This action might not be possible to undo. Are you sure you want to continue?

# February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Publishers’ page

i

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Publishers’ page

ii

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Publishers’ page

iii

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Publishers’ page

iv

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

To my parents

v

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

vi

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Preface

It is known that the balance between nonlinearity and dispersion may result

in an appearance of localized long bell-shaped strain waves of permanent

form (solitary waves or solitons) which may propagate and transfer energy

over the long distance along. Starting with the ﬁrst documented water

surface solitary wave observation, made by J.Scott Russell yet in 1834,

solitons in ﬂuids were observed and generated many times. It was the

most surprising fact, however, that despite of almost similar description of

stresses in ﬂuids and solids, bulk longitudinal strain solitary waves have not

been observed in nonlinearly elastic wave guides.

One of reason of the lack of the results on nonlinear wave in solids is

that the complete description of a three-dimensional (3-D) nonlinear con-

tinuum is a diﬃcult problem. That is why initial 3-D problems are usually

reduced to the one-dimensional (1-D) form in order to clarify the simplest

but qualitatively new analytical solutions. Certainly the cylindrical elastic

rod seems to be a suitable real-life 1-D wave guide. Recently, the theory has

been developed to account for long longitudinal strain solitary waves prop-

agating in a free lateral surface elastic rod with permanent cross section.

The nonlinearity, caused by both the ﬁnite stress values and elastic mate-

rial properties, and the dispersion resulting from the ﬁnite transverse size

of the rod, when in balance allow the propagation of the bulk strain soli-

tary waves. Motivated by analytical theoretical predictions, there has been

successful experimental generation of strain solitary waves in a polystyrene

free lateral surface rod using the holographic interferometry. Hence it was

proven that bulk long localized nonlinear strain waves of permanent form

really exist.

However, presence of a dissipation (accumulation) destroys the balance

between nonlinearity and dispersion, and nonlinear strain wave in the rod

vii

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

viii Book Title

may attenuate or amplify. One possibility occurs when the radius of the

rod varies. Dissipative (active) eﬀects may be caused by internal features

of the elastic material, hence, an irreversible part should be included into

the stress tensor in addition to the reversible one depending only upon the

density of the Helmholtz energy. Dissipation (accumulation) may also come

in an elastic wave guide through phenomena occurring at its lateral surface.

Among the volume sources of dissipation/accumulation one can mention a

microstructure, point moving defects, thermal eﬀects. When dissipation

and activation act together there may be another balance resulted in a for-

mation of a bell-shaped wave with the amplitude and the velocity prescribed

by the condition of the dissipative/active balance. Hence the wave is se-

lected. Note that there exist another kind of nonlinear wave of a permanent

shape sustained either by a balance between nonlinearity and dissipation or

by a balance between nonlinearity, dispersion and dissipation. This wave

has the form of a shock and is often called kink-shaped wave or simply kink.

The ampliﬁcation of the waves (i.e., growth of the amplitude) may cause

the appearance of plasticity zones or microcracks in a wave guide. This is

of importance for an assessment of durability of elastic materials and struc-

tures, methods of non-destructive testing, determination of the physical

properties of elastic materials, particularly, polymeric solids, and ceramics.

Bulk waves provide better suited detection requirements than surface strain

waves in setting up a valuable non-destructive test for pipelines.

Inclusion of dissipation (accumulation) yields nonlinear dispersive- dis-

sipative governing equations that are nonintegrable as a rule. Hence, only

particular, usually travelling wave, solutions may be obtained analytically.

Certainly, these solutions require speciﬁc initial conditions. Moreover, they

usually have no free parameters, and additional relationships between the

equation coeﬃcients are required for the existence of the solutions. That is

why the obtaining of the exact solutions is often considered as useless by

many authors preferring to apply numerical methods only.

I would like to achieve two tasks in this monograph. First, it is planned

to provide the sequential analytical consideration of the strain waves am-

pliﬁcation/attenuation and selection in solids, mainly in an elastic rod. It

may be of interest for those working in the ﬁeld of solid mechanics. Another

task is to demonstrate the use of even particular analytical solutions for the

description of unsteady nonlinear wave processes. It may attract the atten-

tion of the specialists in various ﬁelds since the structure of the governing

equations is rather universal. The content is essentially based on the author

previous research. However, many works were done in a collaboration. The

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Preface ix

author thanks a lot Profs. G.V. Dreiden, I.L. Kliakhandler, G.A. Maugin,

F. Pastrone, D.F. Parker, A.M. Samsonov, M.G. Velarde; Mr. V.V. Gursky

and Mrs. I.V. Semenova for a long time fruitful collaboration. The book

preparation has been supported by the INTAS grant 99-0167 and by the

RFBR under Grant 2000-01-00482.

I dedicate this book to my parents. They always believe in my eﬀorts

and expected this book more than whoever it may be.

Saint-Petersburg, December, 2002 A.V. Porubov

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

x Book Title

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Contents

Preface vii

1. Basic concepts 1

1.1 Single nonlinear waves of permanent shape . . . . . . . . . 2

1.1.1 Monotonic bell-shaped solitary waves . . . . . . . . . 2

1.1.2 Oscillatory bell-shaped solitary waves . . . . . . . . . 6

1.1.3 Kink-shaped waves . . . . . . . . . . . . . . . . . . . 9

1.1.4 Periodic nonlinear waves . . . . . . . . . . . . . . . . 12

1.2 Formation of nonlinear waves of permanent shape from an

arbitrary input . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.2.1 Bell-shaped solitary wave formation from an initial

localized pulse . . . . . . . . . . . . . . . . . . . . . . 15

1.2.2 Kink-shaped and periodic waves formation . . . . . . 26

1.3 Ampliﬁcation, attenuation and selection of nonlinear waves 27

2. Mathematical tools for the governing equations analysis 31

2.1 Exact solutions . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.1.1 Direct methods and elliptic functions . . . . . . . . . 31

2.1.2 Painlev´e analysis . . . . . . . . . . . . . . . . . . . . 35

2.1.3 Single travelling wave solutions . . . . . . . . . . . . 36

2.1.4 Exact solutions of more complicated form . . . . . . 42

2.2 Asymptotic solutions . . . . . . . . . . . . . . . . . . . . . . 47

2.3 Numerical methods . . . . . . . . . . . . . . . . . . . . . . . 51

2.3.1 Nonlinear evolution equations . . . . . . . . . . . . . 52

2.3.2 Nonlinear hyperbolic equations . . . . . . . . . . . . 57

2.4 Use of Mathematica . . . . . . . . . . . . . . . . . . . . . . 59

xi

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

xii Book Title

3. Strain solitary waves in an elastic rod 63

3.1 The sources of nonlinearities . . . . . . . . . . . . . . . . . . 64

3.2 Modelling of nonlinear strain waves in a free lateral surface

elastic rod . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3.2.1 Statement of the problem . . . . . . . . . . . . . . . 66

3.2.2 Derivation of the governing equation . . . . . . . . . 69

3.3 Double-dispersive equation and its solitary wave solution . . 70

3.4 Observation of longitudinal strain solitary waves . . . . . . 75

3.5 Reﬂection of solitary wave from the edge of the rod . . . . . 79

4. Ampliﬁcation of strain waves in absence of external en-

ergy inﬂux 87

4.1 Longitudinal strain solitary wave ampliﬁcation in a narrow-

ing elastic rod . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.1.1 Governing equation for longitudinal strain waves

propagation . . . . . . . . . . . . . . . . . . . . . . . 87

4.1.2 Evolution of asymmetric strain solitary wave . . . . . 90

4.1.3 Experimental observation of the solitary wave ampli-

ﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . 93

4.2 Strain solitary waves in an elastic rod embedded in another

elastic external medium with sliding . . . . . . . . . . . . . 95

4.2.1 Formulation of the problem . . . . . . . . . . . . . . 96

4.2.2 External stresses on the rod lateral surface . . . . . . 99

4.2.3 Derivation of strain-displacement relationships inside

the rod . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.2.4 Nonlinear evolution equation for longitudinal strain

waves along the rod and its solution . . . . . . . . . 102

4.2.5 Inﬂuence of the external medium on the propagation

of the strain solitary wave along the rod . . . . . . . 104

4.2.6 Numerical simulation of unsteady strain wave propa-

gation . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4.3 Strain solitary waves in an elastic rod with microstructure . 114

4.3.1 Modelling of non-dissipative elastic medium with mi-

crostructure . . . . . . . . . . . . . . . . . . . . . . . 114

4.3.2 Nonlinear waves in a rod with pseudo-continuum

Cosserat microstructure . . . . . . . . . . . . . . . . 117

4.3.3 Nonlinear waves in a rod with Le Roux continuum

microstructure . . . . . . . . . . . . . . . . . . . . . . 120

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Front Matter xiii

4.3.4 Concluding remarks . . . . . . . . . . . . . . . . . . 121

5. Inﬂuence of dissipative (active) external medium 123

5.1 Contact problems: various approaches . . . . . . . . . . . . 123

5.2 Evolution of bell-shaped solitary waves in presence of ac-

tive/dissipative external medium . . . . . . . . . . . . . . . 125

5.2.1 Formulation of the problem . . . . . . . . . . . . . . 125

5.2.2 Dissipation modiﬁed double dispersive equation . . 126

5.2.3 Exact solitary wave solutions of DMDDE . . . . . . 128

5.2.4 Bell-shaped solitary wave ampliﬁcation and selection 130

5.2.5 Concluding remarks . . . . . . . . . . . . . . . . . . 134

5.3 Strain kinks in an elastic rod embedded in an ac-

tive/dissipative medium . . . . . . . . . . . . . . . . . . . . 136

5.3.1 Formulation of the problem . . . . . . . . . . . . . . 137

5.3.2 Combined dissipative double-dispersive equation . . 138

5.3.3 Exact solutions . . . . . . . . . . . . . . . . . . . . . 141

5.3.4 Weakly dissipative (active) case . . . . . . . . . . . . 143

5.3.5 Weakly dispersive case . . . . . . . . . . . . . . . . . 145

5.3.6 Summary of results and outlook . . . . . . . . . . . . 150

5.4 Inﬂuence of external tangential stresses on strain solitary

waves evolution in a nonlinear elastic rod . . . . . . . . . . 152

5.4.1 Formulation of the problem . . . . . . . . . . . . . . 152

5.4.2 Derivation of the governing equation . . . . . . . . . 153

5.4.3 Symmetric strain solitary waves . . . . . . . . . . . . 155

5.4.4 Evolution of asymmetric solitary waves . . . . . . . . 159

6. Bulk active or dissipative sources of the ampliﬁcation and selection163

6.1 Nonlinear bell-shaped and kink-shaped strain waves in mi-

crostructured solids . . . . . . . . . . . . . . . . . . . . . . . 164

6.1.1 Modelling of a microstructured medium with dissipa-

tion/accumulation . . . . . . . . . . . . . . . . . . . 165

6.1.2 Bell-shaped solitary waves . . . . . . . . . . . . . . . 169

6.1.3 Kink-shaped solitary waves . . . . . . . . . . . . . . 172

6.1.4 Concluding remarks . . . . . . . . . . . . . . . . . . 176

6.2 Nonlinear seismic solitary waves selection . . . . . . . . . . 178

6.2.1 Modelling of nonlinear seismic waves . . . . . . . . . 178

6.2.2 Asymptotic solution of the governing equation . . . . 180

6.2.3 Numerical simulations . . . . . . . . . . . . . . . . . 183

6.3 Moving defects induced by external energy ﬂux . . . . . . . 188

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

xiv Book Title

6.3.1 Basic concepts and derivation of governing equations 188

6.3.2 Nonlinear waves in a medium . . . . . . . . . . . . . 190

6.3.3 Nonlinear waves in a plate . . . . . . . . . . . . . . . 192

6.4 Thermoelastic waves . . . . . . . . . . . . . . . . . . . . . . 193

6.4.1 Nonlinear waves in thermoelastic medium . . . . . . 195

6.4.2 Longitudinal waves in thermoelastic rod . . . . . . . 196

Bibliography 199

Index 211

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 1

Basic concepts

This chapter is focused on some features of nonlinear waves to be used

further in the book. Linear waves are accounted for the linear equations,

they have inﬁnitesimal amplitude. Nonlinear waves are described by non-

linear equations. In contrast to the linear waves, an amplitude, a veloc-

ity and a wave number of the nonlinear waves are connected to one an-

other. More general information about nonlinear waves may be found in

numerous special books, like Ablowitz and Segur (1981); Bhatnagar (1979);

Calogero and Degasperis (1982); Newell (1985); Sachdev (1987); Whitham

(1974) etc.

The governing equations for the nonlinear strain waves to be considered

are nonitegrable by the inverse scattering transform method, and only par-

ticular exact solutions may be obtained. Of special interest are the single

travelling wave solutions that keep their shapes on propagation. This is a

result of the balances between various factors aﬀecting the wave behaviour.

There are two main types of the nonlinear travelling solitary waves which

could propagate keeping its shape, bell-shaped and kink-shaped solitary

waves. The bell-shaped solitary wave usually appears as a result of a bal-

ance between nonlinearity and dispersion. The kink-shaped wave may be

sustained by diﬀerent balances, one possibility occurs when nonlinearity is

balanced by dissipation (or accumulation) , another case corresponds to the

simultaneous balance between dispersion, nonlinearity and dissipation (or

accumulation). The single travelling wave solution requires speciﬁc initial

conditions. However, one can show that these solutions account for the

ﬁnal quasistationary part of an arbitrary initial pulse evolution. This un-

steady process may be described analytically for the integrable equations

or numerically for others. We illustrate all mentioned above further in this

Chapter.

1

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

2 Ampliﬁcation of Nonlinear Strain Waves in Solids

1.1 Single nonlinear waves of permanent shape

1.1.1 Monotonic bell-shaped solitary waves

The simplest celebrated model equation containing nonlinear and dispersive

terms is the well-known Korteweg- de Vries (KdV) equation Korteweg and

de Vries (1895),

u

t

+ 2b uu

x

+d u

xxx

= 0, (1.1)

whose exact one-parameter single solitary wave solution is

u = 6

d

b

k

2

cosh

−2

k(x −4dk

2

t). (1.2)

The wave amplitude A = 6dk

2

/b and the velocity V = 4dk

2

depend

upon the wave number k which is a free parameter. One can call the solution

(1.2) travelling solitary wave one since it depends upon the phase variable

θ = x−V t only, and monotonic solitary wave since it decays monotonically

when [θ[ → ∞. Typical shape of the wave is shown in Fig. 1.1 where one

can see also that the wave is symmetric with respect to its maximum.

-30 -20 -10 10 20 30

x

0.2

0.4

0.6

0.8

u

Fig. 1.1 Monotonic solitary wave (solid line) and its ﬁrst derivative (dashed line)

Sometimes there is a need for the inclusion of higher- order deriva-

tive (dispersion) or nonlinear terms into Eq.(1.1). A particular case arises

for water waves when surface tension suppresses coeﬃcient d Hunter and

Scheurle (1988) and ﬁfth-order dispersion u

5x

is added in Eq.(1.1).

Also higher- order derivative terms model weak nonlocality Engelbrecht

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 3

and Braun (1998), provide an improvement of bad dispersive properties

Christov et. al (1996); Maugin and Muschik (1994), account for a contin-

uum limit of discrete models with far neighbour interactions Kosevich and

Savotchenko (1999), to say nothing of dissipative (active) generalizations.

An example of the inclusion of higher- order nonlinearity is the Sawada-

Kotera equation Sawada and Kotera (1974).

Hence, the following nonlinear equation may be considered:

u

t

+ 2b uu

x

+ 3c u

2

u

x

+r uu

xxx

+s u

x

u

xx

+d u

3x

+f u

5x

= 0, (1.3)

We get from Eq.(1.3) a ﬁfth-order (in space derivatives) KdV equation

Hunter and Scheurle (1988) when c = r = s = 0. This equation was studied

in many papers, see, e.g., Karpman (1993); Karpman (1998); Karpman and

Vanden-Broeck (1995); Kawahara (1972); Kawahara and Takaoka (1988);

Benilov et. al (1993); Grimshaw et. al (1994). When, in addition d = 0,

the resulting equation models the LC ladder electrical transmittion lines.

Its solutions were obtained in Nagashima and Kuwahara (1981); Kano and

Nakayama (1981). A special integrable case corresponds to the Sawada-

Kotera equation with b = d = 0, c = −r = −s/2 = 10, f = 1 Sawada

and Kotera (1974). Its solitary wave solutions may be found in Parkes and

Duﬀy (1996), see also references therein.

Equation (1.3) is obviously nonintegrable by the Inverse Scattering

Transform method, and only particular exact solutions may be obtained.

Let us consider an exact solution vanishing at inﬁnity. In case of the ﬁfth-

order KdV equation it has the form Kano and Nakayama (1981):

u =

210d

13b

k

2

cosh

−4

k(x −V t), (1.4)

with k

2

= −d/(52f). Hence the width of the wave is prescribed by the

dispersion coeﬃcients d and f which should be of opposite sign. For the

wave velocity we have V = 144dk

2

/13 = −36d

2

/(169f). Hence, simultane-

ous triggering of the signs of b, d and f results in changing only the wave

propagation direction.

In the general case the exact solitary- wave solution has a form similar

to the KdV soliton (1.2),

u = Acosh

−2

k(x −V t). (1.5)

Important particular cases are:

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

4 Ampliﬁcation of Nonlinear Strain Waves in Solids

(i) In presence of only cubic nonlinear term, r = s = 0, we get the solution

with ﬁxed parameters,

A = 2

_

−

30f

c

k

2

, V = 4k

2

(2b

_

−

30f

c

−5d−116f k

2

), k

2

=

b

_

−30f/c −3d

60f

.

(1.6)

In this case the existence of solution vanishing at inﬁnity is provided by

the linear ﬁfth order term f u

5x

. Indeed at f = 0 we get from Eq.(1.3) the

Gardner equation whose solution is Grimshaw et. al (1999):

u =

A

1

coshm(x −V t) +B

1

, (1.7)

where

A

1

=

3

√

2d m

2

√

2b

2

+ 9cd m

2

, B

1

=

√

2b

√

2b

2

+ 9cd m

2

, V = dm

2

. (1.8)

An important feature of the solution is the existence of the ﬁnite limiting

amplitude when B is large Slyunyaev and Pelinovsky (1999). However, at

nonzero f a substitution of Eq.(1.7) into Eq.(1.3) yields B

1

= 1, cosh m(x−

V t) + 1 = 2 cosh

2

m(x − V t)/2, and we get the solution (1.5), (1.6) with

m = 2k.

(ii) When only quadratic higher- order term r uu

xxx

is taken into account,

c = s = 0, the ﬁxed parameters of the solution (1.5) are

A =

30fk

2

r

, V =

2(50b

2

f

2

+ 5bdfr −3d

2

r

2

)

25f r

2

, k

2

=

5bf −dr

10fr

. (1.9)

Note that the solution exists at d = 0.

(iii) In case c = r = 0 the parameter k is free but an additional restriction

on the equation coeﬃcients holds,

A =

60fk

2

s

, V = 4k

2

(d + 4f k

2

), 10b f = d s. (1.10)

One can see that s may be excluded from the amplitude expression using

the third formula from (1.10). Then the amplitude coincides with that of

the KdV soliton (1.2). We also see that the wave velocity consists of two

parts, V

1

= 4dk

2

and V

2

= 16fk

4

, the ﬁrst of which being exactly the KdV

soliton velocity. Let us rewrite the ODE reduction of the equation (1.3) in

the form (

= ∂/∂θ, θ = x −V t):

b u

2

+d u

−V

1

u +

s

2

u

2

+f u

−V

2

u = 0.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 5

One can check that the solution (1.5), (1.10) satisﬁes separately

b u

2

+d u

−V

1

u = 0, (1.11)

s

2

u

2

+f u

−V

2

u = 0,

where the ﬁrst of these equations is the ODE reduction of the KdV equation

(1.1) having a one-parameter solitary- wave solution.

(iv) Higher order nonlinear terms may support the existence of solitary-

wave solutions even in absence of the linear dispersion terms. Higher- order

nonlinear terms provide bounded localized solutions at d = 0 in contrast to

the case c = r = s = d = 0 Kano and Nakayama (1981). The parameters

of the solution (1.5) are

A =

120fk

4

b + 2k

2

(r +s)

, V = 16fk

4

,

while k satisﬁes the equation

4[30cf −r(r +s)]k

4

+ 2bsk

2

+b

2

= 0.

Thanks to the higher order terms the solitary- wave solution may exist

even in the absence of the KdV’s nonlinear term, b = 0, provided that the

restriction 30cf −r(r +s) = 0 is satisﬁed. Then k may be a free parameter.

When f = 0 we have

A =

3cd −2b(2r +s)

c(r +s)

, V =

2d[3cd −2b(2r +s)]

(r +s)(2r +s)

, k

2

=

3cd −2b(2r +s)

2(r +s)(2r +s)

.

There is no exact solution vanishing at inﬁnity in the case d = f = 0.

Instead the solution in the form of a solitary wave on an ”pedestal” may

be obtained as

u = Acosh

−2

k(x −V t) +B, (1.12)

with

A =

2k

2

(2r +s)

c

, B = −

(2r +s)[b + 2k

2

(r +s)]

3c(r +s)

,

V =

s(2r +s)[4k

4

(r +s)

2

−b

2

]

3c(r +s)

2

.

Even equations with dissipation may possess bell-shaped solitary wave

solution. In particular, it was recently found Garazo and Velarde (1991);

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

6 Ampliﬁcation of Nonlinear Strain Waves in Solids

Rednikov et. al (1995) that appropriately heating a shallow horizontal liq-

uid layer long free surface waves u(x, t) can be excited whose evolution is

governed by a dissipation-modiﬁed Korteweg- de Vries (DMKdV) equation

u

t

+ 2α

1

uu

x

+

∼

α

2

u

xx

+α

3

u

xxx

+

∼

α

4

u

xxxx

+

∼

α

5

(uu

x

)

x

= 0. (1.13)

The coeﬃcients in Eq.(1.13) depend upon parameters characterizing the

liquid (Prandtl number etc.), temperature gradient across the layer, and its

depth. The exact travelling bell-shaped solitary wave solution have been

obtained in the form (1.12) Lou et. al (1991); Porubov (1993) with

A = 12

∼

α

4

k

2

/

∼

α

5

, B = −(

∼

α

2

+4

∼

α

4

k

2

)/

∼

α

5

,

V = −2α

1

∼

α

2

/

∼

α

5

, α

3

= 2α

1

∼

α

4

/

∼

α

5

. (1.14)

The meaning of the last expression in (1.14) is similar to that in case c =

r = 0 for Eq.(1.3). Indeed, when the relationships for V and α

3

hold, the

ODE reduction Eq.(1.13) may be rewritten as

(

∼

α

2

∂

∂θ

+

2α

1

∼

α

2

∼

α

5

)

_

u

θ

+

∼

α

4

∼

α

2

u

θθθ

+

∼

α

5

2

∼

α

2

(u

2

)

θ

_

= 0, (1.15)

The restrictions on the equation coeﬃcients do not necessary provide

an evidence of the KdV ODE reduction like Eqs.(1.11), (1.15). Particular

case corresponds to the Kawahara equation (Eq.(1.13) with

∼

α

5

= 0) whose

exact solution is Kudryashov (1988):

u =

15α

3

3

128α

1

∼

α

2

4

cosh

−2

(

α

3

8

∼

α

4

θ)(1 −tanh(

α

3

8

∼

α

4

θ)), (1.16)

where V = 5α

3

3

,

∼

α

2

= α

2

3

/(16

∼

α

4

).

All solutions (1.4), (1.5), (1.7) account for monotonic and symmetric

solitary waves. Despite diﬀerence in their functional form they have one and

the same shape shown in Fig. 1.1. In contrast to them exact solitary wave

solution (1.16) of the Kawahara equation is monotonic but asymmetric, see

Fig. 1.2.

1.1.2 Oscillatory bell-shaped solitary waves

The solitary wave does not decay necessarily in a monotonic manner. Thus

Kawahara (1972) studied decay at inﬁnity of the wave solution of the ﬁfth-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 7

-30 -20 -10 10 20 30

x

-0.1

-0.05

0.05

u

b

-30 -20 -10 10 20 30

x

0.2

0.4

0.6

0.8

u

a

Fig. 1.2 Symmetric (solid line) vs asymmetric monotonic solitary wave (dashed line).

a) proﬁles; b)their ﬁrst derivatives.

order KdV equation using linearized equation analysis phase analysis. It

was found that the wave decays monotonically or oscillatory depending

upon the parameter ε, which is proportional (in our notations) to d and

is inverse proportional by the product of f and the wave velocity. The

same technique has been used in Karpman (2001) when nonlinear term in

the ﬁfth-order KdV equation is of the form u

p

u

x

. The oscillatory solitary-

wave solution is shown in Fig. 1.3. The proﬁle of the ﬁrst derivative with

respect to the phase variable reveals its symmetric nature.

Eq.(1.3) does not possess exact oscillatory solitary wave solution. How-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

8 Ampliﬁcation of Nonlinear Strain Waves in Solids

-4 -2 2 4

x

-0.5

-0.25

0.25

0.5

0.75

1

u

Fig. 1.3 Oscillatory solitary wave (solid line) and its ﬁrst derivative (dashed line)

ever, it may be described asymptotically. Certainly the ﬁfth-order KdV

equation is often considered as a perturbed KdV equation. First the asymp-

totic solution is obtained which consists of the KdV solitary wave solution

and a small perturbation that oscillates but does not vanish at inﬁnity

or a non-local solution Benilov et. al (1993); Hunter and Scheurle (1988);

Karpman (1993); Karpman (1998). Let us consider the ﬁfth- order deriva-

tive term and higher- order nonlinear terms as small perturbations assum-

ing f = δF, r = δR, s = δS, c = δC, δ << 1. The asymptotic solution

u = u(θ), θ = x −V t, is sought in the form

u = u

0

(θ) +δ u

1

(θ) +..., (1.17)

with u

i

→ 0 at [θ[ → ∞. In the leading order we get the KdV equation

(1.1) for the function u

0

whose travelling solitary wave solution is (1.2). In

the next order an inhomogeneous linear equation results for u

1

,

2b (u

0

u

1

)

θ

+d u

1,θθθ

−4dk

2

u

1,θ

= −F u

0,θθθθθ

−3C u

2

0

u

0,θ

−Ru

0

u

0,θθθ

−

S u

0,θ

u

0,θθ

. (1.18)

Its solution vanishing at inﬁnity is

u

1

= (

3C

4b

+

5bF

2d

2

−

2R +S

4d

) u

2

0

+(

S

b

+

4R

b

−

14F

d

−

9Cd

b

2

)k

2

u

0

−

2k

2

F

d

θ u

0,θ

.

(1.19)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 9

The shape of the solution u = u

0

(θ) +δ u

1

(θ) depends upon the values

of the coeﬃcients of Eq.(1.3). It may account for an oscillatory solitary

wave solution ﬁrst obtained numerically in Kawahara (1972), see Fig. 1.3.

In the case of the ﬁfth-order KdV equation, this proﬁle exists at positive d

and f, what corresponds to the Case IV in Kawahara (1972).

Finally, there exist nonlinear equations having exact travelling wave

solutions in the form of an oscillatory solitary wave. In particular, an

equation

u

t

+a u

x

+ 2b uu

x

+ 3c u

2

u

x

+d u

3x

+f u

5x

+g u

7x

= 0, (1.20)

possesses the exact solution

u =

_

35g

c

k

3

cosh

−1

k(x −V t)

_

24 cosh

−2

k(x −V t) −

288

17

_

,

where b = 0, k

2

= 17f/(581g), V = a + 102825k

6

g/289. The equation

coeﬃcients should be connected by d g = 37.405 f

2

.

Oscillatory solitary waves of permanent shape arise also in dissipative

problems, however, usually they are asymmetric and may be found only

numerically.

1.1.3 Kink-shaped waves

The celebrated Burgers’ equation Burgers (1948) is the simplest equation

that models the balance between nonlinearity and dissipation,

u

t

+ 2b uu

x

+g u

xx

= 0, (1.21)

In particular, it possesses the exact travelling solution of permanent form,

u = Am tanh(mθ) +B, (1.22)

where

A =

g

b

, B =

V

2b

, m−free.

If the boundary conditions are

u →h

1

at θ →∞, u →h

2

at θ →−∞, (1.23)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

10 Ampliﬁcation of Nonlinear Strain Waves in Solids

then

m =

(h

1

−h

2

)b

2g

, V = b(h

1

+h

2

).

-4 -2 2 4

x

-1

1

2

3

4

5

u

Fig. 1.4 Burgers’ kink-shaped wave

The shape of the solution (1.22), called kink, is shown in Fig. 1.4. Kinks

may arise also due to the balance between nonlinearity, dispersion and dissi-

pation like in the case of the Korteweg-de Vries-Burgers equation (KdVB),

u

t

+ 2b uu

x

+g u

xx

+du

xxx

= 0, (1.24)

whose exact solution was obtained independently by many authors Vlieg-

Hultsman and Halford (1991)

u = A tanh(mθ)sech

2

(mθ) + 2A tanh(mθ) +C, (1.25)

with

A =

6g

2

50V d

, C =

V

2b

, m =

g

10d

.

It follows from the boundary conditions (1.23) that

h

+

−h

−

= 2B, V = b(h

+

+h

−

),

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 11

and the solution exists under

h

2

+

−h

2

−

=

12g

2

25bd

.

Equations without dissipative terms may also possess the kink-shaped

solutions. Thus the modiﬁed Korteweg-de Vries equation (MKdV),

u

t

+ 3c u

2

u

x

+du

xxx

= 0, (1.26)

has an exact solution

u =

_

−

2d

c

m tanh(mθ), (1.27)

where V = −dm

2

. Note that the MKdV equation has both the bell-shaped

and the kink-shaped solutions. Dissipative equation may possess the same

property. In particular, the DMKdV equation (1.13) has exact kink-shaped

solution Lou et. al (1991); Porubov (1993)

u = C +

36

∼

α

4

∼

α

5

m cosh

−2

(

√

−3mθ)+

12(α

3

∼

α

5

−2α

1

∼

α

4

)

5

∼

α

5

√

−3mtanh(

√

−3mθ), (1.28)

where

C =

12

∼

α

4

∼

α

5

m−

∼

α

2

∼

α

5

+

2α

1

(α

3

∼

α

5

−2α

1

∼

α

4

)

5

∼

α

3

5

+

(α

3

∼

α

5

−2α

1

∼

α

4

)

2

25

∼

α

4

∼

α

3

5

,

V = 2α

1

C +

24α

1

∼

α

4

∼

α

5

m,

while m is a free parameter if α

3

= 12α

1

∼

α

4

/

∼

α

5

or

m = −

(α

3

∼

α

5

−2α

1

∼

α

4

)

2

300

∼

α

2

4

∼

α

2

5

.

The shapes of the kink solutions (1.25), (1.27) are similar to that shown

in Fig. 1.4, while the kink (1.28) may have also diﬀerent proﬁle shown by

solid and dashed lines in Fig. 1.5.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

12 Ampliﬁcation of Nonlinear Strain Waves in Solids

-4 -2 2 4

x

-1.5

-1

-0.5

0.5

1

u

Fig. 1.5 Kink-shaped waves with a ”hat”

1.1.4 Periodic nonlinear waves

Usually single bell-shaped solitary wave solutions are the particular cases of

more general periodic solutions. Thus Korteweg and de Vries (1895) found

the periodic solution of the KdV equation (1.1),

u = 6

d

b

k

2

_

1 −κ

2

+

E

K

+κ

2

cn

2

(kθ, κ)

_

(1.29)

where K and E are the complete elliptic integrals of the ﬁrst and the

second kind respectively , κ is the modulus of the Jacobian elliptic function

Bateman and Erdelyi (1953-54); Byrd and Friedman (1954); Newille (1951).

They called Eq.(1.29) the cnoidal wave solution since it is expressed through

the Jacobi elliptic function cn . Cnoidal wave is not a linear superposition of

the bell-shaped solitary waves. It tends to the single solitary wave solution

(1.2) at κ → 1 as shown in the right column in Fig. 1.6

1

. Exact periodic

and bell-shaped solitary wave solutions correspond in the same manner

in case of the generalized 5th-order KdV equation (1.3) and the DMKdV

equation (1.13).

Although many equations, like Burgers’, BKdV and DMKdV equations

have not exact bounded periodic solutions that transform into the kink-

shaped ones, there exist exceptions. Thus the MKdV periodic solution

1

Reprinted with permission from Elsevier Science

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 13

Ablowitz and Segur (1981)

u =

_

−

2d

c

msn(mθ, κ), (1.30)

transforms into the kink solution (1.27) at κ → 1. Another example is a

dissipative nonlinear equation

u

t

+ 2buu

x

+ 3c u

2

u

x

+du

xxx

+f(u

2

)

xx

+gu

xx

= 0, (1.31)

This evolution equation represents an analog of the hyperbolic equation

to be derived further in Sec. 5.3. Its bounded periodic solution Porubov

(1996) is

u =

m

√

−c

cn(mθ, κ) sn(mθ, κ) dn(mθ, κ)

C

1

+cn

2

(mθ, κ)

−

b

3c

. (1.32)

with

C

1

=

1 −2κ

2

+

√

κ

4

−κ

2

+ 1

3κ

2

, m

2

=

3g

2

−V

4

√

κ

4

−κ

2

+ 1

,

and the following restrictions on the coeﬃcients:

f = −

1

2

√

−c, b = 3g

√

−c.

The periodic wave solution (1.32) has a functional form diﬀerent from both

the KdV cnoidal wave and the MKdV bounded periodic solution. When

κ = 1 we have C

1

= 0, and the solution (1.31) tends to the kink-shaped

solution (1.27) as it is shown in the left column in Fig. 1.6 in comparison

with the transformation of the KdV cnoidal wave solution to the bell-shaped

solitary wave.

1.2 Formation of nonlinear waves of permanent shape from

an arbitrary input

All solutions presented in previous section require speciﬁc initial conditions.

In practice more important is to know how an arbitrary ﬁnite amplitude

input evolves. Analytical solutions of unsteady problems may be obtained

if governing nonlinear equations are integrable Ablowitz and Segur (1981);

Bhatnagar (1979); Calogero and Degasperis (1982); Dodd et. al (1982);

Newell (1985), otherwise only numerical solutions are available. Often ini-

tial input transforms into the stable quasistationary wave structures of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

14 Ampliﬁcation of Nonlinear Strain Waves in Solids

d

-10 -5 5 10

x

-2

-1

1

2

u

-10 -5 5 10

x

-1

1

2

3

u

c

-10 -5 5 10

x

-2

-1

1

2

u

-10 -5 5 10

x

-1

1

2

3

u

b

-10 -5 5 10

x

-2

-1

1

2

u

-10 -5 5 10

x

-1

1

2

3

u

a

-10 -5 5 10

x

-0.4

-0.2

0.2

0.4

u

-10 -5 5 10

x

-0.4

-0.2

0.2

0.4

u

Fig. 1.6 Comparison of the periodic solution (1.32) (left column) and the KdV cnoidal

wave (1.29)(right column) for diﬀerent values of the Jacobi modulus: (a)κ

2

= 0.25, (b)

κ

2

= 0.995, (c)κ

2

= 0.99995, (d)κ

2

= 1. After Porubov and Velarde (2002).

permanent form which may be described by the analytical solutions. Also

the analysis gives the conditions when the formation of them is possible.

In this section we illustrate it using some instructive examples.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 15

1.2.1 Bell-shaped solitary wave formation from an initial

localized pulse

The ﬁfth-order KdV was extensively studied numerically. The oscil-

latory travelling solitary- wave solutions were found in Boyd (1991);

Nagashima and Kuwahara (1981). The evolution of the initial monotonic

solitary wave into radiating or oscillatory solitary waves was simulated in

Benilov et. al (1993); Karpman and Vanden-Broeck (1995). In a series of

papers Salupere et. al (1997); Salupere et. al (2001) the solitary wave forma-

tion from a periodic input was studied for an equation similar to Eq.(1.3).

We shall study the evolution of a localized initial pulse. Previously, local-

ized pulse evolution into an oscillatory solitary wave was considered in

Nagashima and Kuwahara (1981) for the equation u

t

+ uu

x

−γ

2

u

5x

= 0.

Below we consider the formation of solitary waves in the systems governed

by Eq.(1.3). Following Porubov et. al (2002) we use two methods for com-

putations, ﬁnite-diﬀerence and pseudo-spectral, see Sec. 2.3.2. Below only

those results are shown that were obtained using both numerical methods.

We have tried various shapes of the initial localized pulses, rectangular,

Gaussian distribution etc.

Inﬂuence of the ﬁfth-order dispersive term. First the 5th-order KdV

equation was studied. Since the role of the ﬁfth-order derivative term is of

interest the coeﬃcients b and d in Eq.(1.3) were ﬁxed for all computations,

b = 1, d = 0.5. We found that the rectangular initial pulse splits into a

sequence of solitary waves when the coeﬃcients of dispersive terms, d and

f, are of opposite sign. For both coeﬃcients positive the initial rectan-

gular proﬁle is dispersed without formation of any localized waves. The

dependence upon the sign of the ratio d/f is in agreement with the exact

solitary- wave solution (1.4) and the analysis of the dispersion relation

Karpman (1993). However, more smooth Gaussian initial proﬁles provide

the appearance of solitary waves even for positive coeﬃcients when f is

rather small, e.g., f = 0.01.

The next result we have obtained is the dependence of the number of

solitary waves upon the value of f when d/f < 0.

Shown in Fig. 1.7 is the formation of the train of solitary waves from a

Gaussian initial pulse in the KdV case, f = 0.

Figs. 1.8-1.10 demonstrate the decrease of the solitary waves for f =

−1, −10, −50 respectively. Both the amplitude and the velocity decrease

with the increase of the absolute value of f in qualitative agreement with

the exact solution (1.4).The ratio between the amplitude and the velocity

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

16 Ampliﬁcation of Nonlinear Strain Waves in Solids

40 80 120 160 200

x

0

0.2

0.4

u

Fig. 1.7 Evolution of the initial Gaussian proﬁle in the KdV case, f = 0.

of each solitary wave in Fig. 1.7 is equal to 1.5 just as for the KdV soliton

(1.2) . This ratio (and the amplitude) decreases with the decrease of f, from

1.43 at f = −1 to 1.33 at f = −50. A similar tendency is revealed by the

phase analysis of single travelling wave solutions, cf. Fig. 2 in Kawahara

(1972). The ratio for the 5th- order KdV exact solution (1.4) is 1.46, the

amplitude and the velocity for given b and d are −105/(1352f), −9/(169f),

respectively. Only at small f = −0.1 ÷−0.15 are the numerical results for

the leading solitary wave in quantitative agreement with the exact solution

(1.4). More important is that the decrease of f aﬀects the solitary- wave

transformation from monotonic KdV solitons (1.2) to the oscillatory soli-

tary waves. For convenience the last stages from Figs. 1.7-1.10 are collected

in Fig. 1.11. In case f = −1, Fig. 1.11(B), the higher leading solitary wave

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 17

60 120 180 240 300

x

0

0.2

0.4

u

Fig. 1.8 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case, f = −1.

is oscillatory while other solitary waves remain monotonic, then the trans-

formation occurs also for the second solitary wave, Fig. 1.11(C) . Alternate

transformation of the solitary waves conﬁrms the dependence of the kind

of solitary wave upon the value of the product of f and the wave amplitude

(hence, its velocity) found in Kawahara (1972).

Finally, we have found that simultaneous triggering of the signs of b, d,

f doesn’t aﬀect the shapes of the solitary waves. They simply evolve to the

opposite direction according to the analysis of the exact solution (1.4).

Inﬂuence of the cubic nonlinearity. Now we add cubic nonlinearity,

c ,= 0, while r = s = 0. The analytical solutions predict an action of

the cubic nonlinear term depends upon the sign of c. Also the solution is

sensitive to the ratio between nonlinear terms contributions, b/c, and the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

18 Ampliﬁcation of Nonlinear Strain Waves in Solids

80 160 240 320 400

x

0

0.2

0.4

u

Fig. 1.9 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case, f = −10.

value of the amplitude of an initial pulse. Indeed we have found that at

b = 1, c = 1, d = 0.5, f = −1 the train of solitary waves arises only from

positive initial pulse with amplitude equal to 0.5 while a negative one is

dispersed. Dependence of the amplitude on the sign of b is very typical for

the exact solutions of the equation with quadratic nonlinearity, see (1.2),

(1.4). However, at smaller b, b = 0.2 or b = −0.2, the formation of solitary

waves no longer depends upon the sign of the initial pulse amplitude. A

similar tendency is observed when the initial amplitude becomes close to 1

or higher when predominant cubic nonlinearity excludes an inﬂuence of the

quadratic one on the sign of the wave amplitude like in the exact solution

(1.5), (1.6).

At the same time, the stage of forthcoming evolution of already gen-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 19

100 200 300 400 500

x

0

0.2

0.4

u

Fig. 1.10 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case, f = −50.

erated solitary waves is not so sensitive to the value of b, while other an-

alytical restrictions on the coeﬃcients become more important. Besides

the condition d/f < 0 following from the linear analysis Karpman (1993);

Kawahara (1972), there is f/c < 0 resulted from the nonlinear exact so-

lution (1.5), (1.6). Moreover, at small f one can anticipate an evidence of

the condition d/c > 0 given by Eq.(1.8).

When c > 0, f < 0 all above mentioned inequalities are satisﬁed. The

number of solitary waves generated from the initial localized pulse increases

with the increase of the value of c and ﬁxed values of d = 0.5, f = −1 and

also b = 1 or b = 0.2. The velocity of the waves increases also, the width

(proportional to 1/k) decreases, while the amplitude remains practically one

and the same. We also observed the alternate transformation of the solitary

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

20 Ampliﬁcation of Nonlinear Strain Waves in Solids

140 180 220 260 300 340

x

0

0.05

0.1

0.15

0.2

u

C

160 210 260 310 360 410

x

0

0.1

0.2

0.3

0.4

u

D

70 90 110 130 150 170

x

0

0.05

0.1

0.15

0.2

u

A

100 130 160 190 220 250

x

0

0.02

0.04

0.06

0.08

0.1

u

B

Fig. 1.11 Transformation of the kind of solitary waves in the 5th-order KdV case. (A)

f = 0, (B) f = −1, (C) f = −10, (D) f = −50.

waves from monotonic to oscillatory when c increases for both values of b.

Independence of the amplitude of c doesn’t follow from the exact solution

(1.6) as well as from the asymptotic solution. The third formula in (1.6)

predicts growth of positive values of k

2

only for b = 0.2 giving negative

values for b = 1. However, let us express k through the amplitude A and

substitute it into the expression for the velocity V . Then one can exhibit for

both values of b the similarity of the variation of the velocity with respect

to c with that obtained in numerics.

As found in previous subsection, the decrease of the negative f values

aﬀects the decrease in the number of solitary waves. Assume b = 1, d = 0.5,

we have tried simultaneous variations of c and f in order to sustain one

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 21

100 150 200 250 300 350

x

0

0.05

0.1

0.15

0.2

u

B

100 150 200 250 300 350 400 450

x

0

0.05

0.1

0.15

0.2

u

C

40 60 80 100 120 140 160

x

0

0.05

0.1

0.15

0.2

u

A

Fig. 1.12 Conservation of the number of solitary wave thanks to the simultaneous pres-

ence of cubic nonlinearity and ﬁfth-order derivative term. (A) c = 0, f = 0, (B) c = 15,

f = −10, (C) c = 75, f = −50.

and the same number, see Fig. 1.12. We see that the wave amplitude

keeps its value from Fig. 1.12(A) to Fig. 1.12(C) while the velocity growths.

This conﬁrms that the amplitude depends upon the ratio f/c but velocity

is proportional to f. The kind of solitary wave alters from monotonic

Fig. 1.12(A) to oscillatory Fig. 1.12(B, C).

At c < 0, d > 0, f > 0 only one inequality, f/c < 0, is satisﬁed, and

solitary wave formation is observed only for small f , otherwise the initial

pulse is dispersed in this case. On the contrary small absolute values of c

provides the solitary waves formation at c < 0, d > 0, f < 0. When both c

and f are positive no solitary waves appear.

Inﬂuence of nonlinearities s u

x

u

xx

and r uu

xxx

. Suppose c = r = 0

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

22 Ampliﬁcation of Nonlinear Strain Waves in Solids

80 100 120 140 160

x

0

0.1

0.2

0.3

0.4

0.5

0.6

u

r=-1.60

80 100 120 140 160

x

0

0.1

0.2

0.3

0.4

0.5

0.6

u

r=-1.50

80 100 120 140 160

x

0

0.1

0.2

0.3

0.4

0.5

0.6

u

r=-1.57

Fig. 1.13 Equalization of the ﬁrst and the second solitary waves and subsequent ex-

ceeding of the second wave due to the alteration of the negative values of the coeﬃcient

r.

and vary s at ﬁxed b, d and f, that we choose b = 1, d = 0.5, f = −1 . It

is found that the amount of solitary waves and its transition from mono-

tonic to oscillatory don’t depend upon the value of s. The wave amplitude

decreases with the increase of s while the velocity keeps its value. Wave

behavior is not sensitive to the sign of s. The condition for the solitary

wave formation d/f < 0 remains valid.

The fact the velocity doesn’t depend upon s is in agreement with the

exact solution. Certainly, solitary waves exist outside the restriction from

(1.10). We also used numerical values of the amplitude to deﬁne k and

then V using (1.10). A comparison of the velocities with those obtained

numerically demonstrates the more agreement the less is the value of b.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 23

Asymptotic solution (1.17), (1.19) also predicts the decrease of amplitude at

permanent velocity. Indeed, we get that u

max

= u

0

(0)+δu

1

(0) = 6dk

2

/b(1+

k

2

[f/d −s/(2b)]). At coeﬃcient values we used the value of u

max

decreases

with the increase of s (s is not large in the asymptotic solution), while the

exact solution predicts the same behavior only for positive values of s.

When s = c = 0 the behavior of the solution diﬀers from the previous

one. Having the same values for b, d, f we obtain that increase of positive

values of r yields a decrease in the velocity and an increase in the amplitude

of the solitary waves. The number of solitary waves decreases. However,

at negative values of r we found that at the initial stage of the splitting

of the Gaussian proﬁle the amplitude of the second solitary wave becomes

equal to that of the ﬁrst one at r = −1.57, see Fig. 1.13. At lesser r second

solitary wave becomes higher, and two solitary waves form a two-humps

localized structure shown in Fig. 1.14.

It is no longer quasistationary since amplitudes of the humps vary in

time. It looks like an interaction of two solitary waves when the second

higher solitary wave surpasses the ﬁrst one, then it becomes lower, and

the process repeats. Decreasing r we achieve formation of a three-humps

localized structure shown in Fig. 1.15. Its evolution is similar to those

presented in Fig. 1.14. Finally, only multi-humps localized structure arises

from an initial pulse as shown in Fig. 1.16. The localized multi-humps

structures in Figs. 1.14-1.16 keep their width, while their shapes vary in

time.

Certainly, unsteady multi-humps localized structures are not governed

by the ODE reduction of Eq.(1.3) and, hence cannot be explained either by

the phase portraits analysis or by the exact travelling wave solution (1.5),

(1.9). Moreover, at negative values of f the exact solution doesn’t predict

propagation to the right of the solitary wave with positive amplitude.

Absence of linear dispersive terms. We have found an exact solitary

wave solution that may be supported by higher -order nonlinear terms even

without linear dispersive terms, at d = 0 or f = 0. Numerical simulations

show that there are no solitary waves at both zero d and f. Some solutions

from previous subsections keep their features at d = 0, in particular, this

relates to the case r ,= 0. At the same time cubic nonlinearity at d = r =

s = 0 supports two-humps localized waves for c > 0. At negative c the wave

picture is similar to those at d ,= 0. No stable solitary waves propagate in

absence of only the ﬁfth-derivative term, d ,= 0, f = 0 with the exception of

the Gardner equation case where Slyunyaev and Pelinovsky (1999) found

generation of the limiting amplitude solitons.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

24 Ampliﬁcation of Nonlinear Strain Waves in Solids

60 120 180 240 300

x

0

0.2

0.4

0.6

u

Fig. 1.14 Two-humps solitary wave formation at r = −1.6.

To sum up, both higher order nonlinear and dispersive terms aﬀect the

formation of localized nonlinear waves their shape and their parameters.

Thus, the number of solitary waves and the transition from monotonic to

oscillatory wave are under responsibility of both 5th- order linear dispersive

term, cubic nonlinearity and higher- order quadratic nonlinearity r u u

xxx

.

More important is the formation of an unsteady but localized multi-humps

wave structure thanks to r u u

xxx

and cubic nonlinearity at d = 0. The sign

of the coeﬃcient b of the KdV quadratic nonlinear term is important for

choosing the sign of the input amplitude. At the same time the nonlinearity

s u

x

u

xx

doesn’t aﬀect the formation and behavior of solitary waves.

Certainly, the shapes of the resulting solitary waves are not obviously

governed by the exact and asymptotic travelling wave solutions. Some

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 25

60 120 180 240 300

x

0

0.2

0.4

0.6

u

Fig. 1.15 Three-humps solitary wave formation at r = −3.2.

other features of numerical solutions, like the dependence of the number of

solitary waves upon the values of the equation coeﬃcients or a transition

from monotonic wave to an oscillatory one, are not predicted by analytical

solutions. However, the combinations of equation coeﬃcients required for

the existence of solitary wave are realized in numerics. Also numerical

wave amplitude and velocity relate like in the analysis. Evidence of all

these predictions even qualitatively is very important for a justiﬁcation of

the numerical results.

Formation of the bell-shaped solitary waves in presence of dissipation

or an energy inﬂux will be considered further in the book.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

26 Ampliﬁcation of Nonlinear Strain Waves in Solids

60 120 180 240 300

x

0

0.2

0.4

0.6

u

Fig. 1.16 No solitary waves other than multi-humps one at r = −12.

1.2.2 Kink-shaped and periodic waves formation

The formation of the kink-shaped waves was studied considering the

evolution of the Taylor shock from discontinuous (step) initial condi-

tions under the governance of the Burgers equation Sachdev (1987);

Whitham (1974). It was found the appearance in due time the steady state

kink solution (1.22). A quasihyperbolic analog of the Burgers equation was

studied in Alexeyev (1999) where it was found that kink may be formed

from suitable initial conditions. The Korteweg-de Vries-Burgers equation

(1.24) also extensively studied but mainly in the direction of generation of

the triangle proﬁles and oscillating wave packets Berezin (1987), see also

references therein. Of special interest is the formation of the kinks with a

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 27

”hat” shown in Fig. 1.5. One possibility will be considered in Sec. 5.3.

Usually periodic waves are generated in ﬁnite domains from a harmonic

input. Thus the KdV cnoidal waves (1.29) are realized numerically and in

experiments in a paper by Bridges (1986), also Kawahara (1983) obtained

numerically periodic wave structures in a system governed by Eq.(1.13)

with

∼

α

5

= 0, while at nonzero coeﬃcient similar results were found in

Rednikov et. al (1995). Note that harmonic input in the ﬁnite domains is

used also for the study of the bell-shaped solitary waves interactions where

no periodic wave structure of permanent shape arises Salupere et. al (1994);

Salupere et. al (1997); Salupere et. al (2001).

1.3 Ampliﬁcation, attenuation and selection of nonlinear

waves

As already noted the bell-shaped solitary wave is sustained by a balance

between nonlinearity and dispersion. What happens with the wave when

dissipation/accumulation destroys this balance? It was shown in Sec. 1.1

that the bell-shaped wave may exist even in presence of dissipation but un-

der strong restrictions on the equation coeﬃcients. Assume the inﬂuence

of dissipation/ accumulation is weak and is characterized by a small pa-

rameter ε << 1. It turns out that an asymptotic solution may be found in

this case whose leading order part is deﬁned as a solitary wave with slowly

varying parameters. Depending on the problem either slow time, T = εt,

or slow coordinate, X = εx, may be used.

In the former case the solitary wave solution is

u(θ, T) = A(T) cosh

−2

(k(T)θ), (1.33)

where θ

x

= 1, θ

t

= −V (T). In the latter case we have

u(θ, X) = A(X) cosh

−2

(k(X)θ), (1.34)

where θ

x

= P(X), θ

t

= −1. Next order solutions give us the functional

form of the dependence of the wave parameters upon the slow variable.

Derivation of the asymptotic solution will be described in Sec. 2.2. Now only

general features of the wave behaviour are considered. When A(T) increases

in time k(T) usually increases also, hence the width of the wave inverse

proportional to k(T), decreases. This is an ampliﬁcation of the solitary

wave. On the contrary, we have an attenuation of the solitary wave when its

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

28 Ampliﬁcation of Nonlinear Strain Waves in Solids

-10 10 20 30 40 50

x

0.5

1

1.5

2

2.5

3

3.5

4

u

b

-10 10 20 30 40 50

x

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

u

a

Fig. 1.17 Temporal evolution of an initial solitary wave resulting in a selection: a) from

below, b) from above.

amplitude decreases while its width increases. Sometimes it happens that

the increase/decrease of A(T) takes place not up to inﬁnity/zero but to

the ﬁnite value A

∗

. Usually this value is deﬁned by the governing equation

coeﬃcients, hence, by the physical parameters of the problem. To put

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Basic concepts 29

-20 -10 10 20

x

0.2

0.4

0.6

0.8

u

Fig. 1.18 Solitary wave (1.33) (dashed line) vs solitary wave (1.34) ( solid line).

this another way, the parameters of the resulted steady wave are selected.

Selection provided by an ampliﬁcation of an initial wave, will be called

selection from below, see Fig. 1.17(a), while selection from above happens

as a result of an attenuation of an initial wave, see Fig. 1.17(b).

Shown in Fig. 1.18 is the proﬁle of the wave (1.34) in comparison with

the symmetric solitary wave solution (1.33) at t = 0. One can see that the

wave (1.34) is asymmetric with respect to its core (or maximum). However,

only initial stages of the temporal evolution of (1.34) diﬀers from that of

Eq.(1.33). As follows from Fig. 1.19, the ﬁnal stage of the selection both

from below and above, is the symmetric bell-shaped solitary wave like shown

in the last stages in Fig. 1.17.

The ampliﬁcation/attenuation of the kink may be described asymp-

totically and numerically Sachdev (1987), it will be shown in Sec. 5.3.

Cnoidal wave evolution may be accounted for an asymptotic solution

similar to that of the bell-shaped solitary waves Rednikov et. al (1995);

Svendsen and Buhr-Hansen (1978).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

30 Ampliﬁcation of Nonlinear Strain Waves in Solids

-20 -10 10 20 30 40 50

x

0.5

1

1.5

2

2.5

3

3.5

u

b

-20 -10 10 20 30 40 50

x

0.5

1

1.5

2

u

a

Fig. 1.19 Selection of the asymmetric monotonic solitary wave: a) from below, b) from

above.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 2

Mathematical tools for the governing

equations analysis

As a rule governing equations for nonlinear strain waves are nonitegrable

by the inverse scattering transform method, and only particular analytical

solutions may be obtained. Hence the study of real physical processes

requires a combined analytico-numerical approach. The aim of this chapter

is to describe methods to be used in this book. The choice of the analytical

and numerical procedures is based on an experience of the author and does

not claim a completeness.

2.1 Exact solutions

2.1.1 Direct methods and elliptic functions

Most of the mathematical work in the realm of nonlinear phenomena

refers to integrable equations and their exact solutions, particularly, pe-

riodic. Among the recently developed general methods the algebroge-

ometrical approach may be used in an eﬃcient way to ﬁnd such solu-

tions. Not only the numerical realization and graphical representation of

the solution is provided by this method but also multi phase quasiperi-

odic solutions as well as purely periodic ones may be represented using

the algebrogeometrical approach as illustrated in Belokolos et. al (1994).

When we are interested in a self-similar solution of a partial diﬀerential

equation one can use well developed theory of the solutions of ordinary

diﬀerential equations, see, e.g., Ince (1964). Exact solutions of nonlin-

ear nonintegrable partial diﬀerential equations are obtained usually us-

ing various direct methods. The signiﬁcant point in direct methods is to

build in advance the appropriate functional form of the solution (ansatz)

of the equation studied. For example, the usage of ansatz in the form

31

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

32 Ampliﬁcation of Nonlinear Strain Waves in Solids

of a hyperbolic tangent (tanh) power series resulted in ﬁnding of new

exact travelling wave solutions (see, e.g., Korpel and Banerjee (1984);

Parkes and Duﬀy (1996) and references therein). The choice of tanh

is caused by the fact that any derivatives of tanh may be expressed as

a polynomial with respect to the tanh itself. Then the equation stud-

ied becomes polynomial of the tanh after substituting the ansatz, and

solution parameters are obtained from the algebraic equations appear-

ing after equating to zero coeﬃcients at each order of tanh . One would

like to apply the same procedure to ﬁnd more general periodic solutions.

First of all another appropriate ”basic” function is required instead of the

tanh . For this purpose various elliptic functions were proposed recently,

the most popular were theta functions (see, e.g.,Chow (1995), Nakamura

(1979)), Jacobian elliptic functions (see, e.g.,Kostov and Uzunov (1992);

Parker and Tsoy (1999)) and the Weierstrass elliptic function Kascheev

(1990); Porubov (1993); Porubov (1996); Porubov and Parker (2002);

Samsonov (1995); Samsonov (2001). In principle, periodical solutions could

be obtained in terms of any of these functions. The eﬃcient choice is caused

by the simplest procedure of the ansatz construction and the least compli-

cated algebra for determining solution parameters. It is well known that

theta functions may be included in the Hirota bilinear method in order to

get N-periodical solutions Nakamura (1979). However most of dissipative

equations cannot be transformed to the bilinear form. At the same time

single travelling wave solution derivation looks very complicated even for

non-dissipative equations Chow (1995), Nakamura (1979). Moreover we

have to deal with four theta functions that results in additional diﬃcul-

ties for the ansatz construction. Explicit periodic travelling wave solutions

may be found for many nonintegrable equations and systems by using an

ansatz in terms of ℘, with appropriate forms for the ansatz suggested by

information about the poles of the solution. When compared to the use

of theta functions or Jacobi elliptic functions, a prime advantage of using

the function ℘ is that the algebra is drastically simpliﬁed. The ansatz for

the solution involves only one Weierstrass function ℘(ζ, g

2

, g

3

), instead of

four theta functions or three Jacobi elliptic functions cn(ζ, k), sn(ζ, k)

and dn(ζ, k). Another advantage is that two apparently distinct solutions

are readily recognized as equivalent. In order to see it let us ﬁrst give

some properties of the Weierstrass elliptic function ℘ to be used below.

According to its deﬁnition Whittaker and Watson (1927), the Weierstrass

function is analytical in the complex plane other than in the points where

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 33

it has double poles. The governing equation for the function ℘ is:

¦℘

(ζ)¦

2

= 4℘

3

−g

2

℘ −g

3

(2.1)

where g

2

and g

3

are constant parameters. Remarkable features of the func-

tion ℘ are that all of its derivatives can be written by means of itself,

and that any elliptic function f may be expressed using ℘ it and its ﬁrst

derivative as Whittaker and Watson (1927)

f = A(℘) +B(℘)℘

, (2.2)

where A and B are rational functions with respect to ℘. Depending on

the ratio between g

2

and g

3

the Weierstrass function may be bounded or

unbounded inside the domain under study. The bounded periodic solutions

are more conveniently expressed by writing them in terms of the Jacobi

elliptic functions cn, sn and dn which are bounded on the real axis. For

this purpose the relation between the Weierstrass function and the Jacobian

functions is used as a special case of (2.2). Indeed, the familiar link is

obtained in Whittaker and Watson (1927) but using the singular function

sn

−2

,

℘(ζ, g

2

, g

3

) = e

3

+ (e

1

−e

3

)sn

−2

(

√

e

1

−e

3

ζ, k). (2.3)

However, following the method introduced in Whittaker and Watson (1927)

one can check that the following formula is valid:

℘(ζ, g

2

, g

3

) = e

2

−(e

2

−e

3

)cn

2

(

√

e

1

−e

3

ζ, k), (2.4)

connecting the Weierstrass function with the Jacobi function cn, regular

along the real axes. Here k =

_

(e

2

−e

3

)/(e

1

−e

3

) is the modulus of the

Jacobian elliptic function, while τ = e

m

( m = 1, 2, 3 , e

3

≤ e

2

≤ e

1

) are

the real roots of the cubic equation

4τ

3

−g

2

τ −g

3

= 0. (2.5)

Expressing these results in terms of an appropriate choice of parameters,

the wave number κ =

√

e

1

−e

3

and the Jacobian elliptic modulus k, we

have

e

3

= −

1 +k

2

3

κ

2

, e

2

=

2k

2

−1

3

κ

2

, e

1

=

2 −k

2

3

κ

2

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

34 Ampliﬁcation of Nonlinear Strain Waves in Solids

g

2

=

8

3

κ

4

(1 −k

2

+k

4

), g

3

=

4

27

κ

6

(k

2

+ 1)(2 −k

2

)(1 −2k

2

). (2.6)

The localized both the bells-shaped and the kink-shaped solitary wave so-

lutions appear in the limit k →1 of the Jacobi elliptic functions.

Now consider an instructive example Porubov and Parker (2002). In

Parker and Tsoy (1999), solutions were sought in terms of powers and

products of Jacobi functions and thereby two solutions were obtained

z

1

= e

2

−(e

2

−e

3

) cn

2

(

√

e

1

−e

3

ζ, k), (2.7)

z

2

=

ˆ r

2

m

2

2

sn

2

(ˆ rζ, m) ±

ˆ r

2

m

2

cn(ˆ rζ, m)dn(ˆ rζ, m) −

ˆ r

2

(1 +m

2

)

12

, (2.8)

which appear diﬀerent. Obviously, the solution (2.7) is a representation of

the Weierstrass function (2.3). However, one can check by direct substitu-

tion that the solution (2.8) also satisﬁes equation (2.1) when the parameters

ˆ r and m are deﬁned as solutions to

ˆ r

4

12

(1 + 14m

2

+m

4

) = g

2

,

ˆ r

6

216

(1 −33m

2

−33m

4

+m

6

) = g

3

. (2.9)

Therefore z

2

is also a solution satisfying the same governing equation

(2.1) deﬁning the Weierstrass function. The two expressions (2.7) and (2.8)

are essentially equivalent solutions, provided that (ˆ r, m) are appropriately

related to (r, k). Accordingly, working with ℘ reveals links between seem-

ingly distinct forms of solution.

It is to be noted that the ﬁrst Weierstrass function derivative ℘

can-

not be expressed as a polynomial of the Weierstrass function itself Whit-

taker and Watson (1927), and we have to equate zero separately coeﬃ-

cients at each order of ℘ and at products of ℘

**and corresponding orders
**

of ℘ Kascheev (1990); Porubov (1993); Porubov (1996); Samsonov (1995);

Samsonov (2001). Therefore we really deal with two functions, and it is

unlikely to get the solution using the ansatz proposed in the form of power

series with respect both of the ℘ and ℘

**as it was done for the tanh. An-
**

other idea may be used, based on the singular point analysis of the possible

solution and the well known fact that the Weierstrass function ℘ has the

second order poleWhittaker and Watson (1927). In order to check the poles

of a solution we shall use the WTK method Weiss et. al (1983) looking for

the solution in the form of Laur´ent- type series about the singular manifold.

It will be explained in details in Sec. 2.1.2.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 35

The exact solutions obtained in this manner, usually belong to the class

of travelling wave solutions which require special initial conditions. In case

of the solitary wave solution the initial condition should be have the shape

of the solitary wave itself. Moreover, travelling wave solutions for the dis-

sipative equations usually have not free parameters, and additional rela-

tionships on the equation coeﬃcients are required for the existence of the

solutions Porubov (1993); Porubov (1996); Porubov and Velarde (1999);

Porubov and Parker (1999); Porubov and Parker (2002); Samsonov (1995);

Samsonov (2001).

2.1.2 Painlev´e analysis

Recently it was developed the theory of nonlinear ordinary diﬀerential

equations whose solutions have not movable singularities, other than poles.

Then the theory has been extended to partial diﬀerential equations. Usually

these equations are called ”equations with the Painlev´e property”. The

achievements of the theory may be found in Cariello and Tabor (1989);

Conte (1989); Conte et. al (1993); Levi and Winternitz (1992); Newell et. al

(1987); Weiss et. al (1983). Here we concentrate on the one aspect of the

theory-the singular manifold method or WTC method for partial diﬀeren-

tial equations Newell et. al (1987); Weiss et. al (1983). Let ϕ(x, t) = 0 is

the ”singular” or ”pole” manifold on which a solution u(x, t) is singular.

The main idea of the WTC is to demonstrate that the expansion

u(x, t) =

1

ϕ

α

∞

j=0

q

j

(x, t)ϕ

j

(2.10)

is single valued. This requires (i) α is an integer, (ii) ϕ is analytic in x

and t and (iii) the equations for the coeﬃcients q

j

have self-consistent

solutions. When all these conditions are satisﬁed the equation under study

has the Laur´ent property. Also it is necessary to assume that neither ϕ

x

nor ϕ

t

vanish on ϕ(x, t) = 0. Let us illustrate how the methods works

on an example of the KdV equation (1.1). Following Newell et. al (1987);

Weiss et. al (1983) we assume b = 3, d = 1. Substituting the ansatz (2.10)

into Eq.(1.1) one can ﬁnd α = 2 and q

0

= −2ϕ

2

x

. Recursion relations for

the q

j

are

(j + 1)(j −4)(j −6)q

j

= F(ϕ

t

, ϕ

x

, ..., q

k

; k < j). (2.11)

The values of j, j = −1, 4 and 6 are called resonances. At each such

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

36 Ampliﬁcation of Nonlinear Strain Waves in Solids

resonance the right-hand side of Eq.(2.11) vanishes thus ensuring the inde-

terminancy of the corresponding q

j

. Moreover, the expansion (2.10) may

be truncated at O(ϕ

0

). As a result we obtain using Eq.(2.11) an auto-

B¨acklund transformation for the solution of Eq.(1.1),

u(x, t) = 2

∂

∂x

2

log ϕ +q

2

,

where q

2

satisﬁes the KdV equation. Also the Lax pair for the KdV equation

follows from the solution of Eq.(2.11) Newell et. al (1987); Weiss et. al

(1983).

However, the Lax pair cannot be obtained for nonitegrable equations as

opposed to a truncated expansion that carries an information about possible

pole orders of a solution. Using this information the anzats for the solution

may be proposed, in particular, in terms of the Weierstrass elliptic function

℘ and ℘

**. Substituting the proposed form of the solution into the equation
**

under study and equating to zero coeﬃcients at each order of ℘ and at prod-

ucts of ℘

**and corresponding orders of ℘ one can get the algebraic equations
**

on the solution parameters, the phase velocity and the Weierstrass function

parameters g

2

, g

3

. Certainly this procedure is of phenomenological kind

but it allows to obtain the solutions of nonitegrable nonlinear equations in

an explicit form. Some examples are presented below.

2.1.3 Single travelling wave solutions

First we consider exact solutions of DMKdV Eq.(1.13) obtained in Porubov

(1993). It was found there the following auto-B¨acklund transformation for

the solution u(x, t)

u =

12

∼

α

4

∼

α

5

(log ϕ)

xx

+

12

5

∼

α

5

(α

3

−

2α

1

∼

α

4

∼

α

5

)(log ϕ)

x

+

∼

u, (2.12)

where

∼

u (x, t) satisﬁes Eq.(1.13). Concerning only tavelling wave solutions,

one can reduce Eq.(1.13) to the third-order ODE of the form:

∼

α

4

u

+α

3

u

+

∼

α

2

u

+

∼

α

5

uu

+α

1

u

2

−V u +P = 0, (2.13)

where P = const, prime denotes diﬀerentiation with respect to θ, θ = x−

V t. Based on Eq.(2.12) possible solution may contain simple and second-

order poles that may be modelled in terms of ℘ as Porubov (1993):

u = A℘ +

B℘

℘ +C

+D. (2.14)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 37

Substituting Eq.(2.14) into Eq.(2.13) one can derive a system of alge-

braic equations in A, B, C, D, phase velocity V and Weierstrass function

parameters g

2

, g

3

:

(g

2

C −g

3

−4C

3

)B = 0, (12C

2

−g

2

)B = 0,

P = V D + 8α

1

B

2

C −α

1

D

2

−α

3

g

2

A/2 −2

∼

α

2

BC + 12

∼

α

4

BC

2

−

∼

α

5

(2ABC

2

+g

2

AB/2 −2BCD),

2α

1

(2B

2

+AD) −V A+ 2

∼

α

2

B + 2

∼

α

5

(BD −ABC) = 0,

α

1

A

2

+ 6α

3

A+ 12

∼

α

4

B + 6

∼

α

5

AB = 0,

B(2α

1

(D −AC) −V ) = 0, A(12

∼

α

4

+

∼

α

5

A) = 0,

2α

1

AB + 2α

3

B+

∼

α

2

A+

∼

α

5

(AD + 2

∼

α

5

B

2

) = 0.

The solutions of these equations are:

(i) when g

2

, g

3

are free parameters and α

3

= 2α

1

∼

α

4

/

∼

α

5

;

A = −

12

∼

α

4

∼

α

5

, B = 0, D = −

∼

α

2

∼

α

5

, V = −

2α

1

∼

α

2

∼

α

5

.

(ii) when either

C = −

1

300

∼

α

4

2

(α

3

−

2α

1

∼

α

4

∼

α

5

)

2

,

or α

3

= 12α

1

∼

α

4

/

∼

α

5

, C is a free parameter

A = −

12

∼

α

4

∼

α

5

, B = −

6

5

∼

α

5

(α

3

−

2α

1

∼

α

4

∼

α

5

),

D = −

∼

α

2

∼

α

5

+

2α

1

∼

α

5

2

(α

3

−

2α

1

∼

α

4

∼

α

5

) +

1

25

∼

α

4

∼

α

5

(α

3

−

2α

1

∼

α

4

∼

α

5

)

2

,

V =

24α

1

∼

α

4

∼

α

5

C −

2α

1

∼

α

2

∼

α

5

+

4α

2

1

∼

α

5

2

(α

3

−

2α

1

∼

α

4

∼

α

5

) +

2α

1

25

∼

α

4

∼

α

5

(α

3

−

2α

1

∼

α

4

∼

α

5

)

2

, g

2

= 12C

2

, g

3

= 8C

3

.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

38 Ampliﬁcation of Nonlinear Strain Waves in Solids

Using (2.4) the solution (2.14) with parameters deﬁned by (i) may de-

scribe a particular bounded cnoidal wave, propagating with ﬁxed phase

velocity, of the form:

u =

12

∼

α

4

∼

α

5

k

2

κ

2

cn

2

(kθ, κ) −

∼

α

2

∼

α

5

−

4

∼

α

4

∼

α

5

k

2

(2κ

2

−1) (2.15)

where k is a free parameter. When the Jacobian elliptic functions modulus

κ → 1 the solution (2.15) transforms to the solitary wave solution (1.12).

The solution (2.14) with parameters deﬁned by (ii) accounts for a bounded

kink-shaped solution (1.28). When C is a free parameter, kink propagates

with any phase velocity value.

Besides bounded solutions (2.15) and (1.28), the solution (2.14) also

describes unbounded ones in the form of localized and periodic discontinu-

ities. Finally it is to be noted that the functional form (2.14) in terms of

the Weierstrass function is not unique. One can see that there exist at least

one more solution of the form:

u = −

12

25

∼

α

4

∼

α

5

(α

3

−

2α

1

∼

α

4

∼

α

5

)

2

exp(2y)℘(exp(y) +G, 0, g

3

),

where

y = exp(γθ), γ = −

1

5

∼

α

4

(α

3

−

2α

1

∼

α

4

∼

α

5

),

G and g

3

are free parameters. This solution allows to describe only the

bounded kink-shaped wave (1.28). At the same time it accounts for a new

periodically discontinious solution.

In previous Chapter the bounded periodic solution (1.32) of the equa-

tion (1.31) was considered. When studying travelling wave solutions, i.e.

solutions depending only on the phase variable θ = x −V t, this equation

may be transformed into the O.D.E., which results in the following equation

after integrating once with respect to θ:

dη

+gη

−V η

+b(η

2

)

+f(η

2

)

+c(η

3

)

= N. (2.16)

Further we follow the results obtained in Porubov (1996). The transforma-

tion is obtained here using the WTK method of the form

u =

f ±

_

f

2

−2cd

c

(log ϕ)

+ ¯ u. (2.17)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 39

Therefore possible solution should contain ﬁrst order pole. The Weierstrass

elliptic function ℘ possesses second order pole, and we shall propose three

solution forms. The ﬁrst of them is

u =

A℘

℘ +C

+B. (2.18)

In order to ﬁnd the solution parameters the formula (2.18) is substituted

into the Eq.(2.16). Then equating to zero coeﬃcients at each order of ℘+C

and ℘

**one can derive the algebraic equations on A, B, C, phase velocity V
**

and the Weierstrass function parameters g

2

, g

3

:

(℘ +C)

−4

: (g

2

C −g

3

−4C

3

)

2

(6Af −3A

2

c −6d) = 0,

(℘ +C)

−3

: (g

2

C −g

3

−4C

3

)(12C

2

−g

2

)(14Af −9A

2

c −12d) = 0,

(℘ +C)

−2

: 4(g

2

C −g

3

−4C

3

)(2bB −V + 48ACf + 3c(12A

2

C −

B

2

) −48Cd) −3(12C

2

−g

2

)

2

(2f −2Ac −d) = 0,

(℘ +C)

−1

: 4(g

2

C −g

3

−4C

3

)(10f −4f + 3Ac + 6d) −(12C

2

−

g

2

)(2Bb −V + 24ACf + 3(B

2

−12A

2

C)c −12Ad) = 0,

℘

0

: 4(12C

2

−g

2

)qA

2

= N,

(℘ +C) : 2Bb −V + 24ACf + 6(B

2

−12A

2

C)c −12Cd = 0,

(℘ +C)

2

: 2Af +d + 2A

2

c = 0,

℘

(℘ +C)

−3

: (g

2

C −g

3

−4C

3

)(g + 2Bf −Ab −3ABc) = 0,

℘

(℘ +C)

−2

: (12C

2

−g

2

)(2Ab −g −2Bf + 6ABc) = 0,

℘

**: 2Ab +g + 2Bf + 6ABc = 0.
**

Three solutions of algebraic equations are obtained. The ﬁrst appears

when

12C

2

−g

2

= 0, g

2

C −g

3

−4C

3

= 0.

In this case two of three roots e

i

of Eq.(2.5) are equal to one another, and

no periodical solution exists. Then the solution (2.18) will have the form

of localized discontinuity under positive C values. When the parameter C

is negative we get κ = 1, and the bounded kink-shaped solitary wave

solution follows from Eq.(2.18):

u = γ tanh(mθ) +u

0

. (2.19)

For the wave amplitude γ two formulaes are valid

γ = A

1

m, A

1

= (f + (f −2cd)

1/2

)/(6c),

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

40 Ampliﬁcation of Nonlinear Strain Waves in Solids

γ = A

2

m, A

2

= (f −(f −2cd)

1/2

)/(6c).

Then for u

0

we have

u

0

=

2bA

j

−g

2(3cA

j

+f)

, j = 1, 2,

and phase velocities V

1

, V

2

are

V

1

= 2u

01

b + 6c(u

2

01

+ 4A

2

1

m

2

) −4m

2

(2fA

1

−d),

V

2

= 2u

02

b + 6c(u

2

02

+ 4A

2

2

m

2

) −4m

2

(2fA

2

−d),

while m

2

= −3C = 3e

1

is a free parameter.

The second solution corresponds to the situation when

g

2

C −g

3

−4C

3

= 0, 12C

2

−g

2

,= 0. (2.20)

In this case the solution may exist under additional conditions on the equa-

tion coeﬃcients f = g = 0, that results in absence of the dissipative terms

in Eq.(2.16). Hence it becomes now the O.D.E. reduction of the Gardner

equation. The bounded cnoidal wave solution arises when C = −e

1

and

has the form

u = ±

_

−

2d

c

κ

2

m

cn(mθ, κ)sn(mθ, κ)

dn(mθ, κ)

−

b

3c

, (2.21)

where m

2

= e

1

− e

3

. It governs the travelling cnoidal wave, propagating

with the phase velocity V = −b

2

/(3c) − 6e

1

d. that transforms into the

kink-shaped soluiton (2.19) when κ = 1.

Finally, the third solution arises when

12C

2

−g

2

= 0, g

2

C −g

3

−4C

3

,= 0.

In this case the bounded cnoidal wave solution (1.32) holds. When κ = 1

we have C

∗

= 0, and it transforms into the kink-shaped solution (2.19), see

Fig. 1.6.

Now we shall consider the second possible solution’s form in terms of

the Weierstrass elliptic function ℘:

η =

_

A℘ +B. (2.22)

Substitution Eq.(2.22) into the Eq.(2.16) allows us to conclude that solution

may now exist only when f = g = 0. Then one can get the algebraic

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 41

equations for the solution parameters equating to zero coeﬃcients at the

terms ℘

℘

k

, k = 0 ÷3, and (A℘ +B)

5/2

:

℘

℘

3

: cA+ 2d = 0

℘

℘

2

: 9cAB −V A+ +15dB = 0

℘

**℘ : 9cAB −2V A+ +12dB = 0
**

℘

: 12cB

3

−4V B

2

+dA(Bg

2

−Ag

3

) = 0

(A℘ +B)

5/2

: N = 0

One can obtain the following solution of these equations:

A = −

2d

c

, B =

2de

i

c

, i = 1 ÷3, V = −

1

2

cB,

where e

i

are the real roots of Eq. (2.5). We again deal with the O.D.E.

reduction of the Gardner equation, however now one can obtain its another

bounded solutions. Thus for d/fc > 0 we ﬁnd cnoidal wave solutions of

the form:

(I) for B = 2de

2

/c

u =

_

2d

c

m κ cn(mθ, κ), (2.23)

where m

2

= e

1

−e

3

and for the phase velocity we have V = dm

2

(2κ

2

−1).

(II) for B = 2de

1

/c we obtain

u =

_

2d

c

m κ dn(mθ, κ), (2.24)

Solution (2.24) represents cnoidal wave propagating with the phase velocity

V = dm

2

(2 − κ

2

). It is to be noted that at κ = 1 both the solutions

(2.23) and (2.24) transform to the bell-shaped solitary wave

u =

_

2d

c

m ch

−1

(mθ, κ),

propagating with the velocity V = dm

2

.

When d/c > 0 the bounded solution appears from (2.22) if B = 2de

3

/c

and has the form

u =

_

2d

c

m κ sn(mθ, κ), (2.25)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

42 Ampliﬁcation of Nonlinear Strain Waves in Solids

For the phase velocity we ﬁnd V = −dm

2

(1+κ

2

). In contrast to solutions

(2.23), (2.24) the solution (2.25) transforms to the kink-shaped solution

(2.19) when κ = 1.

Finally one can construct the third possible solution in terms of the

Weierstrass elliptic function ℘ depending now on new independent variable

y = exp(kθ). Changing variable in auto-B¨acklund transformation (2.17)

one can see that the following solution may be proposed:

u = A

y℘

y

℘ +C

+B, (2.26)

Substituting it into Eq.(2.16) one can ﬁnd that solution may exist under

f = 0. Therefore we now deal with the O.D.E. reduction of the Gardner

equation with linear dissipation. The solution of the form (2.26) for the

last equation was already obtained previously in Samsonov (1995). It was

found that this solution may correspond either to discontinuous periodical

solution or to the bounded kink-shaped solution of the form similar to the

solution (2.19).

2.1.4 Exact solutions of more complicated form

The procedure of the obtaining exact solutions is based on the reduction

to the ODE, hence only travelling wave solutions may be found. However,

sometimes this procedure is applied to obtain only the part of a solution,

and more complicated proﬁles are decsribed. Consider some examples.

Recently the coupled nonlinear Schr¨odinger equations (CNLS)

iW

t

+sW

xx

+ (ηWW

∗

+σUU

∗

)W = 0,

iρU

t

+rU

xx

+ (η

−1

UU

∗

+σWW

∗

)U = 0, (2.27)

with r, s = ±1; ρ, σ, η real, have attracted considerable interest because

of their role in governing various physical wave-guiding systems. These

equations are not generally integrable by the inverse scattering transform

method, except in cases of high symmetry (e.g. ρ = 1, η = σ = ±1). The

machinery of the Weierstrass function is applied in the search for travelling

wave solutions of equations (2.27) in the form

W = w(ζ) e

iζ

, U = u(ζ) e

iφ

, (2.28)

where θ = x −c t, ζ = ζ(θ, t), φ = φ(θ, t), with u, w, ζ and φ real. Substitu-

tion (2.28) into equations (2.27) allows to separate the real and imaginary

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 43

parts. The solutions for u and w are sought in the form

w =

_

F℘

2

+A℘ +B, u =

_

G℘

2

+D℘ +E, (2.29)

while for ζ and φ we have

ζ =

1

2

s c θ −s C

1

_

dθ

w

2

+ζ

0

(t) ,

φ =

1

2

r c ρ θ −r C

2

_

dθ

u

2

+φ

0

(t) , (2.30)

where C

i

(i = 1, 2) are free parameters. The most notable diﬀerence from

previous solutions is the dependence of the “frequency” and “phase” on w

and u, when C

i

,= 0. This introduces signiﬁcant features into the proﬁles

of the real and imaginary parts of W and of U. Thus, we ﬁnd for W of the

form:

W = w exp ı Y, (2.31)

where w is deﬁned by Porubov and Parker (1999)

w =

¸

A

_

2k

2

−1

3

κ

2

−k

2

κ

2

cn

2

(κθ, k) +B/A

_

, (2.32)

while for the phase function Y we have

Y =

csθ

2

−

sC

1

Π[ϕ, n, k]

κA(H

1

−e

3

)

+ (γ +

c

2

s

4

)t, (2.33)

Π[ϕ, n, k] is the elliptic integral of the third kind, ϕ = arcsin(sn(κθ)), n =

−κ

2

k

2

/(B/A+e

3

) and C

1

is deﬁned by

A =

2(σ r η −s)

η(1 −σ

2

)

, C

2

1

=

A

2

4

_

4

_

B

A

_

3

−g

2

B

A

+g

3

_

, (2.34)

γ =

σ(E A−BD) −3s B

A

, D =

2(s σ −r η)

1 −σ

2

,

where B, E, g

2

and g

3

are free parameters. The shape of the wave (2.31)

depends strongly on the values of the parameters κ, k and H

1

. An example

of the behavior of Re W is shown in Fig.2.1(a) for the case H

1

∈ [e

2

, e

3

], c ∼

κ and at t = 0. We get an interesting wave, consisting of the carrier wave

slightly modulated and with superposed periodic disturbances. Therefore,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

44 Ampliﬁcation of Nonlinear Strain Waves in Solids

20 40 60 80 100

x

-2

-1

1

2

Re W

b

20 40 60 80 100

x

-2

-1

1

2

Re W

a

Fig. 2.1 Exact solutions of the CNLS. a) proﬁle with moving disturbances; b) almost

envelope wave solution.

the wave shape is not determined solely by the amplitude wave shape.

When c considerably exceeds the wave number κ we get the proﬁle closer

to the usual envelope wave solution, see Fig. 2.1(b).

Another interesting proﬁles correspond to the exact solutions of the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 45

complex Ginzburg-Landau equation, CGLE,

ı u

t

+p u

xx

+q [u[

2

u = ı γ u (2.35)

where the constant coeﬃcients are p = p

r

+ı p

i

, q = q

r

+ı q

i

, with p

j

, q

j

,= 0,

(u, p, q) ∈ C, γ ∈ R. The subscripts t and x denote temporal and spatial

derivatives, respectively. This equation appears in the description of a large

variety of physical phenomena, e.g., in nonlinear optics, non-equilibrium

pattern formation, lasers, superconductivity etc. Like in case of CNLS

equations, ﬁrst of all, we decompose the solution u(x, t) in its amplitude,

y, and phase, θ, both real,

u = y(ζ) e

ıθ

, (2.36)

where ζ = x −c t, θ = θ(ζ, t). Substituting (2.36) into (2.35) and equating

to zero the real and imaginary parts one obtains two coupled equations for

the functions y and z ≡ θ

ζ

. Then periodic and pulse solutions of the CGLE

may be found Porubov and Velarde (1999). The ﬁrst derivative of the real

part of the periodic solution with respect to x, v = (Re u)

x

, is

v =

Z

y

sin

_

θ + arcsin(

y y

x

Z

)

_

, (2.37)

with

y =

_

2q

r

p

r

_

k

2

1

dn

2

(k

1

ζ, κ) −δ

2

1

_

, Z =

_

y

2

y

2

x

+C

2

. (2.38)

When C = 0, the zeroes of the ﬁrst derivative (2.37) are deﬁned by

the zeroes of the function y

x

and correspond to the zeroes of the Jacobi

functions cn and sn. Their positions do not change in time, and for u we

have harmonic temporal oscillations of the spatially periodic state deﬁned

by the amplitude parts of the solutions (2.36). However, the situation

changes dramatically when C ,= 0. In this case Z never vanishes, and the

zeroes of the ﬁrst derivative are deﬁned by the zeroes of the sin function

only, whose position vary in time. A typical situation is shown in Fig. 2.2.

Fig. 2.2(a) shows a structure with four spatially repeated parts. During half

of the time period the shapes of these parts vary, and we get in Fig. 2.2(d)

a proﬁle which is practically the mirror image to Fig. 2.2(a). Qualitatively

this evolution does not depend upon the value of the modulus κ of Jacobi

functions.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

46 Ampliﬁcation of Nonlinear Strain Waves in Solids

20 40 60 80 100 120

x

-2

-1.5

-1

-0.5

0.5

1

1.5

2

Re u

d

20 40 60 80 100 120

x

-2

-1.5

-1

-0.5

0.5

1

1.5

2

Re u

c

20 40 60 80 100 120

x

-2

-1.5

-1

-0.5

0.5

1

1.5

2

Re u

b

20 40 60 80 100 120

x

-2

-1.5

-1

-0.5

0.5

1

1.5

2

Re u

a

Fig. 2.2 Evolution of the periodic solution (2.36) Re u vs x for times t = mp

i

π/(3prγ),

0 < m < 3 with k = 0.9, δ

1

= 0.5. a)m = 0, b)m = 1, c)m = 2, d)m = 3.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 47

Consider now the pulse solution Porubov and Velarde (1999),

u =

_

6Al

1

l

2

k cosh

−1

(kζ) exp ıθ, (2.39)

with

κ

2

= −

γ

2(p

i

+ 4Ap

r

−4A

2

p

i

)

, l

1

= p

2

r

+p

2

i

, l

2

= p

r

q

i

−p

i

q

r

.

The ﬁrst derivative for the real part of (2.39) is

(Re u)

x

=

¸

6Al

1

(1 + 4A

2

)

l

2

k

2

cosh

−1

(kζ) tanh(kζ) sin

_

θ −arctan(

1

2A

)

_

.

(2.40)

Thus, from (2.40) it follows that additional zeroes of the ﬁrst derivative

may appear if

k >

_

l

2

6Al

1

exp

_

arctan(1/(2A))

2 A

_

. (2.41)

The evolution of the real part of the solution (2.39) is illustrated in Fig. 2.3.

Again we see that two initial maxima in Fig. 2.3(a) disappear, Fig. 2.3(e),

then an initial minimum at ζ = 0 is changed into a maximum, while two

minima arise, Fig. 2.3(f-h). Therefore, our solution is breather- like. If

(2.41) is not satisﬁed, there is a pulse solution whose spatial behavior is

determined by the function cosh

−1

(kζ) only with one extremum at ζ = 0.

2.2 Asymptotic solutions

Particular exact solutions are insuﬃcient for understanding physical pro-

cesses. One can see that many equations consist in generalisations of the

integrable equations like the KdV equation. Sometimes additional terms

may be considered as small perturbations. Often straight asymptotic ex-

pansions are incorrect Cole (1968); Jeﬀrey and Kawahara (1982); Nayfeh

(1973), and the matching asymptotic procedure Ablowitz and Segur (1981);

Kodama and Ablowitz (1981) may be applied to ﬁnd a solution in this

case. The basic idea of the perturbation approach is to look for a solu-

tion of a perturbed nonlinear equation in terms of certain natural fast and

slow variables. Often there is a need in only one fast variable, such as

phase variable θ in the unperturbed problem. Depending upon the prob-

lem either slow time, T = ε t, or slow coordinate, X = ε x, ε << 1 are

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

48 Ampliﬁcation of Nonlinear Strain Waves in Solids

-0.04 0.04

x

-4

-2

2

4

Re u

g

-0.04 0.04

x

-4

-2

2

4

Re u

h

-0.04 0.04

x

-4

-2

2

4

Re u

e

-0.04 0.04

x

-4

-2

2

4

Re u

f

-0.04 0.04

x

-4

-2

2

4

Re u

c

-0.04 0.04

x

-4

-2

2

4

Re u

d

-0.04 0.04

x

-4

-2

2

4

Re u

a

-0.04 0.04

x

-4

-2

2

4

Re u

b

Fig. 2.3 Evolution of the pulse solution (2.39) Re u vs x for times t = mπ/β, 0 < m < 8.

a)m = 0, b)m = 1, c)m = 2, d)m = 4, e)m = 5, f)m = 6, g)m = 7, h)m = 8.

introduced. Fast variable is generalized in a perturbed problem assuming

either θ

x

= 1, θ

t

= −V (T) or θ

x

= P(X), θ

t

= −1. The functions V (T)

or P(X) are deﬁned to remove secular terms. Some of the appropriate sec-

ularity conditions are formed from Green’s identity as follows. Assume a

solution u = u(θ, T) is of the form

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 49

u(θ, T) = u

0

(θ, T) +ε u

1

(θ, T) +... (2.42)

The leading order problem is nonlinear equation for u

0

whose solution

is known. Then in the next order an inhomogeneous linear equation for u

1

holds

L(u

1

) = F(u

0

).

Here L(u

1

) is a linearized leading-order equation operator. Denoting by

v

i

(i = 1, ..., M) the M solutions of the homogeneous adjoint problem,

L

A

(v

i

) = 0, i = 1, ..., M,

where L

A

is the adjoint operator to L, we obtain that

L(u

1

)v

i

−L

A

(v

i

)u

1

= Fv

i

is always a divergence. Using the boundary conditions it may be integrated

to give the secularity conditions. In particular, when v

i

→ 0 as [θ[ → ∞

while u

1

is bounded the secularity condition is:

_

∞

−∞

v

i

F dθ = 0. (2.43)

Then either V (T) or P(X) are obtained from Eq.(2.43). Method may

be used for ﬁnding perturbed cnoidal wave solutions. In this case in-

tegration in (2.43) is carrying over the period Rednikov et. al (1995);

Svendsen and Buhr-Hansen (1978). It is to be noted that a simple qua-

sistationary expansion is not uniformly valid on [x[ → ∞. Complete

solutions is obtained using matching quasistationary solution to a non-

stationary one for large [θ[, e.g. [θ[ ∼ 1/ε Ablowitz and Segur (1981);

Kodama and Ablowitz (1981).

Let us apply this method for a perturbed solitary wave solution of the

DMKdV equation (1.13). Following Kliakhandler et. al (2000); Rednikov

et. al (1995) assume that

∼

α

i

= ε α

i

, ε << 1, and suppose that u depends

upon a fast variable θ and a slow time T, such as

θ

x

= 1, θ

t

= −V (T), T = ε t.

Then equation (1.13) becomes

α

3

u

θθθ

−V u

θ

+ 2α

1

uu

θ

+ε [u

T

+α

2

u

θθ

+α

4

u

θθθθ

+α

5

(uu

θ

)

θ

] = 0.

(2.44)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

50 Ampliﬁcation of Nonlinear Strain Waves in Solids

The solution of Eq.(2.44) is sought in the form (2.42). In the leading

order we get

α

3

u

0,θθθ

−V u

0,θ

+ 2α

1

u

0

u

0,θ

= 0. (2.45)

We are interested in studying localized solutions vanishing together with

its derivatives at [θ[ →∞. Then the leading order solution is

u

0

=

6 α

3

α

1

b(T)

2

cosh

−2

(b(T) θ) (2.46)

with V = 4α

3

b

2

. At order ε, we have

α

3

u

1,θθθ

−V u

1,θ

+ 2α

1

(u

0

u

1

)

θ

= F, (2.47)

F = −

_

u

0,T

+α

2

u

0,θθ

+α

4

u

0,θθθθ

+α

5

(u

0

u

0,θ

)

θ

¸

.

The operator acting on the function u

1

in the lhs of Eq.(2.47) is adjoint to

that of the Eq.(2.45). Then the secularity condition (2.43) with v

i

= u

0

yields the equation for b:

b

T

=

8

15

b

3

_

A+Bb

2

_

, (2.48)

with

A = α

2

, B =

4

7

_

6α

3

α

5

α

1

−5α

4

_

. (2.49)

The behavior of b (or the sign of b

T

) depends on the signs of A and

B and on the value of b

0

≡ b(T = 0). Indeed, when both A and B

are positive b diverges while for both negative values it will vanish. For

A < 0, B > 0 the parameter b vanishes if b

0

<

_

−A/B while it diverges

if b

0

>

_

−A/B. The most interesting case occurs when A > 0, B < 0.

Here b tends to

_

−A/B independent of b

0

. A quantitative description of

the variation of b can be given. Equation (2.48) may be directly integrated

giving the implicit dependence of b on T:

T =

4 B

15A

2

ln

¸

¸

¸

¸

b

2

0

(A+Bb

2

)

b

2

(A+Bb

2

0

)

¸

¸

¸

¸

−

4(b

2

0

−b

2

)

15Ab

2

b

2

0

. (2.50)

We see that b tends to inﬁnity in ﬁnite time, T = T

∗

for A > 0, B > 0,

T

∗

=

4 B

15A

2

ln

b

2

0

B

(A+Bb

2

0

)

+

4

15A b

2

0

, (2.51)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 51

A similar scenario occurs when A < 0, B > 0, b

0

<

_

−A/B. Hence,

with our asymptotic approximation we can predict when blow-up could

occur. On the other hand, for both A < 0, B < 0 and A < 0, B > 0,

the quantity b

0

>

_

−A/B, b vanishes when T tends to inﬁnity. Finally,

when A > 0, B < 0, the quantity b approaches

_

−A/B when T tends

to inﬁnity, and expression (2.50) provides an analytical description of the

time-dependent process of the parameter-value selection of the solitary wave

(2.46), see Fig. 1.17. In the last case, if we additionally assume α

3

=

2α

1

α

4

/α

5

, the asymptotic ”dissipative” solitary wave (2.46) will tend to

the exact travelling solitary wave solution (1.12).

Function u

1

(θ, T) is a solution of a linear inhomogeneous O.D.E., hence,

it will contain free parameters depending on T like b(T) in the leading or-

der problem. Its deﬁnition allows to satisfy solvability condition in the

next order problem and to avoid secular terms in the asymptotic expan-

sion. Higher order approximations may be studied similarly. The ﬁrst-order

solution u

1

has been obtained in Rednikov et. al (1995) of the form:

u

1

= A

1

u

0,θ

+A

2

θ cosh

−2

(b(T) θ) +A

3

u

0,θ

log(cosh (b(T) θ))+

A

4

(1 −tanh(b(T) θ)),

where A

i

may be found in Rednikov et. al (1995). One can see it predicts

a plateau behind the solitary wave and does not vanish at minus inﬁnity.

Uniformly valid solution is found using the matching asymptotic procedure

may be found in Rednikov et. al (1995). Perturbations of cnoidal wave

solutions were studied, e.g., in Rednikov et. al (1995); Svendsen and Buhr-

Hansen (1978). In this case the modulus of the Jacobian elliptic functions

should be a function of a slow variable, corresponding expressions for the

derivatives of the elliptic functions with respect to modulus may be found

in Byrd and Friedman (1954); Newille (1951).

2.3 Numerical methods

One can see that both exact nor asymptotic solutions have severe limi-

tations. In particular they satisfy speciﬁc initial conditions and do not

allow to account for an arbitrary initial disturbance evolution. Therefore

numerical simulations should be used. Two kinds of nonlinear equations,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

52 Ampliﬁcation of Nonlinear Strain Waves in Solids

considered here, may be written either

u

t

= L(u) (2.52)

or

u

tt

= L(u) (2.53)

where L(u) is a certain nonlinear diﬀerential operator. Equations of the

kind (2.52) will be called nonlinear evolution equations while nonlinear

hyperbolic equations correspond to the class (2.53). Further only nu-

merical methods are described which are used by the author and his co-

workers in solving the problems considered in the book. Two main schemes

are used, pseudo-spectral approach and ﬁnite-diﬀerence approach. More

detailed information about numerical modelling of nonlinear wave equa-

tions may be found, in particular, in Berezin (1987); Dodd et. al (1982);

Fletcher (1984); Mayer (1995); Sachdev (1987); Zwillenger (1989), see also

references therein.

2.3.1 Nonlinear evolution equations

Among the equations of the kind (2.52) main attention has been paid to

numerical solutions of the KdV equation, see Berezin (1987); Dodd et. al

(1982) where various diﬀerence schemes are discussed. As noted in Berezin

(1987) implicit three-levels schemes are rather simple and suitable for a

realization. However, some of the straight ﬁnite-diﬀerence methods re-

quire rather small time step for stability, while long-time evolution should

be studied. It becomes the smaller the higher is the order of a highest

derivative term in NEE. In particular, it takes place for a simulation of the

generalized ﬁfth-order KdV equation (1.3), another example has been stud-

ied in Christov et. al (1997). The problem of an increase of the time step

may be solved either by a modiﬁcation of a diﬀerence scheme or by using

a more eﬀective solving procedure, in particular, the fourth-order Runge-

Kutta ﬁnite-diﬀerence method. Besides simulation of Eq.(1.3) in Porubov

et. al (2002), this method was eﬀectively applied to study the evolution

and interaction of Marangoni-B´enard solitary waves governed by Eq.(1.13)

in Marchant (1996). The same equation has been numerically solved in

Christov and Velarde (1995) using a four-stage scheme providing third-

order approximation in time. The implicit predictor-corrector method has

been applied to nonlinear diﬀusion equations in Sachdev (1987).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 53

Pseudo-spectral methods have been adopted recently, there already exist

monographs Fletcher (1984); Sachdev (1987) where they are explained in

details, see also Zwillenger (1989) and references therein. In a pseudo-

spectral scheme the space derivatives are approximated very accurately by

means of the Fourier transforms, in particular, by the fast Fourier transform

(FFT) algorithm. The fourth-order Runge-Kutta scheme may be used for

the time derivative Kliakhandler (1999). It provides very mild stability

restrictions on the time step. The pseudo-spectral numerical scheme has

proved very eﬃcient in solving dispersive and dissipative equations. The

evolution of the initial monotonic solitary wave into radiating or oscillatory

solitary waves was simulated in Benilov et. al (1993) for the ﬁfth-order

KdV equation. In a series of papers Salupere et. al (1997); Salupere et. al

(2001) the solitary wave formation from a periodic input was studied for

an equation similar to Eq.(1.3), while the paper Salupere et. al (1994) is

devoted to the KdV soliton detection from a harmonic input. The use of the

pseudo-spectral approach to account for an evolution of an initial localized

pulse in framework of Eq.(1.3) is demonstrated in Sec. 1.2.1. Numerical

solutions of the nonlinear diﬀusion equations are considered in Fletcher

(1984); Sachdev (1987).

