Introduction to Soliton Theory: Applications to Mechanics Fundamental Theories of Physics
An International Book Series on The Fundamental Theories of Physics: Their Clarification, Development and Application
Editor: ALWYN VAN DER MERWE, University of Denver, U.S.A.
Editorial Advisory Board: GIANCARLO GHIRARDI, University of Trieste, Italy LAWRENCE P. HORWITZ, Tel-Aviv University, Israel BRIAN D. JOSEPHSON, University of Cambridge, U.K. CLIVE KILMISTER, University of London, U.K. PEKKA J. LAHTI, University of Turku, Finland ASHER PERES, Israel Institute of Technology, Israel EDUARD PRUGOVECKI, University of Toronto, Canada FRANCO SELLERI, Università di Bara, Italy TONY SUDBURY, University of York, U.K. HANS-JÜRGEN TREDER, Zentralinstitut für Astrophysik der Akademie der Wissenschaften, Germany
Volume 143 Introduction to Soliton Theory: Applications to Mechanics
by Ligia Munteanu Institute of Solid Mechanics, Romanian Academy, Bucharest, Romania and Stefania Donescu Technical University of Civil Engineering, Department of Mathematics, Bucharest, Romania
KLUWER ACADEMIC PUBLISHERS NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW eBook ISBN: 1-4020-2577-7 Print ISBN: 1-4020-2576-9
©2005 Springer Science + Business Media, Inc.
Print ©2004 Kluwer Academic Publishers Dordrecht
All rights reserved
No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher
Created in the United States of America
Visit Springer's eBookstore at: http://ebooks.kluweronline.com and the Springer Global Website Online at: http://www.springeronline.com Contents
Preface ix
Part 1. INTRODUCTION TO SOLITON THEORY 1. MATHEMATICAL METHODS 1
1.1 Scope of the chapter 1 1.2 Scattering theory 1 1.3 Inverse scattering theory 12 1.4 Cnoidal method 17 1.5 Hirota method 25 1.6 Linear equivalence method (LEM) 31 1.7 Bäcklund transformation 39 1.8 Painlevé analysis 46
2. SOME PROPERTIES OF NONLINEAR EQUATIONS 53 2.1 Scope of the chapter 53 2.2 General properties of the linear waves 53 2.3 Some properties of nonlinear equations 59 2.4 Symmetry groups of nonlinear equations 62 2.5 Noether theorem 66 2.6 Inverse Lagrange problem 69 2.7 Recursion operators 73
3. SOLITONS AND NONLINEAR EQUATIONS 78 3.1 Scope of the chapter 78 3.2 Korteweg and de Vries equation (KdV) 78 3.3 Derivation of the KdV equation 86 vi INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS
3.4 Scattering problem for the KdV equation 90 3.5 Inverse scattering problem for the KdV equation 95 3.6 Multi-soliton solutions of the KdV equation 101 3.7 Boussinesq, modified KdV and Burgers equations 107 3.8 The sine-Gordon and Schrödinger equations 112 3.9 Tricomi system and the simple pendulum 115 121 Part 2. APPLICATIONS TO MECHANICS 4. STATICS AND DYNAMICS OF THE THIN ELASTIC ROD 121 4.1 Scope of the chapter 121 4.2 Fundamental equations 122 4.3 The equivalence theorem 132 4.4 Exact solutions of the equilibrium equations 134 4.5 Exact solutions of the motion equations 146
5. VIBRATIONS OF THIN ELASTIC RODS 149 5.1 Scope of the chapter 149 5.2 Linear and nonlinear vibrations 149 5.3 Transverse vibrations of the helical rod 155 5.4 A special class of DRIP media 159 5.5 Interaction of waves 163 5.6 Vibrations of a heterogeneous string 166
6. THE COUPLED PENDULUM 173 6.1 Scope of the chapter 173 6.2 Motion equations. Problem E1 173 6.3 Problem E2 177 6.4 LEM solutions of the system E2 180 6.5 Cnoidal solutions 185 6.6 Modal interaction in periodic structures 191
7. DYNAMICS OF THE LEFT VENTRICLE 197 7.1 Scope of the chapter 197 7.2 The mathematical model 198 7.3 Cnoidal solutions 206 7.4 Numerical results 209 7.5 A nonlinear system with essential energy influx 213 CONTENTS vii
8. THE FLOW OF BLOOD IN ARTERIES 220 8.1 Scope of the chapter 220 8.2 A nonlinear model of blood flow in arteries 221 8.3 Two-soliton solutions 228 8.4 A micropolar model of blood flow in arteries 235
9. INTERMODAL INTERACTION OF WAVES 242 9.1 Scope of the chapter 242 9.2 A plate with Cantor-like structure 243 9.3 The eigenvalue problem 248 9.4 Subharmonic waves generation 249 9.5 Internal solitary waves in a stratified fluid 255 9.6 The motion of a micropolar fluid in inclined open channels 259 9.7 Cnoidal solutions 265 9.8 The effect of surface tension on the solitary waves 269
10. ON THE TZITZEICA SURFACES AND SOME RELATED 273 PROBLEMS 10.1 Scope of the chapter 273 10.2 Tzitzeica surfaces 273 10.3 Symmetry group theory applied to Tzitzeica equations 276 10.4 The relation between the forced oscillator and a Tzitzeica 283 curve 10.5 Sound propagation in a nonlinear medium 285 10.6 The pseudospherical reduction of a nonlinear problem 291 References 298 Index 305 Preface
This monograph is planned to provide the application of the soliton theory to solve certain practical problems selected from the fields of solid mechanics, fluid mechanics and biomechanics. The work is based mainly on the authors’ research carried out at their home institutes, and on some specified, significant results existing in the published literature. The methodology to study a given evolution equation is to seek the waves of permanent form, to test whether it possesses any symmetry properties, and whether it is stable and solitonic in nature. Students of physics, applied mathematics, and engineering are usually exposed to various branches of nonlinear mechanics, especially to the soliton theory. The soliton is regarded as an entity, a quasi-particle, which conserves its character and interacts with the surroundings and other solitons as a particle. It is related to a strange phenomenon, which consists in the propagation of certain waves without attenuation in dissipative media. This phenomenon has been known for about 200 years (it was described, for example, by the Joule Verne's novel Les histoires de Jean Marie Cabidoulin, Éd. Hetzel), but its detailed quantitative description became possible only in the last 30 years due to the exceptional development of computers. The discovery of the physical soliton is attributed to John Scott Russell. In 1834, Russell was