz/2
is responsible for the flow stability.

Theorem 3.1 [26] If Ri.z/ > 1=4 for all z 2 .0; h0 /, then the plane shear flow under
a rigid lid is stable.
Proof Introduce the function G D .u0 c/1=2 . Then the equation in (3.37) takes
the form
1 N 2 14 u20z
. 0 .u0 c/Gz /z . 0 u0z /z G k2 0 .u0 c/G C 0 G D 0:
2 u0 c
Multiplying this equation by the complex-conjugate function G, integrating over

the interval .0; h0 /, taking into account the homogeneous boundary conditions for
G, and separating the imaginary part, we obtain the relation
8h 9
<Z 0 Zh0 2 =
1 G
ci 0 .jGz j2 C k2 jGj2 /dz C 0 N 2 u20z dz D 0:
: 4 u0 c ;
0 0
Thus, if Ri.z/ > 1=4 for all z 2 .0; h0 /, then ci D 0. t

u
The Howard semicircle theorem specifies the location of the spectrum of phase
velocities in the unstable case.
Theorem 3.2 [21] If N 2 .z/ > 0 for z 2 .0; h0 /, then the complex phase velocities
of all unstable modes are located in the semicircle of diameter .b a/=2 and center
..a C b/=2; 0/ at the real axis, where
a D min u0 .z/; b D max u0 .z/:

z20;h0 z20;h0
Proof Multiplying the equation in (3.37) by , integrating over .0; h0 /, taking into
account the boundary conditions for , and separating the real and imaginary parts,
we obtain two integral relations
Zh0
.u20 2cr u0 C c2r c2i /Q dz D B;
0
Zh0
u0 Q dz D cr A;
0
3.10 Stationary Internal Waves 109
where
Zh0 Zh0
Q D jz j2 C k2 j j2 ; AD Q dz; BD 0 N 2 j j2 dz:
0 0
From these integral relations we find
Zh0
.u0 a/.u0 b/dz D .c2r .a C b/cr C c2i C ab/A C B:
0
Since A > 0, B > 0, and the integral on the left-hand side is not positive, we obtain
the inequality
2
aCb 2 ab
.cr / C c2i 6 :
2 2
The theorem is proved. t

u
3.10 Stationary Internal Waves
We consider two-dimensional internal waves propagating along the x-axis with

constant velocity u0 in a stratified fluid layer located between the flat bottom y D 0
and the rigid lid y D h0 . The flow is stationary in the reference frame moving with
the wave. In the stationary case, the equations of motion (3.1) take the form
ux C vy D 0;
u x C v y D 0;
.uux C vuy / C px D 0;
.uvx C vvy / C py D g:
We introduce the stream function by the equalities u D y , v D x . Then

the continuity equation div u D 0 is satisfied identically and the equation for the
density u r D 0 can be integrated, which yields the dependence D . /. In
view of this property, the momentum equations admit the Bernoulli integral written
in terms of the stream function as
1 1
jr j2 C p C gy D b. /;
2 . /
110 3 Water Waves
where b. / is the Bernoulli function. Eliminating the pressure p from the equations
of motion and the above integrals, we obtain the second order quasilinear elliptic
equation for the stream function

d . / 1 dH. /
. / C gy C jr j2 D ; (3.38)
d 2 d
where H. / D . /b. /. Equation (3.38) is called the Dubreil-JacotinLong

equation. It is possible to specify . / and b. / for flows of solitary wave type
by conditions at infinity. In particular, if the flow converges to the uniform flow with
the stream function 1 .y/ D u0 y and known density profile 1 .y/ as jxj ! 1,
then
. / D 1 . =u0 /;

dH. / d . / g 1 2
D C u0 :
d d u0 2
In this case, the boundary conditions on the bottom and lid take the form

D 0; D u0 h0 : (3.39)
yD0 yDh0
Example 3.3 Let a fluid in an unperturbed state has the exponential profile of
density
1 .y/ D 0 exp.N 2 y=g/ .0 < y < h0 /;
where the constant N > 0 is the BruntVisl frequency. In this case, the Dubreil-
JacotinLong equation (3.38) takes the form
N2 N2
C . u0 y/ D .jr j2 u20 /:
u20 2gu0
With a uniform flow we associate the exact solution D u0 y to this equation. We

introduce the dimensionless Boussinesq parameter and the (densimetric) Froude
number F by the equalities
N 2 h0 u0
D ; FD p :
g gh0
The parameter characterizes the vertical gradient density of the fluid. It is small in
the case of weak stratification. For small the finite values of the Froude number F
correspond to small velocities u0 , which is a typical situation for internal waves. To
3.10 Stationary Internal Waves 111
construct p
the long-wave approximation, we proceed with dimensionless variables
.x; y/ D . x=h0 ; y=h0 /; D =.u0 h0 /, in which the equation is written as
1
xx C yy C F 2 . y/ D . 2
C 2
1/;
2 x y
.x; 0/ D 0; .x; 1/ D 1
(the bar is omitted). This problem is a nonlinear problem on eigenvalues with the
spectral parameter D F 2 . A solution to this problem is looked for as power series
D yC 0 C 1 C 2 2 C :::;
2
D 0 C 1 C 2 C : : : ;
whose coefficients should satisfy the equations
0yy C 0 0 D 0;
1yy C 0 1 D f1 . 0 /;
:::::::::::::::::::::
with the homogeneous boundary conditions
j .x; 0/ D j .x; 1/ D 0 .j D 0; 1; : : :/
2
and right-hand side f1 D 0xx C 0y C 0y =2 1 0 : Hence for the internal
waves of the principal mode we have 0 .x; y/ D a.x/ sin y, 0 D 2 , where the
amplitude factor a.x/ remains still undefinite. The equation for the amplitude factor
is obtained from the orthogonality conditions
Z1
f1 . 0 / sin y dy D0
0
which implies the nonlinear ordinary differential equation
d2 a 2 2
D a 1 a:
dx2 3
A solitary wave type solution satisfying the condition a.x/ ! 0 as jxj ! 1 is

obtained for 1 < 0 and has the form
91
a.x/ D p :
2 j1 jx
4 cosh 2
112 3 Water Waves
Fig. 3.5 Streamline patterns representing solitary internal waves can be quite complicated. They
essentially depend on the density profile at infinity
By the definition of the Froude number F, the velocity of this wave u0 is

connected with the amplitude parameter 1 in the approximate formula u20 D
gh0 =. 2 C 1 /. The obtained solution is similar to a solitary wave type
solution to the Korteweg-de Vries equation (2.13) and the BoussinesqRayleigh
equation (3.24). In the general case, the form of the equation for the amplitude
function a.x/ essentially depends on the density profile 1 .y/ which can cause more
complicated forms of solitary internal waves (cf. Fig. 3.5).
3.11 Problems
1. A float rises and falls with a wave 15 times per minute. Find the wavelength
L and the wave velocity c if the wave amplitude is small and the fluid depth is
infinitely large.
Answer: L D 24:98 m, c D 6:25 m/sec.
2. Consider a plane travelling wave in a fluid of finite depth with the free surface
y D h0 C a cos.kx !t/ (jaj < h0 and the parameter ! is connected with k by
the dispersion relation ! 2 D gk tanh kh0 ). Find the particle trajectories x D x.t/
by looking for an approximate solution x.t/ D .x.t/; y.t// to the equations
dx gak cosh ky
D cos.kx !t/;
dt ! cosh kh0
dy gak sinh ky (3.40)
D sin.kx !t/;
dt ! cosh kh0
.x; y/jtD0 D .
; /
in the form
x.t/ D x0 .t/ C x1 .t/ C O. 2 /;
where D ak is a small parameter. Along what curve in the .x; y/-plane does
the point x .t/ D x0 .t/ C x1 .t/ move? Is the motion of the point periodic in t?
3.11 Problems 113
Answer:
cosh k
x.t/ D
C a .sin k
sin.k
!t// C O.a2 k2 /;
sinh kh0
sinh k
y.t/ D a .cos k
cos.k
!t// C O.a2 k2 /:
sinh kh0
The point x .t/ rotates with period T D 2=! along an ellipse.

