8/23/2019 Applied Solid Mechanics
1/468
http://www.cambridge.org/97805218548948/23/2019 Applied Solid Mechanics
2/468
This page intentionally left blank
8/23/2019 Applied Solid Mechanics
3/468
A P P L I E D S O L I D M E C H A N I C S
Much of the world around us, both natural and man-made, is built from and held
together by solid materials. Understanding how they behave is the task of solid
mechanics, which can in turn be applied to a wide range of areas from earthquake
mechanics and the construction industry to biomechanics. The variety of materials
(such as metals, rocks, glasses, sand, flesh and bone) and their properties (such as
porosity, viscosity, elasticity, plasticity) are reflected by the concepts and techniques
needed to understand them, which are a rich mixture of mathematics, physics, ex-
periment and intuition. These are all brought to bear in this distinctive book, which
is based on years of experience in research and teaching. Theory is related to
practical applications, where surprising phenomena occur and where innovative
mathematical methods are needed to understand features such as fracture. Starting
from the very simplest situations, based on elementary observations in engineer-
ing and physics, models of increasing sophistication are derived and applied. The
emphasis is on problem solving and on building an intuitive understanding, rather
than on a technical presentation of theoretical topics. The text is complemented by
over 100 carefully chosen exercises, and the minimal prerequisites make it an ideal
companion for mathematics students taking advanced courses, for those undertak-
ing research in the area or for those working in other disciplines in which solid
mechanics plays a crucial role.
8/23/2019 Applied Solid Mechanics
4/468
Cambridge Texts in Applied Mathematics
Editorial Board
Mark Ablowitz, University of Colorado, Boulder
S. Davis, Northwestern UniversityE. J. Hinch, University of Cambridge
Arieh Iserles, University of Cambridge
John Ockendon, University of Oxford
Peter Olver, University of Minnesota
8/23/2019 Applied Solid Mechanics
5/468
A P P L I E D S O L I D M E C H A N I C S
P E T E R H O W E L LUniversity of Oxford
G R E G O R Y K O Z Y R E F FFonds de la Recherche ScientifiqueFNRS
and Universit e Libre de Bruxelles
J O H N O C K E N D O N
University of Oxford
8/23/2019 Applied Solid Mechanics
6/468
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
Cambridge University PressThe Edinburgh Building, Cambridge CB2 8RU, UK
First published in print format
ISBN-13 978-0-521-85489-4
ISBN-13 978-0-521-67109-5
ISBN-13 978-0-511-50639-0
P. D. Howell, G. Kozyreff and J. R. Ockendon 2009
2008
Information on this title: www.cambrid e.or /9780521854894
This publication is in copyright. Subject to statutory exception and to theprovision of relevant collective licensing agreements, no reproduction of any partmay take place without the written permission of Cambridge University Press.
Cambridge University Press has no responsibility for the persistence or accuracyof urls for external or third-party internet websites referred to in this publication,and does not guarantee that any content on such websites is, or will remain,accurate or appropriate.
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
a erback
eBook (EBL)
hardback
http://www.cambridge.org/9780521854894http://www.cambridge.org/http://www.cambridge.org/9780521854894http://www.cambridge.org/8/23/2019 Applied Solid Mechanics
7/468
Contents
List of illustrations page viiiPrologue xiii
Modelling solids 1
1.1 Introduction 1
1.2 Hookes law 2
1.3 Lagrangian and Eulerian coordinates 3
1.4 Strain 4
1.5 Stress 71.6 Conservation of momentum 10
1.7 Linear elasticity 11
1.8 The incompressibility approximation 13
1.9 Energy 14
1.10 Boundary conditions and well-posedness 16
1.11 Coordinate systems 19
Exercises 24
Linear elastostatics 28
2.1 Introduction 28
2.2 Linear displacements 29
2.3 Antiplane strain 37
2.4 Torsion 39
2.5 Multiply-connected domains 42
2.6 Plane strain 47
2.7 Compatibility 682.8 Generalised stress functions 70
2.9 Singular solutions in elastostatics 82
2.10 Concluding remark 93
Exercises 93
v
8/23/2019 Applied Solid Mechanics
8/468
vi Contents
Linear elastodynamics 103
3.1 Introduction 103
3.2 Normal modes and plane waves 104
3.3 Dynamic stress functions 121
3.4 Waves in cylinders and spheres 1243.5 Initial-value problems 132
3.6 Moving singularities 138
3.7 Concluding remarks 143
Exercises 143
Approximate theories 150
4.1 Introduction 150
4.2 Longitudinal displacement of a bar 1514.3 Transverse displacements of a string 152
4.4 Transverse displacements of a beam 153
4.5 Linear rod theory 158
4.6 Linear plate theory 162
4.7 Von Karman plate theory 172
4.8 Weakly curved shell theory 177
4.9 Nonlinear beam theory 187
4.10 Nonlinear rod theory 1954.11 Geometrically nonlinear wave propagation 198
4.12 Concluding remarks 204
Exercises 205
Nonlinear elasticity 215
5.1 Introduction 215
5.2 Stress and strain revisited 216
5.3 The constitutive relation 221
5.4 Examples 233
5.5 Concluding remarks 239
Exercises 239
Asymptotic analysis 245
6.1 Introduction 245
6.2 Antiplane strain in a thin plate 246
6.3 The linear plate equation 248
6.4 Boundary conditions and Saint-Venants principle 2536.5 The von Karman plate equations 261
6.6 The EulerBernoulli plate equations 267
6.7 The linear rod equations 273
6.8 Linear shell theory 278
8/23/2019 Applied Solid Mechanics
9/468
This page intentionally left blank
8/23/2019 Applied Solid Mechanics
10/468
Contents vii
6.9 Concluding remarks 282
Exercises 283
Fracture and contact 287
7.1 Introduction 287
7.2 Static brittle fracture 2887.3 Contact 309
7.4 Concluding remarks 320
Exercises 321
Plasticity 328
8.1 Introduction 328
8.2 Models for granular material 330
8.3 Dislocation theory 3378.4 Perfect plasticity theory for metals 344
8.5 Kinematics 358
8.6 Conservation of momentum 360
8.7 Conservation of energy 360
8.8 The flow rule 362
8.9 Simultaneous elasticity and plasticity 364
8.10 Examples 365
8.11 Concluding remarks 370Exercises 372
More general theories 378
9.1 Introduction 378
9.2 Viscoelasticity 379
9.3 Thermoelasticity 388
9.4 Composite materials and homogenisation 391
9.5 Poroelasticity 408
9.6 Anisotropy 413
9.7 Concluding remarks 417
Exercises 417
Epilogue 426
Appendix Orthogonal curvilinear coordinates 428
References 440
Index 442
8/23/2019 Applied Solid Mechanics
11/468
Illustrations
1.1 A reference tetrahedron. page 81.2 The forces acting on a small two-dimensional element. 91.3 A small pill-box-shaped region at the boundary between two elastic
solids. 181.4 Forces acting on a polar element of solid. 221.5 A system of masses connected by springs. 252.1 A unit cube undergoing (a) uniform expansion, (b) one-dimensional
shear, (c) uniaxial stretching. 302.2 A uniform bar being stretched under a tensile force. 32
2.3 A paper model with negative Poissons ratio. 332.4 A strained plate. 342.5 A bar in a state of antiplane strain. 382.6 A twisted bar. 392.7 A uniform tubular torsion bar. 432.8 The cross-section of (a) a circular cylindrical tube; (b) a cut tube. 442.9 The unit normal and tangent to the boundary of a plane region. 492.10 A plane annulus being inflated by an internal pressure. 532.11 A plane rectangular region subject to tangential tractions on its faces. 572.12 The tractions applied to the edge of a semi-infinite strip. 592.13 The surface displacement of a half-space and corresponding surface
pressure. 652.14 A family of functions (x) that approach a delta-function as 0. 832.15 Contours of the maximum shear stress created by a point force acting
at the origin. 852.16 Four point forces. 913.1 Plots of the first three Bessel functions. 108
3.2 A P-wave reflecting from a rigid boundary. 1163.3 A layered elastic medium. 1173.4 Dispersion relation for symmetric and antisymmetric Love waves. 1203.5 Illustration of flexural waves. 1283.6 The one-dimensional fundamental solution. 1343.7 The two-dimensional fundamental solution. 135
viii
8/23/2019 Applied Solid Mechanics
12/468
List of illustrations ix
3.8 The cone x2 + y2 = c2t2 tangent to the plane k1x + k2y = t. 1383.9 The two-sheeted characteristic cone for the Navier equation. 1383.10 The response of a string to a point force moving at speed V. 1393.11 Wave-fronts generated by a moving force on an elastic membrane. 1413.12 P-wave- and S-wave-fronts generated by a point force moving at speed
V in plane strain. 1423.13 Group velocity versus wave-number for symmetric and antisymmetric
Love waves. 1464.1 The forces acting on a small length of a uniform bar. 1514.2 The forces acting on a small length of an elastic string. 1534.3 The forces and moments acting on a small segment of an elastic beam. 1544.4 The end of a beam under clamped, simply supported and free conditions.1554.5 The first three buckling modes of a clamped elastic beam. 1574.6 The internal force components in a thin elastic rod. 1594.7 Cross-section through a rod showing the bending moment components. 1594.8 Examples of cross-sections in the (y, z)-plane and their bending
stiffnesses. 1614.9 The forces acting on a small section of an elastic plate. 1634.10 The bending moments acting on a section of an elastic plate. 1644.11 The displacement of a simply supported rectangular plate sagging
under gravity. 1694.12 (a) A cylinder, (b) a cone, (c) another developable surface, (d) a
hyperboloid. 1754.13 Typical surface shapes with (a) zero, (b) negative and (c) positiveGauss 179
4.14 Deformations of a cylindrical shell. 1844.15 Deformations of an anticlastic shell. 1854.16 Deformations of a synclastic shell. 1864.17 A beam (a) before and (b) after bending; (c) a close-up of the
displacement field. 1874.18 (a) The forces and moments acting on a small segment of a beam.
