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

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

R

A

F

T

ME185

Introduction to Continuum Mechanics

Panayiotis Papadopoulos

Department of Mechanical Engineering, University of California, Berkeley

Copyright c ( 2008 by Panayiotis Papadopoulos

i

D

R

A

F

T

Introduction

This set of notes has been written as part of teaching ME185, an elective senior-year

undergraduate course on continuum mechanics in the Department of Mechanical Engineering

at the University of California, Berkeley.

Berkeley, California

P. P.

August 2008

ii

D

R

A

F

T

Contents

1 Introduction 1

1.1 Solids and ﬂuids as continuous media . . . . . . . . . . . . . . . . . . . . . . 1

1.2 History of continuum mechanics . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Mathematical preliminaries 3

2.1 Elements of set theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2 Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.3 Points, vectors and tensors in the Euclidean 3-space . . . . . . . . . . . . . . 8

2.4 Vector and tensor calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Kinematics of deformation 19

3.1 Bodies, conﬁgurations and motions . . . . . . . . . . . . . . . . . . . . . . . 19

3.2 The deformation gradient and other measures of deformation . . . . . . . . . 28

3.3 Velocity gradient and other measures of deformation rate . . . . . . . . . . . 48

3.4 Superposed rigid-body motions . . . . . . . . . . . . . . . . . . . . . . . . . 53

4 Basic physical principles 61

4.1 The divergence and Stokes’ theorems . . . . . . . . . . . . . . . . . . . . . . 61

4.2 The Reynolds’ transport theorem . . . . . . . . . . . . . . . . . . . . . . . . 63

4.3 The localization theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.4 Mass and mass density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

4.5 The principle of mass conservation . . . . . . . . . . . . . . . . . . . . . . . 69

4.6 The principles of linear and angular momentum balance . . . . . . . . . . . . 70

4.7 Stress vector and stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 73

iii

D

R

A

F

T

4.8 The transformation of mechanical ﬁelds under superposed rigid-body motions 85

4.9 The Theorem of Mechanical Energy Balance . . . . . . . . . . . . . . . . . . 88

4.10 The principle of energy balance . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.11 The Green-Naghdi-Rivlin theorem . . . . . . . . . . . . . . . . . . . . . . . . 95

5 Inﬁnitesimal deformations 99

5.1 The Gˆateaux diﬀerential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5.2 Consistent linearization of kinematic and kinetic variables . . . . . . . . . . 101

6 Mechanical constitutive theories 109

6.1 General requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

6.2 Inviscid ﬂuids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

6.3 Viscous ﬂuids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.4 Non-linearly elastic solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

6.5 Linearly elastic solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6.6 Viscoelastic solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

7 Boundary- and Initial/boundary-value Problems 135

7.1 Incompressible Newtonian viscous ﬂuid . . . . . . . . . . . . . . . . . . . . . 135

7.1.1 Gravity-driven ﬂow down an inclined plane . . . . . . . . . . . . . . . 135

7.1.2 Couette ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

7.1.3 Poiseuille ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

7.2 Compressible Newtonian viscous ﬂuids . . . . . . . . . . . . . . . . . . . . . 140

7.2.1 Stokes’ First Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

7.2.2 Stokes’ Second Problem . . . . . . . . . . . . . . . . . . . . . . . . . 141

7.3 Linear elastic solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

7.3.1 Simple tension and simple shear . . . . . . . . . . . . . . . . . . . . . 143

7.3.2 Uniform hydrostatic pressure . . . . . . . . . . . . . . . . . . . . . . 144

7.3.3 Saint-Venant torsion of a circular cylinder . . . . . . . . . . . . . . . 144

7.4 Non-linearly elastic solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

7.4.1 Rivlin’s cube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

7.5 Multiscale problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

iv

D

R

A

F

T

7.5.1 The virial theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

7.6 Expansion of the universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

Appendix A A.1

A.1 Cylindrical polar coordinate system . . . . . . . . . . . . . . . . . . . . . . . A.1

v

D

R

A

F

T

vi

D

R

A

F

T

List of Figures

2.1 Schematic depiction of a set . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Example of a set that does not form a linear space . . . . . . . . . . . . . . . 5

2.3 Mapping between two sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.1 A body B and its subset o. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.2 Mapping of a body B to its conﬁguration at time t. . . . . . . . . . . . . . . . 20

3.3 Mapping of a body B to its reference conﬁguration at time t

0

and its current

conﬁguration at time t. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.4 Schematic depiction of referential and spatial mappings for the velocity v. . . 22

3.5 Particle path of a particle which occupies X in the reference conﬁguration. . 25

3.6 Stream line through point x at time t. . . . . . . . . . . . . . . . . . . . . . . 26

3.7 Mapping of an inﬁnitesimal material line elements dX from the reference to

the current conﬁguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.8 Application of the inverse function theorem to the motion χ at a ﬁxed time t. 30

3.9 Interpretation of the right polar decomposition. . . . . . . . . . . . . . . . . . 35

3.10 Interpretation of the left polar decomposition. . . . . . . . . . . . . . . . . . . 36

3.11 Interpretation of the right polar decomposition relative to the principal direc-

tions M

A

and associated principal stretches λ

A

. . . . . . . . . . . . . . . . . 38

3.12 Interpretation of the left polar decomposition relative to the principal directions

RM

i

and associated principal stretches λ

i

. . . . . . . . . . . . . . . . . . . . 39

3.13 Geometric interpretation of the rotation tensor R by its action on a vector x. 42

3.14 Spatially homogeneous deformation of a sphere. . . . . . . . . . . . . . . . . 44

3.15 Image of a sphere under homogeneous deformation. . . . . . . . . . . . . . . 46

vii

D

R

A

F

T

3.16 Mapping of an inﬁnitesimal material volume element dV to its image dv in

the current conﬁguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.17 Mapping of an inﬁnitesimal material surface element dA to its image da in

the current conﬁguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.18 Conﬁgurations associated with motions χ and χ

+

diﬀering by a superposed

rigid motion ¯ χ

+

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.1 A surface / bounded by the curve (. . . . . . . . . . . . . . . . . . . . . . . 63

4.2 A region { with boundary ∂{ and its image {

0

with boundary ∂{

0

in the

reference conﬁguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.3 A limiting process used to deﬁne the mass density ρ at a point x in the current

conﬁguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.4 Setting for a derivation of Cauchy’s lemma. . . . . . . . . . . . . . . . . . . 73

4.5 The Cauchy tetrahedron. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4.6 Interpretation of the Cauchy stress components on an orthogonal parallelepiped

aligned with the ¦e

i

¦-axes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

4.7 Projection of the traction to its normal and tangential components. . . . . . 80

6.1 Traction acting on a surface of an inviscid ﬂuid. . . . . . . . . . . . . . . . . 111

6.2 A ball of ideal ﬂuid in equilibrium under uniform pressure. . . . . . . . . . . 115

6.3 Orthogonal transformation of the reference conﬁguration. . . . . . . . . . . . 122

6.4 Static continuation E

δ

t

(s) of E

t

(s) by δ. . . . . . . . . . . . . . . . . . . . . . 131

6.5 An interpretation of relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . 132

7.1 Flow down an inclined plane . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

7.2 Couette ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

7.3 Semi-inﬁnite domain for Stokes’ Second Problem . . . . . . . . . . . . . . . . 141

7.4 Function f(λ) in Rivlin’s cube . . . . . . . . . . . . . . . . . . . . . . . . . . 148

viii

D

R

A

F

T

List of Tables

ix

D

R

A

F

T

Chapter 1

Introduction

1.1 Solids and ﬂuids as continuous media

All matter is inherently discontinuous, as it is comprised of distinct building blocks, the

molecules. Each molecule consists of a ﬁnite number of atoms, which in turn consist of ﬁnite

numbers of nuclei and electrons.

Many important physical phenomena involve matter in large length and time scales.

This is generally the case when matter is considered at length scales much larger than

the characteristic length of the atomic spacings and at time scales much larger than the

characteristic times of atomic bond vibrations. The preceding characteristic lengths and

times can vary considerably depending on the state of the matter (e.g., temperature, precise

composition, deformation). However, one may broadly estimate such characteristic lengths

and times to be of the order of up to a few Angstroms (1

˚

A= 10

−10

m) and a few femtoseconds

(1 femtosecond = 10

−15

sec), respectively. As long as the physical problems of interest occur

at length and time scales of several orders of magnitude higher that those noted previously,

it is possible to consider matter as a continuous medium, namely to eﬀectively ignore its

discrete nature without introducing any even remotely signiﬁcant errors.

A continuous medium may be conceptually deﬁned as a ﬁnite amount of matter whose

physical properties are independent of its actual size or the time over which they are mea-

sured. As a thought experiment, one may choose to perpetually dissect a continuous medium

into smaller pieces. No matter how small it gets, its physical properties remain unaltered.

1

D

R

A

F

T

2 History of continuum mechanics

Mathematical theories developed for continuous media (or “continua”) are frequently referred

to as “phenomenological”, in the sense that they capture the observed physical response

without directly accounting for the discrete structure of matter.

Solids and ﬂuids (including both liquids and gases) can be accurately viewed as continuous

media in many occasions. Continuum mechanics is concerned with the response of solids

and ﬂuids under external loading precisely when they can be viewed as continuous media.

1.2 History of continuum mechanics

Continuum mechanics is a modern discipline that uniﬁes solid and ﬂuid mechanics, two of

the oldest and most widely examined disciplines in applied science. It draws on classical

scientiﬁc developments that go as far back as the Hellenistic-era work of Archimedes (287-

212 A.C.) on the law of the lever and on hydrostatics. It is motivated by the imagination

and creativity of da Vinci (1452-1519) and the rigid-body experiments of Galileo (1564-

1642). It is founded on the laws of motion put forward by Newton (1643-1727), later set on

ﬁrm theoretical ground by Euler (1707-1783) and further developed and reﬁned by Cauchy

(1789-1857).

Continuum mechanics as taught and practiced today emerged in the latter half of the

20th century. This “renaissance” period can be attributed to several factors, such as the

ﬂourishing of relevant mathematics disciplines (particularly linear algebra, partial diﬀerential

equations and diﬀerential geometry), the advances in materials and mechanical systems tech-

nologies, and the increasing availability (especially since the late 1960s) of high-performance

computers. A wave of gifted modern-day mechanicians, such as Rivlin, Truesdell, Ericksen,

Naghdi and many others, contributed to the rebirth and consolidation of classical mechan-

ics into this new discipline of continuum mechanics, which emphasizes generality, rigor and

abstraction, yet derives its essential features from the physics of material behavior.

ME185

D

R

A

F

T

Chapter 2

Mathematical preliminaries

2.1 Elements of set theory

This section summarizes a few elementary notions and deﬁnitions from set theory. A set X

is a collection of objects referred to as elements. A set can be deﬁned either by the properties

of its elements or by merely identifying all elements. For example,

X = ¦1, 2, 3, 4, 5¦ (2.1)

or

X = ¦all integers greater than 0 and less than 6¦ . (2.2)

Two sets of particular interest in the remainder of the course are:

N = ¦all positive integers¦ (2.3)

and

R = ¦all real numbers¦ (2.4)

If x is an element of the set X, one writes x ∈ X. If not, one writes x / ∈ X.

Let X, Y be two sets. The set X is a subset of the set Y (denoted as X ⊆ Y or Y ⊇ X) if

every element of X is also an element of Y . The set X is a proper subset of the set (denoted

as X ⊂ Y or Y ⊃ X) if every element of X is also an element of Y , but there exists at least

one element of Y that does not belong to X.

3

D

R

A

F

T

4 Vector spaces

The union of sets X and Y (denoted by X ∪ Y ) is the set which is comprised of all

elements of both sets. The intersection of sets X and Y (denoted by X ∩ Y ) is a set which

includes only the elements common to the two sets. The empty set (denoted by ∅) is a set

that contains no elements and is contained in every set, therefore, X ∪ ∅ = X.

The Cartesian product X Y of sets X and Y is a set deﬁned as

X Y = ¦(x, y) such that x ∈ X, y ∈ Y ¦ . (2.5)

Note that the pair (x, y) in the preceding equation is ordered, i.e., the element (x, y) is, in

general, not the same as the element (y, x). The notation X

2

, X

3

, . . ., is used to respectively

denote the Cartesian products X X, X X X, . . ..

2.2 Vector spaces

Consider a set 1 whose members (typically called “points”) can be scalars, vectors or func-

tions, visualized in Figure 2.1. Assume that 1 is endowed with an addition operation (+)

and a scalar multiplication operation (), which do not necessarily coincide with the classical

addition and multiplication for real numbers.

A "point" that

belongs to

V

V

Figure 2.1: Schematic depiction of a set

A linear (or vector) space ¦1, +; R, ¦ is deﬁned by the following properties for any

u, v, w ∈ 1 and α, β ∈ R:

(i) α u + β v ∈ 1 (closure),

(ii) (u +v) +w = u + (v +w) (associativity with respect to + ),

ME185

D

R

A

F

T

Mathematical preliminaries 5

(iii) ∃ 0 ∈ 1 [ u +0 = u (existence of null element),

(iv) ∃ −u ∈ 1 [ u + (−u) = 0 (existence of negative element),

(v) u +v = v +u (commutativity),

(vi) (αβ) u = α (β u) (associativity with respect to ),

(vii) (α + β) u = α u + β u (distributivity with respect to R),

(viii) α (u +v) = α u + α v (distributivity with respect to 1),

(ix) 1 u = u (existence of identity).

Examples:

(1) 1 = P

2

:= ¦all second degree polynomials ax

2

+ bx + c¦ with the standard polynomial

addition and scalar multiplication.

It can be trivially veriﬁed that ¦P

2

, +; R, ¦ is a linear function space. P

2

is also

“equivalent” to an ordered triad (a, b, c) ∈ R

3

.

(2) Deﬁne 1 = ¦(x, y) ∈ R

2

[ x

2

+ y

2

= 1¦ with the standard addition and scalar multi-

plication for vectors. Notice that given u = (x

1

, y

1

) and v = (x

2

, y

2

) as in Figure 2.2,

x

y

1 u

v

u+v

Figure 2.2: Example of a set that does not form a linear space

property (i) is violated, i.e., since, in general, for α = β = 1,

u +v = (x

1

+ x

2

, y

1

+ y

2

) ,

ME185

D

R

A

F

T

6 Vector spaces

and (x

1

+ x

2

)

2

+ (y

1

+ y

2

)

2

= 1. Thus, ¦1, +; R, ¦ is not a linear space.

Consider a linear space ¦1, +; R, ¦ and a subset | of 1. Then | forms a linear sub-space

of 1 with respect to the same operations (+) and (), if, for any u, v ∈ | and α, β, ∈ R

α u + β v ∈ | ,

i.e., closure is maintained within |.

Example:

(a) Deﬁne the set P

n

of all algebraic polynomials of degree smaller or equal to n > 2 and

consider the linear space ¦P

n

, +; R, ¦ with the usual polynomial addition and scalar

multiplication. Then, P

2

is a linear subspace of ¦P

n

, +; R, ¦.

In order to simplify the notation, in the remainder of these notes the symbol used in

scalar multiplication will be omitted.

Let v

1

v

2

, ..., v

p

be elements of the vector space ¦1, +; R, ¦ and assume that

α

1

v

1

+ α

2

v

2

+ . . . + α

p

v

p

= 0 ⇔ α

1

= α

2

= ... = α

p

= 0 . (2.6)

Then, ¦v

1

, v

2

, . . . , v

p

¦ is a linearly independent set in 1. The vector space ¦1, +; R, ¦ is

inﬁnite-dimensional if, given any n ∈ N, it contains at least one linearly independent set

with n + 1 elements. If the above statement is not true, then there is an n ∈ N, such that

all linearly independent sets contain at most n elements. In this case, ¦1, +; R, ¦ is a ﬁnite

dimensional vector space (speciﬁcally, n-dimensional).

A basis of an n-dimensional vector space ¦1, +; R, ¦ is deﬁned as any set of n linearly

independent vectors. If ¦g

1

, g

2

, ..., g

n

¦ form a basis in ¦1, +; R, ¦, then given any non-zero

v ∈ 1,

α

1

g

1

+ α

2

g

2

+ . . . + α

n

g

n

+ βv = 0 ⇔ not all α

1

, . . . , α

n

, β equal zero . (2.7)

More speciﬁcally, β = 0 because otherwise there would be at least one non-zero α

i

, i = 1, . . . , n,

which would have implied that ¦g

1

, g

2

, ..., g

n

¦ are not linearly independent.

Thus, the non-zero vector v can be expressed as

v = −

α

1

β

g

1

−

α

2

β

g

2

−. . . −

α

n

β

g

n

. (2.8)

ME185

D

R

A

F

T

Mathematical preliminaries 7

The above representation of v in terms of the basis ¦g

1

, g

2

, ..., g

n

¦ is unique. Indeed, if,

alternatively,

v = γ

1

g

1

+ γ

2

g

2

+ . . . + γ

n

g

n

, (2.9)

then, upon subtracting the preceding two equations from one another, it follows that

0 =

γ

1

+

α

1

β

g

1

+

γ

2

+

α

2

β

g

2

+ . . . +

γ

n

+

α

n

β

g

n

, (2.10)

which implies that γ

i

= −

α

i

β

, i = 1, 2, . . . , n, since ¦g

1

, g

2

, ..., g

n

¦ are assumed to be a linearly

independent.

Of all the vector spaces, attention will be focused here on the particular class of Euclidean

vector spaces in which a vector multiplication operation () is deﬁned, such that for any

u, v, w ∈ 1 and α ∈ R,

(x) u v = v u (commutativity with respect to ),

(xi) u (v +w) = u v +u w (distributivity),

(xii) (αu) v = u (αv) = α(u v) (associativity with respect to )

(xiii) u u ≥ 0 and u u = 0 ⇔ u = 0.

This vector operation is referred to as the dot-product. An n-dimensional vector space

obeying the above additional rules is referred to as a Euclidean vector space and is denoted

by E

n

.

Example:

The standard dot-product between vectors in R

n

satisﬁes the above properties.

The dot-product provide a natural means for deﬁning the magnitude of a vector as

|u| = (u u)

1/2

. (2.11)

Two vectors u, v ∈ E

n

are orthogonal, if u v = 0. A set of vectors ¦u

1

, u

2

, ...¦ is called

orthonormal, if

u

i

u

j

=

0 if i = j

1 if i = j

= δ

ij

, (2.12)

where δ

ij

is called the Kronecker delta symbol.

ME185

D

R

A

F

T

8 Points, vectors and tensors in the Euclidean 3-space

Every orthonormal set ¦e

1

, e

2

...e

k

¦, k ≤ n in E

n

is linearly independent. This is because,

if

α

1

e

1

+ α

2

e

2

+ . . . + α

k

e

k

= 0 , (2.13)

then, upon taking the dot-product of the above equation with e

i

, i = 1, 2, . . . , k, and invoking

the orthonormality of ¦e

1

, e

2

...e

k

¦,

α

1

(e

1

e

i

) + α

2

(e

2

e

i

) + . . . + α

k

(e

k

e

i

) = α

i

= 0 . (2.14)

Of particular importance to the forthcoming developments is the observation that any

vector v ∈ E

n

can be resolved to an orthonormal basis ¦e

1

, e

2

, ..., e

n

¦ as

v = v

1

e

1

+ v

2

e

2

+ . . . + v

n

e

n

=

n

¸

i=1

v

i

e

i

, (2.15)

where v

i

= v e

i

. In this case, v

i

denotes the i-th component of v relative to the orthonormal

basis ¦e

1

, e

2

, ..., e

n

¦.

2.3 Points, vectors and tensors in the Euclidean 3-

space

Consider the Cartesian space E

3

with an orthonormal basis ¦e

1

, e

2

, e

3

¦. As argued in the

previous section, a typical vector vǫE

3

can be written as

v =

3

¸

i=1

v

i

e

i

; v

i

= v e

i

. (2.16)

Next, consider points x, y in the Euclidean point space c

3

, which is the set of all points in the

ambient three-dimensional space, when taken to be devoid of the mathematical structure of

vector spaces. Also, consider an arbitrary origin (or reference point) O in the same space.

It is now possible to deﬁne vectors x, y ∈ E

3

, which originate at O and end at points x and

y, respectively. In this way, one makes a unique association (to within the speciﬁcation of

O) between points in c

3

and vectors in E

3

. Further, it is possible to deﬁne a measure of

distance between x and y, by way of the magnitude of the vector v = y −x, namely

d(x, y) = [x −y[ = [(x −y) (x −y)]

1/2

. (2.17)

ME185

D

R

A

F

T

Mathematical preliminaries 9

In E

3

, one may deﬁne the vector product of two vectors as an operation with the following

properties: for any vectors u, v and w, and any scalar α,

(a) u v = −v u,

(b) (u v) w = (v w) u = (w u) v, or, equivalently [uvw] = [vwu] = [wuv],

where [uvw] = (u v) w is the triple product of vectors u, v, and w,

(c) [u v[ = [u[[v[ sinθ , cos θ =

u v

(u u)

1/2

(v v)

1/2

, 0 ≤ θ ≤ π.

Appealing to property (a), it is readily concluded that u u = 0. Likewise, properties

(a) and (b) can be used to deduce that (uv) u = (u v) v = 0, namely that the vector

u v is orthogonal to both u and v.

By deﬁnition, for a right-hand orthonormal coordinate basis ¦e

1

, e

2

, e

3

¦, the following

relations hold true:

e

1

e

2

= e

3

, e

2

e

3

= e

1

, e

3

e

1

= e

2

. (2.18)

These relations, together with the implications of property (a)

e

1

e

1

= e

2

e

2

= e

3

e

3

= 0 (2.19)

and

e

2

e

1

= −e

3

, e

3

e

2

= −e

1

, e

1

e

3

= −e

2

(2.20)

can be expressed compactly as

e

i

e

j

=

3

¸

k=1

ǫ

ijk

e

k

, (2.21)

where ǫ

ijk

is the permutation symbol deﬁned as

ǫ

ijk

=

1 if (i, j, k) = (1,2,3), (2,3,1), or (3,1,2)

−1 if (i, j, k) = (2,1,3), (3,2,1), or (1,3,2)

0 otherwise

. (2.22)

It follows that

u v = (

3

¸

i=1

u

i

e

i

) (

3

¸

j=1

v

j

e

j

) =

3

¸

i=1

3

¸

j=1

u

i

v

j

e

i

e

j

=

3

¸

i=1

3

¸

j=1

3

¸

k=1

u

i

v

j

e

ijk

e

k

. (2.23)

ME185

D

R

A

F

T

10 Points, vectors and tensors in the Euclidean 3-space

Let |, 1 be two sets and deﬁne a mapping f from | to 1 as a rule that assigns to each

point u ∈ | a unique point v = f(u) ∈ 1, see Figure 2.3. The usual notation for a mapping

is: f : | → 1 , u → v = f(u) ∈ 1. With reference to the above setting, | is called the

domain of f, whereas 1 is termed the range of f.

u

v

f

U V

Figure 2.3: Mapping between two sets

A mapping T : E

3

→ E

3

is called linear if it satisﬁes the property

T(αu + βv) = αT(u) + βT(v) , (2.24)

for all u, v ∈ E

3

and α, β ∈ N. A linear mapping is also referred to as a tensor.

Examples:

(1) T : E

3

→ E

3

, T(v) = v for all v ∈ E

3

. This is called the identity tensor, and is

typically denoted by T = I.

(2) T : E

3

→ E

3

, T(v) = 0 for all v ∈ E

3

. This is called the zero tensor, and is typically

denoted by T = 0.

The tensor product between two vectors v and w in E

3

is denoted by v ⊗w and deﬁned

according to the relation

(v ⊗w)u = (w u)v , (2.25)

for any vector u ∈ E

3

. This implies that, under the action of the tensor product v ⊗w, the

vector u is mapped to the vector (w u)v. It can be easily veriﬁed that v ⊗ w is a tensor

according to the previously furnished deﬁnition. Using the Cartesian components of vectors,

ME185

D

R

A

F

T

Mathematical preliminaries 11

one may express the tensor product of v and w as

v ⊗w = (

3

¸

i=1

v

i

e

i

) ⊗(

3

¸

j=1

w

j

e

j

) =

3

¸

i=1

3

¸

j=1

v

i

w

j

e

i

⊗e

j

. (2.26)

It will be shown shortly that the set of nine tensor products ¦e

i

⊗e

j

¦, i, j = 1, 2, 3, form a

basis for the space L(E

3

, E

3

) of all tensors on E

3

.

Before proceeding further with the discussion of tensors, it is expedient to introduce a

summation convention, which will greatly simplify the component representation of both

vectorial and tensorial quantities and their algebra. This originates with A. Einstein, who

employed it ﬁrst in his relativity work. The summation convention has three rules, which,

when adapted to the special case of E

3

, are as follows:

Rule 1. If an index appears twice in a single component term or in a product term, the

summation sign is omitted and summation is automatically assumed from value 1 to

3. Such an index is referred to as dummy.

Rule 2. An index which appears once in a single component or in a product expression is

not summed and is assumed to attain a single value (1, 2, or 3). Such an index is

referred to free.

Rule 3. No index can appear more than twice in a single component or in a product term.

Examples:

1. The vector representation u =

3

¸

i=1

u

i

e

i

is replaced by u = u

i

e

i

and it involves the

summation of three terms.

2. The tensor product u ⊗ v =

3

¸

i=1

3

¸

j=1

u

i

v

j

e

i

⊗ e

j

is equivalently written as u ⊗ v =

u

i

v

j

e

i

⊗e

j

and it involves the summation of nine terms.

3. The term u

i

v

j

is a single term with two free indices i and j.

4. It is easy to see that δ

ij

u

i

= δ

1j

u

1

+δ

2j

u

2

+δ

3j

u

3

= u

j

. This index substitution property

is frequently used in component manipulations.

ME185

D

R

A

F

T

12 Points, vectors and tensors in the Euclidean 3-space

5. A similar index substitution property applies in the case of a two-index quantity,

namely δ

ij

a

ik

= δ

1j

a

1k

+ δ

2j

a

2k

+ δ

3j

a

3k

= a

jk

.

With the summation convention in place, take a tensor T ∈ L(E

3

, E

3

) and deﬁne its

components T

ij

, such that Te

j

= T

ij

e

i

. It follows that

(T−T

ij

e

i

⊗e

j

)v = (T−T

ij

e

i

⊗e

j

)v

k

e

k

= Te

k

v

k

−T

ij

v

k

(e

i

⊗e

j

)e

k

= T

ik

e

i

v

k

−T

ij

v

k

(e

j

e

k

)e

i

= T

ik

e

i

v

k

−T

ij

v

k

δ

jk

e

i

= T

ik

e

i

v

k

−T

ik

v

k

e

i

= 0 , (2.27)

hence,

T = T

ij

e

i

⊗e

j

. (2.28)

This derivation demonstates that any tensor T can be written as a linear combination of the

nine tensor product terms ¦e

i

⊗e

j

¦. The components of the tensor T relative to ¦e

i

⊗ e

j

¦

can be put in matrix form as

[T

ij

] =

T

11

T

12

T

13

T

21

T

22

T

23

T

31

T

32

T

33

¸

¸

¸

¸

. (2.29)

The transpose T

T

of a tensor T is deﬁned by the property

u Tv = v T

T

u , (2.30)

for any vectors u and v in E

3

. Using components, this implies that

u

i

T

ij

v

j

= v

i

A

ij

u

j

= v

j

A

ji

u

i

, (2.31)

where A

ij

are the components of T

T

. It follows that

u

i

(T

ij

−A

ji

)v

j

= 0 . (2.32)

ME185

D

R

A

F

T

Mathematical preliminaries 13

Since u

i

and v

j

are arbitrary, this implies that A

ij

= T

ji

, hence the transpose of T can be

written as

T

T

= T

ij

e

j

⊗e

i

. (2.33)

A tensor T is symmetric if T

T

= T or, when both T and T

T

are resolved relative to the

same basis, T

ji

= T

ij

. Likewise, a tensor T is skew-symmetric if T

T

= −T or, again, upon

resolving both on the same basis, T

ji

= −T

ij

. Note that, in this case, T

11

= T

22

= T

33

= 0.

Given tensors T, S ∈ L (E

3

, E

3

), the multiplication TS is deﬁned according to

(TS)v = T(Sv) , (2.34)

for any v ∈ E

3

.

In component form, this implies that

(TS)v = T(Sv) = T[(S

ij

e

i

⊗e

j

)(v

k

e

k

)]

= T(S

ij

v

k

δ

jk

e

i

)

= T(S

ij

v

j

e

i

)

= T

ki

S

ij

v

j

e

i

= (T

ki

S

ij

e

k

⊗e

j

)(v

l

e

l

) , (2.35)

which leads to

(TS) = T

ki

S

ij

e

k

⊗e

j

. (2.36)

The trace tr T : L(E

3

, E

3

) →R of the tensor product of two vectors u ⊗v is deﬁned as

tr u ⊗v = u v , (2.37)

hence, the trace of a tensor T is deduced from equation (2.37) as

tr T = tr(T

ij

e

i

⊗e

j

) = T

ij

e

i

e

j

= T

ij

δ

ij

= T

ii

. (2.38)

The contraction (or inner product) T S : L(E

3

, E

3

) L(E

3

, E

3

) →R of two tensors T and

S is deﬁned as

T S = tr(TS

T

) . (2.39)

Using components,

tr(TS

T

) = tr(T

ki

S

ji

e

k

⊗e

j

) = T

ki

S

ji

e

k

e

j

= T

ki

S

ji

δ

kj

= T

ki

S

ki

. (2.40)

ME185

D

R

A

F

T

14 Vector and tensor calculus

A tensor T is invertible if, for any w ∈ E

3

, the equation

Tv = w (2.41)

can be uniquely solved for v. Then, one writes

v = T

−1

w , (2.42)

and T

−1

is the inverse of T. Clearly, if T

−1

exists, then

T

−1

w−v = 0

= T

−1

(Tv) −v

= (T

−1

T)v −v

= (T

−1

T−I)v , (2.43)

hence T

−1

T = I and, similarly, TT

−1

= I.

A tensor T is orthogonal if

T

T

= T

−1

, (2.44)

which implies that

T

T

T = TT

T

= I . (2.45)

It can be shown that for any tensors T, S ∈ L(E

3

, E

3

),

(S +T)

T

= S

T

+T

T

, (ST)

T

= T

T

S

T

. (2.46)

If, further, the tensors T and S are invertible, then

(ST)

−1

= T

−1

S

−1

. (2.47)

2.4 Vector and tensor calculus

Deﬁne scalar, vector and tensor functions of a vector variable x and a real variable t. The

scalar functions are of the form

φ

1

: R →R , t → φ = φ

1

(t)

φ

2

: E

3

→R , x → φ = φ

2

(x) (2.48)

φ

3

: E

3

R →R , (x, t) → φ = φ

3

(x, t) ,

ME185

D

R

A

F

T

Mathematical preliminaries 15

while the vector and tensor functions are of the form

v

1

: R → E

3

, t → v = v

1

(t)

v

2

: E

3

→ E

3

, x → v = v

2

(x) (2.49)

v

3

: E

3

R → E

3

, (x, t) → v = v

3

(x, t)

and

T

1

: R → L(E

3

, E

3

) , t → T = T

1

(t)

T

2

: E

3

→ L(E

3

, E

3

) , x → T = T

2

(x) (2.50)

T

3

: E

3

R → L(E

3

, E

3

) , (x, t) → T = T

3

(x, t) ,

respectively.

The gradient gradφ(x) (otherwise denoted as ∇φ(x) or

∂φ(x)

∂x

) of a scalar function

φ = φ(x) is a vector deﬁned by

(grad φ(x)) v =

¸

d

dw

φ(x + wv)

w=0

, (2.51)

for any v ∈ E

3

. Using the chain rule, the right-hand side of equation (2.51) becomes

¸

d

dw

φ(x + wv)

w=0

=

¸

∂φ(x + wv)

∂(x

i

+ wv

i

)

d(x

i

+ wv

i

)

dw

w=0

=

∂φ(x)

∂x

i

v

i

. (2.52)

Hence, in component form one may write

gradφ(x) =

∂φ(x)

∂x

i

e

i

. (2.53)

In operator form, this leads to the expression

grad = ∇ =

∂

∂x

i

e

i

. (2.54)

Example: Consider the scalar function φ(x) = [x[

2

= x x. Its gradient is

gradφ =

∂

∂x

(x x) =

∂(x

j

x

j

)

∂x

i

e

i

=

∂x

j

∂x

i

x

j

+ x

j

∂x

j

∂x

i

e

i

= (δ

ij

x

j

+ x

j

δ

ij

)e

i

= 2x

i

e

i

= 2x . (2.55)

ME185

D

R

A

F

T

16 Vector and tensor calculus

Alternatively, using directly the deﬁnition

(grad φ) v =

¸

d

dw

¦(x + wv) (x + uv)¦

w=0

=

¸

d

dw

¦x x + 2wx v + w

2

v v¦

w=0

= [2x v + 2wv v]

w=0

= 2x v (2.56)

**The gradient gradv(x) (otherwise denoted as ∇v(x) or
**

∂v(x)

∂x

) of a vector function

v = v(x) is a tensor deﬁned by the relation

(gradv(x))w =

¸

d

dw

v(x + ww)

w=0

, (2.57)

for any w ∈ E

3

. Again, using chain rule, the right-hand side of equation (2.57) becomes

¸

d

dw

v(x + ww)

w=0

=

¸

∂v

i

(x + ww)

∂(x

j

+ ww

j

)

d(x

j

+ ww

j

)

dw

w=0

e

i

=

∂v

i

(x)

∂x

j

w

j

e

i

, (2.58)

hence, using components,

grad(v(x) =

∂v

i

(x)

∂x

j

e

i

⊗e

j

. (2.59)

In operator form, this leads to the expression

grad = ∇ =

∂

∂x

i

⊗e

j

. (2.60)

Example: Consider the function v(x) = αx. Its gradient is

grad v =

∂(αx)

∂x

=

∂(αx

i

)

∂x

j

e

i

⊗e

j

= αδ

ij

e

i

⊗e

j

= αe

i

⊗e

i

= αI , (2.61)

since (e

i

⊗e

i

)v = (v e

i

)e

i

= v

i

e

i

= v. Alternatively, using directly the deﬁnition,

(grad v)w =

¸

d

dw

(v + ww)

w=0

= αw , (2.62)

hence gradv = αI.

The divergence div v(x) (otherwise denoted as ∇ v(x)) of a vector function v = v(x) is

a scalar deﬁned as

div v(x) = tr(gradv(x)) , (2.63)

ME185

D

R

A

F

T

Mathematical preliminaries 17

on, using components,

div v(x) = tr

∂v

i

∂x

j

e

i

⊗e

j

=

∂v

i

∂x

j

e

i

e

j

=

∂v

i

∂x

j

δ

ij

=

∂v

i

∂x

i

= v

i,i

. (2.64)

In operator form, one writes

div = ∇ =

∂

∂x

i

e

i

. (2.65)

Example: Consider again the function v(x) = αx. Its divergence is

div v(x) =

∂(αx

i

)

∂x

i

= α

∂x

i

∂x

i

= αδ

ii

= 3α . (2.66)

**The divergence div T(x) (otherwise denoted as ∇ T(x)) of a tensor function T = T(x)
**

is a vector deﬁned by the property that

(div T(x)) c = div

(T

T

(x))c

, (2.67)

for any constant vector c ∈ E

3

.

Using components,

div T) = div(T

T

c)

= div[(T

ij

e

j

⊗e

i

)(c

k

e

k

)]

= div[T

ij

c

k

δ

ik

e

j

]

= div[T

ij

c

i

e

j

]

= tr

¸

∂T

ij

c

i

∂x

k

e

j

⊗e

k

=

∂(T

ij

c

i

)

∂x

k

δ

jk

=

∂(T

ij

c

i

)

∂x

j

=

∂T

ij

∂x

j

c

i

, (2.68)

hence,

div T =

∂T

ij

∂x

j

e

i

. (2.69)

ME185

D

R

A

F

T

18 Vector and tensor calculus

In the case of the divergence of a tensor, the operator form becomes

div = ∇ =

∂

∂x

i

e

i

. (2.70)

Finally, the curl curl v(x) of a vector function v(x) is a vector deﬁned as

curl v(x) = ∇v(x) , (2.71)

which translates using components to

curl v(x) =

∂

∂x

i

e

i

(v

j

e

j

) =

∂v

j

∂x

i

e

i

e

j

=

∂v

j

∂x

i

e

ijk

e

k

= e

ijk

∂v

k

∂x

j

e

i

. (2.72)

In operator form, the curl is expressed as

curl = ∇ =

∂

∂x

i

e

i

. (2.73)

ME185

D

R

A

F

T

Chapter 3

Kinematics of deformation

3.1 Bodies, conﬁgurations and motions

Let a continuum body B be deﬁned as a collection of material particles, which, when con-

sidered together, endow the body with local (pointwise) physical properties which are inde-

pendent of its actual size or the time over which they are measured. Also, let a typical such

particle be denoted by P, while an arbitrary subset of B be denoted by o, see Figure 3.1.

o

B

P

Figure 3.1: A body B and its subset o.

