E Thesis

Asymptotic Models for Long Wave Motion in a Pre-stressed Incompressible Elastic Plate
Aleksey V. Pichugin
Research Centre for Applied and Discrete Mathematics University of Salford, Salford, UK
Submitted in Partial Fulllment of the Requirements of the Degree of Doctor of Philosophy, November 2001
Contents
Introduction Notes on notation 1 Mathematical preliminaries 1.1 Congurations of a pre-stressed body . . . . . . . . . . . . . . . . . 1.2 Motion of a pre-stressed incompressible elastic body . . . . . . . . 1.3 Propagation of plane waves, strong ellipticity . . . . . . . . . . . . 1.4 Explicit necessary and sucient conditions for the strong ellipticity 1.4.1 Necessary and sucient conditions for particular directions 1.4.2 General conditions: necessity . . . . . . . . . . . . . . . . . 1.4.3 General conditions: suciency . . . . . . . . . . . . . . . . 1.5 Separation of variables: secular equation . . . . . . . . . . . . . . . 1.6 Some specic material models for rubber-like materials . . . . . . . 1.6.1 Neo-Hookean materials . . . . . . . . . . . . . . . . . . . . 1.6.2 Mooney-Rivlin materials . . . . . . . . . . . . . . . . . . . . 1.6.3 Varga materials, separability of the strain energy function . 1.6.4 Ogden materials . . . . . . . . . . . . . . . . . . . . . . . . 1.7 Measure of incremental surface traction . . . . . . . . . . . . . . . 2 Dispersion of harmonic waves in a layer with free faces 2.1 Harmonic wave propagation in a pre-stressed incompressible elastic 2.2 The dispersion relations . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Extensional (symmetric) motion . . . . . . . . . . . . . . . 2.2.2 Flexural (anti-symmetric) motion . . . . . . . . . . . . . . . 2.3 Neutral curves, stability . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Propagation of long waves . . . . . . . . . . . . . . . . . . . . . . . 2.5 Long wave high frequency approximations of the secular equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 8 8 9 11 13 13 15 17 17 20 21 21 22 22 23 25 25 28 29 30 31 35 37 39 39 40 41 42 42 44 45 47 48 49 50 51
layer . . . . . . . . . . . . . . . . . .
3 Layer with free faces: Long wave high frequency extensional motion 3.1 Analysis of the dispersion relation . . . . . . . . . . . . . . . . . . . . . . 3.1.1 Approximation near cut-o frequencies . . . . . . . . . . . . . . . . 3.1.2 Relative asymptotic orders of displacements . . . . . . . . . . . . . 3.1.3 Physical interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Asymptotically approximate equations: rst family of shear resonances . . 3.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Asymptotically approximate equations: second family of shear resonances 3.3.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . . . . i
CONTENTS 3.4 Illustrative numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ii 52 56 56 57 59 59 60 61 63 65 66 67 71 71 72 73 74 74 75 76 79 82 84 84 85 86 86 87 88 89 91 92 93 94 95 97 97 98 99 100 101 103 103 104 106 107 108
4 Layer with free faces: Long wave high frequency exural motion 4.1 Analysis of the dispersion relation . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Approximation near shear resonance frequencies . . . . . . . . . . 4.1.2 Relative asymptotic orders of displacements . . . . . . . . . . . . . 4.1.3 Physical interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Asymptotically approximate equations: rst family of shear resonances . . 4.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Asymptotically approximate equations: second family of shear resonances 4.4 Numerical results and discussion . . . . . . . . . . . . . . . . . . . . . . . 5 Layer with free faces: Long wave low frequency extensional motion 5.1 The dispersion in long wave low frequency regime . . . . . . . . . . . . . 5.1.1 Leading order approximation and incremental stability . . . . . . 5.1.2 Second order approximation . . . . . . . . . . . . . . . . . . . . . 5.1.3 Relative orders of displacements . . . . . . . . . . . . . . . . . . 5.2 Asymptotically approximate equations . . . . . . . . . . . . . . . . . . . 5.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Illustrative numerical results . . . . . . . . . . . . . . . . . . . . . . . . . 6 Layer with free faces: Long wave low frequency exural motion 6.1 Analysis of the dispersion relation . . . . . . . . . . . . . . . . . . . 6.1.1 Leading order approximation and incremental stability . . . . 6.1.2 Second order approximation . . . . . . . . . . . . . . . . . . . 6.1.3 Relative asymptotic orders of displacements . . . . . . . . . . 6.2 Asymptotically approximate equations . . . . . . . . . . . . . . . . . 6.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . 6.2.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . 6.2.4 Analogy with a membrane . . . . . . . . . . . . . . . . . . . . 6.3 Model problem: shock edge loading . . . . . . . . . . . . . . . . . . . 6.3.1 Shock edge loading . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Typical wave proles . . . . . . . . . . . . . . . . . . . . . . . 7 Dispersion of harmonic waves in a layer with xed faces 7.1 Harmonic wave motion in a pre-stressed incompressibleelastic layer 7.2 The dispersion relations . . . . . . . . . . . . . . . . . . . . . . . . 7.2.1 Symmetric motion . . . . . . . . . . . . . . . . . . . . . . . 7.2.2 Anti-symmetric motion . . . . . . . . . . . . . . . . . . . . 7.3 Propagation of long waves . . . . . . . . . . . . . . . . . . . . . . . 8 Layer with xed faces: Long wave symmetric motion 8.1 The dispersion relation in long wave regime . . . . . . . . . . . 8.1.1 Leading order approximation . . . . . . . . . . . . . . . 8.1.2 Second order approximation . . . . . . . . . . . . . . . . 8.1.3 Relative orders of displacements, physical interpretation 8.2 Asymptotically approximate equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS
iii
8.3
8.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 8.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 Illustrative numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 116 117 119 119 120 122 124 125 126 127
9 Layer with xed faces: Long wave anti-symmetric motion 9.1 Analysis of the dispersion relation . . . . . . . . . . . . . . . . . . . . . . . 9.1.1 Approximation in low wave number limit . . . . . . . . . . . . . . . 9.1.2 Relative asymptotic orders of displacements . . . . . . . . . . . . . . 9.2 Asymptotically approximate equations: rst family of cut-o frequencies . . 9.2.1 Leading order problem . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.2 Second order problem . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.3 Third order problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Asymptotically approximate equations: second family of cut-o frequencies 9.4 Illustrative numerical examples . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography
EHE
Acknowledgement
My presence in Salford became possible in rst place only due to active collaboration between my supervisor Dr. Graham A. Rogerson (The University of Salford) and my former supervisor Prof. Leonid Yu. Kossovich (Saratov State University, Russia). A U. K. Overseas Research Student Award and an award from the University of Salford provided the necessary funding, which is gratefully appriciated. I would like to thank my supervisor Dr. Graham A. Rogerson for introducing me to research in the eld of mechanics, supporting and encouraging me during my work and sharing numerous ideas. I would also like to thank his collaborators, especially Dr. Evgenia Nolde and Prof. Julius D. Kaplunov for many fruitful discussions. Last, but by no means least, I would like to thank my parents for their support and my friendsKevin for introducing me to the essentials of English culture, Aso for showing me the way friendship must be, Rinat for setting an excellent example of hard work and Danila for living six oors up.
Abstract
Many modern materials may support large deformation prior to failure. In particular, the rubber-like components used as vibration isolators in bridges and tall buildings, oer a modern method of earthquake protection. The conventional mathematical modelling of such structures is algebraically cumbersome and computationally inecient, especially for dynamic response. In static problems these complications were tackled by use of lower-dimensional structural theories, e.g. Kirchho plate theory or Kirchho-Love shell theory. These were based on ad hoc hypotheses and are not consistent with exact three-dimensional equations. Rigorous asymptotic analysis was later applied to rene these hypotheses and ensure consistency, but only recently were these ideas extended to dynamic problems. This is partially explained by the fact that in the case of static problems, only one type of asymptotic approximation is required, which is associated with long wave motion of the fundamental mode in the associated dynamic problem. In the dynamic case, high frequency motion will be an additional feature of the problem, in general consisting of both long and short wave contributions. The task of this thesis is therefore twofold. First, it aims to understand important features of wave propagation in pre-stressed incompressible elastic layers with various (xed and free faces) boundary conditions. The second goal is to establish asymptotically consistent two-dimensional models for long wave motion in such layers. Specically, Chapter 1 is concerned with the derivation of underlying theory for incremental motion in a pre-stressed incompressible elastic layer. The dispersion relation for a layer with free faces is presented in Chapter 2, with four associated asymptotic models developed in Chapters 36 (for long wave low and high frequency, symmetric and anti-symmetric motions). In Chapter 7 the dispersion in a layer with xed faces is considered, with associated long wave symmetric and anti-symmetric models obtained in Chapters 8 and 9.
vi
Introduction
The use of rubber, and rubber-like material, has become one of the popular trends in modern construction. The ability of rubber to withstand large strains and subsequently recover elastically has found use in many important industrial applications, including transport suspensions, bridge bearings and spring mountings, see Hirst (1969), Torr (1969) and Crawford (1985). Another rapidly developing application of rubber-like materials is so-called base isolation, which is a technique for protecting buildings and bridges from earthquakes with highly elastic bearings, see e.g. Tyler (1991), Sheridan et al. (1992) and Prendergast (1995). This technique is widely recognized as a cost-eective and ecient method for earthquake protection, with hundreds of base isolated buildings already built around the world, see Tyler (1991). However, the practical justication for use of such structures still heavily relies on the results of experiments on scaled models. This is largely related to the fact that the conventional mathematical modelling of such structures is algebraically cumbersome and computationally inecient, especially for elucidating the dynamic response of the geometrically complex structures used by engineers. The mechanical properties of rubbers and rubber-like materials are quite simple. The elastic response may usually be regarded as homogeneous and isotropic, although it is possible to produce rubbers with certain anisotropic features. The volume changes are very small for most deformations, see Ogden (1984, p. 488), hence rubber is often considered to be slightly compressible or incompressible. This is usually modelled by introducing an appropriate kinematic constraint, see an account of the general theory of internally constrained body presented in a paper by Scott (1975). The constraint of incompressibility is especially popular because it brings substantial simplications into the analysis, see e.g. Scott (1986). It is ironic that the most attractive feature of rubbers and rubber-like materials, which is their high exibility and softness, is at the same time the main source of diculties for a mathematical analysis. Deformations of up to 700% are not unusual in practice, see Treloar (1944), and the classical theory of elasticity can not be used in such situations, because the assumption of small deformations made in classical elasticity to linearise stress and strain measures is no longer valid. This poses two tasks for investigators. Firstly, it is necessary to model the essentially non-linear elastic response typical for rubbers, which is a well studied area nowadays, see Treloar (1975) for a detailed description of experimental and analytical studies of rubber-like materials and Ogden (1982) for a more up to date review. Secondly, in order to account for the highly non-linear elastic response of rubbers at large strains it is necessary to use a theory for nite deformations, see for example Green & Zerna (1954) and Green & Adkins (1960). Although this approach is generally possible and mathematically consistent, the dynamic analysis of even relatively simple objects like plates and shells may quickly become very algebraically complicated. Fortunately, there is often no practical need in dynamic modelling of nite deformations. The dynamic component of strain is often associated with innitesimal deformations, whereas the nite deformation is static and homogeneous. This is in particular the case for both the base isolators and bridge bearings mentioned earlier. The ideas within the previous paragraph were seemingly rst utilised by Biot (1965) to
The cited monograph combines and extends materials of earlier papers by the same author, see e.g. Biot (1934) and Biot (1939).
INTRODUCTION
introduce the concept of incremental deformation, which may be used to describe innitesimal incremental motion of nitely and statically deformed bodies. The main idea of the approach lies in separating the large static (usually assumed homogeneous) component of the deformation and subsequently analysing the innitesimal time-dependent deformations around this statically deformed state. This allows linearisation of the equations of motion while keeping the underlying elastic response non-linear, thus supporting the term non-linear elasticity often used in this context. Biot obtained incremental governing equations for both an initially isotropic and an orthotropic media, mostly in the context of plane strain. Similar ideas were independently developed in an important paper by Green et al. (1952) where the incremental equations were formulated in modern tensor notation for the rst time, however the scope of all of these works was limited to consideration of static problems. The study of dynamic properties of a nitely deformed media began with the analysis of surface waves rst discovered by Lord Rayleigh (1885). Specically, surface waves were examined both for a compressible and an incompressible half-space in the plane strain context for a general material by Hayes & Rivlin (1961), this result being later extended to a three dimensional incompressible half-space by restricting both the material and the deformation to neo-Hookean and bi-axial, respectively, see Flavin (1963). Some additional work was done by Willson (1973) and Willson (1974) to establish the surface wave speed equations for a variety of primary deformations and by Chadwick & Jarvis (1979) for a two-dimensional compressible half-space. In the excellent treatment of the theory of nite elasticity of an initially isotropic body by Ogden (1984), the problem of incremental stability was considered and subsequently, for both compressible and incompressible half-spaces, linked to the surface wave equation, see Dowaikh & Ogden (1990) and Dowaikh & Ogden (1991b). Similar results were also established in respect of interfacial (Stoneley) waves propagating between two nitely deformed incompressible elastic halfspaces, see Dowaikh & Ogden (1991a). More elaborate aspects of wave propagation including wave reection from a plane boundary in a nitely deformed elastic solid were examined by Ogden & Sotiropoulos (1997) and Ogden & Sotiropoulos (1998). However, since a (nitely deformed) half-space does not have a characteristic length all of these studies feature no dispersion. Dispersion is a property of certain types of media, in which the phase speed depends on the wave number. Two distinct types of the dispersion may be distinguished, these are physical and geometric. Whereas physical dispersion is typical for media with absorption, our interest is with geometrical dispersion, typical for structures with a fundamental length (for example, the thickness of an innite plate or a thin shell). By satisfying the equations of motion and boundary conditions simultaneously, it is possible to derive the (often implicit) relationship between phase speed and wave number, usually referred to as the dispersion relation or frequency equation. Typically, the dispersion relation is a complex transcendental equation and may feature an innite number of solution branches or modes. The branches which have a nite wave speed in the low wave number limit are termed as fundamental modes. Although the typical solution of impact problem will generally involve contributions from each mode of the dispersion relation, see Jones (1964) or Rogerson (1992), it was until very recently implicitly assumed in many models that only contributions of the fundamental modes are signicant. However recent studies by Kaplunov & Markushevich (1993) have shown that for the certain types of vibrations the contributions of higher modes may become signicant. Moreover, certain boundary value problems (e.g. an elastic layer with zero displacement boundary conditions) do not feature fundamental modes at all, ensuring that energy will propagate only with higher modes, see Kaplunov & Nolde (2001). One of the rst attempts to consider the propagation of waves associated with higher (non-fundamental) modes was made by Lamb (1917) for a plane section of an isotropic plate. The complexity of the associated dispersion relation, rst derived by Lord Rayleigh (1889), was tackled in the long and short wave regions by use of asymptotic analysis, however big advances in this area were only made with the advent of
INTRODUCTION
digital computers. Much more complete studies of higher modes of an isotropic plate were presented by Tolstoy & Usdin (1957) and Mindlin (1960). This approach was later extended to more complex cases, including analysis of a transversely isotropic elastic layer by Green (1982), of a multilayered bre-reinforced laminate by Baylis & Green (1986) and of a structure periodically layered with monoclinic layers by Nayfeh (1991). Because of the very complicated structure of the underlying dispersion relations, most of the results in these papers are obtained through use of numerical computations. Asymptotic analysis was shown to be an invaluable instrument in analytical investigation of the behaviour of various dispersion relations, see for example Kossovitch & Rogerson (1999) for a study of waves propagating in a plane section of an elastic layer composed of transversely isotropic material or Kossovitch et al. (2001) for the similar treatment of a three-dimensional transversely isotropic incompressible elastic plate. It was a natural extension to apply similar numerical and asymptotic techniques to the analysis of incremental motion in nitely deformed or pre-stressed media. We remark that hereafter we will often use the term pre-stress when referring to the initial nite deformation, to indicate the presence of stress prior to wave propagation. This term is sometimes also used to describe stress induced by a certain manufacturing process, for example rolling or fabrication, however we normally envisage stress introduced by high external loads. Seemingly, the rst study of a pre-stressed structure featuring dispersion was performed for a nite incompressible elastic plate in plane strain, see Ogden & Roxburgh (1993). This study was mainly concerned with aspects of surface wave propagation and incremental stability, but it was soon extended to higher modes in works by Rogerson & Fu (1995) and Rogerson (1997). The power of asymptotic analysis was demonstrated by Fu & Rogerson (1994) who studied non-linear post-buckling behaviour of a nitely deformed incompressible plate composed of a neo-Hookean material. The asymptotic methods for analysing dispersion relations were also generalised for the certain types of multi-layered media, see Rogerson & Sandiford (1996) and Rogerson & Sandiford (2002), slightly compressible, see Sandiford & Rogerson (2000), and three-dimensional plates, see Rogerson & Sandiford (1999). The algebraic complexity of such problems quickly becomes enormous, for example even the plane strain dispersion relation for two perfectly bonded incompressible elastic layers will be associated with the vanishing of the determinant of coecients associated with a system of 8 linear homogeneous equations. Only very limited analytical analysis is possible in such contexts and there is a need for more radical simplications than those provided by asymptotic approximating of the exact dispersion relations. In static problems algebraic complications were tackled by use of lower-dimensional structural theories, e.g. Kirchho and Mindlin theories of plates, Kirchho-Love shells theory and the rened Timoshenko-Reissner theories, see for example Lord Rayleigh (1877, ch. X), Love (1927), Timoshenko (1938), Mindlin (1955), Mindlin (1960), Reissner (1964). These theories were usually based on ad hoc hypotheses and although their predictions were correct to leading order, generally the results were not completely consistent with exact three-dimensional equations and the method of rening these theories to higher order was not available. Rigorous asymptotic analysis was later applied to rene these hypotheses and ensure consistency, see for example how the method of asymptotic integration of full three-dimensional equations of elasticity was applied to derive the asymptotically consistent theory of shells by Goldenveizer (1961) (this work is also briey reviewed by Goldenveiser (1966)). However, it was not until quite recently that these theories were extended to dynamic problems. This is partially explained by the fact that in the case of static problems only one type of asymptotic approximation is required, which is associated with long wave motion of the fundamental mode in the associated dynamic problem. In the dynamic case, high frequency motion will be an additional feature of the problem, in general consisting of both long and short wave contributions. Seemingly the rst attempt to utilise the method of asymptotic integration for analysing long wave high frequency phenomena was made by Achenbach (1969) in the context of stationary vibrations of an isotropic elastic layer. The extension of these techniques to the non-stationary phenomena was made much later,
INTRODUCTION
see Berdichevskii (1977), Kaplunov (1989) and Kaplunov (1990). The power of the method was demonstrated in an extensive monograph by Kaplunov et al. (1998), which covers many aspects of three-dimensional wave propagation in both plates and shells composed of an isotropic elastic material. These ideas are now being extended in various directions. Firstly, the work is being done to include the eects of anisotropy into simplied asymptotic models, see for example Kaplunov et al. (2000a) where long wave high frequency analysis was performed for a three-dimensional transversely isotropic elastic plate. Secondly, some work has been done on the asymptotic modelling of the eects of pre-stress on plane strain dynamic response of an incompressible initially isotropic layer with free faces, see Kaplunov et al. (2000b), Kaplunov et al. (2001a) and Kaplunov et al. (2001b), and with xed faces, see Nolde & Rogerson (2001). The task of this thesis is therefore twofold. First, it aims to extend our understanding of various aspects of wave propagation in pre-stressed incompressible elastic layers with various (xed and free faces) boundary conditions to the full three-dimensional case. The second goal is to establish simplied asymptotically consistent two-dimensional models for long wave motion in such layers. The layout of the thesis is as follows. The basic underlying theory of innitesimal timedependent motion superimposed upon a pre-stressed incompressible elastic body is summarised in Chapter 1. The elastic body under consideration is homogeneously deformed from its initial isotropic unstressed state to a nitely deformed equilibrium state. Small amplitude motion superimposed on this statically deformed state is then the object of our investigation. The linearised equations of motion for an incompressible body are rst obtained. The propagation condition for plane waves is analysed in order to try to ensure physically realistic response, which leads to the formulation of the strong ellipticity conditions. Although explicit strong ellipticity conditions for incompressible isotropic bodies have previously been obtained by Zee & Sternberg (1983), their notation seemed inappropriate for our purposes and new conditions were derived using a slightly modied procedure. Separation of the variables is then performed, giving the cubic secular equation as the necessary and sucient condition for such separation to be possible. Several conditions for factorising of this cubic are then obtained, which may serve as the basis for seriously simplied model problems or numerical tests. Since throughout this thesis we will often compare numerical and asymptotic solutions, a review of some of the existing material models is made and specic materials to be used are presented. Finally, a linearised measure of incremental surface traction is introduced for future formulation of the boundary-value problems associated with an elastic layer. In Chapter 2 we focus on a specic type of body, which is a pre-stressed incompressible elastic layer with free faces, i.e. with zero incremental traction on the surfaces. We assume that the nite deformation is homogeneous and one of the principal axes is orthogonal to the plane of the layer. A Cartesian system of coordinates is then introduced for the layer, with the coordinate axes along the principal axes of deformation. This enables us to employ the governing equations established in the Chapter 1. In particular, the solutions of the secular equation may be used to represent general wave solutions as superpositions of linearly independent functions. When these representations are substituted into the boundary conditions a system of linear homogeneous equations is obtained. The condition for the existence of non-trivial solution is then synonymous with vanishing of the determinant of associated matrix of coetients. In view of the symmetry of the plate about the mid-plane, this condition may be decoupled into the symmetric and the anti-symmetric counterparts, rst obtained in this context by Rogerson & Sandiford (1999) and usually referred to as the extensional and the exural dispersion relations. These relations are utilised to formulate the incremental stability criteria, which is analysed for certain states of the pre-stress. The dispersion relations are then analysed numerically to identify their principal properties as well as to distinguish between dierent asymptotic regimes of wave propagation. The derivation of asymptotic models for long wave high frequency extensional and exural motion is performed for a pre-stressed incompressible elastic plate in Chapter 3 and Chapter 4,
INTRODUCTION
respectively. In both chapters the analysis begins with an asymptotic investigation of the corresponding dispersion relation. Two asymptotic regimes are shown to exist in each case, associated with two families of cut-o frequencies. Third order asymptotic expansions, giving frequency as a function of wave number, are then obtained. We utilise these expansions to analyse the asymptotic structure of the waves propagating in the long wave high frequency regime and to show that the families of cut-o frequencies are in fact families of the shear resonance frequencies. Scalings of space and time variables are then introduced to represent the equations of motion, incompressibility condition and boundary conditions in such form that the contributions of various terms are made explicit. This allows a hierarchical system of boundary value problems at various asymptotic orders to be obtained in each case. Solving these systems, we establish the solutions for all of the displacement components and incremental pressure in terms of single function, referred to as the long wave amplitude. This function itself may be determined as the solution of a two-dimensional governing equation, obtained as an existence condition within the process of derivation. The high level of consistency of this approach is demonstrated by the fact that the dispersion relation of the asymptotic governing equation matches the corresponding approximation of the exact dispersion relation exactly. Some illustrative numerical examples are then presented. The material of these chapters served as the base for three papers, Pichugin & Rogerson (1999), Pichugin & Rogerson (2001a) and Pichugin & Rogerson (2001b). Long wave low frequency extensional wave propagation in a pre-stressed incompressible elastic layer, a problem analysed in Chapter 5, can not in general be reduced to a lower-dimensional theory with a single governing equation. Nevertheless, using our knowledge of asymptotic behaviour of the solutions, it is still possible to build a consistent simplied asymptotic theory, more appropriate for practical applications than the full three-dimensional equations. The analysis is begun by deriving an appropriate asymptotic representation of the dispersion relation, which is subsequently studied at various orders. We observe in the numerical solutions of the dispersion relation in Chapter 2 that extensional motion exhibits two nite limits of phase speed at in the low wave number limit. The associated squared wave speeds are shown to be solutions of a quadratic equation. Generally, these solutions are irrational and can not be represented without radicals, which makes this problem perhaps the most algebraically complicated in this thesis. The implication is that the second order coecients for the frequency expansions can only be obtained in terms of the leading order coecient. The second order frequency expansion coecients are then obtained. Asymptotically approximate equations are again derived to establish an asymptotic model for this type of motion, which in this case is shown to be a system of two two-dimensional governing equations, corresponding to the two associated so-called quasi-fronts. The analysis of exural (bending) motion of an innitely thin layer subjected to uniform tension (usually referred to as a membrane) is a very classical problem of dynamic theory of elasticity. One of the major results presented in Chapter 6 is a demonstration of an asymptotically consistent method for deriving the classical membrane equation, as well as generalisation of available theory to the case of nitely thin membranes. We begin our investigation with an asymptotic analysis of the exural dispersion relation in the long wave low frequency regime. An appropriate approximation for scaled wave speed is established and used to demonstrate the necessity of the incremental stability criteria presented in Chapter 2. Our attention then turns to asymptotically approximate equations. These are derived in the usual manner and leading order and rened second order governing equations are obtained as a result. It is demonstrated that the classical membrane equation is a particular case of our leading order governing equation and that our second order governing equation may be used to model a membrane of nite thickness. A model problem of shock edge loading of a semi-innite plate is then solved to demonstrate some new features of the derived asymptotic model. For instance, the existence of two possible types of quasi-fronts is shown in this problem, one of which having no analogue in isotropic case. A further novel feature of the model lies with the fact that type of quasi-front
INTRODUCTION
may change for various angles of wave propagation, thus greatly enriching the variety of material responses for line- and point-loading problems. Chapter 6 provided material for the recent paper by Pichugin & Rogerson (2001c). The nal three chapters of the thesis are concerned with the propagation of harmonic waves in a pre-stressed incompressible elastic layer with xed faces, i.e. with zero displacement boundary conditions. This type of boundary condition has particular practical signicance in the modelling of a coal layer, which is usually surrounded by a much stier rock, see Liang et al. (1993). We again assume that nite primary deformation is homogeneous, with one of the principal axes normal to the plane of the layer. An appropriate Cartesian system of coordinates is then introduced, the axes of which coincide with the principal axes of primary deformation. The results of the analysis of the equations of motion and secular equation performed in Chapter 1 and Chapter 2 are again applicable. Hence, in Chapter 7 we use the representations for the displacement components obtained in Chapter 2 to formulate the boundary conditions. As with the free faces problem, these comprise of a system of six linear homogeneous equations which by the symmetry of the layer may be decomposed into two systems of three linear homogeneous equations. The associated existence conditions are obtained in the form of symmetric and antisymmetric dispersion relations, which are rst analysed numerically. The numerical investigation reveals an interesting characteristic of the dispersion relations for a layer with xed faces, namely the lack of any fundamental mode. This means that nite wave speed limits do not exist in the low wave number limit for either symmetric or anti-symmetric motion. Long wave high frequency symmetric motion in a layer with xed faces, analysed in Chapter 8, deviates signicantly from other long wave high frequency models considered in this thesis. An appropriate approximation of the dispersion relation is rst obtained and it exhibits the absence of two distinct families of cut-o frequencies. The cut-o frequencies themselves are shown to be solutions of a transcendental equation and depend on the direction of wave propagation, which is highly unusual. An asymptotic analysis of this equation reveals both a connection with the shear resonance frequencies of a layer with free faces as well as providing estimates of cut-o frequencies for high harmonic numbers. A second order frequency correction is then obtained in terms of the associated cut-o frequency. Asymptotic solutions for all of the displacement components and pressure increment may then be derived in terms of a single function, termed the long wave pressure. This function is shown to be the solution of the two-dimensional governing equation, obtained as an existence condition in the process of the derivation of the asymptotic solution. The consistency of the approach is again established by comparing the dispersion relation of the asymptotically approximate governing equation with the appropriate approximation of the exact dispersion relation. Whereas the long wave symmetric motion analysed in Chapter 8 features many novel features, the long wave anti-symmetric motion in a layer with xed faces, studied in Chapter 9, demonstrate no signicant departures from the analogous long wave high frequency motion of a layer with free faces. Two families of cut-o frequencies are shown to exist. However for completeness, we establish third order approximations of the exact dispersion relation, which are then used to appropriately rescale time and space variables. The asymptotically approximate equations are derived using previously established procedures and the associated governing equations are obtained. Illustrative numerical examples are provided at the end of the chapter.
Notes on notation
Although a care has been taken to ensure that the notation throughout this thesis is standard and consistent, it seemed that some points deserve special explanation. Firstly, four dierent df kinds of equality symbol (=, , = and ) are used. While rst two of these are standard and mean equal to and identically equal to, respectively, the two others are less common. We use df = when the denition of a new quantity is implied. In many texts symbol is utilised for this purpose but it is felt that although denition implies identity, the meaning is not quite the same. The symbol denotes an asymptotic equality i.e. the equality of asymptotic orders. Therefore ab means the same as a = O(b) or b = O(a) .
The denitions of many quantities are valid throughout the thesis. However, to keep the notation concise and clear we have chosen to have some locally dened quantities, that is, quantities valid only within a single chapter. For example, parameters E (m) and F (m) (with various subscripts added) will denote coecients of asymptotic expansions specic for a chapter where they appear. Parameters e(m) and f (m) will serve the similar (chapter-specic) role for intermediate quantities. It is hoped that it will be clear from the context which of the quantities are local and which are global.
Chapter 1
Mathematical preliminaries
This rst chapter is devoted to the derivation of the basic underling theory of innitesimal time-dependent motion superimposed upon a pre-stressed elastic body. Although the term pre-stress is sometimes used to describe stresses induced by a manufacturing process (such as stamping or rolling), in the present study we envisage stresses arising from a (possibly large) static external load. Motivated by the industrial use of rubber-like materials in load-supporting structures, we focus our attention on a body composed of incompressible isotropic material. The elastic body under consideration is homogeneously deformed from its initial unstressed state to a nitely deformed equilibrium state. Small amplitude motion superimposed on this statically deformed state is then the object of our investigation. The linearised equations of motion for an incompressible body are rst obtained. The propagation condition for plane waves is analysed in order to try to ensure physically realistic response, which leads to the formulation of the strong ellipticity condition. The separation of the variables is then performed, bringing the secular equation as the necessary and sucient condition for such separation to be possible. Since throughout this thesis we will often compare numerical and asymptotic solutions, a review of some of the existing material models is made and specic materials to be used are presented. Finally, the linearised measure of incremental surface traction is introduced for future formulation of the boundary-value problems associated with an elastic layer.
1.1
Congurations of a pre-stressed body
We shall consider a homogeneous body B, composed of a non-heat-conducting isotropic elastic material which possesses an initial unstressed conguration B0 . A purely homogeneous static deformation is imposed upon B0 , to produce a nitely deformed equilibrium conguration Be , which we will use as our reference conguration of B. Finally, an innitesimal time-dependent motion is superimposed on Be , with the resulting conguration termed the current and denoted by Bt . The position vectors of a representative particle in congurations B0 , Be and Bt are denoted by X = (XA )T , x = (xm (XA ))T and x = (xi (XA ))T respectively, see Figure 1.1. Note that we use capital letters to specify indices of the components of vectors (or tensors) in the unstressed conguration B0 . The deformation gradients associated with the deformations B0 Be , Be Bt and B0 Bt are given in component form by xm FmA = , XA xi Fim = , xm FiA = xi = Fim FmA , XA (1.1)
where the tensor FmA is constant. Generally, we will use an over-bar and an over-tilde to denote quantities evaluated in the equilibrium and the current congurations, Be and Bt respectively. The summation convention over repeated suces is employed in (1.1)3 and will also be tacitly assumed throughout the Chapter 1 unless otherwise stated. Let u(x, t) denote the vector eld 8
CHAPTER 1. MATHEMATICAL PRELIMINARIES
F X x(X)
B0 F x(X, t) F
Be
Bt
Figure 1.1: Congurations of a pre-stressed body.
of innitesimal time-dependent deformation Be Bt , so that xi (XA ) = xi (XA ) + ui (xm (XA ), t) . (1.2)
Equation (1.1) together with (1.2) imply that the tensor of deformation gradient associated with the deformation Be Bt may be given in component form by FiA = (im + ui,m )FmA , (1.3)
in which a comma indicates dierentiation with respect to the implied spatial coordinate in Be and ij is the Kronecker delta.
1.2
Motion of a pre-stressed incompressible elastic body
We now restrict our attention to elastic bodies composed of incompressible materials, i.e. materials which are able to undergo only isochoric deformations. The requirement of incompressibility may be shown to imply J 1 = 0, J = det F , (1.4) throughout every possible material motion, see e.g. Chadwick (1999, p. 90). This kind of kinematic limitation is usually referred to as an internal constraint. It was shown by Truesdell & Noll (1965) and Ogden (1974b, p. 318) that an incompressible isotropic elastic solid is Green-elastic to second order. Within the range of validity of the theory that we are going to derive, this means that there exists an elastic potential energy function, which we shall usually refer to as a strain-energy function. In order to maintain the internal constraint (1.4), the material body B must generate reaction forces, which can do no work because of the constraint. This may be achieved by introducing the so-called pseudo strain-energy function of the form W (F , p ) = W0 (F ) p (J 1) . (1.5)
The rst term on the right hand side of (1.5) is associated with the constitutive (determined by deformation) part of the stress and the second, restricted to be zero for all motions satisfying (1.4), generates a workless reaction stress. The scalar p, usually referred to as a pressure,
10
plays role of a Lagrange multiplier and must ultimately be chosen to satisfy equations of motion and boundary conditions. We remark here that the pressure may be decomposed into a static pressure p in Be and a small time-dependent pressure increment pt = pt (t), thus p = p + pt . (1.6) We now proceed to the derivation of equations of innitesimal incremental motion. If we consider the law of conservation of linear momentum, see e.g. Spencer (1980, p. 134), these equations may be represented in the absence of body forces as Ai = i , u XA (1.7)
within which = (F , p ) is the nominal stress tensor, is the material density in the undistorted conguration B0 and a dot denotes dierentiation with respect to time. Equations (1.7) are evaluated in the natural conguration B0 and for our choice of Be as the reference conguration it is more convenient to use equations of motion in the form FmA Ai = i . u xm (1.8)
Expanding the nominal stress in a Taylor series around the nitely deformed state Be we obtain that to second order Ai Ai = Ai + (FkB FkB ) FkB +(p p ) Ai p +... .
Be
(1.9)
Be
The linear function on the right hand side of (1.9) will oer adequate representation of the nominal stress only provided the higher order terms of the expansion are negligibly small. In view of the relations (1.3) and (1.6) this suggests that a formal denition of the concept of small amplitude motion around Be may be given by FkB FkB = uk,l FlB 1 , |p p | = |pt | 1 , (1.10) see also Ogden (1984, p. 328) and Chadwick et al. (1985, p. 344). Motions that satisfy (1.10) are our exclusive concern from now on. The constitutive equations for a nitely deformed elastic body may be written in terms of the nominal stress and pseudo strain-energy function as Ai = W W0 1 = pJFAi , FiA FiA (1.11)
in which we utilised (1.5). Equations (1.11) are then employed to calculate the derivatives of the components of nominal stress in the expansions (1.9), yielding Ai 1 = JFAi , p Ai 2 W0 1 1 1 1 = pJ FBk FAi FBi FAk . FkB FkB FiA (1.12)
Now the expansions (1.9) may be substituted into the equations of motion (1.8), which together with (1.12) gives, in component form, the linearised equations of motion as Bmilk uk,lm pt,i = i , u where B is the fourth order elasticity tensor, for which the components are dened by Bmilk = FmA FlB 2 W0 FkB FiA .
Be
(1.13)
Essentially, the assumption of incompressibility renders the density constant and we shall therefore not distinguish between densities in various material congurations.
11
The equations of motion (1.13) must be solved in conjunction with the appropriately linearised incompressibility condition, which may be shown to have form ui,i = 0 , (1.14)
see e.g. Spencer (1980, p. 93). It is interesting to remark that the elasticity tensor B, dened immediately after (1.13), does not involve any terms associated with the incompressibility constraint, which is seemingly peculiar to this particular constraint. For an isotropic body B, the components of B allow especially simple representation relative to axes coincident with the principal axes of the primary deformation. The only non-zero components of B have the form Biijj , Bijij or Bijji , i, j {1, 2, 3}, and these may be given in terms of the principal stretches m , m {1, 2, 3}, of the static deformation as 2 W0 Biijj = i j , i j 2 W0 W0 i 2 2 i j i j i j Bijij = W0 1 Biiii Biijj + i 2 i W0 i =j, Bijji = Bijij i i (1.15) i=j, i=j, i = j , (1.16) i = j , (1.17)
within which no summation over repeated suces is assumed, see Ogden (1984, p. 412). We conclude this section by providing an explicit form of the linearised equations of motion (1.13) for an incompressible isotropic body. These are given by B1111 u1,11 + (B1122 + B2112 )u2,12 +(B1133 + B3113 )u3,13 + B2121 u1,22 + B3131 u1,33 pt,1 = 1 , u (B2211 + B1221 )u1,12 + B2222 u2,22 + (B2233 + B3223 )u3,23 + B1212 u2,11 + B3232 u2,33 pt,2 = 2 , u (B3311 + B1331 )u1,13 + (B3322 + B2332 )u2,23 + B3333 u3,33 + B1313 u3,11 + B2323 u3,22 pt,3 = 3 , u (1.20) (1.19) (1.18)
which must be solved in conjunction with the appropriate linearised incompressibility condition (1.14), given explicitly by u1,1 + u2,2 + u3,3 = 0 . (1.21)
1.3
Propagation of plane waves, strong ellipticity
Before continuing the derivation of the necessary basic equations, we need to study some aspects of the propagation of small amplitude plane waves in an incompressible pre-stressed elastic body. Specically, we seek solutions of the governing equations (1.13) and (1.14) in the form u = Ee(x n vt) , pt = Es (x n vt) , (1.22)
where dash denotes dierentiation with respect to the argument and s is to be determined. The (real) scalar v is the speed of propagation of a disturbance with prole specied by a function (), which we assume to have a non-zero second derivative. When v is real and positive,
12
expressions (1.22) characterise a homogeneous plane wave propagating in the direction dened by the unit vector n and polarised along the unit vector e, with the scalar E playing role of the wave amplitude. What will follow next is the specialisation to the case of incompressibility of the general analysis of plane wave propagation in constrained elastic bodies due to Scott (1975) and Chadwick et al. (1985). When solutions of the form (1.22) are substituted into the equations of motion (1.13), and incompressibility condition (1.14), we obtain Qik ek sni = v 2 ei , ei ni = 0 , (1.23)
df df where the tensor Qik = Bmilk nl nm is usually referred to as the acoustic tensor and v = v is convenient shorthand, which we term the scaled wave speed. The condition (1.23)2 allows us to make a preliminary observation, namely that for plane waves propagating in an incompressible elastic body B, the polarisation is always orthogonal to the direction of wave propagation. This prohibits the propagation of longitudinal waves in B, as well as implying that the only type of plane waves that can propagate in B are shear waves. We may now use (1.23) to establish an expression for the parameter s in the form
s = n (Qe) .
(1.24)
Inserting (1.24) into the equation of motion (1.23) we arrive at the following propagation condition (2 1 P Q) e = 0 , v (1.25) within which P = 1 n n is a symmetric tensor. Equation (1.25) determines the possible wave speeds and polarisations for any given direction and is essentially an eigenvalue problem for P Q. Since the system of linear homogeneous equations P m = 0 (m is arbitrary) possesses non-trivial solution m = n, then det P = 0 and det(P Q) = 0, so one of the eigenvalues of P Q must be zero, which may in fact be shown to be a spurious solution, see Scott (1975, p. 68). Hence the characteristic equation for the eigenvalue problem (1.25) may be represented in the following form 24 2 tr(P Q)2 + (tr P Q)2 tr(P Q)2 = 0 , v v (1.26) where tr A denotes the trace of a matrix A, this result is due to Ericksen (1953). Although the acoustic tensor Q is symmetric for Green-elastic materials, see Ogden (1984, p. 474), the tensor P Q is not in general symmetric. However it is possible to show, that when the acoustic tensor Q is symmetric, all eigenvalues of the tensor P Q are real and the two eigenvectors associated with the non-zero eigenvalues are mutually orthogonal (which entails orthogonality of shear wave polarisations). In fact, Scott & Hayes (1985) provided two dierent proofs of this statement as well as generalising this result to the case of an arbitrary singly constrained material. Now we turn our attention back to the constitutive law (1.11). The way it was introduced imposed only certain basic restrictions on the variety of possible responses (e.g. strain energy must vanish in the absence of a deformation), thus allowing some potentially non-physical responses. As an example of this we remark that generally it is possible to choose such strainenergy function and direction of propagation, that the square of the corresponding wave speed, found from (1.25) or (1.26), will be negative. Thus we need to nd requirements which will ensure physically realistic responses and we remark that the broad review of possible approaches is given in Wang & Truesdell (1973, chap. III, sect. 79). In this thesis we will require that solutions of the characteristic equation (1.26), which are squared speeds of waves, must be positive for all directions of propagation. This requirement is usually referred to as the strong ellipticity condition and in our case may be shown to be equivalent to tr(P Q) > 0 , (tr P Q)2 tr(P Q)2 > 0 . (1.27)
df
13
Note that this condition is less restrictive than the requirement of positive-deniteness of the acoustic tensor Q, which usually serves as strong ellipticity condition for unconstrained materials. We also remark that replacing > by in inequalities (1.27) we arrive at the Hadamard condition, which is a necessary condition for stability in any equilibrium conguration of an elastic body subject to mixed boundary conditions, see Wang & Truesdell (1973, p. 220). While the strong ellipticity condition may perhaps be too strong to allow every physically realistic response, see Ogden (1984, p. 394), the Hadamard stability is the least we should require.
1.4
Explicit necessary and sucient conditions for the strong ellipticity
The necessary and sucient conditions (1.27) must hold for all directions n implicitly contained in the acoustic tensor Q. This makes conditions (1.27) unsuitable for practical testing of the strong ellipticity of particular materials. Zee & Sternberg (1983) rst derived explicit conditions for strong ellipticity of an incompressible elastic body in terms of the derivatives of the strain energy function and the principal stretches of primary deformation. Although compact, these conditions are not very convenient for analysing the parameters naturally occuring in the problems we consider. Hence, we reproduce here their result using more conventional notation. When the coordinate axes of an isotropic body B are coincident with the principal axes of the primary deformation, the only non-zero components of the elasticity tensor Biijj , Bijij and Bijji , i, j {1, 2, 3}, possess representations (1.15)(1.17). Keeping in mind the denitions of the acoustic tensor Qik = Bmilk nl nm and the constraint tensor Pik = ik ni nk , we may formulate the conditions (1.27) explicitly as (12 + 13 )m2 + (212 + 13 + 23 )m1 m2 + (21 + 23 )m2 1 2 + (213 + 12 + 32 )m1 m3 + (223 + 21 + 31 )m2 m3 + (31 + 32 )m2 > 0 , 3 12 13 m3 + (221 23 + 23 31 )m2 m3 + (231 23 + 21 32 )m2 m2 1 2 3 + 21 23 m3 + (212 13 + 13 32 )m2 m3 + (232 13 + 12 31 )m1 m2 2 1 3 + 31 32 m3 + (213 12 + 12 23 )m2 m2 + (223 12 + 13 21 )m1 m2 3 1 2 + (412 23 2 + 12 21 + 13 31 + 23 32 )m1 m2 m3 > 0 , 13
df
(1.28)
(1.29)
within which mi = n2 , i {1, 2, 3}, are always non-negative and the material parameters ij , i ij and ij , i, j {1, 2, 3}, are dened through the components of the elasticity tensor B as follows ij = Bijij , ij = ij ik jk , bij = Biiii Biijj Bijji , i<j, 2ij = bij + bji , i=j, k {i, j} , / i, j, k {1, 2, 3} , (1.30)
with no summation over repeated suces. Thus we may summarise our task as: obtaining the restrictions on the material parameters, which are equivalent to the satisfying of the inequalities (1.29) and (1.28) for every vector m {m | mi 0, m1 + m2 + m3 = 1}.
1.4.1
Necessary and sucient conditions for particular directions
First we consider vectors m coinciding with the coordinate unit vectors. Thus substituting m = (1, 0, 0), m = (0, 1, 0) and m = (0, 0, 1) into the conditions (1.28) and (1.29), we arrive at the following inequalities ij + ik > 0 , ij ik > 0 , i = j = k = i, i, j, k {1, 2, 3} , (1.31)
The method of the derivation that we are going to use is very similar to the method employed by Zee & Sternberg (1983) except for some minor deviations.
14
where the summation convention is not assumed. The necessary and sucient conditions for (1.31) to hold are ij > 0 , i=j, i, j {1, 2, 3} , (1.32) which seemingly constitutes six inequalities. We remark however that after referring to the denitions (1.16) of the elasticity tensor components, it is possible to show that ij j i df = ji = k , i j i = j = k = i, i, j, k {1, 2, 3} , (1.33)
where the parameter k is introduced for subsequent utilisation and no summation is assumed over repeated suces. The equality (1.33) is a generalisation of the equality 12 2 = 21 2 (often 2 1 written as 2 = 2 ), used in the analysis of the plane strain problems, see e.g. Dowaikh & Ogden 2 1 (1990). When all of the stretches of primary deformation are distinct, (1.33) may easily be checked by direct substitution. If two of the stretches are equal, say 1 = 2 = , we obtain that (1.33) is equivalent to the following condition W1 W2 + (W11 W22 + W12 W21 ) = 0 , within which Wi = W0 , i Wij = 2 W0 , i j i, j {1, 2, 3} . (1.34)
(1.35)
It is obvious that (1.34) is satised for any isotropic body, hence we can reduce (1.32) to the three inequalities ij > 0 , i<j, i, j {1, 2, 3} , (1.36) which can alternatively be formulated as i > 0, i {1, 2, 3}, see (1.33). Our next step is to analyse the conditions (1.28) and (1.29) for vectors m lying in one of the coordinate planes, i.e. when one of the components of m is equal to zero. To be specic, we choose m3 = 0 and therefore consider vectors m {(m1 , m2 , 0) | mi 0, m1 + m2 = 1}. For such m the inequality (1.29) may be rewritten as 12 13 m3 + (213 12 + 12 23 )m2 m2 + (223 12 + 13 21 )m1 m2 + 21 23 m3 > 0 . 1 1 2 2 To analyse (1.37) we employ the following lemma of Zee & Sternberg (1983, p. 67) Lemma. Let a, b, c and d be the real constants, with a > 0 and d > 0. Then ax3 + bx2 y + cxy 2 + d > 0 if and only if either 4c3 a + 4b3 d b2 c2 18abcd + 27a2 d2 > 0 or b > 0, c > 0. (1.39) x 0, y 0 : x + y = 1, (1.38) (1.37)
Because of the already established necessary conditions (1.36) this lemma is obviously applicable to (1.37). In terms of the coecients of (1.37) the left hand side of the inequality (1.39)1 may be factorised as 2 2 2 2 4(12 12 21 ) 12 23 212 23 13 + 21 13 > 0 , (1.40) which is equivalent to the following condition 12 21 < 12 < 12 21 . (1.41)
On the other hand, for (1.37) the inequalities (1.39)2 take form 12 23 < 12 , 2 13 21 13 < 12 . 2 23 (1.42)
15
The lemma thus implies that the condition (1.37) is satised provided either (1.41) or (1.42) is satised. It is seemingly more natural to represent (1.41) and (1.42) in terms of the parameters k , k {1, 2, 3}, introduced in (1.33), yielding that either 3 < 12 < 3 or 3 1 < 12 , 2 2 3 2 < 12 . 2 1 (1.43)
Since min{1 /2 , 2 /1 }
1, and in view of (1.36), we conclude that (1.43) is equivalent to 3 < 12 . (1.44)
This procedure may be repeated for m = (m1 , 0, m3 ) and m = (0, m2 , m3 ), which after referring back to the parameters ij enables us to establish the following three conditions ij ji < ij , i<j, i, j {1, 2, 3} . (1.45)
We derived (1.36) as the necessary and sucient conditions for (1.28) and (1.29) to hold, provided m coincides with one of the coordinate unit vectors. Conditions (1.45) were only shown to be necessary and sucient for (1.29) when at least one of the components of m is equal to zero. However it is possible to prove that (1.36) together with (1.45) are also sucient conditions for (1.28) to hold for any m. Indeed, we may use inequalities (1.45) together with the appropriately rearranged condition (1.28) to obtain (12 + 13 )m2 + (212 + 13 + 23 )m1 m2 + (21 + 23 )m2 1 2 + (213 + 12 + 32 )m1 m3 + (223 + 21 + 31 )m2 m3 + (31 + 32 )m2 > 3 2 2 12 m1 2 12 21 m1 m2 + 21 m2 + (13 + 23 )m1 m2 + 13 m2 2 13 31 m1 m3 + 31 m2 + (12 + 32 )m1 m3 1 3 2 + 23 m2 2 23 32 m2 m3 + 32 m2 + (21 + 31 )m2 m3 > 0 . 3
(1.46)
The braces in (1.46) contain complete squares and therefore (1.36) necessitate that the expression 0 and at the right-hand side of (1.46) is strictly positive for any m (we reiterate that mi m1 + m2 + m3 = 1). Thus, the conclusion can be made that the conditions (1.36) and (1.45) are necessary and sucient to satisfy (1.28) and (1.29) when at least one of the components of m is equal to zero. Therefore, the conditions (1.45) are not only necessary for (1.29) but also sucient to satisfy (1.28) for any m. Thus the condition (1.28) need not be considered anymore. We remark nally, that one of mi , i {1, 2, 3}, must be zero in the plane strain case and therefore (1.36) and (1.45) are the necessary and sucient for the strong ellipticity of the body subject to plane strain, which matches the result obtained by Ogden (1984, p. 432).
1.4.2
General conditions: necessity
Now having studied the cases when one or two components of m are equal to zero the general case may be considered. First we note that for an arbitrary vector m {m | mi > 0, m1 +m2 +m3 = 1} the condition (1.29) may be represented in the following form 1 2 2 2 m1 m2 m3 (1 + 2 + 3 21 2 21 3 22 3 ) > 0 , 4 within which k = 2ij + ki mj mk mk mi + kj + ij + ji , mi mj mj mi (1.47)
(1.48)
and no summation over repeated suces is assumed. Since all components of m are positive, the inequality (1.47) is equivalent to (3 1 2 )2 < 41 2 . (1.49)
CHAPTER 1. MATHEMATICAL PRELIMINARIES Now it is convenient to introduce new variables such that l1 = 1 m1 , 3 m3 l2 = 2 m2 . 3 m3
16
(1.50)
Keeping in mind that m1 + m2 + m3 = 1 we derive the inverse formulae for (1.50), yielding mi = i li l1 l2 1 + + 1 2 3 , m3 = 3 1 l1 l2 1 + + 1 2 3 , i {1, 2} . (1.51)
The convenience of the one-to-one transformation (1.50) (or (1.51)) lies in the fact that instead of analysing a constrained problem in m, we study an unconstrained problem in the quadrant li > 0, i {1, 2}. We may now use identities (1.33) to recast denitions (1.48) in terms of new variables as (l2 1)2 l1 1 + l1 2 + 3 , l2 l2 2 (l1 1) l2 2 2 = 22 + l2 1 + 2 + 3 , l1 l1 1 1 (l1 l2 )2 2 3 = 23 + 1 + 2 + 3 , l2 l1 l1 l2
2 1 = 21 + df
(1.52) (1.53) (1.54)
2 where k = ij + k , k = i = j = k, i, j, k {1, 2, 3}. It is evident from (1.52)(1.54) that all of i , i {1, 2, 3}, are strictly positive. Therefore the conclusion can be made that the condition (1.49) necessitates 3 < 1 + 2 + 2 1 2 . (1.55)
Since our current task is the derivation of necessary conditions, we will not solve the optimisation problem for (1.55). Instead, we will only nd implications of (1.55) for a certain pair (l1 , l2 ). Specically, we note that for any xed (l1 , l2 ) there exists a number such that 1 2 = 2 . Then we may introduce a weight parameter W , so that 1 = /W and 2 = W . For this W it is true that 1 W 2 2 = 0, which in terms of our variables implies that
2 2 2(1 W 2 2 ) + 2 2 (l2 1)2 W 2 l2 l2 W 2 (l1 1)2 l2 W 2 l2 1 + 1 2 + 1 3 = 0 . l2 l1 l1 l2
(1.56)
It is easy to see that this condition is in particular satised by l1 = 1 , W +1 l2 = W , W +1 W = 2 , 1 (1.57)
which after substituting into the condition (1.55) yields