(b) The sign convention for the forces at the ends of the beam. 1884.19 (a) Final angle of a diving board versus applied force parameter.(b) Deflection of a diving board for various values of the force parameter.191
4.20 (a) Response diagram of the amplitude of the linearised solution for abuckling beam versus the force parameter. (b) Corresponding responseof the weakly nonlinear solution. 193
4.21 (a) Pitchfork bifurcation diagram of leading-order amplitude versusforcing parameter. (b) The corresponding diagram when asymmetry isintroduced. 195
4.22 A system of pendulums attached to a twisting rubber band. 1994.23 A kink propagating along a series of pendulums attached to a rod. 2004.24 Travelling wave solution of the nonlinear beam equations. 2014.25 A beam clamped near the edge of a table. 2064.26 A beam supported at two points. 207
iancurvature.
4.27 The first three buckling modes of a vertically clamped beam. 213
8/23/2019 Applied Solid Mechanics
13/468
x List of illustrations
5.1 The deformation of a small scalene cylinder. 2185.2 Typical forcestrain graphs for uniaxial tests on various materials. 2335.3 A square membrane subject to an isotropic tensile force. 2345.4 Response diagrams for a biaxially-loaded incompressible sheet of
MooneyRivlin material. 2355.5 Scaled pressure inside a balloon as a function of the stretch for various
values of the MooneyRivlin parameter. 2365.6 Gas pressure inside a cavity as a function of inflation coefficient for
various values of the MooneyRivlin parameter. 2386.1 The edge of a plate subject to tractions. 2546.2 The geometry of a deformed two-dimensional plate. 2697.1 Definition sketch of a thin crack. 2887.2 Definition sketch for contact between two solids. 288
7.3 (a) A Mode III crack. (b) A cross-section in the (x, y)-plane. 2907.4 Definition sketch for the function
z2 c2 . 292
7.5 Displacement field for a Mode III crack. 2937.6 (a) A planar Mode II crack. (b) The regularised problem of a thin
elliptical crack. 2977.7 Contour plot of the maximum shear stress around a Mode II crack. 3017.8 The displacement of a Mode II crack under increasing shear stress. 3037.9 A Mode I crack. 3047.10 Contour plot of the maximum shear stress around a Mode I crack. 306
7.11 The displacement of a Mode I crack under increasing normal stress. 3077.12 Solution for the contact between a string and a level surface. 3107.13 Three candidate solutions for a contact problem. 3117.14 The contact between a beam and a horizontal surface under a uniform
pressure. 3147.15 Contact between a rigid body and an elastic half-space. 3177.16 The penetration of a quadratic punch into an elastic half-space. 3197.17 A flexible ruler flattened against a table. 3267.18 A wave travelling along a rope on the ground. 3268.1 A typical stressstrain relationship for a plastic material. 3298.2 The stressstrain relationship for a perfectly plastic material. 3308.3 The forces acting on a particle at the surface of a granular material. 3318.4 The normal force and frictional force acting on a surface element inside
a granular material. 3328.5 The Mohr circle. 3338.6 The triaxial stress factor versus angle of friction. 3368.7 An antiplane cut-and-weld operation. 339
8.8 The displacement field in an edge dislocation. 3408.9 An edge dislocation in a square crystal lattice. 3418.10 A moving edge dislocation. 3428.11 The normalised torque versus twist applied to an elastic-plastic
cylindrical bar. 3478.12 The normalised torque versus twist applied to an elastic-plastic
cylindrical bar, showing the recovery phase. 348
8/23/2019 Applied Solid Mechanics
14/468
List of illustrations xi
8.13 The free-boundary problem for an elastic-perfectly plastic torsion bar. 3498.14 Residual shear stress in a gun barrel versus radial distance for different
values of the maximum internal pressurisation. 3528.15 The Tresca yield surface. 3558.16 The von Mises yield surface. 3568.17 The Coulomb yield surface. 3588.18 Luders bands in a thin sheet of metal. 3698.19 The Mohr surface for three-dimensional granular flow. 3738.20 The normalised torque versus twist applied to an elastic-plastic
cylindrical bar undergoing a loading cycle. 3759.1 (a) A spring; (b) a dashpot; (c) a spring and dashpot connected in
parallel; (d) a spring and dashpot connected in series. 3809.2 (a) Applied tension as a function of time. (b) Resultant displacement
of a linear elastic spring. (c) Resultant displacement of a linear dashpot.3819.3 Displacement of a Voigt element due to the applied tension shown in
Figure 9.2(a). 3829.4 Displacement of a Maxwell element due to the applied tension shown
in Figure 9.2(a). 3839.5 (a) The variation of Youngs modulus with position in a bar. (b) The
corresponding longitudinal displacement. 3929.6 (a) The variation of Youngs modulus with position in a bar. (b) The
corresponding longitudinal displacement. 3959.7 A periodic microstructured shear modulus. 3969.8 A symmetric, piecewise constant shear modulus distribution. 4009.9 Some modulus distributions that are antisymmetric about the diago-
nals of a square. 4029.10 Dimensionless wavenumber versus the Youngs modulus non-uniformity
parameter. 4079.11 The one-dimensional squeezing of a sponge. 4119.12 Dimensionless stress applied to a sponge versus dimensionless time for
different values of the Peclet number. 4129.13 A Jeffreys viscoelastic element. 418
9.14 A system of masses connected by springs and dashpots in parallel. 4189.15 A system of masses connected by springs and dashpots in series. 4199.16 Dimensionless wavenumber versus Youngs modulus contrast for a
piecewise uniform bar. 424A1.1 A small reference box. 432A1.2 Cylindrical polar coordinates. 437A1.3 Spherical polar coordinates. 438
8/23/2019 Applied Solid Mechanics
15/468
8/23/2019 Applied Solid Mechanics
16/468
Prologue
Although solid mechanics is a vitally important branch of applied mechan-ics, it is often less popular, at least among students, than its close relative,
fluid mechanics. Several reasons can be advanced for this disparity, such as
the prevalence of tensors in models for solids or the especial difficulty of han-
dling nonlinearity. Perhaps the most daunting prospect for the student is the
multitude of different behaviours that can occur and cause elementary theo-
ries of elasticity to become irrelevant in practice. Examples include fracture,
buckling and plasticity, and these pose intellectual challenges in solid me-
chanics that are every bit as fascinating as concepts like flight, shock waves
and turbulence in fluid dynamics. Our principal objective in this book is to
demonstrate this fact to undergraduate and beginning graduate students.
We aim to give the subject as wide an accessibility as possible to math-
ematically-minded students and to emphasise the interesting mathematical
issues that it raises. We do this by relating the theory to practical applica-
tions where surprising phenomena occur and where innovative mathematical
methods are needed.Our layout is essentially pragmatic. Although more advanced texts in solid
mechanics often begin with quite general theories founded on basic mechan-
ical and thermodynamic principles, we start from the very simplest models,
based on elementary observations in engineering and physics, and build our
way towards models that are the basis for current applied research in solid
mechanics. Hence, we begin by deriving the basic Navier equations of linear
elasticity, before illustrating the mathematical techniques that allow these
equations to be solved in many different practically relevant situations, bothstatic and dynamic. We then proceed to describe some approximate theories
for the elastic deformation of thin solids, namely bars, strings, beams, rods,
plates and shells. We soon discover that many everyday phenomena, such as
the buckling of a beam under a compressive load, cannot be fully described
xiii
8/23/2019 Applied Solid Mechanics
17/468
xiv Prologue
using linear theories. We therefore give a brief exposition of the general the-
ory of nonlinear elasticity, and then show how formal asymptotic methods
allow simplified linear and weakly nonlinear models to be systematically de-
duced. Although we regard such asymptotic techniques as invaluable to any
applied mathematician, these last two topics may both be omitted on a firstreading without loss of continuity. We go on to present simple models for
fracture and contact, comparing and contrasting these apparently similar
phenomena. Next, we show how plasticity theory can be used to describe
situations where a solid yields under a sufficiently high stress. Finally, we
show how elasticity theory may be generalised to include further physical ef-
fects, such as thermal stresses, viscoelasticity and porosity. These combined
fields of solid mechanics are increasingly finding applications in industrial
and medical processes, and pose ever more elaborate modelling questions.Despite the breadth of the models and relevant techniques that will emerge
in this book, we will usually try to present the theoretical developments
ab initio. Nonetheless, the book is very far from being self-contained. Any
student who aspires to becoming a solid mechanics specialist will have to
delve further into the literature, and we will provide references to help with
this.
We assume only that the reader has a reasonable familiarity with the
calculus of several variables. Fluency with the more advanced techniques
required for Chapters 6 and 7, in particular, will readily be acquired by
a student who works through the exercises in the early Chapters, espe-
cially those cited in the text. Indeed, we firmly believe that solid mechanics
provides a wonderful arena in which to build an understanding of such im-
portant mathematical areas as linear algebra, partial differential equations,
complex variable theory, differential geometry and the calculus of variations.
Our hope is that, having read this book, a student should be able to confrontany practical problem that may be encountered in everyday solid mechanics
with at least some idea of the basic mathematical modelling that will be
required.