Let x be the point in c

3

occupied by a particle P of the body B at time t, and let x be

its associated position vector relative to the origin O of an orthonormal basis in the vector

space E

3

. Then, deﬁne by ¯ χ : (P, t) ∈ BR → E

3

the motion of B, which is a diﬀerentiable

mapping, such that

x = ¯ χ(P, t) = ¯ χ

t

(P) . (3.1)

In the above, ¯ χ

t

: B → c

3

is called the conﬁguration mapping of B at time t. Given ¯ χ,

19

D

R

A

F

T

20 Bodies, conﬁgurations and motions

the body B may be mapped to its conﬁguration 1 = ¯ χ

t

(B, t) with boundary ∂1 at time t.

Likewise, any part o ⊂ B can be mapped to its conﬁguration { = ¯ χ

t

(o, t) with boundary

∂{ at time t, see Figure 3.2. Clearly, 1 and { are point sets in c

3

.

o

B

P

{

1

∂{

∂1

x

¯ χ

Figure 3.2: Mapping of a body B to its conﬁguration at time t.

The conﬁguration mapping ¯ χ

t

is assumed to be invertible, which means that, given any

point x ∈ {,

P = ¯ χ

−1

t

(x) . (3.2)

The motion ¯ χ of the body is assumed to be twice-diﬀerentiable in time. Then, one may

deﬁne the velocity and acceleration of any particle P at time t according to

v =

∂ ¯ χ(P, t)

∂t

, a =

∂

2

¯ χ(P, t)

∂t

2

. (3.3)

The mapping ¯ χ represents the material description of the body motion. This is because

the domain of ¯ χ consists of the totality of material particles in the body, as well as time. This

description, although mathematically proper, is of limited practical use, because there is no

direct quantitative way of tracking particles of the body. For this reason, two alternative

descriptions of the body motion are introduced below.

Of all conﬁgurations in time, select one, say 1

0

= ¯ χ(B, t

0

) at a time t = t

0

, and refer

to it as the reference conﬁguration. The choice of reference conﬁguration may be arbitrary,

although in many practical problems it is guided by the need for mathematical simplicity.

Now, denote the point which P occupies at time t

0

as X and let this point be associated

with position vector X, namely

X = ¯ χ(P, t

0

) = ¯ χ

t

0

(P) . (3.4)

ME185

D

R

A

F

T

Kinematics of deformation 21

Thus, one may write

x = ¯ χ(P, t) = ¯ χ(χ

−1

t

0

(X), t) = χ(X, t) . (3.5)

The mapping χ : E

3

R → E

3

, where

x = χ(X, t) = χ

t

(X) (3.6)

represents the referential or Lagrangian description of the body motion. In such a descrip-

tion, it is implicit that a reference conﬁguration is provided. The mapping χ

t

is the placement

of the body relative to its reference conﬁguration, see Figure 3.3.

B

P

1 1

0

x

X

¯ χ

t

0

¯ χ

t

χ

t

Figure 3.3: Mapping of a body B to its reference conﬁguration at time t

0

and its current

conﬁguration at time t.

Assume now that the motion of the body B is described with reference to the conﬁguration

1

0

deﬁned at time t = t

0

and let the conﬁguration of B at time t be termed the current

conﬁguration. Also, let ¦E

1

, E

2

, E

3

¦ and ¦e

1

, e

2

, e

3

¦ be ﬁxed right-hand orthonormal bases

associated with the reference and current conﬁguration, respectively. With reference to the

preceding bases, one may write the position vectors X and x corresponding to the points

occupied by the particle P at times t

0

and t as

X = X

A

E

A

, x = x

i

e

i

, (3.7)

respectively. Hence, resolving all relevant vectors to their respective bases, the motion χ

may be expressed using components as

x

i

e

i

= χ

i

(X

A

E

A

, t)e

i

, (3.8)

ME185

D

R

A

F

T

22 Bodies, conﬁgurations and motions

or, in pure component form,

x

i

= χ

i

(X

A

, t) . (3.9)

The velocity and acceleration vectors, expressed in the referential description, take the

form

v =

∂χ(X, t)

∂t

, a =

∂

2

χ(X, t)

∂t

2

, (3.10)

respectively. Likewise, using the orthonormal bases,

v = v

i

(X

A

, t)e

i

, a = a

i

(X

A

, t)e

i

. (3.11)

Scalar, vector and tensor functions can be alternatively expressed using the spatial or

Eulerian description, where the independent variables are the current position vector x and

time t. Indeed, starting, by way of an example, with a scalar function f =

ˇ

f(P, t), one may

write

f =

ˇ

f(P, t) =

ˇ

f(χ

−1

t

(x), t) =

˜

f(x, t) . (3.12)

In analogous fashion, one may write

f =

˜

f(x, t) =

˜

f(χ

t

(X), t) =

ˆ

f(X, t) . (3.13)

The above two equations can be combined to write

f =

ˇ

f(P, t) =

˜

f(x, t) =

ˆ

f(X, t) . (3.14)

(X, t) (x, t)

ˆ v

˜ v

v

χ

Figure 3.4: Schematic depiction of referential and spatial mappings for the velocity v.

Any function (not necessarily scalar) of space and time can be written equivalently in

material, referential or spatial form. Focusing speciﬁcally on the referential and spatial

ME185

D

R

A

F

T

Kinematics of deformation 23

descriptions, it is easily seen that the velocity and acceleration vectors can be equivalently

expressed as

v = ˆ v(X, t) = ˜ v(x, t) , a = ˆ a(X, t) = ˜ a(x, t) , (3.15)

respectively, see Figure 3.4. In component form, one may write

v = ˆ v

i

(X

A

, t)e

i

= ˜ v

i

(x

j

, t)e

i

, a = ˆ a

i

(X

A

, t)e

i

= ˜ a

i

(x

j

, t)e

i

. (3.16)

Given a function f =

ˆ

f(X, t), deﬁne the material time derivative of f as

˙

f =

∂

ˆ

f(X, t)

∂t

. (3.17)

From the above deﬁnition it is clear that the material time derivative of a function is the

rate of change of the function when keeping the referential position X (therefore also the

particle P associated with this position) ﬁxed.

If, alternatively, f is expressed in spatial form, i.e., f =

˜

f(x, t), then

˙

f =

∂

˜

f(x, t)

∂t

+

∂

˜

f(x, t)

∂x

∂χ(X, t)

∂t

=

∂

˜

f(x, t)

∂t

+

∂

˜

f(x, t)

∂x

v

=

∂

˜

f(x, t)

∂t

+ grad

˜

f v . (3.18)

The ﬁrst term on the right-hand side of (3.18) is the spatial time derivative of f and corre-

sponds to the rate of change of f for a ﬁxed point x in space. The second term is called the

convective rate of change of f and is due to the spatial variation of f and its eﬀect on the

material time derivative as the material particle which occupies the point x at time t tends

to travel away from x with velocity v. A similar expression for the material time derivative

applies for vector functions. Indeed, take, for example, the velocity v = ˜ v(x, t) and write

˙ v =

∂˜ v(x, t)

∂t

+

∂˜ v(x, t)

∂x

∂χ(X, t)

∂t

=

∂˜ v(x, t)

∂t

+

∂˜ v(x, t)

∂x

v

=

∂˜ v(x, t)

∂t

+ (grad ˜ v)v . (3.19)

ME185

D

R

A

F

T

24 Bodies, conﬁgurations and motions

A volume/surface/curve which consists of the same material points in all conﬁgurations

is termed material. By way of example, consider a surface in three dimensions, which can

be expressed in the form F(X) = 0. This is clearly a material surface, because it contains

the same material particles at all times, since its referential representation is independent

of time. On the other hand, a surface described by the equation F(X, t) = 0 is generally

non material, because the locus of its points contains diﬀerent material particles at diﬀerent

times. This distinction becomes less apparent when a surface is deﬁned in spatial form, i.e.,

by an equation f(x, t) = 0. In this case, one may employ Lagrange’s criterion of materiality,

which states the following:

Lagrange’s criterion of materiality (1783)

A surface described by the equation f(x, t) = 0 is material if, and only if,

˙

f = 0.

A sketch of the proof is as follows: if the surface is assumed material, then

f(x, t) = F(X) = 0 , (3.20)

hence

˙

f(χ, t) =

˙

F(X) = 0 . (3.21)

Conversely, if the criterion holds, then

˙

f(χ, t) =

˙

f(χ(X, t), t) =

∂F

∂t

(X, t) = 0 , (3.22)

which implies that F = F(X), hence the surface is material.

A similar analysis applies to curves in Euclidean 3-space. Speciﬁcally, a material curve

can be be viewed as the intersection of two material surfaces, say F(X) = 0 and G(X) = 0.

Switching to the spatial description and expressing these surfaces as

F(X) = F(χ

−1

t

(x)) = f(x, t) = 0 (3.23)

and

G(X) = G(χ

−1

t

(x)) = g(x, t) = 0 , (3.24)

it follows from Lagrange’s criterion that a curve is material if

˙

f = ˙ g = 0. It is easy to show

that this is a suﬃcient, but not a necessary condition for the materiality of a curve.

ME185

D

R

A

F

T

Kinematics of deformation 25

Some important deﬁnitions regarding the nature of the motion χ follow. First, a motion χ

is steady at a point x, if the velocity at that point is independent of time. If this is the case

at all points, then v = v(x) and the motion is called steady. If a motion is not steady, then

it is called unsteady. A point x in space where v(x, t) = 0 at all times is called a stagnation

point.

Let χ be the motion of body B and ﬁx a particle P, which occupies a point X in the

reference conﬁguration. Subsequently, trace its successive placements as a function of time,

i.e., ﬁx X and consider the one-parameter family of placements

x = χ(X, t) , (X ﬁxed) . (3.25)

The resulting parametric equations (with parameter t) represent in algebraic form the particle

path of the given particle, see Figure 3.5. Alternatively, one may express the same particle

path in diﬀerential form as

dx = ˆ v(X, τ)dτ , x(t

0

) = X , (X ﬁxed) , (3.26)

or, equivalently,

dy = ˜ v(y, τ)dτ , y(t) = x , (x ﬁxed) . (3.27)

X

x

v

1

0

Figure 3.5: Particle path of a particle which occupies X in the reference conﬁguration.

Now, let v = ˜ v(x, t) be the velocity ﬁeld at a given time t. Deﬁne a stream line through x

at time t as the space curve that passes through x and is tangent to the velocity ﬁeld v at

all of its points, i.e., it is deﬁned according to

dy = ˜ v(y, t)dτ , y(τ

0

) = x , (t ﬁxed) , (3.28)

ME185

D

R

A

F

T

26 Bodies, conﬁgurations and motions

where τ is a scalar parameter and τ

0

some arbitrarily chosen value of τ, see Figure 3.6. Using

components, the preceding deﬁnition becomes

dy

1

˜ v

1

(y

j

, t)

=

dy

2

˜ v

2

(y

j

, t)

=

dy

3

˜ v

3

(y

j

, t)

= dτ , y

i

(τ

0

) = x

i

, (t ﬁxed) . (3.29)

x

y

˜ v(y, t)

dy

Figure 3.6: Stream line through point x at time t.

The streak line through a point x at time t is deﬁned by the equation

y = χ(χ

−1

τ

(x), t) , (3.30)

where τ is a scalar parameter. In diﬀerential form, this line can be expressed as

dy = ˜ v(y, t)dt , y(τ) = x . (3.31)

It is easy to show that the streak line through x at time t is the locus of placements at time t

of all particles that have passed or will pass through x. Equation (3.31) can be derived from

(3.30) by noting that ˜ v(y, t) is the velocity at time τ of a particle which at time τ occupies

the point x, while at time t it occupies the point y.

Note that given a point x at time t, then the path of the particle occupying x at t and

the stream line through x at t have a common tangent. Indeed, this is equivalent to stating

that the velocity at time t of the material point associated with X has the same direction

with the velocity of the point that occupies x = χ(X, t).

In the case of steady motion, the particle path for any particle occupying x coincides with

the stream line and streak line through x at time t. To argue this property, take a stream

line (which is now a ﬁxed curve, since the motion is steady). Consider a material point P

which is on the curve at time t. Notice that the velocity of P is always tangent to the stream

line, this its path line coincides with the stream line through x. A similar argument can be

made for streak lines.

ME185

D

R

A

F

T

Kinematics of deformation 27

In general, path lines can intersect, since intersection points merely mean that diﬀerent

particles can occupy the same position at diﬀerent times. However, stream lines do not inter-

sect, except at points where the velocity vanishes, otherwise the velocity at an intersection

point would have two diﬀerent directions.

Example: Consider a motion χ, such that

χ

1

= χ

1

(X

A

, t) = X

1

e

t

χ

2

= χ

2

(X

A

, t) = X

2

+ tX

3

(3.32)

χ

3

= χ

3

(X

A

, t) = −tX

2

+ X

3

,

with reference to ﬁxed orthonormal system ¦e

i

¦. Note that x = X at time t = 0, i.e., the

body occupies the reference conﬁguration at time t = 0.

The inverse mapping χ

−1

t

is easily obtained as

X

1

= χ

−1

t

1

(x

j

) = x

1

e

−t

X

2

= χ

−1

t

2

(x

j

) =

x

2

−tx

3

1 +t

2

(3.33)

X

3

= χ

−1

t

3

(x

j

) =

tx

2

+ x

3

1 +t

2

.

The velocity ﬁeld, written in the referential description has components ˆ v

i

(X

A

, t) =

∂χ

i

(X

A

, t)

∂t

, i.e.,

ˆ v

1

(X

A

, t) = X

1

e

t

ˆ v

2

(X

A

, t) = X

3

(3.34)

ˆ v

3

(X

A

, t) = −X

2

,

while in the spatial description has components ˜ v

i

(χ

j

, t) given by

˜ v

1

(χ

j

, t) = (x

1

e

−t

)e

t

= x

1

˜ v

2

(χ

j

, t) =

tx

2

+ x

3

1 +t

2

(3.35)

˜ v

3

(χ

j

, t) = −

x

2

−tx

3

1 +t

2

.

Note that x = 0 is a stagnation point and, also, that the motion is steady on the x

1

-axis.

ME185

D

R

A

F

T

28 Deformation gradient and other measures of deformation

The acceleration in the referential description has components ˆ a

i

(X

A

, t) =

∂

2

χ

i

(X

A

, t)

∂t

2

,

hence,

ˆ a

1

(X

A

, t) = X

1

e

t

ˆ a

2

(X

A

, t) = 0 (3.36)

ˆ a

3

(X

A

, t) = 0 ,

while in the spatial description the components ˜ a

i

(χ

j

, t) are given by

˜ a

1

(x

j

, t) = x

1

˜ a

2

(x

j

, t) = 0 (3.37)

˜ a

3

(x

j

, t) = 0 .

**3.2 The deformation gradient and other measures of
**

deformation

Consider a body B which occupies its reference conﬁguration 1

0

at time t

0

and the current

conﬁguration 1 at time t. Also, let ¦E

A

¦ and ¦e

i

¦ be two ﬁxed right-hand orthonormal

bases associated with the reference and current conﬁguration, respectively.

Recall that the motion χ is deﬁned so that x = χ(X, t) and consider the deformation of an

inﬁnitesimal material line element dX located at the point X of the reference conﬁguration.

This material element is mapped into another inﬁnitesimal line element dx at point x in the

current conﬁguration at time t, see Figure 3.7.

It follows from chain rule that

dx =

¸

∂χ

∂X

(X, t)dX = FdX , (3.38)

where F is the deformation gradient tensor, deﬁned as

F =

∂χ(X, t)

∂X

. (3.39)

ME185

D

R

A

F

T

Kinematics of deformation 29

x

dx

X

dX

1 1

0

Figure 3.7: Mapping of an inﬁnitesimal material line elements dX from the reference to the

current conﬁguration.

Using components, the deformation gradient tensor can be expressed as

∂χ

i

(X

B

, t)

∂X

A

e

i

⊗E

A

= F

iA

e

i

⊗E

A

. (3.40)

It is clear from the above that the deformation gradient is a two-point tensor which has one

“leg” in the reference conﬁguration and the other in the current conﬁguration. It follows

from (3.38) that the deformation gradient F provides the rule by which inﬁnitesimal line

element are mapped from the reference to the current conﬁguration. Using (3.40), one may

rewrite (3.38) in component form as

dx

i

= F

iA

dX

A

= χ

i,A

dX

A

. (3.41)

Recall now that the motion χ is assumed invertible for ﬁxed t. Also, recall the inverse

function theorem of real analysis, which, in the case of the mapping χ can be stated as

follows: For a ﬁxed time t, let χ

t

: 1

0

→ 1 be continuously diﬀerentiable (i.e.,

∂χ

t

∂X

exists

and is continuous) and consider X ∈ 1

0

, such that det

∂χ

t

∂X

(X) = 0. Then, there is an open

neighborhood {

0

of X in 1

0

and an open neighborhood { of 1, such that χ

t

({

0

) = {

and χ

t

has a continuously diﬀerentiable inverse χ

−1

t

, so that χ

−1

t

({) = {

0

, as in Figure 3.8.

Moreover, for any x ∈ {, X = χ

−1

t

(x) and

∂χ

−1

t

(x)

∂x

= (F(X, t))

−11

.

1

This means that the derivative of the inverse motion with respect to x is identical to the inverse of the

gradient of the motion with respect to X.

ME185

D

R

A

F

T

30 Deformation gradient and other measures of deformation

x X

1 1

0

{ {

0

χ

Figure 3.8: Application of the inverse function theorem to the motion χ at a ﬁxed time t.

The inverse function theorem states that the mapping χ is invertible at a point X for a

ﬁxed time t, if the Jacobian determinant J = det

∂χ(X, t)

∂X

= det F satisﬁes the condition

J = 0. The condition det J = 0 will be later amended to be det J > 0.

Generally, the inﬁnitesimal material line element dX stretches and rotates to dx under

the action of F. To explore this, write dX = MdS and dx = mds, where M and m are unit

vectors (i.e., M M = m m = 1) in the direction of dX and dx, respectively and dS > 0

and ds > 0 are the inﬁnitesimal lengths of dX and dx, respectively.

Next, deﬁne the stretch λ of the inﬁnitesimal material line element dX as

λ =

ds

dS

, (3.42)

and note that

dx = FdX = FMdS

= mds , (3.43)

hence

λm = FM . (3.44)

Since det F = 0, it follows that λ = 0 and, in particular, that λ > 0, given that m is chosen

to reﬂect the sense of x. In summary, the preceding arguments imply that λ ∈ (0, ∞).

To determine the value of λ, take the dot-product of each side of (3.44) with itself, which

ME185

D

R

A

F

T

Kinematics of deformation 31

leads to

(λm) (λm) = λ

2

(m m) = λ

2

= (FM) (FM)

= M F

T

(FM)

= M (F

T

F)M

= M CM , (3.45)

where C is the right Cauchy-Green deformation tensor, deﬁned as

C = F

T

F (3.46)

or, upon using components,

C

AB

= F

iA

F

iB

. (3.47)

Note that, in general C = C(X, t), since F = F(X, t). Also, it is important to observe

that C is deﬁned with respect to the basis in the reference conﬁguration. Further, C is a

symmetric and positive-deﬁnite tensor. The latter means that, M CM > 0 for any M = 0

and MCM = 0, if, and only if M = 0, both of which follow clear from (3.45). To summarize

the meaning of C, it can be said that, given a direction M in the reference conﬁguration, C

allows the determination of the stretch λ of an inﬁnitesimal material line element dX along

M when mapped to the line element dx in the current conﬁguration.

Recall that the inverse function theorem implies that, since det F = 0,

dX =

∂χ

−1

t

(x)

∂x

dx = F

−1

dx . (3.48)

Then, one may write, in analogy with the preceding derivation of C, that

dX = F

−1

dx = F

−1

mds

= MdS , (3.49)

hence

1

λ

M = F

−1

m . (3.50)

ME185

D

R

A

F

T

32 Deformation gradient and other measures of deformation

Again, taking the dot-products of each side with itself leads to

(

1

λ

M) (

1

λ

M) =

1

λ

2

(M M) =

1

λ

2

= (F

−1

m) (F

−1

m)

= m F

−T

(F

−1

m)

= M (F

−T

F

−1

)m

= m B

−1

m , (3.51)

where F

−T

= (F

−1

)

T

= (F

T

)

−1

and B is the left Cauchy-Green tensor, deﬁned as

B = FF

T

(3.52)

or, using components,

B

ij

= F

iA

F

jA

. (3.53)

In contrast to C, the tensor B is deﬁned with respect to the basis in the current conﬁgu-

ration. Like C, it is easy to establish that the tensor B is symmetric and positive-deﬁnite.

To summarize the meaning of B, it can be said that, given a direction m in the current

conﬁguration, B allows the determination of the stretch λ of an inﬁnitesimal element dx

along m which is mapped from an inﬁnitesimal material line element dX in the reference

conﬁguration.

Consider now the diﬀerence in the squares of the lengths of the line elements dX and dx,

namely, ds

2

−dS

2

and write this diﬀerence as

ds

2

−dS

2

= (dx dx) −(dX dX)

= (FdX) (FdX) −(dX dX)

= dX F

T

(FdX) −(dX dX)

= dX (CdX) −(dX dX)

= dX (C−I)dX

= dX 2EdX , (3.54)

where

E =

1

2

(C−I) =

1

2

(F

T

F −I) (3.55)

ME185

D

R

A

F

T

Kinematics of deformation 33

is the (relative) Lagrangian strain tensor. Using components, the preceding equation can be

written as

E

AB

=

1

2

(F

iA

F

iB

−δ

AB

) , (3.56)

which shows that the Lagrangian strain tensor E is deﬁned with respect to the basis in the

reference conﬁguration. In addition, E is clearly symmetric and satisﬁes the property that

E = 0 when the body undergoes no deformation between the reference and the current

conﬁguration.

The diﬀerence ds

2

−dS

2

may be also written as

ds

2

−dS

2

= (dx dx) −(dX dX)

= (dx dx) −(F

−1

dx) (F

−1

dx)

= (dx dx) −dx F

−1

(F

−1

dx)

= dx dx −(dx B

−1

dx)

= dx (i −B

−1

)dx

= dX 2edX , (3.57)

where

e =

1

2

(i −B

−1

) (3.58)

is the (relative) Eulerian strain tensor or Almansi tensor, and i is the identity tensor in the

current conﬁguration. Using components, one may write

e

ij

=

1

2

(δ

ij

−B

−1

ij

) . (3.59)

Like E, the tensor e is symmetric and vanishes when there is no deformation between the

current conﬁguration remains undeformed relative to the reference conﬁguration. However,

unlike E, the tensor e is naturally resolved into components on the basis in the current

conﬁguration.

While, in general, the inﬁnitesimal material line element dXis both stretched and rotated

due to F, neither C (or B) nor E (or e) yield any information regarding the rotation of dX.

To extract rotation-related information fromF, recall the polar decomposition theorem, which

states that any invertible tensor F can be uniquely decomposed into

F = RU = VR , (3.60)

ME185

D

R

A

F

T

34 Deformation gradient and other measures of deformation

where R is an orthogonal tensor and U, V are symmetric positive-deﬁnite tensors. In com-

ponent form, the polar decomposition is expressed as

F

iA

= R

iB

U

BA

= V

ij

R

jA

. (3.61)

The tensors (R, U) or (R, V) are called the polar factors of F.

Given the polar decomposition theorem, it follows that

C = F

T

F = (RU)

T

(RU) = U

T

R

T

RU = UU = U

2

(3.62)

and, likewise,

B = FF

T

= (VR)(VR)

T

= VRR

T

V = VV = V

2

. (3.63)

The tensors U and V are called the right stretch tensor and the left stretch tensor, respec-

tively. Given their relations to C and B, it is clear that these tensors may be used to deter-

mine the stretch of the inﬁnitesimal material line element dX, which explains their name.

Also, it follows that the component representation of the two tensors is U = U

AB

E

A

⊗ E

B

and V = V

ij

e

i

⊗e

j

, i.e., they are resolved naturally on the bases of the reference and current

conﬁguration, respectively. Also note that the relations U = C

1/2

and V = B

1/2

hold true,

as it is possible to deﬁne the square-root of a symmetric positive-deﬁnite tensor.

Of interest now is R, which is a two-point tensor, i.e., R = R

iA

e

i

⊗ E

A

. Recalling

equation (3.38), attempt to provide a physical interpretation of the decomposition theorem,

starting from the right polar decomposition F = RU. Using this decomposition, and taking

into account (3.38), write

dx = FdX = (RU)dX = R(UdX) . (3.64)

This suggests that the deformation of dX takes place in two steps. In the ﬁrst one, dX is

deformed into dX

′

= UdX, while in the second one, dX

′

is further deformed into RdX

′

= dx.

ME185

D

R

A

F

T

Kinematics of deformation 35

Note that, letting dX

′

= M

′

dS

′

, where M

′

M

′

= 1 and dS

′

is the magnitude of dX

′

,

dX

′

dX

′

= (M

′

dS

′

) (M

′

dS

′

) = dS

′

2

= (UdX) (UdX)

= dX (CdX)

= (MdS) (CMdS)

= dS

2

M CM

= dS

2

λ

2

, (3.65)

which, since λ =

ds

dS

, implies that dS

′

= ds. Thus, dX

′

is stretched to the same diﬀerential

length as dx due to the action of U. Subsequently, write

dx dx = (RdX

′

) (RdX

′

) = dX

′

(R

T

RdX

′

) = dX

′

dX

′

, (3.66)

which conﬁrms that R induces a length-preserving transformation on X

′

. In conclusion,

F = RU implies that dXis ﬁrst subjected to a stretch U (possibly accompanied by rotation)

to its ﬁnal length ds, then is reoriented to its ﬁnal state dx by R, see Figure 3.9.

dx dX dX

′

U R

Figure 3.9: Interpretation of the right polar decomposition.

Turning attention to the left polar decomposition F = VR, note that

dx = FdX = (VR)dX = V(RdX) . (3.67)

This, again, implies that the deformation of dX takes place in two steps. Indeed, in the

ﬁrst one, dX is deformed into dx

′

= RdX, while in the second one, dx

′

is mapped into

Vdx

′

= dx. For the ﬁrst step, note that

dx

′

dx

′

= (RdX) (RdX) = dX (R

T

RdX) = dX dX , (3.68)

ME185

D

R

A

F

T

36 Deformation gradient and other measures of deformation

which means that the mapping from dX to dx

′

is length-preserving. For the second step,

write

dx

′

dx

′

= dX dX = dS

2

= (V

−1

dx) (V

−1

dx)

= dx (B

−1

dx)

= (mds) (B

−1

mds)

= ds

2

m B

−1

m

=

1

λ

2

ds

2

, (3.69)

which implies that V induces the full stretch λ during the mapping of dx

′

to dx.

Thus, the left polar decomposition F = VR means that the inﬁnitesimal material line

element dX is ﬁrst subjected to a length-preserving transformation, followed by stretching

(with further rotation) to its ﬁnal state dx, see Figure 3.10.

dx dX dx

′

V R

Figure 3.10: Interpretation of the left polar decomposition.

It is conceptually desirable to decompose the deformation gradient into a pure rotation

and a pure stretch (or vice versa). To investigate such an option, consider ﬁrst the right polar

decomposition of equation (3.66). In this case, for the stretch U to be pure, the inﬁnitesimal

line elements dX and dX

′

need to be parallel, namely

dX

′

= UdX = λdX , (3.70)

or, upon recalling that dX = MdS,

UM = λM . (3.71)

ME185

D

R

A

F

T

Kinematics of deformation 37

Equation (3.71) represents a linear eigenvalue problem. The eigenvalues λ

A

> 0 of (3.71)

are called the principal stretches and the associated eigenvectors M

A

are called the principal

directions. When λ

A

are distinct, one may write

UM

A

= λ

(A)

M

(A)

UM

B

= λ

(B)

M

(B)

, (3.72)

where the parentheses around the subscripts signify that the summation convention is not

in use. Upon premultiplying the preceding two equations with M

B

and M

A

, respectively,

one gets

M

B

(UM

A

) = λ

(A)

M

B

M

(A)

(3.73)

M

A

(UM

B

) = λ

(B)

M

A

M

(B)

. (3.74)

Subtracting the two equations from each other leads to

(λ

(A)

−λ

(B)

)M

(A)

M

(B)

= 0 , (3.75)

therefore, since, by assumption, λ

A

= λ

B

, it follows that

M

A

M

B

= δ

AB

, (3.76)

i.e., that the principal directions are mutually orthogonal and that ¦M

1

, M

2

, M

3

¦ form an

orthonormal basis on E

3

.

It turns out that regardless of whether Uhas distinct or repeated eigenvalues, the classical

spectral representation theorem guarantees that there exists an orthonormal basis M

A

of E

3

consisting entirely of eigenvectors of U and that, if λ

A

are the associated eigenvalues,

U =

3

¸

A=1

λ

(A)

M

(A)

⊗M

(A)

. (3.77)

ME185

D

R

A

F

T

38 Deformation gradient and other measures of deformation

Thus,

C = U

2

= (

3

¸

A=1

λ

(A)

M

(A)

⊗M

(A)

)(

3

¸

B=1

λ

(B)

M

(B)

⊗M

(B)

)

=

3

¸

A=1

3

¸

B=1

λ

(A)

λ

(B)

(M

(A)

⊗M

(A)

)(M

(B)

⊗M

(B)

)

=

3

¸

A=1

3

¸

B=1

λ

(A)

λ

(B)

(M

(A)

M

(B)

)(M

(A)

⊗M

(B)

)

=

3

¸

A=1

λ

2

(A)

M

(A)

⊗M

(A)

(3.78)

and, more generally,

U

m

=

3

¸

A=1

λ

m

(A)

M

(A)

⊗M

(A)

, (3.79)

for any integer m.

Now, attempt a reinterpretation of the right polar decomposition, in light of the discussion

of principal stretches and directions. Indeed, when U applies on inﬁnitesimal material line

elements which are aligned with the principal directions M

A

, then it subjects them to a pure

stretch. Subsequently, the stretched elements are reoriented to their ﬁnal direction by the

action of R, see Figure 3.11.

x

X

M

1

M

2

M

3

λ

1

M

1

λ

2

M

2

λ

3

M

3

λ

1

RM

1

λ

2

RM

2

λ

3

RM

3

U

R

Figure 3.11: Interpretation of the right polar decomposition relative to the principal directions

M

A

and associated principal stretches λ

A

.

Following an analogous procedure for the left polar decomposition, note for the stretch

V to be pure it is necessary that

dx = Vdx

′

= λdx

′

(3.80)

ME185

D

R

A

F

T

Kinematics of deformation 39

or, recalling that dx

′

= RMdS,

VRM = λRM . (3.81)

This is, again, a linear eigenvalue problem that can be solved for the principal stretches λ

i

and the (rotated) principal directions m

i

= RM

i

. Appealing to the spectral representation

theorem, one may write

V =

3

¸

i=1

λ

(i)

(RM

(i)

) ⊗(RM

(i)

) =

3

¸

i=1

λ

(i)

m

(i)

⊗m

(i)

. (3.82)

A reinterpretation of the left polar decomposition can be aﬀorded along the lines of

the earlier corresponding interpretation of the right decomposition. Speciﬁcally, here the

inﬁnitesimal material line elements that are aligned with the principal stretches M

A

are ﬁrst

reoriented by R and subsequently subjected to a pure stretch to their ﬁnal length by the

action of V, see Figure 3.12.

x

X

M

1

M

2

M

3

RM

1

RM

2

RM

3

λ

1

RM

1

λ

2

RM

2

λ

3

RM

3

V

R

Figure 3.12: Interpretation of the left polar decomposition relative to the principal directions

RM

i

and associated principal stretches λ

i

.

Returning to the discussion of the polar factor R, recall that it is an orthogonal tensor.

This, in turn, implies that (det R)

2

= 1, or det R = ±1. An orthogonal tensor R is termed

proper (resp. improper) if det R = 1 (resp. det R = −1).

Assuming, for now, that R is proper orthogonal, and invoking elementary properties of

ME185

D

R

A

F

T

40 Deformation gradient and other measures of deformation

determinant, it is seen that

R

T

R = I ⇒R

T

R − R

T

= I − R

T

⇒R

T

(R − I) = −(R − I)

T

⇒det R

T

det(R − I) = −det(R − I)

T

⇒det(R − I) = −det(R − I)

⇒det(R − I) = 0 , (3.83)

so that R has at least one unit eigenvalue. Denote by p a unit eigenvector associated with

the above eigenvalue (there exist two such unit vectors), and consider two unit vectors q

and r = p q that lie on a plane normal to p. It follows that ¦p, q, r¦ form a right-hand

orthonormal basis of E

3

and, thus, R can be expressed with reference to this basis as

R = R

pp

p ⊗p + R

pq

p ⊗q + R

pr

p ⊗r + R

qp

q ⊗p + R

qq

q ⊗q + R

qr

q ⊗r

+ R

rp

r ⊗p + R

rq

r ⊗q + R

rr

r ⊗r . (3.84)

Note that, since p is an eigenvector of R,

Rp = p ⇒ R

pp

p + R

qp

q + R

rp

r = p , (3.85)

which implies that

R

pp

= 1 , R

qp

= R

rp

= 0 . (3.86)

Moreover, given that R is orthogonal,

R

T

p = R

−1

p = p ⇒ R

pp

p + R

pq

q + R

pr

r = p , (3.87)

therefore

R

pq

= R

pr

= 0 . (3.88)

Taking into account (3.86) and (3.88), the orthogonality of R can be expressed as

(p ⊗p + R

qq

q ⊗q + R

qr

r ⊗q + R

rq

q ⊗r + R

rr

r ⊗r)

(p ⊗p + R

qq

q ⊗q + R

qr

q ⊗r + R

rq

r ⊗q + R

rr

r ⊗r)

= p ⊗p + q ⊗q + r ⊗r . (3.89)

ME185

D

R

A

F

T

Kinematics of deformation 41

and, after reducing the terms on the left-hand side,

p ⊗p + (R

2

qq

+ R

2

rq

)q ⊗q + (R

2

rr

+ R

2

qr

)r ⊗r

+ (R

qq

R

qr

+ R

rq

R

rr

)q ⊗r + (R

rr

R

rq

+ R

qr

R

qq

)r ⊗q

= p ⊗p + q ⊗q + r ⊗r . (3.90)

The above equation implies that

R

2

qq

+ R

2

rq

= 1 , (3.91)

R

2

rr

+ R

2

qr

= 1 , (3.92)

R

qq

R

qr

+ R

rq

R

rr

= 0 , (3.93)

R

rr

R

rq

+ R

qr

R

qq

= 0 . (3.94)

Equations (3.91) and (3.92) imply that there exist angles θ and φ, such that

R

qq

= cos θ , R

rq

= sin θ , (3.95)

and

R

rr

= cos φ , R

qr

= sin φ . (3.96)

It follows from (3.93) (or (3.94)) that

cos θ sin φ + sin θ cos φ = sin (φ + θ) = 0 , (3.97)

thus

φ = −θ + 2kπ or φ = π − θ + 2kπ , (3.98)

where k is an arbitrary integer. It can be easily seen that the latter choice yields an improper

orthogonal tensor R, thus φ = −θ + 2kπ, and, consequently,

R

qq

= cos θ , (3.99)

R

rq

= sin θ , (3.100)

R

rr

= cos θ , (3.101)

R

qr

= −sin θ . (3.102)

ME185

D

R

A

F

T

42 Deformation gradient and other measures of deformation

From (3.86), (3.88) and (3.99-3.102), it follows that R can be expressed as

R = p ⊗p + cos θ (q ⊗q + r ⊗r) − sin θ (q ⊗r − r ⊗q) . (3.103)

The angle θ that appears in (3.103) can be geometrically interpreted as follows: let an

arbitrary vector x be written in terms of ¦p, q, r¦ as

x = pp + qq + rr , (3.104)

where

p = p x , q = q x , r = r x , (3.105)

and note that

Rx = pp + (q cos θ − r sin θ)q + (q sin θ + r cos θ)r . (3.106)

Equation (3.106) indicates that, under the action of R, the vector x remains unstretched

and it rotates by an angle θ around the p-axis, where θ is assumed positive when directed

from q to r in the sense of the right-hand rule.

q

r

θ

Rx

p

x

Figure 3.13: Geometric interpretation of the rotation tensor R by its action on a vector x.