Finally, this method allows to study an evidence of the selected solitary

waves Kliakhandler et. al (2000) predicted on the basis of the asymptotic

analysis of Eq.(1.13) in previous section. To reveal the expected behaviour

of the solutions, the following numerical technique has been employed.

Since at the very late stage of the evolution the unstable waves are ex-

pected to be controlled by the ﬁnite-amplitude waves found in Kawahara

(1983), we deﬁne the typical wavenumber k

c

=

_

α

2

/2α

4

corresponding

to the most unstable linear mode in Eq.(1.13). The corresponding wave-

length is λ

c

= 2π/k

c

. The length of spatial domain was chosen to be 256λ

c

,

i.e. rather long. At the same time, the number of discretization points

was chosen to be 4096, i.e. λ

c

is covered by 16 points. The latter ensures

fair resolution of the whole solutions computed. Periodic boundary condi-

tions have been used for simulations. The pseudo-spectral technique was

employed for the spatial discretization and the Runge-Kutta fourth order

scheme for the time advance. The time step was chosen to be 0.01. The

tests with smaller time steps and better resolution gave indistinguishable

results. The control of the simulations in the Fourier space shows the good

resolution of the computed solutions. First, it has been checked that the

tendency to blow-up, Fig. 2.4(a), the damping, Fig. 2.4(b), and the selection

have been observed at the α

i

values prescribed by the theory in previous

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

54 Ampliﬁcation of Nonlinear Strain Waves in Solids

t

i

m

e

t

i

m

e

Damping (b)

distance

distance

(a) Blow-up

Fig. 2.4 Blow-up (a) and damping (b) of the initial conditions.

section. In the case of ”blow-up”, we found that the pulse tends to grow

rapidly at the time t

∗

rather close to the predicted t

∗

= T

∗

/ε from (2.51).

This observation allows to ﬁnd analytically “time-of-life” of the blowing

solutions of Eq. (1.13).

Following Kliakhandler et. al (2000) let us discuss in detail the selection

of ”dissipative” solitary waves occurring at A > 0, B < 0. Choosing the pa-

rameter values ε = 0.1, α

1

= 1, α

2

= 1, α

3

= 1, α

4

= 6/5, α

5

= −2, the

resulting amplitude of the selected solitary wave is obtained using (2.46),

(2.49) and b =

_

−A/B. The amplitude, 6α

3

b

2

/α

1

= 0.583 and the veloc-

ity V = 4α

3

b

2

= 0.389. We consider both the selection occurring from ”be-

low” when the magnitude of an initial Gaussian pulse is smaller than that of

the eventually selected solitary waves and the selection from ”above” when

the selected solitary wave amplitude is smaller than that of the initial pulse

magnitude. This would permit to separate the selection mechanism from

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 55

200 400 600 800

0

240

480

720

Distance

T

i

m

e

Fig. 2.5 ”Dissipative” solitary wave selection from an initial Gaussian proﬁle with am-

plitude 0.3 and width 36 units.

those originating from the growing unstable disturbances Kawahara (1983);

Oron and Rosenau (1997). One can see in Fig. 2.5 that up to the time

t ∼ 120 an initial Gaussian pulse with the magnitude 0.3 < 0.583 and

width 36 breaks into a train of three localized pulses aligned in row of

decreasing magnitude. Due to smallness of ε, the inﬂuence of the dissipa-

tive non-KdV terms is small at this stage. This may be seen by compar-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

56 Ampliﬁcation of Nonlinear Strain Waves in Solids

200 400 600 800 1000

0

60

120

180

240

300

360

420

480

T

i

m

e

Distance

Fig. 2.6 ”Dissipative” solitary wave selection from an initial Gaussian proﬁle with am-

plitude 1 and width 12 units.

ison of solutions of Eq.(1.13), shown as solid lines, with pure KdV case,

α

2

= α

4

= α

5

= 0, shown by dashed lines. At later stages the initial pulse

transforms into a train of the solitary waves. At nonzero ε each solitary

wave amplitude and velocity tend to the values 0.585 and 0.38 in agreement

with the theory of single solitary wave selection, while each of three KdV

solitons continues propagation with its own amplitude and velocity. ”Dis-

sipative” solitary waves form a bound state Christov and Velarde (1995);

Nekorkin and Velarde (1994) whose unequal spacing between equally high

crests reﬂects the original separation of the solitary waves in the KdV stage

when higher ideal solitons travel faster. The tail behind the train of soli-

tary waves appear as a result of short wave instability similar to Kawahara

(1983). However, the solitary waves have higher velocity than the velocity

of growing wave packets. As a result, the solitary waves bound state es-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 57

capes the destruction induced by the radiation that lags behind. A similar

ﬁnding for the GKS equation was reported in Chang et. al (1995). We see

on the last stages that the magnitude of the tail saturates. All these struc-

tures are quite well resolved and similar to those found in Kawahara (1983);

Oron and Rosenau (1997). They are robust and remain the same under the

mesh reﬁnement and smaller time steps.

The selection process realized from ”above” is shown in Fig. 2.6 when

initial Gaussian pulse has the magnitude 1 > 0.583 and the width 12.

Two equal solitary waves with amplitude 0.585 and velocity 0.38 appear

as a result of decrease of the magnitude of the initial pulse. Again the

comparison with pure KdV case is shown by dashed lines. All features of

the selection process are similar to the selection from ”below”.

Simulations of Eq.(1.13) with other values of parameters ε, α

i

such that

A > 0, B < 0, which are not reported here, show the same features of the

selection process as described above.

2.3.2 Nonlinear hyperbolic equations

Among the equations of kind (2.53) we are especially interested in various

Boussinesq-like long waves equations, that may be written in the form of

so-called double-dispersive equation Erofeev and Klyueva (2002); Samsonov

(2001)

u

tt

−α

1

u

xx

−α

2

( u

2

)

xx

−α

3

u

xxtt

+α

4

u

xxxx

= 0. (2.54)

At α

3

= 0 it corresponds to the classic Boussinesq equation. Besides

quadratic nonlinear term, α

2

( u

2

)

xx

, cubic or higher-order nonlinearities

may be considered Christou and Christov (2002); Soerensen et. al (1984);

Soerensen et. al (1987); also higher-order derivative terms may be incor-

porated in Eq.(2.54) Christou and Christov (2000); Christov et. al (1996).

A Fourier-Galerkin method were applied to computing localized solutions

in Christou and Christov (2000); Christou and Christov (2002). An im-

plicit diﬀerence scheme was developed in Christov and Maugin (1995);

Christov and Velarde (1994); Christov and Velarde (1995) to account for nu-

merical solutions of the Boussinesq equation and its generalizations where

Newton’s quasi-linearization of the nonlinear terms is employed. The so-

lution in Soerensen et. al (1984); Soerensen et. al (1987) is obtained using

the Gauss elimination.

There is an another method developed independently the USSR in 1953

Godunov and Ryaben’kii (1987); Samarskii and Nikolaev (1989) and in the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

58 Ampliﬁcation of Nonlinear Strain Waves in Solids

USA by L.H. Thomas (1949). That is why it is known in the West as the

Thomas method Morton and Mayers (1994); Richtmyer and Morton (1967).

As noted in Godunov and Ryaben’kii (1973), this method is justiﬁed for

the solutions of linear problems. However, the author applied it more than

ten years ago for a numerical solution of Eq.(2.54). One can see further in

the book that it gives numerical results in a good agreement with analytical

predictions. Previously the Thomas method was used for both nonlinear

evolution equationsBerezin (1987) and hyperbolic equationAlexeyev (1999).

Recently it was successfully applied in Bukhanovsky and Samsonov (1998);

Porubov et. al (1998) for computing rather complicated hyperbolic nonlin-

ear elastic systems.

Since the Thomas algorithm is not widely used for solutions of the non-

linear waves problems, it is useful to describe it in a more elaborate manner.

Following Godunov and Ryaben’kii (1987); Godunov and Ryaben’kii (1973)

consider a boundary problem:

a

n

u

n−1

+b

n

u

n

+c

n

u

n+1

= f

n

, 0 < n < N, (2.55)

u

0

= ϕ, u

N

= ψ. (2.56)

Let us assume

u

0

= L

1/2

u

1

+K

1/2

,

where L

1/2

= 0, K

1/2

= ϕ. It allows to exclude u

0

from the equation (2.55)

at n = 1, giving

u

0

= L

3/2

u

2

+K

3/2

,

with L

3/2

= −c

1

/b

1

, K

1/2

= (f

1

− a

1

ϕ)/b

1

. Following this procedure one

can obtain

u

n

= L

n+1/2

u

n+1

+K

n+1/2

, (2.57)

with

L

n+1/2

= −

c

n

b

n

+a

n

L

n−1/2

, K

n+1/2

=

f

n

−a

n

L

n−1/2

b

n

+a

n

L

n−1/2

.

Hence going from 1 to N one can calculate the coeﬃcients L

n+1/2

,

K

n+1/2

. Using (2.57) and the boundary conditions (2.56) at n = N, one

can obtain u

N−1

. Then coming back from n = N −1 to n = 1 one can ﬁnd

every u

n

from (2.57) using already known L

n+1/2

, K

n+1/2

. It was noted

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 59

in Godunov and Ryaben’kii (1987); Godunov and Ryaben’kii (1973) that

small amount of arithmetic operations together with a weak sensitivity to

the calculation errors are the main advantages of the method.

Assume u

n

= u(x

n

, t

j

) is a mesh function of a solution of Eq.(2.54)

while v

n

= u(x

n

, t

j−1

), w

n

= u(x

n

, t

j−2

) correspond to the mesh functions

on previous time steps. Then we have for discretization of Eq.(2.54)

a

n

= c

n

= α

3

, b

n

= −´x

2

−2α

3

,

f

n

= ´x

2

(w

n

−2v

n

)−α

1

´t

2

(v

n−1

−2v

n

+v

n+1

)+α

3

(2v

n−1

−4v

n

+2v

n+1

−

w

n−1

+ 2w

n

−w

n+1

) −α

4

´t

2

/´x

2

(v

n−2

−4v

n−1

+ 6v

n

−4v

n+1

+

v

n+2

) −α

2

´t

2

(v

2

n−1

−2v

2

n

+v

2

n+1

),

where ´t, ´x are time and space steps respectively. Uniqueness condition,

[b

n

[ > [a

n

[ + [c

n

[ Godunov and Ryaben’kii (1973), is always satisﬁed at

positive α

3

. The diﬀerence scheme is similar to those used in Soerensen

et. al (1984); Soerensen et. al (1987) in the case α

4

= 0 and with higher

order nonlinearities being taken into account.

2.4 Use of Mathematica

Recently various symbolic mathematical programs were developed to pro-

vide huge analytical operations. One of the most powerful is the Math-

ematica developed by Wolfram (1999) that is used by the author on all

stages of his studies. This section is not focused on the detailed description

of the magniﬁcent abilities of the program. Only some important features

are considered in relation to the problems studied here.

Among the advantages of the Mathematica one can mention variety

of the build-in mathematical functions. In particular, it works eﬃciently

with the Weierstrass elliptic function ℘ that may be used for obtaining ex-

act solutions. Indeed, the procedure described in Sec. 2.1 is based on the

transformation of the problem of a solution of a PDE to the problem of the

solution of algebraic equations for the ansatz parameters. These equations

may be too complicated for manual operations, and use of the Mathemat-

ica allows to obtain solutions avoiding errors. Design of the Mathematica

package provides an automatic ﬁnding of a solution. An example may be

found in Parkes and Duﬀy (1996) where a package is presented to obtain

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

60 Ampliﬁcation of Nonlinear Strain Waves in Solids

2 4 6 8 10

x

10

20

30

y

b

2 4 6 8 10

x

-6

-4

-2

2

4

6

y

a

Fig. 2.7 a) Incorrect and b) correct representations of the function y.

solutions in terms of the hyperbolic tanh-function. However, it is unlikely

that an automated method may be applied for ﬁnding periodic solutions.

First, it is necessary to introduce the rules for the elliptic functions deriva-

tives. Second, sometimes combinations like g

2

C − g

3

− 4C

3

or 12C

2

− g

2

should be kept in the solution for its convenient analysis, see Sec. 2.1.3.

The Mathematica cannot do it automatically. Finally, it works simulta-

neously with all possible solutions of the algebraic equations that yields

huge expressions at the intermediate stages of a solution and may result in

falling down the evaluation. Hence, it is better to manage the substitution

procedure step by step manually introducing the commands that provide

the most eﬃcient line of attack on the problem.

The algebraic manipulations should be used also for a check of the

asymptotic solutions and for derivation of the governing nonlinear equa-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Mathematical tools for the governing equations analysis 61

tions. Numerical abilities may be eﬃciently used for the analysis of ana-

lytical relationships, while Fortran, C++ or other languages are preferable

for numerical simulations of nonlinear PDEs. The Mathematica possesses

high-level graphic facilities, it admits representation of graphics in many

formats including PostScript and Encapsulate PostScript. Almost all ﬁg-

ures in the book are prepared using the Mathematica. Like analytical

procedures graphics may be automatized. In particular, numerical data

obtained using Fortran or C++ programs, may be represented by an auto-

mated procedure, see, e.g., Kliakhandler (1999).

Sometimes, the Mathematica gives incorrect results. In particular,

the construction of the proﬁles shown in Figs. 2.1 and 2.2 requires cal-

culation of the function y = Π[ϕ, n, κ], ϕ = arcsin(sn(k x)) expressed

through the elliptic integral of the third kind. Direct Mathematica com-

mand Plot[y, ¦x, 0, 10¦] yields the proﬁle shown in Fig. 2.7(a). One can see

it contains the points where there is no ﬁrst derivative of y. In order to

obtain correct smooth proﬁle it is necessary to deﬁne y as y = Π[ψ, n, κ],

where ψ is obtained using the Mathematica commands

ψ[x /; 0 ≤ x ≤ x

0

] = ϕ; ψ[x /; x

0

≤ x ≤ 3 x

0

] = π −ϕ;

ψ[x /; 3 x

0

≤ x ≤ 5 x

0

] = 2π +ϕ;

ψ[x /; 5 x

0

≤ x ≤ 7 x

0

] = 3π −ϕ; etc.

where x

0

= K(κ)/k, K(κ) is the complete elliptic integral of the ﬁrst kind, κ

is the elliptic functions modulus. Then we obtain required smooth proﬁle for

y shown in Fig. 2.7(b) around which the proﬁles are developed in Figs. 2.1,

2.2.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

62 Ampliﬁcation of Nonlinear Strain Waves in Solids

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 3

Strain solitary waves in an elastic rod

The theory of strain waves in solids began to develop over two hundred

years ago, see about it Love (1927); McNiven and McCoy (1974). During

the long period only linear theory of elasticity was considered since because

of the engineering needs and poor experimental facilities. Now the study of

the material properties Lurie (1990); Murnaghan (1951), acoustic signals

Biryukov et. al (1991); Oliner (1978); Parker and Maugin (1987) etc. require

mathematical models based on the nonlinear elasticity. Recent develop-

ments in general elastic theory may be found in Bland (1960); Lurie (1990);

Maugin (1993); Maugin (1995); Murnaghan (1951), while nonlinear waves

in solids were considered in Engelbrecht (1983); Engelbrecht (1997);

Erofeev (2002); Erofeev and Klyueva (2002); Jeﬀrey and Engelbrecht(1994);

Mayer (1995); Parker (1994); Parker and Maugin (1987); Samsonov (2001).

In order to go further it is necessary to deﬁne a notion of the word

solid. In macroelasticity solid may be deﬁned as a substance having a deﬁ-

nite volume and shape and resisting forces that tend to alter its volume or

shape. From the point of view of the theory of discrete media, solid may

be considered as a crystalline material in which the constituent atoms are

arranged in a 3D lattice with certain symmetries. As noted previously

Jeﬀrey and Engelbrecht(1994); Samsonov (2001), these deﬁnitions comple-

ment each other allowing to take into account a model of elastic potential

of atomic interactions, and to cover amorphous, porous or granular media.

When elastic features are one and the same in any direction we have

isotropic solids. Sometimes anisotropy is important, e.g., in design of non-

linear acoustic devices Oliner (1978); Parker and Maugin (1987). Strain

waves in solids may be classiﬁed as follows. Comparing the direction of the

wave propagation with the particle motion one can distinguish longitudinal

and shear waves. The former wave propagates along the direction of the

63

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

64 Ampliﬁcation of Nonlinear Strain Waves in Solids

particle motion, the latter -perpendicular to it. Waves propagating inside

solid are called bulk waves. In presence of a lateral surface, surface strain

waves are possible. In this book main attention is paid to the longitudinal

bulk waves in isotropic media and wave guides.

3.1 The sources of nonlinearities

Among the possible sources of nonlinearity we brieﬂy consider so-called

geometrical and physical nonlinearities since they aﬀect the strain wave

propagation to a greater extent. Other kinds of nonlinearities are consid-

ered in Engelbrecht (1997); Jeﬀrey and Engelbrecht(1994). The geometrical

nonlinearity is described by the exact expression of the strain tensor always

used in the theory of large deformations. Initially the position of a particle

is accounted for a vector-radius

− →

r , or an initial or reference conﬁguration

is deﬁned. Loading forces provide the displacement of particle yielding the

current or actual conﬁguration characterized by an another vector-radius

−→

R. Then the movement is described by the displacement vector,

V =

−→

R -

−→

r . However, not only the positions of the particles vary during the

deformation but also the distances between them. In order to describe the

alteration of the distance a deformation or strain tensor is introduced. It

is obtained from the diﬀerence between the squares of the arc length in the

deformed (actual) and undeformed (reference) conﬁguration. In the refer-

ence conﬁguration the Cauchy-Green ﬁnite deformation tensor C is deﬁned

Lurie (1990), whose general form is

C =

_

→

∇

V + (

→

∇

V )

T

+

→

∇

V (

→

∇

V )

T

_

/2

(written in terms of a vector gradient

→

∇

V and its transpose (

→

∇

V )

T

). In

rectangular Cartesian coordinates x

i

the components of C may be written

in a more familiar form

C

ik

=

1

2

(

∂u

i

∂x

k

+

∂u

k

∂x

i

+

∂u

l

∂x

i

∂u

l

∂x

k

),

where u

k

are the components of the displacement vector

V .

As a result of a deformation process there appear stresses. Certainly

they should be connected with the strains, and this is a source of the phys-

ical nonlinearity. It was found more than hundred years ago that Hook’s

linear law of elasticity is insuﬃcient, see about it in Jeﬀrey and Engel-

brecht(1994); Samsonov (2001). Since Piola-Kirchoﬀ stress tensor is deﬁned

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 65

in the reference conﬁguration, through the volume density of the internal

energy Π in adiabatic processes (or through the Helmgholtz free energy in

thermoelastic processes),

P

ik

=

∂Π

∂C

ik

,

one can say the physical nonlinearity depends upon the structure of the

internal (or free) strain energy density Π. The energy of deformation must

be insensitive to the rotation of the reference frame. It was Murnaghan

(1951) who supposed to develop the energy as a power series in the three

invariants of the strain tensor,

Π =

λ + 2µ

2

I

2

1

−2µI

2

+

l + 2m

3

I

3

1

−2mI

1

I

2

+nI

3

, (3.1)

where I

k

, k = 1, 2, 3 are the invariants of tensor C:

I

1

(C) = trC, I

2

(C) = [(trC)

2

−trC

2

]/2, I

3

(C) = det C. (3.2)

The ﬁrst two terms in (3.1)account for linear elasticity, hence the second or-

der elastic moduli, or the Lam´e coeﬃcients (λ, µ), characterize linear elastic

properties of the isotropic material. Other terms in (3.1) describe material

or physical nonlinearity Lurie (1990); Murnaghan (1951). Accordingly, the

third order elastic moduli, or the Murnaghan moduli (l, m, n) account for

nonlinear elastic properties of the isotropic material. The energy may be

written using another set of invariants,

J

1

(C) = trC, J

2

(C) = trC

2

, J

3

(C) = trC

3

,

in the form Lurie (1990)

Π =

λ

2

J

2

1

+µJ

2

+

ν

1

6

J

3

1

+ν

2

J

1

J

2

+ν

3

J

3

. (3.3)

Since

J

1

(C) = I

1

(C) , J

2

(C) = I

2

1

(C) −2I

2

(C),

J

3

(C) = I

3

1

(C) −3I

1

(C)I

2

(C) + 3I

3

(C),

Eqs.(3.1) and (3.3) coincide if ν

1

= 2l −2m + n, ν

2

= m−n/2, ν

3

= n/4.

In some cases there is a need in more terms in Eq.(3.1):

Π =

λ + 2µ

2

I

2

1

−2µI

2

+

l + 2m

3

I

3

1

−2mI

1

I

2

+nI

3

+

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

66 Ampliﬁcation of Nonlinear Strain Waves in Solids

Table 3.1 Lame’s and Murnaghan’s modulii, ∗10

−9

N/m

2

material λ µ l m n

Polystyrene 1.71 0.95 -18.9 -13.3 -10

Steel Hecla 37 111 82.1 -459 -461 -358

Aluminium 2S 57 27.6 -299 -311 -228

Pyrex glass 13.5 27.5 14 92 420

SiO

2

melted 15.9 31.3 129 71 -44

a

1

I

4

1

+a

2

I

2

1

I

2

+a

3

I

1

I

3

+a

4

I

2

2

, (3.4)

The fourth order moduli (a

1

, a

2

, a

3

, a

4

) can be either positive or negative.

However, mainly the third-order moduli data may be found in the literature

Frantsevich et. al (1982); Lurie (1990), some of them are collected in Table

3.1.

The series expansions (3.1), (3.4) are convenient to account for the de-

formation of compressible materials (metals, polymers, etc.), whose yield

point is small. For incompressible materials, like rubber, there exist an-

other models, e.g., the Mooney model Lurie (1990). Its generalization, the

Mooney-Rivlin model, may be used even for compressible materials, its

application to nonlinear strain waves in a rod may be found, e.g., in Dai

(1998).

3.2 Modelling of nonlinear strain waves in a free lateral

surface elastic rod

3.2.1 Statement of the problem

Let us consider an isotropic, axially inﬁnitely extended, free lateral surface

cylindrical elastic rod, see Fig. 3.1. Axi-symmetry leads to using cylin-

drical Langrangian coordinates (x, r, ϕ), where x is the axis of the rod,

ϕ [0, 2π], 0 ≤ r ≤ R. When torsions are neglected, the displacement

vector is

V = (u, w, 0). We shall consider the propagation of longitudinal

strain waves of small but ﬁnite amplitude in the rod. Once the reference

conﬁguration is deﬁned we use Hamilton’s variation principle to obtain the

governing equations together with the boundary conditions, setting to zero

the variation of the action functional,

δS = δ

_

t

1

t0

dt2π

_

∞

−∞

dx

_

R

0

r Ldr = 0, (3.5)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 67

Fig. 3.1 Free lateral surface cylindrical rod

where, L is the Lagrangian density per unit volume, L=K − Π, with Π

deﬁned by Eq.(3.1). For the kinetic energy density K we have

K =

ρ

0

2

_

_

∂u

∂t

_

2

+

_

∂w

∂t

_

2

_

(3.6)

where ρ

0

is the rod material density at t = t

0

. The integration in brackets in

(3.5) is carried out at the initial time t = t

0

. Initially, the rod is supposed

to be in its natural, equilibrium state. In absence of torsions non-zero

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

68 Ampliﬁcation of Nonlinear Strain Waves in Solids

components of the Cauchy-Green deformation tensor C, are

C

xx

= u

x

+

1

2

(u

2

x

+w

2

x

), C

rr

= w

r

+

1

2

(u

2

r

+w

2

r

), C

ϕϕ

=

1

r

w +

1

2r

2

w

2

,

C

rx

=

1

2

(u

r

+w

x

+u

x

w

r

+w

x

w

r

).

Hence the invariants in (3.2) may be written as

I

1

(C) = C

xx

+C

rr

+C

ϕϕ

,

I

2

(C) = C

xx

C

rr

+C

xx

C

ϕϕ

+C

rr

C

ϕϕ

−C

2

rx

,

I

3

(C) = C

ϕϕ

(C

xx

C

rr

−C

2

rx

).

The following boundary conditions (b.c.) are imposed for a free lateral

surface rod:

w → 0, at r → 0, (3.7)

P

rr

= 0 , at r = R, (3.8)

P

rx

= 0, at r = R, (3.9)

where the components P

rr

, P

rx

of the Piola - Kirchhoﬀ stress tensor P

are deﬁned as

P

rr

= (λ + 2µ) w

r

+λ

w

r

+ λ u

x

+

λ + 2µ +m

2

u

2

r

+

3λ + 6µ + 2l + 4m

2

w

2

r

+ (λ + 2l) w

r

w

r

+

λ + 2l

2

w

2

r

2

+

(λ + 2l) u

x

w

r

+ (2l −2m+n) u

x

w

r

+

λ + 2l

2

u

2

x

+

λ + 2µ +m

2

w

2

x

+ (µ +m) u

r

w

x

, (3.10)

P

rx

= µ (u

r

+w

x

) + (λ + 2µ +m) u

r

w

r

+

2λ + 2m−n

2

u

r

w

r

+

(λ + 2µ +m) u

x

u

r

+

2m−n

2

w

x

w

r

+ (µ +m) w

x

w

r

+

(µ +m) u

x

w

x

. (3.11)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 69

3.2.2 Derivation of the governing equation

Exception of torsions provides transformation of the initial 3D problem

into a 2D one. Subsequent simpliﬁcation is caused by the consideration of

only long elastic waves with the ratio R/L ¸ 1 between the rod radius R

and typical wavelength L. The typical elastic strain magnitude B does not

exceed the yield point of the material. Since Murnaghan’s material have

small yield points, one can assume B ¸ 1. The Hamilton principle (3.5)

yields a set of coupled equations for u and w together with the b.c. (3.8),

(3.9). To obtain a solution in universal way one usually proceeds to the di-

mensionless form of the equations and looks for the unknown displacement

vector components in the form of power series in the small parameters of

the problem (for example R/L), hence, leading to an asymptotic solution

of the problem. However, this procedure has some disadvantages. In par-

ticular, comparison of the predictions from the dimensionless solution to

the experiments suﬀers from the fact that both B and L, are not well de-

ﬁned. In particular, solitary wave has an inﬁnite wave length. Further, the

coeﬃcients of the nonlinear terms usually contain combinations of elastic

moduli which may be also small in addition to the smallness of B something

not predicted beforehand. Finally, this procedure gives equations of only

ﬁrst order in time, t, while general equations for displacements u and w are

of the second order in time. Therefore the solution of the model equation

will not satisfy two independent initial conditions on longitudinal strains

or displacements Samsonov (2001).

An alternative is to simplify the problem making some assumptions

about the behavior of longitudinal and/or shear displacements and/or

strains in the elastic wave-guide. Referring to the elastic rod these rela-

tionships give explicit dependence of u and w upon the radius, while their

variations along the rod axis are described by some unknown function and

its derivatives along the axis of the rod. Then the application of Hamilton’s

principle (3.5) yields the governing equation in dimensional form for this

function. This equation is of the second order of time, hence its solution

can satisfy two independent initial conditions. Any combinations of elastic

moduli appear in the coeﬃcients of the equation, hence, subsequent scaling

may take into account their orders when introducing small parameters.

For an elastic rod, the simplest assumption is the plane cross section

hypothesis McNiven and McCoy (1974): the longitudinal deformation pro-

cess is similar to the beards movement on the thread. Then every cross

section of the rod remains ﬂat, hence, u = U(x, t) does not change along

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

70 Ampliﬁcation of Nonlinear Strain Waves in Solids

the radius r. However, this assumption is not enough due to the Poisson

eﬀect, i.e., longitudinal and shear deformations are related. That is why

Love (1927) proposed to use a relationship between w and u: w = −r ν U

x

,

with ν the Poisson coeﬃcient. Unfortunately, the plane cross-section hy-

pothesis and Love’s hypothesis do not satisfy the boundary conditions that

demand vanishing of both the normal and tangential stresses, P

rr

and P

rx

,

at the lateral surface of the rod with prescribed precision.

Another theory has been proposed in Porubov and Samsonov (1993) to

ﬁnd the relationships between displacement vector components satisfying

b.c. on the lateral surface of the rod (3.8), (3.9) as well as the condition

for w (3.7). Later it was developed in Porubov and Velarde (2000).

Since pure elastic wave are studied, B ¸ 1, the ”linear” and ”non-

linear” parts of the relationships may be obtained separately. A power

series approximations is used, as generally done for long wave processes.

Accordingly, the longitudinal and shear displacement in dimensional form

are:

u = u

L

+u

NL

, u

L

= u

0

(x, t) +r u

1

(x, t) + r

2

u

2

(x, t) +...,

u

NL

= u

NL0

(x, t) +r u

NL1

(x, t) +..., (3.12)

w = w

L

+w

NL

, w

L

= w

0

(x, t) +r w

1

(x, t) + r

2

w

2

(x, t) +....,

w

NL

= w

NL0

(x, t) +r w

NL1

(x, t) +... (3.13)

Substituting the linear parts u

L

and w

L

(3.12), (3.13) into the b.c. (3.7) and

in the linear parts of b.c. (3.8), (3.9), and equating to zero terms at equal

powers of r one obtains u

k

and w

k

. Using these results the nonlinear parts

u

NL

, w

NL

are similarly obtained from the full b.c. We get u

0

(x, t) = U(x, t)

while other u

k

and w

k

are expressed through U and its derivatives. Then

(3.12), (3.13) are substituted into (3.1) and (3.6). Running the Hamilton

principle (3.5) we obtain the single governing equation for the unknown

function U(x, t). In contrast to the theory based on the plane cross section

hypothesis and the Love hypothesis, the present theory allows to account

for nonzero b.c. for the stresses on the lateral surface of the rod.

3.3 Double-dispersive equation and its solitary wave solu-

tion

In order to derive the governing equation for longitudinal strain waves in

a free lateral surface rod we assume that B ∼ R

2

/L

2

. This assumption

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 71

provides a balance between nonlinearity and dispersion required for exis-

tence of the bell-shaped solitary waves of permanent shape. Substituting

power series (3.12), (3.13) into the boundary conditions and equating to

zero combinations at each power of the radius, we get

u = U +a

2

r

2

U

xx

, (3.14)

w = b

1

r U

x

+b

3

r

3

U

xxx

+B

1

r U

2

x

. (3.15)

where

a

2

=

ν

2

, b

1

= −ν, b

3

=

ν

2

2(3 −2ν)

,

B

1

=

ν(1 +ν)

2

+

(1 −2ν)(1 +ν)

E

[l(1 −2ν)

2

+ 2m(1 +ν) −nν].

Then kinetic and potential energy truncated approximations are

K =

ρ

0

2

(U

2

t

+νr

2

[U

t

U

xxt

+νU

2

xt

]), (3.16)

Π =

1

2

_

EU

2

x

+

β

3

U

3

x

+νEr

2

U

x

U

xxx

_

(3.17)

Here ν and E are the Poisson ratio and the Young modulus correspondingly,

ν =

λ

2(λ +µ)

, E =

µ(3λ + 2µ)

λ +µ

,

while β is a nonlinear coeﬃcient, β = 3E + l(1 − 2ν)

3

+ 4m(1 − 2ν)(1 +

ν) + 6nν

2

. We have to truncate the approximations (3.14), (3.15), (3.16),

(3.17) in order to be in an agreement with the ﬁve-constant Murnaghan

approximation (3.1), where cubic nonlinear terms are neglected.

Comparing these relations with those obtained using cross-section and

Love’s hypothesis, a

2

= 0, b

3

= 0, B

1

= 0, one can see that only the

term a

2

r

2

U

xx

makes its contribution into Eqs.(3.16), (3.17), while terms

b

3

r

3

U

xxx

and B

1

r U

2

x

are needed to satisfy the boundary conditions on the

lateral surface with prescribed accuracy.

Substituting (3.16), (3.17) into (3.5) one can obtain the so-called double-

dispersive equation, DDE, for a strain function v = U

x

,

v

tt

−α

1

v

xx

−α

2

( v

2

)

xx

−α

3

v

xxtt

+α

4

v

xxxx

= 0. (3.18)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

72 Ampliﬁcation of Nonlinear Strain Waves in Solids

where

α

1

=

E

ρ

0

, α

2

=

β

2ρ

0

, α

3

=

ν(ν −1)R

2

2

, α

4

= −

νER

2

2ρ

0

.

For the ﬁrst time DDE was derived independently by some authors, in

framework of the plane cross section and Love’s hypothesis, see about it

Erofeev and Klyueva (2002); Samsonov (2001). The important diﬀerence

is only in the values of the dispersive terms coeﬃcients α

3

and α

4

, in

Samsonov (1988) it was α

3

= −ν

2

R

2

/2, α

4

= µν

2

R

2

/(2ρ

0

).

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

Fig. 3.2 Formation of solitary waves from an initial rectangular tensile pulse.

Eq.(3.18) possesses exact travelling solitary wave solution that may be

obtained by direct integration,

v = Acosh

−2

(k (x −V t)), (3.19)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 73

with

A =

3(ρ

0

V

2

−E)

β

, k

2

=

E −ρ

0

V

2

2νR

2

(E +ρ

0

(ν −1)V

2

)

. (3.20)

Accordingly, V

2

lies either inside the interval

E

ρ

0

(1 −ν)

< V

2

< c

2

∗

=

E

ρ

0

, if −1 < ν < 0. (3.21)

or in

c

2

∗

< V

2

<

E

ρ

0

(1 −ν)

, if ν > 0, (3.22)

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

50 100 150 200

-0.5

0.5

1

Fig. 3.3 Formation of a train of solitary waves from an initial rectangular tensile pulse.

When positive Poisson ratio is positive, more precisely, 1/2 ≥ ν > 0,

the sign of the amplitude is deﬁned by the sign of β. The nonlinearity coef-

ﬁcient is the only coeﬃcient carrying an information about the Murnaghan

moduli. Hence, they deﬁne whether tensile or compression solitary wave

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

74 Ampliﬁcation of Nonlinear Strain Waves in Solids

may propagate for a given elastic material of the rod. We see that it is the

mixed dispersive term α

3

v

xxtt

, who establishes the permitted ﬁnite interval

for the wave velocity.

Exact solution requires speciﬁc initial conditions. It is of interest to

know how rather arbitrary localized initial pulse evolves. To this purpose

numerical simulation of the equation has been performed using numerical

method explained in Sec. 2.3.2.

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

50 100 150 200

-0.5

0.5

Fig. 3.4 Delocalization of an initial compression pulse.

It was found that rather arbitrary initial pulse splits into the train of

solitary waves or evolves into small amplitude oscillating wave-packet ac-

cording to the predictions about wave parameters done on the basis of exact

single travelling solitary wave solution. Indeed, assume the material elastic

features of the rod yield β > 0. Shown in Fig. 3.2 is the evolution of an

initial tensile rectangular pulse into single solitary wave, while massive ini-

tial pulse splits into the sequence of solitary waves, see Fig. 3.3. However,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 75

each solitary wave is accounted for the exact single travelling wave solution

(3.19). Figure 3.4 demonstrates no solitary wave generation from the initial

compression pulse. However, when the material of the rod possesses β < 0,

only compressive solitary wave arises from an initial compression localized

input. To sum up, single travelling wave exact solution provides correct

predictions about solitary waves formation in an unsteady process of an

arbitrary localized input evolution.

Finally, let us estimate the deviation from the plane cross section caused

by the second terms in the expression (3.14). Indeed, the curvature h

of the proﬁle of longitudinal strain u along the radius of the rod is h =

[u

rr

[ /(1 +u

2

r

)

3/2

= [νU

xx

[ /(1 +ν

2

r

2

U

2

xx

)

3/2

. It was found in Porubov and

Samsonov (1993)that the variation of longitudinal strain u along the radius

is negligibly small when U

x

is the solitary wave (3.19) that is important for

the experimental observation of the waves.

3.4 Observation of longitudinal strain solitary waves

We brieﬂy consider the recent successful experiments on solitary wave ob-

servation in a transparent rod, details may be found in Dreiden et. al (1995);

Samsonov (2001). The optical methods were used because they are prefer-

able to study transparent optical phase inhomogeneities. They allow not

only to visualize inhomogeneity but also to determine its parameters, and

on the other side, being contactless, they do not introduce any disturbances

in an object under study. All optical methods record the alterations of re-

fractive index in an object, when studying optically transparent phase in-

homogeneities. Shadowgraphy is more convenient to record a considerable

refractive index gradient, for example, caused by a strong shock wave prop-

agation. It was shown theoretically in our case that a strain solitary wave

is a propagating long density wave of small amplitude. Interferometry is

the most appropriate for such waves study because it allows to observe and

measure with suﬃcient accuracy even small refractive indices alterations.

Holographic interferometry method, used in the experiments Dreiden

et. al (1995), has several advantages in comparison with the conventional

optical interferometry. In particular, limitations to the optics quality are

considerably lower because wave fronts to be compared pass through the

same optical path. For this reason both waves are distorted to the same

extent and possible defects in optical elements and experimental cell do not

aﬀect the resulting interference pattern. However, the choice of an optical

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

76 Ampliﬁcation of Nonlinear Strain Waves in Solids

recording method allows to study, in general, only elastic materials, that

are transparent for the given light wave length.

To check that the excited strain wave possesses indeed the solitary wave

feature to conserve its shape, it is necessary to follow in observations a

propagation along an extended elastic wave guide. However, in a wave

guide made of material highly absorbing linear elastic waves, it will be

suﬃcient to detect the constant shape wave propagation at much shorter

distance.

Based on the results of the analysis presented above the transparent

polystyrene SD-3 has been chosen as an appropriate material for a wave

guide manufacturing. The elastic properties of it are given by a set of

parameters ν = 0.35; β = − 6 10

10

N/m

2

; c

∗

= 1.8 10

3

m/sec, see

Frantsevich et. al (1982).

The experimental set-up used to generate and observe the strain soli-

tons, consists of a basin where the rod is submerged into the water, a

device to produce the initial shock wave, a holographic interferometer for

the recording of a wave pattern, a synchronizer and a laser radiation energy

meter. The waves inside the rod are generated from a primary shock wave

produced in the water near the edge of the rod by laser vaporization of a

metal target. The ﬁrst exposure of the hologram is carried out to obtain

the hologram of undisturbed wave guide. The second exposure is produced

by a laser pulse synchronized with the prescribed stage of the wave propa-

gation. Observations are carried out in the transversal direction, and two

cut-oﬀ were made parallel to the rod axis in order to make transparent

the central part of the rod. The carrier fringes on interferograms, obtained

due to the reconstruction of doubly exposed holograms, occur due to the

wedge turn between the exposures. The longitudinal strain wave patterns

are recorded at various distances from the input edge of the rod, that are

attained by the cell displacement along the axis of wave propagation.

The solitary wave parameters are calculated based on the data of the

holography interferograms obtained. Typical interferogram is shown in

Fig .3.5. Only central part of the PS rod is transparent thanks to the

vertical cuts. Curved interference fringe is extracted from the area inside

the rod and placed below the interferogram. Note that the interferometric

pattern does not exhibit a standard bell-shaped image of a shallow water

solitary wave since the strain solitary wave is, in fact, a longitudinal density

wave in solids.

The solitary wave amplitude can be calculated using the interferometry

fringe shift ∆K measured in the interference pattern, see Fig. 3.5. Let 2h

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 77

Fig. 3.5 Experimental observation of longitudinal strain solitary wave in an elastic

rod(after Dreiden et. al (1995)).

be a distance passed by the recording light across the rod, i.e., precisely,

the distance between two longitudinal cut-oﬀ. Before the deformation the

phase variation ∆φ

1

of the light wave having the length λ, is caused by the

laser light propagation along the distance q − 2h through the water and

the distance 2h through the rod ( where q is the distance between the cell

walls) Samsonov et. al (1998):

n

0

(q −2h) + 2hn

1

=

λ

2π

∆φ

1

. (3.23)

Here n

0

, n

1

are the refraction indices of water and the elastic material

before deformation. After the deformation the refraction index value of

the rod changes to n

2

. Moreover, the distance, which light passes through

the rod and water, varies due to the deformation of the rod. As a result

we obtain the formula for the magnitude of the light wave phase variation

∆φ

2

:

n

0

(q −2h −2∆h) +n

2

(2h + 2∆h) =

λ

2π

∆φ

2

. (3.24)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

78 Ampliﬁcation of Nonlinear Strain Waves in Solids

Evidently, the interferometry fringe shift ∆K is deﬁned as

∆K =

∆φ

2

−∆φ

1

2π

. (3.25)

The new value of the refraction index of the deformed rod n

2

is caused by

the local density variation:

∆ρ

ρ

=

n

2

−n

1

n

1

−1

. (3.26)

On the other side, one can obtain the density variation from the solution

of a static linear problem on uniaxial compression (or tension), see Lurie

(1990), and as a result the following relationship is valid:

∆ρ

ρ

= U

x

(2ν −1). (3.27)

Then we get ﬁnally from (3.25):

U

x

= −

λ∆K

2h[(n

1

−1)(1 −2ν) +ν(n

1

−n

0

)]

. (3.28)

The amplitude is determined by the maximal fringe shift value. Deriva-

tion of the relationship (3.28) shows that the length L of the solitary pulse

may be directly determined from the interferogram as the length of the

fringe shift perturbation between two undisturbed areas.

In our experiments Dreiden et. al (1995) we used a long polystyrene rod,

149 mm long. The nonlinearity parameter β < 0 for polystyrene, and we

can anticipate the appearance of only compression solitary wave. It is fairly

long and keeps its shape on propagation. As follows from (3.28) the wave

shown under interferogram in Fig. 3.5 is indeed a longitudinal compression

wave. There is no any tensile wave around the observed wave, moreover, in

the water surrounding the rod the interference patterns remains horizontal,

i.e., undisturbed, that proves the wave that is measured propagates inside

the rod. Therefore, the theory developed for an inﬁnite cylindrical rod

allows to predict an existence of solitary waves in a ﬁnite-length rod with

vertical cuts.

Unfortunately only the wave amplitude may be measured more or less

precisely, hence full quantitative comparison with the theory is impossible.

The unsteady process of the solitary wave generation cannot be accounted

for the theory. The propagation of a shock wave before the solitary wave

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 79

contradicts numerical results shown in Fig. 3.2. For a long time solitary

waves were observed only in a wave guide made of polystyrene. Very re-

cently they were generated also in a plexiglas rod Samsonov et. al (2003).

3.5 Reﬂection of solitary wave from the edge of the rod

Following Dreiden et. al (2001) let us consider a semi-inﬁnite homogeneous

rod, −∞< x < X. The procedure developed in section 3.2 may be applied

but now the elementary work done by the external forces should be included

into the Hamilton principle, and the integration with respect to x is carried

out over −∞< x < X,

δ

_

t

1

t

0

dt

_

2π

_

X

−∞

dx

_

R

0

r Ldr

_

+

_

t

1

t

0

δAdt = 0. (3.29)

The Lagrangian density per unit volume, L=K − Π, is deﬁned as before.

According to Lurie (1990) we cannot call Eq.(3.29) variation principle since

external forces are not necessary potential ones. Use of the Hamilton prin-

ciple yields the DDE (3.18) as a governing equation for longitudinal waves

together with the boundary conditions requiring zero values for v and its

derivatives as x → −∞, while on the rod end, x = X, they depend upon

the type of clamping. If the rod end is assumed free, the elementary work

of external forces at the end of the rod, x = X, is zero, δA = 0, and Eq.

(3.29) yields

v = 0, v

xx

= 0. (3.30)

When the end is clamped, the elementary work is not determined, but

obvious kinematic reasons require zero displacement and its velocity,

U = 0, U

t

= 0, (3.31)

that may be rewritten in terms of strains,

v

x

= 0, v

xt

= 0. (3.32)

Since DDE does not possess an exact solution describing interacting waves

moving in opposite directions, it is necessary to apply an asymptotic tech-

nique, reducing DDE to a nondimensional form. Introducing the scales, L

for x, L/c

0

for t, and B/L for v, where c

0

=

_

E/ρ

0

is the so-called rod

velocity. A small parameter ε is chosen as ε = B = R

2

/L

2

<< 1, to

balance nonlinearity and dispersion.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

80 Ampliﬁcation of Nonlinear Strain Waves in Solids

Assume the asymptotic solution v depends, in addition to x, t, on a

slow time τ = εt. The solution is sought in power series in ε:

v = v

0

+εv

1

+. . . . (3.33)

Substituting (3.33) into dimensionless DDE, and equating the terms with

the same power of ε we obtain the D’Alembert solution for v

0

:

v

0

= v

01

(θ, τ) +v

02

(ψ, τ), (3.34)

where θ = x +t, ψ = x −t. At order ε there is a linear equation for v

1

,

2v

1,θψ

= 2v

01,θτ

−2v

02,ψτ

+

β

2E

_

(v

2

01

)

θθ

+ 2v

01,θ

v

02,ψ

+ (v

2

02

)

ψψ

_

+

ν

2

2

(v

01,θθθθ

+v

02,ψψψψ

) (3.35)

The absence of secular terms leads to two uncoupled KdV equations for the

functions v

01

and v

02

,

2v

01,τ

−

β

2E

(v

2

01

)

θ

−

ν

2

2

v

01,θθθ

= 0, (3.36)

2v

02,τ

+

β

2E

(v

2

02

)

ψ

+

ν

2

2

v

02,ψψψ

= 0, (3.37)

and the bounded solution of Eq. (3.35) is

v

1

=

β

E

v

01

v

02

+v

11

(θ, τ) +v

12

(ψ, τ). (3.38)

Substituting the soliton solutions of the KdV equations (see Chapter 1)

into the leading order solution (3.34), we obtain

v

0

=

6Eν

2

β

k

2

[cosh

−2

_

k[x + (1 +εν

2

k

2

)t −x

01

_

+

cosh

−2

_

k[x −(1 +εν

2

k

2

)t −x

02

]

_

]. (3.39)

where x

01

are the constant phase shifts. It follows from (3.39) that the type

of the strain wave depends upon the sign of β like in previous section. Both

conditions (3.32) on the clamped end are satisﬁed if x

02

= 2X−x

01

. In this

case reﬂection of the solitary wave occurs with no change of its shape. On

the contrary, one cannot satisfy boundary conditions (3.30) at a free end of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 81

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

5 10 15 20

-0.5

0.5

Fig. 3.6 Reﬂection of the solitary wave from a free end of the rod.

the rod. It means that the reﬂected solitary wave does not propagate but

disappears due to dispersion.

It may be clearly seen from numerical simulation of the reﬂection Drei-

den et. al (2001). Again the implicit ﬁnite-diﬀerence scheme explained in

Sec. 2.3.2, is used. Implementation of the boundary conditions at the end

of the rod is eﬀected by means of symmetric continuation of the calculation

area beyond the real rod end, so that the area of calculation would occupy

the interval 0 < x < 2X. In case of free end of the rod the sign of β in

the interval X < x < 2X is chosen opposite to that used in the interval

0 < x < X. Initial pulses are assumed to be the equal magnitude KdV

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

82 Ampliﬁcation of Nonlinear Strain Waves in Solids

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

5 10 15 20

0.25

0.5

Fig. 3.7 Reﬂection of the solitary wave from a ﬁxed end of the rod.

solitons located centrally symmetric relative to the genuine end x = X.

The type of an initial solitary wave depends upon the sign of β. In par-

ticular, if β > 0 the initial solitary wave in the interval 0 < x < X is a

tensile wave, while in the interval 0 < x < 2X we impose the same tensile

solitary wave for the clamped end or the compression solitary wave for the

free end. Initial velocities of the solitary waves are chosen equal and taken

to be directed towards each other.

The reﬂection from the free end is shown in Fig. 3.6. The right-hand

side of the ﬁgure corresponds to the free end. It is seen that the amplitude

of the solitary wave propagating from left to right decreases as it reaches

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 83

Fig. 3.8 Observation of a reﬂection of the solitary wave from a free end of the rod.

the end of the rod. The reﬂected wave is of opposite, ﬁrst, its amplitude

grows, then the wave is dispersed, and no localized strain wave is observed

near the input end of the rod.

In case of the clamped end numerical results are shown in Fig. 3.7. In

agreement with the theory the amplitude of the solitary wave is seen to

be nearly twice as large when it reaches the end of the rod. The reﬂected

wave has the same amplitude and velocity as the incident one. Keeping its

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

84 Ampliﬁcation of Nonlinear Strain Waves in Solids

Fig. 3.9 Reﬂection of the solitary wave from a clamped end of the rod.

shape, the reﬂected wave propagates towards the input end of the rod.

Experimental observation of the solitary strain wave reﬂection were car-

ried out in Dreiden et. al (2001) using the same technique as for the study

of the solitary wave propagation. Shown in Fig. 3.8 is the reﬂection from

the free end. Footnotes in Fig. 3.8(a) and Fig. 3.8(b) demonstrate the

decrease of the amplitude of the initially generated compression localized

wave shown in Fig. 3.8(a), cf. with the ﬁrst and the ﬁfth stages in Fig. 3.7.

It was found in experiments the absence of any reﬂected localized waves

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Strain solitary waves in an elastic rod 85

Fig. 3.10 Reﬂected wave comes back to the input end of the rod.

moving towards the input end of the rod, cf. with the last stage of Fig. 3.6.

Hence all observations verify theoretical predictions.

Figures 3.9, 3.10 demonstrate the reﬂection of the strain solitary wave

from the rod end attached to the right to a massive brass plate. The

material of the plate is chosen so as to avoid penetration of the wave energy

outside the rod. It was found experimentally in Dreiden et. al (2001) that

the acoustic resistance of the plate should be much greater than that of

the material of the rod (polystyrene). It is seen that the amplitude of the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

86 Ampliﬁcation of Nonlinear Strain Waves in Solids

incident solitary wave, Fig. 3.9(a), is almost doubled by the ﬁxed end of

the rod, Fig. 3.9(b)in an agreement with the ﬁrst and the ﬁfth stages in

Fig. 3.7. Figure 3.10 shows the reﬂected solitary wave moving to the left

at a distance of 140 mm from the ﬁxed end. It has the same amplitude

and velocity as the incident one, cf. the last stages in Fig. 3.7. In order to

see it is necessary to subtract the fringe shift outside the rod (marked by

B in Fig. 3.10) from the fringe shift measured inside the rod (marked by

A in Fig. 3.10). The wave observed in experiments demonstrates the main

feature of the solitary waves to keep their shapes after collisions, hence it

conﬁrms also the fact that the observed localized wave is indeed the strain

solitary wave predicted by the theory.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 4

Ampliﬁcation of strain waves in

absence of external energy inﬂux

As we seen in the previous chapter, strain solitary wave propagates without

change of the shape in an uniform rod with a free lateral surface. However,

the wave may still exist even in presence of an external medium or when

a material of the rod is microstructured. At the same time, the shape of

the wave varies when the balance between nonlinearity and dispersion is

destroyed. The simplest reason is the varying cross section of the rod.

4.1 Longitudinal strain solitary wave ampliﬁcation in a nar-

rowing elastic rod

The section is devoted to the theoretical and experimental description of

the propagation and ampliﬁcation of the strain solitary wave (soliton) in

a cylindrical nonlinearly elastic rod with varying cross section. We call it

an inhomogeneous rod in the following for convenience, while the rod with

permanent cross section will be called the homogeneous one. We follow

Samsonov et. al (1998) where the results presented below were published

for the ﬁrst time.

4.1.1 Governing equation for longitudinal strain waves

propagation

Let us consider the wave propagation problem for an isotropic inﬁnite non-

linearly elastic compressible rod with varying cross section, see Fig. 4.1.

Introducing the cylindrical Lagrangian coordinate system (x, r, ϕ), where x

is the axis along the rod, ϕ [0, 2π], 0 ≤ r ≤ R(x) ≤ R

0

, R

0

− const, one

can write the displacement vector

V = (u, w, 0), if torsions are neglected.

Basic equations, describing the nonlinear wave propagation in the initial

87

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

88 Ampliﬁcation of Nonlinear Strain Waves in Solids

Fig. 4.1 Free lateral surface cylindrical rod with varying cross section.

conﬁguration, are obtained from the Hamilton variation principle (3.5).

We are studying long nonlinear longitudinal strain waves (density

waves), that allows to do some simpliﬁcations, namely, the relationships

between longitudinal and transversal displacements u and w. To ﬁnd them

one needs to satisfy the boundary conditions on the free lateral rod surface

r = R(x), namely, the absence of both the normal and tangential stresses

at every moment. We introduce the small parameter ε, taking into account

that the waves under study should be elastic waves with suﬃciently small

magnitude B, B << 1, as well as suﬃciently long waves with the length

L, so as the ratio R

0

/L << 1, where R

0

is the maximal value of r(x)

along the rod . The most important case occurs when both nonlinear and

dispersive features are in balance and small enough:

ε = B =

_

R

0

L

_

2

<< 1. (4.1)

The boundary conditions have the form (3.7)-(3.9). The unknown func-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 89

tions u, w are expanded in a power series (3.12), (3.13). Substituting

them into the boundary conditions and following the procedure explained

in Sec. 3.2, one can obtain

u

L

= U(x, t) +r

2

ν

2

U

xx

, w

L

= −r νU

x

−r

3

ν

2

2(3 −2ν)

U

xxx

. (4.2)

w

NL

= r

_

ν(1 +ν)

2

+

(1 −2ν)(1 +ν)

E

_

l(1 −2ν)

2

+ 2m(1 +ν) −nν

_

_

U

2

x

,

(4.3)

Higher order terms may be obtained by the same way, however, they will

be omitted here because of no inﬂuence on the ﬁnal model equation for

the longitudinal strain waves equation to be derived using the Hamilton

principle (3.5). Substituting expansions (3.12), (3.13) with (4.2), (4.3) into

the formulas for K and Π, one can ﬁnd respectively:

K =

ρ

0

2

_

U

2

t

+νr

2

_

U

t

U

xxt

+νU

2

xt

¸_

, (4.4)

Π =

1

2

_

EU

2

x

+

β

3

U

3

x

+νEr

2

U

x

U

xxx

_

. (4.5)

where β = 3E+2l(1−2ν)

3

+4m(1+ν)

2

(1−2ν) +6nν

2

becomes the only

coeﬃcient depending on the nonlinear elasticity of the rod, it coincides

with that obtained in Sec. 3.2.3. Substituting (4.4), (4.5) into (3.5) and

calculating δS = 0, one can obtain the following nonlinear equation:

U

tt

−

c

2

∗

R

2

∂

∂x

_

R

2

U

x

¸

=

1

R

2

∂

∂x

_

β

2ρ

0

R

2

U

2

x

−

ν

4

∂

∂x

(R

4

U

tt

) +

ν

2

2

R

4

U

xtt

_

+

1

R

2

∂

∂x

_

νc

2

∗

4

_

R

4

U

xxx

+

∂

2

∂x

2

_

R

4

U

x

_

__

−

νR

2

4

U

xxtt

, (4.6)

where c

∗

is the so-called ”rod” wave velocity, c

2

∗

= E/ρ

0

.

Let us consider now the rod which cross section varies slowly along the

x− axis, which is described by a function R = R(γx), γ << 1. Introducing

the notation: v = U

x

, τ = tc

∗

and diﬀerentiating the equation (4.6) on

x, we obtain an equation

v

ττ

−

∂

∂x

1

R

2

∂

∂x

_

R

2

v +

βR

2

2E

(v

2

) +aR

4

v

ττ

−bR

4

v

xx

−4bR

3

R

x

v

x

_

= 0,

(4.7)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

90 Ampliﬁcation of Nonlinear Strain Waves in Solids

with the dispersion terms coeﬃcients a and b, a = −[ν(1 − ν)]/2, b =

−ν/2. Two ﬁrst terms here describe a common linear wave, the third

governs the nonlinearity, and the two following terms are responsible for

dispersive features of the rod, while the last term, being of the same order,

looks like a dissipative one, it occurs due to the cross section variation.

4.1.2 Evolution of asymmetric strain solitary wave

To describe the evolution of a travelling strain wave v we introduce the

phase variable θ and the slow variable X ≡ γx, as follows:

θ

τ

= −1, θ

x

= A(X). (4.8)

Using the asymptotic method explained in Sec. 2.2, the solution of Eq.

(4.7) is obtained in new variables in the power series in γ:

v = v

0

+γv

1

+. . . . (4.9)

Substitution (4.9) into the equation (4.7) gives in leading order of γ the

ODE reduction of the nonlinear double dispersive equation (3.18) for v

0

:

(1 −A

2

)v

0,θθ

−A

2

β

2E

(v

2

0

)

θθ

−R

2

A

4

(a −b)v

0,θθθθ

= 0,

whose solitary wave solution is Samsonov (2001):

v

0

=

3E

β

α cosh

−2

(k(X)[θ −θ

0

(X)]) , (4.10)

depending upon the varying parameter α = α(X), α > 0, while A, k are

expressed through it:

A

2

=

1

1 +α

, k

2

=

α(1 +α)

4R

2

[a(1 +α) −b]

. (4.11)

Since the parameters of the solitary wave depends upon the slow coordinate,

the wave is asymmetric like shown in Fig. 1.18.

Both A and k will be real in Eq.(4.11) for most standard elastic materials

(having the Poisson coeﬃcient ν > 0) if the value of the function α is inside

an interval :

0 < α <

ν

1 −ν

, (4.12)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 91

Then the type of the strain wave (4.10) (compression or tensile one) is

deﬁned only by the sign of the nonlinear coeﬃcient β, which depends on

the elasticity of the rod material, respectively.

Let us study a distortion of the solitary strain wave due to the ”geomet-

rical” inhomogeneity considered. Indeed an inmogeneous linear equation

holds for v

1

at order O(γ)

(1 −A

2

)v

1,θθ

−A

2

β

E

(v

1

v

0

)

θθ

−R

2

A

4

(a −b)v

1,θθθθ

= F, (4.13)

where

F =

R

X

R

A(2v

0,θ

+

β

E

(v

2

0

)

θ

+νR

2

[3(ν −1) + 5A

2

]v

0,θθθ

) +

A

X

(v

0,θ

+

β

2E

(v

2

0

)

θ

+νR

2

[

ν −1

2

+ 3A

2

]v

0,θθθ

) +

A(2v

0,θX

+

β

E

(v

2

0

)

θX

+νR

2

[

ν −1

2

+ 2A

2

]v

0,θθθX

).

Then the equation for the amplitude α arises from the secular term absence

condition (2.43):

_

ln

R

2

α

2

2kA

3

_

X

+

4bk

2

R

2

A

4

5

_

ln 2R

4

α

2

Ak

_

X

= 0, (4.14)

that after use of Eq.(4.11) is reduced to a nonlinear ﬁrst order ODE for an

amplitude variation

R

X

R

= α

X

_

1

6(1 −D +α)

−

1

2α

−

1

3(1 −D

1

+α)

−

1

3(1 −D

2

+α)

_

,

(4.15)

where:

D =

b

a

=

1

(1 −ν)

, D

1,2

=

2 ±

√

9 −5ν

5(1 −ν)

.

Taking the restrictions for α, Eq.(4.12), into account, we conclude that the

expansion in the brackets on the right- hand side of Eq.(4.15) is always

positive. Therefore the unbounded growth of the amplitude of the solitary

wave occurs with the radius decrease, while the increase of the radius is

accompanied by the decay of the amplitude. When the radius no longer

alters at X > X

∗

, see Fig. 4.1, the amplitude and the velocity of the solitary

wave remain constant. The solitary wave evolution in this case is similar

to that shown in Fig. 1.19. However, there is no tendency to a ﬁnite value

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

92 Ampliﬁcation of Nonlinear Strain Waves in Solids

α

∗

in the solutions of Eq.(4.15), and the values of the wave parameters at

X > X

∗

are not ﬁxed by the equation coeﬃcients, they are deﬁned by the

initial condition. Hence this is not a selection of the solitary wave when any

initial conditions provide one and the same solitary waves, see Sec. 1.3.

Direct integration of Eq.(4.15) yields

R

6

α

3

_

ν +α(2ν −6/5) −α

2

(1 −ν)

¸

2

(1 −ν)[ν −α(1 −ν)]

= const. (4.16)

Routine analysis of the functions v

0

, Eq.(4.10), and v

0,x

shows that the

distortion of the wave shape takes place apart from the amplitude variation.

When the bell-shaped solitary wave propagates along the narrowing rod,

its front side becomes steeper while the back one becomes smoother. Vice

versa, the front side of the solitary wave, moving along the expanding rod,

becomes smoother, while the back one - steeper. The equation for the

determination of an extremum of a derivative v

0,x

γ

R

X

R

+[k(1 −γθ

0,X

) +γk

X

[θ −θ

0

(X)]] tanh (k[θ −θ

0

(X)]) = 0. (4.17)

shows that for the wave propagation along the narrowing rod (R

X

< 0)

the extremum is achieved for θ − θ

0

(X) > 0, while in an expanding rod

(R

X

> 0) - for an inverse sign. Then the asymmetric solitary wave

accelerates in the narrowing rod and decelerates in the expanding one in

comparison with the same solitary wave moving along a uniform (homoge-

neous) rod.

The exact formulas (4.11), (4.16) may be easily simpliﬁed to describe

the wave parameters variations. The range of the strain wave amplitude is

restricted by a physical condition of the strain elasticity:

[

_

1 + 2C

xx

−1 [ < e

0

, (4.18)

where e

0

is the yield point of a material, and for most of elastic materials

its value lies in the interval 10

−4

−10

−3

Frantsevich et. al (1982). Therefore

α will have to be small enough, and the following approximations follow

from Eqs. (4.11), (4.16):

A = 1, k

2

=

α

4R

2

(a −b)

,

α

α

0

=

_

R

0

R

_

2

. (4.19)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 93

The most important feature of the solution of Eq.(4.13) is in the appear-

ance of a plateau, propagating behind the solitary wave (4.10) with much

less velocity Samsonov (2001). The amplitude of the plateau is negative

for the narrowing rod and positive when the wave propagates along the

expanding rod.

4.1.3 Experimental observation of the solitary wave ampli-

ﬁcation

We used the same experimental technique as used for the study of the strain

solitary wave propagation in a homogeneous rod, see Sec. 3.3. The choice

of the rod cross sections variation is caused by two reasons. First, we were

going to observe a geometrical inhomogeneity inﬂuence just on the strain

solitary wave, secondly, the experimental setup limitations should be taken

into account. Measurements of the solitary wave amplitude in a homoge-

neous rod resulted in an estimation of the parameter ε = O(10

−3

). When

the inhomogeneity parameter γ is chosen to be γ << ε, then the possible

variation of the initial rod radius (R

0

= 5 mm) at the distance 100 mm

along the axis will be of order 0.1 mm or 2 % from the initial value. The

estimation of the amplitude change in this case by means of an approxima-

tion (4.19) shows that such a magnitude corresponds to the oscillations of

the observed solitary wave front Dreiden et. al (1995). So it seems hardly

possible to detect such a deviation using our experimental setup. Therefore

the inhomogeneity parameter should be chosen as γ >> ε.

It has to be noted, however, that an unsteady process takes place in

experiments in contrast to the quasistationary process governed by the

asymptotic solution obtained above. When γ >> ε the inhomogeneity

will change the initial pulse earlier than both nonlinearity and dispersion,

and the strain solitary wave will hardly appear from an initial shock. Thus

the rod cross section should remain constant at the distance required for

the solitary wave generation and separation, and begin vary only after it.

Experiments on the solitary wave generation in a homogeneous rod, see

Sec. 3.4, showed that a solitary wave appears even at the distance of 60

mm (ca. 10 R

0

) approximately from the input edge of the rod. Based on

this analysis, a rod of 140 mm long was made of polystyrene with uniform

and narrowing parts, as is shown in Fig. 4.2, and two cut oﬀs were made on

the lateral surface for the observation purposes. The rod radius decreases

linearly from the value R

0

= 5 mm to the value R = 2.75 mm along the

distance 70 mm. In this case the inhomogeneity parameter γ = 0.032 is

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

94 Ampliﬁcation of Nonlinear Strain Waves in Solids

Fig. 4.2 Polysteryne rod with variable cross section and cut-oﬀs.

much greater than the typical solitary wave amplitude ≈ 10

−4

Dreiden et. al

(1995) for the homogeneous rod.

The shape of the strain wave was reconstructed by means of Eq.(3.28)

using the following values of parameters: n

0

= 1.33, n

1

= 1.6, Λ = 7

10

−7

m, ν = 0.35. It must be taken into account that light passes the

diﬀerent distances 2h in diﬀerent cross sections. At the interval where the

cross section remains uniform , we have 2h = 2h

0

= 7.75 10

−3

m,

while the measured cross section for the tapered rod part as well as the

inteferograms may be found in Samsonov et. al (1998). The enlargement

Fig. 4.3 Ampliﬁcation of longitudinal strain solitary wave. Two graphs ”strain v vs.

solitary pulse width L” are drawn after interpolation. Solid circles • and the dashed

interpolative line both correspond to experimental data measured on a 40-60 mm interval

of the rod length; open triangles and the solid interpolative line correspond to them

on a 75 - 125 mm interval.

of the amplitude scale allows to visualize the main features of the solitary

wave in the tapered rod, see Fig. 4.3. Some features predicted by our theory

appear in experiments, namely, the increase of the amplitude, the steepness

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 95

of the wave front and smoothness of its back, i.e., asymmetric deformation

of the bell-shaped solitary wave. Moreover, the characteristic width of the

pulse, L

1

= 25, 2 mm, in the homogeneous part of the rod at the one-half

amplitude level is visibly greater than the similar value, L

2

= 22, 3 mm,

in the narrowing part, hence the width of the localized strain solitary pulse

decreases along the tapered rod.

The measurement of the wave amplitude is supposed to be quite plau-

sible for the comparison with the theory. One can see that the maximal

amplitude of the strain solitary wave is achieved at the distances 60 and

95 mm from the rod input edge, respectively. Then from the estimation

(3.28) we obtain the solitary wave magnitudes equal to 3.29 10

−4

in

the interval 40-90 mm, and to 3.83 10

−4

for the interval 75-125 mm.

Therefore the solitary wave magnitude increases 1.16 times. The estima-

tion using the simpliﬁed formulas (4.19), and a length dependence of the

kind R = R

0

−γ(x −70) gives the ampliﬁcation as 1.31 times, which is in

a good agreement with the experimental data.

However, some new theoretical results cannot be checked in experi-

ments, namely:

-The experimental setup does not allow to measure directly the solitary

wave acceleration caused by the narrowing cross section along the rod.

-Like in case of the homogeneous rod, the precise measurement of the

wave width is impossible.

-There is no observation of a plateau.

4.2 Strain solitary waves in an elastic rod embedded in an-

other elastic external medium with sliding

Stresses on the lateral surface of an elastic wave guide, e.g., an elastic rod,

may appear due to its interaction with the surrounding external medium,

as in some technological devices. Various types of contact models can be

used at the interface between the rod and the external medium. The full

(strong) contact model is used when there is continuity of both normal and

shear stresses, and displacements. Alternatively, in a weak contact, friction

may appear at the interface, hence a discontinuity in the shear stresses.

Slippage provides another form of contact at the interface, in which only

the continuity of the normal stresses and displacements is assumed. Surface

stresses may also arise due to the imperfect manufacturing of the lateral

surface of the wave guide and are formally like the ”surface tension” on the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

96 Ampliﬁcation of Nonlinear Strain Waves in Solids

free surface of a liquid Biryukov et. al (1991), Nikolova (1977).

The analytical solution of the contact problem is rather diﬃcult even

in the framework of the linear elasticity theory, see Kerr (1964) and refer-

ences therein. However, considerable progress has been achieved to account

for short nonlinear surface acoustic waves propagating along the interface

between elastic media Parker (1994); Parker and Maugin (1987).

Recently, in the studies of strain waves in a rod interacting with an

elastic external medium, attention was mostly focused on the propagation

ofsurface strain waves along the lateral rod surface perpendicular to its

axis (see, e.g., Gulyaev and Polsikova (1978); Shevyakhov (1977)). Here,

however, we shall consider bulk density strain waves, propagating along the

rod axis. Although rather useful in the study of free lateral surface rods,

the so- called plane cross section hypothesis and Love’s relationship fail

to properly account for contact problems, because they rule out normal

stresses at the rod lateral surface, hence there is discontinuity of normal

stresses at the interface of the rod and the external medium. Most of the

results in this section were ﬁrst published in Porubov et. al (1998).

4.2.1 Formulation of the problem

Let us consider an isotropic, axially inﬁnitely extended, elastic rod sur-

rounded by another albeit diﬀerent elastic medium, in which it may slide

without friction, see Fig. 4.4. We shall consider the propagation of longitu-

dinal strain waves of small but ﬁnite amplitude in the rod. Axi-symmetry

leads to using cylindrical Lagrangian coordinates (x, r, ϕ), where x is the

axis of the rod, ϕ [0, 2π], 0 ≤ r ≤ R. When torsions are neglected, the

displacement vector is

V = (u, w, 0). We choose Murnaghan’s approxima-

tion (3.1) for deformation energy for the rod. The displacement vector for

the linearly elastic external medium may be written as

V

1

= (u

1

, w

1

, 0).

Its density is noted by ρ

1

, and its elastic properties are characterized by

the Lam´e coeﬃcients (λ

1

, µ

1

). Any disturbances due to the wave prop-

agation inside the rod are transmitted to the external medium through

displacements and stresses normal to the rod surface only when contact

with slippage is considered. Disturbances are assumed to decay to zero

in the external medium far from the rod. The normal strains as well as

the displacements inside the rod are smaller than those along the rod axis.

Thus we assume that displacements and strains are inﬁnitesimal in the ex-

ternal medium, hence as already said it is a linear elastic one. Then for the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 97

Fig. 4.4 Cylindrical rod surrounded by an external elastic medium.

external medium we have:

ρ

1

u

1,tt

−(λ

1

+ 2µ

1

) u

1,zz

−(λ

1

+µ

1

)

_

w

1,rz

+

w

1,x

r

_

−

λ

1

_

u

1,rr

+

u

1,r

r

+w

1,rx

+

w

1,x

r

_

= 0 (4.20)

ρ

1

w

1,tt

−(λ

1

+ 2µ

1

)

_

w

1,rr

+

w

1,r

r

−

w

1

r

2

_

−µ

1

w

1,xx

−

(λ

1

+µ

1

) u

1,rx

= 0 (4.21)

The following boundary conditions (b.c.) are imposed:

w → 0, at r → 0, (4.22)

w = w

1

, at r = R, (4.23)

P

rr

= σ

rr

, at r = R, (4.24)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

98 Ampliﬁcation of Nonlinear Strain Waves in Solids

P

rx

= 0, σ

rx

= 0, at r = R, (4.25)

u

1

→ 0, w

1

→ 0 at r → ∞. (4.26)

where P

rr

, P

rx

denote the components of the Piola - Kirchhoﬀ stress

tensor P, see Eqs. (3.10), (3.11). The quantities σ

rr

and σ

rz

are the

corresponding components of the linear stress tensor in the surrounding,

external medium:

σ

rr

= (λ

1

+ 2µ

1

) w

1,r

+λ

1

w

1

r

+ λ

1

u

1,x

(4.27)

σ

rx

= µ

1

(u

1,r

+w

1,x

) (4.28)

The conditions (4.23)-(4.25) deﬁne the sliding contact, while the longitu-

dinal displacements u and u

1

are left free at the interface r = R.

The Piola- Kirchhoﬀ tensor coincides with the linear stress tensor for

inﬁnitesimally small strains. Note that the coeﬃcients in P

rr

and P

rx

depend upon both the second order Lam´e coeﬃcients λ and µ and the

Murnaghan moduli, l, m, n. Hence the tensor P takes into account both

the geometrical and material nonlinearities.

The linear equations (4.20) and (4.21) are solved together with the

boundary conditions (4.23), (4.25), (4.26), assuming that the displacement

w at the interface is a given function of x and t, hence w(x, t, R) ≡ W(x, t).

Then the linear shear stress σ

rr

at the interface r = R is obtained as a

function of W and its derivatives, thus providing the dependence only on

the rod characteristics in the right hand side of the b.c. (4.24). The same

is valid for the elementary work done by external forces at r = R:

δA = 2π

_

∞

−∞

σ

rr

δwdx. (4.29)

Satisfaction of the b.c. on the rod lateral surface yields the relationships

between displacements and strains inside the rod, see Sec. 3.2, allowing

to derive only one nonlinear equation for long longitudinal waves using

Hamilton’s principle (3.29) with the Lagrangian density per unit volume,

L=K − Π, with Π, K and δA deﬁned by Eqs. (3.1), (3.6) and (4.29)

correspondingly.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 99

4.2.2 External stresses on the rod lateral surface

First, the linear problem (4.20), (4.21) will be solved with the boundary

conditions (4.23), (4.25), (4.26). As we focus attention on travelling waves

along the axis of the rod we assume that all variables depend only upon

the phase variable θ = x − ct, where c is the phase velocity of the wave.

Assuming that the unknown functions u

1

, w

1

are

u

1

= Φ

θ

+ Ψ

r

+

Ψ

r

, w

1

= Φ

r

−Ψ

θ

, (4.30)

then Φ and Ψ satisfy the equation:

Φ

rr

+

1

r

Φ

r

+ (1 −

c

2

c

2

l

)Φ

θθ

= 0, (4.31)

Ψ

rr

+

1

r

Ψ

r

−

1

r

2

Ψ + (1 −

c

2

c

2

τ

) Ψ

θθ

= 0, (4.32)

where c

l

and c

τ

are the velocities of the bulk longitudinal and shear linear

waves in the external medium, respectively. They depend on the density

and the Lam´e coeﬃcients, c

2

l

= (λ

1

+ 2µ

1

)/ρ

1

, and c

2

τ

= µ

1

/ρ

1

.

To solve equations (4.31), (4.32)we introduce the Fourier transforms of

Φ and Ψ:

∼

Φ=

_

∞

−∞

Φexp(−k θ) dθ,

∼

Ψ=

_

∞

−∞

Ψexp(−k θ) dθ

that reduces Eqs. (4.31), (4.32) to the Bessel equations :

∼

Φrr

+

1

r

∼

Φr

−k

2

α

∼

Φ= 0, (4.33)

∼

Ψrr

+

1

r

∼

Ψr

−

1

r

2

∼

Ψ −k

2

β

∼

Ψ= 0, (4.34)

with α = 1−c

2

/c

2

l

, and β = 1−c

2

/c

2

τ

. The ratios between c, c

l

and c

τ

deﬁne

the signs of α and β, hence three possible sets of solutions to the equations

(4.33), (4.34) appear, vanishing at inﬁnity due to b.c. (4.26). Using the

boundary conditions (4.23), (4.25),we obtain the following relationships for

the Fourier images of normal stresses at the lateral surface r = R :

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

100 Ampliﬁcation of Nonlinear Strain Waves in Solids

I) when 0 < c < c

τ

:

∼

σ

rr

=

µ

1

∼

W

1 −β

_

2(β −1)

R

+

k(1 +β)

2

K

0

(

√

αkR)

√

α K

1

(

√

αkR)

−

4k

√

βK

0

(

√

βkR)

K

1

(

√

βkR)

_

(4.35)

II) when c

τ

< c < c

l

∼

σ

rr

=

µ

1

∼

W

1 −β

_

2(β −1)

R

+

k(1 +β)

2

K

0

(

√

αkR)

√

α K

1

(

√

αkR)

−

4k

√

βJ

0

(

√

−βkR)

J

1

(

√

−βkR)

_

(4.36)

III) when c > c

l

∼

σ

rr

=

µ

1

∼

W

1 −β

_

2(β −1)

R

+

k(1 +β)

2

J

0

(

√

−αkR)

√

−α J

1

(

√

−αkR)

−

4k

√

βJ

0

(

√

−βkR)

J

1

(

√

−βkR)

_

(4.37)

where J

i

and K

i

(i = 0, 1) denote the corresponding Bessel functions.

We shall see in the next section that in the long wave limit the normal

stress σ

rr

has one and the same functional form at the lateral surface of

the rod in all three cases (4.35)- (4.37). The main diﬀerence in the stress

(and strain) ﬁelds in the external medium is how they vanish at inﬁnity.

This depends on the monotonic decay of K

i

and the oscillatory decay of J

i

when R →∞. Note that the dependence of the strain wave behavior on the

velocities of bulk linear waves, c

l

, c

τ

, is known, in particular, for acoustic

transverse Love waves propagating in an elastic layer superimposed on an

elastic half-space Jeﬀrey and Engelbrecht(1994); Parker and Maugin (1987).

4.2.3 Derivation of strain-displacement relationships in-

side the rod

To solve the nonlinear problem inside the elastic rod, we have to simplify the

relationships between longitudinal and shear displacements u and w follow-

ing the procedure explained in Sec. 3.2. These relationships are obtained,

using conditions on the free lateral surface r = R, namely, the simulta-

neous absence of the tangential stresses and the continuity of the normal

ones. We search for elastic strain waves with suﬃciently small magnitude

B << 1, and a long wavelength relative to the rod radius R, R/ L << 1.

L scales the wavelength along the rod. An interesting case appears when

there is balance between (weak) nonlinearity and (weak) dispersion as for

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 101

a rod with free lateral surface. Then

ε = B =

_

R

L

_

2

<< 1. (4.38)

is the smallness parameter of the problem. The linear part of longitudinal

strain along the rod axis, C

xx

, is u

x

. Then choosing L as a scale along x,

one gets BL as a scale for the displacement u. Similarly, the linear part of

transverse strain, C

rr

, is w

r

. We use the scale BR for the displacement w,

by choosing R as a length scale along the rod radius. Then with [kR[ << 1

in (4.35) - (4.37), we have a power series expansion in kR. It allows to obtain

analytically an inverse Fourier transform for σ

rr

= k

1

r

2

w+k

2

r

−1

w

xx

and

to write the conditions (4.24), (4.25) in dimensionless form at the lateral

surface r = 1 as:

(λ + 2µ) w

r

+ (λ −k

1

) w +λu

x

+

λ + 2µ +m

2

u

2

r

+

ε(

3λ + 6µ + 2l + 4m

2

w

2

r

+ (λ + 2l) ww

r

+

λ + 2l

2

w

2

+ (λ + 2l) u

x

w

r

+

(2l −2m+n) u

x

w + (µ +m) u

r

w

x

+

λ + 2l

2

u

2

x

−k

2

w

xx

) +

ε

2

λ + 2µ +m

2

w

2

x

= O(ε

3

), (4.39)

µu

r

+ε(µw

x

+ (λ + 2µ +m) u

r

w

r

+ 0.5(2λ + 2m−n) u

r

w +

(λ + 2µ +m) u

x

u

r

) +ε

2

(0.5(2m−n) ww

x

+

(µ +m) (w

x

w

r

+u

x

w

x

)) = O(ε

3

) (4.40)

At the rod lateral surface W ≡ w, W

xx

≡ w

xx

. Moreover, for 0 < c < c

τ

:

k

1

= −2µ

1

, k

2

=

µ

1

c

2

(γ −log 2)

c

2

τ

, (4.41)

while for c

τ

< c < c

l

:

k

1

=

2µ

1

(4c

2

τ

−c

2

)

c

2

, k

2

=

µ

1

c

2

τ

c

2

_

1 −

c

2

c

2

τ

+ (2 −

c

2

c

2

τ

)

2

(γ −log 2)

_

,

(4.42)

and for c > c

l

:

k

1

=

2µ

1

[c

2

(c

2

τ

−c

2

l

) + 3c

2

l

c

2

τ

−4c

4

τ

]

c

2

τ

(c

2

l

−c

2

)

, k

2

=

µ

1

c

2

4c

2

τ

. (4.43)

with γ = 0.5772157 Euler’s constant.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

102 Ampliﬁcation of Nonlinear Strain Waves in Solids

The unknown functions u, w will be found in power series of ε:

u = u

0

+εu

1

+ε

2

u

2

+. . . , w = w

0

+εw

1

+ε

2

w

2

+. . . . (4.44)

Substituting (4.44) in (4.39), (4.40), and equating to zero all terms of the

same order of ε, we ﬁnd that the plane cross-section hypothesis and Love’s

relation are valid in the leading order only:

u

0

= U(x, t), w

0

= r C U

x

, (4.45)

with

C =

λ

k

1

−2(λ +µ)

. (4.46)

To order O(ε) we get:

u

1

= − r

2

C

2

U

xx

, w

1

= r

3

DU

xxx

+r QU

2

x

, (4.47)

with coeﬃcients

D =

λ(λ + 2k

2

)

2(k

1

−2(λ +µ))(2(2λ + 3µ) −k

1

)

(4.48)

Q =

λ + 2l + 2C(λ + 4l −2m+n) + 2C

2

(3λ + 3µ + 4l + 2m)

2(k

1

−2(λ +µ))

(4.49)

The higher order terms in the series (4.44) may be obtained in a similar way,

but are omitted here being unnecessary to obtain an evolution equation for

the strain waves.

4.2.4 Nonlinear evolution equation for longitudinal strain

waves along the rod and its solution

Now we can derive the equation for the strain waves along the rod. First

of all, substituting (4.44) into the potential deformation energy density Π

(3.1), one can get in dimensionless form that

Π = a

1

U

2

x

+ε

_

a

2

r

2

U

x

U

xxx

+a

3

U

3

x

¸

+O(ε

2

), (4.50)

with

a

1

=

λ + 2µ

2

+ 2λC + 2(λ +µ)C

2

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 103

a

2

= −

λ + 2µ

2

C −λC

2

+ 4λD + 8(λ +µ) C D,

a

3

=

λ + 2µ

2

+λC +λC

2

+ 2(λ +µ)C

3

+ 2Q[λ + 2(λ +µ)C] +

l

_

1

3

+ 2C + 4C

2

+

8

3

C

3

_

+m

_

2

3

−2C

2

+

4

3

C

3

_

+nC

2

.

For the kinetic energy we have:

K =

ρ

0

2

_

U

2

t

−ε r

2

C(U

t

U

xxt

−CU

2

xt

)

¸

+O(ε

2

) (4.51)

Substituting (4.50), (4.51) and (4.29) into (3.29) and using Hamilton’s vari-

ational principle, we obtain the following equation for a longitudinal strain

wave, v = U

x

:

v

tt

−b

1

v

xx

−ε

_

b

2

v

xxtt

+b

3

v

xxxx

+b

4

(v

2

)

xx

_

= 0, (4.52)

with

b

1

=

2(a

1

−k

1

C

2

)

ρ

0

, b

2

=

C(1 +C)

2

,

b

3

=

a

2

−2C(k

2

C + 2 k

1

D)

ρ

0

, b

4

=

3(a

3

−k

1

C Q)

ρ

0

. (4.53)

Equation (4.52) is nothing but the double dispersive equation (3.18), it

admits, in particular, exact travelling solitary wave solution. Note that

the coeﬃcients depend now upon the wave velocity, c, due to (4.41)- (4.43).

The terms of order O(ε

2

) have been neglected, when deriving equation

(4.52). Therefore we assume c

2

= c

2

0

+ε c

1

+... and consider the coeﬃcients

b

2

− b

4

depending on c

0

only, while the coeﬃcient b

1

may depend also on

c

1

as b

1

= b

10

(c

0

) + ε b

11

(c

0

, c

1

). Then the solitary wave solution has the

form:

v = A m

2

cosh

−2

(mθ) , (4.54)

with

A =

6(b

10

b

2

+b

3

)

b

4

. (4.55)

To leading order the phase velocity is obtained from the equation

c

2

0

= b

10

(c

0

), (4.56)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

104 Ampliﬁcation of Nonlinear Strain Waves in Solids

Table 4.1 Phase velocities of waves in a polystyrene rod embedded indiﬀerent media.

All velocities are measured in 10

−3 m

sec

material c

τ

c

l

c

01

c

02

c

03

model

Quartz 3.78 6.02 2.06 2.1 or 7.15 2.13 or 5.77 I

Iron 3.23 5.85 2.08 2.1 or 6.32 2.11 or 5.15 I

Copper 2.26 4.7 2.07 2.11 or 4.33 2.12 or 3.68 I, II

Brass 2.12 4.43 2.06 2.11 or 4.02 2.12 or 3.45 I, II

Aluminium 3.08 6.26 2.05 2.11 or 5.75 2.13 or 4.97 I, II

Lead 1.09 2.41 2.01 − 1.83 or 2.06 -

and for the function c

1

we get the equation

c

1

= b

11

+ 4k

2

(b

10

b

2

+b

3

), (4.57)

where the wave number k remains a free parameter.

4.2.5 Inﬂuence of the external medium on the propagation

of the strain solitary wave along the rod

Let us estimate the inﬂuence of the external medium on the solitary wave

propagation along the rod. First of all, we have to solve Eq.(4.56) for all

three possible cases (4.41) -(4.43). As ε must not exceed the yield point of

the elastic material (its usual value is less than 10

−3

) we have to compare

with c

l

and c

τ

the values obtained for c

0

, rather than for c.

For the case (4.41), the velocity c

0

is obtained from (4.57) as

c

2

0

=

(3λ + 2µ)µ +µ

1

(λ + 2µ)

ρ

0

(λ +µ +µ

1)

. (4.58)

It appears always higher than the wave velocity in a free rod.

For the model (4.42), Eq. (4.56) yields

c

4

0

−A

1

c

2

0

+A

2

= 0 (4.59)

where

A

1

=

(3λ + 2µ)µ +µ

1

(λ + 2µ) + 4µ

1

ρ

0

c

2

τ

ρ

0

(λ +µ +µ

1)

, A

2

=

4µ

1

c

2

τ

(λ + 2µ)

ρ

0

(λ +µ +µ

1)

.

Finally, for the model (4.43), Eq. (4.56) provides

c

4

0

−B

1

c

2

0

+B

2

= 0 (4.60)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 105

Table 4.2 Phase velocities of waves in a lead rod embedded in diﬀerentexternal media.

All velocities are measured in 10

−3 m

sec

material c

τ

c

l

c

01

c

02

c

03

model

Quartz 3.78 6.02 2.06 2.55 or 4.39 7.51 I,II,III

Iron 3.23 5.85 2.2 2.47 or 4.91 2.73 or4.81 I, II

Copper 2.26 4.7 2.11 − − I

Brass 2.12 4.43 2.08 − − I

Aluminium 3.08 6.26 2.03 − − I

Polystyrene 1.01 2.1 1.83 0.38 or 1.81 1.84 or 2.06 II, III

with

B

1

=

(3λ + 2µ)µc

2

τ

+ (c

2

τ

−c

2

l

)µ

1

(λ + 2µ) + 4µ

1

ρ

0

c

4

τ

+c

2

τ

c

2

l

ρ

0

(λ +µ −3µ

1

)

ρ

0

(c

2

l

µ

1

−c

2

τ

(λ +µ +µ

1

))

B

2

=

c

2

τ

c

2

l

[3µ

1

(λ + 2µ) −µ(3λ + 2µ)] −4µ

1

c

4

τ

(λ + 2µ)

ρ

0

(c

2

l

µ

1

−c

2

τ

(λ +µ +µ

1

))

.

Table 4.1 contains some quantitative estimates for a polystyrene rod and

Table 4.2 for a lead rod, respectively, both embedded in diﬀerent external

media. The quantities c

01

, c

02

and c

03

denote velocities calculated from

Eqs.(4.58), (4.59) and (4.60), respectively. Comparing velocities c

0i

relative

to c

τ

and c

l

we can justify the applicability of cases (4.41)- (4.43). This is

noted by symbols I-III, respectively, in the last column of Tables 4.1 and

4.2. Indeed, the model (4.41) is better for the contact with a polystyrene

rod, while no solitary wave may propagate when the external medium is

lead. However, a solitary wave may propagate along a lead rod embedded

in a polystyrene external medium, as it follows from Table 4.2. Note that

there exist pairs of materials, for which two or even all three models of

sliding contact allow a solitary wave propagation. Thus the balance between

nonlinearity and dispersion may be achieved at diﬀerent phase velocities of

the strain nonlinear waves. This result is of importance when generating

strain solitary waves in a rod embedded in an external elastic medium.

Therefore, strain solitary waves can propagate only with velocities from

the intervals around c

0i

. Note that the solitary wave is a bulk (density)

wave inside the rod and, simultaneously, it is a surface wave for the exter-

nal medium. Then, an important diﬀerence appears relative to long non-

linear Rayleigh surface waves in Cartesian coordinates: in our case more

than one velocity interval exists where solitary waves may propagate. The

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

106 Ampliﬁcation of Nonlinear Strain Waves in Solids

main diﬀerence between modes lies in the diﬀerent rate of wave decay in

the external medium, that follows from the diﬀerent behavior of Bessel’s

functions at large values of their arguments.

Now let us consider the inﬂuence of the type of external medium on

the existence of either compression or tensile longitudinal strain localized

waves. Using the data from Table 4.1 to compute the value of A (4.55) for

a polystyrene rod, it yields that its sign may change according to the values

of the parameters of the material used for the external medium. Therefore

the solitary wave (strain!) amplitude (4.54) may change its sign. The

amplitude is negative for a free lateral surface rod and it remains negative

if the external medium is, say, quartz, brass, copper or iron. However, the

sign changes if c

0

= c

02

and the external medium is aluminium. Therefore,

one can anticipate, in particular, that for a rod embedded in aluminium

an initial pulse with velocity close to c

02

may transform only into a tensile

solitary wave while an initial pulse with velocity close to c

01

evolves to

become a compression solitary wave.

Finally, let us consider the inﬂuence on the sign of c

1

(4.57). For case I,

b

11

= 0, hence the sign is deﬁned by the sign of the quantity (b

10

b

2

+b

3

)/b

4

.

For polystyrene it is, generally, negative for all the external media in Table

4.1, while for a free lateral surface it is positive. Thus, the velocity, c, of

a nonlinear wave in a rod embedded in an external medium is lower than

the linear wave velocity, c

0

, while for a free surface rod nonlinear waves

propagate faster than linear waves. On the other hand, the nonlinear wave

velocity, c, in a polystyrene rod embedded in external medium is higher than

the linear wave velocity for a rod with free lateral surface, c

∗

=

_

E/ρ

0

.

4.2.6 Numerical simulation of unsteady strain wave prop-

agation

Numerical simulation of unsteady nonlinear wave processes in elastic rods

with free lateral surface shows that for A < 0 only initial compression

pulses provide a solitary wave (4.54) or a wave train, while tensile initial

pulses do not become localized and are destroyed by dispersion. On the

contrary, for A > 0 only tensile strain solitary waves may appear, see

Figs. 3.2, 3.3 , and initial compression pulses are destroyed like in Fig. 3.4.

Let us consider now the case when the rod lateral surface is partly free along

the axis and the other part is subjected to a sliding contact with an external

elastic medium, as it is shown in Fig. 4.5. Then the nonlinear strain wave

propagation is described in each part by its own equation (4.52). Matching

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 107

Fig. 4.5 Cylindrical rod partly embedded into an external elastic medium with sliding.

is provided by the continuity of strains and its derivatives. Assume that

for the free surface part ( k

1

= 0, k

2

= 0) A = A

1

, m = m

1

, while for

the embedded one, A = A

2

, m = m

2

. Let the initial solitary wave (4.54)

moves from left to right in Fig. 4.5 far from the embedded part, which is

supposed to be undeformed at the initial time. It was found in Samsonov

(1988) that the mass M conservation in the form

d

dt

M = 0, M =

∞

_

−∞

v dx (4.61)

is satisﬁed by equation (4.52). Then using Eqs.(4.54) and (4.55) we get for

the mass M

1

M

1

= 2 A

1

m

1

, (4.62)

The wave evolution along the embedded part, depends on the ratio

between A

1

and A

2

. Similar to the unsteady processes inside a rod with

the free lateral surface, see Sec. 3.2, an initial strain solitary wave will be

destroyed in the embedded part, if sgnA

2

diﬀers from sgnA

1

. Otherwise

another solitary wave or a wave train will appear. When the initial pulse is

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

108 Ampliﬁcation of Nonlinear Strain Waves in Solids

Fig. 4.6 Ampliﬁcation and reconstruction of a strain solitary wave.

not massive enough one can see in Fig. 3.2, that only one new solitary wave

appears but there is an oscillatory decaying tail. However, the contribution

of the tail to the mass M is negligibly small relative to the solitary wave

contribution, hence

M

2

= 2 A

2

m

2

. (4.63)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 109

Comparing M

1

and M

2

, according to Eq.(4.61) it follows

A

1

m

1

= A

2

m

2

. (4.64)

Fig. 4.7 Attenuation and reconstruction of a strain solitary wave.

Therefore, if A

2

< A

1

the amplitude of the solitary wave increases while

its width, proportional to m

−1

, decreases, hence there is focusing of the soli-

tary wave. On the contrary, when A

2

> A

1

attenuation of the solitary wave

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

110 Ampliﬁcation of Nonlinear Strain Waves in Solids

is provided by the simultaneous decrease of the amplitude and the increase

of the wave width. Numerical simulations conﬁrm our theoretical estimates.

Fig. 4.8 Delocalization and reconstruction of a strain solitary wave.

In Fig. 4.6 the evolution of a strain tensile solitary wave is shown in a rod,

having a central part embedded in an external medium. The value of A

in the central part II, A

2

, is positive but smaller than the value of A

1

in

the surrounding free lateral surface parts I and III, A

1

> A

2

> 0. In the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 111

embedded part II (Fig. 4.6(b)) the solitary wave amplitude exceeds the am-

plitude of the initial solitary wave in Fig. 4.6(a), while its width becomes

narrower than that of the initial wave. Therefore an increase in amplitude

of the elastic strain solitary wave is possible even in an uniformly elastic

rod. This may overtake the yield point inside the elastically deformed rod,

hence the possible appearance of cracks or plasticity zones. In our case

the deformations of the wave front and rear are equal. Hence, the solitary

wave remains symmetric on ampliﬁcation at variance with the strain soli-

tary wave ampliﬁcation in a rod with diminishing cross section. Moreover,

a plateau develops in the tail of the solitary wave in geometrically inhomo-

geneous rod. These diﬀerences could be caused by the absence of mass (and

energy) conservation for strain solitary waves in a narrowing (expanding)

rod.

In the case treated here, the solitary wave does not loose mass, M, hence

its original shape is recovered when traversing part III in Fig. 4.6(c,d). One

can see that an oscillatory tail of the solitary wave in Fig. 4.6(d) is less

pronounced than the tail in Fig. 4.6(c), in agreement with (4.64). Again

there is no solitary wave selection since the parameters of the recovered

wave depends upon that of the original one.

When A

2

> A

1

> 0, an initial tensile strain solitary wave, Fig. 4.7(a), is

drastically attenuated as soon as it enters the embedded area, Fig. 4.7(b),

and its amplitude decreases while its width becomes larger. Again both

the reconstruction of the initial wave proﬁle and the damping of its tail

are observed in the third part of a rod with free lateral surface, part III in

Fig. 4.7(c,d).

Consider now the case of diﬀerent signs of A

i

and assume that A

1

> 0

on both free surface parts. One can see in Fig.4.8 how an initial tensile

solitary wave, Fig. 4.8(a), is destroyed in the embedded part II, Fig. 4.8(b),

in agreement with our previous results on the unsteady processes occurring

for a free surface rod. However, a strain wave is localized again in the third

part of a rod with free lateral surface, Fig. 4.8(c), part III , and ﬁnally

recovers its initial shape in Fig. 4.8(d). Again damping of the tail behind

the solitary wave is observed. Accordingly, both compression and tensile

initial pulses may produce localized strain solitary waves in a rod partly

embedded in an external elastic medium with sliding.

Moreover, the amplitude of the solitary wave generated in such a man-

ner, may be greater than the magnitude of the initial pulse. This case is

shown in Figs. 4.9, 4.10 where A

1

< 0, A

2

> 0 and [A

1

[ < A

2

. One can see

in Fig. 4.9 how an initially localized rectangular tensile pulse, Fig. 4.9(a), is

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

112 Ampliﬁcation of Nonlinear Strain Waves in Solids

Fig. 4.9 Generation of a tensile strain solitary wave train in a rod. The elastic properties

of the rod are chosen such that tensile wave propagation cannot occur in the absence of

contact with an external medium.

destroyed in the free surface part I, Fig.4.9(b). However, a wave train of soli-

tary waves appears, when a destroyed strain wave comes to the embedded

part, Figs. 4.9(c,d). The amplitude of the ﬁrst solitary wave in Fig. 4.9(d)

exceeds the magnitude of the initial rectangular pulse in Fig. 4.9(a). In

the absence of surrounding external medium this rod wave-guide does not

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 113

Fig. 4.10 Delocalization of a strain solitary wave in the absence of external medium.

support tensile solitary wave propagation, and a strain wave is delocalized

as shown in Fig. 4.10.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

114 Ampliﬁcation of Nonlinear Strain Waves in Solids

4.3 Strain solitary waves in an elastic rod with microstruc-

ture

Classic elastic theory cannot account for phenomena caused by the mi-

crostructure of a material. A particular case is a dispersion of strain waves

in an elastic medium. The inﬂuence of microstructure may provide dissipa-

tive eﬀects Cermelli and Pastrone (1997); Engelbrecht and Braun (1998);

Maugin and Muschik (1994), it will be studied in Chapter 6, now consid-

eration is restricted by the non-dissipative case.

The present section refers to the study of nonlinear solitary waves inside

cylindrical rod with microstructure following Porubov (2000). The problem

is solved using the ”pseudo-continuum” Cosserat model and the Le Roux

continuum model. A procedure is developed for derivation of the model

equation for long longitudinal strain waves inside the rod. The inﬂuence

of the microstructure on the solitary wave propagation and ampliﬁcation is

studied.

4.3.1 Modelling of non-dissipative elastic medium with mi-

crostructure

The theory of microstructure has been developed recently, see Engel-

brecht and Braun (1998); Eringen (1968); Mindlin (1964); Nowacki (1975);

Nowacki (1986a) and references therein. Most of results belong to the linear

theory of elasticity, however, there are ﬁndings in the ﬁeld of the nonlinear

theory Engelbrecht and Braun (1998); Eringen (1968). Strain waves were

studied mainly in the linear approximation Eringen (1968); Mindlin (1964);

Nowacki (1975); Nowacki (1986a). Only a few works are devoted to the

nonlinear waves in microstructured non-dissipative media Engelbrecht and

Braun (1998); Erbay et. al (1991); Erofeev and Potapov (1993); Erofeev

(2002); Savin et. al (1973a); Savin et. al (1973b). Waves in elastic wave

guides with microstructure were out of considerable investigation. Also

the values of the parameters characterizing microstructure, are unknown

as a rule, only a few experiments may be mentioned, Savin et. al (1973b);

Potapov and Rodyushkin (2001).