observing a boat being drawn along a narrow channel by a pair of horses. He followed it on horseback and observed an amazing phenomenon: when the boat suddenly stopped, a bow wave detached from the boat and rolled forward with great velocity, having the shape of a large solitary elevation, with a rounded well-defined heap of water. The solitary wave continued its motion along the channel without change of form or velocity. The scientist followed it on horseback as it propagated at about eight or nine miles an hour, but after one or two miles he lost it. Russell was convinced that he had observed an important phenomenon, and he built an experimental tank in his garden to continue the studies of what he named the wave of translation. The wave of translation was regarded as a curiosity until the 1960s, when scientists began to use computers to study nonlinear wave propagation. The discovery of mathematical solutions started with the analysis of nonlinear partial differential equations, such as the work of Boussinesq and Rayleigh, independently, in the 1870s. Boussinesq and Rayleigh explained theoretically the Russell observation and later reproduction in a laboratory experiment. Korteweg and de Vries derived in 1895 the equation for water waves in shallow channels, and confirmed the existence of solitons. x INTRODUCTION TO SOLITON THEORY : APPLICATIONS TO MECHANICS
An explosion of works occurred when it was discovered that many phenomena in physics, electronics, mechanics and biology might be described by using the theory of solitons. Nonlinear mechanics is often faced with the unexpected appearance of chaos or order. Within this framework the soliton plays the role of order. The discovery of orderly stable pulses as an effect of nonlinearity is surprising. The results obtained in the linear theory of waves, by ignoring the nonlinear parts, are most frequently too far from reality to be useful. The linearisation misses an important phenomenon, solitons, which are waves, which maintain their identity indefinitely just when we most expect that dispersion effects will lead to their disappearance. The soliton as the solution of the completely integrable partial differential equations are stable in collision process even if interaction between the solitons takes place in a nonlinear way. The unexpected results obtained in 1955 by Fermi, Pasta and Ulam in the study of a nonlinear anharmonic oscillator, generate much of the work on solitons. Their attempt to demonstrate that the nonlinear interactions between the normal modes of vibrations lead to the energy of the system being evenly distributed throughout all the modes, as a result of the equipartition of energy, failed. The energy does not spread throughout all the modes but recollect after a time in the initial mode where it was when the experiment was started. In 1965, Zabusky and Kruskal approached the Fermi, Pasta and Ulam problem from the continuum point of view. They rederived the Korteweg and de Vries equation and found its stable wave solutions by numerical computation. They showed that these solutions preserve their shape and velocities after two of them collide, interact and then spread apart again. They named such waves solitons. Gardner, Green, Kruskal and Miura introduced in 1974 the Inverse Scattering Transform to integrate nonlinear evolution equations. The conserved features of solitons become intimately related to the notion of symmetry and to the construction of pseudospherical surfaces. The Gauss–Weingarten system for the pseudospherical surfaces yields sine-Gordon equation, providing a bridge to soliton theory. A privileged surface related to the certain nonlinear equations that admit solitonic solutions, is the Tzitzeica surface (1910). Developments in the geometry of such surface gave a gradual clarification of predictable properties in natural phenomena. A remarkable number of evolution equations (sine-Gordon, Korteweg de Vries, Boussinesq, Schrödinger and others) considered by the end of the 19th century, radically changed the thinking of scientists about the nature of nonlinearity. These equations admit solitonic behavior characterized by an infinite number of conservation laws and an infinite number of exact solutions. In 1973, Wahlquist and Estabrook showed that these equations admit invariance under a Bäcklund transformation, and possess multi-soliton solutions expressed as simple superposition formulae relating explicit solutions among themselves. The theory of soliton stores the information on some famous equations: the Korteweg de Vries equation, the nonlinear Schrödinger equation, the sine-Gordon equation, the Boussinesq equation, and others. This theory provides a fascinating glimpse into studying the nonlinear processes in which the combination of dispersion and nonlinearity together lead to the appearance of solitons.
This book addresses practical and concrete resolution methods of certain nonlinear equations of evolution, such as the motion of the thin elastic rod, vibrations of the initial deformed thin elastic rod, the coupled pendulum oscillations, dynamics of the left PREFACE xi ventricle, transient flow of blood in arteries, the subharmonic waves generation in a piezoelectric plate with Cantor-like structure, and some problems of deformation in inhomogeneous media strongly related to Tzitzeica surfaces. George Tzitzeica is a great Romanian geometer (1873–1939), and the relation of his surfaces to the soliton theory and to certain nonlinear mechanical problems has a long history, owing its origin to geometric investigations carried out in the 19th century. The present monograph is not a simple translation of its predecessor which appeared at the Publishing House of the Romanian Academy in 2002. Major improvements outline the way in which the soliton theory is applied to solve some engineering problems. In each chapter a different problem illustrates the common origin of the physical phenomenon: the existence of solitons in a solitonic medium.