3. Find an exact solution x.t/ D .x.t/; 0/ to the system of equations (3.40)
describing the horizontal motion of a particle at the bottom y D 0 with the
initial condition x.0/ D .0; 0/ at t D 0. Compute the coefficients of the power
expansion of this solution
x.t/ D x0 .t/ C ax1 .t/ C a2 x2 .t/ C O.a3 /
with respect to the amplitude a.

Answer:
1
x.t/ D .!t C f 1 .!t//;
k
where
Zz
ds ak
f .z/ D D ;
cos s 1 sinh kh0
0
x0 .t/ 0;
sin !t
x1 .t/ D ;
sinh kh0
1 2!t sin 2!t
x2 .t/ D k :
4 sinh 2 kh0
4. In the linear approximation, consider the three-dimensional wave motion of an

infinitely deep fluid with a real velocity potential ' D Im.'1 C '2 /, where 'j
are complex wave packets
'j .x; z; t/ D bemzCi.kjx!t/ ; x D .x; y/T
with different wave vectors k1 D .k; l/T , k2 D .k; l/T , but the same frequency
p
! D gm (m D jk1 j D jk2 j) and amplitude b 2 R. Show that the potential
' satisfies the impermeability condition 'x D 0 on the vertical wall x D 0 and
find the shape of the contact line of the free surface z D .x; y; t/ with this wall
at each time moment.
114 3 Water Waves
Answer:
2b!
.0; y; t/ D cos.ly !t/:
g
5. Within the framework of the linear wave theory, find frequencies of eigenoscil-
lations of a fluid with the velocity potential ' D ei!t Y.x; y; z/ in a rectangular
reservoir of height h0 , length a, and width b. What is the minimum frequency?
Answer:
2
!nm D gnm tanh nm h0 ;
2nm D 2 ..n=a/2 C .m=b/2 /;
where n; m D 0; 1; 2; : : :, n C m 0;
2 g h0
min !nm D tanh :
nm a a
6. Consider the linear CauchyPoisson problem describing two-dimensional
waves on the surface of an infinitely deep fluid (small perturbations of the state
at rest)
xx C yy D 0 .1 < y < 0/;

r ! 0 .y ! 1/;
t D y ; t C g D 0 .y D 0/;
.x; y; 0/ D 0 .x; y/; .x; 0/ D 0 .x/:
Show that the function describing the shape of the free surface y D .x; t/
satisfies the integro-differential equation
Z
C1
g x .x0 ; t/ 0
tt .x; t/ C v:p: dx D 0
x x0
1
with the initial condition
.x; 0/ D 0 .x/; t .x; 0/ D 0y .x; 0/:
Hint. Use the Cauchy integral formula for the complex-valued function
f .z; t/ D xt iyt which is analytic with respect to z D x C iy in the
half-plane Im z < 0.
3.11 Problems 115
7. Under the conditions of the previous problem, construct the solution .x; t/ with
the initial data
2a
.x; 0/ D 0; t .x; 0/ D .a > 0 const /:
x2 C a 2
For this purpose represent as the Fourier integral.

Answer: D fC .x; t/ C f .x; t/, where
Z
C1
p dk
f .x; t/ D eak sin.kx gkt/ p :
gk
0
8. The frequency ! in a group of waves coming from a distant storm area varied
as a linear function and rose by for the observation time period . Using
the dispersion relation for deep water waves, find the distance r from the storm
area.
g
Answer: r D .
2
9. Show that the phase and group velocities of linear surface waves propagating
through a fluid at rest of depth h0 satisfy the equality

1 2kh0
cg .k/ D cp .k/ 1 C :
2 sinh 2kh0
Using this equality, show that the ratio cg .k/=cp .k/ is a strictly monotonically
decreasing function of the wavenumber k on the half-axis .0; C1/. What is
this ratio in the long (k ! 0) and short (k ! C1) wave limit?
Answer:
cg .k/ cg .k/ 1
lim D 1; lim D :
k!0 cp .k/ k!C1 cp .k/ 2
10. Derive the dispersion relation for the plane problem of the linear wave
propagation in water of finite depth with the surface tension forces taken
into account. The boundary condition for the pressure p on the free surface
y D h0 C .x; t/ in the original nonlinear equations has the form p D K,
where > 0 is the surface tension coefficient and K is the curvature of the free
surface, expressed by the equality K D xx .1 C x2 /3=2 .
Answer:

2 2
! D gk 1 C k tanh kh0 :
g
11. Consider linear surface waves in an infinitely deep fluid with the surface tension
forces taken into account.
116 3 Water Waves
a. Find the value of the wavenumber k D kp at which the phase velocity cp .k/
has a minimum.
b. Find the value of the wavenumber k D kg at which the group velocity cg .k/
has a minimum.
c. Find the value of the wavenumber k D kr at which the equality cp D cg
holds (the resonance of phase and group velocities).
d. Find relations between the group velocity cg and the phase velocity cp in the
short wave limit (jkj ! C1).
Answer:
p
a. kp D sg =.
2
b. kg D kp p 1.
3
c. kr D kp .
d. cg D .3=2/cp .
12. Consider the dispersion relation for three-dimensional linear stationary waves
on the surface of an infinitely deep fluid
p
u0 k C gm D 0;
p
where u0 is the velocity of the fluid flow and m D k2 C l2 is the modulus
of the wave vector k D .k; l/T (this relation is obtained from (3.6) in the case
of stationary waves (! D 0) as h0 ! 1). The wave phase is equal to D
kx C ly.
a. Passing to the polar coordinates in the plane of the wave vector k D
m.cos ; sin /T and using the dispersion relation, find the dependence m D
m./.
b. Find the ratio y=x as a function of the polar angle for which the phase is
stationary: d=d D 0.
c. Find the parameter representation x D x./; y D y./ of the curve L.0 /
along which the wave phase is simultaneously constant .x; y/ D 0 D
const and stationary. Construct this curve.
d. Find the locus of cusps of the curve L.0 / for different values of the
phase 0 .
Remark The family of curves L.0 / yields a three-dimensional pattern of waves
behind the point source moving with the constant velocity u0 (ship waves).
Answer:
g
a. m./ D .
u20
cos2
y sin cos ;
b. D .
x 1 C sin2
3.11 Problems 117
u20 0 u 2 0
c. x D cos .1 C sin2 /; y D 0 cos2 sin .
g g
d. y D p x
.
2 2
13. Construct self-similar solutions to the shallow water equations
ht C .uh/x D 0;
ut C uux C ghx D 0
that are continuous for t > 0 and depend on x=t.