Recall some basic ideas following Eringen (1968). Suppose the macroele-

ment of an elastic body contains discrete micromaterial elements. At any

time the position of a material point of the αth microelement may be

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 115

expressed as

x

(α)

= x +ξ

(α)

,

where x is the position vector of the center of mass of the macroelement,

ξ

(α)

is the position of a point in the microelement relative to the center of

mass. The motion of the center of mass depends upon the initial position

X and time t, x = x(X, t), while for ξ

(α)

the axiom of aﬃne motion is

assumed,

ξ

(α)

= χ

K

(X, t) Ξ

(α)

K

,

where Ξ

(α)

characterizes initial position of a point relative to the center

of mass. Then the square of the arc length is (ds

(α)

)

2

= dx

(α)

dx

(α)

, and

the diﬀerence between the squares of arc length in the deformed and unde-

formed body is

(ds

(α)

)

2

−(dS

(α)

)

2

= (x

k,K

x

k,L

−δ

KL

+ 2x

k,K

χ

kM,L

Ξ

M

+

χ

kM,K

χ

kN,L

Ξ

M

Ξ

N

)dX

K

dX

L

+

2 (x

k,K

χ

kL

−δ

KL

+χ

kL

χ

kM

Ξ

M

) dX

K

dΞ

L

+

χ

kK

χ

kL

dΞ

K

dΞ

L

. (4.65)

where δ

KL

is the Kronecker delta. Let us introduce vector of macrodis-

placements, U(X, t) and tensor of microdisplacements, Φ(X, t),

x

k,K

= (δ

LK

+U

L,K

)δ

kL

,

χ

kK

= (δ

LK

+ Φ

LK

)δ

kL

Then three tensors characterizing the behavior of microstructured

medium follow from (4.65),

C

KL

=

1

2

(U

K,L

+U

L,K

+U

M,K

U

M,L

) ,

E

KL

= Φ

KL

+U

L,K

+U

M,K

Φ

ML

,

Γ

KLM

= Φ

KL,M

+U

N,K

Φ

NL,M

,

where C

KL

is the Cauchy-Green macrostrain tensor , E

KL

is the tensor

of a reference distortion, Γ

KLM

is the tensor of microdistortion. Tensor of

the second rank E

KL

accounts for the microelements motion relative to the

center of mass of the macroelement, while tensor of the third rank Γ

KLM

characterizes relative motion of the microelements of one another.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

116 Ampliﬁcation of Nonlinear Strain Waves in Solids

The density of the potential energy Π should be the function of these

tensors, Π = Π(C

KL

, E

KL

, Γ

KLM

), more precisely upon the invariants of

them. The bulk density of the kinetic energy has the form Mindlin (1964)

K =

1

2

ρ

0

_

U

2

M,t

+J

KN

Φ

KM,t

Φ

NM,t

_

, (4.66)

where ρ

0

is macrodensity of the elastic material, J

KN

is the inertia tensor.

Elastic media with central symmetry posses simpler representation, J

KN

=

J

∗

δ

KN

.

One of the main problem is to deﬁne integrity basis of three tensors

C

KL

, E

KL

, Γ

KLM

Spencer (1971); Eringen and Suhubi (1964). Moreover,

the basic invariants of the third and higher rank tensors have not been

studied. That is why the models were developed based on the additional

assumption on a relationship between U and Φ. One of them is the pseudo-

continuum Cosserat model . According to it

Φ

KL

= −ε

KLM

Φ

M

, Φ

M

=

1

2

ε

MLK

U

K,L

, (4.67)

where ε

KLM

is the alternating tensor. The ﬁrst relationship represents

the classic Cosserat model when only rotations of solid microelements are

possible. The last expression in (4.67) accounts for the pseudo-continuum

Cosserat model when micro-rotation vector Φ coincides with the macro-

rotation vector. In this case the density of the potential energy my be

either Π = Π(C

KL

, Γ

KLM

) or Π = Π(C

KL

, Φ

K,L

) Nowacki (1975); Nowacki

(1986a); Savin et. al (1973b). Tensor E

KL

has the form

E

KL

=

1

2

(U

K,L

+U

L,K

+U

M,K

U

M,L

−U

M,K

U

L,M

) ,

and only linear part of E

KL

coincides with those of C

KL

. Assume the mi-

crostructure is suﬃciently weak to be considered in the linear approximation

Nowacki (1975); Nowacki (1986a); Savin et. al (1973b), and the Murnaghan

model is valid for the macro-motion. Then the density of the potential en-

ergy may be written as

Π =

λ + 2µ

2

I

2

1

−2µI

2

+

l + 2m

3

I

3

1

−2mI

1

I

2

+nI

3

+ 2µM

2

(Φ

K,L

Φ

K,L

+

ηΦ

K,L

Φ

L,K

+βΦ

K,K

Φ

L,L

), (4.68)

where λ and µ are the Lam´e coeﬃcients, (l, m, n) are the third order elastic

moduli, or the Murnaghan moduli, M, η and β are the microstructure

constants, I

p

, p = 1, 2, 3 are the invariants of the tensor C, see Eq.(3.2).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 117

Another simpliﬁed microstructure model was used by some authors, see

Erofeev and Potapov (1993); Mindlin (1964); Savin et. al (1973a). Some-

times it is referred to as the Le Roux continuum Erofeev (2002). According

to it

Φ

KL

= −U

L,K

, Γ

KLM

= −U

L,KM

.

When microstructure is weak and may be considered in the linear ap-

proximation the linear part of E

KL

is zero tensor. It means that there

is no diﬀerence between deformation of elastic microelement and elastic

macrostructure. In this case Π = Π(C

KL

, Γ

KLM

). Assume again the Mur-

naghan model for the macro part of the energy density and use the linear

Mindlin’s model Mindlin (1964) for its micro part one can obtain

Π =

λ + 2µ

2

I

2

1

−2µI

2

+

l + 2m

3

I

3

1

−2mI

1

I

2

+nI

3

+a

1

Γ

KKM

Γ

MLL

+

a

2

Γ

KLL

Γ

KMM

+a

3

Γ

KKM

Γ

LLM

+a

4

Γ

2

KLM

+a

5

Γ

KLM

Γ

MLK

.

(4.69)

where a

i

, i = 1 −5, are the constant microstructure parameters.

4.3.2 Nonlinear waves in a rod with pseudo-continuum

Cosserat microstructure

Let us consider the propagation of a longitudinal strain wave in an isotropic

cylindrical compressible nonlinearly elastic rod, see Fig. 3.1. We take cylin-

drical Lagrangian coordinates (x, r, ϕ) where x is directed along the axis of

the rod, −∞< x < ∞; r is the coordinate along the rod radius, 0 ≤ r ≤ R;

ϕ is a polar angle, ϕ [0, 2π]. Neglecting torsions the displacement vector

is U = (u, w, 0). Then nonzero components of the macrostrain tensor C

are

C

xx

= u

x

+

1

2

(u

2

x

+w

2

x

), C

rr

= w

r

+

1

2

(u

2

r

+w

2

r

), C

ϕϕ

=

w

r

+

w

2

2r

2

,

C

rx

= C

xr

=

1

2

(u

r

+w

x

+u

x

u

r

+w

x

w

r

) . (4.70)

while nonzero components of the rotation tensor Φ

K,L

are

Φ

ϕ,x

= w

xx

−u

rx

, Φ

ϕ,r

= w

xr

−u

rr

. (4.71)

The governing equations together with the boundary conditions are

obtained using the Hamilton variation principle (3.5), where the Lagrangian

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

118 Ampliﬁcation of Nonlinear Strain Waves in Solids

density per unit volume, L=K − Π, with K and Π deﬁned by Eqs.(4.66)

(4.68) correspondingly.

The following boundary conditions (b.c.) are imposed:

w → 0, at r → 0, (4.72)

P

rr

= 0 , at r = R, (4.73)

P

rx

= 0, at r = R, (4.74)

where the components P

rr

, P

rx

of the modiﬁed Piola - Kirchhoﬀ stress

tensor P are deﬁned from (3.5) with (4.66), (4.68), (4.70) and (4.71) being

taken into account:

P

rr

= (λ + 2µ) w

r

+λ

w

r

+ λ u

x

+

λ + 2µ +m

2

u

2

r

+ (λ + 2l) w

r

w

r

+

λ + 2l

2

w

2

r

2

+ (λ + 2l) u

x

w

r

+ (2l −2m+n) u

x

w

r

+

λ + 2l

2

u

2

x

+

λ + 2µ +m

2

w

2

x

+ (µ +m) u

r

w

x

+

3λ + 6µ + 2l + 4m

2

w

2

r

+ 4µM

2

(u

rrx

−w

xxr

), (4.75)

P

rx

= µ (u

r

+w

x

) + (λ + 2µ +m) u

r

w

r

+ (2λ + 2m−n) u

r

w

2r

+

(λ + 2µ +m) u

x

u

r

+

2m−n

2

w

x

w

r

+ (µ +m) w

x

w

r

+

(µ +m) u

x

w

x

+ 4µM

2

[w

xxx

−u

xxr

+

1

r

(r(w

xr

−u

rr

))

r

−

1

2

J

∗

(u

rtt

−w

xtt

)]. (4.76)

Exception of torsions provides transformation of the initial 3D problem

into a 2D one. Subsequent simpliﬁcation is caused by the consideration of

only long elastic waves with the ratio between the rod radius R and typical

wavelength L is R/L ¸ 1. The typical elastic strain magnitude B is also

small, B ¸ 1. Then the procedure from Sec. 3.2.2 is applied to ﬁnd the

relationships between displacement vector components satisfying b.c. on

the lateral surface of the rod (4.73), (4.74) as well as the condition for w

(4.72). An additional parameter γ = M

2

/R

2

is introduced to characterize

the microstructure contribution. Accordingly, the longitudinal and shear

displacement in dimensional form are sought in the form (3.12), (3.13).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 119

Substituting the linear parts u

L

and w

L

(3.12), (3.13) into the b.c. (4.72)

and in the linear parts of b.c. (4.73), (4.74), and equating to zero terms at

equal powers of r one obtains u

k

and w

k

. Using these results the nonlinear

parts u

NL

, w

NL

are similarly obtained from the full b.c. We get

u = U(x, t) +

νr

2

2

1 + 4γ

1 −4γ

U

xx

, (4.77)

w = −νrU

x

−

ν

2(3 −2ν)(1 −4γ)

[ν + 4γ(2 +ν)] r

3

U

xxx

−

(1 +ν)

_

ν

2

+

(1 −2ν)

E

_

l(1 −2ν)

2

+ 2m(1 +ν) −nν

_

_

rU

2

x

, (4.78)

where ν is the Poisson ratio, E is the Young modulus. Other terms from the

series (3.12), (3.13) for i > 3 may be found in the same way, however, they

are omitted here because of no inﬂuence on the ﬁnal model equation for

the strain waves. Substituting Eqs.(4.77), (4.78) into Eq.(3.5), and using

Hamilton’s principle we obtain that longitudinal strains, v = U

x

, obey a

double dispersive nonlinear equation:

v

tt

−α

1

v

xx

−α

2

( v

2

)

xx

+α

3

v

xxtt

−α

4

v

xxxx

= 0, (4.79)

where α

1

= c

2

∗

= E/ρ

0

, α

2

= β/(2ρ

0

), β = 3E + 2l(1 − 2ν)

3

+ 4m(1 +

ν)

2

(1 −2ν) + 6nν

2

, α

3

= ν(1 −ν)R

2

/2,

α

4

=

νER

2

2ρ

0

1 + 4γ

1 −4γ

.

Hence the microstructure aﬀects only dispersion in Eq.(4.77). The soli-

tary wave solution of Eq.(4.77) is

v =

3E

β

_

V

2

c

2

∗

−1

_

cosh

−2

(k (x −V t)), (4.80)

where V is a free parameter while the wave number k is deﬁned by

k

2

=

ρ

0

(V

2

−c

2

∗

)

2νER

2

_

1+4γ

1−4γ

−

(1−ν)V

2

c

2

∗

_. (4.81)

Therefore the contribution of the microstructure results in the widening

of the permitted solitary wave velocities,

1 <

V

2

c

2

∗

<

1

1 −ν

1 + 4γ

1 −4γ

.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

120 Ampliﬁcation of Nonlinear Strain Waves in Solids

Also the characteristic width of the solitary wave proportional to 1/k

becomes larger relative to the wave width in pure elastic case, γ = 0. We

consider γ to be rather small due to the experimental data from Savin

et. al (1973b). Then the type of the solitary wave (compression/tensile) is

deﬁned by the sign of the nonlinearity parameter β like in the case without

microstructure, see Sec. 3.3.

4.3.3 Nonlinear waves in a rod with Le Roux continuum

microstructure

The procedure of obtaining the governing equations is similar to those used

in previous subsection. The nonzero components of the tensor Γ

KLM

are

Γ

xxx

= −u

xx

, Γ

xxr

= Γ

rxx

= −u

xr

, Γ

xrx

= −w

xr

,

Γ

xrr

= Γ

rrx

= −w

xr

, Γ

rxr

= −u

rr

, Γ

rrr

= −w

rr

.

The b.c. (4.73), (4.74) are satisﬁed for the strain tensor components

P

rr

= (λ + 2µ) w

r

+λ(

w

r

+ u

x

) +

λ + 2µ +m

2

u

2

r

+ (λ + 2l) w

r

w

r

+

+

λ + 2l

2

w

2

r

2

+ (λ + 2l) u

x

w

r

+ (2l −2m+n) u

x

w

r

+

λ + 2l

2

u

2

x

+

λ + 2µ +m

2

w

2

x

+ (µ +m) u

r

w

x

+

3λ + 6µ + 2l + 4m

2

w

2

r

+ 2J

∗

(2u

xtt

+w

rtt

) −2a

1

u

xxx

−

2(a

1

+ 2a

2

)w

xxr

−2(a

1

+a

2

)

1

r

(r(w

rr

))

r

−a

1

1

r

(r(u

xr

))

r

, (4.82)

P

rx

= µ (u

r

+w

x

) + (λ + 2µ +m) u

r

w

r

+ (2λ + 2m−n) u

r

w

2r

+

(λ + 2µ +m) u

x

u

r

+

2m−n

2

w

x

w

r

+ (µ +m) w

x

w

r

+

(µ +m) u

x

w

x

+ 2J

∗

u

rtt

−a

1

w

xrr

−2(a

1

+ 2a

2

)u

xxr

−

2a

2

1

r

(r(u

rr

))

r

. (4.83)

Then the approximations for the components of the displacement vector

have the form

u = U(x, t) +

νr

2

2

1

1 −N

U

xx

, (4.84)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux 121

w = −νrU

x

−

4J

∗

(2 −ν)(1 +ν)(1 −2ν)

E(3 −2ν)R

2

r

3

U

xtt

−

ν

2

−(1 −2ν)(1 −N)(G(1 −ν) −2νN)

2(3 −2ν)(1 −N)

r

3

U

xxx

−

(1 +ν)

_

ν

2

+

(1 −2ν)

E

_

l(1 −2ν)

2

+ 2m(1 +ν) −nν

_

_

rU

2

x

, (4.85)

where G = 2a

1

/µR

2

, N = 2a

2

/µR

2

. Like in previous section the governing

equation for longitudinal strain v = U

x

is the double dispersive equation

(4.79) whose coeﬃcients are deﬁned now as

α

1

= c

2

∗

, α

2

=

β

2ρ

0

, α

3

=

νR

2

2(1 −N)

−

ν

2

R

2

2

+ 2J

∗

ν(2 −ν), α

4

=

νc

2

∗

R

2

2(1 −N)

,

while the solitary wave solution has the form

v =

3E

β

_

V

2

c

2

∗

−1

_

cosh

−2

(k (x −V t)), (4.86)

where V is a free parameter, and the wave number k is deﬁned by

k

2

=

(1 −N)ρ

0

(V

2

−c

2

∗

)

2νER

2

[c

2

∗

−V

2

(1 −ν(1 −N) + 4J

∗

(1 −N)(2 −ν)/R

2

)]

. (4.87)

Physically reasonable case corresponds to rather small N, N < 1. Then

the inﬂuence of the microstructure yields an alteration of the permitted

solitary wave velocities interval,

1 <

V

2

c

2

∗

<

1

1 −ν(1 −N) + 4J

∗

(1 −N)(2 −ν)/R

2

.

The widening or narrowing of the interval depends upon the relationship

between N and the parameter of microinertia J

∗

. Again the type of the

solitary wave is governed by the sign of the nonlinearity parameter β. At

the same time the characteristic width of the solitary wave proportional to

1/k turns out smaller than the wave width in a pure macroelastic case,

N = 0, J

∗

= 0.

4.3.4 Concluding remarks

It is to be noted that the assumption of the linear contribution of the

microstructure is correct since its nonlinear contribution, being weaker,

may provide alterations only in the neglected higher-order nonlinear and

dispersion terms in the governing equation both the Cosserat and the Le

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

122 Ampliﬁcation of Nonlinear Strain Waves in Solids

Roux models. Hence we don’t need in the additional nonlinear terms in

the density of the potential energy Π thus avoiding the additional un-

known parameters (like Murnaghan’s third order moduli) describing the

nonlinear contribution of the microstructure Eringen and Suhubi (1964);

Erofeev (2002).

The alterations of the amplitude and the wave width, caused by the

microstructure, have been found in both case under study. The important

result is in the opposite changing of the wave width which gives a possibility

to distinguish the Cosserat and the Le Roux models in possible experiments.

The dispersion caused by the microstructure may be observed experi-

mentally, and numerical data on microstructure parameters my be obtained

Savin et. al (1973b). In experiments on the solitary waves propagation, see

Sec. 3.4, the amplitude and the width of the wave may be measured. There-

fore expressions (4.80), (4.81) provide possible estimation of the parameter

M in the pseudo-continuum Cosserat model. In case of the LeRoux contin-

uum there is an extra parameter J

∗

, see Eqs.(4.86), (4.87), and parameters

N and J

∗

cannot be estimated separately.

The microstructure and the surrounding medium provide similar devi-

ations in the governing double dispersive equation. Hence the analysis in

Sec. 4.2.6 may be used if we consider a rod only part of which contains the

microstructure. Then the ampliﬁcation/attenuation (but not a selection)

of the strain solitary wave occurs similar to that shown in Figs. 4.6-4.10.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 5

Inﬂuence of dissipative (active)

external medium

Now we study the role played by dissipation/energy inﬂux often present in a

realistic case. Dissipative/active eﬀects may be caused by internal features

of the elastic material, hence, an irreversible part should be included into

the stress tensor in addition to the reversible one depending only upon the

density of the Helmholtz energy. Accordingly, the governing equations for

nonlinear strains will contain dissipative/active terms. Dissipation/energy

inﬂux may also occur in an elastic wave guide through phenomena occurring

at or through its lateral surface, and this case is considered further in this

Chapter. Presence of external medium makes a problem more complicated.

However, the Hamiltonian formalism described in Chapter 3 may be applied

since a wave guide remains pure elastic, all dissipative/active factors come

through the elementary work done by the external forces, and therefore

Eq.(3.29) may be used.

5.1 Contact problems: various approaches

External medium aﬀects the lateral surface of a wave guide through the nor-

mal and tangential stresses. In some cases only normal stresses may act like

in the slippage contact. Various contact problems are widely considered,

see, e.g., Galin (1980); Goryacheva (1998); Goryacheva (2001); H¨ahner

and Spencer (1998); Johnson (1985); Kalker et. al (1997); Kerr (1964);

Nikitin (1998). Both elastic and viscoelastic interactions are studied their

in the linear approximation. Main attention is paid to the static loading and

to the interaction forces caused by the relative movement of the contacting

bodies. Of special interest if the contact with friction Goryacheva (1998);

Goryacheva (2001); H¨ahner and Spencer (1998); Kalker et. al (1997);

Nikitin (1998); Stefa´ nski et. al (2000) . Various generalizations of the

123

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

124 Ampliﬁcation of Nonlinear Strain Waves in Solids

Coulomb-Amonton law (dependence on the sliding velocity and the ap-

parent area of contact) are studied Elmer (1997); Goryacheva (2001);

H¨ahner and Spencer (1998); Nikitin (1998); P¨oschel and Herrmann (1993);

Stefa´ nski et. al (2000). However, it was noted in H¨ahner and Spencer

(1998); Nikitin (1998); Stefa´ nski et. al (2000) that the classical Coulomb-

Amonton approach is suﬃcient when the relative velocity of the sliding

bodies is small.

Sometimes the problem of the interaction with an external medium

may be solved directly. It means that we formulate the equation for a

wave guide and the medium and impose the corresponding continuity con-

ditions on the lateral surface of the wave guide. An example may be found

in Sec. 4.2 where dissipationless sliding contact with an elastic external

medium is considered. The diﬃculties of elastic contact stress theory may

arise because the displacement at any point in the contact surface depends

upon the distribution of pressure throughout the whole contact Johnson

(1985). In this case the solution of an integral equation for the pressure

is required. Another problem arises when an external medium is not elas-

tic, and its behavior cannot be described by the equations of elasticity.

The diﬃculties mentioned above may be avoided if the response of a wave

guide is more interesting then the displacements and stresses distribution

in the external medium. In this case the problem reduces to a develop-

ment of a relatively simple foundation models to account for an inﬂuence

of the external medium in terms of the wave guide displacements and/or

strains at the lateral surface. Variety of the foundation models are col-

lected in Kerr (1964). The models are designed replacing an external

medium with interacting spring and dissipative elements. In particular,

when only springs are considered and their shear interactions are assumed

the so-called Winkler-Pasternak model holds Kerr (1964); Pasternak (1954);

Winkler (1867). According to it the pressure p is expressed through the

shear displacement w

p = kw −G∇

2

w,

where ∇

2

is the Laplace operator in x and y, k and G are the constant

foundation moduli. One can see its similarity with the response of the

external medium in case of the sliding contact, see Sec. 4.2.3. Hence foun-

dation models are physically reasonable since they correspond to the results

obtained from the contact problem solutions.

It was Kerr (1964) who developed a viscoelastic foundation model.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 125

Based on the results of footing load tests performed on a snow base

he proposed a viscoelastic model for the interaction between an elas-

tic body and the external snow (or permafrost) medium. He assumed

Newton’s law for the viscous behavior thus including a dissipative ele-

ments besides springs. This model will be studied in the following two

sections where its mathematical expression is presented. Now we only

mention it provides the inﬂuence on a wave guide only through the nor-

mal stresses. Classic case of a dissipative action through the tangential

stresses corresponds to the dry friction contact H¨ahner and Spencer (1998);

Nikitin (1998). Here the foundation model is proposed in Sec. 5.4 when the

external medium aﬀects a lateral surface of an elastic rod both by the nor-

mal and the tangential stresses but dissipative (active) inﬂuence is provided

by the tangential stresses.

5.2 Evolution of bell-shaped solitary waves in presence of

active/dissipative external medium

5.2.1 Formulation of the problem

Let us consider the propagation of a longitudinal strain wave in an isotropic

cylindrical elastic rod embedded in an external medium subjected to Kerr’s

viscoelastic contact model. Most of results were ﬁrst obtained by Porubov

and Velarde (2000)

1

. We take cylindrical Lagrangian coordinates (x, r, ϕ)

where x is directed along the axis of the rod, −∞< x < ∞, r is the radial

coordinate, 0 ≤ r ≤ R, ϕ is a polar angle, ϕ [0, 2π]. Assuming that

torsions can be neglected, then the displacement vector is

V = (u, w, 0).

According to Kerr (1964) the external medium yields a normal stress P

∗

rr

on the lateral surface of the rod r = R :

P

∗

rr

= −

k

r

w −η w

t

+χr

2

w

xxt

, (5.1)

where t is a time, k is the stiﬀness coeﬃcient of the medium, η is the

viscocompressibility coeﬃcient of the external medium, χ is the viscosity

coeﬃcient of the external medium. All three coeﬃcients, k, η and χ, are

positive and constant in framework of the Kerr model, however, we consider

a more general case with the coeﬃcients of either sign.

The evolution of nonlinear waves is obtained in the reference conﬁgu-

ration using Hamilton’s principle Eq.(3.29)), where the Lagrangian density

1

Reprinted with permission from Elsevier Science

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

126 Ampliﬁcation of Nonlinear Strain Waves in Solids

per unit volume, L, is deﬁned as before in the book. Again we choose Mur-

naghan’s approximation (3.1) for the deformation energy. Due to (5.1) the

elementary work, δA, is:

δA = 2π

_

∞

−∞

P

∗

rr

δwdx. (5.2)

The boundary conditions (b.c.) are:

w → 0, at r → 0, (5.3)

P

rr

= P

∗

rr

, at r = R, (5.4)

P

rx

= 0, at r = R, (5.5)

where the components P

rr

, P

rx

of the Piola - Kirchhoﬀ stress tensor P

are deﬁned by Eqs. (3.10), (3.11).

5.2.2 Dissipation modiﬁed double dispersive equation

For the longitudinal and shear displacements inside the rod we assume that:

(i) pure elastic waves have strain magnitude B << 1;

(ii) a characteristic strain wave length L is greater then the rod radius

R, R/L << 1.

Let us obtain the approximations satisfying b.c. (5.3)- (5.5). Accord-

ing to the procedure developed in Chapter 3 the unknown functions u, w

are expanded in a power series (3.12), (3.13). Substituting them into the

boundary conditions and following the approach from Sec. 3.2 we get

u = U +a

2

r

2

U

xx

+a

3

r

3

U

xxt

, (5.6)

w = b

1

r U

x

+b

2

r

2

U

xt

+ r

3

(b

31

U

xxx

+b

32

U

xtt

) +

r

4

(b

41

U

xxxt

+b

42

U

xttt

) +B

1

r U

2

x

+B

2

r

2

U

x

U

xt

. (5.7)

where the explicit forms for the coeﬃcients are given by

a

2

=

λ

2(2(λ +µ) +k)

, a

3

= −

λη

3(2(λ +µ) +k)(3λ + 4µ +k)

,

b

1

= −2 a

2

, b

2

= −3 a

3

, b

31

= −

λa

2

2(2λ + 3µ) +k

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 127

b

32

=

3η a

3

2(2λ + 3µ) +k

, b

41

= −

2χa

2

+η b

31

+λa

3

5λ + 8µ +k

, b

42

= −

η b

32

5λ + 8µ +k

,

B

1

= −

2 b

1

(λ + 4l −2m+n) + 8a

2

2

[3(λ +µ) + 2(2l +m)] +λ + 2l

2(2(λ +µ) +k)

,

B

2

= −

2η B

1

+ 12a

2

a

3

[3(λ +µ) + 2(2l +m)] −3 a

3

(λ + 4l −2m+n)

3λ + 4µ +k

.

Note that a

2

, b

2

, b

42

are always positive, while a

3

, b

1

, b

31

, b

32

are always

negative, while the remaining coeﬃcients may have either sign. Due to (3.1)

we have to truncate the series (5.6), (5.7), hence, omitting nonlinear terms

of order three and higher. Neglecting cubic nonlinear terms we have to

neglect simultaneously the ”corresponding” higher order linear derivative

terms. It was shown in that solitary waves appear as a result of the bal-

ance between nonlinearity and dispersion when B is of order (R/L)

2

. The

”largest” of the cubic terms is rU

3

x

∼ RB

3

∼ R(R/L)

6

. Then, the ”corre-

sponding” comparable linear term is r

5

U

xxxxx

, while quadratic terms are

r

3

U

x

U

xxx

and r

3

(U

xx

)

2

. Similar terms with spatio-temporal mixed deriva-

tives are of the same order. Higher order terms may be added if terms like

I

4

1

are taken into account, see (3.4).

Using Hamilton’s principle we obtain that longitudinal strains, v =

U

x

, obey a dissipation modiﬁed double dispersive nonlinear equation

(DMDDE):

v

tt

−α

1

v

xx

−α

2

v

xxt

−α

3

( v

2

)

xx

−α

4

v

xxxx

+α

5

v

xxtt

−

α

6

(v

2

)

xxt

−α

7

v

xxxxt

+α

8

v

xxttt

= 0. (5.8)

The explicit forms of the coeﬃcients α

i

, i = 1, ..., 8, are

α

1

=

8a

2

2

(2(λ +µ) +k) −8λa

2

+λ + 2µ

ρ

0

, α

2

=

8Rη a

2

2

ρ

0

,

α

3

=

3

ρ

0

[λ

_

1

2

+ 2B

1

−2a

2

(1 + 4B

1

) + 4a

2

2

(1 −4a

2

)

_

+

µ

_

1 −8a

2

B

1

−16a

3

2

_

−4ka

2

B

1

+

l

_

1

3

−4a

2

(1 −4a

2

+

16

3

a

2

2

)

_

+

2

3

m(1 −4a

2

2

(3 + 4a

2

)) + 4na

2

2

],

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

128 Ampliﬁcation of Nonlinear Strain Waves in Solids

α

4

=

R

2

ρ

0

[(λ + 2µ −4λa

2

) a

2

+ 4(λ −2(2λ + 2µ +k) a

2

) b

31

] ,

α

5

=

R

2

2ρ

0

[2ρ

0

a

2

(1 −2a

2

) −8λb

32

+ 9(9λ + 10µ + 4k)a

2

3

+

16(2λ + 2µ +k)a

2

b

32

], α

6

= −

8Ra

2

ηB

1

ρ

0

, α

7

= −

8R

3

a

2

ρ

0

[η b

31

+χa

2

] ,

α

8

=

4R

3

η

ρ

0

_

4a

2

b

32

+ 9a

2

3

¸

.

The coeﬃcients in Eq.(5.8) may take either sign depending upon the mate-

rial properties of the rod and the values of k, η and χ. Higher order terms,

( v

3

)

xx

, v

6x

etc., have been neglected when deriving Eq.(5.8) in accordance

with the early given arguments.

5.2.3 Exact solitary wave solutions of DMDDE

In the moving frame, θ = x−V t, we assume that v = v(θ) and obtain from

(5.8) the following ODE:

(V

2

−α

1

)v+α

2

V v

−α

3

v

2

−(α

4

−α

5

V

2

)v

+α

6

V (v

2

)

+V (α

7

−α

8

V

2

)v

+

P +P

1

θ = 0, (5.9)

where a dash,

, denotes ∂/∂θ; P and P

1

are constants of integration. We

are interested in bell- shaped or kink- shaped solitary waves when v →

const at [ θ [→ ∞, hence, P

1

= 0. It is seen that Eq.(5.9) coincides

with the ODE reduction of the DMKdV Eq.(1.13) obtained in the study of

long waves in surface tension gradient-driven ﬂows Nekorkin and Velarde

(1994); Nepomnyashchy and Velarde (1994). Its most general exact solution

expressed in terms of the Weierstrass elliptic function has been obtained in

Sec. 2.1.3. Here we restrict consideration to its bell-shaped solitary wave

limit,

v = Ak

2

cosh

−2

(k θ) −B, (5.10)

with

A =

6(α

6

α

7

−α

1

α

6

α

8

+α

2

α

3

α

8

)

α

2

6

, B =

A

3

k

2

+

α

2

2α

6

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 129

V

2

= α

1

−

α

2

α

3

α

6

. (5.11)

provided

α

6

(α

4

α

6

−α

3

α

7

) = (α

1

α

6

−α

2

α

3

)(α

5

α

6

−α

3

α

8

). (5.12)

Thus, the elastic features of the rod and the values of the parameters

k, η and χ of the external medium determine the existence of exact bell-

shaped strain wave solution (5.10) as well as the sign of its amplitude,

A, hence, the propagation of either compression or tensile strain solitary

waves. The relationship (5.12) provides simultaneously a balance between

nonlinearity ( v

2

) and dispersion (v

**) and another balance between nonlin-
**

ear active/dissipative ((v

2

)

) and linear active/dissipative (v

) terms .

With Eqs.(5.11), (5.12) taken into account, Eq.(5.9) may be written as

(V

∂

∂θ

−

α

3

α

6

)

_

α

2

v +α

6

v

2

+

α

6

(α

4

−α

5

V

2

)

α

3

v

−

α

6

P

α

3

_

= 0. (5.13)

The ODE reduction of the Boussinesq (or KdV) equation appears in brack-

ets. However, here the wavenumber k is not a free parameter of the solution

(5.10), as it is prescribed by the behaviour as [ θ [→∞. In particular, when

a solution decaying at inﬁnity is considered, then

k

2

= −

α

2

4(α

7

−α

8

V

2

)

, (5.14)

thus implying yet another restriction on the coeﬃcients,

α

8

_

α

1

−

α

2

α

3

α

6

_

−α

7

> 0. (5.15)

Eq. (5.8) also admits kink-type solutions of the form

v = Ak

2

cosh

−2

(kθ) +Dk tanh(kθ) +C, (5.16)

with

A =

6(α

7

−α

8

V

2

)

α

6

, D =

6(α

3

(α

7

−α

8

V

2

) −α

6

(α

4

−α

5

V

2

))

5α

6

,

C =

V

2

−α

1

2α

3

. (5.17)

There are two possibilities for k and V . On the one hand

k =

1

10

¸

¸

¸

¸

α

4

−α

5

V

2

V (α

7

−α

8

V

2

)

−

α

3

α

6

¸

¸

¸

¸

, (5.18)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

130 Ampliﬁcation of Nonlinear Strain Waves in Solids

where V is a solution of the cubic equation,

α

3

3

α

8

V

3

−α

6

(α

2

6

−α

3

3

α

5

)V

2

−(α

3

3

α

7

+α

2

α

3

α

2

6

)V +α

6

(α

1

α

2

6

+α

2

3

α

4

) = 0.

(5.19)

On the other hand,

k

2

=

1

4

_

6α

2

3

α

2

6

−

(V

2

−α

1

)α

6

+α

2

α

3

V

α

3

V (α

7

−α

8

V

2

)

_

, (5.20)

where V is a solution of another cubic equation,

6α

3

α

8

V

3

+α

6

α

7

V

2

−6α

3

α

7

V +α

4

α

6

= 0. (5.21)

Therefore, we have in both cases the kink-shaped solution (5.16) with

prescribed parameters. An initial prestressed state of the rod may trans-

form into a diﬀerent one as the wave passes. The loading or unloading of

the rod depends upon the sign of D, hence on the elastic features of the

rod and on the parameter values of the external medium.

5.2.4 Bell-shaped solitary wave ampliﬁcation and selection

Let us consider the case when the inﬂuence of the external medium is weak,

and all coeﬃcients of the active/dissipative terms in DMDDE (5.8) are

small relative to the other coeﬃcients, i.e., α

2

= ε β

2

, α

6

= ε β

6

, α

7

= ε β

7

,

α

8

= ε β

8

, ε << 1. Then Eq.(5.8) may be written as

v

tt

−α

1

v

xx

−α

3

( v

2

)

xx

−α

4

v

xxxx

+α

5

v

xxtt

=

ε

_

β

2

v

xxt

+β

6

(v

2

)

xxt

+β

7

v

xxxxt

−β

8

v

xxttt

_

. (5.22)

We see that the left hand side of Eq.(5.22) is the double dispersive

equation. Thus at ε = 0, Eq.(5.22) admits an exact bell-shaped solitary

wave solution. For nonzero ε we assume that a solution of Eq.(5.22) is a

function of the phase variable θ and the slow time T, v = v(θ, T), with

θ

x

= 1, θ

t

= −V (T), T = ε t.

Then from (5.22) we get that

(V

2

−α

1

)v

θ

−α

3

( v

2

)

θ

−(α

4

−α

5

V

2

) v

θθθ

=

ε(2V [v

T

+α

5

v

θθT

] +V

T

[v +α

5

v

θθ

] −V

∂

2

∂θ

2

[β

2

v +β

6

v

2

+

(β

7

−β

8

V

2

) v

θθ

]) +O(ε

2

). (5.23)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 131

The solution of Eq.(5.23) is sought using the method explained in

Sec. 2.2.2 in the form

v = v

0

+εv

1

+... (5.24)

Substituting (5.24) into (5.23) in the leading order we have

(V

2

−α

1

)v

0,θ

−α

3

( v

2

0

)

θ

−(α

4

−α

5

V

2

) v

0,θθθ

= 0 (5.25)

The exact solitary wave solution of Eq.(5.25) has the form

v

0

= A(T) cosh

−2

(k(T) θ), (5.26)

with

A =

3(V

2

−α

1

)

2 α

3

, k

2

=

V

2

−α

1

4(α

4

−α

5

V

2

)

. (5.27)

Accordingly, s = V

2

may lie either inside the interval

α

1

< s <

α

4

α

5

, if α

4

−α

1

α

5

> 0, (5.28)

or in

α

4

α

5

< s < α

1

, if α

4

−α

1

α

5

< 0. (5.29)

Only the interval (5.28) is acceptable for a free rod with positive Poisson

ratio. The interval (5.29) exists if the viscocompressibility coeﬃcient is

greater than a given value, η > η

∗

, with

η

∗

=

√

ρ

0

(k + 3λ + 4µ)

√

4k + 9λ + 10µ

.

The correction v

1

(5.24) obeys the inhomogeneous linear equation

(V

2

−α

1

)v

1,θ

−2α

3

(v

0

v

1

)

θ

−(α

4

−α

5

V

2

) v

1,θθθ

= F, (5.30)

where F is

F = 2V [v

0,T

+α

5

v

0,θθT

] +V

T

[v

0

+α

5

v

0,θθ

]−

V

∂

2

∂θ

2

_

β

2

v

0

+β

6

v

2

0

+ (β

7

−β

8

V

2

) v

0,θθ

¸

.

The operator M acting on the function v

1

in Eq.(5.30) is adjoint to the

operator

M

A

= (α

1

−V

2

)∂

θ

+ 2α

3

v

0

∂

θ

+ (α

4

−α

5

V

2

) ∂

3

θ

.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

132 Ampliﬁcation of Nonlinear Strain Waves in Solids

Table 5.1 Selection of solitary wave velocity for the case α

4

−α

1

α

5

> 0.

q

1

s

1q

s

2q

s

0

s

∗

> 0 (α

1

; s

Q

) (α

4

/α

5

; ∞) (α

1

; s

Q

) s

1q

> 0 (α

1

; s

2q

) (s

1q

; s

Q

) (α

1

; s

2q

) s

1q

> 0 (α

1

; s

Q

) (s

Q

; α

4

/α

5

) (α

1

; s

Q

) s

1q

> 0 (α

1

; s

Q

) (s

Q

; α

4

/α

5

) (s

Q

; α

4

/α

5

) s

2q

> 0 (s

Q

; s

2q

) (s

1q

; α

4

/α

5

) (s

1q

; α

4

/α

5

) s

2q

> 0 (0; α

1

) (s

Q

; α

4

/α

5

) (s

Q

; α

4

/α

5

) s

2q

< 0 (0; α

1

) (α

1

; s

Q

) (α

1

; s

Q

) s

2q

< 0 (α

1

; s

2q

) (s

1q

; s

Q

) (s

1q

; s

Q

) s

2q

< 0 (s

Q

; s

2q

) (s

1q

; α

4

/α

5

) (s

Q

; s

2q

) s

1q

< 0 (s

Q

; α

4

/α

5

) (α

4

/α

5

; ∞) (s

Q

; α

4

/α

5

) s

1q

Then using Eq.(5.25) one can obtain the solvability condition (2.43) for

Eq.(5.30),

_

∞

−∞

v

0

F dθ = 0, (5.31)

which yields the equation for the function s,

s

T

Q

3

(s) = s (s −α

1

)

2

(q

1

s

2

+q

2

s +q

3

), (5.32)

with

Q

3

(s) = α

1

α

4

(5α

4

−α

1

α

5

) + 2α

4

(11α

1

α

5

−10α

4

) s−

3α

5

(17α

4

−5α

1

α

5

) s

2

−30α

2

5

s

3

, (5.33)

q

1

= 12α

5

β

6

−5α

3

β

8

, q

2

= 12β

6

(α

4

−α

1

α

5

) +α

3

(7β

2

α

5

−5β

7

+ 5α

1

β

8

),

(5.34)

q

3

= α

3

(7β

2

α

4

+ 5α

1

β

7

) −12α

1

α

4

β

6

.

Important features of the behavior of s may be established analyzing

(5.32) without integration. Note that Q

3

(α

1

) = −15α

1

(α

4

− α

1

α

5

)

2

while

Q

3

(α

4

/α

5

) = α

4

/α

5

(α

4

− α

1

α

5

)

2

, and Q

3

always changes its sign inside

the interval (α

1

, α

4

/α

5

) permitted for s. The most interesting evolution of

s is realized when s tends to the ﬁnite constant value s

∗

as T → ∞. The

values of s

∗

are the solutions of equation

q

1

s

2

+q

2

s +q

3

= 0. (5.35)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 133

First, consider the case α

4

−α

1

α

5

> 0. Then α

1

< s < α

4

/α

5

, and Q

3

cannot have more than one root inside the interval (5.28). Assume that

the root of Q

3

is s

Q

while real roots of Eq.(5.35) are s

1q

< s

2q

. We denote

by s

0

the initial value of s. All possibilities of the solitary wave velocity

selection are collected in Table 5.1. The most interesting case corresponds

to q

1

positive when s

1q

∈ (α

1

; s

Q

), s

2q

∈ (s

Q

; α

4

/α

5

). Table 5.1 shows (see

third and fourth rows) that two diﬀerent solitary waves may be selected

depending on the initial value s

0

. Moreover, any initially initial solitary

wave with velocity s

0

from the permitted interval (5.28) transforms into

a dissipative solitary wave (5.26). Here we are dealing with the selection

of a symmetric solitary wave. It occurs according to Fig. 1.17, the only

diﬀerence is the presence of two thresholds, s

1q

and s

2q

.

When Eq.(5.35) has no multiple root, s

1q

= s

2q

= s

q

, s tends to s

∗

= s

q

at q

1

> 0 if

α

1

< s

q

< s

Q

, α

1

< s

0

< s

q

, or s

Q

< s

q

< α

4

/α

5

, s

q

< s

0

< α

4

/α

5

.

For negative values of q

1

we also have two possibilities

α

1

< s

q

< s

Q

, α

1

< s

0

< s

Q

, or s

Q

< s

q

< α

4

/α

5

, s

Q

< s

0

< s

q

.

The same analysis may be performed for the case α

4

−α

1

α

5

< 0. Under

conditions (5.12) the asymptotic solution (5.26) coincides with the exact

solution (5.10).

We consider here in details only the cases when the velocity tends to

one or another root of Eq.(5.35). Other possibilities corresponding to the

blow-up or the damping of the solitary wave (5.26) may be similarly studied.

There remains the problem of whether solitary wave selection is achieved

in ﬁnite or inﬁnite time. Eq. (5.32) may be integrated in the general case

giving the implicit dependence of s on T. In order to avoid cumbersome

algebra we consider one particular case only, α

1

= 1, β

2

= 1, α

3

= 1, α

4

=

2, α

5

= 1, β

6

= 2, β

7

= 1, β

8

= 1, ε = 0.1. Then the permitted interval for

s is (1, 2) , and Q

3

(s) has only one root, s

Q

= 1.588, inside this interval. For

the roots of Eq.(5.35) we have s

1q

= 1.324, s

2q

= 1.745. Therefore, this is

the case of complete selection when all initially dissipationless solitons with

initial velocity from the interval 1 < s

0

< 1.588 transform into a dissipative

solitary wave with velocity s

∗

= 1.588. When 1.588 < s

0

< 2 they go to a

dissipative solitary wave with s

∗

= 1.745. Integration of Eq. (5.32) yields

exp T =

¸

¸

¸

¸

s

0

−s

1q

s −s

1q

¸

¸

¸

¸

p

1

¸

¸

¸

¸

s

0

−s

2q

s −s

2q

¸

¸

¸

¸

p

2

_

s −1

s

0

−1

_

p

3 _

s

0

s

_

p

4

exp

_

p

5

(s −s

0

)

(s −1)(s

0

−1)

_

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

134 Ampliﬁcation of Nonlinear Strain Waves in Solids

with p

1

= 7.904, p

2

= 0.420, p

3

= 8.980, p

4

= 0.657, p

5

= 4.286. One can

see that s → s

1q

or s → s

2q

at T → ∞. Hence the wave amplitudes tend

to A

1

= 3(s

1q

−α

1

)/2 α

3

or to A

2

= 3(s

2q

−α

1

)/2 α

3

.

Of special interest is the evolution of an arbitrary initial pulse. It cannot

be described by the asymptotic solution. We already noted the similarity

between the governing equation (5.8) and the DMKdV equation (1.13).

For the last equation it was found that the single wave asymptotic solution

accounts for a behaviour of the solitary waves generated from an arbitrary

input, see Sec. 2.3.1. Hence we can anticipate the evolution of the initial

localized strains according to that shown in Figs. 2.4-2.6.

5.2.5 Concluding remarks

We have obtained a nonlinear equation, DMDDE (5.8), governing the evo-

lution of bulk longitudinal long nonlinear strain waves in an elastic rod

immersed inside an active/ dissipative medium. There is an interesting

similarity with the results found for free surface shear long waves in a ther-

moconvective liquid layer described by a dissipation modiﬁed Korteweg-

de Vries equation (DMKdV), see Christov and Velarde (1995); Garazo

and Velarde (1991); Nekorkin and Velarde (1994); Velarde et. al (1995);

Nepomnyashchy and Velarde (1994). Indeed, in the ”travelling wave” limit

the ODE reduction of our DMDDE (5.9) is functionally identical to the cor-

responding ODE reduction of the DMKdV equation Christov and Velarde

(1995); Nekorkin and Velarde (1994). Consequently, all exact travelling

wave solutions for the latter equation are valid in our case. The dynamical

system representation of Eq.(5.9),

.

v = y,

.

y

= z,

.

z = −β z −αv y −ν y −G(v) +P,

with

α =

2α

6

α

7

−α

8

V

2

, β = −

α

4

−α

5

V

2

V (α

7

−α

8

V

2

)

, ν =

α

2

α

7

−α

8

V

2

,

G(v) =

α

1

−V

2

V (α

7

−α

8

V

2

)

v +

α

3

V (α

7

−α

8

V

2

)

v

2

.

coincides with the system studied by Nekorkin and Velarde (1994);

Velarde et. al (1995) when P = 0. Therefore, we can transfer to the lon-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 135

gitudinal strain waves all the results about the existence of pulses, ”bound

solitons” and ”chaotic states” found for the DMKdV equation in Christov

and Velarde (1995); Nekorkin and Velarde (1994); Velarde et. al (1995).

In order to get conservation laws for DMDDE (5.8) we have to rewrite

it as

v

t

= −g

x

,

g

t

= α

1

v

x

+α

2

v

xt

+α

3

( v

2

)

x

+α

4

v

xxx

−α

5

v

xtt

+

α

6

(v

2

)

xt

+α

7

v

xxxt

−α

8

v

xttt

.

Then the time evolution of the wave energy for the solutions vanishing

at ±∞ is governed by the equation

∂

∂t

∞

_

−∞

v g dx = α

2

∞

_

−∞

v

x

g

x

dx +α

6

∞

_

−∞

v

2

x

g

x

dx−

α

7

∞

_

−∞

v

xx

g

xx

dx +α

8

∞

_

−∞

v

xt

g

xt

dx. (5.36)

Thus, instead of energy conservation we have an input-output energy

balance that at the steady state gives a vanishing l.h.s. Indeed, the ﬁrst

term in the right-hand side of (5.36) accounts for the energy input while

energy output is provided by the third term. The second term in (5.36)

may play a stabilizing or a destabilizing role depending on the sign of

α

6

. The last term in (5.36) is absent in the corresponding balance law for

the DMKdV equation Garazo and Velarde (1991); Nepomnyashchy and

Velarde (1994). Here it diminishes the role of the third term in (5.36).

These linear mixed derivative terms in (5.8) appear due to the inﬂuence of

the Poisson eﬀect on the kinetic energy density, K, and on the work, A,

done by external forces. These terms decide the existence of either exact

compression or tensile solitary wave solutions. Moreover, due to Eq.(5.15)

there is no exact solution decaying at [θ[ →∞if α

8

= 0. The corresponding

variation of velocity of the kink-shaped solution depends upon α

8

due to

Eqs.(5.19), (5.21). Both the compression and tensile asymptotic solutions

occur due to the mixed terms with the velocities from the intervals (5.28),

(5.29). Finally, two sets of the selected solitary wave parameters result from

the nonzero coeﬃcient q

1

(5.34) in the equation (5.35).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

136 Ampliﬁcation of Nonlinear Strain Waves in Solids

Our predictions about strain solitary wave selection may help to the

possible experimental generation of active/dissipative solitary waves in a

rod partly embedding in an external medium with the Kerr contact. Indeed,

strain solitary waves may be eﬀectively generated inside a rod with a free

boundary based on the analysis of the exact travelling solitary wave solution

of the governing double dispersive equation ( Eq.(5.8) with k, η and χequal

to zero). Also we have shown that the external medium, e.g., the permafrost

may be responsible for large wave ampliﬁcation. Thus our results permit to

delineate the yield point of the material. The domain of validity of Kerr’s

model could be estimated comparing theory with experiments.

5.3 Strain kinks in an elastic rod embedded in an ac-

tive/dissipative medium

There are two main types of nonlinear solitary waves which could propa-

gate keeping its shape, bell-shaped and kink-shaped solitary waves. The

bell-shaped solitary wave usually appears as a result of a balance be-

tween nonlinearity and dispersion. Nonlinearity in a pure elastic rod is

caused by the ﬁnite stress values and the elastic material properties while

dispersion results from the ﬁnite transverse size of the rod. They are

in balance when, for instance, the strain wave magnitude, B, and the

wave length, L, are such that B = O(R

2

/L

2

) << 1, where R is the

rod radius. Here we address the question of whether besides bell-shaped

solitary waves also kink-shaped waves may propagate in an elastic rod.

The kink-shaped localized traveling structure may be sustained by dif-

ferent balances. There is a balance between cubic nonlinearity and dis-

persion, e.g., resulted in the kink solution of the modiﬁed Korteweg-de

Vries (MKdV) equation case Ablowitz and Segur (1981). Another pos-

sibility occurs when nonlinearity is balanced by dissipation (or accumu-

lation), e.g., the kink solution of the Burgers equation Sachdev (1987);

Whitham (1974). The inclusion of cubic nonlinearity requires to extend the

widely used so-called ”ﬁve constants” Murnaghan energy model, Eq.(3.1),

to a more general ”nine constants” Murnaghan model, Eq.(3.4). In view of

a possible experimental test of our predictions we consider dissipative (ac-

tive) phenomena occurring at the lateral boundary of an otherwise purely

elastic and hence non-dissipative rod in the bulk Kerr (1964). It allows us

to cover both possibilities for longitudinal strain kink propagation. There

is also interest in the analytical study of the simultaneous inﬂuence of dis-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 137

persion, nonlinearity and dissipation (or accumulation) on the evolution,

particularly, ampliﬁcation, of kinks. As we seen before, mathematically the

description of these processes requires inclusion of derivatives of high order

in the model equation. Below we shall use the results obtained in Porubov

and Velarde (2002)

2

.

5.3.1 Formulation of the problem

Again we consider the propagation of a longitudinal strain wave in an

isotropic cylindrical compressible elastic rod embedded in an external

medium subjected to Kerr’s viscoelastic contact on the lateral surface of

the rod r = R :

P

∗

rr

= −

k

r

w −η w

t

, (5.37)

As will be seen later, the third term in Eq.(5.1) now does nor aﬀect the

wave behavior. Again both k and η are assumed to be of either sign.

We choose nine constants Murnaghan’s approximation for the density of

the potential energy, Eq.(3.4). Thus besides the third order elastic moduli,

or the Murnaghan moduli (l, m, n) we are now dealing with the fourth order

moduli (a

1

, a

2

, a

3

, a

4

) account also for nonlinear elastic properties of the

isotropic material. Like Murnaghan’s third order moduli moduli, they can

be either positive or negative.

Otherwise the statement of the problem is similar to that of the previous

section with the exception of the components P

rr

, P

rx

of the Piola -

Kirchhoﬀ stress tensor P that are now written in framework of the nine-

constants theory, Eq.(3.4),

P

rr

= (λ + 2µ) w

r

+λ (u

x

+

w

r

) +

λ + 2µ +m

2

(u

2

r

+ w

2

x

) +

3λ + 6µ + 2l + 4m

2

w

2

r

+ (2l −2m+n) u

x

w

r

+

λ + 2l

2

(u

2

x

+

w

2

r

2

+ 2w

r

w

r

+ 2u

x

w

r

) + (µ +m) u

r

w

x

+

(l + 4a

1

+a

2

)

_

w

3

r

3

+

3ww

2

r

r

+u

3

x

+ 3u

x

w

2

r

_

+ (3m−a

2

)u

r

w

r

w

x

+

2l −2m+n + 24a

1

+ 10a

2

+ 2a

3

+ 4a

4

2

_

u

x

w

2

r

2

+

u

2

x

w

r

+

2u

x

ww

r

r

_

+

2

Reprinted with permission from Elsevier Science

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

138 Ampliﬁcation of Nonlinear Strain Waves in Solids

(2l −µ + 12a

1

+ 4a

2

+ 2a

4

)

_

u

2

x

w

r

+

w

2

w

r

r

2

_

+

(2m−n −2a

2

−a

3

−2a

4

)

u

r

ww

r

2r

+ 2(l + 2m+ 2a

1

)w

3

r

+

(2m+µ −a

2

−a

4

)u

r

u

x

w

x

+

4l −2a

2

−a

3

−2a

4

4

_

u

2

r

w

r

+

ww

2

x

r

_

+

2l + 5m−a

2

2

_

u

2

r

w

r

+w

r

w

2

x

_

+

2l + 2m−a

2

−a

4

2

_

u

2

r

u

x

+u

x

w

2

x

_

,

(5.38)

P

rx

= µ (u

r

+w

x

) + (λ + 2µ +m) (u

r

w

r

+u

x

u

r

) + (2λ + 2m−n) u

r

w

2r

+

2m−n

2

w

x

w

r

+ (µ +m) (w

x

w

r

+u

x

w

x

) +

2m+a

4

4

_

3u

2

r

w

x

+w

3

x

_

+

4l + 2m−n −4µ −2a

2

−2a

3

4

u

r

w

2

+

2m−n −2a

2

−2a

3

4

w

2

w

x

r

2

+

2m−n −2a

2

−a

3

−2a

4

2

_

u

x

w

x

w

r

+

w

r

w

x

w

r

_

+

4m+a

4

4

u

3

r

+

(2m+µ −a

2

−a

4

)u

x

w

r

w

x

+

4m−4µ + 3a

4

4

u

r

w

2

x

+

4l −2a

2

−a

3

−2a

4

2

_

u

x

u

r

w

r

+

u

r

w

r

w

r

_

+ (2l + 2m−a

2

−a

4

)u

x

u

r

w

r

+

2l + 5m−a

2

2

(u

r

u

2

x

+u

r

w

2

r

) +

3m−a

2

2

_

w

x

w

2

r

+u

2

x

w

x

_

. (5.39)

5.3.2 Combined dissipative double-dispersive equation

Besides assumptions (i), (ii) from Sec. 5.2.2, we now assume that (iii)

B ∼ R/L to provide a balance between nonlinearity and dissipation (or

accumulation).

Like before we ﬁrst obtain the relationships between the longitudinal

and shear displacements,

u = U +q

2

r

2

U

xx

, (5.40)

w = b

1

r U

x

+b

2

r

2

U

xt

+ r

3

(b

31

U

xxx

+b

32

U

xtt

) +

+B

1

r U

2

x

+B

2

r U

3

x

+B

3

r

2

U

x

U

xt

. (5.41)

where the explicit forms for the coeﬃcients are given by

q

2

=

λ

2(2(λ +µ) +k)

, b

1

= −2 q

2

, b

2

=

2 q

2

η

3λ + 4µ +k

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 139

b

31

= −

λq

2

2(2λ + 3µ) +k

, b

32

= −

η b

2

2(2λ + 3µ) +k

,

B

1

= −

2 b

1

(λ + 4l −2m+n) + 2b

2

1

[3(λ +µ) + 2(2l +m)] +λ + 2l

2(2(λ +µ) +k)

,

B

2

= −

1

(2(λ +µ) +k)

¦B

1

[λ + 4l −2m+n + 2b

1

(3λ + 3µ + 4l + 2m)] +

1 + 4a

1

+a

2

+

1

2

b

1

[6l −2µ −2m+n + 48a

1

+ 18a

2

+ 2a

3

+ 8a

4

] +

3

2

b

2

1

[2 + 2l −2m+n + 32a

1

+ 12a

2

+ 2a

3

+ 4a

4

] +

b

3

1

[4 −µ + 4l + 4n + 32a

1

+ 8a

2

+ 2a

4

]¦,

B

3

= −

2η B

1

−b

2

¦2q

2

[3(λ +µ) + 2(l +m)] −(λ + 4l −2m+n)¦

3λ + 4µ +k

.

Note that q

2

, b

2

are always positive, while b

1

, b

31

, b

32

are always negative

when the coeﬃcients in (5.37) are positive (Kerr’s model), and the other

coeﬃcients B

i

, i = 1 ÷ 3, have diﬀerent signs. Due to the chosen nine-

constant model we have to truncate the series (5.40), (5.41), hence, omitting

higher order nonlinear terms and the ”corresponding” higher order linear

derivative terms due to the assumption (iii). The ”largest” of the quartic

terms is rU

4

x

∼ RB

4

∼ R(R/L)

4

. Then, the ”corresponding” comparable

linear term is r

4

U

xxxx

, while the cubic term is r

3

U

2

x

U

xxx

. Similar terms

with spatio-temporal mixed derivatives are of the same order.

Substituting (5.40), (5.41) into (3.29), and using Hamilton’s principle

(3.29) we obtain that longitudinal strains, v = U

x

, obey a combined dissi-

pative double-dispersive (CDDD) nonlinear equation:

v

tt

−α

1

v

xx

−α

2

v

xxt

−α

3

( v

2

)

xx

−α

4

v

xxxx

+α

5

v

xxtt

−

α

6

(v

2

)

xxt

−α

7

( v

3

)

xx

= 0. (5.42)

The explicit forms of the coeﬃcients α

i

, i = 1 ÷7, are

α

1

=

8q

2

2

(2(λ +µ) +k) −8λq

2

+λ + 2µ

ρ

0

, , α

2

=

8Rη q

2

2

ρ

0

,

α

3

=

3

ρ

0

[λ

_

1

2

+ 2B

1

−2q

2

(1 + 4B

1

) + 4q

2

2

(1 −4q

2

)

_

+

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

140 Ampliﬁcation of Nonlinear Strain Waves in Solids

µ

_

1 −8q

2

B

1

−16q

3

2

_

−4kq

2

B

1

+

l

_

1

3

−4q

2

(1 −4q

2

+

16

3

q

2

2

)

_

+

2

3

m(1 −4q

2

2

(3 + 4q

2

)) + 4nq

2

2

],

α

4

=

R

2

ρ

0

[(λ + 2µ −4λq

2

) q

2

+ 4b

31

(λ −2(2λ + 2µ +k) q

2

) ] ,

α

5

=

R

2

ρ

0

[ρ

0

q

2

(1 −2q

2

) + 2k(b

2

2

+ 4q

2

b

32

) + (9λ + 10µ)b

2

2

/2 −4¦λ−

4q

2

(λ +µ)¦b

32

],

α

6

= −

8Rq

2

ηB

1

ρ

0

α

7

=

4 kB

2

ρ

0

[B

1

−4q

2

] +

4

ρ

0

¦l[1/2 + 4q

2

(16q

3

2

−8q

2

2

+ 3q

2

−1) +

2B

1

(1 −4q

2

)

2

] +m[1 −4q

2

2

+ 8q

2

(B

1

+ 2q

2

2

)(1 + 2q

2

)] +

2nq

2

[q

2

−4q

2

2

−2B

1

] +λ[1/8 + 2q

2

2

(1 + 4q

2

2

) +B

1

(1 + 2B

1

+ 8q

2

2

) +

2B

2

(1 −4q

2

)] +µ[1/4 + 8q

4

2

+ 2B

1

(B

1

−4q

2

−4q

2

2

) −8q

2

B

2

] +

a

1

[1 −16q

2

+ 96q

2

2

−256q

3

2

+ 256q

4

2

] + 4a

2

q

2

[16q

3

2

−24q

2

2

+ 9q

2

−1] +

4a

3

q

2

2

[1 −4q

2

] + 16a

4

q

2

2

[1 −q

2

]

2

¦.

All coeﬃcients in Eq.(5.42) are always positive in framework of the Kerr

model, with the exception of α

3

and α

6

and α

7

that can be of diﬀerent signs

depending upon the material properties of the rod.

Note that we have obtained Eq.(5.42) in dimensional form without use

of the multiple scales method and hence terms of diﬀerent orders may occur,

simultaneously. Indeed, due to the above given assumptions, (i) and (ii), in

general, the last four terms in (5.42) are smaller than the others and hence

are considered small perturbations to the other four terms. We shall refer

to this case as the weakly dispersive limit. Note that the coeﬃcients α

3

,

α

6

and α

7

depend on the third and fourth elastic moduli. In contrast to

the second order moduli (Lam´e coeﬃcients) they may be of diﬀerent signs

Lurie (1990). Accordingly, their combination in α

3

may be quantitatively

small while not so in α

7

, and then the terms α

3

( v

2

)

xx

and α

7

( v

3

)

xx

may,

quantitatively, be of the same order. Then the dissipative term, α

2

v

xxt

,

α

2

= O(1), will overcome the nonlinearity and drastically alter the wave

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 141

shape before the nonlinearity comes to play. However, if the inﬂuence of the

external medium is weak enough, k and η are small, α

2

<< 1, the signiﬁcant

balance will be between the quadratic-cubic nonlinearities and dispersion,

slightly perturbed by the inﬂuence of dissipative terms. We shall call it

the weakly dissipative limit. Therefore, the advantage of equation (5.42)

is that it embraces diﬀerent important cases, for which we shall below give

exact, asymptotic and numerical solutions.

5.3.3 Exact solutions

Assuming that the solution of Eq.(5.42) depends only upon the phase vari-

able θ = x− c t, then in the moving frame Eq.(5.42) becomes the O.D.E.

v

+β

1

v

+β

2

v +β

3

v

2

+β

4

(v

2

)

+β

5

v

3

= N, (5.43)

where a dash denotes diﬀerentiation with respect to θ; N is a constant, and

β

1

=

α

2

c

α

5

c

2

−α

4

, β

2

=

c

2

−α

1

α

5

c

2

−α

4

, β

3

=

α

3

α

4

−α

5

c

2

,

β

4

=

α

6

c

α

5

c

2

−α

4

, β

5

=

α

7

α

4

−α

5

c

2

.

Eq.(5.43) is a particular case of the equation (2.16) studied in Sec. 2.1.3.

Among exact solutions obtained there we consider two bounded solutions:

(i) kink-shaped solitary wave solution

v = Am tanh(mθ) +B. (5.44)

with

A =

β

4

±

_

β

2

4

−2β

5

β

5

, B = −

β

1

−Aβ

3

2β

4

−3Aβ

5

,

m

2

=

(3A

2

β

5

−4Aβ

4

)(3β

2

β

5

−β

2

3

) + 4β

4

(β

2

β

4

−β

1

β

3

) + 3β

2

1

β

5

2(1 −Aβ

4

)(2β

4

−3Aβ

5

)

2

. (5.45)

(ii) bounded periodic solution

v =

m

√

−β

5

cn(mθ, κ) sn(mθ, κ) dn(mθ, κ)

C

1

+cn

2

(mθ, κ)

−

β

3

3β

5

. (5.46)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

142 Ampliﬁcation of Nonlinear Strain Waves in Solids

with

C

1

=

1 −2κ

2

+

√

κ

4

−κ

2

+ 1

3κ

2

, m

2

=

β

2

+ 3β

2

1

4

√

κ

4

−κ

2

+ 1

,

and the following restrictions on the coeﬃcients:

β

4

= −

1

2

_

−β

5

, β

3

= 3β

1

_

−β

5

. (5.47)

The periodic wave solution (5.46) has a functional form diﬀerent from both

the KdV cnoidal wave and the MKdV bounded periodic solutionAblowitz

and Segur (1981). Note also that the solution (5.46) exists only for non

vanishing β

4

, hence only in the presence of the nonlinear dissipative term

β

4

(v

2

)

**in Eq.(5.43). Equation (5.42), or its equivalent dynamical system,
**

exhibits a more complicated balance between nonlinearity, dispersion and

(linear and nonlinear) dissipation required for the periodic nonlinear wave

than the standard balance between nonlinearity and dispersion that sup-

plies both the KdV and the MKdV periodic solutions. When κ = 1 we

have C

1

= 0, and the solution (5.43) tends to the kink-shaped solution

(5.44) like the MKdV periodic solution, while the KdV cnoidal wave so-

lution becomes the bell-shaped or solitary wave solution in the analogous

limit, see Fig. 1.6.

The dissipationless limit of Eq.(5.43), β

1

= β

4

= 0, is

v

+β

2

v +β

3

v

2

+β

5

v

3

= N. (5.48)

It corresponds to the O.D.E. reduction of the combined KdV-MKdV

(CKdV) equation. The CKdV equation possesses the one-parameter kink

solution (5.44) with (5.45) if β

1

= β

4

= 0. However, a periodic solution of

Eq.(5.48), having a kink limit at κ = 1, has a form diﬀerent from (5.46):

v =

√

2mκ

√

−β

5

sn(mθ, κ) −

β

3

3β

5

,

with m

2

= (3β

2

β

5

− β

2

3

)/(3β

5

(1 + κ

2

)). Moreover, Eq.(5.48) has a variety

of bell-shaped solitary wave solutions.

If α

4

÷α

7

are equal to zero we get from Eq.(5.42) the O.D.E. reduction

of the Burgers equation,

α

2

c v

+ (c

2

−α

1

)v −α

3

v

2

= N. (5.49)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 143

Its kink solution has the functional form like (5.44) but with

A = −

α

2

c

α

3

≡

β

1

β

3

, B =

c

2

−α

1

2α

3

≡ −

β

2

2β

3

, m−free. (5.50)

There are two free parameters, the phase velocity c and the wave number

m. If the boundary conditions are

v →h

1

at θ →∞, v →h

2

at θ →−∞,

then for the kink-shaped solution of Eq. (5.43)

β

4

±

_

β

2

4

−2β

5

β

5

m =

h

1

−h

2

2

,

β

1

−Aβ

3

2β

4

−3Aβ

5

= −

h

1

+h

2

2

while the Burgers model (5.49) gives

m =

(h

1

−h

2

)β

3

2β

1

,

β

2

2β

3

= −

h

1

+h

2

2

.

The coeﬃcients β

i

depend upon the phase velocity c, and the elastic fea-

tures of the material of the rod and the parameters of the external medium.

Hence, for the Burgers model any pair of h

j

deﬁne phase velocity and the

wave number, while in general for the one-parameter solution (5.44), (5.45)

as well as for the MKdV kink solution the boundary conditions imply ad-

ditional restrictions on the parameters of the problem.

5.3.4 Weakly dissipative (active) case

When the viscocompressibility coeﬃcient η is small, i.e. when the external

medium is of little inﬂuence we can take α

2

= δ¯ α

2

, α

6

= δ¯ α

6

, δ << 1.

Then Eq.(5.42) is the perturbed combined double-dispersive equation,

v

tt

−α

1

v

xx

−α

3

( v

2

)

xx

−α

7

( v

3

)

xx

−α

4

v

xxxx

+α

5

v

xxtt

=

δ(¯ α

2

v

xxt

+ ¯ α

6

(v

2

)

xxt

) (5.51)

For nonzero δ we assume that a solution of Eq.(5.51) is a function of

the phase variable θ = x −ct , v = v(θ). Then from Eq.(5.51) we get that

(c

2

−α

1

)v

θ

−α

3

( v

2

)

θ

−α

7

( v

3

)

θ

−(α

4

−α

5

c

2

) v

θθθ

=

−δ c

∂

2

∂θ

2

_

¯ α

2

v + ¯ α

6

v

2

¸

+O(δ

2

). (5.52)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

144 Ampliﬁcation of Nonlinear Strain Waves in Solids

The solution of Eq.(5.52) is sought in the form

v = v

0

+δv

1

+... (5.53)

Substituting (5.53) into (5.52) in the leading order we have the O.D.E.

underlying the CKdV equation,

(c

2

−α

1

)c

0,θ

−α

3

( v

2

0

)

θ

−α

7

( v

3

0

)

θ

−(α

4

−α

5

c

2

) v

0,θθθ

= 0. (5.54)

We look for a solution satisfying the boundary conditions

v

0

→h

±

, v

i

→0, i > 0, at θ →±∞,

and with all derivatives of v

i

with respect to θ vanishing at inﬁnity. For a

kink h

+

,= h

−

.

The exact kink solution of Eq.(5.54) has the form

v

0

= Amtanh(mθ) −α

3

/(3α

7

), (5.55)

with

A =

¸

2(α

5

c

2

−α

4

)

α

7

m, m

2

=

3α

1

α

7

−α

2

3

−3α

7

c

2

6(α

4

−α

5

c

2

)

. (5.56)

The correction v

1

(5.53) obeys the inhomogeneous linear equation

(c

2

−α

1

)v

1,θ

−2α

3

(v

0

v

1

)

θ

−3α

7

(v

2

0

v

1

)

θ

−(α

4

−α

5

c

2

) v

1,θθθ

=

−c

∂

2

∂θ

2

_

¯ α

2

v

0

+ ¯ α

6

v

2

0

¸

, (5.57)

whose solution satisﬁes the b.c. only when

3¯ α

2

α

7

−2α

3

¯ α

6

= 0 (5.58)

and has the form

v

1

=

¯ α

6

mA

2(α

5

c

2

−α

4

)

θ v

0,θ

.

Numerical integration of Eq.(5.51) allows exploration of the kink evo-

lution outside the range imposed by condition (5.58). Case ¯ α

2

< 0, corre-

sponding to dominating damping in the linearized Eq.(5.51) is depicted in

Fig. 5.1 while Fig. 5.2 shows the evolution when ¯ α

2

> 0, hence correspond-

ing to accumulation. In both ﬁgures the steepness of the wave front (i.e.

m = const) and the phase velocity of the initial MKdV kink (5.55), left

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 145

20 40 60 80 100 120 140

x

-1.5

-1

-0.5

0.5

1

v

Fig. 5.1 Numerical simulation of the MKdV kink evolution with weak dissipation.

proﬁle, remain one and the same. The inﬂuence of dissipation (accumu-

lation) shows in the growth experienced by the shelves before and behind

wave fronts, with more dramatic eﬀect in the wave front. The length of

the shelf increases like the length of the shelf behind the perturbed KdV

soliton Ablowitz and Segur (1981) but the height also increases. As a result

no quasistationary proﬁle is possible during the evolution of the perturbed

MKdV kink at variance with the result found for the bell-shaped strain

solitary wave in Porubov and Velarde (2000).

5.3.5 Weakly dispersive case

Let us now take into account all assumptions from Sec. 5.3.2 to obtain

the dimensionless form of Eq.(5.42). Assume the scale for v is B, for x

is L, and for t is L/c

0

where c

0

is a characteristic velocity of the wave.

The small parameter of the problem is ε = B = R/L. Suppose that our

dimensionless solution v depends upon the phase variable θ = x − ct and

that c = 1 +εc

1

+ε

2

c

2

+.... Then from Eq.(5.42) we get

(c

2

0

−α

1

)v

θθ

+ε

_

2c

2

0

c

1

v

θθ

−α

3

( v

2

)

θθ

+α

2

c

0

v

θθθ

_

+

ε

2

(c

2

0

c

1

[2c

2

+c

1

]v

θθ

+α

2

c

0

c

1

v

θθθ

+

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

146 Ampliﬁcation of Nonlinear Strain Waves in Solids

20 40 60 80 100 120 140

x

-1.5

-1

-0.5

0.5

1

v

Fig. 5.2 Numerical simulation of the MKdV kink evolution with weak accumulation

[α

5

c

2

0

−α

4

]v

θθθθ

−α

7

v

3

θθ

+α

6

c

0

v

2

θθθ

) = O(ε

3

). (5.59)

with α

2

= α

2

/R, α

4

= α

4

/R

2

, α

5

= α

5

/R

2

, and α

6

= α

6

/R. The solution

is sought in the form

v = v

0

+ε v

1

+.... (5.60)

The boundary conditions are the same as for the kink solution in the

weakly dissipative case. In the leading order we have c

0

=

√

α

1

, while the

next order yields the equation

∂

2

∂θ

2

_

2α

1

c

1

v

0

−α

3

v

2

0

+α

2

√

α

1

v

0,θ

_

= 0. (5.61)

In brackets we have the O.D.E. reduction of Burgers equation whose

kink-shaped solution is

v

0

= −

α

2

√

α

1

α

3

m tanh(mθ) +

α

1

c

1

α

3

, (5.62)

In the following higher order we get an inhomogeneous linear ordinary dif-

ferential equation for v

1

(θ),

∂

2

∂θ

2

(2α

1

c

1

v

1

−2α

3

v

0

v

1

+α

2

√

α

1

v

1,θ

) =

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 147

[α

4

−α

5

α

1

]v

0,θθθθ

+α

7

v

3

0,θθ

−α

6

√

α

1

v

2

0,θθθ

−α

1

c

1

[2c

2

+c

1

]v

0,θθ

−

α

2

√

α

1

c

1

v

0,θθθ

,

whose solution, decaying at inﬁnity, is

v

1

= v

0,θ

[b

1

θ +b

2

Log(cosh(mθ)) +b

3

] , (5.63)

provided that

c

2

=

1

2α

2

3

c

1

_

α

7

[3α

1

c

2

1

−α

2

2

m] −c

2

1

_

,

where b

3

= const, while b

i

, i = 1, 2, depend upon the coeﬃcients of

Eq.(5.59),

b

1

=

c

1

α

2

α

2

3

_

α

1

[α

2

α

7

−2α

3

α

6

] −α

2

α

2

3

_

,

b

2

=

1

√

α

1

α

2

α

2

3

_

2[α

1

α

5

−α

4

]α

2

3

+α

1

α

2

[2α

3

α

6

−α

2

α

7

]

_

.

We see that b

1

, b

2

remain nonzero even when α

6

= α

7

= 0. The inﬂuence

of dispersion is provided by b

2

only. If b

2

= b

3

= 0 then v may be written

as

v = −

α

2

√

α

1

α

3

m tanh(m[1 +εb

1

]θ) +

α

1

c

1

α

3

+O(ε

2

). (5.64)

Then the ﬁrst term in (5.63) aﬀects the smoothness of the wave front in the

solution (5.60). Shown in Fig. 5.3(a) is the case b

1

> 0, while Fig. 5.3(b)

accounts for negative values of b

1

. The dashed line in Fig. 5.3 accounts for

the unperturbed Burgers kink (5.44), (5.50), while solution (5.64) is shown

with a solid line. Correspondingly, in Fig. 5.4 it is shown the inﬂuence

of the second term (b

1

= b

3

= 0) on the shape of the solution v (5.60)

vs Burgers kink (dashed line) for b

2

< 0 in Fig. 5.4(a) and for b

2

> 0 in

Fig. 5.4(b). Note the asymmetric disturbance of the proﬁle near the upper

and lower states of the solution that are exchanged with the opposite sign

of b

2

. The case b

1

= b

2

= 0 corresponds to a constant phase shift of the

unperturbed Burgers kink solution.

The features of a quasistationary asymptotic solution can be observed

when studying the time-dependent process of the kink formation. Numeri-

cal integration of Eq.(5.42) with an initial condition in the form of a Burgers

kink-shaped wave (5.44), (5.50) shows that the wave attracts the proﬁle de-

scribed by the asymptotic solution even at moderate ε. The time evolution

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

148 Ampliﬁcation of Nonlinear Strain Waves in Solids

5 10 15 20 25 30

x

0.5

1

1.5

2

2.5

v

b

5 10 15 20 25 30

x

0.5

1

1.5

2

2.5

v

a

Fig. 5.3 Smoothness of the Burgers kink proﬁle due to the ﬁrst term in the ﬁrst order

asymptotic solution: (a) b

1

> 0, (b) b

1

< 0.

of the wave at ε = O(1) is shown in Figs. 5.5, 5.6 where the propagation of

the undisturbed Burgers kink is shown with a dashed line; the left proﬁle

corresponds to the initial Burgers kink. The values of the coeﬃcients in

Eq.(5.42) have been chosen such that b

2

, is negative for Fig. 5.5 and positive

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 149

5 10 15 20 25 30

x

0.5

1

1.5

2

2.5

v

b

5 10 15 20 25 30

x

0.5

1

1.5

2

2.5

v

a

Fig. 5.4 Disturbances on the Burgers kink proﬁle due to the second term in the ﬁrst

order asymptotic solution: (a) b

2

< 0, (b) b

2

> 0.

for Fig. 5.6. The three most right solid line proﬁles show that the quasista-

tionary perturbed kink-shaped waves are rather close to the corresponding

asymptotic proﬁles shown in Figs. 5.5, 5.6. The phase velocity remains one

and the same during the whole time and practically equal to the velocity of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

150 Ampliﬁcation of Nonlinear Strain Waves in Solids

20 40 60 80 100 120

x

0.1

0.2

0.3

v

Fig. 5.5 Numerical simulation of the Burgers kink evolution into the quasistationary

proﬁle in the weakly dispersive case at b

2

< 0.

the initial Burgers kink. Any other initial kink diﬀerent from the Burgers

one, diverges even when dispersion is small.

5.3.6 Summary of results and outlook

We have shown that kinks can propagate in a compressible elastic rod, of

radius R, when the wave characteristics, amplitude and wavelength, are

such that B ∼ R/L << 1, a cubic nonlinearity of the elastic material is

taken into account and dissipation exists e.g. due to the rod being em-

bedded in an external viscoelastic medium. We have found in the weakly

dispersive limit how the kink-shaped wave is selected. In contrast to the

bell-shaped solitary wave selection in previous section now only the wave

number tends to a prescribed ﬁnite value, as Eq.(5.64) indicates. Other

possibilities exist for an exact travelling wave solution (5.44) of Eq.(5.42)

as well for the perturbed MKdV kink when the additional condition (5.58)

is satisﬁed.

High order terms do not signiﬁcantly alter the wave structure in the

weakly dispersive case. In such circumstance Eq.(5.42) can be approxi-

mated with α

6

= α

7

= 0 and hence without using the fourth-order elastic

moduli, a

i

. This is very important from the point of view of application

because for most materials the values of a

i

are unknown. However, if dis-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 151

20 40 60 80 100 120

x

0.1

0.2

0.3

v

Fig. 5.6 Numerical simulation of the Burgers kink evolution into the quasistationary

proﬁle in the weakly dispersive case at b

2

> 0.

persion becomes signiﬁcant those terms must be taken into consideration

for a correct description of the strain kinks. In particular they provide the

nontrivial condition 3α

2

α

7

− 2α

3

α

6

= 0, that allows the propagation of a

quasistationary perturbed kink in the weakly dissipative case.

The inﬂuence of dissipation on the balance between nonlinearity and dis-

persion appears aﬀecting wave evolution much more drastically than what

dispersion does to the Burgers nonlinearity-dissipation balance. However,

we have no kink selection in the weakly dissipative (active) case when all

wave parameters tend to ﬁnite values. This seems due to the absence of ”dis-

sipative (active)” cubic nonlinear terms in the r.h.s. of Eq.(5.51). However,

adding a term like (v

3

)

xxt

is not permitted by the assumptions in Sec. 5.3.2.

Indeed, (v

3

)

xxt

and (v

4

)

xx

will be of the same order, and thus the corre-

sponding ”dissipative” quartic nonlinearity demands a simultaneous quintic

nonlinear term, and so on.

Exact solutions obtained account for the case when the dispersion, dis-

sipation and nonlinearity are of the same order. The existence of these

solutions require additional restrictions studied in Sec. 5.3.3. Moreover nu-

merical simulations in the weakly dispersive case point to a formal validity

of the results from Sec. 5.3.5 when dispersion is not small.

As a useful aspect of our analytical study is that obtaining exact and

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

152 Ampliﬁcation of Nonlinear Strain Waves in Solids

asymptotic solutions allow reliable testing points for time-dependent nu-

merical integration. Analytic solutions could be also the starting points for

the numerical search of homoclinic and heteroclinic trajectories yielding

solutions of possibly more complicated form than those described in this

report. On the other hand, we have shown that an external medium, e.g.,

permafrost, may be responsible for the ampliﬁcation of both the strain kink

and bell-shaped solitary waves, see Sec. 5.2. This may be used to locate

zones of potential plasticity. In view of the lack of experimental data on the

fourth order moduli our results may help in ﬁnding these moduli using the

linear algebraic relationships between them and the kink wave amplitude

and velocity obtained using B

2

.

Finally, note that our solutions may also be used to account for the

evolution of surface waves in a heated liquid layer subjected to variations

of surface tension. Indeed, when A/H = H/L << 1, the evolution of

surface waves with surface deformation η of amplitude A and wavelength

L in a layer of depth H is governed by the equation Porubov (1995):

η

t

+γ

1

(η

2

)

x

+γ

2

η

xx

+ε

_

γ

3

η

xxx

+γ

4

(η

2

)

xx

+γ

5

(η

3

)

x

_

= 0, (5.65)

Eq.(5.65) describes waves past an instability threshold. Its O.D.E. reduc-

tion coincides with Eq.(5.43). Note that Eq.(5.65) accounts only for the

weakly dispersive case. Contrary to the elastic bulk dissipation–free rod

inside the liquid layer here the dissipation is caused by ﬂuid viscosity and

heat diﬀusion.

5.4 Inﬂuence of external tangential stresses on strain soli-

tary waves evolution in a nonlinear elastic rod

5.4.1 Formulation of the problem

Let us use the same notations as in previous sections. The elementary work,

δA, done by normal and tangential external forces at the lateral surface of

the rod, r = R, now is:

δA = 2π

_

∞

−∞

(P

∗

rr

δw +P

∗

rx

δu)dx. (5.66)

We assume the external medium yields a normal stress P

∗

rr

,

P

∗

rr

= −

k

1

r

w +k

2

r w

xx

, (5.67)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 153

where k

i

are the stiﬀness coeﬃcients of the external medium, both k

1

and

k

2

are positive and constant. It corresponds to the Pasternak (1954) model

based on the representation of the contact by means of the interacting

spring elements, see also Kerr (1964) and references therein. Tangential

stresses on the lateral surface are assumed in the form:

P

∗

rx

= k

d

P

∗

rr

, (5.68)

that relates to the Coulomb-Amonton law H¨ahner and Spencer (1998);

Nikitin (1998) when k

d

> 0 is a friction coeﬃcient. We consider a more

general model, k

d

is of either sign, in order to account for the inﬂuence of

an active external medium providing an energy inﬂux .

Hence the boundary conditions (b.c.) are:

w → 0, at r → 0, (5.69)

P

rr

= P

∗

rr

, at r = R, (5.70)

P

rx

= P

∗

rx

, at r = R, (5.71)

where the components P

rr

, P

rx

of the Piola - Kirchhoﬀ stress tensor P

are deﬁned by Eqs. (3.10), (3.11).

5.4.2 Derivation of the governing equation

Simpliﬁcations follows from the natural assumptions are similar to those

used in previous sections:

(i) pure elastic strain waves have magnitude B << 1;

(ii) long elastic strain waves have a characteristic length L such that

relative to the rod radius R, R/L << 1.

(iii)B ∼ R

2

/L

2

.

Let us obtain the relationships between longitudinal and shear displace-

ments satisfying b.c. (5.69)- (5.71). According to the procedure developed

in Sec. 3.2 the unknown functions u, w are expanded in a power series

(3.12), (3.13) that yields in our case:

u = U +k

d

a

1

r U

x

+a

2

r

2

U

xx

+k

d

A

1

r U

2

x

, (5.72)

w = b

1

r U

x

+k

d

b

2

r

2

U

xx

+b

3

r

3

U

xxx

+B

1

r U

2

x

. (5.73)

where

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

154 Ampliﬁcation of Nonlinear Strain Waves in Solids

a

1

= −

k

1

µ

b

1

, a

2

= −

[µ

2

(3λ + 4µ +k

1

) + 2 k

2

d

k

2

1

λ]b

1

2µ

2

(3λ + 4µ +k

1

)

,

b

1

= −

λ

2(λ +µ) +k

1

, b

2

=

k

1

λb

1

µ(3λ + 4µ +k

1

)

, b

3

=

(k

2

b

1

−λa

2

)

2(2λ + 3µ) +k

1

,

A

1

= −

k

1

µ

B

1

,

B

1

= −

λ + 2l + 2(λ + 4l −2m+n)b

1

+ 2[3(λ +µ) + 4l + 2m]b

2

1

2(2(λ +µ) +k

1

)

.

Note that a

1

, a

2

, are always positive, b

1

- b

3

are always negative, while

nonlinear term coeﬃcients A

1

, B

1

have diﬀerent but opposite signs. Due to

the chosen ﬁve-constant theory (3.1) we have to truncate the series (5.6),

(5.7) like before in order to omit negligibly small higher-order nonlinear

terms and the ”corresponding” higher order linear derivative terms due to

the assumption (iii).

Substituting Eqs.(5.72), (5.73) into Eq.(3.29) we obtain that longitudi-

nal strains, v = U

x

, obey a nonlinear dispersive-dissipative equation

v

tt

−α

1

v

xx

−α

2

( v

2

)

xx

−α

3

v

xxtt

+α

4

v

xxxx

= k

d

(β

1

v

x

+β

2

(v

2

)

x

+β

3

v

xxx

),

(5.74)

where

α

1

=

λ + 2µ + 4[λ +b

1

(λ +µ)]b

1

+ 2k

1

(b

2

−b

2

1

)

ρ

0

+

3 k

2

1

k

2

d

b

2

1

µρ

0

,

β

1

= −

2k

1

b

1

ρ

0

R

, β

2

= −

2k

1

B

1

ρ

0

R

α

2

=

3

ρ

0

[

1

2

(1 + 2b

1

)(1 + 2b

2

1

) +µ

_

1 + 2b

3

1

_

+

1

3

l

_

1 + 2b

3

1

_

+

2

3

m(1 + 2b

1

)(1 −b

1

)

2

+nb

2

1

+¦λ + 2b

1

(λ +µ) + 2k

1

b

1

¦B

1

+

k

2

1

k

2

d

b

2

1

µ

2

_

(λ +

1

2

m)(1 + 2b

1

) +µ(1 +b

1

) −

1

4

nb

1

−

µ

b

1

B

1

_

],

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 155

α

3

=

R

2

2

(b

2

1

−2a

2

)+

R

2

k

2

1

k

2

d

b

2

1

2µ

2

, β

3

=

2R[k

2

µb

1

−k

1

µ(a

2

b

1

+b

3

) − k

2

1

k

2

d

b

1

b

2

]

µρ

0

α

4

=

R

2

2ρ

0

[a

2

¦4µa

2

−2(λ+2µ) −4b

1

(λ−µ)¦+µb

2

1

−8b

3

¦λ+2b

1

(λ+µ)¦]+

2R

2

b

1

ρ

0

[k

2

µb

1

−k

1

b

3

(µ −k

1

)] +

R

2

k

2

d

2µ

2

ρ

0

[b

2

2

(9λ + 10µ)µ

2

+

k

1

¦b

2

[4µ

2

(a

2

+ 1) + 2µ(µ −3λ)] −4k

2

µb

2

1

¦ +k

2

1

b

1

¦4µb

3

+ (λ + 2µ)b

1

¦].

When k

d

= 0 Eq.(5.74) becomes the double dispersive equation (DDE)

accounting for nonlinear strain waves in a free lateral surface rod.

5.4.3 Symmetric strain solitary waves

Even at k

d

= 0 Eq.(5.74) is nonitegrable by the Inverse Scattering Trans-

form method Ablowitz and Segur (1981). Hence, only particular, usually

travelling wave, exact solutions may be obtained. There exist diﬀerent

approaches based on the assumptions of the appropriate ansatz for a solu-

tions. More general solutions are found in terms of the elliptic functions see

Sec. 2.1. However, up to now only localized strain waves were observed in

experiments, see Sec. 3.4, and main attention is paid here on the exact so-

lutions vanishing at inﬁnity. Assume the solution of Eq.(5.74) depends only

upon the phase variable θ = x− c t, then in the moving frame Eq.(5.74)

becomes the O.D.E.

v

+k

d

γ

1

v

+γ

2

v

+γ

3

vv

+k

d

γ

4

v

2

+k

d

γ

5

v = 0, (5.75)

where a dash denotes diﬀerentiation with respect to θ, and

γ

1

=

β

3

α

3

c

2

−α

4

, γ

2

=

α

1

−c

2

α

3

c

2

−α

4

, γ

3

=

2α

2

α

3

c

2

−α

4

,

γ

4

=

β

2

α

3

c

2

−α

4

, γ

5

=

β

1

α

3

c

2

−α

4

.

Following the procedure from Sec. 2.1. we compare the leading-order

derivative term, v

, and the nonlinear term, vv

**. One can see they are in
**

balance when a solution has the second order pole. Hence, general periodic

solution may be expressed using either the Weierstrass elliptic function , v =

a

1

℘(θ, g

1

, g

2

) + b

1

, or the Jacobian elliptic function, v = a

2

cn

2

(pθ, κ) +b

2

.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

156 Ampliﬁcation of Nonlinear Strain Waves in Solids

Localized solitary wave solution are of interest here. Since they correspond

to the limit κ →1, one can assume the solution of the form:

v = Gcosh

−2

(pθ). (5.76)

Substituting (5.76) into Eq.(5.75) and equating to zero coeﬃcients at

corresponding powers of tanh(pθ) one obtains

p

2

=

c

2

−α

1

4(α

3

c

2

−α

4

)

, G =

3(c

2

−α

1

)

2α

2

, (5.77)

2γ

4

= γ

1

γ

3

, γ

5

= γ

1

γ

2

. (5.78)

The ﬁrst of the conditions (5.78) deﬁnes the phase velocity,

c

2

=

α

4

β

2

+α

2

β

3

α

3

β

2

,

while the second one imposes a restriction on the equation coeﬃcients,

α

2

α

3

β

1

+ (α

4

−α

1

α

3

)β

2

+α

2

β

3

= 0.

When conditions (5.78) hold Eq. (5.75) may be rewritten as

[

∂

∂θ

+k

d

γ

1

](v

+γ

2

v +

1

2

γ

3

v

2

) = 0,

whose solution vanishing at inﬁnity is obtained from the equation

v

+γ

2

v +

1

2

γ

3

v

2

= 0. (5.79)

Eq. (5.79) is nothing but ODE reduction of the DDE integrated twice.

Its solitary wave solution is (5.76), (5.77), while the phase velocity c remains

a free parameter. Hence the general solution (5.76), (5.77), (5.78) is sup-

ported simultaneously by a balance between nonlinearity, γ

3

vv

, and disper-

sion, v

**(like DDE’s solution), and by a balance between active/dissipative
**

terms, k

d

γ

1

v

, k

d

γ

4

v

2

, k

d

γ

5

v. The last balance is realized under conditions

(5.78).

The coeﬃcients α

i

, β

j

depend on the elastic features of the material

of the rod and the parameters of the external medium. They deﬁne the

sign of the wave amplitude G, hence either compression, G < 0, or tensile,

G > 0, wave may propagate in the rod.

Solitary wave solution exists under speciﬁc initial condition in the form

of (5.76) at t = 0. It is known that in non-dissipative case even rather

arbitrary initial pulse splits into the train of solitary waves each being

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 157

accounted for the travelling solitary wave exact solution of the DDE, see

Sec. 3.3. When dissipation/accumulation predominates it destroys initial

pulse before balance between nonlinearity and dispersion become to play,

and no localized waves appear.

Consider now a weakly dissipative case, k

d

<< 1, when nonlinearity and

dispersion dominate over dissipation/accumulation. Assume a solution of

Eq.(5.74) depends upon the phase variable θ and the slow time T, v =

v(θ, T), with

θ

x

= 1, θ

t

= −c(T), T = k

d

t.

Then asymptotic solution of Eq.(5.74) is sought in the form

v = v

0

(θ) +k

d

v

1

(θ) +...

In the leading order we get

(c

2

−α

1

)v

0,θ

−(α

3

c

2

−α

4

)v

0,θθθ

−2α

2

v

0

v

0,θ

= 0, (5.80)

whose one-parameter solution has the form of (5.76), but c now depends

upon T. At order O(k

d

) there is an inhomogeneous linear equation for v

1

,

(c

2

−α

1

)v

1,θ

−(α

3

c

2

−α

4

)v

1,θθθ

−2α

2

(v

0

v

1

)

θ

= F, (5.81)

where

F = 2cv

0,T

+c

T

v

0

−2α

3

cv

0,θθT

−α

3

c

T

v

0,θθθ

+β

1

v

0

+β

2

v

2

0

+β

3

v

0

.

The operator acting on the function v

1

in Eq.(5.81) is adjoint to that in

Eq.(5.80). We look for a solution satisfying the boundary conditions

v

i

→0, i _ 0, at [θ[ →∞, (5.82)

Then the Fredholm alternative,

_

∞

−∞

v

0

F dθ = 0, (5.83)

provides the absence of secular terms. It yields the equation deﬁning the

phase velocity c,

α

2

c

T

Q

6

(c) = (α

1

−c

2

)(α

3

c

2

−α

4

)(6α

3

β

2

c

4

+ [5α

2

α

3

β

1

−

6β

2

(α

4

+α

1

α

3

) −α

2

β

3

]c

2

−5α

2

α

4

β

1

+

6α

1

α

4

β

2

+α

1

α

2

β

3

). (5.84)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

158 Ampliﬁcation of Nonlinear Strain Waves in Solids

where

Q

6

(c) = 30α

2

3

c

6

−3α

3

(11α

4

+ 10α

1

α

3

)c

4

+ 2α

4

(11α

1

α

3

+ 10α

4

)c

2

−

α

2

1

α

3

α

4

−5α

1

α

2

4

.

Eq. (5.84) may be integrated giving the dependence c on T in an implicit

form. However, important features of the evolution of c may be established

studying the sign of c

T

without integration. The stationary solutions of

(5.84) c

1

=

√

α

1

and c

2

=

_

α

4

/α

3

correspond to p = 0 or p → ∞ in

(5.77). Other stationary solutions are deﬁned from the equation

6α

3

β

2

c

4

+ [5α

2

α

3

β

1

−6β

2

(α

4

+α

1

α

3

) −α

2

β

3

]c

2

−5α

2

α

4

β

1

+

6α

1

α

4

β

2

+α

1

α

2

β

3

= 0. (5.85)

Due to (5.77) (α

1

− c

2

)(α

3

c

2

− α

4

) < 0, and the sign of c

T

depends upon

the signs of the polynomial Q

6

(c) and the quartic polynomial in the r.h.s.

of Eq. (5.84). When c

T

keeps its sign as time passes possible scenarios for

evolution of c are either vanishing or diverging to inﬁnity. More interesting

case is realized when, in particular, real roots of Eq.(5.85), c

∗

1

< c

∗

2

, are

located in the interval, q

1

< c

∗

1

< c

∗

2

< q

2

, where q

1

, q

2

are real neighboring

roots of Q

6

(c). If α

3

β

2

> 0 and Q

6

(c) > 0 at q

1

< c < q

2

, we have c

T

> 0

when c

∗

1

< c < c

∗

2

and c

T

< 0 when q

1

< c < c

∗

1

or c

∗

2

< c < q

2

. It means

that if an initial velocity, c(T = 0) ≡ c

0

lies in the intervals c

∗

1

< c

0

< c

∗

2

or

c

∗

2

< c < q

2

, the velocity c(T) tends to the ﬁnite value c

s

= c

∗

2

at T → ∞.

If Q

6

(c) < 0, c tends to c

s

= c

∗

1

. To put this another way, the value of

velocity is selected according to the governing equation coeﬃcients. The

wave amplitude G(c) is selected like velocity c according to (5.77). It was

called in Sec. 1.3 the selection from below when c

0

< c

s

, Fig. 1.17 (a), while

the case c

0

> c

s

is referred to by the selection from above, Fig. 1.17(b).

We see the waves remain symmetric with respect to their maximums, their

amplitudes increase (decrease) while their widths decrease (increase).

Based on Eq. (5.84) the solution of Eq.(5.81) is

v

1

=

C(α

1

−c

2

)

2α

2

[1 −tanh(pθ)] + [C +

(α

1

α

3

−α

4

)c

2

c

T

(α

1

−c

2

)

2

(α

3

c

2

−α

4

)

θ]v

0

+

[C

1

+

C

2

θ −

(α

1

α

3

−α

4

)c

2

c

T

(α

1

−c

2

)

2

(α

3

c

2

−α

4

)

θ

2

]v

0,θ

,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 159

where C

1

is a constant,

C = 12p[α

2

3

β

2

c

8

+α

3

(5α

2

α

3

c

T

−2α

1

α

3

β

2

−2α

4

β

2

−α

2

β

3

)c

6

+

(β

2

¦α

2

1

α

2

3

+ 4α

1

α

3

α

4

+α

2

4

¦ +α

2

β

3

¦2α

1

α

3

+α

4

¦ −

11α

2

α

3

α

4

c

T

)c

4

+ (α

2

α

4

c

T

¦2α

1

α

3

+ 5α

4

¦ −

α

1

¦2α

4

β

2

[α

1

α

3

+α

4

] +α

2

β

3

[α

1

α

3

+ 2α

4

]¦)c

4

−

α

2

1

α

4

(α

4

β

2

+α

2

β

3

)]/¦5α

2

(α

1

−c

2

)

3

(α

3

c

2

−α

4

)¦

One can check that evolution of the solution v = v

0

(θ, T) + k

d

v

1

(θ, T)

does not signiﬁcantly diﬀer from that of the solution v = v

0

(θ, T). How-

ever, it does not vanish at θ → −∞, and a plateau appears behind the

solitary wave. The standard matching asymptotic procedure should be

used to complete the uniformly valid asymptotic solution satisfying bound-

ary conditions (5.82), see Ablowitz and Segur (1981) for details. Absence

of plateau requires additional restrictions on the equation coeﬃcients. Note

that neither exact solitary wave solution (5.76) nor asymptotic selected soli-

tary wave solution exists at β

2

= 0, β

3

= 0, and two values of the selected

wave parameters exist thanks to the mixed dispersion term in Eq.(5.74),

when α

3

,= 0.