2 3 23 < (1 + 2 )2 .
(1.58)
Repeating the procedure for another permutations of indices in (1.55), and returning back to the conventional material parameters dened in (1.30), we obtain three necessary conditions for strong ellipticity, given by ij
ij ji <
ik +
ik ki +
jk +
jk kj
i = j = k = i.
(1.59)
Zee & Sternberg (1983) used a conventional procedure for nding the critical point of (1.55). However, the algebraical complexity of this procedure prevented them from actually proving that their solution delivers a local minimum to (1.55). Thus their procedure only served to nd the critical point in a constructive way (as opposed to just specing this point). We use another (simpler) constructive procedure to nd a solution, which eventually is shown to be the local minima.
17
1.4.3
General conditions: suciency
We have already mentioned that conditions (1.36) and (1.45) are sucient for strong ellipticity provided one of components mi , i {1, 2, 3}, is equal to zero. We also mentioned that these conditions are sucient to satisfy (1.28) for any m, therefore we only need to prove suciency of (1.36), (1.45) and (1.59) for satisfying the inequality (1.29) for an arbitrary m. We now suppose that (1.36), (1.45) and (1.59) hold. The denitions (1.52)(1.54) may be utilised to prove the following identity 1 2 = {21 2 + (1 l2 )1 + (1 l1 )2 + 3 }2 +2
(l2 1 + (l2 1)2 )2 ((l1 1)1 + l1 2 )2 (l2 1 l1 2 )2 1 + 2 2 + 2 3 l2 l1 l1 l2 1 2 1 3 2 3 + (l1 + l2 1)2 + + , l1 l2 l1 l2
(1.60)
which was rst obtained in dierent notation by Zee & Sternberg (1983, p. 76). Conditions (1.36) imply that i > 0, i {1, 2, 3}, and since l1 > 0 and l2 > 0 this shows that all terms in (1.60) are non-negative. Therefore (1.60) necessitates that 1 2 {21 2 + (1 l2 )1 + (1 l1 )2 + 3 }2 , (1.61)
which upon substitution into (1.55) gives one of the conditions (1.59). Since we assume that all conditions (1.59) hold, the conclusion can be made that our inequalities are sucient to satisfy (1.55). Repeating this procedure for another permutations of indices in (1.60) we obtain three conditions k < i + j + 2 i j = ( i + j )2 , k = i = j = k. (1.62)
Two of these inequalities may be used to demonstrate that ( 1 2 )2 < 3 , (1.63)
which together with the already proven result (1.55) is equivalent to the condition (1.49), and in turn is equivalent to (1.29). Therefore we may conclude that the conditions (1.36), (1.45) and (1.59) are necessary and sucient for the strong ellipticity of an incompressible isotropic elastic body B. It must be remarked, that the strong ellipticity condition only ensures some plausible properties of the material response. It is in a sense a local condition that takes no account of the specic shape of the body B. Therefore, after specialising the boundary conditions we will need to nd restrictions, additional to the strong ellipticity, which will ensure physically realistic response of any specic body B.
1.5
Separation of variables: secular equation
The specic type of the plane waves we will study in this thesis are harmonic waves. In view of the homogeneity of the underlying static deformation F we introduce in Be a global Cartesian system of coordinates Ox1 x2 x3 with axes Ox1 , Ox2 and Ox3 along the principal axes of this deformation. Now we perform separation of variables and seek a solution of the equations of motion (1.18)(1.20) in form of a travelling harmonic wave (u1 , u2 , u3 , pt ) = (U1 , U2 , U3 , kP )ekqx2 eik(x1 c +x3 s vt) ,
df
(1.64)
in which k is the wave number, v the phase speed, the vector (c , 0, s ) = (cos , 0, sin ) is the projection of wave normal n onto the coordinate plane Ox1 x3 and the parameter q is to be
18
determined from the governing equations. When either c = 0 or s = 0 our problem reduces to the plane strain case, in which the system of equations of motion decouples and generally requires dierent treatment to the three-dimensional case, see for example Dowaikh & Ogden (1990), Dowaikh & Ogden (1991a) or Ogden & Roxburgh (1993). However the solution of the plane strain problem can always be obtained as the corresponding limit of the full three-dimensional solution and we therefore consider only the three-dimensional case and tacitly assume that c = 0 and s = 0. Substituting the solutions (1.64) into the equations of motion (1.18)(1.20) and the incompressibility condition (1.21) we obtain a system of 4 linear homogeneous equations. This system possesses a non-trivial solution provided the determinant of its coecients is equal to zero, which yields a necessary and sucient condition, usually referred to as the secular equation 21 23 q 6 + (21 + 23 )2 c1 q 4 + (4 c2 v 2 + c3 )q 2 (2 c4 )(2 c5 ) = 0 , v v v v in which we re-introduce v =
df
(1.65)
v as the scaled wave speed and
c1 = (223 21 + 23 31 )s2 + (212 23 + 21 13 )c2 , c2 = (223 + 21 + 31 )s2 + (212 + 23 + 13 )c2 , c3 = (412 23 + 21 12 + 23 32 + 13 31 2 )s2 c2 13 c4 = 32 s2 + 12 c2 , + (223 31 + 21 32 )s4 + (212 13 + 23 12 )c4 , c5 = 31 s4 + 213 s2 c2 + 13 c4 ,
where material parameters ij , ij and ij , i, j {1, 2, 3} are dened through (1.30). We remark 2 2 2 here, that the three roots of the cubic (in q 2 ) secular equation (1.65), denoted by q1 , q2 and q3 , must always satisfy the following equalities (21 + 23 )2 c1 v , 21 23 v 4 c2 v 2 + c3 2 2 2 2 2 2 q1 q2 + q1 q3 + q2 q3 , 21 23 (2 c4 )(2 c5 ) v v 2 2 2 q1 q2 q3 , 21 23
2 2 2 q1 + q2 + q3
(1.66) (1.67) (1.68)
and may be used to represent equation (1.65) as the following product

2 2 2 21 23 (q 2 q1 )(q 2 q2 )(q 2 q3 ) = 0 .
(1.69)
The secular equation (1.65) completely determines the types of waves propagating in the body, that is why it is essential to have complete information about its roots and their dependence on the wave speed. This analysis is overly complicated for the cubic equation (1.65), however we will show now that under some simpling assumptions it may be factorised into the product of linear and quadratic (in q 2 ) equations, which may prove indispensable in analysis of complicated geometric structures, such as multi-layered media, or as a possible test case for numerical computations. First we note that in view of the relations (1.66)(1.68), one of roots of the seqular equation (1.65) is constant whereas two other roots are quadratic functions in v . Thus we may try to seek a solution of the form q2 = C , (1.70)
with the constant C and the conditions under which this solution exists to be determined. Comparing the coecients of secular equations (1.65) and (1.69) at various powers of q and v
Note that in the case of plane strain the secular equation is quadratic in q 2 .
19
we obtain that the only possible solution is associated with C = 1, thus yielding the factorised secular equation (q 2 1) 21 23 q 4 + ((21 + 23 )2 c6 )q 2 + (2 c4 )(2 c7 ) = 0 , v v v c6 = (21 13 + 23 12 )c2 + (21 32 + 23 31 )s2 , (1.71)
c7 = 13 c2 + 31 s2 ,
and since this factorisation must hold for all speeds and angles of propagation we derive additional restrictions on the material and pre-stress parameters, which may be given by ij + ji = 2ij , i<j, (1.72)
12 21 13 + 31 + 23 32 = 0 .
These conditions are symmetric and may therefore be considered as restrictions on the strainenergy function and, consequently, on the class of elastic materials. The rst of (1.72) is a threedimensional generalisation of the well-known plane strain simplifying condition 12 + 21 = 212 (often written as + = 2), see for example Rogerson & Fu (1995, p. 65). Let us take a closer look at the second condition of (1.72). When all of the principal stretches of primary deformation 1 , 2 and 3 are distinct, substituting the relations (1.16) into (1.72)2 immediately conrms it is identically zero. Suppose now that two of the principal stretches are equal, let us for example take 1 = 2 . If we recall the denitions (1.35) then (1.16) reveal that the condition (1.72)2 implies W1 = W2 , W21 W12 + W11 W22 = 0 , (1.73) which is always satised for any isotropic material. The same is true for any other pair of equal principal stretches and we note, that the case of uniaxial deformation corresponds to no primary deformation at all for an incompressible body B. Therefore, the second condition of (1.72) holds for any incompressible isotropic material and can therefore be safely ignored. Another possible form of the solution of (1.65) is given by q 2 = C2 v 2 + C0 , (1.74)
where constants C2 and C0 are to be determined. Comparing the secular equations (1.65) and (1.69), we deduce that there are two possible factorisations of (1.65) with solution of the form (1.74). In the rst case C2 = 1/23 and C0 = c4 /23 , yielding the following secular equation (23 q 2 + v 2 c4 ) 21 q 4 + (2 212 c2 (21 + 31 )s2 )q 2 + c5 v 2 = 0 , v which is valid provided 23 + 32 = 223 , 23 32 = 2(12 13 ) , 12 = 13 . (1.76) (1.75)
Alternatively we may choose C2 = 1/21 and C0 = c4 /21 , to obtain the factorisation of (1.65) given by (21 q 2 + v 2 c4 ) 23 q 4 + (2 (13 + 23 )c2 223 s2 )q 2 + c5 v 2 = 0 , v which holds when the following conditions are satised 12 + 21 = 212 , 21 12 = 2(23 13 ) , 31 = 32 . (1.78) (1.77)
Conditions (1.76) and (1.78) are not symmetric, hence in general they can not be satised by the appropriate choice of strain energy function. Therefore we consider these conditions as specic restrictions on the variety of possible deformations. It is possible to show that the only deformation which can satisfy all of the conditions (1.76) or (1.78) is the bi-axial deformation
20
with the axis of symmetry along Ox1 (2 = 3 ) or Ox3 (1 = 2 ), respectively. Indeed, if we assume that 2 = 3 , denitions (1.15)(1.17) may be used to show that conditions (1.76) imply W2 = W3 , W12 = W13 , W23 W32 + W22 W33 = 0 , (1.79)
and these obviously hold for incompressible isotropic materials. The same apporoach may be used in analysing the set of conditions (1.78), which under assumption 1 = 2 , delivers upon use of relations (1.15)(1.17) and (1.35) that W1 = W2 , W13 = W23 , W12 W21 + W11 W22 = 0 . (1.80)
The equalities (1.80) hold for any isotropic material. Thus the conclusion can be made, that in case of a bi-axial static deformation with axis of symmetry along Ox1 or Ox3 , the secular equation (1.65) may be factorised to obtain (1.75) or (1.77), respectively. We remark, that decoupled solutions q 2 = (c4 v 2 )/23 or q 2 = (c4 v 2 )/21 correspond to one of two shear waves propagating in B. This situation is reminiscent to the case of transversely isotropic body with no pre-stress, see Kaplunov et al. (2000a). Thus a bi-axially deformed body B demonstrates some of the properties specic for transversely isotropic body. To conclude this section we note that the conditions (1.72)1 , (1.76) and (1.78) provide two principally dierent choices of simplifying conditions for the three-dimensional problem. Conditions (1.72)1 restrict the type of material but allow arbitrary (homogeneous) deformations. Conditions (1.76) and (1.78) on the other hand restrict the variety of deformations while allowing use of a general material. In certain situations (the analytical test case for the numerical routine, for instance) the advantages of the analytical simplicity may outweight those of generality of approach and in this case we may even consider restrictions on both material and deformation type. Specically, when both (1.72)1 and (1.76)3 hold, the secular equation (1.71) factorises further yielding (q 2 1)(21 q 2 c7 + v 2 )(23 q 2 c4 + v 2 ) = 0 , (1.81)
with the functions of material parameters and angle of propagation c4 and c7 dened immediately after (1.65) and (1.71), respectively. On the other hand, we may assume that (1.72)1 is satised together with (1.78)3 , thus obtaining (q 2 1)(21 q 2 c4 + v 2 )(23 q 2 c7 + v 2 ) = 0 . (1.82)
Therefore, both choosing a material that satises (1.72)1 and assuming a bi-axial deformation, with symmetry axis along Ox1 or Ox3 , we can obtain exact solutions of the secular equation (1.81) or (1.82), respectively.
1.6
Some specic material models for rubber-like materials
The elastic properties of a body B may be characterised in terms of the strain energy function W0 (F ) through the constituitive equations (1.11). All of the derivations in this thesis will be performed for the most general form of W0 , however we will need to specialise it when comparing numerical and asymptotic solutions. In this section we briey review some of the well-known forms of strain-energy functions employed to model rubber-like materials as well as choosing some particular materials for future use. It needs to be remarked that the strain energy function for an isotropic material may in general be treated as a function of any three independent invariants of the deformation. It is common in the literature to use the principal invariants of the right Cauchy-Green tensor C = F F T , given in terms of the principal stretches of deformation by I = 2 + 2 + 2 , 1 2 3 II = 2 2 + 2 2 + 2 2 , 1 2 1 3 2 3 J = 2 2 2 , 1 2 3 (1.83)
see e.g. Ogden (1984, p. 218) or Spencer (1980, p. 130). But J = 1 for incompressible materials and therefore we may regard the strain energy of an incompressible isotropic body B as a function of two invariants of the deformation W0 = W0 (I, II).
21
1.6.1
Neo-Hookean materials
One of the simplest forms for the strain energy function of a rubber-like material is the neoHookean strain energy function. It was derived within the framework of Gaussian molecular statistical theory of rubber elasticity and may be represented as W = (I 3) , 2 (1.84)
within which I is the invariant dened by (1.83)1 and is the shear modulus in the ground state, see Treloar (1943). The use of this function greatly symplies the analysis, which for example was successfully expoloited to tackle problems of the instability of a non-linearly elastic plate, see Fu & Rogerson (1994), or impact waves in an elastic plate, see Erbay et al. (2000). These simplications result from the fact that neo-Hookean strain energy function satises the conditions (1.72)1 , as well as permitting complete factorisation of the secular equation (1.71). Unfortunately this factorisation is related with the degeneration of two families of shear waves into one, which makes the dynamic response of neo-Hookean material somewhat limited in three dimensions. On the other hand, the series of experiments carried out by Treloar (1944) show that this model provides an adequate approximation to the (static) behaviour of rubber-like solids in the range of small strains. It must nally be noted that any neo-Hookean material is strongly elliptic for all deformations provided > 0. This assertion can easily be veried with the help of conditions (1.36), (1.45) and (1.59).
1.6.2
Mooney-Rivlin materials
Mooney (1940) derived the general form of the strain energy function for which the shear is proportional to the shearing stress. This function may be written as W = 1 2 (I 3) + (II 3) , 2 2 (1.85)
in which 1 is the shear (rigidity) modulus in the ground state and 2 characterises the departure from the symmetric model (1.84). The function (1.85), usually referred as the Mooney-Rivlin strain energy function, also provides serious simplications of the analysis, for instance Boulanger & Hayes (1992) have shown that the propagation of nite elastic waves in a pre-stressed Mooney-Rivlin material is governed by a linear equation (no linearisation is necessary). Materials characterised by (1.85) satisfy the conditions (1.72)1 , thus making possible the factorisation of the secular equation. The experiments of Rivlin & Saunders (1951) indicated that the range of applicability of (1.85) is similar to that of neo-Hookean function (1.84), providing only marginally better t of experimental data. We will sometimes illustrate our results for the Mooney-Rivlin material introduced on the basis of the experiments by Jones & Treloar (1975) with the material parameters 1 = 3.0 kg/cm2 and 2 = 1.1 kg/cm2 . The components of the elasticity tensor associated with a Mooney-Rivlin material may be obtained from the expressions (1.15)(1.17), yielding Biiii = (1 + 2 (2 + 2 ))2 , j i k Bijij = (1 + 2 2 )2 , k i Biijj = 22 2 2 , i j i = j = k = i, Bijji = 2 2 2 , i j (1.86)
i, j, k {1, 2, 3} ,
with no summation over repeated suces. The substituting of the denitions (1.86) into the conditions (1.36), (1.45) and (1.59) reveals that Mooney-Rivlin materials are strongly elliptic for
For the rubbers used in the experiments of Treloar (1944) good agreement with theory was achieved for the range of principal stratches about 0.75 < i < 1.4, i {1, 2, 3} (note that Treloar experiments did not involve contraction and therefore the lower bound for i is purely assumed to be bigger then the inverse of the upper bound). All of the material parameters in the paper by Jones & Treloar (1975, p. 1297) were given in N/mm2 .
22
all deformations provided one of the parameters 1 and 2 is strictly positive and another nonnegative. In order to maintain a neo-Hookean material as a particular case of the Mooney-Rivlin material it seems reasonable to assume 1 > 0 , 2 0,
as the sucient condition for strong ellipticity of Mooney-Rivlin material.
1.6.3
Varga materials, separability of the strain energy function

W = (1 + 2 + 3 3) , (1.87)
The class of materials with the strain energy function
where is a shear modulus, was rst introduced by Varga (1966) specically as a rst approximation to the behaviour of rubber-like solids. The Varga strain energy function does not satisfy the conditions (1.72)1 , so the factorisation of the secular equation is only possible for the aforementioned bi-axial deformations. However, the dynamic response of a Varga material is richer then that of neo-Hookean or Mooney-Rivlin material, compare for example Rogerson & Fu (1995) and Rogerson (1997). The components of the elasticity tensor corresponding to the Varga material may be given by Biiii = 0 , Bijij = 2 i , i + j Biijj = 0 , Bijji = i j , i + j (1.88)
i=j,
i, j {1, 2, 3} ,
where summation over repeated suces is not assumed. It may then be shown with a help of the conditions (1.36), (1.45) and (1.59), that a Varga material is strongly elliptic for all deformations provided > 0. This in particular guarantees the strong ellipticity of the Varga material with = 8.0 kg/cm2 , introduced by Ogden (1972, p. 573) to t the data of Treloar (1944). We will use this material in our numerical computations. Although the range of applicability of the Varga model is similar to that of a neo-Hookean or a Mooney-Rivlin material (only relatively small deformations are allowed), it is principally dierent from both of them since the strain energy function (1.87) does not possess a (simple) representation in terms of the invariants of the deformation (1.83). This indicates that in general the strain energy function is not a simple (e.g. polynomial) function of the invariants. Valanis & Landel (1967) proposed as a hypothesis that the strain energy function of a rubberlike material may be expressible as the sum of three independent functions of the principal stretches, i.e. W = w(1 ) + w(2 ) + w(3 ) , (1.89) in which function w(1) = 0 and w() must by symmetry of W be the same for each of the stretches. The theoretical justication for use of the separable form of the strain energy function (1.89) has been provided by Ogden (1974a) who shown that the Taylor series expansion about 1 = 2 = 3 = 1 is of separable form up to the fth order in i 1, i {1, 2, 3}. We remark that all of the previously considered forms of the strain energy function may be represented in separable form.
1.6.4
Ogden materials
In an attempt to t the data of Treloar (1944), Ogden (1972) proposed the separable multi-term strain energy function of the form
N
W =
k=1
k k ( + k + k 3) , 2 3 k 1
(1.90)
23
in which k are constants and k are possibly non-integer parameters. It is easy to check that (1.90) encloses all of the previously considered forms of the strain energy function. For example, in view of the incompressibility condition J = 1, the case of N = 2, 1 = 2 and 2 = 2 corresponds to the Mooney-Rivlin material. The material model (1.90) with N = 3 was shown to provide good approximation to the (static) behaviour of rubber-like solids. It was sucient to t the data of Treloar (1944) for the complete range of the deformations, and this result was reiterated in the series of additional experiments by Jones & Treloar (1975). In fact, even the single term function (1.90) may serve as a good approximation to the experimental results for small and moderate strains (up to i 2.0), see Ogden (1972, p. 575). Therefore we will sometimes use the one term strain energy function given by W = ( + + 3) , (1.91) 2 3 1 with = 6.8 kg/cm2 and = 1.2. The components of the elasticity tensor B associated with the strain energy function (1.91) may be written as Biiii = ( 1)2 , Biijj = 0 , i 2 2 j j i 2 i i j i 2 , i = j , , i = j , 2 2 i 2 i j = Bijji = j 1 1 , = , ( 2) , i = j , i j i i 2 2 i=j, i, j {1, 2, 3} ,
Bijij
(1.92)
with no summation over repeated suces. The conditions (1.36), (1.45) and (1.59) may then be used in conjunction with the relations (1.92) to establish the necessary and sucient strong ellipticity conditions for the Ogden material (1.91). The general analysis of these conditions will involve many dierent cases and will not be performed here. For our purposes however it is sucient to note that when 1, the condition > 0 is sucient to ensure the strong ellipticity of the Ogden material (1.92) for any deformation, see Zee & Sternberg (1983, p. 85). We conclude the discussion of the models of rubber-like materials providing a Table 1.1 (p. 23) which containes the abbreviated names and the parameter values of the materials to be used in subsequent numerical computations. Abbreviated name MRM OM1 VM Material Mooney-Rivlin Ogden (one term) Varga Material parameters 1 = 3.0 kg/cm2 , 2 = 1.1 kg/cm2 = 6.8 kg/cm2 , = 1.2 = 8.0 kg/cm2
Table 1.1: Specic materials to be used in generating graphs.
1.7
Measure of incremental surface traction
In the course of our study we will need to prescribe incremental tractions, i.e. tractions associated with the secondary deformation Be Bt , along plane material surfaces. We consider an elementary surface (dS, N ) with area dS and outward normal N in the unstressed conguration B0 . The deformations F and F are then imposed on B0 , transforming (dS, N ) into (d, n) and s (ds, n), respectively, see Figure 1.2. The contact force on the surface (d, n) is given by T nd s s and the contact force on (ds, n) is given by T nds, hence we consider the increment of contact force associated with the secondary deformation Be Bt written as f c = T nds T nd . s (1.93)
These parameters were taken from Ogden (1972, p. 574).
24
N F dS
n F d s
ds
Figure 1.2: Deformation of an elementary surface.
By considering the volumes formed by the elementary surfaces and their associated normals before and after deformation, we derive Nansons formulae, see Ogden (1984, p. 88), which in our case take form T s T T N dS = F nd = F F nds = F T nds . (1.94) We can now use (1.94) to represent increment of contact force (1.93) with respect to the surface in the statically deformed conguration Be , yielding f c = 1 F
T
n d = n d , s s
df
(1.95)
where n is the increment of surface traction associated with the deformation Be Bt . On the other hand, we may use connection between the Cauchy and nominal stress, see e.g. Chadwick (1999, p. 99), to conclude T nd = N dS , s
T
T nds = T N dS ,
(1.96)
which in view of (1.94) and (1.95) gives, when referred to the equilibrium conguration Be , the measure of the incremental surface traction in the form T T n = T F n , (1.97)
where n assumed to be the outward normal to the surface in the equilibrium conguration Be . Once again expanding the nominal stress tensor in a Taylor series about the equilibrium conguration Be , we may obtain the representation of the surface traction increment in coordinate form through ni = Bmilk uk,l nm + p um,i nm pt ni . (1.98) When the coordinate axes coincide with the principal axes of the deformation in the equilibrium conguration Be , the static pressure p possesses a simple interpretation in terms of diagonal components of the Cauchy stress tensor. Bearing in mind that in general the Cauchy stress T = F /J and in view of the relations (1.5) and (1.11) we obtain p = Bijij Bijji i , i=j, i, j {1, 2, 3} , (1.99)
which will often be used to eliminate p in favour of one of the principal Cauchy stress components i , i {1, 2, 3}.
Chapter 2
Dispersion of harmonic waves in a layer with free faces

Having obtained the equations governing the propagation of small amplitude waves in prestressed incompressible elastic media in Chapter 1, we will now focus on a specic type of body B. Our concern in this thesis is a nitely deformed layer, composed of such elastic material. Specically, this and the following four chapters are concerned with the propagation of harmonic waves in a layer with free surfaces, i.e. with zero incremental traction on the surfaces. We assume that the nite deformation is homogeneous and one of its principal axes is orthogonal to the plane of the layer. A Cartesian system of coordinates is then introduced for the layer, with the coordinate axes along the principal axes of deformation. This enables us to employ the governing equations established in Chapter 1 for the analysis of harmonic wave propagation in the specied layer. In particular, the solutions of the secular equation (1.65) may be used to represent general wave solutions as a superpositions of linearly independent functions. When these representations are substituted into the boundary conditions, formulated on the basis of the incremental traction measure (1.98), a system of linear homogeneous equations is obtained. The condition for the existence of non-trivial solution is then synonymous with vanishing of the determinant of associated matrix of coetients. Because of the symmetry of the plate this condition may be decoupled into the symmetric and the anti-symmetric counterparts, usually referred to as the extensional and the exural dispersion relations. These relations are utilised to formulate the global stability criteria, which is analysed (mostly numerically) for certain states of the pre-stress. The dispersion relations are then analysed numerically to identify their principal properties as well as to distinguish between dierent asymptotic regimes of wave propagation. In this thesis our attention is entirely focused on the propagation of long waves, i.e. waves with wave length which is much greater than layer thickness. Therefore we conclude this chapter with an asymptotic analysis of the secular equation (1.65) in one of the long wave regimes.
2.1
Harmonic wave propagation in a pre-stressed incompressible elastic layer
Our concern in this thesis is a nitely deformed layer of thickness 2h, composed of an incompressible elastic material. We place the origin O of a Cartesian coordinate system Ox1 x2 x3 in the mid-plane of the layer with axis Ox2 orthogonal to the layer, see Figure 2.1, and focus attention on homogeneous primary deformations of the form x1 = 1 (X1 cos + X3 sin ) ,
x2 = 2 X2 ,
x3 = 3 (X3 cos X1 sin ) ,
(2.1)
As opposed to the local stability criteria, which we termed the strong ellipticity condition.
25
CHAPTER 2. WAVES IN A LAYER WITH FREE FACES
26
where is an arbitrary angle. The important property of the deformation (2.1) is that every material vector orthogonal to the plane of the layer will remain orthogonal after deformation. The principal axes of this deformation coincide with the chosen coordinate axes and parameters m , m {1, 2, 3}, are the principal stretches. Thus we are justied to use the explicit form of the linearised equations of motion (1.18)(1.20), which must be solved in conjunction with the corresponding linearised incompressibility condition (1.21). We may now consider the propagation of a harmonic wave of the specic form (1.64). The separation of variables technique demonstrated in the Section 1.5 then yields the necessary and sucient condition for the existence of non-trivial solution (1.64) in the form of the secular equation (1.65). If the three generally distinct (non-zero) roots of the bi-cubic equation (1.65) are denoted 2 2 2 by q1 , q2 and q3 , any solution for u1 , u2 , u3 or pt can be represented as a superposition of the six linearly independent functions exp(kqm x2 ) and exp(kqm x2 ), m {1, 2, 3} (in order to avoid possible ambiguity, we hereafter assume that the real part of qm is positive). These superpositions may be expressed in the form
3
uj =
m=1 3
Uj
(2m1) kqm x2
+ Uj
(2m) kqm x2
eik(x1 c +x3 s vt) , eik(x1 c +x3 s vt) ,
j {1, 2, 3} ,
(2.2) (2.3)
pt = k
m=1 (m)
Pt
(2m1) kqm x2
+ Pt
(2m) kqm x2
within which Uj and Pt , j {1, 2, 3}, m {1, 2, 3, 4, 5, 6}, are twenty four disposable constants. These constants are not all independent as they are connected through the equations of motion (1.18)(1.20) and the incompressibility condition (1.21). We may therefore utilise (1.18) (1.20) and (1.21) to obtain the solutions for the displacements and incremental pressure in terms of only six disposable constants, yielding
3
(m)
u1 =
m=1 3
iqm U1 (qm , v ) (2m1) kqm x2 (2m) U2 e U2 ekqmx2 c eik(x1 c +x3 s vt) , V(qm , v ) U2
(2m1) kqm x2
(2.4) (2.5) (2.6) (2.7)
u2 =
m=1 3
+ U2
(2m) kqm x2
eik(x1 c +x3 s vt) ,
u3 =
iqm U3 (qm , v ) (2m1) kqm x2 (2m) U2 e U2 ekqmx2 s eik(x1 c +x3 s vt) , V(qm , v ) m=1 qm P(qm , v ) (2m1) kqm x2 (2m) U2 e U2 ekqm x2 eik(x1 c +x3 s vt) , V(qm , v ) m=1 x2
3
pt = k
h x1 O h
h h x3
Figure 2.1: Cartesian coordinate system for a layer with free faces.
CHAPTER 2. WAVES IN A LAYER WITH FREE FACES with the functions U1 (qm , v ), V(qm , v ), U3 (qm , v ) and P(qm , v ) dened by
2 U1 (qm , v ) = 23 qm + 12 s2 13 c2 + v 2 , 2 V(qm , v ) = (21 s2 + 23 c2 )qm + v 2 c5 ,
27
(2.8) (2.9) (2.10) (2.11)
P(qm , v ) = U1 (qm , v )U3 (qm , v ) + (b31 b32 )U1 (qm , v )c2 + (b13 b12 )U3 (qm , v )s2 ,
2 U3 (qm , v ) = 21 qm 31 s2 + 23 c2 + v 2 ,
where we utilised the denitions of the material parameters (1.30) and parameter c5 was dened immediately after (1.65). It must be remarked that the above representation of P(qm , v ) can only be obtained through use of the following identity
2 2 qm U1 (qm , v )U3 (qm , v ) = (13 qm + c4 v 2 )V(qm , v ) ,
(2.12)
which is veried with help of the secular equation (1.65). We also note that the function V(q, v ) in the denominators of the displacement solutions (2.4), (2.5) and (2.7) may vanish for certain combinations of the material parameters and wave speed. However, in view of the denitions (2.8)(2.10) the conclusion can be made that V(qm , v ) = U1 (qm , v )c2 + U3 (qm , v )s2 , (2.13)
which together with the identity (2.12) and the previous agreement on c = 0 and s = 0 means that vanishing of V(qm , v ) is always simultaneous with the vanishing of U1 (qm , v ), U3 (qm , v ) and, see (2.11), P(qm , v ). Therefore the situation when V(qm , v ) 0 corresponds to the propagation of waves with u1 u3 pt 0, which can in our present context be shown to be spurious solutions. To determine the possible wave speeds of these spurious waves we rst note that when V(qm , v ) = 0, the associated wave speed may be given by v 2 = c5 q 2 (23 c2 + 21 s2 ) ,
2
(2.14)
see (2.9). When the wave speed satises (2.14) the secular equation (1.65) factorises, yielding q 2 (23 21 )q 2 (23 + 13 )c2 + (12 + 31 )s2 = 0. (2.15)
There are two possible solutions of the equation (2.15), which after referring to (2.14) give two possible wave speeds of the spurious solutions as v 2 = c5 , v2 = (12 21 + 23 31 )s2 (23 23 + 13 21 )c2 . 23 21 (2.16)
These exact values of the wave speed may be used in the numerical routine to dispose of the spurious solutions. It is interesting to note that the nature of degenerations occuring at these two wave speeds is quite dierent in every case. Specically, the wave speed given by (2.16)1 is associated with double zero root in q, which invalidates the form of the solution (2.2), (2.3). However, the wave speed given by (2.16)2 does not invalidate the solution (2.2), (2.3) (the double root in q 2 evident from (2.15) only corresponds to situation (2.14)). It is also worth remarking that the solutions associated with (2.16) will not be presented on any plots demonstrated in this thesis.
We will concern ourselves in this thesis with two types of the boundary conditions. The free faces (zero incremental surface traction) boundary conditions will be shown to have form (2.20)(2.22) and therefore be automatically satised when V(qm , v ) U1 (qm , v ) U3 (qm , v ) 0. Four of the xed faces (zero displacement) (2m1) (2m) boundary conditions will clearly be satised for u1 and u3 , whereas two parameters U2 and U2 provide enough exibility to accommodate the last two xed faces conditions for u2 . Thus for the both considered types of the boundary conditions these solutions correspond to the wave whose phase speed is independent of the wave number, i.e. non-dispersive wave.
28
2.2
The dispersion relations
The upper and lower surfaces of the layer are dened in the specied system of coordinates by the outward normals su = (0, 1, 0) and sl = (0, 1, 0), respectively. Substituting the components of these vectors into the expressions for the measure of incremental traction (1.98) we obtain the following explicit representations for the traction increment on the surface of the layer s1 = B2121 u1,2 + (B2112 + p )u2,1 , s2 = B2211 u1,1 + (B2222 + p )u2,2 + B2233 u3,3 pt , s3 = (B2332 + p )u2,3 + B2323 u3,2 . (2.17) (2.18) (2.19)
The solutions for the displacement components and incremental pressure (2.4)(2.7) may then be inserted into the relations (2.17)(2.19) to derive the measure of incremental surface traction as a linear combination of the functions exp(kqm x2 ) and exp(kqm x2 ), m {1, 2, 3}. It may be given in the component form by s1 = s2 =
m=1 3
iT1 (qm , v ) (2m1) kqm x2 (2m) e + U2 ekqm x2 c eik(x1 c +x3 s vt) , U2 V(qm , v ) m=1
3
(2.20) (2.21) (2.22)
qm T2 (qm , v ) (2m1) kqm x2 (2m) e U2 ekqmx2 eik(x1 c +x3 s vt) , U2 V(qm , v ) iT3 (qm , v ) (2m1) kqm x2 (2m) U2 e + U2 ekqm x2 s eik(x1 c +x3 s vt) , V(qm , v )
s3 =
m=1
where T1 (qm , v ) = 21 U1 (qm , v )qm + g1 V(qm , v ) , 2 (2.23) 13 )U3 (qm , v )s2 U1 (qm , v )U3 (qm , v ) , (2.24) (2.25)
T2 (qm , v ) = T3 (qm , v ) =
(g1 13 )U1 (qm , v )c2 + (g3 2 23 U3 (qm , v )qm + g3 V(qm , v ) .
In the expressions (2.23)(2.25) we introduced the new parameters gi = 2i 2 ,

df
Gi = 22i 2 ,
df
i {1, 3} ,
(2.26)
which were dened using the identities (1.99) to eliminate p in favour of 2 in the expres sions (2.20)(2.22). It is worth remarking that parameters Gi , i {1, 3}, were not used in (2.20) (2.22) and dened for future convenience. In the course of this and the following four chapters we will focus our attention on the analysis of a layer with the free faces (zero incremental surface traction) boundary conditions, which may be formulated through si
x2 =h
= 0,
i {1, 2, 3} .
(2.27)
Inserting the denitions (2.20)(2.22) into these boundary conditions we arrive at a homogeneous system of six linear equations in six unknowns, given by
3 m=1 3 m=1
T1 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v ) T1 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v )
(2.28)
(2.29)

3 m=1 3
29
qm T2 (qm , v ) (2m1) kqm h (2m) U2 e U2 ekqmh = 0 , V(qm , v )
(2.30)
qm T2 (qm , v ) (2m1) kqm h (2m) U2 e U2 ekqm h = 0 , V(qm , v ) m=1

3 m=1 3
(2.31)
T3 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v )
(2.32)
T3 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v ) m=1
(2.33)
from which we omitted common non-vanishing factors. The symmetry of the layer about the midplane permits decoupling of this system into two homogeneous linear systems of three equations in three unknowns. The rst of these systems may be revealed by subtracting equation (2.29) from (2.28), adding (2.30) to (2.31), and subtracting (2.33) from (2.32), to obtain T1 (qm , v ) (m) Sm (h)U = 0 , V(qm , v ) m=1
3 m=1 3 m=1 df 3
(2.34) (2.35) (2.36)

(2m1) (2m)
qm T2 (qm , v ) (m) Cm (h)U = 0 , V(qm , v ) T3 (qm , v ) (m) Sm (h)U = 0 , V(qm , v )

df (m) df
in which Cm (h) = cosh(kqm h), Sm (h) = sinh(kqm h) and U = U2 U2 . The second system is similarly obtained by adding (2.29) to (2.28), subtracting equation (2.30) from (2.31), and adding (2.33) to (2.32), to yield T1 (qm , v ) (m) Cm (h)U+ = 0 , V(qm , v ) m=1 qm T2 (qm , v ) (m) Sm (h)U+ = 0 , V(qm , v ) m=1
3 m=1 3 3
(2.37) (2.38) (2.39)
T3 (qm , v ) (m) Cm (h)U+ = 0 , V(qm , v )
where U+ = U2 + U2 best analysed separately.
(m) df
(2m1)
(2m)
. These two systems of the boundary conditions for the layer are
2.2.1
Extensional (symmetric) motion

(m) (2m1) (2m)
We rst note that if U+ = 0, m {1, 2, 3}, then U2 = U2 , m {1, 2, 3}, and it is easy to see from (2.4)(2.7) that all of the displacement components and incremental pressure are symmetric functions with regard to x2 except u2 . This type of motion is usually referred to as extensional or symmetric motion. The boundary conditions (2.37)(2.39) are then satised identically and we only need to consider the homogeneous system of linear equations (2.34) (2.36). This system has non-trivial solution provided the determinant of the associated matrix of the coecients is equal to zero. We therefore require T1 (q1 , v )S1 (h) T1 (q2 , v )S2 (h) T1 (q3 , v )S3 (h) q1 T2 (q1 , v )C1 (h) q2 T2 (q2 , v )C2 (h) q3 T2 (q3 , v )C3 (h) = 0 , T3 (q1 , v )S1 (h) T3 (q2 , v )S2 (h) T3 (q3 , v )S3 (h) (2.40)
30
in which common non-zero multipliers have been omitted. Evaluating the determinant in (2.40) and dividing by C1 (h)C2 (h)C3 (h) leads to an explicit representation of the dispersion relation for extensional waves which, after a little algebraic manipulation, is given by
2 2 (q2 q3 )T2 (q1 , v )H(q2 , q3 , v )q1 T2 (h)T3 (h)
2 2 (q1 q3 )T2 (q2 , v )H(q1 , q3 , v )q2 T1 (h)T3 (h)
2 2 +(q1 q2 )T2 (q3 , v )H(q1 , q2 , v )q3 T1 (h)T2 (h) = 0 ,
(2.41)
where Tm (h) = tanh(kqm h), m {1, 2, 3}, and