Answer:
2x 1x 2
u.x; t/ D C C; h.x; t/ D C ; C D const :
3t g 3t
14. For the shallow water equations find all conservation laws Pt C Qx D 0 with
polynomial densities P.u; h/ of degree at most 3.
Answer:

1 2 1
P.u; h/ D C1 u C C2 h C C3 uh C C4 u h C gh2 ;
2 2

1 2
Q.u; h/ D C1 u C gh C C2 uh
2

2 1 2 1 3 2
C C3 u h C gh C C4 u h C guh :
2 2
15. Derive the shock adiabat equation in the .u; h/-plane with the origin .u0 ; h0 / for
a strong discontinuity described by the laws of conservation of mass and total
momentum (3.10) of the shallow water equations.
Answer:
g.h C h0 /
.u u0 /2 D .h h0 /2 :
2hh0
16. Under the conditions of the previous problem, find the expressions for the fluid
relative velocities v D uD and v0 D u0 D on the different sides of the strong
discontinuity (D is the discontinuity velocity) in terms of the depths h and h0
(h > h0 ). Find out whether these velocities are subcritical or supercritical.
Answer:
r s
gh0 .h C h0 / gh.h C h0 /
vD ; v0 D ;
2h 2h0
p p
jvj < gh; jv0 j > gh0 :
118 3 Water Waves
17. Consider the hydraulic jump (a stationary strong discontinuity for the shallow
water equations with the velocity D D 0) at which the laws of conservation of
mass and total momentum are valid. Show that the energy flux
1 3
QD u h C guh2
2
has the jump
gm
Q D h3
4hh0
with the mass flow m D uh D u0 h0 and the fluid depths h and h0 satisfying the
relation
1 p
hD h0 . 1 C 8F 2 1/;
2
p
where F D u0 = gh0 is the Froude number.
18. Owing to the upstream propagation of a tidal bore along a river, the water level
rises by 10% and the flow velocity u0 is halved. Compute the bore velocity.
Answer: D D 92 u0 .
19. Two fluids layers of depth h0 for x < 0 and h1 for x > 0 (h0 > h1 ) are separated
by a rigid wall at x D 0. At the time t D 0, the wall is removed and, as a
result, the fluid begins to move. In the shallow water approximation, find the
flow velocity u.x; t/ and the free surface shape h.x; t/ for t > 0 in the class of
self-similar solutions.
Answer:
x p
u D 0; h D h0 1 < < gh0 ;
t

2 x p 1 x p 2
uD C gh0 ; h D 2 gh0
3 t 9g t
p x p
gh0 < < u2 gh2 ;
t
p x
u D u2 ; h D h2 u2 gh2 < < D ;
t
x
u D 0; h D h1 D < < C1 ;
t
where D D u2 h2 =.h2 h1 / and the constants u2 > 0 and h2 > 0 are found from
the relations
p p
u2 C 2 gh2 D 2 gh0 ;
g.h1 C h2 /
u22 D .h2 h1 /2 :
2h1 h2
3.11 Problems 119
20. Show that the linearization of the Benney equations (3.12) at the constant
solution u D 0; v D 0; h D h0 D const leads to the equations
0h 1
Z0
@h @ @
C u.x; y0 ; t/dy0 A D 0;
@t @x
0
@u @h
Cg D 0;
@t @x
Zy
v D ux .x; y0 ; t/dy0 :
0
21. Show that the linearization of the Benney equations (cf. the previous problem)
admits the Riemann invariants
Zh0 p

R D u.x; y0 ; t/dy0 gh0 h;
0
@ p @
gh0 R D 0;
@t @x
@
.uy / D 0;
@t
0 1
Zh0
@ @ 1
u u.x; y0 ; t/dy0 A D 0:
@t h0
0
Give an interpretation of these invariants.

22. Show that the Benney equations (3.12) admit the Riemann invariants
Rit C ci Rix D 0; i D 1; 2;
where
Zh
dy0
R .x; t/ D ci g
i
;
u.x; y0 ; t/ ci
0
and ci are the roots of the equation
Zh
dy0
1g D 0;
.u.x; y0 ; t/ ci /2
0
120 3 Water Waves
satisfying the inequalities
c1 .x; t/ < min u.x; y; t/ 6 max u.x; y; t/ < c2 .x; t/:

06y6h 06y6h
23. Verify that the velocity curl ! D uy in the long-wave approximation and the
invariant
Zh
dy0
RDug (3.41)
u.x; y0 ; t/ u.x; y; t/
0
are conserved along the trajectories dx=dt D u.x; y; t/, dy=dt D v.x; y; t/:
!t C u!x C v!y D 0;
Rt C uRx C vRy D 0:
The integral in (3.41) is understood in the Cauchy principal value sense.

24. Show that the Benney equations (3.12) have a class of exact solutions with
constant vorticity
u.x; y; t/ D U.x; t/ C y; D const ;
if the functions U.x; t/ and h.x; t/ satisfy the system of equations
@h @ h2
C Uh C D 0;
@t @x 2
(3.42)
@U @U @h
CU Cg D 0:
@t @x @x
Show that the system (3.42) is hyperbolic. Find Riemann invariants of this
system.
25. Under the conditions of the previous problem, find an explicit form of the
Riemann invariant (3.41) for the Benney equations (3.12).
Answer:

g hy
RDu ln :
y
26. For solutions to the system of equations (3.42) we introduce the depth-averaged
fluid velocity W D U C h2 . Show that the equations for the functions W and
h are written in the conservative form
@h @
C .hW/ D 0;
@t @x
3.11 Problems 121
@hW @
C .hW 2 C P/ D 0;
@t @x
gh2 2 h3
PD C :
2 12
This system is similar to the equations governing the motion of a barotropic,
but not polytropic gas. Show that the specific internal energy of such a gas has
the form
gh 2 h2
eD C :
2 24
27. Verify that the nonlinear kinematic condition for the velocity potential ' on the
free boundary z D h.x; y; t/ in the three-dimensional CauchyPoisson problem
is equivalent to the conservative equation
Zh
ht C div.x;y/ r.x;y/ '.x; y; z; t/dz D 0:
0
28. Consider the two-dimensional CauchyPoisson problem. Show that, the tangent
velocity u.x; t/ D .'x C hx 'y /yDh and the normal velocity v.x; t/ D .'y
hx 'x /yDh of the fluid particles on the free boundary y D h.x; t/ satisfy the
system of equations
ht D v;

1 @ u2 2hx uv v 2
ut C C ghx D 0:
2 @x 1 C h2x
29. Show that the two-dimensional CauchyPoisson problem admits the integro-
differential laws of conservation
0 h 1 0 h 1
Z Z
@ @ @ 1 2 1 2 1 2A
'x dyA C @ 'x 'x 't dy gh D 0;
@t @x 2 2 2
0 0
0 1 0 h 1
Zh Z
@ @1 1 @
.'x2 C 'y2 /dy C gh2 A @ 'x 't dyA D 0;
@t 2 2 @x
0 0
where the first equality is the depth-averaged differential law of conservation

of horizontal momentum, whereas the second one is the law of conservation of
energy.
122 3 Water Waves
30. Show that the system of equations (3.23) describing plane nonlinear stationary
surface waves has the first integral
Zh.x/
2 2
. x .x; y/ y .x; y//dy C gh2 .x/ 2bh.x/ D c .c D const /:
0
31. Show that the system of equations (3.30) describing plane nonlinear stationary
waves in a two-layer fluid under a rigid lid has the first integral
Zh.x/ ZH
2 2 2 2
1 . 1x .x; y/ 1y .x; y//dy C 2 . 2x .x; y/ 2y .x; y//dy
0 h.x/
C g. 1 2 /h2 .x/ 2. 1 b1 2 b2 /h.x/ D c .c D const /:
32. Using the method of separation of variables, find eigenfunctions of the two-
dimensional problem describing stationary surface waves p linearized on the
supercritical flows with constant depth h0 and velocity u0 > gh0 :
xx C yy D 0 .0 < y < h0 /;
y D 0 .y D 0/;
u20 xx C gy D 0 .y D h0 /:
Answer: n .x; y/ D en x=h0 cos.n y=h0 / .n D 0; 1; 2; : : :/, where n D