5.4.4 Evolution of asymmetric solitary waves

Another asymptotic solution may be found when k

d

<< 1 is considered but

now it is assumed that a solution of Eq.(5.74) is a function of the phase

variable θ and the slow coordinate X, v = v(θ, X), with

θ

x

= P(X), θ

t

= −1, X = k

d

x.

Then from (5.74) we get that

(1 −α

1

P

2

)v

θ

−2α

2

v v

θ

−(α

3

−α

4

P

2

) v

θθθ

= k

d

(2P[α

1

v

X

+α

2

v

2

X

+

(α

3

−2α

4

P

2

)v

θθX

]+

P

X

[α

1

v+α

2

v

2

θ

+(α

3

−6α

4

P

2

)v

θθ

]+P[β

1

v+β

2

v

2

+p

2

β

3

v

θθ

])+O(k

2

d

). (5.86)

The solution of Eq.(5.86) is sought in the form

v = v

0

(θ, X) +k

d

v

1

(θ, X) +..., (5.87)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

160 Ampliﬁcation of Nonlinear Strain Waves in Solids

and satisﬁes boundary conditions (5.82). Substituting (5.87) into (5.86) we

have a nonlinear O.D.E. for v

0

in the leading order,

(1 −α

1

P

2

)v

0,θ

−2α

2

v

0

v

0,θ

−(α

3

−α

4

P

2

) v

0,θθθ

= 0. (5.88)

Its solution,

v

0

=

3(1 −α

1

P

2

)

2α

2

P

2

cosh

−2

(pθ), p

2

=

1 −α

1

P

2

4P

2

(α

3

−α

4

P

2

)

(5.89)

now accounts for a solitary wave with parameters varying with respect to

X. Depending upon the function P(X) the solitary wave (5.89) may be

symmetric or asymmetric with respect to its core (or maximum) at diﬀerent

points in time, see Fig. 1.18. The evolution of the solitary wave may be

described solving the next order problem where an inhomogeneous linear

equation holds,

(1 −α

1

P

2

)v

1,θ

−2α

2

(v

0

v

1

)

θ

−(α

3

−α

4

P

2

) v

1,θθθ

= F, (5.90)

with

F = 2P[α

1

v

0,X

+α

2

v

2

0,X

+ (α

3

−2α

4

P

2

)v

0,θθX

] +

P

X

[α

1

v

0

+α

2

v

2

0,θ

+ (α

3

−6α

4

P

2

)v

0,θθ

] +

P[β

1

v

0

+β

2

v

2

0

+P

2

β

3

v

0,θθ

].

Like in previous section the solvability condition (5.83) yields the equation

for the function s = P

2

,

α

2

H

3

(s)s

X

= s (α

1

s −1)(α

3

−α

4

s)(6α

3

β

2

+ [5α

2

α

3

β

1

−

6β

2

(α

4

+α

1

α

3

) −α

2

β

3

]s −[5α

2

α

4

β

1

−6α

1

α

4

β

2

−α

1

α

2

β

3

]s

2

), (5.91)

where

H

3

(s) = α

1

α

4

(α

1

α

3

+ 5α

4

)s

3

−2α

4

(11α

1

α

3

+ 10α

4

)s

2

+

3α

3

(5α

1

α

3

+ 17α

4

)s −30α

2

3

.

Analysis of the solutions of Eq.(5.91) may be done similar to that in the

previous section. Again besides decay or an inﬁnite growth of s one can

describe its selection (from below and from above) to the values s

s

obtained

from the equation

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Inﬂuence of dissipative (active) external medium 161

-20 -10 10 20 30 40 50

x

0.25

0.5

0.75

1

1.25

1.5

v

-20 -10 10 20 30 40 50

x

-0.2

0.2

0.4

0.6

v

Fig. 5.7 Selection of the solitary waves governed by the solution (5.89), (5.91): (a) from

below (b) from above.

[5α

2

α

4

β

1

−6α

1

α

4

β

2

−α

1

α

2

β

3

]s

2

−[5α

2

α

3

β

1

−6β

2

(α

4

+α

1

α

3

) −α

2

β

3

]s−

6α

3

β

2

= 0. (5.92)

We see that after substitution s = 1/c

2

we get H

3

(s) = Q

6

(c)/c

6

while

Eqs. (5.92) coincides with Eq. (5.85). It means that parameters of the se-

lected solitary waves (5.76) and (5.89) are the same as well as the conditions

required for the selection.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

162 Ampliﬁcation of Nonlinear Strain Waves in Solids

In Fig. 5.7 it is shown the temporal evolution of the solitary wave (5.89)

whose amplitude and velocity vary according to Eq.(5.91). The parameters

are chosen so as an initial proﬁle is symmetric. We see that the initial

wave increases, Fig. 5.7(a), or decreases, Fig. 5.7(b), into the asymmetric

solitary wave. However, it transforms back into the symmetric wave during

the selection both from below and from above as shown in the last two

stages in Fig. 5.7(a) and Fig. 5.7(b).

We have studied how the external tangential dissipative (active) forces

may support an evolution of the long strain solitary waves in an elastic rod.

The conditions are found analytically when the selection of the solitary wave

occurs both from below and from above. It is shown that both asymptotic

solutions predict evolution into the symmetric selected solitary waves.

Previously, it was found analytically and in experiments the formation

of asymmetric strain solitary wave in a narrowing rod, see Sec. 4.1 where

the governing equation corresponds to Eq.(5.74) with β

1

= β

1

(x), β

2

= 0,

β

3

= 0. In experiments only narrowing part of the rod was observed where

a conversion to an asymmetric wave occurs, and there is no solitary wave

selection.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter 6

Bulk active or dissipative sources of

the ampliﬁcation and selection

Now the attention is paid on the situations when active/dissipative forces

act inside a solid body. It is known that variation formulations used be-

fore in the book cannot be applied in the presence of volume dissipative

eﬀects unless some modiﬁcations are proposed. Thus Lord Rayleigh (1945)

involved so-called dissipation function describing forces which depend upon

temporal derivatives of the variables. Dissipation function is included into

variation formulation through its elementary work. Another idea has been

used in Maugin (2000) to develop a variation formulation in thermoelastic-

ity. New variable has been introduced, called thermacy, into the relation

for the free energy density.

Another approach is of phenomenological kind. The coupled govern-

ing equations are derived in the form of a hyperbolic equation of mo-

tion (or the equation with main hyperbolic part) and the equation for a

variable responsible for dissipation (e.g., microdisplacement, temperature

etc.). The last equation is usually parabolic and is obtained phenomenolog-

ically. This modelling is connected with the concept of internal variables

Engelbrecht and Braun (1998); Maugin (1990); Maugin (1999); Maugin

and Muschik (1994). Application of the internal variables to the wave

problems may be found in Engelbrecht (1997); Engelbrecht et. al (1999);

Engelbrecht and Maugin (1996); Maugin (1999). According to this theory,

elastic stress or strain may be considered as observable variables in the

usual sense of the word. Its behavior is described by the hyperbolic equa-

tion. In order to compensate our lack of description of the internal structure

of material or the heat processes, another variables are introduced called

internal variables. They are usually governed by the equations of parabolic

type (i.e., diﬀusive). Various phenomenologocal model for viscoelastic bi-

ologocal media are discussed in Alekseev and Rybak (2002).

163

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

164 Ampliﬁcation of Nonlinear Strain Waves in Solids

Below some active/dissipative problems are considered, namely, waves

in microstructured solids, seismic waves, waves in a medium with moving

defects, and thermoelastic waves. All processes are governed by coupled

dissipative nonlinear partial diﬀerential equations. We do not consider here

viscoelastic materials describing by the integro- diﬀerential equations, see

about it, e.g., Christensen (1971); Engelbrecht (1979).

6.1 Nonlinear bell-shaped and kink-shaped strain waves in

microstructured solids

The classical theory of elasticity cannot account for eﬀects caused by the

microstructure of a material. The theory of microstructures has been

developed recently, see Capriz (1989); Engelbrecht and Braun (1998);

Eringen (1968); Maugin and Muschik (1994); Mindlin (1964) and refer-

ences therein quoted. Most results belong to the linear theory of elastic-

ity, however, there are ﬁndings in the ﬁeld of the nonlinear theory En-

gelbrecht and Braun (1998); Eringen (1968); Eringen and Suhubi (1964);

Erofeev and Potapov (1993); Maugin and Muschik (1994). An impor-

tant problem is the lack of data on the microstructure parameters, but

a few works can be mentioned Savin et. al (1973b) where attempts to mea-

sure them were done. Also experiments were performed in Potapov and

Rodyushkin (2001). Strain waves may be useful in developing a possible

method to estimate the microparameters since shape, amplitude and veloc-

ity of the strain wave can carry informations about the microstructure.

Strain waves in microstructured medium were studied mainly in the

linear approximation Eringen (1968); Mindlin (1964). Only a few works

are devoted to the nonlinear waves in microstructured non-dissipative me-

dia Engelbrecht and Braun (1998); Erofeev and Potapov (1993); Sillat

(1999) while the inﬂuence of the dissipative microstructure on the evo-

lution of non-linear waves has been discussed in Cermelli and Pastrone

(1997); Engelbrecht (1997); Engelbrecht and Braun (1998); Engelbrecht

and Khamidullin (1988) The inﬂuence of dissipation/accumulation may

be described by various methods, see Engelbrecht (1983) and references

therein. Recently in a series of papers Cermelli and Pastrone (1997);

Engelbrecht et. al (1999) the inﬂuence of dissipation on a shock propagation

was studied in the one dimensional case while inertia of the microstructure

was neglected into account together with the gradient of the microﬁeld .

Later numerical simulations were performed to account for the evolution of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 165

periodic waves Sillat (1999).

Below we follow Porubov and Pastrone (2001). The model discussed in

Cermelli and Pastrone (1997); Engelbrecht et. al (1999) is modiﬁed includ-

ing both the inertia of the microstructure and the gradient of the microﬁeld

. As a result the nonlinear PDE with dispersion and dissipative (active)

terms is derived. It is shown that it is able to describe both the bell-shaped

and kink-shaped longitudinal strain solitary waves. The simultaneous in-

ﬂuence of the accumulation/dissipation on the evolution is studied, and the

selection of quasistationary dissipative solitary waves is found. A possibil-

ity of the estimation of the microstructure parameters is proposed on the

basis of the results obtained here.

6.1.1 Modelling of a microstructured medium with dissipa-

tion/accumulation

Let us recall some basic ideas following Eringen (1968); Mindlin (1964). The

motion of a material particle is characterized by the displacement vector

with components U

J

(x

J

, t). We suppose that the material particle contains

discrete micromaterial elements whose displacements are accounted by the

microdisplacement vector with components U

J

(x

J

, t). Following Eringen

(1968); Mindlin (1964), we assume the microdisplacement depends linearly

on the microcoordinates x

J

: U

J

(x

J

, t) = x

K

ψ

KJ

(x, t). Hence the microdis-

placement gradient is given by ∂

I

U

J

= ψ

IJ

, thus avoiding dependence upon

microcoordinates. In a reference conﬁguration, the fundamental strains are

given by: the Cauchy-Green macrostrain tensor , C

IJ

, the distortion tensor,

E

IJ

= ∂

I

U

J

− ψ

IJ

, and the microdisplacement gradient, Γ

IJK

= ∂

I

ψ

JK

.

This kinematical model is valid for particular families of microstruc-

tures only; if one needs a model to describe wider class of phenomena

(solids with aﬃne microstructures, liquid crystals, ﬂuid with micro-bubbles,

etc.), we refer to the references given in Cermelli and Pastrone (1997);

Engelbrecht et. al (1999).

Let us assume the following:

(i) the movement is one dimensional, all variables depend upon the coordi-

nate x and the time t;

(ii) the macromotion if small but ﬁnite. Then the geometrical nonlinear-

ity is described by the only macrostrain C

XX

= U

x

+ 1/2U

2

x

, while the

Murnaghan model is valid to account for the physical nonlinearity;

(iii) the microstructure is suﬃciently weak to be considered in the linear

approximation, E

XX

= U

x

− ψ, ψ = ψ

XX

, Γ

XXX

= ψ

x

. Now and in the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

166 Ampliﬁcation of Nonlinear Strain Waves in Solids

following small lower indices denote diﬀerentiation.

(iv) the external forces are negligible.

Then the one dimensional governing equations may be written as in

Mindlin (1964),

ρU

tt

= σ

x

+τ

x

, (6.1)

I ψ

tt

= η

x

+τ, (6.2)

where ρ is the macro-density, I characterizes micro inertia. In dissipation-

less case components σ

eq

, τ

eq

and η

eq

of the tensors σ

IJ

, τ

IJ

and η

IJK

are

deﬁned through the derivatives of the free energy W,

σ

IJ

=

∂W

∂C

IJ

, τ

IJ

=

∂W

∂E

IJ

, η

IJK

=

∂W

∂Γ

IJK

,

which in the one dimensional limit Mindlin (1964); Murnaghan (1951) is

reduced to:

σ

eq

= (λ + 2µ)U

x

+βU

2

x

, τ

eq

= D(U

x

−ψ), η

eq

= Gψ

x

. (6.3)

Here λ, µ are the Lame coeﬃcients, β = 3/2(λ + 2µ) + l + 2m, D and

G are constant parameters. Assume the following representations in the

general case,

σ = σ

eq

+AU

xt

+aψ

t

, τ = τ

eq

+BU

xt

+bψ

t

, η = η

eq

+FU

xxt

+f ψ

xt

. (6.4)

Note that Aand a, B and b, F and f have the same dimensions. Dissipation

inequality imposes some restrictions on the involved parameters Cermelli

and Pastrone (1997); Engelbrecht (1983); Engelbrecht et. al (1999); Maugin

and Muschik (1994). The reason of our assumption may be seen considering

the linearized case in absence of the microstructure. Then we have for σ:

σ = (λ + 2µ)U

x

+AU

xt

,

that relates to the Voigt model, see, e.g., Bland (1960), the simplest exten-

sion of the Hook law to the viscoelastic media Alekseev and Rybak (2002).

Our model may be considered as a generalization of the Voigt model of

microstructured solids. Let us remark that the Voigt model accounts for

the inﬂuence of the dissipation only. We would like to consider a more

general case to account also for the energy inﬂux to the strain wave caused

by the microstructure. Hence the coeﬃcients in Eq.(6.4) may be of diﬀer-

ent signs. Certainly, constitutive equations (6.4) are of phenomenological

kind, since strictly speaking, they are not rigorously deduced within the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 167

general framework of Rational Continuum Mechanics, but we do not want

to go further in details here. In the following a more general constitutive

equations will be derived according the basic assumptions and methods of

Continuum Mechanics. Equation (6.4) can be seen as a particular case,

obtained via proper additional assumptions.

Substituting Eqs.(6.3), (6.4) into Eqs. (6.1), (6.2) and introducing the

functions v = U

x

and ψ as unknown variables, we obtain the coupled equa-

tions

ρv

tt

= (λ+2µ+D)v

xx

−Dψ

xx

+β( v

2

)

xx

+(A+B)v

xxt

+(a+b)ψ

xxt

, (6.5)

I ψ

tt

= D(v −ψ) +Bv

t

+bψ

t

+Gψ

xx

+F v

xxt

+f ψ

xxt

. (6.6)

Further simpliﬁcations follow considering only long waves with charac-

teristic length L >> 1. Moreover, the macro strain v is elastic and does

not exceed the yield point. Since the Murnaghan model is chosen, we are

dealing with those elastic materials whose yield points are small, hence,

the characteristic strain magnitude V is also small, V << 1. Let us in-

troduce L as a scale for x, V - as a scale for v and ψ, L/c

0

, as a scale for

t, c

2

0

= (λ + 2µ)/ρ as a characteristic velocity. The microinertia term I

depends upon the square of a typical size p of a microstructure element.

Then I may be rewritten as I = ρp

2

I

∗

, I

∗

being dimensionless. Using di-

mensional analysis, one can assume G = p

2

G

∗

, G

∗

having the dimension of

stress. The inﬂuence of dissipation/accumulation may be conveniently de-

scribed supposing A = dA

∗

, a = da

∗

, B = dB

∗

, b = db

∗

, F = dF

∗

, f = df

∗

,

where d has the dimension of a length. Then three positive dimensionless

parameters appear in Eqs.(6.5), (6.6):

(a) ε = V << 1 accounting for elastic strains;

(b) δ = p

2

/L

2

<< 1, characterizing the ratio between the microstruc-

ture size and the wave length;

(c) γ = d/L, characterizing the inﬂuence of the dissipation.

Then the dimensionless governing equations are (we keep the notations for

variables):

v

tt

−(1 +

D

λ + 2µ

)v

xx

+

D

λ + 2µ

ψ

xx

= ε

β

λ + 2µ

( v

2

)

xx

+

γ

_

(A

∗

+B

∗

)c

0

λ + 2µ

v

xxt

+

(a

∗

+b

∗

)c

0

λ + 2µ

ψ

xxt

_

, (6.7)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

168 Ampliﬁcation of Nonlinear Strain Waves in Solids

Dψ = Dv +γc

0

[B

∗

v

t

+b

∗

ψ

t

] +δ [G

∗

ψ

xx

−(λ + 2µ)I

∗

ψ

tt

] +

γδc

0

[F

∗

v

xxt

+f

∗

ψ

xxt

] . (6.8)

In absence of the microstructure, B

∗

, D, F

∗

, G, a

∗

, b

∗

, and δ are equal

to zero, the only equation for v is of the form

v

tt

−v

xx

−ε

β

λ + 2µ

( v

2

)

xx

−γ

(A

∗

+B

∗

)c

0

λ + 2µ

v

xxt

= 0. (6.9)

Assume dissipation is weak. If we expand the solution of Eq.(6.8) in the

form :

ψ = ψ

0

+γ ψ

1

+δ ψ

2

+γδ ψ

3

+ γ

2

ψ

4

+..., (6.10)

with

ψ

0

= v, ψ

1

=

(B

∗

+b

∗

)c

0

D

v

t

, ψ

2

=

G

∗

D

v

xx

−

(λ + 2µ)I

∗

D

v

tt

,

ψ

3

=

[(F

∗

+f

∗

)D + (B

∗

+ 2b

∗

)G

∗

]c

0

D

2

v

xxt

−

(λ + 2µ)(B

∗

+ 2b

∗

)c

0

I

∗

D

2

v

ttt

,

ψ

4

=

b

∗

(B

∗

+b

∗

)c

2

0

D

2

v

tt

. (6.11)

we obtain, by substituting (6.10), (6.11) into Eq.(6.7), the governing non-

linear PDE for the macrostrain v(x, t),

v

tt

− v

xx

−εα

1

( v

2

)

xx

−γα

2

v

xxt

+δ(α

3

v

xxxx

−α

4

v

xxtt

)+

γδ(α

5

v

xxxxt

+α

6

v

xxttt

) +γ

2

α

7

v

xxtt

= 0, (6.12)

where

α

1

=

β

λ + 2µ

, α

2

=

(A

∗

+a

∗

)c

0

λ + 2µ

, α

3

=

G

∗

λ + 2µ

, α

4

= I

∗

,

α

5

=

[(F

∗

+f

∗

)D + (B

∗

+b

∗

−a

∗

)G

∗

]c

0

(λ + 2µ)D

,

α

6

=

I

∗

(a

∗

−B

∗

−b

∗

)c

0

D

, α

7

=

a

∗

(B

∗

+b

∗

)

ρD

.

Comparing Eq.(6.12) with Eq.(6.9) we see that the inclusion of the gra-

dient of microdistortion provides the dispersion v

xxxx

, while the inertia of

the microstructure gives us mixed dissipation and dispersion terms. The

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 169

evolution of nonlinear strain wave depends upon the ratio between param-

eters ε, γ and δ. This is the reason we retain terms quadratic in these

parameters in the expansion of the solution (6.10). Sometimes they can

be considered ” negligibly small”, sometimes-not, according to the diﬀerent

eﬀects we want to point out, as it will be explained in the following.

There are two main types of nonlinear travelling solitary waves which

could propagate keeping its shape, bell-shaped and kink-shaped solitary

waves. The bell-shaped solitary wave usually appears as a result of a

balance between nonlinearity and dispersion. The kink-shaped localized

travelling structure may be sustained by diﬀerent balances, one possibil-

ity occurs when nonlinearity is balanced by dissipation (or accumulation),

another case corresponds to the simultaneous balance between dispersion,

nonlinearity and dissipation (or accumulation). Typical shapes of the waves

are shown in Figs. 1.1–1.5.

6.1.2 Bell-shaped solitary waves

The balance between nonlinearity and dispersion takes place when δ =

O(ε). If in addition γ = 0, we have the nondissipative case governed by the

double dispersive equation,

v

tt

− v

xx

−ε(α

1

( v

2

)

xx

−α

3

v

xxxx

+α

4

v

xxtt

) = 0. (6.13)

Its exact bell-shaped travelling solitary wave solution arises as a result of

balance between nonlinear and dispersive terms. It satisﬁes the boundary

conditions

∂

k

∂x

k

v →0 for [x[ →∞, k = 0, 1, 2, 3, (6.14)

and takes the form (see section 3):

v =

6k

2

(α

4

c

2

−α

3

)

α

1

cosh

−2

(k θ), (6.15)

where θ = x −ct, c is a free parameter,

k

2

=

c

2

−1

4ε(α

4

c

2

−α

3

)

. (6.16)

Hence, the solitary wave (6.15) exists when c

2

> max¦1, α

3

/α

4

¦ or

when 0 < c

2

< min¦1, α

3

/α

4

¦. In the ﬁrst case longitudinal tensile waves

propagate, while in the second case only compressive waves propagate.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

170 Ampliﬁcation of Nonlinear Strain Waves in Solids

In general, Eq.(6.12) possesses an exact travelling bell-shaped solution

vanishing at inﬁnity Kudryashov (1988); Parkes and Duﬀy (1996),

v = −

60ck

3

γδ

εα

1

(α

5

+α

6

c

2

) cosh

−2

(k (x −ct))[tanh(k (x −ct)) ±1], (6.17)

where

k = ±

2γcα

2

δ(α

3

− ¯ α

4

c

2

)

,

¯ α

4

= α

4

+ε

2

/δ α

7

, the phase velocity c is deﬁned from the overdetermined

system of bi-quadratic equations,

(δα

4

−γ

2

α

7

)c

4

−[δ(α

3

+ ¯ α

4

) + 24γ

2

α

2

2

]c

2

+δα

3

= 0,

(δ¯ α

4

2

+ 16γ

2

α

2

α

6

)c

4

+ (16γ

2

α

2

α

5

−2δα

3

¯ α

4

)c

2

+δα

2

3

= 0,

hence, the exact solution (6.17) exists only for particular values of the

coeﬃcients in such equations.

When δ = O(ε), γ << 1, Eq.(6.12) is considered as dissipation per-

turbed double dispersive equation (6.13). The asymptotic solution is sought

as a function of the phase variable θ and the slow time T, v = v(θ, T), with

θ

x

= 1, θ

t

= −c(T), T = γ t.

Then from (6.12) we get that

(c

2

−1)v

θ

−εα

1

( v

2

)

θ

+ε(α

3

−α

4

c

2

) v

θθθ

=

γ

_

2c[v

T

−εα

4

v

θθT

] +c

T

[v −εα

4

v

θθ

] −c

∂

2

∂θ

2

_

α

2

v −ε(α

5

+α

6

c

2

) v

θθ

¸

_

+

O(γ

2

). (6.18)

The solution of Eq.(6.18) is sought in the form

v = v

0

(θ, T) +γv

1

(θ, T) +... (6.19)

where v

i

, i = 0, 1, ..., satisfy boundary conditions (6.14) at [θ[ → ∞. Sub-

stituting (6.19) into (6.18) we have in the leading order

(c

2

−α

1

)v

0,θ

−εα

1

( v

2

0

)

θ

+ε(α

3

−α

4

c

2

) v

0,θθθ

= 0. (6.20)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 171

The exact solitary wave solution of Eq.(6.20) has the form (6.15) with

c = c(T). The ﬁrst order term v

1

in the solution (6.19) obeys the inhomo-

geneous linear equation

(c

2

−α

1

)v

1,θ

−2εα

1

(v

0

v

1

)

θ

+ε(α

3

−α

4

c

2

) v

1,θθθ

= F, (6.21)

where F is

F = 2c[v

0,T

−εα

4

v

0,θθT

] +c

T

[v

0

−εα

4

v

0,θθ

]−

c

∂

2

∂θ

2

_

α

2

v

0

−ε(α

5

+α

6

c

2

) v

0,θθ

¸

.

The operator M acting on the function v

1

in Eq.(6.21) is adjoint to the

operator

M

A

= (c

2

−1)∂

θ

−2εα

1

v

0

∂

θ

+ε(α

3

−α

4

c

2

) ∂

3

θ

.

Then using (6.20) and boundary conditions at inﬁnity one can obtain

the solvability condition for Eq.(6.21),

_

∞

−∞

v

0

F dθ = 0, (6.22)

which yields the equation for the function s = c

2

,

7ε s

T

Q

3

(s) = 2s (s −α

1

)

2

(q

1

s

2

+q

2

s +q

3

), (6.23)

with

Q

3

(s) = 30α

2

4

s

3

+3α

4

(17α

3

+5α

4

) s

2

+2α

3

(10α

3

+11α

4

) s −α

3

(α

4

+5α

3

),

q

1

= 5α

6

, q

2

= 5α

5

−7α

2

α

4

−5α

6

, q

3

= 7α

2

α

3

−5α

5

. (6.24)

Important features of the behaviour of s may be established analyzing

Eq.(6.23) without integration. There may be decay or inﬁnite growth of the

initial velocity and the amplitude of the solitary wave (6.15). However, the

most interesting evolution of s is realized when s (and hence the amplitude

of the solitary wave) tends to the ﬁnite constant value s

∗

as T →∞. Hence

the solitary wave in microstructured solids may be also selected. Indeed,

the values of s

∗

are the real positive solutions of equation

q

1

s

2

+q

2

s +q

3

= 0. (6.25)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

172 Ampliﬁcation of Nonlinear Strain Waves in Solids

Assume real roots of Eq.(6.25) are s

1q

< s

2q

. We denote by s

0

the initial

value of s while the real root of Q

3

is s

Q

. The sign of s

T

needed for s →

s

nq

depends upon the sign of Q

3

(s) around s = s

nq

, the sign of q

1

, and

the permitted interval deﬁned from (6.16). When s

Q

is the only real root

of Q

3

, and s

Q

< s

1q

the ampliﬁcation of the solitary wave with s

∗

= s

1q

occurs at q

1

> 0 if s

Q

< s

0

< s

1q

while at q

1

< 0 it ampliﬁes by s

∗

= s

2q

if s

1q

< s

0

< s

2q

. The attenuation of the wave to s

∗

= s

1q

happens when

q

1

> 0, s

1q

< s

0

< s

2q

, in case q

1

< 0 wave with initial velocity s

0

> s

2q

decreases to s

∗

= s

2q

. When s

1q

< s

Q

< s

2q

there is no selection for

q

1

> 0, while for negative q

1

the double selection is realized. Thus, waves

with s

0

< s

1q

amplify up to s

∗

= s

1q

, but waves with s

Q

< s

0

< s

2q

increase

up to s

∗

= s

2q

. Similarly, the attenuation to s

∗

= s

1q

happens for the waves

with s

1q

< s

0

< s

Q

, while the waves with s

0

> s

2q

attenuate by s

∗

= s

2q

.

If s

Q

> s

2q

there is no double selection, the wave evolution is similar to

the case s

Q

< s

1q

. The situation when Q

3

has three real roots within or

outside the interval [s

1q

, s

2q

] may be analyzed in the same manner. Thus

the stationary values of the solitary wave parameters are prescribed by the

equation coeﬃcients.

Ampliﬁcation, attenuation and selection of the bell-shaped solitary wave

v = v

0

is shown in Fig. 1.17. One can see that the solitary wave keeps its

localized bell-shaped form. In the case of ampliﬁcation, the wave ampli-

tude increases, while the width decreases; the attenuation is provided by

simultaneous decrease of the amplitude and the increase of the width. The

addition of correction γv

1

does not change signiﬁcantly the proﬁle of v = v

0

.

It is relevant to notice that the solution (6.19) is not uniformly valid: the

matching asymptotic expansions method Ablowitz and Segur (1981) should

be applied to complete the solution, even if it does not modify the behaviour

of the wave near its core.

6.1.3 Kink-shaped solitary waves

The equation (6.12) possesses also exact travelling kink-shaped solution

Kudryashov (1988); Parkes and Duﬀy (1996), in the form

v =

3

i=0

A

i

tanh

i

(k θ), (6.26)

with three possible sets of parameters A

i

, k . There are no free parameters

in the solution, and additional restrictions on the equation coeﬃcients are

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 173

needed.

When dispersion is weak, i.e., ε

2

< δ < ε and γ = O(ε), the nonlinearity

is balanced by the dissipation/accumulation only. Then Eq.(6.12) may be

rewritten as

v

tt

− v

xx

−ε(α

1

( v

2

)

xx

+α

2

v

xxt

) = −δ(α

3

v

xxxx

− ¯ α

4

v

xxtt

)−

εδ(α

5

v

xxxxt

+α

6

v

xxttt

), (6.27)

whose solution is sought in the form

v = v

0

+δv

1

+... (6.28)

where v

i

= v

i

(θ = x −ct) satisﬁes the boundary conditions

v

0

→h

±

, v

i

→0, i > 0, for θ →±∞, (6.29)

and all derivatives of v

i

with respect to θ vanish at inﬁnity. For a kink

h

+

,= h

−

. In the leading order the kink solution has the form

v = Am tanh(mθ) +B, (6.30)

with

A = −

α

2

c

α

1

, B =

c

2

−1

2εα

1

. (6.31)

There are two free parameters, the phase velocity c and the wave number

m which are deﬁned from the boundary conditions,

m =

(h

−

−h

+

)α

1

2cα

2

, c

2

= 1 +εα

1

(h

+

+h

−

).

Next order solution v

1

consists of two parts, v

1

= v

1d

+ v

1a

, where dispersive

perturbation of the kink vanishing at inﬁnity is accounted for the solution,

v

1d

=

2(¯ α

4

c

2

−α

3

)m

2

εα

1

cosh

−2

(mθ) log sech(mθ), (6.32)

while higher order dissipative/active terms contribution is

v

1a

=

2(α

5

+α

6

c

2

)cm

3

α

1

cosh

−2

(mθ) (3 tanh(mθ) −2mθ) . (6.33)

The alterations of the kink shape in absence of higher order dissipa-

tive/active terms, α

5

= α

6

= 0, depend upon the sign of (α

4

c

2

− α

3

)/α

1

.

Typical proﬁles of v = v

0

+ δv

1d

with v

1d

deﬁned by (6.32) are shown

in Fig. 6.1 for diﬀerent values of α

3

with other parameters values ﬁxed:

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

174 Ampliﬁcation of Nonlinear Strain Waves in Solids

-6 -4 -2 2 4 6

x

-1

1

2

3

v

e

-6 -4 -2 2 4 6

x

-1

1

2

3

v

c

-6 -4 -2 2 4 6

x

-1

1

2

3

v

d

-6 -4 -2 2 4 6

x

-1

1

2

3

4

v

a

-6 -4 -2 2 4 6

x

-1

1

2

3

v

b

Fig. 6.1 Inﬂuence of weak dispersion on the shape of the kink-shaped wave. Shown by

dashed line is the unperturbed kink.

α

1

= 1, α

2

= −1, ¯ α

4

= 1, c = 1.5, m = 1.5, ε = 0.5, δ = 0.1. The ”non-

symmetric” disturbances of the kink shape are seen in Fig. 6.1(a), where

α

3

= −2.5; they become weaker when α

3

tends to zero. In Fig. 6.1(b) the

value is α

3

= −0.5, while undisturbed kink appears in Fig. 6.1(c) where

α

3

= 2.25. Then the disturbances develop on another ”side” of the kink,

as shown in Figs. 6.1(d,e) with α

3

= 5, 7, correspondingly, and we have the

mirror proﬁle of those shown in Fig. 6.1(a, b). Fig. 6.1 shows the ampliﬁ-

cation of the kink-shaped wave since the diﬀerence between its maximum

and its minimum is larger than in the undisturbed case. In contrast to the

bell-shaped wave now the ampliﬁcation is accompanied by the alteration

of the wave proﬁle. One can note the similarity with the proﬁle in Fig. 1.5

of the exact kink-shaped solution of DMKdV equation. In Fig. 6.2 it

is shown what happens with the kink (6.30) when there is no dispersion,

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 175

-6 -4 -2 2 4 6

x

-2

-1

1

2

3

4

v

e

-6 -4 -2 2 4 6

x

-1

1

2

3

v

c

-6 -4 -2 2 4 6

x

-1

1

2

3

v

d

-6 -4 -2 2 4 6

x

-1

1

2

3

v

a

-6 -4 -2 2 4 6

x

-1

1

2

3

v

b

Fig. 6.2 nﬂuence of weak higher order dissipation on the shape of the kink-shaped wave

. Shown by dashed line is the unperturbed kink.

α

3

= α

4

= 0. The proﬁles of v = v

0

+ δv

1a

with v

1a

deﬁned by (6.33)

correspond to α

1

= 1, α

2

= −1, α

5

= 1.5, c = 1.5, m = 1.5, ε = 0.5,

δ = 0.1. In Fig. 6.2(a) α

6

= −2.5, it varies from α

6

= −1.5, Fig. 6.2(b) to

α

6

= −0.5, Fig. 6.2(c) where there are no disturbances. We have α

6

= 0.5

in Fig. 6.2(d), while Fig. 6.2(e) corresponds to α

6

= 2.5. Two main dis-

tinctions may be mentioned in comparison with proﬁles shown in Fig. 6.1.

First, the disturbances of the kink shape are ”symmetric” independently of

the sign of α

6

. Second, there are no mirror proﬁles arising at distinct signs

of α

6

.

The case δ = O(ε), γ = O(ε) corresponds to the simultaneous balance

between nonlinearity, dispersion and dissipation/accumulation, perturbed

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

176 Ampliﬁcation of Nonlinear Strain Waves in Solids

by the higher order dissipative/active terms,

v

tt

− v

xx

−ε[α

1

( v

2

)

xx

+α

2

v

xxt

−α

3

v

xxxx

+ ¯ α

4

v

xxtt

] =

−ε

2

(α

5

v

xxxxt

+α

6

v

xxttt

), (6.34)

The solution of Eq.(6.34) is sought in the form

v = v

0

+εv

1

+ε

2

v

2

... (6.35)

where boundary conditions (6.29) hold. Substituting (6.35) into (6.34)

in the leading order we obtain the D’Alembert equation. Consider only

one wave travelling to the left and assume v depends upon phase variable

θ = x − ct with c = 1 + εc

1

+ ε

2

c

2

.... Then at order ε the ODE equation

for v

0

is,

2c

1

v

0,θ

−α

1

( v

2

0

)

θ

+α

2

v

0,θθ

+ (α

3

− ¯ α

4

) v

0,θθθ

= 0. (6.36)

Equation (6.36) possesses the exact kink-shaped solution Vlieg-

Hultsman and Halford (1991),

v

0

= A tanh(mθ)sech

2

(mθ) +B tanh(mθ) +C, (6.37)

with

A =

3α

2

2

50α

1

(¯ α

4

−α

3

)

, B =

3α

2

2

25α

1

(¯ α

4

−α

3

)

, C =

c

1

α

1

, m =

α

2

10(α

3

− ¯ α

4

)

.

It follows from (6.29) that

h

+

−h

−

= 2B, c

1

=

α

1

2

(h

+

+h

−

),

and the solution exists under special boundary conditions. The inﬂuence of

higher order terms may be studied in a similar way as done for the solution

(6.28).

6.1.4 Concluding remarks

We have found that various features of the microstructure provide corre-

sponding terms in the governing equation (6.12). Thus inertia yields mixed

derivative terms v

xxtt

, v

xxttt

, while dispersion v

xxxx

and higher order dissi-

pative/active term v

xxxxt

are due to the micro-deformation gradient. Dis-

persion is required for the existence of the bell-shaped solitary waves in an

elastic microstructured medium. The dispersive term v

xxtt

determines, in

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 177

particular, the existence of either exact compression or tensile solitary wave

solutions (see Eqs.(6.15), (6.16)), while the higher order dissipative/active

term α

6

v

xxttt

provides nonzero q

1

in Eq.(6.25), namely two stationary ﬁ-

nite velocities of the solitary wave. Dispersion terms also account for the

alterations in the kink-shaped wave proﬁle, see Fig. 6.1, while higher or-

der dissipative/active terms are responsible for a saturation which prevents

unbounded growth of the bell-shaped solitary wave.

We see that pairs B

∗

, b

∗

and F

∗

, f

∗

appear in the expressions for the

equation coeﬃcients only in combinations B

∗

+ b

∗

, F

∗

+ f

∗

. Hence, one

could reduce the number of the microstructure parameters in the model

(6.4). At the same time there is a need of both A and a, since a

∗

appears

in the expression for α

4

independently fromA

∗

. The ratio between the small

parameters, ε, δ and γ, allows to estimate the size p of the microstructure

and the parameter d which inﬂuence the dissipation/accumulation eﬀects.

According to this ratio the governing equation (6.12) may describe either

bell-shaped or kink- shaped solitary waves.

The solutions of Eq.(6.12) allow to describe in an explicit form the

ampliﬁcation of both types of the waves, as well as the selection of the

solitary wave, when its parameters tend to the ﬁnite values prescribed by

the coeﬃcients of the governing equation. The relationships among these

parameters deﬁne the thresholds that separate the parameters of the initial

solitary waves which will be ampliﬁed or attenuated. The wave amplitude

and velocity depend upon macro- and micro-properties of the microstruc-

tured medium through the analytical relationships, explicitly given above.

An application of the results obtained here consists in a possible esti-

mation of the microstructure parameters on the basis of the strain wave

behaviour. In principle, the measurements of the solitary wave amplitude

and velocity allow to obtain the parameters of the microstructure using

elastic macro-moduli known beforehand.

The asymptotic solution (6.19) describing bell-shaped solitary wave se-

lection may help to explain transfer of the strain energy by the microstruc-

ture. Let us assume one single solitary wave with initial velocity

√

s

01

,

so that it will be attenuated propagating in the microstructured medium.

Hence, it looses its energy which is absorbed by the microstructure. If an-

other wave travels with an initial velocity

√

s

02

, it will be ampliﬁed and

it means that we need a source of energy to justify the ampliﬁcation. A

possible explanation could be that the energy stored by the microstructure

is released because of the passing wave. Certainly the solution cannot de-

scribe an energy exchange between the waves but it gives us the range of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

178 Ampliﬁcation of Nonlinear Strain Waves in Solids

the microstructure parameters when the energy transmission is possible.

The predictions of the asymptotic solution may be valid even in a more

complicated unsteady process of the formation of solitary waves from an

arbitrary input, see, e.g., Sec. 2.3.1. where the equation is considered rather

similar to ours. Like here its single solitary wave asymptotic solution has

been obtained in Sec. 2.2, and the conditions were found for the decay or a

selection of a single solitary wave. Then the evolution of an initial arbitrary

pulse has been studied numerically. It was found that the initial pulse splits

into a sequence of solitary waves but each of them evolves according to the

single solitary wave analytical solution.

6.2 Nonlinear seismic solitary waves selection

6.2.1 Modelling of nonlinear seismic waves

The inﬂuence of microstructure may explain phenomena caused by the

energy input/output. Thus, recently the phenomenological theory has been

developed in Engelbrecht (1997); Engelbrecht and Khamidullin (1988) to

account for the seismic waves propagation in a horizontal layer. It was

proposed to describe longitudinal strain waves evolution by the nonlinear

equation,

u

t

+uu

x

+d u

xxx

= εf(u), (6.38)

where f is the body force related to the so-called dilation mechanism,

f(u) = −

_

a

1

u −a

2

u

2

+a

3

u

3

_

, (6.39)

a

1

, a

2

, a

3

are positive constants and ε is a small parameter. Eq. (6.38)

may describe an appearance of microseisms. The internal energy is stored

in a geophysical medium, while propagating seismic waves may release the

locked-in internal energy. Additional energy inﬂux yields an ampliﬁcation

of the wave.

The basic idea of the seismic waves modelling is originated from the di-

lation theory in fracture mechanics Zhurkov (1983). It was assumed there

that negative density ﬂuctuations play an essential role in the strength of

solids. These ﬂuctuations are called dilatons. They may be considered

as short-lived objects which are able to absorb energy from a surround-

ing medium. Accumulation of the energy may happen only up to a cer-

tain threshold value, then it is released, and the dilaton breaks generating

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 179

a crack. Qualitatively similar phenomena were recognized in Koz´ak and

ˇ

Silen´ y (1985) studying seismic energy release to explain the earthquake

mechanism. The necessary condition for the fracturing of the medium

under load is the existence of an inhomogeneity like a tectonic fault, an

inclusion etc. Hence it was assumed in Koz´ak and

ˇ

Silen´ y (1985) to consider

a medium as a two-dimensional homogeneous space containing a linear in-

homogeneity compressed uniaxially, which is the structure that simulates

commonly occurring geological faults subjected to tectonic stress with a

predominant orientation. The area, aﬀected by the loading, increases until

the stress ﬁeld achieves a threshold. Then a seismic-energy-releasing events

occur. A similar dilatancy model has been proposed in Gusev (1988) to

explain the nature of earthquake precursors. In particular, it was assumed

that the mechanism of seismic radiation is connected with rapid dilatancy

variations.

The theory developed in Gusev (1988); Koz´ak and

ˇ

Silen´ y (1985) is

linear. Preliminary results, mainly qualitative, were obtained in Niko-

laev (1989) to clarify the role of the simultaneous inﬂuence of nonlinear-

ity and dissipation on the seismic waves evolution. However, the most

important contribution to nonlinear description of the seismic waves has

been done in Engelbrecht (1997); Engelbrecht and Khamidullin (1988).

In order to govern a medium that may store and release the energy it

was proposed in Engelbrecht (1997); Engelbrecht and Khamidullin (1988)

to consider the Earth crust as a certain hierarchy of elastic blocks con-

nected by thin interface layers. The layers are inhomogeneities where

the energy is pumped, stored and released. Hence the interface lay-

ers behave like dilatons. Derivation of Eq.(6.38) in Engelbrecht (1997);

Engelbrecht and Khamidullin (1988) is based on a model where the classic

elasticity basic equations are complemented by the inclusion of the body

force to account for the dilaton mechanism, and the phenomenological ex-

pression for the body force (6.39) closes the basic equations.

In absence of the body force, f = 0, Eq.(6.38) is the celebrated

Korteweg-de Vries equation, whose exact travelling one-parameter solitary

wave solution arises as a result of a balance between nonlinearity, uu

x

, and

dispersion, d u

xxx

. Body force f plays a dissipative/active role destroy-

ing this balance. When all terms in the expression for f are dissipative,

the solitary wave decays, while there is an inﬁnite growth in a pure ac-

tive case. The most interesting scenario happens in the mixed dissipative-

active case. In particular, numerical results in Engelbrecht (1997);

Engelbrecht and Khamidullin (1988) demonstrate transformation of an ini-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

180 Ampliﬁcation of Nonlinear Strain Waves in Solids

tial KdV soliton into a new stable localized bell-shaped wave with the

amplitude and velocity prescribed by the equation coeﬃcients.

The nature of the terms in f depend upon the values of the coeﬃcients

a

1

, a

2

, a

3

, but numerical simulations cannot describe the intervals of their

values required for the appearance of the stable localized waves. In order to

obtain this information a procedure is developed below. Most of the results

were ﬁrst published in Porubov et. al (2003). First, the unsteady process

of the transformation of the KdV soliton into the solitary wave with pre-

scribed parameter values is described analytically. Then it is demonstrated

numerically that solitary waves selection

(i) in presence of the solitary waves interactions;

(ii) when an initial proﬁle is arbitrary;

(iii) when the parameter ε is not small

happens in quantitative agreement with asymptotic solution.

6.2.2 Asymptotic solution of the governing equation

Let us assume that ε << 1. Furthermore the function u depends upon a

fast variable ξ and a slow time T , such as

ξ

x

= 1, ξ

t

= −V (T), T = ε t.

Then equation (6.38) becomes

d u

ξξξ

−V u

ξ

+uu

ξ

+ε

_

u

T

+a

1

u −a

2

u

2

+a

3

u

3

¸

= 0. (6.40)

The solution u of Eq.(6.40) is sought in the form:

u(ξ, T) = u

0

(ξ, T) +ε u

1

(ξ, T) +... (6.41)

In the leading order we have

d u

0,ξξξ

−V u

0,ξ

+ u

0

u

0,ξ

= 0. (6.42)

Equation (6.42) contains coeﬃcients V = V (T), hence, its exact solitary

wave solution will have slowly varying parameters,

u

0

= 12 d k(T)

2

cosh

−2

(k(T) ξ) (6.43)

with V = 4dk

2

.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 181

In the next order an inhomogeneous linear diﬀerential equation for u

1

appears,

d u

1,ξξξ

−V u

1,ξ

+ (u

0

u

1

)

ξ

= F, (6.44)

with

F = −

_

u

0,T

+a

1

u

0

−a

2

u

2

0

+a

3

u

3

0

¸

.

Due to (6.43)

u

0,T

=

2k

T

k

u

0

+

k

T

k

ξ u

0,ξ

.

The solvability condition for Eq.(6.44) is

∞

_

−∞

u

0

F dξ = 0. (6.45)

Then it follows from (6.45) that k obeys the equation

k

T

= −

2

105

k

_

3456a

3

d

2

k

4

−336a

2

dk

2

+ 35a

1

_

, (6.46)

that may be rewritten in terms of the solitary wave amplitude Q =

12 d k(T)

2

as

Q

T

= −

4

105

Q(24a

3

Q

2

−28a

2

Q+ 35a

1

). (6.47)

The roots of the equation,

24a

3

Q

2

−28a

2

Q+ 35a

1

= 0,

are

Q

1

=

14a

2

−2

_

49a

2

2

−210a

3

a

1

24a

3

, Q

2

=

14a

2

+ 2

_

49a

2

2

−210a

3

a

1

24a

3

.

(6.48)

The behavior of the solitary wave amplitude, Q, depends on the value of

Q

0

≡ Q(T = 0). Indeed, Q will diverge at Q

0

< Q

1

, when Q

1

< Q

0

< Q

2

parameter Q will grow up to Q

2

, while if Q

0

> Q

2

it will decrease by Q

2

.

Hence parameters of the solitary wave tends to the ﬁnite values prescribed

by the equation coeﬃcients a

i

and is selected.

A more quantitative description of the variation of Q can be given in

order to see at what time the selected values are achieved. Equation (6.47)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

182 Ampliﬁcation of Nonlinear Strain Waves in Solids

may be directly integrated over the range (0, T) giving the implicit depen-

dence of Q on T:

T =

35

32a

3

Q

1

Q

2

(Q

2

−Q

1

)

_

Q

2

log

(Q−Q

1

)

(Q

0

−Q

1

)

−Q

1

log

(Q−Q

2

)

(Q

0

−Q

2

)

+ (Q

2

−Q

1

) log

Q

Q

0

_

(6.49)

One can see that T tends to inﬁnity when Q → Q

2

, and expression

(6.49) provides an analytical description of the time-dependent process of

the parameter-value selection of the solitary wave (6.43).

With Eq.(6.47) being taken into account, the solution for u

1

is

u

1

= A

1

[tanh(kξ) −1] + [3A

1

+ 2A

2

ξ] cosh

−2

(kξ) + [C −3kA

1

ξ −

A

2

ξ

2

−A

3

log(cosh(kξ))] tanh(kξ) cosh

−2

(kξ), (6.50)

where C = const,

A

1

=

1152a

3

d

2

k

4

−168a

2

dk

2

+ 35a

1

35k

,

A

2

=

3456a

3

d

2

k

4

−336a

2

dk

2

+ 35a

1

35

, A

3

=

1728a

3

d

2

k

3

35

.

We see that u

1

does not vanish at ξ →−∞, and a plateau appears behind

a solitary wave. It may be of negative or positive amplitude depending

upon the sign of A

1

. A uniformly valid solution vanishing at ξ →−∞ may

be obtained by the standard procedure described in Ablowitz and Segur

(1981).

One can make now some important conclusions. If we formally assume

a

2

= 0, a

3

= 0, both the behavior of the solitary wave parameters and

the sign of the amplitude of plateau are deﬁned by the sign of a

1

. Indeed

when a

1

> 0, the amplitude and velocity of the wave decreases in time

according to Eq.(6.47), while A

1

> 0, and the plateau is negative. On the

contrary, at negative a

1

we have an increase of the wave amplitude and

positive plateau. In general case the plateau may be negative both in case

of an increase and a decrease of the solitary wave. We also see that the

increase of the amplitude is accompanied by the decrease of the wave width.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 183

50 100 150 200 250 300

x

0

5

10

u

0

5

10

Fig. 6.3 Evolution of three solitary waves in absence of their interaction.

6.2.3 Numerical simulations

An asymptotic solution requires speciﬁc initial conditions while an evolution

of an arbitrary initial disturbance as well as interactions between nonlinear

localized waves are of practical interest. It may be described only numeri-

cally, however, it is important to know whether analytical predictions may

be used for a design of numerics, since the behavior of the waves is sensitive

to the values of the equation coeﬃcients and the initial conditions.

We use for computations a pseudo-spectral method whose computation

code was designed in Kliakhandler (1999). The program computes solutions

of 1D scalar PDEs with periodic boundary conditions. It evaluates spatial

derivatives in Fourier space by means of the Fast Fourier Transform, while

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

184 Ampliﬁcation of Nonlinear Strain Waves in Solids

50 100 150 200 250

x

0

10

u

0

10

Fig. 6.4 Selection of two solitary in presence of their interaction.

the time discretization is performed using the fourth-order Runge–Kutta

method. This scheme appears to have a good stability with respect to

the time step and was already successfully used for the modelling of the

solitary wave selection in a convective ﬂuid, see Sec. 2.3.1. More detailed

information about the code may be found in Kliakhandler (1999).

We choose the parameter values identical to that used in numerics in

Engelbrecht (1997): a

1

= 1, a

2

= 0.5, a

3

= 0.0556, d = 0.5, ε = 0.1.

Following the analysis from the previous section one obtains Q

1

= 4.11,

Q

2

= 6.38, and the selection occurs for single solitary waves with initial

amplitudes from the interval 4.11 < Q

0

< 6.38. Numerical results for

the single wave evolution conﬁrm analytical solutions and agree with the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 185

50 100 150 200 250

x

0

10

u

0

10

Fig. 6.5 Selection of one solitary wave and decay of another one in presence of their

interaction.

numerical results in Engelbrecht (1997).

Then the initial conditions are changed to the proﬁle containing three

solitary waves each accounting for Eq.(6.43) at T = 0. First, the waves

are located so as to avoid their interactions, see the ﬁrst stage in Fig. 6.3.

The initial amplitudes are chosen so as the values of the amplitudes of

the ﬁrst two solitary waves are brought into the selection interval, while

the amplitude of the last one is below Q

1

= 4.11. For convenience here

and in the following ﬁgures thresholds 4.11 and 6.38 are shown by dashed

lines at each stage. One can see in Fig. 6.3 that the amplitudes of the

ﬁrst two solitary waves tend to the value Q

2

= 6.38, while the last solitary

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

186 Ampliﬁcation of Nonlinear Strain Waves in Solids

50 100 150 200 250 300 350 400

x

0

5

10

u

0

5

10

Fig. 6.6 Evolution of an initial Gaussian proﬁle and formation of two selected solitary

waves.

wave decays. Hence each solitary wave evolves according to the asymptotic

solution.

Let us re-arrange the initial positions of the solitary waves in order to

include their interactions. First we take two larger initial solitary waves.

One can see in Fig. 6.4 that the interaction does not aﬀect the selection,

and again both solitary waves evolve in agreement with the theory. When

the second initial solitary wave from Fig. 6.3 is moved behind the third one,

its selection occurs despite the interaction, while the smaller wave decays.

The process is shown in Fig. 6.5.

One can see in Fig. 6.6 that an initial Gaussian pulse produces a train

of solitary waves of diﬀerent magnitude in agreement with the KdV theory.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 187

1 3 5 7 9

x

0

1

2

3

4

5

6

7

u

1 3 5 7 9

c

1 3 5 7 9

x

0

1

2

3

4

5

6

7

u

1 3 5 7 9

d

1 3 5 7 9

x

0

1

2

3

4

5

6

7

u

1 3 5 7 9

a

1 3 5 7 9

x

0

1

2

3

4

5

6

7

u

1 3 5 7 9

b

Fig. 6.7 Comparison of asymptotic (dashed line) and numerical (solid line) solutions at

various ε: a)ε = 0.1, b)ε = 0.3, c)ε = 0.5, d)ε = 1.

Then, the selection of those solitary waves occurs whose amplitudes come

to the selection interval prescribed by the theory. Note that two leading

solitary waves are selected from below while other solitary waves generated

from the input, vanish.

Finally, the inﬂuence of the small parameter value is studied. In Fig. 6.7

we see that the solitary waves continue to evolve according to the asymp-

totic solution with growth in ε. Two main deviations are observed. First is

the diﬀerence in amplitudes, it is caused by the increase of the contribution

of u

1

into the asymptotic solution. The diﬀerence in the shape of plateau is

shown in Fig. 6.8 where we see the decrease of its length. At larger ε = 10

the plateau almost disappears, see Fig. 6.9, however, the initial solitary

wave amplitude Q

0

= 5 still tends to the value Q

2

= 6.38. Selection from

above still occurs for Q

0

> Q

2

. At the same time we observe a decay of

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

188 Ampliﬁcation of Nonlinear Strain Waves in Solids

1 3 5 7 9 11 13

x

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

u

1 3 5 7 9 11 13

c

1 3 5 7 9 11 13 15

x

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

u

1 3 5 7 9 11 13 15

d

1 3 5 7 9 11 13 15

x

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

u

1 3 5 7 9 11 13 15

a

1 3 5 7 9 11 13 15

x

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

u

1 3 5 7 9 11 13 15

b

Fig. 6.8 Comparison of plateau in asymptotic (dashed line) and numerical (solid line)

solutions at various ε: a)ε = 0.1, b)ε = 0.3, c)ε = 0.5, d)ε = 1.

the initial solitary wave with the amplitude less than Q

0

= 3 that already

diﬀers from the theoretical predictions, Q

0

< Q

1

= 4.11.

6.3 Moving defects induced by external energy ﬂux

6.3.1 Basic concepts and derivation of governing equations

Recently it was found that point defects may be generated in a solid sub-

jected to the laser radiation, see Mirzoev et. al (1996) and references therein.

Point defect is described as a distortion in a crystal lattice in the area equal

to the atomic volume. The simplest point defects in the crystal are the

interstitial atom, or an atom occupying a position among the atomic equi-

librium positions in an ideal lattice, and the vacancy, or a speciﬁc site in the

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 189

16 18 20 22 24

x

0

1

2

3

4

5

6

u

Fig. 6.9 Evolution of an initial KdV soliton (dashed line) at ε = 10.

lattice of the crystal where atoms are absent Kosevich (1981). Generation

of defects due to the laser radiation may be explained using interaction of

strain, temperature and defect-concentration ﬁelds Mirzoev et. al (1996).

Let us consider an isotropic solid where a concentrated energy inﬂux

(e.g., laser beam) produces moving point defects. Following Mirzoev et. al

(1996) assume n

j

(x, t) be the concentration of defects of the kind j (j = v

for vacancies, j = i for interstitial atoms) at a point r(x, y, z) at time t.

The main processes responsible for the temporal evolution of a defect are

generation, recombination and diﬀusion. Then the kinetics of the point

defects is governed by the equation

n

j,t

= q

0

+q

ε

U

x

+D

j

n

j,xx

−β

j

n

j

, (6.51)

where q

0

is a velocity of the defects generation in absence of the strain, next

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

190 Ampliﬁcation of Nonlinear Strain Waves in Solids

term in the r.h.s. of Eq.(6.51) accounts for a contribution of a strain into

the defects generation, D

j

is the diﬀusion coeﬃcient of the defect of the

kind j, β

j

is a recombination velocity at sinks Mirzoev and Shelepin (2001).

Volume mutual recombination of the defects of diﬀerent kind is neglected.

It is important that point defects cause deformation of a medium at

macroscopic distances. In elasticity single defect is described by the volume

density of the force, f ,

f (r) = −K Ω

0

∇δ(r),

where K is the bulk modulus, Ω

0

is a dilatation parameter representing

the change in the volume of a crystal as a result of a formation of the one

defect, while defect is located in the coordinate origin Kosevich (1981). For

vacancies we have Ω

0

< 0, while Ω

0

> 0 corresponds to the interstitial

atoms. Combinations of the point defects yield defect of the dipole kind

that may be accounted for

f (r) = −K Ω

ik

∇

k

δ(r),

where Ω

ik

is a symmetric tensor. When dipole is axi-symmetric, we have

Ω

ik

= Ω

0

δ

ik

+ Ω

1

(l

i

l

k

−1/3δ

ik

),

where l is a unit vector of the dipole axis, Ω

1

deﬁnes deviator of the tensor

Ω

ik

Kosevich (1981). The average parameters of the crystal specimen may

be introduced if a typical distance between defects is considerably less than

the gradient of a strain ﬁeld. Then we have in one-dimensional case

f = −K Ω

j

n

j,x

, (6.52)

where evolution of the concentration of defects is accounted for Eq. (6.51).

6.3.2 Nonlinear waves in a medium

One-dimensional governing equation in an elastic medium with moving de-

fects may be obtained similar to Eq.(6.12) for a microstructured medium.

We start with the coupled equations for n

j

, Eq.(6.51), and for nonlinear

longitudinal displacement U(x, t),

ρU

tt

= σ

x

, (6.53)

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 191

where σ,

σ = (λ + 2µ)U

x

+ (3/2(λ + 2µ) +l + 2m)U

2

x

−K Ω

j

n

j

.

is written with Eq. (6.52) being taken into account. Equations (6.53),

(6.51) are simply uncoupled yielding the governing equation for the strain

waves v = U

x

,

v

tt

−a

1

v

xx

−a

2

( v

2

)

xx

−a

3

v

xxt

+a

4

v

ttt

−a

5

v

xxtt

+a

6

v

xxxx

−

a

7

( v

2

)

xxt

+a

8

( v

2

)

xxxx

= 0, (6.54)

where

a

1

=

λ + 2µ

ρ

−

K Ω

j

q

ε

ρβ

j

, a

2

=

3(λ + 2µ) + 2(l +m) + 3n

4ρ

, a

3

=

(λ + 2µ)

β

j

ρ

,

a

4

=

1

β

j

, a

5

=

D

j

β

j

, a

6

=

(λ + 2µ)D

j

ρβ

j

, a

7

=

a

2

β

j

, a

8

=

a

2

D

j

β

j

.

We see from Eq. (6.54) that presence of the moving defects provides

a dispersion in an elastic medium, a

5

v

xxtt

and a

6

v

xxxx

, as well as dissipa-

tive/active terms. The coeﬃcients in Eq.(6.54) do not depend upon K Ω

j

with the exception of a

1

. Equation (6.54) is similar to Eq.(5.42) in Sec. 5.3

describing the inﬂuence of an external dissipative/active medium on the

waves evolution in a rod. Various simpliﬁed cases may be considered:

(i) β

j

>> 1;

(ii) D

j

<< 1;

(iii) β

j

>> 1, D

j

<< 1, D

j

= O(1/β

j

).

In case (i) recombinations at sinks are strong, quadratic nonlinearity in

Eq.(6.54) predominates, and the formation of shocks is possible. Weakly

dispersive case (ii) corresponds to the weak defects diﬀusion. An analy-

sis is already done in Sec. 5.3 where stable kink waves with a ”hat” are

found, see Figs. 5.5, 5.6. Simultaneous weak diﬀusion and strong recombi-

nations in case (iii) provide weakly dissipative/active case which has been

also studied in Sec. 5.3. When dissipation is negligibly small, the inﬂu-

ence of the higher-order nonlinear term a

8

( v

2

)

xxxx

may aﬀect the wave

behavior. Nondissipative limit of Eq.(6.54) is similar to a particular case

of Eq.(1.3) that possesses oscillatory and multi-humps solitary wave solu-

tions. However, it was shown in Sec. 1.2 that no stable bell-shaped solitary

waves arise in absence of higher-order derivative terms (like a

9

v

6x

or cor-

responding mixed derivative terms) in Eq.(6.54).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

192 Ampliﬁcation of Nonlinear Strain Waves in Solids

6.3.3 Nonlinear waves in a plate

Equation for longitudinal displacements in a plate already includes a dis-

persion even in absence of defects Mirzoev and Shelepin (2001):

U

tt

−c

2

s

U

xx

−β

N

U

x

U

xx

−l

2

(U

ttxx

−c

2

τ

U

xxxx

) = −

K Ω

j

ρ

n

j,x

(6.55)

where

c

2

s

=

E

ρ(1 −ν

2

)

, β

N

=

3E

ρ(1 −ν

2

)

+

2(1 −2ν)

ρ(1 −ν)

3

[(1 −2ν)

2

l +2(1 −ν +ν

2

)m],

l

2

=

ν

2

h

2

12(1 −ν)

2

, c

2

τ

=

µ

ρ(1 −ν

2

)

,

where h is the width of the plate. Again the evolution of the defects is

accounted for Eq.(6.51). Note that nonlinearity parameter β

N

does not

depend upon the Murnaghan modulus n.

Uncoupling of Eqs. (6.51), (6.55) in case of the elastic plate is per-

formed similar to the previous case giving the governing equation for the

longitudinal strains of the form

v

tt

−a

1

v

xx

−a

2

( v

2

)

xx

−a

3

v

xxtt

+a

4

v

xxxx

−

b

1

v

xxt

+b

2

v

ttt

−b

3

( v

2

)

xxt

−b

4

v

xxttt

+b

5

v

xxxxt

+

c

1

( v

2

)

xxxx

+c

2

v

xxxxtt

−c

3

v

xxxxxx

= 0, (6.56)

where

a

1

= c

2

s

−

K Ω

j

q

ε

ρβ

j

, a

2

=

β

N

2

, a

3

=

D

j

β

j

+l

2

, a

4

=

c

2

s

D

j

β

j

+c

2

τ

l

2

,

b

1

=

c

2

s

β

j

, b

2

=

1

β

j

, b

3

=

β

N

2β

j

, b

4

=

l

2

β

j

, b

5

=

c

2

τ

l

2

β

j

,

c

1

=

β

N

D

j

2β

j

, c

2

=

l

2

D

j

β

j

, c

3

=

c

2

τ

l

2

D

j

β

j

.

Equation (6.56) is close to Eq. (5.8) when the c

i

’s terms in Eq.(6.56)

are negligibly small. Now dispersion appears as a result of the plate ﬁnite

width also, and no weakly dispersive case is considered. Case β

j

>> 1

now corresponds to the perturbed double-dispersive equation. There are

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 193

two kinds of perturbations. Active/dissipative perturbations are accounted

for the b

i

’s terms while c

i

’s terms in Eq.(6.56) play the role of higher-

order nonlinearity and dispersion. Hence weakly active/dissipative case is

realized if additionally D

j

<< 1. Then we can use the results obtained

in Sec. 5.2 to account for the bell-shaped solitary wave selection. When

D

j

>> 1, c

i

’s terms dominate over b

i

’s terms, and we obtain from Eq.

(6.56) the hyperbolic analog of Eq.(1.3) studied in Chapter 1. Thanks to

nonzero dispersion, now stable oscillatory and multi-humps strain solitary

wave solutions are realized, see Figs. 1.7–1.16.

Eqs. (6.51), (6.55) are uncoupled in a diﬀerent manner if β

j

<< 1. In

this case Mirzoev and Shelepin (2001) derived the governing equation for

the strain v = U

x

,

v

t

+c

s

v

x

+β

d

v

xxx

+γ

N

v

2

x

= gv −ζv

xx

+µv

xxxx

, (6.57)

where

β

d

=

l

2

(c

2

s

−c

2

τ

)

2

, γ

N

=

β

N

ρc

s

, g =

q

ε

K Ω

j

ρc

2

s

, ζ = β

j

l

2

, µ = D

j

l

2

.

Equation (6.57) may be considered as a perturbed KdV equation. Then

its asymptotic solution is obtained using the procedure explained in Chap-

ter 2. The most interesting case is realized when small parameters q

ε

, D

j

and β

j

are of the same order. Then one can write g = εg

∗

, ζ = εζ

∗

, µ = εµ

∗

,

ε << 1. The fast variable θ and the slow time T are introduced as before,

θ

x

= 1, θ

t

= −V (T), T = εt, and the solution of Eq.(6.57) may be obtained

using the procedure from Chapter 2 to describe the selection of localized

longitudinal strain wave as a result of the interactions with moving defects.

6.4 Thermoelastic waves

In thermoelasticity the deformation and temperature ﬁelds aﬀect each

other. As a result governing equations for the strains and the tempera-

ture should be coupled. Very often the inﬂuence of the temperature on

the strains is negligibly small. However, thermoelasticity is important to

study attenuation of the waves that is observed in experiments. Of special

interest are the processes in polymers arising due to the laser irradiation

Kartashov and Bartenev (1988). Linear thermoelastic waves were studied

extensively, analytical results for a medium may be found in Kartashov and

Bartenev (1988); Nowacki (1975); Nowacki (1986b), while numerical simu-

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

194 Ampliﬁcation of Nonlinear Strain Waves in Solids

lations are performed in Berezovski et. al (2000); Berezovski and Maugin

(2001). Nonlinear surface wave attenuation is considered in Mayer (1990);

Mayer (1995), nonlinear bulk waves are studied in a medium in Engelbrecht

and Maugin (1996) and in a rod in Potapov and Semerikova (1988).

Derivation of the equations may be done by various approaches. First,

the balance laws are used. In contrast to liquid, the transfer of heat in

solids is caused by the heat conduction only. Usually the equation of the

heat conduction is obtained from the energy conservation law Engelbrecht

(1983); Engelbrecht and Nigul (1981); Landau et. al (1987); Maugin (1995);

Maugin (1999); Nowacki (1975); Nowacki (1986b):

T

∂S

∂t

= div(κ∇T), (6.58)

where S is the entropy per unit volume, T is the absolute temperature,

κ is the thermal conductivity. Usually the last coeﬃcients is assumed to

be constant while the entropy is expressed through the temperature and

the displacements u

i

. Then Eq. (6.58) is rewritten in the form, see, e.g.,

Landau et. al (1987),

C

v

∂T

∂t

+

C

p

−C

v

α

T

∂

∂t

div(u) = κ∇T, (6.59)

where C

p

, C

v

are the speciﬁc heat per unit volume at ﬁxed density and

volume respectively, α

T

is the heat extension coeﬃcient. The Fourier law

of thermal conduction, Q = −κ∇T, is used to obtain Eqs. (6.58), (6.59).

However, it predicts an inﬁnite velocity of a thermal wave. In order to

describe the ﬁnite velocity, the modiﬁcation of the law is needed. In par-

ticular, the generalized law has been proposed in Likov (1967)

Q = −κ∇T −τ

∗

Q

t

,

that takes into account an inertia of the heat ﬂux, τ

∗

is a time of the heat

ﬂux relaxation.

Balance of linear momentum provides the second equation of thermoe-

lasticity,

ρ

0

u

tt

= DivP

∗

, (6.60)

where P

∗

is the Piola-Kirchoﬀ tensor expressed through the free energy

density W(u,T), P

∗

= ∂W/∂∇u. Usually only the term linear in ∇u,

α

T

(T − T

0

)∇u, is included into W. Hence Eq.(6.60) contains the linear

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 195

term depending on the temperature while nonlinearity is caused by the

same reasons as in pure elastic case, see Chapter 3.

Equations of thermoelasticity may be derived using the variational

methods Fares (2000); Maugin (2000); Nowacki (1975); Nowacki (1986b);

Rayleigh (1945). However, usually only Eq.(6.60) is derived from the

Hamilton principle while the heat equation is introduced additionally Fares

(2000). It was Maugin (2000) who proposed the modiﬁcation of the free

energy density so as to get all ﬁeld equations, balance laws and constitutive

relations for the theory of materially inhomogeneous, ﬁnitely deformable,

thermoelastic conductors of heat, in the same manner. The main idea is to

assume a dependence upon an additional variable called thermacy whose

temporal derivative is the temperature. As a result all equations obtained

turn out strict conservation laws. Classical dissipative thermoelastic equa-

tions are obtained by isolation the contribution of the thermacy.

6.4.1 Nonlinear waves in thermoelastic medium

Below we follow Engelbrecht and Maugin (1996) where one-dimensional

model is considered using the notion of thermodynamic internal variables.

The evolution nonlinear equation is obtained for the observable variable

(longitudinal strain v = U

x

(x, t)) in a medium when temperature T(x, t)

eﬀects are considered as internal process. The equation of motion,

ρ

0

U

tt

−(λ+2µ)U

xx

−[3(λ+2µ)+2l+4m]U

x

U

xx

= −(3λ+2µ)α

T

T

x

, (6.61)

is coupled with the energy equation for the temperature,

ρ

0

C

v

T

t

+T

0

(3λ + 2µ)α

T

U

xt

= κT

xx

. (6.62)

Single nonlinear model equation may be obtained in case of a weak coupling,

ε = (3λ + 2µ)α

T

/ρ

0

, ε << 1. Introducing fast and slow variables, ξ =

c

0

t −X, τ = εX, one obtains for w = U

t

w

t

+a

1

ww

ξ

+ Λ w = 0, (6.63)

where

a

1

=

3(λ + 2µ) + 2l + 4m

(λ + 2µ)εc

0

, Λ =

T

0

((3λ + 2µ)α

T

)

2

2ερ

0

c

0

κ

, c

2

0

=

λ + 2µ

ρ

0

.

Exact solution of Eq.(6.63) is known Whitham (1974), it may include shocks

when the gradient of the initial excitation is large enough.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

196 Ampliﬁcation of Nonlinear Strain Waves in Solids

When the coupling is not weak, Eqs. (6.61), (6.62) are transformed by

another way, the case was not considered in Engelbrecht and Maugin (1996).

First, the temperature gradient, T

x

, is expressed through the displacement

using Eq.(6.61). Then it is substituted into Eq.(6.62) diﬀerentiated with

respect to x, yielding nonlinear equation for U

(ρ

0

C

v

c

2

0

+Λ

1

)U

xxt

−ρ

0

C

v

U

ttt

+a

2

(U

2

x

)

xt

+κ(U

xxtt

−c

2

0

U

xxxx

−a

2

(U

2

x

)

xxx

) = 0,

(6.64)

where

a

2

=

3(λ + 2µ) + 2l + 4m

2ρ

0

, Λ

1

=

T

0

((3λ + 2µ)α

T

)

2

ρ

0

.

In the reference frame, ξ = x −V t, the ODE reduction of Eq.(6.64) is

written for the function v = U

ξ

∂

2

∂ξ

2

_

b

1

v +b

2

v

2

+b

3

v

ξ

+b

4

(v

2

)

ξ

_

= 0,

with

b

1

= V (ρ

0

C

v

V

2

−ρ

0

C

v

c

2

0

−Λ

1

), b

2

= −a

2

V,

b

3

= κ(V

2

−c

2

0

), b

4

= −a

2

κ.

There is a similarity with the Burgers equation model if the last non-

linear term is negligibly small. The Burgers model has been developed in

Engelbrecht (1983), with the relaxation tame taken into account. Compar-

ison of the Burgers equation model and the model governed by Eq.(6.63)

is done in Engelbrecht and Maugin (1996).

6.4.2 Longitudinal waves in thermoelastic rod

As already noted, thermal conduction weakly aﬀects strain waves in solids.

More promising looks the inﬂuence of a heat transfer through the lateral

surface of a wave guide, especially caused by a laser irradiation, see Kar-

tashov and Bartenev (1988); Mirzoev et. al (1996) and references therein.

The inﬂuence of the external heat transfer on nonlinear longitudinal strain

waves in a rod has been studied in Potapov and Semerikova (1988). Esti-

mations done there demonstrate the dominant role of the heat transfer in

comparison with the thermal conduction mechanism.

The coupled equations of thermoelasticity are used in Potapov and Se-

merikova (1988) with the heat transfer boundary conditions on the rod

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection 197

lateral surface being taken into account:

U

tt

−c

2

s

U

xx

−β U

x

U

xx

+ 0.5R

2

(ν(1 −ν)U

ttxx

−νc

2

s

U

xxxx

) = −α

T

c

2

s

T

x

,

(6.65)

ρ

0

C

v

R T

t

−κR T

xx

+ 2h(T −T

0

) = −α

T

ET

0

R U

xt

, (6.66)

where β is a nonlinear coeﬃcient, β = ρ

−1

0

(3E+l(1−2ν)

3

+4m(1−2ν)(1+

ν) + 6nν

2

), R is a radius of the rod, T

0

is a constant temperature of an

external medium, h is a heat transfer coeﬃcient, c

2

s

= E/ρ

0

is a velocity of

the linear waves in a rod. Note that the dispersion terms coeﬃcient from

Potapov and Semerikova (1988) are corrected here in accordance with the

procedure from Chapter 3.

Let us diﬀerentiate Eq.(6.66) with respect to x and substitute T

x

from

Eq.(6.65). Then the governing equation for longitudinal displacements

holds,

∂

∂t

¦ρ

0

U

tt

−E

_

1 +

α

2

T

ET

0

ρ

0

C

v

_

U

xx

−β U

x

U

xx

+

ρ

0

ν(1 −ν)R

2

2

U

ttxx

−

νER

2

2

U

xxxx

¦ =

_

κ

ρ

0

C

v

∂

2

∂x

2

−

2h

ρ

0

C

v

R

_

¦ρ

0

U

tt

−EU

xx

−

β U

x

U

xx

+

ρ

0

ν(1 −ν)R

2

2

U

ttxx

−

νER

2

2

U

xxxx

¦. (6.67)

We are dealing with elastic strain waves whose magnitude B is small,

B << 1, and with long waves with typical length L, so as R/L << 1. As

usual we consider the case of a balance between nonlinearity and dispersion

introducing small parameter ε = B = (R/L)

2

. Contribution of the terms

in the r.h.s. of Eq.(6.67) is weak, they may be considered as small pertur-

bations. Then nonlinear and dispersive terms may be omitted in the r.h.s.

Obviously,

U

xx

=

ρ

2

0

C

v

E(ρ

0

C

v

+α

2

T

ET

0

)

U

tt

+O(ε).

Hence the simplest equation accounting for the inﬂuence of both the

heat transfer and thermal conduction on longitudinal strain waves v = U

x

is

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

198 Ampliﬁcation of Nonlinear Strain Waves in Solids

v

tt

−c

2

s

(1 +

α

2

T

c

2

s

T

0

C

v

)v

xx

−

β

2ρ

0

(v

2

)

xx

+

ν(1 −ν)R

2

2

v

ttxx

−

νc

2

s

R

2

2

v

xxxx

=

α

2

T

c

2

s

T

0

C

v

+α

2

T

c

2

s

T

0

_

κ

ρ

0

C

v

v

ttx

−

2h

ρ

0

C

v

R

v

t

_

. (6.68)

Certainly Eq.(6.68) may be studied by the methods developed in Chap-

ter 2 to account for an ampliﬁcation, attenuation and selection of longitu-

dinal thermoelastic waves.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Bibliography

Ablowitz, M. J. and Segur, H.(1981) Solitons and the Inverse Scattering Trans-

form, SIAM, Philadelphia.

Akhmediev, N.N. and Ankiewicz, A. (1997) Solitons. Nonlinear pulses and beams,

Chapman & Hall, London.

Alekseev, V.N. and Rybak S.A. (2002) ”Equations of state for viscoelastic bio-

logical media”, Acoust. J. 48, 511.

Alexeyev, A.A.(1999) ”Classical and non-classical interactions of kinks in some

bubbly medium”, J. Phys. A 32, 4419.

Bateman, G. and Erdelyi, A. (1953-54). Higher Transcendental Functions, vol. 3,

McGraw Hill, New York.

Belokolos, A., Bobenko, A., Enol’skij, V., Its, V. and Matveeev, V. (1994)

Algebro- Geometrical Approach to Nonlinear Integrable Equations, Springer,

Berlin.

Benilov, E.S., Grimshaw, R., Kuznetsova, E.P.(1993) ” The generation of radiat-

ing waves in a singularly-perturbed KdV equation”, Physica D 69 270.

Berezin, Ya. A. (1987) Modelling Non-Linear Wave Processes. VNU Science

Press, Utrecht.

Berezovski, A., Engelbrecht, J., Maugin, G.A. (2000) ”Thermoelastic wave prop-

agation in inhomogeneous media”, Arch. Appl. Mech. 70, 694.

Berezovski, A. and Maugin, G.A. (2001) ”Simulation of Thermoelastic Wave

Propagation by Means of a Composite Wave-Propagation Algorithm”, J.

Compt. Phys. 168, 249.

Bhatnagar, P.L. (1979) Nonlinear Waves in One-Dimensional Dispersive Sys-

tems, Claredon Press, Oxford.

Biryukov, S.V., Gulyaev, V.V., Krylov, V.V. and Plessky, V.P. (1991) Surface

acoustic waves in inhomogeneous media, Nauka, Moscow (in Russian).

Bland, D. R. (1960) The Theory of Linear Viscoelasticity, Oxford .

Bland, D.R.(1969) Nonlinear Dynamic Elasticity, Blaisdell, Waltham, Mass et al.

Boyd, J. P. (1991) ”Weakly non-local solitons for capillary-gravity waves: the

ﬁfth-degree Korteweg-de Vries equation”, Physica D 48, 129.

Bridges, T. (1986) ”Cnoidal standing waves and the transition to the travelling

hydravlic jump”, Phys. Fluids 29, 2819.

199

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

200 Book Title

Bukhanovsky, A.V. and Samsonov, A.M. (1998) ”Numerical Simulation of Nonlin-

ear Elastic waves in Inhomogeneous Wave Guides”, St. Petersburg, Preprint

of the Institute for High performance Computing and Data Bases No 6-98

(in Russian).

Bullough, R.K. and Caudrey, P.J. eds., (1980) Solitons, Springer, Berlin.

Burgers, J.M. (1948) ”A Mathematical model illustrating the theory of turbu-

lence”, Adv. Appl. Mech. 1, 171.

Byrd, P.F. and Friedman, M.D. (1954) Handbook of Elliptic Integrals for Engi-

neers and Physicists, Springer, Berlin.

Calogero, F. and Degasperis, A. (1982) Spectral Transform and Solitons, North-

Holland, Amsterdam.

Cariello, F. and Tabor, M.(1989) ”Painlev´e expansions for nonintegrable evolution

equations”, Physica D 39, 77.

Capriz, G. (1989) Continua with microstrucure, Springer Tracts in Nat. Phil.,

V.35, Springer, Berlin.

Cermelli, P. and Pastrone, F. (1997) ”Growth and decay of waves in microstruc-

tured solids”, Proc. Estonian Acad. Sci. Phys. Math. 46, 32.

Chang, H.-C., Demekhin, E.A., and Kopelevich, D.I. (1995) ”Stability of a Soli-

tary Pulse against Wave Packet Disturbances in an Active Media”, Phys.

Rev. Lett. 75, 1747.

Chow, K.W. (1995) ”A class of exact, periodic solutions of nonlinear envelope

equations”, J. Math. Phys. 36, 4125.

Christensen, R.M. (1971) Theory of Viscoelasticity, Academic Press, New York.

Christiansen, P.L., Eilbeck, J.C., Enolskii, V.Z., Kostov, N.A. (1995) ”Quasi-

periodic solutions of the coupled nonlinear Schr¨ odinger equations”, Proc.

R. Soc. Lond. A 451, 685.

Christou, M.A. and Christov, C.I. (2000) ”Fourier-Galerkin Method for Local-

ized Solutions of the sixth-order Generalized Boussinesq Equation”, Proc.

Intern. Conf. on Dynamical Systema and Diﬀerential Equations, May 18-

21, Atlanta, USA, 121.

Christou, M.A. and Christov, C.I. (2002) ”Fourier-Galerkin Method for Localized

Solutions of the Equations with Cubic Nonlinearity”, J. Comp. Anal. Appl.

4, 63.

Christov, C.I. (1994) ”Numerical Investigation of the long-time evolution and

interaction of localized waves”, in Fluid Physics, eds. Velarde, M.G. &

Christov, C.I., World Scientiﬁc, Singapore, 353.

Christov, C.I. and Maugin, G.A. (1995) ”An Implicit Diﬀerence Scheme for the

Long-Time Evolution of Localized Solutions of a Generalized Boussinesq

System”, J. Compt. Phys. 116, 39.

Christov, C.I., Maugin, G.A., Velarde, M.G. (1996) ”Well-posed Boussinesq

paradigm with purely spatial higher-order derivatives”, Phys.Rev. E 54,

3621.

Christov, C.I., Pontes, J., Walgraef, D., and Velarde, M.G. (1997) ”Implicit

time splitting for fourth-order parabolic equations”, Comput. Methods Appl.

Mech. Engrg. 148, 209.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 201

Christov, C.I. and Velarde, M.G. (1994) ”Inelastic Interactions of Boussinesq

Solitons”, Intern. J. Bif. Chaos 4, 1095.

Christov, C.I. and Velarde, M.G. (1995) ”Dissipative solitons”, Physica D 86,

323.

Cole, J.D. (1968) Perturbation Methods in Applied Mathematics, Blaisdell,

Waltham, Mass.

Conte, R. (1989) ”Invariant Painlev´e analysis of partial diﬀerential equations”,

Phys. Lett. A 140, 383.

Conte, R., Fordy, A., Pickering, A. (1993) ”A perturbative Painlev´e approach to

nonlinear diﬀerential equations”, Physica D 69, 33.

Crighton, D.G. (1995) ”Applications of KdV”, Acta Applicandae Mathematicae

39, 39.

Dai, H.-H. (1998) ”Model equations for nonlinear dispersive waves in a compress-

ible Mooney-Rivlin rod”, Acta Mech. 127, 193.

Dodd, R.K., Eilbeck, J.C., Gibbon, J.D. and Morris, H.C. (1982) Solitons and

Nonlinear Wave Equations, Academic Press, London et al.

Drazin, P.G. and Johnson, R.S. (1989) Solitons: an Introduction, Cambridge

Univ. Press, Cambridge.

Dreiden, G.V., Porubov, A.V., Samsonov, A.M., Semenova, I.V., and Sokurin-

skaya, E.V. (1995) ”Experiments in the propagation of longitudinal strain

solitons in a nonlinearly elastic rod”, Tech. Phys. Lett 21, 415.

Dreiden, G.V. , Porubov, A.V., Samsonov, A.M., Semenova, I.V. (2001) ”Reﬂec-

tion of a longitudinal strain solitary wave from the end face of a nonlinearly

elastic rod”, Tech.Phys. 46, no 5, 505.

Elmer, F.-J. (1997) ”Nonlinear dynamics of dry friction”, J. Phys. A: Math. Gen.,

30, 6057.

Engelbrecht, J. (1979) ”One-Dimensional Deformation waves in Nonlinear Vis-

coelastic Media”, Wave Motion 1, 65.

Engelbrecht, J. (1983) Nonlinear wave processes of deformation in solids, Pitman,

Boston.

Engelbrecht, J. ed. (1989) Nonlinear Waves in Active media, Springer-Verlag,

Berlin.

Engelbrecht, J. (1997) Nonlinear Wave Dynamics. Complexity and Simplicity,

Kluwer, The Netherlands.

Engelbrecht, J. and Braun, M. (1998) ”Nonlinear waves in nonlocal media”, Appl.

Mech. Rev. 51, No 8, 475.

Engelbrecht,J., Cermelli, P. and Pastrone, F. (1999) ”Wave hierarchy in mi-

crostructured solids”, In: Geometry, Continua and Microstructure. Ed. by

G.A. Maugin, Herman Publ., Paris, 99.

Engelbrecht, J. and Khamidullin, Y. (1988) ”On the possible ampliﬁcation of

nonlinear seismic waves”, Phys. Earth Planet. Inter. 50, 39.

Engelbrecht, J. and Maugin, G.A. (1996) ”Deformation waves in thermoelastic

media and the concept of internal variables”, Arch. Appl. Mech. 66, 200.

Engelbrecht, J. and Nigul, U. (1981) Nonlinear Deformation Waves, Nauka,

Moscow (in Russian).

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

202 Book Title

Erbay, S., Erbay, H.A., and Dost, S. (1991) ”Nonlinear wave modulation in mi-

cropolar elastic media-I. longitudinal waves” Int. J. Engng. Sci. 29, 845.

Eringen, A. C. (1968). Theory of micropolar elasticity. P. 622. In: Fracture. An

advanced treatise. V.2. Mathematical fundamentals. Ed. by H. Liebowitz,

Academic Press, New York.

Eringen, A.C. and Suhubi, E.S. (1964) ”Nonlinear theory of micro-elastic solids”,

Intern. J. Eng. Sci. 2, Part I, 189, Part II, 389.

Erofeev, V.I. (2002) Wave processes in solids with microstructure, World Scien-

tiﬁc, Singapore.

Erofeev, V.I. and Klyueva, N.V. (2002) ”Solitons and nonlinear periodic strain

waves in rods, plates and shells”, Acoust. J. 48, 725 (in Russian).

Erofeev, V.I. and Potapov, A.I. (1993) ”Longitudinal strain waves in non-linearly

elastic media with couple stresses”, Int. J. Nonl. Mech. 28, 483.

Fares, M.E. (2000) ”Generalized non-linear thermoelasticity for composite lami-

nated structures using a mixed variational approach”, Int. J. Nonl. Mech.

35, 439.

Feng, B.F. and Kawahara, T. (2000) ”Stationary travelling-wave solutions of an

unstable KdV-Burgers equation”, Physica D 137, 228.

Fletcher, C.A.J. (1984) Computational Galerkin Methods, Springer, Berlin.

Fowler, A.C. (1997) Mathematical Models in the Applied Sciences, Cambridge

Univ. Press, Cambridge.

Frantsevich, I.N., Voronov, F.F., and Bakuta, S.A. (1982) Elastic Constants and

Moduli of Metals and Nonmetals, Naukova Dumka, Kiev, (in Russian).

Galin, L.A. (1980) Contact problems of elasticity and viscoelasticity, Nauka,

Moscow (in Russian).

Garazo, A.N. and Velarde, M.G. (1991) ”Dissipative Korteweg-de Vries descrip-

tion of Marangoni-B´enard oscillatory convection”, Phys. Fluids A 3, 2295.

Godunov, S.K. and Ryaben’kii, V.S. (1973) Diﬀerence Schemes, Nauka, Moscow,

(in Russian).

Godunov, S.K. and Ryaben’kii, V.S. (1987) Diﬀerence Schemes: introduction to

the underlying theory of, North-Holland, Amsterdam.

Godunov, S.K. (1997) Ordinary Diﬀerential Equations with Constant Coeﬃcients,

AMS.

Goryacheva, I.G. (1998) Contact mechanics in tribology, Dordrecht etc., Kluwer.

Goryacheva, I.G. (2001) Mechanics of friction interaction, Nauka, Moscow (in

Russian).

Grimshaw, R., Malomed, B. and Benilov, E. (1994) ” Solitary waves with damped

oscillatory tails: an analysis of the ﬁfth-order Korteweg-de Vries equation”,

Physica D 77, 4783.

Grimshaw R, Pelinovsky E. and Talipova T. (1999) ”Solitary wave transformation

in a medium with sign-variable quadratic nonlinearity and cubic nonlinear-

ity”, Physica D 132, 40.

Gulyaev, Yu.V. and Polsikova, N.I. (1978) ”Shear surface acoustic waves on a

cylindriucal surface of a solid body covered by a layer of a foreign material”,

Sov. Phys- Acoust. R. 24, 287.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 203

Gusev, A. A. (1988) Two dilatancy-based models to explain coda-wave precursors

and P/S spectral ratio. Tectonophysics 152, 227.

H¨ahner, G. and Spencer, N. (1998) ”Rubbing and scrubbing”, Physics Today,

No 9, 22.

Harker, A.N. (1988) Elastic Waves in Solids with Application to Nondestructive

Testing of Pipelines, Hilger & Bristol, Philadelphia.

Hunter, J.K. and Scheurle, J. (1988) ”Existence of perturbed solitary wave solu-

tions to a model equation for water waves”, Physica D 32, 253.

Ince, E. (1964) Ordinary Diﬀerential Equations, Dover, New York.

Jeﬀrey, A. and Kawahara, T. (1982) Asymptotic Methods in Nonlinear Wave

Theory, Pitman, London.

Jeﬀrey, A. and Engelbrecht, J., eds. (1994) Nonlinear Waves in Solids, Springer,

Wien.

Johnson, K.L. (1985) Contact Mechanics, Cambridge Univ. Press, Cambridge.

Kalker, J.J., Dekking, F.M., and Vollebregt, E.A.H. (1997) ”Simulation of Rough,

Elastic Contacts”, J. Appl. Mech. 64, 361.

Kano, K. and Nakayama, T. (1981) ”An Exact Solution of the Wave Equation

u

t

+ uu

x

− u

5x

= 0”, J. Phys. Soc. Jpn. 50, 361.

Karpman, V. I. (1993) ”Radiation by solitons due to higher-order dispersion”,

Phys.Rev. E 47, 2073.

Karpman, V.I. and Vanden-Broeck, J.-M. (1995) ”Stationary solitons and stabi-

lization of the collapse described by KdV-type equations with high nonlin-

earities and dispersion”, Phys. Lett. A 200, 423.

Karpman, V.I. (1998) ”Evolution of radiating solitons described by the ﬁfth order

Korteweg-de Vries type equations”, Phys. Lett. A 244, 394.

Karpman, V.I. (2001) ”Stability of solitons described by KdV-type equations with

higher-order dispersion and nonlinearity (analytical results)”, Phys. Lett. A

284, 246.

Kartashov, E.M. and Bartenev, G.M. (1988) ”Dynamic eﬀects in solid under

interactions with intensive heat ﬂuxes”, Itogi Nauki i tekhniki, Ser.: Khimia

i tekhnologia visokomolekuliarnikh soedinenii” 25, 3 (in Russian).

Kascheev, V.N. (1990) Heuristic Approach to Obtain Solutions to Nonlinear Equa-

tions of Solitonics, Zinatne, Riga (in Russian).

Kawahara, T. (1972) ”Oscillatory Solitary Waves in Dispersive Media”, J. Phys.

Soc. Jpn. 33, 260.

Kawahara, T. (1983) ” Formation of saturated solitons in a nonlinear dispersive

system with instability and dissipation” Phys. Rev. Lett. 51, 381.

Kawahara, T. and Takaoka, M. (1988) ”Chaotic Motions in an Oscillatory Soliton

Lattice”, J. Phys. Soc. Jpn. 57, 3714.

Kerr, A.D. (1964) ”Elastic and viscoelastic foundation models”, J. Appl. Mech.

31, 491.

Kliakhandler, I.L. (1999) http://www.math.mtu.edu/˜igor

Kliakhandler, I.L. , Porubov, A.V. and Velarde, M.G. (2000) ”Localized ﬁnite-

amplitude disturbances and selection of solitary waves”, Phys. Rev. E 62,

4959.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

204 Book Title

Kodama, J. and Ablowitz, M. (1981) ”Perturbation of solitons and solitary

waves”, Stud. Appl. Math. 64, 225.

Korpel, A. and Banerjee, P. (1984) ”Heuristic guide to nonlinear dispersive wave

equations and soliton-type solutions”, IEEE 72, 1109.

Korteweg, D.J. and de Vries, G. (1895) ”On the change of the form of long waves

advancing in a rectangular channel and on a new type of long stationary

waves”, Phil. Mag. 5, 422.

Kosevich, A.M. (1981) Physical Mechanics of Real Crystals, Naukova Dumka,

Kiev (in Russian).

Kosevich, A.M. and Savotchenko, S.E. (1999) ”Features of dynamics of one-

dimensional discrete systems with non-neighbouring interactions, and role

of higher-order dispersion in solitary wave dynamics”, Low Temp. Phys. 25,

737 (in Russian).

Kostov, N.A. and Uzunov, I.M. (1992) ”New kinds of periodical waves in bire-

fringent optical ﬁbers”, Opt. Comm. 89, 389.

Koz´ak, J. and

ˇ

Silen´ y, J. (1985) ”Seismic events with non-shear component. I.

Shallow earthquakes with a possible tensile source component”. PAGEOPH

123, 1.

Kudryashov, N.A. (1988)” Exact solitonic solutions of the generalized evolution

equation of wave dynamics”, J. Appl. Math. Mech. 52, 361.

Landau L.D., Lifshitz E.M., Kosevich, A.M., and Pitaevsky, L.P. (1987) Theory

of Elaticity, Pergamon Press, Oxford.

Levi, D. and Winternitz, P. , eds. (1992) Painlev´e Transcendents: Their Asymp-

totics and Physical Applications, Plenum Press, New York and London.

Likov, B. Ya. (1967) Theory of thermal conduction, Nauka, Moscow (in Russian).

Lou, S., Huang, G. and Ruan, H. (1991) ”Exact solitary waves in a convecting

ﬂuid”, J. Phys. A 24, L587.

Love, A.E.H. (1927) A Treatise on the Mathematical Theory of Elasticity, Uni-

versity Press, Cambridge.

Lurie, A.I. (1990) Nonlinear Theory of elasticity, Amsterdam, Elsevier.

Marchant, T.R. (1996) ”The evolution and interaction of Marangoni-B´enard soli-

tary waves”, Wave Motion 23, 307

Maugin, G.A. (1990) ”Internal Variables and Dissipative Structures”, J. Non-

Equilib. Thermodyn. 15, 173.

Maugin, G.A. (1993) Material Inhomogeneities in Elasticity, Chapman & Hall,

London.

Maugin, G.A. (1995) ”Material forces: Concepts and applications”, Appl. Mech.

Rev. 48, 213.

Maugin, G.A. (1999) The Thermomechanics of Nonlinear Irreversible Behaviors.

An Introduction., World Scientiﬁc, Singapore.

Maugin, G.A. (2000) ”Towards an analytical mechanics of dissipative materi-

als” Rendiconti del Seminario Matematico dell’Universita’ e Politecnico di

Torino 58, 171.

Maugin, G.A. and Muschik, W. (1994) ”Thermodynamics with internal variables”

J.Non-Equil. Thermodyn. 19, 217.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 205

Mayer, A (1990) ”Thermoelastic attenuation of surface acoustic waves in coated

elastic media”, J. Appl. Phys. 68, 5913.

Mayer, A (1995) ”Surface Acoustic Waves in Nonlinear Elastic Media”, Phys.

Reports 256, 257.

Miloserdova, I.V. and Potapov, A.I. (1983) ”Nonlinear standing waves in a rod

of ﬁnite length”, Sov. Phys. Acoust., 29, 306.

Mindlin, R.D. (1964) ”Microstructure in linear elasticity”, Arch. Rat. Mech. Anal.

No1, 51.

Mirzoev, F. Kh., Panchenko, V. Ya, Shelepin, L.A. (1996) ”Laser Control of

Processes in Solids”, Physics- Uspekhi 39, 1.

Mirzoev, F. Kh., and Shelepin, L.A. (2001) ”Nonlinear Strain Waves and Densities

of Defects in Metal Plates Induced by External Energy Fluxes”, Tech. Phys.

46, 952.

Morton, K.W. and Mayers, D.F. (1994) Numerical Solution of Partial Diﬀerential

Equations, Cambridge University Press, Cambridge.

Murnaghan, F.D. (1951) Finite Deformations of an Elastic Solid, J. Wiley, New

York.

Nakamura, A. (1979) ”A direct method of calculating periodic wave solutions

to nonlinear evolution equations. I. exact two-periodic wave solution”, J.

Phys. Soc. Jpn. 47, 1701.

Nagashima, H. and Kuwahara, M. (1981) ”Computer Simulation of Solitary

Waves of the Nonlinear Wave Equation ut + uux − γ

2

u5x = 0” J. Phys.

Soc. Jpn. 50, 3792.

Nayfeh, A.H. (1973) Perturbation Methods, J. Wiley & Sons, New York.

Nekorkin, V.I. and Velarde, M.G. (1994) ”Solitary waves, soliton bound states

and chaos in a dissipative Korteweg-de Vries equation”, Int. J. Bif. Chaos

4, 1135.

Nepomnyashchy, A.A. and Velarde, M.G. (1994) ”A three-dimensional description

of solitary waves and their interaction in Marangoni-B´enard layers”, Phys.

Fluids A 6, 187.

Newell, A (1985) Solitons in Mathematics and Physics, SIAM, Philadelphia.

Newell, A., Tabor, M., Zeng, Y.B. (1987) ”A Uniﬁed Approach to Painlev´e Ex-

pansions”, Physica D 29, 1.

Newille, E.H. (1951) Jacobian Elliptic Functions, Claredon Press, Oxford.

Nikitin, L.V. (1998) Statics and dynamics of solids with external dry friction,

Moscovskii Licei, Moscow (in Russian).

Nikolaev, A.V. (1989) ”Scattering and dissipation of seismic waves in presence of

nonlinearity”, PAGEOPH 131, 687.

Nikolova, E.G. (1977) ”On the eﬀective surface tension of solid bodies”, Sov.

Phys. JETP 45, 285.

McNiven, H.D. and McCoy, J.J. (1974) ”Vibrations and wave propagation in

rods”, in: G. Herrmann ed., R. Mindlin and Applied Mechanics, Pergamon,

New York , p. 197.

Nowacki, W. (1975) Theory of Elasticity, Mir, Moscow (in Russian).

Nowacki, W. (1986a) Theory of Asymmetric Elasticity, Pergamon, Oxford.

Nowacki, W. (1986b) Thermoelasticity, Pergamon, Oxford.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

206 Book Title

Oliner, A.A. ed. (1978) Surface Acoustic Waves. Springer-Verlag, Berlin.

Oron, A. and Rosenau, P. (1997) ”Evolution and formation of dispersive-

dissipative patterns”, Phys. Rev. E 55, R1267.

Ostrovsky, L.A. and Potapov, A.I. (1999) Modulated waves in linear and nonlinear

media, The John Hopkins University Press, Baltimore.

Parker, D.F. (1994) ”Nonlinear Surface Acoustic Waves and Waves on Stratiﬁed

Media”, In: A.Jeﬀrey and J. Engelbrecht, eds. , Nonlinear Waves in Solids,

Springer, Berlin.

Parker, D.F. and Maugin, G.A. , eds. (1987) Recent Developments in Surface

Acoustic waves, Springer, Berlin.

Parker, D.F. and Mayer, A. (1991) ” Dissipation of Surface Acoustic Waves”,

In: D.Fusco and A. Jeﬀrey, eds., Nonlinear Waves and Dissipative Eﬀects,

Longman, London.