During the writing of this book, we received a great deal of help and inspi-
ration as a result of discussions with David Allwright, Jon Chapman, Sam
Howison, L. Mahadevan, Roman Novokshanov and Domingo Salazar, as well
as many other colleagues and students too numerous to thank individually.
We would like to express our particular gratitude to Gareth Jones, HilaryOckendon and Tom Witelski who gave invaluable advice on draft Chapters.
We are also indebted to David Tranah and his colleagues at Cambridge
University Press for helping to make this book a reality.
8/23/2019 Applied Solid Mechanics
18/468
1
Modelling solids
1.1 Introduction
In everyday life we regularly encounter physical phenomena that apparently
vary continuously in space and time. Examples are the bending of a paper
clip, the flow of water or the propagation of sound or light waves. Such phe-
nomena can be described mathematically, to lowest order, by a continuum
model, and this book will be concerned with that class of continuum models
that describes solids. Hence, at least to begin with, we will avoid all consid-eration of the atomistic structure of solids, even though these ideas lead
to great practical insight and also to some beautiful mathematics. When
we refer to a solid particle, we will be thinking of a very small region of
matter but one whose dimension is nonetheless much greater than an atomic
spacing.
For our purposes, the diagnostic feature of a solid is the way in which it
responds to an applied system of forces and moments. There is no hard-and-fast rule about this but, for most of this book, we will say that a continuum is
a solid when the response consists of displacements distributed through the
material. In other words, the material starts at some reference state, from
which it is displaced by a distance that depends on the applied forces. This is
in contrast with a fluid, which has no special rest state and responds to forces
via a velocity distribution. Our modelling philosophy is straightforward. We
take the most fundamental pieces of experimental evidence, for example
Hookes law, and use mathematical ideas to combine this evidence withthe basic laws of mechanics to construct a model that describes the elastic
deformation of a continuous solid. Following this simple approach, we will
find that we can construct solid mechanics theories for phenomena as diverse
as earthquakes, ultrasonic testing and the buckling of railway tracks.
1
8/23/2019 Applied Solid Mechanics
19/468
2 Modelling solids
By basing our theory on Hookes law, the simplest model of elasticity,
for small enough forces and displacements, we will first be led to a system
of differential equations that is both linear, and therefore mathematically
tractable, and reversible for time-dependent problems. By this we mean
that, when forces and moments are applied and then removed, the systemeventually returns to its original state without any significant energy being
lost, i.e. the system is not dissipative.
Reversibility may apply even when the forces and displacements are so
large that the problem ceases to be linear; a rubber band, for example,
can undergo large displacements and still return to its initial state. How-
ever, nonlinear elasticity encompasses some striking new behaviours not
predicted by linear theory, including the possibility of multiple steady states
and buckling. For many materials, experimental evidence reveals that evenmore dramatic changes can take place as the load increases, the most strik-
ing phenomenon being that of fracture under extreme stress. On the other
hand, as can be seen by simply bending a metal paper clip, irreversibility
can readily occur and this is associated with plastic flow that is significantly
dissipative. In this situation, the solid takes on some of the attributes of a
fluid, but the model for its flow is quite different from that for, say, water.
Practical solid mechanics encompasses not only all the phenomena men-
tioned above but also the effects of elasticity when combined with heat
transfer (leading to thermoelasticity) and with genuine fluid effects, in cases
where the material flows even in the absence of large applied forces (leading
to viscoelasticity) or when the material is porous (leading to poroelastic-
ity). We will defer consideration of all these combined fields until the final
chapter.
1.2 Hookes lawRobert Hooke (1678) wrote
it is . . . evident that the rule or law of nature in every springing body is that theforce or power thereof to restore itself to its natural position is always proportionateto the distance or space it is removed therefrom, whether it be by rarefaction, orseparation of its parts the one from the other, or by condensation, or crowding ofthose parts nearer together.
Hookes observation is exemplified by a simple high-school physics experi-
ment in which a tensile force T is applied to a spring whose natural length
is L. Hookes law states that the resulting extension of the spring is propor-
tional to T: if the new length of the spring is , then
T = k( L), (1.2.1)
where the constant of proportionality k is called the spring constant.
8/23/2019 Applied Solid Mechanics
20/468
1.3 Lagrangian and Eulerian coordinates 3
Hooke devised his law while designing clock springs, but noted that it
appears to apply to all springy bodies whatsoever, whether metal, wood,
stones, baked earths, hair, horns, silk, bones, sinews, glass and the like. In
practice, it is commonly observed that k scales with 1/L; that is, everything
else being equal, a sample that is initially twice as long will stretch twiceas far under the same force. It is therefore sensible to write (1.2.1) in the
form
T = k L
L, (1.2.2)
where k is the elastic modulus of the spring, which will be defined more
rigorously in Chapter 2. The dimensionless quantity ( L)/L, measuring
the extension relative to the initial length, is called the strain.Equation (1.2.2) is the simplest example of the all-important constitutive
law relating the force to displacement. As shown in Exercise 1.3, it is possible
to construct a one-dimensional continuum model for an elastic solid from
this law, but, to generalise it to a three-dimensional continuum, we first need
to generalise the concepts of strain and tension.
1.3 Lagrangian and Eulerian coordinatesSuppose that a three-dimensional solid starts, at time t = 0, in its rest
state, or reference state, in which no macroscopic forces exist in the solid
or on its boundary. Under the action of any subsequently applied forces
and moments, the solid will be deformed such that, at some later time t, a
particle in the solid whose initial position was the point X is displaced
to the point x (X, t). This is a Lagrangian description of the continuum: if
the independent variable X is held fixed as t increases, then x(X, t) labels
a material particle. In the alternative Eulerian approach, we consider the
material point which currently occupies position x at time t, and label its
initial position by X(x, t). In short, the Eulerian coordinate x is fixed in
space, while the Lagrangian coordinate X is fixed in the material.
The displacementu(X, t) is defined in the obvious way to be the difference
between the current and initial positions of a particle, that is
u(X, t) = x(X, t) X. (1.3.1)
Many basic problems in solid mechanics amount to determining the dis-
placement field u corresponding to a given system of applied forces.
The mathematical consequence of our statement that the solid is a con-
tinuum is that there must be a smooth one-to-one relationship between X
and x, i.e. between any particles initial position and its current position.
8/23/2019 Applied Solid Mechanics
21/468
4 Modelling solids
This will be the case provided the Jacobian of the transformation from X
to x is bounded away from zero:
0 < J < , where J = det xiXj
. (1.3.2)The physical significance of J is that it measures the change in a small
volume compared with its initial volume:
dx1dx2dx3 = JdX1dX2dX3, or dx = JdX (1.3.3)
as shorthand. The positivity ofJ means that we exclude the possibility that
the solid turns itself inside-out.
We can use (1.3.3) to derive a kinematic equation representing conserva-tion of mass. Consider a moving volume V(t) that is always bounded by
the same solid particles. Its mass at time t is given, in terms of the density
(X, t), by
M(t) =
V(t)
dx =
V(0)
JdX. (1.3.4)
Since V(t) designates a fixed set of material points, M(t) must be a constant,namely its initial value M(0):
V(0)JdX= M(t) = M(0) =
V(0)
0 dX, (1.3.5)
where 0 is the density in the rest state. Since V is arbitrary, we deduce that
J = 0. (1.3.6)
Hence, we can calculate the density at any time t in terms of 0 and the
displacement field. The initial density 0 is usually taken as constant, but
(1.3.6) also applies if 0 = 0(X).
1.4 Strain
To generalise the concept of strain introduced in Section 1.2, we consider thedeformation of a small line segment joining two neighbouring particles with
initial positions X and X+ X. At some later time, the solid deforms such
that the particles are displaced to X+u(X, t) and X+ X+u(X+ X, t)
respectively. Thus we can use Taylors theorem to show that the line element
8/23/2019 Applied Solid Mechanics
22/468
1.4 Strain 5
X that joins the two particles is transformed to
x = X+u(X+ X, t)u(X, t) = X+ (X)u(X, t) + , (1.4.1)
where
(X ) = X1 X1
+ X2 X2
+ X3 X3
. (1.4.2)
Let L = |X| and = |x| denote the initial and current lengths respectively
of the line segment; the difference L is known as the stretch. Then, to
lowest order in L,
2 = |X+ (X )u(X, t)|2. (1.4.3)
Although we will try in subsequent chapters to minimise the use ofsuffices, it is helpful at this stage to introduce components so that
X= (Xi) = (X1, X2, X3)T and similarly for u. Then (1.4.3) may be written
in the form
2 L2 = 23
i,j =1
Eij XiXj , (1.4.4)
where
Eij =1
2
uiXj
+ujXi
+3
k=1
ukXi
ukXj
. (1.4.5)
By way of introduction to some notation that will be useful later, we point
out that (1.4.4) may be written in at least two alternative ways. First, we
may invoke the summation convention, in which one automatically sums over
any repeated suffix. This avoids the annoyance of having to write explicit
summation, so (1.4.4) is simply
2 = L2 + 2Eij XiXj , where Eij =1
2
uiXj
+ujXi
+ukXi
ukXj
.
(1.4.6)
Second, we note that 2 L2 is a quadratic form on the symmetric matrix
E whose components are (Eij ):
2 L2 = 2 XTEX. (1.4.7)
It is clear from (1.4.4) that the stretch is measured by the quantities Eij ;
in particular, the stretch is zero for all line elements if and only ifEij 0. It
is thus natural to identify Eij with the strain. Now let us ask: what happens
when we perform the same calculation in a coordinate system rotated by an
8/23/2019 Applied Solid Mechanics
23/468
6 Modelling solids
orthogonal matrix P = (pij )? Intuitively, we might expect the strain to be
invariant under such a rotation, and we can verify that this is so as follows.