2
n .F/ .n < n < =2 C pn) is the root of the equation tan D F with the
Froude number F D u0 = gh0 .
33. The LeviCivita variables .'; /; .'; / are introduced for the velocity field
u D .'x ; 'y /T of the plane stationary potential fluid motion by the formula
u D e .cos ; sin /T ;
where the potential ' and the stream function are independent variables.
Show that the nonlinear dynamic condition on the free boundary
'x2 C 'y2 C 2F 2 y D const .y D h.x//
is written in the LeviCivita variables as
D F 2 e3 sin . D 0 /;
where F is the Froude number and 0 is the value of the stream function on the
free boundary: .x; h.x// D 0 .
3.11 Problems 123
34. Consider the plane potential flow of a heavy fluid inside the angle z D rei
(r > 0, 5=6 < < =6) with a complex potential w D ' C i of the
form w.z/ D Az3=2 . Find a constant A for which the kinematic and dynamic
conditions on the free boundary are satisfied on the angle sides:
2
dw
Im w D 0; C 2F 2 Im z D 0: (3.43)
dz
Find the modulus of the fluid velocity on the free boundary and derive the
equation for the stream line L passing through the point z D i in the polar
coordinates .r; /.
Remark This flow is called the Stokes flow.
Answer:

2 i=4 dw p
AD e ; D r=F . D =6; D 5=6/ ;
3F dz
2
L W r3 D :
1 C sin 3
35. Consider the plane potential flow of a heavy fluid with free boundary having a
corner point with the unknown opening angle 0 (without loss of generality we
can assume that the corner point is located at the origin). The complex potential
of the flow in a neighborhood of the corner point is looked for in the form
w.z/ D Czn C o.zn / .C D const n > 1/:
Using the conditions (3.43) on the free boundary, find n and the opening angle
at the corner point.
Answer: n D 3=2, 0 D 2=3.
36. Consider a solitary wave with profile y D h.x/ possessing a corner point at
the wave crest with the ordinate y D hc . It is known that the fluid velocity u
vanishes at the corner point. Using the Bernoulli integral juj2 C 2gh D u20 C
2gh0 (here, u0 and h0 are the velocity and
p depth of the fluid at infinity) and the
value of the Froude number F D u0 = gh0 D 1:290, found by a numerical
calculation for the flow under consideration, find the dimensionless amplitude
a D .hc h0 /=h0 of the cusped solitary wave.
Answer: a D 0:832.
37. Show that, during the time of motion 1 < t < C1, the solitary surface wave
transfers in the direction of its propagation through each vertical cross-section
of the layer the fluid mass
Z
C1
mD .
/d
;
1
124 3 Water Waves
where D const is the fluid density and y D h0 C .x u0 t/ is the free surface
shape in the travelling wave (h0 is the depth of the fluid at rest and u0 is the
wave velocity).
38. The momentum I of a travelling solitary wave type is defined by
Z Zh.
/
C1
ID u.
; y/dy d
;
1 0
where u D 'x .x u0 t; y/ is the horizontal fluid velocity and y D h.x u0 t/ D

h0 C .x u0 t/ is the free boundary shape. Show that I D mu0 , where m is the
fluid mass transferred by the wave.
39. Show that the SerreSuGardner system of equations
ht C .uh/x D 0;
1 2 2
ut C uux C ghx C .h dt h/x D 0;
3h
dt D @t C u@x
is invariant under the Galilean transformatione t D t,e u D uu0, e

x D xu0 t,e hD
h. Do the Boussinesq equations (3.18) and (3.19) possess a similar property?
40. Within the framework of the second shallow water approximation (the Serre
SuGardner model) find an expression for the hydrodynamic pressure p.x; y; t/
inside the fluid layer in terms of the functions h and u.
Hint. Integrate the vertical momentum equation and ignore terms of order
higher than O."2 /.
Answer:
1 h 2 y2 2
p D g.h y/ C dt h:
2 h
41. Establish the law of conservation of energy

1 1 2 1
@t hjuj2 C q C div ..e
p C q/u/ D 0; qD gh C h.dt h/2 ;
2 2 6
for the GreenNaghdi system (3.20). Here, the function e p is the same as in the
law of conservation of momentum (3.21).
42. Let h.x; t/ and u.x; t/ be solutions to the GreenNaghdi system (3.20). Tra-
jectories of fluid particles are defined as the integral curves of the system of
ordinary differential equations
dx
D u.x; t/:
dt
3.11 Problems 125
Show that for an arbitrary closed contour C.t/ in the plane x D .x; y/ consisting
of the same particles the following relation holds:
Z Z T !
d @u
v dx D dt v C v dx;
dt @x
C.t/ C.t/
where
1
vDuC r.h2 dt h/; dt D @t C u r:
3h
43. Under the conditions of the previous problem, show that the scalar =h, where
(a counterpart of the curl for a medium with dispersion) is defined by ez D
curl v, is conserved along the particle trajectories. Here, ez is the unit vector of
the z-axis directed along the direction of the force of gravity.
44. Consider the linearized two-dimensional problem describing waves on the
interface y D .x; t/ of two layers of a heavy fluid of infinite depth with the
surface tension forces taken into account
1xx C 1yy D 0 .1 < y < 0/;

2xx C 2yy D 0 .0 < y < C1/
with the boundary conditions at y D 0 (the unperturbed level of the interface)
t C u1 x D 1y ; t C u2 x D 2y ;
1 .1t C u1 1x C g/ 2 .2t C u2 2x C g/ D xx
and the decay condition
1 .x; y; t/ ! 0 .y ! 1/;
2 .x; y; t/ ! 0 .y ! C1/;
where the constants 1 and 2 < 1 are the fluid densities in the layers, u1
and u2 are the velocities the piecewise constant flow where the linearization is
performed, and > 0 is the surface tension coefficient. Find the dependence
of the phase velocity cp .k/ on the wavenumber k. For what velocities u1 and u2
are all wave modes stable?
126 3 Water Waves
q
Answer: cp .k/ D cm c20 .k/ c2 , where
1 u1 C 2 u2 g. 1 2 / k
cm D ; c20 .k/ D C ;
1 C 2 . 1 C 2 /k 1 C 2
1 2 .u1 u2 /2
c2 D :
. 1 C 2 /2
The stability condition:
p
2 1 1
.u1 u2 / 6 2 g. 1 2 / C :
1 2
45. Consider the problem describing waves in a two-layer fluid of finite depth with
the free upper boundary:
'1xx C '1yy D 0; 0 < y < h1 C 1 .x; t/;

'2xx C '2yy D 0; h1 C 1 .x; t/ < y < h1 C h2 C 2 .x; t/;
'1y D 0; y D 0;
9
1t C 'jx 1x 'jy D 0 .j D 1; 2/ >>
>
>
1 2 1 2 >
>
=
1 '1t C '1x C '1y C g1
2 2 y D h1 C 1 .x; t/;
>
>
>
>
1 2 1 2 >
D 2 '2t C '2x C '2y C g1 ;> ;
2 2
9
2t C '2x 2x '2y D 0; =
1 2 1 2 y D h1 C h2 C 2 .x; t/;
'2t C '2x C '2y C g2 D 0;;
2 2
where 2 < 1 are the fluid densities of the layers, h1 and h2 are the layer depths
at the equilibrium state. Linearize the equations at the state at rest '1 D 0, '2 D
0, 1 D 2 D 0 and derive the dispersion relation. What is the asymptotics of
wave modes in the limit of close fluid densities in the layers . 1 2 /= 1 ! 0?
Answer:
.! 2 gk tanh kh1 /.! 2 gk tanh kh2 /