Parker, D.F. and Tsoy, E.N. (1999) ”Explicit Solitary and Periodic Solutions for

Optical Cascading”, J. Eng. Math. 36, 149.

Parkes, E.J. and Duﬀy, B.R. (1996) ”An automated tanh-function method for

ﬁnding solitary wave solutions to nonlinear evolution equations”, Computer

Phys. Comm. 98, 288.

Pasternak, P.L. (1954) On a New Method of Analysis of an elastic Foundation

be Means of Two Foundation Constants (in Russian), Gosudarstvennoe Iz-

datel’stvo Litearturi po Stroitel’stvu i Arkhitekture, Moscow, USSR (in

Russian).

Porubov, A.V. (1993) ”Exact travelling wave solutions of nonlinear evolution

equation of surface waves in a convecting ﬂuid”, J. Phys. A: Math. Gen.

26, L797.

Porubov, A.V.(1995) Free Surface Nonlinear Waves on a viscous inhomogeneous

liquid layer. Ph.D. Thesis, State Technical University, Saint Petersburg ( in

Russian).

Porubov, A.V. (1996) ”Periodical solution to the nonlinear dissipative equation

for surface waves in a convective liquid layer”, Phys. Lett. A 221, 391.

Porubov, A.V. (2000) ”Strain solitary waves in an elastic rod with microstruc-

ture”, Rendiconti del Seminario Matematico dell’Universita’ e Politecnico

di Torino 58, 189.

Porubov, A.V., Gursky, V.V. and Maugin, G.A. (2003) ”Selection of localized

nonlinear seismic waves”, Proc. Estonian Acad. Sci., Phys. Math. 52, No

1.

Porubov, A.V., Maugin, G.A., Krzhizhanovskaya, V.V. and Gursky, V.V. (2002)

”Inﬂuence of higher- order dispersion and nonlinearity on solitary wave

formation from an initial localized pulse”, submitted for publication

Porubov, A.V. and Parker, D.F. (1999) ”Some General Periodic Solutions to

Coupled Nonlinear Schr¨ odinger Equations”, Wave Motion 29, 97.

Porubov, A.V. and Parker, D.F. (2002) ”Some General Periodic Solutions for

Optical Cascading”, Proc. Roy. Soc. A 458, 2139.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 207

Porubov, A.V. and Pastrone, F. (2001) ”Inﬂuence of macro- and microdissipation

on nonlinear strain solitary wave evolution”. Quaderni del Dipartimento di

Matematica N. 56, University of Turin, Turin.

Porubov, A.V. and Samsonov, A.M. (1993) ”Reﬁnement of the model for the

propagation of longitudinal strain waves in a rod with nonlinear elasticity”,

Tech. Phys. Lett. 19, 365.

Porubov, A.V. , Samsonov, A.M. , Velarde, M.G. and Bukhanovsky, A.V. (1998)

”Strain solitary waves in an elastic rod embedded in another elastic external

medium with sliding”, Phys.Rev. E 58, 3854 .

Porubov, A.V. and Velarde, M.G. (1999) ”Exact Periodic Solutions of the Com-

plex Ginzburg-Landau Equation”, J. Math. Phys. 40, 884.

Porubov, A.V. and Velarde, M.G. (2000) ”Dispersive-dissipative solitons in non-

linear solids”, Wave Motion 31, 197.

Porubov, A.V. and Velarde, M.G. (2002) ”Strain kinks in an elastic rod embedded

in a viscoelastic medium”, Wave Motion 35, 189.

P¨oschel, T. and Herrmann, H.J. (1993) ”A simple geometrical model for solid

friction”, Physica A 198, 441.

Potapov, A.I. and Rodyushkin, V.M. (2001) ”Experimental study of strain waves

in materials with microstructure”, Acoust. J. 47, 347.

Potapov, A.I. and Semerikova, N.P. (1988) ”Nonlinear longitudinal waves in a rod

taking into account the interaction between strain and temperature ﬁelds”,

Sov. Appl. Mech. and Tech. Phys., No1, 54.

Qaisar, M. (1989) ”Attenuation Properties of Viscoelastic Material” PAGEOPH

131, 703.

Rayleigh Lord (J.W. Strutt) (1945) Theory of Sound, Vol. 1, Reprint, Dover, New

York.

Rednikov, A. Ye., Velarde, M.G., Ryazantsev, Yu. S., Nepomnyashchy, A.A.,

and Kurdyumov V.N., (1995) ”Cnoidal Wave Trains and Solitary Waves

in a Dissipation Modiﬁed Korteweg- de Vries Equation”, Acta Applicandae

Mathematicae 39, 457.

Richtmyer, R.D. and Morton, K.W. (1967) Diﬀerence Methods for Initial-Value

Problems, Wiley & Sons, New York.

Sachdev, P.L. (1987) Nonlinear diﬀusive waves, Cambridge Univ. Press, Cam-

bridge.

Salupere, A., Maugin, G.A., Engelbrecht, J. (1994) ”Korteweg-de Vries soliton

detection from a harmonic input”, Phys. Lett. A 192, 5.

Salupere, A. , Maugin, G.A. and Engelbrecht, J. (1997) ”Solitons in systems with

a quartic potential and higher-order dispersion”, Proc. Estonian Acad. Sci.,

Phys. Math. 46, 118.

Salupere, A. , Engelbrecht, J. and Maugin, G.A. (2001) ”Solitonic structures in

KdV-based higher-order systems”, Wave Motion 34, 51.

Samarskii, A.A. and Nikolaev, E.S. (1989) Numerical Methods for Grid Equations,

Birkh¨ auser, Basel.

Samsonov, A.M. (1988) ”On existence of longitudinal strain solitons in an inﬁnite

nonlinearly elastic rod”, Sov. Phys.-Doklady, 33, 298.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

208 Book Title

Samsonov, A.M. (1995) ”Travelling wave solutions for nonlinear waves with dis-

sipation”, Appl.Anal. 57, 85.

Samsonov, A.M. (2001) Strain soltons in solids and how to construct them, Chap-

man & Hall/CRC.

Samsonov, A.M., Dreiden, G.V., Porubov, A.V. and Semenova, I.V. (1998)

”Longitudinal-strain soliton focusing in a narrowing nonlinearly elastic

rod”, Phys. Rev. B 57, 5778.

Samsonov, A.M., Dreiden, G.V., and Semenova, I.V. (2003) ”On existence of bulk

solitary waves in plexiglas.” Proc. Estonian Acad. Sci., Phys. Math. 52,

No 1.

Savin, G.N., Lukashev, A.A. and Lysko, E.M. (1973) ”Elastic wave propagation

in a solid with microstructure” Soviet Appl. Mech. 15, 725.

Savin, G.N., Lukashev, A.A., Lysko, E.M. , Veremeeenko, S.V., and Agasiev,

G.G. (1973) ”Elastic wave propagation in the Cosserat continuum with

constrained particle rotation”, Soviet Appl. Mech. 15.

Sawada, K. and Kotera, T. (1974) ”A method for ﬁnding N-soliton solutions of

the KdV equation and KdV-like equation”, Prog. Theor. Phys. 51 , 1355.

Shevyakhov, N.S. (1977) ”Love waves on the surface of cylinder covered by a

layer”, Sov. Phys. -Acoust. R. 23, 86.

Sillat, T. (1999) ”Wave propagation in dissipative microstructured materials”.

Thesis of Master of Science. Technical University, Tallinn.

Slyunyaev, A.V. and Pelinovsky, E. N. (1999) ”Dynamics of large amplitude

solitons” JETP 89 173.

Soerensen, M.P., Christiansen, P.L., Lomdahl, P.S. (1984) ”Solitary waves in

nonlinear elastic rods, I”, J. Acoust. Soc. Amer. 76, 871.

Soerensen, M.P., Christiansen, P.L., Lomdahl, P.S., and Skovgaard, O. (1987)

”Solitary waves in nonlinear elastic rods, II”, J. Acoust. Soc. Amer. 81, 6.

Spencer, A.J.M. (1971) In ”Treatise on Continuum Physics” (A.C. Eringen, ed.),

Vol. I, Theory of invariants, Chap.3. Academic, New York.

Sokurinskaya, E.V. (1991) ”Study of nonlinear travelling strain waves in one-

dimensional elastic wave guide”. Ph.D. Dissertation, Technical University,

St.Petersburg (in Russian).

Stefa´ nski, A., Wojewoda, J. and Wiercigroch, M. (2000) ”Numerical analysis of

duﬃng oscillator with dry friction damper”, Mech. & Mech. Ing. 4, No2,

127.

Svendsen, I. and Buhr-Hansen, J.B. (1978) ”On the deformation of periodic waves

over a gently sloping bottom”, J. Fluid Mech. 87, 433.

Thomas, L.H. (1949) ”Elliptic problems in linear diﬀerence equations over a net-

work”, Watson Sci. Comput. Lab. Rept., Columbia University, New York.

Velarde, M.G., Nekorkin, V.I. and Maksimov, A.G. (1995) ”Further results on the

evolution of solitary waves and their bound states of a dissipative Korteweg-

de Vries equation”, Int. J. Bif. Chaos 5, 831.

Vlieg-Hultsman, M. and Halford, W. (1991) ”The Kotreweg-de Vries-Burgers

equation: a reconstruction of exact solutions”, Wave Motion 14, 267.

Weiss, J. , Tabor, M. and Carnevale, G. (1983) ”The Painlev´e property for partial

diﬀerential equations”, J. Math. Phys. 24, 522.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 209

Whitham, G. B. (1974) Linear and Nonlinear Waves, Wiley, New York .

Whittaker, E.T. and Watson, G.N. (1927) A Course of Modern Analysis, Cam-

bridge, University Press.

Winkler, E. (1867) Die Lehre von der Elasticitaet und Festigkeit, Prag, Domini-

cicus.

Wolfram, S. (1999) The Mathematica book. Fourt Edition, Addison-Wesley.

Zhurkov, S.N. (1983) ”Dilaton Mechanism of the Strength of Solids”, Sov. Phys.

Solid State 25, 1797.

Zwillenger, D. (1989) Handbook of Diﬀerential Equations, Acad. Press, Boston.

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

210 Book Title

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Index

acoustic resistance, 85

action functional, 66

analysis of singular points, 34

ansatz, 31, 34

auto-B¨acklund transformation, 36

balance

between nonlinearity and

dispersion, 1, 27, 71, 79, 87,

100, 105, 127, 129, 136, 142,

156, 157, 169, 197

between nonlinearity and

dissipation, 1, 9, 129, 136,

138, 156, 169, 173

between nonlinearity, dispersion

and dissipation, 1, 10, 142,

169, 175

of linear momentum, 194

balance laws, 194

blow-up, 51, 53, 133

body force, 178

Boussinesq equation, 57, 129

breather-like solution, 47

Burgers

equation, 9, 12, 26, 136, 142, 146,

196

kink, 147, 148

perturbed kink, 149

Cauchy-Green tensor, 64, 115, 165

cnoidal wave, 12, 13, 27, 38, 41, 49

conservation laws, 135, 194

Cosserat model, 116

Coupled nonlinear Schr¨odinger

equations, 42

D’Alembert solution, 80

dilation mechanism, 178

dilaton, 178

dissipation-modiﬁed

double dispersive equation, 127

Korteweg-de Vries equation, 6, 11,

12, 27, 36, 49, 52, 54, 134,

174

dissipative elements, 124

double-dispersive equation, 57, 71,

79, 90, 103, 155, 156

dynamical system, 134, 142

elementary work, 79, 98, 123, 126, 152

elliptic

integrals, 12, 43, 61

Jacobi functions, 12, 32, 51, 155

Weierstrass function, 32, 36, 42, 59,

128, 155

energy

free, 65, 163, 194

inﬂux, 25, 123, 153, 166, 178, 189

internal, 65, 178

kinetic, 67, 71, 103, 116, 135

potential, 71, 102, 116, 122, 137

seismic, 179

entropy, 194

envelope wave solution, 44

211

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

212 Book Title

experiments on

strain solitary wave ampliﬁcation,

93

strain solitary wave propagation,

75

strain solitary wave reﬂection, 84

ﬁfth-order KdV equation, 3, 7, 15, 53

generalized, 3, 15, 52

ﬁnite-diﬀerence methods, 52

foundation models, 124

Fredholm alternative, 157

friction contact, 123, 153

fringe shift, 76

geophysical medium, 178

Ginzburg-Landau equation, 45

Hamilton principle, 66, 70, 79, 88,

117, 195

heat conduction, 194

heat transfer, 196, 197

holographic interferometry method,

75

inertia of the microstructure, 165, 168

internal variables, 163, 195

interstitial atom, 188

invariants of the strain tensor, 65

isotropic, 63, 65, 87, 96, 117, 125,

137, 189

Kawahara equation, 6

Kerr model, 124, 125, 136, 137, 140

Korteweg-de Vries equation, 2, 12,

179

modiﬁed, 11, 136

Korteweg-de Vries-Burgers equation,

10, 26

laser radiation, 189

Le Roux model, 114, 117

Love hypothesis, 70

matching asymptotic procedure, 47

microdisplacement, 165

microﬁeld, 164

gradient, 165

Mindlin model, 117

Mooney-Rivlin model, 66

movable singularities, 35

Murnaghan

ﬁve constants model, 65, 71, 96,

116, 117, 126, 165

moduli, 65, 73, 98, 116, 137, 192

nine constants model, 65, 137

nonlinearity

cubic, 136, 150

geometrical, 64

physical, 64

permafrost, 125, 136, 152

Piola-Kirchoﬀ stress tensor, 64, 68,

98, 126, 137, 153, 194

in presence of microstructure, 118

plane cross section hypothesis, 69

plate, 192

plateau, 51, 93, 111, 159, 182

point defects, 188

kinetics, 189

Poisson ratio, 71

pseudo-spectral methods, 15, 52, 53,

183

reference conﬁguration, 64, 65, 125,

165

reference distorsion, 115

rod

clamped end, 79, 83

free end, 79, 82

free lateral surface, 66

semi-inﬁnite, 79

with varying cross section, 87

Runge-Kutta method, 52, 184

secularity conditions, 48

shock, 26, 75, 78, 93, 164, 191, 195

sliding contact, 98, 105, 124

solitary waves interaction, 186

soliton, 3, 16, 53

spring elements, 124, 153

February 11, 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6

Chapter Title for Bibliography 213

surface tension, 95

thermacy, 195

thermal conduction, 196, 197

Fourier law, 194

thermoelasticity, 193

coupled equations, 196

Thomas method, 58

thresholds, 133, 152, 177, 185

truncated expansion, 36

uniformly valid solution, 49

vacancy, 188

variables

fast, 47

slow, 47

viscoelastic, 123, 124, 137, 150, 164,

166

Voigt, 166

Young modulus, 71

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Publishers’ page

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Publishers’ page

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Publishers’ page

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 To my parents .February 11.

February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 vi .

the theory has been developed to account for long longitudinal strain solitary waves propagating in a free lateral surface elastic rod with permanent cross section. Recently. presence of a dissipation (accumulation) destroys the balance between nonlinearity and dispersion. The nonlinearity. that despite of almost similar description of stresses in ﬂuids and solids. It was the most surprising fact. made by J. However. Hence it was proven that bulk long localized nonlinear strain waves of permanent form really exist. however. One of reason of the lack of the results on nonlinear wave in solids is that the complete description of a three-dimensional (3-D) nonlinear continuum is a diﬃcult problem. Certainly the cylindrical elastic rod seems to be a suitable real-life 1-D wave guide. solitons in ﬂuids were observed and generated many times. Motivated by analytical theoretical predictions. when in balance allow the propagation of the bulk strain solitary waves. bulk longitudinal strain solitary waves have not been observed in nonlinearly elastic wave guides.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Preface It is known that the balance between nonlinearity and dispersion may result in an appearance of localized long bell-shaped strain waves of permanent form (solitary waves or solitons) which may propagate and transfer energy over the long distance along.Scott Russell yet in 1834. and the dispersion resulting from the ﬁnite transverse size of the rod. caused by both the ﬁnite stress values and elastic material properties. there has been successful experimental generation of strain solitary waves in a polystyrene free lateral surface rod using the holographic interferometry. and nonlinear strain wave in the rod vii . Starting with the ﬁrst documented water surface solitary wave observation. That is why initial 3-D problems are usually reduced to the one-dimensional (1-D) form in order to clarify the simplest but qualitatively new analytical solutions.

they usually have no free parameters. dispersion and dissipation. Dissipative (active) eﬀects may be caused by internal features of the elastic material. growth of the amplitude) may cause the appearance of plasticity zones or microcracks in a wave guide. many works were done in a collaboration. Hence the wave is selected. The . Dissipation (accumulation) may also come in an elastic wave guide through phenomena occurring at its lateral surface. I would like to achieve two tasks in this monograph. hence.dissipative governing equations that are nonintegrable as a rule. However.e. It may attract the attention of the specialists in various ﬁelds since the structure of the governing equations is rather universal. One possibility occurs when the radius of the rod varies. Another task is to demonstrate the use of even particular analytical solutions for the description of unsteady nonlinear wave processes. First. an irreversible part should be included into the stress tensor in addition to the reversible one depending only upon the density of the Helmholtz energy. solutions may be obtained analytically. it is planned to provide the sequential analytical consideration of the strain waves ampliﬁcation/attenuation and selection in solids. methods of non-destructive testing. polymeric solids. Hence. these solutions require speciﬁc initial conditions. Inclusion of dissipation (accumulation) yields nonlinear dispersive. The content is essentially based on the author previous research. and ceramics. Moreover. usually travelling wave.February 11. only particular. The ampliﬁcation of the waves (i. point moving defects. thermal eﬀects. This wave has the form of a shock and is often called kink-shaped wave or simply kink. mainly in an elastic rod. When dissipation and activation act together there may be another balance resulted in a formation of a bell-shaped wave with the amplitude and the velocity prescribed by the condition of the dissipative/active balance. Among the volume sources of dissipation/accumulation one can mention a microstructure. This is of importance for an assessment of durability of elastic materials and structures. and additional relationships between the equation coeﬃcients are required for the existence of the solutions. particularly. Note that there exist another kind of nonlinear wave of a permanent shape sustained either by a balance between nonlinearity and dissipation or by a balance between nonlinearity. It may be of interest for those working in the ﬁeld of solid mechanics. determination of the physical properties of elastic materials. Certainly.. That is why the obtaining of the exact solutions is often considered as useless by many authors preferring to apply numerical methods only. Bulk waves provide better suited detection requirements than surface strain waves in setting up a valuable non-destructive test for pipelines. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 viii Book Title may attenuate or amplify.

L.V.M.V.G. The book preparation has been supported by the INTAS grant 99-0167 and by the RFBR under Grant 2000-01-00482. A.A. Velarde. December. Gursky and Mrs.V.F. Saint-Petersburg. G. Maugin. D. I dedicate this book to my parents. I. They always believe in my eﬀorts and expected this book more than whoever it may be. Dreiden. G. Kliakhandler. M. Semenova for a long time fruitful collaboration.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Preface ix author thanks a lot Profs. Pastrone. V. Samsonov.V. F. I. Parker. Mr. Porubov . 2002 A.

February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 x Book Title .

. . . 2. . . . . . . . . . . . . . . . . . . . .1 Nonlinear evolution equations . . . . . . . . . . . . . . . 1. . . . . . . . . . . .2 Kink-shaped and periodic waves formation . . 1. . . . . . 1. . . . . . . . 1. . . . . .3 Single travelling wave solutions . . . . . attenuation and selection of nonlinear waves Mathematical tools for the governing equations analysis 2. . . . . . .3 Kink-shaped waves . . .1. . . . . . . . . e 2. 1. . . . .4 Exact solutions of more complicated form . . 2. . .1. . . . . . . .3. . . . . . . . . . . . . .3 Ampliﬁcation. . . . . . . .2 Asymptotic solutions . . 2. . . . . .4 Periodic nonlinear waves . . . . . . . . . . . .1. . . . 1. . . . . . . . . .3 Numerical methods . . . . .1 Single nonlinear waves of permanent shape . . .1. .2 Oscillatory bell-shaped solitary waves . . . . . . . . . . . . . . . . . . 2. .2. . . . . . . . xi vii 1 2 2 6 9 12 13 15 26 27 31 31 31 35 36 42 47 51 52 57 59 2. . . . . . . . . .1. . . . . . . . 2. . . . . . . . . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Contents Preface 1. . . .2. . . . . . . . . . . .1. . 2. . . . .2 Formation of nonlinear waves of permanent shape from an arbitrary input . . . . . 1.2 Nonlinear hyperbolic equations . . . . Basic concepts 1. . . . 2.February 11. . .1. . . .1.1 Bell-shaped solitary wave formation from an initial localized pulse . . .2 Painlev´ analysis . . . . . .1 Monotonic bell-shaped solitary waves . . . .4 Use of Mathematica .1 Exact solutions . . . . . .3. . . . . . . 2.1 Direct methods and elliptic functions . 1. . .

3. . .2. . . . . . . .2.5 Reﬂection of solitary wave from the edge of the rod . .3 Double-dispersive equation and its solitary wave solution . . . . 4. . 3. . . . . 3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 xii Book Title 3. . .3. . .5 Inﬂuence of the external medium on the propagation of the strain solitary wave along the rod .1 Formulation of the problem . . . . . . 4. . . . . .1 Governing equation for longitudinal strain waves propagation . . 4. . . . . . . . . . . .1 Statement of the problem .1. . . 4. . . . . .2. . . . . . . . 3. . .1 Longitudinal strain solitary wave ampliﬁcation in a narrowing elastic rod . . . . . . .2 Modelling of nonlinear strain waves in a free lateral surface elastic rod . . . 4. . . .2. . . . . . . . . . . . 4. . Strain solitary waves in an elastic rod 3. .3 Strain solitary waves in an elastic rod with microstructure . . 3. 87 87 87 90 93 95 96 99 100 102 104 106 114 114 4. . . . . .4 Observation of longitudinal strain solitary waves . . . . .2 Evolution of asymmetric strain solitary wave . .3. . . . . . .2 Nonlinear waves in a rod with pseudo-continuum Cosserat microstructure . . . . 4. . .2 Derivation of the governing equation . . . . .3 Experimental observation of the solitary wave ampliﬁcation . . . . . . . . . . . . . . . . . . . . . .3.4 Nonlinear evolution equation for longitudinal strain waves along the rod and its solution . . . . Ampliﬁcation of strain waves in absence of external energy inﬂux 4. . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. .1 The sources of nonlinearities . . . . . . . . . . . . . . . 117 4. . . . . . . . . . . 63 64 66 66 69 70 75 79 4. 4. . . . . . . . . . . . . . . . . . . . . . . . . . .1 Modelling of non-dissipative elastic medium with microstructure . .3 Derivation of strain-displacement relationships inside the rod . . . . . . . . . . . . . .3 Nonlinear waves in a rod with Le Roux continuum microstructure . . . . . . . . . . . . .2. 4. . . . . . 4. .February 11. . . . . 120 . . . . .1. 4. . . . . . . . . . . . . . . . .2 External stresses on the rod lateral surface . .6 Numerical simulation of unsteady strain wave propagation . 4. . . .2.2 Strain solitary waves in an elastic rod embedded in another elastic external medium with sliding . . .1. .2. . . . .

. .4. . 180 6. . . 183 6. . . . . . .3 Numerical simulations . . .3. . .3. .1 Nonlinear bell-shaped and kink-shaped strain waves in microstructured solids . . . . . . . . . . . 5. . . .2 Evolution of bell-shaped solitary waves in presence of active/dissipative external medium . . . . . . . . . . . . . .1 Modelling of nonlinear seismic waves . . . .1 Contact problems: various approaches . 5. . . .4 Inﬂuence of external tangential stresses on strain solitary waves evolution in a nonlinear elastic rod .3 Exact solutions . . . . . . . .2.3 Kink-shaped solitary waves . . . . . . . . . . . . . . 5. . . 5.2 Combined dissipative double-dispersive equation . . .1 Modelling of a microstructured medium with dissipation/accumulation . . . . . . . . . . .2. . . . .1 Formulation of the problem . . . . . 178 6. . . .2 Bell-shaped solitary waves .3 Moving defects induced by external energy ﬂux . . . . 5. . .1. . .5 Concluding remarks . . . .2 Derivation of the governing equation .February 11. . 5. .3 Strain kinks in an elastic rod embedded in an active/dissipative medium . . . . . . . 5. . . .4. . 169 6. . . . 5.5 Weakly dispersive case . . . . . . . . . . . 5. . .3. .4 Weakly dissipative (active) case . . . . . . . . . .3. . . . . . . . . 5. . 188 . . . . . . . . . 5. . . . . . .2. . . . . . . . . . . . . . . . . . . 5. . . . . 121 123 123 125 125 126 128 130 134 136 137 138 141 143 145 150 152 152 153 155 159 Inﬂuence of dissipative (active) external medium 5. . . . . .2.6 Summary of results and outlook . . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Front Matter xiii 4. . . . . . 172 6. . . . . . . . . . . . 5. .2. . . . . . . .4 Concluding remarks . . . . . . . . . .1 Formulation of the problem . . . . . . . . .2. . . . . . . . . .4 Evolution of asymmetric solitary waves . . . . 5. . .2 Asymptotic solution of the governing equation . . . . . . . . . .3 Symmetric strain solitary waves . . . .4 Bell-shaped solitary wave ampliﬁcation and selection 5. . . 165 6.1 Formulation of the problem . 5. . .3. . . . . . . . . . . .2 Nonlinear seismic solitary waves selection . . . . .4. . . . . 164 6. . . 6. 176 6. . . . . .4. . . . . . . . . .1. . . . . . . . Bulk active or dissipative sources of the ampliﬁcation and selection163 6. . . .3. . . .2. . . . .1. . . . . . . . . .3.2 Dissipation modiﬁed double dispersive equation .2.3 Exact solitary wave solutions of DMDDE . . . . 5. . . . . . . . 5. . 178 6. . . . . . .1. . . . .4 Concluding remarks 5.

. . . . . 6. . . . .3 Nonlinear waves in a plate . . . . .2 Longitudinal waves in thermoelastic rod . .3.4. . . . . . . . . . . 6.4. . 188 190 192 193 195 196 199 211 . . Bibliography Index equations .February 11. . . . . . . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 xiv Book Title 6. . .2 Nonlinear waves in a medium . . . . . .3. . . . . . . . . . . .1 Nonlinear waves in thermoelastic medium 6. . . 6. . .1 Basic concepts and derivation of governing 6. . . .3.4 Thermoelastic waves . .

This unsteady process may be described analytically for the integrable equations or numerically for others. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter 1 Basic concepts This chapter is focused on some features of nonlinear waves to be used further in the book. they have inﬁnitesimal amplitude. like Ablowitz and Segur (1981). The governing equations for the nonlinear strain waves to be considered are nonitegrable by the inverse scattering transform method. The bell-shaped solitary wave usually appears as a result of a balance between nonlinearity and dispersion. another case corresponds to the simultaneous balance between dispersion. a velocity and a wave number of the nonlinear waves are connected to one another. More general information about nonlinear waves may be found in numerous special books. nonlinearity and dissipation (or accumulation). and only particular exact solutions may be obtained. Linear waves are accounted for the linear equations. We illustrate all mentioned above further in this Chapter. Whitham (1974) etc. bell-shaped and kink-shaped solitary waves. Calogero and Degasperis (1982). one possibility occurs when nonlinearity is balanced by dissipation (or accumulation) . Sachdev (1987). Newell (1985). There are two main types of the nonlinear travelling solitary waves which could propagate keeping its shape. In contrast to the linear waves. 1 . However. The kink-shaped wave may be sustained by diﬀerent balances.February 11. This is a result of the balances between various factors aﬀecting the wave behaviour. an amplitude. one can show that these solutions account for the ﬁnal quasistationary part of an arbitrary initial pulse evolution. The single travelling wave solution requires speciﬁc initial conditions. Bhatnagar (1979). Of special interest are the single travelling wave solutions that keep their shapes on propagation. Nonlinear waves are described by nonlinear equations.

de Vries (KdV) equation Korteweg and de Vries (1895).(1.1).2) travelling solitary wave one since it depends upon the phase variable θ = x − V t only. Typical shape of the wave is shown in Fig. ut + 2b u ux + d uxxx = 0. 1. A particular case arises for water waves when surface tension suppresses coeﬃcient d Hunter and Scheurle (1988) and ﬁfth-order dispersion u5x is added in Eq.1 1. u 0.February 11.6 0.(1. and monotonic solitary wave since it decays monotonically when |θ| → ∞. b (1.1.1).1 Single nonlinear waves of permanent shape Monotonic bell-shaped solitary waves The simplest celebrated model equation containing nonlinear and dispersive terms is the well-known Korteweg.order derivative (dispersion) or nonlinear terms into Eq.8 0.1) The wave amplitude A = 6dk 2 /b and the velocity V = 4dk 2 depend upon the wave number k which is a free parameter.1 Monotonic solitary wave (solid line) and its ﬁrst derivative (dashed line) Sometimes there is a need for the inclusion of higher.order derivative terms model weak nonlocality Engelbrecht . whose exact one-parameter single solitary wave solution is d u = 6 k 2 cosh−2 k(x − 4dk 2 t).2 x -30 -20 -10 10 20 30 Fig.4 0. Also higher. 1. One can call the solution (1.2) (1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 2 Ampliﬁcation of Nonlinear Strain Waves in Solids 1.1 where one can see also that the wave is symmetric with respect to its maximum.

An example of the inclusion of higher. In the general case the exact solitary.. Benilov et. the resulting equation models the LC ladder electrical transmittion lines. see also references therein. d and f results in changing only the wave propagation direction. Kawahara (1972). Kano and Nakayama (1981).3) is obviously nonintegrable by the Inverse Scattering Transform method.wave solution has a form similar to the KdV soliton (1. Its solutions were obtained in Nagashima and Kuwahara (1981). Karpman and Vanden-Broeck (1995). Grimshaw et. Hence the width of the wave is prescribed by the dispersion coeﬃcients d and f which should be of opposite sign. to say nothing of dissipative (active) generalizations.3) a ﬁfth-order (in space derivatives) KdV equation Hunter and Scheurle (1988) when c = r = s = 0. al (1994). Maugin and Muschik (1994). the following nonlinear equation may be considered: ut + 2b u ux + 3c u2 ux + r u uxxx + s ux uxx + d u3x + f u5x = 0. For the wave velocity we have V = 144dk 2 /13 = −36d2 /(169f ). Hence.2). A special integrable case corresponds to the SawadaKotera equation with b = d = 0. e. f = 1 Sawada and Kotera (1974). in addition d = 0. see. (1.(1. When. c = −r = −s/2 = 10. Hence. Important particular cases are: (1.3) We get from Eq. and only particular exact solutions may be obtained. Karpman (1998). Karpman (1993). This equation was studied in many papers. u = A cosh−2 k(x − V t). Let us consider an exact solution vanishing at inﬁnity. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 3 and Braun (1998). provide an improvement of bad dispersive properties Christov et. simultaneous triggering of the signs of b.5) . Its solitary wave solutions may be found in Parkes and Duﬀy (1996). Kawahara and Takaoka (1988). In case of the ﬁfthorder KdV equation it has the form Kano and Nakayama (1981): u= 210d 2 k cosh−4 k(x − V t).4) with k 2 = −d/(52f ).g. al (1993). Equation (1.February 11. account for a continuum limit of discrete models with far neighbour interactions Kosevich and Savotchenko (1999). 13b (1. al (1996).order nonlinearity is the SawadaKotera equation Sawada and Kotera (1974).

θ = x − V t): b u2 + d u − V1 u + s 2 u +fu 2 − V2 u = 0.3) the Gardner equation whose solution is Grimshaw et. V1 = 4dk 2 and V2 = 16f k 4 .order term r uuxxx is taken into account.(1.February 11.7) into Eq.k = .5) are A= 30f k 2 2(50b2 f 2 + 5bdf r − 3d2 r2 ) 2 5bf − dr . we get the solution with ﬁxed parameters. s (1. We also see that the wave velocity consists of two parts. Then the amplitude coincides with that of the KdV soliton (1. V = 4k 2 (d + 4f k 2 ). b −30f /c − 3d 30f −5d−116f k 2 ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 4 Ampliﬁcation of Nonlinear Strain Waves in Solids (i) In presence of only cubic nonlinear term.10) One can see that s may be excluded from the amplitude expression using the third formula from (1. cosh m(x− V t) + 1 = 2 cosh2 m(x − V t)/2.3) yields B1 = 1. Let us rewrite the ODE reduction of the equation (1. (iii) In case c = r = 0 the parameter k is free but an additional restriction on the equation coeﬃcients holds.5).6) with m = 2k. B1 = √ . r = s = 0. at nonzero f a substitution of Eq. (ii) When only quadratic higher. al (1999): A=2 − 30f 2 k . and we get the solution (1. However. A1 = √ 2 + 9cd m2 2 + 9cd m2 2b 2b (1.9) Note that the solution exists at d = 0.8) An important feature of the solution is the existence of the ﬁnite limiting amplitude when B is large Slyunyaev and Pelinovsky (1999).2). V = 4k 2 (2b c − u= where A1 . . (1.6) In this case the existence of solution vanishing at inﬁnity is provided by the linear ﬁfth order term f u5x .7) √ √ 2b 3 2d m2 . 2 r 25f r 10f r (1. Indeed at f = 0 we get from Eq. A= 60f k 2 . k 2 = .3) in the form ( = ∂/∂θ.V = . the ﬁxed parameters of the solution (1. V = dm2 . cosh m(x − V t) + B1 (1.(1.(1. c = s = 0. the ﬁrst of which being exactly the KdV soliton velocity. 10b f = d s.10). c 60f (1.

10) satisﬁes separately b u2 + d u − V1 u = 0.1) having a one-parameter solitary. The parameters of the solution (1. provided that the restriction 30cf − r(r + s) = 0 is satisﬁed.k = .5). c 3c(r + s) V = s(2r + s)[4k 4 (r + s)2 − b2 ] . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 5 One can check that the solution (1. b + 2k 2 (r + s) while k satisﬁes the equation 4[30cf − r(r + s)]k 4 + 2bsk 2 + b2 = 0. 3c(r + s)2 (1. s 2 u + f u − V2 u = 0.order nonlinear terms provide bounded localized solutions at d = 0 in contrast to the case c = r = s = d = 0 Kano and Nakayama (1981). it was recently found Garazo and Velarde (1991). V = 16f k 4 .February 11.5) are A= 120f k 4 . In particular.B = − . . b = 0. Higher. Thanks to the higher order terms the solitary. Then k may be a free parameter.V = .wave solution may exist even in the absence of the KdV’s nonlinear term. 2 (1. Instead the solution in the form of a solitary wave on an ”pedestal” may be obtained as u = A cosh−2 k(x − V t) + B. When f = 0 we have A= 3cd − 2b(2r + s) 2d[3cd − 2b(2r + s)] 2 3cd − 2b(2r + s) .11) where the ﬁrst of these equations is the ODE reduction of the KdV equation (1. (iv) Higher order nonlinear terms may support the existence of solitarywave solutions even in absence of the linear dispersion terms. c(r + s) (r + s)(2r + s) 2(r + s)(2r + s) There is no exact solution vanishing at inﬁnity in the case d = f = 0. with A= 2k 2 (2r + s) (2r + s)[b + 2k 2 (r + s)] .12) Even equations with dissipation may possess bell-shaped solitary wave solution. (1.wave solution.

2. al (1991).4).(1. (1.13) with α5 = 0) whose exact solution is Kudryashov (1988): u= 3 15α3 ∼2 128α1 α4 cosh−2 ( ∼ α3 8 α4 ∼ θ)(1 − tanh( α3 8 α4 ∼ θ)). The exact travelling bell-shaped solitary wave solution have been obtained in the form (1. al (1995) that appropriately heating a shallow horizontal liquid layer long free surface waves u(x.16) of the Kawahara equation is monotonic but asymmetric. Particular ∼ case corresponds to the Kawahara equation (Eq.13) may be rewritten as ∼ (α2 α4 α5 ∂ 2α1 α2 + ∼ ) uθ + ∼ uθθθ + ∼ (u2 )θ ∂θ α5 α2 2 α2 ∼ ∼ ∼ = 0. Indeed.14) The meaning of the last expression in (1. (1.de Vries (DMKdV) equation ut + 2α1 u ux + α2 uxx + α3 uxxx + α4 uxxxx + α5 (u ux )x = 0.11). see Fig.1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 6 Ampliﬁcation of Nonlinear Strain Waves in Solids Rednikov et. (1.16) 3 2 where V = 5α3 .14) is similar to that in case c = r = 0 for Eq.February 11. (1.(1. temperature gradient across the layer.(1. and its depth.2 Oscillatory bell-shaped solitary waves The solitary wave does not decay necessarily in a monotonic manner.). α3 = 2α1 α4 / α5 . Porubov (1993) with A = 12 α4 k 2 / α5 . ∼ ∼ ∼ (1. ∼ 1. V = −2α1 α2 / α5 .1.13) depend upon parameters characterizing the liquid (Prandtl number etc.(1. the ODE reduction Eq. t) can be excited whose evolution is governed by a dissipation-modiﬁed Korteweg.3).5). when the relationships for V and α3 hold.7) account for monotonic and symmetric solitary waves. All solutions (1. (1.13) The coeﬃcients in Eq. 1. In contrast to them exact solitary wave solution (1.12) Lou et. Despite diﬀerence in their functional form they have one and the same shape shown in Fig. B = −(α2 +4 α4 k 2 )/ α5 . ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ (1. 1.15) The restrictions on the equation coeﬃcients do not necessary provide an evidence of the KdV ODE reduction like Eqs.(1. α2 = α3 /(16 α4 ).15). Thus Kawahara (1972) studied decay at inﬁnity of the wave solution of the ﬁfth- .

05 -0. 1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 7 u a 0. order KdV equation using linearized equation analysis phase analysis.(1. which is proportional (in our notations) to d and is inverse proportional by the product of f and the wave velocity.3) does not possess exact oscillatory solitary wave solution. It was found that the wave decays monotonically or oscillatory depending upon the parameter ε.8 0. The oscillatory solitarywave solution is shown in Fig. How- . b)their ﬁrst derivatives. Eq.6 0.1 Fig. The proﬁle of the ﬁrst derivative with respect to the phase variable reveals its symmetric nature.05 -30 -20 -10 10 20 30 x -0.2 Symmetric (solid line) vs asymmetric monotonic solitary wave (dashed line).4 0.February 11. The same technique has been used in Karpman (2001) when nonlinear term in the ﬁfth-order KdV equation is of the form up ux . 1.3. a) proﬁles.2 -30 -20 -10 10 20 30 x u b 0.

order nonlinear terms as small perturbations assuming f = δF .1) for the function u0 whose travelling solitary wave solution is (1. 2b (u0 u1 )θ + d u1.θ = −F u0. is sought in the form u = u0 (θ) + δ u1 (θ) + . c = δC. First the asymptotic solution is obtained which consists of the KdV solitary wave solution and a small perturbation that oscillates but does not vanish at inﬁnity or a non-local solution Benilov et.18) .February 11. s = δS.19) (1. r = δR.3 Oscillatory solitary wave (solid line) and its ﬁrst derivative (dashed line) ever. it may be described asymptotically.θ − R u0 u0. The asymptotic solution u = u(θ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 8 Ampliﬁcation of Nonlinear Strain Waves in Solids u 1 0.25 2 4 -0.17) with ui → 0 at |θ| → ∞. In the next order an inhomogeneous linear equation results for u1 .θθ . Karpman (1993). Karpman (1998). 4b 2d 4d b b d b d (1. In the leading order we get the KdV equation (1.5 Fig.order derivative term and higher. δ << 1..θ .2). 1. Certainly the ﬁfth-order KdV equation is often considered as a perturbed KdV equation.75 0.θθθθθ − 3C u2 u0.25 x -4 -2 -0..5 0. Hunter and Scheurle (1988). θ = x − V t. (1.θ u0. Let us consider the ﬁfth.θθθ − 0 S u0..θθθ − 4dk 2 u1. al (1993). Its solution vanishing at inﬁnity is u1 = ( 3C 5bF 2R + S 2 S 4R 14F 9Cd 2 2k 2 F + 2− ) u0 +( + − − 2 )k u0 − θ u0.

23) .3 Kink-shaped waves The celebrated Burgers’ equation Burgers (1948) is the simplest equation that models the balance between nonlinearity and dissipation. usually they are asymmetric and may be found only numerically. (1. Finally.3. The equation coeﬃcients should be connected by d g = 37. B = .1. what corresponds to the Case IV in Kawahara (1972). ut + 2b u ux + g uxx = 0. it possesses the exact travelling solution of permanent form. 1. (1. In the case of the ﬁfth-order KdV equation. In particular. It may account for an oscillatory solitary wave solution ﬁrst obtained numerically in Kawahara (1972). however. m − f ree.21) In particular. this proﬁle exists at positive d and f .3).February 11. there exist nonlinear equations having exact travelling wave solutions in the form of an oscillatory solitary wave. an equation ut + a ux + 2b u ux + 3c u2 ux + d u3x + f u5x + g u7x = 0. (1. possesses the exact solution u= 35g 3 288 k cosh−1 k(x − V t) 24 cosh−2 k(x − V t) − c 17 . Oscillatory solitary waves of permanent shape arise also in dissipative problems. b 2b (1. k 2 = 17f /(581g).405 f 2 . see Fig. V = a + 102825k 6 g/289.(1.20) where b = 0. 1.22) If the boundary conditions are u → h1 at θ → ∞. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 9 The shape of the solution u = u0 (θ) + δ u1 (θ) depends upon the values of the coeﬃcients of Eq. where A= V g . u = A m tanh(m θ) + B. u → h2 at θ → −∞.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 10 Ampliﬁcation of Nonlinear Strain Waves in Solids then m= (h1 − h2 )b .24) whose exact solution was obtained independently by many authors VliegHultsman and Halford (1991) u = A tanh(m θ)sech2 (m θ) + 2A tanh(m θ) + C.25) It follows from the boundary conditions (1. 50V d 2b 10d (1. m= . is shown in Fig. C= . V = b(h1 + h2 ).22). with A= V g 6g 2 . 1.4 Burgers’ kink-shaped wave The shape of the solution (1. (1. Kinks may arise also due to the balance between nonlinearity. called kink. dispersion and dissipation like in the case of the Korteweg-de Vries-Burgers equation (KdVB). ut + 2b u ux + g uxx + duxxx = 0. . V = b(h+ + h− ).4.February 11. 1. 2g u 5 4 3 2 1 x -4 -2 -1 2 4 Fig.23) that h+ − h− = 2B.

while the kink (1. Thus the modiﬁed Korteweg-de Vries equation (MKdV).February 11. has an exact solution u= − 2d m tanh(m θ). (1. Dissipative equation may possess the same property. V = 2α1 C + 24α1 α4 while m is a free parameter if α3 = 12α1 α4 / α5 or m=− (α3 α5 −2α1 α4 )2 300 α4 α5 The shapes of the kink solutions (1. ∼2 ∼2 ∼ ∼ . In particular. 25bd Equations without dissipative terms may also possess the kink-shaped solutions.28) may have also diﬀerent proﬁle shown by solid and dashed lines in Fig.27) (1.13) has exact kink-shaped solution Lou et. 1. ut + 3c u2 ux + duxxx = 0. al (1991). c (1.25).5. Note that the MKdV equation has both the bell-shaped and the kink-shaped solutions. (1. 1. .4. Porubov (1993) u = C+ 36 α4 α5 ∼ ∼ ∼ √ m cosh−2 ( −3m θ)+ 12(α3 α5 −2α1 α4 ) √ 5 α5 where C= 12 α4 α5 ∼ ∼ ∼ ∼ √ −3m tanh( −3m θ).26) where V = −dm2 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 11 and the solution exists under h2 − h2 = + − 12g 2 . the DMKdV equation (1.27) are similar to that shown in Fig. ∼ ∼ ∼3 ∼ ∼ + (α3 α5 −2α1 α4 )2 25 α4 α5 ∼ ∼3 ∼ ∼ .28) m− α2 α5 α5 ∼ ∼ ∼ + ∼ 2α1 (α3 α5 −2α1 α4 ) 5 α5 m.

13). Thus Korteweg and de Vries (1895) found the periodic solution of the KdV equation (1.4 Periodic nonlinear waves Usually single bell-shaped solitary wave solutions are the particular cases of more general periodic solutions.29) where K and E are the complete elliptic integrals of the ﬁrst and the second kind respectively .1).6 1 . 1. They called Eq. Thus the MKdV periodic solution 1 Reprinted with permission from Elsevier Science . Newille (1951). d E u = 6 k 2 1 − κ2 + + κ2 cn2 (kθ. BKdV and DMKdV equations have not exact bounded periodic solutions that transform into the kinkshaped ones. 1.29) the cnoidal wave solution since it is expressed through the Jacobi elliptic function cn . Exact periodic and bell-shaped solitary wave solutions correspond in the same manner in case of the generalized 5th-order KdV equation (1. It tends to the single solitary wave solution (1.3) and the DMKdV equation (1. like Burgers’. there exist exceptions.February 11. κ is the modulus of the Jacobian elliptic function Bateman and Erdelyi (1953-54).(1.5 Fig. Byrd and Friedman (1954). Although many equations.2) at κ → 1 as shown in the right column in Fig. κ) b K (1.5 Kink-shaped waves with a ”hat” 1.1.5 -1 -1. Cnoidal wave is not a linear superposition of the bell-shaped solitary waves.5 x -4 -2 2 4 -0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 12 Ampliﬁcation of Nonlinear Strain Waves in Solids u 1 0.

3. 1.32) has a functional form diﬀerent from both the KdV cnoidal wave and the MKdV bounded periodic solution. c (1. κ) b u= √ − . Analytical solutions of unsteady problems may be obtained if governing nonlinear equations are integrable Ablowitz and Segur (1981). and the solution (1. 2 3κ 4 κ4 − κ2 + 1 √ 1√ −c. In practice more important is to know how an arbitrary ﬁnite amplitude input evolves.31) tends to the kink-shaped solution (1. Often initial input transforms into the stable quasistationary wave structures of . C1 + cn2 (m θ. m2 = √ .2 Formation of nonlinear waves of permanent shape from an arbitrary input All solutions presented in previous section require speciﬁc initial conditions. Calogero and Degasperis (1982). (1.30) transforms into the kink solution (1.31) This evolution equation represents an analog of the hyperbolic equation to be derived further in Sec. otherwise only numerical solutions are available. Its bounded periodic solution Porubov (1996) is m cn(m θ.6 in comparison with the transformation of the KdV cnoidal wave solution to the bell-shaped solitary wave. 2 √ (1. al (1982). b = 3g −c. Newell (1985). κ) sn(m θ. Bhatnagar (1979). κ).32) and the following restrictions on the coeﬃcients: f =− The periodic wave solution (1.27) at κ → 1. Dodd et. 1. Another example is a dissipative nonlinear equation ut + 2bu ux + 3c u2 ux + duxxx + f (u2 )xx + guxx = 0. κ) 3c −c with C1 = 1 − 2κ2 + 3g 2 − V κ4 − κ2 + 1 .27) as it is shown in the left column in Fig.February 11. When κ = 1 we have C1 = 0. 5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 13 Ablowitz and Segur (1981) u= − 2d m sn(m θ. κ) dn(m θ.

After Porubov and Velarde (2002).995.32) (left column) and the KdV cnoidal wave (1.4 0. (d)κ2 = 1.99995.2 -10 -5 -0.25.4 0. (b) κ2 = 0.6 Comparison of the periodic solution (1.4 5 10 x u 0. 1. .February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 14 Ampliﬁcation of Nonlinear Strain Waves in Solids a u 0. Also the analysis gives the conditions when the formation of them is possible. In this section we illustrate it using some instructive examples.2 -0.29)(right column) for diﬀerent values of the Jacobi modulus: (a)κ2 = 0.2 -0. (c)κ2 = 0. permanent form which may be described by the analytical solutions.2 -10 -5 -0.4 b u 2 1 -10 -5 -1 -2 c u 2 1 -10 -5 -1 -2 d u 2 1 -10 -5 -1 -2 5 10 x -10 -5 -1 u 3 2 1 5 10 x 5 10 x -10 -5 -1 u 3 2 1 5 10 x 5 10 x -10 -5 -1 u 3 2 1 5 10 x 5 10 x Fig.

(1.7 is the formation of the train of solitary waves from a Gaussian initial pulse in the KdV case. e. localized pulse evolution into an oscillatory solitary wave was considered in Nagashima and Kuwahara (1981) for the equation ut + u ux − γ 2 u5x = 0. The evolution of the initial monotonic solitary wave into radiating or oscillatory solitary waves was simulated in Benilov et. al (1997). Nagashima and Kuwahara (1981). f = 0. rectangular. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 15 1. Inﬂuence of the ﬁfth-order dispersive term.5. −50 respectively.01.wave solution (1. The oscillatory travelling solitary. Shown in Fig. d = 0.. b = 1. Below only those results are shown that were obtained using both numerical methods.10 demonstrate the decrease of the solitary waves for f = −1. Karpman and Vanden-Broeck (1995). al (2001) the solitary wave formation from a periodic input was studied for an equation similar to Eq. Both the amplitude and the velocity decrease with the increase of the absolute value of f in qualitative agreement with the exact solution (1.4). Salupere et. In a series of papers Salupere et. Gaussian distribution etc.3. We have tried various shapes of the initial localized pulses.1 Bell-shaped solitary wave formation from an initial localized pulse The ﬁfth-order KdV was extensively studied numerically. −10. 1. ﬁnite-diﬀerence and pseudo-spectral. Following Porubov et. Figs.February 11.4) and the analysis of the dispersion relation Karpman (1993). However. 2. Since the role of the ﬁfth-order derivative term is of interest the coeﬃcients b and d in Eq.8-1. Previously.2.(1. f = 0. For both coeﬃcients positive the initial rectangular proﬁle is dispersed without formation of any localized waves. Below we consider the formation of solitary waves in the systems governed by Eq. d and f . are of opposite sign.3).3) were ﬁxed for all computations. We shall study the evolution of a localized initial pulse. The dependence upon the sign of the ratio d/f is in agreement with the exact solitary. al (1993).2. al (2002) we use two methods for computations.(1. see Sec.3).g. more smooth Gaussian initial proﬁles provide the appearance of solitary waves even for positive coeﬃcients when f is rather small. We found that the rectangular initial pulse splits into a sequence of solitary waves when the coeﬃcients of dispersive terms.wave solutions were found in Boyd (1991).The ratio between the amplitude and the velocity . First the 5th-order KdV equation was studied. 1. The next result we have obtained is the dependence of the number of solitary waves upon the value of f when d/f < 0.

Fig.4) is 1.4 0.11(B).order KdV exact solution (1. respectively.33 at f = −50. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 16 Ampliﬁcation of Nonlinear Strain Waves in Solids u 0. Only at small f = −0.1 ÷ −0. 1. from 1. A similar tendency is revealed by the phase analysis of single travelling wave solutions. 1.7 is equal to 1. the amplitude and the velocity for given b and d are −105/(1352f ). 1. The ratio for the 5th.7 Evolution of the initial Gaussian proﬁle in the KdV case.7-1.2 0 40 80 x 120 160 200 Fig. This ratio (and the amplitude) decreases with the decrease of f .2) to the oscillatory solitary waves. In case f = −1.43 at f = −1 to 1.February 11.11. For convenience the last stages from Figs. the higher leading solitary wave . −9/(169f ). f = 0.4).5 just as for the KdV soliton (1.15 are the numerical results for the leading solitary wave in quantitative agreement with the exact solution (1. 2 in Kawahara (1972). More important is that the decrease of f aﬀects the solitary.2) . 1. Fig. of each solitary wave in Fig. cf. 1.10 are collected in Fig.wave transformation from monotonic KdV solitons (1.46.

c = 0. Finally.8 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case. d.2 0 60 120 x 180 240 300 Fig. is oscillatory while other solitary waves remain monotonic. They simply evolve to the opposite direction according to the analysis of the exact solution (1. its velocity) found in Kawahara (1972). Also the solution is sensitive to the ratio between nonlinear terms contributions.February 11. we have found that simultaneous triggering of the signs of b. Inﬂuence of the cubic nonlinearity. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 17 u 0.4). 1. while r = s = 0.4 0. Now we add cubic nonlinearity. The analytical solutions predict an action of the cubic nonlinear term depends upon the sign of c. and the . 1. f = −1. Fig. then the transformation occurs also for the second solitary wave.11(C) . f doesn’t aﬀect the shapes of the solitary waves. Alternate transformation of the solitary waves conﬁrms the dependence of the kind of solitary wave upon the value of the product of f and the wave amplitude (hence. b/c.

(1.9 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case. (1. c = 1. at smaller b. f = −10.4 0.2). 1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 18 Ampliﬁcation of Nonlinear Strain Waves in Solids u 0.5). Indeed we have found that at b = 1. the stage of forthcoming evolution of already gen- . However. b = 0.2 0 80 160 x 240 320 400 Fig.5. f = −1 the train of solitary waves arises only from positive initial pulse with amplitude equal to 0. A similar tendency is observed when the initial amplitude becomes close to 1 or higher when predominant cubic nonlinearity excludes an inﬂuence of the quadratic one on the sign of the wave amplitude like in the exact solution (1. At the same time.6). Dependence of the amplitude on the sign of b is very typical for the exact solutions of the equation with quadratic nonlinearity.5 while a negative one is dispersed. the formation of solitary waves no longer depends upon the sign of the initial pulse amplitude.4). see (1.2 or b = −0. value of the amplitude of an initial pulse. d = 0.2.February 11.

6). The number of solitary waves generated from the initial localized pulse increases with the increase of the value of c and ﬁxed values of d = 0. there is f /c < 0 resulted from the nonlinear exact solution (1. f = −50. Moreover.8). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 19 u 0.February 11. while other analytical restrictions on the coeﬃcients become more important.5. Kawahara (1972). the width (proportional to 1/k) decreases.2. The velocity of the waves increases also.5).4 0. at small f one can anticipate an evidence of the condition d/c > 0 given by Eq. f < 0 all above mentioned inequalities are satisﬁed. erated solitary waves is not so sensitive to the value of b.2 0 100 200 x 300 400 500 Fig. while the amplitude remains practically one and the same.(1. When c > 0. (1. We also observed the alternate transformation of the solitary .10 Evolution of the initial Gaussian proﬁle in the 5th-order KdV case. f = −1 and also b = 1 or b = 0. 1. Besides the condition d/f < 0 following from the linear analysis Karpman (1993).

The third formula in (1.04 0. waves from monotonic to oscillatory when c increases for both values of b. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 20 Ampliﬁcation of Nonlinear Strain Waves in Solids 0. Assume b = 1.February 11.1 u 0.02 0. Independence of the amplitude of c doesn’t follow from the exact solution (1. we have tried simultaneous variations of c and f in order to sustain one .2 giving negative values for b = 1. 1. (C) f = −10.15 0.6) predicts growth of positive values of k 2 only for b = 0.05 0 70 90 110 0 x 130 150 170 100 130 160 x 190 220 250 0. the decrease of the negative f values aﬀects the decrease in the number of solitary waves.5.08 0.1 0 140 180 220 0 260 300 340 160 210 260 310 360 410 x x Fig. As found in previous subsection. (B) f = −1.1 u 0.2 A 0.15 0.11 Transformation of the kind of solitary waves in the 5th-order KdV case.06 u 0. d = 0.6) as well as from the asymptotic solution.2 C 0.2 0. However. (D) f = −50.4 D 0.1 B 0. Then one can exhibit for both values of b the similarity of the variation of the velocity with respect to c with that obtained in numerics. (A) f = 0.3 u 0. let us express k through the amplitude A and substitute it into the expression for the velocity V .05 0.

(A) c = 0. Suppose c = r = 0 . (C) c = 75. (B) c = 15.15 u 0.1 0. and solitary wave formation is observed only for small f .05 0. f /c < 0. 1.05 0 40 60 80 100 x 120 140 160 0.12(B. This conﬁrms that the amplitude depends upon the ratio f /c but velocity is proportional to f .12. We see that the wave amplitude keeps its value from Fig. When both c and f are positive no solitary waves appear.15 0.1 u 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 21 0. 1.2 C 0. and the same number. f = −10. f = −50. f = 0. Inﬂuence of nonlinearities s ux uxx and r uuxxx . is satisﬁed.12(A) to Fig.February 11.05 0 100 150 200 0 x 250 300 350 100 150 200 250 300 350 400 450 x Fig. At c < 0. see Fig.1 0.2 B 0. f > 0 only one inequality.12(A) to oscillatory Fig.15 u 0.12 Conservation of the number of solitary wave thanks to the simultaneous presence of cubic nonlinearity and ﬁfth-order derivative term.2 A 0. d > 0. 1. 1. C). otherwise the initial pulse is dispersed in this case. 1. The kind of solitary wave alters from monotonic Fig. 1. On the contrary small absolute values of c provides the solitary waves formation at c < 0. f < 0. d > 0.12(C) while the velocity growths.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 22 Ampliﬁcation of Nonlinear Strain Waves in Solids r=-1.3 0. . The wave amplitude decreases with the increase of s while the velocity keeps its value.4 r=-1.60 0.6 0.4 u 0.10).4 0. f = −1 . Certainly. and vary s at ﬁxed b. solitary waves exist outside the restriction from (1. that we choose b = 1.50 0.6 0. Wave behavior is not sensitive to the sign of s.5 0. d = 0. It is found that the amount of solitary waves and its transition from monotonic to oscillatory don’t depend upon the value of s.5.5 0.1 0 80 100 120 140 160 x Fig.57 u 0. The fact the velocity doesn’t depend upon s is in agreement with the exact solution. 1.5 0.2 0.13 Equalization of the ﬁrst and the second solitary waves and subsequent exceeding of the second wave due to the alteration of the negative values of the coeﬃcient r.6 0.10). The condition for the solitary wave formation d/f < 0 remains valid.1 0 80 100 120 140 160 u 0. d and f .3 0.2 0.2 0.1 0 80 100 120 140 160 x x r=-1. A comparison of the velocities with those obtained numerically demonstrates the more agreement the less is the value of b. We also used numerical values of the amplitude to deﬁne k and then V using (1.February 11.3 0.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Basic concepts

23

Asymptotic solution (1.17), (1.19) also predicts the decrease of amplitude at permanent velocity. Indeed, we get that umax = u0 (0)+δu1 (0) = 6dk 2 /b(1+ k 2 [f /d − s/(2b)]). At coeﬃcient values we used the value of umax decreases with the increase of s (s is not large in the asymptotic solution), while the exact solution predicts the same behavior only for positive values of s. When s = c = 0 the behavior of the solution diﬀers from the previous one. Having the same values for b, d, f we obtain that increase of positive values of r yields a decrease in the velocity and an increase in the amplitude of the solitary waves. The number of solitary waves decreases. However, at negative values of r we found that at the initial stage of the splitting of the Gaussian proﬁle the amplitude of the second solitary wave becomes equal to that of the ﬁrst one at r = −1.57, see Fig. 1.13. At lesser r second solitary wave becomes higher, and two solitary waves form a two-humps localized structure shown in Fig. 1.14. It is no longer quasistationary since amplitudes of the humps vary in time. It looks like an interaction of two solitary waves when the second higher solitary wave surpasses the ﬁrst one, then it becomes lower, and the process repeats. Decreasing r we achieve formation of a three-humps localized structure shown in Fig. 1.15. Its evolution is similar to those presented in Fig. 1.14. Finally, only multi-humps localized structure arises from an initial pulse as shown in Fig. 1.16. The localized multi-humps structures in Figs. 1.14-1.16 keep their width, while their shapes vary in time. Certainly, unsteady multi-humps localized structures are not governed by the ODE reduction of Eq.(1.3) and, hence cannot be explained either by the phase portraits analysis or by the exact travelling wave solution (1.5), (1.9). Moreover, at negative values of f the exact solution doesn’t predict propagation to the right of the solitary wave with positive amplitude. Absence of linear dispersive terms. We have found an exact solitary wave solution that may be supported by higher -order nonlinear terms even without linear dispersive terms, at d = 0 or f = 0. Numerical simulations show that there are no solitary waves at both zero d and f . Some solutions from previous subsections keep their features at d = 0, in particular, this relates to the case r = 0. At the same time cubic nonlinearity at d = r = s = 0 supports two-humps localized waves for c > 0. At negative c the wave picture is similar to those at d = 0. No stable solitary waves propagate in absence of only the ﬁfth-derivative term, d = 0, f = 0 with the exception of the Gardner equation case where Slyunyaev and Pelinovsky (1999) found generation of the limiting amplitude solitons.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

24

Ampliﬁcation of Nonlinear Strain Waves in Solids

u

0.6 0.4 0.2 0 60 120 180 240 300

x

Fig. 1.14

Two-humps solitary wave formation at r = −1.6.

To sum up, both higher order nonlinear and dispersive terms aﬀect the formation of localized nonlinear waves their shape and their parameters. Thus, the number of solitary waves and the transition from monotonic to oscillatory wave are under responsibility of both 5th- order linear dispersive term, cubic nonlinearity and higher- order quadratic nonlinearity r u uxxx . More important is the formation of an unsteady but localized multi-humps wave structure thanks to r u uxxx and cubic nonlinearity at d = 0. The sign of the coeﬃcient b of the KdV quadratic nonlinear term is important for choosing the sign of the input amplitude. At the same time the nonlinearity s ux uxx doesn’t aﬀect the formation and behavior of solitary waves. Certainly, the shapes of the resulting solitary waves are not obviously governed by the exact and asymptotic travelling wave solutions. Some

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Basic concepts

25

u

0.6 0.4 0.2 0 60 120

x

180

240

300

Fig. 1.15

Three-humps solitary wave formation at r = −3.2.

other features of numerical solutions, like the dependence of the number of solitary waves upon the values of the equation coeﬃcients or a transition from monotonic wave to an oscillatory one, are not predicted by analytical solutions. However, the combinations of equation coeﬃcients required for the existence of solitary wave are realized in numerics. Also numerical wave amplitude and velocity relate like in the analysis. Evidence of all these predictions even qualitatively is very important for a justiﬁcation of the numerical results. Formation of the bell-shaped solitary waves in presence of dissipation or an energy inﬂux will be considered further in the book.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

26

Ampliﬁcation of Nonlinear Strain Waves in Solids

u

0.6 0.4 0.2 0 60 120

x

180

240

300

Fig. 1.16

No solitary waves other than multi-humps one at r = −12.

1.2.2

Kink-shaped and periodic waves formation

The formation of the kink-shaped waves was studied considering the evolution of the Taylor shock from discontinuous (step) initial conditions under the governance of the Burgers equation Sachdev (1987); Whitham (1974). It was found the appearance in due time the steady state kink solution (1.22). A quasihyperbolic analog of the Burgers equation was studied in Alexeyev (1999) where it was found that kink may be formed from suitable initial conditions. The Korteweg-de Vries-Burgers equation (1.24) also extensively studied but mainly in the direction of generation of the triangle proﬁles and oscillating wave packets Berezin (1987), see also references therein. Of special interest is the formation of the kinks with a

hence the width of the wave inverse proportional to k(T ). It turns out that an asymptotic solution may be found in this case whose leading order part is deﬁned as a solitary wave with slowly varying parameters. or slow coordinate. X = εx. decreases. Usually periodic waves are generated in ﬁnite domains from a harmonic input. 1. we have an attenuation of the solitary wave when its . Salupere et. also Kawahara (1983) obtained numerically periodic wave structures in a system governed by Eq.1 that the bell-shaped wave may exist even in presence of dissipation but under strong restrictions on the equation coeﬃcients. Thus the KdV cnoidal waves (1.2. What happens with the wave when dissipation/accumulation destroys this balance? It was shown in Sec.13) ∼ with α5 = 0. Assume the inﬂuence of dissipation/ accumulation is weak and is characterized by a small parameter ε << 1. 2. Now only general features of the wave behaviour are considered. al (1995). In the former case the solitary wave solution is u(θ. Depending on the problem either slow time. al (1997). al (1994). 5. may be used. Note that harmonic input in the ﬁnite domains is used also for the study of the bell-shaped solitary waves interactions where no periodic wave structure of permanent shape arises Salupere et. In the latter case we have u(θ. θt = −V (T ).3. On the contrary. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 27 ”hat” shown in Fig.33) where θx = P (X).34) (1. One possibility will be considered in Sec.(1. T = εt.February 11. attenuation and selection of nonlinear waves As already noted the bell-shaped solitary wave is sustained by a balance between nonlinearity and dispersion. This is an ampliﬁcation of the solitary wave. (1. θt = −1. while at nonzero coeﬃcient similar results were found in Rednikov et. T ) = A(T ) cosh−2 (k(T )θ). where θx = 1. When A(T ) increases in time k(T ) usually increases also. Salupere et. 1.29) are realized numerically and in experiments in a paper by Bridges (1986).3 Ampliﬁcation. Next order solutions give us the functional form of the dependence of the wave parameters upon the slow variable. Derivation of the asymptotic solution will be described in Sec. al (2001). 1.5. X) = A(X) cosh−2 (k(X)θ).

6 0. To put . 1.5 3 2. hence. amplitude decreases while its width increases.6 1.5 -10 10 20 10 a 20 30 40 50 x b 30 40 50 x Fig.5 1 0.February 11.17 Temporal evolution of an initial solitary wave resulting in a selection: a) from below.2 -10 u 4 3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 28 Ampliﬁcation of Nonlinear Strain Waves in Solids u 1.4 0. b) from above.2 1 0.5 2 1.4 1.8 0. Sometimes it happens that the increase/decrease of A(T ) takes place not up to inﬁnity/zero but to the ﬁnite value A∗ . by the physical parameters of the problem. Usually this value is deﬁned by the governing equation coeﬃcients.

Shown in Fig. One can see that the wave (1.18 is the proﬁle of the wave (1. Selection provided by an ampliﬁcation of an initial wave. 1. . 1. will be called selection from below. Cnoidal wave evolution may be accounted for an asymptotic solution similar to that of the bell-shaped solitary waves Rednikov et.8 0. see Fig. al (1995). only initial stages of the temporal evolution of (1. see Fig.34) in comparison with the symmetric solitary wave solution (1. the ﬁnal stage of the selection both from below and above.2 x -20 -10 10 20 Fig. 1.33). As follows from Fig.34) diﬀers from that of Eq. is the symmetric bell-shaped solitary wave like shown in the last stages in Fig. However. this another way.17(b). 5.17(a).33) (dashed line) vs solitary wave (1.17. the parameters of the resulted steady wave are selected. 1. it will be shown in Sec.19. while selection from above happens as a result of an attenuation of an initial wave.6 0.34) ( solid line).February 11.33) at t = 0.34) is asymmetric with respect to its core (or maximum). Svendsen and Buhr-Hansen (1978). 1. 1.18 Solitary wave (1.4 0.3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Basic concepts 29 u 0.(1. The ampliﬁcation/attenuation of the kink may be described asymptotically and numerically Sachdev (1987).

5 2 1. b) from .5 -20 -10 10 b 20 30 40 50 x Fig.19 above.5 1 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 30 Ampliﬁcation of Nonlinear Strain Waves in Solids u 2 a 1.February 11.5 1 0.5 -20 -10 10 20 30 40 50 x u 3.5 3 2. 1. Selection of the asymmetric monotonic solitary wave: a) from below.

The choice of the analytical and numerical procedures is based on an experience of the author and does not claim a completeness. When we are interested in a self-similar solution of a partial diﬀerential equation one can use well developed theory of the solutions of ordinary diﬀerential equations. periodic. Not only the numerical realization and graphical representation of the solution is provided by this method but also multi phase quasiperiodic solutions as well as purely periodic ones may be represented using the algebrogeometrical approach as illustrated in Belokolos et.1 2. The signiﬁcant point in direct methods is to build in advance the appropriate functional form of the solution (ansatz) of the equation studied.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter 2 Mathematical tools for the governing equations analysis As a rule governing equations for nonlinear strain waves are nonitegrable by the inverse scattering transform method. Hence the study of real physical processes requires a combined analytico-numerical approach. Among the recently developed general methods the algebrogeometrical approach may be used in an eﬃcient way to ﬁnd such solutions. Ince (1964).. e. Exact solutions of nonlinear nonintegrable partial diﬀerential equations are obtained usually using various direct methods. see. For example. The aim of this chapter is to describe methods to be used in this book.g. the usage of ansatz in the form 31 . and only particular analytical solutions may be obtained. al (1994).1. 2. particularly.1 Exact solutions Direct methods and elliptic functions Most of the mathematical work in the realm of nonlinear phenomena refers to integrable equations and their exact solutions.

According to its deﬁnition Whittaker and Watson (1927). However most of dissipative equations cannot be transformed to the bilinear form.g. Korpel and Banerjee (1984).Chow (1995). k). For this purpose various elliptic functions were proposed recently.g. In order to see it let us ﬁrst give some properties of the Weierstrass elliptic function ℘ to be used below. e. g2 . Samsonov (2001). with appropriate forms for the ansatz suggested by information about the poles of the solution. One would like to apply the same procedure to ﬁnd more general periodic solutions. sn(ζ. When compared to the use of theta functions or Jacobi elliptic functions. Another advantage is that two apparently distinct solutions are readily recognized as equivalent. Parkes and Duﬀy (1996) and references therein). Jacobian elliptic functions (see. a prime advantage of using the function ℘ is that the algebra is drastically simpliﬁed. the most popular were theta functions (see.. and solution parameters are obtained from the algebraic equations appearing after equating to zero coeﬃcients at each order of tanh . Samsonov (1995). At the same time single travelling wave solution derivation looks very complicated even for non-dissipative equations Chow (1995).Kostov and Uzunov (1992). First of all another appropriate ”basic” function is required instead of the tanh . the Weierstrass function is analytical in the complex plane other than in the points where . Moreover we have to deal with four theta functions that results in additional diﬃculties for the ansatz construction.. The eﬃcient choice is caused by the simplest procedure of the ansatz construction and the least complicated algebra for determining solution parameters. It is well known that theta functions may be included in the Hirota bilinear method in order to get N-periodical solutions Nakamura (1979).February 11. Porubov (1993). k) and dn(ζ. e. Nakamura (1979)). g3 ).. Porubov and Parker (2002). The ansatz for the solution involves only one Weierstrass function ℘(ζ. periodical solutions could be obtained in terms of any of these functions.g. Explicit periodic travelling wave solutions may be found for many nonintegrable equations and systems by using an ansatz in terms of ℘. In principle. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 32 Ampliﬁcation of Nonlinear Strain Waves in Solids of a hyperbolic tangent (tanh) power series resulted in ﬁnding of new exact travelling wave solutions (see. Then the equation studied becomes polynomial of the tanh after substituting the ansatz. e. instead of four theta functions or three Jacobi elliptic functions cn(ζ. k). Parker and Tsoy (1999)) and the Weierstrass elliptic function Kascheev (1990). Porubov (1996). The choice of tanh is caused by the fact that any derivatives of tanh may be expressed as a polynomial with respect to the tanh itself. Nakamura (1979).

3 e1 = 2 − k2 2 κ . sn and dn which are bounded on the real axis.4) connecting the Weierstrass function with the Jacobi function cn. g3 ) = e2 − (e2 − e3 )cn2 ( e1 − e3 ζ. the familiar link is obtained in Whittaker and Watson (1927) but using the singular function sn−2 .2) where A and B are rational functions with respect to ℘.February 11. (2. 2. g2 . (2.3) However. The bounded periodic solutions are more conveniently expressed by writing them in terms of the Jacobi elliptic functions cn. 3 e2 = 2k 2 − 1 2 κ . e3 ≤ e2 ≤ e1 ) are the real roots of the cubic equation 4τ 3 − g2 τ − g3 = 0. regular along the real axes. we have e3 = − 1 + k2 2 κ . (2. g2 . √ ℘(ζ. 3 . The governing equation for the function ℘ is: {℘ (ζ)}2 = 4℘3 − g2 ℘ − g3 (2. Indeed. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 33 it has double poles. k). √ the wave number κ = e1 − e3 and the Jacobian elliptic modulus k. Depending on the ratio between g2 and g3 the Weierstrass function may be bounded or unbounded inside the domain under study. (2. following the method introduced in Whittaker and Watson (1927) one can check that the following formula is valid: √ ℘(ζ. and that any elliptic function f may be expressed using ℘ it and its ﬁrst derivative as Whittaker and Watson (1927) f = A(℘) + B(℘)℘ . k). For this purpose the relation between the Weierstrass function and the Jacobian functions is used as a special case of (2. Remarkable features of the function ℘ are that all of its derivatives can be written by means of itself. Here k = (e2 − e3 )/(e1 − e3 ) is the modulus of the Jacobian elliptic function. g3 ) = e3 + (e1 − e3 )sn−2 ( e1 − e3 ζ. while τ = em ( m = 1.1) where g2 and g3 are constant parameters.2). 3 .5) Expressing these results in terms of an appropriate choice of parameters.

3).7) and (2.8) are essentially equivalent solutions. m) − r r . 12 216 (2. m) ± r cn(ˆζ. Porubov (1993). . Therefore we really deal with two functions.7) z2 = r2 m ˆ r2 (1 + m2 ) ˆ r2 m2 2 ˆ sn (ˆζ. Porubov (1996). In Parker and Tsoy (1999). and it is unlikely to get the solution using the ansatz proposed in the form of power series with respect both of the ℘ and ℘ as it was done for the tanh. Accordingly. the solution (2.6) The localized both the bells-shaped and the kink-shaped solitary wave solutions appear in the limit k → 1 of the Jacobi elliptic functions. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 34 Ampliﬁcation of Nonlinear Strain Waves in Solids g2 = 8 4 4 6 2 κ (1 − k 2 + k 4 ). m) are appropriately r related to (r.type series about the singular manifold. However. Samsonov (1995). In order to check the poles of a solution we shall use the WTK method Weiss et. Another idea may be used. e It will be explained in details in Sec.1.1) when the parameters r and m are deﬁned as solutions to ˆ r4 ˆ r6 ˆ (1 + 14m2 + m4 ) = g2 . al (1983) looking for the solution in the form of Laur´nt.9) Therefore z2 is also a solution satisfying the same governing equation (2. m)dn(ˆζ.7) is a representation of the Weierstrass function (2. (1 − 33m2 − 33m4 + m6 ) = g3 . Samsonov (2001). k). provided that (ˆ. 3 27 (2. Obviously. working with ℘ reveals links between seemingly distinct forms of solution. The two expressions (2.1) deﬁning the Weierstrass function. based on the singular point analysis of the possible solution and the well known fact that the Weierstrass function ℘ has the second order poleWhittaker and Watson (1927). g3 = κ (k + 1)(2 − k 2 )(1 − 2k 2 ). It is to be noted that the ﬁrst Weierstrass function derivative ℘ cannot be expressed as a polynomial of the Weierstrass function itself Whittaker and Watson (1927). and we have to equate zero separately coeﬃcients at each order of ℘ and at products of ℘ and corresponding orders of ℘ Kascheev (1990). solutions were sought in terms of powers and products of Jacobi functions and thereby two solutions were obtained √ z1 = e2 − (e2 − e3 ) cn2 ( e1 − e3 ζ.8) which appear diﬀerent. 2 2 12 (2. k). (2. 2.February 11.8) also satisﬁes equation (2. Now consider an instructive example Porubov and Parker (2002).2. one can check by direct substitution that the solution (2.

Then the theory has been extended to partial diﬀerential equations. Weiss et. t) = 0 is the ”singular” or ”pole” manifold on which a solution u(x. Recursion relations for x the qj are (j + 1)(j − 4)(j − 6)qj = F (ϕt . 4 and 6 are called resonances..10) into Eq. qk . al (1993). t) = 1 ϕα ∞ qj (x. When all these conditions are satisﬁed the equation under study has the Laur´nt property. Usually these equations are called ”equations with the Painlev´ property”. Here we concentrate on the one aspect of the theory-the singular manifold method or WTC method for partial diﬀerential equations Newell et. Substituting the ansatz (2. Conte et. Also it is necessary to assume that neither ϕx e nor ϕt vanish on ϕ(x. al (1987). Weiss et. ϕx . (ii) ϕ is analytic in x and t and (iii) the equations for the coeﬃcients qj have self-consistent solutions. 2. d = 1.1) one can ﬁnd α = 2 and q0 = −2ϕ2 . Let ϕ(x. Following Newell et.(1. t)ϕj j=0 (2.. k < j). Porubov and Parker (1999). Conte (1989). Newell et. and additional relationships on the equation coeﬃcients are required for the existence of the solutions Porubov (1993). This requires (i) α is an integer. Moreover. al (1983) we assume b = 3.10) is single valued. Levi and Winternitz (1992). Weiss et. At each such . al (1987). usually belong to the class of travelling wave solutions which require special initial conditions.1. . al (1987). j = −1. In case of the solitary wave solution the initial condition should be have the shape of the solitary wave itself. The e achievements of the theory may be found in Cariello and Tabor (1989). Porubov (1996). The main idea of the WTC is to demonstrate that the expansion u(x.February 11. Porubov and Velarde (1999). other than poles. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 35 The exact solutions obtained in this manner. travelling wave solutions for the dissipative equations usually have not free parameters. Samsonov (2001). (2. al (1983).. Porubov and Parker (2002). Samsonov (1995).11) The values of j.2 Painlev´ analysis e Recently it was developed the theory of nonlinear ordinary diﬀerential equations whose solutions have not movable singularities. t) is singular. t) = 0.1). al (1983). Let us illustrate how the methods works on an example of the KdV equation (1.

∼ ∼ ∼ (2.(1.11) vanishes thus ensuring the indeterminancy of the corresponding qj . in terms of the Weierstrass elliptic function ℘ and ℘ .13) to the third-order ODE of the form: α4 u + α3 u + α2 u + α5 u u + α1 u2 − V u + P = 0. θ = x − V t. one can reduce Eq. Some examples are presented below. al (1983). 2. the phase velocity and the Weierstrass function parameters g2 .11) Newell et.(2.(2. ℘+C (2. Certainly this procedure is of phenomenological kind but it allows to obtain the solutions of nonitegrable nonlinear equations in an explicit form.1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 36 Ampliﬁcation of Nonlinear Strain Waves in Solids resonance the right-hand side of Eq.(1. As a result we obtain using Eq. Based on Eq. However. a u(x.13) obtained in Porubov (1993). Substituting the proposed form of the solution into the equation under study and equating to zero coeﬃcients at each order of ℘ and at products of ℘ and corresponding orders of ℘ one can get the algebraic equations on the solution parameters. t) = 2 ∂ log ϕ + q2 .12) possible solution may contain simple and secondorder poles that may be modelled in terms of ℘ as Porubov (1993): u = A℘ + B℘ + D. t) u= ∼ 12 α4 α5 ∼ ∼ (log ϕ)xx + 12 5 α5 ∼ (α3 − 2α1 α4 α5 ∼ ∼ )(log ϕ)x + u.11) an autoB¨cklund transformation for the solution of Eq. Weiss et. g3 . ∼ (2. Also the Lax pair for the KdV equation follows from the solution of Eq.February 11.(1.13) where P = const. Concerning only tavelling wave solutions. the expansion (2.12) where u (x.(2. ∂x2 where q2 satisﬁes the KdV equation.(1. the Lax pair cannot be obtained for nonitegrable equations as opposed to a truncated expansion that carries an information about possible pole orders of a solution.(2.14) . al (1987). in particular.10) may be truncated at O(ϕ0 ). Using this information the anzats for the solution may be proposed. prime denotes diﬀerentiation with respect to θ.1). Moreover.13). t) satisﬁes Eq. It was found there the following auto-B¨cklund transformation for a the solution u(x.3 Single travelling wave solutions First we consider exact solutions of DMKdV Eq.

A(12 α4 + α5 A) = 0.(2. The solutions of these equations are: ∼ ∼ (i) when g2 . D. phase velocity V and Weierstrass function parameters g2 . α1 A2 + 6α3 A + 12 α4 B + 6 α5 AB = 0. . B(2α1 (D − AC) − V ) = 0. (α3 − ∼ + ∼ (α3 − ∼2 α5 C− α5 ∼ 2α1 α4 α5 ∼ ∼ )+ 2α1 α4 α5 ∼ ∼ 25 α4 α5 − ∼ ∼ ∼ )2 . C is a free parameter A=− D=− V = 12 α4 α5 α2 α5 ∼ ∼ ∼ ∼ ∼ . A=− (ii) when either C=− ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼ 12 α4 α5 ∼ ∼ .14) into Eq. V =− 2α1 α2 α5 ∼ ∼ . D = − α2 α5 ∼ ∼ . 24α1 α4 α5 2α1 25 α4 α5 ∼ ∼ 2α1 α2 ∼ + 2 4α1 α5 (α3 ∼2 2α1 α4 α5 )+ (α3 − 2α1 α4 α5 ∼ )2 . B=− 2α1 6 5 α5 ∼ (α3 − ∼ 2α1 α4 α5 1 ∼ ∼ ). C. g3 = 8C 3 . g3 are free parameters and α3 = 2α1 α4 / α5 . B.February 11. 2α1 (2B 2 + AD) − V A + 2 α2 B + 2 α5 (BD − ABC) = 0. or α3 = 12α1 α4 / α5 . 1 300 α4 (α3 ∼2 − 2α1 α4 α5 ∼ ∼ )2 . (12C 2 − g2 )B = 0.13) one can derive a system of algebraic equations in A. B = 0. g3 : (g2 C − g3 − 4C 3 )B = 0. 2α1 AB + 2α3 B+ α2 A+ α5 (AD + 2 α5 B 2 ) = 0. g2 = 12C 2 . P = V D + 8α1 B 2 C − α1 D2 − α3 g2 A/2 − 2 α2 BC + 12 α4 BC 2 − α5 (2ABC 2 + g2 AB/2 − 2BCD). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 37 Substituting Eq.(2.

15) where k is a free parameter.February 11.14) with parameters deﬁned by (ii) accounts for a bounded kink-shaped solution (1. the solution (2. this equation may be transformed into the O.31) was considered. i. κ) − α2 α5 ∼ ∼ − 4 α4 α5 ∼ ∼ k 2 (2κ2 − 1) (2. One can see that there exist at least one more solution of the form: u=− where y = exp(γθ).28). The solution (2.28).. solutions depending only on the phase variable θ = x − V t. Besides bounded solutions (2. When the Jacobian elliptic functions modulus κ → 1 the solution (2. propagating with ﬁxed phase velocity.14) with parameters deﬁned by (i) may describe a particular bounded cnoidal wave. kink propagates with any phase velocity value. g3 ). G and g3 are free parameters.12). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 38 Ampliﬁcation of Nonlinear Strain Waves in Solids Using (2.e.17) . The transformation is obtained here using the WTK method of the form u = f± f 2 − 2cd (log ϕ) + u.14) also describes unbounded ones in the form of localized and periodic discontinuities. which results in the following equation after integrating once with respect to θ: dη + gη − V η + b(η 2 ) + f (η 2 ) + c(η 3 ) = N. γ = − 1 5 α4 ∼ 12 25 ∼ ∼ α4 α5 (α3 − 2α1 α4 α5 ∼ ∼ )2 exp(2y)℘(exp(y) + G.D.4) the solution (2.28).32) of the equation (1. (2. When C is a free parameter.16) Further we follow the results obtained in Porubov (1996).15) and (1. In previous Chapter the bounded periodic solution (1.E. of the form: u= 12 α4 α5 ∼ ∼ k 2 κ2 cn2 (kθ. Finally it is to be noted that the functional form (2. At the same time it accounts for a new periodically discontinious solution. c (2. (α3 − 2α1 α4 α5 ∼ ∼ ). When studying travelling wave solutions. 0. This solution allows to describe only the bounded kink-shaped wave (1.15) transforms to the solitary wave solution (1.14) in terms of the Weierstrass function is not unique.

The ﬁrst of them is u = A℘ + B.(2. For the wave amplitude γ two formulaes are valid γ = A1 m.18) In order to ﬁnd the solution parameters the formula (2. and the bounded kink-shaped solitary wave solution follows from Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 39 Therefore possible solution should contain ﬁrst order pole.18) will have the form of localized discontinuity under positive C values. B. A1 = (f + (f − 2cd)1/2 )/(6c). phase velocity V and the Weierstrass function parameters g2 . g3 : (℘ + C)−4 : (g2 C − g3 − 4C 3 )2 (6Af − 3A2 c − 6d) = 0.(2. (2. When the parameter C is negative we get κ = 1. The ﬁrst appears when 12C 2 − g2 = 0. (℘ + C)−2 : 4(g2 C − g3 − 4C 3 )(2bB − V + 48ACf + 3c(12A2 C − B 2 ) − 48Cd) − 3(12C 2 − g2 )2 (2f − 2Ac − d) = 0.18): u = γ tanh(mθ) + u0 . C. (℘ + C) : 2Bb − V + 24ACf + 6(B 2 − 12A2 C)c − 12Cd = 0.16). and no periodical solution exists. In this case two of three roots ei of Eq. The Weierstrass elliptic function ℘ possesses second order pole. Then the solution (2. Then equating to zero coeﬃcients at each order of ℘+C and ℘ one can derive the algebraic equations on A.19) .5) are equal to one another. (℘ + C)−1 : 4(g2 C − g3 − 4C 3 )(10f − 4f + 3Ac + 6d) − (12C 2 − g2 )(2Bb − V + 24ACf + 3(B 2 − 12A2 C)c − 12Ad) = 0. ℘+C (2. ℘ : 2Ab + g + 2Bf + 6ABc = 0. ℘ (℘ + C)−3 : (g2 C − g3 − 4C 3 )(g + 2Bf − Ab − 3ABc) = 0. and we shall propose three solution forms. g2 C − g3 − 4C 3 = 0.February 11. ℘ (℘ + C)−2 : (12C 2 − g2 )(2Ab − g − 2Bf + 6ABc) = 0.18) is substituted into the Eq. (℘ + C)2 : 2Af + d + 2A2 c = 0. Three solutions of algebraic equations are obtained. (℘ + C)−3 : (g2 C − g3 − 4C 3 )(12C 2 − g2 )(14Af − 9A2 c − 12d) = 0. ℘0 : 4(12C 2 − g2 )qA2 = N.(2.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 40 Ampliﬁcation of Nonlinear Strain Waves in Solids γ = A2 m. V2 are V1 = 2u01 b + 6c(u2 + 4A2 m2 ) − 4m2 (2f A1 − d). The bounded cnoidal wave solution arises when C = −e1 and has the form u = ± − 2d 2 cn(mθ.(2. Then for u0 we have u0 = 2bAj − g . 1. see Fig. In this case the bounded cnoidal wave solution (1. that transforms into the kink-shaped soluiton (2.16).6. (2.22) Substitution Eq.21) where m2 = e1 − e3 . κ)sn(mθ. and it transforms into the kink-shaped solution (2.February 11. 2. g2 C − g3 − 4C 3 = 0. 2(3cAj + f ) and phase velocities V1 .19) when κ = 1.D.E.(2. Now we shall consider the second possible solution’s form in terms of the Weierstrass elliptic function ℘: η= A℘ + B.16) allows us to conclude that solution may now exist only when f = g = 0. κ) 3c (2. j = 1.20) In this case the solution may exist under additional conditions on the equation coeﬃcients f = g = 0.(2. (2. A2 = (f − (f − 2cd)1/2 )/(6c). propagating with the phase velocity V = −b2 /(3c) − 6e1 d. reduction of the Gardner equation. Hence it becomes now the O. 12C 2 − g2 = 0. Finally. 2 02 while m2 = −3C = 3e1 is a free parameter. c dn(mθ.22) into the Eq. It governs the travelling cnoidal wave.32) holds. that results in absence of the dissipative terms in Eq.19). When κ = 1 we have C ∗ = 0. 01 1 V2 = 2u02 b + 6c(u2 + 4A2 m2 ) − 4m2 (2f A2 − d). Then one can get the algebraic . κ) b κ m − . the third solution arises when 12C 2 − g2 = 0. The second solution corresponds to the situation when g2 C − g3 − 4C 3 = 0.

c (2.22) if B = 2de3 /c and has the form u = 2d m κ sn(mθ. κ).D. (2.23) where m2 = e1 − e3 and for the phase velocity we have V = dm2 (2κ2 − 1).24) Solution (2. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 41 equations for the solution parameters equating to zero coeﬃcients at the terms ℘ ℘k . reduction of the Gardner equation. however now one can obtain its another bounded solutions. Thus for d/f c > 0 we ﬁnd cnoidal wave solutions of the form: (I) for B = 2de2 /c u = 2d m κ cn(mθ. It is to be noted that at κ = 1 both the solutions (2. c (2. i = 1 ÷ 3.25) . κ). When d/c > 0 the bounded solution appears from (2.February 11.E. V = − cB. c (2. and (A℘ + B)5/2 : ℘ ℘3 : cA + 2d = 0 ℘ ℘2 : 9cAB − V A + +15dB = 0 ℘ ℘ : 9cAB − 2V A + +12dB = 0 ℘ : 12cB 3 − 4V B 2 + dA(Bg2 − Ag3 ) = 0 (A℘ + B)5/2 : N = 0 One can obtain the following solution of these equations: A = − 2d 2dei 1 .24) transform to the bell-shaped solitary wave u= 2d m ch−1 (mθ. B = . (II) for B = 2de1 /c we obtain u = 2d m κ dn(mθ. c propagating with the velocity V = dm2 . κ).23) and (2. κ).5). We again deal with the O.24) represents cnoidal wave propagating with the phase velocity V = dm2 (2 − κ2 ). k = 0 ÷ 3. c c 2 where ei are the real roots of Eq.

(2. hence only travelling wave solutions may be found. η = σ = ±1).February 11.(2. (2. However.23). ρ. Recently the coupled nonlinear Schr¨dinger equations (CNLS) o iWt + sWxx + (ηW W ∗ + σU U ∗ )W = 0.28) into equations (2.25) transforms to the kink-shaped solution (2.27) with r. ζ = ζ(θ.19).24) the solution (2. Changing variable in auto-B¨cklund transformation (2.19) when κ = 1.g. have attracted considerable interest because of their role in governing various physical wave-guiding systems.27) allows to separate the real and imaginary . σ. t).26) Substituting it into Eq.27) in the form W = w(ζ) eiζ . These equations are not generally integrable by the inverse scattering transform method. and more complicated proﬁles are decsribed. iρUt + rUxx + (η −1 U U ∗ + σW W ∗ )U = 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 42 Ampliﬁcation of Nonlinear Strain Waves in Solids For the phase velocity we ﬁnd V = − dm2 (1 + κ2 ). The solution of the form (2. w. 2.26) for the last equation was already obtained previously in Samsonov (1995). In contrast to solutions (2.1. The machinery of the Weierstrass function is applied in the search for travelling wave solutions of equations (2.16) one can ﬁnd that solution may exist under f = 0.4 Exact solutions of more complicated form The procedure of the obtaining exact solutions is based on the reduction to the ODE. Substitution (2. s = ±1. sometimes this procedure is applied to obtain only the part of a solution. ℘+C (2. except in cases of high symmetry (e. Consider some examples. ζ and φ real. Therefore we now deal with the O. t).E. ρ = 1. φ = φ(θ. η real. U = u(ζ) eiφ . with u. reduction of the Gardner equation with linear dissipation. (2.D.17) a one can see that the following solution may be proposed: u = A y℘y + B. Finally one can construct the third possible solution in terms of the Weierstrass elliptic function ℘ depending now on new independent variable y = exp(kθ).28) where θ = x − c t. It was found that this solution may correspond either to discontinuous periodical solution or to the bounded kink-shaped solution of the form similar to the solution (2.

An example of the behavior of Re W is shown in Fig. A 1 − σ2 where B. we ﬁnd for W of the form: W = w exp ı Y.34) 2(s σ − r η) σ(E A − B D) − 3s B . We get an interesting wave.32) (2. ϕ = arcsin(sn(κθ)).D = . c ∼ κ and at t = 0.February 11. (2. n. Therefore.1(a) for the case H1 ∈ [e2 . k) + B/A . The solutions for u and w are sought in the form w= F ℘2 + A℘ + B. w2 dθ + φ0 (t) . consisting of the carrier wave slightly modulated and with superposed periodic disturbances. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 43 parts.31) while for the phase function Y we have Y = csθ sC1 Π[ϕ. The shape of the wave (2. when Ci = 0. C1 = 2) η(1 − σ 4 γ= 4 B A 3 − g2 B + g3 A . 3 (2. (2. This introduces signiﬁcant features into the proﬁles of the real and imaginary parts of W and of U . The most notable diﬀerence from previous solutions is the dependence of the “frequency” and “phase” on w and u. Thus. k] c2 s − + (γ + )t. e3 ].2.33) Π[ϕ. n. g2 and g3 are free parameters. E. u2 (2. 2 κ A(H1 − e3 ) 4 (2. n = −κ2 k 2 /(B/A + e3 ) and C1 is deﬁned by A= 2(σ r η − s) A2 2 . where w is deﬁned by Porubov and Parker (1999) w= A 2k 2 − 1 2 κ − k 2 κ2 cn2 (κθ. k and H1 . .30) where Ci (i = 1. k] is the elliptic integral of the third kind.29) while for ζ and φ we have ζ = 1 s c θ − s C1 2 1 φ = 2 r c ρ θ − r C2 dθ + ζ0 (t) . 2) are free parameters.31) depends strongly on the values of the parameters κ. u= G℘2 + D℘ + E.

1 Exact solutions of the CNLS. 2.1(b). b) almost envelope wave solution. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 44 Ampliﬁcation of Nonlinear Strain Waves in Solids Re W 2 a 1 20 40 60 80 100 x -1 -2 Re W 2 b 1 20 40 60 80 100 x -1 -2 Fig.February 11. When c considerably exceeds the wave number κ we get the proﬁle closer to the usual envelope wave solution. the wave shape is not determined solely by the amplitude wave shape. a) proﬁle with moving disturbances. Another interesting proﬁles correspond to the exact solutions of the . 2. see Fig.

g. 2.36) where ζ = x − c t. we decompose the solution u(x. 2. θ. both real. Substituting (2. with pj . Fig. q) ∈ C. p. This equation appears in the description of a large variety of physical phenomena. Like in case of CNLS equations. κ) − δ1 . θ = θ(ζ. . the situation changes dramatically when C = 0. whose position vary in time. y Z (2. Then periodic and pulse solutions of the CGLE may be found Porubov and Velarde (1999). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 45 complex Ginzburg-Landau equation.38) When C = 0. ﬁrst of all.. (u. u = y(ζ) eıθ . A typical situation is shown in Fig.February 11. γ ∈ R. lasers. non-equilibrium pattern formation. In this case Z never vanishes. 2. Z = pr 2 y 2 yx + C 2 .2. and the zeroes of the ﬁrst derivative are deﬁned by the zeroes of the sin function only. CGLE. and for u we have harmonic temporal oscillations of the spatially periodic state deﬁned by the amplitude parts of the solutions (2. qj = 0. During half of the time period the shapes of these parts vary. t). and we get in Fig.37) (2. is v= with y= 2qr 2 2 2 k1 dn (k1 ζ.36) into (2. respectively. t) in its amplitude.37) are deﬁned by the zeroes of the function yx and correspond to the zeroes of the Jacobi functions cn and sn. Their positions do not change in time. and phase.36).2(a). q = qr +ı qi . However.35) and equating to zero the real and imaginary parts one obtains two coupled equations for the functions y and z ≡ θζ . (2. y. 2 ı ut + p uxx + q |u| u = ı γ u (2.2(d) a proﬁle which is practically the mirror image to Fig. Z y yx sin θ + arcsin( ) . Qualitatively this evolution does not depend upon the value of the modulus κ of Jacobi functions. e.35) where the constant coeﬃcients are p = pr +ı pi . v = (Re u)x . 2. The ﬁrst derivative of the real part of the periodic solution with respect to x. the zeroes of the ﬁrst derivative (2. in nonlinear optics. The subscripts t and x denote temporal and spatial derivatives. superconductivity etc.2(a) shows a structure with four spatially repeated parts.

5 1 0.5 1 0.5 1 0.2 Evolution of the periodic solution (2. c)m = 2.5 20 -0.5 -2 Re u 2 1. d)m = 3.5 -1 -1. δ1 = 0.5 20 -0.5 -1 -1.5 -2 Re u 2 1.5 -2 Re u 2 1.5 -1 -1. a)m = 0.5 -2 40 60 40 60 40 60 40 60 a 80 100 120 x b 80 100 120 x c 80 100 120 x d 80 100 120 x Fig.5 20 -0.5 20 -0.36) Re u vs x for times t = m pi π/(3pr γ). b)m = 1.9.5 1 0. 0 < m < 3 with k = 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 46 Ampliﬁcation of Nonlinear Strain Waves in Solids Re u 2 1.5 -1 -1. .February 11. 2.5.

Therefore. such as phase variable θ in the unperturbed problem. l2 (2. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 47 Consider now the pulse solution Porubov and Velarde (1999). Often straight asymptotic expansions are incorrect Cole (1968). (2.39) is 6Al1 (1 + 4A2 ) 2 1 k cosh−1 (kζ) tanh(kζ) sin θ − arctan( ) . The basic idea of the perturbation approach is to look for a solution of a perturbed nonlinear equation in terms of certain natural fast and slow variables. there is a pulse solution whose spatial behavior is determined by the function cosh−1 (kζ) only with one extremum at ζ = 0. 2.3(a) disappear. Often there is a need in only one fast variable. or slow coordinate.41) The evolution of the real part of the solution (2. 2. our solution is breather. l2 = pr qi − pi qr . r i 2(pi + 4A pr − 4A2 pi ) 6Al1 k cosh−1 (kζ) exp ıθ.40) it follows that additional zeroes of the ﬁrst derivative may appear if (Re u)x = k> l2 exp 6A l1 arctan(1/(2A)) 2A . Jeﬀrey and Kawahara (1982).like. while two minima arise. and the matching asymptotic procedure Ablowitz and Segur (1981). Sometimes additional terms may be considered as small perturbations.41) is not satisﬁed. l1 = p2 + p2 . One can see that many equations consist in generalisations of the integrable equations like the KdV equation.39) is illustrated in Fig. Kodama and Ablowitz (1981) may be applied to ﬁnd a solution in this case. from (2. If (2. T = ε t. l2 2A (2. 2.40) Thus. Fig. 2. Again we see that two initial maxima in Fig. Depending upon the problem either slow time. Fig.39) The ﬁrst derivative for the real part of (2. then an initial minimum at ζ = 0 is changed into a maximum.February 11. ε << 1 are . u= with κ2 = − γ . 2.3. Nayfeh (1973).2 Asymptotic solutions Particular exact solutions are insuﬃcient for understanding physical processes.3(f-h). X = ε x.3(e).