The vectors X and u are transformed to X and u in the new coordinate
system, where
X = PX, u = Pu. (1.4.8)
Since P is orthogonal, (1.4.8) may be inverted to give X= PTX. Alterna-
tively, using suffix notation, we have
X = pjX
j , ui = pi u. (1.4.9)
The strain in the new coordinate system is denoted by
Eij =1
2
uiXj
+ujXi
+ukXi
ukXj
, (1.4.10)
which may be manipulated using the chain rule, as shown in Exercise 1.4,
to give
Eij = pipjE. (1.4.11)
In matrix notation, (1.4.11) takes the form
E = PEPT, (1.4.12)
so the 3 3 symmetric array (Eij ) transforms exactly like a matrix repre-
senting a linear transformation of the vector space R3. Arrays that obey the
transformation law (1.4.11) are called second-rank Cartesian tensors, and
E= (Eij ) is therefore called the strain tensor.
Almost as important as the fact that E is a tensor is the fact that itcan vanish without u vanishing. More precisely, if we consider a rigid-body
translation and rotation
u = c + (Q I)X, (1.4.13)
where I is the identity matrix while the vector c and orthogonal matrixQ are constant, then E is identically zero. This result follows directly from
substituting (1.4.13) into (1.4.6) and using the fact that QQT = I, andconfirms our intuition that a rigid-body motion induces no deformation.
The word tensor as used here is effectively synonymous with matrix, but it is easy togeneralise (1.4.11) to a tensor with any number of indices. A vector, for example, is a tensorwith just one index.
8/23/2019 Applied Solid Mechanics
24/468
1.5 Stress 7
1.5 Stress
In the absence of any volumetric (e.g. gravitational or electromagnetic) ef-
fects, a force can only be transmitted to a solid by being applied to its
boundary. It is, therefore, natural to consider the force per unit area or
stress applied at that boundary. To do so, we now analyse an infinitesimalsurface element, whose area and unit normal are da and n respectively. If it
is contained within a stressed medium, then the material on (say) the side
into which n points will exert a force df on the element. (By Newtons third
law, the material on the other side will also exert a force equal to df.) In
the expectation that the force should be proportional to the area da, we
write
df = da, (1.5.1)
where is called the traction or stress acting on the element.
Perhaps the most familiar example is that of an inviscid fluid, in which
the stress is related to the pressure p by
= pn. (1.5.2)
This expression implies that (i) the stress acts only in a direction normal
to the surface element, (ii) the magnitude of the stress (i.e. p) is indepen-dent of the direction ofn. In an elastic solid, neither of these simplifying
assumptions holds; we must allow for stress which acts in both tangential
and normal directions and whose magnitude depends on the orientation of
the surface element.
First consider a surface element whose normal points in the x1-direction,
and denote the stress acting on such an element by 1 = (11 , 21, 31)T. By
doing the same for elements with normals in the x2- and x3-directions, we
generate three vectors j (j = 1, 2, 3), each representing the stress acting
on an element normal to the xj -direction. In total, therefore, we obtain nine
scalars ij (i, j = 1, 2, 3), where ij is the i-component ofj , that is
j = ijei, (1.5.3)
where ei is the unit vector in the xi-direction.
The scalars ij may be used to determine the stress on an arbitrary surface
element by considering the tetrahedron shown in Figure 1.1. Here ai denotesthe area of the face orthogonal to the xi-axis. The fourth face has area
a =
a21 + a22 + a
23; in fact if this face has unit normal n as shown, with
components (ni), then it is an elementary exercise in trigonometry to show
that ai = ani.
8/23/2019 Applied Solid Mechanics
25/468
8 Modelling solids
a2
a3
x1
x3
x2
n
a1
Fig. 1.1 A reference tetrahedron; ai is the area of the face orthogonal to the xi -axis.
The outward normal to the face with area a1 is in the negative x1-direction
and the force on this face is thus a11. Similar expressions hold for the
faces with areas a2 and a3. Hence, if the stress on the fourth face is denoted
by , then the total force on the tetrahedron is
f = a ajj . (1.5.4)
When we substitute for aj and j , we find that the components of f aregiven by
fi = a (i ij nj ) . (1.5.5)
Now we shrink the tetrahedron to zero volume. Since the area a scales
with 2, where is a typical edge length, while the volume is proportional
to 3, if we apply Newtons second law and insist that the acceleration be
finite, we see that f/a must tend to zero as 0.
Hence we deduce an
Readers of a sensitive disposition may be slightly perturbed by our glibly letting the dimensionalvariable tend to zero: if is reduced indefinitely then we will eventually reach an atomic scaleon which the solid can no longer be treated as a continuum. We reassure such readers that(1.5.6) can be more rigorously justified provided the macroscopic dimensions of the solid arelarge compared to any atomistic length-scale.
8/23/2019 Applied Solid Mechanics
26/468
1.5 Stress 9
G11
11
22
21
22
12
12
21
x2
x1
x2
x1
Fig. 1.2 The forces acting on a small two-dimensional element.
expression for :
i = ij nj , or = n. (1.5.6)
This important result enables us to find the stress on any surface element
in terms of the nine quantities (ij ) = .
Now let us follow Section 1.4 and examine what happens to ij when we
rotate the axes by an orthogonal matrix P. In the new frame, (1.5.6) will
become
= n (1.5.7)
where, since and n are vectors, they transform according to
= P, n = Pn. (1.5.8)
It follows that n = (P PT)n and so, since n is arbitrary,
= P PT, or ij = pipj. (1.5.9)
Thus ij , like Eij , is a second-rank tensor, called the Cauchy stress tensor.
We can make one further observation about ij by considering the angular
momentum of the small two-dimensional solid element shown in Figure 1.2.The net anticlockwise moment acting about the centre of mass G is (per
unit length in the x3-direction)
2 (21x2)x1
2 2 (12 x1)
x22
,
8/23/2019 Applied Solid Mechanics
27/468
10 Modelling solids
where 21 and 12 are evaluated at G to lowest order. By letting the rectangle
shrink to zero (see again the footnote on page 8), and insisting that the
angular acceleration be finite, we deduce that 12 = 21 . This argument can
be generalised to three dimensions (see Exercise 1.5) and it shows that
ij ji (1.5.10)
for all i and j, i.e. that ij , like Eij , is a symmetric tensor.
1.6 Conservation of momentum
Now we derive the basic governing equation of solid mechanics by apply-
ing Newtons second law to a material volume V(t) that moves with the
deforming solid:
d
dt
V(t)
uit
dx =
V(t)
gi dx +
V(t)
ij nj da. (1.6.1)
The terms in (1.6.1) represent successively the rate of change of momentum
of the material in V(t), the force due to an external body force g, such as
gravity, and the traction exerted on the boundary of V, whose unit normal
is n, by the material around it. We differentiate under the integral (using
the fact that dx = 0 dX is independent of t) and apply the divergencetheorem to the final term to obtain
V(t)
2uit2
dx =
V(t)
gi dx +
V(t)
ijxj
dx. (1.6.2)
Assuming each integrand is continuous, and using the fact that V(t) is ar-
bitrary, we arrive at Cauchys momentum equation:
2u
it2 = g
i +
ijxj
. (1.6.3)
This may alternatively be written in vector form by adopting the following
notation for the divergence of a tensor: we define the ith component of
to be
( )i =jixj
. (1.6.4)
Since is symmetric, we may thus write Cauchys equation as
2u
t2= g + . (1.6.5)
This equation applies to any continuous medium for which a displacement
u and stress tensor can be defined. The distinction between solid, fluid
8/23/2019 Applied Solid Mechanics
28/468
1.7 Linear elasticity 11
or some other continuum comes when we impose an empirical constitutive
relation between and u.
For solids, (1.6.5) already confronts us with a distinctive fundamental
difficulty. The most obvious generalisation of Hookes law is to suppose that
a linear relationship exists between the stress and the strain E. But wenow recall that E was defined in Section 1.4 in terms of the Lagrangian
variables X; indeed, the time derivative in (1.6.5) is taken in a Lagrangian
frame, with Xfixed. On the other hand, the stress tensor has been defined
relative to Eulerian coordinates and is differentiated in (1.6.5) with respect
to the Eulerian variable x. It is not immediately clear, therefore, how the
stress and strain, which are defined in different frames of reference, may be
self-consistently related. We will postpone the full resolution of this difficulty
until Chapter 5 and, for the present, restrict our attention to linear elasticityin which, as we shall see, the two frames are essentially identical.
1.7 Linear elasticity
The theory of linear elasticity follows from the assumption that the dis-
placement u is small relative to any other length-scale. This assumption
allows the theory developed thus far to be simplified in several ways. First,it means that ui/Xj is small for all i and j. Second, we note from (1.3.1)
that x and X are equal to lowest order in u. Hence, if we only consider
leading-order terms, there is no need to distinguish between the Eulerian
and Lagrangian variables: we can simply replace X by x and ui/Xj by
ui/xj throughout. A corollary is that the Jacobian J is approximately
equal to one, so (1.3.6) tells us that the density is fixed, to leading order,
at its initial value 0. Finally, we can use the smallness of ui/xj to neglect
the quadratic term in (1.4.5) and hence obtain the linearised strain tensor
Eij eij =1
2
uixj
+ujxi
. (1.7.1)
Much of this book will be concerned with this approximation. Therefore,
and with a slight abuse of notation, we will write E= (eij ).