2 2 H(qi , qj , v ) = 21 23 H1 ()qi qj + (2 c5 ) 21 23 (23 21 )(qi + qj ) H2 () , v 2 2 v v
(2.42)
H2 () = g1 23 U3 (0, v ) g3 21 U1 (0, v ) , v
H1 () = (23 v
21 )(23 c2
21 s2 )
+ 23 U3 (0, v ) 21 U1 (0, v ) ,
with T2 (qm , v ) dened in (2.24). We remark that the dispersion relation (2.41) was seemingly rst derived, in slightly dierent notation, by Rogerson & Sandiford (1999). We also note that the components of the displacement (2.4)(2.6) and incremental pressure (2.7) associated with extensional motion may be written as
3
u1 =
m=1 3
iqm U1 (qm , v ) (m) U Cm (x2 )c eik(x1 c +x3 s vt) , V(qm , v ) U Sm (x2 )eik(x1 c +x3 s vt) ,
(m)
(2.43) (2.44) (2.45) (2.46)
u2 =
m=1 3
u3 =
iqm U3 (qm , v ) (m) U Cm (x2 )s eik(x1 c +x3 s vt) , V(qm , v ) m=1 qm P(qm , v ) (m) U Cm (x2 )eik(x1 c +x3 s vt) , V(qm , v ) m=1
df 3
pt = k
df
within which Cm (x2 ) = cosh(kqm x2 ) and Sm (x2 ) = sinh(kqm x2 ). Provided the extensional dispersion relation (2.41) is satised, the homogeneous system of boundary conditions (2.34) (2.36) possesses a non-trivial solution. The relations (2.43)(2.46) may then be inserted into (m) the boundary conditions (2.34) and (2.36) to establish representations of the coecients U , m {1, 2, 3}, in terms of the single parameter U , yielding
(k) U 2 2 (qi qj )H(qi , qj , v )V(qk , v ) = (1) U , Sk (h) i <j, k {i, j} , / i, j, k {1, 2, 3} . k
(2.47)
We will use these relationships in due course, when analysing the relative asymptotic orders of the displacements in long wave regimes.
2.2.2
Flexural (anti-symmetric) motion

(m) (2m1) (2m)
The condition U = 0, m {1, 2, 3}, implies that U2 = U2 , m {1, 2, 3}, and therefore u1 , u3 and pt are anti-symmetric functions of x2 and u2 is a symmetric function, see (2.4)(2.7). This type of motion is usually termed as exural or anti-symmetric motion and it is clear that for such motions the boundary conditions (2.34)(2.36) are satised identically.
31
We therefore consider here the homogeneous system of linear equations (2.37)(2.39), which possesses non-trivial solution provided T1 (q1 , v )C1 (h) T1 (q2 , v )C2 (h) T1 (q3 , v )C3 (h) q1 T2 (q1 , v )S1 (h) q2 T2 (q2 , v )S2 (h) q3 T2 (q3 , v )S3 (h) = 0 , T3 (q1 , v )C1 (h) T3 (q2 , v )C2 (h) T3 (q3 , v )C3 (h) (2.48)
within which we omitted common non-vanishing factors. Evaluating the determinant in (2.48) we obtain the exural dispersion relation
2 2 (q2 q3 )T2 (q1 , v )H(q2 , q3 , v )q1 T1 (h)
2 2 (q1 q3 )T2 (q2 , v )H(q1 , q3 , v )q2 T2 (h)
2 2 +(q1 q2 )T2 (q3 , v )H(q1 , q2 , v )q3 T3 (h) = 0 ,
(2.49)
which was seemingly rst derived, in slightly dierent notation, by Rogerson & Sandiford (1999). The denitions of the functions T2 (q1 , v ) and H(qi , qj , v ) were given by (2.24) and (2.42), respec tively. The displacement components (2.4)(2.6) and the incremental pressure (2.7) may be represented for exural motion through u1 = u2 =
m=1 3
iqm U1 (qm , v ) (m) U+ Sm (x2 )c eik(x1 c +x3 s vt) , V(qm , v ) m=1

3
(2.50) (2.51) (2.52) (2.53)

(m)
U+ Cm (x2 )eik(x1 c +x3 s vt) , iqm U3 (qm , v ) (m) U+ Sm (x2 )s eik(x1 c +x3 s vt) , V(qm , v ) qm P(qm , v ) (m) U+ Sm (x2 )eik(x1 c +x3 s vt) . V(qm , v )
(m)
u3 =
m=1 3
pt = k
m=1
Making use of the boundary conditions (2.37) and (2.39) the coecients U+ , m {1, 2, 3}, may be represented in terms of the single parameter U+ independent of x2 , yielding U+ = (1)k
(k) 2 2 (qi qj )H(qi , qj , v )V(qk , v ) U+ , Ck (h) i <j, k {i, j} , / i, j, k {1, 2, 3} .
(2.54)
2.3
Neutral curves, stability
Before we begin analysing the dispersion relations, both numerically and asymptotically, attention must be paid to stability considerations. It has previously been mentioned that the local (material) stability of the body, ensured by the strong ellipticity conditions (1.36), (1.45) and (1.59), does not in general imply the (global) stability of a pre-stressed plate. In view of the fact that the stability condition may be dened as the requirement of positiveness of the square of wave speed, loss of the stability is then associated with vanishing of the square of wave speed (or frequency). The dispersion relations (2.41) and (2.49) provide implicit relationships between the scaled wave number kh and the scaled phase speed v (or scaled frequency = v kh) for a given state of the pre-stress. We therefore may insert v = 0 into (2.41) and (2.49), yielding the
df
CHAPTER 2. WAVES IN A LAYER WITH FREE FACES relations

2 2 (q2 q3 )T2 (q1 , 0)H(q2 , q3 , 0)q1 T2 (h)T3 (h)
32
2 2 (q1 q3 )T2 (q2 , 0)H(q1 , q3 , 0)q2 T1 (h)T3 (h)
2 2 +(q1 q2 )T2 (q3 , 0)H(q1 , q2 , 0)q3 T1 (h)T2 (h) = 0 ,
(2.55)
and
2 2 (q2 q3 )T2 (q1 , 0)H(q2 , q3 , 0)q1 T1 (h)
2 2 (q1 q3 )T2 (q2 , 0)H(q1 , q3 , 0)q2 T2 (h)
2 2 +(q1 q2 )T2 (q3 , 0)H(q1 , q2 , 0)q3 T3 (h) = 0 ,
(2.56)
associated with extensional and exural motions, respectively. It worth noting that the parameters q1 , q2 and q3 here are the solutions of the accordingly simplied secular equation (1.65), which may be written as 21 23 q 6 c1 q 4 + c3 q 2 c4 c5 = 0 , (2.57) with the parameters c1 , c3 , c4 and c5 dened immediately after (1.65). Equations (2.55) and (2.56) are usually referred to as bifurcation conditions and may be interpreted as restrictions on the range of pre-stresses for every specic kh. Keeping in mind the denitions (2.24), (2.26) and (2.42) the conclusion can be made that the implicit conditions (2.55) and (2.56) are in fact quadratic equations in the normal component of the Cauchy stress 2 . These equations are very algebraically complicated and the general analytic investigation of their solutions remains an unsolved problem, which contrasts the plane strain case thoroughly analysed by Ogden & Roxburgh (1993). However we may establish some properties of their solutions using the combination of the numerical and asymptotic methods as well as specialisation of the material model.
15 10 5 15 10 5
0 -5 -10 -15 0 1 2 3 4 5 6 7 2=0.6 2=1.1 2=1.8 8 9 10
0 -5 -10 -15 0 1 2 3 4 5 6 7 2=0.6 2=1.1 2=1.8 8 9 10
(a) Extensional motion, kh
(b) Flexural motion, kh
Figure 2.2: Neutral curves, showing Cauchy stress 2 against scaled wave number kh shown for the (a) extensional and (b) exural motion with an angle of propagation = 45 . The layer is composed of OM1, see Table 1.1 (p. 23), and subjected to the primary deformation with 1 = 0.9 and various values of 2 .
The numerical solutions of the relations (2.55) and (2.56) for the single term Ogden material are presented at the Figure 2.2 in the form of the neutral curves of Cauchy stress component 2 against scaled wave number kh. Although generated only for a few specic pre-stress congurations these graphs demonstrate the major features of the neutral curves for a pre-stressed plate. First, we remark that the square of the wave speed will be positive provided the normal stress lies in between of the two roots of the equations (2.55) and (2.56). Since we decomposed
This fact is easy to verify numerically and will also be indirectly conrmed by the asymptotic analysis in Chapters 5 and 6. Therefore, we defer further discussion until the later chapters to avoid repeating the material.
33
a general motion of the plate into the symmetric (extensional) and anti-symmetric (exural) motions, to ensure the global stability of the plate we must require stability of both extensional and exural motions. So we introduce the condition for the global stability of the plate in the following form e(l) f (l) e(u) f (u) max 2 , 2 < 2 < min 2 , 2 , (2.58)
kh[0,) e(u) kh[0,)
within which 2 and 2 are the lower and upper solutions of the equation (2.55) and 2 f (u) and 2 are the lower and upper solutions of the equation (2.56). It is worth noting, that generally these solutions depend on the angle of wave propagation . For some kinds of problems the material response is essentially three-dimensional and can not be reduced to plane strain situation (e.g. line loading of the edge of an semi-innite plate). To ensure the incremental stability of such problems it is necessary to verify the inequalities (2.58) for all angles involved in the generated response. However, when we talk about the incremental stability of a certain conguration in this thesis, we usually use it in the limited sense that this conguration is incrementally stable for responses only involving waves propagating at the angle to the axis Ox1 . Thus the stability region is bounded by the neutral curves of 2 . It may then be observed from the Figure 2.2 that the range of the stable values of 2 is wider for the extensional motion. Furthermore, a nite limit of the lower neutral curve at kh 0 does not exist for extensional motion. To conrm this observation analytically we consider the limit of the relation (2.55) as kh 0, which may be factorised and written in the form
2 2 2 2 2 2 21 23 q1 q2 q3 (q1 q2 )(q1 q3 )(q2 q3 ) 2c5 2 213 (12 + 21 )c4 231 (23 + 23 )s4
e(l)
f (l)
+ (2 412 23 + 221 12 + 223 23 + (21 23 )2 13 31 )c2 s2 = 0 , 13
(2.59)
with the parameter c5 dened immediately after (1.65). It is clear from (2.59) that the term 2 with 2 has cancelled out and the only nite limit available at kh = 0 is the upper limit. Thus a necessary condition for stability of the pre-stressed plate in extensional motion may be formulated as 2c5 2 < 213 (12 + 21 )c4 231 (23 + 23 )s4
+ (2 412 23 + 221 12 + 223 23 + (21 23 )2 13 31 )c2 s2 , 13
(2.60)
where we keep in mind that c5 is bound to be positive by the strong ellipticity conditions (1.45). Similar analysis may be performed for the exural relation (2.56). The limit at kh 0 may be factorised with the help of equalities (1.66)(1.68) to obtain
2 2 2 2 2 2 2 c5 (q1 q2 )(q1 q3 )(q2 q3 ) c8 2 221 23 2 + c9
c8 =
23 c2
(23 23 + 13 21 )c2 (21 12 + 23 31 )s2 = 0 , + 21 s2 , c9 = 21 23 (21 c2 + 23 s2 c4 ) .
(2.61)
Therefore the limits of the neutral curves for exural motion at kh = 0 can be found as zeros of a quadratic in 2 . For exural stability of the pre-stressed plate it is then required that 21 23 c8 21 23 c8
2
c9 21 23 < 2 < + c8 c8
21 23 c8
c9 . c8
(2.62)
Another interesting feature of the graphs depicted on both plots in Figure 2.2 is apparent independence of the upper neutral curve on kh. However, numerical analysis shows that this curve is not generally constant and does depend on kh, but with only a small variation. We will conclude this section by providing some additional remarks for specic material models. In Figure 2.3 the neutral curves for a Mooney-Rivlin material are presented. This model
34
10
10
-5 2=0.8 2=1.1 2=1.4 0 1 2 3 4 5 6 7 8 9 10
-5 2=0.8 2=1.1 2=1.4 0 1 2 3 4 5 6 7 8 9 10
-10
-10
Figure 2.3: Neutral curves of Cauchy stress component 2 against scaled wave number kh shown for the (a) extensional and (b) exural motion with an angle of propagation = 15 . The layer is composed of MRM, see Table 1.1 (p. 23), and subjected to the primary deformation with 1 = 0.9 and various values of 2 .
does not provide a good representation of the elastic behaviour of rubber-like materials at large deformations, these were therefore appropriately reduced in comparison with the Figure 2.2. The major properties of the neutral curves which we were able to observe previously are all kept intact. Our only comment is that the numerical procedure for plotting the neutral curves may be seriously simplied for this material because of the complete factorisation of the secular equation occurring at the zero wave speed. Thus the analogous secular equation to (2.57) may be given for a Mooney-Rivlin material by (q 2 1)(2 q 2 2 c2 2 s2 ) 2 1 3
2 (1 + 2 2 )(1 + 2 2 )q 2 (1 + 2 2 )(1 (2 c2 + 2 s2 ) + 2 2 2 ) = 0 , 2 1 3 2 1 3 1 3
(2.63)
in which 1 and 2 are the material parameters and m , m {1, 2, 3}, are the principal stretches 2 of primary deformation. Note that all roots qm , m {1, 2, 3}, of this equation are positive ensuring that there are no complex or purely imaginary solutions and the functions Tm (h), m {1, 2, 3}, behave as hyperbolic tangents.
10 10
-5
-5
-10
-15 0 1 2 3 4 5 6 7
2=0.8 2=1.1 2=1.4 8 9 10
-10
-15 0 1 2 3 4 5 6 7
2=0.8 2=1.1 2=1.4 8 9 10
Figure 2.4: Neutral curves of Cauchy stress 2 against scaled wave number kh shown for the (a) extensional and (b) exural motion with an angle of propagation = 75 . The layer is composed of VM, see Table 1.1 (p. 23), and subjected to the primary deformation with 1 = 0.9 and various values of 2 .
Further information is obtainable for Varga materials. The typical plots of the neutral curves shown in Figure 2.4 do not demonstrate signicant departures from the previous material models.

4.5 4 70 3.5 60 90 80
35
3 2.5 2
50 40 30 20 10
1.5 5 10 15 20 25 30 35 40 45 50
0 0 5 10 15 20 25 30
kh
kh
Figure 2.5: Scaled wave speed v and frequency against scaled wave number kh for extensional motion in the MRM with the angle of propagation = 45 . The pre-stressed state is specialised to 1 = 0.9, 2 = 1.4 and 2 = 5.0 (see Figure 2.3 for the associated neutral curves).
However, for the Varga material the cubic (in q 2 ) secular equation (2.57) possesses a double root and may be represented as 2 2 q 2 1 c2 3 s2
2
2 (1 + 3 )q 2 2 (2 + 3 )c2 2 (1 + 2 )s2 2 1 3 = 0, (1 + 2 )(1 + 3 )(2 + 3 )
(2.64)
2 which clearly possesses three positive solutions qm , m {1, 2, 3}, similarly with Mooney-Rivlin materials. Strictly speaking, the relations (2.55) and (2.56), derived with the assumption that all qm , m {1, 2, 3}, are distinct, are not therefore valid for the Varga materials. Nevertheless the continuity of all of the functions in (2.55) and (2.56) ensures that limits of these relations as the material parameters tend to those of a Varga material coincide with the correct relations for it. Thus we can utilise here the results of the general analysis of neutral curves that we obtained previously. In doing so we immediately obtain that for the Varga materials the necessary conditions (2.60) and (2.62) recast to 2 < 2 , (2.65)
and 2 22
1 c2 + 3 s2 < 2 < 2 , 1 c2 + 3 s2 + 2
(2.66)
respectively. We nally remark that the upper neutral curve is actually constant for the Varga material, which is related to the fact that T2 (q3 , 0) = 0 when 2 = 2 . Therefore, the upper stability limit for a Varga material is known exactly and independent of kh. The general stability analysis for the pre-stressed three-dimensional plate remains an open task. However, the value of the normal component 2 of the Cauchy stress can always be veried against the numerical solutions of the relations (2.55) and (2.56) to ensure the stability of the chosen pre-stress conguration. On the other hand, the necessary stability conditions that we obtained in this section will prove useful in later parameter analysis.
2.4
Propagation of long waves
In Figure 2.5 some typical solutions of the extensional dispersion relation (2.41) are depicted. df df The scaled wave speed v = v and the scaled frequency = kh of the rst twenty solution v branches are shown against the scaled wave number kh for a Mooney-Rivlin material. It is now possible to distinguish between dierent asymptotic regimes of the solutions of (2.41), following the classication suggested by Kaplunov et al. (1998). It is worth remarking that since this

4
36
80 70 60
3.5
50
v
2.5
40
30
20 10
1.5
0
5 10 15 20 25 30 35 40 45 50
10
15
20
25
30
kh
kh
Figure 2.6: Scaled wave speed v and frequency against scaled wave number kh for exural motion in the MRM with the angle of propagation = 45 . The pre-stressed state is specialised to 1 = 0.9, 2 = 1.4 and 2 = 5.0 (see Figure 2.3 for the associated neutral curves).
thesis is only concerned with the propagation of waves for which kh 1, we will not discuss here the motions associated with kh , usually referred to as short wave motions. As kh 0, the wave number is small and the wave length is therefore large, so we will term this type of motion as long wave motion. For such motion the wave length is much larger than the typical length scale of the structure (i.e. the thickness of the plate). The subsequent analysis is therefore particularly applicable to thin structures. It is clear from Figure 2.5 that the wave speeds of most of the solution branches tend to innity as kh 0. The corresponding frequencies have nite limits at kh = 0 which we will usually refer to as cut-o frequencies. This regime of wave propagation will therefore be termed as long wave high frequency motion and the corresponding solution branches will be termed harmonics. We remark that in our case it is possible to distinguish two families of cut-o frequencies, with the frequencies within each of the families separated by constant intervals. These two intervals are very similar for the solutions shown on the righthand plot on Figure 2.5, which gives an impression of couples of harmonics coming closer together as kh 0. On the other hand, the rst two modes on the Figure 2.5 have nite phase speeds at kh = 0 and their frequencies are O(kh). Hence we will refer to this type of motion as long wave low frequency motion terming the corresponding solution modes as fundamental. The observations just made for extensional motion are also generally valid for the solutions of the exural dispersion relation (2.49) shown in Figure 2.6. The only signicant dierence is that for the exural waves there is only one nite limit of phase speed at kh = 0. The existence of two such limits for extensional waves is a direct result of solving the full three-dimensional equations of elasticity. In the analogous plane strain problems there is exactly one fundamental mode for both exural and extensional motions, see Rogerson & Fu (1995) and Rogerson (1997). It worth noting that in general there is nothing unusual in having three nite speed limits at kh = 0 (two extensional and one exural). This was for example observed by Nayfeh (1991) for a layer composed of strongly anisotropic (monoclinic) media. We have already mentioned that Mooney-Rivlin materials do not provide as rich a dynamic
In the plane strain case a comprehensive numerical and asymptotic analysis of short wave motions in a pre-stressed incompressible elastic layer was performed by Rogerson & Fu (1995) and Rogerson (1997). To the best of the authors knowledge, the only (incomplete) numerical and asymptotic investigation of short wave motion in a three-dimensional pre-stressed incompressible layer was provided by Rogerson & Sandiford (1999). In the forthcoming paper by Kaplunov et al. (2001b) some plane strain asymptotic models for short wave motion in a pre-stressed incompressible elastic plate will be presented. This does not imply that the second fundamental mode of the extensional motion degenerates for waves propagating along one of the principal directions of a pre-stress. It was shown by Rogerson & Sandiford (1999) that the second extensional fundamental mode arises as the solution of the so-called anti-plane problem.

4 3.8 3.6
10 5 0
37
v 3.4
3.2 3 2.8 5 10 15 20 25 30 35 40 45 50
-5 -10 -15 -20 0 1 2 3 4 5 6 7 2=2.0 8 9 10
kh
kh
Figure 2.7: Scaled wave speed v against the scaled wave number kh shown together with the corresponding neutral curves for extensional motion in the OM1 with the angle of propagation = 15 . The pre-stressed state is specialised to 1 = 2.0, 2 = 0.7 and 2 = 2.0.
4 3.8 3.6
0 10
v 3.4
3.2 3 2.8 5 10 15 20 25 30 35 40 45 50
2
-5
-10 2=2.0 0 1 2 3 4 5 6 7 8 9 10
-15
kh
kh
Figure 2.8: Scaled wave speed v against the scaled wave number kh shown together with the corresponding neutral curves for exural motion in the OM1 with the angle of propagation = 15 . The pre-stressed state is specialised to 1 = 2.0, 2 = 0.7 and 2 = 2.0.
response as Varga or Ogden materials. This point is demonstrated for extensional and exural dispersion relations by Figure 2.7 and Figure 2.8, respectively, along with the corresponding neutral curves. The band of the oscillatory behaviour evident from the left hand plots of both gures can not exist for the Mooney-Rivlin material, see for example Rogerson (1997). However, in the long wave regime the principal properties of the solutions of the dispersion relations (2.41) and (2.49) do not change. Specically, for extensional motion there are two nite wave speed limits at kh 0, while all other branches possess speeds of O((kh)1 ). For exural motion there is only one nite limit of the phase speed, for all other solutions v (kh)1 as kh 0.
2.5
Long wave high frequency approximations of the secular equation
We have observed that as the scaled wave number kh 0, v for all harmonics of the dispersion relations (2.41) and (2.49), with the corresponding limiting (cut-o) frequencies remaining nite and non-zero. Thus, to nd appropriate asymptotic approximations of the secular equation we assume that v as kh 0. Analysis of the identities (1.66)(1.68), connecting 2 the roots of the secular equation with its coecients, suggests then that two of its roots (q1 and 2 ) are O(2 ), whereas the third root q 2 is O(1). Expanding these roots as power series in v 2 , q3 v 2
38
and substituting them into the secular equation, the following approximations may be derived
2 q1 = 2 q2
Q s2 + Q1c c2 c2 v2 (2) (2) + 1s Q1s s2 + Q1c c2 + O(4 ) , v 21 21 v2

(2) (2)
(0)
(0)
(2.67) (2.68) (2.69)
s2 + Q2c c2 c5 Q = 1 + 2s + O(4 ) , v v2 (0) (0) Q s2 + Q3c c2 s2 v2 (2) (2) 2 + 3s Q3s s2 + Q3c c2 + O(4 ) , v q3 = 23 23 v2
in which Q1c = 212 21 , Q1s

(2) (0)
= 31 32 + =
(0) Q1c
(2) Q1c (2) Q2c
= 212 21 12 , +
(0) Q3c
(13 + 21 )2 , 23 21
Q1s = 31 ,
(0)
Q3c = 13 , Q3c
(2)
(0)
Q3s = 223 23 , (13 + 23 )2 , 23 21 (2.70)
(0)
= 13 12
(0) (0)
12 ,
(2) Q3s (2) Q2s
= 223 23 32 ,
= Q1s + Q3s 32 .
It might be helpful to explain the rationale behind chosen notation for material and pre-stress parameters. Generally, we use superscripts in brackets to indicate the power of any associated large (or small) factor whereas subscripts are used to distinguish between specic parameters. For example, superscripts of the parameters dened in (2.70) refer to the associated power of v and subscripts ic and is , i {1, 2, 3}, indicate that the parameter specied is a coecient of 2 long wave high frequency expansion for qi , multiplied onto c or s , respecively. Corresponding expansions for q1 , q2 and q3 are given by i Q1s s2 + Q1c c2 i v + O(3 ) , v q1 = 21 2 21 v q2 = 1 + Q2s s2 + Q2c c2 c5 + O(4 ) , v 22 v
(0) (0) (2) (2) (0) (0)
(2.71) (2.72) (2.73)
i Q3s s2 + Q3c c2 i v q3 = + O(3 ) . v 23 2 23 v
The secular equation (1.65) may also be treated as quadratic in v 2 . For long wave high frequency 2 2 motion the wave speed must tend to innity as kh 0 and since v 2 = O(q1 ) (or O(q3 )), only 2 2 the wave speeds associated with q1 and q3 are of interest, with appropriate expansions given by
2 v1 = 21 q1 + Q1s s2 + Q1c c2 + Q1s s2 + Q1c c2 2 2 v3 = 23 q3 + Q3s s2 + Q3c c2 + Q3s s2 + Q3c c2 2 (0) (0) (2) (2) (0) (0) (2) (2)
c2 4 2 + O(q1 ) , q1 s2 4 2 + O(q3 ) . q3
(2.74) (2.75)
2 The two wave speeds associated with q2 , are of O(1) and are therefore not relevant for high frequency motion. We remark that the lack of a third large wave speed associated with q2 is a consequence of imposing the incompressibility constraint, which disabled propagation of a longitudinal wave. In an unconstrained material there is a third possible large speed associated with q2 , see for example Kaplunov et al. (2000a) in respect of a compressible transversely isotropic elastic plate.
Chapter 3
Layer with free faces: Long wave high frequency extensional motion
The rst asymptotic model that we derive is for long wave high frequency extensional motion of a pre-stressed incompressible elastic plate, which is associated with the extensional harmonics in the vicinity of the cut-o frequencies. The analysis begins with an asymptotic investigation of the corresponding dispersion relation. Two asymptotic regimes are shown to exist, those associated with two families of cut-o frequencies, and third order asymptotic expansions are obtained. We utilise these expansions to analyse the asymptotic structure of the waves propagating in the long wave high frequency regime and to show that the two families of cut-o frequencies are in fact two families of so-called shear resonance frequencies. Scalings of space and time variables are then introduced to represent the equations of motion, incompressibility condition and boundary conditions in such form that the contributions of various terms are made explicit. This allows a hierarchical system of boundary value problems at various asymptotic orders to be obtained. Solving this system we establish the solutions for all of the displacement components and incremental pressure in terms of a single function, referred to as the long wave amplitude. This function itself may be determined as the solution of one of the two-dimensional governing equations, obtained as the existence conditions in the process of derivation. The high level of consistency of this approach is demonstrated by the fact that the dispersion relation of the asymptotic governing equation matches the corresponding approximation of the exact dispersion relation. The chapter is concluded by some illustrative numerical examples.
3.1
Analysis of the dispersion relation
To begin our analysis we need to obtain the long wave high frequency approximation of the dispersion relation. At this stage it is convenient to introduce a small non-dimensional parameter df , dened as the ratio of a layer half-thickness h to a typical wavelength l, so = h/l = kh. The observations made in the previous chapter indicate that in the long wave high frequency df regime, as 0, v . Moreover, the scaled frequency is dened by = v kh and is O(1) for any high frequency motion. By inserting approximations (2.67)(2.69) into the dispersion relation (2.41), we obtain the leading order approximation of the dispersion relation, namely iA1 T2 (h)T3 (h) v 3 T1 (h)T3 (h) + iA3 T1 (h)T2 (h) 0 , A1 = G2 c2 1 21 , A3 = G2 s2 3 23 . (3.1)
with parameters Gi , i {1, 3}, dened in (2.26). Hereafter, it is assumed that the functions of material parameters A1 and A3 are generally O(1). Since q2 = 1 + O(2 ), see (2.72), it is v clear that T2 (h) = O(1 ). This observation suggests that terms on the left-hand side of the v 39
CHAPTER 3. HIGH FREQUENCY EXTENSIONAL MOTION
40
v v relation (3.1) can be asymptotically balanced only if T1 (h) = O(4 ) or T3 (h) = O(4 ). In fact, each of these asymptotic relations corresponds to the harmonics associated with one of the two possible families of cut-o frequencies.
3.1.1
Approximation near cut-o frequencies
We shall consider rst the case in which T1 (h) = O(4 ), the asymptotic order of T1 (h) then v implying the following expansions kq1 h = in + i1
e(4) 4
+ O( 6 ) ,
e(4)
T1 (h) = i1
e(4) 4
+ O( 6 ) ,
(3.2)
in which the function of material parameters 1 is to be determined. In this case an approximation for the wave speed is given by (2.74). We may invoke relations (2.71)(2.73), and the denition of Tm (h) given immediately after equation (2.41), to show that ie q1 = 1 + O( 3 ) , q2 = 1 + O( 2 ) , 21 e v = 1 + O() , T2 (h) = + O( 3 ) , within which e = 1
df
ie q3 = 1 + O() , 23 T3 (h) = O(1) ,
(3.3)
21 n, n = 1, 2, 3, . . . are the associated cut-o frequencies. It is now

e(4)
possible to determine an explicit expression for the function 1 by utilising expansions (3.2) and (3.3), together with the leading order approximation of the dispersion relation (3.1), to obtain A1 e(4) 1 = e 3 . (3.4) (1 ) The desired approximation of scaled frequency = v of long wave high frequency motion is nally derived from the wave speed approximation (2.74) and the expansion for q1 given in (3.2), yielding 2 = (e )2 + E1s s2 + E1c c2 2 + E1s s2 + E1c c2 c2 4 + O( 6 ) , 1 E1s = Q1s ,
(2) (0) (2) (2) (4) (4) df
(3.5)
(2)
E1c = Q1c ,
(2)
(0)
E1s =
(4)
2G2 21 Q1s 1 (e )2 1
(2)
E1c =
(4)
2G2 21 Q1c 1 (e )2 1
We shall now consider the second case T3 (h) = O(4 ). Following the same procedure we v expand q3 to obtain kq3 h = in + i3
e(4) 4
+ O( 6 ) ,
T3 (h) = i3
e(4) 4
+ O( 6 ) ,
(3.6)
with the wave speed given by expansion (2.75). Introducing the second family of cut-o fredf quencies e = 23 n, n = 1, 2, 3, . . . , and utilising relations (2.71)(2.73), the following ap3 proximations can be derived ie q3 = 3 + O( 3 ) , q2 = 1 + O( 2 ) , 23 e v = 3 + O() , T2 (h) = + O( 3 ) , The function 3
e(4)
ie q1 = 3 + O() , 21 T1 (h) = O(1) .
(3.7)
is now determined from (3.1), (3.6) and (3.7), thus 3

e(4)
A3 . (e )3 3
(3.8)
41
Finally, through use of (2.75) and (3.6), the approximation of the scaled frequency associated with the second family of cut-o frequencies is given by 2 = (e )2 + E3s s2 + E3c c2 2 + E3s s2 + E3c c2 s2 4 + O( 6 ) , 3
(2) E3s (2) (2) (4) (4)
(3.9)
(2)
(0) Q3s
(2) E3c
(0) Q3c
(4) E3s
2G2 23 Q3s 3 = (e )2 3
(2)
(4) E3c
2G2 23 Q3c 3 = (e )2 3
3.1.2
Relative asymptotic orders of displacements
Construction of our asymptotically consistent lower dimensional model requires knowledge of the relative orders of particle displacements for long wave high frequency motion. These may be obtained with the help of the representations of the general solutions for displacement components and pressure (2.43)(2.46) derived in Chapter 2. The coecients of these representations are not independent and related through (2.47). Substituting (2.47) into the displacement relations (2.43)(2.46) we may compare the relative orders of displacement components. To do this we will have to estimate the orders of all functions occurring in (2.43)(2.46) and (2.47), and we begin by noting that S2 (x2 ) = x2 + O( 3 ) , h C2 (x2 ) = 1 + O( 2 ) . (3.10)
The remaining functions must be analysed separately for the motions in the vicinity of each set of cut-o frequencies. It is also worth noting that we assume the value x2 /h to be O(1). In the rst case ( = e ) the expansions (3.3) and (3.10), supplemented by 1 S1 (x2 ) = i sin S3 (x2 ) = i sin e x2 1 21 h e x2 1 23 h + O( 2 ) , + O( 2 ) , C1 (x2 ) = i cos C3 (x2 ) = i cos
e(4) 4
e x2 1 21 h e x2 1 23 h
+ O( 2 ) , + O( 2 ) ,
S1 (h) = i(1)n 1
+ O( 6 ) ,
reveal that the orders of displacements are distributed as follows u1 O(pt ) , u2 O(pt ) , u3 2 O(pt ) . (3.11)
We repeat this routine for the second case ( = e ), taking together 3 S1 (x2 ) = i sin S3 (x2 ) = i sin e x2 3 21 h e x2 3 23 h + O( 2 ) , + O( 2 ) , C1 (x2 ) = i cos C3 (x2 ) = i cos
e(4) 4
e x2 3 21 h e x2 3 23 h
+ O( 2 ) , + O( 2 ) ,
S3 (h) = i(1)n 3
+ O( 6 ) ,
and approximations (3.7) to indicate that for the second family of cut-o frequencies u1 2 O(pt ) , u2 O(pt ) , u3 O(pt ) . (3.12)
The results (3.11) and (3.12) clearly indicate that in each case the leading order displacement is one of the in-plane displacements.
42
3.1.3
Physical interpretation
The chosen form of the wave solution (1.64) together with the long wave high frequency approximations (2.71) or (2.73) imply that to leading order the wave normal is collinear with a vector of the form (c , O( 1 ), s ). This means that for both families of cut-o frequencies the leading order direction of wave propagation is orthogonal to the layer. The corresponding wave polarisation directions can be estimated with help of the relative orders of the displacement components (3.11) and (3.12). These imply that the polarisation directions will coincide with non-normalised vectors of the form (O(1), O(), O( 2 )) and (O( 2 ), O(), O(1)). Therefore, in the long wave limit these waves degenerate into two waves, rapidly ( = O( 1 )) travelling v along the normal direction and polarised along one of the in-plane axes of primary deformation. What we have is actually a description of two shear waves, which complements our previous note that only two shear waves can propagate in any direction in an incompressible elastic solid, see (1.23)2 and the associated discussion. The previously introduced parameters e = 2m n, m m {1, 3}, n = 1, 2, 3, . . . , are the leading order terms of the scaled frequency expansions (3.5) and (3.9) respectively. They dene two innite families of cut-o frequencies, frequency limits as 0. Furthermore, the large propagation speed of two shear waves indicates that the param12 12 12 12 12 12 12 12
Figure 3.1: Symmetric shear resonance of an innitesimally thin transverse bre of the layer.
eters e , m {1, 3}, n = 1, 2, 3, . . . , are in fact natural thickness shear resonance frequencies m of an innitesimally thin transverse bre of the layer, see Figure 3.1, which satisfy one of the eigenvalue problems 2m um,22 + 2 um = 0 , um,2 |x2 =h = 0 , m {1, 3} . (3.13) We remark, that the existence of thickness stretch resonance (associated with normal displacement) is obviously precluded by the incompressibility constraint. A similar asymptotic analysis performed for the compressible transversely isotropic layer demonstrated the existence of the thickness stretch resonance and the associated family of cut-o frequency, see Kaplunov et al. (2000a).
3.2
Asymptotically approximate equations: rst family of shear resonance frequencies
The deriving of the asymptotically approximate equations requires appropriate rescaling of the spatial and time variables as well as particle displacements and pressure increment. We choose scalings of the spatial coordinates so as to balance x2 with half-thickness h and x1 , x3 with a typical wave length l x1 = l1 , x2 = h = l , x3 = l3 . (3.14) Corresponding rescaling of time, displacements and pressure is best done separately for each family of shear resonance frequencies. We shall consider both families in turn in this chapter. For motion in the vicinity of the rst family of shear resonance frequencies a typical wave travels with the scaled speed v e /, see appropriate approximation given in (3.3), covering 1 the distance of a typical wave length l in a time t l/e . Upon invoking the denition of 1 the e , given immediately after (3.3), we therefore introduce the following scaling of time 1 t = l . 21 (3.15)
43
The displacement components and incremental pressure are scaled according to their relative orders, indicated in (3.11), yielding um (x1 , x2 , x3 , t) = l m1 u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = m {1, 2, 3} , (3.16)
21 p (1 , , 3 , ) , t
in which the superscript denotes non-dimensional quantities of the same asymptotic order. The equations of motion (1.18)(1.20) can now be formulated in terms of the rescaled nondimensional variables, so that 21 u + (e )2 u (e )2 u + 21 u + 2 B1111 u 1 1 + 31 u 3 3 1 1 1 1 1, 1, 1, 1,
+ (B1122 + B1221 )u 1 21 p 1 + 4 B1133 + B1331 u 1 3 = 0 , 2, t, 3,
(3.17)
B2222 u + (e )2 u (e )2 u + 21 u + (B1122 + B1221 )u 1 1 2 1 2 2, 2, 1,
21 p + 2 12 u 1 1 + 32 u 3 3 + (B2233 + B2332 )u 3 = 0 , t, 2, 2, 3,
(3.18)
23 u + (e )2 u (e )2 u + 21 u + (B1133 + B1331 )u 1 3 1 3 1 3 3, 3, 1,
+ (B2233 + B2332 )u 3 21 p 3 + 2 13 u 1 1 + B3333 u 3 3 = 0 , 2, t, 3, 3,
(3.19)
in which we use ,1 , , , ,3 and , to denote dierentiation with respect to 1 , , 3 and , respectively. These equations must be solved in conjunction with the appropriate form of the incompressibility condition, namely u 1 + u + 2 u 3 = 0 , 1, 2, 3, and the incremental surface traction boundary conditions 2, 21 u + 2 (B1221 + p )u 1 = 0 1, B1122 u 1 1, + (B2222 + 23 u 3, + p )u 2, 21 p + t (B2332 + p )u 3 2,
2
(3.20)
B2233 u 3 3,
at = 1 , =0
(3.21) (3.22) (3.23)
=0
at = 1 .
at = 1 ,
We will seek solutions of the boundary value problem given by equations (3.17)(3.19) and (3.20), subject to (3.21)(3.23), in a form of the power series
m
(u , u , u , p ) = 1 2 3 t
n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(3.24)
The result of substituting solutions (3.24) into the equations (3.17)(3.23) is a hierarchical system of equations and boundary conditions at various orders. Keeping in mind that the resulting response must be compatible with both (3.5) and (3.13) we also require that for the motion in the vicinity of the shear resonance frequencies 21 u + (e )2 u 2 u , m, 1 m m m {1, 2, 3} , (3.25)
which can alternatively be veried by the direct substitution of the wave solution (1.64). This allows us to conclude that the quantities within braces in the equations of motion (3.17)(3.19) are, in fact, O( 2 ). Consequently, the asymptotic solutions to the system (3.17)(3.23) are obtainable by solving a sequence of systems of essentially ordinary dierential equations.
44
3.2.1
Leading order problem
At leading order our problem is given by the equations of motion 21 u1, + (e )2 u1 1

(0) B2222 u2, (0) 23 u3, (0) (0)
= 0, + (B1122 + + (B1133 +
(0) B1221 )u1,1 (0) B1331 )u1,1 3 (0) 21 pt,
(3.26) = 0, (3.27)
+ +
(0) (e )2 u2 1 (0) (e )2 u3 1
+ (B2233 + B2332 )u2,3 21 pt,3 = 0 , and incompressibility condition u1,1 + u2, = 0 , which must be solved subject to leading order boundary conditions 21 u1, = 0
(0) (0) (0) 23 u3, (0) (0) (0)
(0)
(0)
(3.28)
(3.29)
at = 1 ,
(0)
(3.30) at = 1 , (3.31) (3.32)
(B2222 + p )u2, + B1122 u1,1 21 pt +

(0) (B2332 + p )u2,3
=0
=0
at = 1 .
The solution of the boundary value problem (3.26), (3.30) may be represented in the form u1
(0)
= u1
(0,0)
cos
e 1 21
(3.33)
where the function with double superscript is independent of . Hereafter, all functions with a double superscript will be assumed independent of and therefore functions only of 1 , 3 and . Specically, the rst superscript refers to the order of the approximation, explanation (0) of the second is deferred until later. Substituting the solution for u1 from (3.33) into the (0) incompressibility condition (3.29) we obtain a dierential equation for u2 , indicating that 21 (0,0) e (0) (0,0) (0,0) (0,0) u2 = u2 sin 1 + U2 , u2 = e u1,1 . (3.34) 21 1 We now substitute the functions u1 and u2 , as dened in (3.33) and (3.34), into equation (3.27), which when solved subject to (3.31) yields 21 pt within which
(0) (0) (0)
= pt
(0,0)
cos
e 1 21
+ Pt
(0,1)
+ Pt
(0,0)
(3.35)
pt
(0,0)
= (21 b21 )u1,1 ,
(0,0)
Pt
(0,1)
= (e )2 U2 1
(0,0)
with the boundary condition (3.31) immediately revealing that U2

(0,0)
= 0,
Pt
(0,1)
= 0,
Pt
(0,0)
= G1 u1,1 cos
(0)
(0,0)
e 1 21
The leading order term of the third displacement component u3 is governed by equation (3.28), subject to the boundary condition (3.32). In view of results (3.33), (3.34) and (3.35), we seek the solution of (3.28) in the form u3
(0)
= v3
(0,0)
cos
e 1 23
+ v3
(0,1)
sin
e 1 23
+ u3
(0,0)
cos
e 1 21
+ U3
(0,0)
(3.36)
CHAPTER 3. HIGH FREQUENCY EXTENSIONAL MOTION in which u3

(0,0)
45
21 (21 + 13 ) (0,0) u , (21 23 )(e )2 1,1 3 1 v3

(0,0)
U3
(0,0)
221 2 cos (e )2 1
e 1 21
u1,1 3 ,
(0,0)
and a glance at the boundary condition (3.32) shows that it can only be satised when = 0, v3
(0,1)
= 0.
Solution (3.36) is the rst relation in which a function with non-zero second superscript occurs. The second superscript may now be dened as the power of any associated multiplier. We note that for extensional (symmetric) motion the in-plane displacements u , u and pressure p t 1 3 vary only as symmetric functions of the normal coordinate , whereas the normal displacement u is an anti-symmetric function. Our leading order solutions (3.33)(3.36) conrm this, and 2 for the sake of brevity we will henceforth ignore terms of the solutions which do not comply to these requirements. In this section we have obtained leading order solutions for all displacement components (0,0) and pressure increment in terms of the function u1 , or more specically, as linear functions (0,0) of it and its derivatives. Since knowledge of u1 completely determines the long wave high frequency motion at leading order, it is usually referred to as the leading order long wave amplitude, this being independent of . However, this function itself can not be determined without resorting to the higher order.
3.2.2
Second order problem

21 u1, + (e )2 u1 1
(2) (2) (0) (0)
The second order equations of motion are given by = B1111 u1,1 1 31 u1,3 3
(0) (0) (2) (2)
(B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (e )2 u1 1 B2222 u2, + (e )2 u2 1

(2) (2) (0)
(0)
(0)
(3.37)
+ (B1122 + B1221 )u1,1 21 pt, = 12 u2,1 1

(0) (0)
(0)
32 u2,3 3 (B2233 + B2332 )u3,3 + 2 21 u2, + (e )2 u2 1
(0)
(3.38)
where we remark that the nal terms containing large brackets in each equation are of the same order as other terms, see (3.25). We need not consider the third equation of motion (3.19) to (0,0) derive a governing equation for u1 , or its second order correction, which is the reason why it is omitted here. Equations (3.37) and (3.38) must be solved subject to the second order incompressibility condition (2) (2) (0) u1,1 + u2, = u3,3 , (3.39) and boundary conditions 21 u1, = (B1221 + p )u2,1 (B2222 + p )u2, + B1122 u1,1 21 pt
(2) (2) (2) (2) (0)
at = 1 ,
(0)
(3.40) at = 1 . (3.41)
= B2233 u3,3
The rst equation of motion (3.37), when examined in conjunction with the leading order solutions (3.33)(3.36) and subject to the boundary condition (3.40), indicates that u1
(2)
= u1
(2,0)
cos
e 1 21
+ U1
(2,0)
U1
(2,0)
G1 cos (e )2 1
e 1 21
u1,1 1 ,
(0,0)
(3.42)
within which the function u1 can not be determined without considering the third order problem. The solution (3.42) is only valid provided 21 u1, + (e )2 u1 1
(0,0) (0,0)
(2,0)
2 E1c u1,1 1 + E1s u1,3 3 = 0 .
(2) (0,0)
(2) (0,0)
(3.43)

(0) (0,0)
46
As the displacement component u2 is a linear function of u1 (and its derivatives), see (3.34), it will also satisfy equation (3.43). This fact may be used to dispose of the term of (0) (0) O( 2 ) in the second equation of motion (3.38) and determine solutions for u2 and pt , these being given by u2 21 pt where u2
(2,0) (2) (2)
e 1 21 e (2,0) = pt cos 1 21 = u2
(2,0)
sin
+ U2
(2,1)
, + Pt
(2,0)
(3.44) , (3.45)
+ 2 Pt
(2,2)
(0,0) (2,0) u1,1 21 (21 + 13 ) u1,1 3 3 , = 21 23 21 (e )3 e 1 1 pt

(2,0)
U2
(2,1)
2e1 , (e )2 1
(2)
= (21 b21 )u1,1 21 (e )2 1
21 Q1c (0,0) u (e )2 1,1 1 1 1 (21 + 13 )(21 b23 ) (0,0) + 31 32 u1,1 3 3 , 23 21

(2,0)
(2)
Pt
(2,0)
g1 + B2222 B1221 1 (2,0) 1 e 2 (e )2 G1 u1,1 1 e )2 (1 (1 ) (21 + 13 )(23 + 21 2 ) + B2233 G1 21 + 31 32 23 21 e (2) (0,0) + B2211 G1 21 Q1c u1,1 1 1 cos 1 , 21 = e1
(2)
u1,1 3 3
(0,0)
and e1 = Pt
(2) df (2,2)
G1 (0,0) (0,0) u1,1 1 1 + u1,1 3 3 cos 2

df (0,0)
e 1 21
We shall now take a closer look at the equation (3.43). Returning back to the original non-scaled variables and introducing u(0) (x1 , x3 , t) = u1 h2
(2)
(1 , 3 , ) yields
(2) 2 u(0)
2 u(0) + (e )2 u(0) h2 1 t2
E1c
(2)
2 u(0)
x2 1
+ E1s
x2 3
= 0,
(3.46)
in which E1c and E1s are the functions of material parameters and pre-stress dened immediately after expansion (3.5). It was mentioned previously that the solution of the two-dimensional equation (3.46), termed the leading order long wave amplitude, will completely determine the extensional long wave high frequency motion of the layer at leading order. Moreover, if this equation is satised, then the second order problem will also be satised. We also remark that the associated dispersion relation, derived by substituting a solution of the form (1.64) into equation (3.46), will match the rst two orders of the frequency expansion (3.5), demonstrating a high level of consistency of the model. (2) (2) It is worth noting that when the coecients E1c and E1s are positive, equation (3.46) is (2) hyperbolic. However, it is possible to nd stable pre-stressed states for which the coecient E1c (2) becomes negative, note that the strong ellipticity conditions (1.36) necessitate that E1s must
(2)
47
always be positive. In such a case equation (3.46) still remains hyperbolic, but it is not a proper wave-like equation, in the sense that time t and third spatial variable x3 swap roles. This result complements a previous observation in a plane strain case, for which a similar change of the coecient sign leads to loss of hyperbolicity, see Kaplunov et al. (2001a). This behaviour is also related to the possible existence of harmonics with negative group velocity at low wave number; a point we shall discuss in greater detail later in this chapter.
3.2.3
Third order problem
We will not solve the third order problem completely, but merely derive a rened governing equation for the long wave amplitude. Accordingly, we only consider the third order approximation of the rst equation of motion, namely 21 u1, + (e )2 u1 1
(4) (4)
= (B1133 + B1331 )u3,1 3 B1111 u1,1 1 31 u1,3 3