The representation (3.103) of a proper orthogonal tensor R is often referred to as the

Rodrigues formula. If R is improper orthogonal, the alternative solution in (3.98

2

in connec-

tion with the negative unit eigenvalue p implies that, Rx rotates by an angle θ around the

p-axis and is also reﬂected relative to the origin of the orthonormal basis ¦p, q, r¦.

ME185

D

R

A

F

T

Kinematics of deformation 43

A simple counting check can be now employed to assess the polar decomposition (3.60).

Indeed, F has nine independent components and U (or V) has six independent components.

At the same time, R has three independent components, for instance two of the three

components of the unit eigenvector p and the angle θ.

Example: Consider a motion χ deﬁned in component form as

χ

1

= χ

1

(X

A

, t) = (

√

a cos ϑX

1

) −(

√

a sin ϑ)X

2

χ

2

= χ

2

(X

A

, t) = (

√

a sin ϑ)X

1

+ (

√

a cos ϑ)X

2

χ

3

= χ

3

(X

A

, t) = X

3

,

where a = a(t) > 0 and ϑ = ϑ(t).

This is clearly a planar motion, speciﬁcally independent of X

3

.

The components F

iA

= χ

i,A

of the deformation gradient can be easily determined as

[F

iA

] =

√

a cos ϑ −

√

a sin ϑ 0

√

a sin ϑ

√

a cos ϑ 0

0 0 1

¸

¸

¸

¸

.

This is clearly a spatially homogeneous motion, i.e., the deformation gradient is independent

of X. Further, note that det(F

iA

) = a > 0, hence the motion is always invertible.

The components C

AB

of C and the components U

AB

of U can be directly determined as

[C

AB

] =

a 0 0

0 a 0

0 0 1

¸

¸

¸

¸

and

[U

AB

] =

√

a 0 0

0

√

a 0

0 0 1

¸

¸

¸

¸

.

Also, recall that

CM = λ

2

M ,

which implies that λ

1

= λ

2

=

√

a and λ

3

= 1.

ME185

D

R

A

F

T

44 Deformation gradient and other measures of deformation

Given that U is known, one may apply the right polar decomposition to determine the

rotation tensor R. Indeed, in this case,

[R

iA

] =

√

a cos ϑ −

√

a sin ϑ 0

√

a sin ϑ

√

a cos ϑ 0

0 0 1

¸

¸

¸

¸

1

√

a

0 0

0

1

√

a

0

0 0 1

¸

¸

¸

¸

=

cos ϑ −sin ϑ 0

sin ϑ cos ϑ 0

0 0 1

¸

¸

¸

¸

.

Note that this motion yield pure stretch for ϑ = 2kπ, where k = 0, 1, 2, . . ..

Example: Sphere under homogeneous deformation

Consider the part of a deformable body which occupies a sphere {

0

of radius σ centered

at the ﬁxed origin O of c

3

. The equation of the surface ∂{

0

of the sphere can be written as

Π Π = σ

2

, (3.107)

where

Π = σM , (3.108)

and M M = 1. Assume that the body undergoes a spatially homogeneous deformation with

deformation gradient F(t), so that

π = FΠ , (3.109)

where π(t) is the image of Π in the current conﬁguration. Recall that λm = FM, where

λ(t) is the stretch of a material line element that lies along M in the reference conﬁguration,

O O

M

m

{ {

0

Π

π

Figure 3.14: Spatially homogeneous deformation of a sphere.

and m(t) is the unit vector in the direction this material line element at time t, as in the

ﬁgure above. Then, it follows from (3.107) and (3.109) that

π = σλm . (3.110)

ME185

D

R

A

F

T

Kinematics of deformation 45

Also, since λ

2

m B

−1

m = 1, where B(t) is the left Cauchy-Green deformation tensor,

equation (3.110) leads to

π B

−1

π = σ

2

. (3.111)

Recalling the spectral representation theorem, the left Cauchy-Green deformation tensor

can be resolved as

B =

3

¸

i=1

λ

2

(i)

m

(i)

⊗m

(i)

, (3.112)

where (λ

2

i

, m

i

), i = 1, 2, 3, are respectively the eigenvalues (squares of the principal stretches)

and the associated eigenvectors (principal directions) of B. In addition, note that m

i

, i =

1, 2, 3, are orthonormal and form a basis of E

3

, therefore the position vector π can be

uniquely resolved in this basis as

π = π

i

m

i

. (3.113)

Starting from equation (3.112), one may write

B

−1

=

3

¸

i=1

λ

−2

(i)

m

(i)

⊗m

(i)

, (3.114)

and using (3.113) and (3.114), deduce that

π B

−1

π = λ

−2

i

π

2

i

. (3.115)

It is readily seen from (3.115) that

π

2

1

λ

2

1

+

π

2

2

λ

2

2

+

π

2

3

λ

2

3

= σ

2

,

which demonstrates that, under a spatially homogeneous deformation, the spherical region {

0

is deformed into an ellipsoid with principal semi-axes of length σλ

i

along the principal

directions of B, see Figure 3.15.

Consider now the transformation of an inﬁnitesimal material volume element dV of the

reference conﬁguration to its image dv in the current conﬁguration due to the motion χ.

The referential volume element is deﬁned as an inﬁnitesimal parallelepiped with sides dX

1

,

dX

2

, and dX

3

. Likewise, its spatial counterpart is the inﬁnitesimal parallelepiped with sides

dx

1

, dx

2

, and dx

3

, where each dx

i

is the image of dX

i

due to χ, see Figure 3.15.

ME185

D

R

A

F

T

46 Deformation gradient and other measures of deformation

m

1

m

2

m

3

Figure 3.15: Image of a sphere under homogeneous deformation.

x X

1

0 1

dX

1

dX

2

dX

3

dx

1

dx

2

dx

3

Figure 3.16: Mapping of an inﬁnitesimal material volume element dV to its image dv in the

current conﬁguration.

First, note that

dV = dX

1

(dX

2

dX

3

) = dX

2

(dX

3

dX

1

) = dX

3

(dX

1

dX

2

) , (3.116)

where each of the representation of dV in (3.116) corresponds to the triple product [dX

1

dX

2

dX

3

]

of the vectors dX

1

, dX

2

and dX

3

. Recalling that the triple product satisﬁes the property

[dX

1

dX

2

dX

3

] = det [[dX

1

A

], [dX

2

A

], [dX

3

A

]], it follows that in the current conﬁguration

dv = dx

1

(dx

2

dx

3

)

= (FdX

1

)

(FdX

2

) FdX

3

)

= det

[F

iA

dX

1

A

], [F

iA

dX

2

A

], [F

iA

dX

3

A

]

= det

F

iA

[dX

1

A

, dX

2

A

, dX

3

A

]

= (det F) det

[dX

1

A

], [dX

2

A

], [dX

3

A

]

= JdV . (3.117)

ME185

D

R

A

F

T

Kinematics of deformation 47

A motion for which dv = dV (i.e. J = 1) for all inﬁnitesimal material volumes dV at all

times is called isochoric (or volume preserving).

Here, one may argue that if, by convention, dV > 0 (which is true as long as dX

1

, dX

2

and dX

3

observe the right-hand rule), then the relative orientation of line elements should

be preserved during the motion, i.e., J > 0 everywhere and at all time. Indeed, since the

motion is assumed smooth, any changes in the sign of J would necessarily imply that there

exists a time t at which J = 0 at some material point(s), which would violate the assumption

the assumption of invertibility. Based on the preceding observation, the Jacobian J will be

taken to be positive at all time. Consider next the transformation of an inﬁnitesimal material

surface element dA in the reference conﬁguration to its image da in the current conﬁguration.

The referential surface element is deﬁned as the parallelogram formed by the inﬁnitesimal

material line elements dX

1

and dX

2

, such that

dA = dX

1

dX

2

= NdA , (3.118)

where N is the unit normal to the surface element consistently with the right-hand rule, see

Figure 3.16. Similarly, in the current conﬁguration, one may write

da = dx

1

dx

2

= nda , (3.119)

where n is the corresponding unit normal to the surface element deﬁned by dx

1

and dx

2

.

x X

1

0 1

dX

1

dX

2

dx

1

dx

2

Figure 3.17: Mapping of an inﬁnitesimal material surface element dA to its image da in the

current conﬁguration.

Let dX be any inﬁnitesimal material line element, such that N dX > 0 and consider

the inﬁnitesimal volumes dV and dv formed by ¦dX

1

, dX

2

, dX¦ and ¦dx

1

, dx

2

, dx = FdX¦,

ME185

D

R

A

F

T

48 Velocity gradient and other measures of deformation rate

respectively. It follows from (3.117) that

dv = dx (dx

1

dx

2

) = dx nda = (FdX) nda

= JdV

= JdX (dX

1

dX

2

) = JdX NdA , (3.120)

which implies that

(FdX) nda = JdX NdA , (3.121)

for any inﬁnitesimal material line element dX. This, in turn, leads to

nda = JF

−T

NdA . (3.122)

Equation (3.122) is known as Nanson’s formula. Dotting each side in (3.122) with itself and

taking square roots yields

[da[ = J

N (C

−1

N)[dA[ . (3.123)

As argued in the case of the inﬁnitesimal volume transformations, if an inﬁnitesimal material

line element satisﬁes dA > 0, then da > 0 everywhere and at all time, since J > 0 and C

−1

is positive-deﬁnite.

3.3 Velocity gradient and other measures of deforma-

tion rate

In this section, interest is focused to measures of the rate at which deformation occurs in

the continuum. To this end, deﬁne the spatial velocity gradient tensor L as

L = gradv =

∂v

∂x

. (3.124)

This tensor is naturally deﬁned relative to the basis ¦e

i

¦ in the current conﬁguration. Next,

recall any such tensor can be uniquely decomposed into a symmetric and a skew-symmetric

part, so that L can be written as

L = D+W , (3.125)

where

D =

1

2

(L +L

T

) (3.126)

ME185

D

R

A

F

T

Kinematics of deformation 49

is the rate-of-deformation tensor, which is symmetric, and

W =

1

2

(L −L

T

) (3.127)

is the vorticity (or spin) tensor, which is skew-symmetric.

Consider now the rate of change of the deformation gradient for a ﬁxed particle associated

with point X in the reference conﬁguration. To this end, write the material time derivative

of F as

˙

F =

˙

∂χ(X, t)

∂X

=

∂

∂X

˙

χ(X, t) =

∂ˆ v(X, t)

∂X

=

∂˜ v(x, t)

∂x

∂χ(X, t)

∂X

= LF , (3.128)

where use is made of the chain rule. Also, in the above derivation the change in the order

of diﬀerentiation between the derivatives with respect to X and t is allowed under the

assumption that the mixed second derivative

∂

2

χ

∂X∂t

is continuous.

Given (3.128), one may express with the aid of the product rule the rate of change of the

right Cauchy-Green deformation tensor C for a ﬁxed particle X as

˙

C =

˙

F

T

F =

˙

F

T

F +F

T

˙

F = (LF)

T

F +F

T

(LF) = F

T

(L

T

+L)F = 2F

T

DF . (3.129)

Likewise, for the left Cauchy-Green deformation tensor, one may write

˙

B =

˙

FF

T

=

˙

FF

T

+F

˙

F

T

= (LF)F

T

+F(LF)

T

= LB+BL

T

. (3.130)

Similar results may be readily obtained for the rates of the Lagrangian and Eulerian strain

measures. Speciﬁcally, it can be shown that

˙

E = F

T

DF (3.131)

and

˙ e =

1

2

(B

−1

L +L

T

B

−1

) . (3.132)

Proceed now to discuss physical interpretations for the rate tensors D and W. Starting from

(3.44), take the material time derivatives of both sides to obtain the relation

˙

λm+ λ ˙ m =

˙

FM+F

˙

M

= LFM = L(λm) = λLm . (3.133)

ME185

D

R

A

F

T

50 Velocity gradient and other measures of deformation rate

Note that

˙

M = 0, since M is a ﬁxed vector in the ﬁxed reference conﬁguration, hence does

not vary with time. Upon taking the dot-product of each side with m, it follows that

˙

λm m+ λ ˙ m m = λ(Lm) m . (3.134)

However, since m is a unit vector, it follows that ˙ m m = 0, so that the preceding equation

simpliﬁes to

˙

λ = λm Lm . (3.135)

Further, since the skew-symmetric part W of L satisﬁes

m Wm = m (−W

T

)m = −m Wm , (3.136)

hence m Wm = 0, one may rewrite (3.135) as

˙

λ = λm Dm (3.137)

or, alternatively, as

˙

ln λ = m Dm = D (m⊗m) . (3.138)

Thus, the tensor D fully determines the material time derivative of the logarithmic stretch

ln λ for a material line element along a direction m in the current conﬁguration.

Before proceeding with the geometric interpretation of W, some preliminary development

on skew-symmetric tensors is required. First, note that any skew-symmetric tensor in E

3

,

such as W, has only only three independent components. This suggests that there exists a

one-to-one correspondence between skew-symmetric tensors and vectors in E

3

. To establish

this correspondence, observe that

W =

1

2

(W−W

T

) , (3.139)

so that, when operating on any vector z ∈ E

3

,

Wz =

1

2

W

ij

(e

i

⊗e

j

−e

j

⊗e

i

)z

=

1

2

W

ij

[(z e

j

)e

i

−(z e

i

)e

j

] . (3.140)

ME185

D

R

A

F

T

Kinematics of deformation 51

Recalling the standard identity u (v w) = (u w)v − (v u)w, the preceding equation

can be rewritten as

Wz =

1

2

W

ij

[z (e

i

e

j

)]

= −

1

2

W

ij

[(e

i

e

j

) z]

=

1

2

W

ji

[(e

i

e

j

) z]

=

¸

1

2

W

ji

e

i

e

j

]

z

= wz , (3.141)

where the vector w is deﬁned as

w =

1

2

W

ji

e

i

e

j

(3.142)

and is called the axial vector of the skew-symmetric tensor W. Using components, one may

represent W in terms of w and vice-versa. Speciﬁcally, starting from (3.142),

w = w

k

e

k

=

1

2

W

ji

e

i

e

j

=

1

2

W

ji

e

ijk

e

k

, (3.143)

hence,

w

k

=

1

2

e

ijk

W

ji

. (3.144)

Conversely, starting from (3.141),

W

ij

z

j

e

i

= e

ijk

w

j

z

k

e

i

= e

ikj

w

k

z

j

e

i

, (3.145)

so that

W

ij

= e

ikj

w

k

= e

jik

w

k

. (3.146)

The above relations apply to any skew-symmetric tensor. If W is speciﬁcally the vorticity

ME185

D

R

A

F

T

52 Velocity gradient and other measures of deformation rate

tensor, then the associated axial vector satisﬁes the relation

w =

1

4

(v

j,i

−v

i,j

)e

i

e

j

=

1

4

(v

j,i

−v

i,j

)e

ijk

e

k

=

1

4

(e

ijk

v

j,i

−e

ijk

v

i,j

)e

k

=

1

4

(e

ijk

v

j,i

−e

jik

v

i,j

)e

k

=

1

2

e

ijk

v

j,i

e

k

=

1

2

curl v . (3.147)

In this case, the axial vector w is called the vorticity vector. Also, a motion is termed

irrotational if W = 0 (or, equivalently, w = 0).

Let w = ˜ w(x, t) be the vorticity vector ﬁeld at a given time t. The vortex line through x

at time t is the space curve that passes through x and is tangent to the vorticity vector ﬁeld

˜ w at all of its points, hence is deﬁned as

dy = ˜ w(y, t)dτ , y(τ

0

) = x , (t ﬁxed) . (3.148)

For an irrotational motion, any line passing through x at time t is a vortex line.

Returning to the geometric interpretation of W, take ¯ m to be a unit vector that lies

along a principal direction of D in the current conﬁguration, namely

(D−γi) ¯ m = 0 . (3.149)

It follows from (3.149) that

(D¯ m) ¯ m = γ ¯ m ¯ m = γ =

˙

¯

λ

¯

λ

=

˙

ln

¯

λ, (3.150)

i.e., the eigenvalues of Dare equal to the material time derivatives of the logarithmic stretches

ln

¯

λ of line elements along the eigendirections ¯ m in the current conﬁguration.

Recalling the identity (3.133), write

˙ m = Lm−

˙

λ

λ

m =

L −

˙

λ

λ

i

m

=

D−

˙

λ

λ

i

m+Wm , (3.151)

ME185

D

R

A

F

T

3.4. SUPERPOSED RIGID-BODY MOTIONS 53

which holds for any direction m in the current conﬁguration. Setting in the above equation

m = ¯ m, it follows that

˙

¯ m = W¯ m = w ¯ m . (3.152)

Therefore, the material time derivative of a unit vector ¯ m along a principal direction of D

is determined by (3.152). Recalling from rigid-body dynamics the formula relating linear to

angular velocities, one may conclude that w plays the role of the angular velocity of a line

element which, in the current conﬁguration, lies along the principal direction m of D.

3.4 Superposed rigid-body motions

A rigid motion is one in which the distance between any two material points remains constant

at all times. Denoting by X and Y the position vectors of two material points on the ﬁxed

reference conﬁguration, deﬁne the distance function as

d(X, Y) = [X−Y[ . (3.153)

According to the preceding deﬁnition, a motion is rigid if, for any points X and Y,

d(X, Y) = d(χ(X, t), χ(Y, t)) = d(x, y) , (3.154)

for all times t.

Now, introduce the notion of a superposed rigid-body motion. To this end, take a motion

χ of body B such that, as usual, x = χ(X, t). Then, let another motion χ

+

be deﬁned for

B, such that

x

+

= χ

+

(X, t) , (3.155)

so that χ and χ

+

diﬀer by a rigid-body motion. Then, with reference to Figure 3.18, one

may write

x

+

= χ

+

(X, t) = χ

+

(χ

−1

t

(x), t) = ¯ χ

+

(x, t) . (3.156)

Said diﬀerently,

x

+

= χ

+

t

(X) = ¯ χ

+

t

(x) = ¯ χ

+

t

(χ

t

(X)) (3.157)

or, equivalently,

χ

+

t

= ¯ χ

+

t

◦ χ

t

. (3.158)

ME185

D

R

A

F

T

54 Superposed rigid-body motions

x

x

+

X

y

y

+

Y

1

0

1

1

+

χ

χ

+

¯ χ

+

Figure 3.18: Conﬁgurations associated with motions χ and χ

+

diﬀering by a superposed rigid

motion ¯ χ

+

.

Clearly, the superposed motion ¯ χ

+

(x, t) is invertible for ﬁxed t, since χ

+

is assumed invert-

ible for ﬁxed t.

Similarly, take a second point Y in the reference conﬁguration, so that y = χ(Y, t) and

write

y

+

= χ

+

(Y, t) = χ

+

(χ

−1

t

(y), t) = ¯ χ

+

(y, t) . (3.159)

Recalling that 1 and 1

+

diﬀer only by a rigid transformation, one may conclude that

(x −y) (x −y) = (x

+

−y

+

) (x

+

−y

+

)

=

¯ χ

+

(x, t) − ¯ χ

+

(y, t)

¯ χ

+

(x, t) − ¯ χ

+

(y, t)

, (3.160)

for all x, y in the region 1 at time t. Since x and y are chosen independently, one may

diﬀerentiate equation (3.160) ﬁrst with respect to x to get

x −y =

¸

∂ ¯ χ

+

(x, t)

∂x

T

(¯ χ

+

(x, t) − ¯ χ

+

(y, t)) . (3.161)

Then, equation (3.161) may be diﬀerentiated with respect to y, which leads to

i =

¸

∂ ¯ χ

+

(x, t)

∂x

T

¸

∂ ¯ χ

+

(y, t)

∂y

. (3.162)

Since the motion ¯ χ

+

is invertible, equation (3.162) can be equivalently written as

¸

∂ ¯ χ

+

(x, t)

∂x

T

=

¸

∂ ¯ χ

+

(y, t)

∂y

−1

. (3.163)

ME185

D

R

A

F

T

Kinematics of deformation 55

Then, the left- and right-hand side should be necessarily functions of time only, hence

¸

∂χ

+

(x, t)

∂x

T

=

¸

∂χ

+

(y, t)

∂y

−1

= Q

T

(t) . (3.164)

Since equation (3.164) implies that

∂χ

+

(x, t)

∂x

=

∂χ

+

(y, t)

∂y

= Q(t), it follows immediately

that Q

T

(t)Q(t) = I, i.e., Q(t) is an orthogonal tensor. Further, note that using the chain

rule the deformation gradient F

+

of the motion χ

+

is written as

F

+

=

∂χ

+

∂X

=

∂χ

+

∂x

∂χ

∂X

= QF . (3.165)

Since, by assumption, both motions χ and χ

+

lead to deformation gradients with positive

Jacobians, equation (3.165) implies that det Q > 0, hence det Q = 1 and Q is proper

orthogonal.

Since Q is a function of time only, equation (3.164)

1

can be directly integrated with

respect to x, leading to

x

+

= ¯ χ

+

(x, t) = Q(t)x +c(t) , (3.166)

where c(t) is a vector function of time. Equation (3.166) is the general form of the superposed

rigid motion ¯ χ on the original motion χ.

Given (3.165)

3

and recalling the right polar decomposition of F, write

F

+

= R

+

U

+

= QF = QRU ,

(3.167)

where R, R

+

are proper orthogonal tensors and U, U

+

are symmetric positive-deﬁnite

tensors. Since, clearly,

(QR)

T

(QR) = (R

T

Q

T

)(QR) = R

T

(Q

T

Q)R = R

T

R = I (3.168)

and also det (QR) = (det Q)(det R) = 1, therefore QR is proper orthogonal, the uniqueness

of the polar decomposition, in conjunction with (3.167), necessitates that

R

+

= QR (3.169)

and

U

+

= U . (3.170)

ME185

D

R

A

F

T

56 Superposed rigid-body motions

Similarly, the left decomposition of F yields

F

+

= V

+

R

+

= V

+

(QR)

= QF = Q(VR) ,

(3.171)

which implies that

V

+

(QR) = Q(VR) , (3.172)

hence,

V

+

= QVQ

T

. (3.173)

It follows readily from (3.165)

3

that

C

+

= F

+

T

F

+

= (QF)

T

(QF) = (F

T

Q

T

)(QF) = F

T

F = C (3.174)

and

B

+

= F

+

F

+

T

= (QF)(QF)

T

= (QF)(F

T

Q

T

) = QBQ

T

. (3.175)

Likewise,

E

+

=

1

2

(C

+

−I) =

1

2

(C−I) = E (3.176)

and

e

+

=

1

2

(I −B

+−1

) =

1

2

[I −(QBQ

T

)

−1

]

=

1

2

(I −Q

−T

B

−1

Q

−1

)

=

1

2

(I −QB

−1

Q

T

)

=

1

2

Q(I −B

−1

)Q

T

= QeQ

T

. (3.177)

A vector or tensor is called objective if it transforms under superposed rigid motions

in the same manner as its natural basis. Physically, this means that the components of

an objective tensor relative to its basis are unchanged if both the tensor and its natural

basis are transformed due to a rigid motion superposed to the original motion. Speciﬁcally,

a referential tensor is objective when it does not change under superposed rigid motions,

ME185

D

R

A

F

T

Kinematics of deformation 57

since its natural basis ¦E

A

⊗ E

B

¦ remains untransformed under superposed rigid motions,

as the latter do not aﬀect the reference conﬁguration. Hence, referential tensors such as

C, U and E are objective. Correspondingly a spatial tensor is objective if it transforms

according to ( )

+

= Q( )Q

T

. This is because its tensor basis ¦e

i

⊗ e

j

¦ transforms as

¦(Qe

i

) ⊗(Qe

j

)¦ = Q¦e

i

⊗e

j

¦Q

T

. Hence, spatial tensors such as B, V and e are objective.

It is easy to deduce that two-point tensors are objective if they transform as ( )

+

= Q( ) or

( )

+

= ( )Q

T

depending on whether the ﬁrst or second leg of the tensor lies in the current

conﬁguration, respectively. Analogous conclusions may be drawn for vectors, which, when

objective, remain untransformed if they are referential or transform as ( )

+

= Q( ) if they

are spatial.

Examine next the transformation of the velocity vector v under a superposed rigid-body

motion. In this case

v

+

= ˙ χ

+

(X, t)

=

˙

χ

+

(x, t) =

˙

[Q(t)x +c(t)] =

˙

Q(t)x +Q(t)v + ˙ c(t) . (3.178)

Since QQ

T

= I,

˙

QQ

T

=

˙

QQ

T

+Q

˙

Q

T

= 0 . (3.179)

Setting

Ω(t) =

˙

Q(t)Q

T

(t) , (3.180)

it follows from (3.179) that Ω is skew-symmetric, hence there exists an axial vector ω such

that Ωz = ω z, for any z ∈ E

3

. Returning to (3.178), write, with the aid of (3.166) and

(3.180),

v

+

= ΩQ(t)x +Q(t)v + ˙ c(t) = Ω(x

+

−c(t)) +Q(t)v + ˙ c(t) . (3.181)

Invoking the deﬁnition of the axial vector ω, one may rewrite (3.181) as

v

+

= ω (x

+

−c) +Qv + ˙ c . (3.182)

Equation (3.181) reveals that the velocity is not an objective vector.

ME185

D

R

A

F

T

58 Superposed rigid-body motions

Next, examine how the various tensorial measures of deformation rate transform under

superposed rigid motions. Starting with the spatial velocity gradient, write

L

+

=

∂v

+

∂x

+

=

∂

∂x

+

[Ω(x

+

−c) +Qv + ˙ c]

= Ω+

∂(Q˜ v)

∂x

+

= Ω+

∂(Q˜ v)

∂x

∂χ

∂x

+

= Ω+Q

∂˜ v

∂x

∂

∂x

+

[Q

T

(x

+

−c)]

= Ω+QLQ

T

, (3.183)

where use is made of (3.181), the deﬁnition of L and the chain rule. It follows from (3.183)

that L is not objective. Also, the rate-of-deformation tensor D transforms according to

D

+

=

1

2

(L

+

+L

+T

)

=

1

2

(Ω+ QLQ

T

) +

1

2

(Ω+QLQ

T

)

T

=

1

2

(Ω+ Ω

T

) +Q

1

2

(L +L

T

)Q

T

= QDQ

T

, (3.184)

which implies that D is objective. Turning to the vorticity tensor W, one may write

W

+

=

1

2

(L

+

−L

+T

)

=

1

2

(Ω+QLQ

T

) −

1

2

(Ω+QLQ

T

)

T

=

1

2

(Ω−Ω

T

) +Q

1

2

(L −L

T

)Q

T

= Ω+QWQ

T

, (3.185)

from where it is seen that W is not objective.

Objectivity conclusions can be drawn for other kinematic quantities of interest by appeal-

ing to results already in place. For instance, inﬁnitesimal material line elements transform

as

dx

+

= F

+

dX = (QF)dX = Q(FdX) = Qdx , (3.186)

ME185

D

R

A

F

T

Kinematics of deformation 59

hence dx is objective. Similarly, inﬁnitesimal material volume elements transform as

dv

+

= J

+

dV = det(QF)dV = (det Q)(det F)dV = (det F)dV = JdV = dv , (3.187)

hence dv is an objective scalar quantity. For inﬁnitesimal material area elements, write

da

+

= n

+

da

+

= J

+

F

+−T

NdA

= J(QF)

−T

NdA = J(Q

−T

F

−T

)NdA = JQF

−T

NdA = Qnda = Qda ,

(3.188)

hence the vector da is objective. Now, taking the dot-product of each side of (3.188) with

itself yields

(n

+

da

+

) (n

+

da

+

) = (Qnda) (Qnda) , (3.189)

therefore (da

+

)

2

= da

2

, hence also da

+

= da (as long as da is taken positive from the outset)

and n

+

= Qn.

In closing, note that the superposed rigid motion operation ( )

+

commutes with the

transposition ( )

T

, inversion ( )

−1

and material time derivative

˙

( ) operations.

ME185

D

R

A

F

T

60 Superposed rigid-body motions

ME185

D

R

A

F

T

Chapter 4

Basic physical principles

4.1 The divergence and Stokes’ theorems

By way of background, ﬁrst review the divergence theorem for scalar, vector and tensor

functions. To this end, let { be a bounded closed region with smooth boundary ∂P in the

Euclidean point space c

3

. Recall that { is bounded if it can be enclosed by a sphere of

ﬁnite radius, closed if it contains its boundary and smooth if the normal n to its boundary

is uniquely deﬁned at all boundary points.

Next, deﬁne a scalar function φ : { → R, a vector function v : { → E

3

, and a tensor

function T : { → L(E

3

, E

3

). All three functions are assumed continuously diﬀerentiable.

Then, the gradients of φ and v satisfy

P

grad φ dv =

∂P

φnda , (4.1)

and

P

grad v dv =

∂P

v ⊗nda . (4.2)

In addition, the divergences of v and T satisfy

P

div v dv =

∂P

v nda , (4.3)

and

P

div Tdv =

∂P

Tnda . (4.4)

61

D

R

A

F

T

62 The divergence and Stokes’ theorems

Equation (4.3) expresses the classical divergence theorem. The other three identities can be

derived from this theorem. Indeed, the identity (4.4) can be derived by dotting the left-hand

side by a constant vector c and using (4.3) and the deﬁnition of the divergence of a tensor.

This leads to

P

div Tdv c =

P

div T c dv =

P

div(T

T

c) dv =

∂P

(T

T

c) nda

=

∂P

(Tn) c da =

∂P

Tnda c . (4.5)

Since c is chosen arbitrarily, equation (4.4) follows immediately. Equation (4.1) can be

deduced from (4.4) by setting T = φI, in which case, for any constant vector c,

P

div(φI) dv c =

P

div(φc) dv =

P

tr grad(φc) dv =

P

tr(c ⊗gradφ) dv

=

P

grad φ c dv =

P

grad φ dv c =

P

gradφ dv c

=

∂P

φnda c , (4.6)

which, again due to the arbitrariness of c, implies (4.1). Lastly, (4.2) is obtained from (4.1)

by taking any constant c and writing

¸

P

gradv dv

T

c =

P

(gradv)

T

c dv =

P

grad(v c) dv

=

∂P

(v c)nda =

∂P

(n ⊗v)c da =

¸

∂P

(n ⊗v) da

c , (4.7)

which proves the identity.

Consider a closed non-intersecting curve ( which is parametrized by a scalar τ, 0 ≤ τ ≤ 1,

so that the position vector of a typical point on ( is c(τ). Also, let / be an open surface

bounded by (, see Figure 4.1. Clearly, any point on / possesses two equal and opposite

unit vectors, each pointing outward to one of the two sides of the surface. To eliminate the

ambiguity, choose one of the sides of the surface and denote its outward unit normal by n.

This side is chosen so that c(¯ τ) c(¯ τ + dτ) points toward it, for any ¯ τ ∈ [0, 1). If now v is

a continuously diﬀerentiable vector ﬁeld, then Stokes’ theorem states that

A

curl v ndA =

C

v dx . (4.8)

ME185

D

R

A

F

T

4.2. THE REYNOLDS’ TRANSPORT THEOREM 63

This integral on the right-hand side of (4.8) is called the circulation of the vector ﬁeld v

around (. The circulation is the (inﬁnite) sum of the tangential components of v along (. If

v is identiﬁed as the spatial velocity ﬁeld, then Stokes’ theorem states that the circulation of

the velocity around ( equals to twice the integral of the normal component of the vorticity

vector on any open surface that is bounded by (.

(

/

Figure 4.1: A surface / bounded by the curve (.

4.2 The Reynolds’ transport theorem

Let { be a closed and bounded region in c

3

with smooth boundary ∂{ and assume that the

particles which occupy this region at time t occupy a closed and bounded region {

0

with

smooth boundary ∂{

0

at another time t

0

, see Figure 4.2.

χ

{ ∂{

{

0

∂{

0

1

∂1

1

0

∂1

0

Figure 4.2: A region { with boundary ∂{ and its image {

0

with boundary ∂{

0

in the reference

conﬁguration.

Further, let a scalar ﬁeld φ be deﬁned by a referential function

ˆ

φ or a spatial function

˜

φ,

such that

φ =

ˆ

φ(X, t) =

˜

φ(x, t) . (4.9)

ME185

D

R

A

F

T

64 The Reynolds’ transport theorem

Both

ˆ

φ and

˜

φ are assumed continuously diﬀerentiable in both of their variables. In the

forthcoming discussion of balance laws, it is important to be able to manipulate integrals of

the form

d

dt

P

˜

φdv , (4.10)

namely, material time derivatives of volume integrals deﬁned in some subset of the current

conﬁguration.

Example: Consider the integral in (4.10) for φ = 1. Here,

d

dt

P

dv =

d

dt

vol¦({)¦, which

is the rate of change at time t of the total volume of the region occupied by the material

particles that at time t occupy {.

In evaluating (4.10), one ﬁrst notes that the diﬀerentiation and integration operations

cannot be directly interchanged, because the region { over which the integral is evaluated

is itself a function of time. Therefore, one needs to ﬁrst map the integral to the (ﬁxed)

reference conﬁguration, interchange the diﬀerentiation and integration operations, evaluate

the derivative of the integrand, and then map the integral back to the current conﬁguration.

Taking this approach leads to

d

dt

P

˜

φ dv =

d

dt

P

0

ˆ

φJ dV (pull-back to {

0

)

=

P

0

d

dt

[

ˆ

φJ] dV (interchange

d

dt

and

)

=

P

0

¸

∂

ˆ

φ

∂t

J +

ˆ

φ

∂J

∂t

¸

dV

=

P

0

(

˙

φJ +

ˆ

φJ div v) dV (

˙

J = J div v)

=

P

0

(

˙

φ +

ˆ

φ div v)J dV

=

P

(

˙

φ +

˜

φ div v) dv (push-forward to {) . (4.11)

This result is known as the Reynolds’ transport theorem. Note that the theorem applies also

to vector and tensor functions without any modiﬁcations.

To interpret the Reynolds’ transport theorem, note that the left-hand side of (4.11) is

the rate of change of the integral of φ over the region {, when following the set of particles

that happen to occupy { at time t. The right-hand side of (4.11) consists of the sum of two

ME185

D

R

A

F

T

Basic physical principles 65

terms. The ﬁrst one is the integral of the rate of change of φ for all particles that happen to

occupy { at time t. The second one is due to the rate of change of the volume occupied by

the same particles assuming that

˜

φ remains constant in time for those particles.

The Reynolds’ transport theorem can be restated in a number of equivalent forms. To

this end, write with the aid of the divergence theorem (4.3),

d

dt

P

˜

φdv =

P

(

˙

φ + φ div v) dv

=

P

¸

∂

˜

φ

∂t

+

∂

˜

φ

∂x

v + φ div v

¸

dv

=

P

¸

∂

˜

φ

∂t

+ div(

˜

φv)

¸

dv

=

P

∂

˜

φ

∂t

dv +

∂P

˜

φv nda . (4.12)

An alternative interpretation of the theorem is now in order. Here, the right-hand side of

(4.12) consists of the sum of two terms. The ﬁrst term is the integral of the rate of change

of φ at time t for all points that form the ﬁxed region {. The second term is the rate at

which the volume of { weighted by φ changes as particles exit { across ∂{.

Example: Consider again the special case

˜

φ(x, t) = 1, which corresponds to the transport of

volume. Here,

d

dt

P

dv =

P

div v dv =

∂P

v nda . (4.13)

This means that the rate of change of the volume occupied by the same material particles

equals the boundary integral of the normal component of the velocity v n of ∂{, i.e., the

rate at which the volume of { changes as the particles exit across the boundary ∂

¯

{ of the

ﬁxed region

¯

{ which equals to { at time t.

Starting from (4.12), note that

P

∂

˜

φ

∂t

dv =

∂

∂t

P

˜

φdv, since

∂

˜

φ

∂t

is the rate of change of φ

for ﬁxed position in space (hence, { is also ﬁxed in space). Therefore, one may alternatively

express (4.12) as

d

dt

P

˜

φ dv =

∂

∂t

P

˜

φ dv +

∂P

˜

φv nda . (4.14)

ME185

D

R

A

F

T

66 The localization theorem

4.3 The localization theorem

Another important result with implications in the study of balance laws is presented here by

way of background. Let

˜

φ : 1R →R be a function such that φ =

˜

φ(x, t), where 1 ⊂ E

3

.

Also, let

˜

φ be continuous function in the spatial argument. Then, assume that

P

˜

φ dv = 0 , (4.15)

for all { ⊂ 1 at a given time t. The localization theorem states that this is true if, and only

if,

˜

φ = 0 everywhere in 1 at time t.