C ! 2 tanh kh1 .! 2 tanh kh2 gk/ D 0;
! 2 .k/ D gk tanh k.h1 C h2 / C O./ (modes of surface waves);
gk
! 2 .k/ D C O.2 / (modes of internal waves);
coth kh1 C coth kh2
3.11 Problems 127
where
1 2 2
D ; D D 1 :
1 1
46. Consider linear stationary waves (i.e., wave packets with a given frequency
! D 0 and the sought wavenumber k) on the piecewise constant flow of a two-
layer fluid under a rigid lid with the density j ( 2 < 1 ), velocity uj , and depths
h1 and h2 D H h1 in the lower (j D 1) and upper (j D 2) layers. Introduce the
dimensionless parameters: the dimensionless wavenumber D kh1 , the ratio
of unperturbed depths of the layers r D h2 =h1 , and the densimetric Froude
numbers
1 u21 2 u22
F12 D ; F22 D :
g. 1 2 /h1 g. 1 2 /h2
Verify that, in the case of stationary waves, the dispersion relation (3.25) is
written in the above dimensionless variables as
D.I F; r/ D F12 coth C F22 r coth r 1 D 0:

def
Show that the dispersive function D.I F; r/ possesses the following proper-
ties.
a. All roots of the analytic function D.zI F; r/ of the complex variable z D
C i are located on the coordinate axes Im z D 0 and Re z D 0.
b. There are only two real roots with the multiplicity taken into account.
These roots exist if and only if F12 CF22 6 1 (the case of subcritical piecewise
constant flows).
c. The function D.zI F; r/ has countably many roots on the imaginary axis.
Among these roots, there are only two roots z D i lying in the strip
j Im zj < minf1; r1 g for F12 C F22 > 1 (the case of supercritical flows).
Moreover, is the minimal positive root of the real equation
F12 cotan C F22 r cotan r D 1
and ! 0 as F12 C F22 ! 1.

Remark The imaginary roots i of the dispersive function characterize the
exponential asymptotics exp.jxj/ as jxj ! 1 for solitary waves in a
two-layer fluid.
47. Using the integral laws of conservation of mass, momentum, energy and the
Bernoulli integral for a piecewise potential flow, find out for what layer depths
H1 , H2 , h1 , h2 (H1 C H2 D h1 C h2 D H) and velocities U1 , U2 , u1 , u2 taken
as x ! 1 there exists a stationary two-layer flow under a rigid lid, as in
Fig. 3.3, in the form of a smooth bore.
128 3 Water Waves
Answer:
h2 H1 H2 ra
D r; D 1 C a; D ;
h1 h1 h2 r
U1 1 U2 r
D ; D ;
u1 1Ca u2 ra
1 u21 .1 C a/2 2 u22 .r a/2
F12 D F22 D
def def
D ; D ;
g. 1 2 /h1 1Cr g. 1 2 /h2 r.1 C r/
where 0 < r < C1, 1 < a < r.

48. The equation (3.33) describing solitary waves and smooth bores in a two-layer
fluid under a rigid lid has the form
d 2 P. I F; r/
D 3 2
dx Q. I F; r/
with polynomials
Q. I F; r/ D F12 .r / C r3 F22 .1 C /;
P. I F; r/ D F12 .r / C rF22 .1 C / .1 C /.r /;
where Fj .j D 1; 2/ are the densimetric Froude numbers and r D h2 =h1 is the

ratio of unperturbed layer depths. Prove the following assertions.
a. The discriminant d.F; r/ D .F12 rF22 C r 1/2 C 4r.1 F12 F22 / of the
quadratic polynomial P. I F; r/ in pis positivepif the point F D .F1 ; F2 / is
located inside the rhombuspjF1 j C p rjF2 j < 1 C r and vanishes on the
rhombus boundary jF1 j C rjF2 j D 1 C r.
b. The roots D a .F; r/ of the polynomial P. I F; r/, where
1 p
a .F; r/ D .rF22 F12 C 1 r d.F; r//;
2
satisfy the inequalities 0 < a .F; r/ < aC .F; r/ if the point F D .F1 ; F2 /
2
is located inside the lateral curvilinear
2
pbounded by the circle F1 C
p triangles
F2 D 1 and the rhombus jF1 j C rjF2 j D 1 C r (cf. Fig. 3.4) and the
inequalities a .F; r/ < aC .F; r/ < 0 if the point F D .F1 ; F2 / lies in an
analogous upper or lower triangle.
49. Construct a solitary wave type solution to the following approximate version
of (3.33) describing nonlinear stationary waves in a two-layer fluid under a
rigid lid:
2
d 3 2 fF12 .r / C rF22 .1 C / .1 C /.r /g
D
dx rF12 C r3 F22
3.11 Problems 129
which is obtained for waves of small amplitude if the linear denominator

depending on of the fraction on the right-hand side of (3.33) is approximated
by a constant, namely, by its value at D 0.
Answer:
r.F12 C F22 1/
.x/ D p ;
a .F; r/ d.F; r/ cosh 2 kx
where
3.F12 C F22 1/
k2 D :
4.F12 C r2 F22 /
The upper sign (plus or minus) is taken for 0 < a < aC (waves of elevation
type) and the lower sign is taken for a < aC < 0 (waves of depression type).
The quantities a .F; r/ and d.F; r/ are defined in the previous problem.
50. Consider the two-layer shallow water equations with a lid in the Boussinesq
approximation:
At C ..A2 1/B/x D 0;
(3.44)
Bt C ..B2 1/A/x D 0:
Show that these equations are hyperbolic in the domain jAj < 1, jBj < 1. Find
characteristics and Riemann invariants.
Answer:
dx p
D 2AB .1 A2 /.1 B2 /;
dt
p
r D AB .1 A2 /.1 B2 /:
51. Under the conditions of the previous problem, show that the pair of functions
S D .1 A2 /.1 B2 / and F D 2ABS forms the conservation law
S.A; B/t C F.A; B/x D 0:
Find a subdomain of the hyperbolicity domain where S is convex.

Answer:
jBj < 1; .1 A2 /.1 B2 / > 4A2 B2 :
52. Show that all conservation laws S.A; B/t C F.A; B/x D 0 for the system (3.44)
are exhausted by functions S.A; B/ such that .1 B2 /SBB .1 A2 /SAA D 0.
130 3 Water Waves
53. Show that the change of variables u D 2AB, h D .1 A2 /.1 B2 / transforms

the system (3.44) to the shallow water equations
ht C .uh/x D 0;
ut C uux C hx D 0:
What is the Jacobian of the transformation?