Remembering (1.4.13), we note that it is possible to approximate E by
(1.7.1) even when u is not small compared with X, just as long as u is closeto a rigid-body translation and rotation. This situation is called geometric
nonlinearity and we will encounter it frequently in Chapters 4 and 6. It
occurs because Eij is identically zero for rigid-body motions of the solid
given by (1.4.13); however eij does not vanish for such rigid-body motions,
8/23/2019 Applied Solid Mechanics
29/468
12 Modelling solids
but rather for displacements of the form
u = c + x, (1.7.2)
where c and are constant (see Exercise 1.6).
Assuming the validity of (1.7.1), we can now generalise Hookes law bypostulating a linear relationship between the stress and strain tensors. We
assume that is zero when E is; in other words the stress is zero in the
reference state. This is not the case for pre-stressed materials, and we will
consider some of the implications of so-called residual stress in Chapter 8.
Even with this assumption, we apparently are led to the problem of defining
81 material parameters Cijk (i,j,k, = 1, 2, 3) such that
ij = Cijkek . (1.7.3)
The symmetry of ij and eij only enables us to reduce the number of
unknowns to 36. This can be reduced to a more manageable number by
assuming that the solid is isotropic, by which we mean that it behaves the
same way in all directions. This implies that Cijk must satisfy
Cijkpiipjjpkk p Cijk (1.7.4)
for all orthogonal matrices P = (pij ). It can be shown (see, for example,Ockendon & Ockendon, 1995, pp. 79) that this is sufficient to reduce the
specification of Cijk to just two scalar quantities and , such that
Cijk = ij k + 2ik j, (1.7.5)
where ij is the usual Kronecker delta, which represents the identity matrix;
consequently,
ij = (ekk ) ij + 2eij . (1.7.6)This relation can also be inverted to give the strain corresponding to a given
stress, that is
eij =1
2
ij
(kk )
(3 + 2)ij
. (1.7.7)
In Chapter 9 we will consider solids, such as wood or fibre-reinforced mate-
rials, that are not isotropic, and for which (1.7.6) must be generalised.
The material parameters and are known as the Lame constants, and is called the shear modulus.As we shall see in Chapter 2, and measure a
materials ability to resist elastic deformation. They have the units of pres-
sure; typical values for a few familiar solid materials are given in Table 1.1.
It will be observed that these values may be very large for relatively hard
8/23/2019 Applied Solid Mechanics
30/468
1.8 The incompressibility approximation 13
(GPa) (GPa)
Cartilage 3 105 9 105
Rubber 0.04 0.003Polystyrene 2.3 1.2Granite 10 30
Glass 28 28Copper 86 37Steel 100 78Diamond 270 400
Table 1.1 Typical values of the Lame constants and for some everyday
materials (1 GPa = 109 N m2 = 104 atmospheres; a typical car tyre
pressure is two atmospheres).
materials, the significance being that tractions much less than these values
will result in small deformations, so that linear elasticity is valid.
Now we substitute our linear constitutive relation (1.7.6) into the momen-
tum equation (1.6.3) and replace X with x to obtain the Navier equation,
also known as the Lame equation,
2u
t2 = g + ( + )graddivu +
2
u. (1.7.8)Recall that does not vary to leading order, so (1.7.8) comprises three
equations for the three components ofu. It may alternatively be written in
component form
2uit2
= gi + ( + )2uj
xixj+
2uix2j
, (1.7.9)
where the final x
2
j is treated as a repeated suffix, or
2u
t2= g + ( + 2)graddivu curl curlu, (1.7.10)
where we have used the well-known vector identity
del squared equals grad div minus curl curl. (1.7.11)
1.8 The incompressibility approximationThere is an interesting and important class of materials that, although elas-
tic, are virtually incompressible, so they may be sheared elastically but are
highly resistant to tension or compression. In linear elasticity, this amounts
to saying that the Lame constant is much larger than the shear modulus .
8/23/2019 Applied Solid Mechanics
31/468
14 Modelling solids
The values given in Table 1.1 show that rubber has this property, as do many
biomaterials such as muscle.
If a material is almost incompressible, we can set
=
1
, (1.8.1)
where is a small parameter. From (1.7.8), we expect that, in the limit
0, divu will be of order . Hence, if we define a scalar function p such
that
p =
divu, (1.8.2)
then p will approach a finite limit p as 0.
When we now substitute (1.8.1) and (1.8.2) into the Navier equation(1.7.8) and let 0, we obtain
2u
t2= g p + 2u, (1.8.3a)
along with the limit of (1.8.2), that is
divu = 0. (1.8.3b)
The condition (1.8.3b) means that each material volume is conserved duringthe deformation, and it imposes an extra constraint on the Navier equation.
The extra unknown p, representing the isotropic pressure in the medium,
gives us the extra freedom we need to satisfy this constraint.
1.9 Energy
We can obtain an energy equation from (1.6.3) by taking the dot product
with u/t and integrating over an arbitrary volume V:V
2uit2
uit
dx =
V
giuit
dx +
V
ijxj
uit
dx. (1.9.1)
The final term may be rearranged, using the divergence theorem, toV
ijxj
uit
dx =
V
uit
ij nj da
V
ijeijt
dx. (1.9.2)
8/23/2019 Applied Solid Mechanics
32/468
1.9 Energy 15
Hence (1.9.1) may be written in the form
d
dt
V
1
2
u
t
2
dx +
VWdx
=
V
giuit
dx +
V
uit
ij nj da, (1.9.3)
where W is a scalar function of the strain components that is chosen to
satisfy
W
eij= ij . (1.9.4)
With ij given by (1.7.6), we can integrate (1.9.4) to determine W up to anarbitrary constant as
W=1
2ij eij =
1
2 (ekk )
2 + (eij eij ). (1.9.5)
Here the summation convention is invoked such that (ekk )2 is the square of
the trace of E, while (eij eij ) is the sum of the squares of the components
ofE.
The first term in braces in (1.9.3) is the net kinetic energy in V, while theterms on the right-hand side represent the rate of working of the external
body force g and the tractions on V respectively. Hence, in the absence of
other energy sources resulting from, say, chemical or thermal effects, we can
interpret equation (1.9.3) as a statement of conservation of energy. The dif-
ference between the rate of working and the rate of change of kinetic energy
is the rate at which elastic energy is stored in the material as it deforms; W
is therefore called the strain energy density. This is analogous to the energy
stored in a stretched spring (see Exercise 1.1) and, at a fundamental scale,
is a manifestation of the energy stored in the bonds between the atoms. If
, > 0, we can easily see from (1.9.5) that W is a non-negative function
of the strain components, whose unique global minimum is attained when
eij = 0. In fact, Exercise 1.7 demonstrates that it is only necessary to have
, ( + 2/3) > 0.
The net conservation of energy implied by (1.9.3) reflects the fact that the
Navier equation is not dissipative. Furthermore, even without the constitu-tive relation (1.7.6), the steady Navier equation is a necessary condition for
the net gravitational and strain energy in an elastic body D, namely
U =
DW g udx, (1.9.6)
8/23/2019 Applied Solid Mechanics
33/468
16 Modelling solids
to be minimised, as shown in Exercise 1.8. However, the situation changes
when thermal effects are important, as we will see in Chapter 9.
1.10 Boundary conditions and well-posednessSuppose that we wish to solve (1.7.8) for u(x, t) when t is positive and x
lies in some prescribed domain D. We now ask: what sort of boundary
conditions may be imposed on D to obtain a well-posed mathematical
problem, in other words, one for which a solution u exists, is unique and
depends continuously on the boundary data? For boundary-value problems
in linear elasticity, it is generally far easier to discuss questions of uniqueness
than it is to prove existence. Hence in this section we will focus only on
establishing uniqueness.
In elastostatic problems, in which the left-hand side of (1.7.8) is zero,
the Navier system is, roughly speaking, a generalisation of a scalar elliptic
equation. By analogy, it seems appropriate for either u or three linearly
independent scalar combinations ofu and u/n to be prescribed on D.
In many physical problems, we specify either the displacement u or the
traction n everywhere on the boundary, and we will now examine each of
these in turn.First consider a solid body D on whose boundary the displacement is
prescribed, that is
u = ub(x) on D. (1.10.1)
Inside D, u satisfies the steady Navier equation
ijxj
+ gi = 0, (1.10.2)
and we will now show that, if a solution u of (1.7.6), (1.10.2) with the
boundary condition (1.10.1) exists, then it is unique.
Suppose that two solutions u(1) and u(2) exist and let u = u(1)u(2) . Thus
u satisfies the homogeneous problem, with ub = g = 0. Now, by multiplying
(1.10.2) by ui, integrating over D and using the divergence theorem, we
obtain
D
uiij nj da =
Deij ij dx = 2
DWdx, (1.10.3)
where W is given by (1.9.5). The left-hand side of (1.10.3) is zero by the
boundary conditions, while the integrand W on the right-hand side is non-
negative and must, therefore, be zero. It follows that the strain tensor eij
8/23/2019 Applied Solid Mechanics
34/468
1.10 Boundary conditions and well-posedness 17
is identically zero in D, and the displacement can therefore only be a rigid-
body motion (i.e. a uniform translation and rotation; see Exercise 1.6). Since
u is zero on D, we deduce that it must be zero everywhere and, hence, that
u(1) u(2) .