To prove this result, ﬁrst note that the “if” portion of the theorem is straightforward,

since, if

˜

φ = 0 in 1, then (4.15) holds trivially true for any subset { of 1. To prove the

converse, note that continuity of

˜

φ in the spatial argument x at a point x

0

∈ 1 means that

for any time t and every ε > 0, there is a δ = δ(ε) > 0, such that

[

˜

φ(x, t) −

˜

φ(x

0

, t)[ < ε , (4.16)

provided that

[x −x

0

)[ < δ(ε) . (4.17)

Now proceed by contradiction: assume that there exists a point x

0

∈ 1, such that, at any

ﬁxed time t,

˜

φ(x

0

, t) = φ

0

> 0. Then, invoking continuity of

˜

φ, there exists a δ = δ(

φ

0

2

),

such that

[

˜

φ(x, t) −

˜

φ(x

0

, t)[ = [

˜

φ(x, t) −φ

0

[ <

φ

0

2

, (4.18)

whenever

[x −x

0

)[ < δ(

φ

0

2

) . (4.19)

Now, deﬁne the region {

δ

that consists of all points of 1 for which [x − x

0

[ < δ(

φ

0

2

). This

is a sphere of radius δ in c

3

with volume vol({

δ

) =

P

δ

dv > 0. It follows from (4.18) that

˜

φ(x, t) >

φ

0

2

everywhere in P

δ

. This, in turn, implies that

P

δ

˜

φ dv >

P

δ

φ

0

2

dv =

φ

0

2

vol({

δ

) > 0 , (4.20)

which constitutes a contradiction. Therefore, the localization theorem holds.

The localization theorem can be also proved with ease for vector and tensor functions.

ME185

D

R

A

F

T

4.4. MASS AND MASS DENSITY 67

4.4 Mass and mass density

Consider a body B and take any arbitrary part o ⊆ B. Deﬁne a set function m : o → R

with the following properties:

(i) m(o) ≥ 0, for all o ⊆ B (i.e., m non-negative)

(ii) m(∅) = 0.

(iii) m(∪

∞

i=1

o

i

) =

∞

¸

i=1

m(o

i

), where o

i

⊂ B, i = 1, 2, . . . , and o

i

∩ o

j

= ∅, if i = j (i.e., m

countably additive).

A function m with the preceding properties is called a measure on B. Assume here that

there exists such a measure m and refer to m(B) as the mass of body B and m(o) as the

mass of the part o of B. In other words, consider the body as a set of particles with positive

mass.

Recall now that at time t the body B occupies a region 1 ⊂ c

3

and the part o occupies a

region {. Assuming that m is an absolutely continuous measure, it can be established that

there exists a unique function ρ = ρ(x, t), such that, for any function f =

ˇ

f(P, t) =

˜

f(x, t),

one may write

B

ˇ

f dm =

R

˜

fρ dv (4.21)

and

S

ˇ

f dm =

P

˜

fρ dv . (4.22)

The function ρ > 0 is termed the mass density. Its existence is a direct consequence of a

classical result in measure theory, known as the Radon-Nikodym theorem.

As a special case, one may consider the function f = 1, so that (4.21) and (4.22) reduce

to

B

dm =

R

ρ dv = m(B) (4.23)

and

S

dm =

P

ρ dv = m(o) . (4.24)

ME185

D

R

A

F

T

68 Mass and mass density

The mass density of a particle P occupying point x in the current conﬁguration may be

deﬁned by a limiting process as

ρ = lim

δ→0

m(o

δ

)

vol({

δ

)

, (4.25)

where {

δ

⊂ c

3

denotes a sphere of radius δ > 0 centered at x and o

δ

the part of the body

that occupies {

δ

at time t, see Figure 4.3.

{

δ

1

o

δ

B

x

δ

Figure 4.3: A limiting process used to deﬁne the mass density ρ at a point x in the current

conﬁguration.

An analogous deﬁnition of mass density can be furnished in the reference conﬁguration,

where, for any function f =

ˇ

f(P, t) =

ˆ

f(X, t),

B

ˇ

f dm =

R

0

ˆ

fρ

0

dV (4.26)

and

S

ˇ

f dm =

P

0

ˆ

fρ dV . (4.27)

Here, the mass density ρ

0

= ρ

0

(X, t) in the reference conﬁguration may be again deﬁned by

a limiting process, such that at a given point X,

ρ

0

= lim

δ→0

m(o

δ

)

vol({

0,δ

)

, (4.28)

where {

0,δ

⊂ c

3

denotes a sphere of radius δ > 0 centered at X and o

δ

the part of the body

that occupies {

0,δ

at time t

0

. Also, as in the spatial case, one may write

B

dm =

R

0

ρ

0

dV = m(B) (4.29)

and

S

dm =

P

0

ρ

0

dV = m(o) . (4.30)

The continuum hypothesis is crucial in establishing the existence of ρ and ρ

0

.

ME185

D

R

A

F

T

4.5. THE PRINCIPLE OF MASS CONSERVATION 69

4.5 The principle of mass conservation

The principle of mass conservation states that the mass of any material part of the body

remains constant at all times, namely that

d

dt

m(o) = 0 (4.31)

or, upon recalling (4.24),

d

dt

P

ρ dv = 0 . (4.32)

The preceding is an integral form of the principle of mass conservation in the spatial descrip-

tion. Using the Reynolds’ transport theorem, the above equation may be written as

P

[ ˙ ρ + ρ div v] dv = 0 . (4.33)

Assuming that the integrand in (4.33) is continuous and recalling that o (hence, also {) is

arbitrary, it follows from the localization theorem that

˙ ρ + ρ div v = 0 . (4.34)

Equation (4.34) constitutes the local form of the principle of mass conservation in the spatial

description.

1

It can be readily rewritten as

∂ρ

∂t

+

∂ρ

∂x

v + ρ div v = 0 (4.35)

hence, also as

∂ρ

∂t

+ div (ρv) = 0 . (4.36)

Example: In a volume-preserving ﬂow of a material with uniform density, conservation of

mass reduces to

∂ρ

∂t

= 0. Hence, recalling (4.14), one may write

d

dt

P

ρ dv =

P

∂ρ

∂t

dv +

∂P

ρv nda =

∂P

ρv nda . (4.37)

1

This is also sometimes referred to as the “continuity” equation, with reference to the continuity of ﬂuid

ﬂow in a control volume {.

ME185

D

R

A

F

T

70 The principles of linear and angular momentum balance

An alternative form of the mass conservation principle can be obtained by recalling

equations (4.24) and (4.30), from which it follows that

m(o) =

P

ρ dv =

P

0

ρ

0

dV . (4.38)

Recalling also (3.117), one concludes that

P

0

ρJ dV =

P

0

ρ

0

dV . (4.39)

This is an integral form of the principle of mass conservation in referential description. From

it, one ﬁnds that

P

0

(ρJ −ρ

0

) dV = 0 . (4.40)

Taking into account the arbitrariness of {

0

, the localization theorem may be invoked to yield

a local form of mass conservation in referential description as

ρ

0

= ρJ . (4.41)

4.6 The principles of linear and angular momentum

balance

Once mass conservation is established, the principles of linear and angular momentum are

postulated to describe the motion of continua. These two principles originate from the work

of Newton and Euler.

By way of background, review Newton’s three laws of motion, as postulate for particles in

1687. The ﬁrst law says that a particle stays at rest or continues at constant velocity unless

an external force acts on it. The second law says that the total external force on a particle is

proportional to the rate of change of the momentum of the particle. The third law says that

every action has an equal and opposite reaction. As Euler recognized, Newton’s three laws

of motion, while suﬃcient for the analysis of particles, are not strictly appropriate for the

study of rigid and deformable continua. Rather, he postulate a linear momentum balance

law (akin to Newton’s second law) and an angular momentum balance law. The latter can

be easily motivated from the analysis of systems of particles.

ME185

D

R

A

F

T

Basic physical principles 71

To formulate Euler’s two balance laws, ﬁrst deﬁne the linear momentum of the part of

the body that occupies the inﬁnitesimal volume element dv at time t as dmv, where dm is

the mass of dv. Also, deﬁne the angular momentum of the same part relative to the origin

of the ﬁxed basis ¦e

i

¦ as x (dmv), where x is the position vector associated with the

inﬁnitesimal volume element. Similarly, deﬁne the linear and angular momenta of the part

o which occupies a region { at time t as

S

v dm and

S

x v dm, respectively.

Next, admit the existence of two types of external forces acting on B at any time t.

There are: (a) body forces per unit mass (e.g., gravitational, magnetic) b = b(x, t) which

act on the particles which comprise the continuum, and (b) contact forces per unit area

t = t(x, t; n) = t

(n)

(x, t), which act on the particles which lie on boundary surfaces and

depend on the orientation of the surface on which they act through the outward unit normal

n to the surface. The force t

(n)

is alternatively referred to as the stress vector or the traction

vector.

The principle of linear momentum balance states that the rate of change of linear mo-

mentum for any region { occupied at time t by a part o of the body equals the total external

forces acting on this part. In mathematical terms, this means that

d

dt

S

v dm =

S

bdm+

∂P

t

(n)

da (4.42)

or, equivalently,

d

dt

P

ρv dv =

P

ρbdv +

∂P

t

(n)

da . (4.43)

Using conservation of mass, the left-hand side of the equation can be written as

d

dt

P

ρv dv =

P

d

dt

(ρv) dv +

P

(ρv) div v dv

=

P

( ˙ ρv + ρ ˙ v) dv +

P

(ρv) div v dv

=

P

[( ˙ ρ + ρ div v)v + ρ ˙ v] dv

=

P

ρa dv , (4.44)

hence, the principle of linear momentum balance can be also expressed as

P

ρa dv =

P

ρbdv +

∂P

t

(n)

da . (4.45)

ME185

D

R

A

F

T

72 The principles of linear and angular momentum balance

The principle of angular momentum balance states that the rate of change of angular

momentum for any region { occupied at t by a part o of the body equals the moment of all

external forces acting on this part. Again, this principle can be expressed mathematically as

d

dt

S

x v dm =

S

x bdm+

∂P

x t

(n)

da (4.46)

or, equivalently,

d

dt

P

x ρv dv =

P

x ρbdv +

∂P

x t

(n)

da . (4.47)

Again, appealing to conservation of mass, one may easily rewrite the term on the left-hand

side of (4.47) as

d

dt

P

x ρv dv =

P

d

dt

(x ρv) + (x ρv) div v

dv

=

P

¦[ ˙ x ρv +x ˙ ρv +x ρ ˙ v] + (x ρv div v)¦ dv

=

P

¦x ( ˙ ρ + ρ div v)v +x ρa¦ dv

=

P

x ρa dv . (4.48)

As a result, the principle of angular momentum balance may be also written as

P

x ρa dv =

P

x ρbdv +

∂P

x t

(n)

da . (4.49)

The preceding two balance laws are also referred to as Euler’s laws. They are termed

“balance” laws because they postulate that there exists a balance between external forces

(and their moments) and the rate of change of linear (and angular) momentum. Euler’s laws

are independent axioms in continuum mechanics.

In the special case where b = 0 in { and t

(n)

= 0 on ∂{, then equations (4.43) and

(4.47) readily imply that the linear and the angular momentum are conserved quantities in {.

Another commonly encountered special case is when the velocity v vanishes identically. In

this case, equations (4.43) and (4.47) imply that the sum of all external forces and the some

of all external moments vanish, which gives rise to the classical equilibrium equations of

statics.

ME185

D

R

A

F

T

4.7. STRESS VECTOR AND STRESS TENSOR 73

4.7 Stress vector and stress tensor

As in the case of mass balance, it is desirable to obtain local forms of linear and angular

momentum balance. Recalling the corresponding integral statements (4.43) and (4.47), it

is clear that the Reynolds’ transport theorem may be employed in the rate of change of

momentum terms to yield volume integrals, while the body force terms are already in the

form of volume integral. Therefore, in order to apply the localization theorem, it is essential

that the contact form terms (presently written as surface integrals) be transformed into

equivalent volume integral terms.

{

1

{

1

{

2

{

2

∂{ ∂{

′

∂{

′′

σ

σ

σ

n

1

= n n

2

Figure 4.4: Setting for a derivation of Cauchy’s lemma.

By way of background, consider some properties of the traction vector t

(n)

. To this end,

take an arbitrary region { ⊆ 1 and divide it into two mutually disjoint subregions {

1

and

{

2

by means of an arbitrary smooth surface σ, namely { = {

1

∪ {

2

and {

1

∩ {

2

= ∅, see

Figure 4.4. Also, note that the boundaries ∂{

1

and ∂{

2

of {

1

and {

2

, respectively, can be

expressed as ∂{

1

= ∂{

′

∪ σ and ∂{

2

= ∂{

′′

∪ σ. Now, enforce linear momentum balance

separately in {

1

and {

2

to ﬁnd that

d

dt

P

1

ρv dv =

P

1

ρbdv +

∂P

1

t

(n)

da (4.50)

and

d

dt

P

2

ρv dv =

P

2

ρbdv +

∂P

2

t

(n)

da (4.51)

Subsequently, add the two equations together to ﬁnd that

d

dt

P

1

∪P

2

ρv dv =

P

1

∪P

2

ρbdv +

∂P

1

∪∂P

2

t

(n)

da (4.52)

ME185

D

R

A

F

T

74 Stress vector and stress tensor

or

d

dt

P

ρv dv =

P

ρbdv +

∂P

1

∪∂P

2

t

(n)

da . (4.53)

Further, enforce linear momentum balance on the union of {

1

and {

2

to ﬁnd that

d

dt

P

ρv dv =

P

ρbdv +

∂P

t

(n)

da . (4.54)

Subtracting (4.54) from (4.53) leads to

∂P

1

∪∂P

2

t

(n)

da =

∂P

t

(n)

da . (4.55)

Recalling the decomposition of ∂{

1

and ∂{

2

, the preceding equation may be also expressed

as

∂P

′

∪σ

t

(n)

da +

∂P

′′

∪σ

t

(n)

da =

∂P

t

(n)

da (4.56)

or

∂P

′

∪∂P

′′

t

(n)

da +

σ

t

(n

1

)

da +

σ

t

(n

2

)

da =

∂P

t

(n)

da , (4.57)

so, ﬁnally,

σ

t

(n

1

)

da +

σ

t

(n

2

)

da = 0 , (4.58)

which can be also written as

σ

t

(n)

da +

σ

t

(−n)

da = 0 . (4.59)

Since σ is an arbitrary surface, assuming that t depends continuously on n and x along σ,

the localization theorem yields the condition

t

(n)

+t

(−n)

= 0 . (4.60)

This result is called Cauchy’s lemma on t

(n)

. It states that the stress vectors acting at x on

opposite sides of the same smooth surface are equal and opposite, namely that

t(x, t; n) = −t(x, t; −n) . (4.61)

It is important to recognize here that in continuum mechanics Cauchy’s lemma is not a

principle. Rather, it is derivable from linear momentum balance, as above. This is in stark

contrast with particle mechanics, where action-reaction is a principle, also known as Newton’s

Third Law.

ME185

D

R

A

F

T

Basic physical principles 75

At this stage, consider the following problem, originally conceived by Cauchy: take a

tetrahedral region { ⊂ 1, such that three edges are parallel to the axes of ¦e

i

¦ and meet

at a point x, as in Figure 4.5. Let σ

i

be the face with unit outward normal −e

i

, and σ

0

the

(inclined) face with outward unit normal n. Denote by A the area of σ

0

, so that the area

vector ndA can be resolved as

nA = (n

i

e

i

)A = An

i

e

i

= A

i

e

i

, (4.62)

where A

i

= An

i

is the area of the face σ

i

. In addition, the volume V of the tetrahedron can

be written as V =

1

3

Ah, where h is the distance of x from face σ

0

.

3

2 1

1

e

1

e

2

e

3

x

Figure 4.5: The Cauchy tetrahedron.

Now apply balance of linear momentum to the tetrahedral region { in the form of equa-

tion (4.45) and concentrate on the surface integral term, which, in this case, becomes

∂P

t

(n)

da =

σ

1

t

(−e

1

)

da +

σ

2

t

(−e

2

)

da +

σ

3

t

(−e

3

)

da +

σ

0

t

(n)

da . (4.63)

Upon using Cauchy’s lemma, this becomes

∂P

t

(n)

da = −

σ

1

t

(e

1

)

da −

σ

2

t

(e

2

)

da −

σ

3

t

(e

3

)

da +

σ

0

t

(n)

da . (4.64)

Returning now to the balance of linear momentum statement (4.45), it can be written as

P

ρ(a −b) dv =

σ

0

t

(n)

da −

σ

1

t

(e

1

)

da −

σ

2

t

(e

2

)

da −

σ

3

t

(e

3

)

da . (4.65)

Assuming that ρ, a, and b are bounded, one can obtain an upper-bound estimate for the

domain integral on the left-hand side as

P

ρ(a −b) dv

≤

P

[ρ(a −b) dv[ =

P

K(x, t) dv = K

∗

V = K

∗

1

3

Ah , (4.66)

ME185

D

R

A

F

T

76 Stress vector and stress tensor

where K(x, t) = [ρ(a −b)[ and K

∗

= K(x

∗

, t), with x

∗

being some interior point of {. The

preceding derivation made use of the mean-value theorem for integrals.

2

Assuming that t(e

i

) are continuous in x, apply the mean value theorem for integrals

component-wise to get

σ

i

t

(e

i

)

da = t

∗

i

A

i

, (4.67)

so that summing up all three like equations

3

¸

i=1

σ

i

t

(e

i

)

da = t

∗

i

A

i

= t

∗

i

An

i

. (4.68)

Also, for the inclined face

σ

0

t

(n)

da = t

∗

(n)

A . (4.69)

In the preceding two equations, t

∗

i

and t

∗

(n)

are the traction vectors at some interior points

of σ

i

and σ

0

. Recalling from (4.65) and (4.66) that

σ

0

t

(n)

da −

3

¸

i=1

σ

i

t

(e

i

)

da

≤

1

3

K

∗

Ah , (4.70)

write

σ

0

t

(n)

da −

3

¸

i=1

σ

i

t

(e

i

)

da

= [t

∗

A−t

∗

i

An

i

[ = A[t

∗

−t

∗

i

n

i

[ ≤

1

3

K

∗

Ah , (4.71)

which simpliﬁes to

[t

∗

(n)

−t

∗

i

n

i

[ ≤

1

3

K

∗

h . (4.72)

Now, upon applying the preceding analysis to a sequence of geometrically similar tetra-

hedra with heights h

1

> h

2

> . . ., where lim

i→∞

h

i

= 0, one ﬁnds that

[t

∗

(n)

−t

∗

i

n

i

[ ≤ 0 , (4.73)

where, obviously, all stress vectors are evaluated exactly at x. It follows from (4.73) that at

point x

t

(n)

= t

i

n

i

, (4.74)

2

Mean-value Theorem for Integrals: If { has positive volume (vol({) > 0) and is compact and connected,

and f is continuous in ε

3

, then there exists a point x

∗

∈ { for which

P

f(x) dv = f(x

∗

) vol({).

ME185

D

R

A

F

T

Basic physical principles 77

where the superscript has been dropped, since, now, x

∗

= x.

With the preceding development in place, deﬁne the tensor T as

T = t

i

⊗e

i

, (4.75)

so that, when operating on n,

Tn = (t

i

⊗e

i

)n = t

i

(e

i

n) = t

i

n

i

= t

(n)

, (4.76)

as seen from (4.74). The tensor T is called the Cauchy stress tensor.

It can be readily seen from (4.75) that

T = T

ki

e

k

⊗e

i

= t

i

⊗e

i

, (4.77)

hence

t

i

= T

ki

e

k

. (4.78)

Also, since

t

i

e

j

= T

ki

e

k

e

j

= T

ji

, (4.79)

it is immediately seen that

T

ij

= e

i

t

j

. (4.80)

Note that, from its deﬁnition in (4.75), it is clear that the Cauchy stress tensor does not

depend on the normal n.

Return now to the integral statement of linear momentum balance and, taking into

account (4.76), apply the divergence theorem to the boundary integral term. This leads to

P

ρa dv =

P

ρbdv +

∂P

t

(n)

da

=

P

ρbdv +

∂P

Tnda

=

P

ρbdv +

P

div Tdv . (4.81)

It follows from the preceding equation that the condition

P

(ρa −ρb −div T) dv = 0 , (4.82)

ME185

D

R

A

F

T

78 Stress vector and stress tensor

holds for an arbitrary area {, which, with the aid of the localization theorem leads to a local

form of linear momentum balance in the form

div T+ ρb = ρa . (4.83)

An alternative statement of linear momentum balance can be obtained by noting from

(4.75) that

P

ρa dv =

P

ρbdv +

∂P

t

(n)

da

=

P

ρbdv +

∂P

t

i

n

i

da

=

P

ρbdv +

P

t

i,i

dv . (4.84)

Again, appealing to the localization theorem, this leads to

t

i,i

+ ρb = ρa . (4.85)

Turn attention next to the balance of angular momentum and examine the boundary integral

term in (4.49). This can be written as

∂P

x t

(n)

da =

∂P

x t

i

n

i

da =

P

(x t

i

)

,i

dv =

P

(x

,i

t

i

+x t

i,i

) dv . (4.86)

Substituting the preceding equation into (4.49) yields

P

(x ρa) dv =

P

(x ρb) dv +

P

(x

,i

t

i

+x t

i,i

) dv (4.87)

or, upon rearranging the terms,

P

[x (ρa −ρb −t

i,i

) +x

,i

t

i

] dv = 0 . (4.88)

Recalling the local form of linear momentum balance equation (4.85), the above equation

reduces to

P

x

,i

t

i

dv = 0 . (4.89)

The localization theorem can be invoked again to conclude that

x

,i

t

i

= 0 (4.90)

ME185

D

R

A

F

T

Basic physical principles 79

or

e

i

t

i

= 0 . (4.91)

In component form, this condition can be expressed as

e

i

(T

ji

e

j

) = T

ji

e

i

e

j

= T

ji

e

ijk

e

k

= 0 , (4.92)

which means that T

ij

= T

ji

, i.e., the Cauchy stress tensor is symmetric (or, said diﬀerently,

the skew-symmetric tensor T−T

T

vanishes identically). Hence, angular momentum balance

restricts the Cauchy stress tensor to be symmetric.

e

1

e

2

e

3

T

11

T

21

T

31

T

12

T

13

T

22

T

23

T

32

T

33

Figure 4.6: Interpretation of the Cauchy stress components on an orthogonal parallelepiped

aligned with the ¦e

i

¦-axes.

An interpretation of the components of T on an orthogonal parallelepiped is shown in

Figure 4.6. Indeed, recalling (4.78), it follows that

t

1

= T

11

e

1

+ T

21

e

2

+ T

31

e

3

, (4.93)

which means that T

i1

is the i-th component of the traction that acts on the plane with

outward unit normal e

1

. More generally, T

ij

is the i-th component of the traction that acts

on the plane with outward unit normal e

j

. The components T

ij

of the Cauchy stress tensor

can be put in matrix form as

[T

ij

] =

T

11

T

12

T

13

T

21

T

22

T

23

T

31

T

32

T

33

¸

¸

¸

¸

, (4.94)

ME185

D

R

A

F

T

80 Stress vector and stress tensor

where [T

ij

] is symmetric. The linear eigenvalue problem

(T−Ti)n = 0 (4.95)

yields three real eigenvalues T

1

≥ T

2

≥ T

3

, which are solutions of the characteristic polyno-

mial equation

T

3

−I

T

T

2

+ II

T

T −III

T

= 0 , (4.96)

where I

T

, II

T

and III

T

are the three principal invariants of the tensor T, deﬁned as

I

T

= tr(T)

II

T

=

1

2

[(tr(T))

2

−tr T

2

] (4.97)

III

T

=

1

6

[(tr(T))

3

−3 tr T(tr T

2

) + 2 tr T

3

] = det T . (4.98)

As is well-known, the associated unit eigenvectors n

(1)

, n

(2)

and n

(3)

of T are mutually

orthogonal provided the eigenvalues are distinct. Also, whether the eigenvalues are distinct

or not, there exists a set of mutually orthogonal eigenvectors for T.

The traction vector t

(n)

can be generally decomposed into normal and shearing compo-

nents on the plane of its action. To this end, the normal traction (i.e., the projection of t

(n)

along n) is given by

(t

(n)

n)n = (n ⊗n)t

(n)

, (4.99)

as in Figure 4.7. Then, the shearing traction is equal to

n

t

(n)

(t

(n)

n)n

Figure 4.7: Projection of the traction to its normal and tangential components.

ME185

D

R

A

F

T

Basic physical principles 81

t

(n)

−(t

(n)

n)n = t

(n)

−(n ⊗n)t

(n)

= (i −n ⊗n)t

(n)

. (4.100)

If n is a principal direction of T, equations (4.95) and (4.100) imply that

(i −n ⊗n)t

(n)

= (i −n ⊗n)Tn = (i −n ⊗n)Tn = 0 , (4.101)

namely that the shearing traction vanishes on the plane with unit normal n.

Next, consider three special forms of the stress tensor that lead to equilibrium states in

the absence of body forces.

(a) Hydrostatic pressure

In this state, the stress vector is always pointing in the direction normal to any plane

that it is acting on, i.e.,

t

(n)

= −pn , (4.102)

where p is called the pressure. It follows from (4.76) that

T = −pi . (4.103)

(b) Pure tension along the e-axis

Without loss of generality, let e = e

1

. In this case, the traction vectors t

i

are of the

form

t

1

= Te

1

, t

2

= t

3

= 0 . (4.104)

Then, it follows from (4.78) that

T = T(e

1

⊗e

1

) = T(e ⊗e) . (4.105)

(c) Pure shear on the (e, k)-plane

Here, let e and k be two orthogonal vectors of unit magnitude and, without loss of

generality, set e

1

= e and e

2

= k. The tractions t

i

are now given by

t

1

= Te

2

, t

2

= Te

1

, t

3

= 0 . (4.106)

Appealing, again, to (4.78), it is easily seen that

T = T(e

1

⊗e

2

+e

2

⊗e

1

) = T(e ⊗k +k ⊗e) . (4.107)

ME185

D

R

A

F

T

82 Stress vector and stress tensor

It is possible to resolve the stress vector acting on a surface of the current conﬁguration

using the geometry of the reference conﬁguration. This is conceivable when, for example,

one wishes to measure the internal forces developed in the current conﬁguration per unit

area of the reference conﬁguration. To this end, start by letting df be the total force acting

on the diﬀerential area da with outward unit normal n on the surface ∂P in the current

conﬁguration, i.e.,

df = t

(n)

da . (4.108)

Also, let dA be the image of da in the reference conﬁguration under χ

−1

t

and assume that

its outward unit is N. Then, deﬁne p

(N)

to be the traction vector resulting from resolving

the force df, which acts on ∂P, on the surface ∂P

0

, namely,

df = p

(N)

dA . (4.109)

Clearly, t and p are parallel, since they are both parallel to df.

Returning to the integral statement of linear momentum balance in (4.45), note that this

can be now readily “pulled-back” to the reference conﬁguration, hence taking the form

P

0

ρ

0

a dV =

P

0

ρ

0

bdV +

∂P

0

p

(N)

dA . (4.110)

Upon applying the Cauchy tetrahedron argument to {

0

, it is immediately concluded that

p

(N)

= p

A

N

A

, (4.111)

where p

A

are the tractions developed in the current conﬁguration, but resolved on the

geometry of the reference conﬁguration on surfaces with outward unit normals E

A

. Further,

let the tensor P be deﬁned as

P = p

A

⊗E

A

, (4.112)

so that, when operating on N,

PN = (p

A

⊗E

A

)N = p

A

N

A

, (4.113)

thus, from (4.111), it is seen that

p = PN . (4.114)

ME185

D

R

A

F

T

Basic physical principles 83

The tensor P is called the ﬁrst Piola-Kirchhoﬀ stress tensor and it is naturally unsymmetric,

since it has a mixed basis, i.e.,

P = P

iA

e

i

⊗E

A

. (4.115)

It follows from (4.112) that

P = p

A

⊗E

A

= P

iA

e

i

⊗E

A

, (4.116)

which implies that

p

A

= P

iA

e

i

. (4.117)

Further, since

p

A

e

j

= P

iA

e

i

e

j

= P

jA

, (4.118)

it is clear that

P

iA

= e

i

p

A

. (4.119)

Turning attention to the integral statement (4.110), it is concluded with the aid of (4.114)

and the divergence theorem that

P

0

ρ

0

a dV =

P

0

ρ

0

bdV +

∂P

0

p

(N)

dA

=

P

0

ρ

0

bdV +

∂P

0

PNdA

=

P

0

ρ

0

bdV +

P

0

Div PdA (4.120)

which, upon using the localization theorem, results in

ρ

0

a = ρ

0

b + Div P , (4.121)

This is the local form of linear momentum balance in the referential description.

Alternatively, equation (4.113) and the divergence theorem can be invoked to show that

P

0

ρ

0

a dV =

P

0

ρ

0

bdV +

∂P

0

p

(N)

dA

=

P

0

ρ

0

bdV +

∂P

0

p

A

N

A

dA

=

P

0

ρ

0

bdV +

P

0

p

A,A

dA (4.122)

ME185

D

R

A

F

T

84 Stress vector and stress tensor

from which another version of the referential statement of linear momentum balance can be

derived in the form

ρ

0

a = ρ

0

b +p

A,A

, (4.123)

Starting from the integral form of angular momentum balance in (4.49) and pulling it

back to the reference conﬁguration, one ﬁnds that

P

0

x ρ

0

a dV =

P

0

x ρ

0

bdV +

∂P

0

x p

(N)

dA . (4.124)

Using (4.113) and the divergence theorem on the boundary term gives rise to

P

0

x ρ

0

a dV =

P

0

x ρ

0

bdV +

∂P

0

x p

A

N

A

dA

=

P

0

x ρ

0

bdV +

P

0

(x p

A

)

,A

dV . (4.125)

Expanding and appropriately rearranging the terms of the above equation leads to

P

0

[x (ρ

0

a −ρ

0

b −P

A,A

) +x

,A

p

A

] dV = 0 . (4.126)

Appealing to the local form of linear momentum balance in (4.123) and, subsequently, the

localization theorem, one concludes that

x

,A

p

A

= 0 . (4.127)

With the aid of (4.117) and the chain rule, the preceding equation can be rewritten as

x

,A

p

A

= F

iA

P

jA

e

i

e

j

= F

iA

P

jA

e

ijk

e

k

= 0 , (4.128)

which implies that FP

T

= PF

T

. This is a local form of angular momentum balance in the

referential description.

Recalling (4.108) and (4.108), one may conclude with the aid of (4.76), (4.114) and

Nanson’s formula (3.122) that

PNdA = Tnda = TJF

−T

NdA , (4.129)

so that

T =

1

J

PF

T

. (4.130)

ME185

D

R

A

F

T

Basic physical principles 85

Clearly, the above relation is consistent with the referential and spatial statements of an-

gular momentum balance, namely (4.130) can be used to derive the local form of angular

momentum balance in spatial form from the referential statement and vice-versa. Likewise,

it is possible to derive the local linear momentum balance statement in the referential (resp.

spatial) form from its corresponding spatial (resp. referential) counterpart.

Note that there is no approximation or any other source of error associated with the

use the balance laws in the referential vs. spatial description. Indeed, the invertibility of

the motion at any ﬁxed time implies that both forms of the balance laws are completely

equivalent.

Other stress tensors beyond the Cauchy and ﬁrst Piola-Kirchhoﬀ tensors are frequently

used. Among them, the most important are the nominal stress tensor Π deﬁned as the

transpose of the ﬁrst Piola-Kirchhoﬀ stress, i.e.,

Π = P

T

= JF

−1

T , (4.131)

the Kirchhoﬀ stress tensor τ, deﬁned as

τ = JT = PF

T

, (4.132)

and the second Piola-Kirchhoﬀ stress S, deﬁned as

S = F

−1

P = JF

−1

TF

−T

. (4.133)

It is clear from (4.133) that S has both “legs” in the reference conﬁguration and is symmetric.

All the stress measures deﬁned here are obtainable from one another assuming that the

deformation is known.

4.8 The transformation of mechanical ﬁelds under su-

perposed rigid-body motions

In this section, the transformation under superposed rigid motions is considered for mechan-

ical ﬁelds, such as the stress vectors and tensors. To this end, start with the stress vector

t = t(x, t; n), and recalling the general form of the superposed rigid motion in (3.166), write

ME185

D

R

A

F

T

86 Mechanical ﬁelds under superposed rigid motions

t

+

= t

+

(x

+

, t; n

+

). To argue how t and t

+

are related, ﬁrst recall that n

+

= Qn and that

t is linear in n, as established in (4.76). Since the two motions give rise to the same defor-

mation, it is then reasonable to assume that, under a superposed rigid motion, t

+

will not

change in magnitude relative to t and will have the same orientation relative to n

+

as t has

relative to n. Therefore, it is postulated that

t

+

= Qt . (4.134)

The above transformation indeed implies that [t

+

[ = [t[ and t

+

n

+

= t n.

Unlike the transformation of kinematic terms, which is governed purely by geometry, in

the case of kinetic terms (both mechanical and thermal) the transformation is governed by

the principle of invariance under superposed rigid motion. This, eﬀectively, states that the

balance laws are invariant under superposed rigid motions, in the sense that their mathemat-

ical representation remains unchanged under such motions. To demonstrate this principle,

consider the transformation of the Cauchy stress tensor. Taking taking into account (4.76),

invariance under superposed rigid motions implies that t

+

= T

+

n

+

. This means that if two

motions diﬀer by a mere rigid motion, then the relation between the stress vector t on a

plane with an outward unit normal n and the Cauchy stress tensor T is the same with the

relation of the stress vector t

+

on a plane with outward unit normal n

+

with the Cauchy

stress tensor T

+

. Invariance of (4.76) under superposed rigid motions in conjunction with

(4.134) implies that

t

+

= Qt = QTn

= T

+

n

+

= T

+

Qn ,

(4.135)

from where it is concluded that

(QT−T

+

Q)n = 0 , (4.136)

hence

T

+

= QTQ

T

. (4.137)

Equation (4.137) implies that T is an objective spatial tensor.

Recall next the relation between the Cauchy and the ﬁrst Piola-Kirchhoﬀ stress tensor in

(4.130). Given that this relation is itself invariant under superposed rigid motions it follows

ME185

D

R

A

F

T

Basic physical principles 87

that

P

+

= J

+

T

+

(F

−T

)

+

= J(QTQ

T

)(QF

−T

) = Q(JTF

−T

) = QP , (4.138)

where the kinematic transformations (3.165) and (3.187) are employed.

3

Equation (4.138)

implies that P is an objective two-point tensor. Similarly, turning to the second Piola-

Kirchhoﬀ stress tensor, it follows from (4.133) that

S

+

= J

+

(F

−1

)

+

T

+

(F

−T

)

+

= J(F

−1

Q

T

)(QTQ

T

)(QF

−T

)

= JF

−1

TF

−T

= S , (4.139)

which implies that S is an objective referential tensor.

Since (4.139) holds true, it follows also that

˙

S

+

=

˙

S , (4.140)

namely that the rate of S is objective. However, from (4.137) and the relation (3.180) it can

be seen that

˙

T

+

=

˙

QTQ

T

+Q

˙

TQ

T

+QT

˙

Q

T

= (ΩQ)TQ

T

+Q

˙

TQ

T

+QT(ΩQ)

T

= Ω(QTQ

T

) +Q

˙

TQ

T

+ (QTQ)

T

Ω

T

= ΩT

+

+Q

˙

TQ

T

−T

+

Ω , (4.141)

which shows that, unlike T, the spatial tensor

˙

T is not objective. A similar conclusion may

be drawn for the rate

˙

P of the ﬁrst Piola-Kirchhoﬀ stress tensor.

Invariance under superposed rigid motions applies to the principle of mass conservation.

Indeed, using the local referential form (4.41) of this principle gives rise to

ρ

0

= ρ

+

J

+

= ρ

+

J

= ρJ ,

(4.142)

which imply that

ρ

+

= ρ . (4.143)

3

Also, note that, by its deﬁnition, (F

−1

)

+

= (F

+

)

−1

.

ME185

D

R

A

F

T

88 Mechanical energy balance theorem

Finally, admitting invariance under superposed rigid motions of the local spatial form (4.83)

of linear momentum balance leads to

div

+

T

+

+ ρ

+

b

+

= ρ

+

a

+

. (4.144)

Resorting to components, note that

∂T

+

ij

∂x

+

j

=

∂(Q

ik

T

kl

Q

jl

)

∂x

m

∂χ

m

∂x

+

j

= Q

ik

∂T

kl

∂x

m

Q

jl

Q

jm

= Q

ik

∂T

kl

∂x

m

δ

lm

= Q

ik

∂T

kl

∂x

l

, (4.145)

where it is recognized from (3.166) that

∂χ

∂x

+

= Q

T

, therefore, in components,

∂x

m

∂x

+

j

= Q

jm

.