Answer:
@.u; h/=@.A; B/ D 4.A2 B2 /:
54. Show that the system of equation (3.27) describing two-layer shallow water
under a lid is hyperbolic if and only if the inequality (3.28) holds.
55. Consider the system of equations describing two-layer shallow water with a
free surface
ut C A.u/ux D 0
for vector-valued functions u D .h1 ; h2 ; u1 ; u2 /T , where hj > 0 are the layer

depths, uj are the fluid velocities in the layers (j D 1; 2), and the matrix A has
the form
0 1
u1 0 h1 0
B0 u2 0 h2 C
A.u/ D B
@g
C;
g u1 0A
g g 0 u2
where D 2 = 1 . 2 < 1 / is the ratio of densities. Find the characteristic

determinant .c/ D det .A cI/ of this system. Transform the characteristic
equation .c/ Dp0 to an equivalent system
p of equations for the Froude numbers
F1 D .u1 c/= gh1 , F2 D .u2 c/= gh2 . How many real roots does the
equation .c/ D 0 have if
(a) u2 D u1 , p p
(b) u2 D u1 C gh1 C gh2 .
Answer:
.c/ D ..u1 c/2 gh1 /..u2 c/2 gh2 / g2 h1 h2 ;

p p
.F12 1/.F22 1/ D ; F2 D h1 =h2 F1 C .u2 u1 /= gh2 ;
(a) four roots (the system is hyperbolic),

(b) two roots (the system is not hyperbolic).
3.11 Problems 131
56. Construct an exact solution of the form h1 D h1 .t/, h2 D const , u1 D a.t/x,

u2 D 0 to the two-layer shallow water equations with a free surface (cf. the
previous problem) satisfying the initial conditions h1 .0/ D h0 , a.0/ D a0 > 0.
Show that there exist points .x; t/ in the half-plane t > 0 for which there are
complex roots of the characteristic polynomial .c/ on this solution (points of
nonhyperbolicity).
Answer:
a0 x h0
u1 .x; t/ D ; h1 .t/ D :
1 C a0 t 1 C a0 t
57. Show that the two-layer shallow water equations with a free surface can be
represented in the form
ut C .Rru E.u//x D 0
for u D .h1 ; h2 ; u1 ; u2 /T , where R is the symmetric matrix of the form

0 1
0 0 11 0
B0 0 0 1 C
B 2 C
RDB1 C;
@ 1 0 0 0 A
0 12 0 0
and E.u/ is the energy function

1 1 1 1
E.u/ D 1 h1 u21 C 2 h2 u22 C 1 gh21 C 2 gh1 h2 C 2 gh22 :
2 2 2 2
58. The Eulerian coordinates x and the Lagrangian coordinates of fluid particles
are connected by the relations
dx
D u.x; t/; xjtD0 D :
dt
Show that the system (3.1) describing the motion of an inhomogeneous fluid is
written in the Lagrangian coordinates .; t/ as
det M D 1; 0 M T .xtt g/ C r
p D 0;
where 0 D 0 ./ is the initial density field and M D @x=@ is the Jacobi
matrix of the mapping ! x.; t/.
59. Show that the plane motion of a homogeneous fluid is described in the
Lagrangian coordinates by the system of equations
x
y x y
D 1;
(3.45)
x x
t x
x t C y y
t y
y t D !0 .
; /;
where !0 .
; / is the initial vorticity.
132 3 Water Waves
60. Show that if the initial vorticity is identically constant: !0 .

; / const , then
the transformation
!0 t !0 t
x0 D x cos C y sin ;
2 2
!0 t !0 t
y0 D x sin C y cos
2 2
transforms the system (3.45) to a system of the same form for x0 .

; ; t/ and
y0 .
; ; t/ with vorticity !0 D 0.
61. Consider the solid body rotation of an ideal incompressible fluid with constant
angular velocity !0 about the Oz-axis with the particle trajectories
x D
cos !0 t sin !0 t;
y D
sin !0 t C cos !0 t;
z D :
Find the expression in the Eulerian coordinates for the stream function of the
flow in the Oxy-plane (u D y , v D x ) and the vorticity vector ! D curl u
corresponding to the velocity field u D .u; v; 0/ of this motion.
Answer:
1
.x; y/ D !0 .x2 C y2 /; ! D .0; 0; 2!0 /:
2
62. Consider the two-dimensional motion of a homogeneous fluid in the .x; y/-
plane with the particle trajectories
1
x D a C ekb sin k.a ct/;
k
1
y D b ekb cos k.a ct/;
k
where k D const and c D const, whereas the parameters a and b (b 6 0)
are constant on each fixed trajectory, but vary from one trajectory to another
(the Gerstner trochoidal waves). Show that the pressure is constant along each
trajectory if and only if c2 D g=k; moreover, in this case, the pressure p and
vorticity ! have the form
1 2kce2kb
p D p0 0 gb C 0 c2 .e2kb 1/; !D ;
2 1 e2kb
where 0 D const is the fluid density and p0 D const .
3.11 Problems 133
63. Show that the law of conservation of energy

1 1
@t juj2 C gz C div u juj2 C p C gz D 0
2 2
is a consequence of the system (3.1) governing the motion of a stratified fluid.
64. At the time t D 0, the layer 0 < z < h0 of a stratified fluid has the density
distribution D 0 .z/ with a smooth monotone function 0 ( 00 .z/ < 0) and
the constant pressure p D p0 on the free boundary z D h0 . Construct a solution
to the system (3.1) describing the spreading of the layer over the bottom z D 0
with the linear velocity field u.x; t/ D a.t/x, v.y; t/ D 0, w.z; t/ D a.t/z
(a.0/ D a0 > 0) under the action of the force of gravity. Find trajectories of
fluid particles.
Answer:
a0
a.t/ D I
1 C a0 t
trajectories of particles:
x D x0 .1 C a0 t/; y D 0; z D z0 =.1 C a0 t/I D 0 ..1 C a0 t/z/;

Zh0 Zh0
g 2a20
p D p0 0 .s/ds s 0 .s/ds:
1 C a0 t .1 C a0 t/4
.1Ca0 t/z .1Ca0 t/z
65. Consider the atmosphere with pressure p and density connected by the
equation of state of an ideal gas p D RT, where R D const is the gas constant
and T is the absolute temperature. Using the hydrostatic law dp=dz D g .z/,
find the dependence of the density of a gas at rest on the vertical variable z for
the following temperature distributions:
(a) T D T0 ,
(b) T D T0 .1 z=h0 /, where T0 D const.
Answer:
(a) .z/ D 0 ez=h0 ,
(b) .z/ D 0 .1 z=h0 /1 ( D g 0 h0 =p0 , p0 D 0 RT0 ).
66. Find the density .z/ and pressure p.z/ of a stratified fluid at rest in the layer
h0 6 z < 0 if we know the values of the density 0 and pressure p0 on the
p z D 0 and the dependence of the BruntVisl frequency on z:
layer surface
N.z/ D g=.h0 z/ . > 0 const /.
Answer:

z z
.z/ D 0 1 ; p.z/ D p0 g 0 z 1 :
h0 2h0
134 3 Water Waves
67. Consider the SturmLiouville problem (3.36) governing linear internal waves
in a layer of a fluid under a rigid lid with exponential stratification
2
0 .z/ D eN0 z=g . D const > 0; N0 D const /:
Find the spectrum of frequencies and the corresponding eigenfunctions.