Now we attempt the same calculation when the surface traction, ratherthan the displacement, is specified:
n = (x) on D. (1.10.4)
Like the Neumann problem for a scalar elliptic partial differential equation
(Ockendon et al., 2003, p. 154), the Navier equation only admits solutions
satisfying (1.10.4) if so-called solvability conditions are satisfied. If we inte-
grate (1.10.2) over D and use the divergence theorem, we find thatD
ij nj da +
D
gi dx = 0 (1.10.5)
and hence that D
da +
D
g dx = 0. (1.10.6)
This represents a net balance between the forces, namely surface traction
and gravity, acting on D. An analogous balance between the moments actingon D may also be obtained by taking the cross product of x with (1.10.2)
before integrating, to giveD
x da +
D
xg dx = 0, (1.10.7)
as shown in Exercise 1.9. As well as representing physical balances on the
system, (1.10.6) and (1.10.7) may be interpreted as instances of the Fredholm
Alternative (see Ockendon et al., 2003, p. 43).Now suppose the solvability conditions (1.10.6) and (1.10.7) are satisfied
and that two solutions u(1) and u(2) of (1.10.2) and the boundary condition
(1.10.4) exist. As before, the difference u = u(1) u(2) satisfies the homo-
geneous version of the problem, with g and set to zero. By an argument
analogous to that presented above, we deduce that the strain tensor eij must
be identically zero. However, since u is now not specified on D, we can only
infer from this that the displacement is a rigid-body motion, as shown in
Exercise 1.6. Thus the solution of (1.10.2) subject to the applied traction(1.10.4) is determined only up to the addition of an arbitrary translation
and rotation.
As well as the boundary conditions (1.10.1) and (1.10.4), there are gener-
alisations in which the traction is specified on some parts of the boundary
8/23/2019 Applied Solid Mechanics
35/468
18 Modelling solids
solid 1
solid 2
n
Fig. 1.3 A small pill-box-shaped region at the boundary between two elastic solids.
and the displacement on others, for example in contact problems and in frac-
ture, as described in Chapter 7. Another common generalisation of (1.10.1)
and (1.10.4) occurs when two solids with different elastic moduli are bondedtogether across a common boundary D, as shown in Figure 1.3. Then the
displacement vectors are the same on either side of D and, by balancing
the stresses on the small pill-box-shaped region shown in Figure 1.3, we see
that
(1)n = (2)n, (1.10.8)
where (1) and (2) are the values of on either side of the boundary. Thus
there are six continuity conditions across such a boundary.
On the other hand, if two unbonded solids are in smooth contact, only
the normal displacement is continuous across D. However, this loss of in-
formation is compensated by the fact that the four tangential components
of (1)n and (2)n are zero and the normal components of these tractions
are continuous. Frictional contact between rough unbonded surfaces poses
serious modelling challenges, as we will see in Chapter 7.For elastodynamic problems, we may anticipate that (1.7.8) admits wave-
like solutions. It may, therefore, be viewed as a generalisation of a scalar
wave equation, such as the familiar equation
2w
t2= T
2w
x2(1.10.9)
which describes small transverse waves on a string with tension T and linedensity (see Section 4.3). We will examine elastic waves in more detail
in Chapter 3 but, in the meantime, we expect to prescribe Cauchy initial
conditions for u and u/t at t = 0, as well as elliptic boundary conditions
such as (1.10.1) or (1.10.4).
8/23/2019 Applied Solid Mechanics
36/468
1.11 Coordinate systems 19
1.11 Coordinate systems
In the next two chapters, we will construct some elementary solutions of the
Navier equation (1.7.8). In doing so, it is often useful to employ coordinate
systems particularly chosen to fit the geometry of the problem being consid-
ered. A detailed derivation of the Navier equation in an arbitrary orthogonal
coordinate system may be found in the Appendix. Here we state the main
results that will be useful in subsequent chapters for the three most popular
coordinate systems, namely Cartesian, cylindrical polar and spherical polar
coordinates.
All three of these coordinate systems are orthogonal; in other words the
tangent vectors obtained by varying each coordinate in turn are mutually
perpendicular. This means that the coordinate axes at any fixed point areorthogonal and may thus be obtained by a rotation of the usual Cartesian
axes. Under the assumptions of isotropic linear elasticity, the Cartesian stress
and strain components are related by (1.7.6), which is invariant under any
such rotation. Hence the constitutive relation (1.7.6) applies literally to any
orthogonal coordinate system.
1.11.1 Cartesian coordinates
First we write out in full the results derived thus far using the usual Carte-
sian coordinates (x,y,z). To avoid the use of suffices, we will denote the
displacement components by u = (u, v, w)T. It is also conventional to la-
bel the stress components by {xx, xy , . . .} rather than {11 , 12 , . . .}, and
similarly for the strain components. The linear constitutive relation (1.7.6)
gives
xx = ( + 2)exx + eyy + ezz , xy = 2exy ,
yy = exx + ( + 2)eyy + ezz , xz = 2exz ,
zz = exx + eyy + ( + 2)ezz , yz = 2eyz , (1.11.1)
where
exx =
u
x , 2exy =
u
y +
v
x ,
eyy =v
y, 2eyz =
v
z+
w
x,
ezz =w
z, 2exz =
u
z+
w
x, (1.11.2)
8/23/2019 Applied Solid Mechanics
37/468
20 Modelling solids
and the three components of Cauchys momentum equation are
2u
t2= gx +
xxx
+xy
y+
xzz
,
2
vt2
= gy + xyx
+ yyy
+ yzz
,
2w
t2= gz +
xzx
+yzy
+zzz
, (1.11.3)
where the body force is g = (gx, gy , gz )T. In terms of the displacements, the
Navier equation reads (assuming that and are constant)
2
ut2 = gx + ( + ) x ( u) +
2u,
2v
t2= gy + ( + )
y( u) + 2v,
2w
t2= gz + ( + )
z( u) + 2w. (1.11.4)
1.11.2 Cylindrical polar coordinatesWe define cylindrical polar coordinates (r,,z) in the usual way and de-
note the displacements in the r-, - and z-directions by ur, u and uz re-
spectively. The stress components are denoted by ij where now i and j
are equal to either r, or z and, as in Section 1.5, ij is defined to be the
i-component of stress on a surface element whose normal points in the j-
direction. As noted above, the constitutive relation (1.7.6) applies directly
to this coordinate system, so that
rr = ( + 2)err + e + ezz , r = 2er ,
= err + ( + 2)e + ezz , rz = 2erz ,
zz = err + e + ( + 2)ezz , z = 2ez , (1.11.5)
where the strain components are now given by
err =
ur
r , 2er =
1
r
ur
+
u
r
u
r ,
e =1
r
u
+ ur
, 2erz =
urz
+uzr
,
ezz =uzz
, 2ez =uz
+1
r
uz
. (1.11.6)
8/23/2019 Applied Solid Mechanics
38/468
1.11 Coordinate systems 21
The three components of Cauchys momentum equation (1.6.3) read
2urt2
= gr +1
r
r(rrr ) +
1
r
r
+rzz
r
,
2
ut2 = g + 1r r (r
r ) + 1r
+ z
z + rr ,
2uzt2
= gz +1
r
r(rrz ) +
1
r
z
+zzz
, (1.11.7)
where the body force is g = grer + ge + gzez . Written out in terms of
displacements, these become
2ur
t2
= gr + ( + )
r
( u) + 2ur urr
2
2
r2
u
,
2ut2
= g +( + )
r
( u) +
2u
ur2
+2
r2ur
,
2uzt2
= gz + ( + )
z( u) + 2uz , (1.11.8)
where
u =
1
r
r (rur) +
1
r
u
+
uz
z ,
2ui =1
r
r
r
uir
+
1
r22ui2
+2uiz2
(1.11.9)
are the divergence ofu and the Laplacian of ui respectively, expressed in
cylindrical polars.
Detailed derivations of (1.11.6) and (1.11.7) are given in the Appendix.
Notice the undifferentiated terms proportional to 1/r which are not present
in the corresponding Cartesian expressions (1.11.2) and (1.11.3). The originof these terms may be understood in two dimensions (r, ) by considering
the equilibrium of a small polar element as illustrated in Figure 1.4, in which
= (r + r,) = + r
r+ ,
= (r, + ) = +
+ , (1.11.10)
when we expand using Taylors theorem. Summing the resultant forces inthe r- and -directions to zero results in
rr (r + r) rr r r sin + r r cos r r = 0,
r cos r + r r sin + r (r + r) r r = 0. (1.11.11)
8/23/2019 Applied Solid Mechanics
39/468
22 Modelling solids
r
rrr
e
r
rr r
r r
er
Fig. 1.4 Forces acting on a polar element of solid.
Now letting ,r 0 and using (1.11.10), we obtain
rrr +
1
r
r +
rr
r = 0,
rr +1
r
+
2rr = 0, (1.11.12)
which are the components of the two-dimensional steady Navier equation in
plane polar coordinates with no body force; cf (1.11.7). The stress component
is the so-called hoop stress in the -direction that results from inflating
an elastic object radially; we will see an explicit example of hoop stress in
Section 2.6.
1.11.3 Spherical polar coordinates
The spherical polar coordinates (r,,) are defined in the usual way, such
that the position vector of any point is given by
r(r,,) =
r sin cos r sin sin
r cos
. (1.11.13)
Again, we can apply the constitutive relation (1.7.6) literally, to obtainrr = ( + 2)err + e + e , r = 2er ,
= err + ( + 2)e + e , r = 2er ,
= err + e + ( + 2)e , = 2e . (1.11.14)
8/23/2019 Applied Solid Mechanics
40/468
1.11 Coordinate systems 23
The linearised strain components are now given by
err =urr
, 2er =1
r
ur
+ur
ur
,
e = 1r
u
+ ur
, 2er = 1r sin
ur
+ ur
ur
,
e =1
r sin
u
+urr
+u cot
r, 2e =
1
r sin
u
+1
r
u
u cot
r.