Equation (4.145) can be written using direct notation as

div

+

T

+

= Qdiv T . (4.146)

Using (4.83), (4.144) and (4.146), one concludes that

div

+

T

+

= ρ

+

(a

+

−b

+

) = ρ(a

+

−b

+

)

= Qdiv T = Qρ(a −b) ,

from where it follows that

a

+

−b

+

= Q(a −b) . (4.147)

This means that, under superposed rigid motions, the body forces transform as

b

+

= Qb +a

+

−Qa . (4.148)

4.9 The Theorem of Mechanical Energy Balance

Consider again the body B in the current conﬁguration 1 and take an arbitrary material

region { with smooth boundary ∂{.

ME185

D

R

A

F

T

Basic physical principles 89

Recalling the integral statement of linear momentum balance (4.43), deﬁne the rate at

which the external body forces b and surface tractions t do work in { and on ∂{, respectively,

as

R

b

({) =

P

ρb v dv (4.149)

and

R

c

({) =

∂P

t v da , (4.150)

respectively. Also, deﬁne the rate of work done by all external forces as

R({) = R

b

({) + R

c

({) . (4.151)

In addition, deﬁne the total kinetic energy of the material points contained in { as

K({) =

P

1

2

v vρ dv . (4.152)

Starting from the local spatial form of linear momentum balance (4.83), one may dot both

sides with the velocity v to obtain

ρ ˙ v v = ρb v + div T v . (4.153)

Now, note that

(div T) v = div(T

T

v) −T gradv

= div(Tv) −T (D+W)

= div(Tv) −T D , (4.154)

where angular momentum balance is invoked in the form of the symmetry of T. Equation

(4.154) can be used to rewrite (4.153) as

1

2

ρ

d

dt

(v v) = ρb v + div(Tv) −T D . (4.155)

Integrating (4.155) over { leads to

P

1

2

ρ

d

dt

(v v) dv =

P

ρb v dv +

P

div(Tv) dv −

P

T Ddv . (4.156)

or, upon using conservation of mass and the divergence theorem,

d

dt

P

1

2

ρ(v v) dv +

P

T Ddv =

P

ρb v dv +

∂P

(Tv) nda . (4.157)

ME185

D

R

A

F

T

90 Mechanical energy balance theorem

Recalling (4.76), the preceding equation can be rewritten as

d

dt

P

1

2

ρ(v v) dv +

P

T Ddv =

P

ρb v dv +

∂P

t v da . (4.158)

The second term on the right-hand side of (4.158),

S({) =

P

T Ddv , (4.159)

is called the stress power and it represents the rate at which the stresses do work in {.

Taking into account (4.149), (4.150), (4.152), (4.159) and (4.158), it is seen that

d

dt

K({) + S({) = R

b

({) + R

c

({) = R({) . (4.160)

Equation (4.160) (or, equivalently, equation (4.158)) states that, for any region {, the rate

of change of the kinetic energy and the stress power are balanced by the rate of work done

by the external forces. This is a statement of the mechanical energy balance theorem. It

is important to emphasize that mechanical energy balance is derivable from the three basic

principles of the mechanical theory, namely conservation of mass and balance of linear and

angular momentum.

Returning to the stress power term S({), note that

P

T Ddv =

P

T Ldv

=

P

1

J

PF

T

Ldv

=

P

0

(PF

T

) LdV

=

P

0

tr[(PF

T

)L

T

] dV

=

P

0

tr[P(LF)

T

] dV

=

P

0

tr[P

˙

F

T

]dV

=

P

0

P

˙

FdV , (4.161)

ME185

D

R

A

F

T

Basic physical principles 91

where use is made of (4.130) and (3.128). Further,

P

T Ddv =

P

0

P

˙

FdV

=

P

0

(FS)

˙

FdV

=

P

0

S (F

T

˙

F) dV

=

P

0

S

1

2

(

˙

F

T

F +F

T

˙

F) dV

=

P

0

S

˙

EdV , (4.162)

where, this time use is made of (4.133) and the deﬁnition of the Lagrangian strain.

Equations (4.161) and (4.162) reveal that P is the work-conjugate kinetic measure to F

in {

0

and, likewise, S is work-conjugate to E. These equations appear to leave open the

question of work-conjugacy for T.

4.10 The principle of energy balance

The physical principles postulated up to this point are incapable of modeling the intercon-

vertibility of mechanical work and heat. In order to account for this class of (generally

coupled) thermomechanical phenomena, one needs to introduce an additional primitive law

known as balance of energy.

Preliminary to stating the balance of energy, deﬁne a scalar ﬁeld r = r(x, t) called the

heat supply per unit mass (or speciﬁc heat supply), which quantiﬁes the rate at which heat

is supplied (or absorbed) by the body. Also, deﬁne a scalar ﬁeld h = h(x, t; n) = h

(n)

(x, t)

called the heat ﬂux per unit area across a surface ∂{ with outward unit normal n. Now,

given any region { ⊆ 1, deﬁne the total rate of heating H({) as

H({) =

P

ρr dv −

∂P

h da , (4.163)

where the negative sign in front of the boundary integral signiﬁes the fact that the heat ﬂux

is assumed positive when it exits the region {.

Next, assume the existence of a scalar function ε = ε(x, t) per unit mass, called the

(speciﬁc) internal energy. This function quantiﬁes all forms of energy stored in the body

ME185

D

R

A

F

T

92 The principle of energy balance

other than the kinetic energy. Examples of stored energy include strain energy (i.e., energy

due to deformation), chemical energy, etc. The internal energy stored in { is denoted by

U({) =

P

ρε dv . (4.164)

The principle of balance of energy is postulated in the form

d

dt

P

¸

1

2

ρv v + ρε

dv =

P

ρb v dv +

∂P

t v da +

P

ρr dv −

∂P

h da . (4.165)

This is also sometimes referred to as a statement of the “First Law of Thermodynamics”.

Equivalently, equation (4.165) can be written as

d

dt

[K({) + U({)] = R({) + H({) . (4.166)

Subtracting (4.158) from (4.165) leads to a statement of thermal energy balance in the

form

d

dt

P

ρε dv =

P

T Ddv +

P

ρr dv −

∂P

h da . (4.167)

Returning to the heat ﬂux h = h(x, t; n), and note that one can apply a standard ar-

gument to ﬁnd the exact dependence of h on n, as already done with the stress vector

t = t(x, t; n). Indeed, one may apply thermal energy balance to a region { with boundary

∂{ and to each of two regions {

1

and {

2

with boundaries ∂{

1

and ∂{

2

, where {

1

∪{

2

= {

and {

1

∪ {

2

= ∅. Also, the boundaries ∂{

1

= ∂{

′

∪ σ, ∂{

2

= ∂{

′′

∪ σ have a common

surface σ and ∂{

′

∪ ∂{

′′

= ∂{. It follows that

d

dt

P

ρε dv =

P

T Ddv +

P

ρr dv −

∂P

h da . (4.168)

and, also,

d

dt

P

1

ρε dv =

P

1

T Ddv +

P

1

ρr dv −

∂P

1

h da (4.169)

and

d

dt

P

2

ρε dv =

P

2

T Ddv +

P

2

ρr dv −

∂P

2

h da (4.170)

Adding the last two equations leads to

d

dt

P

1

∪P

2

ρε dv =

P

1

∪P

2

T Ddv +

P

1

∪P

2

ρr dv −

∂P

1

∪∂P

2

h da (4.171)

ME185

D

R

A

F

T

Basic physical principles 93

or, equivalently,

d

dt

P

ρε dv =

P

T Ddv +

P

ρr dv −

∂P

1

∪∂P

2

h da . (4.172)

Subtracting (4.168) from (4.172) results in

∂P

1

∪∂P

2

h da −

∂P

h da = 0 , (4.173)

or, equivalently,

∂P

′

∪σ

h da +

∂P

′′

∪σ

h da =

∂P

h da . (4.174)

As in the case of the stress vector, the preceding equation may be expanded to

∂P

′

∪∂P

′′

h da +

σ

h

(n

1

)

da +

σ

h

(n

2

)

da =

∂P

h da (4.175)

or

σ

(h

(n)

−h

(−n)

)da = 0 , (4.176)

where n

1

= n and n

2

= −n. Since σ is an arbitrary surface and h is assumed to depend

continuously on n and x along σ, the localization theorem yields the condition

h(n) = −h(−n) . (4.177)

or, more explicitly,

h(x, t; n) = −h(x, t; −n) . (4.178)

This states that the ﬂux of heat exiting a body across a surface with outward unit normal

n at a point x is equal to the ﬂux of heat entering a neighboring body at the same point

across the same surface.

Using the tetrahedron argument in connection with the thermal energy balance equation

(4.167) and the ﬂux continuity equation (4.178), gives rise to

h = h

i

n

i

(4.179)

where h

i

are the ﬂuxes across the faces of the tetrahedron with outward unit normals e

i

.

Thus, one may write

h = q n , (4.180)

ME185

D

R

A

F

T

94 The principle of energy balance

where q is the heat ﬂux vector with components q

i

= h

i

.

Now, returning to the integral statement of energy balance in (4.165), use mass conser-

vation to rewrite it as

P

(ρv ˙ v + ρ ˙ ε) dv =

P

ρb v dv +

∂P

t v da +

P

ρr dv −

∂P

h da . (4.181)

Using (4.76) and (4.180), the above equation may be put in the form

P

(ρv ˙ v + ρ ˙ ε) dv =

P

ρb v dv +

∂P

(Tn) v da +

P

ρr dv −

∂P

q nda . (4.182)

Upon invoking the divergence theorem, it is easily seen that

∂P

(Tn) v da =

P

[(div T) v +T D] dv (4.183)

and

∂P

q nda =

P

div qdv . (4.184)

If the last two equations are substituted in (4.182), one ﬁnds that

P

¦ρ ˙ v −ρb −div T¦ v + ρ ˙ ε −T D−ρr + div q

dv = 0 . (4.185)

Upon observing linear momentum balance and invoking the localization theorem, the pre-

ceding equation gives rise to the local form of energy balance as

ρ ˙ ε = T D+ ρr −div q . (4.186)

This equation could be also derived along the same lines from the integral statement of

thermal energy balance (4.167).

4

Regarding invariance under superposed rigid-body motions, it is postulated that

ε

+

= ε (4.187)

and

r

+

= r , q

+

= Qq . (4.188)

4

The energy equation is frequently quoted in elementary thermodynamics texts as “dU = δQ + δW”,

where dU corresponds to ρ ˙ ε, δQ to ρr −div q, and δW to T D.

ME185

D

R

A

F

T

Basic physical principles 95

Equation (4.188)

2

and invariance of the thermal energy balance imply that

h

+

= q

+

n

+

= (Qq) (Qn) = q n = h . (4.189)

Example: Consider a rigid heat conductor, where rigidity implies that F = R (i.e., U = I).

This means that

˙

F =

˙

R = ΩR (Ω =

˙

RR

T

)

= LF = LR ,

(4.190)

which implies that L = Ω, so that D = 0. Further, assume that Fourier’s law holds, namely

that

q = −k grad T , (4.191)

where T is the empirical temperature and k > 0 is the (isotropic) conductivity. These

conditions imply that the balance of energy (4.186) reduces to

ρ ˙ ε −div(k gradT) −ρr = 0 . (4.192)

Further, assume that the internal energy depends exclusively on T and that this dependence

is linear, hence

dε

dt

= c, where c is termed the heat capacity. It follows from (4.192) that

ρc

˙

T = div(k gradT) + ρr , (4.193)

which is the classical equation of transient heat conduction.

4.11 The Green-Naghdi-Rivlin theorem

Assume that the principle of energy balance remains invariant under superposed rigid mo-

tions. With reference to (4.165), this means that

d

dt

P

+

ρ

+

ε

+

+

1

2

ρ

+

v

+

v

+

dv

+

=

P

+

ρ

+

b

+

v

+

dv

+

+

∂P

+

t

+

v

+

da

+

+

P

+

ρ

+

r

+

dv

+

−

∂P

+

h

+

da

+

. (4.194)

Now, choose a special superposed rigid motion, which is a pure rigid translation, such

that

Q = I ,

˙

Q = 0 , c = c

0

t , (4.195)

ME185

D

R

A

F

T

96 The Green-Naghdi-Rivlin theorem

where c

0

is a constant vector in E

3

. It follows immediately from (3.178) and (4.195) that

v

+

= v +c

0

, ˙ v

+

= ˙ v . (4.196)

Moreover, it is readily concluded from (4.148), (4.134), (4.195) and (4.196) that under this

superposed rigid translation

b

+

= b , t

+

= t . (4.197)

It follows from (4.143), (4.196) and (4.197) that (4.194) takes the form

d

dt

P

ρε +

1

2

ρ(v +c

0

) (v +c

0

)

dv

=

P

ρb (v +c

0

) dv +

∂P

t (v +c

0

) da +

P

ρr dv −

∂P

h da . (4.198)

Upon subtracting (4.165) from (4.198), it is concluded that

c

0

d

dt

P

ρv dv −

P

ρbdv −

∂P

t da

+

1

2

(c

0

c

0

)

d

dt

P

ρ dv

= 0 . (4.199)

Since c

0

is an arbitrary constant vector, one may set c

0

= γ¯ c

0

, where γ is an arbitrary

non-vanishing scalar constant. Substituting γ with −γ in (4.199), it is proved that

d

dt

P

ρ dv = 0 (4.200)

hence, also,

d

dt

P

ρv dv =

P

ρbdv +

∂P

t da . (4.201)

This, in turn, means that conservation of mass and balance of linear momentum hold, re-

spectively.

Next, a second special superposed rigid-body motion is chosen, such that

Q = I ,

˙

Q = Ω

0

, c = 0 , (4.202)

where Ω

0

is a constant skew-symmetric tensor. Given (4.202), it can be easily seen from

(3.166) and (3.178) that

v

+

= v +Ω

0

x = v +ω

0

x (4.203)

and

˙ v

+

= ˙ v + 2Ω

0

v +Ω

2

0

x = ˙ v +ω

0

(2v +ω

0

x) , (4.204)

ME185

D

R

A

F

T

Basic physical principles 97

where ω

0

is the (constant) axial vector of Ω

0

. Equations (4.203) and (4.204) imply that

the superposed motion is a rigid rotation with constant angular velocity ω

0

on the original

current conﬁguration of the continuum. Taking into account (4.134) (4.148), (4.202) and

(4.204), it is established that in this case

b

+

= b + 2Ω

0

v +Ω

2

0

x = b +ω

0

(2v +ω

0

x) (4.205)

and

t

+

= t . (4.206)

In addition, equations (4.203)

1

and (4.205)

1

lead to

v

+

v

+

= v v + 2(Ω

0

x) v + (Ω

0

x) (Ω

0

x) (4.207)

and

b

+

v

+

= b v +b (Ωx) + 2(Ω

0

v) (Ωx) + 2(Ω

0

v) v

+ (Ω

2

0

x) (Ω

0

x) + (Ω

2

0

x) v

= b v +b (Ωx) + (Ω

0

v) (Ωx) + (Ω

2

0

x) (Ω

0

x) , (4.208)

where the readily veriﬁable identities

(Ω

0

v) v = 0 , (Ω

0

v) (Ω

0

x) + (Ω

2

0

x) v = 0 (4.209)

are employed. Similarly, using (4.203)

1

and (4.205)

1

, it is seen that

1

2

d

dt

(v

+

v

+

) = ˙ v v + ˙ v (Ω

0

x) + 2(Ω

0

v) (Ω

0

x) + 2(Ω

0

v) v

+ (Ω

2

0

x) (Ω

0

x) + (Ω

2

0

x) v

= ˙ v v + ˙ v (Ω

0

x) + (Ω

0

v) (Ω

0

x) + (Ω

2

0

x) (Ω

0

x) . (4.210)

Invoking now invariance of the energy equation under the superposed rigid rotation, it

can be concluded from (4.194), as well as (4.207) (4.208) and (4.210), that

d

dt

P

ρε dv +

P

ρ

˙ v v + ˙ v (Ω

0

x) + (Ω

0

v) (Ω

0

x) + (Ω

2

0

x) (Ω

0

x)

dv

=

P

ρ

b v +b (Ω

0

x) + (Ω

0

v) (Ω

0

x) + (Ω

2

0

x) (Ω

0

x)

dv

+

∂P

t (v +Ω

0

x) da +

P

ρr dv −

∂P

h da . (4.211)

ME185

D

R

A

F

T

98 The Green-Naghdi-Rivlin theorem

After subtracting (4.165) from (4.211) and simplifying the resulting equation, it follows that

P

ρ ˙ v (Ω

0

x) dv =

P

ρb (Ω

0

x) dv +

∂P

t (Ω

0

x) da . (4.212)

Observing that for any given vector z in E

3

z (Ω

0

x) = z (ω

0

x) = ω

0

(x z) , (4.213)

and recalling that ω

0

is a constant vector, equation (4.212) takes the form

ω

0

P

ρx ˙ v dv −

P

ρx bdv −

∂P

x t da

. (4.214)

This, in turn, implies that

d

dt

P

ρx v dv =

P

ρx bdv +

∂P

x t da , (4.215)

namely that balance of angular momentum holds.

The preceding analysis shows that the integral forms of conservation of mass, and balance

of linear and angular momentum are directly deduced from the integral form of energy bal-

ance and the postulate of invariance under superposed rigid-body motions. This remarkable

result is referred to as the Green-Naghdi-Rivlin theorem.

The Green-Naghdi-Rivlin theorem can be viewed as an implication of the general covari-

ance principle proposed by Einstein. According to this principle, all physical laws should

be invariant under any smooth time-dependent coordinate transformation (including, as a

special case, rigid time-dependent transformations). This far-reaching principle stems from

Einstein’s conviction that physical laws are oblivious to speciﬁc coordinate systems, hence

should be expressed in a covariant manner, i.e., without being restricted by speciﬁc choices

of coordinate systems. In covariant theories, the energy equation plays a central role, as

demonstrated by the Green-Naghdi-Rivlin theorem.

ME185

D

R

A

F

T

Chapter 5

Inﬁnitesimal deformations

The development of kinematics and kinetics presented up to this point does not require any

assumptions on the magnitude of the various measures of deformation. In many realistic

circumstances, solids and ﬂuids may undergo “small” (or “inﬁnitesimal”) deformations. In

these cases, the mathematical representation of kinematic quantities and the associated

kinetic quantities, as well as the balance laws, may be simpliﬁed substantially.

In this chapter, the special case of inﬁnitesimal deformations is discussed in detail. Pre-

liminary to this discussion, it is instructive to formally deﬁne the meaning of “small” or

“inﬁnitesimal” changes of a function. To this end, consider a scalar-valued real function

f = f(x), which is assumed to possess continuous derivatives up to any desirable order. To

analyze this function in the neighborhood of x = x

0

, one may use a Taylor series expansion

at x

0

in the form

f(x

0

+ v) = f(x

0

) + vf

′

(x

0

) +

v

2

2!

f

′′

(x

0

) +

v

3

3!

f

′′′

(x

0

) + . . .

= f(x

0

) + vf

′

(x

0

) +

v

2

2!

f

′′

(¯ x) ,

(5.1)

where v is a change to the value of x

0

and ¯ x ∈ (x

0

, x

0

+v). Denoting by ε the magnitude of

the diﬀerence between x

0

+ v and x

0

, i.e., ε = [v[, it follows that as ε → 0 (i.e., as v → 0),

the scalar f(x

0

+ v) is satisfactorily approximated by the linear part of the Taylor series

expansions in (5.1), namely

f(x

0

+ v)

.

= f(x

0

) + vf

′

(x

0

) . (5.2)

99

D

R

A

F

T

100 The Gˆateaux diﬀerential

Recalling the expansion (5.1)

2

, one says that ε = [v[ is “small”, when the term

v

2

2!

f

′′

(¯ x)

can be neglected in this expansion without appreciable error, i.e. when

v

2

2!

f

′′

(¯ x)

≪ [f(x

0

+ v)[ , (5.3)

assuming that f(x

0

+ v) = 0.

5.1 The Gˆateaux diﬀerential

Linear expansions of the form (5.2) can be readily obtained for vector or tensor functions

using the Gˆateaux diﬀerential. In particular, given F = F(X), where F is a scalar, vector or

tensor function of a scalar, vector or tensor variable X, the Gˆateaux diﬀerential DF(X

0

, V)

of F at X = X

0

in the direction V is deﬁned as

DF(X

0

, V) =

¸

d

dω

F(X

0

+ ωV)

w=0

, (5.4)

where ω is a scalar. Then,

F(X

0

+V) = F(X

0

) + DF(X

0

, V) + o([V[

2

) , (5.5)

such that

lim

|V|→0

o([V[

2

)

[V[

= 0 . (5.6)

The linear part L[F; V]

X

0

of F at X

0

in the direction V is then deﬁned as

L[F; V]

X

0

= F(X

0

) + DF(X

0

, V) . (5.7)

Examples:

(1) Let F(X) = f(x) = x

2

. Using the deﬁnition in (5.4),

Df(x

0

, v) =

¸

d

dω

f(x

0

+ ωv)

ω=0

=

¸

d

dω

(x

0

+ ωv)

2

ω=0

=

¸

d

dω

(x

2

0

+ 2x

0

ωv + ω

2

v

2

)

ω=0

=

2x

0

v + 2ωv

2

ω=0

= 2x

0

v .

ME185

D

R

A

F

T

Inﬁnitesimal deformations 101

Hence,

L[f; v]

x

0

= x

2

0

+ 2x

0

v .

**(2) Let F(X) = T(x) = x ⊗x. Using, again, the deﬁnition in (5.4),
**

DT(x

0

, v) =

¸

d

dω

T(x

0

+ ωv)

ω=0

=

¸

d

dω

¦(x

0

+ ωv) ⊗(x

0

+ ωv)¦

ω=0

=

¸

d

dω

¦x

0

⊗x

0

+ ω(x

0

⊗v +v ⊗x

0

) + ω

2

v ⊗v¦

ω=0

= [(x

0

⊗v +v ⊗x

0

) + 2ωv ⊗v]

ω=0

= x

0

⊗v +v ⊗x

0

.

It follows that

L[T; v]

x

0

= x

0

⊗x

0

+x

0

⊗v +v ⊗x

0

.

**5.2 Consistent linearization of kinematic and kinetic
**

variables

Start by noting that the position vector x of a material point P in the current conﬁguration

can be written as the sum of the position vector X of the same point in the reference conﬁg-

uration plus the displacement u of the point from the reference to the current conﬁguration,

namely

x = X+u . (5.8)

As usual, the displacement vector ﬁeld can be expressed equivalently in referential or spatial

form as

u = ˆ u(X, t) = ˜ u(x, t) . (5.9)

ME185

D

R

A

F

T

102 Consistent linearization of kinematic and kinetic variables

It follows that the deformation gradient can be written as

F =

∂χ

∂X

=

∂(X+ ˆ u)

∂X

= I +

∂ˆ u

∂X

= I +H , (5.10)

where H is the relative displacement gradient tensor deﬁned by

H =

∂ˆ u

∂X

. (5.11)

Clearly, H quantiﬁes the deviation of F from the identity tensor.

Recalling the discussion in Section 5.1, a linearized counterpart of a given kinematic

measure is obtained by ﬁrst expressing the kinematic measure as a function of H, i.e., as

F(H) and, then, by expanding F(H) around the reference conﬁguration, where H = 0. This

leads to

F(H) = F(0) + DF(0, H) + o(|H|

2

) , (5.12)

where |H| is the two-norm of H deﬁned at a point X and time t as |H| = (H H)

1/2

.

Taking into account (5.7) and (5.12), the linearized counterpart of F(H) is the linear part

L(F; H)

0

of F around the reference conﬁguration in the direction of H, given by

L(F; H)

0

= F(0) + DF(0, H) . (5.13)

Here, a suitable scalar measure of magnitude for the deviation of F from identity can be

deﬁned as

ε = ε(t) = sup

X∈R

0

[[H(X, t)[[ , (5.14)

where “sup” denotes the least upper bound over all points X in the reference conﬁguration.

Now one may say that the deformations are “small” (of “inﬁnitesimal”) at a given time t if

ε is small enough so that the term o([[H[[

2

) can be neglected when compared with F(H).

Now, proceed to obtain inﬁnitesimal counterparts of some standard kinematic ﬁelds,

starting with the deformation gradient F. To this end, ﬁrst recall that F =

¯

F(H) = I +H.

Then, note that the Gˆateaux diﬀerential

DF(0, H) =

¸

d

dω

¯

F(0 + ωH)

ω=0

=

¸

d

dω

(I + ωH)

ω=0

= H . (5.15)

ME185

D

R

A

F

T

Inﬁnitesimal deformations 103

Hence,

L[F; H]

0

=

¯

F(0) + DF(0, H)

= I +H . (5.16)

Eﬀectively, (5.16) shows that the linear part of F is F itself, which should be obvious from

(5.10).

Next, starting from FF

−1

= I, take the linear part of both sides in the direction of H.

This leads to

L[FF

−1

; H]

0

=

¯

F(0)

¯

F

−1

(0) + D(FF

−1

)(0, H) = L[I; H]

0

= I , (5.17)

where

D(FF

−1

)(0, H) = DF(0, H)

¯

F

−1

(0) +

¯

F(0)DF

−1

(0, H)

= H+ DF

−1

(0, H) = 0 .

(5.18)

The preceding equation implies that

DF

−1

(0, H) = −H . (5.19)

Hence, the linear part of F

−1

at H = 0 in the direction H is

L[F

−1

; H]

0

= I −H . (5.20)

Next, consider the linear part of the spatial displacement gradient grad u. First, observe

that, using the chain rule, one ﬁnds that

gradu = (Gradu)F

−1

= (F −I)F

−1

= I −F

−1

, (5.21)

therefore

gradu = gradu(H) = I −(I +H)

−1

. (5.22)

This means that, taking into account (5.19),

D(gradu)(0, H) = −DF

−1

(0, H) = H . (5.23)

As a result,

L[gradu; H]

0

= gradu(0) + D(gradu)(0, H) = 0 +H = H . (5.24)

ME185

D

R

A

F

T

104 Consistent linearization of kinematic and kinetic variables

The last result shows that the linear part of the spatial displacement gradient gradu coincides

with the referential displacement gradient Gradu(= H). This, in turn, implies that, within

the context of inﬁnitesimal deformations, there is no diﬀerence between the partial derivatives

of the displacement u with respect to X or x. This further implies that the distinction

between spatial and referential description of kinematic quantities becomes immaterial in

the case of inﬁnitesimal deformations.

To determine the linear part of the right Cauchy-Green deformation tensor C, write

C =

¯

C(H) = (I +H)

T

(I +H) = I +H+H

T

+H

T

H . (5.25)

Then,

DC(0, H) =

¸

d

dw

¯

C(0 + wH)

w=0

=

¸

d

dw

¸

I + w(H+H

T

) + w

2

H

T

H

¸

w=0

=

H+H

T

+ 2wH

T

H

w=0

= H+H

T

. (5.26)

Consequently, the linear part of C at H = 0 in the direction H is

L[C; H]

0

=

¯

C(0) + DC(0, H) = I + (H+H

T

) . (5.27)

Using (5.27), it can be immediately concluded that the linear part of the Lagrangian

strain tensor E is

L [E; H]

0

=

1

2

(H+H

T

) . (5.28)

At the same time, the Eulerian strain tensor e can be written as

e = ¯ e(H) =

1

2

i −

¯

F

−T

(H)

¯

F

−1

(H)

, (5.29)

hence its Gˆateaux diﬀerential is given, with the aid of (5.19), by

De(0, H) = −

1

2

DF

−T

(0, H) + DF

−1

(0, H)

=

1

2

(H+H

T

) . (5.30)

This means that the linear part of e is equal to

L[e; H]

0

= ¯ e(0) + De(0, H) =

1

2

(H+H

T

) . (5.31)

ME185

D

R

A

F

T

Inﬁnitesimal deformations 105

It is clear from (5.28) and (5.31) that the linear parts of the Lagrangian and Eulerian strain

tensors coincide. Hence, under the assumption of inﬁnitesimal deformations, the distinction

between the two strains ceases to exist and one writes that

L[E; H]

0

= L[e; H]

0

= ε , (5.32)

where ε is the classical inﬁnitesimal strain tensor, with components ε

ij

=

1

2

(u

i,j

+ u

j,i

).

Proceed next with the linearization of the right stretch tensor U. To this end, write

U

2

= C = I +H+H

T

+H

T

H , (5.33)

so that, with the aid of (5.26),

DU

2

(0, H) = DU(0, H)

¯

U(0) +

¯

U(0)DU(0, H) = 2DU(0, H)

= H+H

T

,

(5.34)

hence,

L[U; H]

0

=

¯

U(0) + DU(0, H) = I +

1

2

(H+H

T

) . (5.35)

Repeating the procedure used earlier in this section to determine the Gˆateaux diﬀerential

of F

−1

, one easily ﬁnds that

DU

−1

(0, H) = −

1

2

(H+H

T

) (5.36)

and

L

U

−1

; H

0

= I −

1

2

(H+H

T

) . (5.37)

It is now possible to determine the linear part of the rotation tensor R, written as

R =

¯

R(H) =

¯

F(H)

¯

U

−1

(H) , (5.38)

by ﬁrst obtaining the Gˆateaux diﬀerential as

DR(0, H) = DF(0, H)

¯

U

−1

(0) +

¯

F(0)DU

−1

(0, H) = H−

1

2

(H+H

T

) =

1

2

(H−H

T

) ,

(5.39)

and then writing

L[R; H]

0

=

¯

R(0) + DR(0, H) = I +

1

2

(H−H

T

) . (5.40)

ME185

D

R

A

F

T

106 Consistent linearization of kinematic and kinetic variables

When H is small, the tensor

ω =

1

2

(H−H

T

) (5.41)

is called the inﬁnitesimal rotation tensor and has components ω

ij

=

1

2

(u

i,j

−u

j,i

).

Next, derive the linear part of the Jacobian J of the deformation gradient. To this end,

observe that

D(det F)(0, H) =

¸

d

dω

det

¯

F(ωH)

ω=0

=

¸

d

dω

det(I + ωH)

ω=0

=

¸

d

dω

det¦ω(H−(−

1

ω

)I)¦

ω=0

=

¸

d

dω

ω

3

−(−

1

ω

)

3

+ I

H

(−

1

ω

)

2

−II

H

(−

1

ω

) + III

H

ω=0

=

¸

d

dω

¸

1 +ωI

H

+ ω

2

II

H

+ ω

3

III

H

¸

ω=0

= I

H

= tr H , (5.42)

where I

H

, II

H

, and III

H

are the three principal invariants of H. This, in turn, leads to

L[det F; H]

0

= det

¯

F(0) + D(det F)(0, H) = 1 + tr H = 1 + tr ε . (5.43)

The balance laws are also subject to linearization. For instance, the conservation of mass

statement (4.41) can be linearized to yield

L[ρ

0

; H]

0

= L[ρJ; H]

0

. (5.44)

This means that

ρ

0

= ¯ ρ(0)

¯

J(0) + Dρ(0, H)

¯

J(0) + ¯ ρ(0)DJ(0, H) . (5.45)

Since conservation of mass is assumed to hold in all conﬁgurations therefore also including

the reference conﬁguration, it follows that

ρ

0

= ¯ ρ(0)

¯

J(0) = ¯ ρ(0) , (5.46)

thus equation (5.45), with the aid of (5.42) results in

Dρ(0, H) + ¯ ρ(0) tr ε = 0 , (5.47)

ME185

D

R

A

F

T

Inﬁnitesimal deformations 107

or, equivalently,

Dρ(0, H) = −̺

0

tr ε . (5.48)

The linear part of the mass density relative to the reference conﬁguration now takes the form

L[ρ; H]

0

= ¯ ρ(0) + Dρ(0, H) = ρ

0

(1 −tr ε) . (5.49)

Equation (5.49) reveals that the linearized mass density does not coincide with the mass

density of the reference conﬁguration.

ME185

D

R

A

F

T

108 Consistent linearization of kinematic and kinetic variables

ME185

D

R

A

F

T

Chapter 6

Mechanical constitutive theories

6.1 General requirements

The balance laws furnish a total of seven equations (one from mass balance, three from

linear momentum balance and three from angular momentum balance) to determine thirteen

unknowns, namely the mass density ρ, the position x (or velocity v) and the stress tensor

(e.g., the Cauchy stress T). Clearly, without additional equations this system lacks closure.

The latter is established by constitutive equations, which relate the stresses to the mass

density and the kinematic variables.

Before accounting for any possible restrictions or reductions, a general constitutive equa-

tion for the Cauchy stress may be written as

T =

ˆ

T(x, v, . . . , F,

˙

F, . . . , GradF, . . . , ρ, ˙ ρ, . . .) (6.1)

or, in rate form, as

˙

T =

ˆ

˙

T(x, v, . . . , F,

˙

F, . . . , GradF, . . . , ρ, ˙ ρ, . . .) . (6.2)

Analogous general functional representations may be written for other stress measures.

A number of restrictions may be placed to the preceding equations on mathematical or

physical grounds. Some of these restrictions appear to be universally agreed upon, while

others tend to be less uniformly accepted. Three of these restrictions are reviewed below.

First, all constitutive laws need to be dimensionally consistent. This simply means that

the units of the left- and right-hand side of (6.1) or (6.2) should be the same. For example,

109

D

R

A

F

T

110 Inviscid ﬂuids

when applied to a constitutive law of the form

˙

T = αB, this would necessitate that the

parameter α have units of stress (since B is unitless). Second, the same laws need to be

consistent in their tensorial representation. This means that the right-hand side of (6.1)

and (6.2) should have a tensorial representation in terms of the Eulerian basis. This is to

maintain consistency with the left-hand sides, which are naturally resolved on this basis.

This restriction would disallow, for example, a constitutive law of the form T = βF, where

β is a constant.

The third source of restrictions is the postulate of invariance under superposed rigid

motions (also referred to as objectivity), which is assumed to apply to constitutive laws just

as it applies to balance laws. This postulate requires the functions

ˆ

T and

ˆ

˙

T in (6.1) and

(6.2) to remain unaltered under superposed rigid motions. By way of example, applying the

postulate of invariance under superposed rigid motions to the constitutive law

T =

ˆ

T(F) (6.3)

necessitates that

T

+

=

ˆ

T(F

+

) . (6.4)

Taking into account (3.165) and (4.137), equations (6.3) and (6.4) lead to

Q

ˆ

T(F)Q

T

=

ˆ

T(QF) , (6.5)

for all proper orthogonal tensors Q. Equation (6.5) places a restriction on the choice of

ˆ

T.

An identical restriction is placed on the function

ˆ

˙

T, where, for example,

◦

T =

ˆ

T(F) , (6.6)

and

◦

T is the Jaumann rate of the Cauchy stress tensor. The same conclusion is reached

when using any other objective rate of the Cauchy stress tensor in (6.6).

6.2 Inviscid ﬂuids

An inviscid ﬂuid is deﬁned by the property that the stress vector t acting on any surface

is always opposite to the outward normal n to the surface, regardless of whether the ﬂuid

ME185

D

R

A

F

T

Mechanical constitutive theories 111

is stationary or ﬂowing. Said diﬀerently, an inviscid ﬂuid cannot sustain shearing tractions

under any circumstances. This means that

t = Tn = −pn , (6.7)

hence

T = −pi , (6.8)

see Figure 6.1.

n

t

Figure 6.1: Traction acting on a surface of an inviscid ﬂuid.

On physical grounds, one may assume that the pressure p depends on the density ρ, i.e.,

T = −p(ρ)i . (6.9)

This constitutive relation deﬁnes a special class of inviscid ﬂuids referred to as elastic ﬂuids.

It is instructive here to take an alternative path for the derivation of (6.9). In particular,

suppose that one starts from the more general constitutive assumption

T =

ˆ

T(ρ) . (6.10)

Upon invoking invariance under superposed rigid motions, it follows that

T

+

=

ˆ

T(ρ

+

) , (6.11)

which, with the aid of (4.137) and (4.143) leads to

Q

ˆ

T(ρ)Q

T

=

ˆ

T(ρ) , (6.12)

for all proper orthogonal Q. Furthermore, substituting −Q for Q in (6.12), it is clear that

(6.12) holds for all improper orthogonal tensors Q as well, hence it holds for all orthogonal

tensors. A tensor function

ˆ

T(φ) of a scalar variable is termed isotropic when

Q

ˆ

T(φ)Q

T

=

ˆ

T(φ) , (6.13)

ME185

D

R

A

F

T

112 Inviscid ﬂuids

for all orthogonal Q. This condition may be interpreted as meaning that the components of

the tensor function remain unaltered when resolved on any two orthonormal bases. Clearly,

the constitutive function

ˆ

T in (6.10) is isotropic.