Answer:
N02 m2
!n2 .m/ D ;
2 n2 N04
m2 C h20
C 4g2
2 n
Wn .z/ D eN0 z=.2g/ sin z
h0
for n D 1; 2; 3; : : :
68. Consider the spectral SturmLiouville problem
N2
. 0 Wz /z D 0 k2 1 02 W .0 < z < h0 /;
!
k2
W.0/ D 0; W 0 .h0 / D g W.h0 /
!2
describing linear internal waves in a layer of a stratified fluid with a free
surface in the Boussinesq approximation. Here, 0 D const , N0 D const , and
k is a wavenumber. Find the spectrum of frequencies and the corresponding
eigenfunctions.
Answer:
N02 k2
!02 .k/ D ; W0 .z/ D sinh 0 .k/z;
k2 20 .k/
where 0 .k/ > 0 is a root of the equation
N02 coth h0 D g.k2 2 /I
N02 k2
!n2 .k/ D ; Wn .z/ D sin n .k/z .n D 1; 2; 3; : : :/;
k2 C 2n .k/
where 0 < 1 .k/ < 2 .k/ < : : : are roots of the equation
N02 cotan h0 D g.k2 C 2 /:

3.11 Problems 135
69. Linear long waves in the shear flow of a stratified fluid under a rigid lid are
described in the Boussinesq approximation by the spectral problem
..u0 .z/ c/2 'z /z C N02 ' D 0;

'.0/ D '.h0 / D 0;
where c is the phase velocity and N0 D const is the BruntVisl frequency.

Find the spectrum of phase velocities and the eigenfunctions for the flow with
linear shear of velocity u0 .z/ D az (a D const, a 0) satisfying the stability
condition a2 < 4N02 .
Answer:
ah0
cn D ;
1 en=
n o
sin ln 1 az
cn
'n .z/ D q ;
1 cazn
s
N02 1
D ; n D 1; 2; : : : :
a2 4
70. Show that eigenfunctions Wi and Wj of the problem (3.36) corresponding to
any two real wave modes !i .m/ and !j .m/ with i j satisfy the orthogonality
conditions
Zh0
0z Wi Wj dz D 0;
0
Zh0
0 .Wiz Wjz C m2 Wi Wj /dz D 0:
0
71. Prove that for the spectrum of the SturmLiouville problem (3.36) the following
estimate holdsW ! 2 < gm.
72. Show that the group and phase velocities of each mode of linear internal waves
described by the SturmLiouville problem (3.36) are connected by the relation
Zh0
Wz2 .z/dz
0
cg .k/ D cp .k/ ;
Zh0 Zh0
Wz2 .z/dz C k2 W 2 .z/dz
0 0
where k is the wavenumber and W is the corresponding eigenfunction.

136 3 Water Waves
73. We consider the two-dimensional stationary flow of a continuously stratified

fluid in the gravity field. The stream function is defined by the relations
p
y D = 0 u;
p
x D = 0 v;
where u and v are components of the velocity vector, is the fluid density, and
0 D const is some characteristic value of the density. Show that the function
satisfies the DubreilJacotinLong equation in the form due to Yih
gy d . / 1 dH. /
xx C yy C D ;
0 d 0 d
where H is the same as in Eq. (3.38).

74. Show that the DubreilJacotinLong equation

d . / 1
. / C g y C jr j2 u20 D 0
d u0 2
is the EulerLagrange equation with the Lagrangian
Z
1 2 g
L D . /jr. u0 y/j C . ./ . // d:
2 u0
u0 y
References
1. Brekhovskikh, L. M., Goncharov, V. V.: Mechanics of Continua and Wave Dynamics. Springer,
Berlin etc. (1985)
2. Lighthill, J.: Waves in Fluids. Cambridge University Press, Cambridge (2001)
3. Whitham, G. B.: Linear and Nonlinear Waves. Wiley, New York, NY (1999)
Supplementary References for Chapter 1
4. Godunov, S. K., Romenskii, E. I.: Elements of Continuum Mechanics and Conservation Laws.
Kluwer, New York (2003)
5. Kulikovskii, A. G., Sveshnikova, E. I.: Nonlinear Waves in Elastic Media. CRC Press, Boca
Raton, FL (1995)
6. Landa, P. S.: Nonlinear Oscillations and Waves in Dynamical Systems. Kluwer, Dordrecht
(1996)
7. Rozhdestvenskij, B. L., Yanenko, N. N.: Systems of Quasilinear Equations and Their Applica-
tions to Gas Dynamics. Am. Math. Soc., Providence, RI (1983)
8. Serre, D.: Systems of Conservation Laws. I: Hyperbolicity, Entropies, Shock Waves. Cam-
bridge University Press, Cambridge (1999)
9. Smoller, J.: Shock Waves and Reaction-Diffusion Equations. Springer, New York (1994)
10. Achenbach, J. D.: Wave Propagation in Elastic Solids. North-Holland, Amsterdam etc. (1973)
11. Karpman, V. I.: Nonlinear Waves in Dispersive Media. Elsevier (1974)
12. Nesterenko, V. F.: Dynamics of Heterogeneous Materials. Springer, New York (2001)
13. Newell, A. C.: Solitons in Mathematics and Physics. SIAM (1987)
14. Royer, D., Dieulesaint, E.: Elastic Waves in Solids I. Free and Guided Propagation. Springer,
Berlin etc. (2000)
15. Sneddon, I. N.: Fourier Transforms. Courier Corporation (1995)

Mathematics and Computing, DOI 10.1007/978-3-319-49277-3
138 References
16. Babenko, K. I. 1987 Some remarks on the theory of surface waves of finite amplitude. Sov.
Math. Dokl. 35, 599603
17. Benjamin, T. B. 1968, Gravity currents and related phenomena. J. Fluid Mech. 31, part 2,
209248
18. Debnath, L.: Nonlinear Water Waves. Academic Press, Boston, MA (1994)
19. Drazin, P. G.: Introduction to Hydrodynamic Stability. Cambridge University Press, Cambridge
(2002)
20. Grimshaw, R. (Ed.): Enviromental Stratified Flows. Kluwer Academic Publishers, London
(2001)
21. Howard, L. N. 1961 Note on a paper of John W. Miles. J. Fluid Mech. 10, 509512
22. Johnson, R. S.: A Modern Introduction to the Mathematical Theory of Water Waves.
Cambridge University Press, Cambridge (1997)
23. Kochin, N. E., Kibel, I. A., Roze, N. V.: Theoretical Hydromechanics. John Wiley & Sons,
New York etc. (1964)
24. Lamb, H.: Hydrodynamics. Cambridge University Press, Cambridge (1993)
25. Milne-Thomson, L. M.: Theoretical Hydrodynamics. Mac. Millan and Co. Ltd., London (1968)
26. Miles, J. W. 1961 On the stability of heterogeneous shear fows. J. Fluid Mech. 10, 496508
27. Novikov, S. P., Manakov, S. V., Pitaevskii, L. P., Zakharov, V. E.: Theory of Solitons. The
Inverse Scattering Methods. Plenum Publ., New York (1984)
28. Pedlosky, J.: Geophysical Fluid Dynamics. Springer, New York etc. (1987)
29. Stoker, J. J.: Water Waves. The Mathematical Theory with Applications. Wiley, New York,
NY (1992)
30. Sutherland, B. R.: Internal Gravity Waves. Cambridge University Press, Cambridge (2010)
31. Teshukov, V. M. 1985 On the hyperbolicity of the long wave equations. Sov. Math. Dokl. 32
469437
32. Turner, J. S.: Buoyancy Effects in Fluids. Cambridge University Press, Cambridge etc. (1979)
33. Yih, C. S.: Stratified Flows. Academic Press, New York etc. (1980)
Index
A D
acoustic equations, 15 DAlembert formula, 45
Airy function, 55 de Broglie wave, 69
dependence, domain of, 4
determinacy, domain of, 4
direction, characteristic, 1, 13
B Dirichlet problem, 95
BenjaminOno equation, 76 dispersion relation, 44, 46
Benney equations, 88 domain, subcritical, 102
Bernoulli constant, 80 domain, supercritical, 102
Bernoulli function, 110 Doppler effect, 47
Bernoulli integral, 80 Doppler shift, 47
Boussinesq approximation, 104 DubreilJacotinLong equation, 136
Boussinesq equation, 49
Boussinesq equations, 92
Boussinesq parameter, 110 E
BoussinesqRayleigh equation, 96 eikonal equation, 16
BruntVisl frequency, 104 Euler identity, 40
BuckleyLeverett equation, 28 EulerBernoulli beam equation, 53
buoyancy frequency, 104 EulerLagrange equation, 136
Burgers equation, 28 Eulerian coordinates, 78
F
C field, linearly degenerate in the sense of Lax,
Cauchy formula, 79 31
Cauchy problem, 4 fluid, stratified, 103
CauchyLagrange integral, 79, 86 Fourier integral, 51
CauchyPoisson problem, 82 Fourier transform, 53
CauchyRiemann equations, 90 Froude number, 83
Chaplygin isentropic gas, 24
characteristic, 1, 14 G
circulation, generalized, 94 Galilean transformation, 124
ColeHopf transform, 29 GardnerMiura transformation, 69
conservation laws, 7, 77