θt = −1.04 x c Re u 4 2 0. g)m = 7. h)m = 8. Fast variable is generalized in a perturbed problem assuming either θx = 1.04 x -0. 2.04 x e Re u 4 2 0. The functions V (T ) or P (X) are deﬁned to remove secular terms.04 -2 -4 d 0.04 -2 -4 Re u 4 2 -0. Some of the appropriate secularity conditions are formed from Green’s identity as follows. Assume a solution u = u(θ.February 11.04 -2 -4 Re u 4 2 -0.04 x Fig.04 -2 -4 Re u 4 2 -0.04 -2 -4 a Re u 4 2 0. T ) is of the form . c)m = 2. θt = −V (T ) or θx = P (X).04 x -0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 48 Ampliﬁcation of Nonlinear Strain Waves in Solids Re u 4 2 -0.04 -2 -4 b 0. d)m = 4.04 -2 -4 h 0.04 x -0. b)m = 1.39) Re u vs x for times t = mπ/β.3 Evolution of the pulse solution (2. introduced. 0 < m < 8.04 x g Re u 4 2 0.04 -2 -4 f 0. f)m = 6. e)m = 5.04 x -0. a)m = 0.

. LA (vi ) = 0.43) is carrying over the period Rednikov et. In this case integration in (2.43).. i = 1. T ) + ε u1 (θ. al (2000). |θ| ∼ 1/ε Ablowitz and Segur (1981). we obtain that L(u1 )vi − LA (vi )u1 = F vi is always a divergence.44) . Following Kliakhandler et.43) Then either V (T ) or P (X) are obtained from Eq. Method may be used for ﬁnding perturbed cnoidal wave solutions. and suppose that u depends upon a fast variable θ and a slow time T . Here L(u1 ) is a linearized leading-order equation operator..February 11.(2.. Using the boundary conditions it may be integrated to give the secularity conditions. (2. T ) + .. In particular. ε << 1. Kodama and Ablowitz (1981)..g. Complete solutions is obtained using matching quasistationary solution to a nonstationary one for large |θ|. M ) the M solutions of the homogeneous adjoint problem. al (1995) assume that αi = ε α i . Let us apply this method for a perturbed solitary wave solution of the DMKdV equation (1. Then in the next order an inhomogeneous linear equation for u1 holds L(u1 ) = F (u0 ). when vi → 0 as |θ| → ∞ while u1 is bounded the secularity condition is: ∞ vi F dθ = 0. Then equation (1.. Denoting by vi (i = 1. (2.. such as θx = 1. M. where LA is the adjoint operator to L. al (1995). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 49 u(θ. It is to be noted that a simple quasistationary expansion is not uniformly valid on |x| → ∞.. θt = −V (T ). T ) = u0 (θ. T = ε t.13) becomes α3 uθθθ − V uθ + 2α1 u uθ + ε [uT + α2 uθθ + α4 uθθθθ + α5 (u uθ )θ ] = 0.13).42) The leading order problem is nonlinear equation for u0 whose solution is known. Svendsen and Buhr-Hansen (1978). e. . Rednikov ∼ et. −∞ (2.

Then the leading order solution is u0 = 6 α3 b(T )2 cosh−2 (b(T ) θ) α1 (2.(2.45).47) The behavior of b (or the sign of bT ) depends on the signs of A and B and on the value of b0 ≡ b(T = 0). α1 (2.θ = 0. Here b tends to −A/B independent of b0 .θθθ − V u1.T + α2 u0. The most interesting case occurs when A > 0.50) We see that b tends to inﬁnity in ﬁnite time.47) is adjoint to that of the Eq. Equation (2.θθθθ + α5 (u0 u0. F = − u0. For A < 0.(2.(2.49) 8 3 b A + B b2 .48) may be directly integrated giving the implicit dependence of b on T : T = b2 (A + B b2 ) 4B 4(b2 − b2 ) 0 ln 0 .44) is sought in the form (2. B = 4 7 6α3 α5 − 5α4 .February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 50 Ampliﬁcation of Nonlinear Strain Waves in Solids The solution of Eq. Indeed. T∗ = 4B b2 B 4 0 ln + . 2 15A (A + B b2 ) 15A b2 0 0 (2. (2.42).θ + 2α1 u0 u0. B > 0.θ )θ .θ + 2α1 (u 0 u1 )θ = F.43) with vi = u0 yields the equation for b: bT = with A = α2 .45) We are interested in studying localized solutions vanishing together with its derivatives at |θ| → ∞.48) (2. B < 0. In the leading order we get α3 u0.θθ + α4 u0. 15 (2. T = T ∗ for A > 0. Then the secularity condition (2. At order ε. B > 0 the parameter b vanishes if b0 < −A/B while it diverges if b0 > −A/B. The operator acting on the function u1 in the lhs of Eq. we have α3 u1.51) . − 2 2 (A + B b2 ) 15A b 15A b2 b2 0 0 (2.θθθ − V u0. when both A and B are positive b diverges while for both negative values it will vanish. A quantitative description of the variation of b can be given.46) with V = 4α3 b2 .

corresponding expressions for the derivatives of the elliptic functions with respect to modulus may be found in Byrd and Friedman (1954). the asymptotic ”dissipative” solitary wave (2. and expression (2. al (1995).46) will tend to the exact travelling solitary wave solution (1.g. In this case the modulus of the Jacobian elliptic functions should be a function of a slow variable. Higher order approximations may be studied similarly. b0 < −A/B. the quantity b approaches −A/B when T tends to inﬁnity. On the other hand. B > 0. al (1995) of the form: u1 = A1 u0. in Rednikov et. Perturbations of cnoidal wave solutions were studied. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 51 A similar scenario occurs when A < 0.θ + A2 θ cosh−2 (b(T ) θ) + A3 u0. al (1995). with our asymptotic approximation we can predict when blow-up could occur. Its deﬁnition allows to satisfy solvability condition in the next order problem and to avoid secular terms in the asymptotic expansion. 1. if we additionally assume α3 = 2α1 α4 /α5 .February 11. Therefore numerical simulations should be used. see Fig. e.. Finally.12). Two kinds of nonlinear equations. Svendsen and BuhrHansen (1978). hence. Uniformly valid solution is found using the matching asymptotic procedure may be found in Rednikov et..E. In particular they satisfy speciﬁc initial conditions and do not allow to account for an arbitrary initial disturbance evolution. . One can see it predicts a plateau behind the solitary wave and does not vanish at minus inﬁnity. Newille (1951). the quantity b0 > −A/B.50) provides an analytical description of the time-dependent process of the parameter-value selection of the solitary wave (2.17. Function u1 (θ. al (1995). 2. B > 0. B < 0 and A < 0. B < 0.3 Numerical methods One can see that both exact nor asymptotic solutions have severe limitations. for both A < 0. where Ai may be found in Rednikov et. The ﬁrst-order solution u1 has been obtained in Rednikov et. when A > 0. it will contain free parameters depending on T like b(T ) in the leading order problem. T ) is a solution of a linear inhomogeneous O. In the last case.D. Hence.θ log(cosh (b(T ) θ))+ A4 (1 − tanh(b(T ) θ)). b vanishes when T tends to inﬁnity.46).

al (1997). pseudo-spectral approach and ﬁnite-diﬀerence approach. 2. may be written either ut = L(u) or utt = L(u) (2. Sachdev (1987). Equations of the kind (2. see also references therein. Dodd et. it takes place for a simulation of the generalized ﬁfth-order KdV equation (1. It becomes the smaller the higher is the order of a highest derivative term in NEE. this method was eﬀectively applied to study the evolution and interaction of Marangoni-B´nard solitary waves governed by Eq. As noted in Berezin (1987) implicit three-levels schemes are rather simple and suitable for a realization. . The implicit predictor-corrector method has been applied to nonlinear diﬀusion equations in Sachdev (1987). see Berezin (1987). Further only numerical methods are described which are used by the author and his coworkers in solving the problems considered in the book.52) where L(u) is a certain nonlinear diﬀerential operator. al (2002). Besides simulation of Eq. Fletcher (1984).(1. Two main schemes are used. the fourth-order RungeKutta ﬁnite-diﬀerence method. while long-time evolution should be studied. some of the straight ﬁnite-diﬀerence methods require rather small time step for stability. in particular. al (1982).53). The problem of an increase of the time step may be solved either by a modiﬁcation of a diﬀerence scheme or by using a more eﬀective solving procedure. Dodd et. However.53) (2. al (1982) where various diﬀerence schemes are discussed.3) in Porubov et. In particular. Mayer (1995). Zwillenger (1989).52) will be called nonlinear evolution equations while nonlinear hyperbolic equations correspond to the class (2.3).February 11.(1. The same equation has been numerically solved in Christov and Velarde (1995) using a four-stage scheme providing thirdorder approximation in time.1 Nonlinear evolution equations Among the equations of the kind (2. More detailed information about numerical modelling of nonlinear wave equations may be found.52) main attention has been paid to numerical solutions of the KdV equation. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 52 Ampliﬁcation of Nonlinear Strain Waves in Solids considered here. in Berezin (1987). in particular. another example has been studied in Christov et.13) e in Marchant (1996).3.

in particular.13). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 53 Pseudo-spectral methods have been adopted recently. The corresponding wavelength is λc = 2π/kc . we deﬁne the typical wavenumber kc = α2 /2α4 corresponding to the most unstable linear mode in Eq. Finally. The use of the pseudo-spectral approach to account for an evolution of an initial localized pulse in framework of Eq. 1.3). and the selection have been observed at the αi values prescribed by the theory in previous . The fourth-order Runge-Kutta scheme may be used for the time derivative Kliakhandler (1999). al (1994) is devoted to the KdV soliton detection from a harmonic input. The control of the simulations in the Fourier space shows the good resolution of the computed solutions.e. al (2001) the solitary wave formation from a periodic input was studied for an equation similar to Eq. The latter ensures fair resolution of the whole solutions computed. The pseudo-spectral numerical scheme has proved very eﬃcient in solving dispersive and dissipative equations.(1. 2. see also Zwillenger (1989) and references therein. rather long.e. At the same time. Salupere et.4(a). this method allows to study an evidence of the selected solitary waves Kliakhandler et. Numerical solutions of the nonlinear diﬀusion equations are considered in Fletcher (1984). it has been checked that the tendency to blow-up. i. al (1993) for the ﬁfth-order KdV equation. The pseudo-spectral technique was employed for the spatial discretization and the Runge-Kutta fourth order scheme for the time advance. by the fast Fourier transform (FFT) algorithm. al (2000) predicted on the basis of the asymptotic analysis of Eq. al (1997). Fig.13) in previous section.(1. Fig. the following numerical technique has been employed. i. 2.3) is demonstrated in Sec. The tests with smaller time steps and better resolution gave indistinguishable results. there already exist monographs Fletcher (1984).(1.(1. To reveal the expected behaviour of the solutions. Periodic boundary conditions have been used for simulations.February 11. the number of discretization points was chosen to be 4096. It provides very mild stability restrictions on the time step.4(b).01. λc is covered by 16 points. while the paper Salupere et. the damping.2. In a pseudospectral scheme the space derivatives are approximated very accurately by means of the Fourier transforms. The time step was chosen to be 0. The length of spatial domain was chosen to be 256λc . The evolution of the initial monotonic solitary wave into radiating or oscillatory solitary waves was simulated in Benilov et. Sachdev (1987). In a series of papers Salupere et. First. Since at the very late stage of the evolution the unstable waves are expected to be controlled by the ﬁnite-amplitude waves found in Kawahara (1983). Sachdev (1987) where they are explained in details.1.

51). α5 = −2.46). This observation allows to ﬁnd analytically “time-of-life” of the blowing solutions of Eq. In the case of ”blow-up”. section.583 and the veloc(2. −A/B. we found that the pulse tends to grow rapidly at the time t∗ rather close to the predicted t∗ = T ∗ /ε from (2. We consider both the selection occurring from ”below” when the magnitude of an initial Gaussian pulse is smaller than that of the eventually selected solitary waves and the selection from ”above” when the selected solitary wave amplitude is smaller than that of the initial pulse magnitude. (1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 54 Ampliﬁcation of Nonlinear Strain Waves in Solids (a) time Blow-up (b) time Damping distance distance Fig. B < 0. 6α3 b2 /α1 = 0. The amplitude.February 11.389.49) and b = ity V = 4α3 b2 = 0. Choosing the parameter values ε = 0. α1 = 1.13). α4 = 6/5. α3 = 1.4 Blow-up (a) and damping (b) of the initial conditions. α2 = 1.1. the resulting amplitude of the selected solitary wave is obtained using (2. al (2000) let us discuss in detail the selection of ”dissipative” solitary waves occurring at A > 0. Following Kliakhandler et. 2. This would permit to separate the selection mechanism from .

February 11. Due to smallness of ε.5 that up to the time t ∼ 120 an initial Gaussian pulse with the magnitude 0.3 < 0.5 ”Dissipative” solitary wave selection from an initial Gaussian proﬁle with amplitude 0. 2.583 and width 36 breaks into a train of three localized pulses aligned in row of decreasing magnitude. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 55 720 Time 480 240 0 200 400 600 800 Distance Fig.3 and width 36 units. Oron and Rosenau (1997). This may be seen by compar- . One can see in Fig. those originating from the growing unstable disturbances Kawahara (1983). the inﬂuence of the dissipative non-KdV terms is small at this stage. 2.

As a result. shown as solid lines. ”Dissipative” solitary waves form a bound state Christov and Velarde (1995). while each of three KdV solitons continues propagation with its own amplitude and velocity. with pure KdV case.6 ”Dissipative” solitary wave selection from an initial Gaussian proﬁle with amplitude 1 and width 12 units. At later stages the initial pulse transforms into a train of the solitary waves.(1.38 in agreement with the theory of single solitary wave selection.February 11. At nonzero ε each solitary wave amplitude and velocity tend to the values 0. Nekorkin and Velarde (1994) whose unequal spacing between equally high crests reﬂects the original separation of the solitary waves in the KdV stage when higher ideal solitons travel faster. The tail behind the train of solitary waves appear as a result of short wave instability similar to Kawahara (1983). However. α2 = α4 = α5 = 0. 2.585 and 0.13). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 56 Ampliﬁcation of Nonlinear Strain Waves in Solids 480 420 360 Time 300 240 180 120 60 0 200 400 600 800 1000 Distance Fig. the solitary waves bound state es- . ison of solutions of Eq. shown by dashed lines. the solitary waves have higher velocity than the velocity of growing wave packets.

2 Nonlinear hyperbolic equations Among the equations of kind (2.585 and velocity 0. Christov and Velarde (1995) to account for numerical solutions of the Boussinesq equation and its generalizations where Newton’s quasi-linearization of the nonlinear terms is employed. Again the comparison with pure KdV case is shown by dashed lines. They are robust and remain the same under the mesh reﬁnement and smaller time steps. The selection process realized from ”above” is shown in Fig. al (1987) is obtained using the Gauss elimination.583 and the width 12. An implicit diﬀerence scheme was developed in Christov and Maugin (1995). al (1996).38 appear as a result of decrease of the magnitude of the initial pulse. that may be written in the form of so-called double-dispersive equation Erofeev and Klyueva (2002).3. All features of the selection process are similar to the selection from ”below”. Samsonov (2001) utt − α1 uxx − α2 ( u2 )xx − α3 uxxtt + α4 uxxxx = 0. which are not reported here. Two equal solitary waves with amplitude 0.13) with other values of parameters ε. al (1987). αi such that A > 0.53) we are especially interested in various Boussinesq-like long waves equations. There is an another method developed independently the USSR in 1953 Godunov and Ryaben’kii (1987). A similar ﬁnding for the GKS equation was reported in Chang et.(2. We see on the last stages that the magnitude of the tail saturates. A Fourier-Galerkin method were applied to computing localized solutions in Christou and Christov (2000). (2. Christou and Christov (2002).54) At α3 = 0 it corresponds to the classic Boussinesq equation. also higher-order derivative terms may be incorporated in Eq.February 11.54) Christou and Christov (2000). 2. Christov et.(1. cubic or higher-order nonlinearities may be considered Christou and Christov (2002). The solution in Soerensen et. Christov and Velarde (1994). Besides quadratic nonlinear term. B < 0. Soerensen et. Samarskii and Nikolaev (1989) and in the . α2 ( u2 )xx . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 57 capes the destruction induced by the radiation that lags behind. 2. al (1984). Simulations of Eq. Oron and Rosenau (1997). show the same features of the selection process as described above. al (1984). All these structures are quite well resolved and similar to those found in Kawahara (1983). Soerensen et.6 when initial Gaussian pulse has the magnitude 1 > 0. al (1995). Soerensen et.

However. Then coming back from n = N − 1 to n = 1 one can ﬁnd every un from (2. Kn+1/2 . with Ln+1/2 = − fn − an Ln−1/2 cn . Thomas (1949). Let us assume u0 = L1/2 u1 + K1/2 .54). Using (2. one can obtain uN −1 . Since the Thomas algorithm is not widely used for solutions of the nonlinear waves problems. One can see further in the book that it gives numerical results in a good agreement with analytical predictions. It was noted . uN = ψ. with L3/2 = −c1 /b1 .(2. u0 = ϕ. K1/2 = (f1 − a1 ϕ)/b1 . where L1/2 = 0. Kn+1/2 = . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 58 Ampliﬁcation of Nonlinear Strain Waves in Solids USA by L. Richtmyer and Morton (1967). It allows to exclude u0 from the equation (2. Kn+1/2 . Following this procedure one can obtain un = Ln+1/2 un+1 + Kn+1/2 . Godunov and Ryaben’kii (1973) consider a boundary problem: an un−1 + bn un + cn un+1 = fn . it is useful to describe it in a more elaborate manner.57) and the boundary conditions (2.57) (2. giving u0 = L3/2 u2 + K3/2 .H. Recently it was successfully applied in Bukhanovsky and Samsonov (1998). That is why it is known in the West as the Thomas method Morton and Mayers (1994). this method is justiﬁed for the solutions of linear problems. al (1998) for computing rather complicated hyperbolic nonlinear elastic systems. 0 < n < N.55) at n = 1.56) Hence going from 1 to N one can calculate the coeﬃcients Ln+1/2 . As noted in Godunov and Ryaben’kii (1973). bn + an Ln−1/2 bn + an Ln−1/2 (2. Porubov et.55) (2. Following Godunov and Ryaben’kii (1987). the author applied it more than ten years ago for a numerical solution of Eq.February 11.57) using already known Ln+1/2 . K1/2 = ϕ. Previously the Thomas method was used for both nonlinear evolution equationsBerezin (1987) and hyperbolic equationAlexeyev (1999).56) at n = N .

These equations may be too complicated for manual operations.54) while vn = u(xn .(2. 2. Among the advantages of the Mathematica one can mention variety of the build-in mathematical functions. Then we have for discretization of Eq. where t. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 59 in Godunov and Ryaben’kii (1987). x are time and space steps respectively.4 Use of Mathematica Recently various symbolic mathematical programs were developed to provide huge analytical operations. Indeed. tj−2 ) correspond to the mesh functions on previous time steps. The diﬀerence scheme is similar to those used in Soerensen et. fn = x2 (wn −2vn )−α1 t2 (vn−1 −2vn +vn+1 )+α3 (2vn−1 −4vn +2vn+1 − wn−1 + 2wn − wn+1 ) − α4 t2 / x2 (vn−2 − 4vn−1 + 6vn − 4vn+1 + 2 2 2 vn+2 ) − α2 t2 (vn−1 − 2vn + vn+1 ). One of the most powerful is the Mathematica developed by Wolfram (1999) that is used by the author on all stages of his studies. is always satisﬁed at positive α3 . This section is not focused on the detailed description of the magniﬁcent abilities of the program. Assume un = u(xn . tj−1 ).1 is based on the transformation of the problem of a solution of a PDE to the problem of the solution of algebraic equations for the ansatz parameters. Soerensen et.(2. Godunov and Ryaben’kii (1973) that small amount of arithmetic operations together with a weak sensitivity to the calculation errors are the main advantages of the method. Only some important features are considered in relation to the problems studied here. An example may be found in Parkes and Duﬀy (1996) where a package is presented to obtain . 2. Design of the Mathematica package provides an automatic ﬁnding of a solution.54) an = cn = α3 . al (1984). Uniqueness condition. wn = u(xn . In particular. tj ) is a mesh function of a solution of Eq. the procedure described in Sec. al (1987) in the case α4 = 0 and with higher order nonlinearities being taken into account. it works eﬃciently with the Weierstrass elliptic function ℘ that may be used for obtaining exact solutions. and use of the Mathematica allows to obtain solutions avoiding errors. |bn | > |an | + |cn | Godunov and Ryaben’kii (1973).February 11. bn = − x2 − 2α3 .

3. 2. solutions in terms of the hyperbolic tanh-function. Hence. The Mathematica cannot do it automatically. First.1. However.7 a) Incorrect and b) correct representations of the function y.February 11. it is necessary to introduce the rules for the elliptic functions derivatives. see Sec. Second. it is better to manage the substitution procedure step by step manually introducing the commands that provide the most eﬃcient line of attack on the problem. The algebraic manipulations should be used also for a check of the asymptotic solutions and for derivation of the governing nonlinear equa- . it works simultaneously with all possible solutions of the algebraic equations that yields huge expressions at the intermediate stages of a solution and may result in falling down the evaluation. Finally. sometimes combinations like g2 C − g3 − 4C 3 or 12C 2 − g2 should be kept in the solution for its convenient analysis. it is unlikely that an automated method may be applied for ﬁnding periodic solutions. 2. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 60 Ampliﬁcation of Nonlinear Strain Waves in Solids y 6 4 2 2 -2 -4 -6 4 a 6 8 10 x y b 30 20 10 2 4 6 8 10 x Fig.

2.2.7(b) around which the proﬁles are developed in Figs.2 requires calculation of the function y = Π[ϕ. where x0 = K(κ)/k. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Mathematical tools for the governing equations analysis 61 tions. Sometimes. 5 x0 ≤ x ≤ 7 x0 ] = 3π − ϕ. 0 ≤ x ≤ x0 ] = ϕ. Numerical abilities may be eﬃciently used for the analysis of analytical relationships. 0. 2. e. Kliakhandler (1999). In particular. see. n. κ]. C++ or other languages are preferable for numerical simulations of nonlinear PDEs. the construction of the proﬁles shown in Figs. numerical data obtained using Fortran or C++ programs.7(a). Direct Mathematica command Plot[y. The Mathematica possesses high-level graphic facilities. In order to obtain correct smooth proﬁle it is necessary to deﬁne y as y = Π[ψ. where ψ is obtained using the Mathematica commands ψ[x /. x0 ≤ x ≤ 3 x0 ] = π − ϕ. it admits representation of graphics in many formats including PostScript and Encapsulate PostScript. ψ[x /. K(κ) is the complete elliptic integral of the ﬁrst kind. n.February 11. ψ[x /. the Mathematica gives incorrect results..1. κ].1 and 2. One can see it contains the points where there is no ﬁrst derivative of y. Almost all ﬁgures in the book are prepared using the Mathematica. Then we obtain required smooth proﬁle for y shown in Fig. may be represented by an automated procedure. ψ[x /. 2. . etc. 10}] yields the proﬁle shown in Fig. 3 x0 ≤ x ≤ 5 x0 ] = 2π + ϕ. κ is the elliptic functions modulus. while Fortran. ϕ = arcsin(sn(k x)) expressed through the elliptic integral of the third kind. {x. In particular. Like analytical procedures graphics may be automatized. 2.g. 2.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 62 Ampliﬁcation of Nonlinear Strain Waves in Solids .February 11.

Parker and Maugin (1987) etc. solid may be considered as a crystalline material in which the constituent atoms are arranged in a 3D lattice with certain symmetries. these deﬁnitions complement each other allowing to take into account a model of elastic potential of atomic interactions. Jeﬀrey and Engelbrecht(1994). e. in design of nonlinear acoustic devices Oliner (1978). From the point of view of the theory of discrete media.. During the long period only linear theory of elasticity was considered since because of the engineering needs and poor experimental facilities. Samsonov (2001). Murnaghan (1951). Engelbrecht (1997). Now the study of the material properties Lurie (1990). Parker and Maugin (1987). acoustic signals Biryukov et. Maugin (1993). Lurie (1990). Oliner (1978). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter 3 Strain solitary waves in an elastic rod The theory of strain waves in solids began to develop over two hundred years ago. As noted previously Jeﬀrey and Engelbrecht(1994).February 11. Erofeev and Klyueva (2002). Murnaghan (1951). Strain waves in solids may be classiﬁed as follows. Mayer (1995). Comparing the direction of the wave propagation with the particle motion one can distinguish longitudinal and shear waves. and to cover amorphous. When elastic features are one and the same in any direction we have isotropic solids. In macroelasticity solid may be deﬁned as a substance having a deﬁnite volume and shape and resisting forces that tend to alter its volume or shape. Sometimes anisotropy is important. al (1991). The former wave propagates along the direction of the 63 .g. while nonlinear waves in solids were considered in Engelbrecht (1983). see about it Love (1927). Maugin (1995). McNiven and McCoy (1974). Erofeev (2002). In order to go further it is necessary to deﬁne a notion of the word solid. require mathematical models based on the nonlinear elasticity. Samsonov (2001). porous or granular media. Recent developments in general elastic theory may be found in Bland (1960). Parker (1994). Parker and Maugin (1987).

not only the positions of the particles vary during the deformation but also the distances between them. Waves propagating inside solid are called bulk waves. 3. Samsonov (2001). However. Other kinds of nonlinearities are considered in Engelbrecht (1997). In rectangular Cartesian coordinates xi the components of C may be written in a more familiar form Cik = 1 ∂ui ∂uk ∂ul ∂ul ( + + ). As a result of a deformation process there appear stresses. Loading forces provide the displacement of particle yielding the current or actual conﬁguration characterized by an another vector-radius − → R . It was found more than hundred years ago that Hook’s linear law of elasticity is insuﬃcient.r . In this book main attention is paid to the longitudinal bulk waves in isotropic media and wave guides. The geometrical nonlinearity is described by the exact expression of the strain tensor always used in the theory of large deformations. 2 ∂xk ∂xi ∂xi ∂xk where uk are the components of the displacement vector V . In order to describe the alteration of the distance a deformation or strain tensor is introduced. whose general form is C= → V +( → V )T + → → V ·( → V )T /2 → (written in terms of a vector gradient V and its transpose ( V )T ). Since Piola-Kirchoﬀ stress tensor is deﬁned . the latter -perpendicular to it. see about it in Jeﬀrey and Engelbrecht(1994). Initially the position of a particle − is accounted for a vector-radius →. Jeﬀrey and Engelbrecht(1994). V = − − → → R . Then the movement is described by the displacement vector.February 11. In the reference conﬁguration the Cauchy-Green ﬁnite deformation tensor C is deﬁned Lurie (1990). and this is a source of the physical nonlinearity. In presence of a lateral surface. It is obtained from the diﬀerence between the squares of the arc length in the deformed (actual) and undeformed (reference) conﬁguration. surface strain waves are possible. or an initial or reference conﬁguration r is deﬁned.1 The sources of nonlinearities Among the possible sources of nonlinearity we brieﬂy consider so-called geometrical and physical nonlinearities since they aﬀect the strain wave propagation to a greater extent. Certainly they should be connected with the strains. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 64 Ampliﬁcation of Nonlinear Strain Waves in Solids particle motion.

hence the second order elastic moduli. It was Murnaghan (1951) who supposed to develop the energy as a power series in the three invariants of the strain tensor. J1 (C) = trC. 2 1 6 (3. µ). I2 (C) = [(trC)2 − trC2 ]/2. J3 (C) = trC3 .(3. in the form Lurie (1990) Π = Since 2 J1 (C) = I1 (C) . ν3 = n/4. ν1 3 λ 2 J + µJ2 + J1 + ν2 J1 J2 + ν3 J3 . Accordingly. ν2 = m − n/2. k = 1. In some cases there is a need in more terms in Eq.1) where Ik .3) coincide if ν1 = 2l − 2m + n. Π = λ + 2µ 2 l + 2m 3 I1 − 2µI2 + I1 − 2mI1 I2 + nI3 . or the Lam´ coeﬃcients (λ.1): Π= l + 2m 3 λ + 2µ 2 I1 − 2µI2 + I1 − 2mI1 I2 + nI3 + 2 3 .1)account for linear elasticity. The energy of deformation must be insensitive to the rotation of the reference frame. Other terms in (3. 2 3 (3.February 11. Pik = ∂Π . 3 are the invariants of tensor C: I1 (C) = trC. The energy may be written using another set of invariants.2) The ﬁrst two terms in (3. the third order elastic moduli. J2 (C) = trC2 . 3 J3 (C) = I1 (C) − 3I1 (C)I2 (C) + 3I3 (C). ∂Cik one can say the physical nonlinearity depends upon the structure of the internal (or free) strain energy density Π. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 65 in the reference conﬁguration. through the volume density of the internal energy Π in adiabatic processes (or through the Helmgholtz free energy in thermoelastic processes).3) Eqs. Murnaghan (1951).1) and (3. 2.(3. (3. I3 (C) = det C. n) account for nonlinear elastic properties of the isotropic material. m. J2 (C) = I1 (C) − 2I2 (C). or the Murnaghan moduli (l. characterize linear elastic e properties of the isotropic material.1) describe material or physical nonlinearity Lurie (1990).

4) are convenient to account for the deformation of compressible materials (metals. When torsions are neglected. (3.2. there exist another models. like rubber.1 material Polystyrene Steel Hecla 37 Aluminium 2S Pyrex glass SiO2 melted Lame’s and Murnaghan’s modulii. 3. ϕ [0.1 27..5 15. the displacement vector is V = (u. whose yield point is small. a3 .3 l -18. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 66 Ampliﬁcation of Nonlinear Strain Waves in Solids Table 3. However. free lateral surface cylindrical elastic rod.). mainly the third-order moduli data may be found in the literature Frantsevich et. Once the reference conﬁguration is deﬁned we use Hamilton’s variation principle to obtain the governing equations together with the boundary conditions. the Mooney-Rivlin model. t1 ∞ R δS = δ t0 dt2π −∞ dx 0 r Ldr = 0.2 Modelling of nonlinear strain waves in a free lateral surface elastic rod Statement of the problem 3.5) .1. (3. a2 . a4 ) can be either positive or negative. r. ϕ). Axi-symmetry leads to using cylindrical Langrangian coordinates (x. in Dai (1998).g.1 Let us consider an isotropic.. the Mooney model Lurie (1990).3 -461 -311 92 71 n -10 -358 -228 420 -44 4 2 2 a1 I1 + a2 I1 I2 + a3 I1 I3 + a4 I2 .g. w. see Fig.9 µ 0. e. may be used even for compressible materials. The series expansions (3. setting to zero the variation of the action functional. e. etc.6 27. axially inﬁnitely extended. its application to nonlinear strain waves in a rod may be found.71 111 57 13. (3.1. 3. where x is the axis of the rod.1).5 31. 2π]. al (1982). 0 ≤ r ≤ R. For incompressible materials. 0). polymers. ∗10−9 N/m2 λ 1. Its generalization. Lurie (1990).4) The fourth order moduli (a1 .95 82. some of them are collected in Table 3.February 11.9 -459 -299 14 129 m -13. We shall consider the propagation of longitudinal strain waves of small but ﬁnite amplitude in the rod.

5) is carried out at the initial time t = t0 .February 11.1). In absence of torsions non-zero . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 67 Fig. 3. equilibrium state. L=K − Π.6) where ρ0 is the rod material density at t = t0 . L is the Lagrangian density per unit volume. For the kinetic energy density K we have ρ0 2 ∂u ∂t 2 K= + ∂w ∂t 2 (3.1 Free lateral surface cylindrical rod where. The integration in brackets in (3. the rod is supposed to be in its natural.(3. Initially. with Π deﬁned by Eq.

2 x 2 r r 2r 1 Crx = (ur + wx + ux wr + wx wr ).2) may be written as I1 (C) = Cxx + Crr + Cϕϕ . (3. The following boundary conditions (b. at r → 0.Kirchhoﬀ stress tensor P are deﬁned as Prr = (λ + 2µ) wr + λ w λ + 2µ + m 2 + λ ux + ur + r 2 3λ + 6µ + 2l + 4m 2 w λ + 2l w2 wr + (λ + 2l) wr + + 2 r 2 r2 w λ + 2l 2 ux + (λ + 2l) ux wr + (2l − 2m + n) ux + r 2 λ + 2µ + m 2 wx + (µ + m) ur wx . at r = R.7) (3. 2 I2 (C) = Cxx Crr + Cxx Cϕϕ + Crr Cϕϕ − Crx . Crr = wr + (u2 + wr ). Prx = 0. (3.c.8) (3. Prr = 0 . 2 I3 (C) = Cϕϕ (Cxx Crr − Crx ).February 11. are 1 1 1 1 2 2 Cxx = ux + (u2 + wx ). 2m − n w wx + (µ + m) wx wr + 2 r (3.11) . at r = R.) are imposed for a free lateral surface rod: w → 0.9) where the components Prr . Prx of the Piola . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 68 Ampliﬁcation of Nonlinear Strain Waves in Solids components of the Cauchy-Green deformation tensor C. Cϕϕ = w + 2 w2 .10) 2 2λ + 2m − n w ur + 2 r Prx = µ (ur + wx ) + (λ + 2µ + m) ur wr + (λ + 2µ + m) ux ur + (µ + m) ux wx . 2 Hence the invariants in (3.

t) does not change along .9). Finally. An alternative is to simplify the problem making some assumptions about the behavior of longitudinal and/or shear displacements and/or strains in the elastic wave-guide. Then every cross section of the rod remains ﬂat. one can assume B 1. t. the coeﬃcients of the nonlinear terms usually contain combinations of elastic moduli which may be also small in addition to the smallness of B something not predicted beforehand. are not well deﬁned. while general equations for displacements u and w are of the second order in time. leading to an asymptotic solution of the problem. This equation is of the second order of time. hence. this procedure gives equations of only ﬁrst order in time.2 Derivation of the governing equation Exception of torsions provides transformation of the initial 3D problem into a 2D one. Further. solitary wave has an inﬁnite wave length. Referring to the elastic rod these relationships give explicit dependence of u and w upon the radius. In particular. the simplest assumption is the plane cross section hypothesis McNiven and McCoy (1974): the longitudinal deformation process is similar to the beards movement on the thread. For an elastic rod.2. this procedure has some disadvantages. Therefore the solution of the model equation will not satisfy two independent initial conditions on longitudinal strains or displacements Samsonov (2001). Subsequent simpliﬁcation is caused by the consideration of only long elastic waves with the ratio R/L 1 between the rod radius R and typical wavelength L.c. u = U (x.8).5) yields the governing equation in dimensional form for this function. while their variations along the rod axis are described by some unknown function and its derivatives along the axis of the rod. hence. Since Murnaghan’s material have small yield points. The typical elastic strain magnitude B does not exceed the yield point of the material. comparison of the predictions from the dimensionless solution to the experiments suﬀers from the fact that both B and L.February 11. subsequent scaling may take into account their orders when introducing small parameters. Any combinations of elastic moduli appear in the coeﬃcients of the equation. In particular.5) yields a set of coupled equations for u and w together with the b. hence. (3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 69 3. To obtain a solution in universal way one usually proceeds to the dimensionless form of the equations and looks for the unknown displacement vector components in the form of power series in the small parameters of the problem (for example R/L). However. Then the application of Hamilton’s principle (3. hence its solution can satisfy two independent initial conditions. The Hamilton principle (3. (3.

t). In contrast to the theory based on the plane cross section hypothesis and the Love hypothesis.8). uL = u0 (x. (3. longitudinal and shear deformations are related.. t) + . uN L = uN L0 (x. for the stresses on the lateral surface of the rod.c. However.. with ν the Poisson coeﬃcient.12) w = wL + wN L . (3.. Unfortunately. t) + .3 Double-dispersive equation and its solitary wave solution In order to derive the governing equation for longitudinal strain waves in a free lateral surface rod we assume that B ∼ R2 /L2 .. We get u0 (x.12). Later it was developed in Porubov and Velarde (2000).9) as well as the condition for w (3. the present theory allows to account for nonzero b.8).7). t) + r2 w2 (x..13) Substituting the linear parts uL and wL (3. t) = U (x..c. on the lateral surface of the rod (3.. the ”linear” and ”nonlinear” parts of the relationships may be obtained separately. this assumption is not enough due to the Poisson eﬀect. B 1.. (3.1) and (3. (3.. wN L = wN L0 (x. t) + .. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 70 Ampliﬁcation of Nonlinear Strain Waves in Solids the radius r. i.c.c.e.. Accordingly.6).. This assumption . t) + r uN L1 (x. t) + r u1 (x.12).13) into the b. t) + .February 11. and equating to zero terms at equal powers of r one obtains uk and wk . (3.c. A power series approximations is used. the plane cross-section hypothesis and Love’s hypothesis do not satisfy the boundary conditions that demand vanishing of both the normal and tangential stresses. t) while other uk and wk are expressed through U and its derivatives. (3. (3.9). the longitudinal and shear displacement in dimensional form are: u = uL + uN L . t) + r wN L1 (x.7) and in the linear parts of b. Another theory has been proposed in Porubov and Samsonov (1993) to ﬁnd the relationships between displacement vector components satisfying b. (3. at the lateral surface of the rod with prescribed precision. Prr and Prx . That is why Love (1927) proposed to use a relationship between w and u: w = −r ν Ux . Since pure elastic wave are studied. wN L are similarly obtained from the full b. wL = w0 (x. as generally done for long wave processes.. t) + r w1 (x. Using these results the nonlinear parts uN L . 3.13) are substituted into (3. t) + r2 u2 (x. Then (3.5) we obtain the single governing equation for the unknown function U (x. Running the Hamilton principle (3.

17) Here ν and E are the Poisson ratio and the Young modulus correspondingly. while terms 2 b3 r3 Uxxx and B1 r Ux are needed to satisfy the boundary conditions on the lateral surface with prescribed accuracy.15) where a2 = ν ν2 .13) into the boundary conditions and equating to zero combinations at each power of the radius. 2 2(3 − 2ν) B1 = ν(1 + ν) (1 − 2ν)(1 + ν) + [l(1 − 2ν)2 + 2m(1 + ν) − nν]. B1 = 0. β = 3E + l(1 − 2ν)3 + 4m(1 − 2ν)(1 + ν) + 6nν 2 . (3. (3. DDE. (3.5) one can obtain the so-called doubledispersive equation. 2(λ + µ) λ+µ while β is a nonlinear coeﬃcient.17) in order to be in an agreement with the ﬁve-constant Murnaghan approximation (3.(3. 2 w = b1 r Ux + b3 r3 Uxxx + B1 r Ux .1). vtt − α1 vxx − α2 ( v 2 )xx − α3 vxxtt + α4 vxxxx = 0. 2 t 2 EUx + Then kinetic and potential energy truncated approximations are K= 1 2 (3.E = .18) . (3. a2 = 0.16). one can see that only the term a2 r2 Uxx makes its contribution into Eqs. Substituting power series (3. we get u = U + a2 r2 Uxx .15).February 11. Substituting (3. We have to truncate the approximations (3. (3. where cubic nonlinear terms are neglected.16) Π= β 3 U + νEr2 Ux Uxxx 3 x (3. (3.14) (3. b3 = 0.17). b3 = . 2 E ρ0 2 2 (U + νr2 [Ut Uxxt + νUxt ]).12).14). Comparing these relations with those obtained using cross-section and Love’s hypothesis. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 71 provides a balance between nonlinearity and dispersion required for existence of the bell-shaped solitary waves of permanent shape. for a strain function v = Ux . b1 = −ν. (3.16).17) into (3. (3.16). ν= λ µ(3λ + 2µ) .

3.5 1 0. The important diﬀerence is only in the values of the dispersive terms coeﬃcients α3 and α4 .5 -0. see about it Erofeev and Klyueva (2002).2 Formation of solitary waves from an initial rectangular tensile pulse.5 1 0.February 11. α4 = µν 2 R2 /(2ρ0 ).(3. 1 0. α3 = . v = A cosh−2 (k (x − V t)).19) . ρ0 2ρ0 2 2ρ0 For the ﬁrst time DDE was derived independently by some authors.5 -0.5 -0. Samsonov (2001).5 -0.5 1 0.5 1 0.5 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 Fig.5 -0. Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 72 Ampliﬁcation of Nonlinear Strain Waves in Solids where α1 = β ν(ν − 1)R2 νER2 E . α4 = − . α2 = . in framework of the plane cross section and Love’s hypothesis.18) possesses exact travelling solitary wave solution that may be obtained by direct integration.5 -0. in Samsonov (1988) it was α3 = −ν 2 R2 /2.5 1 0.5 -0. (3.5 -0.5 1 0.5 1 0.

The nonlinearity coefﬁcient is the only coeﬃcient carrying an information about the Murnaghan moduli. V 2 lies either inside the interval E E < V 2 < c2 = . they deﬁne whether tensile or compression solitary wave .5 1 0.5 -0.5 -0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 73 with A= 3(ρ0 V 2 − E) 2 E − ρ0 V 2 . if − 1 < ν < 0.5 -0. ∗ ρ0 (1 − ν) ρ0 or in c2 < V 2 < ∗ E .5 1 0.5 1 0. When positive Poisson ratio is positive.5 1 0.5 1 0. Hence.20) Accordingly. the sign of the amplitude is deﬁned by the sign of β.3 Formation of a train of solitary waves from an initial rectangular tensile pulse.5 1 0.February 11.5 -0. k = .5 -0.5 1 0.5 -0.5 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 50 100 150 200 Fig. if ν > 0.5 -0.22) (3. more precisely.21) 1 0. β 2νR2 (E + ρ0 (ν − 1)V 2 ) (3. 3. 1/2 ≥ ν > 0.5 -0. ρ0 (1 − ν) (3.

4 Delocalization of an initial compression pulse. 2.5 0.5 50 -0.3. To this purpose numerical simulation of the equation has been performed using numerical method explained in Sec. Shown in Fig.5 50 -0. We see that it is the mixed dispersive term α3 vxxtt .5 50 -0. It was found that rather arbitrary initial pulse splits into the train of solitary waves or evolves into small amplitude oscillating wave-packet according to the predictions about wave parameters done on the basis of exact single travelling solitary wave solution. while massive initial pulse splits into the sequence of solitary waves. 0. who establishes the permitted ﬁnite interval for the wave velocity. It is of interest to know how rather arbitrary localized initial pulse evolves.5 0.5 0.5 50 -0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 74 Ampliﬁcation of Nonlinear Strain Waves in Solids may propagate for a given elastic material of the rod.February 11.5 50 -0. 3.2 is the evolution of an initial tensile rectangular pulse into single solitary wave.5 0.5 100 150 200 100 150 200 100 150 200 100 150 200 100 150 200 100 150 200 100 150 200 100 150 200 Fig.5 50 -0.5 50 -0. see Fig. Exact solution requires speciﬁc initial conditions. Indeed. 3. However.5 50 -0.3.5 0.5 0. .5 0.2. 3. assume the material elastic features of the rod yield β > 0.

when studying optically transparent phase inhomogeneities. the choice of an optical . the curvature h of the proﬁle of longitudinal strain u along the radius of the rod is h = 2 |urr | /(1 + u2 )3/2 = |νUxx | /(1 + ν 2 r2 Uxx )3/2 . Interferometry is the most appropriate for such waves study because it allows to observe and measure with suﬃcient accuracy even small refractive indices alterations. Finally. caused by a strong shock wave propagation. Figure 3.4 demonstrates no solitary wave generation from the initial compression pulse. In particular. being contactless. It was found in Porubov and r Samsonov (1993)that the variation of longitudinal strain u along the radius is negligibly small when Ux is the solitary wave (3.February 11.4 Observation of longitudinal strain solitary waves We brieﬂy consider the recent successful experiments on solitary wave observation in a transparent rod. However. They allow not only to visualize inhomogeneity but also to determine its parameters. has several advantages in comparison with the conventional optical interferometry.19). For this reason both waves are distorted to the same extent and possible defects in optical elements and experimental cell do not aﬀect the resulting interference pattern. al (1995). Holographic interferometry method. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 75 each solitary wave is accounted for the exact single travelling wave solution (3. All optical methods record the alterations of refractive index in an object. single travelling wave exact solution provides correct predictions about solitary waves formation in an unsteady process of an arbitrary localized input evolution.19) that is important for the experimental observation of the waves.14). details may be found in Dreiden et. used in the experiments Dreiden et. let us estimate the deviation from the plane cross section caused by the second terms in the expression (3. Samsonov (2001). To sum up. It was shown theoretically in our case that a strain solitary wave is a propagating long density wave of small amplitude. limitations to the optics quality are considerably lower because wave fronts to be compared pass through the same optical path. 3. Indeed. The optical methods were used because they are preferable to study transparent optical phase inhomogeneities. when the material of the rod possesses β < 0. for example. Shadowgraphy is more convenient to record a considerable refractive index gradient. they do not introduce any disturbances in an object under study. al (1995). and on the other side. only compressive solitary wave arises from an initial compression localized input. However.

The solitary wave parameters are calculated based on the data of the holography interferograms obtained. The experimental set-up used to generate and observe the strain solitons. it will be suﬃcient to detect the constant shape wave propagation at much shorter distance.February 11. Curved interference fringe is extracted from the area inside the rod and placed below the interferogram. see Fig. The waves inside the rod are generated from a primary shock wave produced in the water near the edge of the rod by laser vaporization of a metal target. al (1982). and two cut-oﬀ were made parallel to the rod axis in order to make transparent the central part of the rod. The elastic properties of it are given by a set of parameters ν = 0.5. that are attained by the cell displacement along the axis of wave propagation. To check that the excited strain wave possesses indeed the solitary wave feature to conserve its shape. it is necessary to follow in observations a propagation along an extended elastic wave guide. Observations are carried out in the transversal direction. Only central part of the PS rod is transparent thanks to the vertical cuts. consists of a basin where the rod is submerged into the water. Let 2h .35. β = − 6 · 1010 N/m2 . a device to produce the initial shock wave. The ﬁrst exposure of the hologram is carried out to obtain the hologram of undisturbed wave guide. The carrier fringes on interferograms. in a wave guide made of material highly absorbing linear elastic waves. 3. only elastic materials. a holographic interferometer for the recording of a wave pattern.8 · 103 m/sec. a longitudinal density wave in solids. obtained due to the reconstruction of doubly exposed holograms. that are transparent for the given light wave length.3. a synchronizer and a laser radiation energy meter. in general. in fact. c∗ = 1. The longitudinal strain wave patterns are recorded at various distances from the input edge of the rod. Typical interferogram is shown in Fig . However. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 76 Ampliﬁcation of Nonlinear Strain Waves in Solids recording method allows to study. The solitary wave amplitude can be calculated using the interferometry fringe shift ∆K measured in the interference pattern.5. Based on the results of the analysis presented above the transparent polystyrene SD-3 has been chosen as an appropriate material for a wave guide manufacturing. The second exposure is produced by a laser pulse synchronized with the prescribed stage of the wave propagation. Note that the interferometric pattern does not exhibit a standard bell-shaped image of a shallow water solitary wave since the strain solitary wave is. see Frantsevich et. occur due to the wedge turn between the exposures.

23) Here n0 . is caused by the laser light propagation along the distance q − 2h through the water and the distance 2h through the rod ( where q is the distance between the cell walls) Samsonov et. al (1995)). 3. the distance. precisely.5 Experimental observation of longitudinal strain solitary wave in an elastic rod(after Dreiden et. the distance between two longitudinal cut-oﬀ. which light passes through the rod and water. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 77 Fig. As a result we obtain the formula for the magnitude of the light wave phase variation ∆φ2 : λ ∆φ2 .February 11.e. varies due to the deformation of the rod. Moreover. i. n1 are the refraction indices of water and the elastic material before deformation. 2π n0 (q − 2h − 2∆h) + n2 (2h + 2∆h) = (3. be a distance passed by the recording light across the rod. al (1998): λ ∆φ1 .. After the deformation the refraction index value of the rod changes to n2 .24) . 2π n0 (q − 2h) + 2hn1 = (3. Before the deformation the phase variation ∆φ1 of the light wave having the length λ.

and as a result the following relationship is valid: ∆ρ = Ux (2ν − 1). that proves the wave that is measured propagates inside the rod. undisturbed.27) The amplitude is determined by the maximal fringe shift value. hence full quantitative comparison with the theory is impossible. the theory developed for an inﬁnite cylindrical rod allows to predict an existence of solitary waves in a ﬁnite-length rod with vertical cuts. 2h[(n1 − 1)(1 − 2ν) + ν(n1 − n0 )] (3. As follows from (3. 149 mm long.25): Ux = − λ∆K .28) shows that the length L of the solitary pulse may be directly determined from the interferogram as the length of the fringe shift perturbation between two undisturbed areas. The unsteady process of the solitary wave generation cannot be accounted for the theory. The nonlinearity parameter β < 0 for polystyrene. see Lurie (1990).e. moreover. one can obtain the density variation from the solution of a static linear problem on uniaxial compression (or tension)..28) (3.25) The new value of the refraction index of the deformed rod n2 is caused by the local density variation: ∆ρ n2 − n1 = . Therefore. It is fairly long and keeps its shape on propagation. Derivation of the relationship (3. Unfortunately only the wave amplitude may be measured more or less precisely.26) On the other side. The propagation of a shock wave before the solitary wave . i.5 is indeed a longitudinal compression wave. 3.28) the wave shown under interferogram in Fig. 2π ∆K = (3.February 11. al (1995) we used a long polystyrene rod. ρ n1 − 1 (3. the interferometry fringe shift ∆K is deﬁned as ∆φ2 − ∆φ1 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 78 Ampliﬁcation of Nonlinear Strain Waves in Solids Evidently. ρ Then we get ﬁnally from (3. In our experiments Dreiden et. and we can anticipate the appearance of only compression solitary wave. in the water surrounding the rod the interference patterns remains horizontal. There is no any tensile wave around the observed wave.

it is necessary to apply an asymptotic technique. 3. The procedure developed in section 3. U = 0. A small parameter ε is chosen as ε = B = R2 /L2 << 1.February 11. (3. (3. δA = 0. If the rod end is assumed free. L for x. 3. vx = 0. the elementary work of external forces at the end of the rod.5 Reﬂection of solitary wave from the edge of the rod Following Dreiden et.29) variation principle since external forces are not necessary potential ones. al (2001) let us consider a semi-inﬁnite homogeneous rod. while on the rod end.32) (3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 79 contradicts numerical results shown in Fig.29) The Lagrangian density per unit volume. is deﬁned as before. . According to Lurie (1990) we cannot call Eq. L/c0 for t. Ut = 0. L=K − Π. t1 X R t1 δ t0 dt 2π −∞ dx 0 r Ldr + t0 δA dt = 0. vxx = 0.29) yields v = 0.(3. and B/L for v. but obvious kinematic reasons require zero displacement and its velocity. x = X.2 may be applied but now the elementary work done by the external forces should be included into the Hamilton principle. they depend upon the type of clamping. Very recently they were generated also in a plexiglas rod Samsonov et. is zero.2. and Eq. −∞ < x < X. reducing DDE to a nondimensional form. For a long time solitary waves were observed only in a wave guide made of polystyrene. Introducing the scales. where c0 = E/ρ0 is the so-called rod velocity.30) When the end is clamped. x = X.31) Since DDE does not possess an exact solution describing interacting waves moving in opposite directions. Use of the Hamilton principle yields the DDE (3. the elementary work is not determined. al (2003). that may be rewritten in terms of strains.18) as a governing equation for longitudinal waves together with the boundary conditions requiring zero values for v and its derivatives as x → −∞. (3. vxt = 0. to balance nonlinearity and dispersion. and the integration with respect to x is carried out over −∞ < x < X. (3.

on a slow time τ = εt. 2E 2 β 2 ν2 (v02 )ψ + v02. in addition to x.32) on the clamped end are satisﬁed if x02 = 2X −x01 .ψτ + β 2 2 (v01 )θθ + 2v01.θθθθ + v02. τ ) + v02 (ψ. It follows from (3. 2v1. one cannot satisfy boundary conditions (3. (3. τ ) + v12 (ψ.37) and the bounded solution of Eq. .38) Substituting the soliton solutions of the KdV equations (see Chapter 1) into the leading order solution (3.34) where θ = x + t. τ ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 80 Ampliﬁcation of Nonlinear Strain Waves in Solids Assume the asymptotic solution v depends.30) at a free end of . At order ε there is a linear equation for v1 . The solution is sought in power series in ε: v = v0 + εv1 + . E (3.ψψψψ ) 2 The absence of secular terms leads to two uncoupled KdV equations for the functions v01 and v02 .33) Substituting (3. and equating the terms with the same power of ε we obtain the D’Alembert solution for v0 : v0 = v01 (θ. ψ = x − t. 2E 2 (3. Both conditions (3.33) into dimensionless DDE.36) 2v02.ψ + (v02 )ψψ + 2E (3.February 11. (3.39) that the type of the strain wave depends upon the sign of β like in previous section. we obtain v0 = 6Eν 2 2 k [cosh−2 k[x + (1 + εν 2 k 2 )t − x01 + β cosh−2 k[x − (1 + εν 2 k 2 )t − x02 ] ].ψψψ = 0.θθθ = 0. (3. .θτ − 2v02.35) ν2 (v01.θ v02. t.τ + (3. 2v01.θψ = 2v01. (3. In this case reﬂection of the solitary wave occurs with no change of its shape.35) is v1 = β v01 v02 + v11 (θ. τ ).τ − β 2 ν2 (v01 )θ − v01.34).39) where x01 are the constant phase shifts. On the contrary. .

5 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20 Fig.5 -0.5 0.5 0. so that the area of calculation would occupy the interval 0 < x < 2X.3. Initial pulses are assumed to be the equal magnitude KdV . 2. It means that the reﬂected solitary wave does not propagate but disappears due to dispersion. It may be clearly seen from numerical simulation of the reﬂection Dreiden et.5 0. Again the implicit ﬁnite-diﬀerence scheme explained in Sec.2.5 0. al (2001).5 -0.5 -0.5 -0.5 0. is used.5 -0.5 0.5 0.5 -0.February 11.5 -0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 81 0.5 -0. In case of free end of the rod the sign of β in the interval X < x < 2X is chosen opposite to that used in the interval 0 < x < X.5 -0. 3.5 0. Implementation of the boundary conditions at the end of the rod is eﬀected by means of symmetric continuation of the calculation area beyond the real rod end.5 -0.6 Reﬂection of the solitary wave from a free end of the rod.5 0. the rod.

The type of an initial solitary wave depends upon the sign of β. In particular. Initial velocities of the solitary waves are chosen equal and taken to be directed towards each other. 3.7 Reﬂection of the solitary wave from a ﬁxed end of the rod. while in the interval 0 < x < 2X we impose the same tensile solitary wave for the clamped end or the compression solitary wave for the free end.25 5 0. solitons located centrally symmetric relative to the genuine end x = X.5 0.February 11.5 0.25 5 0. The reﬂection from the free end is shown in Fig. It is seen that the amplitude of the solitary wave propagating from left to right decreases as it reaches . 3.25 5 0.25 5 0.5 0.5 0.5 0. if β > 0 the initial solitary wave in the interval 0 < x < X is a tensile wave.5 0.5 0.5 0. The right-hand side of the ﬁgure corresponds to the free end.25 5 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 10 15 20 Fig.5 0.25 5 0.25 5 0.6.25 5 0.25 5 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 82 Ampliﬁcation of Nonlinear Strain Waves in Solids 0.5 0.25 5 0.

In agreement with the theory the amplitude of the solitary wave is seen to be nearly twice as large when it reaches the end of the rod. 3. The reﬂected wave has the same amplitude and velocity as the incident one. Keeping its . then the wave is dispersed. its amplitude grows. and no localized strain wave is observed near the input end of the rod. ﬁrst. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 83 Fig.February 11. 3.8 Observation of a reﬂection of the solitary wave from a free end of the rod. The reﬂected wave is of opposite.7. the end of the rod. In case of the clamped end numerical results are shown in Fig.

It was found in experiments the absence of any reﬂected localized waves .7. Footnotes in Fig. Shown in Fig. al (2001) using the same technique as for the study of the solitary wave propagation. shape.8 is the reﬂection from the free end. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 84 Ampliﬁcation of Nonlinear Strain Waves in Solids Fig. 3. 3. 3.9 Reﬂection of the solitary wave from a clamped end of the rod. the reﬂected wave propagates towards the input end of the rod.8(a). cf. 3. with the ﬁrst and the ﬁfth stages in Fig.8(b) demonstrate the decrease of the amplitude of the initially generated compression localized wave shown in Fig. 3. 3.February 11. Experimental observation of the solitary strain wave reﬂection were carried out in Dreiden et.8(a) and Fig.

10 demonstrate the reﬂection of the strain solitary wave from the rod end attached to the right to a massive brass plate.10 Reﬂected wave comes back to the input end of the rod. cf. moving towards the input end of the rod. It is seen that the amplitude of the . It was found experimentally in Dreiden et.9.6. 3. The material of the plate is chosen so as to avoid penetration of the wave energy outside the rod. Figures 3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Strain solitary waves in an elastic rod 85 Fig. al (2001) that the acoustic resistance of the plate should be much greater than that of the material of the rod (polystyrene).February 11. Hence all observations verify theoretical predictions. 3. 3. with the last stage of Fig.

The wave observed in experiments demonstrates the main feature of the solitary waves to keep their shapes after collisions. Fig. is almost doubled by the ﬁxed end of the rod. the last stages in Fig.10 shows the reﬂected solitary wave moving to the left at a distance of 140 mm from the ﬁxed end. 3. 3. hence it conﬁrms also the fact that the observed localized wave is indeed the strain solitary wave predicted by the theory.7.9(b)in an agreement with the ﬁrst and the ﬁfth stages in Fig.10). 3.10) from the fringe shift measured inside the rod (marked by A in Fig. Fig. Figure 3. It has the same amplitude and velocity as the incident one. 3.7.February 11. cf. 3.9(a). In order to see it is necessary to subtract the fringe shift outside the rod (marked by B in Fig. 3. . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 86 Ampliﬁcation of Nonlinear Strain Waves in Solids incident solitary wave.

ϕ). one can write the displacement vector V = (u. 0). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter 4 Ampliﬁcation of strain waves in absence of external energy inﬂux As we seen in the previous chapter. The simplest reason is the varying cross section of the rod. However. We call it an inhomogeneous rod in the following for convenience. Basic equations.1 Governing equation for longitudinal strain waves propagation Let us consider the wave propagation problem for an isotropic inﬁnite nonlinearly elastic compressible rod with varying cross section.1 Longitudinal strain solitary wave ampliﬁcation in a narrowing elastic rod The section is devoted to the theoretical and experimental description of the propagation and ampliﬁcation of the strain solitary wave (soliton) in a cylindrical nonlinearly elastic rod with varying cross section. ϕ [0.February 11. r.1. the shape of the wave varies when the balance between nonlinearity and dispersion is destroyed. describing the nonlinear wave propagation in the initial 87 . where x is the axis along the rod.1. We follow Samsonov et. w. 4. if torsions are neglected. Introducing the cylindrical Lagrangian coordinate system (x. 4. al (1998) where the results presented below were published for the ﬁrst time. 0 ≤ r ≤ R(x) ≤ R0 . At the same time. 2π]. while the rod with permanent cross section will be called the homogeneous one. 4. the wave may still exist even in presence of an external medium or when a material of the rod is microstructured. see Fig. R0 − const. strain solitary wave propagates without change of the shape in an uniform rod with a free lateral surface.

conﬁguration.7)-(3. B << 1. We are studying long nonlinear longitudinal strain waves (density waves).1 Free lateral surface cylindrical rod with varying cross section. The most important case occurs when both nonlinear and dispersive features are in balance and small enough: R0 L 2 ε = B = << 1. as well as suﬃciently long waves with the length L.5). the relationships between longitudinal and transversal displacements u and w. To ﬁnd them one needs to satisfy the boundary conditions on the free lateral rod surface r = R(x). where R0 is the maximal value of r(x) along the rod . the absence of both the normal and tangential stresses at every moment. 4. that allows to do some simpliﬁcations.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 88 Ampliﬁcation of Nonlinear Strain Waves in Solids Fig. are obtained from the Hamilton variation principle (3. (4. namely.9). The unknown func- . so as the ratio R0 /L << 1. namely. taking into account that the waves under study should be elastic waves with suﬃciently small magnitude B. We introduce the small parameter ε.1) The boundary conditions have the form (3.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 89 tions u.12).5) where β = 3E + 2l(1 − 2ν)3 + 4m(1 + ν)2 (1 − 2ν) + 6nν 2 becomes the only coeﬃcient depending on the nonlinear elasticity of the rod.2.5) and calculating δS = 0. γ << 1. however. (3.3) into the formulas for K and Π. (4.2) 2 Ux .6) R ∂x 4 ∂x2 4 where c∗ is the so-called ”rod” wave velocity. 3. (4.6) on x. (3.3) Higher order terms may be obtained by the same way.13) with (4.7) . w are expanded in a power series (3.2.13). Substituting them into the boundary conditions and following the procedure explained in Sec.5). (4. one can ﬁnd respectively: K = 1 2 ρ0 2 Ut2 + νr2 Ut Uxxt + νUxt 2 2 EUx + . Substituting (4.5) into (3. t) + r2 Uxx . they will be omitted here because of no inﬂuence on the ﬁnal model equation for the longitudinal strain waves equation to be derived using the Hamilton principle (3. (4. one can obtain the following nonlinear equation: Utt − c2 ∂ 1 ∂ β 2 2 ν ∂ ν2 ∗ R2 Ux = 2 R Ux − (R4 Utt ) + R4 Uxtt R2 ∂x R ∂x 2ρ0 4 ∂x 2 1 ∂ νc2 ∂2 νR2 ∗ + 2 R4 Uxxx + R4 Ux − Uxxtt . ∗ Let us consider now the rod which cross section varies slowly along the x− axis.February 11. Substituting expansions (3. 3 x (4.12). 2 2(3 − 2ν) wN L = r ν(1 + ν) (1 − 2ν)(1 + ν) + l(1 − 2ν)2 + 2m(1 + ν) − nν 2 E (4. Introducing the notation: v = Ux . which is described by a function R = R(γx). 3.2). one can obtain ν ν2 uL = U (x. (4.4) Π = β 3 U + νEr2 Ux Uxxx . wL = −r νUx − r3 Uxxx .4). we obtain an equation ∂ 1 ∂ ∂x R2 ∂x βR2 2 (v ) + aR4 vτ τ − bR4 vxx − 4bR3 Rx vx 2E vτ τ − R2 v + = 0. τ = tc∗ and diﬀerentiating the equation (4. c2 = E/ρ0 .3. it coincides with that obtained in Sec. (4.

the wave is asymmetric like shown in Fig. 2E 0 whose solitary wave solution is Samsonov (2001): 3E α cosh−2 (k(X)[θ − θ0 (X)]) . while the last term. 1+α 4R2 [a(1 + α) − b] (4.θθ − A2 β 2 (v )θθ − R2 A4 (a − b)v0.8) Using the asymptotic method explained in Sec. . . as follows: θτ = − 1. . while A.9) into the equation (4. 1.11) Since the parameters of the solitary wave depends upon the slow coordinate.February 11.12) . 4. looks like a dissipative one. it occurs due to the cross section variation. and the two following terms are responsible for dispersive features of the rod.9) Substitution (4. k2 = .2. θx = A(X).18.2 Evolution of asymmetric strain solitary wave To describe the evolution of a travelling strain wave v we introduce the phase variable θ and the slow variable X ≡ γx. α > 0. b = −ν/2.7) is obtained in new variables in the power series in γ: v = v0 + γv1 + . being of the same order.10) depending upon the varying parameter α = α(X). (4.18) for v0 : (1 − A2 )v0.11) for most standard elastic materials (having the Poisson coeﬃcient ν > 0) if the value of the function α is inside an interval : 0 < α < ν .θθθθ = 0. k are expressed through it: A2 = α(1 + α) 1 .1. (4. the solution of Eq.(4. a = −[ν(1 − ν)]/2. 1−ν (4. the third governs the nonlinearity. β v0 = (4.7) gives in leading order of γ the ODE reduction of the nonlinear double dispersive equation (3. Both A and k will be real in Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 90 Ampliﬁcation of Nonlinear Strain Waves in Solids with the dispersion terms coeﬃcients a and b. (4. 2. Two ﬁrst terms here describe a common linear wave.

the amplitude and the velocity of the solitary wave remain constant.θθθX ).43): ln R 2 α2 2kA3 + X 4bk 2 R2 A4 ln 2R4 α2 Ak 5 X = 0. see Fig.(4. a (1 − ν) 5(1 − ν) Taking the restrictions for α.15) is always positive.10) (compression or tensile one) is deﬁned only by the sign of the nonlinear coeﬃcient β.θθθ ) + 2E 0 2 β 2 ν−1 A(2v0.θθθθ = F. which depends on the elasticity of the rod material.February 11.θθθ ) + R E β 2 ν−1 AX (v0.(4. (4. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 91 Then the type of the strain wave (4.19.(4.12). Indeed an inmogeneous linear equation holds for v1 at order O(γ) (1 − A2 )v1. 4. into account. we conclude that the expansion in the brackets on the right.θX + (v0 )θX + νR2 [ + 2A2 ]v0.θθ − A2 where F = RX β 2 A(2v0. while the increase of the radius is accompanied by the decay of the amplitude. E 2 β (v1 v0 )θθ − R2 A4 (a − b)v1. However. Eq. E (4.θ + (v )θ + νR2 [ + 3A2 ]v0. respectively.θ + (v0 )θ + νR2 [3(ν − 1) + 5A2 ]v0. The solitary wave evolution in this case is similar to that shown in Fig. Let us study a distortion of the solitary strain wave due to the ”geometrical” inhomogeneity considered. When the radius no longer alters at X > X ∗ . there is no tendency to a ﬁnite value .15) √ b 1 2 ± 9 − 5ν D = = . D1.2 = . Therefore the unbounded growth of the amplitude of the solitary wave occurs with the radius decrease.1.11) is reduced to a nonlinear ﬁrst order ODE for an amplitude variation RX = αX R where: 1 1 1 1 − − − . 6(1 − D + α) 2α 3(1 − D1 + α) 3(1 − D2 + α) (4.14) that after use of Eq.13) Then the equation for the amplitude α arises from the secular term absence condition (2. 1.hand side of Eq.

its front side becomes steeper while the back one becomes smoother.17) R shows that for the wave propagation along the narrowing rod (RX < 0) the extremum is achieved for θ − θ0 (X) > 0. (4. while the back one . Then the asymmetric solitary wave accelerates in the narrowing rod and decelerates in the expanding one in comparison with the same solitary wave moving along a uniform (homogeneous) rod. Direct integration of Eq. (4. The equation for the determination of an extremum of a derivative v0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 92 Ampliﬁcation of Nonlinear Strain Waves in Solids α∗ in the solutions of Eq.February 11. and the values of the wave parameters at X > X ∗ are not ﬁxed by the equation coeﬃcients.(4. al (1982). (4.3. (4. k 2 = α α . while in an expanding rod (RX > 0) .11).16): A = 1.x γ RX + [k(1 − γθ0. and the following approximations follow from Eqs.18) where e0 is the yield point of a material.15). see Sec. becomes smoother.X ) + γkX [θ − θ0 (X)]] tanh (k[θ − θ0 (X)]) = 0.19) . Hence this is not a selection of the solitary wave when any initial conditions provide one and the same solitary waves.(4.steeper.10).16) Routine analysis of the functions v0 . and v0. Eq.x shows that the distortion of the wave shape takes place apart from the amplitude variation. 1.16) may be easily simpliﬁed to describe the wave parameters variations. Therefore α will have to be small enough. moving along the expanding rod. and for most of elastic materials its value lies in the interval 10−4 −10−3 Frantsevich et.for an inverse sign. = 4R2 (a − b) α0 R0 R 2 . Vice versa. When the bell-shaped solitary wave propagates along the narrowing rod.15) yields R6 α3 ν + α(2ν − 6/5) − α2 (1 − ν) (1 − ν)[ν − α(1 − ν)] 2 = const.11). (4. the front side of the solitary wave. The exact formulas (4.(4. The range of the strain wave amplitude is restricted by a physical condition of the strain elasticity: | 1 + 2Cxx − 1 | < e0 . (4. (4. they are deﬁned by the initial condition.

Based on this analysis. showed that a solitary wave appears even at the distance of 60 mm (ca. Experiments on the solitary wave generation in a homogeneous rod. secondly. Measurements of the solitary wave amplitude in a homogeneous rod resulted in an estimation of the parameter ε = O(10−3 ). Therefore the inhomogeneity parameter should be chosen as γ >> ε.10) with much less velocity Samsonov (2001). see Sec. 3. the experimental setup limitations should be taken into account. and the strain solitary wave will hardly appear from an initial shock. 3.2. we were going to observe a geometrical inhomogeneity inﬂuence just on the strain solitary wave.February 11.1. then the possible variation of the initial rod radius (R0 = 5 mm) at the distance 100 mm along the axis will be of order 0. 4. however. When γ >> ε the inhomogeneity will change the initial pulse earlier than both nonlinearity and dispersion. a rod of 140 mm long was made of polystyrene with uniform and narrowing parts. as is shown in Fig. The amplitude of the plateau is negative for the narrowing rod and positive when the wave propagates along the expanding rod.3.(4.4.1 mm or 2 % from the initial value. First. 4. In this case the inhomogeneity parameter γ = 0. Thus the rod cross section should remain constant at the distance required for the solitary wave generation and separation. It has to be noted. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 93 The most important feature of the solution of Eq. and begin vary only after it. The rod radius decreases linearly from the value R0 = 5 mm to the value R = 2.032 is .13) is in the appearance of a plateau. and two cut oﬀs were made on the lateral surface for the observation purposes. 10 R0 ) approximately from the input edge of the rod. that an unsteady process takes place in experiments in contrast to the quasistationary process governed by the asymptotic solution obtained above. When the inhomogeneity parameter γ is chosen to be γ << ε. The choice of the rod cross sections variation is caused by two reasons. al (1995). propagating behind the solitary wave (4. So it seems hardly possible to detect such a deviation using our experimental setup.3 Experimental observation of the solitary wave ampliﬁcation We used the same experimental technique as used for the study of the strain solitary wave propagation in a homogeneous rod.19) shows that such a magnitude corresponds to the oscillations of the observed solitary wave front Dreiden et. see Sec. The estimation of the amplitude change in this case by means of an approximation (4.75 mm along the distance 70 mm.

At the interval where the cross section remains uniform . Some features predicted by our theory appear in experiments. 4.3.125 mm interval. 4. the steepness .February 11. ν = 0.(3. 4. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 94 Ampliﬁcation of Nonlinear Strain Waves in Solids Fig.6. al (1998). The shape of the strain wave was reconstructed by means of Eq.33.75 · 10−3 m. while the measured cross section for the tapered rod part as well as the inteferograms may be found in Samsonov et. The enlargement Fig. the increase of the amplitude. of the amplitude scale allows to visualize the main features of the solitary wave in the tapered rod. al (1995) for the homogeneous rod. Two graphs ”strain v vs. solitary pulse width L” are drawn after interpolation. n1 = 1.28) using the following values of parameters: n0 = 1. we have 2h = 2h0 = 7. Solid circles • and the dashed interpolative line both correspond to experimental data measured on a 40-60 mm interval of the rod length.2 Polysteryne rod with variable cross section and cut-oﬀs. namely. much greater than the typical solitary wave amplitude ≈ 10−4 Dreiden et. Λ = 7 · 10−7 m. It must be taken into account that light passes the diﬀerent distances 2h in diﬀerent cross sections.35. open triangles and the solid interpolative line correspond to them on a 75 . see Fig.3 Ampliﬁcation of longitudinal strain solitary wave.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux

95

of the wave front and smoothness of its back, i.e., asymmetric deformation of the bell-shaped solitary wave. Moreover, the characteristic width of the pulse, L1 = 25, 2 mm, in the homogeneous part of the rod at the one-half amplitude level is visibly greater than the similar value, L2 = 22, 3 mm, in the narrowing part, hence the width of the localized strain solitary pulse decreases along the tapered rod. The measurement of the wave amplitude is supposed to be quite plausible for the comparison with the theory. One can see that the maximal amplitude of the strain solitary wave is achieved at the distances 60 and 95 mm from the rod input edge, respectively. Then from the estimation (3.28) we obtain the solitary wave magnitudes equal to 3.29 · 10−4 in the interval 40-90 mm, and to 3.83 · 10−4 for the interval 75-125 mm. Therefore the solitary wave magnitude increases 1.16 times. The estimation using the simpliﬁed formulas (4.19), and a length dependence of the kind R = R0 − γ(x − 70) gives the ampliﬁcation as 1.31 times, which is in a good agreement with the experimental data. However, some new theoretical results cannot be checked in experiments, namely: -The experimental setup does not allow to measure directly the solitary wave acceleration caused by the narrowing cross section along the rod. -Like in case of the homogeneous rod, the precise measurement of the wave width is impossible. -There is no observation of a plateau.

4.2

Strain solitary waves in an elastic rod embedded in another elastic external medium with sliding

Stresses on the lateral surface of an elastic wave guide, e.g., an elastic rod, may appear due to its interaction with the surrounding external medium, as in some technological devices. Various types of contact models can be used at the interface between the rod and the external medium. The full (strong) contact model is used when there is continuity of both normal and shear stresses, and displacements. Alternatively, in a weak contact, friction may appear at the interface, hence a discontinuity in the shear stresses. Slippage provides another form of contact at the interface, in which only the continuity of the normal stresses and displacements is assumed. Surface stresses may also arise due to the imperfect manufacturing of the lateral surface of the wave guide and are formally like the ”surface tension” on the

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

96

Ampliﬁcation of Nonlinear Strain Waves in Solids

free surface of a liquid Biryukov et. al (1991), Nikolova (1977). The analytical solution of the contact problem is rather diﬃcult even in the framework of the linear elasticity theory, see Kerr (1964) and references therein. However, considerable progress has been achieved to account for short nonlinear surface acoustic waves propagating along the interface between elastic media Parker (1994); Parker and Maugin (1987). Recently, in the studies of strain waves in a rod interacting with an elastic external medium, attention was mostly focused on the propagation ofsurface strain waves along the lateral rod surface perpendicular to its axis (see, e.g., Gulyaev and Polsikova (1978); Shevyakhov (1977)). Here, however, we shall consider bulk density strain waves, propagating along the rod axis. Although rather useful in the study of free lateral surface rods, the so- called plane cross section hypothesis and Love’s relationship fail to properly account for contact problems, because they rule out normal stresses at the rod lateral surface, hence there is discontinuity of normal stresses at the interface of the rod and the external medium. Most of the results in this section were ﬁrst published in Porubov et. al (1998).

4.2.1

Formulation of the problem

Let us consider an isotropic, axially inﬁnitely extended, elastic rod surrounded by another albeit diﬀerent elastic medium, in which it may slide without friction, see Fig. 4.4. We shall consider the propagation of longitudinal strain waves of small but ﬁnite amplitude in the rod. Axi-symmetry leads to using cylindrical Lagrangian coordinates (x, r, ϕ), where x is the axis of the rod, ϕ [0, 2π], 0 ≤ r ≤ R. When torsions are neglected, the displacement vector is V = (u, w, 0). We choose Murnaghan’s approximation (3.1) for deformation energy for the rod. The displacement vector for the linearly elastic external medium may be written as V1 = (u1 , w1 , 0). Its density is noted by ρ1 , and its elastic properties are characterized by the Lam´ coeﬃcients (λ1 , µ1 ). Any disturbances due to the wave prope agation inside the rod are transmitted to the external medium through displacements and stresses normal to the rod surface only when contact with slippage is considered. Disturbances are assumed to decay to zero in the external medium far from the rod. The normal strains as well as the displacements inside the rod are smaller than those along the rod axis. Thus we assume that displacements and strains are inﬁnitesimal in the external medium, hence as already said it is a linear elastic one. Then for the

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux

97

Fig. 4.4

Cylindrical rod surrounded by an external elastic medium.

external medium we have: ρ1 u1,tt − (λ1 + 2µ1 ) u1,zz − (λ1 + µ1 ) w1,rz + λ1 w1,x r u1,r w1,x u1,rr + + w1,rx + r r − = 0 (4.20)

ρ1 w1,tt − (λ1 + 2µ1 ) w1,rr +

w1,r w1 − 2 − µ1 w1,xx − r r (λ1 + µ1 ) u1,rx = 0

(4.21)

The following boundary conditions (b.c.) are imposed: w → 0, w = w1 , Prr = σrr , at r → 0, at r = R, at r = R, (4.22) (4.23) (4.24)

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

98

Ampliﬁcation of Nonlinear Strain Waves in Solids

Prx = 0, σrx = 0,

at r = R,

(4.25)

u1 → 0, w1 → 0

at r → ∞.

(4.26)

where Prr , Prx denote the components of the Piola - Kirchhoﬀ stress tensor P, see Eqs. (3.10), (3.11). The quantities σrr and σrz are the corresponding components of the linear stress tensor in the surrounding, external medium: σrr = (λ1 + 2µ1 ) w1,r + λ1 w1 + λ1 u1,x r (4.27)

σrx = µ1 (u1,r + w1,x )

(4.28)

The conditions (4.23)-(4.25) deﬁne the sliding contact, while the longitudinal displacements u and u1 are left free at the interface r = R. The Piola- Kirchhoﬀ tensor coincides with the linear stress tensor for inﬁnitesimally small strains. Note that the coeﬃcients in Prr and Prx depend upon both the second order Lam´ coeﬃcients λ and µ and the e Murnaghan moduli, l, m, n. Hence the tensor P takes into account both the geometrical and material nonlinearities. The linear equations (4.20) and (4.21) are solved together with the boundary conditions (4.23), (4.25), (4.26), assuming that the displacement w at the interface is a given function of x and t, hence w(x, t, R) ≡ W (x, t). Then the linear shear stress σrr at the interface r = R is obtained as a function of W and its derivatives, thus providing the dependence only on the rod characteristics in the right hand side of the b.c. (4.24). The same is valid for the elementary work done by external forces at r = R:

∞

δA = 2π

−∞

σrr δw dx .

(4.29)

Satisfaction of the b.c. on the rod lateral surface yields the relationships between displacements and strains inside the rod, see Sec. 3.2, allowing to derive only one nonlinear equation for long longitudinal waves using Hamilton’s principle (3.29) with the Lagrangian density per unit volume, L=K − Π, with Π, K and δA deﬁned by Eqs. (3.1), (3.6) and (4.29) correspondingly.

23). respectively. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 99 4. (4. r cl 1 1 c2 Ψrr + Ψr − 2 Ψ + (1 − 2 ) Ψθθ = 0. (4.c. (4. (4. (4.34) with α = 1−c2 /c2 . vanishing at inﬁnity due to b.we obtain the following relationships for the Fourier images of normal stresses at the lateral surface r = R : . r u1 = Φθ + Ψr + w1 = Φr − Ψθ . They depend on the density and the Lam´ coeﬃcients. Assuming that the unknown functions u1 .25). r r cτ (4. where c is the phase velocity of the wave.33) Ψrr + ∼ ∼ 1 ∼ 1 ∼ 2 Ψr − 2 Ψ −k β Ψ= 0. c2 = (λ1 + 2µ1 )/ρ1 .25). (4. (4.32) where cl and cτ are the velocities of the bulk longitudinal and shear linear waves in the external medium.2.32) to the Bessel equations : Φrr + ∼ ∼ 1 ∼ 2 Φr −k α Φ= 0.21) will be solved with the boundary conditions (4.26).32)we introduce the Fourier transforms of Φ and Ψ: Φ= −∞ ∼ ∞ Φ exp(−k θ) dθ. hence three possible sets of solutions to the equations (4.2 External stresses on the rod lateral surface First.February 11. Using the boundary conditions (4.26). (4. w1 are Ψ . and c2 = µ1 /ρ1 . (4. r (4.31). and β = 1−c2 /c2 . As we focus attention on travelling waves along the axis of the rod we assume that all variables depend only upon the phase variable θ = x − ct.20). r r (4. e τ l To solve equations (4. the linear problem (4.31).23).34) appear.33).31) (4. (4. The ratios between c.30) then Φ and Ψ satisfy the equation: 1 c2 Φrr + Φr + (1 − 2 )Φθθ = 0. cl and cτ deﬁne τ l the signs of α and β. Ψ= −∞ ∼ ∞ Ψ exp(−k θ) dθ that reduces Eqs.

37) where Ji and Ki (i = 0. Note that the dependence of the strain wave behavior on the velocities of bulk linear waves. for acoustic transverse Love waves propagating in an elastic layer superimposed on an elastic half-space Jeﬀrey and Engelbrecht(1994). namely.February 11. 1) denote the corresponding Bessel functions. Parker and Maugin (1987). and a long wavelength relative to the rod radius R. is known. An interesting case appears when there is balance between (weak) nonlinearity and (weak) dispersion as for . L scales the wavelength along the rod. cl .(4. We shall see in the next section that in the long wave limit the normal stress σrr has one and the same functional form at the lateral surface of the rod in all three cases (4. using conditions on the free lateral surface r = R.37).3 Derivation of strain-displacement relationships inside the rod To solve the nonlinear problem inside the elastic rod. we have to simplify the relationships between longitudinal and shear displacements u and w following the procedure explained in Sec.2. 3.35) < c < cl √ √ √ 2(β − 1) k(1 + β)2 K0 ( αkR) 4k βJ0 ( −βkR) √ √ √ + − R α K1 ( αkR) J1 ( −βkR) (4.36) III) when c > cl µ1 W 1−β ∼ √ √ √ 2(β − 1) k(1 + β)2 J0 ( −αkR) 4k βJ0 ( −βkR) √ √ + √ − R −α J1 ( −αkR) J1 ( −βkR) (4. We search for elastic strain waves with suﬃciently small magnitude B << 1. This depends on the monotonic decay of Ki and the oscillatory decay of Ji when R → ∞. R / L << 1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 100 Ampliﬁcation of Nonlinear Strain Waves in Solids I) when 0 < c < cτ : ∼ σ rr = µ1 W 1−β ∼ II) when cτ ∼ σ rr = √ √ √ 2(β − 1) k(1 + β)2 K0 ( αkR) 4k βK0 ( βkR) √ √ √ + − R α K1 ( αkR) K1 ( βkR) (4. in particular. These relationships are obtained. ∼ σ rr = µ1 W 1−β ∼ 4. the simultaneous absence of the tangential stresses and the continuity of the normal ones. The main diﬀerence in the stress (and strain) ﬁelds in the external medium is how they vanish at inﬁnity. cτ .35).2.

The linear part of longitudinal strain along the rod axis. Cxx .5(2λ + 2m − n) ur w + (λ + 2µ + m) ux ur ) + ε2 (0. (4. 2 (c2 − c2 ) cτ l 4c2 τ (4. (4.February 11.43) with γ = 0.39) 2 (λ + 2µ) wr + (λ − k1 ) w + λ ux + µ ur + ε(µ wx + (λ + 2µ + m) ur wr + 0.37). Then with |kR| << 1 in (4. Then ε = B = R L 2 << 1.40) At the rod lateral surface W ≡ w. We use the scale B R for the displacement w.41) and for c > cl : k1 = 2µ1 [c2 (c2 − c2 ) + 3c2 c2 − 4c4 ] µ1 c2 τ τ l l τ . Wxx ≡ wxx . is wr . is ux . Crr . the linear part of transverse strain. k2 = . for 0 < c < cτ : k1 = − 2µ1 .5(2m − n) w wx + (µ + m) (wx wr + ux wx )) = O(ε3 ) (4. c2 τ (4. by choosing R as a length scale along the rod radius. (4. k2 = while for cτ < c < cl : k1 = 2µ1 (4c2 − c2 ) µ1 c2 τ τ . we have a power series expansion in kR. one gets B L as a scale for the displacement u.(4. Moreover.38) is the smallness parameter of the problem.35) . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 101 a rod with free lateral surface. Then choosing L as a scale along x. k2 = c2 c2 1− c2 c2 + (2 − 2 )2 (γ − log 2) .5772157 Euler’s constant. . It allows to obtain analytically an inverse Fourier transform for σrr = k1 r2 w + k2 r−1 wxx and to write the conditions (4.42) µ1 c2 (γ − log 2) . Similarly. c2 cτ τ (4.24).25) in dimensionless form at the lateral surface r = 1 as: λ + 2µ + m 2 ur + 2 3λ + 6µ + 2l + 4m 2 λ + 2l 2 ε( wr + (λ + 2l) w wr + w + (λ + 2l) ux wr + 2 2 λ + 2l 2 ux − k2 wxx ) + (2l − 2m + n) ux w + (µ + m) ur wx + 2 λ + 2µ + m 2 ε2 wx = O(ε3 ).

45) Q= λ + 2l + 2C(λ + 4l − 2m + n) + 2C 2 (3λ + 3µ + 4l + 2m) 2(k1 − 2(λ + µ)) (4.39). w0 = r C Ux . t). .4 Nonlinear evolution equation for longitudinal strain waves along the rod and its solution Now we can derive the equation for the strain waves along the rod. 2 (4. k1 − 2(λ + µ) (4.44) Substituting (4. .44) into the potential deformation energy density Π (3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 102 Ampliﬁcation of Nonlinear Strain Waves in Solids The unknown functions u.40).46) (4. 4.February 11.48) C 2 Uxx . (4. w will be found in power series of ε: u = u0 + εu1 + ε2 u2 + . substituting (4. . w1 = r3 D Uxxx + r Q Ux . w = w0 + εw1 + ε2 w2 + .44) in (4. we ﬁnd that the plane cross-section hypothesis and Love’s relation are valid in the leading order only: u0 = U (x. First of all. .44) may be obtained in a similar way.50) with a1 = λ + 2µ + 2λC + 2(λ + µ)C 2 .49) The higher order terms in the series (4.2. one can get in dimensionless form that 2 3 Π = a1 Ux + ε a2 r2 Ux Uxxx + a3 Ux + O(ε2 ). .1). and equating to zero all terms of the same order of ε. . (4. 2 .47) λ . but are omitted here being unnecessary to obtain an evolution equation for the strain waves. with C= To order O(ε) we get: u1 = − r2 with coeﬃcients D= λ(λ + 2k2 ) 2(k1 − 2(λ + µ))(2(2λ + 3µ) − k1 ) (4. (4.

due to (4.February 11.41).55) (4. Then the solitary wave solution has the form: v = A m2 cosh−2 (m θ) . and consider the coeﬃcients 0 b2 − b4 depending on c0 only.29) and using Hamilton’s variational principle. Note that the coeﬃcients depend now upon the wave velocity.51) Substituting (4. v = Ux : vtt − b1 vxx − ε b2 vxxtt + b3 vxxxx + b4 (v 2 )xx = 0.53) Equation (4. b4 = . The terms of order O(ε2 ) have been neglected. with A= 6(b10 b2 + b3 ) . c. we obtain the following equation for a longitudinal strain wave. in particular.52). b2 = .54) To leading order the phase velocity is obtained from the equation c2 = b10 (c0 ).50).(4. exact travelling solitary wave solution. when deriving equation (4.56) . 0 (4..52) is nothing but the double dispersive equation (3. with 2(a1 − k1 C 2 ) C(1 + C) .. (4. ρ0 ρ0 b1 = (4. while the coeﬃcient b1 may depend also on c1 as b1 = b10 (c0 ) + ε b11 (c0 . it admits.29) into (3.52) (4. 3 3 3 3 ρ0 2 Ut2 − ε r2 C(Ut Uxxt − CUxt ) + O(ε2 ) 2 For the kinetic energy we have: K= (4. b4 (4.18).43). Therefore we assume c2 = c2 + ε c1 + . 2 a3 = λ + 2µ + λ C + λ C 2 + 2(λ + µ)C 3 + 2Q [λ + 2(λ + µ)C] + 2 1 8 2 4 l + 2C + 4C 2 + C 3 + m − 2C 2 + C 3 + n C 2 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 103 a2 = − λ + 2µ C − λ C 2 + 4λ D + 8(λ + µ) C D.51) and (4. ρ0 2 a2 − 2C(k2 C + 2 k1 D) 3(a3 − k1 C Q) b3 = . c1 ).

43).33 2.13 2. II I. A2 = . rather than for c.68 3.43 6. Eq.78 3.11 or 4.45 4. 4. First of all.58) It appears always higher than the wave velocity in a free rod.08 1.43).09 cl 6.February 11. ρ0 (λ + µ + µ1) ρ0 (λ + µ + µ1) (4.12 2.75 − c03 2. the velocity c0 is obtained from (4.56) provides c4 − B1 c2 + B2 = 0 0 0 (4.12 3.(4.42).59) Finally.56) for all three possible cases (4.1 Phase velocities of waves in a polystyrene rod embedded indiﬀerent media.83 or or or or or or 5.97 2. For the case (4.02 5.11 or 5.11 or 4.41).06 model I I I.41 c01 2. m All velocities are measured in 10−3 sec material Quartz Iron Copper Brass Aluminium Lead cτ 3.15 2. where the wave number k remains a free parameter.01 c02 2. As ε must not exceed the yield point of the elastic material (its usual value is less than 10−3 ) we have to compare with cl and cτ the values obtained for c0 . we have to solve Eq.13 1.56) yields c4 − A1 c2 + A2 = 0 0 0 where A1 = 4µ1 c2 (λ + 2µ) (3λ + 2µ)µ + µ1 (λ + 2µ) + 4µ1 ρ0 c2 τ τ .12 2.2.57) Let us estimate the inﬂuence of the external medium on the solitary wave propagation along the rod.7 4. (4.23 2. II I. For the model (4.32 2. (4.08 2.06 2.57) as c2 = 0 (3λ + 2µ)µ + µ1 (λ + 2µ) .26 2.15 3. for the model (4.41) -(4.5 Inﬂuence of the external medium on the propagation of the strain solitary wave along the rod (4. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 104 Ampliﬁcation of Nonlinear Strain Waves in Solids Table 4.1 or 6.1 or 7.07 2.26 2.60) .06 2.85 4.11 2. ρ0 (λ + µ + µ1) (4.05 2. Eq.02 2.77 5. II - and for the function c1 we get the equation c1 = b11 + 4k 2 (b10 b2 + b3 ).

simultaneously. Indeed.2. Then. respectively.06 model I.06 2.55 or 4.59) and (4.83 c02 2.2 for a lead rod. The . an important diﬀerence appears relative to long nonlinear Rayleigh surface waves in Cartesian coordinates: in our case more than one velocity interval exists where solitary waves may propagate. This result is of importance when generating strain solitary waves in a rod embedded in an external elastic medium. a solitary wave may propagate along a lead rod embedded in a polystyrene external medium. Comparing velocities c0i relative to cτ and cl we can justify the applicability of cases (4. for which two or even all three models of sliding contact allow a solitary wave propagation.2 2.73 or4.12 3. II I I I II.11 2.2.43).39 2. However. strain solitary waves can propagate only with velocities from the intervals around c0i . Thus the balance between nonlinearity and dispersion may be achieved at diﬀerent phase velocities of the strain nonlinear waves.1 and 4. respectively. as it follows from Table 4. Note that the solitary wave is a bulk (density) wave inside the rod and.41) is better for the contact with a polystyrene rod. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 105 Table 4. m All velocities are measured in 10−3 sec material Quartz Iron Copper Brass Aluminium Polystyrene cτ 3.43 6. This is noted by symbols I-III. in the last column of Tables 4.60).1 contains some quantitative estimates for a polystyrene rod and Table 4.85 4.51 2.38 or 1.23 2.7 4.2 Phase velocities of waves in a lead rod embedded in diﬀerentexternal media. respectively. both embedded in diﬀerent external media.(4.08 1.08 2. ρ0 (c2 µ1 − c2 (λ + µ + µ1 )) τ l Table 4. while no solitary wave may propagate when the external medium is lead.81 − − − 1. (4.1 c01 2.February 11.02 5. c02 and c03 denote velocities calculated from Eqs. it is a surface wave for the external medium. Note that there exist pairs of materials.47 or 4. III with B1 = (3λ + 2µ)µc2 + (c2 − c2 )µ1 (λ + 2µ) + 4µ1 ρ0 c4 + c2 c2 ρ0 (λ + µ − 3µ1 ) τ τ τ τ l l ρ0 (c2 µ1 − c2 (λ + µ + µ1 )) τ l B2 = c2 c2 [3µ1 (λ + 2µ) − µ(3λ + 2µ)] − 4µ1 c4 (λ + 2µ) τ l τ .(4. The quantities c01 .III I.01 cl 6.81 c03 7. the model (4.26 2.84 or 2.91 − − − 0. Therefore.26 2.58).41).78 3.II.03 1.

For case I. Thus.57). in a polystyrene rod embedded in external medium is higher than the linear wave velocity for a rod with free lateral surface. Then the nonlinear strain wave propagation is described in each part by its own equation (4. However. 3.2. hence the sign is deﬁned by the sign of the quantity (b10 b2 + b3 )/b4 . 4. 4.52). the nonlinear wave velocity.1 to compute the value of A (4. copper or iron. and initial compression pulses are destroyed like in Fig. the velocity. Therefore the solitary wave (strain!) amplitude (4. On the other hand.February 11. let us consider the inﬂuence on the sign of c1 (4. b11 = 0. Matching . quartz. c. The amplitude is negative for a free lateral surface rod and it remains negative if the external medium is. while for a free lateral surface it is positive.54) may change its sign. brass. Now let us consider the inﬂuence of the type of external medium on the existence of either compression or tensile longitudinal strain localized waves. c. Using the data from Table 4. in particular. 3. while for a free surface rod nonlinear waves propagate faster than linear waves.5.1. Finally. c0 . that for a rod embedded in aluminium an initial pulse with velocity close to c02 may transform only into a tensile solitary wave while an initial pulse with velocity close to c01 evolves to become a compression solitary wave.2. For polystyrene it is. c∗ = E/ρ0 . Therefore. the sign changes if c0 = c02 and the external medium is aluminium. it yields that its sign may change according to the values of the parameters of the material used for the external medium. as it is shown in Fig. for A > 0 only tensile strain solitary waves may appear. Let us consider now the case when the rod lateral surface is partly free along the axis and the other part is subjected to a sliding contact with an external elastic medium. 3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 106 Ampliﬁcation of Nonlinear Strain Waves in Solids main diﬀerence between modes lies in the diﬀerent rate of wave decay in the external medium.54) or a wave train. generally. that follows from the diﬀerent behavior of Bessel’s functions at large values of their arguments. while tensile initial pulses do not become localized and are destroyed by dispersion. of a nonlinear wave in a rod embedded in an external medium is lower than the linear wave velocity. one can anticipate.55) for a polystyrene rod.6 Numerical simulation of unsteady strain wave propagation Numerical simulation of unsteady nonlinear wave processes in elastic rods with free lateral surface shows that for A < 0 only initial compression pulses provide a solitary wave (4. see Figs. negative for all the external media in Table 4. say. On the contrary.4.3 .

5 Cylindrical rod partly embedded into an external elastic medium with sliding. an initial strain solitary wave will be destroyed in the embedded part. if sgnA2 diﬀers from sgnA1 .54) moves from left to right in Fig.5 far from the embedded part.2. Let the initial solitary wave (4. When the initial pulse is . 4.(4.February 11. Assume that for the free surface part ( k1 = 0. while for the embedded one. see Sec.52).62) The wave evolution along the embedded part. (4. Similar to the unsteady processes inside a rod with the free lateral surface. 4. is provided by the continuity of strains and its derivatives. k2 = 0) A = A1 . which is supposed to be undeformed at the initial time. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 107 Fig. Otherwise another solitary wave or a wave train will appear.54) and (4. depends on the ratio between A1 and A2 . m = m1 . A = A2 .61) is satisﬁed by equation (4. It was found in Samsonov (1988) that the mass M conservation in the form d M = 0. 3.55) we get for the mass M1 M1 = 2 A1 m1 . m = m2 . M = dt ∞ v dx −∞ (4. Then using Eqs.

(4.2. hence M2 = 2 A2 m2 . 3. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 108 Ampliﬁcation of Nonlinear Strain Waves in Solids Fig. that only one new solitary wave appears but there is an oscillatory decaying tail. the contribution of the tail to the mass M is negligibly small relative to the solitary wave contribution. not massive enough one can see in Fig. However. 4.6 Ampliﬁcation and reconstruction of a strain solitary wave.February 11.63) .

hence there is focusing of the solitary wave. when A2 > A1 attenuation of the solitary wave .64) Fig. Therefore. according to Eq. On the contrary. if A2 < A1 the amplitude of the solitary wave increases while its width.(4. 4.February 11.7 Attenuation and reconstruction of a strain solitary wave. proportional to m−1 .61) it follows A1 m1 = A2 m2 . (4. decreases. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 109 Comparing M1 and M2 .

8 Delocalization and reconstruction of a strain solitary wave.6 the evolution of a strain tensile solitary wave is shown in a rod. Numerical simulations conﬁrm our theoretical estimates. 4. A1 > A2 > 0. 4. The value of A in the central part II. A2 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 110 Ampliﬁcation of Nonlinear Strain Waves in Solids is provided by the simultaneous decrease of the amplitude and the increase of the wave width. Fig. In the . is positive but smaller than the value of A1 in the surrounding free lateral surface parts I and III. In Fig. having a central part embedded in an external medium.February 11.

is . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 111 embedded part II (Fig. Fig. 4. A2 > 0 and |A1 | < A2 .d). When A2 > A1 > 0. and ﬁnally recovers its initial shape in Fig. In our case the deformations of the wave front and rear are equal.8(a).8(c). Hence.6(c. 4. Fig. One can see in Fig.9 how an initially localized rectangular tensile pulse. Moreover. However. the solitary wave does not loose mass. Therefore an increase in amplitude of the elastic strain solitary wave is possible even in an uniformly elastic rod. 4. hence the possible appearance of cracks or plasticity zones. Again both the reconstruction of the initial wave proﬁle and the damping of its tail are observed in the third part of a rod with free lateral surface. a strain wave is localized again in the third part of a rod with free lateral surface.d).8(b). part III . 4.9.8 how an initial tensile solitary wave.6(d) is less pronounced than the tail in Fig.7(c. the amplitude of the solitary wave generated in such a manner.6(b)) the solitary wave amplitude exceeds the amplitude of the initial solitary wave in Fig. while its width becomes narrower than that of the initial wave.4.9(a). an initial tensile strain solitary wave. may be greater than the magnitude of the initial pulse. Again damping of the tail behind the solitary wave is observed.7(a). Moreover. Again there is no solitary wave selection since the parameters of the recovered wave depends upon that of the original one. is drastically attenuated as soon as it enters the embedded area. M . in agreement with (4. Accordingly. One can see that an oscillatory tail of the solitary wave in Fig. in agreement with our previous results on the unsteady processes occurring for a free surface rod.February 11. 4. 4. is destroyed in the embedded part II. 4. and its amplitude decreases while its width becomes larger. Fig. 4. both compression and tensile initial pulses may produce localized strain solitary waves in a rod partly embedded in an external elastic medium with sliding. This may overtake the yield point inside the elastically deformed rod. 4. Fig. In the case treated here. part III in Fig.7(b). 4. This case is shown in Figs. Consider now the case of diﬀerent signs of Ai and assume that A1 > 0 on both free surface parts. 4. a plateau develops in the tail of the solitary wave in geometrically inhomogeneous rod. 4. hence its original shape is recovered when traversing part III in Fig.6(a).64). the solitary wave remains symmetric on ampliﬁcation at variance with the strain solitary wave ampliﬁcation in a rod with diminishing cross section. Fig. One can see in Fig. Fig. 4.8(d). These diﬀerences could be caused by the absence of mass (and energy) conservation for strain solitary waves in a narrowing (expanding) rod. 4. 4. 4.6(c).10 where A1 < 0.