The representation theorem for isotropic tensor functions of a scalar variable states that

a tensor function of a scalar variable is isotropic if, and only if, it is a scalar multiple of

the identity tensor. In the case of

ˆ

T in (6.10), this immediately leads to the constitutive

equation (6.9).

To prove the preceding representation theorem, ﬁrst note that the suﬃciency argument

is trivial. The necessity argument can be made by setting

Q = Q

1

= e

1

⊗e

1

−e

2

⊗e

3

+e

3

⊗e

2

, (6.14)

which, recalling the Rodrigues formula (3.103), corresponds to p = e

1

, q = e

2

, r = e

3

, and

θ = π/2, i.e., to a rigid rotation of π/2 with respect to the axis of e

1

. It is easy to verify

that, in this case, equation (6.13) yields

T

11

−T

13

T

12

−T

31

T

33

−T

32

T

21

−T

23

T

22

¸

¸

¸

¸

=

T

11

T

12

T

13

T

21

T

22

T

23

T

31

T

32

T

33

¸

¸

¸

¸

. (6.15)

This, in turn, means that

T

22

= T

33

, T

12

= T

21

= T

13

= T

31

= 0 , T

23

= −T

32

. (6.16)

Next, set

Q = Q

2

= e

2

⊗e

2

−e

3

⊗e

1

+e

1

⊗e

3

, (6.17)

which corresponds to p = e

2

, q = e

3

, r = e

1

, and θ = π/2. This is a rigid rotation of π/2

with respect to the axis of e

2

. Again, upon using this rotation in (6.13), it follows that

T

33

T

32

−T

31

T

23

T

22

−T

21

−T

13

−T

12

T

11

¸

¸

¸

¸

=

T

11

T

12

T

13

T

21

T

22

T

23

T

31

T

32

T

33

¸

¸

¸

¸

, (6.18)

which leads to

T

11

= T

33

, T

32

= T

12

= −T

12

, T

23

= T

21

. (6.19)

ME185

D

R

A

F

T

Mechanical constitutive theories 113

One may combine the results in (6.16) and (6.19) to deduce that

T = T I , (6.20)

where T = T

11

= T

22

= T

33

, which completes the proof.

Invariance under superposed rigid motions may be also used to exclude certain functional

dependencies in the constitutive assumption for T. Indeed, assume that

T =

ˆ

T(x, ρ) , (6.21)

namely that the Cauchy stress tensor depends explicitly on the position. Invariance of

ˆ

T

under superposed rigid motions implies that

T

+

=

ˆ

T(x

+

, ρ

+

) (6.22)

hence

Q

ˆ

T(x, ρ)Q

T

=

ˆ

T(Qx +c, ρ) , (6.23)

for all proper orthogonal tensors Q and vectors c. Now, choose a superposed rigid motion,

such that Q = I and c = c

0

, where c

0

is constant. It follows from (6.23) that

ˆ

T(x, ρ) =

ˆ

T(x +c

0

, ρ) . (6.24)

However, given that c

0

is arbitrary, the condition (6.24) can be met only if

ˆ

T is altogether

independent of x.

A similar derivation can be followed to argue that a constitutive law of the form

T =

ˆ

T(v, ρ) (6.25)

violates invariance under superposed rigid motions. Indeed, in this case, invariance implies

that

T

+

=

ˆ

T(v

+

, ρ

+

) , (6.26)

which readily translates to

Q

ˆ

T(v, ρ)Q

T

=

ˆ

T(ΩQx +Qv + ˙ c, ρ) . (6.27)

ME185

D

R

A

F

T

114 Inviscid ﬂuids

Now, choose Q = I, Ω = 0 and c = c

0

t, where c

0

is, again, a constant. It follows that for

this particular choice of a superposed motion

ˆ

T(v, ρ) =

ˆ

T(v +c

0

, ρ) , (6.28)

which leads to the elimination of v as an argument in

ˆ

T.

Returning to the balance laws for the elastic ﬂuid, note that angular momentum balance

is satisﬁed automatically by the constitutive equation (6.8) and the non-trivial equations

that govern its motion are written in Eulerian form (which is suitable for ﬂuids) as

˙ ρ + ρ div v = 0

−grad p(ρ) + ρb = ρa

(6.29)

or, upon expressing the acceleration in Eulerian form in terms of the velocities

˙ ρ + ρ div v = 0

−gradp(ρ) + ρb = ρ(

∂v

∂t

+Lv) .

(6.30)

Equations (6.30)

2

are referred to as the compressible Euler equations. Equations (6.30) form

a set of four coupled non-linear partial diﬀerential equations in x and t, which, subject to

the speciﬁcation of suitable initial and boundary conditions and a pressure law p = p(ρ),

can be solved for ρ(x, t) and ˜ v(x, t).

Recall the deﬁnition of an isochoric (or volume-preserving) motion, and note that for such

a motion conservation of mass leads to ρ

0

(X) = ρ(x, t) for all time. Then, upon appealing

to the local statement of mass conservation (4.34), it is seen that div v = 0 for all isochoric

motions.

A material is called incompressible if it can only undergo isochoric motions. If the invis-

cid ﬂuid is assumed incompressible, then the constitutive equation (6.9) loses its meaning,

because the function p(ρ) does not make sense as the density ρ is not a variable quantity.

Instead, the constitutive equation T = −pi holds with p being the unknown. In summary,

the governing equations for an incompressible inviscid ﬂuid (ofter also referred to as an ideal

ﬂuid) are

div v = 0

−grad p + ρ

0

b = ρ

0

(

∂v

∂t

+Lv) ,

(6.31)

ME185

D

R

A

F

T

6.3. VISCOUS FLUIDS 115

where now the unknowns are p and v.

Notice that if a set (p, v) satisﬁes the above equations, then so does another set of the form

(p + c, v), where c is a constant. This suggests that the pressure ﬁeld in an incompressible

elastic ﬂuid is not uniquely determined by the equations of motion. The indeterminacy is

broken by specifying the value of the pressure on some part of the boundary of the domain.

This point is illustrated by way of an example: consider a ball composed of an ideal ﬂuid,

which is in equilibrium under uniform time-independent pressure p. The same “motion” of

the ball can be also sustained by any pressure ﬁeld p+c, where c is a constant, see Figure 6.2.

p

Figure 6.2: A ball of ideal ﬂuid in equilibrium under uniform pressure.

6.3 Viscous ﬂuids

All real ﬂuids exhibit some viscosity, i.e., some ability to sustain shearing forces. It is easy

to conclude on physical grounds that the resistance to shearing is related to the gradient of

the velocity. Therefore, it is sensible to postulate a general constitute law for viscous ﬂuids

in the form

T =

ˆ

T(ρ, L) (6.32)

or, recalling the unique additive decomposition of L in (3.125), more generally as

T =

ˆ

T(ρ, D, W) . (6.33)

The explicit dependence of the Cauchy stress tensor on W can be excluded by invoking

invariance under superposed rigid-body motions. Indeed, invariance requires that

T

+

=

ˆ

T(ρ

+

, D

+

, W

+

) . (6.34)

ME185

D

R

A

F

T

116 Viscous ﬂuids

Now, consider a special superposed rigid motion for which Q(t) = I,

˙

Q(t) = Ω

0

(Ω

0

being

constant skew-symmetric tensor), c(t) = 0 and ˙ c(t) = 0. This is a superposed rigid rotation

with constant angular velocity deﬁned by the skew-symmetric tensor Ω

0

(or, equivalently,

its axial vector ω

0

). Recalling (3.184), (3.185) and (4.143), equation (6.34) takes the form

Q

ˆ

T(ρ, D, W)Q

T

=

ˆ

T(ρ, QDQ

T

, QWQ

T

+ Ω) , (6.35)

for all proper orthogonal Q. Given the special form of the chosen superposed rigid motion,

equation (6.35) leads to

ˆ

T(ρ, D, W) =

ˆ

T(ρ, DQ

T

, W+Ω

0

) , (6.36)

which needs to hold for any skew-symmetric tensor Ω

0

. This implies that the constitutive

function

ˆ

T cannot depend on W, thus it reduces to

T =

ˆ

T(ρ, D) . (6.37)

Invariance under superposed rigid motions for the constitutive function in (6.37) gives rise

to the condition

T

+

=

ˆ

T(ρ

+

, D

+

) , (6.38)

which, in turn, necessitates that

Q

ˆ

T(ρ, D)Q

T

=

ˆ

T(ρ, QDQ

T

) , (6.39)

for all proper orthogonal tensors Q. In fact, since both sides of (6.39) are even functions of

Q, it is clear that (6.39) must hold for all orthogonal tensors Q.

Neglecting, for a moment, the dependence of

ˆ

T on ρ in equation (6.39), note that a tensor

function

ˆ

T of a tensor variable S is called isotropic if

Q

ˆ

T(S)Q

T

=

ˆ

T(QSQ

T

) , (6.40)

for all orthogonal tensors Q. It can be proved following the process used earlier for isotropic

tensor functions of a scalar variable that a tensor function

ˆ

T of a tensor variable S is isotropic

if, and only if, it can be written in the form

ˆ

T(S) = a

0

I + a

1

S + a

2

S

2

, (6.41)

ME185

D

R

A

F

T

Mechanical constitutive theories 117

where a

0

, a

1

, and a

2

are scalar functions of the three principal invariants I

S

, II

S

and III

S

of S, i.e.,

a

0

= ˆ a

0

(I

S

, II

S

, III

S

)

a

1

= ˆ a

1

(I

S

, II

S

, III

S

) . (6.42)

a

2

= ˆ a

2

(I

S

, II

S

, III

S

)

The above result is known as the ﬁrst representation theorem for isotropic tensors. Using

this theorem, it is readily concluded that the Cauchy stress for a ﬂuid that obeys the general

constitutive law (6.37) is of the form

T = a

0

i + a

1

D+ a

2

D

2

, (6.43)

where a

0

, a

1

and a

2

are functions of I

D

, II

D

, III

D

and ρ. The preceding equation character-

izes what is known as the Reiner-Rivlin ﬂuid. Materials that obey (6.43) are also generally

referred to as non-Newtonian ﬂuids.

At this stage, introduce a physically plausible assumption by way of which the Cauchy

stress T reduces to hydrostatic pressure −p(ρ)i when D = 0. Then, one may rewrite the

constitutive function (6.43) as

T = (−p(ρ) + a

∗

0

)i + a

1

D+ a

2

D

2

, (6.44)

where, in general, a

∗

0

= ˆ a

∗

0

(ρ, I

D

, II

D

, III

D

). Clearly, when a

∗

0

= a

1

= a

2

= 0, the viscous

ﬂuid degenerates to an inviscid one.

From the above general class of viscous ﬂuids, consider the sub-class of those which are

linear in D. To preserve linearity in D, the constitutive function in (6.44) is reduced to

T = (−p(ρ) + a

∗

0

)I + a

1

D , (6.45)

where, a

∗

0

= λI

D

and a

1

= 2µ, where λ and µ depend, in general, on ρ. This means that the

Cauchy stress tensors takes the form

T = −p(ρ)i + λ tr Di + 2µD . (6.46)

Viscous ﬂuids that obey (6.46) are referred to as Newtonian viscous ﬂuids or linear viscous

ﬂuids. The functions λ and µ are called the viscosity coeﬃcients.

ME185

D

R

A

F

T

118 Viscous ﬂuids

With the constitutive equation (6.46) in place, consider the balance laws for the Newto-

nian viscous ﬂuid. Clearly, angular momentum balance is satisﬁed at the outset, since T in

(6.46) is symmetric. Conservation of mass and linear momentum balance can be expressed

as

˙ ρ + ρ div v = 0

div(−p(ρ)i + λ tr Di + 2µD) + ρb = ρ ˙ v .

(6.47)

Assuming that either λ and µ are independent of ρ (which is very common) or that ρ is

spatially homogeneous, one may write balance take the form

div(−p(ρ)i + λ tr Di + 2µD) = −gradp + λ graddiv v + µ(div gradv + grad div v)

= −gradp + (λ + µ) graddiv v + µdiv grad v . (6.48)

This implies that for this special case the governing equations of motions (6.47) can be recast

as

˙ ρ + ρ div v = 0

−gradp + (λ + µ) graddiv v + µdiv grad v + ρb = ρ ˙ v .

(6.49)

Equations (6.49)

2

are known as the Navier-Stokes equations for the compressible Newtonian

viscous ﬂuid. As in the case of the compressible inviscid ﬂuid, there are four coupled non-

linear partial diﬀerential equations in (6.49) and four unknowns, namely the velocity v and

the mass density ρ.

If the Newtonian viscous ﬂuid is incompressible, the Cauchy stress is given by

T = −pi + 2µD , (6.50)

hence, the governing equations of (6.49) become

div v = 0

−gradp + µdiv grad v + ρb = ρ ˙ v .

(6.51)

The former equation is a local statement of the constraint of incompressibility, while the

latter is the reduced statement of linear momentum balance that reﬂects incompressibility.

As in the inviscid case, the four unknowns now are the velocity v and the pressure p.

ME185

D

R

A

F

T

6.4. NON-LINEARLY ELASTIC SOLID 119

The Navier-Stokes equations (compressible or incompressible) are non-linear in v due to

the acceleration term, which is expanded in the form ˙ v =

∂˜ v

∂t

+

∂˜ v

∂x

v. In the special case of

very slow and nearly steady ﬂow, referred to as creeping ﬂow or Stokes ﬂow, the acceleration

term may be ignored, giving rise to a system of linear partial diﬀerential equations.

6.4 Non-linearly elastic solid

Recalling the deﬁnition of stress power in the mechanical energy balance theorem of equation

(4.158), deﬁne the non-linearly elastic solid by admitting the existence of a strain energy

function Ψ =

ˆ

Ψ(F) per unit mass, such that

T D = ρ

˙

Ψ . (6.52)

It follows that the stress power in the region { takes the form

P

T Ddv =

P

ρ

˙

Ψdv =

d

dt

P

ρΨdv =

d

dt

W({) , (6.53)

where W({) =

P

ρΨdv is the total strain energy of the material occupying the region {.

As a result, the mechanical energy balance theorem for this class of materials takes the form

d

dt

[K({) + W({)] = R

b

({) + R

c

({) = R({) . (6.54)

In words, equation (6.54) states that the rate of change of the kinetic and strain energy (which

together comprise the total internal energy of the non-linearly elastic material) equals the

rate of work done by the external forces.

Recall that the strain energy function at a given time depends exclusively on the defor-

mation gradient. With the aid of the chain rule, this leads to

˙

Ψ =

∂

ˆ

Ψ

∂F

˙

F , (6.55)

so that, upon recalling (6.52),

T D = ρ

˙

Ψ = ρ

∂

ˆ

Ψ

∂F

(LF) , (6.56)

ME185

D

R

A

F

T

120 Non-linearly elastic solid

which, in turn, leads to

T L = ρ

∂

ˆ

Ψ

∂F

(LF) = ρ

∂

ˆ

Ψ

∂F

F

T

L . (6.57)

The preceding equation can be also written as

(T−ρ

∂

ˆ

Ψ

∂F

F

T

) L = 0 . (6.58)

Observing that this equation holds for any L, it is immediately concluded that

T = ρ

∂

ˆ

Ψ

∂F

F

T

. (6.59)

Upon enforcing angular momentum balance, equation (6.59) leads to

∂

ˆ

Ψ

∂F

F

T

= F

∂

ˆ

Ψ

∂F

T

. (6.60)

This places a restriction on the form of the strain energy function

ˆ

Ψ. Instead of explicitly

enforcing this restriction, one may simply write the Cauchy stress as

T =

1

2

ρ

∂

ˆ

Ψ

∂F

F

T

+F

∂

ˆ

Ψ

∂F

T

¸

¸

. (6.61)

Alternative expressions for the strain energy of the non-linearly elastic solid can be ob-

tained by invoking invariance under superposed rigid motions. Speciﬁcally, invariance of the

function

ˆ

Ψ implies that

Ψ

+

=

ˆ

Ψ(F

+

) =

ˆ

Ψ(QF)

= Ψ =

ˆ

Ψ(F) ,

(6.62)

for all proper orthogonal tensors Q. Selecting Q = R

T

, where R is the rotation stemming

from the polar decomposition of F, it follows from (6.62) that

ˆ

Ψ(F) =

ˆ

Ψ(QF) =

ˆ

Ψ(R

T

RU) =

ˆ

Ψ(U) . (6.63)

Therefore, one may write

Ψ =

ˆ

Ψ(F) =

ˆ

Ψ(U) =

¯

Ψ(C) =

ˇ

Ψ(E) , (6.64)

ME185

D

R

A

F

T

Mechanical constitutive theories 121

by merely exploiting the one-to-one relations between tensors U, C and E. Then, the

material time derivative of Ψ can be expressed as

˙

Ψ =

∂

¯

Ψ

∂C

˙

C =

∂

¯

Ψ

∂C

(2F

T

DF) , (6.65)

where (3.129) is invoked. It follows from (6.52) that

T D = ρ

∂

¯

Ψ

∂C

(2F

T

DF) = 2ρF

∂

¯

Ψ

∂C

F

T

D , (6.66)

which readily leads to

T−2ρF

∂

¯

Ψ

∂C

F

T

D = 0 . (6.67)

Given the arbitrariness of D, it follows that

T = 2ρF

∂

¯

Ψ

∂C

F

T

. (6.68)

Using an analogous procedure, one may also derive a constitutive equation for the Cauchy

stress in terms of the strain energy function

ˇ

Ψ as

T = ρF

∂

ˇ

Ψ

∂E

F

T

. (6.69)

Now, consider a body made of non-linearly elastic material that undergoes a special

motion χ, for which there exist times t

1

and t

2

(> t

1

), such that

x = χ(X, t

1

) = χ(X, t

2

)

v = ˙ χ(X, t

1

) = ˙ χ(X, t

2

) .

(6.70)

for all X. This motion is referred to as a closed cycle. Recall next the theorem of mechanical

energy balance in the form of (6.54) and integrate this equation in time between t

1

and t

2

to ﬁnd that

[K({) + W({)]

t

2

t

1

=

t

2

t

1

[R

b

({) + R

c

({)] dt . (6.71)

However, since the motion is a closed cycle, it is immediately concluded from (6.70) that

[K({) + W({)]

t

2

t

1

=

¸

P

1

2

ρv v dv +

P

ρ

ˆ

Ψ(F) dv

t

2

t

1

= 0 , (6.72)

thus, also

t

2

t

1

[R

b

({) + R

c

({)] dt =

t

2

t

1

¸

P

ρb v dv +

∂P

t v da

dt = 0 . (6.73)

ME185

D

R

A

F

T

122 Non-linearly elastic solid

This proves that the work done on a non-linearly elastic solid by the external forces during

a closed cycle is equal to zero.

Equation (6.54) further implies that

[K({) + W({)]

t

t

1

=

t

t

1

¸

P

ρb v dv +

∂P

t v da

dt . (6.74)

This means that the work done by the external forces taking the body from its conﬁguration

at time t

1

to time t(> t

1

) depends only on the end states at t and t

1

and not on the path

connecting these two states. This is the sense in which the non-linearly elastic material is

characterized as path-independent.

The preceding class of non-linearly elastic materials for which there exists a strain energy

function

ˆ

Ψ is sometimes referred to as Green-elastic or hyperelastic materials. A more general

class of non-linearly elastic materials is deﬁned by the constitutive relation

T =

ˆ

T(F) . (6.75)

Such materials are called Cauchy-elastic and, in general, do not satisfy the condition of

worklessness in a closed cycle. Recalling the constitutive equation (6.61), it is clear that any

Green-elastic material is also Cauchy-elastic.

P

P

{

0

{

′

0

Figure 6.3: Orthogonal transformation of the reference conﬁguration.

The concept of material symmetry is now introduced for the class of Cauchy-elastic ma-

terials. To this end, let P be a material particle that occupies the point X in the reference

conﬁguration. Also, take an inﬁnitesimal volume element {

0

which contains X in the ref-

erence conﬁguration. Since the material is assumed to be Cauchy-elastic, it follows that

the Cauchy stress tensor for P at time t is given by (6.75). Now, subject the reference

conﬁguration to an orthogonal transformation characterized by the orthogonal tensor Q, see

ME185

D

R

A

F

T

Mechanical constitutive theories 123

Figure 6.3. This transforms the region {

0

to {

′

0

. Note, however, that the stress at a point

is independent of the speciﬁc choice of reference conﬁguration. Hence, when expressed in

terms of the deformation relative to the orthogonally transformed reference conﬁguration,

the Cauchy stress at point P is, in general, given by

T =

ˆ

T

′

(FQ) . (6.76)

The preceding analysis shows that the constitutive law depends, in general, on the choice of

reference conﬁguration. For this reason, one may choose, at the expense of added notational

burden, to formally write (6.75) and (6.76) as

T =

ˆ

T

P

0

(F) . (6.77)

and

T =

ˆ

T

P

′

0

(FQ) , (6.78)

respectively.

By way of background, recall here that a group ( is a set together with an operation ∗,

such that the following properties hold for all elements a, b, c of the set:

(i) a ∗ b belongs to the set (closure),

(ii) (a ∗ b) ∗ c = a ∗ (b ∗ c) (associativity),

(iii) There exists an element i, such that i ∗ a = a ∗ i = a (existence of identity),

(iv) For every a, there exists an element −a, such that a ∗ (−a) = (−a) ∗ a = i (existence

of inverse).

It is easy to conﬁrm that the set of all orthogonal transformations Q of the original refer-

ence conﬁguration forms a group under the usual tensor multiplication, called the orthogonal

group or O(3). In this group, the identity element is the identity tensor I and the inverse

element is the inverse (or transpose) Q

−1

(or Q

T

) of any given element Q. The subgroup

1

1

A subset of the group set together with the group operation is called a subgroup if it satisﬁes the closure

property within the subset.

ME185

D

R

A

F

T

124 Non-linearly elastic solid

(

P

0

⊆ O(3) is called a symmetry group for the Cauchy-elastic material with respect to the

reference conﬁguration {

0

if

ˆ

T

P

0

(F) =

ˆ

T

P

0

(FQ) , (6.79)

for all Q ∈ (

P

0

. Physically, equation (6.79) identiﬁes orthogonal transformation Q which

produce the same stress at P under two diﬀerent loading cases. The ﬁrst one subjects the

reference conﬁguration to a deformation gradient F. The second one subjects the reference

conﬁguration to an orthogonal transformation Q and then to the deformation gradient F. If

the stress in both cases is the same, then the orthogonal transformation characterizes the ma-

terial symmetry of the body in the neighborhood of P relative to the reference conﬁguration

{

0

.

If equation (6.79) holds for all Q ∈ O(3), then the Cauchy-elastic material is termed

isotropic relative to the conﬁguration {

0

. An isotropic material is insensitive to any orthog-

onal transformation of its reference conﬁguration. Choosing Q = R

T

and recalling the left

polar decomposition of the deformation gradient, equation (6.79) implies that

T =

ˆ

T

P

0

(F) =

ˆ

T

P

0

(FR

T

) =

ˆ

T

P

0

(VRR

T

) =

ˆ

T

P

0

(V) . (6.80)

In addition, invariance of

ˆ

T

P

0

under superposed rigid motions leads to the condition

Q

ˆ

T

P

0

(V)Q

T

=

ˆ

T

P

0

(QVQ

T

) , (6.81)

for all proper orthogonal tensors Q (hence, given that (6.80) is quadratic in Q, all orthog-

onal Q). This means that

ˆ

T

P

0

is an isotropic tensor function of V. Invoking the ﬁrst

representation theorem for isotropic tensor functions, it follows that

T =

ˆ

T

P

0

(V) = a

0

i + a

1

V+ a

2

V

2

, (6.82)

where a

0

, a

1

, and a

2

are functions of the three principal invariants of V. An alternative

representation of the Cauchy stress of a Cauchy-elastic material is

T =

¯

T

P

0

(B) = b

0

i + b

1

B+ b

2

B

2

, (6.83)

where, now, b

0

, b

1

, and b

2

are functions of the three principal invariants of V. This can be

trivially derived by noting that

¯

T

P

0

(B) =

ˆ

T

P

0

(V).

ME185

D

R

A

F

T

6.5. LINEARLY ELASTIC SOLID 125

6.5 Linearly elastic solid

In this section, a formal procedure is followed to obtain the equations of motion and the

constitutive equations for a linearly elastic solid. To this end, start by writing the linearized

version of linear momentum balance as

L[div T; H]

0

+L[ρb; H]

0

= L[ρa; H]

0

. (6.84)

Now, proceed by making three assumptions:

First, let the reference conﬁguration be stress-free, i.e., assume that if T =

ˆ

T(F) =

¯

T(H),

then

ˆ

T(I) =

¯

T(0) = 0 . (6.85)

It follows that

L[T; H]

0

=

¯

T(0) + D

¯

T(0, H) = D

¯

T(0, H) , (6.86)

where

D

¯

T(0, H) =

¸

d

dω

¯

T(0 + ωH)

ω=0

= CCH . (6.87)

The quantity CC is called the elasticity tensor. This is a fourth-order tensor that can be

resolved in components as

CC = C

ijkl

e

i

⊗e

j

⊗e

k

⊗e

l

. (6.88)

The product CCH in (6.87) can be written explicitly as

CCH = (C

ijkl

e

i

⊗e

j

⊗e

k

⊗e

l

)(H

mn

e

m

⊗e

n

)

= C

ijkl

H

mn

e

i

⊗e

j

[(e

k

⊗e

l

) (e

m

⊗e

n

)]

= C

ijkl

H

mn

e

i

⊗e

j

[δ

km

δ

ln

]

= C

ijkl

H

kl

e

i

⊗e

j

. (6.89)

At this state, note that invariance under superposed rigid motions implies that

Q

ˆ

T(F)Q

T

=

ˆ

T(QF) , (6.90)

for all proper orthogonal tensors Q. Taking into account (6.85), it is concluded from the

above equation that

ˆ

T(Q(t

0

)) = 0 , (6.91)

ME185

D

R

A

F

T

126 Linearly elastic solid

namely that any rigid rotation results in no stress. Hence, one may choose a special such

rotation for which Q(t

0

) = I (hence, also H(t

0

) = 0) and

˙

Q(t

0

) = Ω

0

(hence, also

˙

H = Ω

0

),

where Ω

0

is a constant skew-symmetric tensor. In this case, one may write that

¸

d

dω

¯

T(0 + ω

˙

H)

ω=0

= D

¯

T(0,

˙

H) = CC

˙

H = CCΩ

0

= 0 . (6.92)

Since Ω

0

is an arbitrarily chosen skew-tensor, this means that CCΩ = 0 for any skew-

symmetric tensor Ω. Recalling (6.86), (6.87) and that H = ε +ω, it follows that

L[T; H]

0

= CC(ε +ω) = CCε = σ . (6.93)

Here, σ is the stress tensor of the theory of linear elasticity.

Since the distinction between partial derivatives with respect to X and x disappears in

the inﬁnitesimal case, it is clear that so does the distinction between the “Div” and “div”

operators. Therefore,

L[div T; H]

0

= L[Div T; H]

0

= Div L[T; H]

0

= Div σ . (6.94)

By way of a second assumption, write

L[ρa; H]

0

= ¯ ρ(0)¯ a(0) + [Dρ(0, H)]¯ a(0) + ¯ ρ(0)[Da(0, H)] . (6.95)

With reference to (6.95), assume that in the linearized theory the acceleration a satis-

ﬁes ¯ a(0) = 0 (i.e., there is no acceleration in the reference conﬁguration) and also that

[Da(0, H)] = a (i.e., the acceleration is linear in H). It follows from (6.95) and the preced-

ing assumptions that

L[ρa; H]

0

= ρ

0

a . (6.96)

Finally, noting ﬁrst that

L[ρb; H]

0

= ¯ ρ(0)

¯

b(0) + [Dρ(0, H)]

¯

b(0) + ρ(0)[Db(0, H)] (6.97)

assume that in the linearized theory

¯

b(0) = 0 (which, in view of the earlier two assumptions,

is tantamount to admitting that equilibrium holds in the reference conﬁguration) and also

that [Db(0, H)] = b (i.e., the body force is linear in H). With (6.97) and the preceding

assumptions on b in place, it is easily seen that

L[ρb; H]

0

= ρ

0

b . (6.98)

ME185

D

R

A

F

T

Mechanical constitutive theories 127

Taking into account (6.94), (6.96) and (6.98), equation (6.84) reduces to

Div σ + ρ

0

b = ρ

0

a . (6.99)

Of course, neither a nor b are explicit functions of the deformation. The acceleration a

depends on the deformation implicitly in the sense that the latter is obtained from the motion

χ whose second time derivative equals a. On the other hand, b depends on the motion (and

deformation) implicitly through the balance of linear momentum.

In the context of linear elasticity, all measures of stress coincide, namely the distinction

between the Cauchy stress T and other stress tensors such as P, S, etc, disappears. To see

this point, recall the relation between T and P in (4.130) and take the linear part of both

sides to conclude that

L[T; H]

0

= L

¸

1

J

PF

T

; H

0

, (6.100)

which, in light of (6.93), implies that

σ =

1

¯

J(0)

¯

P(0)

¯

F

T

(0)+

¸

D

1

J

(0, H)

¯

P(0)

¯

F

T

(0)+

1

¯

J(0)

[DP(0, H)]

¯

F

T

(0)+

1

¯

J(0)

¯

P(0)[DF

T

(0, H)] .

(6.101)

Recalling that the reference conﬁguration is assumed stress-free, it follows from the above

equation that

σ = DP(0, H) , (6.102)

which implies further that

L[P; H]

0

=

¯

P(0) + DP(0, H) = σ , (6.103)

hence,

L[P; H]

0

= L[T; H]

0

. (6.104)

Similar derivations apply to deduce equivalency of T to other stress tensors in the inﬁnites-

imal theory.

Return now to the constitutive law (6.93) and write it in component form as

σ

ij

= C

ijkl

ε

kl

. (6.105)

ME185

D

R

A

F

T

128 Linearly elastic solid

In general, the fourth-order elasticity tensor possesses 3 3 3 3 = 81 material constants

C

ijkl

. However, since balance of angular momentum implies that σ

ij

= σji and also, by the

deﬁnition of ε in (5.32), ε

ij

= ε

ji

, it follows that

C

ijkl

= C

jikl

= C

ijlk

= C

jilk

, (6.106)

which readily implies that only 6 6 = 36 of these components are independent.

2

Next, note that in the inﬁnitesimal theory, equation (6.93) may be derived from a strain

energy function W = W(ε) per unit volume as

σ =

∂W

∂ε

, (6.107)

where

W =

1

2

ε CCε . (6.108)

It follows from (6.108) and (6.107) that

∂σ

ij

∂ε

kl

=

∂

2

W

∂ε

ij

∂ε

kl

= C

ijkl

, (6.109)

which, in turn, implies that C

ijkl

= C

klij

. The preceding identity reduces the number of

independent material parameters from 36 to 36 −

36

2

+ 3 = 21.

3

The number of independent material parameters can be further reduced by material

symmetry. To see this point, recall the constitutive equations (6.82) or (6.83) for the isotropic

non-linearly elastic solid and note that a corresponding equation can be easily deduced in

terms of the Almansi strain e, namely

T =

˜

T

P

0

(e) = c

0

i + c

1

e + c

2

e

2

, (6.110)

where c

0

, c

1

, and c

2

are function of the three principal invariants I

e

, II

e

and III

e

of e.

Recalling (5.32)

2

and noting that, by assumption, the reference conﬁguration is stress-free,

it is easy to see that the linearization of (6.110) leads to

σ = c

∗

0

I

ε

I + c

1

ε , (6.111)

2

To see this, take each pair (i, j) or (k, l) and use (6.106) to conclude that only 6 combinations of each

pair are independent.

3

To see this, write the 36 parameters as a 6 6 matrix and argue that only the terms on and above (or

below) the major diagonal are independent.

ME185

D

R

A

F

T

6.6. VISCOELASTIC SOLID 129

where c

∗

0

and c

1

are constants. Setting c

∗

0

= λ and c

1

= 2µ, one may rewrite the preceding

equation as

σ = λ tr εI + 2µε . (6.112)

The parameters λ and µ are known as the Lam´e constants. These are related to the Young’s

modulus E and the Poisson’s ratio ν by

λ =

Eν

(1 +ν)(1 −2ν)

, µ =

E

2(1 +ν)

(6.113)

and, inversely,

E =

µ(3λ + 2µ)

λ + µ

, ν =

λ

2(λ + µ)

. (6.114)

6.6 Viscoelastic solid

Most materials exhibit some memory eﬀects, i.e., their current state of stress depends not

only on the current state of deformation, but also on the deformation history.

Consider ﬁrst a broad class of materials with memory, whose Cauchy stress is given by

T(X, t) =

ˆ

T( H

τ≤t

[F(X, τ)]; X) . (6.115)

This means that the Cauchy stress at time t of some material particle P which occupies

point X in the reference conﬁguration depends on the history of the deformation gradient of

that point up to (and including) time t. Materials that satisfy the constitutive law (6.115)

are called simple.

Invoking invariance under superposed rigid motions and suppressing, in the interest of

brevity, the explicit dependence of functions on X, it is concluded that

Q(t)

ˆ

T( H

τ≤t

[F(τ)])Q

T

(t) =

ˆ

T( H

τ≤t

[Q(τ)F(τ)]) , (6.116)

for all proper orthogonal Q(τ), where −∞ < τ ≤ t. Choosing Q(τ) = R

T

(τ), for all

τ ∈ (−∞, t], it follows that

R

T

(t)

ˆ

T( H

τ≤t

[F(τ)])R(t) =

ˆ

T( H

τ≤t

[U(τ)]) , (6.117)

where, as usual, F(τ) = R(τ)U(τ). Equation (6.117) can be readily rewritten as

T(t) = R(t)

ˆ

T( H

τ≤t

[U(τ)])R

T

(t) . (6.118)

ME185

D

R

A

F

T

130 Viscoelastic solid

or, equivalently,

T(t) = F(t)U

−1

(t)

ˆ

T( H

τ≤t

[U(τ)])U

−1

(t)F

T

(t) . (6.119)

Upon recalling (4.133)

2

, this means that

S(t) = J(t)U

−1

(t)

ˆ

T( H

τ≤t

[U(τ)])U

−1

(t) =

ˆ

S( H

τ≤t

[U(τ)]) . (6.120)

As usual, one may alternatively write

S(t) =

¯

S( H

τ≤t

[C(τ)]) =

ˇ

S( H

τ≤t

[E(τ)]) . (6.121)

Now, proceed by distinguishing between the past (τ < t) and the present (τ = t) in the

preceding constitutive laws. To this end, deﬁne

E

t

(s) = E(t −s) −E(t) , (6.122)

where, obviously, E

t

(0) = 0. Clearly, for any given time t, the variable s is probing the

history of the Lagrangian strain looking further in the past as s increases.

Next, rewrite (6.121)

2

as

S(t) =

ˇ

S( H

τ≤t

[E(τ)]) =

ˇ

S(

¯

H

s≥0

[E

t

(s), E(t)]) . (6.123)

Then, deﬁne the elastic response function S

e

as

S

e

(E(t)) =

ˇ

S(

¯

H

s≥0

[0, E(t)]) . (6.124)

and the memory response function S

m

as

S

m

(

¯

H

s≥0

[E

t

(s), E(t)]) =

ˇ

S(

¯

H

s≥0

[E

t

(s), E(t)]) −

ˇ

S(

¯

H

s≥0

[0, E(t)]) . (6.125)

In summary, the stress response is given by

S(t) = S

e

(E(t)) + S

m

(

¯

H

s≥0

[E

t

(s), E(t)]) . (6.126)

The ﬁrst term on the right-hand side of (6.126) is the stress which depends exclusively on

the present state of the Lagrangian strain. The second term on the right-hand side of (6.126)

contains all the dependency of the stress response on past Lagrangian strain states.

ME185

D

R

A

F

T

Mechanical constitutive theories 131

Note that, by deﬁnition, the stress during a state deformation, i.e., which E(t) = E

0

,

where E

0

is a constant, is equal to S(t) = S

e

(E

0

), or, equivalently, S

m

(

¯

H

s≥0

[0, E(t)]) = 0.