(1.11.15)
Cauchys equation of motion leads to the three equations
2
urt2
= gr + 1r2
(r2
rr )r
+ 1r sin
(sin r )
+1
r sin
r
+
r,
2ut2
= g +1
r2(r2r )
r+
1
r sin
(sin )
+1
r sin
+r cot
r,
2ut2
= g +1
r2(r2r )
r+
1
r sin
(sin )
+1
r sin
+r + cot
r, (1.11.16)
where the body force is g = grer +ge +ge. Again, (1.11.15) and (1.11.16)
may be derived using the general approach given in the Appendix or more
directly by analysing a small polar element. In terms of displacements, the
Navier equation reads
2urt2
= gr + ( + )
r( u)
+
2ur
2urr2
2
r2 sin
(u sin )
2
r2 sin
u
,
2ut2
= g +( + )
r
( u)
+
2u +
2
r2ur
u
r2 sin2
2cos
r2 sin2
u
,
8/23/2019 Applied Solid Mechanics
41/468
24 Modelling solids
2ut2
= g +( + )
r sin
( u)
+
2u +
2
r2 sin
ur
+2cos
r2 sin2
u
u
r2 sin2
, (1.11.17)
where
u =1
r2
r
r2ur
+
1
r sin
(sin u) +
1
r sin
u
,
2ui =1
r2
r
r2
uir
+
1
r2 sin
sin
ui
+
1
r2 sin2
2ui2
.
(1.11.18)
Exercises
1.1 A light spring of natural length L and spring constant k hangs freely
with a mass m attached to one end and the other end fixed. Show
that the length of the spring satisfies the differential equation
md2
dt2+ k( L) mg = 0.
Deduce that
1
2m
d
dt
2+
1
2k( L)2 + mg(L ) = const.
and interpret this result in terms of energy.
1.2 A string, stretched to a tension Talong the x-axis, undergoes small
transverse displacements such that its position at time t is given
by the graph z = w(x, t). Given that w satisfies the wave equation
(1.10.9), where is the mass per unit length of the string, show that,if x = a and x = b are any two points along the string,
d
dt
1
2
ba
w
t
2 dx +
ba
1
2
w
x
2Tdx
=
T
w
x
w
t
ba
.
Interpret this result in terms of conservation of energy.
1.3 A system of masses m along the x-axis at positions Xn = nL
(n = 0, 1, 2, . . .) are linked by springs satisfying Hookes law (1.2.1),as shown in Figure 1.5. If each mass is displaced by a distance un(t),
show that the tension Tn joining Xn to Xn+1 satisfies
Tn Tn1 = md2undt2
, where Tn = k (un+1 un).
8/23/2019 Applied Solid Mechanics
42/468
Exercises 25
m mm
Tn1 Tn1 Tn Tn
unun1 un+1
Fig. 1.5 A system of masses connected by springs along the x-axis.
Deduce that
md2undt2
= k (un+1 2un + un1)
and show that this is a spatial discretisation of the partial differential
equation
2ut2
= c2 2u
X2,
where X = nL, un(t) = u(X, t) and c2 = kL2/m.
[It is not so easy to use discrete element models in more than
one dimension; for example, it can be shown that it is impossible to
retrieve the Navier equation as the limit of a lattice of masses joined
by springs aligned along three orthogonal axes.]
1.4 If x, x, u and u are related by (1.4.9), use the chain rule to showthat
uixj
= pipjux
.
Hence establish equation (1.4.11) to show that Eij transforms as a
tensor under a rotation of the coordinate axes.
[Hint: note that, since P is orthogonal, pklpkm lm .]
1.5 Consider a volume V(t) that is fixed in a deforming solid body. Showthat conservation of angular momentum for V(t) leads to the equa-
tion
d
dt
V(t)
xu
t dx =
V(t)
xg dx +
V(t)
x(n) da.
From this and Cauchys momentum equation (1.6.3), deduce that
ij ji.
1.6 Show that the linearised strain eij , given by (1.7.1), is identically zeroif and only ifu = c + x, where c and are spatially-uniform
vectors. Show that this approximates the rigid-body motion (1.4.13)
when P is close to I. Ifu is of this form, and known to be zero at
three non-collinear points, deduce that c = = 0.
8/23/2019 Applied Solid Mechanics
43/468
26 Modelling solids
1.7 By writing the linearised strain tensor eij as the sum of a zero-
trace contribution eij (1/3)ij ekk and a purely diagonal contribu-
tion (1/3)ij ekk , show that W can be rewritten as
W=
2 +
3
(ekk )2
+
eij 1
3 ij ekk
eij 1
3 ij ekk
.
Deduce that, for W to have a single global minimum at eij = 0, it is
sufficient for and +2/3 to be positive. By considering particular
values of eij , show that it is also necessary.
1.8 Suppose a solid body occupies the region D and the displacement u
is prescribed on D. Let
U =
D W g udx,
where W is the strain energy density. Show that, if ui is changed by
a small virtual displacement i, then the corresponding leading-order
change in U is
U =
D
ixj
ij gii
dx.
[Hint: use the fact thatij is symmetric.] Use the divergence theoremto show that
U =
D
iij nj da
D
ijxj
+ gi
i dx,
and deduce that the minimisation of U with respect to all displace-
ments satisfying the given boundary condition leads to the steady
Navier equation. Deduce also that, if no boundary condition is im-
posed, the natural boundary condition is the vanishing of the tractionij nj on D.
Show also that the minimisation of U subject to the constraint
divu 0 leads to the steady incompressibleNavier equation (1.8.3a),
where p is a Lagrange multiplier.
1.9 An elastic body D at rest is subject to a traction n = (x) on its
boundary D. By taking the cross product ofx with (1.10.2) before
integrating over D, derive (1.10.7) and deduce that the net moment
acting on D must be zero.1.10 Suppose that u satisfies the steady Navier equation (1.10.2) in a
region D and the mixed boundary condition (x)u+(x)n = f(x)
on D. Show that, if a solution exists, it is unique provided > 0
and 0.
8/23/2019 Applied Solid Mechanics
44/468
Exercises 27
[If and take different signs, then there is no guarantee of
uniqueness. This is analogous to the difficulty associated with the
Robin boundary condition for scalar elliptic partial differential equa-
tions (Ockendon et al., 2003, p. 154).]
1.11 (a) Show that, in plane polar coordinates (r, ), the basis vectorssatisfy
derd
= e ,ded
= er .
(b) Consider a small line segment joining two particles whose po-
lar coordinates are (r, ) and (r + r, + ). Show that the
vector joining the two particles is given to leading order by
X rer + re .
(c) If a two-dimensional displacement field is imposed, with
u = ur (r, )er () + u(r, )e(), show that the line element
X is displaced to
x = X+
urr
r +ur
u
er
+
ur
r + u
+ ur e .
(d) Hence show that
|x|2 = |X|2 +(r,r)
err erer e
r
r
where, to leading order in the displacements,
err = urr , 2er = 1r u
r + u
r u
r , e
= 1r
u + ur
.
[These are the elements of the linearised strain tensor in plane
polar coordinates, as in (1.11.6).]
8/23/2019 Applied Solid Mechanics
45/468
2
Linear elastostatics
2.1 Introduction
This chapter concerns steady state problems in linear elasticity. This topic
may appear to be the simplest in the whole of solid mechanics, but we will
find that it offers many interesting mathematical challenges. Moreover, the
material presented in this chapter will provide crucial underpinning to the
more general theories of later chapters.
We will begin by listing some very simple explicit solutions which give
valuable intuition concerning the role of the elastic moduli introduced in
Chapter 1. Our first application of practical importance is elastic torsion,
which concerns the twisting of an elastic bar. This leads to a class of exact
solutions of the Navier equation in terms of solutions of Laplaces equation
in two dimensions. However, as distinct from the use of Laplaces equation
in, say, hydrodynamics or electromagnetism, the dependent variable is the
displacement, which has a direct physical interpretation, rather than a po-
tential, which does not. This means we have to be especially careful to ensurethat the solution is single-valued in situations involving multiply-connected
bars.
These remarks remain important when we move on to another class of
two-dimensional problems called plane strain problems. These have even
more general practical relevance but involve the biharmonic equation. This
equation, which will be seen to be ubiquitous in linear elastostatics, poses
significant extra difficulties as compared to Laplaces equation. In particular,
we will find that it is much more difficult to construct explicit solutions using,for example, the method of separation of variables.
An interesting technique to emerge from both these classes of problems is
the use ofstress potentials, which are the elastic analogues of electrostatic or
gravitational potentials, say, or the stream function in hydrodynamics. We
28
8/23/2019 Applied Solid Mechanics
46/468
2.2 Linear displacements 29
will find that a large class of elastostatic problems with some symmetry, for
example two-dimensional or axisymmetric, can be described using a single
scalar potential that satisfies the biharmonic equation.
Fully three-dimensional problems are mostly too difficult to be suitable for
this chapter. Nonetheless, we will be able to provide a conceptual frameworkwithin which to represent the solution by generalising the idea of Greens
functions for scalar ordinary and partial differential equations. The necessary
Greens tensor describes the response of an elastic body to a localised point
force applied at some arbitrary position in the body. This idea opens up one
of the most distinctive and fascinating aspects of linear elasticity: because
of the intricacy of (1.7.8), many different kinds of singular solutions can
be constructed using stress functions and Greens tensors, each being the
response to a different kind of localised forcing, and the catalogue of thesedifferent responses is a very helpful toolkit for thinking about solid mechanics
more generally.