9(a).February 11. The elastic properties of the rod are chosen such that tensile wave propagation cannot occur in the absence of contact with an external medium. Figs. However. Fig.d). a wave train of solitary waves appears.9(c.9(b). The amplitude of the ﬁrst solitary wave in Fig. when a destroyed strain wave comes to the embedded part. 4. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 112 Ampliﬁcation of Nonlinear Strain Waves in Solids Fig.4.9(d) exceeds the magnitude of the initial rectangular pulse in Fig. 4.9 Generation of a tensile strain solitary wave train in a rod. In the absence of surrounding external medium this rod wave-guide does not . destroyed in the free surface part I. 4. 4.

10. .10 Delocalization of a strain solitary wave in the absence of external medium. 4.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 113 Fig. 4. support tensile solitary wave propagation. and a strain wave is delocalized as shown in Fig.

3. 4.1 Modelling of non-dissipative elastic medium with microstructure The theory of microstructure has been developed recently. Savin et. Nowacki (1975). Eringen (1968). now consideration is restricted by the non-dissipative case. Nowacki (1975). only a few experiments may be mentioned. however. Erbay et. The inﬂuence of the microstructure on the solitary wave propagation and ampliﬁcation is studied. see Engelbrecht and Braun (1998). Recall some basic ideas following Eringen (1968). The problem is solved using the ”pseudo-continuum” Cosserat model and the Le Roux continuum model. it will be studied in Chapter 6. Savin et. At any time the position of a material point of the αth microelement may be . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 114 Ampliﬁcation of Nonlinear Strain Waves in Solids 4. al (1973b). A procedure is developed for derivation of the model equation for long longitudinal strain waves inside the rod. Mindlin (1964).3 Strain solitary waves in an elastic rod with microstructure Classic elastic theory cannot account for phenomena caused by the microstructure of a material. Engelbrecht and Braun (1998). Nowacki (1986a) and references therein. al (1973b). al (1991). there are ﬁndings in the ﬁeld of the nonlinear theory Engelbrecht and Braun (1998). Most of results belong to the linear theory of elasticity. Also the values of the parameters characterizing microstructure. Maugin and Muschik (1994). Waves in elastic wave guides with microstructure were out of considerable investigation. Only a few works are devoted to the nonlinear waves in microstructured non-dissipative media Engelbrecht and Braun (1998). A particular case is a dispersion of strain waves in an elastic medium. Eringen (1968). Potapov and Rodyushkin (2001). Erofeev and Potapov (1993). Strain waves were studied mainly in the linear approximation Eringen (1968). Savin et. al (1973a). The inﬂuence of microstructure may provide dissipative eﬀects Cermelli and Pastrone (1997). Mindlin (1964). Erofeev (2002). are unknown as a rule. Suppose the macroelement of an elastic body contains discrete micromaterial elements. Nowacki (1986a).February 11. The present section refers to the study of nonlinear solitary waves inside cylindrical rod with microstructure following Porubov (2000).

Tensor of the second rank EKL accounts for the microelements motion relative to the center of mass of the macroelement. where Ξ(α) characterizes initial position of a point relative to the center of mass. U(X.K ΦM L .K χkM.K xk. where CKL is the Cauchy-Green macrostrain tensor .K )δkL .K χkL − δKL + χkL χkM ΞM ) dXK dΞL + χkK χkL dΞK dΞL .K + UM. where x is the position vector of the center of mass of the macroelement. The motion of the center of mass depends upon the initial position X and time t. ξ (α) = χK (X. x = x(X.L ΞM ΞN )dXK dXL + 2 (xk.M + UN. EKL is the tensor of a reference distortion.K UM. while tensor of the third rank ΓKLM characterizes relative motion of the microelements of one another. t).L − δKL + 2xk.February 11. Φ(X. ΓKLM = ΦKL. Then the square of the arc length is (ds(α) )2 = dx(α) dx(α) . and the diﬀerence between the squares of arc length in the deformed and undeformed body is (ds(α) )2 − (dS (α) )2 = (xk. 2 = ΦKL + UL. while for ξ (α) the axiom of aﬃne motion is assumed. ΓKLM is the tensor of microdistortion.M .L ) .65). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 115 expressed as x(α) = x + ξ (α) .K = (δLK + UL.65) (α) where δKL is the Kronecker delta. t). CKL = EKL 1 (UK. xk.L + UL.K χkN.K + UM.L ΞM + χkM. χkK = (δLK + ΦLK )δkL Then three tensors characterizing the behavior of microstructured medium follow from (4. (4. t) ΞK . Let us introduce vector of macrodisplacements. t) and tensor of microdisplacements.K ΦN L. . ξ (α) is the position of a point in the microelement relative to the center of mass.

more precisely upon the invariants of them.67) accounts for the pseudo-continuum Cosserat model when micro-rotation vector Φ coincides with the macrorotation vector.67) where εKLM is the alternating tensor. n) are the third order elastic e moduli.K ΦL. Savin et. JKN is the inertia tensor. That is why the models were developed based on the additional assumption on a relationship between U and Φ. 2.K + βΦK. Nowacki (1986a). 2 and only linear part of EKL coincides with those of CKL .L + 2 3 ηΦK. Ip . ΦM = 1 εM LK UK. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 116 Ampliﬁcation of Nonlinear Strain Waves in Solids The density of the potential energy Π should be the function of these tensors. Savin et. and the Murnaghan model is valid for the macro-motion. The bulk density of the kinetic energy has the form Mindlin (1964) K= 1 2 ρ0 UM.L − UM. Assume the microstructure is suﬃciently weak to be considered in the linear approximation Nowacki (1975).K UL. Π = Π(CKL .L . ΓKLM ). (4.February 11.68) where λ and µ are the Lam´ coeﬃcients. the basic invariants of the third and higher rank tensors have not been studied. or the Murnaghan moduli. Tensor EKL has the form EKL = 1 (UK. Nowacki (1986a). EKL . Elastic media with central symmetry posses simpler representation. 2 (4.K + UM.L ΦK.t + JKN ΦKM. (l. Moreover. see Eq. One of the main problem is to deﬁne integrity basis of three tensors CKL . One of them is the pseudocontinuum Cosserat model . M . In this case the density of the potential energy my be either Π = Π(CKL .t .M ) . 2 (4. m. Eringen and Suhubi (1964). . p = 1. 3 are the invariants of the tensor C. Then the density of the potential energy may be written as Π= l + 2m 3 λ + 2µ 2 I1 − 2µI2 + I1 − 2mI1 I2 + nI3 + 2µM 2 (ΦK. EKL . The ﬁrst relationship represents the classic Cosserat model when only rotations of solid microelements are possible. ΓKLM Spencer (1971). al (1973b).L ΦL. The last expression in (4.L + UL. al (1973b). According to it ΦKL = −εKLM ΦM . η and β are the microstructure constants.2). ΓKLM ) or Π = Π(CKL .66) where ρ0 is macrodensity of the elastic material. ΦK.L ) Nowacki (1975).K UM. JKN = J ∗ δKN .L ).t ΦN M.(3.

KM .71) The governing equations together with the boundary conditions are obtained using the Hamilton variation principle (3. see Erofeev and Potapov (1993). Savin et. r is the coordinate along the rod radius. i = 1 − 5.2 Nonlinear waves in a rod with pseudo-continuum Cosserat microstructure Let us consider the propagation of a longitudinal strain wave in an isotropic cylindrical compressible nonlinearly elastic rod. When microstructure is weak and may be considered in the linear approximation the linear part of EKL is zero tensor. r. ΓKLM = −UL. ϕ [0. −∞ < x < ∞.x = wxx − urx . al (1973a). Mindlin (1964).70) 2 while nonzero components of the rotation tensor ΦK.3. 3. Crr = wr + (u2 + wr ). Sometimes it is referred to as the Le Roux continuum Erofeev (2002). where the Lagrangian .69) where ai .5). 0).L are Φϕ. In this case Π = Π(CKL . 2π]. Then nonzero components of the macrostrain tensor C are 1 1 w w2 2 2 Cxx = ux + (u2 + wx ). ϕ) where x is directed along the axis of the rod. 0 ≤ r ≤ R. (4. see Fig. Neglecting torsions the displacement vector is U = (u. Cϕϕ = + 2 . w. ΓKLM ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 117 Another simpliﬁed microstructure model was used by some authors. Assume again the Murnaghan model for the macro part of the energy density and use the linear Mindlin’s model Mindlin (1964) for its micro part one can obtain Π = λ + 2µ 2 l + 2m 3 I1 − 2µI2 + I1 − 2mI1 I2 + nI3 + a1 ΓKKM ΓM LL + 2 3 a2 ΓKLL ΓKM M + a3 ΓKKM ΓLLM + a4 Γ2 KLM + a5 ΓKLM ΓM LK . ϕ is a polar angle.1. It means that there is no diﬀerence between deformation of elastic microelement and elastic macrostructure. According to it ΦKL = −UL. Φϕ. (4. x r 2 2 r 2r 1 Crx = Cxr = (ur + wx + ux ur + wx wr ) . 4. (4. We take cylindrical Lagrangian coordinates (x. are the constant microstructure parameters.February 11.r = wxr − urr .K .

Prr = 0 . Then the procedure from Sec.72) (4. The following boundary conditions (b. Subsequent simpliﬁcation is caused by the consideration of only long elastic waves with the ratio between the rod radius R and typical wavelength L is R/L 1. . The typical elastic strain magnitude B is also small.75) 2 w + 2r Prx = µ (ur + wx ) + (λ + 2µ + m) ur wr + (2λ + 2m − n) ur (λ + 2µ + m) ux ur + (µ + m) ux wx 1 ∗ J (urtt − wxtt )].73).13).74) where the components Prr .71) being taken into account: Prr = (λ + 2µ) wr + λ w λ + 2µ + m 2 w + λ ux + ur + (λ + 2l) wr + r 2 r λ + 2l w2 w + (λ + 2l) ux wr + (2l − 2m + n) ux + 2 r2 r λ + 2l 2 λ + 2µ + m 2 ux + wx + (µ + m) ur wx + 2 2 3λ + 6µ + 2l + 4m 2 wr + 4µM 2 (urrx − wxxr ). Accordingly.c.76) Exception of torsions provides transformation of the initial 3D problem into a 2D one. at r = R.) are imposed: w → 0. on the lateral surface of the rod (4. B 1.72).73) (4.2.68). Prx of the modiﬁed Piola . (4.70) and (4. the longitudinal and shear displacement in dimensional form are sought in the form (3.February 11.2 is applied to ﬁnd the relationships between displacement vector components satisfying b.(4.12). 2 w 2m − n wx + (µ + m) wx wr + 2 r 1 2 + 4µM [wxxx − uxxr + (r(wxr − urr ))r − r (4. An additional parameter γ = M 2 /R2 is introduced to characterize the microstructure contribution. at r → 0. (4.5) with (4. (4.74) as well as the condition for w (4. Prx = 0. at r = R. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 118 Ampliﬁcation of Nonlinear Strain Waves in Solids density per unit volume.Kirchhoﬀ stress tensor P are deﬁned from (3.66). (4.68) correspondingly.c.66) (4. (4. 3. L=K − Π. (3. with K and Π deﬁned by Eqs.

(4. (4. 2ρ0 1 − 4γ Hence the microstructure aﬀects only dispersion in Eq. < c2 1 − ν 1 − 4γ ∗ .74).79) where α1 = c2 = E/ρ0 .(4.13) into the b.73). (3. α3 = ν(1 − ν)R2 /2. 1< V2 1 1 + 4γ .13) for i > 3 may be found in the same way.77) w = −νrUx − ν [ν + 4γ(2 + ν)] r3 Uxxx − 2(3 − 2ν)(1 − 4γ) (1 − 2ν) ν 2 + l(1 − 2ν)2 + 2m(1 + ν) − nν rUx .72) and in the linear parts of b. (4.77) is v= 3E β V2 −1 c2 ∗ cosh−2 (k (x − V t)). (4. (4. E is the Young modulus. α4 = νER2 1 + 4γ . α2 = β/(2ρ0 ).80) where V is a free parameter while the wave number k is deﬁned by k2 = ρ0 (V 2 − c2 ) ∗ 2νER2 1+4γ 1−4γ (1−ν)V 2 c2 ∗ − .5).(4. (4. The solitary wave solution of Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Ampliﬁcation of strain waves in absence of external energy inﬂux 119 Substituting the linear parts uL and wL (3. however. Using these results the nonlinear parts uN L .(3.81) Therefore the contribution of the microstructure results in the widening of the permitted solitary wave velocities. t) + νr2 1 + 4γ Uxx .(4.c. and equating to zero terms at equal powers of r one obtains uk and wk . Substituting Eqs. β = 3E + 2l(1 − 2ν)3 + 4m(1 + ∗ ν)2 (1 − 2ν) + 6nν 2 .c.78) (1 + ν) 2 E where ν is the Poisson ratio. (4. 2 1 − 4γ (4.77).February 11. wN L are similarly obtained from the full b.c. v = Ux . (4.77).12). (3.78) into Eq. obey a double dispersive nonlinear equation: vtt − α1 vxx − α2 ( v 2 )xx + α3 vxxtt − α4 vxxxx = 0. We get u = U (x. Other terms from the series (3. they are omitted here because of no inﬂuence on the ﬁnal model equation for the strain waves. and using Hamilton’s principle we obtain that longitudinal strains.12).

The nonzero components of the tensor ΓKLM are Γxxx = −uxx . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 120 Ampliﬁcation of Nonlinear Strain Waves in Solids Also the characteristic width of the solitary wave proportional to 1/k becomes larger relative to the wave width in pure elastic case. 2 1−N (4. Γrxr = −urr .82) r r w + 2r Prx = µ (ur + wx ) + (λ + 2µ + m) ur wr + (2λ + 2m − n) ur (λ + 2µ + m) ux ur + (µ + m) ux wx 1 2a2 (r(urr ))r .84) . Γxxr = Γrxx = −uxr . Then the type of the solitary wave (compression/tensile) is deﬁned by the sign of the nonlinearity parameter β like in the case without microstructure.3. γ = 0. (4.c.83) Then the approximations for the components of the displacement vector have the form u = U (x.3. Γxrr = Γrrx = −wxr .February 11. (4. Γrrr = −wrr . 3. (4. r 2m − n w wx + (µ + m) wx wr + 2 r + 2J ∗ urtt − a1 wxrr − 2(a1 + 2a2 )uxxr − (4.74) are satisﬁed for the strain tensor components Prr = (λ + 2µ) wr + λ( w λ + 2µ + m 2 w + ux ) + ur + (λ + 2l) wr + r 2 r λ + 2l w2 w + + (λ + 2l) ux wr + (2l − 2m + n) ux + 2 r2 r λ + 2l 2 λ + 2µ + m 2 ux + wx + (µ + m) ur wx + 2 2 3λ + 6µ + 2l + 4m 2 wr + 2J ∗ (2uxtt + wrtt ) − 2a1 uxxx − 2 1 1 2(a1 + 2a2 )wxxr − 2(a1 + a2 ) (r(wrr ))r − a1 (r(uxr ))r . We consider γ to be rather small due to the experimental data from Savin et. t) + 1 νr2 Uxx . 4.3 Nonlinear waves in a rod with Le Roux continuum microstructure The procedure of obtaining the governing equations is similar to those used in previous subsection. The b. see Sec. Γxrx = −wxr .73). al (1973b).

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Ampliﬁcation of strain waves in absence of external energy inﬂux

121

w = −νrUx −

4J ∗ (2 − ν)(1 + ν)(1 − 2ν) 3 r Uxtt − E(3 − 2ν)R2 ν 2 − (1 − 2ν)(1 − N )(G(1 − ν) − 2νN ) 3 r Uxxx − 2(3 − 2ν)(1 − N ) ν (1 − 2ν) (1 + ν) + l(1 − 2ν)2 + 2m(1 + ν) − nν 2 E

2 rUx , (4.85)

where G = 2a1 /µR2 , N = 2a2 /µR2 . Like in previous section the governing equation for longitudinal strain v = Ux is the double dispersive equation (4.79) whose coeﬃcients are deﬁned now as α1 = c2 , α2 = ∗ β νR2 ν 2 R2 νc2 R2 ∗ , α3 = − + 2J ∗ ν(2 − ν), α4 = , 2ρ0 2(1 − N ) 2 2(1 − N ) 3E β V2 −1 c2 ∗

while the solitary wave solution has the form v= cosh−2 (k (x − V t)), (4.86)

**where V is a free parameter, and the wave number k is deﬁned by k2 = 2νER2 [c2 ∗ (1 − N )ρ0 (V 2 − c2 ) ∗ . − V (1 − ν(1 − N ) + 4J ∗ (1 − N )(2 − ν)/R2 )]
**

2

(4.87)

Physically reasonable case corresponds to rather small N , N < 1. Then the inﬂuence of the microstructure yields an alteration of the permitted solitary wave velocities interval, 1< V2 1 < . c2 1 − ν(1 − N ) + 4J ∗ (1 − N )(2 − ν)/R2 ∗

The widening or narrowing of the interval depends upon the relationship between N and the parameter of microinertia J ∗ . Again the type of the solitary wave is governed by the sign of the nonlinearity parameter β. At the same time the characteristic width of the solitary wave proportional to 1/k turns out smaller than the wave width in a pure macroelastic case, N = 0, J ∗ = 0. 4.3.4 Concluding remarks

It is to be noted that the assumption of the linear contribution of the microstructure is correct since its nonlinear contribution, being weaker, may provide alterations only in the neglected higher-order nonlinear and dispersion terms in the governing equation both the Cosserat and the Le

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

122

Ampliﬁcation of Nonlinear Strain Waves in Solids

Roux models. Hence we don’t need in the additional nonlinear terms in the density of the potential energy Π thus avoiding the additional unknown parameters (like Murnaghan’s third order moduli) describing the nonlinear contribution of the microstructure Eringen and Suhubi (1964); Erofeev (2002). The alterations of the amplitude and the wave width, caused by the microstructure, have been found in both case under study. The important result is in the opposite changing of the wave width which gives a possibility to distinguish the Cosserat and the Le Roux models in possible experiments. The dispersion caused by the microstructure may be observed experimentally, and numerical data on microstructure parameters my be obtained Savin et. al (1973b). In experiments on the solitary waves propagation, see Sec. 3.4, the amplitude and the width of the wave may be measured. Therefore expressions (4.80), (4.81) provide possible estimation of the parameter M in the pseudo-continuum Cosserat model. In case of the LeRoux continuum there is an extra parameter J ∗ , see Eqs.(4.86), (4.87), and parameters N and J ∗ cannot be estimated separately. The microstructure and the surrounding medium provide similar deviations in the governing double dispersive equation. Hence the analysis in Sec. 4.2.6 may be used if we consider a rod only part of which contains the microstructure. Then the ampliﬁcation/attenuation (but not a selection) of the strain solitary wave occurs similar to that shown in Figs. 4.6-4.10.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Chapter 5

Inﬂuence of dissipative (active) external medium

Now we study the role played by dissipation/energy inﬂux often present in a realistic case. Dissipative/active eﬀects may be caused by internal features of the elastic material, hence, an irreversible part should be included into the stress tensor in addition to the reversible one depending only upon the density of the Helmholtz energy. Accordingly, the governing equations for nonlinear strains will contain dissipative/active terms. Dissipation/energy inﬂux may also occur in an elastic wave guide through phenomena occurring at or through its lateral surface, and this case is considered further in this Chapter. Presence of external medium makes a problem more complicated. However, the Hamiltonian formalism described in Chapter 3 may be applied since a wave guide remains pure elastic, all dissipative/active factors come through the elementary work done by the external forces, and therefore Eq.(3.29) may be used. 5.1 Contact problems: various approaches

External medium aﬀects the lateral surface of a wave guide through the normal and tangential stresses. In some cases only normal stresses may act like in the slippage contact. Various contact problems are widely considered, see, e.g., Galin (1980); Goryacheva (1998); Goryacheva (2001); H¨hner a and Spencer (1998); Johnson (1985); Kalker et. al (1997); Kerr (1964); Nikitin (1998). Both elastic and viscoelastic interactions are studied their in the linear approximation. Main attention is paid to the static loading and to the interaction forces caused by the relative movement of the contacting bodies. Of special interest if the contact with friction Goryacheva (1998); Goryacheva (2001); H¨hner and Spencer (1998); Kalker et. al (1997); a Nikitin (1998); Stefa´ski et. al (2000) . Various generalizations of the n

123

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

124

Ampliﬁcation of Nonlinear Strain Waves in Solids

Coulomb-Amonton law (dependence on the sliding velocity and the apparent area of contact) are studied Elmer (1997); Goryacheva (2001); H¨hner and Spencer (1998); Nikitin (1998); P¨schel and Herrmann (1993); a o Stefa´ski et. al (2000). However, it was noted in H¨hner and Spencer n a (1998); Nikitin (1998); Stefa´ski et. al (2000) that the classical Coulombn Amonton approach is suﬃcient when the relative velocity of the sliding bodies is small. Sometimes the problem of the interaction with an external medium may be solved directly. It means that we formulate the equation for a wave guide and the medium and impose the corresponding continuity conditions on the lateral surface of the wave guide. An example may be found in Sec. 4.2 where dissipationless sliding contact with an elastic external medium is considered. The diﬃculties of elastic contact stress theory may arise because the displacement at any point in the contact surface depends upon the distribution of pressure throughout the whole contact Johnson (1985). In this case the solution of an integral equation for the pressure is required. Another problem arises when an external medium is not elastic, and its behavior cannot be described by the equations of elasticity. The diﬃculties mentioned above may be avoided if the response of a wave guide is more interesting then the displacements and stresses distribution in the external medium. In this case the problem reduces to a development of a relatively simple foundation models to account for an inﬂuence of the external medium in terms of the wave guide displacements and/or strains at the lateral surface. Variety of the foundation models are collected in Kerr (1964). The models are designed replacing an external medium with interacting spring and dissipative elements. In particular, when only springs are considered and their shear interactions are assumed the so-called Winkler-Pasternak model holds Kerr (1964); Pasternak (1954); Winkler (1867). According to it the pressure p is expressed through the shear displacement w p = kw − G

2

w,

where 2 is the Laplace operator in x and y, k and G are the constant foundation moduli. One can see its similarity with the response of the external medium in case of the sliding contact, see Sec. 4.2.3. Hence foundation models are physically reasonable since they correspond to the results obtained from the contact problem solutions. It was Kerr (1964) who developed a viscoelastic foundation model.

2π]. Assuming that torsions can be neglected. −∞ < x < ∞. a Nikitin (1998). k. k is the stiﬀness coeﬃcient of the medium. Here the foundation model is proposed in Sec. we consider a more general case with the coeﬃcients of either sign. w. Classic case of a dissipative action through the tangential stresses corresponds to the dry friction contact H¨hner and Spencer (1998). ϕ is a polar angle. Most of results were ﬁrst obtained by Porubov and Velarde (2000) 1 . r is the radial coordinate.1 Let us consider the propagation of a longitudinal strain wave in an isotropic cylindrical elastic rod embedded in an external medium subjected to Kerr’s viscoelastic contact model. Now we only mention it provides the inﬂuence on a wave guide only through the normal stresses. r. 0). He assumed Newton’s law for the viscous behavior thus including a dissipative elements besides springs. 5.February 11. are positive and constant in framework of the Kerr model. ∗ According to Kerr (1964) the external medium yields a normal stress Prr on the lateral surface of the rod r = R : k ∗ Prr = − w − η wt + χ r2 wxxt .2 Evolution of bell-shaped solitary waves in presence of active/dissipative external medium Formulation of the problem 5. η is the viscocompressibility coeﬃcient of the external medium. η and χ. where the Lagrangian density 1 Reprinted with permission from Elsevier Science . The evolution of nonlinear waves is obtained in the reference conﬁguration using Hamilton’s principle Eq.4 when the external medium aﬀects a lateral surface of an elastic rod both by the normal and the tangential stresses but dissipative (active) inﬂuence is provided by the tangential stresses. This model will be studied in the following two sections where its mathematical expression is presented. 5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 125 Based on the results of footing load tests performed on a snow base he proposed a viscoelastic model for the interaction between an elastic body and the external snow (or permafrost) medium.(3.1) where t is a time. r (5. All three coeﬃcients. χ is the viscosity coeﬃcient of the external medium.29)). ϕ) where x is directed along the axis of the rod. 0 ≤ r ≤ R. ϕ [0. then the displacement vector is V = (u.2. however. We take cylindrical Lagrangian coordinates (x.

δA.c.7) where the explicit forms for the coeﬃcients are given by a2 = λ λη . (3. (5.1) the elementary work. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 126 Ampliﬁcation of Nonlinear Strain Waves in Solids per unit volume.1) for the deformation energy. 5. at r = R.(5. (ii) a characteristic strain wave length L is greater then the rod radius R.6) (5.3). Due to (5.2 Dissipation modiﬁed double dispersive equation For the longitudinal and shear displacements inside the rod we assume that: (i) pure elastic waves have strain magnitude B << 1. 2(2λ + 3µ) + k . at r → 0.c. Let us obtain the approximations satisfying b. ∗ Prr = Prr . where the components Prr . is deﬁned as before in the book. Substituting them into the boundary conditions and following the approach from Sec.5) Prx = 0. is: ∞ δA = 2π −∞ ∗ Prr δw dx . w = b1 r Ux + b2 r2 Uxt + r3 (b31 Uxxx + b32 Uxtt ) + 2 r4 (b41 Uxxxt + b42 Uxttt ) + B1 r Ux + B2 r2 Ux Uxt .5).) are: w → 0. Prx of the Piola . (3. L.2 we get u = U + a2 r2 Uxx + a3 r3 Uxxt . w are expanded in a power series (3.11). Again we choose Murnaghan’s approximation (3.2. at r = R.13). b31 = − λ a2 .Kirchhoﬀ stress tensor P are deﬁned by Eqs. (5. According to the procedure developed in Chapter 3 the unknown functions u. (5.4) (5. (5. a3 = − . b2 = −3 a3 . 3.3) (5. 2(2(λ + µ) + k) 3(2(λ + µ) + k)(3λ + 4µ + k) b1 = −2 a2 .12).February 11.10). R/L << 1. (3.2) The boundary conditions (b.

Neglecting cubic nonlinear terms we have to neglect simultaneously the ”corresponding” higher order linear derivative terms.8) α3 = . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 127 b32 = 3η a3 2χ a2 + η b31 + λ a3 η b32 . b2 . The 3 ”largest” of the cubic terms is rUx ∼ RB 3 ∼ R(R/L)6 . It was shown in that solitary waves appear as a result of the balance between nonlinearity and dispersion when B is of order (R/L)2 . b41 = − . b32 are always negative.. obey a dissipation modiﬁed double dispersive nonlinear equation (DMDDE): vtt − α1 vxx − α2 vxxt − α3 ( v 2 )xx − α4 vxxxx + α5 vxxtt − α6 (v 2 )xxt − α7 vxxxxt + α8 vxxttt = 0. the ”corresponding” comparable linear term is r5 Uxxxxx .6). α2 = . (5. Then. while a3 . ρ0 ρ0 3 [λ ρ0 1 + 2B1 − 2a2 (1 + 4B1 ) + 4a2 (1 − 4a2 ) + 2 2 µ 1 − 8a2 B1 − 16a3 − 4ka2 B1 + 2 l 16 2 1 − 4a2 (1 − 4a2 + a2 ) + m(1 − 4a2 (3 + 4a2 )) + 4na2 ]. 2(2(λ + µ) + k) B1 = − B2 = − 2η B1 + 12a2 a3 [3(λ + µ) + 2(2l + m)] − 3 a3 (λ + 4l − 2m + n) . . are α1 = 8a2 (2(λ + µ) + k) − 8λ a2 + λ + 2µ 8R η a2 2 2 . Similar terms with spatio-temporal mixed derivatives are of the same order. 3λ + 4µ + k Note that a2 . i = 1.. 8. v = Ux .7). see (3. The explicit forms of the coeﬃcients αi . 2 2 3 3 2 3 (5. b42 are always positive. 2(2λ + 3µ) + k 5λ + 8µ + k 5λ + 8µ + k 2 b1 (λ + 4l − 2m + n) + 8a2 [3(λ + µ) + 2(2l + m)] + λ + 2l 2 . Due to (3.4). omitting nonlinear terms of order three and higher. Using Hamilton’s principle we obtain that longitudinal strains.1) we have to truncate the series (5. while quadratic terms are r3 Ux Uxxx and r3 (Uxx )2 . b42 = − . b1 . Higher order terms may be added if terms like 4 I1 are taken into account..February 11. hence. b31 . while the remaining coeﬃcients may have either sign.

Higher order terms. α6 = − α8 = 4R 3 η 4a2 b32 + 9a2 .8) the following ODE: (V 2 −α1 )v+α2 V v −α3 v 2 −(α4 −α5 V 2 )v +α6 V (v 2 ) +V (α7 −α8 V 2 )v + P + P1 θ = 0.2.(5.8) in accordance with the early given arguments. Its most general exact solution expressed in terms of the Weierstrass elliptic function has been obtained in Sec. 5. α7 = − [η b31 + χ a2 ] .10) . Here we restrict consideration to its bell-shaped solitary wave limit. (5. 2.8) may take either sign depending upon the material properties of the rod and the values of k. 3 ρ0 The coeﬃcients in Eq. η and χ. It is seen that Eq. hence.shaped solitary waves when v → const at | θ |→ ∞.1. ( v 3 )xx .(5. have been neglected when deriving Eq.3. B = k2 + 2 α6 3 2α6 (5.(5. we assume that v = v(θ) and obtain from (5. We are interested in bell.(1.February 11. v6x etc. θ = x − V t. P and P1 are constants of integration. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 128 Ampliﬁcation of Nonlinear Strain Waves in Solids α4 = R2 [(λ + 2µ − 4λ a2 ) a2 + 4(λ − 2(2λ + 2µ + k) a2 ) b31 ] .shaped or kink. . . denotes ∂/∂θ. P1 = 0. ρ0 ρ0 α5 = 16(2λ + 2µ + k)a2 b32 ]. ρ0 R2 [2ρ0 a2 (1 − 2a2 ) − 8λ b32 + 9(9λ + 10µ + 4k)a2 + 3 2ρ0 8R a2 ηB1 8R3 a2 . with A= 6(α6 α7 − α1 α6 α8 + α2 α3 α8 ) α2 A . Nepomnyashchy and Velarde (1994).. v = A k 2 cosh−2 (k θ) − B.3 Exact solitary wave solutions of DMDDE In the moving frame.13) obtained in the study of long waves in surface tension gradient-driven ﬂows Nekorkin and Velarde (1994).9) where a dash.9) coincides with the ODE reduction of the DMKdV Eq.

10).14) 4(α7 − α8 V 2 ) thus implying yet another restriction on the coeﬃcients. (5.15) Eq. when a solution decaying at inﬁnity is considered. with 6(α7 − α8 V 2 ) 6(α3 (α7 − α8 V 2 ) − α6 (α4 − α5 V 2 )) .18) (5. then α2 k2 = − .17) 2α3 A= There are two possibilities for k and V . In particular. (5. D= . η and χ of the external medium determine the existence of exact bellshaped strain wave solution (5. (5. the propagation of either compression or tensile strain solitary waves.16) .8) also admits kink-type solutions of the form v = A k 2 cosh−2 (kθ) + D k tanh(kθ) + C. Eq.(5.11). A. α8 α1 − α2 α3 α6 − α7 > 0. α6 5α6 V 2 − α1 C= .13) The ODE reduction of the Boussinesq (or KdV) equation appears in brackets.12) Thus. On the one hand k= 1 α3 α4 − α5 V 2 − .11) α6 (α4 α6 − α3 α7 ) = (α1 α6 − α2 α3 )(α5 α6 − α3 α8 ).10) as well as the sign of its amplitude. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 129 V 2 = α1 − provided α2 α3 . α6 (5. hence. With Eqs.12) taken into account. However. (5. The relationship (5.12) provides simultaneously a balance between nonlinearity ( v 2 ) and dispersion (v ) and another balance between nonlinear active/dissipative ((v 2 ) ) and linear active/dissipative (v ) terms . the elastic features of the rod and the values of the parameters k. (5. (5. as it is prescribed by the behaviour as | θ |→ ∞.February 11.9) may be written as (V ∂ α3 α6 (α4 − α5 V 2 ) α6 P − ) α2 v + α6 v 2 + v − ∂θ α6 α3 α3 = 0. 10 V (α7 − α8 V 2 ) α6 (5. (5. here the wavenumber k is not a free parameter of the solution (5.(5.

Eq.8) are small relative to the other coeﬃcients. 6α3 α8 V 3 + α6 α7 V 2 − 6α3 α7 V + α4 α6 = 0.(5. α7 = ε β7 .21) Therefore.16) with prescribed parameters. with θx = 1.22) we get that (V 2 − α1 )vθ − α3 ( v 2 )θ − (α4 − α5 V 2 ) vθθθ = ε(2V [vT + α5 vθθT ] + VT [v + α5 vθθ ] − V (β7 − β8 V 2 ) vθθ ]) + O(ε2 ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 130 Ampliﬁcation of Nonlinear Strain Waves in Solids where V is a solution of the cubic equation. we have in both cases the kink-shaped solution (5.(5. T ). The loading or unloading of the rod depends upon the sign of D.(5. (5.4 Bell-shaped solitary wave ampliﬁcation and selection Let us consider the case when the inﬂuence of the external medium is weak.2. v = v(θ.22) is the double dispersive equation. and all coeﬃcients of the active/dissipative terms in DMDDE (5. α2 = ε β2 . An initial prestressed state of the rod may transform into a diﬀerent one as the wave passes. Thus at ε = 0. (5.e. (5. hence on the elastic features of the rod and on the parameter values of the external medium. ε << 1. 5. For nonzero ε we assume that a solution of Eq.February 11. θt = −V (T ). k2 = 1 4 2 6α3 (V 2 − α1 )α6 + α2 α3 V − 2 α6 α3 V (α7 − α8 V 2 ) .20) where V is a solution of another cubic equation. ∂2 [β2 v + β6 v 2 + ∂θ2 (5. α6 = ε β6 ..8) may be written as vtt − α1 vxx − α3 ( v 2 )xx − α4 vxxxx + α5 vxxtt = ε β2 vxxt + β6 (v 2 )xxt + β7 vxxxxt − β8 vxxttt . i. Then Eq. (5.22) admits an exact bell-shaped solitary wave solution.(5. 3 2 3 3 2 2 2 α3 α8 V 3 − α6 (α6 − α3 α5 )V 2 − (α3 α7 + α2 α3 α6 )V + α6 (α1 α6 + α3 α4 ) = 0.22) is a function of the phase variable θ and the slow time T .22) We see that the left hand side of Eq. α8 = ε β8 . T = ε t.19) On the other hand. Then from (5.23) .

23) is sought using the method explained in Sec.29) Only the interval (5. if α4 − α1 α5 < 0.23) in the leading order we have 2 (V 2 − α1 )v0. η > η ∗ .T + α5 v0. if α4 − α1 α5 > 0.(5..θθ ]− V ∂2 2 β2 v0 + β6 v0 + (β7 − β8 V 2 ) v0. α5 (5.2 in the form v = v0 + εv1 + .24) (5.θ − α3 ( v0 )θ − (α4 − α5 V 2 ) v0.θθT ] + VT [v0 + α5 v0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 131 The solution of Eq.. s = V 2 may lie either inside the interval α4 α1 < s < . 2 α3 4(α4 − α5 V 2 ) (5.(5. α5 or in α4 < s < α1 . ∂θ2 (5. 4k + 9λ + 10µ The correction v1 (5.2. The interval (5.26) Accordingly.30) The operator M acting on the function v1 in Eq. with √ ρ0 (k + 3λ + 4µ) η∗ = √ .28) is acceptable for a free rod with positive Poisson ratio.θ − 2α3 (v0 v1 )θ − (α4 − α5 V 2 ) v1.24) obeys the inhomogeneous linear equation (V 2 − α1 )v1. .25) has the form v0 = A(T ) cosh−2 (k(T ) θ). where F is F = 2V [v0.28) (5.29) exists if the viscocompressibility coeﬃcient is greater than a given value. k = . 2.θθ .θθθ = 0 (5.February 11.25) The exact solitary wave solution of Eq. Substituting (5.θθθ = F.30) is adjoint to the operator 3 M A = (α1 − V 2 )∂θ + 2α3 v0 ∂θ + (α4 − α5 V 2 ) ∂θ .24) into (5.27) (5.(5. with A= 3(V 2 − α1 ) 2 V 2 − α1 .

α4 /α5 ) (sQ . (5. Important features of the behavior of s may be established analyzing (5.43) for Eq. ∞) s0 (α1 . sQ ) (sQ . α4 /α5 ) Then using Eq.32) (5. α4 /α5 ) permitted for s. sQ ) (α1 .February 11. s2q ) (sQ .33) q1 = 12α5 β6 − 5α3 β8 . ∞) (s1q . α4 /α5 ) (s1q . α4 /α5 ) (sQ . sQ ) (α1 . with Q3 (s) = α1 α4 (5α4 − α1 α5 ) + 2α4 (11α1 α5 − 10α4 ) s− 2 3α5 (17α4 − 5α1 α5 ) s2 − 30α5 s3 . α1 ) (α1 .35) . and Q3 always changes its sign inside the interval (α1 . s2q ) (sQ . s2q (α4 /α5 . α4 /α5 ) (α4 /α5 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 132 Table 5.25) one can obtain the solvability condition (2. α4 /α5 ) (α1 . s2q ) (0. sQ ) (sQ .32) without integration.34) q3 = α3 (7β2 α4 + 5α1 β7 ) − 12α1 α4 β6 . The most interesting evolution of s is realized when s tends to the ﬁnite constant value s∗ as T → ∞.(5. sQ ) (sQ . The values of s∗ are the solutions of equation q1 s2 + q2 s + q3 = 0. (5. (5. sQ ) (sQ . Note that Q3 (α1 ) = −15α1 (α4 − α1 α5 )2 while Q3 (α4 /α5 ) = α4 /α5 (α4 − α1 α5 )2 . sQ ) (s1q . sT Q3 (s) = s (s − α1 )2 (q1 s2 + q2 s + q3 ).30). s2q ) (α1 .1 q1 >0 >0 >0 >0 >0 >0 <0 <0 <0 <0 s1q Ampliﬁcation of Nonlinear Strain Waves in Solids Selection of solitary wave velocity for the case α4 − α1 α5 > 0. α4 /α5 ) (α1 . s2q ) (sQ . sQ ) (s1q . q2 = 12β6 (α4 − α1 α5 ) + α3 (7β2 α5 − 5β7 + 5α1 β8 ). ∞ v0 F dθ = 0. α4 /α5 ) (s1q . −∞ (5. α4 /α5 ) (sQ . s2q ) (α1 .(5. α4 /α5 ) s∗ s1q s1q s1q s2q s2q s2q s2q s2q s1q s1q (α1 .31) which yields the equation for the function s. sQ ) (α1 . α1 ) (0. sQ ) (s1q .

1.32) may be integrated in the general case giving the implicit dependence of s on T . Moreover.745.588. Therefore. s2q = 1. We consider here in details only the cases when the velocity tends to one or another root of Eq.26) coincides with the exact solution (5. The same analysis may be performed for the case α4 − α1 α5 < 0.(5.35) has no multiple root. It occurs according to Fig. inside this interval. Eq. 2) . α3 = 1. α4 /α5 ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 133 First. For negative values of q1 we also have two possibilities α1 < sq < sQ . sQ = 1. the only diﬀerence is the presence of two thresholds. When Eq. s tends to s∗ = sq at q1 > 0 if α1 < sq < sQ . consider the case α4 − α1 α5 > 0. s2q ∈ (sQ . s1q = s2q = sq . Here we are dealing with the selection of a symmetric solitary wave.17. β7 = 1.588 < s0 < 2 they go to a dissipative solitary wave with s∗ = 1. α1 < s0 < sQ . s1q and s2q . or sQ < sq < α4 /α5 .28). Then α1 < s < α4 /α5 . sQ ).February 11. When 1. Assume that the root of Q3 is sQ while real roots of Eq.(5. There remains the problem of whether solitary wave selection is achieved in ﬁnite or inﬁnite time. Under conditions (5. β8 = 1. α1 < s0 < sq . any initially initial solitary wave with velocity s0 from the permitted interval (5. α4 = 2.(5.28) transforms into a dissipative solitary wave (5.12) the asymptotic solution (5.26).324.35) we have s1q = 1.26) may be similarly studied. (5.1 shows (see third and fourth rows) that two diﬀerent solitary waves may be selected depending on the initial value s0 . or sQ < sq < α4 /α5 . Integration of Eq.1.588 transform into a dissipative solitary wave with velocity s∗ = 1. sQ < s0 < sq .588. For the roots of Eq. All possibilities of the solitary wave velocity selection are collected in Table 5.1. We denote by s0 the initial value of s.35). Table 5. The most interesting case corresponds to q1 positive when s1q ∈ (α1 . sq < s0 < α4 /α5 . β2 = 1. β6 = 2.745. In order to avoid cumbersome algebra we consider one particular case only. Then the permitted interval for s is (1. (s − 1)(s0 − 1) .(5.35) are s1q < s2q . Other possibilities corresponding to the blow-up or the damping of the solitary wave (5. (5. and Q3 cannot have more than one root inside the interval (5. and Q3 (s) has only one root. α5 = 1.32) yields exp T = s0 − s1q s − s1q p1 s0 − s2q s − s2q p2 s−1 s0 − 1 p3 s0 s p4 exp p5 (s − s0 ) . this is the case of complete selection when all initially dissipationless solitons with initial velocity from the interval 1 < s0 < 1. α1 = 1.10). ε = 0.

657.8) and the DMKdV equation (1.13). ν= . Nepomnyashchy and Velarde (1994). with α= 2α6 α4 − α5 V 2 α2 . p5 = 4.980. Nekorkin and Velarde (1994).8). Indeed. 5. Garazo and Velarde (1991). in the ”travelling wave” limit the ODE reduction of our DMDDE (5. .2. coincides with the system studied by Nekorkin and Velarde (1994). z = −β z − α v y − ν y − G(v) + P. governing the evolution of bulk longitudinal long nonlinear strain waves in an elastic rod immersed inside an active/ dissipative medium.6. p2 = 0. It cannot be described by the asymptotic solution. p3 = 8. Consequently. 2 α7 − α8 V V (α7 − α8 V 2 ) α7 − α8 V 2 G(v) = α1 − V 2 α3 v+ v2 . The dynamical system representation of Eq. 2.1. all exact travelling wave solutions for the latter equation are valid in our case. β=− .904. For the last equation it was found that the single wave asymptotic solution accounts for a behaviour of the solitary waves generated from an arbitrary input. v = y. There is an interesting similarity with the results found for free surface shear long waves in a thermoconvective liquid layer described by a dissipation modiﬁed Kortewegde Vries equation (DMKdV).(5.February 11.286. Nekorkin and Velarde (1994). Hence we can anticipate the evolution of the initial localized strains according to that shown in Figs. Velarde et. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 134 Ampliﬁcation of Nonlinear Strain Waves in Solids with p1 = 7. Hence the wave amplitudes tend to A1 = 3(s1q − α1 )/2 α3 or to A2 = 3(s2q − α1 )/2 α3 .5 Concluding remarks We have obtained a nonlinear equation. Therefore. al (1995). see Christov and Velarde (1995). Of special interest is the evolution of an arbitrary initial pulse. we can transfer to the lon- . . p4 = 0.3. One can see that s → s1q or s → s2q at T → ∞. al (1995) when P = 0.420.9). 2) V (α7 − α8 V V (α7 − α8 V 2 ) . We already noted the similarity between the governing equation (5.9) is functionally identical to the corresponding ODE reduction of the DMKdV equation Christov and Velarde (1995). DMDDE (5. 2.4-2. see Sec. y = z. Velarde et.

15) there is no exact solution decaying at |θ| → ∞ if α8 = 0. al (1995). (5. These linear mixed derivative terms in (5. A.February 11.8) we have to rewrite it as vt = −gx . Then the time evolution of the wave energy for the solutions vanishing at ±∞ is governed by the equation ∂ ∂t ∞ ∞ ∞ v g dx = α2 −∞ ∞ −∞ vx g x dx + α6 −∞ ∞ 2 vx gx dx− α7 −∞ vxx gxx dx + α8 −∞ vxt gxt dx. ”bound solitons” and ”chaotic states” found for the DMKdV equation in Christov and Velarde (1995).36) is absent in the corresponding balance law for the DMKdV equation Garazo and Velarde (1991).21).(5.35). Both the compression and tensile asymptotic solutions occur due to the mixed terms with the velocities from the intervals (5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 135 gitudinal strain waves all the results about the existence of pulses. Finally. two sets of the selected solitary wave parameters result from the nonzero coeﬃcient q1 (5. and on the work. In order to get conservation laws for DMDDE (5. gt = α1 vx + α2 vxt + α3 ( v 2 )x + α4 vxxx − α5 vxtt + α6 (v 2 )xt + α7 vxxxt − α8 vxttt . The last term in (5. due to Eq. These terms decide the existence of either exact compression or tensile solitary wave solutions.8) appear due to the inﬂuence of the Poisson eﬀect on the kinetic energy density.28). Indeed.34) in the equation (5. done by external forces. (5. the ﬁrst term in the right-hand side of (5.36) Thus. Nekorkin and Velarde (1994).19).(5.36). The second term in (5. . (5.h. Velarde et. Here it diminishes the role of the third term in (5.36) accounts for the energy input while energy output is provided by the third term. instead of energy conservation we have an input-output energy balance that at the steady state gives a vanishing l. K. Moreover.29). Nepomnyashchy and Velarde (1994).36) may play a stabilizing or a destabilizing role depending on the sign of α6 .s. The corresponding variation of velocity of the kink-shaped solution depends upon α8 due to Eqs.

It allows us to cover both possibilities for longitudinal strain kink propagation. There is also interest in the analytical study of the simultaneous inﬂuence of dis- . The inclusion of cubic nonlinearity requires to extend the widely used so-called ”ﬁve constants” Murnaghan energy model. Eq..4). The kink-shaped localized traveling structure may be sustained by different balances.g. the strain wave magnitude.(3. Thus our results permit to delineate the yield point of the material. The domain of validity of Kerr’s model could be estimated comparing theory with experiments.g. Another possibility occurs when nonlinearity is balanced by dissipation (or accumulation). to a more general ”nine constants” Murnaghan model. Eq. Also we have shown that the external medium.. Whitham (1974). B. where R is the rod radius. strain solitary waves may be eﬀectively generated inside a rod with a free boundary based on the analysis of the exact travelling solitary wave solution of the governing double dispersive equation ( Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 136 Ampliﬁcation of Nonlinear Strain Waves in Solids Our predictions about strain solitary wave selection may help to the possible experimental generation of active/dissipative solitary waves in a rod partly embedding in an external medium with the Kerr contact.(5.8) with k. 5. and the wave length. e. are such that B = O(R2 /L2 ) << 1. L. the permafrost may be responsible for large wave ampliﬁcation. η and χ equal to zero). for instance.g.February 11. In view of a possible experimental test of our predictions we consider dissipative (active) phenomena occurring at the lateral boundary of an otherwise purely elastic and hence non-dissipative rod in the bulk Kerr (1964). Here we address the question of whether besides bell-shaped solitary waves also kink-shaped waves may propagate in an elastic rod. resulted in the kink solution of the modiﬁed Korteweg-de Vries (MKdV) equation case Ablowitz and Segur (1981). Indeed. e. the kink solution of the Burgers equation Sachdev (1987). bell-shaped and kink-shaped solitary waves. The bell-shaped solitary wave usually appears as a result of a balance between nonlinearity and dispersion. Nonlinearity in a pure elastic rod is caused by the ﬁnite stress values and the elastic material properties while dispersion results from the ﬁnite transverse size of the rod..(3. They are in balance when.1).3 Strain kinks in an elastic rod embedded in an active/dissipative medium There are two main types of nonlinear solitary waves which could propagate keeping its shape. e. There is a balance between cubic nonlinearity and dispersion.

mathematically the description of these processes requires inclusion of derivatives of high order in the model equation.February 11. r (5.37) As will be seen later. a2 . the third term in Eq. a4 ) account also for nonlinear elastic properties of the isotropic material.(3. n) we are now dealing with the fourth order moduli (a1 . Like Murnaghan’s third order moduli moduli. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 137 persion. Again both k and η are assumed to be of either sign. or the Murnaghan moduli (l. they can be either positive or negative. As we seen before. Eq. of kinks.(5.(3.1 Formulation of the problem Again we consider the propagation of a longitudinal strain wave in an isotropic cylindrical compressible elastic rod embedded in an external medium subjected to Kerr’s viscoelastic contact on the lateral surface of the rod r = R : k ∗ Prr = − w − η wt . particularly.3. Otherwise the statement of the problem is similar to that of the previous section with the exception of the components Prr . Below we shall use the results obtained in Porubov and Velarde (2002) 2 . m. Prr = (λ + 2µ) wr + λ (ux + w λ + 2µ + m 2 2 )+ (ur + wx ) + r 2 3λ + 6µ + 2l + 4m 2 w wr + (2l − 2m + n) ux + 2 r λ + 2l 2 w2 w (ux + 2 + 2wr + 2ux wr ) + (µ + m) ur wx + 2 r r 2 w3 3wwr 2 (l + 4a1 + a2 ) + + u3 + 3ux wr + (3m − a2 )ur wr wx + x 3 r r 2l − 2m + n + 24a1 + 10a2 + 2a3 + 4a4 ux w2 u2 w 2ux wwr + x + + 2 2 r r r with permission from Elsevier Science 2 Reprinted .4). ampliﬁcation. a3 . We choose nine constants Murnaghan’s approximation for the density of the potential energy. Eq. Thus besides the third order elastic moduli. 5.1) now does nor aﬀect the wave behavior.4). nonlinearity and dissipation (or accumulation) on the evolution. Prx of the Piola Kirchhoﬀ stress tensor P that are now written in framework of the nineconstants theory.

3.2 Combined dissipative double-dispersive equation Besides assumptions (i).38) Prx = µ (ur + wx ) + (λ + 2µ + m) (ur wr + ux ur ) + (2λ + 2m − n) ur w + 2r w 2m + a4 2m − n 3 wx + (µ + m) (wx wr + ux wx ) + 3u2 wx + wx + r 2 r 4 4l + 2m − n − 4µ − 2a2 − 2a3 2m − n − 2a2 − 2a3 w2 wx ur w2 + + 4 4 r2 2m − n − 2a2 − a3 − 2a4 ux wx w wr wx w 4m + a4 3 + + ur + 2 r r 4 4m − 4µ + 3a4 2 ur wx + (2m + µ − a2 − a4 )ux wr wx + 4 4l − 2a2 − a3 − 2a4 ux ur w ur wr w + + (2l + 2m − a2 − a4 )ux ur wr + 2 r r 3m − a2 2l + 5m − a2 2 2 (ur u2 + ur wr ) + wx wr + u2 wx .39) x x 2 2 5. we now assume that (iii) B ∼ R/L to provide a balance between nonlinearity and dissipation (or accumulation).40) (5.2. 2 2 (5. (5. 2(2(λ + µ) + k) 3λ + 4µ + k . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 138 Ampliﬁcation of Nonlinear Strain Waves in Solids (2l − µ + 12a1 + 4a2 + 2a4 ) u2 wr + x (2m − n − 2a2 − a3 − 2a4 ) w2 wr r2 + ur wwr 3 + 2(l + 2m + 2a1 )wr + 2r 2 4l − 2a2 − a3 − 2a4 u2 w wwx r (2m + µ − a2 − a4 )ur ux wx + + + 4 r r 2l + 5m − a2 2 2l + 2m − a2 − a4 2 2 2 ur wr + wr wx + ur ux + ux wx . (5. b2 = . 5. Like before we ﬁrst obtain the relationships between the longitudinal and shear displacements.41) where the explicit forms for the coeﬃcients are given by q2 = λ 2 q2 η .2. b1 = −2 q2 . w = b1 r Ux + b2 r2 Uxt + r3 (b31 Uxxx + b32 Uxtt ) + 2 3 +B1 r Ux + B2 r Ux + B3 r2 Ux Uxt . u = U + q2 r2 Uxx . (ii) from Sec.February 11.

have diﬀerent signs. 1 2η B1 − b2 {2q2 [3(λ + µ) + 2(l + m)] − (λ + 4l − 2m + n)} . b2 are always positive. while b1 . Due to the chosen nineconstant model we have to truncate the series (5. i = 1 ÷ 3. b32 = − . and the other coeﬃcients Bi .40). Similar terms with spatio-temporal mixed derivatives are of the same order. (5.41) into (3.40). are α1 = 2 2 8R η q2 8q2 (2(λ + µ) + k) − 8λ q2 + λ + 2µ . Substituting (5. i = 1 ÷ 7. 2(2(λ + µ) + k) B2 = − 1 {B1 [λ + 4l − 2m + n + 2b1 (3λ + 3µ + 4l + 2m)] + (2(λ + µ) + k) 1 1 + 4a1 + a2 + b1 [6l − 2µ − 2m + n + 48a1 + 18a2 + 2a3 + 8a4 ] + 2 3 2 b [2 + 2l − 2m + n + 32a1 + 12a2 + 2a3 + 4a4 ] + 2 1 b3 [4 − µ + 4l + 4n + 32a1 + 8a2 + 2a4 ]}. obey a combined dissipative double-dispersive (CDDD) nonlinear equation: vtt − α1 vxx − α2 vxxt − α3 ( v 2 )xx − α4 vxxxx + α5 vxxtt − α6 (v 2 )xxt − α7 ( v 3 )xx = 0.41). b32 are always negative when the coeﬃcients in (5. .February 11.29) we obtain that longitudinal strains. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 139 b31 = − λ q2 η b2 . while the cubic term is r3 Ux Uxxx .29). 2(2λ + 3µ) + k 2(2λ + 3µ) + k B1 = − 2 b1 (λ + 4l − 2m + n) + 2b2 [3(λ + µ) + 2(2l + m)] + λ + 2l 1 . α2 = . b31 . and using Hamilton’s principle (3. The explicit forms of the coeﬃcients αi . 3λ + 4µ + k B3 = − Note that q2 . ρ0 ρ0 (5. omitting higher order nonlinear terms and the ”corresponding” higher order linear derivative terms due to the assumption (iii). v = Ux . hence.37) are positive (Kerr’s model). Then.42) α3 = 3 [λ ρ0 1 2 + 2B1 − 2q2 (1 + 4B1 ) + 4q2 (1 − 4q2 ) + 2 . (5. the ”corresponding” comparable 2 linear term is r4 Uxxxx . The ”largest” of the quartic 4 terms is rUx ∼ RB 4 ∼ R(R/L)4 .

their combination in α3 may be quantitatively small while not so in α7 . the last four terms in (5.(5.42) in dimensional form without use of the multiple scales method and hence terms of diﬀerent orders may occur. be of the same order. Then the dissipative term. 3 3 3 α4 = R2 [(λ + 2µ − 4λ q2 ) q2 + 4b31 (λ − 2(2λ + 2µ + k) q2 ) ] . (i) and (ii). Note that the coeﬃcients α3 . α2 vxxt . in general. simultaneously. due to the above given assumptions. Accordingly. α2 = O(1).February 11. quantitatively. α6 and α7 depend on the third and fourth elastic moduli.(5. In contrast to the second order moduli (Lam´ coeﬃcients) they may be of diﬀerent signs e Lurie (1990).42) are smaller than the others and hence are considered small perturbations to the other four terms. with the exception of α3 and α6 and α7 that can be of diﬀerent signs depending upon the material properties of the rod. ρ0 α5 = R2 [ρ0 q2 (1 − 2q2 ) + 2k(b2 + 4q2 b32 ) + (9λ + 10µ)b2 /2 − 4{λ− 2 2 ρ0 4q2 (λ + µ)}b32 ].42) are always positive in framework of the Kerr model. Indeed. 8R q2 ηB1 ρ0 4 kB2 4 3 2 α7 = [B1 − 4q2 ] + {l[1/2 + 4q2 (16q2 − 8q2 + 3q2 − 1) + ρ0 ρ0 2 2 2B1 (1 − 4q2 )2 ] + m[1 − 4q2 + 8q2 (B1 + 2q2 )(1 + 2q2 )] + α6 = − 2 2 2 2 2nq2 [q2 − 4q2 − 2B1 ] + λ[1/8 + 2q2 (1 + 4q2 ) + B1 (1 + 2B1 + 8q2 ) + 4 2 2B2 (1 − 4q2 )] + µ[1/4 + 8q2 + 2B1 (B1 − 4q2 − 4q2 ) − 8q2 B2 ] + 2 3 4 3 2 a1 [1 − 16q2 + 96q2 − 256q2 + 256q2 ] + 4a2 q2 [16q2 − 24q2 + 9q2 − 1] + 2 2 4a3 q2 [1 − 4q2 ] + 16a4 q2 [1 − q2 ]2 }. We shall refer to this case as the weakly dispersive limit. Note that we have obtained Eq. will overcome the nonlinearity and drastically alter the wave . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 140 Ampliﬁcation of Nonlinear Strain Waves in Solids 3 µ 1 − 8q2 B1 − 16q2 − 4kq2 B1 + l 16 2 2 1 2 2 − 4q2 (1 − 4q2 + q2 ) + m(1 − 4q2 (3 + 4q2 )) + 4nq2 ]. All coeﬃcients in Eq. and then the terms α3 ( v 2 )xx and α7 ( v 3 )xx may.

2.3 Exact solutions Assuming that the solution of Eq. 5. the signiﬁcant balance will be between the quadratic-cubic nonlinearities and dispersion.1. β3 = . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 141 shape before the nonlinearity comes to play. (5.3. κ) β3 − .(5. (5. However. slightly perturbed by the inﬂuence of dissipative terms.42) depends only upon the phase variable θ = x− c t. β5 = . v + β1 v + β2 v + β3 v 2 + β4 (v 2 ) + β5 v 3 = N. κ) 3β5 −β5 (5. α2 << 1. We shall call it the weakly dissipative limit.45) 2(1 − Aβ4 )(2β4 − 3Aβ5 )2 (ii) bounded periodic solution v=√ m cn(m θ.43) where a dash denotes diﬀerentiation with respect to θ.44) m2 = 2 2 (3A2 β5 − 4Aβ4 )(3β2 β5 − β3 ) + 4β4 (β2 β4 − β1 β3 ) + 3β1 β5 . α5 c2 − α4 α5 c2 − α4 α4 − α5 c2 β4 = α6 c α7 . B=− .(5. asymptotic and numerical solutions. β5 2β4 − 3A β5 (5. Among exact solutions obtained there we consider two bounded solutions: (i) kink-shaped solitary wave solution v = A m tanh(m θ) + B.16) studied in Sec. κ) sn(m θ. with A= β4 ± 2 β4 − 2β5 β1 − A β3 . α5 c2 − α4 α4 − α5 c2 Eq. Therefore.42) is that it embraces diﬀerent important cases.(5. for which we shall below give exact.E. if the inﬂuence of the external medium is weak enough.43) is a particular case of the equation (2. k and η are small.3. and β1 = α2 c c2 − α1 α3 .February 11.42) becomes the O. then in the moving frame Eq.46) . C1 + cn2 (m θ. β2 = . N is a constant. κ) dn(m θ. the advantage of equation (5.D.

(5. see Fig.46) exists only for non vanishing β4 .E. If α4 ÷ α7 are equal to zero we get from Eq.46): √ β3 2mκ sn(m θ.46) has a functional form diﬀerent from both the KdV cnoidal wave and the MKdV bounded periodic solutionAblowitz and Segur (1981). The CKdV equation possesses the one-parameter kink solution (5.43). (5.42). is v + β2 v + β3 v 2 + β5 v 3 = N. α2 c v + (c2 − α1 )v − α3 v 2 = N. dispersion and (linear and nonlinear) dissipation required for the periodic nonlinear wave than the standard balance between nonlinearity and dispersion that supplies both the KdV and the MKdV periodic solutions.44) like the MKdV periodic solution.48) It corresponds to the O. When κ = 1 we have C1 = 0.48) has a variety of bell-shaped solitary wave solutions. hence only in the presence of the nonlinear dissipative term β4 (v 2 ) in Eq.(5. m2 = √ 2 3κ 4 κ4 − κ2 + 1 √ and the following restrictions on the coeﬃcients: β4 = − 1 2 −β5 . κ) − .45) if β1 = β4 = 0. (5.D.E.(5.D. has a form diﬀerent from (5.(5. β1 = β4 = 0.6. and the solution (5. v= √ 3β5 −β5 2 with m2 = (3β2 β5 − β3 )/(3β5 (1 + κ2 )).44) with (5. or its equivalent dynamical system. exhibits a more complicated balance between nonlinearity.43) tends to the kink-shaped solution (5. while the KdV cnoidal wave solution becomes the bell-shaped or solitary wave solution in the analogous limit. .43). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 142 Ampliﬁcation of Nonlinear Strain Waves in Solids with C1 = 1 − 2κ2 + 2 κ4 − κ2 + 1 β2 + 3β1 . 1.47) The periodic wave solution (5. The dissipationless limit of Eq.(5. reduction of the Burgers equation. Eq. reduction of the combined KdV-MKdV (CKdV) equation.48). β3 = 3β1 −β5 .49) . However. Moreover. Note also that the solution (5.February 11.42) the O. having a kink limit at κ = 1. a periodic solution of Eq. Equation (5. (5.

B= ≡− . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 143 Its kink solution has the functional form like (5. ∂θ2 (5. v → h2 at θ → −∞. for the Burgers model any pair of hj deﬁne phase velocity and the wave number. the phase velocity c and the wave number m.3. 5.February 11. i. Then from Eq.4 Weakly dissipative (active) case When the viscocompressibility coeﬃcient η is small. =− .e.51) is a function of the phase variable θ = x − ct .(5.51) For nonzero δ we assume that a solution of Eq.42) is the perturbed combined double-dispersive equation.45) as well as for the MKdV kink solution the boundary conditions imply additional restrictions on the parameters of the problem. while in general for the one-parameter solution (5.50) There are two free parameters. when the external medium is of little inﬂuence we can take α2 = δ α2 . (5. δ << 1. Then Eq. If the boundary conditions are v → h1 at θ → ∞. α6 = δ α6 . 2β1 2β3 2 The coeﬃcients βi depend upon the phase velocity c. (5.44) but with A=− α2 c β1 c2 − α1 β2 ≡ . and the elastic features of the material of the rod and the parameters of the external medium.43) β4 ± 2 β4 − 2β5 h1 − h2 β 1 − A β3 h1 + h2 m= . =− β5 2 2β4 − 3A β5 2 while the Burgers model (5.44). m − f ree. Hence.52) . v = v(θ).(5.51) we get that (c2 − α1 )vθ − α3 ( v 2 )θ − α7 ( v 3 )θ − (α4 − α5 c2 ) vθθθ = −δ c ∂2 α2 v + α6 v 2 + O(δ 2 ). α3 β3 2α3 2β3 (5.49) gives m= (h1 − h2 )β3 β2 h1 + h2 .(5. vtt − α1 vxx − α3 ( v 2 )xx − α7 ( v 3 )xx − α4 vxxxx + α5 vxxtt = δ(α2 vxxt + α6 (v 2 )xxt ) (5. then for the kink-shaped solution of Eq.

55).1 while Fig.θ − α3 ( v0 )θ − α7 ( v0 )θ − (α4 − α5 c2 ) v0.February 11.58).55) (5.52) is sought in the form v = v0 + δv1 + . at θ → ±∞.e. underlying the CKdV equation. ∂θ2 (5. 5. with A= 2 2(α5 c2 − α4 ) 3α1 α7 − α3 − 3α7 c2 m. In both ﬁgures the steepness of the wave front (i.51) is depicted in Fig.E. (5. 2 3 (c2 − α1 )c0. only when 3α2 α7 − 2α3 α6 = 0 and has the form v1 = α6 m A θ v0.57) whose solution satisﬁes the b. vi → 0. left .2 shows the evolution when α2 > 0.. 2(α5 c2 − α4 ) (5. m = const) and the phase velocity of the initial MKdV kink (5. The exact kink solution of Eq.(5.51) allows exploration of the kink evolution outside the range imposed by condition (5.53) into (5.54) We look for a solution satisfying the boundary conditions v0 → h± . Case α2 < 0.52) in the leading order we have the O.(5.54) has the form v0 = A m tanh(m θ) − α3 /(3α7 )..53) obeys the inhomogeneous linear equation 2 (c2 − α1 )v1. For a kink h+ = h− .53) Substituting (5.58) Numerical integration of Eq.(5.θθθ = −c ∂2 2 α2 v0 + α6 v0 . i > 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 144 Ampliﬁcation of Nonlinear Strain Waves in Solids The solution of Eq.56) The correction v1 (5. 5. and with all derivatives of vi with respect to θ vanishing at inﬁnity.D. (5. hence corresponding to accumulation.θ − 2α3 (v0 v1 )θ − 3α7 (v0 v1 )θ − (α4 − α5 c2 ) v1.c. m2 = .θθθ = 0.θ . α7 6(α4 − α5 c2 ) (5.(5. corresponding to dominating damping in the linearized Eq.

remain one and the same.5 20 -0.5 Fig. Then from Eq.February 11.5 40 60 80 100 120 140 x -1 -1.3. The length of the shelf increases like the length of the shelf behind the perturbed KdV soliton Ablowitz and Segur (1981) but the height also increases. The inﬂuence of dissipation (accumulation) shows in the growth experienced by the shelves before and behind wave fronts. The small parameter of the problem is ε = B = R/L.. proﬁle.(5.1 Numerical simulation of the MKdV kink evolution with weak dissipation. Suppose that our dimensionless solution v depends upon the phase variable θ = x − ct and that c = 1 + εc1 + ε2 c2 + .5 Weakly dispersive case Let us now take into account all assumptions from Sec. Assume the scale for v is B. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 145 v 1 0. 5.2 to obtain the dimensionless form of Eq.42) we get (c2 − α1 )v θθ + ε 2c2 c1 v θθ − α3 ( v 2 )θθ + α2 c0 v θθθ + 0 0 ε2 (c2 c1 [2c2 + c1 ]v θθ + α2 c0 c1 v θθθ + 0 . 5. 5. with more dramatic eﬀect in the wave front... As a result no quasistationary proﬁle is possible during the evolution of the perturbed MKdV kink at variance with the result found for the bell-shaped strain solitary wave in Porubov and Velarde (2000).(5. and for t is L/c0 where c0 is a characteristic velocity of the wave. for x is L.42).3.

(5.62) v0 = − α3 α3 In the following higher order we get an inhomogeneous linear ordinary differential equation for v 1 (θ).θ ) = 2 ∂θ . and α6 = α6 /R. reduction of Burgers equation whose kink-shaped solution is √ α2 α1 α1 c1 m tanh(mθ) + . The solution is sought in the form v = v 0 + ε v 1 + .D.59) with α2 = α2 /R..February 11.5 Fig.5 x 20 -0.5 40 60 80 100 120 140 -1 -1.θ = 0. α4 = α4 /R2 . while the next order yields the equation √ ∂2 2α1 c1 v 0 − α3 v 2 + α2 α1 v 0. In the leading order we have c0 = α1 ..2 Numerical simulation of the MKdV kink evolution with weak accumulation [α5 c2 − α4 ]v θθθθ − α7 v 3 + α6 c0 v 2 ) = O(ε3 ). 0 2 ∂θ (5. √ ∂2 (2α1 c1 v 1 − 2α3 v 0 v 1 + α2 α1 v 1. α5 = α5 /R2 .60) The boundary conditions are the same as for the kink solution in the √ weakly dissipative case.61) In brackets we have the O. 5. (5. 0 θθ θθθ (5.E.. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 146 Ampliﬁcation of Nonlinear Strain Waves in Solids v 1 0.

(5. c1 2 2 α1 [α2 α7 − 2α3 α6 ] − α2 α3 .60) vs Burgers kink (dashed line) for b2 < 0 in Fig.θθθ − α1 c1 [2c2 + c1 ]v 0. α2 α3 1 2 b2 = √ 2 2[α1 α5 − α4 ]α3 + α1 α2 [2α3 α6 − α2 α7 ] .64) is shown with a solid line. If b2 = b3 = 0 then v may be written as √ α2 α1 α1 c1 v=− m tanh(m[1 + εb1 ]θ) + + O(ε2 ).44). is v 1 = v 0. (5.θθθθ + α7 v 3 − α6 α1 v 2 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 147 √ [α4 − α5 α1 ]v 0.60). 5. Correspondingly.63) where b3 = const. The time evolution . α1 α2 α3 b1 = We see that b1 . Note the asymmetric disturbance of the proﬁle near the upper and lower states of the solution that are exchanged with the opposite sign of b2 . whose solution. while Fig. The inﬂuence of dispersion is provided by b2 only.3 accounts for the unperturbed Burgers kink (5. in Fig. 5. The dashed line in Fig. provided that c2 = 1 α7 [3α1 c2 − α2 m] − c2 .42) with an initial condition in the form of a Burgers kink-shaped wave (5. decaying at inﬁnity. 5.63) aﬀects the smoothness of the wave front in the solution (5. (5. b2 remain nonzero even when α6 = α7 = 0. 5. The case b1 = b2 = 0 corresponds to a constant phase shift of the unperturbed Burgers kink solution. Numerical integration of Eq. 1 2 1 2 2α3 c1 (5.θθ − √ α2 α1 c1 v 0.4(a) and for b2 > 0 in Fig.θθ 0.February 11. while bi .θ [b1 θ + b2 Log(cosh(mθ)) + b3 ] .50) shows that the wave attracts the proﬁle described by the asymptotic solution even at moderate ε. i = 1.(5. Shown in Fig.4 it is shown the inﬂuence of the second term (b1 = b3 = 0) on the shape of the solution v (5. 5. depend upon the coeﬃcients of Eq. 5.3(b) accounts for negative values of b1 . The features of a quasistationary asymptotic solution can be observed when studying the time-dependent process of the kink formation.(5. 2.θθθ .3(a) is the case b1 > 0.59). while solution (5.64) α3 α3 Then the ﬁrst term in (5.44).50).4(b).

5 b 2 1. the left proﬁle corresponds to the initial Burgers kink.5.(5.5 and positive . (b) b1 < 0.5 5 10 15 20 25 30 x Fig.5 1 0.42) have been chosen such that b2 . 5. The values of the coeﬃcients in Eq.5 5 10 15 20 25 30 x v 2.6 where the propagation of the undisturbed Burgers kink is shown with a dashed line.February 11. 5. 5. is negative for Fig. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 148 Ampliﬁcation of Nonlinear Strain Waves in Solids v 2. of the wave at ε = O(1) is shown in Figs. 5.5 a 2 1.5 1 0.3 Smoothness of the Burgers kink proﬁle due to the ﬁrst term in the ﬁrst order asymptotic solution: (a) b1 > 0.

6. (b) b2 > 0. The three most right solid line proﬁles show that the quasistationary perturbed kink-shaped waves are rather close to the corresponding asymptotic proﬁles shown in Figs. 5. for Fig.5.5 a 5 v 2.5 1 0.February 11. The phase velocity remains one and the same during the whole time and practically equal to the velocity of . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 149 v 2. 5.6.5 10 15 b 20 25 30 x 5 10 15 20 25 30 x Fig.5 1 0.5 2 1.4 Disturbances on the Burgers kink proﬁle due to the second term in the ﬁrst order asymptotic solution: (a) b2 < 0. 5. 5.5 2 1.

3. 5. diverges even when dispersion is small. However. the initial Burgers kink.February 11. as Eq. Other possibilities exist for an exact travelling wave solution (5.(5. due to the rod being embedded in an external viscoelastic medium.6 Summary of results and outlook We have shown that kinks can propagate in a compressible elastic rod. This is very important from the point of view of application because for most materials the values of ai are unknown. are such that B ∼ R/L << 1.g.42) can be approximated with α6 = α7 = 0 and hence without using the fourth-order elastic moduli.(5. High order terms do not signiﬁcantly alter the wave structure in the weakly dispersive case. when the wave characteristics. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 150 Ampliﬁcation of Nonlinear Strain Waves in Solids v 0. ai . Any other initial kink diﬀerent from the Burgers one.3 0. In such circumstance Eq.42) as well for the perturbed MKdV kink when the additional condition (5. In contrast to the bell-shaped solitary wave selection in previous section now only the wave number tends to a prescribed ﬁnite value. if dis- .2 0.(5. 5.5 Numerical simulation of the Burgers kink evolution into the quasistationary proﬁle in the weakly dispersive case at b2 < 0.44) of Eq.1 x 20 40 60 80 100 120 Fig.64) indicates. amplitude and wavelength. We have found in the weakly dispersive limit how the kink-shaped wave is selected.58) is satisﬁed. a cubic nonlinearity of the elastic material is taken into account and dissipation exists e. of radius R.

However.s. However.February 11.3 0. dissipation and nonlinearity are of the same order. The existence of these solutions require additional restrictions studied in Sec. As a useful aspect of our analytical study is that obtaining exact and . and so on.h.2 0. 5. and thus the corresponding ”dissipative” quartic nonlinearity demands a simultaneous quintic nonlinear term.6 Numerical simulation of the Burgers kink evolution into the quasistationary proﬁle in the weakly dispersive case at b2 > 0. adding a term like (v 3 )xxt is not permitted by the assumptions in Sec. Exact solutions obtained account for the case when the dispersion. of Eq. In particular they provide the nontrivial condition 3α2 α7 − 2α3 α6 = 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 151 v 0.3. Moreover numerical simulations in the weakly dispersive case point to a formal validity of the results from Sec. we have no kink selection in the weakly dissipative (active) case when all wave parameters tend to ﬁnite values. that allows the propagation of a quasistationary perturbed kink in the weakly dissipative case.(5. This seems due to the absence of ”dissipative (active)” cubic nonlinear terms in the r. 5.1 x 20 40 60 80 100 120 Fig.5 when dispersion is not small. Indeed.3.51).2.3. (v 3 )xxt and (v 4 )xx will be of the same order. 5.3. persion becomes signiﬁcant those terms must be taken into consideration for a correct description of the strain kinks. 5. The inﬂuence of dissipation on the balance between nonlinearity and dispersion appears aﬀecting wave evolution much more drastically than what dispersion does to the Burgers nonlinearity-dissipation balance.

done by normal and tangential external forces at the lateral surface of the rod. On the other hand. r = R.1 Let us use the same notations as in previous sections.65) accounts only for the weakly dispersive case. the evolution of surface waves with surface deformation η of amplitude A and wavelength L in a layer of depth H is governed by the equation Porubov (1995): ηt + γ1 (η 2 )x + γ2 ηxx + ε γ3 ηxxx + γ4 (η 2 )xx + γ5 (η 3 )x = 0. Its O. reduction coincides with Eq. we have shown that an external medium.(5.(5. In view of the lack of experimental data on the fourth order moduli our results may help in ﬁnding these moduli using the linear algebraic relationships between them and the kink wave amplitude and velocity obtained using B2 .. This may be used to locate zones of potential plasticity. e. may be responsible for the ampliﬁcation of both the strain kink and bell-shaped solitary waves. (5.(5.g.66) ∗ We assume the external medium yields a normal stress Prr .65) Eq.E. see Sec. ∗ Prr = − k1 w + k2 r wxx . Contrary to the elastic bulk dissipation–free rod inside the liquid layer here the dissipation is caused by ﬂuid viscosity and heat diﬀusion. (5.4 Inﬂuence of external tangential stresses on strain solitary waves evolution in a nonlinear elastic rod Formulation of the problem 5. 5.43).D. permafrost. now is: ∞ δA = 2π −∞ ∗ ∗ (Prr δw + Prx δu)dx .February 11.4. δA.2. when A/H = H/L << 1. Note that Eq. 5.65) describes waves past an instability threshold. Analytic solutions could be also the starting points for the numerical search of homoclinic and heteroclinic trajectories yielding solutions of possibly more complicated form than those described in this report. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 152 Ampliﬁcation of Nonlinear Strain Waves in Solids asymptotic solutions allow reliable testing points for time-dependent numerical integration. The elementary work. note that our solutions may also be used to account for the evolution of surface waves in a heated liquid layer subjected to variations of surface tension. r (5.67) . Finally. Indeed.

both k1 and k2 are positive and constant. kd is of either sign.c. 2 w = b1 r Ux + kd b2 r2 Uxx + b3 r3 Uxxx + B1 r Ux .73) where . We consider a more general model. see also Kerr (1964) and references therein. (3.13) that yields in our case: 2 u = U + kd a1 r Ux + a2 r2 Uxx + kd A1 r Ux .72) (5.c. (3. 5. ∗ Prx = Prx .70) (5.) are: w → 0.2 the unknown functions u.11).4. at r = R. Hence the boundary conditions (b. Tangential stresses on the lateral surface are assumed in the form: ∗ ∗ Prx = kd Prr .71). (ii) long elastic strain waves have a characteristic length L such that relative to the rod radius R. Prx of the Piola . at r = R. w are expanded in a power series (3. Let us obtain the relationships between longitudinal and shear displacements satisfying b. (3. It corresponds to the Pasternak (1954) model based on the representation of the contact by means of the interacting spring elements. ∗ Prr = Prr .71) where the components Prr .68) that relates to the Coulomb-Amonton law H¨hner and Spencer (1998). (5. 3. in order to account for the inﬂuence of an active external medium providing an energy inﬂux .2 Derivation of the governing equation Simpliﬁcations follows from the natural assumptions are similar to those used in previous sections: (i) pure elastic strain waves have magnitude B << 1.February 11. (5.12). a Nikitin (1998) when kd > 0 is a friction coeﬃcient. (5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 153 where ki are the stiﬀness coeﬃcients of the external medium.(5. R/L << 1. (iii)B ∼ R2 /L2 .69) (5.Kirchhoﬀ stress tensor P are deﬁned by Eqs. (5. According to the procedure developed in Sec. at r → 0.69).10).

(5. b2 = . µ B1 = − λ + 2l + 2(λ + 4l − 2m + n)b1 + 2[3(λ + µ) + 4l + 2m]b2 1 .29) we obtain that longitudinal strains.7) like before in order to omit negligibly small higher-order nonlinear terms and the ”corresponding” higher order linear derivative terms due to the assumption (iii).(5.72).(3. while nonlinear term coeﬃcients A1 . Substituting Eqs. (5. b3 = . µ 2µ2 (3λ + 4µ + k1 ) b1 = − λ k 1 λ b1 (k2 b1 − λa2 ) . B1 have diﬀerent but opposite signs. β2 = − ρ0 R ρ0 R α2 = 3 1 1 [ (1 + 2b1 )(1 + 2b2 ) + µ 1 + 2b3 + l 1 + 2b3 + 1 1 1 ρ0 2 3 2 m(1 + 2b1 )(1 − b1 )2 + nb2 + {λ + 2b1 (λ + µ) + 2k1 b1 }B1 + 1 3 2 2 k1 kd b2 1 µ2 1 1 µ (λ + m)(1 + 2b1 ) + µ(1 + b1 ) − nb1 − B1 ]. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 154 Ampliﬁcation of Nonlinear Strain Waves in Solids a1 = − 2 2 [µ2 (3λ + 4µ + k1 ) + 2 kd k1 λ ]b1 k1 b1 . obey a nonlinear dispersive-dissipative equation vtt − α1 vxx − α2 ( v 2 )xx − α3 vxxtt + α4 vxxxx = kd (β1 vx + β2 (v 2 )x + β3 vxxx ). are always positive. ρ0 µρ0 β1 = − 2k1 b1 2k1 B1 . Due to the chosen ﬁve-constant theory (3.74) where α1 = 2 2 λ + 2µ + 4[λ + b1 (λ + µ)]b1 + 2k1 (b2 − b2 ) 3 k1 kd b2 1 1 + . 2 4 b1 . b1 .b3 are always negative. a2 . v = Ux . (5.6).February 11.1) we have to truncate the series (5. 2(λ + µ) + k1 µ(3λ + 4µ + k1 ) 2(2λ + 3µ) + k1 A1 = − k1 B1 . a2 = − .73) into Eq. 2(2(λ + µ) + k1 ) Note that a1 .

(5. see Sec. α3 c2 − α4 α3 c2 − α4 α3 c2 − α4 γ4 = β1 β2 .4. vv . Assume the solution of Eq.75) where a dash denotes diﬀerentiation with respect to θ. κ) + b2 . β3 = 2 2 2µ µρ0 α4 = R2 [a2 {4µa2 − 2(λ + 2µ) − 4b1 (λ − µ)} + µ b2 − 8b3 {λ + 2b1 (λ + µ)}]+ 1 2ρ0 R2 k 2 2R2 b1 [k2 µb1 − k1 b3 (µ − k1 )] + 2 d [b2 (9λ + 10µ)µ2 + ρ0 2µ ρ0 2 2 k1 {b2 [4µ2 (a2 + 1) + 2µ(µ − 3λ)] − 4k2 µb2 } + k1 b1 {4µb3 + (λ + 2µ)b1 }]. general periodic solution may be expressed using either the Weierstrass elliptic function .4. More general solutions are found in terms of the elliptic functions see Sec. or the Jacobian elliptic function.(5. 2−α α3 − α4 α3 c 4 c2 Following the procedure from Sec.74) is nonitegrable by the Inverse Scattering Transform method Ablowitz and Segur (1981). 1 When kd = 0 Eq. 5.1.3 Symmetric strain solitary waves Even at kd = 0 Eq. . One can see they are in balance when a solution has the second order pole. γ3 = .1. There exist diﬀerent approaches based on the assumptions of the appropriate ansatz for a solutions.D. usually travelling wave.February 11. 3. v = a2 cn2 (pθ.(5. then in the moving frame Eq. exact solutions may be obtained.(5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 155 α3 = 2 2 2 2 R2 k1 kd b2 R2 2 2R[k2 µb1 − k1 µ(a2 b1 + b3 ) − k1 kd b1 b2 ] 1 (b1 −2a2 )+ . v + kd γ1 v + γ2 v + γ3 vv + kd γ4 v 2 + kd γ5 v = 0. we compare the leading-order derivative term. and main attention is paid here on the exact solutions vanishing at inﬁnity. v = a1 ℘(θ.74) depends only upon the phase variable θ = x− c t. g1 .74) becomes the O. 2. However. only particular.74) becomes the double dispersive equation (DDE) accounting for nonlinear strain waves in a free lateral surface rod. g2 ) + b1 . Hence. (5. and γ1 = β3 α1 − c2 2α2 . γ2 = . v . γ5 = . 2. and the nonlinear term. Hence. up to now only localized strain waves were observed in experiments.E.

(5. (5. kd γ4 v 2 . and by a balance between active/dissipative terms. Solitary wave solution exists under speciﬁc initial condition in the form of (5. Since they correspond to the limit κ → 1. and dispersion. (5. γ3 vv . It is known that in non-dissipative case even rather arbitrary initial pulse splits into the train of solitary waves each being . α3 β2 (5. 2 whose solution vanishing at inﬁnity is obtained from the equation (5.76) Substituting (5. βj depend on the elastic features of the material of the rod and the parameters of the external medium. kd γ1 v . or tensile. They deﬁne the sign of the wave amplitude G. while the phase velocity c remains a free parameter.G= . (5.78) is supported simultaneously by a balance between nonlinearity. (5. 2−α ) 4(α3 c 2α2 4 2γ4 = γ1 γ3 .75) and equating to zero coeﬃcients at corresponding powers of tanh(pθ) one obtains p2 = c2 − α1 3(c2 − α1 ) .78). (5. G < 0. one can assume the solution of the form: v = G cosh−2 (pθ). Its solitary wave solution is (5.77). G > 0.78) deﬁnes the phase velocity. α2 α3 β1 + (α4 − α1 α3 )β2 + α2 β3 = 0.76) into Eq.77) (5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 156 Ampliﬁcation of Nonlinear Strain Waves in Solids Localized solitary wave solution are of interest here. The last balance is realized under conditions (5. γ5 = γ1 γ2 .78) while the second one imposes a restriction on the equation coeﬃcients.76) at t = 0. The coeﬃcients αi .77). When conditions (5.February 11. c2 = α4 β2 + α2 β3 . Hence the general solution (5.79) Eq. kd γ5 v.78) hold Eq. (5.76).75) may be rewritten as [ ∂ 1 + kd γ1 ](v + γ2 v + γ3 v 2 ) = 0. wave may propagate in the rod. ∂θ 2 1 v + γ2 v + γ3 v 2 = 0. v (like DDE’s solution).76). The ﬁrst of the conditions (5. hence either compression.79) is nothing but ODE reduction of the DDE integrated twice.

We look for a solution satisfying the boundary conditions vi → 0. −∞ (5. Assume a solution of Eq.80) whose one-parameter solution has the form of (5. with θx = 1. 3.74) depends upon the phase variable θ and the slow time T .81) The operator acting on the function v1 in Eq.3.(5. (5. but c now depends upon T .84) . ∞ 0. and no localized waves appear. θt = −c(T ).. When dissipation/accumulation predominates it destroys initial pulse before balance between nonlinearity and dispersion become to play. α2 cT Q6 (c) = (α1 − c2 )(α3 c2 − α4 )(6α3 β2 c4 + [5α2 α3 β1 − 6β2 (α4 + α1 α3 ) − α2 β3 ]c2 − 5α2 α4 β1 + 6α1 α4 β2 + α1 α2 β3 ).74) is sought in the form v = v0 (θ) + kd v1 (θ) + .θ = 0.θθθ + β1 v0 + β2 v0 + β3 v0 . (5. At order O(kd ) there is an inhomogeneous linear equation for v1 .θθT − α3 cT v0. T ). (5.83) provides the absence of secular terms.θ − (α3 c2 − α4 )v0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 157 accounted for the travelling solitary wave exact solution of the DDE. (c2 − α1 )v1.θ − (α3 c2 − α4 )v1.θθθ − 2α2 v0 v0. kd << 1..February 11.82) v0 F dθ = 0. It yields the equation deﬁning the phase velocity c. where 2 F = 2cv0. when nonlinearity and dispersion dominate over dissipation/accumulation.76). T = kd t. Then asymptotic solution of Eq.(5. see Sec.T + cT v0 − 2α3 cv0. Consider now a weakly dissipative case. i Then the Fredholm alternative. at |θ| → ∞.81) is adjoint to that in Eq.θθθ − 2α2 (v0 v1 )θ = F.(5. In the leading order we get (c2 − α1 )v0. (5. v = v(θ.(5.80).

However. and the sign of cT depends upon the signs of the polynomial Q6 (c) and the quartic polynomial in the r. Fig. (5. When cT keeps its sign as time passes possible scenarios for evolution of c are either vanishing or diverging to inﬁnity. It was called in Sec. we have cT > 0 when c∗ < c < c∗ and cT < 0 when q1 < c < c∗ or c∗ < c < q2 . 1. Based on Eq. q2 are real neighboring 2 1 roots of Q6 (c).3 the selection from below when c0 < cs . To put this another way.84).(5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 158 Ampliﬁcation of Nonlinear Strain Waves in Solids where 2 Q6 (c) = 30α3 c6 − 3α3 (11α4 + 10α1 α3 )c4 + 2α4 (11α1 α3 + 10α4 )c2 − 2 2 α1 α3 α4 − 5α1 α4 .θ .84) may be integrated giving the dependence c on T in an implicit form. Eq.81) is v1 = C(α1 − c2 ) (α1 α3 − α4 )c2 cT [1 − tanh(pθ)] + [C + θ]v0 + 2α2 (α1 − c2 )2 (α3 c2 − α4 ) C (α1 α3 − α4 )c2 cT [C1 + θ − θ2 ]v0. the value of 1 velocity is selected according to the governing equation coeﬃcients.17(b). (5. while the case c0 > cs is referred to by the selection from above. 2 (α1 − c2 )2 (α3 c2 − α4 ) . where q1 . It means 2 1 2 1 that if an initial velocity. 2 2 If Q6 (c) < 0.77) (α1 − c2 )(α3 c2 − α4 ) < 0.(5. q1 < c∗ < c∗ < q2 . (5.85) Due to (5.h. in particular. c∗ < c∗ . c tends to cs = c∗ . c(T = 0) ≡ c0 lies in the intervals c∗ < c0 < c∗ or 1 2 c∗ < c < q2 . More interesting case is realized when. The wave amplitude G(c) is selected like velocity c according to (5.84) c1 = α1 and c2 = α4 /α3 correspond to p = 0 or p → ∞ in (5.s. 1. (5.17 (a).84) the solution of Eq.February 11.77). Other stationary solutions are deﬁned from the equation 6α3 β2 c4 + [5α2 α3 β1 − 6β2 (α4 + α1 α3 ) − α2 β3 ]c2 − 5α2 α4 β1 + 6α1 α4 β2 + α1 α2 β3 = 0.85). the velocity c(T ) tends to the ﬁnite value cs = c∗ at T → ∞. their amplitudes increase (decrease) while their widths decrease (increase). If α3 β2 > 0 and Q6 (c) > 0 at q1 < c < q2 . Fig. 1. are 2 1 located in the interval. The stationary solutions of √ (5. We see the waves remain symmetric with respect to their maximums. important features of the evolution of c may be established studying the sign of cT without integration. real roots of Eq. of Eq.77).

87) .74) we get that 2 (1 − α1 P 2 )vθ − 2α2 v vθ − (α3 − α4 P 2 ) vθθθ = kd (2P [α1 vX + α2 vX + (α3 − 2α4 P 2 )vθθX ]+ 2 2 PX [α1 v+α2 vθ +(α3 −6α4 P 2 )vθθ ]+P [β1 v+β2 v 2 +p2 β3 vθθ ])+O(kd ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 159 where C1 is a constant. θt = −1. and a plateau appears behind the solitary wave.86) is sought in the form v = v0 (θ.(5. The standard matching asymptotic procedure should be used to complete the uniformly valid asymptotic solution satisfying boundary conditions (5.4 Evolution of asymmetric solitary waves Another asymptotic solution may be found when kd << 1 is considered but now it is assumed that a solution of Eq.(5.4.. (5. Then from (5. X) + kd v1 (θ. T ) + kd v1 (θ. v = v(θ.74) is a function of the phase variable θ and the slow coordinate X. 2 C = 12p[α3 β2 c8 + α3 (5α2 α3 cT − 2α1 α3 β2 − 2α4 β2 − α2 β3 )c6 + 2 2 2 (β2 {α1 α3 + 4α1 α3 α4 + α4 } + α2 β3 {2α1 α3 + α4 } − 11α2 α3 α4 cT )c4 + (α2 α4 cT {2α1 α3 + 5α4 } − α1 {2α4 β2 [α1 α3 + α4 ] + α2 β3 [α1 α3 + 2α4 ]})c4 − 2 α1 α4 (α4 β2 + α2 β3 )]/{5α2 (α1 − c2 )3 (α3 c2 − α4 )} One can check that evolution of the solution v = v0 (θ. it does not vanish at θ → −∞. 5.February 11..86) The solution of Eq. (5. and two values of the selected wave parameters exist thanks to the mixed dispersion term in Eq. β3 = 0.(5.76) nor asymptotic selected solitary wave solution exists at β2 = 0..74). see Ablowitz and Segur (1981) for details. T ) does not signiﬁcantly diﬀer from that of the solution v = v0 (θ. with θx = P (X). However. Absence of plateau requires additional restrictions on the equation coeﬃcients. X = kd x. when α3 = 0. X).82). Note that neither exact solitary wave solution (5. T ). X) + .

(5.θ − (α3 − α4 P 2 ) v0.θθ ] + 2 P [β1 v0 + β2 v0 + P 2 β3 v0. Again besides decay or an inﬁnite growth of s one can describe its selection (from below and from above) to the values ss obtained from the equation .θ + (α3 − 6α4 P 2 )v0.87) into (5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 160 Ampliﬁcation of Nonlinear Strain Waves in Solids and satisﬁes boundary conditions (5. α2 H3 (s)sX = s (α1 s − 1)(α3 − α4 s)(6α3 β2 + [5α2 α3 β1 − 6β2 (α4 + α1 α3 ) − α2 β3 ]s − [5α2 α4 β1 − 6α1 α4 β2 − α1 α2 β3 ]s2 ).18.X + (α3 − 2α4 P 2 )v0.88) now accounts for a solitary wave with parameters varying with respect to X. with 2 F = 2P [α1 v0. 1. (5.86) we have a nonlinear O.February 11.θθX ] + 2 PX [α1 v0 + α2 v0.θθθ = F.θθθ = 0. for v0 in the leading order. Substituting (5.91) Analysis of the solutions of Eq.89) (5.E.θθ ]. see Fig.θ − 2α2 (v0 v1 )θ − (α3 − α4 P 2 ) v1.(5. p2 = 2α2 P 2 4P 2 (α3 − α4 P 2 ) (5. (1 − α1 P 2 )v0.89) may be symmetric or asymmetric with respect to its core (or maximum) at diﬀerent points in time.82). v0 = 3(1 − α1 P 2 ) 1 − α1 P 2 cosh−2 (pθ).X + α2 v0.θ − 2α2 v0 v0.91) may be done similar to that in the previous section. Its solution.83) yields the equation for the function s = P 2 . (1 − α1 P 2 )v1. Depending upon the function P (X) the solitary wave (5. The evolution of the solitary wave may be described solving the next order problem where an inhomogeneous linear equation holds.90) Like in previous section the solvability condition (5.D. where H3 (s) = α1 α4 (α1 α3 + 5α4 )s3 − 2α4 (11α1 α3 + 10α4 )s2 + 2 3α3 (5α1 α3 + 17α4 )s − 30α3 .

7 Selection of the solitary waves governed by the solution (5. (5.February 11.2 v 1.85). 5.91): (a) from below (b) from above.89) are the same as well as the conditions required for the selection.92) coincides with Eq. .25 1 0. [5α2 α4 β1 − 6α1 α4 β2 − α1 α2 β3 ]s2 − [5α2 α3 β1 − 6β2 (α4 + α1 α3 ) − α2 β3 ]s− 6α3 β2 = 0.75 0.6 0. (5.4 0. It means that parameters of the selected solitary waves (5. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Inﬂuence of dissipative (active) external medium 161 v 0.76) and (5.5 1.5 0.92) We see that after substitution s = 1/c2 we get H3 (s) = Q6 (c)/c6 while Eqs.25 10 20 30 40 50 x -20 -10 10 20 30 40 50 x Fig.2 -20 -10 -0.89). (5. (5.

5. Fig.7(a) and Fig. β3 = 0. Fig.1 where the governing equation corresponds to Eq.7(b). 4. 5. The parameters are chosen so as an initial proﬁle is symmetric. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 162 Ampliﬁcation of Nonlinear Strain Waves in Solids In Fig. 5. We have studied how the external tangential dissipative (active) forces may support an evolution of the long strain solitary waves in an elastic rod. The conditions are found analytically when the selection of the solitary wave occurs both from below and from above. In experiments only narrowing part of the rod was observed where a conversion to an asymmetric wave occurs.(5. or decreases. it transforms back into the symmetric wave during the selection both from below and from above as shown in the last two stages in Fig. it was found analytically and in experiments the formation of asymmetric strain solitary wave in a narrowing rod. see Sec. β2 = 0.7 it is shown the temporal evolution of the solitary wave (5. We see that the initial wave increases.91). 5.89) whose amplitude and velocity vary according to Eq. Previously. and there is no solitary wave selection.7(b).(5. 5.February 11. However. It is shown that both asymptotic solutions predict evolution into the symmetric selected solitary waves.74) with β1 = β1 (x). into the asymmetric solitary wave.7(a). .

New variable has been introduced.. Engelbrecht et.February 11. Maugin (1990). temperature etc. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter 6 Bulk active or dissipative sources of the ampliﬁcation and selection Now the attention is paid on the situations when active/dissipative forces act inside a solid body. Dissipation function is included into variation formulation through its elementary work. The last equation is usually parabolic and is obtained phenomenologically. Maugin (1999). This modelling is connected with the concept of internal variables Engelbrecht and Braun (1998). microdisplacement.g. diﬀusive). al (1999). Various phenomenologocal model for viscoelastic biologocal media are discussed in Alekseev and Rybak (2002).). elastic stress or strain may be considered as observable variables in the usual sense of the word.e. According to this theory. Another idea has been used in Maugin (2000) to develop a variation formulation in thermoelasticity. into the relation for the free energy density. Thus Lord Rayleigh (1945) involved so-called dissipation function describing forces which depend upon temporal derivatives of the variables. called thermacy. Its behavior is described by the hyperbolic equation. In order to compensate our lack of description of the internal structure of material or the heat processes. Engelbrecht and Maugin (1996).. Another approach is of phenomenological kind. another variables are introduced called internal variables. 163 . Maugin and Muschik (1994). Maugin (1999). It is known that variation formulations used before in the book cannot be applied in the presence of volume dissipative eﬀects unless some modiﬁcations are proposed. Application of the internal variables to the wave problems may be found in Engelbrecht (1997). They are usually governed by the equations of parabolic type (i. The coupled governing equations are derived in the form of a hyperbolic equation of motion (or the equation with main hyperbolic part) and the equation for a variable responsible for dissipation (e.