(2) (2) (2)
(0)
(2)
(2)
(B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (e )2 u1 1 subject to the appropriate order boundary condition 21 u1, + (B1221 + p )u2,1 = 0 .
(4) (2)
(2)
(3.47)
(3.48)
Substituting the functions on the right hand side of the equation of motion (3.47) with their solutions (3.36), (3.42), (3.44) and (3.45), which were obtained in previous sections, we observe (4) that the solution for u1 has the general form u1
(4)
= u1
(4,0)
cos
e 1 21
+ u1
(4,1)
sin
e 1 21
+ 2 U1
(4,2)
+ U1
(4,0)
(3.49)
Utilising the boundary condition (3.48), and the restriction on u1 following solutions for the coecients of the (3.49) are derived
(4,1) u1
(0,0)
imposed by (3.43), the
G1 (4) = 2 e , 21 (e )3 1 1
(4,2) U1
e = 1e 2 cos (1 )
(4)
e 1 21
U1
(4,0)
1 (2) (0,0) e 2 (4) e )4 21 (2G1 + Q1c )u1,1 1 1 1 (22 + 221 + (1 ) )e1 (1 (21 + 13 )(23 + 21 2 ) (2,0) (e )2 G1 u1,1 1 + 21 + 31 32 1 23 21 e (0,0) + G1 (23 13 ) u1,1 1 3 3 cos 1 , 21 = e1 =
(4) df
G1 (0,0) (0,0) u1,1 1 1 1 + u1,1 1 3 3 . 2 The general solution (3.49) will only satisfy the equation of motion (3.47) provided 21 u1, + (e )2 u1 1
(2,0) (2,0)
within which
2 E1c u1,1 1 + E1s u1,3 3 E1c u1,1 1 1 1 E1s u1,1 1 3 3 = 0 ,

(4) (0,0) (4) (0,0)
(2) (2,0)
(2) (2,0)
(3.50)
this may be used together with equation (3.43) to rene the leading order long wave amplitude u(0) . To do this we multiply equation (3.50) by 2 and add the result to equation (3.43), which upon the introduction of the non-scaled and scaled long wave amplitude expansions u = u(0) + u(2) 2 + O( 4 ) , u = u(0) + u(2) 2 + O( 4 ) , (3.51)
CHAPTER 3. HIGH FREQUENCY EXTENSIONAL MOTION gives the following governing equation 21 u + (e )2 u 2 E1c u 1 1 + E1s u 3 3 , 1 , , + 4 E1c u 1 1 1 1 + E1s u 1 1 3 3 = 0 , , ,
(4) (4) (2) (2)
48
(3.52)
where notation introduced immediately after (3.5) has been utilised. This equation can be expressed in terms of the original (non-scaled) variables, giving h2
2 2 2u (2) u (2) u + (e )2 u h2 E1c + E1s 1 t2 x2 x2 1 3
+ h4 E1c
(4)
4u
x4 1
+ E1s
(4)
4u x2 x2 1 3
= 0.
(3.53)
When equation (3.53) is satised, all the equations of the third order problem will also be satised. It is easy to see from our solutions, that each displacement component and the pressure (0,0) (2,0) increment are represented as linear functions of u1 and u1 and their derivatives. More(2m,0) , m = 0, 1, 2, . . . will all satisfy each of the boundary value problems over, the functions u1 (0,0) (2m2,0) posed for u1 , . . . , u1 , indicating that the displacement components and pressure increment are linear functions of the long wave amplitude and its derivatives. We may, as we did it for the second order governing equation for the long wave amplitude (3.46), substitute the solution of the form (1.64) into the (3.53) and derive the corresponding dispersion relation. The rened result will match our expansion for the scaled frequency expansion (3.5) exactly.
3.3
Asymptotically approximate equations: second family of shear resonance frequencies
It follows from the approximations (3.7), that for motions in the vicinity of second family of cut-o frequencies a typical wave travels the distance of a typical wave length l in a time t l/e , so the suitable rescaling of time is 3 t = l . 23 (3.54)
It must be remarked that the chosen scales of time (3.15) and (3.54) only dier by an O(1) factor and therefore are asymptotically equivalent. The particular choice of scalings was made merely to reduce the algebraical complexity of further derivation. Motivated by the estimates (3.12) we rescale the displacements and pressure so that um (x1 , x2 , x3 , t) = l 3m u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = 23 p (1 , , 3 , ) . t m {1, 2, 3} , (3.55)
When appropriately rescaled, the equations of motions (1.18)(1.20) are given by 21 u + (e )2 u (e )2 u + 23 u + (B1122 + B1221 )u 1 3 1 3 1 1, 1, 2,
+ (B1133 + B1331 )u 1 3 23 p 1 + 2 B1111 u 1 1 + 31 u 3 3 = 0 , 3, t, 1, 1,
(3.56)
B2222 u + (e )2 u (e )2 u + 23 u + (B2233 + B2332 )u 3 2, 3 2 3 2 2, 3,
23 p + 2 (B1122 + B1221 )u 1 + 12 u 1 1 + 32 u 3 3 = 0 , t, 1, 2, 2,
(3.57)
23 u + (e )2 u (e )2 u + 23 u + 2 (B2233 + B2332 )u 3 3 3 3 3 3, 3, 2,
+ 13 u 1 1 + B3333 u 3 3 21 p 3 + 4 B1133 + B1331 u 1 3 = 0 , 3, 3, t, 1,
(3.58)
CHAPTER 3. HIGH FREQUENCY EXTENSIONAL MOTION these must be solved together with the incompressibility condition u + u 3 + 2 u 1 = 0 , 2, 3, 1, subject to the boundary conditions 21 u + (B1221 + p )u 1 = 0 2, 1, 23 u 3, + (B2332 +
2
49
(3.59)
(B2222 + p )u + B2233 u 3 21 p + 2 B1122 u 1 = 0 2, t 3, 1, p )u 3 2, =0 As we did in the previous case, we may utilise the fact that 23 u + (e )2 u 2 u , m, 3 m m
at = 1 ,
(3.60) at = 1 , (3.61) (3.62)
at = 1 .
m {1, 2, 3} ,
(3.63)
to establish the order of terms in the braces in the equations of motion (3.56)(3.58).
3.3.1
For the leading order problem we are picking up the leading order terms of the equations (3.56) (3.62), yielding the equations of motion 21 u1, + (e )2 u1 3 B2222 u2, + (e )2 u2 3 23 u3, + (e )2 u3 3
(0) (0) (0) (0)
+ (B1122 + B1221 )u2,1 + (B1133 + B1331 )u3,1 3 23 pt,1 = 0 , + (B2233 + B2332 )u3,3 23 pt, = 0 , = 0,
(0) (0) (0) (0)
(0)
(3.64) (3.65) (3.66)
(0) (0)
and the leading order incompressibility condition u2, + u3,3 = 0 .

(0) (0)
(3.67)
Equations (3.64)(3.67) must be solved in conjunction with the leading order boundary conditions, given by 21 u1, + (B1221 + p )u2,1 = 0 (B2222 + p )u2, + B2233 u3,3 23 pt 23 u3, = 0 e 3 23 e 3 23
(0) (0) (0) (0) (0) (0)
at = 1 , =0 at = 1 ,
(3.68) (3.69) (3.70)
at = 1 .
(0) (0,0)
The solution for the system (3.64)(3.70) is readily established in the following form u1
(0)
= u1 =
(0,0)
cos cos
+ U1 ,
(0,0)
u2
= u2
sin
(0) u3
(0,0) u3
(0) 21 pt
(0,0) pt cos
e , 3 23 e (0,0) + Pt , 3 23
(3.71)
in which u1 u2 pt
(0,0) (0,0) (0,0)
23 (23 + 13 ) (0,0) G3 e (0,0) (0,0) u3,1 3 , U1 = e 2 cos 3 u3,1 3 , e )2 (23 21 )(3 (3 ) 23 e 23 (0,0) (0,0) (0,0) = e u3,3 , Pt = G3 u3,3 cos 3 , 3 23 = = (23 b23 )u3,3 .
(0,0) (0,0)
To determine the function u3 we have to resort to higher order. For the considered case, (0,0) u3 plays role of the leading order long wave amplitude, so every function in (3.71) is the (0,0) linear function of u3 and its derivatives.
50
3.3.2
In order to derive the leading order governing equation for the long wave amplitude, and its correction at third order, we need only consider the appropriate second and third equations of motions B2222 u2, + (e )2 u2 3
(2) (2)
32 u2,3 3 (B1122 + B1221 )u1,1 + 2 23 u2, + (e )2 u2 3 23 u3, + (e )2 u3 3

(2) (2)
(0)
(B2233 + B2332 )u3,3 23 pt, = 12 u2,1 1

(0) (0)
(2)
(2)
(0)
(0)
(3.72)
= 13 u3,1 1 B3333 u3,3 3

(0) (0)
(0)
(0)
(B2233 + B2332 )u2,3 + 23 pt,3 + 2 23 u3, + (e )2 u3 3 and incompressibility condition u2, + u3,3 = u1,1 ,
(2) (2) (0)
(0)
(0)
(3.73) (3.74)
which are to be considered in conjunction with the second and third second order boundary conditions (B2222 + p )u2, + B2233 u3,3 23 pt
(2) (2) (2) (2) (0)
= B1122 u1,1
(0)
at = 1 ,
(3.75) (3.76)
23 u3, = (B2332 + p )u2,3 The solution for the system (3.72)(3.76) is given by u2 u3 23 pt
(2) (2) (2)
at = 1 .
(2,1)
e 3 23 e (2,0) = u3 cos 3 23 e (2,0) = pt cos 3 23 = u2

(2,0)
sin
+ U2 + U3
, (3.77) + Pt
(2,0)
(2,0)
, ,
+ 2 Pt
(2,2)
this solution only being valid provided 23 u3, + (e )2 u3 3

(0,0) (0,0)
2 E3c u3,1 1 + E3s u3,3 3 = 0 .
(2) (0,0)
(2) (0,0)
(3.78)
Substituting the solutions (3.77) into the equations (3.72)(3.76) the following representations can be derived (0,0) (2,0) u3,1 1 3 u3,3 23 (23 + 13 ) (2,0) , u2 = 23 23 21 (e )3 e 3 3 U2
(2,1)
2e3 , (e )2 3
(2)
U3
(2,0)
G3 cos (e )2 3
e 3 23
u3,3 3 ,
(0,0)
(2,0) pt
23 Q3s (0,0) = (23 e )2 u3,3 3 3 (3 23 (23 + 13 )(23 b21 ) (0,0) e 2 + 12 13 u3,1 1 3 , (3 ) 23 21

(2,0) b23 )u3,3
(2)
Pt
(2,0)
g3 + B2222 B2332 1 (2,0) 1 e 2 (e )2 G3 u3,3 3 (e )2 (3 ) 1 (23 + 13 )(23 + 21 2 ) + B1122 G3 + 23 + 12 13 23 21 e (2) (0,0) + B2233 G3 23 Q3s u3,3 3 3 cos 3 , 23 = e3
(2)
u3,1 1 3
(0,0)
CHAPTER 3. HIGH FREQUENCY EXTENSIONAL MOTION and also e3 = Pt

(2) df (2,2)
51
G3 (0,0) (0,0) u3,1 1 3 + u3,3 3 3 cos 2
e 3 23
.
(0)
This derivation again utilises the fact that every displacement component um (m {1, 3}) is a (0) linear function of the leading order long wave amplitude u3 and its derivatives, which guarantees that equation (3.78) is also valid for each of the displacement components. Rescaling (3.78) back to the original variables we obtain the governing equation for the leading order long wave amplitude 2 (0) 2 (0) 2 u(0) (2) u (2) u + (e )2 u(0) h2 E3c + E3s = 0. (3.79) h2 3 t2 x2 x2 1 3 It is again remarked that the dispersion relation associated with this equation matches the rst two orders of the expansion (3.9).
3.3.3
Third order problem
At third order the third equation of motion is given by 23 u3, + (e )2 u3 3

(4) (4)
13 u3,1 1 B3333 u3,3 3 + 23 pt,3 + 2 23 u3, + (e )2 u3 3 with the associated boundary condition (B2332 + p )u2,3 + 23 u3, = 0 .
(2) (4)
(2)
= (B1133 + B1331 )u1,1 3 (B2233 + B2332 )u2,3

(2) (2) (2)
(0)
(2)
(2)
(3.80)
(3.81)
The problem posed by (3.80) and (3.81) is all that is needed at this order to derive a corrected governing equation for the long wave amplitude. Inserting solutions derived at the previous orders into the right hand side of the equation (3.80) we observe the following form of general solution u3
(4)
= u3
(4,0)
cos
e 3 23
+ u3
(4,1)
sin
e 3 23
+ 2 U3
(4,2)
+ U3
(4,0)
(3.82)
which together with the boundary condition (3.81) yields u3

(4,1)
= 2
G3 (4) e )3 e3 , 23 (3
U3
(4,2)
e3 cos (e )2 3
(4)
e 3 23
U3
(4,0)
1 (2) (0,0) (4) 23 (2G1 + Q3s )u3,3 3 3 3 (22 + 223 + (e )2 )e3 3 (e )4 3 (23 + 13 )(23 + 21 2 ) (2,0) (e )2 G3 u3,3 3 23 + 12 13 3 23 21 e (0,0) G3 (21 13 ) u3,1 1 3 3 cos 3 , 23 = e3 =
(4) df
G3 (0,0) (0,0) u3,1 1 3 3 + u3,3 3 3 3 2 Equation (3.80) may only be satised provided 23 u3, + (e )2 u3 3
(2,0) (2,0)
where
2 E3c u3,1 1 + E3s u3,3 3 E3c u3,1 1 3 3 E3s u3,3 3 3 3 = 0 ,

(4) (0,0) (4) (0,0)
(2) (2,0)
(2) (2,0)
(3.83)
52
which upon multiplying by 2 , adding to (3.78) and rescaling to the original variables gives the governing equation for the long wave amplitude h2
2 2 2u (2) u (2) u + (e )2 u h2 E3c + E3s 3 t2 x2 x2 1 3
+ h4 E3c
(4)
4 4u (4) u + E3s x2 x2 x4 1 3 3
= 0,
(3.84)
in which u is dened by (3.51). The derivation of the leading and second order governing equations for the second family of shear resonance frequencies is very similar to the analogous procedure for the rst family. In the high frequency analysis in some of the future chapters we will only provide detailed derivations for the rst family of shear resonances, mentioning only the principal results for the second.
3.4
Illustrative numerical results
The numerical results presented in this section demonstrate some particular features of the derived asymptotic models. The left hand plot on Figure 3.2 shows the scaled group velocity
0.05 First family Second family
15
0.04
10
0.03
vg
0.02
2
0
0.01
-5
0
-10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Extensional Flexural 2=6.0 0 2 4 6 8 10
(a) Group velocity,
kh
(b) Neutral curves,
kh
Figure 3.2: (a) Scaled group velocity vg against scaled wave number kh shown for the rst six harmonics for extensional motion with an angle of propagation = 15 . (b) The associated neutral curves of Cauchy stress 2 against kh. The layer is composed of VM, see Table 1.1 (p. 23), and nitely deformed with 1 = 0.7, 2 = 1.5 and 2 = 6.0.
vg of the rst six harmonics (three for each of the two families of shear resonance frequencies) against the scaled wave number kh for propagation along = 15 for a Varga material, vg being dened by df vg = , (3.85) (kh) in which is the scaled frequency. An interesting feature of this graph is the strongly linear behaviour of the group velocity associated with each harmonic within the low wave number region, especially noticeable for those associated with the second family of shear resonance frequencies. To analyse this we consider, as an example, the frequency expansion (3.5), which may be represented as 2 = (e )2 + E1 2 + O( 4 ) , 1 Using (3.85) we are able to conclude that E vg = 1 e + O( 3 ) , 1
(2) (2)
E1
(2)
= E1s s2 + E1c c2 .
(2)
(2)
(3.86)
(3.87)
indicating linear variation with an O( 3 ) correction in the long wave high frequency regime.

9.06
53
Numerical Second order Third order
17.978
26.959
9.04
17.976
26.958
9.02
17.974
26.957
17.972
17.97
8.98
26.956
17.968
0 0.3 0.6 0.9
Numerical Second order Third order 0 0.3 0.6 0.9
0.3
0.6
0.9
kh
8.62 Numerical Second order Third order 17.08 Numerical Second order Third order
kh
25.57
kh
25.56 8.58 17.06 25.55 8.54 17.04 25.54
8.5 0 0.3 0.6 0.9
17.02 0 0.3 0.6 0.9
25.53 0 0.3 0.6 0.9
kh
kh
kh
Figure 3.3: Numerical and asymptotic solutions for the scaled frequency of the rst three extensional harmonics, associated with the rst (upper row) and second (lower row) families of shear resonance frequencies and the angle of propagation = 15 . The material of the layer is VM subjected to the primary deformation with 1 = 0.7, 2 = 1.5 and 2 = 6.0.
A further interesting feature of the left hand plot in gure 3.2 is the possible existence of negative group velocity at low wave number. It must be remarked that this behaviour is not associated with any instability, which is demonstrated by the neutral curves on the right hand plot. For the chosen parameters, this is specically associated with the rst family of shear resonance frequencies and is evident in Figure 3.2 for the rst three harmonics. Plots depicting the scaled frequencies of these three harmonics, and showing numerical and asymptotic solutions, are presented in Figure 3.3 (upper row). Negative group velocity corresponds to the decreasing of frequency as the wave number increases and indicates that certain modes may have frequencies (2) below the cut-o. In order for this phenomenon to manifest itself it is necessary that E1 < 0, see expansion (3.86). When this occurs it is clear that the third order correction term must be used to adequately approximate the numerical solution in all but the very lowest wave number region. We note that in general the asymptotic frequency approximations improve as n, the harmonic number, increases, the third order term being O(n2 ). The harmonics associated with the second family of shear resonance frequencies have no negative group velocity. For these the second order approximation is almost indistinguishable from the third order, as illustrated in the lower row in Figure 3.3. It is worth remarking that negative group velocity is impossible in a Mooney-Rivlin material. Indeed, we demonstrated in the Chapter 1 that every Mooney-Rivlin material satises the conditions (1.72)1 , which imply that E1c 212 21 = 12 > 0 ,
(2)
E3s 223 23 = 32 > 0 ,
(2)
(3.88)
with the nal inequalities following from the strong ellipticity conditions (1.36). Thus generally
54
speaking, every material that satises (1.72)1 can not have negative group velocity. We must however note that the existence of negative group velocity is by no means a specic feature of Varga materials, which is demonstrated by the group velocity plot for an Ogden material in Figure 3.4.
0.04 First family Second family
10
0.03
vg
0.02
2
0
0.01
0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
-5 0 2 4 6
Extensional Flexural 2=6.0 8 10
kh
kh
Figure 3.4: (a) Scaled group velocity vg against scaled wave number kh shown for the rst six harmonics for extensional motion with an angle of propagation = 15 . (b) The associated neutral curves of Cauchy stress 2 against kh. The layer is composed of OM1, see Table 1.1 (p. 23), and nitely deformed with 1 = 0.6, 2 = 1.4 and 2 = 6.0.
As previously mentioned, the existence of negative group velocity is connected with the (2) (2) change of sign of E1c or E3s in the governing equation (3.46) or (3.79), respectively. When this occurs there is a signicant change in the character of the appropriate governing equation. In particular, time and one of the in-plane spatial variables swap roles. In fact, it was shown by Kaplunov et al. (2001a) that in the plain strain case a similar change of the coecient sign changes the analogous equation from hyperbolic to elliptic. In the plain strain case this change of sign is always synonymous with the existence of negative group velocity. However, in the present case it is quite possible for time and one of the in-plane spatial variables to swap roles and the group velocity remain positive, see gure 3.5. In this gure the left hand graph shows
5 4 3 2
2=2.0 2=1.5 2=1.0 2=0.5
E1
(2)
1 0
E1c
=15 =30 =45 =60 =75 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(2)
0
-1 -2 -3
-2 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(a)
1
(2)
(b)
Figure 3.5: Dependence of the second order frequency expansion term E1 and the governing equation (2) coecient E1c on the in-plane deformation, shown for (a) various angles of propagation with xed 2 = 2.0 and (b) various principal normal deformations 2 . The plots depict the motions associated with the rst family of shear resonance frequencies in OM1.
The alternative secular equation factorisation conditions (1.76)1 or (1.78)1 will preclude the existence of negative group velocity in the vicinity of second or rst family of shear resonance frequencies, respectively.

4 3 2 8 6 4 2=2.0 2=1.5 2=1.0 2=0.5
55
E3
(2) 1
0 -1 -2 -3 0.6 0.8 1 1.2 1.4 1.6 =15 =30 =45 =60 =75 1.8 2
E3s
(2) 2
0 -2 -4 -6 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(a)
1
(2)
(b)
Figure 3.6: Dependence of the second order frequency expansion term E3 and the governing equation (2) coecient E3s on the in-plane deformation, shown for (a) various angles of propagation with xed 2 = 2.0 and (b) various principal normal deformations 2 . The plots depict the motions associated with the second family of shear resonance frequencies in OM1.
the dependence of the second order term of the frequency expansion on in-plane deformation, with xed normal deformation. The sign of this term indicates the sign of the group velocity, see equation (3.87). We note that this phenomenon is most likely to occur for propagation along a (2) principal direction. The right hand graph shows the behaviour of E1c , the sign of which triggers (2) the role change of the spatial variable and time previously mentioned. Since E1c is independent (2) of , and E1s always positive, a value of may always be chosen such that vg > 0. Similar comments may also be made for the coecients associated with the second family of shear resonance frequencies, corresponding numerical illustrations being shown in gure 3.6. For example, an initial pre-stress arising from the principal stretches 1 = 0.6, 2 = 2.0 renders the (2) coecient E3s negative, see the right hand graph, the governing equation not being a typical hyperbolic equation. However, the left hand graph shows that for the angles of propagation 45 the group velocity of the harmonics still remains positive.
Chapter 4
Layer with free faces: Long wave high frequency exural motion
Having completed the derivation of an asymptotically consistent model for long wave high frequency extensional motion, we turn our attention to a complementary model for long wave high frequency exural (anti-symmetric) motion in a pre-stressed incompressible elastic plate. An asymptotic analysis of the associated dispersion relation once again indicates the presence of two distinct asymptotic cases, corresponding to two families of shear resonance frequencies. Third order approximations for scaled frequency are obtained and used to determine relative orders of displacements. All information necessary for the derivation of asymptotically approximate equations is then available and we introduce appropriate scalings of space and time variables to recast the equations of motion, incompressibility and boundary conditions in non-dimensional form. A subsequent analysis delivers power series representations for all displacement components and incremental pressure in terms of a single function, termed the long wave amplitude. The leading and second order governing equations for the long wave amplitude are obtained during this derivation as necessary and sucient existence conditions. Consistency of the developed theory is again demonstrated by comparing the exact dispersion relation of approximate governing equation with an appropriate approximation of exact dispersion relation. We conclude this chapter by presenting some illustrative numerical examples.
4.1
To begin asymptotic analysis of the dispersion relation (2.49) we re-introduce a small parameter , the ratio of plate half-thickness h and typical wave length l, hence = h/l = kh. Recalling that in long wave high frequency regime v as 0, we insert expansions (2.67)(2.69) and (2.71)(2.73) into the dispersion relation (2.49) to obtain i A1 v 2 + A1
(2) (0)
T1 (h) + v 3 A2 v 2 + A2
(2)
(5)
(3)
T2 (h) + i A3 v 2 + A3
(2)
(2)
(0)
T3 (h) 0 ,
(4.1)
in which the leading order coecients A1 , A2 and A3 are given by A1 =

(2)
(5)
G2 (23 21 )c2 1 , 21
A2 = (23 21 ) ,
(5)
A3 =
(2)
G2 (23 21 )s2 3 , 23
56
CHAPTER 4. HIGH FREQUENCY FLEXURAL MOTION and the second order coecients of the dispersion relation (4.1) have the forms G1 c2 (0) A1 = 2 21 G1 21 (221 612 + 31 ) 23 (423 423 + 421 + 531 )
57
421 (23 21 )(32 31 21 13 ) s2 2G1 (23 21 )c5 + G1 223 (23 + 23 2(12 13 )) (23 21 )(212 + 613 21 ) + 421 (23 21 )Q1c A2 =
(3) (2)
c2 ,
223 (23 + 23 2(12 13 )) (23 21 )(612 + 21 + 713 ) c2 , G3 s2 (0) A3 = 2 23 G3 23 (623 13 223 ) + 21 (423 421 + 412 + 513 )
1 (23 21 )(c4 + c5 42 ) 2 + 221 (12 + 21 2(23 31 )) + (23 21 )(623 + 23 + 731 ) s2
423 (23 21 )(12 13 13 23 ) c2 2G3 (23 21 )c5 G3 221 (12 + 21 2(23 31 )) + (23 21 )(223 + 631 23 ) 423 (23 21 )Q3s
(2)
s2 ,
with parameters gm and Gm , m = {1, 3}, dened in (2.26). Hereafter, we presume that all (2) (0) (5) (3) (2) (0) of A1 , A1 , A2 , A2 , A3 and A3 are generally of O(1). Since q2 is O(1), see (2.72), T2 (h) = O(1 ) and consequently the asymptotic equality (4.1) implies T1 (h) = O(2 ) or, v v alternatively, T3 (h) = O(2 ), which is an indication of the existence of two distinct asymptotic v cases associated with two families of shear resonance frequencies.
4.1.1
Approximation near shear resonance frequencies
Suppose T1 (h) = O(2 ), hence the argument of this hyperbolic tangent must be imaginary and v to leading order equal to i( 1 + n). Thus, we expand the argument in a power series in small 2 as follows: 1 f (2) f (4) kq1 h = i + n + i1 2 + i1 4 + O( 6 ) , (4.2) 2 in which the O(1) parameters 1 for T1 (h) is given by
f (2)
and 1 i
f (4)
are to be determined. The associated expansion + i1

f (4)
T1 (h) =
f (2) 1 2
f (2) (1 )2
+ O( 2 ) .
(4.3)
At this point it is convenient to introduce the parameter f = 21 ( 1 + n), n = 1, 2, 3, . . . , 1 2 the physical interpretation of which is deferred until later. We may now utilise (2.71) and (2.74) to obtain if f (2) q1 = 1 + i1 + O( 3 ) , 21 v= f 1 + 21 1
f (2)
(4.4)
(0)
Q1s s2 + Q1c c2 2f 1
(0)
+ O( 3 ) .
(4.5)
CHAPTER 4. HIGH FREQUENCY FLEXURAL MOTION
58
These may now be inserted into (2.72)(2.73), which allows us to approximate corresponding hyperbolic tangents, thus q2 = 1 + T2 (h) = + Q2s s2 + Q2c c2 c5 2(f )2 1
(2) (2) (2) (2)
2 + O( 4 ) , 1 3 3 + O( 5 ) , f 1 23 + O( 2 ) .
(4.6) (4.7) (4.8)
Q2s s2 + Q2c c2 c5 2(f )2 1
if q3 = 1 + O() , 23
T3 (h) = i tan
Substituting expansions (4.2)(4.8) back into the second order approximation of the dispersion f (2) f (4) relation (4.1), we obtain expressions for 1 and 1 in the form 1
f (2)
G2 c2 1 21 (f )3 1 A1
(2)
(4.9)
f (4) 1
f (2) 1
21 (f )2 A2 1 A1
(0) (2)
(5) f (2) (1 )2
Q1s s2 + Q1c c2 + A1 (f )2 A1 1
(0) (0) (2)
(0)
(0)
(0)
+ f 1 +2
1 2 21 A1 (f )2 1
6Q1s + Q3s 32 s2 + 6Q1c + Q3c 12 c2 c5 A2

(2)
(0)
(5)
A3
(2)
f 1
tan
f 1 23
A2 (3) 1 (f )2 + A2 1 (2) 3 A1
(5)
f (2)
(4.10)
It is now possible to employ (2.74) to obtain the appropriate long wave high frequency approximation of the dispersion relation, in the form of scaled frequency v as a function of for f each n (note that 1 is a function of n, n = 1, 2, 3, . . . ), given by 2 = (f )2 + F1c c2 + F1s s2 2 F1c c2 + F1s s2 c2 4 + O( 6 ) , 1 F1c =
(4) (2) (2) (2) (4) (4)
(4.11)
2G2 1
(f )2 1 G2 1
+ Q1c , 5G2 1
(0)
F1s = Q1s , Q1c

(2)
(2)
(0)
F1c =
(f )4 1
2 4g1 (f )2 f 2 3 1 (1 )
(f )4 1
22 G1 21 (f )2 , 1
2 2 G 21 Q1s 2G2 G2 f (4) 3 F1s = 1 f tan 1 3 1 23 23 (1 )5 (f )2 1 2G1 f G1 (2g3 + 23 31 + 32 ) + 221 D1 f 4 (1 )
(2)
in which
f D1 =
The analogous analysis, applied to the case T3 (h) = O(2 ), delivers another set of frequency v approximations, which may be written as 2 = (f )2 + F3c c2 + F3s s2 2 F3c c2 + F3s s2 s2 4 + O( 6 ) , 3 F3c = Q3c ,
(2) (0) (2) (2) (4) (4)
(21 + 23 2 )(13 + 21 ) + 31 32 . 23 21
(4.12)
F3s =
(2)
2G2 3
(f )2 3
+ Q3s ,
(0)

(4)
59 22 G3 23 (f )2 , 3
F3s =
G2 3 (f )4 3
2 4g3 (f )2 f 2 3 3 (3 )
5G2 3
Q3s
(2)
(f )4 3
2 2 G 23 Q3c 2G2 G2 f (4) 3 3 3 F3c = 1 f tan 3 21 21 (3 )5 (f )2 3 2G3 f f G3 (2g1 + 21 13 + 12 ) 223 D3 4 (3 )
(2)
where f = 3
23 ( 1 + n), n = 1, 2, 3, . . . , and 2
f D3 =
(21 + 23 2 )(13 + 23 ) + 12 13 . 23 21
4.1.2
Comparison of the relative orders of particle displacements not only gives us a clear physical picture of the structure of this type of motion, but also provides the basis for building a lowerdimensional asymptotically consistent model. Relations (2.54) may be used to represent the displacement components and pressure increment (2.50)(2.53) in terms of a single parameter U+ . Thus, to determine the relative asymptotic orders of displacement we only need to establish asymptotic representations for all functions occurring in (2.50)(2.54). First note that it is the consequence of (2.71)(2.73) that S2 (x2 ) = x2 + O( 3 ) , h C2 (x2 ) = 1 + O( 2 ) , (4.13)
in which we assume that x2 /h is O(1). Additionally, our expansions for the rst case of asymptotic balance of the dispersion relation (T1 (h) = O(2 )) also imply v S1 (h) = i(1)n 1
f (4) 4
+ O( 6 ) ,
f x2 Sm (x2 ) = i sin 1 2m h + O( 2 ) ,
+ O( 2 ) ,
f x2 Cm (x2 ) = i cos 1 2m h
m {1, 3} ,
which together with the approximations (4.4)(4.8) yields u1 O(pt ) , u2 O(pt ) , u3 2 O(pt ) . (4.14)
Repeating the procedure for the second case (T3 (h) = O(2 )) one may obtain the following v distribution of relative orders of displacements u1 2 O(pt ) , u2 O(pt ) , u3 O(pt ) . (4.15)
One of the in-plane displacements is therefore the leading order displacement, similarly to the results for long wave high frequency extensional motion obtained in Chapter 3.
4.1.3
Physical interpretation
The asymptotic expansions derived in previous sections show that in both cases (T1 (h) = O(2 ) v 2 )) to leading order the wave normal is given by a non-normalised vector of the or T3 (h) = O( v form (c , O( 1 ), s ), see (4.4). The second component of this vector is large, so the leading order direction of wave propagation is normal to plate. The polarisation directions for each case are given by (non-normalised) vectors of the form (O(1), O(), O( 2 )) and (O( 2 ), O(), O(1)) respectively, which is suggested by the relative orders of the displacements (4.14) and (4.15). As
60
12 12 12 12 12 12 12 12
Figure 4.1: Anti-symmetric shear resonance of an innitesimally thin transverse bre of the layer.
kh 0, to leading order waves travel along the normal direction and are polarised along one of the in-plane axes of the primary deformation. These waves are obviously two shear waves, which reminds us of the discussion on physical signicance of long wave high frequency extensional motion in section 3.1.3 of this thesis. The parameters f = 2m ( 1 + n), m {1, 3}, n = m 2 1, 2, 3, . . . , are the leading orders of the scaled frequency expansions (4.11) and (4.12) respectively (the cut-o frequencies). These are in fact natural thickness shear resonance frequencies of an innitesimally thin transverse bre of the layer, see Figure 4.1, which satisfy one of the eigenvalue problems 2m um,22 + 2 um = 0 , um,2 |x2 =h = 0 , m {1, 3} . (4.16)
It is worth remarking that although the boundary value problem (4.16) appears to be equivalent to (3.13), they are substantially dierent, because in the current context um , m {1, 3}, is assumed to be anti-symmetric function of , whereas um of Chapter 3 was assumed to be symmetric.
4.2
Asymptotically approximate equations: rst family of shear resonances
The information obtained in the previous section may be used to build a lower-dimensional model for long wave high frequency motion. In order to set up the necessary perturbation scheme we rst need to introduce appropriate scales of space and time. Recalling that l denotes a typical wavelength, and keeping in mind that = h/l, we may choose the following spatial scalings x1 = l1 , x2 = h = l , x3 = l3 , (4.17) where 1 , and 3 are new non-dimensional spatial variables. Let us now focus on the rst family of the shear resonance frequencies ( = f ). As expansion (4.5) shows, a typical (long) 1 wave propagates with scaled speed f / and therefore travels the distance of one wave length 1 in time l /f . Hence, it is appropriate to rescale time as 1 t = l . 21 (4.18)
According to the distribution of the relative orders (4.14), the displacement components and pressure increment must have the following asymptotic structure um (x1 , x2 , x3 , t) = l m1 u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = m {1, 2, 3} , (4.19)
21 p (1 , , 3 , ) , t
in which denotes non-dimensional quantities of a same asymptotic order and 21 is introduced purely for algebraic convenience.
61
The system of equations of motion (1.18)(1.20) can now be recast in terms of the nondimensional variables, yielding 21 u + (f )2 u (f )2 u + 21 u + 2 B1111 u 1 1 + 31 u 3 3 1 1 1, 1, 1, 1, 1 1 B2222 u + (f )2 u (f )2 u + 21 u 2, 2 2 2, 1 1 2 21 pt, + 12 u2,1 1 + 32 u2,3 3 + 23 u + (f )2 u (f )2 u + 21 u 3 3 3, 3, 1 1 2 + (B2233 + B2332 )u2,3 + 13 u3,1 1 + (B1122 + B1221 )u 1 21 p 1 + 4 B1133 + B1331 u 1 3 = 0 , 2, t, 3, + (B1122 + B1221 )u 1 1, (B2233 + B2332 )u 3 = 0 , 3, + B3333 u 3 3 = 0 , 3, + (B1133 + B1331 )u 1 3 21 p 3 1, t,
(4.20) (4.21) (4.22)
here comma subscripts denote dierentiation with respect to the indicated scaled (space or time) variable. These equations must be solved in conjunction with the appropriately rescaled incompressibility condition u 1 + u + 2 u 3 = 0 , (4.23) 1, 2, 3, and subject to the zero surface traction boundary conditions 21 u + 2 (B1221 + p )u 1 = 0 2, 1, B1122 u 1 1, + (B2222 + 23 u + 3, p )u 2, 21 p + t (B2332 + p )u 3 2,
2
B2233 u 3 3,
at = 1 , =0
(4.24) (4.25) (4.26)
=0
at = 1 .
at = 1 ,
A glance at the boundary value problem (4.20)(4.26) exposes that in order to ensure response compatible with (4.16), and in view of the approximation (4.11), we require 21 u + (f )2 u 2 u , m, m m 1 m {1, 2, 3} , (4.27)
which can also be veried by direct substitution of the travelling wave solution (1.64). Note, that one of the implications of imposing (4.27) is that all values in braces in system (4.20)(4.22) must be considered as O( 2 ). We now seek the solutions in a form of the power series expansions
m
(u , u , u , p ) 1 2 3 t
=
n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(4.28)
Inserting the solutions (4.28) into equations (4.20)(4.26), we obtain hierarchical systems of essentially ordinary dierential equations at various orders.
4.2.1
The leading order equations of motion 21 u1, + (f )2 u1 1 B2222 u2, + (f )2 u2 1 23 u3, + (f )2 u3 1

(0) (0) (0) (0) (0) (0)
= 0, + (B1122 + B1221 )u1,1 21 pt, = 0 , + (B1133 + B1331 )u1,1 3 + (B2233 + B2332 )u2,3 21 pt,3 = 0 ,
(0) (0) (0) (0) (0)
(4.29) (4.30)
(4.31)
must be solved in conjunction with the leading order incompressibility condition u1,1 + u2, = 0 ,
(0) (0)
(4.32)
CHAPTER 4. HIGH FREQUENCY FLEXURAL MOTION subject to the leading order boundary conditions 21 u1, = 0
(0) (0) (0) (0)
62
at = 1 ,
(0)
(4.33) at = 1 , (4.34) (4.35)
(B2222 + p )u2, + B1122 u1,1 21 pt

(0)
=0
23 u3, + (B2332 + p )u2,3 = 0

(0) (0,0)
at = 1 . ,
The solution of the boundary value problem (4.29), (4.33) is given by u1 = u1 sin e 1 21 (4.36)
in which the function with double superscript does not depend on . Hereafter, double superscripts will denote functions independent of , with the rst superscript referring to the order of the approximation and the second denoting the power of any possible m multiplier. Note, that our choice of scaling ensures that the displacement components and pressure vary appropriately for exural motion, i.e. as an anti-symmetric (u , u and p ) or a symmetric (u ) function of the t 1 3 2 normal coordinate . Thus, for the sake of brevity, we will always omit terms of the solutions which do not comply to this requirement. Substituting the solution (4.36) into the incompressibility condition (4.32) we may establish (0) the form of the expression for u2 u2
(0)
= u2
(0,0)
cos
e 1 21
+ U2
(0,0)
u2
(0,0)
21 f 1
u1,1 .
(0,0)
(4.37)
Solutions (4.36) and (4.37) suggest that the solution of the second equation of motion (4.30) should be sought in the following form 21 pt
(0)
= pt pt
(0,0)
sin
e 1 21
+ Pt
(0,1)
, Pt
(0,1)
(4.38) = (f )2 U2 1
(0,0)
(0,0)
= (21 b21 )u1,1 , = G1 (f )2 1 u1,1 sin

(0,0)
(0,0)
whereas the boundary condition (4.34) yields U2

(0,0)
e 1 21
The leading order problem for u is given by (4.31) and (4.35). The result of substitution of the 3 previously established solutions (4.36), (4.37) and (4.38) into the equation (4.31) indicates that (0) u3 may be represented as u3
(0)
= u3 U3
(0,0)
sin =
(0,1)
e 1 21 G1 (f )2 1
+ v3
(0,0)
(0,0)
sin
e 1 23 .
+ U3
(0,1)
(4.39)
u1,1 3 sin
e 1 21
Finally, we utilise appropriate boundary condition (4.35) to nd u3

(0,0)
= =
21 (13 + 21 ) (23 G1 G3
(0,0) u f 2 1,1 3 21 )(1 )
, sec f 1 23 .
(0,0) v3
(0,0) u f 3 1,1 3 23 (1 )
sin
e 1 21
Thus, the leading order solutions for the displacement components u , i {1, 2, 3} and pressure i are obtained in terms of a function u(0,0) = u(0,0) ( , , ) and its derivatives. We remark pt 1 3 1 1 (0,0) that u1 alone species the long wave high frequency motion at leading order, and term it the leading order long wave amplitude. This function can not be determined without resorting to the higher order.
63
4.2.2
At second order we only consider the rst two second order equations of motion 21 u1, + (f )2 u1 1
(2) (2)
(B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (f )2 u1 1 B2222 u2, + (f )2 u2 1

(0) (2) (2) (2) (2) (0) (0)
= B1111 u1,1 1 31 u1,3 3

(0) (0)
(0)
(0)
(0)
(0)
,
(0)
(4.40)
32 u2,3 3 (B2233 + B2332 )u3,3 + 2 21 u2, + (f )2 u2 1
+ (B1122 + B1221 )u1,1 21 pt, = 12 u2,1 1

(0)
(4.41)
which must be solved in association with the second order incompressibility condition u1,1 + u2, = u3,3 , and the appropriate boundary conditions 21 u1, = (B1221 + p )u2,1 (B2222 + p )u2, + B1122 u1,1 21 pt
(2) (2) (2) (2) (0) (2) (2) (0)
(4.42)
at = 1 ,
(0)
(4.43) at = 1 . (4.44)
= B2233 u3,3
Substituting the leading order displacements and pressure, see (4.36), (4.37) and (4.38), into the rst second order equation of motion (4.40) and satisfying the corresponding boundary condition (4.43), immediately yields u1
(2)
= u1 u1
(2,0)
sin
e 1 21
+ u1
(2,1)
cos U1
e 1 21 =
+ U1 G1 (f )2 1
(2,1)
, e 1 21 .
(4.45)
(2,1)
G2 (0,0) 1 = u , f 3 1,1 1 21 (1 )
(2,1)
u1,1 1 sin
(0,0)
The solution (4.45) is valid provided 21 u1, + (f )2 u1 1

(0,0) (0,0)
2 F1c u1,1 1 + F1s u1,3 3 = 0 .