All viscoelastic materials can be described by the constitutive equation (6.126). For such

materials, S

m

is rate-dependent (i.e., it depends on the rate

˙

E of Lagrangian strain) and

also exhibits fading memory. The latter means that the impact on the stress at time t of

the deformation at time t −s (s > 0) diminishes as s → ∞. This condition can be expressed

mathematically as

lim

δ→∞

S

m

(

¯

H

s≥0

[E

δ

t

(s), E(t)]) = 0 , (6.127)

where

E

δ

t

(s) =

0 if 0 ≤ s < δ

E

t

(s −δ) if δ ≤ s < ∞

(6.128)

is the static continuation of E

t

(s) by δ. With reference to Figure 6.4, it is seen that the static

s

δ

E

t

(s) E

δ

t

(s)

Figure 6.4: Static continuation E

δ

t

(s) of E

t

(s) by δ.

continuation is a rigid translation of the argument E

t

(s) of the memory response function

S

m

by δ. Therefore, the fading memory condition (6.127) implies that, as time elapses,

the eﬀect of earlier Lagrangian strain states on S

m

continuously diminishes and, ultimately,

disappears altogether. The condition (6.127) is often referred to as the relaxation property.

This is because it implies that any time-dependent Lagrangian strain process which reaches

a steady-state results in memory response which ultimately relaxes to zero memory stress

(plus, possibly, elastic stress), see Figure 6.5.

Under special regularity conditions, the memory response function bS

m

can be reduced

to the linear functional form

S

m

(

¯

H

s≥0

[E

t

(s), E(t)]) =

∞

0

LL(E(t), s)E

t

(s) dS , (6.129)

ME185

D

R

A

F

T

132 Viscoelastic solid

s

t

t

1

t

2

E

t

1

(s) E

t

2

(s)

Figure 6.5: An interpretation of relaxation

where LL(E(t), s) is a fourth-order tensor function of E(t) and s. Of course, LL(E(t), s) needs

to be chosen so that S

m

satisfy the relaxation property (6.127).

Upon Taylor expansion of E

t

(s) around t −s, one ﬁnds that

E

t

(s) = E(t −s) −E(t) = −s

˙

E(t −s) + o(s

2

) . (6.130)

Ignoring the second-order term in (6.130), which is tantamount to neglecting long-term

memory eﬀects, one may substitute E

t

(s) in (6.129) to ﬁnd that

S

m

(

¯

H

s≥0

[E

t

(s), E(t)]) =

∞

0

LL(E(t), s)¦−s

˙

E(t −s)¦ ds =

∞

0

¯

LL(E(t), s)

˙

E(t −s) ds .

(6.131)

Alternatively, upon Taylor expansion of E

t

(s) around t, one ﬁnds that

E

t

(s) = E(t −s) −E(t) = −s

˙

E(t) + o(s

2

) , (6.132)

which leads to

S

m

(

¯

H

s≥0

[E

t

(s), E(t)]) =

∞

0

LL(E(t), s)¦−s

˙

E(t)¦ ds =

¸

−

∞

0

LL(E(t), s)s ds

˙

E(t)

= ´´(E(t))

˙

E(t) , (6.133)

where

´´(E(t)) = −

∞

0

LL(E(t), s)s ds . (6.134)

Now, attempt to reconcile the general constitutive framework developed here with the clas-

sical one-dimensional viscoelasticity models of Kelvin-Voigt and Maxwell. Start with the

Kelvin-Voigt model, which is comprised of a spring and a dashpot in parallel, and let the

ME185

D

R

A

F

T

Mechanical constitutive theories 133

spring constant be E and the dashpot constant be η. It follows that the uniaxial stress σ is

related to the uniaxial strain ε by

σ = Eε + η ˙ ε . (6.135)

Clearly, this law is a simple reduction of (6.126), where the memory response is obtained

from a one-dimensional counterpart of (6.133).

Next, consider the Maxwell model of a spring and dashpot in series with material prop-

erties as in the Kelvin-Vogt model. In this case, the constitutive law becomes

˙ ε =

˙ σ

E

+

σ

η

, (6.136)

and take the accompanying initial condition to be σ(0) = 0. The general solution of (6.136)

is

σ(t) = c(t)e

−

E

η

t

, (6.137)

which, upon substituting into (6.136) leads to

˙ c(t) = Ee

E

η

t

˙ ε(t) . (6.138)

This, in turn, may be integrated to yield

c(t) = c(0) +

t

0

Ee

E

η

τ

˙ ε(τ) dτ . (6.139)

Noting that the initial condition results in c(0) = 0, one may write that

σ(t) =

¸

t

0

Ee

E

η

τ

˙ ε(τ) dτ

e

−

E

η

t

=

t

0

Ee

E

η

(τ−t)

˙ ε(τ) dτ

=

t

0

Ee

−

E

η

s

˙ ε(t −s) ds . (6.140)

Clearly, the stress response of the Maxwell model falls within the constitutive framework of

(6.126), where the elastic response function vanishes identically and the memory response

can be deduced from (6.131).

ME185

D

R

A

F

T

134 Viscoelastic solid

ME185

D

R

A

F

T

Chapter 7

Boundary- and Initial/boundary-value

Problems

7.1 Incompressible Newtonian viscous ﬂuid

Parts of this chapter include material taken from the continuum mechanics notes of the late

Professor P.M. Naghdi.

7.1.1 Gravity-driven ﬂow down an inclined plane

Consider an incompressible Newtonian viscous ﬂuid under steady ﬂow down an inclined plan

due to gravity, see Figure 7.1. Let the pressure of the free surface be p

0

and assume that the

ﬂuid region has constant depth h.

g

h

θ

e

1

e

2

e

3

Figure 7.1: Flow down an inclined plane

135

D

R

A

F

T

136 Flow down an inclined plane

Assume that the velocity and pressure ﬁelds are of the form

v = ˜ v(x

2

, x

3

)e

2

(7.1)

and

p = ˜ p(x

1

, x

2

, x

3

) , (7.2)

respectively. Incompressibility implies that

div v =

∂˜ v

∂x

2

= 0 , (7.3)

which means that the velocity ﬁeld is independent of x

2

, namely that

v = ¯ v(x

3

)e

2

. (7.4)

Given the reduced velocity ﬁeld in (7.4), the rate-of-deformation tensor has components

[D

ij

] =

0 0 0

0 0

1

2

∂¯ v

∂x

3

0

1

2

∂¯ v

∂x

3

0

¸

¸

¸

¸

. (7.5)

Recalling the constitutive equation (6.50), the Cauchy stress is given by

[T

ij

] =

−p 0 0

0 −p µ

∂¯ v

∂x

3

0 µ

∂¯ v

∂x

3

−p

¸

¸

¸

¸

. (7.6)

Note that gravity induces a body force per unit mass equal to

b = g(sin θe

2

−cos θe

3

) , (7.7)

where g is the gravitational constant. Given (7.2), (7.4), (7.6) and (7.7), the equations of

linear momentum balance assume the form

−

∂˜ p

∂x

1

= 0

−

∂˜ p

∂x

2

+ µ

d

2

¯ v

dx

2

3

+ ρg sin θ = ρ

¸

∂¯ v

∂t

+

∂¯ v

∂x

k

v

k

. (7.8)

−

∂˜ p

∂x

3

−ρg cos θ = 0

ME185

D

R

A

F

T

Boundary- and initial-value problems 137

It follows from (7.8)

1,3

that

p = ¯ p(x

2

, x

3

) = −ρg cos θx

3

+ f(x

2

) . (7.9)

Imposing the boundary condition on the free surface leads to

¯ p(x

2

, h) = −ρg cos θh + f(x

2

) = p

0

, (7.10)

which implies that the function f(x

2

) is constant and equal to

f(x

2

) = p

0

+ ρg cos θh . (7.11)

Substituting this equation to (7.9) results in an expression for the pressure as

p = p

0

+ ρg cos θ(h −x

3

) . (7.12)

Substituting the pressure from (7.12) into the remaining momentum balance equation (7.8)

2

yields

µ

d

2

¯ v

dx

2

3

+ ρg sin θ = 0 , (7.13)

which may be integrated to

¯ v(x

3

) =

−ρg sin θ

2µ

x

2

3

+ c

1

x + c

2

. (7.14)

Enforcing the boundary conditions ¯ v(0) = 0 (no slip condition on the solid-ﬂuid interface)

and T

23

(h) = 0 (no shear stress on the free surface), gives c

2

= 0 and c

1

=

ρgh sin θ

µ

, which,

when substituted into (7.14) yield

¯ v(x

3

) =

ρg sin θ

µ

x

3

(h −

x

3

2

) . (7.15)

It is seen from (7.15) that the velocity distribution is parabolic along x

3

and attains a

maximum value of v

max

=

ρg sin θ

2µ

h

2

on the free surface.

7.1.2 Couette ﬂow

This is the steady ﬂow between two concentric cylinder of radii R

o

(outer cylinder) and R

i

(inner cylinder) rotating with constant angular velocities ω

o

(outer cylinder) and ω

i

(inner

ME185

D

R

A

F

T

138 Couette ﬂow

r

R

o

R

i

ω

o

ω

i

e

r

e

θ

Figure 7.2: Couette ﬂow

cylinder), see Figure 7.2. The ﬂuid is assumed Newtonian and incompressible. Also, the

eﬀect of body forces is neglected.

The problem lends itself naturally to analysis using cylindrical polar coordinates with

basis vectors ¦e

r

, e

θ

, e

z

¦. The velocity and pressure ﬁelds are assumed axisymmetric and,

using the cylindrical polar coordinate representation, can be expressed as

v = ¯ v(r)e

θ

(7.16)

and

p = ¯ p(r) . (7.17)

Taking into account (A.9), the spatial velocity gradient can be written as

L =

d¯ v(r)

dr

e

θ

⊗e

r

−

¯ v

r

e

r

⊗e

θ

, (7.18)

so that

D =

r

2

d

dr

¯ v(r)

r

(e

r

⊗e

θ

+e

θ

⊗e

r

) . (7.19)

It follows from (7.19) that

T = −pi + µr

d

dr

¯ v(r)

r

(e

r

⊗e

θ

+e

θ

⊗e

r

) . (7.20)

Given the form of the velocity in (7.16), it is clear that div v = 0, hence the incompress-

ibility condition is satisﬁed at the outset. Also, the acceleration becomes

a =

d¯ v(r)

dr

e

θ

⊗e

r

− ¯ v

1

r

e

r

⊗eθ

¯ ve

θ

= −

¯ v

2

r

e

r

. (7.21)

ME185

D

R

A

F

T

Boundary- and initial-value problems 139

Taking into account (7.20) and (7.21) the linear momentum balance equations in the r-

and θ-directions become

−

dp

dr

= −ρ

¯ v

2

(r)

dr

µ

d

dr

¸

r

d

dr

v

r

+ 2µ

d

dr

v

r

= 0 ,

(7.22)

respectively.

The second of the above equations may be integrated twice to give

¯ v(r) = c

1

r +

c

2

r

. (7.23)

The integration constants c

1

and c

2

can be determined by imposing the boundary conditions

¯ v(R

i

) = ω

i

R

i

and ¯ v(R

o

) = ω

o

R

o

. Upon determining these constants, the velocity of the ﬂow

takes the form

¯ v(r) = ω

o

R

o

Ro

R

i

r

R

i

−

R

i

r

+

ω

i

ω

o

R

o

r

−

r

R

o

R

o

R

i

2

−1

. (7.24)

Finally, integrating equation (7.22)

1

and using a pressure boundary condition such as, e.g.,

¯ p(R

0

) = p

0

, leads to an expression for the pressure ¯ p(r).

7.1.3 Poiseuille ﬂow

This is the ﬂow of an incompressible Newtonian viscous ﬂuid through a straight cylindrical

pipe of radius R. Assuming that the pipe is aligned with the e

3

-axis, assume that the velocity

of the ﬂuid of the general form

v = v(r)e

z

, (7.25)

while the pressure is

p = ¯ p(r, z) . (7.26)

Clearly, the assumed velocity ﬁeld satisﬁes the incompressibility condition at the outset.

Taking again into account (A.9), the velocity gradient for this ﬂow is

L =

∂v

∂r

e

z

⊗e

r

, (7.27)

ME185

D

R

A

F

T

140 Stokes’ First Problem

hence the rate-of-deformation tensor is

D =

1

2

∂v

∂r

(e

r

⊗e

z

+e

z

⊗e

r

) . (7.28)

Then, the Cauchy becomes

T = −pi + µ

∂v

∂r

(e

r

⊗e

z

+e

z

⊗e

r

) . (7.29)

Noting that the acceleration vanishes identically, the equations of linear momentum bal-

ance in the absence of body force take the form

−

∂p

∂r

= 0

−

1

r

∂p

∂θ

= 0 , (7.30)

−

∂p

∂z

+ µ

d

2

v

dr

2

+

µ

r

dv

dr

= 0

where (A.11) is employed. Equation (7.30)

2

is satisﬁed identically due to the assumption

(7.26), while equation (7.30)

1

implies that p = p(z). However, given that v depends only on

r, equation (7.30)

3

requires that

dp

dz

= c , (7.31)

where c is a constant. Upon integrating (7.30)

3

in r, one ﬁnds that

v =

cr

2

4µ

+ c

1

ln r + c

2

, (7.32)

where c

1

and c

2

are also constants. Admitting that the solution should remain ﬁnite at r = 0

and imposing the no-slip condition v(R) = 0, it follows that

v(r) =

c

4µ

(r

2

−R

2

0

) . (7.33)

Two additional boundary conditions are necessary (either a velocity boundary condition on

one end and a pressure boundary condition on another or pressure boundary conditions on

both ends of the pipe) in order to fully determine the velocity and pressure ﬁeld.

7.2 Compressible Newtonian viscous ﬂuids

7.2.1 Stokes’ First Problem

See Problem 5 in Homework 11.

ME185

D

R

A

F

T

Boundary- and initial/boundary-value problems 141

7.2.2 Stokes’ Second Problem

Consider the semi-inﬁnite domain 1 = ¦(x

1

, x

2

, x

3

) [ x

3

> 0¦, which contains a Newtonian

viscous ﬂuid, see Figure 7.3. The ﬂuid is subjected to a periodic motion of its boundary

x

3

= 0 in the form

v

p

= U cos ωte

1

, (7.34)

where U and ω are given positive constants. In addition, the body force is zero everywhere

in the domain.

v

p

e

1

e

2

e

3

Figure 7.3: Semi-inﬁnite domain for Stokes’ Second Problem

Adopting a semi-inverse approach, assume a general form of the solution as

v = f(x

3

) cos (ωt −αx

3

)e

1

, (7.35)

where the function f and the constant α are to be determined. The solution (7.35) assumes

that the velocity ﬁeld varies along x

3

and is also phase-shifted by αx

3

relative to the boundary

velocity. Further, note that the assumed solution renders the motion isochoric.

In view of (7.35), the acceleration ﬁeld is

a = −ωf sin (ωt −αx

3

)e

1

. (7.36)

Further, mass conservation implies that ˙ ρ = 0, i.e., if the ﬂuid is initially homogeneous

(which it is assumed to be), then it remains homogeneous for all time.

The only non-vanishing components of the rate-of-deformation tensor are

D

23

= D

32

=

1

2

¸

df

dx

cos (ωt −αx

3

) + αf sin (ωt −αx

3

)

. (7.37)

ME185

D

R

A

F

T

142 Stokes’ Second Problem

Recalling (6.46), with tr D = div v = 0, and noting that p and µ are necessarily constant,

since the mass density is homogeneous, it follows that

[T

ij

] =

−p 0 T

13

0 −p 0

T

31

0 −p

¸

¸

¸

¸

, (7.38)

where

T

13

= T

31

= µ

¸

df

dx

cos (ωt −αx

3

) + αf sin (ωt −αx

3

)

. (7.39)

Taking into account (7.36) and (7.38) it is easy to see that the linear momentum balance

equations in the e

2

and e

3

directions hold identically. In the e

1

direction, the momentum

balance equation reduces to

∂T

13

∂x

3

= ρa

1

(7.40)

or

µ

¸

d

2

f

dx

2

3

cos (ωt −αx

3

) + 2α

df

dx

3

sin (ωt −αx

3

) −α

2

f cos (ωt −αx

3

)

**= ρωf sin (ωt −αx
**

3

) . (7.41)

The preceding equation can be also written as

µ

¸

d

2

f

dx

2

3

−α

2

f

cos (ωt −αx

3

) +

¸

2µα

df

dx

3

+ ρωf

sin (ωt −αx

3

) = 0 . (7.42)

For this equation to be satisﬁed identically for all x

3

and t, it is necessary and suﬃcient that

d

2

f

dx

2

3

−α

2

f = 0 (7.43)

and

2µα

df

dx

3

+ ρωf = 0 . (7.44)

These two equations can be directly integrated to give

f(x

3

) = c

1

e

αx

3

+ c

2

e

−αx

3

(7.45)

and

f(x

3

) = c

3

e

−

ρω

2µα

x

3

, (7.46)

ME185

D

R

A

F

T

7.3. LINEAR ELASTIC SOLIDS 143

respectively. To reconcile the two solutions, one needs to take

f(x

3

) = ce

−

ρω

2µα

x

3

, (7.47)

where α =

ρω

2µ

. With this expression in place, the velocity ﬁeld of equation (7.35) takes

the form

v = ce

−

ρω

2µα

x

3

cos (ωt −αx

3

)e

1

. (7.48)

Applying the boundary condition v(0) = U cos ωte

1

leads to c = U, so that, ﬁnally,

v = Ue

−

ρω

2µα

x

3

cos (ωt −αx

3

)e

1

. (7.49)

In this problem, the pressure p is constitutively speciﬁed, yet is here constant throughout

the semi-inﬁnite domain owing to the homogeneity of the mass density.

7.3 Linear elastic solids

Consider a homogeneous and isotropic linearly elastic solid, and recall that its stress-strain

relation is given by (6.112). Taking the trace of both sides on (6.112) and assuming that

λ +

2

3

µ is non-zero, it is easily seen that

tr ε =

1

3λ + 2µ

tr σ , (7.50)

hence

ε =

1

2µ

¸

σ −

λ

3λ + 2µ

tr σI

(7.51)

or, using the elastic materials parameters E and ν,

ε =

1

E

[(1 +ν)σ −ν tr σI] . (7.52)

where (6.113) is used.

7.3.1 Simple tension and simple shear

In the case of simple tension along the e

3

-axis, σ

33

= 0, while all other components σ

ij

= 0.

It follows from (7.52) that

ε

33

=

σ

33

E

, ε

11

= ε

22

= −

νσ

33

E

, (7.53)

ME185

D

R

A

F

T

144 Saint-Venant torsion of a circular cylinder

while all shearing components of strain vanish. A simple tension experiment can be used to

determine the material constants E and ν as

E =

σ

33

ε

33

, ν = −

ε

11

ε

33

. (7.54)

In the case of simple shear on the plane of e

1

and e

2

, the only non-zero components of

stress is σ

23

= σ

32

. Recalling (7.52), it follows that

ε

12

=

σ

12

2µ

, (7.55)

while all other strain components vanish. The elastic constant µ can be experimentally

measured by arguing that 2ε is the change in the angle between the axes e

1

and e

2

.

7.3.2 Uniform hydrostatic pressure

Suppose that a homogeneous and isotropic linearly elastic solid is subjected to a uniform

hydrostatic pressure σ = −pI. Taking into account (7.51), it follows that

ε = −p

1

3λ + 2µ

I = −p

1

3K

I , (7.56)

where K =

3λ + 2µ

3

=

E

3(1 −2ν)

is the bulk modulus of elasticity. Equation (7.56) can

be used in an experiment to determine the bulk modulus by noting that tr ε = −

p

K

is the

inﬁnitesimal change of volume due to the hydrostatic pressure p.

7.3.3 Saint-Venant torsion of a circular cylinder

Consider an isotropic homogeneous linearly elastic cylinder under quasistatic conditions.

The cylinder has length L and radius R and is ﬁxed on the one end (x

3

= 0), while at the

opposite end (x

3

= L) it is subjected to a resultant moment Me

3

relative to the point with

coordinates (0, 0, L). Also, the lateral sides of the cylinder are assumed traction-free.

Due to symmetry, it is assumed that the cross-section remains circular and that plane

sections of constant x

3

remain plane after the deformation. With these assumptions in place,

assume that the displacement of the cylinder may be written as

u = αx

3

re

θ

, (7.57)

ME185

D

R

A

F

T

Boundary- and initial-value problems 145

where α is the angle of twist per unit x

3

-length. Recalling that e

θ

= −e

1

x

2

r

+e

2

x

1

r

, one may

rewrite the displacement as

u = α(−x

2

x

3

e

1

+ x

1

x

3

e

2

) . (7.58)

It follows that the inﬁnitesimal strain tensor has components

[ε

ij

] =

0 0 −

1

2

αx

2

0 0

1

2

αx

1

−

1

2

αx

2

1

2

αx

1

0

¸

¸

¸

¸

. (7.59)

Hence, the stress tensor has components

[σ

ij

] = µα

0 0 −x

2

0 0 x

1

−x

2

x

1

0

¸

¸

¸

¸

. (7.60)

It can be readily demonstrated that, in the absence of body forces, all the equilibrium

equations are satisﬁed identically. Further, for the lateral surfaces, the tractions vanish, since

[t

i

] = [σ

ij

][n

j

] = µα

0 0 −x

2

0 0 x

1

−x

2

x

1

0

¸

¸

¸

¸

1

r

x

1

x

2

0

¸

¸

¸

¸

=

0

0

0

¸

¸

¸

¸

. (7.61)

On the other hand, the traction at x = L is

[t

i

] = [σ

ij

][n

j

] = µα

0 0 −x

2

0 0 x

1

−x

2

x

1

0

¸

¸

¸

¸

0

0

1

¸

¸

¸

¸

= µα

−x

2

x

1

0

¸

¸

¸

¸

, (7.62)

so that the resultant force is given by

x

3

=L

[t

i

] dA = µα

2π

0

R

0

r

−sin θ

cos θ

0

¸

¸

¸

¸

r drdθ

= µα

R

3

3

2π

0

−sin θ

cos θ

0

¸

¸

¸

¸

dθ = µα

R

3

3

cos θ

sin θ

0

¸

¸

¸

¸

2π

0

=

0

0

0

¸

¸

¸

¸

.

(7.63)

ME185

D

R

A

F

T

146 Rivlin’s cube

Moreover, the resultant moment with respect to (0, 0, L) equals

Me

3

=

x

3

=L

(x

1

e

1

+ x

2

e

2

) µα(−x

2

e

1

+ x

1

e

2

) dA

= µα

x

3

=L

(x

2

1

+ x

2

2

) dA = µα

2π

0

R

0

r

2

rdrdθ = µαI ,

(7.64)

where I =

πR

4

2

is the polar moment of inertia of the circular cross-section.

7.4 Non-linearly elastic solids

7.4.1 Rivlin’s cube

Consider a unit cube made of a homogeneous, isotropic and incompressible non-linearly

elastic material. First, observe that the pressure p is work-conjugate to the volume change

in that

T D = (T

′

+

1

3

tr Ti) (D

′

+

1

3

tr Di) = T

′

D

′

+ (−p) div v , (7.65)

where T

′

and D

′

are the deviatoric parts of T and D, respectively. Next, recall the general

form of the constitutive equations for isotropic non-linearly elastic materials in (6.83) and,

assuming, as a special case, b

2

= 0, write

T = −pi + b

1

B , (7.66)

where b

1

(> 0) is a constant, and p is a Lagrange multiplier to be determined upon enforcing

the incompressibility constraint.

Returning to the unit cube, assume that its edges are aligned with the common orthonor-

mal basis ¦e

1

, e

2

, e

3

¦ of the reference and current conﬁguration, and is loaded by three pairs

of equal and opposite tensile forces, all of equal magnitude, and distributed uniformly on

each face.

Taking into account (4.130) and (7.66), one may write

P = J(−pi + b

1

B)F

−T

= −pF

−T

+ b

1

F , (7.67)

where J = 1 due to the assumption of incompressibility. The tractions, when resolved on

the geometry of the reference conﬁguration, satisfy

Pe

i

= ce

i

, (7.68)

ME185

D

R

A

F

T

Boundary- and initial/boundary-value problems 147

where c > 0 is the magnitude of the tractions per unit area in the reference conﬁguration.

Note that c is the same for all faces of the cube, since, by assumption, the force on each face

is constant and uniform. Therefore, one may also write that

P = c(e

1

⊗e

1

+e

2

⊗e

2

+e

3

⊗e

3

) . (7.69)

Solutions of the form

F = λ

1

e

1

⊗e

1

+ λ

2

e

2

⊗e

2

+ λ

3

e

3

⊗e

3

(7.70)

are sought for this boundary-value problem, subject to the condition λ

1

λ

2

λ

3

= 1. It is clear

from (7.68) that the equilibrium equations are satisﬁed identically in the absence of body

forces. Returning to the constitutive equations, combine (7.67) and (7.69) to conclude that

c = −

p

λ

i

+ b

1

λ

i

, (7.71)

or

b

1

λ

2

i

= cλ

i

+ p , (7.72)

where i = 1, 2, 3. Eliminating the pressure in the preceding equation leads to

b

1

(λ

2

i

−λ

2

j

) = c(λ

i

−λ

j

) , (7.73)

where i = j. This, in turn, means that

λ

1

= λ

2

or b

1

(λ

1

+ λ

2

) = c

λ

2

= λ

3

or b

1

(λ

2

+ λ

3

) = c , (7.74)

λ

3

= λ

1

or b

1

(λ

3

+ λ

1

) = c

subject to λ

1

λ

2

λ

3

= 1, where not all conditions can be chosen from the same column in

(7.74).

One solution of (7.74) is obviously

λ

1

= λ

2

= λ

3

= 1 . (7.75)

This corresponds to the cube remaining rigid under the inﬂuence of the tensile load. Another

possible set of solutions is λ

1

= λ

2

= λ

3

, λ

2

= λ

3

= λ

1

and λ

3

= λ

1

= λ

2

. Explore one of

these solutions, say λ

1

= λ

2

= λ

3

by setting λ

3

= λ and noting from (7.74) that

λ

2

+ λ

3

= λ

3

+ λ

1

=

c

b

1

= η , (7.76)

ME185

D

R

A

F

T

148 Rivlin’s cube

so that

λ

1

λ

2

λ

3

= (η −λ)

2

λ = 1 . (7.77)

The preceding equation may be rewritten as

f(λ) = λ

3

−2ηλ

2

+ η

2

λ −1 = 0 . (7.78)

To examine the roots of f(λ) = 0, note that

f

′

(λ) = 3λ

2

−4ηλ + η

2

, f

′′

(λ) = 6λ −4η , (7.79)

hence the extrema of f occur at

λ

1,2

=

η

3

where f

′′

(

η

3

) = −2η < 0 (maximum)

η where f

′′

(η) = η > 0 (minimum)

(7.80)

and are equal to

f(

η

3

) =

4

27

η

3

−1 , f(η) = −1 . (7.81)

It is also obvious from the deﬁnition of f(λ) that f(0) = −1, f(−∞) = −∞, and f(∞) = ∞.

The plot in Figure 7.4 depicts the essential features of f(λ).

0

−1

η/3

η

λ

f

Case I

Case II

Case III

Figure 7.4: Function f(λ) in Rivlin’s cube

In summary, λ

1

= λ

2

= λ

3

= 1 is always a solution. Further,

1. If

4

27

η

3

< 1, then there are no additional solutions (Case I).

2. If

4

27

η

3

= 1, then there is one set of three additional solutions corresponding to λ =

η

3

(Case II).

ME185

D

R

A

F

T

7.5. MULTISCALE PROBLEMS 149

3. If

4

27

η

3

> 1, then there are two sets of three additional solutions corresponding to the

two roots of f(λ) which are smaller than η (Case III).

Note that the root λ = λ

3

> η is inadmissible because it leads to λ

1

= λ

2

= η −λ < 0.

7.5 Multiscale problems

It is sometimes desirable to relate the theory of continuous media to theories of particle

mechanics. This is, for example, the case, when one wishes to analyze metals and semi-

conductors at very small length and time scales, at which the continuum assumption is not

unequivocally satisﬁed. In such cases, multiscale analyses oﬀer a means for relating kinematic

and kinetic information between the continuum and the discrete system.

7.5.1 The virial theorem

The virial theorem is a central result in the study of continua whose constitutive behavior is

derived from an underlying microscale particle system. Preliminary to the derivation of the

theorem, recall that the mean Cauchy stress T in a material region { can be written as

(vol {)

¯

T =

∂P

t ⊗xda −

P

div T⊗xdv . (7.82)

Taking into account (4.83), the preceding equation may be rewritten as

(vol {)

¯

T =

∂P

t ⊗xda +

P

ρ(b −a) ⊗xdv (7.83)

=

∂P

t ⊗xda +

P

ρb ⊗xdv −

d

dt

P

ρ ˙ x ⊗xdv +

P

ρ ˙ x ⊗ ˙ xdv . (7.84)

Deﬁne next the (long) time-average 'φ` of a time-dependent quantity φ = φ(t) as

'φ` = lim

T→∞

1

T

t

0

+T

t

0

φ(t) dt . (7.85)

and note that

d

dt

P

ρ ˙ x ⊗xdv

= lim

T→∞

1

T

¸

P

ρ ˙ x ⊗xdv [

t=t

0

+T

−

P

ρ ˙ x ⊗xdv [

t=t

0

= 0 . (7.86)

ME185

D

R

A

F

T

150 The virial theorem

The preceding time-average vanishes due to the assumed boundedness of the integral

P

ρ ˙ x⊗

xdv throughout time. Using (7.86), the time-averaged counterpart of the mean-stress for-

mula (7.83) is

'(vol {)

¯

T` =

∂P

t ⊗xda

+

P

ρb ⊗xdv

+

P

ρ ˙ x ⊗ ˙ xdv

. (7.87)

Turn attention now to a system of N particles whose motion is governed by Newton’s

Second Law, namely

m

α

¨ x

α

= f

α

, (7.88)

where m

α

and x

α

are the mass and the current position of particle α, while f

α

is the total

force acting on particle α. It is easy to show that the above equation leads to

m

α

d

dt

( ˙ x

α

⊗x

α

) −m

α

˙ x

α

⊗ ˙ x

α

= f

α

⊗x

α

. (7.89)

Taking time averages of (7.89) for the totality of the particles and assuming boundedness of

the term

¸

N

α=1

m

α

˙ x

α

⊗x

α

, it is seen that

−

N

¸

α=1

m

α

˙ x

α

⊗ ˙ x

α

=

N

¸

α=1

f

α

⊗x

α

. (7.90)

Recognizing that, in each particle, the total force f

α

is comprised of an internal part f

ext,α

(due to inter-particle potentials) and an external part f

int,α

, the preceding equation may be

rewritten as

−

N

¸

α=1

m

α

˙ x

α

⊗ ˙ x

α

=

N

¸

α=1

f

int,α

⊗x

α

+

N

¸

α=1

f

ext,α

⊗x

α

. (7.91)

Upon ignoring the body forces in the continuum problem and comparing (7.87) to (7.91),

it can be argued that there is a one-to-one correspondence between the three terms in each

statement. Speciﬁcally, for the mean stress, one may argue that

'(vol {)

¯

T`

.

= −

N

¸

α=1

f

int,α

⊗x

α

, (7.92)

which leads to an estimate of the mean Cauchy stress in terms of the underlying particle

system dynamics as

'(vol {)

¯

T`

.

=

N

¸

α=1

m

α

˙ x

α

⊗ ˙ x

α

+

N

¸

α=1

f

ext,α

⊗x

α

. (7.93)

Equation (7.93) is a statement of the virial theorem.

ME185

D

R

A

F

T

7.6. EXPANSION OF THE UNIVERSE 151

7.6 Expansion of the universe

ME185

D

R

A

F

T

Appendix A A.1

APPENDIX A

A.1 Cylindrical polar coordinate system

The basis vectors of the cylindrical polar coordinate system are ¦e

r

, e

θ

, e

z

¦, where

e

r

= e

1

cos θ +e

2

sin θ

e

θ

= −e

1

sin θ +e

2

cos θ , (A.1)

e

z

= e

3

(A.2)

Here, θ is the angle formed between the e

1

and e

r

vectors. Conversely,

e

1

= e

r

cos θ −e

θ

sin θ

e

2

= e

r

sin θ +e

θ

cos θ . (A.3)

e

z

= e

3

(A.4)

Further, one may easily conclude that

r =

x

2

1

+ x

2

2

, θ = arctan

x

2

x

1

, z = x

3

(A.5)

and, conversely,

x

1

= r cos θ , x

2

= r sin θ , x

3

= z . (A.6)

Using chain rule, one may express the gradient of a scalar function f in polar coordinates

as

grad f =

∂f

∂r

e

r

+

1

r

∂f

∂θ

e

θ

+

∂f

∂z

e

z

=

∂

∂r

e

r

+

1

r

∂

∂θ

e

θ

+

∂

∂z

e

z

f . (A.7)

When the gradient operator grad is applied on a vector function v = v

r

e

r

+v

θ

e

θ

+v

z

e

z

, then

grad

T

v =

∂

∂r

e

r

+

1

r

∂

∂θ

e

θ

+

∂

∂z

e

z

⊗v , (A.8)

or, upon expanding,

gradv =

∂v

r

∂r

e

r

⊗e

r

+

∂v

θ

∂r

e

θ

⊗e

r

+

∂v

z

∂r

e

z

⊗e

r

+

1

r

∂v

r

∂θ

−v

θ

e

r

⊗e

θ

+

1

r

∂v

θ

∂θ

+ v

r

e

θ

⊗e

θ

+

1

r

∂v

z

∂θ

e

z

⊗e

θ

+

∂v

r

∂z

e

r

⊗e

z

+

∂v

θ

∂z

e

θ

⊗e

z

+

∂v

z

∂z

e

z

⊗e

z

. (A.9)

ME185

D

R

A

F

T

A.2 Cylindrical polar coordinate system

Also, given a symmetric tensor function

T = T

rr

e

r

⊗e

r

+ T

rθ

(e

r

⊗e

θ

+e

θ

⊗e

r

) + T

rz

(e

r

⊗e

z

+e

z

⊗e

r

)+

T

θθ

e

θ

⊗e

θ

+ T

θz

(e

θ

⊗e

z

+e

z

⊗e

θ

) + T

zz

e

z

⊗e

z

, (A.10)

its divergence can be expanded to

div T =

¸

∂T

rr

∂r

+

T

rr

−T

θθ

r

+

1

r

∂T

rθ

∂θ

+

∂T

rz

∂z

e

r

+

¸

∂T

rθ

∂r

+

2T

rθ

r

+

1

r

∂T

θθ

∂θ

+

∂T

θz

∂z

e

θ

+

¸

∂T

rz

∂r

+

T

rz

r

+

1

r

∂T

θz

∂θ

+

∂T

zz

∂z

e

z

. (A.11)

ME185

Introduction

undergraduate course on continuum mechanics in the Department of Mechanical Engineering at the University of California, Berkeley. Berkeley, California August 2008

AF

ii

This set of notes has been written as part of teaching ME185, an elective senior-year

T

P. P.

DR

Contents

1 Introduction 1.1 1.2

AF

iii

T

1 1 2 3 3 4 8 14 19 19 28 48 53 61 61 63 66 67 69 70 73

Solids and ﬂuids as continuous media . . . . . . . . . . . . . . . . . . . . . . History of continuum mechanics . . . . . . . . . . . . . . . . . . . . . . . . .

2 Mathematical preliminaries 2.1 2.2 2.3 2.4

Elements of set theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Points, vectors and tensors in the Euclidean 3-space . . . . . . . . . . . . . . Vector and tensor calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Kinematics of deformation 3.1 3.2 3.3 3.4

Bodies, conﬁgurations and motions . . . . . . . . . . . . . . . . . . . . . . . The deformation gradient and other measures of deformation . . . . . . . . . Velocity gradient and other measures of deformation rate . . . . . . . . . . . Superposed rigid-body motions . . . . . . . . . . . . . . . . . . . . . . . . .