Recently in a series of papers Cermelli and Pastrone (1997). waves in microstructured solids. All processes are governed by coupled dissipative nonlinear partial diﬀerential equations. Maugin and Muschik (1994). Eringen (1968). Engelbrecht (1997). waves in a medium with moving defects. Mindlin (1964). 6. Only a few works are devoted to the nonlinear waves in microstructured non-dissipative media Engelbrecht and Braun (1998). Strain waves in microstructured medium were studied mainly in the linear approximation Eringen (1968). Also experiments were performed in Potapov and Rodyushkin (2001). al (1999) the inﬂuence of dissipation on a shock propagation was studied in the one dimensional case while inertia of the microstructure was neglected into account together with the gradient of the microﬁeld . al (1973b) where attempts to measure them were done. Later numerical simulations were performed to account for the evolution of . e. Erofeev and Potapov (1993). however. Most results belong to the linear theory of elasticity. Engelbrecht (1979). Eringen and Suhubi (1964). Engelbrecht et. there are ﬁndings in the ﬁeld of the nonlinear theory Engelbrecht and Braun (1998). Erofeev and Potapov (1993). The theory of microstructures has been developed recently. Engelbrecht and Braun (1998). Eringen (1968).. see Capriz (1989). Sillat (1999) while the inﬂuence of the dissipative microstructure on the evolution of non-linear waves has been discussed in Cermelli and Pastrone (1997). We do not consider here viscoelastic materials describing by the integro.February 11. Maugin and Muschik (1994). namely. see about it. Christensen (1971). and thermoelastic waves. Strain waves may be useful in developing a possible method to estimate the microparameters since shape. Engelbrecht and Khamidullin (1988) The inﬂuence of dissipation/accumulation may be described by various methods. seismic waves.g. Mindlin (1964) and references therein quoted. see Engelbrecht (1983) and references therein. Engelbrecht and Braun (1998).1 Nonlinear bell-shaped and kink-shaped strain waves in microstructured solids The classical theory of elasticity cannot account for eﬀects caused by the microstructure of a material. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 164 Ampliﬁcation of Nonlinear Strain Waves in Solids Below some active/dissipative problems are considered.diﬀerential equations. An important problem is the lack of data on the microstructure parameters. amplitude and velocity of the strain wave can carry informations about the microstructure. but a few works can be mentioned Savin et.

all variables depend upon the coordinate x and the time t. and the selection of quasistationary dissipative solitary waves is found. Now and in the . The simultaneous inﬂuence of the accumulation/dissipation on the evolution is studied. we assume the microdisplacement depends linearly on the microcoordinates xJ : UJ (xJ . (ii) the macromotion if small but ﬁnite. ΓXXX = ψx . t). EIJ = ∂I UJ − ψIJ .1.). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 165 periodic waves Sillat (1999). Then the geometrical nonlinear2 ity is described by the only macrostrain CXX = Ux + 1/2Ux . Engelbrecht et. al (1999) is modiﬁed including both the inertia of the microstructure and the gradient of the microﬁeld . Engelbrecht et. CIJ . ΓIJK = ∂I ψJK . ψ = ψXX . while the Murnaghan model is valid to account for the physical nonlinearity. ﬂuid with micro-bubbles. the fundamental strains are given by: the Cauchy-Green macrostrain tensor .1 Modelling of a microstructured medium with dissipation/accumulation Let us recall some basic ideas following Eringen (1968). t) = xK ψKJ (x. It is shown that it is able to describe both the bell-shaped and kink-shaped longitudinal strain solitary waves. thus avoiding dependence upon microcoordinates. liquid crystals. Let us assume the following: (i) the movement is one dimensional. Hence the microdisplacement gradient is given by ∂I UJ = ψIJ . We suppose that the material particle contains discrete micromaterial elements whose displacements are accounted by the microdisplacement vector with components UJ (xJ . etc. Mindlin (1964). and the microdisplacement gradient. t). A possibility of the estimation of the microstructure parameters is proposed on the basis of the results obtained here. The motion of a material particle is characterized by the displacement vector with components UJ (xJ . Following Eringen (1968). As a result the nonlinear PDE with dispersion and dissipative (active) terms is derived. 6. In a reference conﬁguration. The model discussed in Cermelli and Pastrone (1997). t).February 11. if one needs a model to describe wider class of phenomena (solids with aﬃne microstructures. the distortion tensor. EXX = Ux − ψ. Below we follow Porubov and Pastrone (2001). This kinematical model is valid for particular families of microstructures only. al (1999). we refer to the references given in Cermelli and Pastrone (1997). (iii) the microstructure is suﬃciently weak to be considered in the linear approximation. Mindlin (1964).

We would like to consider a more general case to account also for the energy inﬂux to the strain wave caused by the microstructure. (6. Maugin and Muschik (1994). al (1999).2) where ρ is the macro-density. (6.(6. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 166 Ampliﬁcation of Nonlinear Strain Waves in Solids following small lower indices denote diﬀerentiation. they are not rigorously deduced within the . Certainly. constitutive equations (6. τeq = D(Ux − ψ). F and f have the same dimensions. (6. the simplest extension of the Hook law to the viscoelastic media Alekseev and Rybak (2002). Let us remark that the Voigt model accounts for the inﬂuence of the dissipation only. Engelbrecht et. e.4) may be of diﬀerent signs. Assume the following representations in the general case. Engelbrecht (1983). see.4) Note that A and a. ρUtt = σx + τx . Hence the coeﬃcients in Eq. I characterizes micro inertia. σ = σeq +AUxt +aψt . Then the one dimensional governing equations may be written as in Mindlin (1964).1) (6. In dissipationless case components σeq . ηIJK = .4) are of phenomenological kind. ηeq = Gψx . β = 3/2(λ + 2µ) + l + 2m. Dissipation inequality imposes some restrictions on the involved parameters Cermelli and Pastrone (1997).g. The reason of our assumption may be seen considering the linearized case in absence of the microstructure. I ψtt = ηx + τ. Bland (1960).February 11. D and G are constant parameters. Murnaghan (1951) is reduced to: 2 σeq = (λ + 2µ)Ux + βUx . Then we have for σ: σ = (λ + 2µ)Ux + AUxt . τIJ and ηIJK are deﬁned through the derivatives of the free energy W . Our model may be considered as a generalization of the Voigt model of microstructured solids. σIJ = ∂W ∂W ∂W . τeq and ηeq of the tensors σIJ . η = ηeq +F Uxxt +f ψxt .. B and b. µ are the Lame coeﬃcients. τ = τeq +BUxt +bψt . ∂CIJ ∂EIJ ∂ΓIJK which in the one dimensional limit Mindlin (1964). (iv) the external forces are negligible. τIJ = . since strictly speaking.3) Here λ. that relates to the Voigt model.

5) I ψtt = D(v − ψ) + Bvt + bψt + G ψxx + F vxxt + f ψxxt . the macro strain v is elastic and does not exceed the yield point.7) γ . but we do not want to go further in details here. a = da∗ .(6. G∗ having the dimension of stress. λ + 2µ λ + 2µ (6. Using dimensional analysis.6): (a) ε = V << 1 accounting for elastic strains. hence.as a scale for v and ψ.4) into Eqs. (6. F = dF ∗ . we obtain the coupled equations ρvtt = (λ+2µ+D)vxx −Dψxx +β( v 2 )xx +(A+B)vxxt +(a+b)ψxxt . one can assume G = p2 G∗ . B = dB ∗ . V << 1.4) can be seen as a particular case. Since the Murnaghan model is chosen. (b) δ = p2 /L2 << 1.February 11. (6. Then the dimensionless governing equations are (we keep the notations for variables): vtt − (1 + D D β )vxx + ψxx = ε ( v 2 )xx + λ + 2µ λ + 2µ λ + 2µ (A∗ + B ∗ )c0 (a∗ + b∗ )c0 vxxt + ψxxt .6) Further simpliﬁcations follow considering only long waves with characteristic length L >> 1.3). we are dealing with those elastic materials whose yield points are small. Then three positive dimensionless parameters appear in Eqs. The microinertia term I 0 depends upon the square of a typical size p of a microstructure element. V . where d has the dimension of a length. (c) γ = d/L. characterizing the ratio between the microstructure size and the wave length. f = df ∗ . (6.1). (6. Then I may be rewritten as I = ρp2 I ∗ . (6. L/c0 . Moreover. Equation (6. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 167 general framework of Rational Continuum Mechanics. obtained via proper additional assumptions.(6.2) and introducing the functions v = Ux and ψ as unknown variables. Substituting Eqs. Let us introduce L as a scale for x. In the following a more general constitutive equations will be derived according the basic assumptions and methods of Continuum Mechanics.5). The inﬂuence of dissipation/accumulation may be conveniently described supposing A = dA∗ . as a scale for t. characterizing the inﬂuence of the dissipation. the characteristic strain magnitude V is also small. c2 = (λ + 2µ)/ρ as a characteristic velocity. (6. b = db∗ . I ∗ being dimensionless.

B ∗ . ψ2 = vxx − vtt . the only equation for v is of the form vtt − vxx − ε β (A∗ + B ∗ )c0 ( v 2 )xx − γ vxxt = 0.. (6.February 11. 2 D D2 b∗ (B ∗ + b∗ )c2 0 vtt . If we expand the solution of Eq. (6.12) with Eq. λ + 2µ λ + 2µ λ + 2µ [(F ∗ + f ∗ )D + (B ∗ + b∗ − a∗ )G∗ ]c0 . α7 = . The . α4 = I ∗ . a∗ . D ρD (6. α2 = .. α3 = . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 168 Ampliﬁcation of Nonlinear Strain Waves in Solids Dψ = Dv + γc0 [B ∗ vt + b∗ ψt ] + δ [G∗ ψxx − (λ + 2µ)I ∗ ψtt ] + γδc0 [F ∗ vxxt + f ∗ ψxxt ] .(6.11) ψ4 = D2 ψ0 = v.8) In absence of the microstructure. by substituting (6. α5 = (λ + 2µ)D I ∗ (a∗ − B ∗ − b∗ )c0 a∗ (B ∗ + b∗ ) α6 = .(6. t). (6.9) we see that the inclusion of the gradient of microdistortion provides the dispersion vxxxx .8) in the form : ψ = ψ0 + γ ψ1 + δ ψ2 + γδ ψ3 + γ 2 ψ4 + . F ∗ . while the inertia of the microstructure gives us mixed dissipation and dispersion terms. where α1 = β (A∗ + a∗ )c0 G∗ . the governing nonlinear PDE for the macrostrain v(x.10).(6.12) (6.9) Assume dissipation is weak. ψ1 = we obtain. and δ are equal to zero.(6. λ + 2µ λ + 2µ (6. with G∗ (λ + 2µ)I ∗ (B ∗ + b∗ )c0 vt . b∗ .7). D. G. vtt − vxx − εα1 ( v 2 )xx − γα2 vxxt + δ(α3 vxxxx − α4 vxxtt )+ γδ(α5 vxxxxt + α6 vxxttt ) + γ 2 α7 vxxtt = 0..11) into Eq.10) Comparing Eq. D D D [(F ∗ + f ∗ )D + (B ∗ + 2b∗ )G∗ ]c0 (λ + 2µ)(B ∗ + 2b∗ )c0 I ∗ ψ3 = vxxt − vttt .

This is the reason we retain terms quadratic in these parameters in the expansion of the solution (6.February 11. another case corresponds to the simultaneous balance between dispersion.1–1. α3 /α4 } or when 0 < c2 < min{1. The kink-shaped localized travelling structure may be sustained by diﬀerent balances.1.5. c is a free parameter. k2 = c2 − 1 . one possibility occurs when nonlinearity is balanced by dissipation (or accumulation). we have the nondissipative case governed by the double dispersive equation.10). vtt − vxx − ε(α1 ( v 2 )xx − α3 vxxxx + α4 vxxtt ) = 0. α1 (6. as it will be explained in the following. 1. 6. There are two main types of nonlinear travelling solitary waves which could propagate keeping its shape. 3. If in addition γ = 0. γ and δ. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 169 evolution of nonlinear strain wave depends upon the ratio between parameters ε.16) Hence. It satisﬁes the boundary conditions ∂k v → 0 for |x| → ∞.14) where θ = x − ct. Sometimes they can be considered ” negligibly small”. while in the second case only compressive waves propagate.13) Its exact bell-shaped travelling solitary wave solution arises as a result of balance between nonlinear and dispersive terms. 4ε(α4 c2 − α3 ) (6. . according to the diﬀerent eﬀects we want to point out.15) (6. In the ﬁrst case longitudinal tensile waves propagate.15) exists when c2 > max{1. ∂xk and takes the form (see section 3): v= 6k 2 (α4 c2 − α3 ) cosh−2 (k θ). (6. Typical shapes of the waves are shown in Figs. k = 0. sometimes-not. bell-shaped and kink-shaped solitary waves. 2.2 Bell-shaped solitary waves The balance between nonlinearity and dispersion takes place when δ = O(ε). The bell-shaped solitary wave usually appears as a result of a balance between nonlinearity and dispersion. the solitary wave (6. nonlinearity and dissipation (or accumulation). α3 /α4 }. 1.

2 (δα4 − γ 2 α7 )c4 − [δ(α3 + α4 ) + 24γ 2 α2 ]c2 + δα3 = 0.. T ) + γv1 (θ. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 170 Ampliﬁcation of Nonlinear Strain Waves in Solids In general.18) we have in the leading order 2 (c2 − α1 )v0.17) exists only for particular values of the coeﬃcients in such equations. Parkes and Duﬀy (1996).February 11. The solution of Eq. with θx = 1. (6.13).(6. Eq.18) is sought in the form v = v0 (θ. γ << 1. Substituting (6. θt = −c(T ).12) possesses an exact travelling bell-shaped solution vanishing at inﬁnity Kudryashov (1988). Then from (6.19) ∂2 α2 v − ε(α5 + α6 c2 ) vθθ ∂θ2 + (6.14) at |θ| → ∞. v=− where k=± 2γcα2 .20) . 1..19) into (6.θ − εα1 ( v0 )θ + ε(α3 − α4 c2 ) v0. hence. Eq.17) εα1 α4 = α4 + ε2 /δ α7 . When δ = O(ε). the exact solution (6. T ). .(6..(6. (6. T ) + . the phase velocity c is deﬁned from the overdetermined system of bi-quadratic equations.18) where vi . (6. i = 0. T = γ t..12) is considered as dissipation perturbed double dispersive equation (6..θθθ = 0. The asymptotic solution is sought as a function of the phase variable θ and the slow time T . 2 (δ α4 2 + 16γ 2 α2 α6 )c4 + (16γ 2 α2 α5 − 2δα3 α4 )c2 + δα3 = 0. satisfy boundary conditions (6.12) we get that (c2 − 1)vθ − εα1 ( v 2 )θ + ε(α3 − α4 c2 ) vθθθ = γ 2c[vT − εα4 vθθT ] + cT [v − εα4 vθθ ] − c O(γ 2 ). v = v(θ. δ(α3 − α4 c2 ) 60ck 3 γδ (α5 + α6 c2 ) cosh−2 (k (x − ct))[tanh(k (x − ct)) ± 1].

23) q1 = 5α6 .23) without integration. Hence the solitary wave in microstructured solids may be also selected.θθθ = F.24) Important features of the behaviour of s may be established analyzing Eq.20) and boundary conditions at inﬁnity one can obtain the solvability condition for Eq. with 2 Q3 (s) = 30α4 s3 + 3α4 (17α3 + 5α4 ) s2 + 2α3 (10α3 + 11α4 ) s − α3 (α4 + 5α3 ).(6. Indeed.21). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 171 The exact solitary wave solution of Eq.25) .(6. (6. the most interesting evolution of s is realized when s (and hence the amplitude of the solitary wave) tends to the ﬁnite constant value s∗ as T → ∞. ∂θ2 (6. the values of s∗ are the real positive solutions of equation q1 s2 + q2 s + q3 = 0.22) which yields the equation for the function s = c2 .21) is adjoint to the operator 3 M A = (c2 − 1)∂θ − 2εα1 v0 ∂θ + ε(α3 − α4 c2 ) ∂θ . 7ε sT Q3 (s) = 2s (s − α1 )2 (q1 s2 + q2 s + q3 ). where F is F = 2c[v0.(6. −∞ (6. ∞ v0 F dθ = 0.θθ . Then using (6.θ − 2εα1 (v0 v1 )θ + ε(α3 − α4 c2 ) v1.θθT ] + cT [v0 − εα4 v0.15).21) c The operator M acting on the function v1 in Eq. The ﬁrst order term v1 in the solution (6.20) has the form (6.19) obeys the inhomogeneous linear equation (c2 − α1 )v1. (6. q3 = 7α2 α3 − 5α5 .February 11. (6. There may be decay or inﬁnite growth of the initial velocity and the amplitude of the solitary wave (6.T − εα4 v0. However. q2 = 5α5 − 7α2 α4 − 5α6 .θθ ]− ∂2 α2 v0 − ε(α5 + α6 c2 ) v0.(6.15) with c = c(T ).

It is relevant to notice that the solution (6. and the permitted interval deﬁned from (6.(6. even if it does not modify the behaviour of the wave near its core. There are no free parameters in the solution.16).1. One can see that the solitary wave keeps its localized bell-shaped form. attenuation and selection of the bell-shaped solitary wave v = v0 is shown in Fig. Thus the stationary values of the solitary wave parameters are prescribed by the equation coeﬃcients. and sQ < s1q the ampliﬁcation of the solitary wave with s∗ = s1q occurs at q1 > 0 if sQ < s0 < s1q while at q1 < 0 it ampliﬁes by s∗ = s2q if s1q < s0 < s2q . Ampliﬁcation. the wave amplitude increases. the attenuation to s∗ = s1q happens for the waves with s1q < s0 < sQ .February 11. while for negative q1 the double selection is realized.19) is not uniformly valid: the matching asymptotic expansions method Ablowitz and Segur (1981) should be applied to complete the solution. If sQ > s2q there is no double selection. in the form 3 v = i=0 A i tanhi (k θ). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 172 Ampliﬁcation of Nonlinear Strain Waves in Solids Assume real roots of Eq. We denote by s0 the initial value of s while the real root of Q3 is sQ .17. (6. the attenuation is provided by simultaneous decrease of the amplitude and the increase of the width. Similarly. the wave evolution is similar to the case sQ < s1q .26) with three possible sets of parameters A i . s2q ] may be analyzed in the same manner. Parkes and Duﬀy (1996). and additional restrictions on the equation coeﬃcients are . When s1q < sQ < s2q there is no selection for q1 > 0. The sign of sT needed for s → snq depends upon the sign of Q3 (s) around s = snq . 6.3 Kink-shaped solitary waves The equation (6. When sQ is the only real root of Q3 .25) are s1q < s2q . but waves with sQ < s0 < s2q increase up to s∗ = s2q . k . The attenuation of the wave to s∗ = s1q happens when q1 > 0. in case q1 < 0 wave with initial velocity s0 > s2q decreases to s∗ = s2q . The situation when Q3 has three real roots within or outside the interval [s1q . Thus. the sign of q1 .12) possesses also exact travelling kink-shaped solution Kudryashov (1988). The addition of correction γv1 does not change signiﬁcantly the proﬁle of v = v0 . while the waves with s0 > s2q attenuate by s∗ = s2q . 1. In the case of ampliﬁcation. while the width decreases. waves with s0 < s1q amplify up to s∗ = s1q . s1q < s0 < s2q .

α1 2εα1 (6. Then Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 173 needed.. the nonlinearity is balanced by the dissipation/accumulation only. α1 (6.27) and all derivatives of vi with respect to θ vanish at inﬁnity.31) (6. For a kink h+ = h− .33) The alterations of the kink shape in absence of higher order dissipative/active terms. m= (h− − h+ )α1 2 .29) (6. v1d = 2(α4 c2 − α3 )m2 cosh−2 (mθ) log sech(mθ). depend upon the sign of (α4 c2 − α3 )/α1 . Typical proﬁles of v = v0 + δv1d with v1d deﬁned by (6. for θ → ±∞. (6. εα1 (6.12) may be rewritten as vtt − vxx − ε(α1 ( v 2 )xx + α2 vxxt ) = −δ(α3 vxxxx − α4 vxxtt )− εδ(α5 vxxxxt + α6 vxxttt ).1 for diﬀerent values of α3 with other parameters values ﬁxed: .e.(6. In the leading order the kink solution has the form v = A m tanh(m θ) + B. 2cα2 Next order solution v1 consists of two parts.February 11.28) (6. where vi = vi (θ = x − ct) satisﬁes the boundary conditions v0 → h± . B= .30) There are two free parameters.32) are shown in Fig.. ε2 < δ < ε and γ = O(ε). v1 = v1d + v1a . i. vi → 0. c = 1 + εα1 (h+ + h− ). α5 = α6 = 0. with A=− α2 c c2 − 1 . i > 0. the phase velocity c and the wave number m which are deﬁned from the boundary conditions. whose solution is sought in the form v = v0 + δv1 + . 6. where dispersive perturbation of the kink vanishing at inﬁnity is accounted for the solution.32) while higher order dissipative/active terms contribution is v1a = 2(α5 + α6 c2 )cm3 cosh−2 (mθ) (3 tanh(mθ) − 2mθ) .. When dispersion is weak.

30) when there is no dispersion. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 174 Ampliﬁcation of Nonlinear Strain Waves in Solids v 4 3 2 1 -6 -4 -2 -1 v 3 2 1 -6 -4 -2 -1 v 3 2 1 -6 -4 -2 -1 2 2 2 a v 3 2 1 b 4 6 x -6 -4 -2 -1 2 4 6 x c v 3 2 1 d 4 6 x -6 -4 -2 -1 2 4 6 x e 4 6 x Fig. as shown in Figs.5. 6.1(b) the value is α3 = −0. 6. correspondingly.1(d. In Fig. 6.February 11.1. One can note the similarity with the proﬁle in Fig. 6. m = 1.5.1(c) where α3 = 2. In contrast to the bell-shaped wave now the ampliﬁcation is accompanied by the alteration of the wave proﬁle. . and we have the mirror proﬁle of those shown in Fig.1 Inﬂuence of weak dispersion on the shape of the kink-shaped wave. In Fig.5. ε = 0. 1. c = 1.2 it is shown what happens with the kink (6. where α3 = −2. while undisturbed kink appears in Fig. Fig.5 of the exact kink-shaped solution of DMKdV equation. α2 = −1.25.1(a. 6. Then the disturbances develop on another ”side” of the kink. 6.1(a).e) with α3 = 5. 6. b). 6. 7.5. α4 = 1. The ”nonsymmetric” disturbances of the kink shape are seen in Fig. Shown by dashed line is the unperturbed kink. they become weaker when α3 tends to zero.1 shows the ampliﬁcation of the kink-shaped wave since the diﬀerence between its maximum and its minimum is larger than in the undisturbed case. α1 = 1.5. δ = 0.

Fig.February 11. γ = O(ε) corresponds to the simultaneous balance between nonlinearity.5.2(c) where there are no disturbances. dispersion and dissipation/accumulation. it varies from α6 = −1. the disturbances of the kink shape are ”symmetric” independently of the sign of α6 . α5 = 1. m = 1. First. α2 = −1.5. 6. 6. c = 1.2 nﬂuence of weak higher order dissipation on the shape of the kink-shaped wave . there are no mirror proﬁles arising at distinct signs of α6 .5. Shown by dashed line is the unperturbed kink.5.1. ε = 0. Fig. We have α6 = 0.5. 6. Two main distinctions may be mentioned in comparison with proﬁles shown in Fig. 6.5. 6. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 175 v 3 2 1 -6 -4 -2 -1 v 3 2 1 -6 -4 -2 -1 v 4 3 2 1 -6 -4 -2 -1 -2 2 2 2 a v 3 2 1 x b 4 6 -6 -4 -2 -1 v 3 2 1 2 4 6 x c d 4 6 x -6 -4 -2 -1 2 4 6 x e 4 6 x Fig.2(a) α6 = −2.1.2(d). δ = 0. 6.5 in Fig.2(b) to α6 = −0. Second. α3 = α4 = 0. perturbed . The case δ = O(ε).5.33) correspond to α1 = 1. The proﬁles of v = v0 + δv1a with v1a deﬁned by (6.2(e) corresponds to α6 = 2.5. 6. while Fig. In Fig.

C= . Then at order ε the ODE equation for v0 is. 2 and the solution exists under special boundary conditions.28).37) It follows from (6. (6. 50α1 (α4 − α3 ) 25α1 (α4 − α3 ) α1 10(α3 − α4 ) (6. 2 2c1 v0.29) hold. v0 = A tanh(m θ)sech2 (m θ) + B tanh(m θ) + C.36) possesses the exact kink-shaped solution VliegHultsman and Halford (1991). Thus inertia yields mixed derivative terms vxxtt .35) (6.12).34) where boundary conditions (6.36) Equation (6.. Substituting (6.34) in the leading order we obtain the D’Alembert equation. Consider only one wave travelling to the left and assume v depends upon phase variable θ = x − ct with c = 1 + εc1 + ε2 c2 .... Dispersion is required for the existence of the bell-shaped solitary waves in an elastic microstructured medium. in . B= .1.θθ + (α3 − α4 ) v0. with A= 2 2 3α2 3α2 c1 α2 . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 176 Ampliﬁcation of Nonlinear Strain Waves in Solids by the higher order dissipative/active terms.February 11.35) into (6. The solution of Eq. c1 = α1 (h+ + h− ). vxxttt . vtt − vxx − ε[α1 ( v 2 )xx + α2 vxxt − α3 vxxxx + α4 vxxtt ] = −ε2 (α5 vxxxxt + α6 vxxttt )..4 Concluding remarks We have found that various features of the microstructure provide corresponding terms in the governing equation (6.(6.34) is sought in the form v = v0 + εv1 + ε2 v2 .θθθ = 0. The dispersive term vxxtt determines.29) that h+ − h− = 2B.θ − α1 ( v0 )θ + α2 v0. (6. The inﬂuence of higher order terms may be studied in a similar way as done for the solution (6. while dispersion vxxxx and higher order dissipative/active term vxxxxt are due to the micro-deformation gradient. 6. m= .

The wave amplitude and velocity depend upon macro.and micro-properties of the microstructured medium through the analytical relationships.(6. If an√ other wave travels with an initial velocity s02 . explicitly given above.12) allow to describe in an explicit form the ampliﬁcation of both types of the waves. (6. Hence. see Fig. Dispersion terms also account for the alterations in the kink-shaped wave proﬁle. Certainly the solution cannot describe an energy exchange between the waves but it gives us the range of . ε. b∗ and F ∗ . namely two stationary ﬁnite velocities of the solitary wave. Hence. The relationships among these parameters deﬁne the thresholds that separate the parameters of the initial solitary waves which will be ampliﬁed or attenuated. f ∗ appear in the expressions for the equation coeﬃcients only in combinations B ∗ + b∗ . one could reduce the number of the microstructure parameters in the model (6. while higher order dissipative/active terms are responsible for a saturation which prevents unbounded growth of the bell-shaped solitary wave. We see that pairs B ∗ .(6. 6. According to this ratio the governing equation (6.19) describing bell-shaped solitary wave selection may help to explain transfer of the strain energy by the microstruc√ ture.4). The solutions of Eq.shaped solitary waves. δ and γ. Let us assume one single solitary wave with initial velocity s01 . when its parameters tend to the ﬁnite values prescribed by the coeﬃcients of the governing equation. since a∗ appears in the expression for α4 independently from A∗ . while the higher order dissipative/active term α6 vxxttt provides nonzero q1 in Eq. An application of the results obtained here consists in a possible estimation of the microstructure parameters on the basis of the strain wave behaviour. At the same time there is a need of both A and a.February 11.15).12) may describe either bell-shaped or kink.16)).1. F ∗ + f ∗ . it looses its energy which is absorbed by the microstructure. it will be ampliﬁed and it means that we need a source of energy to justify the ampliﬁcation. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 177 particular. the existence of either exact compression or tensile solitary wave solutions (see Eqs. as well as the selection of the solitary wave. so that it will be attenuated propagating in the microstructured medium. In principle. allows to estimate the size p of the microstructure and the parameter d which inﬂuence the dissipation/accumulation eﬀects. The asymptotic solution (6.25). the measurements of the solitary wave amplitude and velocity allow to obtain the parameters of the microstructure using elastic macro-moduli known beforehand. The ratio between the small parameters. A possible explanation could be that the energy stored by the microstructure is released because of the passing wave.(6.

recently the phenomenological theory has been developed in Engelbrecht (1997).1 Nonlinear seismic solitary waves selection Modelling of nonlinear seismic waves The inﬂuence of microstructure may explain phenomena caused by the energy input/output. 2. (6. a3 are positive constants and ε is a small parameter. (6. a2 .2. Additional energy inﬂux yields an ampliﬁcation of the wave.. ut + u ux + d uxxx = εf (u).1. They may be considered as short-lived objects which are able to absorb energy from a surrounding medium. Engelbrecht and Khamidullin (1988) to account for the seismic waves propagation in a horizontal layer.g. 6. Then the evolution of an initial arbitrary pulse has been studied numerically. e. and the dilaton breaks generating . Like here its single solitary wave asymptotic solution has been obtained in Sec. while propagating seismic waves may release the locked-in internal energy. and the conditions were found for the decay or a selection of a single solitary wave. 2. These ﬂuctuations are called dilatons.38) where f is the body force related to the so-called dilation mechanism.3. Accumulation of the energy may happen only up to a certain threshold value. It was found that the initial pulse splits into a sequence of solitary waves but each of them evolves according to the single solitary wave analytical solution. then it is released. The basic idea of the seismic waves modelling is originated from the dilation theory in fracture mechanics Zhurkov (1983). Sec. f (u) = − a1 u − a2 u2 + a3 u3 .38) may describe an appearance of microseisms.2.2 6. Eq. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 178 Ampliﬁcation of Nonlinear Strain Waves in Solids the microstructure parameters when the energy transmission is possible.39) a1 . The predictions of the asymptotic solution may be valid even in a more complicated unsteady process of the formation of solitary waves from an arbitrary input. It was assumed there that negative density ﬂuctuations play an essential role in the strength of solids. The internal energy is stored in a geophysical medium. Thus. (6. see. It was proposed to describe longitudinal strain waves evolution by the nonlinear equation.February 11. where the equation is considered rather similar to ours.

Hence it was assumed in Koz´k and Silen´ (1985) to consider a y a medium as a two-dimensional homogeneous space containing a linear inhomogeneity compressed uniaxially. were obtained in Nikolaev (1989) to clarify the role of the simultaneous inﬂuence of nonlinearity and dissipation on the seismic waves evolution. ˇ The theory developed in Gusev (1988). The necessary condition for the fracturing of the medium under load is the existence of an inhomogeneity like a tectonic fault. mainly qualitative. and dispersion. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 179 a crack.38) in Engelbrecht (1997). Engelbrecht and Khamidullin (1988) demonstrate transformation of an ini- . increases until the stress ﬁeld achieves a threshold. while there is an inﬁnite growth in a pure active case. the most important contribution to nonlinear description of the seismic waves has been done in Engelbrecht (1997). it was assumed that the mechanism of seismic radiation is connected with rapid dilatancy variations. an ˇ inclusion etc. Derivation of Eq. Engelbrecht and Khamidullin (1988). In particular. which is the structure that simulates commonly occurring geological faults subjected to tectonic stress with a predominant orientation. stored and released. The area. Qualitatively similar phenomena were recognized in Koz´k and a ˇ Silen´ (1985) studying seismic energy release to explain the earthquake y mechanism.(6.38) is the celebrated Korteweg-de Vries equation. Eq. aﬀected by the loading. whose exact travelling one-parameter solitary wave solution arises as a result of a balance between nonlinearity. Body force f plays a dissipative/active role destroying this balance.February 11. The layers are inhomogeneities where the energy is pumped. f = 0. Koz´k and Silen´ (1985) is a y linear. In absence of the body force. However. d uxxx . Engelbrecht and Khamidullin (1988) is based on a model where the classic elasticity basic equations are complemented by the inclusion of the body force to account for the dilaton mechanism. numerical results in Engelbrecht (1997).(6. u ux . and the phenomenological expression for the body force (6. The most interesting scenario happens in the mixed dissipativeactive case. Then a seismic-energy-releasing events occur. Engelbrecht and Khamidullin (1988) to consider the Earth crust as a certain hierarchy of elastic blocks connected by thin interface layers. When all terms in the expression for f are dissipative.39) closes the basic equations. the solitary wave decays. A similar dilatancy model has been proposed in Gusev (1988) to explain the nature of earthquake precursors. Preliminary results. Hence the interface layers behave like dilatons. In order to govern a medium that may store and release the energy it was proposed in Engelbrecht (1997). In particular.

42) Equation (6.2 Asymptotic solution of the governing equation Let us assume that ε << 1. Furthermore the function u depends upon a fast variable ξ and a slow time T .ξξξ − V u0. the unsteady process of the transformation of the KdV soliton into the solitary wave with prescribed parameter values is described analytically. (iii) when the parameter ε is not small happens in quantitative agreement with asymptotic solution. (6.February 11.41) (6.43) . but numerical simulations cannot describe the intervals of their values required for the appearance of the stable localized waves. In the leading order we have d u0.42) contains coeﬃcients V = V (T ). Then it is demonstrated numerically that solitary waves selection (i) in presence of the solitary waves interactions. T ) = u0 (ξ.(6. such as ξx = 1.40) is sought in the form: (6.. its exact solitary wave solution will have slowly varying parameters.38) becomes d uξξξ − V uξ + u uξ + ε uT + a1 u − a2 u2 + a3 u3 = 0. In order to obtain this information a procedure is developed below. a2 . The nature of the terms in f depend upon the values of the coeﬃcients a1 . (6. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 180 Ampliﬁcation of Nonlinear Strain Waves in Solids tial KdV soliton into a new stable localized bell-shaped wave with the amplitude and velocity prescribed by the equation coeﬃcients. (ii) when an initial proﬁle is arbitrary. The solution u of Eq.. Then equation (6.40) u(ξ. hence. 6. u0 = 12 d k(T )2 cosh−2 (k(T ) ξ) with V = 4dk 2 . T ) + ε u1 (ξ. T ) + .ξ = 0.2. Most of the results were ﬁrst published in Porubov et. First. T = ε t. ξt = −V (T ). al (2003). a3 .ξ + u0 u0.

105 (6. 105 (6.February 11.(6. A more quantitative description of the variation of Q can be given in order to see at what time the selected values are achieved. Hence parameters of the solitary wave tends to the ﬁnite values prescribed by the equation coeﬃcients ai and is selected. when Q1 < Q0 < Q2 parameter Q will grow up to Q2 . depends on the value of Q0 ≡ Q(T = 0).T = kT 2kT u0 + ξ u0. are 14a2 − 2 49a2 − 210a3 a1 14a2 + 2 49a2 − 210a3 a1 2 2 . 24a3 24a3 (6. k k ∞ (6. 0 0 Due to (6. −∞ (6.44) The solvability condition for Eq.45) that k obeys the equation kT = − 2 k 3456a3 d2 k 4 − 336a2 dk 2 + 35a1 . while if Q0 > Q2 it will decrease by Q2 .ξ . Equation (6.ξξξ − V u1.47) The roots of the equation. with F = − u0. Q2 = .43) u0. d u1. Q will diverge at Q0 < Q1 .48) The behavior of the solitary wave amplitude. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 181 In the next order an inhomogeneous linear diﬀerential equation for u1 appears.T + a1 u0 − a2 u2 + a3 u3 . Q.47) Q1 = . 24a3 Q2 − 28a2 Q + 35a1 = 0.44) is u0 F dξ = 0.45) Then it follows from (6.46) that may be rewritten in terms of the solitary wave amplitude Q = 12 d k(T )2 as QT = − 4 Q(24a3 Q2 − 28a2 Q + 35a1 ). Indeed.ξ + (u0 u1 )ξ = F.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

182

Ampliﬁcation of Nonlinear Strain Waves in Solids

may be directly integrated over the range (0, T ) giving the implicit dependence of Q on T : T = 35 32a3 Q1 Q2 (Q2 − Q1 )

Q2 log

(Q − Q2 ) Q (Q − Q1 ) − Q1 log + (Q2 − Q1 ) log (Q0 − Q1 ) (Q0 − Q2 ) Q0

(6.49)

One can see that T tends to inﬁnity when Q → Q2 , and expression (6.49) provides an analytical description of the time-dependent process of the parameter-value selection of the solitary wave (6.43). With Eq.(6.47) being taken into account, the solution for u1 is u1 = A1 [tanh(kξ) − 1] + [3A1 + 2A2 ξ] cosh−2 (kξ) + [C − 3kA1 ξ − A2 ξ 2 − A3 log(cosh(kξ))] tanh(kξ) cosh−2 (kξ), where C = const, A1 = 1152a3 d2 k 4 − 168a2 dk 2 + 35a1 , 35k (6.50)

A2 =

3456a3 d2 k 4 − 336a2 dk 2 + 35a1 1728a3 d2 k 3 , A3 = . 35 35

We see that u1 does not vanish at ξ → −∞, and a plateau appears behind a solitary wave. It may be of negative or positive amplitude depending upon the sign of A1 . A uniformly valid solution vanishing at ξ → −∞ may be obtained by the standard procedure described in Ablowitz and Segur (1981). One can make now some important conclusions. If we formally assume a2 = 0, a3 = 0, both the behavior of the solitary wave parameters and the sign of the amplitude of plateau are deﬁned by the sign of a1 . Indeed when a1 > 0, the amplitude and velocity of the wave decreases in time according to Eq.(6.47), while A1 > 0, and the plateau is negative. On the contrary, at negative a1 we have an increase of the wave amplitude and positive plateau. In general case the plateau may be negative both in case of an increase and a decrease of the solitary wave. We also see that the increase of the amplitude is accompanied by the decrease of the wave width.

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection

183

u

10

10

5

5

0 50 100 150

0

x

200

250

300

Fig. 6.3

Evolution of three solitary waves in absence of their interaction.

6.2.3

Numerical simulations

An asymptotic solution requires speciﬁc initial conditions while an evolution of an arbitrary initial disturbance as well as interactions between nonlinear localized waves are of practical interest. It may be described only numerically, however, it is important to know whether analytical predictions may be used for a design of numerics, since the behavior of the waves is sensitive to the values of the equation coeﬃcients and the initial conditions. We use for computations a pseudo-spectral method whose computation code was designed in Kliakhandler (1999). The program computes solutions of 1D scalar PDEs with periodic boundary conditions. It evaluates spatial derivatives in Fourier space by means of the Fast Fourier Transform, while

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

184

Ampliﬁcation of Nonlinear Strain Waves in Solids

u

10

10

0 50 100

0

x

150

200

250

Fig. 6.4

Selection of two solitary in presence of their interaction.

the time discretization is performed using the fourth-order Runge–Kutta method. This scheme appears to have a good stability with respect to the time step and was already successfully used for the modelling of the solitary wave selection in a convective ﬂuid, see Sec. 2.3.1. More detailed information about the code may be found in Kliakhandler (1999). We choose the parameter values identical to that used in numerics in Engelbrecht (1997): a1 = 1, a2 = 0.5, a3 = 0.0556, d = 0.5, ε = 0.1. Following the analysis from the previous section one obtains Q1 = 4.11, Q2 = 6.38, and the selection occurs for single solitary waves with initial amplitudes from the interval 4.11 < Q0 < 6.38. Numerical results for the single wave evolution conﬁrm analytical solutions and agree with the

February 11, 2004

20:10

WSPC/Book Trim Size for 9in x 6in

ws-book9x6

Bulk active or dissipative sources of the ampliﬁcation and selection

185

u

10

10

0 50 100

0

x

150

200

250

Fig. 6.5 Selection of one solitary wave and decay of another one in presence of their interaction.

numerical results in Engelbrecht (1997). Then the initial conditions are changed to the proﬁle containing three solitary waves each accounting for Eq.(6.43) at T = 0. First, the waves are located so as to avoid their interactions, see the ﬁrst stage in Fig. 6.3. The initial amplitudes are chosen so as the values of the amplitudes of the ﬁrst two solitary waves are brought into the selection interval, while the amplitude of the last one is below Q1 = 4.11. For convenience here and in the following ﬁgures thresholds 4.11 and 6.38 are shown by dashed lines at each stage. One can see in Fig. 6.3 that the amplitudes of the ﬁrst two solitary waves tend to the value Q2 = 6.38, while the last solitary

6. Hence each solitary wave evolves according to the asymptotic solution. 6. One can see in Fig.5. When the second initial solitary wave from Fig.6 waves. Evolution of an initial Gaussian proﬁle and formation of two selected solitary wave decays. First we take two larger initial solitary waves.4 that the interaction does not aﬀect the selection. 6.3 is moved behind the third one.6 that an initial Gaussian pulse produces a train of solitary waves of diﬀerent magnitude in agreement with the KdV theory. Let us re-arrange the initial positions of the solitary waves in order to include their interactions.February 11. while the smaller wave decays. One can see in Fig. 6. 6. The process is shown in Fig. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 186 Ampliﬁcation of Nonlinear Strain Waves in Solids u 10 10 5 5 0 50 100 150 200 0 x 250 300 350 400 Fig. . its selection occurs despite the interaction. and again both solitary waves evolve in agreement with the theory.

At larger ε = 10 the plateau almost disappears. At the same time we observe a decay of . the initial solitary wave amplitude Q0 = 5 still tends to the value Q2 = 6. Two main deviations are observed. vanish. First is the diﬀerence in amplitudes. 6. Finally.7 we see that the solitary waves continue to evolve according to the asymptotic solution with growth in ε. 6. Selection from above still occurs for Q0 > Q2 . it is caused by the increase of the contribution of u1 into the asymptotic solution. 6. The diﬀerence in the shape of plateau is shown in Fig. b)ε = 0. the inﬂuence of the small parameter value is studied. however.February 11.7 Comparison of asymptotic (dashed line) and numerical (solid line) solutions at various ε: a)ε = 0.8 where we see the decrease of its length.3. the selection of those solitary waves occurs whose amplitudes come to the selection interval prescribed by the theory. c)ε = 0. Then.1.38.5. see Fig.9. 6. d)ε = 1. Note that two leading solitary waves are selected from below while other solitary waves generated from the input. In Fig. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 187 a 7 6 5 4 1 3 5 7 9 7 6 5 4 1 3 b 5 7 9 u3 2 1 0 1 3 u3 2 1 0 x 5 7 9 1 3 x 5 7 9 c 7 6 5 4 1 3 5 7 9 7 6 5 4 1 3 d 5 7 9 u3 2 1 0 1 3 u3 2 1 0 x 5 7 9 1 3 x 5 7 9 Fig.

5 x 7 9 11 13 1 3 5 7 x 9 11 13 15 Fig. c)ε = 0. or a speciﬁc site in the .1 0 -0.3.4 -0. al (1996) and references therein.1 1 3 5 7 9 11 13 0.February 11.1 0 -0.2 -0.2 0. The simplest point defects in the crystal are the interstitial atom.1 0 -0. Q0 < Q1 = 4.1 1 3 5 7 d 9 11 13 15 u -0.1 1 3 5 7 9 11 13 15 0.2 -0. and the vacancy.3 -0.11.2 -0.5.4 -0.5 x 9 11 13 15 1 3 5 7 x 9 11 13 15 c 0. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 188 Ampliﬁcation of Nonlinear Strain Waves in Solids a 0. 6.1 0 -0.3 6.5 1 3 5 u -0.2 0.4 -0.2 0. d)ε = 1.8 Comparison of plateau in asymptotic (dashed line) and numerical (solid line) solutions at various ε: a)ε = 0.3 -0. the initial solitary wave with the amplitude less than Q0 = 3 that already diﬀers from the theoretical predictions.3 -0. Point defect is described as a distortion in a crystal lattice in the area equal to the atomic volume.4 -0.1 1 3 5 b 7 9 11 13 15 u -0.2 -0.3 -0. 6. or an atom occupying a position among the atomic equilibrium positions in an ideal lattice.2 0.3.1. b)ε = 0.1 Moving defects induced by external energy ﬂux Basic concepts and derivation of governing equations Recently it was found that point defects may be generated in a solid subjected to the laser radiation.5 1 3 5 7 u -0. see Mirzoev et.

2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 189 6 5 4 u3 2 1 0 16 18 20 x 22 24 Fig.51) where q0 is a velocity of the defects generation in absence of the strain.. laser beam) produces moving point defects.xx − βj nj .9 Evolution of an initial KdV soliton (dashed line) at ε = 10. Following Mirzoev et. t) be the concentration of defects of the kind j (j = v for vacancies. next . temperature and defect-concentration ﬁelds Mirzoev et. recombination and diﬀusion. z) at time t. Let us consider an isotropic solid where a concentrated energy inﬂux (e. y. Then the kinetics of the point defects is governed by the equation nj. Generation of defects due to the laser radiation may be explained using interaction of strain. lattice of the crystal where atoms are absent Kosevich (1981). al (1996).February 11.t = q0 + qε Ux + Dj nj. al (1996) assume nj (x. 6. j = i for interstitial atoms) at a point r(x.g. (6. The main processes responsible for the temporal evolution of a defect are generation.

Combinations of the point defects yield defect of the dipole kind that may be accounted for f (r) = −K Ωik k δ(r). (6. Ω0 is a dilatation parameter representing the change in the volume of a crystal as a result of a formation of the one defect. f (r) = −K Ω0 δ(r). ρUtt = σx . We start with the coupled equations for nj . The average parameters of the crystal specimen may be introduced if a typical distance between defects is considerably less than the gradient of a strain ﬁeld. Dj is the diﬀusion coeﬃcient of the defect of the kind j.h.(6. where K is the bulk modulus. (6. f . where Ωik is a symmetric tensor. t). while defect is located in the coordinate origin Kosevich (1981). Eq.51). where l is a unit vector of the dipole axis.February 11.12) for a microstructured medium.(6.x .53) . while Ω0 > 0 corresponds to the interstitial atoms. Then we have in one-dimensional case f = −K Ωj nj.52) where evolution of the concentration of defects is accounted for Eq. In elasticity single defect is described by the volume density of the force. and for nonlinear longitudinal displacement U (x.s. 6.51) accounts for a contribution of a strain into the defects generation. It is important that point defects cause deformation of a medium at macroscopic distances.3.(6. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 190 Ampliﬁcation of Nonlinear Strain Waves in Solids term in the r.51).2 Nonlinear waves in a medium One-dimensional governing equation in an elastic medium with moving defects may be obtained similar to Eq. of Eq. Volume mutual recombination of the defects of diﬀerent kind is neglected. βj is a recombination velocity at sinks Mirzoev and Shelepin (2001). When dipole is axi-symmetric. we have Ωik = Ω0 δik + Ω1 (li lk − 1/3δik ). Ω1 deﬁnes deviator of the tensor Ωik Kosevich (1981). For vacancies we have Ω0 < 0. (6.

52) being taken into account.(5. (6.(6. it was shown in Sec. Nondissipative limit of Eq. a3 = . as well as dissipative/active terms. Various simpliﬁed cases may be considered: (i) βj >> 1. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 191 where σ. 5.3 where stable kink waves with a ”hat” are found.51) are simply uncoupled yielding the governing equation for the strain waves v = Ux .(6. Weakly dispersive case (ii) corresponds to the weak defects diﬀusion. vtt − a1 vxx − a2 ( v 2 )xx − a3 vxxt + a4 vttt − a5 vxxtt + a6 vxxxx − a7 ( v 2 )xxt + a8 ( v 2 )xxxx = 0. (iii) βj >> 1. a5 vxxtt and a6 vxxxx .6. 5.54) predominates. However. ρ ρβj 4ρ βj ρ a4 = Dj (λ + 2µ)Dj a2 a2 Dj 1 .54) We see from Eq. (6. where a1 = λ + 2µ K Ωj qε 3(λ + 2µ) + 2(l + m) + 3n (λ + 2µ) − . .(6.3) that possesses oscillatory and multi-humps solitary wave solutions. 1.3. Dj << 1.42) in Sec. a6 = .3 describing the inﬂuence of an external dissipative/active medium on the waves evolution in a rod.February 11. 5.(1. a8 = . An analysis is already done in Sec. The coeﬃcients in Eq.(6.54) that presence of the moving defects provides a dispersion in an elastic medium. In case (i) recombinations at sinks are strong. and the formation of shocks is possible. a5 = .2 that no stable bell-shaped solitary waves arise in absence of higher-order derivative terms (like a9 v6x or corresponding mixed derivative terms) in Eq. a7 = . Equations (6.5. 2 σ = (λ + 2µ)Ux + (3/2(λ + 2µ) + l + 2m)Ux − K Ωj nj .53). (ii) Dj << 1. see Figs. Equation (6. a2 = .54). the inﬂuence of the higher-order nonlinear term a8 ( v 2 )xxxx may aﬀect the wave behavior. Dj = O(1/βj ). When dissipation is negligibly small. 5.54) is similar to Eq. is written with Eq.54) do not depend upon K Ωj with the exception of a1 . quadratic nonlinearity in Eq.54) is similar to a particular case of Eq. (6. 5. Simultaneous weak diﬀusion and strong recombinations in case (iii) provide weakly dissipative/active case which has been also studied in Sec. βj βj ρβj βj βj (6.

b5 = τ .51).56) is close to Eq.56) are negligibly small. There are . b4 = .(6. βN = + [(1 − 2ν)2 l + 2(1 − ν + ν 2 )m]. Again the evolution of the defects is accounted for Eq.3 Nonlinear waves in a plate Equation for longitudinal displacements in a plate already includes a dispersion even in absence of defects Mirzoev and Shelepin (2001): Utt − c2 Uxx − βN Ux Uxx − l2 (Uttxx − c2 Uxxxx ) = − s τ where c2 = s E 3E 2(1 − 2ν) .8) when the ci ’s terms in Eq.February 11.x ρ (6. Note that nonlinearity parameter βN does not depend upon the Murnaghan modulus n. c2 = . 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 192 Ampliﬁcation of Nonlinear Strain Waves in Solids 6. βj βj 2βj βj βj c1 = βN Dj l 2 Dj c2 l2 Dj . 2βj βj βj (6. Uncoupling of Eqs. and no weakly dispersive case is considered.56) b1 = Equation (6. (6. c3 = τ .51).3. where a1 = c2 − s K Ω j qε βN Dj c2 Dj . a3 = + l2 . τ 2 12(1 − ν) ρ(1 − ν 2 ) K Ωj nj. (5. b2 = . τ ρβj 2 βj βj 1 βN l2 c2 l2 c2 s . a2 = . Case βj >> 1 now corresponds to the perturbed double-dispersive equation.55) in case of the elastic plate is performed similar to the previous case giving the governing equation for the longitudinal strains of the form vtt − a1 vxx − a2 ( v 2 )xx − a3 vxxtt + a4 vxxxx − b1 vxxt + b2 vttt − b3 ( v 2 )xxt − b4 vxxttt + b5 vxxxxt + c1 ( v 2 )xxxx + c2 vxxxxtt − c3 vxxxxxx = 0. Now dispersion appears as a result of the plate ﬁnite width also.(6. b3 = . (6. a4 = s + c2 l2 . c2 = . 2) 2) ρ(1 − ν ρ(1 − ν ρ(1 − ν)3 l2 = µ ν 2 h2 .55) where h is the width of the plate.

The most interesting case is realized when small parameters qε . and we obtain from Eq. and the solution of Eq. Then one can write g = εg ∗ .57) may be obtained using the procedure from Chapter 2 to describe the selection of localized longitudinal strain wave as a result of the interactions with moving defects. (6.56) play the role of higherorder nonlinearity and dispersion. analytical results for a medium may be found in Kartashov and Bartenev (1988). Thanks to nonzero dispersion. θt = −V (T ). Eqs. T = εt. In this case Mirzoev and Shelepin (2001) derived the governing equation for the strain v = Ux . while numerical simu- . ζ = εζ ∗ . Then its asymptotic solution is obtained using the procedure explained in Chapter 2.4 Thermoelastic waves In thermoelasticity the deformation and temperature ﬁelds aﬀect each other.7–1. thermoelasticity is important to study attenuation of the waves that is observed in experiments. 5. As a result governing equations for the strains and the temperature should be coupled. (6.2 to account for the bell-shaped solitary wave selection.(6.57) where βd = βN qε K Ωj l2 (c2 − c2 ) s τ . Then we can use the results obtained in Sec. now stable oscillatory and multi-humps strain solitary wave solutions are realized.February 11. ci ’s terms dominate over bi ’s terms. . see Figs.(1. Of special interest are the processes in polymers arising due to the laser irradiation Kartashov and Bartenev (1988). 2 vt + cs vx + βd vxxx + γN vx = gv − ζvxx + µvxxxx . Nowacki (1986b).56) the hyperbolic analog of Eq. ζ = βj l2 .55) are uncoupled in a diﬀerent manner if βj << 1. 6. 1. g= 2 ρcs ρc2 s Equation (6. γN = . However. The fast variable θ and the slow time T are introduced as before. When Dj >> 1. Active/dissipative perturbations are accounted for the bi ’s terms while ci ’s terms in Eq. Very often the inﬂuence of the temperature on the strains is negligibly small.57) may be considered as a perturbed KdV equation.(6.51). 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 193 two kinds of perturbations. µ = Dj l2 . Dj and βj are of the same order. (6. Nowacki (1975). (6.16. µ = εµ∗ . ε << 1.3) studied in Chapter 1. θx = 1. Hence weakly active/dissipative case is realized if additionally Dj << 1. Linear thermoelastic waves were studied extensively.

First. (6.60) where P∗ is the Piola-Kirchoﬀ tensor expressed through the free energy density W (u. ρ0 utt = DivP∗ . P∗ = ∂W/∂ u. Nonlinear surface wave attenuation is considered in Mayer (1990). that takes into account an inertia of the heat ﬂux. the generalized law has been proposed in Likov (1967) Q = −κ T − τ ∗ Qt . Usually only the term linear in u. al (1987). it predicts an inﬁnite velocity of a thermal wave. αT (T − T0 ) u. Maugin (1999). Cv ∂T Cp − Cv ∂ + div(u) = κ T. αT is the heat extension coeﬃcient. Balance of linear momentum provides the second equation of thermoelasticity. Mayer (1995). the transfer of heat in solids is caused by the heat conduction only. Landau et.58) is rewritten in the form.60) contains the linear . T is the absolute temperature. Nowacki (1975). al (2000). (6. is used to obtain Eqs. In particular. Nowacki (1986b): T ∂S = div(κ T ). Maugin (1995). Engelbrecht and Nigul (1981).58). Usually the last coeﬃcients is assumed to be constant while the entropy is expressed through the temperature and the displacements ui . see. Then Eq..g. The Fourier law of thermal conduction. (6. τ ∗ is a time of the heat ﬂux relaxation.59). nonlinear bulk waves are studied in a medium in Engelbrecht and Maugin (1996) and in a rod in Potapov and Semerikova (1988). ∂t (6.59) where Cp . κ is the thermal conductivity. In contrast to liquid.February 11. Cv are the speciﬁc heat per unit volume at ﬁxed density and volume respectively.T ). Usually the equation of the heat conduction is obtained from the energy conservation law Engelbrecht (1983). Q = −κ T . the balance laws are used. ∂t αT ∂t (6. the modiﬁcation of the law is needed. However.58) where S is the entropy per unit volume.(6. (6. is included into W . Hence Eq. Berezovski and Maugin (2001). Landau et. In order to describe the ﬁnite velocity. e. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 194 Ampliﬁcation of Nonlinear Strain Waves in Solids lations are performed in Berezovski et. al (1987). Derivation of the equations may be done by various approaches.

c0 = . thermoelastic conductors of heat. ε << 1. (6.February 11. However. The equation of motion. ξ = c0 t − X.4. Nowacki (1975).63) Exact solution of Eq. one obtains for w = Ut wt + a1 wwξ + Λ w = 0. where a1 = 3(λ + 2µ) + 2l + 4m T0 ((3λ + 2µ)αT )2 2 λ + 2µ . τ = εX. ﬁnitely deformable. in the same manner. (λ + 2µ)εc0 2ερ0 c0 κ ρ0 (6. Maugin (2000).(6. Classical dissipative thermoelastic equations are obtained by isolation the contribution of the thermacy. ρ0 Utt −(λ+2µ)Uxx −[3(λ+2µ)+2l+4m]Ux Uxx = −(3λ+2µ)αT Tx .60) is derived from the Hamilton principle while the heat equation is introduced additionally Fares (2000). see Chapter 3. balance laws and constitutive relations for the theory of materially inhomogeneous. t)) in a medium when temperature T (x. .1 Nonlinear waves in thermoelastic medium Below we follow Engelbrecht and Maugin (1996) where one-dimensional model is considered using the notion of thermodynamic internal variables. 6.61) is coupled with the energy equation for the temperature. usually only Eq.(6. Λ= . The main idea is to assume a dependence upon an additional variable called thermacy whose temporal derivative is the temperature.62) Single nonlinear model equation may be obtained in case of a weak coupling.63) is known Whitham (1974). ε = (3λ + 2µ)αT /ρ0 . (6. Equations of thermoelasticity may be derived using the variational methods Fares (2000). As a result all equations obtained turn out strict conservation laws. Rayleigh (1945). The evolution nonlinear equation is obtained for the observable variable (longitudinal strain v = Ux (x. It was Maugin (2000) who proposed the modiﬁcation of the free energy density so as to get all ﬁeld equations. Nowacki (1986b). ρ0 Cv Tt + T0 (3λ + 2µ)αT Uxt = κTxx . Introducing fast and slow variables. t) eﬀects are considered as internal process. it may include shocks when the gradient of the initial excitation is large enough. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 195 term depending on the temperature while nonlinearity is caused by the same reasons as in pure elastic case.

2 Longitudinal waves in thermoelastic rod As already noted.(6.(6. the ODE reduction of Eq.(6. thermal conduction weakly aﬀects strain waves in solids.February 11.63) is done in Engelbrecht and Maugin (1996). 0 There is a similarity with the Burgers equation model if the last nonlinear term is negligibly small. ξ = x − V t. especially caused by a laser irradiation. Comparison of the Burgers equation model and the model governed by Eq. yielding nonlinear equation for U 2 2 (ρ0 Cv c2 +Λ1 )Uxxt −ρ0 Cv Uttt +a2 (Ux )xt +κ(Uxxtt −c2 Uxxxx −a2 (Ux )xxx ) = 0.(6.64) is written for the function v = Uξ ∂2 b1 v + b2 v 2 + b3 vξ + b4 (v 2 )ξ = 0. see Kartashov and Bartenev (1988).4.62) diﬀerentiated with respect to x. is expressed through the displacement using Eq. More promising looks the inﬂuence of a heat transfer through the lateral surface of a wave guide. with the relaxation tame taken into account. 6. Eqs. 2ρ0 ρ0 In the reference frame. Estimations done there demonstrate the dominant role of the heat transfer in comparison with the thermal conduction mechanism. (6. (6. ∂ξ 2 with b1 = V (ρ0 Cv V 2 − ρ0 Cv c2 − Λ1 ). Tx . Then it is substituted into Eq. Mirzoev et. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 196 Ampliﬁcation of Nonlinear Strain Waves in Solids When the coupling is not weak. b2 = −a2 V.61). 0 b3 = κ(V 2 − c2 ).62) are transformed by another way. the temperature gradient. The inﬂuence of the external heat transfer on nonlinear longitudinal strain waves in a rod has been studied in Potapov and Semerikova (1988). 0 0 (6.61). The Burgers model has been developed in Engelbrecht (1983). The coupled equations of thermoelasticity are used in Potapov and Semerikova (1988) with the heat transfer boundary conditions on the rod . al (1996) and references therein. First. b4 = −a2 κ.64) where a2 = 3(λ + 2µ) + 2l + 4m T0 ((3λ + 2µ)αT )2 . the case was not considered in Engelbrecht and Maugin (1996). Λ1 = .

Uxx = ρ2 C v 0 Utt + O(ε). h is a heat transfer coeﬃcient.67) is weak. T0 is a constant temperature of an external medium. 2 2 (6. s s s (6.h. so as R/L << 1.66) where β is a nonlinear coeﬃcient. R is a radius of the rod.s.66) with respect to x and substitute Tx from Eq.(6. B << 1.65). c2 = E/ρ0 is a velocity of s the linear waves in a rod. (6. Then the governing equation for longitudinal displacements holds. of Eq. As usual we consider the case of a balance between nonlinearity and dispersion introducing small parameter ε = B = (R/L)2 . and with long waves with typical length L. they may be considered as small perturbations.5R2 (ν(1 − ν)Uttxx − νc2 Uxxxx ) = −αT c2 Tx .(6. β = ρ−1 (3E + l(1 − 2ν)3 + 4m(1 − 2ν)(1 + 0 ν) + 6nν 2 ). Then nonlinear and dispersive terms may be omitted in the r.s. 2 E(ρ0 Cv + αT ET0 ) Hence the simplest equation accounting for the inﬂuence of both the heat transfer and thermal conduction on longitudinal strain waves v = Ux is . ∂ α2 ET0 {ρ0 Utt − E 1 + T ∂t ρ0 C v νER2 Uxxxx } = 2 Uxx − β Ux Uxx + ρ0 ν(1 − ν)R2 Uttxx − 2 κ ∂2 2h − ρ0 Cv ∂x2 ρ0 C v R {ρ0 Utt − EUxx − β Ux Uxx + ρ0 ν(1 − ν)R2 νER2 Uttxx − Uxxxx }. Contribution of the terms in the r. Note that the dispersion terms coeﬃcient from Potapov and Semerikova (1988) are corrected here in accordance with the procedure from Chapter 3.(6.67) We are dealing with elastic strain waves whose magnitude B is small.February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bulk active or dissipative sources of the ampliﬁcation and selection 197 lateral surface being taken into account: Utt − c2 Uxx − β Ux Uxx + 0. Obviously.65) ρ0 Cv R Tt − κR Txx + 2h(T − T0 ) = −αT ET0 R Uxt .h. Let us diﬀerentiate Eq.

(6.68) Certainly Eq.February 11. ρ0 C v ρ0 C v R (6. . attenuation and selection of longitudinal thermoelastic waves.68) may be studied by the methods developed in Chapter 2 to account for an ampliﬁcation. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 198 Ampliﬁcation of Nonlinear Strain Waves in Solids vtt − c2 (1 + s 2 νc2 R2 αT c2 T0 β ν(1 − ν)R2 s vttxx − s vxxxx = )vxx − (v 2 )xx + Cv 2ρ0 2 2 2 αT c2 T0 s 2 Cv + αT c2 T0 s 2h κ vttx − vt .

N. J. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Bibliography Ablowitz.R. (1991) ”Weakly non-local solitons for capillary-gravity waves: the ﬁfth-degree Korteweg-de Vries equation”. A. 249. Acoust. (1994) Algebro. Engelbrecht. A 32. and Segur. S. V. 129. Bridges. Berlin. 168.A.February 11. (2001) ”Simulation of Thermoelastic Wave Propagation by Means of a Composite Wave-Propagation Algorithm”. Physica D 48. Phys. Bland. G. and Matveeev. Nauka. Higher Transcendental Functions. 2819. Phys. Ya. Krylov. D. Akhmediev. V. Berezovski.(1993) ” The generation of radiating waves in a singularly-perturbed KdV equation”. 4419. A. Grimshaw. G. Maugin. Oxford . J. Phys. Waltham.. Kuznetsova. J.S. 511. M. 199 . and Rybak S. J.L. and Ankiewicz. SIAM. Berezin. G. V. Claredon Press. VNU Science Press. Physica D 69 270. R. Utrecht.A. V. Its. (1987) Modelling Non-Linear Wave Processes.V. Alekseev... 70. A. P. 48. (1953-54). E. V.. (1986) ”Cnoidal standing waves and the transition to the travelling hydravlic jump”. Mass et al. Mech.Geometrical Approach to Nonlinear Integrable Equations. Belokolos.(1999) ”Classical and non-classical interactions of kinks in some bubbly medium”. A.P. Nonlinear pulses and beams. vol. 3. (1991) Surface acoustic waves in inhomogeneous media. Berezovski. A. Alexeyev. Moscow (in Russian). Bhatnagar.A. Compt. Appl. Enol’skij. (1997) Solitons. Arch. Oxford. N. Boyd. (2000) ”Thermoelastic wave propagation in inhomogeneous media”. (2002) ”Equations of state for viscoelastic biological media”. New York. V..(1969) Nonlinear Dynamic Elasticity. London. A.V. P. A. Blaisdell. R. Fluids 29. and Erdelyi. and Maugin.. J. Chapman & Hall..V. A. Bobenko. Bland. J. Springer..N.A. D. H. (1960) The Theory of Linear Viscoelasticity. E. Philadelphia. Benilov. Bateman. McGraw Hill. V. Gulyaev.. and Plessky. T.(1981) Solitons and the Inverse Scattering Transform. Biryukov. (1979) Nonlinear Waves in One-Dimensional Dispersive Systems. 694.P.

St. (1997) ”Implicit time splitting for fourth-order parabolic equations”. 4125.W. A.I. Proc. 685. Appl. Engrg.I.M. Lond. o R. (1989) Continua with microstrucure. Methods Appl.A. E 54.I. P. Physica D 39.K. Intern. D. Compt. and Maugin.I. 63. Cariello. Anal. C. Phys. (2002) ”Fourier-Galerkin Method for Localized Solutions of the Equations with Cubic Nonlinearity”. (1995) ”An Implicit Diﬀerence Scheme for the Long-Time Evolution of Localized Solutions of a Generalized Boussinesq System”. Rev. K. (1998) ”Numerical Simulation of Nonlinear Elastic waves in Inhomogeneous Wave Guides”. Maugin.Z. Berlin. and Velarde. 1747. (1995) ”Quasiperiodic solutions of the coupled nonlinear Schr¨dinger equations”. Proc. Academic Press. Comput. C. and Tabor.M. in Fluid Physics. (1982) Spectral Transform and Solitons. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 200 Book Title Bukhanovsky. Math.J. Walgraef. Springer.. C. Conf. (1995) ”A class of exact. F.February 11.Rev. P. (1995) ”Stability of a Solitary Pulse against Wave Packet Disturbances in an Active Media”. USA. Christensen. (1997) ”Growth and decay of waves in microstructured solids”. Estonian Acad..I. M.. Kostov. 36... Math. (1994) ”Numerical Investigation of the long-time evolution and interaction of localized waves”. 75.G. M.V. (1996) ”Well-posed Boussinesq paradigm with purely spatial higher-order derivatives”. G. 46.. Pontes. Phys.D.. J. Christov. and Degasperis. F. Mech.L. Christov. on Dynamical Systema and Diﬀerential Equations. Chang. NorthHolland. Springer. (1980) Solitons. C. Preprint of the Institute for High performance Computing and Data Bases No 6-98 (in Russian). Adv. Velarde. M. (1948) ”A Mathematical model illustrating the theory of turbulence”.F.-C. J. Petersburg.(1989) ”Painlev´ expansions for nonintegrable evolution e equations”. Enolskii. and Kopelevich. C. and Christov. World Scientiﬁc. 3621. A 451. New York. C.A. Phil. & Christov. .G. eds.. Berlin. F. Cermelli. Proc. D. Springer Tracts in Nat. and Pastrone. Velarde.A. Amsterdam. A. and Caudrey. and Samsonov.A. and Friedman. Christov. (1954) Handbook of Elliptic Integrals for Engineers and Physicists. Christiansen. Phys. Sci.I. Soc. 1.. (1971) Theory of Viscoelasticity. Christov.A. 4. Mech. J. Demekhin. P. May 1821. Christou. V.35. Phys. A. R.M. periodic solutions of nonlinear envelope equations”. E. 32.I. Lett. C. P. and Christov. Bullough. 353. 148. 77. Comp.I. M. Appl. Burgers. J. 209. V. eds. Capriz. 116.. R.. Phys. H. Byrd. (2000) ”Fourier-Galerkin Method for Localized Solutions of the sixth-order Generalized Boussinesq Equation”. G. M. Christou.A.. Atlanta. Chow. N. J. Singapore. J. 39.. Springer. M. 171. Calogero.C. Berlin. M. G. 121. Eilbeck.G.

Wave Motion 1. R. Mech. Engelbrecht. I.G. (1989) Nonlinear Waves in Active media.I.V. G. 475. and Maugin. Crighton. Appl. (1998) ”Nonlinear waves in nonlocal media”.. Complexity and Simplicity. 65. Kluwer. (1983) Nonlinear wave processes of deformation in solids. A. Dreiden. Phys. M. Y. C. Dai. 193. Gen. Tech.M. 51.V. 200. Engelbrecht. I. F. A 140. M. (1981) Nonlinear Deformation Waves. London et al. J. G. (1993) ”A perturbative Painlev´ approach to e nonlinear diﬀerential equations”. 127.G. (1995) ”Dissipative solitons”..V. and Velarde.C. Engelbrecht. J. R. Chaos 4. Lett 21.A. ed. (1988) ”On the possible ampliﬁcation of nonlinear seismic waves”. Pitman. (1968) Perturbation Methods in Applied Mathematics. 505. U. Berlin. (1995) ”Experiments in the propagation of longitudinal strain solitons in a nonlinearly elastic rod”. H. (1989) Solitons: an Introduction.. Pickering. Gibbon. J.. Nauka. Porubov. Cambridge Univ. A.V. J. Bif. Engelbrecht.. and Braun. (1997) ”Nonlinear dynamics of dry friction”. Conte. Porubov. Engelbrecht. (1995) ”Applications of KdV”. G. and Morris. Phys. Press. Drazin. J. Springer-Verlag. Tech. Mech.V. (1997) Nonlinear Wave Dynamics. Cermelli. 46. A. J. (2001) ”Reﬂection of a longitudinal strain solitary wave from the end face of a nonlinearly elastic rod”. M. Intern. R. . D. E. Boston. Arch. e Phys. No 8. Herman Publ. A. Acta Applicandae Mathematicae 39.D.. C. 33. (1989) ”Invariant Painlev´ analysis of partial diﬀerential equations”. A. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 201 Christov. Dodd..K. Academic Press. Eilbeck. Engelbrecht. 6057. by G. Samsonov. Paris. Christov. Mass. Cambridge. J.C. Cole. (1999) ”Wave hierarchy in microstructured solids”. and Nigul. P. Samsonov.A.. 99. Lett. 66. and Sokurinskaya. (1998) ”Model equations for nonlinear dispersive waves in a compressible Mooney-Rivlin rod”. Inter. 30. (1982) Solitons and Nonlinear Wave Equations. Moscow (in Russian). Fordy. (1979) ”One-Dimensional Deformation waves in Nonlinear Viscoelastic Media”.. Dreiden. (1996) ”Deformation waves in thermoelastic media and the concept of internal variables”.J. H. Phys.G. 39. J.M.-H. 323. Engelbrecht. 383.-J. 50. R. (1994) ”Inelastic Interactions of Boussinesq Solitons”.V. Physica D 86. J. J. and Johnson.. Acta Mech. . Appl. and Khamidullin.S. and Velarde. The Netherlands. Blaisdell. Engelbrecht. and Pastrone.D.. Semenova. J. Waltham. Conte. Ed. Elmer. no 5.I. Maugin. 415..February 11.Phys. P. F.. 1095. Semenova. Continua and Microstructure. J.V. Earth Planet. A: Math. In: Geometry. Physica D 69. Rev. Engelbrecht.G. J. 39. A.

Theory of micropolar elasticity.. E.A. Physica D 137. S. Int. (1980) Contact problems of elasticity and viscoelasticity. and Kawahara. Intern. Erofeev. C. Erbay. Nonl.N. New York. Phys. 4783. Moscow (in Russian). (2000) ”Stationary travelling-wave solutions of an unstable KdV-Burgers equation”. J.A. C. (1982) Elastic Constants and Moduli of Metals and Nonmetals. B. (in Russian). 389. Moscow.I. Cambridge. In: Fracture. Sci. M. S. Press. Sov. N. J.I.F. V. and Potapov.E. Eringen. V.K. Goryacheva. (2000) ”Generalized non-linear thermoelasticity for composite laminated structures using a mixed variational approach”. Part II. 845.. Voronov. and Talipova T. Academic Press..V. 28. I. J.G.February 11. Frantsevich. H. longitudinal waves” Int. Nauka. (1997) Ordinary Diﬀerential Equations with Constant Coeﬃcients. Kiev. Godunov. 2. 483.F. by H.K.Acoust. Moscow (in Russian). Mech. Physica D 77. V. Physica D 132. (1999) ”Solitary wave transformation in a medium with sign-variable quadratic nonlinearity and cubic nonlinearity”. plates and shells”. A. Part I. S. World Scientiﬁc.S. 439. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 202 Book Title Erbay. Pelinovsky E.C. 622. and Ryaben’kii. Engng. Dordrecht etc. (1994) ” Solitary waves with damped oscillatory tails: an analysis of the ﬁfth-order Korteweg-de Vries equation”. Galin. e Godunov. Gulyaev.I. Grimshaw. V. Amsterdam. R. Nonl.. North-Holland. Eringen. S. and Velarde. An advanced treatise. I. P. Springer. A.K.V. Fowler. A.A. (1987) Diﬀerence Schemes: introduction to the underlying theory of. and Ryaben’kii. Liebowitz. V.I. and Suhubi. (2002) Wave processes in solids with microstructure. Fluids A 3. 29.. (2001) Mechanics of friction interaction. (in Russian).S. V. (1991) ”Nonlinear wave modulation in micropolar elastic media-I. S. and Klyueva. Eng. F. and Benilov.A. and Dost. (1964) ”Nonlinear theory of micro-elastic solids”. Garazo. Yu. Godunov. Malomed. 48. J. Fares.2. N. Nauka. (1968). Phys. (1978) ”Shear surface acoustic waves on a cylindriucal surface of a solid body covered by a layer of a foreign material”. J. (1984) Computational Galerkin Methods. Acoust. (1997) Mathematical Models in the Applied Sciences. 725 (in Russian). Mathematical fundamentals. Sci.G. AMS. 228.G. Kluwer.S. (1993) ”Longitudinal strain waves in non-linearly elastic media with couple stresses”. Mech. S. 287. 2295. 40. Naukova Dumka. (1991) ”Dissipative Korteweg-de Vries description of Marangoni-B´nard oscillatory convection”. (1998) Contact mechanics in tribology. Feng.N. Erofeev. Cambridge Univ. Int. A. 189. T.. 35. and Polsikova. Ed.C. L. Fletcher. (2002) ”Solitons and nonlinear periodic strain waves in rods. Singapore. Erofeev. Goryacheva. A. M. .J. and Bakuta. R. 24.I. Nauka. B. I. (1973) Diﬀerence Schemes. Grimshaw R. E. Berlin.

J. eds. (1988) ”Chaotic Motions in an Oscillatory Soliton Lattice”. and Kawahara. Kawahara. Mech. 57. Jeﬀrey. Lett.N. (1990) Heuristic Approach to Obtain Solutions to Nonlinear Equations of Solitonics. 253. a No 9. Karpman. and Vollebregt. Physics Today. (1988) ”Dynamic eﬀects in solid under interactions with intensive heat ﬂuxes”. Jpn. 381. Lett. Dover. E 47. Rev. Karpman.mtu. 394. E. Kalker. Soc. T. V. Mech. and Scheurle. A. and Nakayama. J. Itogi Nauki i tekhniki. 4959. N. Jeﬀrey. A. Wien. Karpman.I. (1985) Contact Mechanics. I. J. Kawahara. (2001) ”Stability of solitons described by KdV-type equations with higher-order dispersion and nonlinearity (analytical results)”. Appl. Cambridge. Hunter. Hilger & Bristol..V. 22. A. E. (1994) Nonlinear Waves in Solids.L. M.. A. Phys. and Velarde.M. G. F. A 200. K. J.J. (1993) ”Radiation by solitons due to higher-order dispersion”. A. Ser.M. I. Kartashov. 246. . Phys. V.February 11. I. Phys. (1983) ” Formation of saturated solitons in a nonlinear dispersive system with instability and dissipation” Phys. and Spencer. Johnson. Phys. (1995) ”Stationary solitons and stabilization of the collapse described by KdV-type equations with high nonlinearities and dispersion”. J. 64. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 203 Gusev. J. T. Cambridge Univ. Dekking.K. Porubov.A. Kliakhandler. (2000) ”Localized ﬁniteamplitude disturbances and selection of solitary waves”. T.L. T. Phys. Phys. Physica D 32. (1964) Ordinary Diﬀerential Equations. (1981) ”An Exact Solution of the Wave Equation ut + u ux − u5x = 0”. J. Soc. (1964) ”Elastic and viscoelastic foundation models”. 361. and Bartenev. 33. Jpn. G. (1998) ”Rubbing and scrubbing”.Rev. (1982) Asymptotic Methods in Nonlinear Wave Theory. (1988) ”Existence of perturbed solitary wave solutions to a model equation for water waves”. and Takaoka. Lett. Lett. Zinatne.math. E.-M. 50. Appl. Tectonophysics 152.. . 260. Phys. J. 2073. H¨hner. 491. (1998) ”Evolution of radiating solitons described by the ﬁfth order Korteweg-de Vries type equations”. Elastic Contacts”. (1988) Elastic Waves in Solids with Application to Nondestructive Testing of Pipelines.: Khimia i tekhnologia visokomolekuliarnikh soedinenii” 25.N. Pitman. Riga (in Russian). Kascheev. Karpman. A 284.M. 3714. 51. Kano.H. J. T. Philadelphia. Jpn. A. K. Springer. (1988) Two dilatancy-based models to explain coda-wave precursors and P/S spectral ratio.I. Press. 227. V. and Engelbrecht. London.L. (1997) ”Simulation of Rough. 423. Ince. 3 (in Russian). Rev.G. A. J.I. V.D. A 244. (1999) http://www. 361. (1972) ”Oscillatory Solitary Waves in Dispersive Media”. Phys. and Vanden-Broeck. 31. M. Kerr. E 62. Soc. Kawahara. Harker.edu/˜igor Kliakhandler. V. New York.

(1993) Material Inhomogeneities in Elasticity. Likov.M. Lurie.A.. A. J.A. Lifshitz E. and Silen´. 89. Opt. Kosevich. (1990) Nonlinear Theory of elasticity. (1988)” Exact solitonic solutions of the generalized evolution equation of wave dynamics”. Math.A.I. I. Mag. 173. (1967) Theory of thermal conduction.A. N.February 11. S. G. Appl. London. Amsterdam. Plenum Press. New York and London. Kostov. Naukova Dumka.D. Pergamon Press. World Scientiﬁc. and role of higher-order dispersion in solitary wave dynamics”.Non-Equil. N. Low Temp. Kosevich. D. S. . L. 52. A. 225. T. Phys.H. 1. A.E. a y Shallow earthquakes with a possible tensile source component”. 361. H. A.M. G. 1109. Kudryashov. 48.M. Thermodyn. 217. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 204 Book Title Kodama. 5. J. Math. . Singapore. (1981) ”Perturbation of solitons and solitary waves”. G. 64. ˇ Koz´k. Appl. J.. A. IEEE 72. (1999) ”Features of dynamics of onedimensional discrete systems with non-neighbouring interactions. (1992) ”New kinds of periodical waves in birefringent optical ﬁbers”. (1999) The Thermomechanics of Nonlinear Irreversible Behaviors. Kiev (in Russian). and Uzunov.R. Maugin. G. Maugin. J. Chapman & Hall. W. (1994) ”Thermodynamics with internal variables” J. Rev. Lou. G. Korpel.M. Levi. Comm.. and Ruan. Phys. University Press. Elsevier.A. (1927) A Treatise on the Mathematical Theory of Elasticity.A. and de Vries. Stud. (1991) ”Exact solitary waves in a convecting ﬂuid”. and Banerjee. Landau L. J. Thermodyn. Kosevich. (1895) ”On the change of the form of long waves advancing in a rectangular channel and on a new type of long stationary waves”. 213. D. and Muschik.J. and Winternitz. Nauka. (1981) Physical Mechanics of Real Crystals. (1992) Painlev´ Transcendents: Their Asympe totics and Physical Applications. Appl. I. and Ablowitz. Mech. and Savotchenko. G. (1996) ”The evolution and interaction of Marangoni-B´nard solie tary waves”. Huang. Maugin. (1987) Theory of Elaticity. Phil. 422. eds. Cambridge. (1985) ”Seismic events with non-shear component.. P. An Introduction. Mech. 15. A. M. Love. Korteweg. NonEquilib. (1995) ”Material forces: Concepts and applications”. P. Ya. 19.. Oxford. (1984) ”Heuristic guide to nonlinear dispersive wave equations and soliton-type solutions”. Maugin. (1990) ”Internal Variables and Dissipative Structures”. (2000) ”Towards an analytical mechanics of dissipative materials” Rendiconti del Seminario Matematico dell’Universita’ e Politecnico di Torino 58. Marchant. Moscow (in Russian). 171.P.A. 307 Maugin. G.A. J. G. L587. and Pitaevsky. Wave Motion 23.M.E. 737 (in Russian). PAGEOPH 123. 389. 25. B. A 24. Maugin.

1701.V.D.D. A. and Kuwahara. 197. 68. New York . J. F. Wiley. Pergamon. Jpn. 257.V. M. 46. Mindlin. soliton bound states and chaos in a dissipative Korteweg-de Vries equation”. 285. A (1990) ”Thermoelastic attenuation of surface acoustic waves in coated elastic media”. Mech. R. Appl. Phys. 1135. M. (1975) Theory of Elasticity.V. Nekorkin. Kh. Cambridge. and Potapov.. J. and Velarde. e Fluids A 6. Nikitin. F. I. Newell. Physica D 29. A. Mir. Nowacki. 47. Newell. 306. (1998) Statics and dynamics of solids with external dry friction. W. SIAM. Miloserdova.W. Oxford.A. Cambridge University Press. (1964) ”Microstructure in linear elasticity”. and Velarde. (1977) ”On the eﬀective surface tension of solid bodies”. (1983) ”Nonlinear standing waves in a rod of ﬁnite length”.. Pergamon. Nowacki. (1994) ”A three-dimensional description of solitary waves and their interaction in Marangoni-B´nard layers”. Chaos 4. Ya. PAGEOPH 131.. (1986b) Thermoelasticity. J. Rat. and Shelepin. M. Phys. Oxford. E. Zeng. 51. Phys. Pergamon. Int.G. No1. K. A. Tabor. (1951) Jacobian Elliptic Functions. 3792. (1994) Numerical Solution of Partial Diﬀerential Equations. Moscovskii Licei. (1986a) Theory of Asymmetric Elasticity. 50. (1996) ”Laser Control of Processes in Solids”.D. Bif. Reports 256. p. Acoust. A. W. A. 952. Herrmann ed.H. Phys.J. New York. V. (1951) Finite Deformations of an Elastic Solid. Arch. Phys. (1987) ”A Uniﬁed Approach to Painlev´ Exe pansions”. Claredon Press. and Mayers. V. H. 1. A. Shelepin. Nagashima. Nikolova. Y.I. Anal. exact two-periodic wave solution”. J.February 11. (1989) ”Scattering and dissipation of seismic waves in presence of nonlinearity”. Kh. (1981) ”Computer Simulation of Solitary Waves of the Nonlinear Wave Equation ut + u ux − γ 2 u5x = 0” J. Wiley & Sons. 1.G. Nakamura. (1974) ”Vibrations and wave propagation in rods”. Murnaghan. 5913. 187. L..I.F. (1994) ”Solitary waves. Oxford. Phys. E.H. Mirzoev. Nowacki. Physics. Moscow (in Russian). A (1995) ”Surface Acoustic Waves in Nonlinear Elastic Media”. Jpn. Morton. Tech.. W. Sov. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 205 Mayer. F. D. I.B. Nikolaev. 29. Mindlin and Applied Mechanics. 687. Mirzoev.A. J. Panchenko. (1979) ”A direct method of calculating periodic wave solutions to nonlinear evolution equations. Sov.. Nepomnyashchy. Phys. A (1985) Solitons in Mathematics and Physics. JETP 45. (1973) Perturbation Methods. R.Uspekhi 39. New York. Nayfeh. (2001) ”Nonlinear Strain Waves and Densities of Defects in Metal Plates Induced by External Energy Fluxes”. J. Mayer. M. Soc.A. McNiven. . Moscow (in Russian). in: G. Newille. Phys. H. L.G. Soc. Philadelphia. and McCoy. L.

and Duﬀy.V. 391. 97. A. A. Phys. (1954) On a New Method of Analysis of an elastic Foundation be Means of Two Foundation Constants (in Russian). A.February 11. Berlin. . ed.A. Springer.A.V. A. G. o Porubov.Fusco and A. Oron.F.V. (1999) ”Explicit Solitary and Periodic Solutions for Optical Cascading”. E 55. Roy. Rendiconti del Seminario Matematico dell’Universita’ e Politecnico di Torino 58. and Parker.N. Phys.V.(1995) Free Surface Nonlinear Waves on a viscous inhomogeneous liquid layer. V. Parkes. A: Math. (1996) ”Periodical solution to the nonlinear dissipative equation for surface waves in a convective liquid layer”.D.V. Gosudarstvennoe Izdatel’stvo Litearturi po Stroitel’stvu i Arkhitekture.. A 458. 36. (2003) ”Selection of localized nonlinear seismic waves”. Baltimore.V. Eng. B.A. Lett.. 2139. D. L. (1996) ”An automated tanh-function method for ﬁnding solitary wave solutions to nonlinear evolution equations”. . and Rosenau... and Maugin. Parker. Springer. Thesis. Ph. 98. No 1. Longman. Porubov. Proc.F. J. and Tsoy. A. In: D. V. L797. and Maugin.V. Krzhizhanovskaya. Estonian Acad. Math. Moscow. Nonlinear Waves in Solids. Sci. P.A. Parker. and Gursky.F. Saint Petersburg ( in Russian). V. eds. (1978) Surface Acoustic Waves. (1987) Recent Developments in Surface Acoustic waves. G. (2000) ”Strain solitary waves in an elastic rod with microstructure”. Maugin. Soc.Jeﬀrey and J. A.L. Proc. A 221. 288. Wave Motion 29. State Technical University. (1994) ”Nonlinear Surface Acoustic Waves and Waves on Stratiﬁed Media”. (1997) ”Evolution and formation of dispersivedissipative patterns”. Engelbrecht. A. Porubov. Pasternak.V. A.V. submitted for publication Porubov. (1999) Modulated waves in linear and nonlinear media. J. Rev. D. A. In: A. P. 26. and Mayer. (1999) ”Some General Periodic Solutions to Coupled Nonlinear Schr¨dinger Equations”. (1991) ” Dissipation of Surface Acoustic Waves”. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 206 Book Title Oliner. (2002) ”Inﬂuence of higher. and Potapov.F.I.A. Berlin.order dispersion and nonlinearity on solitary wave formation from an initial localized pulse”. Parker. Parker. G. Gen.. 189.R. Springer-Verlag. Porubov. Porubov. (2002) ”Some General Periodic Solutions for Optical Cascading”. D. . Gursky. USSR (in Russian).F. and Parker. A. D.V. A. Phys. Phys. (1993) ”Exact travelling wave solutions of nonlinear evolution equation of surface waves in a convecting ﬂuid”. 52. 149. E.J. E. eds. Computer Phys. R1267. Porubov. D.V. eds. Comm. Berlin. D. Nonlinear Waves and Dissipative Eﬀects. The John Hopkins University Press. Porubov. Jeﬀrey. Ostrovsky. A. Math.F. London.

S. Estonian Acad. (1988) ”Nonlinear longitudinal waves in a rod taking into account the interaction between strain and temperature ﬁelds”. Quaderni del Dipartimento di Matematica N. M. Sci. J. Ryazantsev.. Porubov.V. A. Reprint.G.G. A. 54. and Samsonov. Phys. 189. (1993) ”Reﬁnement of the model for the propagation of longitudinal strain waves in a rod with nonlinear elasticity”. and Velarde. M.. (1998) ”Strain solitary waves in an elastic rod embedded in another elastic external medium with sliding”. Appl.V. Potapov. Salupere. Phys. Turin. (1997) ”Solitons in systems with a quartic potential and higher-order dispersion”. 40. N. Cambridge Univ. J. 33. Salupere. 118. (1988) ”On existence of longitudinal strain solitons in an inﬁnite nonlinearly elastic rod”. J.. M. E. Wiley & Sons. (1967) Diﬀerence Methods for Initial-Value Problems. 1. (1994) ”Korteweg-de Vries soliton detection from a harmonic input”. M. G. K.M. Rednikov. A.V. . Nepomnyashchy.M. 298. Velarde. A. 5.V. Richtmyer. Lett. . Lett. 703.G. A. P¨schel. Engelbrecht. Birkh¨user. H. (2000) ”Dispersive-dissipative solitons in nonlinear solids”. A. 884. Mech. and Nikolaev.A. Wave Motion 34. a Samsonov.J. Porubov. Phys. A. Salupere. and Bukhanovsky. G. New York. Qaisar. Physica A 198. and Pastrone.V. T. and Rodyushkin.February 11. F. Rayleigh Lord (J.P. A. Math. Phys.W. J. 197.Rev. V. Wave Motion 35.I. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 207 Porubov.A.V. No1. Acta Applicandae Mathematicae 39. and Morton.G. University of Turin. M.. (2002) ”Strain kinks in an elastic rod embedded in a viscoelastic medium”. . A. 441. Sov.M. R. G.A. Sov. and Velarde. A. Yu. Porubov. (2001) ”Experimental study of strain waves in materials with microstructure”. P. Basel. Acoust. 56. Sachdev. A.D. (2001) ”Solitonic structures in KdV-based higher-order systems”.. Samarskii. Dover. and Kurdyumov V. 457. Proc.-Doklady. Wave Motion 31. Press. S. Tech. Phys. Porubov.. (1995) ”Cnoidal Wave Trains and Solitary Waves in a Dissipation Modiﬁed Korteweg. A. Phys. and Maugin.de Vries Equation”..G. A 192. 19. Ye. 51. Porubov. 46. A. and Engelbrecht. Cambridge. A.I. Maugin.N.A. (1999) ”Exact Periodic Solutions of the Complex Ginzburg-Landau Equation”. (1989) Numerical Methods for Grid Equations. Velarde..and microdissipation on nonlinear strain solitary wave evolution”. E 58. A. (1989) ”Attenuation Properties of Viscoelastic Material” PAGEOPH 131. A. Phys. (2001) ”Inﬂuence of macro. 347.L. Potapov. A. J. Maugin. Math. 47. 3854 .W. A.M. Samsonov.V. . and Herrmann. (1993) ”A simple geometrical model for solid o friction”. Engelbrecht. Vol. and Semerikova.A. and Velarde. Strutt) (1945) Theory of Sound. and Tech. (1987) Nonlinear diﬀusive waves.. 365. . New York. M.

K. Christiansen. . 5778. 76. (1971) In ”Treatise on Continuum Physics” (A. 57.. L. Wave Motion 14. V. and Agasiev.M. Theor.. 15. (1949) ”Elliptic problems in linear diﬀerence equations over a network”. 86.. I”.S. R. (1999) ”Wave propagation in dissipative microstructured materials”.L.V. (1995) ”Travelling wave solutions for nonlinear waves with dissipation”. Weiss. Lysko. G. J. Slyunyaev. P. Samsonov. Mech. Chap.M. G. and Lysko.M. Sov. (1977) ”Love waves on the surface of cylinder covered by a layer”. 433. (1991) ”The Kotreweg-de Vries-Burgers equation: a reconstruction of exact solutions”. J. (2000) ”Numerical analysis of n duﬃng oscillator with dry friction damper”. Int. O. A. (1984) ”Solitary waves in nonlinear elastic rods. Lab.V. Technical University. A. . and Maksimov.G. 81. Soc. M.B. and Semenova. Phys. ed. Shevyakhov..H.V. No2. New York. A. 267. 51 . and Pelinovsky. Vol. M. W.” Proc. Svendsen. Estonian Acad. M.3. J. Prog. Dreiden. 52.L. P.D.V. Dissertation. . Dreiden. Tabor. Phys. T. (2001) Strain soltons in solids and how to construct them.V. A. Sokurinskaya.G. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 208 Book Title Samsonov. Spencer. Rept. Amer. Math. (1978) ”On the deformation of periodic waves over a gently sloping bottom”. 127. Soerensen. Samsonov. Phys. 871. St. Sci. New York. Sawada. (1987) ”Solitary waves in nonlinear elastic rods. Columbia University. Thesis of Master of Science. G.. N. Vlieg-Hultsman. P. G. and Halford..M. J. (2003) ”On existence of bulk solitary waves in plexiglas..S. Acoust. Samsonov. Savin. Mech. Tallinn. J. Lomdahl. A.N. Academic. Watson Sci. and Wiercigroch. Lukashev. 24.V. A... I.P. Wojewoda. G. 725.. 6. J. Ph. Rev.. II”. Christiansen. Phys. B 57. Comput. 23. E. I. I. Theory of invariants. Phys. A. J. Bif.M. and Buhr-Hansen.Anal. J. 831.).M. 1355. Amer. A.Petersburg (in Russian).February 11. 85.. Soerensen. Sillat. Soc. Ing.A. 15. N. Velarde. No 1. E. A. Fluid Mech. Nekorkin.A.. Appl. P. 522. (1999) ”Dynamics of large amplitude solitons” JETP 89 173.J. M. Lukashev.S. 87. Chapman & Hall/CRC. Technical University.. (1983) ”The Painlev´ property for partial e diﬀerential equations”. Soviet Appl. Veremeeenko. Mech. (1991) ”Study of nonlinear travelling strain waves in onedimensional elastic wave guide”. Math. A.P. I. Stefa´ ski. (1995) ”Further results on the evolution of solitary waves and their bound states of a dissipative Kortewegde Vries equation”. Thomas. Lomdahl. S. Porubov. 4.M. Chaos 5. & Mech.N. and Kotera. M. (1998) ”Longitudinal-strain soliton focusing in a narrowing nonlinearly elastic rod”. G. and Carnevale.V. A. E. Acoust. Eringen. (1973) ”Elastic wave propagation in a solid with microstructure” Soviet Appl. and Semenova. (1973) ”Elastic wave propagation in the Cosserat continuum with constrained particle rotation”.. M. Savin. (1974) ”A method for ﬁnding N-soliton solutions of the KdV equation and KdV-like equation”.. -Acoust. and Skovgaard. T.C. E.G.I..V.

N. New York . Solid State 25. Press. (1989) Handbook of Diﬀerential Equations. Boston. Acad. (1974) Linear and Nonlinear Waves. 1797. (1983) ”Dilaton Mechanism of the Strength of Solids”. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 209 Whitham. University Press.T. B.February 11. Zwillenger. (1999) The Mathematica book. Whittaker. D.N. E. and Watson. G. Zhurkov. Phys. (1867) Die Lehre von der Elasticitaet und Festigkeit. (1927) A Course of Modern Analysis. Addison-Wesley. Wolfram. Fourt Edition. Wiley. Cambridge. G. S. Dominicicus. . S. Winkler. E. Prag. Sov.

February 11. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 210 Book Title .

6. 42 D’Alembert solution. 138. 197 between nonlinearity and dissipation. 38. 155 energy free. 44 . 71. 127 Korteweg-de Vries equation. 137 seismic.February 11. 136. 98. 155. 27. 142. 135. 12. 34 ansatz. 57. 129 breather-like solution. 157. 105. 67. 124 double-dispersive equation. 80 dilation mechanism. 103. 179 entropy. 156 dynamical system. 42. 65. 12. dispersion and dissipation. 36. 43. 116. 142 elementary work. 194 211 Cosserat model. 142. 51. 49. 36 a balance between nonlinearity and dispersion. 87. 115. 64. 85 action functional. 178 kinetic. 169. 36. 155 Weierstrass function. 173 between nonlinearity. 116. 169. 41. 194 blow-up. 79. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Index acoustic resistance. 79. 11. 27. 1. 136. 103. 12. 90. 129. 178 Boussinesq equation. 100. 51. 12. 102. 134. 59. 9. 49 conservation laws. 169. 128. 134. 142. 26. 71. 194 inﬂux. 116 Coupled nonlinear Schr¨dinger o equations. 166. 146. 136. 156. 149 Cauchy-Green tensor. 13. 54. 147. 1. 32. 178 dilaton. 25. 196 kink. 27. 152 elliptic integrals. 79. 175 of linear momentum. 123. 133 body force. 12. 189 internal. 165 cnoidal wave. 31. 194 balance laws. 71. 61 Jacobi functions. 34 auto-B¨cklund transformation. 194 envelope wave solution. 47 Burgers equation. 178 dissipation-modiﬁed double dispersive equation. 127. 153. 53. 65. 174 dissipative elements. 148 perturbed kink. 163. 126. 123. 178. 135 potential. 1. 122. 10. 52. 71. 32. 9. 129. 66 analysis of singular points. 156. 57.

125. 196. 136. 65 isotropic. 65. 189 Le Roux model. 65. 153. 124. 87. 117 Love hypothesis. 53 spring elements. 116. 88.February 11. 136. 73. 70 matching asymptotic procedure. 15. 79. 189 Poisson ratio. 16. 76 geophysical medium. 125. 194 heat transfer. 65. 75. 150 geometrical. 79 with varying cross section. 6 Kerr model. 114. 2. 45 Hamilton principle. 163. 78. 98. 164. 137. 192 nine constants model. 65. 64. 63. 195 interstitial atom. 71 pseudo-spectral methods. 118 plane cross section hypothesis. 66 movable singularities. 137 nonlinearity cubic. 125. 35 Murnaghan ﬁve constants model. 65. 192 plateau. 164 gradient. 117. 84 ﬁfth-order KdV equation. 124. 47 microdisplacement. 183 reference conﬁguration. 184 secularity conditions. 75 inertia of the microstructure. 117 Mooney-Rivlin model. 51. 136 Korteweg-de Vries-Burgers equation. 153 . 15. 96. 189 Kawahara equation. 7. 64. 188 kinetics. 71. 188 invariants of the strain tensor. 66. 159. 194 in presence of microstructure. 66 semi-inﬁnite. 26 laser radiation. 165 reference distorsion. 105. 152 Piola-Kirchoﬀ stress tensor. 165. 197 holographic interferometry method. 125. 79. 69 plate. 111. 186 soliton. 126. 96. 98. 157 friction contact. 191. 52 foundation models. 64 permafrost. 75 strain solitary wave reﬂection. 52 ﬁnite-diﬀerence methods. 126. 195 sliding contact. 53. 26. 15. 165 moduli. 12. 116. 79. 140 Korteweg-de Vries equation. 53 generalized. 52. 137. 165 Mindlin model. 64 physical. 48 shock. 70. 168 internal variables. 11. 179 modiﬁed. 137. 68. 124 Fredholm alternative. 87 Runge-Kutta method. 136. 165 microﬁeld. 117. 93 strain solitary wave propagation. 137. 153 fringe shift. 93. 93. 82 free lateral surface. 124 solitary waves interaction. 3. 10. 182 point defects. 123. 98. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 212 Book Title experiments on strain solitary wave ampliﬁcation. 3. 83 free end. 3. 195 heat conduction. 117. 52. 115 rod clamped end. 178 Ginzburg-Landau equation.

197 Fourier law. 2004 20:10 WSPC/Book Trim Size for 9in x 6in ws-book9x6 Chapter Title for Bibliography 213 surface tension. 196. 196 Thomas method. 124. 137. 195 thermal conduction. 166 Voigt. 71 . 193 coupled equations.February 11. 58 thresholds. 36 uniformly valid solution. 164. 185 truncated expansion. 177. 166 Young modulus. 95 thermacy. 152. 194 thermoelasticity. 133. 49 vacancy. 123. 188 variables fast. 150. 47 slow. 47 viscoelastic.