(2) (2)
(2) (0,0)
(2) (0,0)
(4.46)
The functions of material parameters and pre-stress F1c and F1s were dened previously and are given directly after the rst scaled frequency expansion (4.11). The incompressibility condition (4.42) is then considered to obtain the form of the solution (2) for u2 , which may be expressed as u2
(2)
= u2
(2,0)
cos
e 1 21
+ u2
(2,1)
sin
e 1 21 + U2
+ v2
(2,0)
cos ,
e 1 23 (4.47)
(2,2) 2
+ U2
(2,0)
where
(2,0) u2
(2,0) u1,1 f 1
21
, G1
2 21 (13 + 21 )
21 (23
(2,0) v2
(0,0) u f 3 1,1 3 3 21 )(1 )
21 G2 1 (f )5 1
u1,1 1 1 , f 1 23 ,
(0,0)
(2,1) u2
(0,0) u f 4 1,1 1 1 (1 )
G2 1
G1 G3
(0,0) u f 4 1,1 3 3 (1 ) (0,0)
sin
e 1 21 e 1 21
sec .
U2
(2,2)
2(f )2 1
u1,1 1 1 + u1,1 3 3 sin
(0,0)
We mention that in fact the leading order of every displacement component and pressure can be (0,0) expressed as a linear function of the leading order long wave amplitude u1 and its derivatives.
64
As a consequence, the equality (4.46) is also valid for u , i {1, 2, 3}, and p , hence the terms t i multiplied by 2 in the equation (4.41) can be represented without time derivatives, which (2) enables us to determine the solution for pt in the form 21 pt
(2)
= pt
(2,0)
sin
e 1 21
+ pt
(2,1)
cos
e 1 21 + Pt
+ pt
(2,0)
sin
e 1 23 . (4.48)
(2,3) 3
+ Pt
(2,1)
The functions pt , pt , pt and Pt second equation of motion (4.41), yielding pt

(2,0)
(2,0)
(2,1)
(2,0)
(2,3)
can now be obtained by inserting (4.48) into the
= (21 b21 )u1,1 pt 21 Q1c

(2)
(2,0)
21 (f )2 1
(13 + 21 )(21 b23 ) (0,0) + 31 32 u1,1 3 3 23 21 pt

(2,1)
(f )2 1 = Pt
u1,1 1 1 ,
(0,0)
G2 (21 b21 ) (0,0) 1 = u1,1 1 1 , 21 (f )3 1 e 1 21 sec e 1 21

(2)
(2,0)
G1 G3 (23 b23 ) (0,0) u1,1 3 3 sin 23 (f )3 1 =
f 1 23 .
(2,3)
G1 (0,0) (0,0) u1,1 1 1 + u1,1 3 3 sin 6
Satisfying the boundary condition (4.44) gives Pt

(2,1)
= sin G1
e 1 21 g3 + b23 (f )2 1
G1 1 6
g1 + b21 (f )2 1
1 6
+ tan
21 Q1c 23
G1 G2 3 23 (f )3 1
(f )2 1 f 1
u1,1 1 1 G1 u1,1 +
f 21 D1
(0,0)
(2,0)
(f )2 1
u1,1 3 3
(2)
(0,0)
which after resorting back to (4.41) returns the last unknown function of the u2 expressed here as U2
(2,0)
representation,
1 (f )2 1 G1
G1
g1 (f )2 1
1 6
2G2 1 (f )4 1
1 6
g1 Q1c tan .
(2)
(f )2 1
u1,1 1 1
(0,0)
+ +
g3 31 + 32 (f )2 1
(0,0)
G1 G2 3 23 (f )3 1 e 1 21
f 1 23
f 21 D1
(f )2 1
u1,1 3 3 G1 u1,1
(2,0)
sin
Let us take a closer look at equation (4.46). Its satisfaction ensures the existence of the (0,0) solution for the second order problem and its solution u1 completely determines the leading order stressed state, see (4.36), (4.37), (4.39) and (4.38). We will refer to equation (4.46) as the leading order governing equation for the long wave amplitude. It can also be rewritten in terms of the original non-scaled variables as h2 2 u1 t2
(0,0)
+ (f )2 u1 1
(0,0)
h2
F1c
(2)
2 u(0,0) 1 x2 1
+ F1s
(2)
2 u(0,0) 1 x2 3
= 0.
(4.49)
within which u1 (x1 , x3 , t) u1 (1 , 3 , ). The solution (1.64) when substituted into (4.49) yields a dispersion relation which matches the expansion (4.11), thus demonstrating a high level of consistency.
(0,0)
(0,0)

(2) (2)
65
When F1c and F1s are positive, the leading order governing equation for the long wave amplitude (4.49) is hyperbolic. Nevertheless, it is possible (and is demonstrated later numerically) (2) to choose such combinations of the material and pre-stress parameters that F1c will become (2) negative. Technically, the governing equation (4.49) will remain hyperbolic for F1c < 0, however it will certainly be non-wave-like, with time and one of the in-plane spatial variables swapping their roles. This behaviour is closely related to the phenomenon of negative group velocity. In the present case this phenomenon is a necessary, but not sucient, condition for a nonwave-like hyperbolic equation. It has previously been remarked that in the plane strain case the existence of negative group velocity is both necessary and sucient to lose hyperbolicity, see Kaplunov et al. (2000a).
4.2.3
Third order problem
Our main purpose here is the derivation of the second order governing equation therefore instead of giving a complete third order solution to the problem we will merely solve the third order problem for the rst third order equation of motion 21 u1, + (f )2 u1 1
(4) (4)
(B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (f )2 u1 1 and the associated boundary condition 21 u1, + (B1221 + p )u2,1 = 0 .
(4) (2)
= (B1133 + B1331 )u3,1 3 B1111 u1,1 1 31 u1,3 3

(2) (2) (2) (2)
(0)
(2)
(2)
(4.50)
(4.51)
In view of the results obtained at previous orders, the solution of (4.50) may be sought in the form u1
(4)
= u1
(4,0)
sin
e 1 21 + v1
+ u1 sin
(4,1)
cos
e 1 21 + U1
+ u1
(4,2) 2
sin .
e 1 21 (4.52)
(4,0)
e 1 23
(4,3) 3
+ U1
(4,1)
Substituting (4.52) into the rst equation of motion (4.50) and comparing the coecients of linearly independent terms it is possible to obtain u1
(4,2)
(0,0) u f 6 1,1 1 1 1 221 (1 ) (0,0)
G4 1
, f 1 23
(4,0) v1
23 G1 G3 (13 + 23 ) (23 21 )(f )5 1 1 G1 g1
u1,1 1 3 3 sin
e 1 21
sec
(4,1) U1
(f )2 1 G1
1 f f 2 4 6 (1 ) (1 ) 1 6
2G2 1
g1 Q1c tan ,
(2)
(f )2 1
u1,1 1 1 1 f 1 23
(0,0)
+ +
g3 13 21 (f )2 1
(0,0)
G1 G2 3 23 (f )3 1 e 1 21
f 21 D1
(f )2 1
u1,1 3 3 G1 u1,1 1
(2,0)
sin
Actually, we have not even completed the solution for second order problem leaving u3
(2)
undened.
CHAPTER 4. HIGH FREQUENCY FLEXURAL MOTION U1

(4,3)
66 e 1 21 .
G1 6(f )2 1
u1,1 1 1 1 + u1,1 1 3 3 sin
(0,0)
(0,0)
The boundary condition (4.51) is then used to establish u1

(4,1)
G1 21 (f )3 1
G1
2g1 (f )2 1
3 +
G2 1 (f )4 1
1 3
2 Q1c tan
(2)
(f )2 1
u1,1 1 1 1 f 1 23
(0,0)
+ +
G1
2g3 + 23 31 + 32 (f )2 1
(0,0)
1 3 .
G1 G2 3 23 (f )3 1
f 221 D1
(f )2 1
u1,1 1 3 3 G1 u1,1 1
(2,0)
As with the second order problem, the solution for u1 an additional condition is satised, which is 21 u1, + (f )2 u1 1
(2,0) (2,0) (2) (2,0)
(4)
given by (4.52) is only valid provided

(2) (2,0)
2 F1c u1,1 1 + F1s u1,3 3 + F1c u1,1 1 1 1 + F1s u1,1 1 3 3 = 0 .

(4) (0,0) (4) (0,0)
(4.53)
Let us introduce a new function u = u(0) + u(2) 2 + O( 4 ) , u = u(0) + u(2) 2 + O( 4 ) , (4.54)
in which u(x1 , x3 , t) = u (1 , 3 , ) and for the rst family of the shear resonance frequencies (2m,0) (2m,0) . The fact that the function u1 is essentially the solution of all we assume u(2m) = u1 (2m2,0) boundary value problems posed for u1 , m = 1, 2, 3, . . . , means that every displacement and pressure is a linear function of u and its derivatives. Therefore, we will term u as the long wave amplitude. Now if we add the leading order governing equation (4.46) to the product of 2 and (4.53) we obtain a second order governing equation, which is given in terms of original variables and long wave amplitude as h2 2u + (f )2 u 1 t2 2 2 (2) u (2) u h2 F1c + F1s x2 x2 1 3
h4 F1c
(4)
4u
x4 1
+ F1s
(4)
4u x2 x2 1 3
= 0.
(4.55)
The dispersion relation associated with this equation matches the third order approximation of the exact dispersion relation (4.11) exactly.
4.3
Asymptotically approximate equations: second family of shear resonances
The analysis of the asymptotic behaviour of a plate in the vicinity of the second family of shear resonance frequencies is very similar to the case just discussed. The space and time coordinates are best re-scaled according to (4.17) and t = l , 23 (4.56)
with displacements, whose scalings are chosen to coincide with (4.15), thus um (x1 , x2 , x3 , t) = l 3m u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = 23 p (1 , , 3 , ) . t m {1, 2, 3} , (4.57)

Numerical Second order Third order 11.34
18.77
67
18.79
4.5
4.3
11.3 4.1
18.75
3.9
11.26
18.73
3.7 0 0.3 0.6 0.9
11.22 0 0.3 0.6 0.9
18.71 0 0.3 0.6 0.9
kh
kh
kh
Figure 4.2: Numerical and asymptotic solutions for the scaled frequency of the rst three exural harmonics, associated with the rst family of shear resonance frequencies and the angle of propagation = 15 . The material of the layer is MRM subjected to a primary deformation with 1 = 0.9, 2 = 1.2 and 2 = 5.5.
The consequent derivation yields the leading and second order governing equations for a long wave amplitude, given by h
2 2 u(0,0) 3 t2
(0,0) (f )2 u3 3
(2) F3c
2 u(0,0) 3 x2 1
(2) F3s
2 u(0,0) 3 x2 3
= 0,
(4.58)
h2
2u + (f )2 u 3 t2 2 2 (2) u (2) u h2 F3c + F3s x2 x2 1 3

(4) (4)
h4 F3c
(4)
4u
x4 1
+ F3s
(4)
4u x2 x2 1 3
= 0,
(4.59)
where F3c , F3s , F3c and F3s were given immediately after the expansion (4.12) and we assume (2m,0) u(2m) = u3 in the denition (4.54). The dispersion relation associated with this equation is consistent with the exact dispersion relation (2.49) in the sense that it matches all three orders of the frequency expansion (4.12) exactly (rst two orders in case of the leading order governing equation (4.58)). As for the existence of negative group velocity, it may also occur for the second (2) family of shear resonance frequencies, with the necessary condition given by F3s < 0.
(2)
(2)
4.4
Numerical results and discussion
The similarities between two families of shear resonance frequencies are numerous and any features established for one of them may always be reproduced for another by the appropriate change of in-plane pre-stress and the angle of wave propagation, namely by swapping values of 1 and 3 and changing to /2 . We therefore will only study here the rst family of shear resonances assuming that all of the results are also applicable to the second family after appropriate change of the problem conguration. In Figure 4.2, is shown as a function of kh. Specically, the numerical solution and both the second and third order approximations are presented in respect of the rst three harmonics associated with rst family of shear resonances in a Mooney-Rivlin material. The associated neutral curves are shown in the left-hand plot in Figure 4.3. It is easy to see that the accuracy of the approximation in general increases considerably for the second (and further) harmonic. This is because the third order term in the frequency expansion is divided by (f )2 , which is 1 O(n2 ). However, the left-most plot in Figure 4.2 requires further investigation since it depicts

0.002
68
10
0.0015
Second order Third order
0.001
-5
0.0005
-10 0 2 4 6
0 0 0.1 0.2 0.3
(a) Neutral curves,
kh
(b) Approximation error,
kh
Figure 4.3: (a) The neutral curves of Cauchy stress 2 against kh for exural motion with an angle of propagation = 15 . (b) An absolute error of second and third order approximations shown at the left-most plot in Figure 4.2 against kh. The layer is composed of MRM, see Table 1.1 (p. 23), and nitely deformed with 1 = 0.9, 2 = 1.2 and 2 = 5.5.
a situation when the second order approximation is apparently better then the third order. This can be explained by comparing the absolute errors of both the second and third order approximations presented in right-hand plot in Figure 4.3. The third order approximation is clearly a better approximation for all kh 0.2 and therefore for all kh 1. It is worth reiterating that all our approximations were obtained for kh 1 and, as any asymptotic expansions, may, but should not be expected to, provide good approximation outside their destined domain of validity. It is only a quirk of the material and pre-stress parameters that made the second order approximation look better then the third order.
Numerical Second order Third order Numerical Second order Third order
23.952
5.2
14.383
5.1
14.38
23.951
14.377
23.95
4.9
14.374
4.8 0 0.3 0.6 0.9
14.371
23.949
0.3
0.6
0.9
kh
kh
kh
Figure 4.4: Numerical and asymptotic solutions for the scaled frequency of the rst three exural harmonics, associated with the rst family of shear resonance frequencies and the angle of propagation = 15 . The material of the layer is OM1, see Table 1.1 (p. 23), subjected to a primary deformation with 1 = 0.7, 2 = 1.6 and 2 = 10.0.
The next Figure 4.4 demonstrates the scaled frequencies of the rst three harmonics for an Ogden material. All of these harmonics belong to the rst family of shear resonance frequencies. A remarkable feature of the right-most plot in Figure 4.4 is the decrease in scaled frequency associated with negative group velocity. We note that it again has no incremental stability repercussions, which is indicated by the neutral curves in Figure 4.5. Negative group velocity is not possible for long wave high frequency exural motion in a Mooney-Rivlin material and the argument is the same as used in (3.88) for extensional motion in Chapter 3. However, the plots in Figure 4.4 demonstrate a new feature in comparison with extensional motion, which is

15
69
10
2
0
-5 Extensional Flexural 2=10.0 0 2 4 6 8 10
-10
kh Figure 4.5: The neutral curves of Cauchy stress 2 against kh for exural motion with an angle of propagation = 15 . The material parameters are chosen to coincide with those in Figure 4.4.
the possibility of having harmonics with both positive and negative group velocity within the same family of shear resonances. This is a clear indication of the dependence of second order frequency expansion terms on the harmonic number n. Indeed, for the rst family of shear resonance frequencies, existence of negative group velocity will arise for small kh whenever F1
(2) (2) df
= F1c c2 + F1s s2 < 0 ,
(2)
(2)
(4.60)
(2)
see (4.11), in which F1c does depend on n. Note, that for physically realistic response F1s is always positive, which immediately follows from the denitions (2.70) and the strong ellipticity conditions (1.36).
=15 =30 =45 =60 =75
25
25
20
20
15
F1
(2) 15
10
F1c
(2)
10
0 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0 0.6 0.8 1 1.2 1.4
2=2.0 2=1.5 2=1.0 2=0.5 1.6 1.8 2
(a)
1
(2)
(b)
Figure 4.6: Dependence of the second order frequency expansion term F1 and the governing equation (2) coecient F1c on the change of in-plane deformation, shown for (a) various angles of propagation with xed 2 = 2.0 and (b) various normal deformations 2 for xed = 15 . The plots depict the motions associated with the rst harmonic from the rst family of shear resonance frequencies in OM1 subject to deformation with 2 = 6.0.
A better insight into the possibilities for negative group velocity may be obtained by direct (2) (2) analysing of the coecients F1 and F1c . We only remark that some of the congurations shown on the following graphs may not be incrementally stable, which is caused by the fact that the Cauchy stress must be uniquely specied for a wide range of primary deformations and it (2) is not always possible. F1c < 0 is the necessary condition for the existence of negative group
This may be demonstrated by comparing non-overlapping stability regions in Figure 2.2. It is clearly not
70
velocity, but it also indicates that equation (4.49) is not wave-like. In order to illustrate possible (2) (2) (2) scenarios for which F1c < 0, Figure 4.6 shows F1 and F1c against kh for a variety of angles of propagation (left-hand plot) and normal stretches (right-hand plot), respectively. Both plots are presented for the rst family of shear resonances, and the harmonic number n = 1, and clearly indicate that it is not easily possible to obtain the negative group velocity for the rst (2) harmonic, which is related to the large positive term 2G2 /(f )2 in F1c . However, analogous 1 1 plots for the second harmonic in Figure 4.7 demonstrate numerous possible congurations for negative group velocity.
10 8 6 =15 =30 =45 =60 =75 10 8 6 2=2.0 2=1.5 2=1.0 2=0.5
(2) F1 4
2 0 -2 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(2) F1c 4
2 0 -2 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(a)
1
(2)
(b)
Figure 4.7: Dependence of the second order frequency expansion term F1 and the governing equation (2) coecient F1c on the change of in-plane deformation, shown for (a) various angles of propagation with xed 2 = 2.0 and (b) various normal deformations 2 for xed = 15 . The plots depict the motions associated with the second harmonic from the rst family of shear resonance frequencies in OM1 subject to deformation with 2 = 6.0.
We nally remark that in Kaplunov et al. (2001a) it is shown that in the analogous plain strain case the existence of negative group velocity also changes the associated governing equation for the long wave amplitude from hyperbolic to elliptic, which is obviously not the case here. It should be stressed that the governing equation (4.49) is only valid in the vicinity of the cut-o frequencies. Further work is therefore required to fully elucidate the implications of a possible non-wave-like governing equation on dynamic response.
possible to choose a value of 2 applicable both for regions corresponding to 2 = 0.6 and 2 = 1.8. This remark is also applicable to the numerical investigation of coecients in Chapter 3, because our notion of incremental stability assumes stability of both extensional and exural motions.
Chapter 5
Layer with free faces: Long wave low frequency extensional motion
Long wave low frequency extensional wave propagation in a pre-stressed incompressible elastic layer can not in general be reduced to a lower-dimensional theory with an associated single governing equation. Nevertheless, using knowledge of the asymptotic behaviour of the solutions, it is still possible to build a consistent simplied asymptotic theory, more appropriate for practical applications than the full three-dimensional equations. The analysis is begun by deriving an appropriate asymptotic representation of the dispersion relation, which is subsequently studied at various orders. It was observed in the numerical solutions of the dispersion relation, presented in Chapter 2, that extensional motion exhibits two nite limits of phase speed at low wave number. These are shown to be solutions of the quadratic (in v 2 ) equation. Generally, these solutions are irrational and can not be represented without radicals, which makes this problem perhaps the most algebraically complicated in this thesis. The implication is that the second order coecients for the frequency expansions can only be obtained in terms of the leading order coecient. The second order frequency expansion coecients are then obtained and the resulting expansions used to determine the relative asymptotic orders of the displacement components. Asymptotically approximate equations are again derived to establish an asymptotic model for this type of motion, which in this case is shown to be a system of two two-dimensional governing equations, associated with the two quasi-fronts. The chapter is concluded by discussion of some illustrative numerical examples.
5.1
The dispersion in long wave low frequency regime
Long wave low frequency extensional motion was analysed numerically in Chapter 2 of this thesis and shown to be associated with solution branches having nite wave speeds in the low wave number limit. Therefore, the variation of wave speed is limited for this mode of propagation 2 2 2 and solutions q1 , q2 and q3 of the secular equation (1.65) may be assumed to be slowly varying quantities generally of O(1). Notwithstanding whether these solutions are real or complex, the hyperbolic tangents Tm (h), m {1, 2, 3}, occurring in the dispersion relation (2.41), may be expanded as kh 0 into the following power series 1 3 Tm (h) = qm qm 3 + O( 5 ) , 3
df
(5.1)
in which we re-introduced the small parameter = h/l = kh, the ratio of layer half-thickness h and typical wave length l. These expansions, when inserted back into the dispersion relation, will deliver the appropriate approximation of the dispersion relation, which at the leading order
71
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION may be factorised and written as

2 2 2 2 2 2 21 23 q1 q2 q3 (q1 q2 )(q1 q3 )(q2 q3 )
72
(23 21 )2 (21 12 + 23 31 )s2 + (23 23 + 13 21 )c2 E0 () O( 2 ) , v v
(5.2)
within which the function E0 () is given by v v E0 () = v 4 (E1c + 13 )c2 + (E3s + 31 )s2 v 2 + 13 E1c c4 + E1c E3s + 13 31 (13 g1 g3 )2 c2 s2 + 31 E3s s4 , E1c = 2(12 + g1 ) ,
(0) df (0) (0) (0) (0) df (0) (0)
(5.3)
E3s = 2(23 + g3 ) ,
(0) df
with the material parameters and pre-stress dened in (1.30) and (2.26). The second order term of the dispersion relation (2.41) is more complex and involves various combinations of solutions 2 2 2 q1 , q2 and q3 of the secular equation (1.65). We therefore use identities (1.66)(1.68) to eliminate qm from our expressions and to establish the second order term of the dispersion relation (2.41) in the following form 1 2 2 2 2 2 2 q1 q2 q3 (q1 q2 )(q1 q3 )(q2 q3 ) 3 v (23 21 )2 (21 12 + 23 31 )s2 + (23 23 + 13 21 )c2 + E0 () (21 + 23 )2 c1 v v .
2 2 (2 c5 ) 21 23 v 2 + 23 (g1 12 21 )c2 + 21 (g3 23 32 )s2 v
(5.4)
We remark that this particular representation will prove indispensable in the following analysis and now turn our attention to a detailed investigation of various orders of approximation of the dispersion relation (2.41).
5.1.1
Leading order approximation and incremental stability
We have already remarked that in the long wave low frequency regime the wave speed remains nite, it is therefore appropriate to seek scaled wave speeds in the form of power series expansions, given by v 2 = v0 + 2 v2 + O( 4 ) . 2 2 (5.5) When this expansion is substituted into the leading order approximation of the dispersion relation (5.2), we obtain an explicit representation of the dispersion relation for the leading order scaled wave speed, yielding
2 2 2 2 2 2 21 23 q1 q2 q3 (q1 q2 )(q1 q3 )(q2 q3 )
(23 21 )0 (21 12 + 23 31 )s2 + (23 23 + 13 21 )c2 E0 (0 ) = 0 . v2 v
(5.6)
The solutions associated with qm = 0 or qi = qj (i = j, i, j, m {1, 2, 3}) are of no interest since they are associated with degenerate cases, leading to non-dispersive solutions (we always 2 assume that qm , m {1, 2, 3} are distinct and non-zero, see Chapter 2). The explicit factor within braces also corresponds to a non-dispersive solution associated with an exceptional wave able to satisfy both equations of motion and boundary conditions, see (2.16). Thus we focus on the factor E0 (0 ), which is given by (5.3). It is a quadratic in v0 (or let us term it biv 2 quadratic equation) which coincides with our previous observation concerning two nite wave speed limits. Generally, solutions of this bi-quadratic equation contain radicals and therefore are irrational, making impossible the substituting of the solutions into subsequent approximation orders. The discriminant of the equation E0 (0 ) = 0 may be analysed to formulate necessary v
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION
73
and sucient conditions on pre-stress and material parameters for which the roots are rational for every direction of propagation, which are either g1 + g3 13 = 0 or (g1 + g3 13 )2 (212 + 2g1 13 )(223 + 2g3 31 ) = 0 . (5.7)
It is easy to see now that it is generally not possible to choose a deformation associated with rational solutions without specifying the normal Cauchy stress on the layer surface, see (2.26). Thus we can only obtain the second order term of the wave speed approximation (5.5) in terms of the leading order. When analysing the bifurcation conditions for a layer with free faces we never justied the choice of interval in the inequality (2.58), see the associated footnote on page 33. Actually, because of the continuity of neutral curves it is sucient to show that the necessary conditions (2.60) and (2.62), valid only for kh 0, both correspond to stability regions (the continuity of neutral curves extending the result to all values of kh). Here we will demonstrate it for the necessary condition of extensional stability (2.60). Indeed, in order to ensure that both solutions of the bi-quadratic equation E0 (0 ) = 0 are positive it is necessary to require v (E1c + 13 )c2 + (E3s + 31 )s2 > 0 ,
(0) 13 E1c c4 (0) (0)
(5.8) (5.9)
+ 2(21 12 + 23 23 ) 4(12 23 2 13 ) 13 31 s2 c2 > 0 ,
(0) 31 E3s s4
(23 21 )2 + 2 13
see (5.3). Upon using the parameter denitions, given immediately after (1.65) and (5.3), it is easy to see that this necessary condition (2.60) is identical to (5.9), thus indicating that the choice of the stability interval for extensional condition (2.60) is valid.
5.1.2
Second order approximation
The leading order term v0 of the expansion for scaled wave speed (5.5) was shown to be a solution 2 of the bi-quadratic equation (5.3). This solution is generally irrational making it ineectual for simplifying the second order expressions. Instead we will only assume that the quantity v0 is a 2 solution of (5.3), indicating that v (5.10) E0 () = O( 2 ) . To establish the second order correction for squared speed v2 we can ignore all terms with factor 2 E0 () in the second order dispersion relation (5.4). Taking into consideration the correction v term associated with the leading order dispersion relation (5.2), at second order we obtain the following relation
2 21 23 20 (E1c + 13 )c2 (E3s + 31 )s2 v2 v2 (0) (0)
1 2 2 2 2 + (0 c5 ) 21 23 vs0 + 23 (g1 12 21 )c2 + 21 (g3 23 32 )s2 = 0 , v 3
(5.11)
Strictly speaking, for the directions of propagation coinciding with the principal axes of deformation, one of the solutions of E0 (0 ) = 0 degenerates into the (rational) solution of so-called anti-plane problem. For bi-axial v deformations with 1 = 3 solutions of E0 (0 ) = 0 are also rational. However in both of these cases the dynamic v response is essentially limited to a plane strain type situation. The proof of the fact that neutral curves are always continuous was not seemingly given previously in the literature. Since this topic is beyond the scope of our work, we limit ourselves with the following simple consideration. The only possibly non-continuous functions that gure in the bifurcation conditions (2.55) and (2.56) are the tangents Tm (h), m {1, 2, 3}. Each of these bifurcation conditions is essentially a quadratic equation in 2 and every tangent function is involved in every coecient of this quadratic. It is the case that solutions of the quadratic will vary continuously when the variation of the coecients is continuous. Let us suppose then that at a certain value of kh one of the tangents becomes discontinuous. This tangent will dominate the coecients of the quadratic in the vicinity of this point, and since it occurs in every coecient, to leading order it will factorise out of the quadratic and not inuence the solutions. Thus possible discontinuity in coecients of the bifurcation conditions (2.55) and (2.56) will never cause associated discontinuity in neutral curves.
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION which may be recast as v2 = 2

2 2 v2 2 1 (0 c5 ) 21 23 v0 + 23 (g1 12 21 )c2 + 21 (g3 23 32 )s2 . (0) 3 2 + (E (0) + )c2 22 21 23 (E3s + 31 )s v0 13 1c
74
(5.12)
We can use (5.12) to determine the second order wave speed expansion coecient whenever we know the leading order expansion term. Thus we have obtained second order expansions for both long wave low frequency extensional modes. We remark that it is generally possible to choose a combination of pre-stress and material parameters for which the denominator in (5.12) vanishes. Comparing the leading order dispersion relation (5.3) and expression in the denominator of (5.12) it is easy to see that this situation corresponds to a double root of the bi-quadratic equation (5.3) and therefore vanishing of the associated discriminant. However, a detailed parameter analysis indicates that such coupling of roots of the equation (5.3) can in general be possible only for certain exceptional combinations of the direction of propagation and normal Cauchy stress 2 . The condition only becomes independent of for a bi-axial deformation with 1 = 3 , which is of no interest for us because of the corresponding degeneration of dynamic response analogous to a plane strain case. In the vast majority of cases, such coupling of roots will not happen at all, expression (5.12) providing satisfactory approximation.
5.1.3
Relative orders of displacements
The relative orders of displacement components may be obtained from general solutions (2.43) (2.45) by substituting the relations (2.47) and subsequently expanding the trigonometric functions occurring in the resulting expressions. In order to do this we rst remark that at low wave number Sm (x2 ) = x2 1 x3 3 3 2 qm + q + O( 5 ) , h 6 h3 m Cm (x2 ) = 1 + 1 x2 2 2 2 q + O( 4 ) , 2 h2 m (5.13)
within which we assume that ratio x2 /h is of O(1). We therefore insert (2.47) and (5.13) into the general solutions (2.43)(2.45) to establish that u1 O(pt ) , u2 O(pt ) , u3 O(pt ) . (5.14)
The in-plane displacement is dominant for this type of motion. It is worth remarking that in contrast to the high frequency cases analysed in Chapter 3 and Chapter 4, we can not describe 2 the wave structure precisely. This is because both the wave speed and qm , m {1, 2, 3}, are all quantities of O(1).
5.2
Asymptotically approximate equations
In a similar manner to Chapter 3 and Chapter 4, we begin the derivation of asymptotically approximate equations by changing the scale of space and time variables. Specically, the inplane coordinates x1 and x3 are balanced with a typical wave length l and the normal coordinate x2 with the half-thickness h, thus x1 = l1 , x2 = h = l , x3 = l3 . (5.15)
Expansion (5.5) indicates that to leading order any long wave low frequency extensional wave propagates with the scaled phase velocity v v v0 , in which v0 is one of the solutions associated with E0 (0 ) = 0, see (5.3). The distance of a typical wave length l is covered in time v t = l/v, therefore we rescale the time variable as t=l . (5.16) v0
75
We also know that the relative orders of displacement components and pressure are distributed for long wave low frequency according to (5.14). We therefore introduce u1 (1 , , 3 , ) = lu (1 , , 3 , ) , 1 u2 (1 , , 3 , ) = lu (1 , , 3 , ) , 2 u3 (1 , , 3 , ) = lu (1 , , 3 , ) , 3 pt (1 , , 3 , ) = v0 p (1 , , 3 , ) , 2 t (5.17)
where the superscript denotes non-dimensional quantities of the same asymptotic order. We can therefore use (5.15), (5.16) and (5.17) to recast the equations of motion (1.18)(1.20) in non-dimensional form, yielding 21 u + 2 B1111 u 1 1 + 31 u 3 3 v0 (u + p 1 ) 2 1, 1, 1, 1, t,
+ (B1122 + B1221 )u 1 + (B1133 + B1331 )u 3 1 = 0 , 2, 3,
(5.18)
(B1122 + B1221 )u 1 + B2222 u + (B2233 + B2332 )u 3 1, 2, 3, v0 p + 2 12 u 1 1 + 32 u 3 3 v0 u = 0 , 2 t, 2 2, 2, 2, 23 u + 2 (B1133 + B1331 )u 3 1 + (B2233 + B2332 )u 3 3, 1, 2, + 13 u 1 1 + B3333 u 3 3 v0 (p 3 + u ) = 0 , 2 t, 3, 3, 3, (5.20) (5.19)
where comma subscripts denote dierentiation with respect to the indicated space or time variable. The rescalings of space and time variables were chosen to ensure that derivatives of all of the displacement components and incremental pressure, with respect to scaled variables, remain quantities of the same asymptotic order. This enables us to determine the contribution of various terms directly by comparing powers of multipliers. The system of equations of motion (5.18)(5.20) must be solved in conjunction with the rescaled incompressibility condition, given by u 1 + u + u 3 = 0 , (5.21) 1, 2, 3, subject to the appropriately rescaled zero incremental traction boundary conditions 21 u + 2 (B1221 + p)u 1 = 0 2, 1, at = 1 , at = 1 , (5.22) (5.23) (5.24)
B1122 u 1 + (B2222 + p)u + B2233 u 3 v0 p = 0 2, 2 t 1, 3, 23 u + 2 (B2332 + p)u 3 = 0 2, 3,

m
at = 1 .
The solution of (5.18)(5.24) is now sought in a form of the power series (u , u , u , p ) = 1 2 3 t

n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(5.25)
After substitution of the solutions (5.25) back into equations (5.18)(5.24), the problem reduces to a hierarchical system of ordinary dierential equations at various orders.
5.2.1
At the leading order, the problem (5.18)(5.24) reduces to the following combination of leading order equations of motion (0) 21 u1, = 0 , (5.26) (B1122 + B1221 )u1,1 + B2222 u2, + (B2233 + B2332 )u3,3 v0 pt, = 0 , 2 23 u3, = 0 ,
(0) (0) (0) (0) (0)
(5.27) (5.28)
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION and leading order incompressibility condition u1,1 + u2, + u3,3 = 0 , which must be solved subject to the appropriate boundary conditions 21 u1, = 0
(0) (0) (0) (0) (0) (0)
76
(5.29)
at = 1 ,
(0) (0)
(5.30) =0 at = 1 , (5.31) (5.32)
B1122 u1,1 + (B2222 + p )u2, + B2233 u3,3 v0 pt 2 23 u3, = 0

(0)
at = 1 .
A curious departure from the situation observed for the previous asymptotic models may be observed for boundary value problems (5.26), (5.30) and (5.28), (5.32). These problems possess separate and independent solutions, which indicates the existence of two independent motions to leading order, each governed by two independent functions. Specically, the general solutions of (5.26), (5.30) and (5.28), (5.32) are given by u1
(0)
= U1
(0,0)
u3
(0)
= U3
(0,0)
(5.33)
We remark that hereafter we use double superscripts to denote functions independent of . When solutions (5.33) are inserted into the appropriate incompressibility condition (5.29), the (0) general solution for u2 can be established, yielding u2
(0)
= U1,1
(0,0)
+ U3,3
(0)
(0,0)
(5.34)
in which we used the fact that normal displacement u2 is an anti-symmetric function of for extensional motion. The boundary value problem (5.27), (5.31) is nally used to establish a solution for incremental pressure in the form pt
(0)
= Pt
(0,0)
Pt
(0,0)
1 (0,0) (0,0) (b21 + g1 )U1,1 + (b23 + g3 )U3,3 , v0 2
(5.35)
with the parameters b21 and b23 dened as in (1.30). We have therefore obtained leading order solutions for the displacement components and pressure increment in terms of two functions (0,0) (0,0) U1 and U3 , which will subsequently be referred to as leading order governing extensions. (0,0) (0,0) Since all of the leading order equations are already satised, the functions U1 and U3 can not be determined without resorting to higher order.
5.2.2
The second order problem comprises of the appropriate equations of motion 21 u1, = B1111 u1,1 1 31 u1,3 3 + v0 u1, 2
(0) (2) (0) (0) (0)
(B1122 + B1221 )u2,1 (B1133 + B1331 )u3,3 1 + v0 pt,1 , 2

(2) (2)
(0)
(0)
(5.36)
(B1122 + B1221 )u1,1 + (B2233 + B2332 )u3,3 + B2222 u2, v0 pt, = 12 u2,1 1 32 u2,3 3 + v0 u2, , 2 2 23 u3, = (B1133 + B1331 )u1,3 1 13 u3,1 1
(0) (0) (2) (0) (0) (2) (2) (0) (0) (0)
(5.37)
(B2233 + B2332 )u2,3 B3333 u3,3 3 + v0 u3, + v0 pt,3 , 2 2
(0)
(0)
(5.38)
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION solved subject to the incompressibility condition u1,1 + u2, + u3,3 = 0 , and second order boundary conditions 21 u1, = (B1221 + p)u2,1
(2) (2) (2) (2) (0) (2) (2) (2)
77
(5.39)
at = 1 ,
(2)
(5.40) at = 1 , (5.41) (5.42)
B1122 u1,1 + (B2222 + p)u2, + B2233 u3,3 v0 pt 2 23 u3, = (B2332 + p)u2,3 u1

(2,2) (2) (2) (0)
=0
at = 1 .
(2,0)
A general solution to the rst equation of motion (5.36) can be represented in the form = 2 U1
(2,2)
+ U1
(5.43)
in which the function U1 may be obtained from (5.36). In order for this general solution to satisfy the rst boundary condition (5.40) the following condition must be met
(0) v0 U1, E1c U1,1 1 E1s U1,3 3 + Ecs U3,1 3 = 0 , 2 (0,0) (0) (0,0) (0) (0,0) (0,0)
E1s = 31 ,
(0) df
(5.44)
where the parameter E1c is dened immediately after (5.3) and

(0) Ecs = 13 g1 g3 . df
(0)
When the condition (5.44)1 is satised, the function U1 U1

(2,2)
(2,2)
possesses the explicit representation (5.45)
g1 (0,0) (0,0) U1,1 1 + U3,1 3 , 221
where we remark that the function U1 can not be determined at the second order. A similar train of thought may be applied to the boundary value problem (5.38), (5.42) yielding the (2) solution for the second order displacement u3 in the form u3
(2)
(2,0)
= U3 U3
(2,2) 2
(2,2)
+ U3 , g3 (0,0) (0,0) = U3,3 3 + U1,1 3 . 223
(2,0)
(5.46)
The specied solution is only valid provided

(0) v0 U3, E3c U3,1 1 E3s U3,3 3 + Ecs U1,1 3 = 0 , 2 (0,0) (0) (0,0) (0) (0,0) (0,0)
E3c = 13 ,
(0) df
(5.47)
which was established upon use of the third boundary condition (5.42). The incompressibility condition may now be utilised to determine the second order solution for normal displacement, yielding u2
(2)
= 3 U2
(2,1) U2 (2,3)
(2,3)
+ U2
(2,1)
(5.48)
U2
g1 g3 (0,0) (0,0) (0,0) (0,0) = U1,1 1 1 + U3,1 1 3 U3,3 3 3 + U1,1 3 3 621 623
(2,0) U1,1
(2,0) U3,3 ,
The only quantity we have not yet obtained at the second order is an appropriate correction for incremental pressure. The second equation of motion (5.37) may now be considered as an (2) ordinary dierential equation for pt with solution given by pt
(2)
= 2 Pt
(2,2)
+ Pt
(2,0)
(5.49)
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION within which Pt

(2,2)
78
1 g1 b21 g3 b23 (0) (0,0) (0,0) (0) E1c 12 U1,1 1 1 + Ecs 31 + 32 U1,1 3 3 2 21 23 20 v g1 b21 g3 b23 (0,0) (0) (0,0) (0) + Ecs + 12 13 U3,1 1 3 + E3s 32 U3,3 3 3 . (5.50) 21 23
(2,0)
The function Pt Pt
is determined from boundary condition (5.41) and represented as
(2,0)
2 1 g3 (2,0) (0,0) (0) Ecs + 31 32 U1,1 3 3 2(b21 + g1 )U1,1 + 23 20 v2 2 g1 (2,0) (0,0) (0) + 2(b23 + g3 )U3,3 + Ecs 12 + 13 U3,1 1 3 21 2 G G2 (2) (0,0) (2) (0,0) + Q1c + 1 U1,1 1 1 + Q3s + 3 U3,3 3 3 , 21 23
(5.51)
with parameters Q1c and Q3s dened in (2.70). The two existence conditions (5.44)1 and (5.47)1 , obtained in the process of solving the second order problem, are in fact two-dimensional governing equations for the leading order (0,0) (0,0) extensions U1 and U3 . Therefore, in contrast to asymptotic models of Chapter 3 and Chapter 4, a system of two simultaneous dierential equations governs long wave low frequency df (0,0) extensional motion of a plate. It is convenient to introduce new functions u(0) = U1 and v (0) = U3
df (0,0)
(2)
(2)
, in terms of which the governing equations (5.44)1 and (5.47)1 take form
(0) v0 u(0) E1c u,1 1 E1s u,3 3 + Ecs v,3 1 = 0 , 2 , (0) v0 v, E3c v,1 1 E3s v,3 3 + Ecs u,3 1 = 0 . 2 (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)
(5.52) (5.53)
These equations may be rescaled back into original dimensional variables, yielding
2 (0) 2 (0) 2 (0) 2 u(0) (0) u (0) u (0) v E1c E1s + Ecs = 0, t2 x1 x3 x2 x2 1 3 2 (0) 2 (0) 2 (0) 2 v (0) (0) v (0) v (0) u E3s E3c + Ecs = 0, t2 x1 x3 x2 x2 3 1
(5.54) (5.55)
where we assume that u(0) (x1 , x3 , t) u(0) (1 , 3 , ) and v (0) (x1 , x3 , t) v (0) (1 , 3 , ). Introducing the travelling wave solutions, which according to (1.64) have the form (u(0) , v (0) ) = (U (0) , V (0) )eik(x1 c +x3 s vt) , (5.56)
we may transform the governing equations (5.54) and (5.55) into a system of two homogeneous linear equations in two variables, given by
(0) v 2 E1c c2 E1s s2 U (0) + Ecs c s V (0) = 0 , (0) Ecs c s U (0) + v 2 E3c c2 E3s s2 V (0) = 0 , (0) (0) (0) (0)
(5.57) (5.58)
where U (0) and V (0) are the amplitudes of the travelling waves. The condition for the existence of non-trivial solutions of this system is the vanishing of the determinant of associated matrix of coecients, specically v 4 (E1c + E3c )c2 + (E1s + E3s )s2 v 2
(0) + E1c E3c c4 + E1c E3s + E1s E3c (Ecs )2 c2 s2 + E1s E3s s4 = 0 , (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)
(5.59)
79
v which may be shown to coincide with E0 () = 0, see (5.3). Thus, the dispersion relation of the system of leading order governing equations (5.54), (5.55) has an identical set of non-dispersive solutions to the leading order approximation of the dispersion relation (5.2). It is interesting to remark that for certain congurations of material parameters and prestress, it is possible to pick particular values of normal Cauchy stress 2 for which the coecient (0) (0) Ecs vanishes, uncoupling the system (5.54), (5.55). Moreover, the vanishing of Ecs makes solutions of the leading order bi-quadratic equation E0 (0 ) = 0 rational, which can be demonstrated v (0) by comparing (5.7) with the denition of Ecs given after (5.44). Notwithstanding the concerns of possible physical non-reality, or even incremental instability of such congurations, we believe they may help elucidate processes involved in long wave low frequency extensional motion of pre-stressed incompressible elastic plates.
5.2.3
Third order problem
It is not our intention to establish a complete third order solution for our problem, therefore only the steps necessary to derive a rened lower-dimensional model will be performed. Specically, we focus our attention on the rst and third third order equations of motion 21 u1, = B1111 u1,1 1 31 u1,3 3 + v0 u1, 2
(2) (4) (2) (2) (2)
(B1122 + B1221 )u2,1 (B1133 + B1331 )u3,3 1 + v0 pt,1 , 2

(4) (2) (2)
(2)
(2)
(5.60)
23 u3, = (B1133 + B1331 )u1,3 1 (B2233 + B2332 )u2,3

(2) (2) (2)
13 u3,1 1 B3333 u3,3 3 + v0 u3, + v0 pt,3 , 2 2
(2)
(5.61)
considered in conjunction with the appropriate third order boundary conditions 21 u1, = (B1221 + p)u2,1 23 u3, = (B2332 + p)u2,3
(4) (4) (4) (2) (2)
at = 1 , at = 1 .
(4,0)
(5.62) (5.63)
A general solution to the rst equation of motion (5.60) may be represented as u1

(4,4)
= 4 U1
(4,4)
+ 2 U1
(4,2)
+ U1
(5.64)
U1
2 2g1 G1 (0) (0,0) (4) (0,0) U1,1 1 1 1 + e1 Ecs U1,1 1 3 3 21 21 g1 (23 + g1 + g3 + 13 ) (0,0) (0) Ecs + 12 13 U3,1 1 1 3 21 G1 (0) (4) (0,0) + (E3s 31 ) + e1 U3,1 3 3 3 , 21
1 2421
E1c 12 +
(0)
(5.65)
in which
e1 =
(4) df
g3 13 + 31 32 . 23
This solution can only satisfy the corresponding boundary condition (5.62) when
(0) v0 U1, E1c U1,1 1 E1s U1,3 3 + Ecs U3,1 3 2 (2) (0,0) (2) (0,0) (2) (2,0) (0) (2,0) (0) (2,0) (2,0) (0,0) (2) (0,0)
E1c U1,1 1 1 1 + E3 31 U3,1 1 1 3 + E1s U1,1 1 3 3 + E3 13 U3,1 3 3 3 = 0 ,
(5.66)
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION in which we introduced the following functions of material and pre-stress parameters E1c = E1s
(2) (2) df
80
E3 31 E3 13
(2)
(2)
1 3 df 1 = 3 df 1 = 3 df 1 = 6
Q1c
G2 1 , 21 1 g3 g1 g2 (0) 2+ Ecs 3 31 + 32 , 2 23 21 23 1 g1 (23 + 13 + 23 21 2g1 ) (0) + Ecs + 12 13 , 2 21 g1 (0) g3 (0) (0) (E 2g3 ) + (E 31 ) 2(E3s 32 ) . 23 cs 21 3s
(2)
When the condition (5.66) is satised, an expression for the function U1 given by U1
(4,2)
(4,2)
may be established,
1 G1 (0) g1 g3 (2,0) (2,0) (4) (0,0) 6g1 U1,1 1 + U3,1 3 + Ecs + e1 U1,1 1 3 3 1221 21 23 2 G1 G1 (0) g1 g3 (2) (0,0) (4) (0,0) + Q1c U1,1 1 1 1 (E 31 ) + e1 U3,1 3 3 3 21 21 3s 23 g1 (23 + g3 + 13 ) (0,0) (0) U3,1 1 1 3 . + Ecs + 12 13 21 =
(5.67)
The method of solving the boundary value problem (5.61), (5.63) is very similar and delivers the following solution (4) (4,4) (4,2) (4,0) u3 = 4 U3 + 2 U3 + U3 , (5.68) U3
(4,4) 2 2g3 G3 (0) (0,0) (4) (0,0) U3,3 3 3 3 + e3 E U3,1 1 3 3 23 23 cs g3 (12 + g1 + g3 + 31 ) (0,0) (0) Ecs 31 + 32 U1,1 3 3 3 23 G3 (0) (4) (0,0) + (E 13 ) + e3 U1,1 1 1 3 , 23 1c
1 2423
E3s 32 +
(0)
(5.69)
U3
(4,2)
1 G3 (0) g1 g3 (2,0) (2,0) (4) (0,0) 6g3 U3,3 3 + U1,1 3 + E + e3 U3,1 1 3 3 1223 23 cs 21 G2 G3 (0) g1 g3 (2) (0,0) (4) (0,0) 3 + Q3s U3,3 3 3 3 (E1c 13 ) + e3 U1,1 1 1 3 23 23 21 g3 (12 + g1 + 31 ) (0,0) (0) + Ecs 31 + 32 U1,1 3 3 3 , 23 = e3 =
(4)
(5.70)
within which This solution is valid provided
13 g1 12 + 13 . 21
(5.71)
(0) v0 U3, E3c U3,1 1 E3s U3,3 3 + Ecs U1,1 3 2 (2) (0,0) (2) (0,0) (2)
(2,0)
(0)
(2,0)
(0)
(2,0)
(2,0) (0,0) (2) (0,0)
E3s U3,3 3 3 3 + E1 13 U1,1 3 3 3 + E3c U3,1 1 3 3 + E1 31 U1,1 1 1 3 = 0 ,
(5.72)

Numerical Asymptotic
3.525
81
3.524
2.96
v
2.92
3.523
3.522
2.88
3.521
Numerical Asymptotic 0 0.3 0.6 0.9
0.3
0.6
0.9
kh
kh
Figure 5.1: Numerical and asymptotic solutions for scaled wave speed v against scaled wave number kh, shown for both fundamental branches at low wave numbers. Waves propagate at the angle = 15 in OM1, see Table 1.1 (p. 23), subject to deformation with 1 = 2.0, 2 = 0.7 and 2 = 2.0.
-5
-10 Extensional Flexural 2=2.0 0 2 4 6 8 10
-15
kh Figure 5.2: Neutral curves of 2 against scaled wave number kh for the same pre-stress conguration as depicted in Figure 5.1.
where E3s = E3c