Mathematics and Computing, DOI 10.1007/978-3-319-49277-3
140 Index
genuine nonlinearity in the sense of Lax, 31 P

Gibbs identity, 23 Planck constant, 69
GodunovFriedrichsLax theorem, 16 Poisson coefficient, 75
GreenNaghdi equations, 93
R
H RankineHugoniot conditions, 9
Hamilton system, 14 Rayleigh wave, 73
HamiltonJacobi equation, 14 RayleighTaylor instability, 98
Hamiltonian formulation, 68 Riccati equation, 6
Helmholtz equation, 78 Riemann invariant, 3
Helmholtz representation, 72 Riemann problem, 11
Hertz law, 75 RiemannLebesgue lemma, 65
Hessian matrix, 17
Hilbert transform, 76
Howard semicircle theorem, 108 S
Schrdinger equation, 56
SerreSuGardner equations, 91
J shallow water equations, 3, 85
Jacobi amplitude, 58 shear flow, 103
Jacobian, 17 shock adiabat, 9
SmirnovSobolev functional-invariant
solution, 38
K Stokes flow, 123
KelvinFoigt medium, 63 stratification, 103
KelvinHelmholtz instability, 98 SturmLiouville problem, 68
Korteweg-de Vries equation, 57 system, t-hyperbolic Friedrichs, 16
Korteweg-de VriesBurgers equation, 45 system, conservative, 7
Kronecker symbol, 37 system, hyperbolic, 1, 14
system, strictly hyperbolic, 2
L
Lagrange series, 90 T
Lagrange theorem, 79 telegraph equation, 44
Lagrangian coordinates, 78 transport equation, 6
Lam constants, 32
Lam equations, 48
Langmuir adsorption equations, 36
Langmuir isotherm, 35 U
Laplace equation, 83 UmovPoynting vector, 39
Laplace operator, 69
Legendre transform, 17
LeviCivita variables, 122 V
velocity, critical, 83
velocity, group, 49
M velocity, phase, 43
Madelung transformation, 70 Vieta formula, 96
Maxwell equations, 39 von Mises yield criterion, 34
Maxwell medium, viscoelastic, 64
Miles theorem, 108
W
wave packet, elementary, 43, 47
O wave, cnoidal, 58
OstrogradskiiLiouville formula, 79 wave, deep water, 83
Index 141
wave, dispersive, 44 wavelength, 43

wave, internal, 103 Whitham equation, 75
wave, kinematic, 10
wave, long, 83
wave, solitary, 97 Y
wave, stationary, 83 Young modulus, 75

Waves in Continuous Media PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Waves in Continuous Media PDF

Caricato da

Copyright:

Formati disponibili

Lecture Notes in Geosystems Mathematics

Waves in Continuous Media

Lecture Notes in Geosystems Mathematics and Computing

Library of Congress Control Number: 2017930812

Springer International Publishing AG 2017

Printed on acid-free paper

This book is published under the trade name Birkhuser, www.birkhauser-science.com

Marseille, France S.L. Gavrilyuk

3.10 Stationary Internal Waves . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109

1.1 Hyperbolic Systems

We consider the quasilinear system of first order equations

ut C A.u; x; t/ux C b.u; x; t/ D 0; (1.1)

where the n n-matrix A and vector b depend on x, t and u D .u1 ; : : : ; un /T . A

det .A cI/ D 0: (1.2)

where dt D @t C c@x is the operator of differentiation along the characteristic.

Springer International Publishing AG 2017 1

Fig. 1.1 Location of

@t .u C f.u// C v@x u D 0; (1.4)

where u D .u1 ; : : : ; un /T are the concentrations of the separated substances passing

are real, positive, and distinct:

0 < 1 < : : : < n :

A cI D ..v c/I cf0 .u//.I C f0 .u//1 ;

then the relation (1.3) is equivalent to the equation

1.2 Propagation of Weak Discontinuities

Fig. 1.2 The solution to the

Fig. 1.3 The weak

Consequently, from the system (1.1) we have

with the initial conditions

obtain the transport equation

1.3 Motion with Strong Discontinuities

@t '.x; t; u/ C @x .x; t; u/ D f.x; t; u/; (1.7)

where ' D .'1 ; : : : ; 'n /T , D . 1 ; : : : ; n /T , and f D . f1 ; : : : ; fn /T are n-

Zx2 xDx2 Zx2

Systems of the form (1.1) representable as systems of independent conservation

Solution According to the definition of a conservation law, we have

In this problem, it is required to find all solutions of the form

' D .u/v 2 C .u/v C .u/

' D C1 '1 C : : : C C5 '5 ;

'1 D u; '2 D 0 v; '3 D uv;

Fig. 1.4 The solutions can

0 .v v0 /2 D f.u/ .u0 /g.u u0 /: (1.10)

obtain the expression

1.4 Kinematic Waves

By kinematic waves we mean the class of one-dimensional motions of a continuous

where c. / D Q0 . / is the characteristic velocity.

(a) Q. / > 0 .0 < < /;

A typical graph of Q. / is shown in Fig. 1.5.

Fig. 1.5 Flow-density

where i D const, 1 2 . This problem has a solution in the class of self-similar

which implies c1 > D > c2 , ci D Q0 . i /, provided that c. / is monotonically

.x; 0/ D 0 .x < 0/; .0; t/ D :

Fig. 1.6 .x; t/ - diagram

u D u0 ; D 0 .x < D0 t/; u D 0; D .D0 t < x < 0/;

1.5 Multi-dimensional Wave Fronts

Consider the linear system of equations

in R4 with normal vector D .

The system (1.14) is said to be hyperbolic at a point .x; t/ if for any D .

has n linearly independent vectors l 2 Rn such that l A./ D 0.

't D H.rx 'I x; t/: (1.16)

Characteristics of this equation are called bicharacteristics (rays in wave theory) of

where i D 1; 2; 3, pi D @'=@xi . For the linear hyperbolic system (1.14) with

Problem 1.5 The front of a two-dimensional acoustic wave propagating through a

where i D 1; 2; 3, pi D @ =@xi , and s is the parameter along the bicharacteristics.

' D .u/v 2 C .u/v C .u/

0 .v v0 /2 D f.u/ .u0 /g.u u0 /: (1.10)

(the Chaplygin isentropic gas with polytropic exponent D 1).

ut C uux D uxx . D const > 0/

@v1 @11 @12

@v2 @21 @22

where R3 is an arbitrary domain with piecewise smooth boundary S and

where s D eb is the UmovPoynting vector (the energy flux vector), R3