The book requires as preliminaries only the mathematical knowledge acquired by a student in a technical university. It is addressed to both beginner and advanced practitioners interested in using the soliton theory in various topics of the physical, mechanical, earth and life sciences. We also hope it will induce students and engineers to read more difficult papers in this field, many of them given in the references.
Authors PART 1
INTRODUCTION TO SOLITON THEORY
Chapter 1
MATHEMATICAL METHODS
1.1 Scope of the chapter This chapter introduces the fundamental ideas underlying some mathematical methods to study a certain class of nonlinear partial differential equations known as evolution equations, which possess a special type of elementary solution. These solutions known as solitons have the form of localized waves that conserve their properties even after interaction among them, and then act somewhat like particles. These equations have interesting properties: an infinite number of local conserved quantities, an infinite number of exact solutions expressed in terms of the Jacobi elliptic functions (cnoidal solutions) or the hyperbolic functions (solitonic solutions or solitons), and the simple formulae for nonlinear superposition of explicit solutions. Such equations were considered integrable or more accurately, exactly solvable. Given an evolution equation, it is natural to ask whether it is integrable, or it admits the exact solutions or solitons, whether its solutions are stable or not. This question is still open, and efforts are made for collecting the main results concerning the analysis of nonlinear equations. Substantial parts of this chapter are based on the monographs of Dodd et al. (1982), Lamb (1980), Drazin (1983), Drazin and Johnson (1989), Munteanu and Donescu (2002), Toma (1995) and on the articles of Hirota (1980) and Osborne (1995).
1.2 Scattering theory Historically, the scattering theory was fairly well understood by about 1850. It took almost one hundred years before the inverse scattering theory could be applied. Since 1951, various types of nonlinear equations with a soliton as a solution have been solved by direct and inverse scattering theories. However, given any evolution equation, it is natural to ask whether it can be solved in the context of the scattering theory. This question is related to the Painlevé property. We may say that a nonlinear 2 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS partial differential equation is solvable by inverse scattering technique if, and only if, every ordinary differential equation derived from it, by exact reduction, satisfies the Painlevé property (Ablowitz et al.). The Painlevé property refers to the absence of movable critical points for an ordinary differential equation. Let us begin with the equation known as a Schrödinger equation, of frequent occurrence in applied mathematics (Lamb)
MOxx [(,)]uxt M 0 , (1.2.1) whereMo:R R is a dimensionless scalar field in one space coordinate x . The potential function uxt(,) contains a parameter t , that may be the temporal variable, t t 0 . At this point, t is only a parameter, so that the shape of ux(,)t varies from t . Subscripts that involve x or t are used to denote partial derivatives, for example wu wu u , u . t wt x wx If the function u depends only on x , axbd d , where a and b can be infinity, the equation (1.2.1) for imposed boundary conditions at x a and b , leads to certain values of the constant O (the eigenvalues O j ) for which the equation has a nonzero solution (the eigenfunctions M j ()x ). For a given function ux() , the determination of the dependence of the solution M on the parameter O and the dependence of the eigenvalues O j on the boundary conditions is known as a Sturm-Liouville problem. The solutions of (1.2.1) exist only if b the function ux() is integrable, that is |()|dux x f. The spectrum of eigenvalues O ³ j a is made up of two cases corresponding to O!0 and O 0 . The case O 0 does not occur if ux()z 0. In particular, for ux( ) 2sech2 x , and the boundary conditions M()0rf leads to the single eigenvalue O 1 with the associated eigenfunction M sech x . The scattering solutions of (1.2.1) are made up of linear combinations of the functions
M1 exp(i Ox )(iO tanhx ) , and M2 exp( i Ox )(i O tanhx ) . The solving of the Schrödinger equation (1.2.1) when the potential function ux() is specified is referred to as the direct scattering problem. If u depends on x and t , uuxt (,), then we expect the values of the O j to depend upon t . It is interesting to ask whether or not there are potential functions uxt(,) for which the O j remain unchanged as the parameter t is varied.
In particular, if uuxt ( ) satisfies the linear partial differential equation uux t , the variation of t has no effect upon the eigenvalues O j . Also, the eigenvalues are invariant to the variation of t , if ux(,)t satisfies the nonlinear partial differential equation
uuuutxxxx 0 , (1.2.2) MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 3 known as the Korteweg–de Vries equation (KdV) . Therefore, solving the KdV equation is related to finding the potentials in a Sturm- Liouville equation, and vice versa. The direct scattering problem is concerned with determining of a wave function M when the potential u is specified. Determination of a potential u from information about the wave function M is referred to as the inverse scattering problem.