2.2 Linear displacements
We will begin by neglecting the body force g so the steady Navier equation
reduces to
= ( + )graddivu + 2u = 0. (2.2.1)This vector partial differential equation for u is the starting point for all
we will say in this chapter. As discussed in Section 1.10, it needs to be
supplemented with suitable boundary conditions, which will vary depending
on the situation being modelled.
If the displacement u is a linear function of position x, then the strain
tensor Eis spatially uniform. It follows that the stress tensor is also uniformand, therefore, trivially satisfies (2.2.1). Such solutions provide considerableinsight into the predictions of (2.2.1) and also give a feel for the significance
of the parameters and . To avoid suffices, we will write u = (u, v, w)T
and x = (x,y,z)T.
2.2.1 Isotropic expansion
As a first example, suppose
u =
3x, (2.2.2)
where is a constant scalar, which must be small for linear elasticity to be
valid. When > 0, this corresponds to a uniform isotropic expansion of the
8/23/2019 Applied Solid Mechanics
47/468
30 Linear elastostatics
(a)(c)(b)
1 + /3
1 + /3
1 + /3
1
11y
x
z
1
1
1 +
Fig. 2.1 A unit cube undergoing (a) uniform expansion (2.2.2), (b) one-dimensionalshear (2.2.6), (c) uniaxial stretching (2.2.8).
medium so that, as illustrated in Figure 2.1(a), a unit cube is transformedto a cube with sides of length 1 + /3. (Of course, if is negative, the
displacement is an isotropic contraction.) Since is small, the relative change
in volume is thus 1 +
3
3 1 . (2.2.3)The strain and stress tensors corresponding to this displacement field are
given by
eij =
3ij and ij =
+
2
3
ij . (2.2.4)
This is a so-called hydrostatic situation, in which the stress is characterised
by a scalar isotropic pressure p, and ij = pij . The pressure is related tothe relative volume change by p = K, where
K = +2
3 (2.2.5)
measures the resistance to expansion or compression and is called the bulk
modulus or modulus of compression; from Exercise 1.7, we know that K is
positive.
8/23/2019 Applied Solid Mechanics
48/468
2.2 Linear displacements 31
2.2.2 Simple shear
As our next example, suppose
u = u
vw = y
00 , (2.2.6)where is again a constant scalar. This corresponds to a simple shear of
the solid in the x-direction, as illustrated in Figure 2.1(b). The strain and
stress tensors are now given by
E= 2
0 1 0
1 0 0
0 0 0
, =
0 1 0
1 0 0
0 0 0
. (2.2.7)
Note that does not affect the stress, so the response to shear is accounted
for entirely by , which is therefore called the shear modulus.
2.2.3 Uniaxial stretching
Our next example is uniaxial stretching in which, as shown in Figure 2.1(c),
the solid is stretched by a factor in (say) the x-direction. We suppose,for reasons that will emerge shortly, that the solid simultaneously shrinks
by a factor in the other two directions. The corresponding displacement,
strain and stress are
u =
xyz
, E= 1 0 00 0
0 0
, (2.2.8)
=
(1 2) + 2 0 00 (1 2) 2 00 0 (1 2) 2
. (2.2.9)This simple solution may be used to describe a uniform elastic bar that is
stretched in the x-direction under a tensile force T, as shown in Figure 2.2.
Notice that, since the bar is assumed not to vary in the x-direction, the
outward normal n to the lateral boundary always lies in the (y, z)-plane.
If the curved surface of the bar is stress-free, then the resulting boundarycondition n = 0 may be satisfied identically by ensuring that yy = zz = 0,
which occurs if
=
2 ( + ). (2.2.10)
8/23/2019 Applied Solid Mechanics
49/468
32 Linear elastostatics
n = 0
z
y
x AT T
Fig. 2.2 A uniform bar being stretched under a tensile force T.
Hence the bar, while stretching by a factor in the x-direction, must shrink
by a factor in the two transverse directions; if happened to be neg-ative, this would correspond to an expansion. The ratio between lateral
contraction and longitudinal extension is called Poissons ratio.
With given by (2.2.10), the stress tensor has just one non-zero element,
namely
xx = E, (2.2.11)
where
E = (3 + 2) +
(2.2.12)
is called Youngs modulus. If the cross-section of the bar has area A, then
the tensile force T applied to the bar is related to the stress by
T = Axx = AE; (2.2.13)
thus AE is the elastic modulus k referred to in (1.2.2). By measuring T, the
corresponding extensional strain and transverse contraction , one may
thus infer the values of E and for a particular solid from a bar-stretching
experiment like that illustrated in Figure 2.2. The Lame constants may then
be evaluated using
=E
(1 + )(1 2) , =E
2(1 + ). (2.2.14)
We note that the constitutive relation (1.7.7) can be written in terms of E
and as
Eeij = (1 + )ij kk ij . (2.2.15)While E is a positive constant with the dimensions of pressure, is dimen-
sionless and constrained on physical grounds to lie in the range
1 < < 1/2. The lower bound for comes from the condition +2/3 > 0
8/23/2019 Applied Solid Mechanics
50/468
2.2 Linear displacements 33
(a)
(b)
Fig. 2.3 A paper model with negative Poissons ratio. Each line segment is a stripof paper viewed end-on.
required for convexity of the strain energy, as shown in Exercise 1.7. The up-
per bound 1/2 is approached as , in other words as the materialbecomes incompressible, as discussed in Section 1.8.
For most solids, is positive, but it is possible (see Figure 2.3) to construct
simple hinged paper models with negative Poissons ratio; try extending acrumpled piece of paper! So-called auxetic materials, in which < 0, have
been developed whose microscopic structure mimics such paper models so
they also display negative values of and they expand in all directions when
pulled in only one (see Lakes, 1987).
2.2.4 Biaxial strain
As a final illustrative example, we generalise Section 2.2.3 and consider an
elastic plate strained in the (x, y)-plane as illustrated in Figure 2.4. Suppose
the plate experiences a linear in-plane distortion while shrinking by a factor
in the z-direction, so the displacement is given by
u = uvw = ax + bycx + dyz , (2.2.16)
and, as in Section 2.2.3, the stress and strain tensors are both constant. Here
we choose to satisfy the condition zz = 0 on the traction-free upper and
8/23/2019 Applied Solid Mechanics
51/468
34 Linear elastostatics
y
x
z
h
Tyx
Txx
Txx
Tyy
TxyTyx
Tyy
Txy
Fig. 2.4 A plate being strained under tensions Txx , Ty y and shear forces Txy , Ty x .
lower surfaces of the plate, so that
=
+ 2
(a + d) =
1
(a + d), (2.2.17)
where again denotes Poissons ratio. With this choice, and with E again
denoting the Youngs modulus, the only non-zero stress components are
xx =
E(a + d)
1 2 , xy =E(b + c)
2(1 + ) , yy =
E(a + d)
1 2 . (2.2.18)
We denote the net in-plane tensions and shear stresses applied to the plate
by Tij = hij , as illustrated in Figure 2.4. We can use (2.2.18) to relate these
to the in-plane strain components by
Txx =Eh
1 2 (exx + eyy ), (2.2.19a)
Txy = Tyx = Eh1 +
exy , (2.2.19b)
Tyy =Eh
1 2 (exx + eyy ). (2.2.19c)
These will provide useful evidence when constructing more general models
for the deformation of plates in Chapter 4.
If no force is applied in the y-direction, that is Txy = Tyy = 0, then (2.2.19)
reproduces the results of uniaxial stretching, with d =
a and Txx = Eha.
On the other hand, it is possible for the displacement to be purely in the
(x, z)-plane, with
b = c = d = 0, yy =Ea
1 2 , xx =Ea
1 2 . (2.2.20)
8/23/2019 Applied Solid Mechanics
52/468
2.2 Linear displacements 35
Thus a transverse stress yy must be applied to prevent the plate from
contracting in the y-direction when we stretch it in the x-direction. Notice
also that the effective elastic modulus E/(1 2) is larger than E whenever is non-zero, which shows that purely two-dimensional stretching is always
more strenuous than uniaxial stretching.
2.2.5 General linear displacements
The simple linear examples considered above shed useful light on more gen-
eral solutions of the Navier equation. Indeed, any displacement field, when
expanded in a Taylor series about some point x0, is linear to leading order:
u(x) = u(x0) + u (x0)T(x x0) + . (2.2.21)Here u is the displacement gradient matrix, whose entries are conven-
tionally defined to be u = (uj /xi), and may be written as the sum of
symmetric and skew-symmetric parts:
ujxi
=1
2
uixj
+ujxi
+
1
2
ujxi
uixj
. (2.2.22)
We recognise the symmetric part ofu as the linear strain tensorE
, while
the skew-symmetric part may be written in the form
1
2
uuT =
0 3 23 0 12 1 0
, (2.2.23)so that, with = (1, 2, 3)
T, (2.2.21) becomes
u(x) = u0
+
(xx
0) +
E(x
x
0) +
. (2.2.24)
The terms on the right-hand side represent a small rigid-body translation
and rotation, followed by a linear deformation characterised by the matrix E.Since Eis real and symmetric, it has real eigenvalues, say {1, 2, 3}, and
orthogonal eigenvectors. These eigenvalues are referred to as the principal
strains, and the directions defined by the eigenvectors as the principal di-
rections. If we use an ortho