(2) (2) df
E1 13 E1 31
(2)
(2)
1 3 df 1 = 3 df 1 = 3 1 df = 6
Q3s
G2 3 , 23 1 g1 g3 g2 (0) 2+ Ecs 1 + 12 13 , 2 21 23 21 1 g3 (12 + 31 + 21 23 2g3 ) (0) + Ecs + 32 31 , 2 23 g1 (0) g3 (0) (0) (Ecs 2g1 ) + (E 13 ) 2(E1c 12 ) . 21 23 1c
(2)
It is worth remarking that the functions U1 and U3 can not be obtained at this order. Let us reconsider the existence conditions (5.66) and (5.72) obtained at this order. If we now introduce the following (non-scaled and scaled) functions u = u(0) + u(2) 2 + O( 4 ) , v=v
(0)
(4,0)
(4,0)
u = u(0) + u(2) 2 + O( 4 ) , v =v
(0)
(5.73) (5.74)
+v
(2) 2
+ O( ) ,
+v
(2) 2
+ O( ) ,
termed governing extensions, it is possible to combine leading order governing equations (5.44) and (5.47) with conditions (5.66) and (5.72) to obtain rened (second order) governing equations
CHAPTER 5. LOW FREQUENCY EXTENSIONAL MOTION in the form

(0) v0 u E1c u 1 1 E1s u 3 3 + Ecs v,1 3 2 , , , (2) (2) (0) (0)
82
2 E1c u 1 1 1 1 E3 31 v,1 1 1 3 E1s u 1 1 3 3 E3 13 v,1 3 3 3 , , (0) (0)
(2)
(2)
= 0,
(5.75)
(0) v0 v, E3s v,3 3 E3c v,1 1 + Ecs u 1 3 2 , (2) (2)
2 E3s v,3 3 3 3 E1 13 u 1 3 3 3 E3c v,1 1 3 3 E1 31 u 1 1 1 3 , , (2m,0) (2m,0)
(2)
(2)
= 0,
(5.76)
where u(2m) U1 and v (2m) U3 , m = 0, 1, 2, . . . . These equations may also be represented in terms of non-scaled variables, yielding
2 2 2 2u (0) u (0) u (0) v E1c E1s + Ecs t2 x1 x3 x2 x2 1 3
h2 E1c
(2)
4u
x4 1
E3 31
(2)
4 4v 4v (2) u (2) E1s E3 13 x3 x3 x2 x2 x1 x3 1 1 3 3
= 0,
(5.77)
2 2 2 2v (0) v (0) v (0) u E3c E3s + Ecs t2 x1 x3 x2 x2 1 3 4v 4u 4 4u (2) (2) (2) v (2) h2 E3s E1 13 E3c E1 31 3 x4 x1 x3 x2 x2 x1 x3 3 3 1 3
= 0.
(5.78)
If travelling harmonic wave solutions are substituted into governing equations (5.77) and (5.78), a system of two linear homogeneous equations is obtained, similar to the second order situation. In order for this system to have a non-trivial solution we require vanishing of the associated determinant. When the phase speed in the expression for this determinant is expanded in a power series (5.5), a second order approximation of the dispersion relation is established, its dispersive solutions matching the dispersive solutions of the second order approximation of the dispersion relation exactly.
5.3

2.7
5
2.69
2.68
-5
2.67
-10
2.66 Numerical Asymptotic 0 0.3 0.6 0.9
2.65
-15 0 2 4 6
(a)
kh
(b)
kh
Figure 5.3: (a) Scaled wave speed v against scaled wave number kh and (b) associated neutral curves of 2 against kh, shown for waves propagating at = 15 in MRM, see Table 1.1 (p. 23), pre-stressed with 1 = 1.5, 3 = 1.4 and 2 = 3.9.
To demonstrate some typical features of the asymptotic solutions, some plots in respect of Ogden and Mooney-Rivlin materials are presented. Plots in Figure 5.1 depict the phase speed
83
associated with both fundamental modes of the extensional dispersion relation (2.41) together with appropriate long wave low frequency approximations. Associated neutral curves are shown in Figure 5.2 and demonstrate incremental stability of this conguration. Although only a second order approximation is available in this case, the quality of approximations is generally good and can be accurately used instead of numerical solution at very low wave numbers. It was previously commented that, for certain material and pre-stress congurations, the second order term (5.12) of the scaled wave speed expansion grows large and may even cause the approximation to break down. We were not able to nd any congurations for which denominator of (5.12) vanishes, however depicted in Figure 5.3 is a situation when this denominator is very small. This pre-stress conguration is outside the range of deformations modelled realistically by the Mooney-Rivlin model, see Section 1.6.2, however it was chosen as the best demonstration of the discussed features. As expected, both nite wave speed limits are very close to each other in this case. It is also interesting to note that the error of approximation grows much faster in such conguration, indicating a change in asymptotic orders of various terms in (5.5).
Chapter 6
Layer with free faces: Long wave low frequency exural motion
The analysis of exural (anti-symmetric or bending) motion of an innitely thin layer subjected to uniform tension (usually referred to as a membrane) is a very special problem of dynamic theory of elasticity. One of the major results presented in this chapter is a demonstration of an asymptotically consistent method for deriving the classical membrane equation, as well as generalisation of available theory to the case of membranes of nite thickness. We begin our investigation with an asymptotic analysis of the exural dispersion relation in the long wave low frequency regime. An appropriate approximation for the scaled wave speed is established and utilised to determine the relative asymptotic orders of displacement components. The discussion of necessary conditions for incremental stability, which began in Chapter 2 and was continued in Chapter 5, is nally closed with the help of the derived asymptotic expansions. Then our attention turns to asymptotically approximate equations. These are derived in the usual manner and leading order and rened second order governing equations are obtained. It is demonstrated that the membrane equation is a particular case of our leading order governing equation, our second order governing equation being appropriate for a membrane of nite thickness. A model problem of shock edge loading of a semi-innite plate is then solved to demonstrate the existence of two possible types of quasi-fronts, one of which has no analogue in the isotropic case. A novel feature of the model lies in the fact that this type of quasi-front may change for various angles of wave propagation, thus greatly enriching the variety of material response for line- and pointloading problems.
6.1
The numerical investigation performed in Chapter 2 indicated that for long wave low frequency anti-symmetric motion, as the scaled wave number kh 0, there exists one nite phase speed limit associated with the fundamental mode. Generally, for the fundamental mode we may assume that as kh 0, the roots of the secular equation (1.65) do not undergo signicant changes and stay O(1). These observations enable us to derive long wave low frequency approximations of the hyperbolic tangents Tm (h) in the form of a power series 1 3 Tm (h) = qm qm 3 + O( 5 ) , 3
df
m {1, 2, 3} ,
(6.1)
in which we reintroduced the small parameter = h/l kh, where l is the layer half-thickness and l is a typical wave length. On substitution of the expansions (6.1) into the dispersion
84
CHAPTER 6. LOW FREQUENCY FLEXURAL MOTION relation (2.49) we obtain an approximation with error O( 4 ), namely
2 2 2 2 (q1 q2 )T2 (q3 , v )H(q1 , q2 , v )q3 (q1 q3 )T2 (q2 , v )H(q1 , q3 , v )q2 2 2 2 2 + (q2 q3 )T2 (q1 , v )H(q2 , q3 , v )q1 2
85
2 2 2 (q1 q2 )T2 (q3 , v )H(q1 , q2 , v )q3 4 3 (6.2)
2 2 2 2 (q1 q3 )T2 (q2 , v )H(q1 , q3 , v )q2 + (q2 q3 )T2 (q1 , v )H(q2 , q3 , v )q1 O( 4 ) . 4 4
Since we are interested in behaviour within the vicinity of a nite wave speed limit, we seek a scaled wave speed approximation in the following form v 2 = v0 + v2 2 + O( 4 ) . 2 2 (6.3)
Inserting the expansion (6.3) into the approximation of the dispersion relation (6.2), we obtain an asymptotic equality which must be satised at each order. Note, that the parameters qm , m {1, 2, 3}, which must satisfy the secular equation (1.65), may be eliminated through use of the relations (1.66)(1.68), taken in the form
2 2 2 q1 + q2 + q3 =
(21 + 23 )0 c1 (21 + 23 )2 2 v2 v2 + O( 4 ) , 21 23 21 23 v0 c2 v0 + c3 (20 c2 )2 2 4 2 v2 v2 + + O( 4 ) , 21 23 21 23
(6.4) (6.5) (6.6)
2 2 2 2 2 2 q1 q2 + q1 q3 + q2 q3 = 2 2 2 q1 q2 q3 =
(0 c4 )(0 c5 ) (20 c4 c5 )2 2 v2 v2 v2 v2 + + O( 4 ) . 21 23 21 23
6.1.1
Leading order approximation and incremental stability

2 2 2 2 2 2 21 23 (q1 q2 )(q1 q3 )(q2 q3 )
The leading order term of the dispersion relation (6.2) may be factorised, yielding
(0 c5 ) (21 23 )0 f1 v2 v2
(0)
(0)
(0) (0) v0 Fc c2 Fs s2 = 0 , 2
(6.7)
in which the equalities (6.4)(6.6) were used and f1 = (23 23 + 13 21 )c2 (12 21 + 23 31 )s2 ,
(0) Fc = 12 2 g1 , 21 (0) Fs = 32 2 g3 . 23
(6.8)
Because all values qm , m {1, 2, 3}, are assumed to be non-zero and distinct, the leading order term (6.7) can be equal to zero only when the leading order scaled wave speed v0 is equal to one of 2 (0) three possible limits. However, two of these limits, specically v0 = c5 and (21 23 )0 = f1 , 2 v2 were previously shown to be non-dispersive, see (2.16) and associated discussion. Thus only one of the three nite wave speed limits, implied by the equation (6.7), is dispersive, which supports our numerical observation. We therefore focus our attention on the only dispersive (0) (0) limit v0 = Fc c2 + Fs s2 . 2 In order to justify the choice for the interval of incremental stability in inequality (2.58), it is still necessary to demonstrate that the necessary condition for exural stability (2.62) is valid (the validity of the necessary condition for extensional stability (2.60) was shown in Chapter 5). It is easy to do this by comparing the leading order approximation for phase speed (0) (0) v0 = Fc c2 + Fs s2 with the quadratic for 2 guring in (2.56). The parameter denitions 2 in (2.26), (2.56) and (6.8) may then be used to establish that this quadratic is in fact the leading order wave speed taken with negative sign, i.e.
2 (0) (0) c8 2 221 23 2 + c9 Fc c2 Fs s2 ,
(6.9)
CHAPTER 6. LOW FREQUENCY FLEXURAL MOTION
86
see (2.56). Therefore the interval of incremental stability coincides with the interval for which the quadratic on the left-hand side of (6.9) remains negative. Since the strong ellipticity conditions (1.36) require that parameter c8 must be positive, the conclusion can be made that stability interval is enclosed between the two roots of the quadratic. This is precisely what condition (2.62) implies, thus proving necessity of (2.62) and, by the continuity of neutral curves, see footnote on page 73, of (2.58).
6.1.2
In view of the equalities (6.4)(6.6), the second order term of the dispersion relation (6.2) (adjusted to reect the eect of use of (6.4)(6.6) at leading order) may also be factorised, yielding
2 2 21 23 (23 21 )(23 g1 c2 + 21 g3 s2 21 23 c4 ) 21 23 f1 (0)
2 2 2 2 2 2 (2) (2) (2) v2 2 (q1 q2 )(q1 q3 )(q2 q3 )(0 c5 ) v2 + Fcc c4 + Fcs s2 c2 + Fss s4 = 0 ,
(6.10)
in which
(2) Fcc = (2) Fcs
1 (Fc 3
(0)
2 212 2g1 )g1 , 2 21
(2) Fss =
1 (Fs 3
(0)
2 223 2g3 )g3 , 2 23 (0)
(6.11) . (6.12)
1 = 3
(Fc
(0)
2 2 2g1 g3 (g1 + g3 13 ) (Fs 31 )g1 13 )g3 + 2 2 21 23 23 21 (2) (2) (2)
It is then deduced that v2 = Fcc c4 Fcs c2 s2 Fss s4 and the second order expansion for 2 the square of scaled wave speed may be written as
(0) (0) (2) (2) (2) v 2 = Fs s2 + Fc c2 Fcc c4 + Fcs c2 s2 + Fss s4 2 + O( 4 ) .
(6.13)
6.1.3
Comparison of the relative orders of displacements and pressure increment is essential for the derivation of the lower dimensional model for long wave low frequency motion. In order to accomplish such comparison it is necessary to represent the displacements and pressure in terms of a single constant, which can be done by inserting relations (2.54) into the solutions (2.50) (2.53). We previously remarked that in the long wave low frequency regime, qm , m {1, 2, 3}, is O(1) as 0. Hence, we may estimate the trigonometric functions occurring in (2.50)(2.53) and (2.54) with the following expansions Sm (x2 ) = qm x2 + O( 3 ) , h Cm (x2 ) = 1 + O( 2 ) , m {1, 2, 3} , (6.14)
in which we tacitly assume that x2 /h O(1). It is now possible to utilise relations (6.4)(6.6), as well as expansions (6.14), in order to determine the relative orders of displacements (2.50) (2.53), yielding u1 O(pt ) , u2 1 O(pt ) , u3 O(pt ) , (6.15) indicating that for this type of motion the leading order displacement is normal to the plane of the layer.
87
6.2
The derivation of asymptotically approximate equations involves the rescaling of space and time variables. The rescaling of space coordinates aims to balance the plate half-thickness h and normal coordinate with a typical wave length l and in-plane coordinates, so that we choose x1 = l1 , x2 = h = l , x3 = l3 , (6.16)
where 1 , and 3 are the new non-dimensional space variables. A typical wave of the considered (0) (0) type propagates with a scaled speed F (0) , where F (0) Fs s2 + Fc c2 , see expansion (6.13), and thus travels the distance of a typical wave length l in time l /F (0) and the appropriate scale for time may be chosen in form l , (6.17) t= F (0)
with denoting non-dimensional time. The distribution of the relative orders of displacements (6.15) dictates the asymptotic structure of the displacement components and pressure, motivating the following rescalings u1 (1 , , 3 , ) = lu (1 , , 3 , ) , 1 u2 (1 , , 3 , ) = lu (1 , , 3 , ) , 2 u3 (1 , , 3 , ) = lu (1 , , 3 , ) , 3 pt (1 , , 3 , ) = F (0) p (1 , , 3 , ) , t (6.18)
in which superscript denotes non-dimensional quantities of a same asymptotic order. We may now reformulate the equations of motion (1.18)(1.20) in terms of the rescaled non-dimensional variables, so that 21 u + (B1122 + B1221 )u 1 + 2 B1111 u 1 1 1, 2, 1, + (B1133 + B1331 )u 1 3 + 31 u 3 3 F (0) (u + p 1 ) = 0 , 1, 3, 1, t, B2222 u + 2 12 u 1 1 + (B1122 + B1221 )u 1 2, 2, 1, + (B2233 + B2332 )u 3 + 32 u 3 3 F (0) (u + p ) = 0 , 3, 2, 2, t, 23 u + (B2233 + B2332 )u 3 + 2 B3333 u 3 3 3, 2, 3, + (B1133 + B1331 )u 1 3 + 13 u 1 1 F (0) (u + p 3 ) = 0 , 3, 1, 3, t, (6.21) in which a comma subscript denotes dierentiation with respect to the indicated (space or time) variable. Equations (6.19)(6.21) must be solved in conjunction with the appropriately rescaled incompressibility condition (6.22) u + 2 (u 1 + u 3 ) = 0 , 2, 1, 3, and subject to the corresponding zero surface traction boundary conditions 21 u + (B1221 + p )u 1 = 0 2, 1, 23 u 3, + (B2332 + p )u 3 2, =0 at = 1 , at = 1 . at = 1 , (6.23) (6.24) (6.25) (6.20) (6.19)
(B2222 + p )u + 2 (B1122 u 1 + B2233 u 3 F (0) p ) = 0 2, t 1, 3,
All information regarding long wave low frequency contributions of dierent terms in equations (6.19)(6.25) is now indicated explicitly through the appearance of 2 . This enables us to set up a perturbation scheme to obtain lower-dimensional governing equations for this type of motion. The solutions will be sought in a form of the power series
m
(u , u , u , p ) 1 2 3 t
=
n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(6.26)
Substituting the solution (6.26) into the problem indicated by (6.19)(6.25) we obtain a hierarchical system at various orders.
88
6.2.1
At leading order, the equations of motion (6.19)(6.21) take the form 21 u1, + (B1122 + B1221 )u2,1 = 0 , B2222 u2, = 0 , 23 u3, + (B2233 + B2332 )u2,3 = 0 , which are to be solved subject to the leading order incompressibility condition u2, = 0 , and appropriate leading order boundary conditions 21 u1, + (B1221 + p )u2,1 = 0 (B2222 + p )u2, = 0 at 23 u3, + (B2332 + p )u2,3 = 0
(0) (0) (0) (0) (0) (0) (0) (0) (0) (0) (0)
(6.27) (6.28) (6.29)
(6.30)
at = 1 , = 1 , at = 1 .
(6.31) (6.32) (6.33)
The general solution of the boundary value problem (6.28), (6.32) may be represented as u2
(0,0) (0,0) (0)
= U2
(0,0)
(6.34)
= U2 (1 , 3 , ). Henceforth, we use double superscripts to denote functions in which U2 independent of , with the rst superscript referring to the approximation order and the second denoting the power of any possible m multiplier. Also note that our choice of scaling parameters ensures that the displacement components and pressure vary appropriately for exural motion, i.e. as an anti-symmetric (u , u and p ) or symmetric (u ) function of the normal coordinate t 1 3 2 . Thus, for the sake of brevity, we will always omit terms of the solutions which do not comply with this requirement. Substituting the solution (6.34) into the boundary value problem (6.27), (6.31) it is possible (0) to obtain the solution for u1 in the form u1
(0)
= U1
(0,1)
U1
(0,1)
g1 (0,0) U , 21 2,1
(6.35)
where we recall that gm = 2m 2 , m {1, 3}. This procedure may be repeated for the problem (6.29), (6.33), to obtain u3
(0)
= U3
(0,1)
U3
(0,1)
g3 (0,0) U . 23 2,3
(6.36)
We have now obtained leading order solutions for the displacement components in terms of (0,0) the function U2 , or more specically, as linear functions of it and its derivatives. We will (0,0) term the leading order normal displacement U2 as the leading order mid-surface deection. Knowledge of this function determines the long wave low frequency motion at leading order. However, this function itself (and the leading order pressure increment) can not be determined without resorting to higher order.
89
6.2.2
The second order problem is comprised of the appropriate equations of motion 21 u1, + (B1122 + B1221 )u2,1 = B1111 u1,1 1 (B1133 + B1331 )u3,1 3 31 u1,3 3 + F (0) u1, + pt,1 F (0) pt, = B2222 u2, + 12 u2,1 1 + 32 u2,3 3
(0) (0) (2) (0) (0) (0) (0) (0) (0) (0) (2) (2)
(6.37)
+ (B1122 + B1221 )u1,1 + (B2233 + B2332 )u3,3 F (0) u2, ,

(2)
(0)
(0)
(6.38)
23 u3, + (B2233 + B2332 )u2,3 = 13 u3,1 1 (B1133 + B1331 )u1,1 3 B3333 u3,3 3 + F (0) u3, + pt,3 together with the second order incompressibility condition u2, = u1,1 u3,3 , solved subject to the second order boundary conditions 21 u1, + (B1221 + p )u2,1 = 0 F (0) pt
(0) (2) (0) (2) (2) (2) (0) (0) (0) (0) (0) (0) (0)
(2)
(6.39)
(6.40)
at = 1 ,
(0)
(6.41) (6.42) (6.43)
= (B2222 + p )u2, + B1122 u1,1 + B2233 u3,3 23 u3, + (B2332 + p )u2,3 = 0

(0) (2) (2)
at = 1 ,
at = 1 .
Note, the leading order pressure pt was deliberately shown on the left hand side of equations (6.38) and (6.42) to signify that it can only be determined at the current order. Upon use of the leading order solutions (6.35) and (6.36), the incompressibility condition (6.40) may be utilised to establish the second order term of normal displacement u2
(2)
= 2 U2
(2,2)
+ U2
(2,0)
(2,0)
U2
(2,2)
1 2
g1 (0,0) g3 (0,0) U2,1 1 + U 21 23 2,3 3
(6.44)
within which the function U2 can only be determined at third order. In view of (6.44), the analysis of the boundary value problem (6.38), (6.42) reveals that it may only be satised by a solution of the following form pt
(0)
= Pt
(0,1)
F (0) Pt
(0,1)
g1 (b21 + g1 ) (0,0) g3 (b23 + g3 ) (0,0) U2,1 1 + U2,3 3 , 21 23
(6.45)
provided
(0) (0) Fc U2,1 1 + Fs U2,3 3 F (0) U2, (0,0) (0,0) (0,0) (0)
= 0,
(0)
(6.46)
with the functions of pre-stress and material parameters Fc and Fs dened in (6.8). The condition (6.46) was introduced to accommodate the second equation of motion (6.38) and (0,0) provides the means to determine the leading order mid-surface deection U2 , hence we will refer to it as to the leading order governing equation for the mid-surface deection. Since all (0,0) leading order displacement components and incremental pressure are linear functions of U2
CHAPTER 6. LOW FREQUENCY FLEXURAL MOTION and its derivatives, equation (6.46) will in fact be valid for all of u1 particular we note that
(0) (0) F (0) u(0) = Fc um,1 1 + Fs um,3 3 , m, (0) (0) (0)
90 , u2
(0)
, u3
(0)
and pt
(0)
, in
m {1, 3} .
(6.47)
Using (6.47) to dispose of time derivatives in the rst equation of motion (6.37), it is possible (2) to conclude that the second order displacement coecient u1 has the form u1
(2)
= 3 U1
(2,3)
+ U1
(2,1)
(6.48)
which, after inserting this into the rst equation of motion (6.37) and the rst boundary condition (6.41), yields U1 U1 where f13 =
(2) (2,3) (2,1) 2 1 g1 g3 g1 (0,0) (0,0) + f2,11 U2,1 3 3 2 + f2,13 U2,1 1 1 + 6 21 21 23 g1 (2,0) 1 (2) (0,0) (2) (0,0) = U + f U + f11 U2,1 3 3 , 21 2,1 2 13 2,1 1 1
3Fcc , g1
(2)
f11 =
(2)
g1 (Fs
(0)
2 21
31 )
g3 (13 g1 g3 ) , 21 23
(2)
(6.49)
with the function of the material parameters and pre-stress Fcc dened immediately after (6.10). The same mathematical manipulation sequence repeated for the boundary value problem (6.39), (6.43) delivers the second order solution for the third displacement component, given by u3 U3 U3 in which f31 =
(2) (2,3) (2,1) (2)
= 3 U3
(2,3)
+ U3
(2,1)
, ,
(6.50)
2 1 g1 g3 g3 (2) (0,0) (2) (0,0) + f31 U2,1 1 3 + U2,3 3 3 2 + f33 6 21 23 23 g3 (2,0) 1 (2) (0,0) (2) (0,0) f U = U + + f33 U2,3 3 3 , 23 2,3 2 31 2,1 1 3
g3 (Fc
(0)
2 23
13 )
g1 (13 g1 g3 ) , 21 23
f33 =
(2)
3Fss , g3
(2)
(6.51)
with the parameter Fss dened immediately after (6.10). We shall now reconsider the leading order equation for a mid-surface deection (6.46), which in terms of the original variables may be recast as
(0) Fc 2 2 u0 2 u0 (0) u0 + Fs 2 = 0, t x2 x2 1 3 (0,0)
(2)
(6.52)
in which we introduced u0 (x1 , x3 , t) U2 (1 , 3 , ). It was mentioned previously that the solution of equation (6.52) enables the leading order solutions for all displacement components and pressure to be determined. Also, the dispersion relation associated with this equation is precisely the leading order approximation for wave speed given immediately after (6.8), which demonstrates the asymptotic consistency of the model. It is interesting to note that the dispersion relation associated with the leading order governing equations for long wave high frequency motion, or motion in the vicinity of the shear resonance frequencies, matches the second order dispersion relation approximation, see Chapter 3 or Chapter 4. However, in the present case the (2) second order pressure pt, can only be obtained at third order, in terms of the leading order mid-surface deection and its second order correction, thus the satisfying of the equation (6.52) does not guarantee the consistency of the model at second order.
91
6.2.3
Third order problem
Instead of solving all of the third order equations, we only consider those necessary to complete the second order solution and obtain a rened second order governing equation. Hence, we examine the second equation of motion F (0) pt, = B2222 u2, + 12 u2,1 1
(2) (2) (4) (2)
+ (B1122 + B1221 )u1,1 + (B2233 + B2332 )u3,3 + 32 u2,3 3 F (0) u2, , u2, = u1,1 u3,3 ,
(4) (2) (2)
(2)
(2)
(2)
(6.53)
subject to the third order incompressibility condition (6.54)

(2)
and the third order form of the second boundary condition, namely F (0) pt
(2)
= (B2222 + p )u2, + B1122 u1,1 + B2233 u3,3
(4)
(2)
at = 1 .
(6.55)
Inserting the solutions (6.48) and (6.50) into the incompressibility condition (6.54) it is possible (4) to deduce a general solution for u2 in the form u2 1 2
(4)
= 4 U2
(4,4)
+ 2 U2
(4,2)
+ U2
(4,0)
(6.56)
which when coupled with the boundary condition (6.55), yields U2

(4,2)
g1 (2,0) g3 (2,0) U + U 21 2,1 1 23 2,3 3 ,
U2
(4,4)
1 (2) (0,0) (2) (2) (0,0) (2) (0,0) f13 U2,1 1 1 1 + f11 + f31 U2,1 1 3 3 + f33 U2,3 3 3 3 4 = 1 g1 g3 (2) (2) (0,0) 2 + f11 + f31 U2,1 1 3 3 24 21 23 2 2 g1 g3 (2) (0,0) (2) (0,0) U2,1 1 1 1 + U2,3 3 3 3 2 + f13 2 + f33 21 23
+
(4,0)
and U2 can only be determined at higher order. The equation of motion (6.53) may then be (2) used to establish a general solution for pt , given by pt where F (0) Pt
2 1 g1 g1 g1 (2) (0,0) (0) (0) b21 + 12 Fc U2,1 1 1 1 + 32 Fs 2 + f13 6 21 21 21 g1 g3 g1 g3 g3 (2) (2) (0,0) (0) +b21 + f11 + b23 + f31 + 12 Fc U2,1 1 3 3 21 23 21 23 23 2 g3 g3 (2) (0,0) (0) + b23 + 32 Fs U2,3 3 3 3 . 2 + f33 23 23 (3,2) (2)
= 3 Pt
(3,2)
+ Pt
(1,2)
(6.57)
Finally, we use the boundary condition (6.55) to obtain F (0) Pt g1 (b21 + g1 ) (2,0) g3 (b23 + g3 ) (2,0) U2,1 1 + U2,3 3 21 23 3 1 g1 g1 g1 (2) (0,0) (0) (0) + 12 Fc f13 (2g1 + 3b21 ) U2,1 1 1 1 32 Fs 2 6 21 21 21 g1 g3 (g1 + g3 ) g3 (2) (2) (0,0) (0) + 12 Fc + f11 (2g1 + 3b21 ) + f31 (2g3 + 3b23 ) U2,1 1 3 3 21 23 23 3 g3 g3 (2) (0,0) (0) + 32 Fs f33 (2g3 + 3b23 ) U2,3 3 3 3 , 2 23 23
(1,2)
CHAPTER 6. LOW FREQUENCY FLEXURAL MOTION which is only valid provided

(0) (0) Fc U2,1 1 + Fs U2,3 3 F (0) U2, (0,0) (2,0) (2,0) (2,0) (0,0) (0,0)
92
(2) (2) (2) + Fcc U2,1 1 1 1 + Fcs U2,1 1 3 3 + Fss U2,3 3 3 3 = 0 ,
(6.58)
within which use has been made of the denitions (6.12), (6.49) and (6.51) to conclude that (2) (2) (2) Fcs = 1 g1 f11 + g3 f31 . Let us introduce a new function u(x1 , x3 , t) = u (1 , 3 , ), dened 3 as u = u(0) + u(2) 2 + O( 4 ) , u = u(0) + u(2) 2 + O( 4 ) , (6.59) where we assume u(m) U2 , m = 1, 2, 3, . . . . Note, when = 0 all of the solutions for (m) (m) (m) (m) u1 , u 3 and pt , m {0, 2, 4}, vanish, except the solutions for u2 . This strongly supports the choice of the term mid-surface deection for u. We also remark, that since every (2m,0) (0,0) (2m2,0) satises each boundary value problem posed for U2 , . . . , U2 , then function U2 every displacement component and pressure may be represented as a linear function of u and its derivatives, thus any governing equation for u is also valid for all of the displacement components and pressure. In view of the denition (6.59), the linear combination (6.46)+ 2 (6.58) gives the rened (second order) governing equation for a mid-surface deection
(0) Fc 2 4 4 4 2 u 2 u (0) u (2) u (2) u (2) u + Fs F (0) + 2 Fcc + Fcs + Fss 2 2 4 2 2 4 2 1 3 1 1 3 3 (m,0)
= 0,
(6.60)
which may also be written in terms of the non-scaled variables as

(0) Fc 2 4 4 4 2u 2u (0) u (2) u (2) u (2) u + Fs 2 + h2 Fcc + Fcs + Fss t x2 x2 x4 x2 x2 x4 1 3 1 1 3 3
= 0.
(6.61)
The dispersion relation associated with equation (6.61) matches the second order wave speed approximation (6.13) exactly.
6.2.4
Analogy with a membrane
The leading order governing equation for the mid-surface deection (6.52) is closely connected with the classic membrane equation. In order to see this we require zero normal pre-stress (2 = 0), implying that p = 21 B1221 = 23 B2332 , see (1.99), whereas the components of the in-plane Cauchy stresses are specied as pure tensions 1 = T1 , 3 = T3 . (6.62)
When these requirements are coupled with the relations (1.99) and (6.8), it is concluded that (0) (0) Fc = T1 and Fs = T3 , thus enabling (6.52) to be recast as a membrane-like equation T1 2 u(0) 2 u(0) 2 u(0) + T3 = 0, x1 2 x3 2 t2 (6.63)
with u(0) playing role of the normal deection of an innite membrane. The classic membrane df equation is usually derived with the assumption of uniform in-plane tension, i.e. T1 = T3 = T , which transforms (6.63) into the well-known form rst obtained by Poisson, see Lord Rayleigh (1877, p. 289). The second order governing equation (6.61) may also be appropriately specialised, yielding the rened membrane equation T1 2u 2u 2 u h2 + T3 2 2 + t 3 x2 x3 1 4u 4u + (T3 223 223 ) 4 x4 x3 1 4u +(T1 13 + T3 31 + 2(13 21 23 )) 2 2 = 0 . x1 x3 (T1 212 221 )
(6.64)
93
We remark that the coecients in the rened membrane equation (6.64) involve both material parameters and thickness, in addition to tensions and material density. This equation is seemingly the rst rened membrane equation derived without utilising any ad hoc assumptions. This result complements Kaplunov et al. (2000b) where a similar plain-strain model mimicked the equation for waves on an innite string.
6.3
Model problem: shock edge loading
As an illustration of the theory, we shall now consider the problem of shock edge loading of a semi-innite plate with free faces. In order to simplify the problem we focus attention to the case when the edge surface is orthogonal to the direction of wave propagation (c , 0, s ). It is then convenient to rotate the in-plane coordinate axes as follows n = c 1 + s 3 , t = s 1 + c 3 , (6.65)
in which coordinates n and t are normal and tangential to the edge n = 0, respectively. For the problem of homogeneous edge loading it is natural to assume no variation of displacement components and pressure with respect to t . Therefore, the coordinate transformation (6.65) enables the governing equation (6.60) to be expressed as F (0)
df
2 u (n , ) 2 u (n , ) 2 n 2
df (2)
+ 2 F (2)
(2)
4 u (n , ) = 0, n 4
(2)
(6.66)
with u (n , ) = u (1 , 3 , ) and F (2) = Fcc c4 + Fcs c2 s2 + Fss s4 . A surface parallel to the edge of the plate is dened by the outward normal (c , 0, s ). Thus, (1.98) gives the measure of incremental surface traction on any surface parallel to the edge in the component form 1 = ((B1111 + p )u1,1 + B1122 u2,2 + B1133 u3,3 pt )c + (B3131 u1,3 + (B3113 + p )u3,1 )s , (6.67) 2 = ((B1221 + p )u1,2 + B1212 u2,1 )c + (B3232 u2,3 + (B3223 + p )u3,2 )s , (6.68) 3 = ((B1331 + p )u1,3 + B1313 u3,1 )c + (B3311 u1,1 + B3322 u2,2 + (B3333 + p )u3,3 pt )s , (6.69) in which m , m {1, 2, 3}, is the projection of traction onto the appropriate coordinate axis. We may determine relative orders of traction components by substituting the leading order solutions for the displacement components (6.34)(6.36) and pressure (6.45) into the expressions (6.67) (6.69) to yield
1 = O() , (0) (0) 2 = Fc U2,1 c + Fs U2,3 s + O( 2 ) , (0,0) (0,0) 3 = O() .
(6.70)
This result clearly indicates that in the long wave low frequency regime the component of traction 2 is asymptotically leading. Therefore, we may utilise denitions (6.65) and (6.59) to construct the boundary condition for the governing equation (6.66) in the form
2 = F (0) df (0) (0)
u (n , ) n
at
n = 0 ,
(6.71)
where F (0) = Fc c2 + Fs s2 . Note that the assumptions made in (6.71) imply a discrepancy of O() in the traction components 1 and 3 . This discrepancy may, for example, be tackled by considering quasi-static boundary layers along the edge, however this discussion is deferred for further study.
94
6.3.1
Shock edge loading
The solution of the shock edge loading problem for the leading order governing equation (6.46) contains a discontinuity at the wave front. The fourth order derivative term in the singularly perturbed governing equation (6.60) smoothes this discontinuity. This phenomenon is known in the theory of thin structures as a quasi-front, see for example Kaplunov et al. (1998). The method of matched asymptotic expansions, see Cole (1968), provides one possible way of analysing this phenomenon. Let us consider the problem of a semi-innite plate subject to the shock edge loading df (6.72) T = 2 = M H( ) at n = 0 , in which M is the magnitude of the load and H( ) is the unit Heaviside function. The governing equation (6.66) is a classical wave equation perturbed by a term of O( 2 ), this term providing signicant contribution only when n varies rapidly. This is of particular importance within a thin boundary layer in the vicinity of quasi-front. In our approach, the solution of (6.66), termed the inner solution, is not subjected to the boundary condition (6.71). Instead, we rst analyse the boundary value problem (6.46), (6.71), with appropriately standardised variables, whose solution will be referred to as the outer solution, and then a matching condition between the inner and outer solutions compensates for the boundary condition required for the inner (boundary layer) problem. The introduction of the inner (boundary layer) coordinates in = 2/3 ( n ) , in = , (6.73)
is aimed at asymptotically balancing time and the fourth order space derivatives in the governing equation (6.66), which then takes the appropriate form 2F (0) 2 u (in , in ) 4 u (in , in ) F (2) = 0. 4 in in in (6.74)
The inner (boundary layer) traction Tin (in , in ) may now be introduced, which upon use of (6.70)2 recasts equation (6.74) as Tin (in , in ) 1 F (2) 3 Tin (in , in ) = 0, 3 in 2 F (0) in
(6.75)
where Tin (in , in ) = 2F (0) u (in , in )/in . Equation (6.75) may be solved using Fourier transforms to obtain the solution, which in terms of original scaled variables takes the form
Tin (n , ) =
1 2
TF ( 2/3 ) exp i
1 F (2) 2 3 2 F (0)
ei( n ) 2/3 d ,
(6.76)
with the function TF ( 2/3 ) to be determined. The outer problem is governed by the leading order governing equation (6.46), which in view of the coordinate rotation (6.65) and the normal traction component denition (6.70)2 , may be written as 2 Tout (n , ) 2 Tout (n , ) = 0, (6.77) 2 n 2 and solved subject to the boundary condition (6.72) given explicitly as
Tout (0, ) = M H( ) .
(6.78)
The boundary value problem (6.77), (6.78) is a standard signalling problem for a semi-innite interval, see e.g. Kevorkian (1989, p. 153). The appropriate solution is given by one of DAlembert

1.4 1.2 1 0.8 1.4 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4
95
=25
=5
T M
0.6 0.4 0.2 0 -0.2
Figure 6.1: The receding wave front proles at = 3, shown for various angles of wave propagation. The layer is composed of VM, see Table 1.1 (p. 23), and subjected to a pre-stress with 1 = 1.2, 2 = 0.9 and 2 = 5.5. It is also assumed that = 0.4.
waves propagating away from the edge, and may upon applying the boundary condition (6.78) be expressed as
Tout (n , ) = M H( n ) =
M 2
()
ei( n ) d ,
(6.79)
with () the Dirac delta function. It is remarked that the value of the improper integral on right-hand side of (6.79) is assumed to be the Cauchy principal value. We may now com pare the denitions of the inner and outer traction Tin (n , ) and Tout (n , ) to deduce that 2/3 T ( , ) = 2T ( , ). The matching procedure requires that the limit of T ( , ) as out n in n in n 0 must be equal to Tout (n , ), which gives TF ( 2/3 ) = 2M () i . (6.80)
Therefore, in view of (6.80) we may integrate (6.76) to obtain T (n , ) = M M + sgn cF + M 2 6

z0 0
Ai (z sgn cF ) dz ,
(6.81)
in which Ai is the Airy function, see Abramowitz & Stegun (1970, p. 446), and z0 = n , cF = 3 F (2) 2 . 2 F (0)
|cF |1/3
6.3.2
Typical wave proles
The numerical evaluation of the integral on the left-hand side of (6.81) can be eciently performed by using standard expansions from Abramowitz & Stegun (1970, p. 447). In order to demonstrate the behaviour of the solution (6.81), plots of typical wave proles were generated in respect of a Varga material. In contrast to the classic isotropic case, see for example Kaplunov et al. (1998), the sign of the third order derivative term in the governing equation (6.75) may change. When sgn cF < 0, the solution (6.81) generates a receding wave front, see gure 6.1. When sgn cF > 0 however, the wave front is advancing, see gure 6.2; this latter case has no analog in the classical linear isotropic theory. Figures 6.1 and 6.2 also show that the type of quasi-front may change as the propagation angle varies. This phenomenon is further

1.2 1 0.8 0.6 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5
96
=65
=45
T M
0.4 0.2 0 -0.2 -0.4
Figure 6.2: The advancing wave front proles at = 3 shown for various angles of wave propagation. The layer is composed of VM and subjected to the same pre-stress as shown in Figure 6.1.
T M
Figure 6.3: The dependence of the quasi-front type on the angle of wave propagation. The layer is composed of VM and subjected to the same pre-stress as shown in Figure 6.1.
illuminated by the diagrams presented in gure 6.3, which show the traction against space coordinate n at = 3. A polar-like system of coordinates was used with the angle equal to the angle of wave propagation and the distance from pole equal to n . It is important to note that these diagrams are not to be considered as the possible traction distributions generated by a point load, although the solution for point load problem will feature similar change of the wave type. We also note that the behaviour at the transition from one type of wave front to another can not be properly predicted by this model as it corresponds to vanishing of F (2) and, consequently, the third order derivative term in the governing equation (6.75). The corresponding values of traction were replaced on the diagrams with 0.5 to clearly distinguish these transition regions. An analysis of within the vicinity of the transition region may only be carried out by resorting to the higher order approximation and calculation of a third order governing equation.
Chapter 7
Dispersion of harmonic waves in a layer with xed faces

The nal three chapters of this thesis are concerned with the propagation of harmonic waves in a pre-stressed incompressible elastic layer with xed faces, i.e. with zero displacement boundary conditions. This type of boundary condition has particular practical signicance in the modelling of a coal layer, which is usually surrounded by much stier rock, see Liang et al. (1993). We again assume that the nite primary deformation is homogeneous, with one of its principal axes normal to the plane of the layer. An appropriate Cartesian system of coordinates is then introduced, the axes of which coincide with the principal axes of the primary deformation. The results of the analysis of the equations of motion and secular equation performed in Chapter 1 and Chapter 2 are again applicable to the current problem. Hence, we may use the representations for the displacement components obtained in Chapter 2 to formulate the boundary conditions. As with the free faces problem, these comprise of a system of six linear homogeneous equations which, by the symmetry of the layer about the mid-plane, may be decomposed into two systems of three linear homogeneous equations. The associated existence conditions are obtained in the form of symmetric and anti-symmetric dispersion relations, which are rst analysed numerically. The numerical investigation reveals an interesting characteristic of the dispersion relations for a layer with xed faces, namely the lack of any fundamental mode. This means that nite wave speed limits do not exist in the low wave number limit for either symmetric or anti-symmetric motion.
7.1
Harmonic wave motion in a pre-stressed incompressible elastic layer
We consider a nitely deformed layer of thickness 2h, composed of an incompressible elastic material. The origin O of a Cartesian coordinate system Ox1 x2 x3 is placed in the mid-plane of the layer with Ox2 orthogonal to the layer, see Figure 7.1. It is again assumed that the primary deformation has the form x1 = 1 (X1 cos + X3 sin ) , x2 = 2 X2 , x3 = 3 (X3 cos X1 sin ) , (7.1)
so that material line elements normal to the plane of the layer remain normal after the static deformation. The principal axes of the deformation (7.1) coincide with the chosen coordinate axes and the parameters m , m {1, 2, 3}, are the principal stretches of the primary deformation. Innitesimal motion of the layer is therefore governed by the linearised equations of motion (1.18)(1.20), solved in conjunction with the corresponding linearised incompressibility condition (1.21). To separate the variables we consider the propagation of harmonic waves of the specic form (1.64). The necessary and sucient condition for the existence of non-trivial solutions of this form is the secular equation (1.65). In the most general case this equation has 97
CHAPTER 7. WAVES IN A LAYER WITH FIXED FACES
98
2 2 2 three distinct non-zero roots q1 , q2 and q3 , which may be utilised to derive formal solutions for the displacement components (2.4)(2.6). We will employ these to formulate the xed faces boundary conditions.
7.2
The dispersion relations
The thickness of our layer is again chosen to be 2h and since the origin O of the Cartesian coordinate system Ox1 x2 x3 lies in the mid-plane of the layer, the xed faces (zero displacement) boundary conditions may be specied by ui
x2 =h
= 0,
i {1, 2, 3} .
(7.2)
The formal solutions for the displacement components (2.4)(2.6) may then be inserted into these boundary conditions to obtain a homogeneous system of six equations in six unknowns, which after omitting common non-vanishing factors, is given by
3 m=1 3
qm U1 (qm , v ) (2m1) (kqm h) (2m) U2 e U2 e(kqm h) = 0 , V(qm , v )
(7.3)
qm U1 (qm , v ) (2m1) (kqm h) (2m) U2 e U2 e(kqm h) = 0 , V(qm , v ) m=1

3 m=1 3 m=1
(7.4)
U2
(2m1) (kqm h)
+ U2
(2m) (kqm h)
= 0,
(7.5)
U2
3
(2m1) (kqm h)
+ U2
(2m) (kqm h)
= 0,
(7.6)
qm U3 (qm , v ) (2m1) (kqm h) (2m) U2 e U2 e(kqm h) = 0 , V(qm , v ) m=1

3 m=1
(7.7)
qm U3 (qm , v ) (2m1) (kqm h) (2m) U2 e U2 e(kqm h) = 0 . V(qm , v )
(7.8)
see denitions of U3 (q, v ), U3 (q, v ) and V(q, v ) in (2.8)(2.10). Because of the symmetry of the layer about the mid-plane, this system of six linear homogeneous equations is equivalent to two homogeneous linear systems of three equations in three unknowns. The rst of these may be x2
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
h x1 O h
h h x3
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Figure 7.1: Cartesian coordinate system for a layer with xed faces.
99
obtained by adding equations (7.4) and (7.3), subtracting (7.6) from (7.5), and adding (7.8) to (7.7), to yield
3 m=1 3 m=1 3
qm U1 (qm , v ) (m) Cm (h)U = 0 , V(qm , v ) Sm (h)U

(m)
(7.9) (7.10) (7.11)