THEOREM 1.2.1 LetS be a pre-hilbertian space of functions y :R2 o R . Let us consider the operators LS: o S , BS: o S having the properties:
a) Ly12,, y y 1 Ly 2, yy12, S. b) L admits only simple eigenvalues, namely O()t is an eigenvalue for L if there exists the function <S , so that Lxttx,(,)()(,< O c) Baty,() at () By , , y S , and at() y S. It follows that the relations LLBBLt 0 , (1.2.4) LS: S , Lyxt , (,) Ly , (,) xt Lyxt , (,), (1.2.5) tto ,t t are verified. Also, it follows that 1. the eigenvalues are constants O()t O R, t R , (1.2.6) 2. the eigenfunctions verify the evolution equation Lxttx,(,)()(,< O LxtLxttxttxttt,<< (,) , (,) O<O< () (,) ()t (,)t , LxtLxttxttxttt,<< (,) , (,) O<O< () (,) ()t (,)t , xt,R. From (1.2.4) and (1.2.5) it results O This implies Ot 0 . The function < < Mt B,M . (1.2.8) To illustrate this, let us consider the example ^`yyyx: R×R+ ooof R, ,x 0, , with the scalar product f yy,(,)( yxtyxt,)dx, 12S ³ 1 2 f and operators LB,: So S Ly,( yxx uxty,) , (1.2.9) B,46(,)3(, y yxxx uxt y x u x xt) y. (1.2.10) According to Lytt,(, uxty) , 2 L, By 4 yxxxxx 10 uy xxx 15 u x y xx 12 u xx y x 6 u y x 3 u xxx y 3 uu x y , 2 B, Ly 4 yxxxxx 10 uy xxx 15 u x y xx 12 u xx y x 6 u y x 4 u xxx y 9 uu x y , it is found that (1.2.3) can be written under the standard form of the KdV equation uutxxxx60 uu . The operators satisfy the properties mentioned in the theorem 1.2.1. For (1.2.5) we find <O Mxx O()u M 0 , (1.2.11) which verifies the equation Mtxxxx 463M uuM xM . (1.2.12) Consequently, finding solutions to the KdV equation is related to solving the Schrödinger equation 2 Mxx [(,)]kuxtM 0 . (1.2.13) Note that in (1.2.13), t is playing the role of a parameter, k is a real or a pure complex number ik ,k ! 0 , and the potential function u has the property uxoof0, . For localized potentials ux() , all solutions of (1.2.13) will reduce to a linear combination of the functions exp(ikx ) , and exp( ikx ) as x of. Following Faddeev (1967), the solutions of the Schrödinger equation are expressed as linear combinations of a solution fxk1 (,) that reduces to exp(ikx ) , as x of, and a solution fxk2 (, ) that reduces to exp(i kx ), as x of. By definition, fxk1 (,) and fxk2 (, ) are fundamental solutions of (1.2.13) and are exact solutions of (1.2.13) and verify fxk1 (,)|of exp(i), kxx , fxk1 (,)exp(io kx ) 1, x of, (1.2.14a) fxk2 (, )| exp(i kxx ), of, fxk2 (,)exp(i) kxoof 1, x . (1.2.14b) THEOREM 1.2.2 Fundamental solutions fxk1 (,) andfxk2 (, ) verify the equations 1 f fxk(, ) exp(i) kx sin( kx DDD )( u ) f (, k )dD , (1.2.15a) 11³ k x 1 x fxk(,) exp(i kx ) sin( kx DDD )( u ) f (,)d kD . (1.2.15b) 21³ k f Proof. The homogeneous equation associated to (1.2.13), M'' k 2M 0 , admits the solutions M()x A ()exp(i) x kx B ()exp(i x kx ), with A, B arbitrary constants. By applying the method of variation of constants, we obtain 1 A'(xuxx ) ( )M ( ) exp( ikx ) , 2ik 6 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS 1 Bx'( ) ux ( ) M ( x ) exp(i kx ) , 2ik and then, by integration, we have 1 x A()xu ()()exp(iDM DDD k )d C , ³ 1 2ik 0 1 x Bx( ) u ( D )M (DDD )exp(i k )d C . ³ 2 2ik 0 The constants CC1, 2 are found from (1.2.14a) 1 f Cu 1()()exp(i DM DD k)dD , 1 ³ 2ik 0 1 f Cu ()()exp(i)dDM DD kD . 2 ³ 2ik 0 Substitution of these expressions into M()x , leads to 1 f M(xk , ) exp(i kx )D³ u ( )M (DD )expi kx ( )dD 2ik x 1 f DMDDD³ukx()()expi( )d. 2ik x The function f2 is derived in an analogous manner. Equations (1.2.15) are the Volterra integral equations, which can be solved by an iteration procedure. More specifically, the substitution of exp(ikx ) into (1.2.15a) yields to the conclusion that the resulting integrals converge for Im(k )! 0 . For integral equations of Volterra, the resulting series expansion is always convergent. Hence, the functions ff12, are analytic in the upper half of the complex k plane. For real ux() and k , we have fii(,xk ) fxk (, ),i 1, 2 , where “ ” is the complex conjugate operator. From (1.2.15) we see that the functions fxk1 (,),fx1 (, k ) are independent. The functionsfxk2 (, ),fxk2 (, ) are also independent. So, there exist the coefficients ckij (), ij,1,2 , depending on k, so that fxk2111121(,) ckfxkckfxk () (,) () (, ), (1.2.16a) fxk1212222(,) c () kfxk (,) c () kfx (, k ). (1.2.16b) From the limiting form of fxi (,r k ), i 1,2, we may write fxk2 (,)| exp(i kxx ), of , MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 7 fxk21112( , )|of ck ( )exp(i kxck ) ( )exp( i kxx ), , which means the solution corresponds to a scattering problem in which the incident wave is coming from f with an amplitude ck12 () , and is reflected with an amplitude ck11 (), and transmitted to f with an amplitude of unity. In particular, the fundamental solutions for the potential ux( ) 2sech2 x , are obtained by solving the equation zkcc (2sech)22 xz , f x f. By using the substitution yx tanh , 11x , we obtain the associated Legendre equation (Drazin and Johnson) ddzk2 [(1yz2 ) ] ( 2 ) 0 , dd1yy y2 whose general solution is given by zA exp( kxkyB )( ) exp( kxky )( ) . From here we obtain the fundamental solutions 1 fxk( , ) exp(i kxk )(i tanh x ) , 1 i1k 1 fxk( , ) exp( i kxk )(i tanh x ) . 