(2m1) (2m)
= 0,
qm U3 (qm , v ) (m) Cm (h)U = 0 , V(qm , v ) m=1

df df (m) df
in which Cm (h) = cosh(kqm h), Sm (h) = sinh(kqm h) and U = U2 U2 . The second system is similarly obtained by subtracting equation (7.4) from (7.3), adding (7.5) and (7.6), and subtracting (7.8) from (7.7). The result may be written as qm U1 (qm , v ) (m) Sm (h)U+ = 0 , V(qm , v ) m=1
3 m=1 3 m=1 (m) df (2m1) (2m) 3
(7.12) (7.13) (7.14)
Cm (h)U+
(m)
= 0,
qm U3 (qm , v ) (m) Sm (h)U+ = 0 , V(qm , v )
where U+ = U2 + U2 . Thus our boundary conditions are represented as two homogeneous systems of linear equations, which are best analysed separately.
7.2.1
Symmetric motion
(m) (2m1) (2m)
When U+ = 0 then U2 = U2 , m {1, 2, 3}, and consequently all of the displacement components (except u2 ) and pressure increment are symmetric functions with regard to x2 , see (2.4)(2.7). We henceforth will refer to this type of motion as symmetric. It is clear that the boundary conditions (7.12)(7.14) are satised identically for such motion, therefore only the system (7.9)(7.11) needs to be considered. This homogeneous system of linear equations has non-trivial solution only provided the determinant of the associated matrix of the coecients is equal to zero, thus we require q1 U1 (q1 , v )C1 (h) q2 U1 (q2 , v )C2 (h) q3 U1 (q3 , v )C3 (h) V(q1 , v )S1 (h) V(q2 , v )S2 (h) V(q3 , v )S3 (h) = 0 , q1 U3 (q1 , v )C1 (h) q2 U3 (q2 , v )C2 (h) q3 U3 (q3 , v )C3 (h) (7.15)
within which common non-zero factors were omitted. Evaluating the determinant in (7.15) leads to an explicit representation of the dispersion relation for symmetric waves, which after dividing by C1 (h)C2 (h)C3 (h) may be written as
2 2 H() q2 q3 (q2 q3 )V(q1 , v )T1 (h) v
2 2 2 2 q1 q3 (q1 q3 )V(q2 , v )T2 (h) + q1 q2 (q1 q2 )V(q3 , v )T3 (h) = 0 ,
(7.16)
Considering a layer with free faces in Chapter 2 we referred to this type of motion as extensional or symmetric. Further investigation will however demonstrate that the term extensional does not seem applicable to symmetric motion in a layer with xed faces and will therefore not be used.
CHAPTER 7. WAVES IN A LAYER WITH FIXED FACES where Tm (h) = tanh(kqm h), m {1, 2, 3}, and H() = (23 21 )2 + (23 23 + 13 21 )c2 (21 12 + 23 31 )s2 , v v
100
(7.17)
It is possible to show that the factor H() is associated with a spurious (non-dispersive) wave, v see (2.16)2 and the relevant discussion in Chapter 2. Hereafter, we shall assume that the symmetric dispersion relation for a layer with xed faces is only that part of (7.16) within the braces. The displacement components and incremental pressure associated with symmetric motion may be determined using relations (2.43)(2.46). When the dispersion relation (7.16) is satised, the system (7.9)(7.11) possesses a non-trivial solution. We may then insert the displacement components and pressure increment (2.43)(2.46) into the boundary conditions (7.9) and (7.11) (m) to derive representations of the coecients U , m {1, 2, 3}, in terms of the single parameter U , yielding U = (1)k
(k) 2 2 (qi qj )V(qk , v )
i <j,
qk Ck (h) k {i, j} , /
U , i, j, k {1, 2, 3} .
(7.18)
As in the free faces case, this will be used to establish the relative asymptotic orders of the displacements in the long wave regime.
7.2.2
Anti-symmetric motion
(m) (2m1) (2m)
If we assume that U = 0 the implication is that U2 = U2 , m {1, 2, 3}. Two of the displacement components u1 and u3 (as well as pressure increment pt ) are then anti-symmetric functions of x2 , see (2.4)(2.7). This type of motion is termed as anti-symmetric and the system of the boundary conditions (7.9)(7.11) is satised identically for such motion. We therefore consider only the homogeneous system of linear equations (7.12)(7.14), which only has nontrivial solution provided the determinant of the matrix of associated coecients is equal to zero. This condition may be written as q1 U1 (q1 , v )S1 (h) q2 U1 (q2 , v )S2 (h) q3 U1 (q3 , v )S3 (h) V(q1 , v )C1 (h) V(q2 , v )C2 (h) V(q3 , v )C3 (h) = 0 , q1 U3 (q1 , v )S1 (h) q2 U3 (q2 , v )S2 (h) q3 U3 (q3 , v )S3 (h) (7.19)
where we omitted common non-dispersive factors. It is now possible to evaluate the determinant (7.19), and after dividing it by C1 (h)C2 (h)C3 (h) and some algebraic transforming it may be recast as
2 2 q2 q3 (q2 q3 )V(q1 , v )T2 (h)T3 (h)
2 2 2 2 q1 q3 (q1 q3 )V(q2 , v )T1 (h)T3 (h) + q1 q2 (q1 q2 )V(q3 , v )T1 (h)T2 (h) = 0 ,
(7.20)
in which the non-dispersive factor H() was omitted, see (2.16)2 . v The relations (2.50)(2.53) describe the displacement components and incremental pressure (m) associated with anti-symmetric motion in terms of three parameters U+ , m {1, 2, 3}. However, when the dispersion relation (7.20) is satised, the boundary conditions (7.12) and (7.14) may be used to represent all of these parameters in terms of the single parameter U+ , given by U+ = (1)k
(k) 2 2 (qi qj )V(qk , v ) U+ , qk Sk (h) i <j, k {i, j} , / i, j, k {1, 2, 3} .
(7.21)
101
7.3
Propagation of long waves

df
Figure 7.2 depicts typical solutions of the symmetric dispersion relation (7.16). The scaled wave speed v and the scaled frequency = kh of the rst twenty solution branches are shown v against the scaled wave number kh for a Mooney-Rivlin material. A particularly striking feature
4.5 4
90 80 70
3.5
60
3 2.5 2
50 40 30 20 10
1.5 5 10 15 20 25 30 35 40 45 50
0 0 5 10 15 20 25 30
kh
kh
Figure 7.2: Scaled wave speed v and frequency against scaled wave number kh for symmetric motion in the MRM with the angle of propagation = 45 . The pre-stressed state is specialised to 1 = 0.9 and 2 = 1.4.
of both plots is the lack of any fundamental mode, which reects the impossibility of surface wave propagation in a layer with xed faces and the lack of any bending or extensional deformation. It is worth remarking, that in contrast with the problem of wave propagation in a layer with free faces, the lack of fundamental mode does not correspond to a layer instability. Actually, the slowest wave to propagate in a layer with xed faces is associated with the short wave limit of harmonics, as depicted on the left hand plot in Figure 7.2. Since the only type of body waves to propagate in an incompressible solid are shear waves, the stability of the layer with xed faces is then guaranteed by the strong ellipticity conditions (1.36), (1.45) and (1.59). On the other hand, the long wave high frequency behaviour of the symmetric dispersion relation is generally similar to the case of layer with free faces. The scaled wave speed of the branches tend to innity as the scaled wave number kh decreases, whereas the scaled frequencies tend to nite limits, so-called cut-o frequencies. The only point worth noting is the apparent non-symmetry in the two families of cut-o frequencies, the rst branch on the right hand plot in Figure 7.2 does not seemingly have a corresponding branch with a similar cut-o frequency from another family. This point will be further discussed in the next chapter. The plots in Figure 7.3 demonstrate the scaled wave speed and scaled frequency of the rst twenty anti-symmetric branches against scaled wave number. There is again no fundamental mode on both plots, which means that the only type of long wave motion possible in a layer with xed faces is long wave high frequency motion. In a similar way to the free faces problem, the existence of two families of cut-o frequencies is apparent from the right hand plot in Figure 7.3. The branches within one family are separated by an apparently constant interval and these two intervals are similar for the specied material parameters. This gives the impression that couples of the solution branches (especially the lowest ones) have the same cut-o frequencies. It is also worth remarking, that some of the branches have negative group velocity despite the fact that Figure 7.3 depicts the wave propagation in a Mooney-Rivlin material. This contrasts with the results of the analysis of a layer with free faces, for which the impossibility of negative group velocity in Mooney-Rivlin materials was demonstrated, see (3.88) and the associated discussion in Chapter 3. We nally note that apart from the lacking fundamental modes, most of the phenomena previously observed in a layer with free faces remain intact in a layer with xed faces. The plots

4
102
80 70 60
3.5
50
v
2.5
40
30 20 10
1.5 5 10 15 20 25 30 35 40 45 50
0 0 5 10 15 20 25 30
kh
kh
Figure 7.3: Scaled wave speed v and frequency against scaled wave number kh for anti-symmetric motion in the MRM with the angle of propagation = 45 . The pre-stressed state is specialised to 1 = 0.9 and 2 = 1.4.
3.9 3.8 3.7 3.9 3.8 3.7
v 3.6
3.5 3.4 3.3 5 10 15 20 25 30 35 40 45 50
v 3.6
3.5 3.4 3.3 5 10 15 20 25 30 35 40 45 50
(a) symmetric motion, kh
(b) anti-symmetric motion, kh
Figure 7.4: Scaled wave speed v against scaled wave number kh for (a) symmetric and (b) anti-symmetric motion in the OM1 with the angle of propagation = 15 . The pre-stressed state is specialised to 1 = 2.0 and 2 = 0.7.
in Figure 7.4 depict the scaled speeds of solutions of the symmetric and anti-symmetric dispersion relations for a single term Ogden material. These plots may be compared with Figure 2.7 and Figure 2.8 to conclude that even complicated short wave behaviour does not change with the change in boundary conditions.
Chapter 8
Layer with xed faces: Long wave symmetric motion

Long wave high frequency symmetric motion in a layer with xed faces deviates signicantly from other long wave high frequency models considered in this thesis. An appropriate approximation of the dispersion relation is rst obtained and it indicates the absence of two distinct families of cut-o frequencies. The cut-o frequencies themselves are shown to be solutions of a transcendental equation and depend on the direction of wave propagation, which is highly unusual. An asymptotic analysis of this equation reveals both a connection with the shear resonance frequencies of a layer with free faces as well as estimates of cut-o frequencies for high harmonic numbers. A second order frequency correction is then obtained in terms of the associated cut-o frequency. We use the resulting frequency expansion to determine the relative orders of the displacement components and introduce appropriate scalings to derive asymptotically approximate equations. Asymptotic solutions for all of the displacement components and pressure increment are then derived in terms of a single function, termed the long wave pressure. This function is shown to be the solution of a two-dimensional governing equation, obtained as an existence condition in the process of the derivation of the asymptotic solution. The consistency of the approach is again established by comparing the dispersion relation of the asymptotically approximate governing equation with the appropriate approximation of the exact dispersion relation. The discussion is concluded with some illustrative numerical examples.
8.1
The dispersion relation in long wave regime
To perform an analysis of long wave high frequency symmetric motion in a pre-stressed incompressible elastic layer with xed faces we rst need to obtain an appropriate asymptotic representation of the associated dispersion relation. The numerical investigation in Chapter 7 indicated that the scaled wave speed v , as 0, whereas the scaled frequency O(1). The approximations (2.67)(2.69) may therefore be used to establish the second order approximation of the dispersion relation (7.16) in the following form i A1 v 2 + A1
(2) (0)
T1 (h) + v A2 v 2 + A2
(3)
(1)
T2 (h) + i A3 v 2 + A3
(2)
(0)
T3 (h) 0 ,
(8.1)
within which the functions of material parameters and pre-stress Ak , k {1, 2, 3}, are all assumed to be O(1). Specically, the leading order coecients of the dispersion relation (8.1) are given by A1 =
(2)
(m)
(21 23 )c2 , 23
21 23 (3) A2 = , 21 23
A3 =
(2)
(21 23 )s2 , 21
(8.2)
103
CHAPTER 8. LONG WAVE SYMMETRIC MOTION and the second order coecients may be represented as A1 =
(0)
104
1 (0) (0) (0) (0) (23 21 ) c5 + Q1c + 2Q3c + 12 c2 + Q1s + 2Q3s + 32 s2 223 2 221 13 c2 + (213 31 )s2 Q1c
(0)
c2 23
1/2
(8.3)
A2 =
(1)
1 (0) (0) (0) (0) (23 21 ) 2c5 + Q1c + Q3c 223 c2 + Q1s + Q3s 221 s2 2 + 223 Q1c c2 + Q1s s2 221 Q3c c2 + Q3s s2
(0) (0) (0) (0)
21
1/2 1/2 23 ,
(8.4)
A3 =
(0)
1 (0) (0) (0) (0) (23 21 ) c5 + 2Q1c + Q3c + 12 c2 + 2Q1s + Q3s + 32 s2 221 2 + 223 (213 13 )c2 + 31 s2 Q3s
(0)
s2 21
1/2
,
(0) (0)
(8.5)
with the parameter c5 given immediately after (1.65) and the functions Qmc and Qms , m {1, 3}, dened in (2.70). We have previously shown that when the wave speed is large, q2 = 1 + O(2 ), see (2.72). Thus, T1 (h) v T2 (h) T3 (h) and all three terms on the left-hand side v of the relation (8.1) have the same asymptotic order. It is not therefore possible to distinguish two families of cut-o frequencies, which contrasts with our previous long wave high frequency analysis for a layer with free faces performed in Chapter 3 and Chapter 4. In the long wave high frequency regime, solutions of the secular equation (1.65) may be expanded into power series, see (2.71)(2.73). Inserting these expansions into the denitions of the functions Tm (h) = tanh(kqm h), m {1, 2, 3}, and using v , we establish the following asymptotic approximations T1 (h) = i tan
(0)
21
(0)
i 2
Q1c c2 + Q1s s2 sec2 2 21

(0) (0)
(0)
(0)
21
+ O( 4 ) ,
(8.6)
T2 (h) = + 3
Q3c + Q1c 12 c2 + Q1s + Q3s 32 s2 c5 1 2 2 3 23 i 2 Q3c c2 + Q3s s2 sec2 2 23

(0) (0)
+ O( 5 ) ,
(8.7)
T3 (h) = i tan
23
+ O( 4 ) ,
df
(8.8)
within which we re-introduced the small non-dimensional parameter = h/l = kh, where h is the plate half-thickness and l is a typical wave length. It is now assumed that scaled frequency possesses a power series representation = e + (2) 2 + O( 4 ) , (8.9)
in which parameter e plays role of (scaled) cut-o frequency and is assumed to be O(1). Expansion (8.9), together with the expansions (8.6)(8.8), may be substituted into the second order approximation of the dispersion relation (8.1) to obtain an asymptotic equality, which must be satised at various orders.
8.1.1
Leading order approximation
At the leading order we obtain the following transcendental equation for e e(0) =
df
21 tan
e 21
e 21
c2 +
23 tan
e 23
e 23
s2 = 0 .
(8.10)
CHAPTER 8. LONG WAVE SYMMETRIC MOTION

25 20 15 10 5
105
e(0)
0 -5 -10 -15 -20 -25 0 5 10 15 20
Figure 8.1: First six solutions of the equation (8.10) (solid lines) along with the discontinuities of tangents (dotted lines) shown for the same set of material parameters as used in Figure 7.2.
This form of representing the leading order dispersion relation is chosen to reect the connection with the analogous plane strain problem analysed by Nolde & Rogerson (2001). We note that the leading order dispersion relation in the plane strain case is given by tan e = e . It is worth remarking that the occurrence of a transcendental equation to dene the cut-o frequency is unusual but not new, see Kaplunov & Nolde (2001) and Nolde & Rogerson (2001). However, the dependence of cut-o frequency on the angle of wave propagation, apparent from (8.10), is novel, although the physical signicance of this phenomenon is not clear. The transcendental equation (8.10) has no analytical solutions except e = 0. However, this solution must be disregarded because our approximations were obtained with the assumption of large wave speed and nite frequency, whereas e = 0 corresponds to both zero frequency and zero wave speed. The non-zero solutions of the leading order dispersion relation (8.10) can only be obtained numerically, for example using the method of bisection. Figure 8.1 demonstrates typical behaviour of the function e(0) , dened in (8.10). It is easy to see that the behaviour is dominated by the tangents occurring in e(0) . Specically, every non-zero solution of (8.10) is e e included in an interval between singularities of tan 21 or tan 23 , shown as dotted lines df in Figure 8.1. We previously dened the frequencies of these singularities as f = 2m ( 1 +n), m 2 m {1, 3}, namely the two families of anti-symmetric shear resonance frequencies of a layer with free faces. As the harmonic number, and hence the corresponding frequency, increases, larger tangent values are required to compensate for the (e ) term in (8.10). Therefore, the symmetric cut-o frequencies e of a layer with xed faces tend from the left to the exural shear resonance frequencies f or f of a layer with free faces, as n . 1 3 The observations above may be used to conclude that for large harmonic numbers, the cut-o frequencies e 1 and they may be represented as e = f 1 or e = f , 3 (8.11)
within which the positive parameter 1. If the rst equality in (8.11) holds, then the tangents occurring in the equation (8.10) may be written as tan e 21 = cot 21 , (8.12)
This property of non-zero solutions of (8.10) may be used to distinguish between two families of cut-o frequencies, dened as those associated with f or with f . It also explains the apparent non-symmetry in two 1 3 families of cut-o frequencies, shown in Figure 7.2, which is related to the lack of solution of (8.10) between zero and the rst singularity of one of the tangents guring in e(0) .
CHAPTER 8. LONG WAVE SYMMETRIC MOTION e13 tan 1 + e13 tan 23 23
106
tan
e 23
e13 = tan
df
f 1 23
(8.13)
within which we tacitly assume that e13 is O(1). Generally, the parameters and are not connected and can not be compared. However, in view of the fact that solutions of equation (8.10) are constant and independent of , we choose a small value of and assume that = O(). The relations (8.12) and (8.13) may then be used to establish the asymptotic representation for equation (8.10), given by 21 c2 1 f + 1 23 e13 s2 + 2 2 c e2 s2 + O(2 ) = 0 . 13 3 (8.14)
Considering the expression on the left-hand side of this equality as a quadratic in , it is now possible to choose a small positive value of which will satisfy three orders of the asymptotic equality (8.14). The result may be substituted back into equation (8.11)1 to obtain the following cut-o frequency approximation
e
f 1
f 3 23 e13 s2 1 + + 2 2c2 3e2 s2 13
9 4
23 e13 s2 f 1 (2c2 3e2 s2 )2 13
321 c2 . 2c2 3e2 s2 13
(8.15)
It is worth reiterating that this approximation is only valid for large harmonic numbers and can not therefore be used as a general replacement for a numerical routine to solve (8.10). The same procedure can be repeated for cut-o frequencies associated with f , to obtain 3 where e31 = tan
df e
f 3
f 3 21 e31 c2 3 + + 2 2s2 3e2 c2 31
9 4
21 e31 c2 f 3 (2s2 3e2 c2 )2 31 f 3 21 .
323 s2 , 2s2 3e2 c2 31
(8.16)
(8.17)
Although approximations (8.15) and (8.16) were derived with the assumption that both the harmonic number and associated cut-o frequency are large, they give quite good approximation even for very low harmonic numbers. Shown in Table 8.1 (p. 107) are the numerically obtained cut-o frequencies e , associated with rst ve harmonics for each family, along with the cor responding approximate solutions e obtained from (8.15) and (8.16) for a Varga material with df the same set of material parameters as in Figure 7.2 and Figure 8.1. The absolute a = |e e |
df and the relative r = |e e |/e errors, also shown in the table, never exceed 0.2 or 2%, respectively. The high precision of these approximations is not however surprising in view of the fact, that even the lowest cut-o frequency, in this case, e = 4.282, is at least twice the e e magnitude of tan 21 or tan 23 on 72% of their domain.
8.1.2
The second order term of the expansion, derived by inserting relations (8.6)(8.8) and (8.9) into (8.1), is now considered. As the leading order solution e of the dispersion relation (8.10) can not be obtained analytically, the second order correction can only be represented in terms of the leading order solution. We therefore make use of the equalities (1.66)(1.68) and relation (8.10)
CHAPTER 8. LONG WAVE SYMMETRIC MOTION e 4.282 12.887 21.5619 30.25787 38.95182 e 4.271 12.879 21.5608 30.25792 38.95195 a 1.2102 8.5103 1.1103 5.4105 1.2104 r 2.7103 6.6104 5.1105 1.8106 3.2106 e 12.21 20.907 29.44009 37.92975 46.4033 e 12.37 20.916 29.44082 37.92986 46.40332 a 1.6101 8.9103 7.3104 1.1104 2.5105 r 1.3102 4.3104 2.5105 2.9106 5.3107
107
Table 8.1: Exact e and approximate e solutions of the leading order dispersion relation (8.10), along with absolute a and relative r errors of the approximation. The left-hand table uses asymptotics (8.15) and the right-hand table uses asymptotics (8.16). The material parameters are the same as in Figure 7.2 and Figure 8.1.
to establish that the second order term may be written as 2e tan2 e 21 c2 + tan2 e 23 2 s2 (2) = (e )2 3
e e (0) (0) (0) (0) + Q3c c2 + Q3s s2 s2 tan2 + Q1c c2 + Q1s s2 c2 tan2 21 23 e e 21 (2) 23 (2) (2) (2) E1 tan E1 + E3 tan E3 c2 s2 , e e 21 23 within which the functions of material parameters E1 E1 E3
(0) (2) (0) (0) (2)
(8.18)
and E3
(2)
are given by (8.19) (8.20)
(2)
21 (13 + 21 ) , 21 23 23 (13 + 23 ) (0) (0) = 3Q3c 3Q1c + 221 + 4 , 21 23 = 3Q1s 3Q3s + 223 4
and parameters Qmc and Qms , m {1, 3}, are dened in (2.70). We note that the relation (8.18) indicates the possibility of the existence of negative group velocity in a layer with xed faces. (0) (0) Indeed, when Q1c < 0 or Q3s < 0 it is always possible to choose an angle such that (2) < 0 2 e )2 may even and the corresponding group velocity is negative. The large negative term 3 ( enable negative group velocity in materials with positive Q1c and Q3s , for example in MooneyRivlin materials, which is not possible in a layer with free faces.
(0) (0)
(0)
8.1.3
Relative orders of displacements, physical interpretation
With the help of the relations (7.18), the displacement components and incremental pressure (2.43)(2.46) may be established in terms of a single constant U . We use these representations to obtain the relative asymptotic orders of displacement components, an essential component in derivation of asymptotically approximate equations. In view of the fact that in the long wave high frequency regime = O(1), as 0, the trigonometric functions occurring in (2.43)(2.46) and (7.18) possess the following asymptotic representations C1 (x2 ) = cos S1 (x2 ) = i sin x2 21 h x2 21 h + O( 2 ) , + O( 2 ) , C3 (x2 ) = cos S3 (x2 ) = i sin S2 (x2 ) = x2 23 h x2 23 h + O( 2 ) , + O( 2 ) , (8.21)
(8.22) (8.23)
C2 (x2 ) = 1 + O( 2 ) ,
x2 + O( 3 ) , h
108
in which we tacitly assume that x2 /h = O(1). Inserting these expansions into the expressions (2.43)(2.46), and making use of relations (7.18), it is possible to obtain that u1 2 O(pt ) , u2 3 O(pt ) , u3 2 O(pt ) . (8.24)
The long wave motion in an incompressible elastic layer with xed faces is very limited, which is indicated by (8.24). The displacements are negligible in comparison with incremental pressure. However, to leading order, the wave normal is parallel to a vector of the form (c , O( 1 ), s ), see (2.71) and (2.73), and the polarisation direction is in the plane generated by vectors of the form (O(1), O(), O(1)). We thus conclude that long wave high frequency motion in a layer with xed faces is again a rapid motion of shear waves along the normal to the plane of a layer. However, physical peculiarities of motion in a layer with xed faces, specically the lack of bending or symmetric deformation and the dependence of cut-o frequency on the direction of wave propagation, preclude us from associating this type of motion with thickness shear resonance.
8.2
We begin the procedure of deriving asymptotically approximate equations by choosing scales of space variables, which will balance in-plane coordinates x1 and x3 with a typical wave length l and normal coordinate x2 with a layer half-thickness h. The scales are therefore chosen as x1 = l1 , x2 = h = l , x3 = l3 . (8.25)
Since the leading order scaled frequency of long wave motion in a layer with xed faces e is generally of O(1) and scaled wave speed v v 1 , a typical wave travels a distance of a wave length l in time t = l /e , thus we introduce scaled non-dimensional time through t = l . (8.26) Finally we rescale the displacement components according to the known distribution of the relative orders of the displacements (8.24), to obtain um (x1 , x2 , x3 , t) = l 2 u (1 , , 3 , ) , m u2 (x1 , x2 , x3 , t) = l 3 u (1 , , 3 , ) , 2 m {1, 3} , (8.27)
pt (x1 , x2 , x3 , t) = p (1 , , 3 , ) , t
where the superscript denotes non-dimensional quantities of the same asymptotic order. Relations (8.25), (8.26) and (8.27) may now be utilised to reformulate the linearised equations of motion (1.18)(1.20) in terms of the rescaled non-dimensional variables, yielding 21 u + (e )2 u (e )2 u + u p 1 + 2 B1111 u 1 1 1 1 1, 1, t, 1,
+31 u 3 3 + (B1122 + B1221 )u 1 + (B1133 + B1331 )u 1 3 = 0 , 1, 2, 3,
(8.28)
p 2 B2222 u + (e )2 u (e )2 u + u + (B1122 + B1221 )u 1 t, 2, 2 2 2, 1,
+(B2233 + B2332 )u 3 4 12 u 1 1 + 32 u 3 3 = 0 , 3, 2, 2,
(8.29)
23 u + (e )2 u (e )2 u + u p 3 + 2 (B1133 + B1331 )u 1 3 3, 3 3 3, t, 1,
+(B2233 + B2332 )u 3 + 13 u 1 1 + B3333 u 3 3 = 0 , 2, 3, 3,
(8.30)
which must be solved in conjunction with the appropriately rescaled linearised incompressibility condition (8.31) u 1 + u + u 3 = 0 , 1, 2, 3,
CHAPTER 8. LONG WAVE SYMMETRIC MOTION subject to the rescaled zero displacement boundary conditions u (1 , , 3 , ) = u (1 , , 3 , ) = u (1 , , 3 , ) = 0 1 2 3 at = 1 .
109
(8.32)
The solution of the boundary value problem given by the equations of motion (8.28)(8.30) and incompressibility condition (8.31), subject to boundary conditions (8.32), is sought in the form of a power series
m
(u , u , u , p ) 1 2 3 t
=
n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(8.33)
In order to produce response which is compatible with the properties of long wave motion in a layer with xed faces already observed in previous sections, we require 21 u + (e )2 u 2 u , m, m m m {1, 2, 3} , (8.34)
where scaled frequency e is assumed to be a solution of the leading order dispersion relation (8.10). We note in passing that the validity of the asymptotic equalities (8.34) may be veried by the direct substitution of the wave solution (1.64) and subsequent comparison with the scaled frequency expansion (8.9). By inserting solutions (8.33) into the equations (8.28) (8.32) we obtain a hierarchical system of boundary value problems at various orders. The asymptotic equalities (8.34) imply that quantities in braces in (8.28)(8.30) are, actually, O( 2 ), which must be taken into consideration when separating asymptotic problems at various orders in (8.28)(8.32).
8.2.1
At the leading order we consider the appropriate equations of motion 21 u1, + (e )2 u1

(0) (0) (0)
pt,1 = 0 ,
(0)
(8.35) (8.36)
pt, = 0 , 23 u3, + (e )2 u3 subject to the incompressibility condition u1,1 + u2, + u3,3 = 0 , and the leading order xed faces boundary conditions u1
(0) (0) (0) (0) (0) (0)
pt,3 = 0 ,
(0)
(8.37)
(8.38)
= u2
(0)
= u3
(0)
=0
at
= 1 .
(8.39)
The general solution of equation (8.36) has the form pt

(0)
= Pt
(0,0)
(8.40)
within which we keep the agreement to use double superscript for functions independent of , with the rst superscript referring to the order of approximation and the second denoting the degree of any possible m factor. In view of (8.40), the solution of the boundary value problem (8.35), (8.39)1 may be given by u1
(0)
= u1 u1
(0,0)
cos
e 21
(0,0)
+ U1
(0,0)
, , U1
(0,0)
(8.41) = Pt,1
(0,0)
(0,0)
Pt,1
(e )2
sec
e 21
(e )2
110
and, by the same argument, the solution of the boundary value problem (8.37), (8.39)3 may be represented as u3
(0)
= u3 u3
(0,0)
cos
e 23
(0,0)
+ U3
(0,0)
, , U3
(0,0)
(8.42) = Pt,3
(0,0)
(0,0)
Pt,3
(e )2
sec
e 23
(e )2
It must be kept in mind that for any symmetric motion in a pre-stressed incompressible elastic layer with xed faces u , u and p vary as symmetric functions and u varies as an antit 1 3 2 symmetric function of the normal coordinate . It can be demonstrated that this property of response is ensured by the relations (8.10), (8.26) and (8.34), however for the sake of brevity we will simply omit terms which do not satisfy the symmetry requirements, as we did in (8.41) and (8.42). Solutions (8.41) and (8.42) may now be inserted in the incompressibility condition, (0) which we use to establish a general solution for u2 in the form u2 where
(0,0) u2 (0)
= u2
(0,0)
sin
e 21
+ v2
(0,0)
sin
e 23
+ U2
(0,1)
(8.43)
21 Pt,1 1 (e )3
(0,0)
sec
e 21
(0,1) U2
(0,0) v2 (0,0)
23 Pt,3 3 (e )3
(0,0)
sec
e 23
(8.44)
Pt,3 3 + Pt,1 1 (e )2
(0,0)
(8.45)
The only relation left to consider at the leading order is the second boundary condition (8.39)2 . Substituting all known functions into it we obtain the following equality 21 tan e 21 e Pt,1 1 +
(0,0)
23 tan
e 23
e Pt,3 3 = 0 ,
(0,0)
(8.46)
which is connected with the leading order approximation of the dispersion relation (8.10). Keeping in mind our assumption that e is a solution of (8.10), it is possible to demonstrate that equality (8.46) is equivalent to (0,0) (0,0) Pt,1 1 s2 = Pt,3 3 c2 . (8.47) This condition is automatically satised for any plane wave propagating along (c , 0, s ) or any other direction whose projection on the plane of a layer is parallel to (c , 0, s ). This is in particular true for the type of harmonic waves we consider, see (1.64). (0,0) All of the leading order displacements are now obtained in terms of the single function Pt , which is the leading order pressure. We term this function as the long wave pressure and remark that a governing equation for it can only be determined at the higher order.
8.2.2
In contrast to other asymptotic models presented in this thesis, the derivation of the leading order governing equation for long wave motion in a layer with xed faces requires considering all of the equations of motion and boundary conditions. The second order problem is given by the appropriate equations of motion 21 u1, + (e )2 u1
(2) (2)
= pt,1 B1111 u1,1 1 31 u1,3 3

(0) (0)
(2)
(0)
(0)
(B1122 + B1221 )u2,1 (B1133 + B1331 )u3,1 3 + 2 u1, + (e )2 u1
(0)
(0)
(8.48)
CHAPTER 8. LONG WAVE SYMMETRIC MOTION pt, = (B1122 + B1221 )u1,1 + B2222 u2, + (e )2 u2 23 u3, + (e )2 u3
(2) (2) (2) (0) (0) (0)
111 + (B2233 + B2332 )u3,3 ,

(0)
(8.49)
= pt,3 (B1133 + B1331 )u1,1 3

(0) (0) (0)
(2)
(0)
(B2233 + B2332 )u2,3 13 u3,1 1 B3333 u3,3 3 + 2 u3, + (e )2 u3
(0)
(0)
(8.50)
in which we remark that quantities in braces are of the same asymptotic order as other terms, see (8.34). Equations (8.48)(8.50) must be solved in conjunction with the second order incompressibility condition (2) (2) (2) (8.51) u1,1 + u2, + u3,3 = 0 , and second order boundary conditions u1
(2)
= u2
(2)
= u3
(2)
=0
at
= 1 .
(8.52)
It is worth noting that the equations of motion (8.48)(8.50) are written in such form that terms on the right hand side are all known at the time we use any of these equations. It is therefore easy to see that the general form of the solution for the second equation of motion is given by pt within which pt
(2,0) (2)
= pt
(2,0)
cos
e 21
+ pt
(2,0)
cos
e 23
(2,0)
+ 2 Pt
(2,2)
+ Pt
(2,0)
(8.53)
(b21 21 )Pt,1 1 (e )2 cos Pt

(2,2) e 21
(0,0)
pt
(b23 23 )Pt,3 3 (e )2 cos

e 23
(0,0)
(8.54) (8.55)
1 (0,0) (0,0) Pt,1 1 + Pt,3 3 . 2 In view of (8.53), the rst equation of motion (8.48) possesses the solution = u1
(2)
= u1
(2,0)
cos
e 21
+ u1
(2,1)
sin
e 21
+ v1
(2,0)
cos
e 23 + U1
(2,0)
+ 2 U1 where u1 v1
(2,0) (2,1) (0,0)
(2,2)
(8.56)
e1
(2,0)
e 1 2 21
(2,2)
sec , U1
e 21 =
, Pt,1 3 3 + Pt,1 1 1 2(e )2 e1

(2,0) (0,0) (0,0)
(8.57) , (8.58) (8.59)
23 (13 + 23 )Pt,1 3 3 (e )4 (21 23 ) = Pt,1

(2,0)
sec +
e 23
(2,2)
(2,0) U1
(e )2
(0)
b21 Pt,1 1 1 (e )4
(0)
(0,0)
(21 b12 + b13 )Pt,1 3 3 (e )4
(0,0)
e1 e
(2,2)
in which e1
(2,0)
Q1c Pt,1 1 1 + Q1s Pt,1 3 3 (e )3
(0,0)
(0,0)
e1
(2,2)
Pt,1 + (e )2 Pt,1 2 (e )3
(2,0)
(0,0)
(0,0)
(8.60)
Inserting these results into the rst boundary condition (8.52)1 we may conclude that
(2,0) u1
Pt,1
(2,0)
(e )2
1 tan 2
e 21
21 e
e1
e1 21
(2,2)
1 b12 b13 21 23 (13 + 23 ) + e 4 2(e )2 (e )4 ( ) (21 23 ) 1 b21 e 4 e )2 2( ( ) Pt,1 1 1

(0,0)
Pt,1 3 3 (8.61)
(0,0)
sec
e 21
112
When the solution (8.53) is substituted into the third equation of motion (8.50), the general (2) solution for u3 is established, yielding u3
(2)
= u3
(2,0)
cos
e 23
+ u3
(2,1)
sin
e 23
+ v3
(2,0)
cos
e 21 + U3
(2,0)
+ 2 U3 Equation (8.50) may then be used to obtain u3

(2,0) v3 (2,1) (2,0) (2,2)
(2,2)
(8.62)
e3
e 3 2 23
sec
e 23
,
(0,0) (0,0)
(8.63)
21 (13 + 21 )Pt,1 1 3 (e )4 (23 21 ) = Pt,3

(2,0)
(0,0)
sec
e 21
(0,0)
(2,2) U3
1 Pt,3 3 3 + Pt,1 1 3 = , 2 (e )2 e3
(2,0)
(8.64) (8.65)
U3
(2,0)
(e )2
(0)
(b31 b32 + 23 )Pt,1 1 3 (e )4
b23 Pt,3 3 3 (e )4
(0,0)
e3 e
(2,2)
within which
(2,0) e3
Q3c Pt,1 1 3 + Q3s Pt,3 3 3 (e )3
(0,0)
(0)
(0,0)
(2,2) e3
Pt,3 + (e )2 Pt,3 2 (e )3
(0,0)
(0,0)
(8.66)
The last unknown function in (8.62) may be derived with the help of the third boundary condition (8.52)3 and is given by
(2,0) u3
Pt,3
(2,0)
(e )2
1 tan 2
e 23
23 e
e3
(2,0)
e3 23
(2,2)
1 b32 b31 23 21 (13 + 21 ) + + e 4 e )2 e )4 2( ( ( ) (21 23 ) 1 b23 e 4 e )2 2( ( ) Pt,3 3 3

(0,0)
Pt,1 1 3 (8.67)
(0,0)
sec
e 23
The solutions (8.56) and (8.62) may now be inserted into the incompressibility condition (8.51). (2) The general solution of the resulting dierential equation for u2 has the following form u2
(2)
= u2 + v2
(2,0)
sin sin
(2,0)
e e (2,1) + u2 cos 21 21 e e (2,1) (2,3) (2,1) + v2 cos + 3 U2 + U2 . 23 23
(8.68)
A series of algebraically complex derivations, aimed at satisfying equation (8.51) with the solution (8.68), yields
(2,1) u2
e1,1 e1,1 2e
(2,0)
(2,2)
sec
(2,0)
e 21
(2,1) v2
e3,3 e3,3 2e
(2,0)
(2,2)
sec
e 23
(8.69)
(2,0) u2
Pt,1 1 (2b21 (e )2 )Pt,1 1 1 1 1 = e 21 + (e )2 2(e )4 Pt,1 1 3 3 (8.70)

(0,0)
(0,0)
1 21 b12 + b13 21 (13 + 21 ) 23 (13 + 23 ) e 4 e 4 e )2 e )4 2( ( ( ) (21 23 ) ( ) (21 23 ) 3 21 1 e e (2,0) (2,2) + tan e1,1 e1,1 sec , 2 21 e 21
CHAPTER 8. LONG WAVE SYMMETRIC MOTION Pt,3 3 (2b23 (e )2 )Pt,3 3 3 3 1 = e 23 + (e )2 2(e )4 Pt,1 1 3 3
(0,0) (2,0) (0,0)
113
(2,0) v2
1 23 + b31 b32 21 (13 + 21 ) 23 (13 + 23 ) + e 4 + e 4 e )2 e )4 2( ( ( ) (21 23 ) ( ) (21 23 ) 23 e 1 e (2,0) (2,2) tan sec + 3 e e3,3 e3,3 , 2 23 23 U2
(2,3)
(8.71)
1 Pt,1 1 1 1 + 2Pt,1 1 3 3 + Pt,3 3 3 3 , 6 (e )2 U2

(2,1)
(0,0)
(0,0)
(0,0)
(8.72)
and
(2,0)
= U1,1
(2,0)
U3,3
(2,0)
(8.73)
in which functions U1 and U3 were specied in (8.59) and (8.65), respectively. Finally, we need to ensure that the second boundary condition (8.52)2 is satised. Inserting the solution for normal displacement component (8.68) into (8.52)2 and rearranging terms with the help of equality (8.46), we establish that if Pt,1 1 s2 = Pt,3 3 c2 , the condition (8.52)2 can only be satised provided tan2 e 21
(0) (2,0) (2,0)
(2,0)
(8.74)
Pt,1 1 + (e )2 Pt,1 1 e 21 e 21
(0,0)
(0,0)
+ tan2
e 23
Pt,3 3 + (e )2 Pt,3 3
(0,0)
(0,0)
2 +
Q1c tan2 21 E1 e
(2)
2 (0,0) (e )2 Pt,1 1 1 1 3 Q3c tan2

(0) (0)
tan
e 23
4 e 2 e (0) ( ) Q1s tan2 3 21 (2) 23 E3 e tan + e 23 = 0, (8.75)
+ E13
(2)
Pt,1 1 3 3 + Q3s tan2

(2)
(0,0)
e 23
2 (0,0) (e )2 Pt,3 3 3 3 3
within which E1 and E3 are the functions of material parameters and primary deformation dened in (8.20). The relation (8.74) plays a role similar to (8.46), see the discussion in the previous section, and always holds for waves propagating in a direction with a wave normal the projection of which on the Ox1 x3 plane is given by (c , 0, s ). Thus we conclude that (8.74) is satised and (8.75) is the leading order governing equation. The function p(x1 , x3 , t) = (0,0) Pt (1 , 3 , ) may now be introduced to represent equation (8.75) in terms of the non-scaled variables, yielding tan2 e 21
(0) df
(2)
h2
2 4p e 2 p 2 t2 + ( ) x2 x1 1
+ tan2 4p x4 1 e 23
e 23
h2
2 4p e 2 p 2 t2 + ( ) x2 x3 3
h2 +
Q1c tan2 21 E1 e
(2)
e 21 e 21
2 (e )2 3
(0)
tan
Q3c tan2 e 23
4 e 2 e (0) ( ) Q1s tan2 3 21 (2) 23 E3 e + tan e 23 4p x4 3 = 0. (8.76)
+ E13
(2)
4p (0) 2 2 x2 + Q3s tan x1 3
2 (e )2 3
114
It is possible to substitute the travelling wave solution (1.64) into the governing equation (8.76) and expand the result using the scaled frequency representation (8.9) and the equality (8.10). The relation we obtain in this way matches the second order approximation of the exact dispersion relation (8.18) exactly, thus demonstrating a high level of consistency of the model. We (2,0) remark that the function Pt can not be determined without resorting to higher order. The governing equation (8.76) appears to be very complex when compared to other asymptotic models obtained in this thesis. This is related to the fact that this governing equation describes the motion, which in the free face problem was in a vicinity of two explicit families of cut-o frequencies. In practical applications an additional asymptotic analysis of the governing equation (8.76) may be performed, using the same ideas as the high harmonic number approximations in Section 8.1.1. Indeed, as the harmonic number increases, terms with one of e e the tangents tan 21 or tan 23 will dominate, thus indicating the evolution of another asymptotic process. It is also interesting to note that the plane strain governing equation derived by Nolde & Rogerson (2001) may be recovered from (8.76) by using (8.10) and assuming no dependence of p(x1 , x3 , t) on one of the in-plane variables x1 or x3 .
8.3
We begin the discussion of the numerical results looking at the most interesting feature of long wave motion in a layer with xed faces, namely the dependence of cut-o frequencies on the
3.29 3.26 9.6 3.23
16 16.4
9.8
16.2
3.2 3.17 3.14 3.11 -90 -60 -30 0 30 60 90
9.4
15.8
9.2
15.6 15.4 15.2 -90
8.8 -90
-60
-30
30
60
90
-60
-30
30
60
90
, degrees
, degrees
, degrees
Figure 8.2: The dependence of the scaled cut-o frequency on the in-plane direction of wave propagation, specied by the angle . First ve solutions of (8.10) are shown for a layer composed of MRM, see Table 1.1 (p. 23), and subjected to the static deformation with 1 = 1.09 and 2 = 1.0.
direction of wave propagation. The plots in Figure 8.2 demonstrate how the frequency of the rst ve solutions of (8.10) varies for dierent directions of wave propagation in a MooneyRivlin material. One of the obvious implications of these plots is the fact that for any scaled frequency satisfying (8.10), say 9.6 at the middle plot in Figure 8.2, there are generally two non-orthogonal directions for which this frequency will be the cut-o. The motion in a vicinity of the shear resonance frequency in a layer with free faces is normally associated with the whole layer resonating. However, in this case the resonance will aect only two particular directions. On the other hand it is easier to achieve the resonance in a layer with xed faces, because instead of xed cut-o frequencies there are ranges of cut-o frequencies at which the layer is able to resonate, see Figure 8.2. In view of the asymptotic dominance of incremental pressure over displacements, demonstrated in relations (8.24), it is not clear how dangerous this resonance may be structurally. Although this topic is beyond the scope of our thesis, it deserves close treatment in future studies.

4.3
115
11.61
11.96
4.2
11.94 11.6
4.1
11.92 11.59 11.9 Numerical Asymptotic 0 0.3 0.6 0.9
3.9 0 0.3 0.6 0.9
11.58
11.88 0 0.3 0.6 0.9
kh
kh
kh
Figure 8.3: Numerical and asymptotic solutions for the scaled frequency of the rst three symmetric harmonics, associated with the angle of propagation = 15 . The material of the layer is MRM subjected to the primary deformation with 1 = 0.8 and 2 = 1.3.
11.975 Numerical Asymptotic 12.76
4.2
11.97 12.72
4.15
11.965

4.1
11.96
12.68
12.64 11.955
4.05
11.95 12.6 0 0.3 0.6 0.9 0 Numerical Asymptotic 0.3 0.6 0.9
0.3
0.6
0.9
kh
kh
kh
Figure 8.4: Numerical and asymptotic solutions for the scaled frequency of the rst three symmetric harmonics, associated with the angle of propagation = 15 . The material of the layer is OM1 subjected to the primary deformation with 1 = 0.6 and 2 = 1.4.
The next set of plots in Figure 8.3 presents the scaled frequency of the numerical and asymptotic solutions, shown for the rst three branches of the extensional dispersion relation (7.16) of a layer composed of a Mooney-Rivlin material. The plot in the middle illustrates the point we made earlier, that negative group velocity is possible in a layer with xed faces composed of a Mooney-Rivlin material. It may be worth noting that the left-most and right-most plots in Fige ure 8.3 are related to the rst family of cut-o frequencies (tan 21 dominating) whereas the
middle plot is associated with the second family of cut-o frequencies (tan 23 dominating). We also remark that the quality of approximation generally improves as the harmonic number increases, which may be attributed to the large (at least O((e )3 )) term in the denominator of the expression for (2) , see the left hand side of the relation (8.18). Similar features may be ob served on the plots in Figure 8.4, which depict the scaled frequency of the rst three branches of the dispersion relation for an Ogden material. It is certainly not surprising to nd the existence of negative group velocity in a layer composed of an Ogden material, see the right hand plot in Figure 8.4. Nevertheless, in contrast to the observed behaviour of a Mooney-Rivlin material, plots in Figure 8.4 exhibit a change of sign of the second order approximation coecient (2) . The left-most and the right-most plots are associated with the same family of cut-o frequencies, yet the group velocity signs, related to increasing or decreasing of frequency, are dierent for them.
e
Chapter 9
Layer with xed faces: Long wave anti-symmetric motion

Whereas the long wave symmetric motion analysed in Chapter 8 exhibits many novel features, the long wave anti-symmetric motion in a layer with xed faces, studied in this chapter, demonstrates no signicant departures from the analogous long wave high frequency motion of a layer with free faces. A second order long wave approximation of the dispersion relation is rst obtained and used to demonstrate the existence of two asymptotic regimes, associated with two families of cut-o frequencies. The third order asymptotic approximations of the squares of scaled frequency are then obtained in terms of the scaled wave number. This enables us to determine the relative orders of displacements in each asymptotic regime and to formulate appropriate rescalings for the derivation of asymptotically approximate equations. These are derived using previously established procedures and the associated governing equations are obtained. The derived models are shown to be asymptotically consistent with the appropriate approximations of the exact dispersion relation. Illustrative numerical examples are provided at the end of the chapter. It is demonstrated that negative group velocity is much easier to obtain in a layer with xed faces, in particular it is shown possible to obtain negative group velocity in a layer composed of a Mooney-Rivlin material, something not possible in a layer with free faces in Chapter 3 and Chapter 4.
9.1
We observed in Chapter 7 that the only type of long wave motion possible in a layer with xed faces is the long wave high frequency motion. It is the case for any long wave high frequency motion that v as kh 0. We may therefore use approximations (2.67)(2.69) and (2.71) (2.73) to obtain an appropriate second order approximation of the anti-symmetric dispersion relation (7.20), yielding i A1 v 2 + A1
(2) (0) (3)
T2 (h)T3 (h)
(1)
+ A2 v 2 + A2
v T1 (h)T3 (h) + i A3 v 2 + A3
(2) (3) (2)
(2)
(0)
T1 (h)T2 (h) 0 ,
(9.1)
in which the leading order parameters A1 , A2 and A3 are given by c2 (2) A1 = , 23 A2 =

(3)
1 21 23
s2 (2) A3 = , 21
116
CHAPTER 9. LONG WAVE ANTI-SYMMETRIC MOTION and the second order parameters dened as c2 (0) A1 = 2 23
(0)
117
221 (0) 13 c2 + (213 31 )s2 Q1c c5 23 21

(0) (0) (0)
Q1s + 2Q3s + 32 s2 Q1c + 2Q3c + 12 c2 + 223 , A2 =

(1)
1 (0) (0) (0) (0) 23 Q1s s2 + Q1c c2 21 Q3s s2 + Q3c c2 21 23 21 23 1 1 (0) (0) (0) (0) Q1s + Q3s 221 s2 Q1c + Q3c 223 c2 c5 , 2 2 223 (0) (213 13 )c2 + 31 s2 Q3s c5 21 23
(0) (0) (0)
s2 (0) A3 = 2 21
(0)
2Q1s + Q3s + 32 s2 2Q1c + Q3c + 12 c2 + 221 . It is worth mentioning that these parameter representations made use of the denitions specied immediately after (1.65) and in (2.70).
9.1.1
Approximation in low wave number limit
Because of the large O() factor in the second term in the dispersion relation approximation (9.1) v and the fact that T2 (h) = O(1 ), see (2.72), the dispersion relation can only be balanced when v v one of the hyperbolic tangents, T1 (h) or T3 (h), is small enough, i.e. is a value of O(2 ). Thus we have two possibilities to consider. In the rst case we assume T1 (h) = O(2 ), which implies v the following expansion for tangents argument kq1 h = in + i1
f (2) 2
+ 1
df
f (4) 4
+ O( 6 ) ,
f (2)
(9.2)
f (4)
where the functions of material parameters and pre-stress 1
and 1
are to be determined.
We again re-introduced the small parameter = h/l = kh where h is a half-thickness of the layer and l is a typical wave length. When the argument of T (h) is given according to (9.2) it follows that T (h) can be expanded as (9.3) It is convenient at this point to introduce the parameter family = 21 n, n = 1, 2, 3, . . . . Using this, in conjunction with the expansions (2.71) and (2.74), it is possible to establish f 1 if f (2) q1 = 1 + i1 + O( 3 ) , 21 v= f 1 + 21 1
f (2)
T1 (h) = i1
f (2) 2
+ i1
f (4) 4
+ O( 6 ) .
(9.4)
(0)
Q1c c2 + Q1s s2 2f 1
(0)
+ O( 3 ) .
(9.5)
The appropriate approximations for q2 and q3 may then easily be derived with the help of (2.72) and (2.73), yielding q2 = 1 + T2 (h) = + Q2s s2 + Q2c c2 c5 2(f )2 1
(2) (2) (2) (2)
2 + O( 4 ) , 1 3 3 + O( 5 ) , f 1 23 + O( 2 ) .
(9.6) (9.7) (9.8)
Q2s s2 + Q2c c2 c5 2(f )2 1
if q3 = 1 + O() , 23
T3 (h) = i tan
CHAPTER 9. LONG WAVE ANTI-SYMMETRIC MOTION
118
We now obtained the long wave high frequency representations for all functions featuring in (9.1). When the expansions (9.3)(9.8) are inserted into the second order approximation of the disf (2) f (4) persion relation (9.1) it is possible to determine the unknown parameters 1 and 1 in the following form 21 c2 f (2) 1 = , (9.9) f 1 1
f (4)
21 c2
(f )3 1
1 f 2 ( ) + (12 212 )c2 3 1 23 f s2 cot 1 f 1 23 , (9.10)
f + 2D1 23 31 + 32 s2 +
in which we introduced
f D1 = df
We can now establish an explicit long wave approximation of the dispersion relation (7.20), in form of scaled frequency v as a function of , which upon use of (2.74) takes the form 2 = (f )2 + F1c c2 + F1s s2 2 21 F1c c2 + F1s s2 c2 4 + O( 6 ) , 1 F1c = Q1c 221 , F1c = F1s =
(4) (4) (2) (0) (2) (2) (4) (4)
21 (13 + 21 ) . 21 23
(9.11)
F1s = Q1s , 2 , 3 2 23 f 1 cot f 1 23 2 . 3
(2)
(0)
1 (f )2 1 2 (f )2 1
3Q1c
(2)
+ 21
1 (2) f Q 2D1 + 23 + 31 32 2 1s
It is worth remarking that expansion (9.11) clearly indicates the signicance of the parameters f as leading order frequencies in the low wave number regime and therefore cut-o frequencies. 1 In the second case (when T3 (h) = O(2 )) an analogous procedure can also be performed. v The parameter f = 23 n, n = 1, 2, 3, . . . , is introduced and associated with the second 3 family of cut-o frequencies. It may than be concluded that an explicit approximation of the dispersion relation is given by 2 = (f )2 + F3c c2 + F3s s2 2 23 F3c c2 + F3s s2 s2 4 + O( 6 ) , 3 F3c = Q3c ,
(4) F3c (4) (2) (0) (2) (2) (4) (4)
(9.12)
F3s = Q3s 223 , 21 f 3 21 2 , 3
(2)
(0)
2 (f )2 3 1 (f )2 3
1 (2) f Q + 2D3 + 21 12 + 13 2 3c 3Q3s

(2)
f 3
cot
F3s = in which
+ 23
2 , 3 23 (13 + 23 ) . 21 23
f D3 =
119
9.1.2
The relative asymptotic orders of the displacement components can be obtained with the help of the general solutions (2.50)(2.53). It was shown previously that these solutions may all be represented in terms of a single constant, see (7.21). Hence, we only need to asymptotically estimate the quantities guring in these formulae. It follows from (2.72) that S2 (x2 ) = x2 + O( 3 ) , h C2 (x2 ) = 1 + O( 2 ) , (9.13)
hereafter it is assumed that x2 /h = O(1). In the rst case (when T1 (h) = O(2 )) we can use v the asymptotic expansions (9.4)(9.8) to yield S1 (h) = i(1)n 1
f (2) 2
+ O( 4 ) ,
f x2 Sm (x2 ) = i sin 1 2m h + O( 2 ) ,
+ O( 2 ) ,
f x2 Cm (x2 ) = i cos 1 2m h
m {1, 3} .
In the second case (T3 (h) = O(2 )) the procedure is very similar and after repeating appropriate v steps it delivers the following distribution u1 2 O(pt ) , u2 O(pt ) , u3 O(pt ) . (9.15)
When these are inserted into the general solution representations (2.50)(2.53), the use of (7.21) reveals that u2 O(pt ) , u3 2 O(pt ) . (9.14) u1 O(pt ) ,
Therefore, in both cases the leading order displacement is one of the in-plane displacements. This reminds us of a situation with both extensional and exural long wave high frequency motion in a layer with free faces.
9.2
Asymptotically approximate equations: rst family of cuto frequencies
The perturbation scheme is based on an appropriate rescaling of space and time variables. The scalings of space are again chosen so as to balance the in-plane coordinates x1 and x3 with a typical wave length l, and the normal coordinate x2 is balanced with a plate half-thickness h. We thus use x1 = l1 , x2 = h = l , x3 = l3 , (9.16) in which 1 , and 3 are the new non-dimensional coordinates. Our attention is now restricted to the rst family of cut-o frequencies ( = f ). It follows from (9.5) that a typical wave 1 propagates with scaled speed v v = f /, covering the distance of one wave length in time 1 t = l /. We therefore choose v t = l , (9.17) 21 within which is the new non-dimensional time variable. The relative orders of particle displacements must be distributed for long wave high frequency motion according to (9.14), thus the non-dimensional displacement components may be introduced as um (x1 , x2 , x3 , t) = l m1 u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = 21 p (1 , , 3 , ) , t m {1, 2, 3} , (9.18)
where quantities with are of the same asymptotic order. We comment that the 21 factor was introduced in scaling for incremental pressure purely for algebraic convenience.
120
The scalings (9.16), (9.17) and (9.18) may now be introduced into the equations of motion (1.18)(1.20), yielding a non-dimensional system of the following form 21 u + (f )2 u (f )2 u + 21 u + 2 B1111 u 1 1 + 31 u 3 3 1 1 1, 1, 1, 1, 1 1 B2222 u + (f )2 u (f )2 u + 21 u + (B1122 + B1221 )u 1 2 2 2, 2, 1, 1 1 23 u 3, + (f )2 u (f )2 u + 21 u 3 3 3, 1 1 21 pt,3 + (B2233 + B2332 )u2,3 + 2 + (B1122 + B1221 )u 1 21 p 1 + 4 B1133 + B1331 u 1 3 = 0 , 2, t, 3, 21 p + 2 12 u 1 1 + 32 u 3 3 + (B2233 + B2332 )u 3 = 0 , t, 2, 2, 3, + (B1133 + B1331 )u 1 3 1, 13 u3,1 1 + B3333 u3,3 3 = 0,
(9.19) (9.20) (9.21)
where comma subscripts denote dierentiation with respect to the indicated scaled variable. The equations of motion (9.19)(9.21) must be solved in conjunction with the appropriately rescaled incompressibility condition u 1 + u + 2 u 3 = 0 , (9.22) 1, 2, 3, subject to displacement free boundary conditions u = u = u = 0 1 2 3 at = 1 . (9.23)
It follows from our previous analysis, that for any long wave high frequency motion response of a layer with xed faces 21 u + (f )2 u 2 u , m, m m 1 m {1, 2, 3} , (9.24)
which may be veried by inserting the travelling wave solutions (1.64) into (9.24) and subsequent comparison with (9.11). The implication of (9.24) is that all quantities in braces within scaled equations of motion (9.19)(9.21) are in fact of O( 2 ). We now seek the solutions of our problem in a form of the power series expansions
m
(u , u , u , p ) = 1 2 3 t
n=0
2n u1
(2n)
, u2
(2n)
, u3
(2n)
, pt
(2n)
+ O( 2m+2 ) .
(9.25)
By substituting these solutions into the equations of the boundary value problem (9.19)(9.23) we obtain hierarchical systems of essentially ordinary dierential equations at various orders.
9.2.1
At the leading order we consider the system of the appropriate equations of motion 21 u1, + (f )2 u1 1 B2222 u2, + (f )2 u2 1 23 u3, + (f )2 u3 1
(0) (0) (0) (0) (0) (0)
= 0, + (B1122 + B1221 )u1,1 21 pt, = 0 , + (B1133 + B1331 )u1,1 3 + (B2233 + B2332 )u2,3 21 pt,3 = 0 ,
(0) (0) (0) (0) (0)
(9.26) (9.27)
(9.28)
subject to the leading order incompressibility condition u1,1 + u2, = 0 , and associated leading order xed faces boundary conditions u1
(0) (0) (0)
(9.29)
= 0,
u2
(0)
= 0,
u3
(0)
=0
at
= 1 .
(9.30)
121
The general solution of the boundary value problem given by the rst equation of motion (9.26) and rst boundary condition (9.30)1 may be represented as u1
(0)
= u1
(0,0)
sin
f 1 21
(9.31)
It is worth commenting, that hereafter the double superscripts indicate functions independent of normal variable . The incompressibility condition (9.29) is then considered in conjunction (0) with the second boundary condition (9.30)2 to determine the solution for u2 in the form
(0) u2
21
f 1
cos
f 1 21
cos
f 1 21
u1,1 .
(0,0)
(9.32)
The general solution for leading order incremental pressure can be obtained from the second equation of motion (9.27) and written as 21 pt in which pt
(0,0) (0)
= pt
(0,0)
sin
f 1 21
+ Pt
(0,1)
(9.33)
= u1,1 (b21 21 ) ,
(0,0)
Pt
(0,1)
(0,0) = 21 f u1,1 cos 1
f 1 21
We remark that similarly with previously considered asymptotic models, the symmetry considerations were taken into account when specifying general solutions for displacement components and incremental pressure. Specically, we only choose anti-symmetric terms in the solutions for u1 , u3 and pt , and symmetric terms in the solutions for u2 . This is not leading to the loss of generality of our solutions, as any other terms will vanish under conditions specied when rescaling the space and time variables. The solution of the third equation of motion (9.28) is now sought as u3 where u3
(0,0) (0)
= u3
(0,0)
sin
f 1 21
+ v3
(0,0)
sin
f 1 23
+ U3
(0,1)
(9.34)
f D1
(f )2 1
u1,1 3 ,
(0,0)
U3
(0,1)
21
f 1
u1,1 3 cos
(0,0)
f 1 21
f with the parameter D1 dened immediately after (9.10). Third boundary condition (9.30)3 is used to establish that 21 (0,0) f f (0,0) v3 = u1,1 3 cos 1 csc 1 , 21 23 f 1
thus completing the solution for u3 . The leading order solution is therefore derived in terms (0,0) of the function u1 and its derivatives. We term this function as the leading order long wave amplitude and will obtain a leading order governing equation for it at the next order.
(0)
122
9.2.2
Since our main interest is to derive leading and second order governing equations for the long wave amplitude, we only need to consider the rst two equations of motion at the second order, these being given by 21 u1, + (f )2 u1 1
(2) (2)
(B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (f )2 u1 1 B2222 u2, + (f )2 u2 1

(0) (2) (2) (2) (2) (0) (0)
= B1111 u1,1 1 31 u1,3 3

(0) (0)
(0)
(0)
(0)
(0)
,
(0)
(9.35)
32 u2,3 3 (B2233 + B2332 )u3,3 + 2 21 u2, + (f )2 u2 1 and solved subject to appropriate incompressibility u1,1 + u2, = u3,3 , and boundary conditions u1
(2) (2) (2) (2) (0)
+ (B1122 + B1221 )u1,1 21 pt, = 12 u2,1 1

(0)
(9.36)
(9.37)
= 0,
u2
=0
at
= 1 .
(9.38)
In view of the solutions obtained at the leading order, the solution for the rst equation of motion (9.35) is now sought in the form u1
(2)
= u1
(2,0)
sin
f 1 21
+ u1
(2,1)
cos
f 1 21
+ U1
(2,1)
(9.39)
which after inserting back into (9.35) and referring to the rst boundary condition (9.38)1 yields
(2,1) u1
(0,0) u1,1 1 f 1
21
(2,1) U1
21 f 1
u1,1 1 cos
(0,0)
f 1 21
provided 21 u1, + (f )2 u1 1
(2) (2) (0,0) (0,0)
2 F1c u1,1 1 + F1s u1,3 3 = 0 ,
(2) (0,0)
(2) (0,0)
(9.40)
with the parameters F1c and F1s dened immediately after (9.11). Equation (9.40) is the leading order governing equation for the long wave amplitude. Its solution completely determines the leading order solution of our problem, moreover, (9.40) plays role of the existence condition for the solution of the second order problem. The incompressibility condition (9.37) is analysed next to conclude that u2
(2)
= u2
(2,0)
cos
f 1 21
+ u2
(2,1)
sin
f 1 21
+ v2
(2,0)
cos
f 1 23 + U2
(2,0)
+ 2 U2 within which
(2,0) u2
(2,2)
(9.41)
21
(f )3 1
f (f )2 u1,1 21 u1,1 1 1 + D1 u1,1 3 3 , 1 (2,0)
(2,0)
(0,0)
(0,0)
u2
(2,1)
21 (f )2 1
u1,1 1 1 ,
(2,2)
(0,0)
v2 21
21 23
(f )2 1
u1,1 3 3 cos
(0,0)
f 1 21 .
csc
f 1 23
U2
2f 1
u1,1 1 1 + u1,1 3 3 cos
(0,0)
(0,0)
f 1 21
CHAPTER 9. LONG WAVE ANTI-SYMMETRIC MOTION The second boundary condition (9.38)2 may then be applied to deduce U2
(2,0)
123
21
23 1 + cot 2 f 1
f 1
u1,1 +
(2,0)
1 21 2 (f )2 1 +
f D1
u1,1 1 1
(0,0) u1,1 3 3
(0,0)
f 1 23
(f )2 1
cos
f 1 21
(9.42)
The second order incremental pressure is determined from the second equation of motion (9.36) in the form 21 pt
(2)
= pt
(2,0)
sin
f 1 21
+ pt
(2,1)
cos
f 1 21
+ pt
(2,0)
sin
(2,3)
f 1 23 + Pt
(2,1)
+ 3 Pt with the implied parameters dened by Pt

(2,1)
(9.43)
(2,0) = 21 f u1,1 + 1 1 f + 23 cot 2 1 Pt

(2,3)
1 f Q1c 2 1 +
(2)
+ b21 21 f 1
u1,1 1 1 u1,1 3 3 f 1 21
(0,0)
(0,0)
f 1 23 21 6f 1
f D1 b23 31 + 32
f 1
cos
f 1 21
u1,1 1 1 + u1,1 3 3 cos

(2,0)
(0,0)
(0,0)
pt
(2,0)
= (b21 21 )u1,1 (f )2 21 Q1c u1,1 1 1 1

f (21 b23 )D1 21 (31 32 ) u1,1 3 3 , (0,0)
(2) (0,0)
(2,1) pt (2,0) pt
= =
21 (b21 21 ) f 1 21 (b23 23 ) f 1
u1,1 1 1 ,
(0,0)
(0,0)
u1,1 3 3 cos
f 1 21
csc
f 1 23
where the parameter denitions (2.70) were utilised. It is worth remarking that we derived all of (0,0) (2,0) our solutions in terms of two functions u1 and u1 and their derivatives. The rst of these may be obtained from the governing equation (9.40) whereas the second can not be determined without resorting to the third order. Let us reconsider the leading order governing equation (9.40). The equation may be rescaled back in terms of the original non-scaled variables, thus recasting it in the form h2 2 u1 t2
(0,0)
+ (f )2 u1 1
(0,0)
h2
F1c
(2)
2 u(0,0) 1 x2 1
+ F1s
(2)
2 u(0,0) 1 x2 3
= 0.
(9.44)
There are obvious formal similarities between this equation and the corresponding long wave high frequency governing equations established in Chapter 3 and Chapter 4. It is also possible (2) df (2) (2) to obtain negative group velocity for this model whenever F1 = F1c c2 + F1s s2 < 0, with (2) (2) the necessary condition given by F1c < 0 (inequality F1s 31 > 0 is required by the strong
124
ellipticity conditions (1.36)). It is actually easier to obtain negative group velocity. Let us consider for example the case of a Mooney-Rivlin material. Because every Mooney-Rivlin material satises conditions (1.72)1 , it is not possible to obtain negative group velocity in a layer with free faces composed of such material, see (3.88). However, for a layer with xed faces composed of a Mooney-Rivlin material it is the case that F1c 212 321 = 12 221 ,
(2)
(9.45)
which can become negative for certain pre-stresses. We will further illusitrate this point in a section containing numerical examples.
9.2.3
Third order problem
In order to obtain a rened second order governing equation for the long wave amplitude it is only necessary to consider the rst third order equation of motion 21 u1, + (f )2 u1 1
(2) (4) (4)
31 u1,3 3 (B1122 + B1221 )u2,1 + 21 pt,1 + 2 21 u1, + (f )2 u1 1 in conjunction with the rst third order boundary condition u1
(4)
= (B1133 + B1331 )u3,1 3 B1111 u1,1 1

(2) (2)
(0)
(2)
(2)
(2)
(9.46)
=0
at
= 1 .
(9.47)
After inserting the solutions obtained at previous orders into the right hand side of (9.46), it is (4) possible to deduce that the function u1 has the form u1
(4)
= u1
(4,0)
sin
f 1 21
+ u1
(4,1)
cos
f 1 21 sin
+ u1
(4,2) 2
sin
(4,3)
f 1 21 + U1
(4,1)
+ v1
(4,0)
f 1 23
+ 3 U1
(9.48)
This solution may be substituted back into (9.46) to establish representations for the functions occuring in (9.48), yielding U1
(4,1)
21
23 1 + cot 2 f 1 21 ,
f 1
cos
f 1 21 f 1 23
u1,1 1 + +
(2,0)
1 212 12 2 (f )2 1 (f )2 1
u1,1 1 1 1 u1,1 1 3 3
(0,0)
(0,0)
f D1 + 13 + 21 31 + 32
(4,2) u1
(0,0) u f 2 1,1 1 1 1 2(1 ) (4,0) v1
(4,3) U1
21
6f 1
(0,0) u1,1 1 1 1
(0,0) u1,1 1 3 3
cos
f 1 21
f 21 D3 (f )3 1
u1,1 1 3 3 cos
(0,0)
f 1 21
csc
f 1 23
The boundary condition (9.47) gives then the second order governing equation in the form 21 u1, + (f )2 u1 1
(2,0) (2,0)
2 F1c u1,1 1 + F1s u1,3 3 + 21 F1c u1,1 1 1 1 + F1s u1,1 1 3 3

(4) (0,0) (4) (0,0)
(2) (2,0)
(2) (2,0)
= 0,
(9.49)

(4) (4)
125
in which we made use of the parameters F1c and F1s dened immediately after (9.11). When this equation is satised it is possible to establish that 21 1 212 12 (4,1) (2,0) (0,0) = u1 u1,1 1 + u1,1 1 1 1 f f 2 3 1 (1 ) f 23 1 f 2D1 23 31 + 32 (0,0) + + cot 1 + u1,1 1 3 3 . (9.50) f f 2 3 23 1 (1 ) For the following discussion it is convenient to introduce a new function dened in the nonscaled and scaled forms by u = u(0) + u(2) 2 + O( 4 ) , u = u(0) + u(2) 2 + O( 4 ) , (9.51) and termed as the long wave amplitude. The linear combination (9.40)+ 2 (9.49) may then be considered to be rened governing equation for long wave amplitude and given by 21 u + (f )2 u 2 F1c u 1 1 + F1s u 3 3 , , , 1 21 4 F1c u 1 1 1 1 + F1s u 1 1 3 3 = 0 . , , When transformed into non-dimensional variables this governing equation takes the form h2
2 2 2u (2) u (2) u + (f )2 u h2 F1c 1 2 + F1s x2 t2 x1 3 (4) (4) (2) (2)
(9.52)
21 h4 F1c
(4)
4u
x4 1
+ F1s
(4)
4u x2 x2 1 3
= 0 . (9.53)
Equation (9.53) is the second order governing equation of our asymptotic model for long wave high frequency motion in a pre-stressed elastic layer with xed faces. The consistency of the derivation may again be demonstrated by inserting the travelling wave solution (1.64) into (9.53). The obtained dispersion relation of the asymptotic model will match the appropriate approximation of the exact dispersion relation (9.11) exactly.
9.3
Asymptotically approximate equations: second family of cut-o frequencies
The derivation of asymptotically approximate equations for the second family of cut-o frequencies ( = f ) follows very similar procedure. We again rescale space and time according 3 to (9.16) and t = l , (9.54) 23 respectively. The displacement components and pressure are rescaled to reect the distribution (9.15), thus choosing um (x1 , x2 , x3 , t) = l 3m u (1 , , 3 , ) , m pt (x1 , x2 , x3 , t) = 23 p (1 , , 3 , ) . t m {1, 2, 3} , (9.55)
The subsequent derivation delivers the leading and second order governing equations for long wave amplitude, given by h2 h2
2 2 2u (2) u (2) u + (f )2 u h2 F3c + F3s 3 t2 x2 x2 1 3
= 0,
(9.56) = 0 , (9.57)
2 2 2u (2) u (2) u + (f )2 u h2 F3c + F3s 3 t2 x2 x2 1 3 (2)
23 h4 F3c
(2) (4)
(4)
4 4u (4) u + F3s x2 x2 x4 1 3 3 (4)
in which we use parameter denitions for F3c , F3s , F3c and F3s given immediately after (9.12). These equations may also generate response with a negative group velocity, the necessary con(2) dition for this is F3s < 0.

8.2
16.4
126
24.56
16.35
8.1
16.3
24.52

16.25
24.48
8
16.2
24.44
7.9 0
Numerical Second order Third order 0.3 0.6 0.9
16.15
24.4 0
Numerical Second order Third order 0.3 0.6 0.9
kh
kh
kh
Figure 9.1: Numerical and asymptotic solutions for the scaled frequency of the rst three anti-symmetric harmonics, associated with the rst family of cut-o frequencies and the angle of propagation = 15 . The material of the layer is MRM, see Table 1.1 (p. 23), subjected to the primary deformation with 1 = 0.8 and 2 = 1.3.
9.4
Illustrative numerical examples
It was felt that the only interesting point worth illustrating in this case was possible existence of negative group velocity in a Mooney-Rivlin material, see (9.45) and the associated discussion. Specically, scaled frequencies of the rst three anti-symmetric harmonics were depicted in Figure 9.1 against scaled wave number, along with the corresponding second and third order approximations (9.11). The negative group velocity, associated with decreasing in frequency as the wave number increases, is evident for all three harmonics. Similar to all previously observed high frequency models, the precision of approximations increases for higher harmonic numbers.
Bibliography
Abramowitz, M. & Stegun, I. A., eds. (1970) Handbook of mathematical functions with formulas, graphs, and mathematical tables, vol. 55. U.S. Department of Commerce, National Bureau of Standards, Applied Mathematics Series. Achenbach, J. D. (1969) An asymptotic method to analyse the vibrations of an elastic layer. Trans. ASME Jl. Appl. Mech., 36, pp. 6572. Baylis, E. R. & Green, W. A. (1986) Flexural waves in bre-reinforced laminated plates. Journal of Sound and Vibration, 110, pp. 126. Berdichevskii, V. L. (1977) Equations of high-frequency long-wave vibrations of plates. Doklady Akad. Nauk SSSR (Soviet Phys. Dokl.), 236, pp. 13191322. In Russian. Biot, M. A. (1934) Sur la stabilit de lquilibre lastique. Equations de llasticit dun milieu e e e e e soumis a tension initiale. Annales de la Socit Scientique de Bruxelles, Ser. B., 54, pp. ` ee 1821. Biot, M. A. (1939) Non-linear theory of elasticity and the linearized case for a body under initial stress. Philosophical magazine, Ser. 7 , 27, pp. 468489. Biot, M. A. (1965) Mechanics of incremental deformations: theory of elasticity and viscoelasticity of initially stressed solids and uids, including thermodynamic foundations and applications to nite strain. New York, London, Sydney: John Wiley & Sons, Inc. Boulanger, P. & Hayes, M. (1992) Finite-amplitude waves in deformed Mooney-Rivlin materials. Q. Jl. Mech. Appl. Math., 45, pp. 575593. Chadwick, P. (1999) Continuum mechanics: Concise theory and problems. Mineola, New York: Dover Publications, Inc. Chadwick, P. & Jarvis, D. A. (1979) Surface waves in a pre-stressed elastic body. Proc. R. Soc. Lond., Ser. A, 366, pp. 517536. Chadwick, P., Whitworth, A. M. & Borejko, P. (1985) Basic theory of small-amplitude waves in a constrained elastic body. Arch. Ration. Mech. Anal., 87, pp. 339354. Cole, J. D. (1968) Perturbation methods in applied mathematics. Blaisdell, Waltham, MA. Crawford, R. J. (1985) Plastics and rubbers: engineering design and applications. London: Mechanical Engineering Publications Ltd. Dowaikh, M. A. & Ogden, R. W. (1990) On surface waves and deformations in a pre-stressed incompressible elastic solid. IMA Jl. Appl. Math., 44, pp. 261284. Dowaikh, M. A. & Ogden, R. W. (1991a) Interfacial waves and deformations in pre-stressed elastic media. Proc. R. Soc. Lond., Ser. A, 433, pp. 313328. 127
BIBLIOGRAPHY
128
Dowaikh, M. A. & Ogden, R. W. (1991b) On surface waves and deformations in a compressible elastic half space. SAACM , 1, pp. 2745. Erbay, H. A., Fu, Y. B. & Rogerson, G. A. (2000) Eects of pre-stress on impact waves in an incompressible elastic plate. Quart. J. Mech. Appl. Math., 53, pp. 531550. Ericksen, J. L. (1953) On the propagation of waves in isotropic, incompressible, perfectly elastic materials. Jl. Rational Mech. Analysis, 2, pp. 329337. Flavin, J. N. (1963) Surface waves in pre-stressed Mooney material. Q. Jl. Mech. Appl. Mech., 16, pp. 441449. Fu, Y. B. & Rogerson, G. A. (1994) A non-linear analysis of instability of a pre-stressed incompressible elastic plate. Proc. R. Soc. Lond., Ser. A, 446, pp. 233254. Goldenveiser, A. L. (1966) The principles of reducing three-dimensional problems of elasticity to two-dimensional problems of the theory of plates and shells. In H. Gortler, ed., Applied Mechanics, Proc. 11th Int. Congr. Appl. Mech., pp. 306311. Berlin: Springer. Goldenveizer, A. L. (1961) Theory of elastic thin shells. Oxford, London, New York, Paris: Pergamon Press. Transl. from Goldenveizer, A. L. (1953) Teoriia uprugykh tonkikh obolochek, Moscow: Gostekhteoretizdat (in Russian). Green, A. E. & Adkins, J. E. (1960) Large elastic deformations. Oxford: Clarendon Press. Green, A. E. & Zerna, W. (1954) Theoretical elasticity. Oxford: Clarendon Press. Green, A. E., Rivlin, R. S. & Shield, R. T. (1952) General theory of small elastic deformations superimposed on nite elastic deformations. Proc. R. Soc. Lond., Ser. A, 211, pp. 128154. Green, W. A. (1982) Bending waves in strongly anisotropic elastic plates. Q. Jl. Mech. Appl. Math., 35, pp. 485507. Hayes, M. A. & Rivlin, R. S. (1961) Surface waves in deformed elastic materials. Arch. Ration. Mech. Anal., 8, pp. 358380. Hirst, A. J. (1969) Rubber suspensions in transport. In L. R. Mernagh, ed., Rubbers handbook , pp. 7986. West Wickham: Morgan-Grampian (Publishers) Ltd. Jones, D. F. & Treloar, L. R. G. (1975) The properties of rubber in pure homogeneous strain. J. Phys. D: Appl. Phys., 8, pp. 12851304. Jones, R. P. N. (1964) Transverse impact waves in a bar under conditions of plane strain elasticity. Q. Jl. Mech. Appl. Math., 17, pp. 401421. Kaplunov, J. D. (1989) Equations for high-frequency long-wave vibrations of an elastic layer lying on an acoustic half-space. Doklady Akad. Nauk SSSR (Soviet Phys. Dokl.), 309, pp. 10771081. Kaplunov, J. D. (1990) High-frequency stress-strain states of small variability in elastic shells. Izv. Akad. Nauk SSSR, Ser. Mekhanika Tverdogo Tela (Mechanics of Solids), 25(5), pp. 147 157. Kaplunov, J. D. & Markushevich, D. G. (1993) Plane vibrations and radiation of an elastic layer lying on a liquid half-space. Wave Motion, 17, pp. 199211.
BIBLIOGRAPHY
129
Kaplunov, J. D., Kossovich, L. Y. & Nolde, E. V. (1998) Dynamics of thin walled elastic bodies. New York: Academic Press. Kaplunov, J. D., Kossovich, L. Y. & Rogerson, G. A. (2000a) Direct asymptotic integration of the equations of transversely isotropic elasticity for a plate near cut-o frequencies. Quart. J. Mech. Appl. Math., 53, pp. 323341. Kaplunov, J. D., Nolde, E. V. & Rogerson, G. A. (2000b) A low frequency model for dynamic motion in pre-stressed incompressible elastic structures. Proc. R. Soc. Lond., Ser. A, 456, pp. 25892610. Kaplunov, J. D., Nolde, E. V. & Rogerson, G. A. (2001a) An asymptotically consistent model for long wave high frequency motion in a pre-stressed elastic plate. Mech. Math. Solids. To appear. Kaplunov, J. D., Nolde, E. V. & Rogerson, G. A. (2001b) Short wave motion in a pre-stressed elastic plate. Submitted for publication. Kaplunov, Y. D. & Nolde, Y. V. (2001) Long-wave vibrations of a nearly incompressible isotropic plate with xed faces. Quart. Jl. Mech. Appl. Math.. To appear. Kevorkian, J. (1989) Partial dierential equations: analytical solution techniques. Pacic Grove, California: Wadsworth & Brooks/Cole Advanced Books & Software. Kossovitch, L. Y. & Rogerson, G. A. (1999) Approximations of the dispersion relation for an elastic plate composed of strongly anisotropic elastic material. Journal of Sound and Vibration, 225, pp. 283305. Kossovitch, L. Y., Moukhomodiarov, R. R. & Rogerson, G. A. (2001) Analysis of the dispersion relation for an incompressible transversely isotropic elastic plate. Submitted for publication. Lamb, H. (1917) On waves in an elastic plate. Proc. R. Soc. Lond., Ser. A, 93, pp. 114128. Liang, C., Lippmann, H. & Najar, J. (1993) Eects of articially induced vibrations on the prevention of coal mine bumps. In R. P. Young, ed., Proceedings of the 3rd International Symposium on Rockbursts and Seismicity in Mines, pp. 9194. Balkema. Lord Rayleigh, J. W. S. (1877) The theory of sound, vol. 1. London: Macmillan and Co. Lord Rayleigh, J. W. S. (1885) On waves propagated along the plane surface of an elastic solid. Proc. Lond. Math. Soc., 17, pp. 411. Lord Rayleigh, J. W. S. (1889) On the free vibrations of an innite plate of homogeneous isotropic elastic matter. Proc. Lond. Math. Soc., 20, pp. 225234. Love, A. E. H. (1927) A treatise on the mathematical theory of elasticity. Cambridge University Press. Mindlin, R. D. (1955) An introduction to the mathematical theory of vibrations of elastic plates. Ft. Monmouth, New Jersey: U.S. Army Signal Corps Eng. Lab. A monograph. Mindlin, R. D. (1960) Waves and vibrations in isotropic, elastic plates. In J. N. Goodier & N. J. Ho, eds., Structural Mechanics, pp. 199232. Oxford, New York: Pergamon Press, Inc. Mooney, M. (1940) A theory of large elastic deformation. Jl. Appl. Phys., 11, pp. 582592.
BIBLIOGRAPHY
130
Nayfeh, A. H. (1991) The general problem of elastic wave propagation in multilayered anisotropic media. J. Acoust. Soc. Am., 89, pp. 15211531. Nolde, E. V. & Rogerson, G. A. (2001) Long wave asymptotic integration of the governing equations for a pre-stressed incompressible elastic layer with xed faces. Submitted for publication. Ogden, R. W. (1972) Large deformation isotropic elasticity - on the correlation of theory and experiment for incompressible rubberlike solids. Proc. R. Soc. Lond., Ser. A, 326, pp. 565584. Ogden, R. W. (1974a) On isotropic tensors and elastic moduli. Proc. Camb. Phil. Soc., 75, pp. 427436. Ogden, R. W. (1974b) On stress rates in solid mechanics with application to elasticity theory. Proc. Camb. Phil. Soc., 75, pp. 303319. Ogden, R. W. (1982) Elastic deformations of rubberlike solids. In H. G. Hopkins & M. J. Sewell, eds., Mechanics of solids. The Rodney Hill 60th anniversary volume, pp. 499537. Pergamon Press. Ogden, R. W. (1984) Non-linear elastic deformations. New York: Ellis Horwood. Ogden, R. W. & Roxburgh, D. G. (1993) The eect of pre-stress on the vibration and stability of elastic plates. Int. J. Engng. Sci., 31, pp. 16111639. Ogden, R. W. & Sotiropoulos, D. A. (1997) The eect of pre-stress on the propagation and reection of plane waves in incompressible elastic solids. IMA Jl. Appl. Math., 59, pp. 95121. Ogden, R. W. & Sotiropoulos, D. A. (1998) Reection of plane waves from the boundary of a pre-stressed compressible elastic half-space. IMA Jl. Appl. Math., 61, pp. 6190. Pichugin, A. V. & Rogerson, G. A. (1999) An asymptotic model for the dynamic motion in an incompressible plate. In Proc. of Fourth Int. Conf. On Vibration Problems, pp. 6469. Calcutta. Pichugin, A. V. & Rogerson, G. A. (2001a) A two-dimensional model for extensional motion of a pre-stressed incompressible elastic layer near cut-o frequencies. IMA Jl. Appl. Math., 66, pp. 357385. Pichugin, A. V. & Rogerson, G. A. (2001b) Dynamic model of a pre-stressed incompressible elastic layer in the vicinity of shear resonance. Submitted for publication. Pichugin, A. V. & Rogerson, G. A. (2001c) An asymptotic membrane-like theory for long wave motion in a pre-stressed elastic plate. Proc. R. Soc. Lond., Ser. A. Accepted for publication. Prendergast, J. (1995) Seismic isolation in bridges. ASCE Civil Engineering, 65, pp. 5861. Reissner, E. (1964) On the derivation of the theory of thin elastic shells. J. Math. Phys., 43, pp. 263277. Rivlin, R. S. & Saunders, D. W. (1951) Large elastic deformations of isotropic materials VII. Experiments on the deformation of rubber. Phil. Trans. R. Soc. Lond., Ser. A, 243, pp. 251288.
BIBLIOGRAPHY
131
Rogerson, G. A. (1992) Penetration of impact waves in a six-ply bre composite laminate. Journal of Sound and Vibration, 158, pp. 105120. Rogerson, G. A. (1997) Some asymptotic expansions of the dispersion relation for an incompressible elastic plate. Int. Jl. Solid. Struct., 34(22), pp. 27852802. Rogerson, G. A. & Fu, Y. B. (1995) An asymptotic analysis of the dispersion relation of a pre-stressed incompressible elastic plate. Acta Mechanica, 111, pp. 5977. Rogerson, G. A. & Sandiford, K. J. (1996) On small amplitude vibrations of pre-stressed laminates. Int. Jl. Eng. Sci., 34(8), pp. 853872. Rogerson, G. A. & Sandiford, K. J. (1999) Harmonic wave propagation along a nonprincipal direction in a pre-stressed elastic plate. Int. Jl. Eng. Sci., 37, pp. 16631691. Rogerson, G. A. & Sandiford, K. J. (2002) Some comments on the dispersion relation for periodically layered pre-stressed elastic media. Int. Jl. Eng. Sci., 40, pp. 2349. Sandiford, K. J. & Rogerson, G. A. (2000) Some dynamic properties of a pre-stressed, nearly incompressible (rubber-like) elastic layer. Int. Jl. Non-linear Mech., 35, pp. 849868. Scott, N. (1975) Acceleration waves in constrained elastic materials. Arch. Rational Mech. Anal., 58, pp. 5775. Scott, N. H. (1986) The slowness surface of incompressible and nearly incompressible elastic materials. J. Elasticity., 16, pp. 239250. Scott, N. H. & Hayes, M. (1985) A note on wave propagation in internally constrained hyperelastic materials. Wave Motion, 7, pp. 601605. Sheridan, P. M., James, F. O. & Miller, T. S. (1992) Design of components. In A. N. Gent, ed., Engineering with Rubber , pp. 209235. Munich: Hanser. Spencer, A. J. M. (1980) Continuum mechanics. Longman Scientic & Technical. Timoshenko, S. P. (1938) Vibration problems in engineering. New York: D. Van Nostrand Company, Inc. Tolstoy, I. & Usdin, E. (1957) Wave propagation in elastic plates: low and high mode dispersion. J. Acoust. Soc. Am., 29, pp. 3742. Torr, R. P. (1969) Bridge bearings. In L. R. Mernagh, ed., Rubbers handbook , pp. 99103. West Wickham: Morgan-Grampian (Publishers) Ltd. Treloar, L. R. G. (1943) The elasticity of a network of long-chain moleculesII. Trans. Faraday Soc., 39, pp. 241246. Treloar, L. R. G. (1944) Stress-strain data for vulcanised rubber under various types of deformation. Trans. Faraday Soc., 40, pp. 5970. Treloar, L. R. G. (1975) The physics of rubber elasticity. Oxford: Clarendon Press. Truesdell, C. & Noll, W. (1965) The non-linear eld theories of mechanics. Handbuch der physik , vol. III/3. Berlin, Heidelberg, New York: Springer. Tyler, R. G. (1991) Rubber bearings in base-isolated structuresa summary paper. Bulletin of the New Zealand National Society for Earthquake Engineering, 24, pp. 251274.
BIBLIOGRAPHY
132
Valanis, K. C. & Landel, R. F. (1967) The strain-energy function of a hyperelastic material in terms of the extension ratios. Jl. Appl. Phys., 38, pp. 29973002. Varga, O. H. (1966) Stress-strain behaviour of elastic materials. New York: John Wiley and Sons. Wang, C.-C. & Truesdell, C. (1973) Introduction to rational elasticity. Leyden: Noordho International Publishing. Willson, A. J. (1973) Surface and plate waves in biaxially-stressed elastic media. Pure Appl. Geophys., 102, pp. 182192. Willson, A. J. (1974) The anomalous surface wave in uniaxially-stressed elastic material. Pure Appl. Geophys., 112, pp. 665674. Zee, L. & Sternberg, E. (1983) Ordinary and strong ellipticity in the equilibrium theory of incompressible hyperelastic solids. Arch. Rational Mech. Anal., 83, pp. 5390.

E Thesis

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

E Thesis

Transféré par

Droits d'auteur :

Formats disponibles

Asymptotic Models for Long Wave Motion in a Pre-stressed Incompressible Elastic Plate

ii 52 56 56 57 59 59 60 61 63 65 66 67 71 71 72 73 74 74 75 76 79 82 84 84 85 86 86 87 88 89 91 92 93 94 95 97 97 98 99 100 101 103 103 104 106 107 108

Congurations of a pre-stressed body

CHAPTER 1. MATHEMATICAL PRELIMINARIES

of innitesimal time-dependent deformation Be Bt , so that xi (XA ) = xi (XA ) + ui (xm (XA ), t) . (1.2)

Motion of a pre-stressed incompressible elastic body

CHAPTER 1. MATHEMATICAL PRELIMINARIES

CHAPTER 1. MATHEMATICAL PRELIMINARIES

Propagation of plane waves, strong ellipticity

CHAPTER 1. MATHEMATICAL PRELIMINARIES

CHAPTER 1. MATHEMATICAL PRELIMINARIES

Explicit necessary and sucient conditions for the strong ellipticity

Necessary and sucient conditions for particular directions

CHAPTER 1. MATHEMATICAL PRELIMINARIES

CHAPTER 1. MATHEMATICAL PRELIMINARIES

1, and in view of (1.36), we conclude that (1.43) is equivalent to 3 < 12 . (1.44)

General conditions: necessity

(1.52) (1.53) (1.54)

It is easy to see that this condition is in particular satised by l1 = 1 , W +1 l2 = W , W +1 W = 2 , 1 (1.57)

which after substituting into the condition (1.55) yields

CHAPTER 1. MATHEMATICAL PRELIMINARIES

General conditions: suciency

(l2 1 + (l2 1)2 )2 ((l1 1)1 + l1 2 )2 (l2 1 l1 2 )2 1 + 2 2 + 2 3 l2 l1 l1 l2 1 2 1 3 2 3 + (l1 + l2 1)2 + + , l1 l2 l1 l2

Two of these inequalities may be used to demonstrate that ( 1 2 )2 < 3 , (1.63)

Separation of variables: secular equation

CHAPTER 1. MATHEMATICAL PRELIMINARIES

v as the scaled wave speed and

(1.66) (1.67) (1.68)

and may be used to represent equation (1.65) as the following product

CHAPTER 1. MATHEMATICAL PRELIMINARIES

CHAPTER 1. MATHEMATICAL PRELIMINARIES

Some specic material models for rubber-like materials

CHAPTER 1. MATHEMATICAL PRELIMINARIES

CHAPTER 1. MATHEMATICAL PRELIMINARIES

as the sucient condition for strong ellipticity of Mooney-Rivlin material.

Varga materials, separability of the strain energy function

The class of materials with the strain energy function

CHAPTER 1. MATHEMATICAL PRELIMINARIES

Table 1.1: Specic materials to be used in generating graphs.

Measure of incremental surface traction

These parameters were taken from Ogden (1972, p. 574).

CHAPTER 1. MATHEMATICAL PRELIMINARIES

Figure 1.2: Deformation of an elementary surface.

Dispersion of harmonic waves in a layer with free faces

Harmonic wave propagation in a pre-stressed incompressible elastic layer

x3 = 3 (X3 cos X1 sin ) ,

CHAPTER 2. WAVES IN A LAYER WITH FREE FACES

eik(x1 c +x3 s vt) , eik(x1 c +x3 s vt) ,

(2.4) (2.5) (2.6) (2.7)

eik(x1 c +x3 s vt) ,

(2.8) (2.9) (2.10) (2.11)

CHAPTER 2. WAVES IN A LAYER WITH FREE FACES

The dispersion relations

(2.20) (2.21) (2.22)

(g1 13 )U1 (qm , v )c2 + (g3 2 23 U3 (qm , v )qm + g3 V(qm , v ) .

In the expressions (2.23)(2.25) we introduced the new parameters gi = 2i 2 ,

CHAPTER 2. WAVES IN A LAYER WITH FREE FACES

qm T2 (qm , v ) (2m1) kqm h (2m) U2 e U2 ekqmh = 0 , V(qm , v )

qm T2 (qm , v ) (2m1) kqm h (2m) U2 e U2 ekqm h = 0 , V(qm , v ) m=1

T3 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v )

T3 (qm , v ) (2m1) kqm h (2m) U2 e + U2 ekqm h = 0 , V(qm , v ) m=1

(2.34) (2.35) (2.36)