2 i1k Let us introduce the reflection and transmission coefficients for an incident wave of unit amplitude (Achenbach). The ratio ck11 () RkR () , (1.2.17) ck12 () is the reflection coefficient at f , and the ratio 1 TkR () , (1.2.18) ck12 () is the transmission coefficient at f . The subscript R refers to a wave incident from the right. Similarly, we have fxk1 (,)|of exp(i), kxx , fxk12122( , )|of c ( k )exp(i kxc ) ( k )exp( i kx ), x , which means the incident wave from f with an amplitude ck21 () is reflected with an amplitude c22 ()k and transmitted to f with an amplitude of unity. The ratio 8 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS ck22 () RkL () , (1.2.19) ck21 () is the reflection coefficient at f , and the ratio 1 TkL () , (1.2.20) ck21 () is the transmission coefficient al f . The Wronskian of any two functions M1 and M2 , is defined as wx[(),()]MM12 x MM 1,2x ()() xx MM 12, () xx () x. (1.2.21) If M1 and M2 are two linearly independent solutions of (1.2.13), then their Wronskian is a constant wxkxkfk[MM12 (;),(;)] (), k C . (1.2.22) The relation (1.2.22) results by adding (1.2.14) written for M1 and multiplied by M2 , to (1.2.13) written for M2 and multiplied by M1 . It results d w[,MM ]0 . dx 12 According to definition of the Wronskian and (1.2.22), the following properties hold wf[(;),(;)]2i11 xk f x k k, wf[(;),(;)]2i22 xk f x k k, (1.2.23) whereff12, are fundamental solutions (1.2.15). Substituting fxk1 (; ) from (1.2.16b) into (1.2.16a), and substituting fxk2 (; ) from (1.2.16a) into (1.2.16b), k C , and taking account of the independency of fxkfx22(; ),(,k) , the following relations are obtained ckck11() 22 () ckc 12 () 21 ( k ) 1, ckck11() 21 () ckc 12 () 22 ( k ) 0, (1.2.24) ckck21() 12 ( ) ckck 22 () 11 () 1, ckck21() 11 ( ) ckck 22 () 12 () 0. The coefficients cij may be written in terms of the Wronskian 1 ck( ) wfxkfxk [ ( ; ), ( ; )], 112ik 2 1 1 ck() wfxkfxk [ (; ),(;)], (1.2.25) 222ik 2 1 MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 9 1 ck( ) ck ( ) wfxkfxk [ ( ; ), ( ; )], 12 212ik 1 2 k C . That yields TkRL() Tk () Tk (). The relations (1.2.25) are obtained from (1.2.16) and (1.2.23). Furthermore, from fxkii(; ) fxki (; ), 1,2, we have RkTkRL()( ) R ( kTk )() 0, k C , (1.2.26) and ckck12() 12 (), ck11() c 22 () k c 22 ( k ), 22 2 ck12() 1 ck 11 () 1 ck 22 () , (1.2.27) 2222 Tk() RRL () k Tk () R () k 1, RkRkRkRkLLRR() (),() (), k R . The location of the poles of the transmission and reflection coefficients in the upper half-plane are important to obtain information about the localized or bound-state solutions. Consider now the poles of Tk() . THEOREM 1.2.3 For real potential functions u :Ro R , any poles of the transmission coefficient in the upper half-plane must be on the imaginary axis. More precisely, if kC0 is a pole for Tk(), then k00 i,N N0 R . 1 Proof. Let k0 C be a pole for Tk() . Then it is a zero for c12 , ck12 () ck12() 0 0. According to (1.2.27) 3 , for k R we have ck12 ()z 0, and then Imk0 ! 0 . Writing (1.2.13) for k0 2 M''M (ku0 ) 0 , and similarly for k0 2 M ''M (ku0 ) 0 , and subtracting them, it follows that 22 MMMM'' '' (kk00 ) MM. Integrating then over x , from f to f , yields f 22 2 ww(,MM ) (, MM ) ( kk00 ) M dx . f f ³ f The Wronskian of MM, being a constant, it follows that 10 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS f 2 Rekk ImM d x 0 , 00³ f and hence, from Im k0 ! 0 , it results Rek0 0 . When ck12() 0 0 ,kC0 , the fundamental solutions are linearly dependent, and then lead to fxk2011010(, ) ckfxk ( ) (, ), (1.2.28) and 1 ck22() 0 . (1.2.29) ck11() 0 This property results from (1.2.16a,b) written for k k0 . Next, we show that the value of the residuum of the function Tk() in every pole klll i,N N! 0 is given by 1i Res(Tk ( ))( k ) . (1.2.30) l ck () f 12 l fxkfxkx(, ) (, )d ³ 12ll f To obtain this, let us differentiate with respect to k , the relation (1.2.25)3 and set kk l . According to (1.2.28) and (1.2.29) we have d1 1 ck() ckwxk ( ) (, ) ckwxk ( ) (, ), (1.2.31) d2i2ikk12 11ll 1 k 22 l 2 l kk l ll where wf wxk(, ) w [i (, xk ) ; fxk (, )], i 1,2. ilwk il kk l To obtain w1 , let us multiply (1.2.13) written for fxk1 (,) withfxk1 (,l ) , then multiply (1.2.13) written for fxk1 (,l ) with fxk1 (,) , and add the results. We have ww w [f (, xk ) f (, xk ) f (, xk ) f (, xk )]( k22 k ) f (, xk ) f (, xk ) 0. wwxx1111ll w x l 11 l Differentiating the above relation with respect to k R , we have ww wfxk[(,); fxk (,) ]2((,)) kfxk 2 . xk11ll 1l wwkk l Integration from x to f , gives w f Awfxk[(,); fxk (,)]2 k ((,))d fDD k 2 , 11llll³ 1 wk x MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 11 w Awfxkfxk lim[11 (,ll ); (, )]0 . xof wk It follows that f wkfk 2[(;)]dDD2 . 11ll³ x In a similar way we obtain x wkfk 2[(;)]dDD2 . 22ll³ f Substitution ofwii ,1, 2 , into (1.2.31) yields f ck11 ()l 2 ck12() [ f 1 (;DD kl )]d kk l ³ i f 1 ffck() [(;)](;)dfkfkDDD 22 l [ f (;)]d.D k2 D ³³12ll 21 l iif f Note that f ck()[(;)]dJ fxk2 x 1, (1.2.32a) ³ 11ll 1 l f f ck()[J fxk (;)]d2 x 1. (1.2.32b) ³ 22ll 2 l f 1/2 1/2 Thus, the quantities [(Jllck11 )] and [(Jllck22 )] are the normalization constants for the bound-state wave functions fxil(,k ),i 1,2 . Using (1.2.30) we may write the normalization constants as ck() f mck J() i11 l {[(;)]d} fxkx2 1 , (1.2.33a) Rl l11 l³ 1 l ck12 ()l f ck() f mck J() i22 l {[ fxkx (;)]d}2 1 , (1.2.33b) Ll l22 l³ 2 l ck12 ()l f wheremmRl, Ll R(,)R,i,1,2 , due to the fact that fxkil k l N l i . Any poles of the transmission coefficient are simple because, if kl is a pole for Tk(), kll iN,N!l 0 , then it has the properties (1.2.32a), (1.2.33a), and it results f ck () i ck ()[(;)]d fxk2 x z 0. 12lll 11³ 1 f 12 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS 1.3 Inverse scattering theory The inverse scattering theory was firstly considered to solve an inverse physical problem of finding the shape of a mechanical object, which vibrates, from the knowledge of the energy or amplitude at each frequencies (Drazin and Johnson). In our terms, the methods consist in determination of the potential function u from given coefficients ckij () , that relate the fundamental solutions of the equation M''M (ku2 ) 0 . (1.3.1) The fundamental solutions of the Schrödinger equation may be written under the form f fxk( , ) exp(i kx ) Axx ( , ')exp(i kxx ')d ' , (1.3.2a) 1 ³ R x x fxk(,) exp(i) kx Axx (, ')exp(i')d kxx' . (1.3.2b) 2 ³ L f Balanis considered these forms in 1972, by solving the elastically braced vibrating string equation yyxx x'' x uxy() 0, (1.3.3) for which the solutions are written as yxx1 (, ') G (' x x ) T (' x xAxx )R (, '), (1.3.4a) yxx2 ( , ') G ( x ' x ) T ( x ' xAx )L ( , x '), (1.3.4b) whereGT is the Dirac function, is the Heaviside function, and AR , AL are functions that describe the scattering or wake. Applying the Fourier transform f F[(,')](,)yxx xk ³ yxx (,')exp(i')d' kx x, f to (1.3.3), we find (Lamb) 2 Fyxx[ ( , ')]xx ( k ux ( )) Fyxx [ ( , ')] 0 . (1.3.5) The equation (1.3.5) admits as solutions the Fourier transform of (1.3.4) f F[yxx ( , ')]( xk , ) exp(i kx ) Axx ( , ')exp(i kxx ')d ' , (1.3.6a) 1 ³ R x x F[yxx ( , ')]( xk , ) exp( i kx ) Ax ( , x ')exp( i kxx ')d ' . (1.3.6b) 2 ³ L f From (1.3.5), we see that the solutions of (1.3.1) take the form (1.3.6) fxkii(, ) Fyxx [ (, ')](, xk ), i 1,2. MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 13 Substituting ff12, into (1.3.1) we derive the conditions to be verified by AR , AL . For this, we writef1 as i1wA fxk(,) exp(i)[1 kx Axx (,') R (,')] xx 1 kxRxk 2 w ' ' x 1 f w2 A R (,xx ')exp(i')d', kx x 22³ kxx w ' then integrate it by parts and introduce into (1.3.1). By imposing the conditions AxxRRx(, '), A' (, xx ')oo 0,for x ' f , we find dA exp(ikx )[2R ( x , x ) u ( x )] dx f ww22AA [RR (,')xx (,')() xx uxA (,')]exp(i')d' xx kx x 0. ³ 22 R x wwxx' Therefore, the equation (1.3.1) is verified for dA ux() 2R (, xxc ), dx AR (,xx ')0, x ' x , (1.3.7a) ww22AA RR(,x xxxuxAxxx ') (, ') () (, ')0, '! x . wwxx22' R Similarly, f2 verifies the equation (1.3.1) for dA ux() 2L (, xxc ), dx AL (,xx ')0, x '! x , (1.3.7b) ww22AA LL(,x xxxuxAxxx ') (, ') () (, ')0, ' x , wwxx22' L and the Faddeev condition is verified (Faddeev 1958) f ³ (1 fxux )()d x . f From (1.3.7) we see that for given AR , AL , we can find the potential function u. Next, we try to determine the functions AR , AL in terms of the coefficients, ckij (),, ij 1,2, considered specified. For this we write (1.2.14a) under the form 14 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS Tk() f211 (, xk ) RR () k f (, xk ) f (, x k ), (1.3.8) and derive the corresponding relation in the time domain. Taking the Fourier transform does this 1 f F 1[(,)](,')fxk xx fxk (,)exp(i kxk ')d. ³ 1 2S f By noting 1 f *()zTkkz ³ [ () 1]exp(i )dk , (1.3.9) 2S f the Fourier transform on the left-hand side of (1.3.8) yields 1 ff F 1[()Tk f (,)](,') xk xx (1* ()exp(i)d) z kzz u 2 ³³ 2S f f f u(yxx ( , '') exp(i kxx '')d '') exp( i kxk ')d ³ 2 f 1 yxx( , ')* ( zyxx ) ( , '') exp[i kzx ( '' x ')]d kxz d ''d 22³³³ 2S f, f G('xx ) T (' xxAxx )L (, ') f (1.3.10) *(x xxxAxx ') * ( ' '') ( , '')dx ''. ³ L x In a similar way, by noting 1 f rz() Rk ()exp(i)d kzk, (1.3.11) RR³ 2S f the Fourier transform on the right-hand side of (1.3.8) leads to 1 FRkfxkfxkxx[R ()11 (,) (, )](,') 1 f Rkfxk() (,)exp(i kxk ')d yxx (, ') ³ R 11 2S f 1 ff f (r ( z )exp(i kz )d z )( y exp(i kx '')d x '')exp( i kx ')d k y ( x , x ') ³³R ³1 1 2S f f f f 1 rzyxx( ) ( , ''){ exp[i k ( z x '' x ')]d k }d x ''d zyx ( , x ') ³³R 11 ³ f, f f 2S f rx(''')(,'')d''(, xyxx x yx x ') ³ R 11 f MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 15 f rx( x ') rx ( '' xAxxx ') ( , '')d '' RRR³ x G('x xxxAxx ) T ( ' )R (, '). (1.3.12) Substituting (1.3.10) and (1.3.12) into (1.3.8) we have f T(')(,')(xxAxx * xx ') * (''')(,'')d'' xxAxx x LL³ x (1.3.13) f rx( x ') rx ( '' xAxxx ') ( , '')d '' T ( x ' xAx ) ( , x '). RRR³ R x We study now the case xx '0. To interpret (1.3.13) we must evaluate the function *()z , z 0 . Case 1. When the transmission coefficient Tk() possesses neither poles nor zeros in the upper half-plane, then *()z 0. (1.3.14) To show this, we consider the closed contour in the complex plane CC R [, RR] where CR a semicircle of radius R . According to the Cauchy theorem we have ³[Tk ( ) 1]exp( i kzk )d 0 , C due to the fact that the integrant is an olomorphic function in the simple convex domain enclosed by the contour C . Then, we can write R ³³[Tk ( ) 1]exp( i kzk )d [ Tk ( ) 1]exp( i kzk )d . (1.3.15) CRR According to the Jordan lemma, if CR is a semi-circle in the upper half-plane, centered in zero and having the radius R, and the function Gk() satisfies the condition Gk()oof 0, k , in the upper half-plane and on the real axis, and m is a positive real number, then we have ³ Gk()exp(i)d kmkoo 0, R f . CR Here mz R ,Gk() Tk () 1 . If Tk()oof 1, k , we may write ³ [Tk ( )oo 1]exp( i kzk )d 0, R f . CR From (1.3.15) it results 16 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS R *()zTkkzk lim[ () 1]exp(i )d 0. Rof ³ R When Tk() contains no poles in the upper half-plane, the equation (1.3.13) becomes f rx( x ') rx ( '' xAxxx ') ( , '')d '' Ax ( , x ') 0, x x ' 0 , RRRR³ x or, denoting x ' y f rx( y ) rx ( '' yAxxx ) ( , '')d '' Axy ( , ) 0, x y. (1.3.16a) RRRR³ x In a similar way we obtain x rx( y ) rx ( '' yAxx ) ( , '')d x '' Axy ( , ) 0, x ! y, (1.3.16b) LLLL³ f where 1 f rz() Rk ()exp(i kzk )d . LL³ 2S f Case 2. If Tk()contains first-order zeros or poles in the upper half-plane, then they are situated on the imaginary axis kllll iN , N! 0, 1,2,..., n . From the residuum theorem we find f 1 n exp( ikz ) *()zTkkzk [ () 1]exp(i )d 2iS l Rez[Tk ( )] ³ ¦ l f 2S l 1 2S nn i exp(kz )i exp( kz ) , (1.3.17) ¦¦llJ llJ ll 11 whereJl is given by (1.2.31). The relation (1.3.13) is written as nnf JJexp[kx ( x ')] exp[ kx ( x '')] Ax ( , x '')d x '' ¦¦ll³ l lL ll11 x (1.3.18) f rxx( ') rx ( '' xAxxx ') ( , '')d '' Ax ( , x '), RRRR³ x forxx '0 . In terms of x ' y , and taking into consideration that MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 17 f fx( ,iN ) c (i N ) fx ( ,i N ) c (i N )[exp( N x ) Axx ( , ')exp( N x '')d x ''], 211111lllllR³ l x the equation (1.3.18) becomes f ::(x y ) ( x '' yAxxx ) ( , '')d '' Axy ( , ) 0 , x y . (1.3.19a) RRRR³ x Similarly, we obtain x ::(x yxyAxxxAxyx ) ( '' ) ( , '')d '' ( , ) 0, !y . (1.3.19b) LLRL³ f The functions :R ()z and :L ()z are defined as n :RR(zrz ) ( ) JN¦ ll c11 (i )exp( kz l ) l 1 (1.3.20a) 1 f ck() n 11 exp(ikz )d kN i m (i )exp(N z ), ³ ¦ Rl l l 2()S f ck12 l 1 1 f ck() n :(zkzkm ) 22 exp( i )d i (i N )exp( Nz ) , (1.3.20b) LL³ ¦ ll l 2()S f ck12 l 1 with mRl and mLl given by (1.2.31) c11 (iNl ) c22 (iNl ) mRl(iN l ) i , mLl(iN l ) i . c12 (iNl ) c12 (iNl ) In this case we have obtained the same integral equations (1.3.19) as in the first case, with the difference that rR is replaced to :R . These equations are known as Marchenko equations (Agranovich and Marchenko), and they can be used to determine AR or AL when one of the reflection coefficients rR or rL is specified. Solutions of Marchenko equations are the functions AR , AL , which allow the determination of the potential function u . The Marchenko equations are also used to determine the reflection coefficients when the potential and hence the fundamental solutions and the functions AR or AL are specified. We can say that determination of RRR , L and rrR , L , is made from (1.2.15) and (1.2.17), and determination of ::R , L from (1.3.20). 1.4 Cnoidal method The inverse scattering theory generally solves certain nonlinear differential equations, which have cnoidal solutions. The mathematical and physical structure of the inverse scattering transform solutions has been extensively studied in both one and two 18 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS dimensions (Osborne, Drazin and Johnson, Ablowitz and Segur, Ablowitz and Clarkson). The theta-function representation of the solutions is describable as a linear superposition of Jacobi elliptic functions (cnoidal functions) and additional terms, which include nonlinear interactions among them. Osborne is suggesting that the method is reducible to a generalization of the Fourier series with the cnoidal functions as the fundamental basis function. This is because the cnoidal functions are much richer than the trigonometric or hyperbolic functions, that is, the modulus m of the cnoidal function, 0 ddm 1 , can be varied to obtain a sine or cosine function (m # 0)), a Stokes function (m # 0.5 or a solitonic function, sech or tanh (m # 1) (Nettel). Since the original paper by Korteweg and DeVries, it remains an open question (Ablowitz and Segur): “if the KdV linearised equation can be solved by an ordinary Fourier series as a linear superposition of sine waves, can the KdV equation itself be solved by a generalization of Fourier series which uses the cnoidal wave as the fundamental basis function?” This method requires brief information necessary to describe the cnoidal waves. The arc length of the ellipse is related to the integral z (1 kx22 )d x Ez() , ³ 2 0 (1 x ) with 0 k 1 . Another elliptical integral is given by z dx Fz() . ³ 222 0 (1x )(1kx ) The integralsEz() and F()z are Jacobi elliptic integrals of the first and the second kinds. Legendre is the first who works with these integrals, being followed by Abel (1802–1829) and Jacobi (1804–1851). Jacobi inspired by Gauss, discovered in 1820 that the inverse of F()z is an elliptical double-periodic integral F 1 ()Z sn(). Z Jacobi compares the integral M dT v , (1.4.1) ³ 21/2 0 (1Tm sin ) where 0 ddm 1 , to the elementary integral M dt w , (1.4.2) ³ 21/2 0 (1 t ) and observed that (1.4.2) defines the inverse of the trigonometric function sin if we use the notations t sin T and \ sin w . He defines a new pair of inverse functions from (1.4.1) snv sin M, cnv cos M. (1.4.3) MATHEMATICAL METHODS FOR NONLINEAR EQUATIONS ANALYSIS 19 These are two of the Jacobi elliptic functions, usually written sn(vm , ) and cn(vm , ) to denote the dependence on the parameter m . The angle M is called the amplitude M amu . We also define the Jacobi elliptic function dnvm (1sin21 M ) / 2 . Form 0 , we have v M, cn(vv ,0) cosM cos , v M sn(vv ,0) sinM sin , dn(v ,0) 1 , (1.4.4) and for m 1 v arcsech(cosM ) , cn(vv ,1) sech , sn(vv ,1) tanh , dn(vv ,1) sech . (1.4.5) The functions sn v and cn v are periodic functions with the period 2/2SSddTT 4. ³³2 1/2 2 1/2 00(1Tmm sin ) (1 T sin ) The later integral is the complete elliptic integral of the first kind S /2 dT Km() . (1.4.6) ³ 21/2 0 (1Tm sin ) The period of the function dn v is 2K . For m 0 we have K(0) S / 2 . For increasing of m , K()m increases monotonically 116 Km()| log . 21 m Thus, this periodicity of sn (v ,1) and cn (vv ,1) sech is lost for m 1 , so Km()of. Some important algebraic and differential relations between the cnoidal functions are given below d cn22 +sn 1, dn22 +m sn 1, cn= sn dn , dv d d sn= cn dn , dn= m sn cn , (1.4.7) dv dv where the argument vm and parameter are the same throughout relations. Now, consider the function ()t introduced by Weierstrass (1815–1897) in 1850, which verifies the equation 23 4g2 g3 , (1.4.8) where the superimposed point means differentiation with respect to t . 20 INTRODUCTION TO SOLITON THEORY: APPLICATIONS TO MECHANICS 3 Ifeee123,, are real roots of the equation 40ygyg 23 with eee12!!3 , then (1.4.8) can be written under the form 2 4( eee12 )( )( 3 ) , (1.4.9) with 222 g212 2(eee3 ) , g312 4eee3 , ee123e 0 . Introducing 3 2 ' g2 27g3 , (1.4.10) when'!0 , equation (1.4.9) admits the elliptic Weierstrass function as a particular solution, which is reducing in this case to the Jacobi elliptic function cn 2 (;,)(tggeeeGcc23 2G 2 3 )cn( eet 1 3 ) , (1.4.11) whereGc is an arbitrary real constant. If we impose initial conditions to (1.4.9) (0) T0 , c (0) Tp0 , (1.4.12) then a linear superposition of cnoidal functions (1.4.11) is also a solution for (1.4.8) n 2cn[;2 tm] Tlin ¦ D k Z k k , (1.4.13) k 0 where the angular frequencies Zk , and amplitudes Dk depend on T0 , T p0 . When ' 0 the solution of (1.4.9) is 1cn(2GtH2 c ) eH22 , 1cn(2GtH2 c ) with 1 3e2 2 g2 m , He22 3 . 24H 2 4 When ' 0 , we have ee12 cc, e3 2 , and the solution of (1.4.9) is 3c c . sinh2 ( 3ct Gc ) Since the calculation of the elliptic functions is very important for practical problems, in Chapter 10, the Shen-Ling method to construct a Weierstrass elliptic function from the solutions of the Van der Pol’s equation is presented. Consider now a generalized Weierstrass equation with a polynomial of n degree in T()t