DR

4 Basic physical principles 4.1 4.2 4.3 4.4 4.5 4.6 4.7

The divergence and Stokes’ theorems . . . . . . . . . . . . . . . . . . . . . . The Reynolds’ transport theorem . . . . . . . . . . . . . . . . . . . . . . . . The localization theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mass and mass density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The principle of mass conservation . . . . . . . . . . . . . . . . . . . . . . . The principles of linear and angular momentum balance . . . . . . . . . . . . Stress vector and stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .1 7. . 115 Non-linearly elastic solid . . . . . . . . . .3 7. . . . .1 7. 119 Linearly elastic solid . .2 The Gˆteaux diﬀerential . . .3 7. . . . . . . . . . . . . . . . . . . 146 DR 7. .1 7. . . . . . . . . 137 Poiseuille ﬂow . . . . . . . . . . . . . . . . . . . . .11 The Green-Naghdi-Rivlin theorem . . .2.1 7. . . . . . .1. . 141 Simple tension and simple shear . . . . . . . .5 6.4. . . . . . . . . . . . . . .3.3 6. . 101 109 6 Mechanical constitutive theories 6. . . . . . . . 135 Gravity-driven ﬂow down an inclined plane . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . 5 Inﬁnitesimal deformations 5. 125 Viscoelastic solid . . . . . . . . . . . . . . . . . . . . . 143 Multiscale problems . . . . . . . . . . . . .4 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 General requirements . . . . . . . 144 Saint-Venant torsion of a circular cylinder . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . .8 4. . . . .5 Non-linearly elastic solids Compressible Newtonian viscous ﬂuids . . . . . . . . . . . . . . . . . . .4.1 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 Viscous ﬂuids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 7. . . . . . . . . . . . . . . . . .9 The transformation of mechanical ﬁelds under superposed rigid-body motions 85 88 91 95 99 4. . . 140 Stokes’ Second Problem . . . . . . . . .1. .2 7. . . . .3 7. . . . . . . . . . . .2 Incompressible Newtonian viscous ﬂuid . . . . . . . . 100 a AF iv Consistent linearization of kinematic and kinetic variables T The Theorem of Mechanical Energy Balance . . . . . . . . . . . 146 Rivlin’s cube . . . . . . . 109 Inviscid ﬂuids . . . . . . . . . . . . . . . . . . .2 7. . . . . .1. . . . . . . . .and Initial/boundary-value Problems 7. . . . . . . .2 6. . . .3. . . . . . . . . . . . 143 Uniform hydrostatic pressure . . . . . .1 7. . . . 140 Linear elastic solids . . . . . . . . . . . . . . . 149 . . . . . . . 129 135 7 Boundary.4 7. . . . . . . . . . . . . . . . . 4. . . . . . . 135 Couette ﬂow . .10 The principle of energy balance . . . . . . . . . . . . . . . . . 144 . . . . . . . . . . . . . . . . . . . 139 Stokes’ First Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .7. . . . . A. . . 151 A. . 149 Appendix A T v Expansion of the universe . .1 Cylindrical polar coordinate system . . . . . . . . . . . . . . . . . . . . .1 DR AF .1 7. . . . . . . . . . . . . . . . . . . . . . . .1 A. .6 The virial theorem . . . . . . . . . . . .5. . . .

DR AF T vi .

. . . . . Mapping of a body B to its conﬁguration at time t. . . .13 Geometric interpretation of the rotation tensor R by its action on a vector x. . . . 3. . . . . . . A body B and its subset S. . . . . . . . . . . . . .4 3. . . . . . . . . . . . . .3 3. . . . . . . . . . . .5 3. . . . . . .10 Interpretation of the left polar decomposition. . Mapping of an inﬁnitesimal material line elements dX from the reference to the current conﬁguration.7 3. 3. . . . . RMi and associated principal stretches λi . .12 Interpretation of the left polar decomposition relative to the principal directions 3. . . . . . . . . . . . . . . . . . . . . . Interpretation of the right polar decomposition. . . . . . . . . . .15 Image of a sphere under homogeneous deformation. . . . . . . . . . . . . . . . . . . . . . . . . . . .1 2. Example of a set that does not form a linear space . . . .2 2. . . Mapping between two sets . . . . . . . . . AF vii T 4 5 10 19 20 21 22 25 26 29 30 35 36 38 39 42 44 46 Schematic depiction of referential and spatial mappings for the velocity v. . . . . . . . . . . . . . . . . . . . DR 3. Application of the inverse function theorem to the motion χ at a ﬁxed time t. . . . . .8 Schematic depiction of a set . .11 Interpretation of the right polar decomposition relative to the principal directions MA and associated principal stretches λA . . . . . . . . . . . . . .3 3. . . . . . . . . 3. . . . . . . . . . . . . 3. . . . . . . . . . . . . . . .6 3. . . . . . . . . . . . . .1 3. .14 Spatially homogeneous deformation of a sphere. . . Mapping of a body B to its reference conﬁguration at time t0 and its current conﬁguration at time t. . . . . . . . . . . . . . . . . . . . . .List of Figures 2. . . . . . . . . . . . . . . . . . Particle path of a particle which occupies X in the reference conﬁguration. . .2 3. . . . . . . . . . . . . . . .9 3. . . . . Stream line through point x at time t. . . . . . . . . . .

.1 4. . . . . . . . . . . . . . . . . . . 54 63 63 68 73 75 47 A region P with boundary ∂P and its image P0 with boundary ∂P0 in the A limiting process used to deﬁne the mass density ρ at a point x in the current Traction acting on a surface of an inviscid ﬂuid. . . .6 4. . 122 Static continuation Etδ (s) of Et (s) by δ. . . 3. . . . . . . . . . .2 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 DR 7. . . . . . . .4 6. . . . . . . . . . . . Setting for a derivation of Cauchy’s lemma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 6. . . . . . . . . . . . . . . . .3 7. . . . . . . . . . . . . 115 Orthogonal transformation of the reference conﬁguration. . . .5 A surface A bounded by the curve C. . . . . . . . . . 79 80 Projection of the traction to its normal and tangential components. . . . . 141 Function f (λ) in Rivlin’s cube . .1 7. . . . . . . .4 4.16 Mapping of an inﬁnitesimal material volume element dV to its image dv in 3. . . . . . . . . reference conﬁguration. . . .3 4. . . . . .17 Mapping of an inﬁnitesimal material surface element dA to its image da in the current conﬁguration. . . . . 135 Couette ﬂow . . . . . . . . . . . . . . 131 An interpretation of relaxation . . . . . . . .2 4.3. . . . . .5 4. . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Interpretation of the Cauchy stress components on an orthogonal parallelepiped aligned with the {ei }-axes. . . . . . . . . . . . . . . . . . . . 111 A ball of ideal ﬂuid in equilibrium under uniform pressure. . . . . conﬁguration. . . . . . . . . . . . . . . . . . 138 Semi-inﬁnite domain for Stokes’ Second Problem . . .4 viii AF T the current conﬁguration. . . . . .3 6. . . . . . . . . . . . The Cauchy tetrahedron. . . . . . . . .2 7. . . . 132 Flow down an inclined plane . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 Conﬁgurations associated with motions χ and χ+ diﬀering by a superposed ¯ rigid motion χ+ . . . . . . . . . . . . . . . . . . .

List of Tables DR AF ix T .

precise composition.. As long as the physical problems of interest occur DR at length and time scales of several orders of magnitude higher that those noted previously. As a thought experiment. which in turn consist of ﬁnite numbers of nuclei and electrons. However.g. deformation). as it is comprised of distinct building blocks. No matter how small it gets. the molecules. its physical properties remain unaltered. Each molecule consists of a ﬁnite number of atoms. temperature. respectively.Chapter 1 Introduction 1. one may broadly estimate such characteristic lengths ˚ and times to be of the order of up to a few Angstroms (1 A= 10−10 m) and a few femtoseconds (1 femtosecond = 10−15 sec). namely to eﬀectively ignore its discrete nature without introducing any even remotely signiﬁcant errors. A continuous medium may be conceptually deﬁned as a ﬁnite amount of matter whose physical properties are independent of its actual size or the time over which they are measured. one may choose to perpetually dissect a continuous medium into smaller pieces. The preceding characteristic lengths and times can vary considerably depending on the state of the matter (e. This is generally the case when matter is considered at length scales much larger than the characteristic length of the atomic spacings and at time scales much larger than the characteristic times of atomic bond vibrations. Many important physical phenomena involve matter in large length and time scales. 1 AF T .1 Solids and ﬂuids as continuous media All matter is inherently discontinuous. it is possible to consider matter as a continuous medium.

and the increasing availability (especially since the late 1960s) of high-performance computers. This “renaissance” period can be attributed to several factors. Naghdi and many others.C.2 History of continuum mechanics Mathematical theories developed for continuous media (or “continua”) are frequently referred without directly accounting for the discrete structure of matter. such as the ﬂourishing of relevant mathematics disciplines (particularly linear algebra. Continuum mechanics as taught and practiced today emerged in the latter half of the 20th century. Continuum mechanics is concerned with the response of solids and ﬂuids under external loading precisely when they can be viewed as continuous media. It is motivated by the imagination and creativity of da Vinci (1452-1519) and the rigid-body experiments of Galileo (15641642). later set on ﬁrm theoretical ground by Euler (1707-1783) and further developed and reﬁned by Cauchy (1789-1857). two of the oldest and most widely examined disciplines in applied science. partial diﬀerential DR ME185 equations and diﬀerential geometry). which emphasizes generality. the advances in materials and mechanical systems technologies. in the sense that they capture the observed physical response . Ericksen. A wave of gifted modern-day mechanicians. AF T to as “phenomenological”. It is founded on the laws of motion put forward by Newton (1643-1727).) on the law of the lever and on hydrostatics. contributed to the rebirth and consolidation of classical mechanics into this new discipline of continuum mechanics.2 History of continuum mechanics Continuum mechanics is a modern discipline that uniﬁes solid and ﬂuid mechanics. It draws on classical scientiﬁc developments that go as far back as the Hellenistic-era work of Archimedes (287212 A. yet derives its essential features from the physics of material behavior. 1. rigor and abstraction. Truesdell. Solids and ﬂuids (including both liquids and gases) can be accurately viewed as continuous media in many occasions. such as Rivlin.

4. / every element of X is also an element of Y . If not. For example. X = {all integers greater than 0 and less than 6} .1 Elements of set theory This section summarizes a few elementary notions and deﬁnitions from set theory. 2. 5} AF N = {all positive integers} R = {all real numbers} 3 T (2.2) (2.4) or DR and one element of Y that does not belong to X.Chapter 2 Mathematical preliminaries 2. one writes x ∈ X. Y be two sets. one writes x ∈ X. but there exists at least . The set X is a proper subset of the set (denoted Let X. Two sets of particular interest in the remainder of the course are: If x is an element of the set X. A set X is a collection of objects referred to as elements. X = {1. 3. The set X is a subset of the set Y (denoted as X ⊆ Y or Y ⊇ X) if as X ⊂ Y or Y ⊃ X) if every element of X is also an element of Y .1) (2.3) (2. A set can be deﬁned either by the properties of its elements or by merely identifying all elements.

.5) Note that the pair (x. y) is.2 Vector spaces and a scalar multiplication operation (·). . β ∈ R: A linear (or vector) space {V. visualized in Figure 2. tions. The empty set (denoted by ∅) is a set that contains no elements and is contained in every set. i. v. w ∈ V and α. ME185 AF Figure 2. Assume that V is endowed with an addition operation (+) Consider a set V whose members (typically called “points”) can be scalars. vectors or func- DR V u. not the same as the element (y. therefore.. .e. The intersection of sets X and Y (denoted by X ∩ Y ) is a set which (2. the element (x. R. . X × X × X. ·} is deﬁned by the following properties for any (i) α · u + β · v ∈ V (closure).. 2. The Cartesian product X × Y of sets X and Y is a set deﬁned as X × Y = {(x. .1: Schematic depiction of a set . T A "point" that belongs to V elements of both sets. (ii) (u + v) + w = u + (v + w) (associativity with respect to + ). X ∪ ∅ = X. The notation X 2 . x). . which do not necessarily coincide with the classical addition and multiplication for real numbers.. in general. y ∈ Y } .4 Vector spaces The union of sets X and Y (denoted by X ∪ Y ) is the set which is comprised of all includes only the elements common to the two sets. is used to respectively denote the Cartesian products X × X. y) in the preceding equation is ordered.1. X 3 . +. y) such that x ∈ X.

It can be trivially veriﬁed that {P2 . y) ∈ R2 | x2 + y 2 = 1} with the standard addition and scalar multiy u+v plication for vectors. c) ∈ R3 . (2) Deﬁne V = {(x.Mathematical preliminaries (iii) ∃ 0 ∈ V | u + 0 = u (existence of null element). (vii) (α + β) · u = α · u + β · u (distributivity with respect to R). ME185 AF v u x T . Examples: (1) V = P2 := {all second degree polynomials ax2 + bx + c} with the standard polynomial addition and scalar multiplication.2. b. y2 ) as in Figure 2. ·} is a linear function space.2: Example of a set that does not form a linear space property (i) is violated. y1 + y2 ) . +. (vi) (αβ) · u = α · (β · u) (associativity with respect to ·). since. (ix) 1 · u = u (existence of identity). u + v = (x1 + x2 . | u + (−u) = 0 (existence of negative element). Notice that given u = (x1 . i. in general. DR 1 Figure 2. for α = β = 1. (viii) α · (u + v) = α · u + α · v (distributivity with respect to V).. y1 ) and v = (x2 . 5 (iv) ∃ − u ∈ V (v) u + v = v + u (commutativity). P2 is also “equivalent” to an ordered triad (a. R.e.

·}. for any u. ·} is a ﬁnite A basis of an n-dimensional vector space {V.. If the above statement is not true. R. {V.. +. {v1 . + αn gn + βv = 0 ⇔ not all α1 . ·} is deﬁned as any set of n linearly DR v ∈ V. R. Then. ∈ R α·u + β·v ∈ U . (2. v ∈ U and α. In order to simplify the notation.. then given any non-zero α1 g1 + α2 g2 + . . R. .8) AF T . +. +. R. ·} is not a linear space. . In this case. . If {g1 . P2 is a linear subspace of {Pn . ·} is inﬁnite-dimensional if. R. . . all linearly independent sets contain at most n elements. Consider a linear space {V. . (2. +. v = − ME185 independent vectors. Example: i. . ·} with the usual polynomial addition and scalar multiplication. = αp = 0 . in the remainder of these notes the symbol · used in Let v1 · v2 . .e.. n-dimensional). . . +. vp be elements of the vector space {V. . .. gn } are not linearly independent. v2 . − gn .6 Vector spaces and (x1 + x2 )2 + (y1 + y2 )2 = 1. given any n ∈ N. n. R.7) More speciﬁcally. β equal zero . such that dimensional vector space (speciﬁcally. ·} and assume that α1 v1 + α2 v2 + . +.. Then. β. R. which would have implied that {g1 . β = 0 because otherwise there would be at least one non-zero αi . . R. The vector space {V. . . i = 1. it contains at least one linearly independent set with n + 1 elements. ·}. +. ·} and a subset U of V. g2 . Thus.. αn . Thus. . {V. closure is maintained within U. gn } form a basis in {V... +.6) scalar multiplication will be omitted. . if. the non-zero vector v can be expressed as α2 αn α1 g1 − g2 − . g2 . + αp vp = 0 ⇔ α1 = α2 = . +. β β β (2. . then there is an n ∈ N. vp } is a linearly independent set in V. R. of V with respect to the same operations (+) and (·)... . Then U forms a linear sub-space (a) Deﬁne the set Pn of all algebraic polynomials of degree smaller or equal to n > 2 and consider the linear space {Pn . ..

since {g1 . v. + γn + αn β gn . w ∈ V and α ∈ R. if..Mathematical preliminaries 7 The above representation of v in terms of the basis {g1 . if ui · uj = The standard dot-product between vectors in Rn satisﬁes the above properties.. . g2 . Of all the vector spaces. i = 1.11) Two vectors u. v ∈ E n are orthogonal. + γn gn . gn } are assumed to be a linearly β . such that for any u... alternatively. u2 . . (x) u · v = v · u (commutativity with respect to ·). it follows that 0 = γ1 + α1 β g1 + γ2 + (2. This vector operation is referred to as the dot-product. An n-dimensional vector space obeying the above additional rules is referred to as a Euclidean vector space and is denoted DR by E n . upon subtracting the preceding two equations from one another.9) then. The dot-product provide a natural means for deﬁning the magnitude of a vector as = (u · u)1/2 .. . . . .} is called (2.10) independent.12) where δij is called the Kronecker delta symbol. if u · v = 0. (2. . . g2 . (xi) u · (v + w) = u · v + u · w (distributivity). (2. (xii) (αu) · v = u · (αv) = α(u · v) (associativity with respect to ·) (xiii) u · u ≥ 0 and u · u = 0 ⇔ u = 0... . . Indeed. 0 if i = j 1 if i = j = δij . A set of vectors {u1 . ME185 AF which implies that γi = − αi . T α2 β g2 + . attention will be focused here on the particular class of Euclidean vector spaces in which a vector multiplication operation (·) is deﬁned. gn } is unique. . 2. v = γ1 g1 + γ2 g2 + . Example: u orthonormal. n..

. vectors and tensors in the Euclidean 3space Consider the Cartesian space E 3 with an orthonormal basis {e1 . T n if (2. . k ≤ n in E n is linearly independent. e2 ..3 Points. 2. where vi = v · ei . Also.14) Of particular importance to the forthcoming developments is the observation that any vector v ∈ E n can be resolved to an orthonormal basis {e1 . Next. α1 e1 + α2 e2 + . Further..15) basis {e1 .ek }. one makes a unique association (to within the speciﬁcation of distance between x and y. e2 . In this case.16) ambient three-dimensional space. + vn en = vi ei . . α1 (e1 · ei ) + α2 (e2 · ei ) + . i = 1. e2 . In this way. and invoking the orthonormality of {e1 . a typical vector vǫE 3 can be written as vi ei .... namely d(x. which is the set of all points in the It is now possible to deﬁne vectors x. y. consider points x. y ∈ E 3 . ..8 Points. .17) ME185 .13) then. . k... upon taking the dot-product of the above equation with ei . . y in the Euclidean point space E 3 . This is because. e3 }.ek }.. . . . (2. . + αk (ek · ei ) = αi = 0 . . it is possible to deﬁne a measure of (2. en }. en } as v = v1 e1 + v2 e2 + . which originate at O and end at points x and O) between points in E 3 and vectors in E 3 . vectors and tensors in the Euclidean 3-space Every orthonormal set {e1 . vi = v · ei . when taken to be devoid of the mathematical structure of vector spaces. As argued in the DR v = i=1 previous section. e2 . e2 . (2. y) = |x − y| = [(x − y) · (x − y)]1/2 . + αk ek = 0 . vi denotes the i-th component of v relative to the orthonormal 2. . consider an arbitrary origin (or reference point) O in the same space. respectively. AF i=1 3 (2. by way of the magnitude of the vector v = y − x.

and any scalar α. v.2.2) . namely that the vector Appealing to property (a). (2. the following e1 × e2 = e3 . . j.1). equivalently [uvw] = [vwu] = [wuv]. or. relations hold true: (a) and (b) can be used to deduce that (u × v) · u = (u × v) · v = 0. properties By deﬁnition.1. for a right-hand orthonormal coordinate basis {e1 .22) 1 0 −1 if (i. k) = (2.21) symbol deﬁned as if (i. (b) (u × v) · w = (v × w) · u = (w × u) · v. or (1. one may deﬁne the vector product of two vectors as an operation with the following (a) u × v = −v × u.2. e3 × e1 = e2 . (3. where [uvw] = (u × v) · w is the triple product of vectors u.1).Mathematical preliminaries 9 In E 3 . k=1 3 3 3 3 properties: for any vectors u.1.3. e2 . it is readily concluded that u × u = 0. and w. e3 }.19) and e2 × e1 = −e3 can be expressed compactly as (2. v and w. Likewise. cos θ = (c) |u × v| = |u||v| sin θ u·v .3). e2 × e3 = e1 .3.3).18) These relations. (2. (2.20) DR ei × ej = It follows that u×v = ( where ǫijk is the permutation ǫijk = 3 3 (2. (2. j. together with the implications of property (a) AF e1 × e1 = e2 × e2 = e3 × e3 = 0 .2) (2. 0 ≤ θ ≤ π.23) i=1 j=1 k=1 ME185 . or (3. otherwise 3 i=1 uiei ) × ( vj ej ) = j=1 i=1 j=1 ui vj ei × ej = ui vj eijk ek . e1 × e3 = −e2 ǫijk ek . k) = (1. e3 × e2 = −e1 3 T (u · u)1/2 (v · v)1/2 . u × v is orthogonal to both u and v.

AF U V Figure 2. and is (2) T : E 3 → E 3 .10 Points. v ∈ E 3 and α.3: Mapping between two sets T v point u ∈ U a unique point v = f (u) ∈ V. This is called the identity tensor. whereas V is termed the range of f . Using the Cartesian components of vectors. vectors and tensors in the Euclidean 3-space Let U. T(v) = v for all v ∈ E 3 . This implies that.25) for any vector u ∈ E 3 . DR typically denoted by T = I. V be two sets and deﬁne a mapping f from U to V as a rule that assigns to each domain of f . according to the relation ME185 (1) T : E 3 → E 3 . β ∈ N. denoted by T = 0. . u → v = f (u) ∈ V. T(v) = 0 for all v ∈ E 3 . With reference to the above setting. U is called the f u A mapping T : E 3 → E 3 is called linear if it satisﬁes the property T(αu + βv) = αT(u) + βT(v) . A linear mapping is also referred to as a tensor.24) Examples: for all u. This is called the zero tensor.3. under the action of the tensor product v ⊗ w. (2. It can be easily veriﬁed that v ⊗ w is a tensor according to the previously furnished deﬁnition. The usual notation for a mapping (2. is: f : U → V . see Figure 2. the vector u is mapped to the vector (w · u)v. and is typically The tensor product between two vectors v and w in E 3 is denoted by v ⊗ w and deﬁned (v ⊗ w)u = (w · u)v .

This index substitution property is frequently used in component manipulations. 4. ME185 AF i=1 3 3 i=1 j=1 summation convention. Such an index is referred to as dummy. Einstein. DR Examples: 3 1. The tensor product u ⊗ v = ui vj ei ⊗ ej and it involves the summation of nine terms. j = 1. Such an index is referred to free. An index which appears once in a single component or in a product expression is not summed and is assumed to attain a single value (1. which will greatly simplify the component representation of both ui ei is replaced by u = ui ei and it involves the ui vj ei ⊗ ej is equivalently written as u ⊗ v = . (2. i. 2. The summation convention has three rules. Rule 3. The vector representation u = summation of three terms.26) It will be shown shortly that the set of nine tensor products {ei ⊗ ej }. It is easy to see that δij ui = δ1j u1 +δ2j u2 +δ3j u3 = uj . If an index appears twice in a single component term or in a product term. This originates with A. E 3 ) of all tensors on E 3 . which. 3. Rule 2. when adapted to the special case of E 3 . 2. The term ui vj is a single term with two free indices i and j. No index can appear more than twice in a single component or in a product term. the summation sign is omitted and summation is automatically assumed from value 1 to 3. who employed it ﬁrst in his relativity work. 3. or 3). form a basis for the space L(E 3 . are as follows: Rule 1. 2.Mathematical preliminaries one may express the tensor product of v and w as 3 11 v⊗w = ( i=1 vi ei ) ⊗ ( T 3 3 3 wj ej ) = j=1 i=1 j=1 vi wj ei ⊗ ej . Before proceeding further with the discussion of tensors. it is expedient to introduce a vectorial and tensorial quantities and their algebra.

A similar index substitution property applies in the case of a two-index quantity.28) This derivation demonstates that any tensor T can be written as a linear combination of the nine tensor product terms {ei ⊗ ej }. It follows that (2. T31 T32 T33 T11 T12 T13 u · Tv = v · TT u .31) where Aij are the components of TT . = Tek vk − Tij vk (ei ⊗ ej )ek T namely δij aik = δ1j a1k + δ2j a2k + δ3j a3k = ajk .32) . [Tij ] = T21 T22 T23 .12 Points. this implies that (2.30) for any vectors u and v in E 3 . (2. (2. ui Tij vj = vi Aij uj = vj Ajiui . T = Tij ei ⊗ ej . Using components. E 3 ) and deﬁne its (T − Tij ei ⊗ ej )v = (T − Tij ei ⊗ ej )vk ek AF = Tik ei vk − Tij vk (ej · ek )ei = Tik ei vk − Tij vk δjk ei = Tik ei vk − Tik vk ei = 0.29) The transpose TT of a tensor T is deﬁned by the property (2. ui (Tij − Aji)vj = 0 .27) hence. It follows that With the summation convention in place. such that Tej = Tij ei . vectors and tensors in the Euclidean 3-space 5. take a tensor T ∈ L(E 3 . The components of the tensor T relative to {ei ⊗ ej } can be put in matrix form as DR ME185 (2. components Tij .

37) hence. written as (2.37) as tr T = tr(Tij ei ⊗ ej ) = Tij ei · ej = Tij δij = Tii . E 3 ) → R of two tensors T and (2. this implies that Aij = Tji . in this case. (2. T = T(Sij vk δjk ei ) = T(Sij vj ei ) = Tki Sij vj ei = (Tki Sij ek ⊗ ej )(vl el ) . (TS)v = T(Sv) = T[(Sij ei ⊗ ej )(vk ek )] (TS) = TkiSij ek ⊗ ej .40) ME185 . a tensor T is skew-symmetric if TT = −T or. Likewise. (2. Note that. this implies that (2. T · S = tr(TST ) . E 3 ) × L(E 3 .35) which leads to (2.38) The contraction (or inner product) T · S : L(E 3 . The trace tr T : L(E 3 . E 3 ) → R of the tensor product of two vectors u ⊗ v is deﬁned as (2. Given tensors T. Tji = −Tij . tr u ⊗ v = u · v .39) tr(TST ) = tr(Tki Sji ek ⊗ ej ) = Tki Sji ek · ej = Tki Sji δkj = Tki Ski .34) for any v ∈ E 3 .36) DR S is deﬁned as Using components. hence the transpose of T can be TT = Tij ej ⊗ ei . upon resolving both on the same basis. (2. the trace of a tensor T is deduced from equation (2. when both T and TT are resolved relative to the same basis. In component form.Mathematical preliminaries 13 Since ui and vj are arbitrary.33) A tensor T is symmetric if TT = T or. again. Tji = Tij . T11 = T22 = T33 = 0. E 3 ). the multiplication TS is deﬁned according to AF (TS)v = T(Sv) . S ∈ L (E 3 .

t → φ = φ1 (t) φ2 : E 3 → R . TT−1 = I. t) → φ = φ3 (x. similarly.46) DR 2. one writes T . Clearly. the tensors T and S are invertible. for any w ∈ E 3 .42) and T−1 is the inverse of T. further.43) hence T−1 T = I and. the equation Tv = w can be uniquely solved for v. TT = T−1 . (2. if T−1 exists. Then. (ST)T = TT ST .44) which implies that (2. t) . x → φ = φ2 (x) φ3 : E 3 × R → R . (ST)−1 = T−1 S−1 . φ1 : R → R . S ∈ L(E 3 . (2. (S + T)T = ST + TT (2.47) Vector and tensor calculus Deﬁne scalar.4 scalar functions are of the form ME185 If. (x. E 3 ). A tensor T is orthogonal if (2. (2. TT T = TTT = I .41) v = T−1 w .48) . then T−1 w − v = 0 AF = T−1 (Tv) − v = (T−1 T)v − v = (T−1 T − I)v . vector and tensor functions of a vector variable x and a real variable t. then (2.14 Vector and tensor calculus A tensor T is invertible if.45) It can be shown that for any tensors T. The (2.

51) for any v ∈ E 3 . Using the chain rule. E 3 ) . Its gradient is ∂xj ∂xj xj + xj ∂xi ∂xi ei = (δij xj + xj δij )ei = 2xi ei = 2x . t → v = v1 (t) 15 v2 : E 3 → E 3 . w=0 (2. ∂xi (2. E 3 ) . x → T = T2 (x) (grad φ(x)) · v = = w=0 T1 : R → L(E 3 . t) → T = T3 (x. this leads to the expression grad = ∇ = ∂ ei . (2. t) and AF T2 : E 3 → L(E 3 .52) DR Hence. t) .49) v3 : E 3 × R → E 3 .55) ME185 . in component form one may write grad φ = ∂ ∂(xj xj ) (x · x) = ei = ∂x ∂xi grad φ(x) = ∂φ(x) ei . (x. (x.54) Example: Consider the scalar function φ(x) = |x|2 = x · x. φ = φ(x) is a vector deﬁned by The gradient grad φ(x) (otherwise denoted as ∇φ(x) or (2.51) becomes d φ(x + wv) dw ∂φ(x + wv) d(xi + wvi ) ∂(xi + wvi) dw = w=0 ∂φ(x) vi . E 3 ) .53) In operator form. t → T = T1 (t) (2. ∂φ(x) ) of a scalar function ∂x respectively. the right-hand side of equation (2.50) T3 : E 3 × R → L(E 3 .Mathematical preliminaries while the vector and tensor functions are of the form v1 : R → E 3 . ∂xi (2. x → v = v2 (x) T d φ(x + wv) dw . t) → v = v3 (x. ∂xi (2.

Its gradient is ∂(αxi ) ∂(αx) = ei ⊗ ej = αδij ei ⊗ ej = αei ⊗ ei = αI . ∂xj ∂ ⊗ ej . using chain rule. using components.57) becomes d v(x + ww) dw = w=0 ∂vi (x + ww) d(xj + wwj ) ∂(xj + wwj ) dw ei = w=0 ∂vi (x) wj ei .59) In operator form.61) since (ei ⊗ ei )v = (v · ei )ei = vi ei = v.16 Vector and tensor calculus Alternatively. ∂xj (2.56) v = v(x) is a tensor deﬁned by the relation (grad v(x))w = The gradient grad v(x) (otherwise denoted as ∇v(x) or d v(x + ww) dw AF . d (v + ww) dw = αw .58) hence.57) for any w ∈ E 3 . w=0 (2. (2. using directly the deﬁnition (grad φ) · v = = = [2x · v + 2wv · v]w=0 = 2x · v T w=0 d {(x + wv) · (x + uv)} dw w=0 d {x · x + 2wx · v + w 2 v · v} dw (2.63) . ∂xi (2. this leads to the expression DR grad v = (grad v)w = hence grad v = αI. Again.60) Example: Consider the function v(x) = αx. the right-hand side of equation (2.62) The divergence div v(x) (otherwise denoted as ∇ · v(x)) of a vector function v = v(x) is div v(x) = tr(grad v(x)) . grad(v(x) = ∂vi (x) ei ⊗ ej . a scalar deﬁned as ME185 grad = ∇ = (2. using directly the deﬁnition. Alternatively. ∂x ∂xj (2. w=0 ∂v(x) ) of a vector function ∂x (2.

div T) = div(TT c) = div[(Tij ej ⊗ ei )(ck ek )] = div[Tij ck δik ej ] = div[Tij ci ej ] = tr ∂Tij ci ej ⊗ ek ∂xk ∂(Tij ci ) = δjk ∂xk ∂(Tij ci ) = ∂xj ∂Tij = ci .64) In operator form. ∂xi ∂xi (div T(x)) · c = div (TT (x))c .65) Example: Consider again the function v(x) = αx. ∂xi 17 T ∂Tij ei . ∂xj (2.Mathematical preliminaries on. one writes div = ∇· = (2.66) is a vector deﬁned by the property that The divergence div T(x) (otherwise denoted as ∇ · T(x)) of a tensor function T = T(x) (2.67) for any constant vector c ∈ E 3 . Its divergence is div v(x) = AF ∂xi ∂(αxi ) = α = αδii = 3α . ∂xj ∂xj ∂xi ∂ · ei . (2. DR hence. using components. ∂xj div T = (2.68) (2.i . div v(x) = tr ∂vi ei ⊗ ej ∂xj = ∂vi ∂vi ∂vi ei · ej = δij = = vi. Using components.69) ME185 .

the curl is expressed as (2. which translates using components to curl v(x) = ∂ ei ∂xi × (vj ej ) = (2.70) Finally.71) AF curl = ∇× = ∂vj ∂vj ∂vk ei × ej = eijk ek = eijk ei .72) In operator form. the curl curl v(x) of a vector function v(x) is a vector deﬁned as curl v(x) = ∇ × v(x) . the operator form becomes T ∂ ei .18 Vector and tensor calculus In the case of the divergence of a tensor. ∂xi ∂xi ∂xj (2.73) DR ME185 . ∂xi ∂ ei × . ∂xi div = ∇· = (2.

1) ¯ ¯ In the above. when conpendent of its actual size or the time over which they are measured.Chapter 3 3. t) ∈ B × R → E 3 the motion of B. (3. Then. see Figure 3. t) = χt (P ) . deﬁne by χ : (P. endow the body with local (pointwise) physical properties which are inde- DR mapping. and let x be ¯ space E 3 .1. P sidered together.1 Bodies. such that its associated position vector relative to the origin O of an orthonormal basis in the vector Let x be the point in E 3 occupied by a particle P of the body B at time t. while an arbitrary subset of B be denoted by S. Also.1: A body B and its subset S. 19 AF S B Figure 3. Given χ. conﬁgurations and motions Let a continuum body B be deﬁned as a collection of material particles. which. which is a diﬀerentiable ¯ ¯ x = χ(P. χt : B → E 3 is called the conﬁguration mapping of B at time t. let a typical such particle be denoted by P . Kinematics of deformation T .

2. is of limited practical use. t) with boundary ∂R at time t. R and P are point sets in E 3 . P ¯ χ Figure 3. S T x ∂R P ∂P ¯ Likewise. conﬁgurations and motions ¯ the body B may be mapped to its conﬁguration R = χt (B. t) . as well as time. and refer although in many practical problems it is guided by the need for mathematical simplicity. given any point x ∈ P. For this reason. Now. t) with boundary (3.4) . namely ME185 ¯ The mapping χ represents the material description of the body motion. denote the point which P occupies at time t0 as X and let this point be associated AF B R ¯ ¯ X = χ(P. This description. t0 ) = χt0 (P ) .20 Bodies. This is because ¯ the domain of χ consists of the totality of material particles in the body. t) ∂t . two alternative descriptions of the body motion are introduced below. ∂P at time t.2) ¯ The motion χ of the body is assumed to be twice-diﬀerentiable in time. say R0 = χ(B. Then. because there is no direct quantitative way of tracking particles of the body. ¯ Of all conﬁgurations in time.3) DR with position vector X. Clearly.2: Mapping of a body B to its conﬁguration at time t. to it as the reference conﬁguration. select one. one may deﬁne the velocity and acceleration of any particle P at time t according to v = ¯ ∂ χ(P. any part S ⊂ B can be mapped to its conﬁguration P = χt (S. The choice of reference conﬁguration may be arbitrary. (3. ¯ The conﬁguration mapping χt is assumed to be invertible. t0 ) at a time t = t0 . which means that. ¯ −1 P = χt (x) . although mathematically proper. a = ¯ ∂ 2 χ(P. see Figure 3. ∂t2 (3.

Also. (3. E2 . E3 } and {e1 .5) x = χ(X.7) associated with the reference and current conﬁguration. respectively. it is implicit that a reference conﬁguration is provided. one may write the position vectors X and x corresponding to the points occupied by the particle P at times t0 and t as (3. With reference to the Assume now that the motion of the body B is described with reference to the conﬁguration respectively. Hence.6) represents the referential or Lagrangian description of the body motion. where 21 T ¯ χt X x χt R0 R . e2 .3: Mapping of a body B to its reference conﬁguration at time t0 and its current DR X = XA E A may be expressed using components as R0 deﬁned at time t = t0 and let the conﬁguration of B at time t be termed the current conﬁguration. The mapping χt is the placement (3.Kinematics of deformation Thus. e3 } be ﬁxed right-hand orthonormal bases preceding bases.3.8) ME185 . t) = χt (X) (3. tion. Figure 3. the motion χ AF ¯ χt0 B xi ei = χi (XA EA . see Figure 3. t) . one may write ¯ ¯ t0 x = χ(P. t) = χ(χ−1 (X). t) = χ(X. In such a descripof the body relative to its reference conﬁguration. x = xi ei . let {E1. resolving all relevant vectors to their respective bases. The mapping χ : E 3 × R → E 3 . t)ei . P conﬁguration at time t.

t). one may write (3. t) = f (χt (X). Likewise. AF a = ai (XA . t) . Any function (not necessarily scalar) of space and time can be written equivalently in material.12) In analogous fashion. t) = f(χ−1 (x). expressed in the referential description. conﬁgurations and motions or. where the independent variables are the current position vector x and ˇ time t. starting. t) χ (x. t) . with a scalar function f = f(P. take the form v = ∂χ(X. t)ei .11) Scalar.14) Figure 3. t) = f (X. t) . one may write (3. t) = f (x. t) ∂t . (X.9) (3. t) . t)ei .22 Bodies. Indeed. The velocity and acceleration vectors. ∂t2 (3.10) respectively. v = vi (XA . by way of an example. t) = f(X. t) = f (x. ˇ ˜ ˆ f = f (P.13) The above two equations can be combined to write DR ME185 (3. Focusing speciﬁcally on the referential and spatial . t) . in pure component form. a = ∂ 2 χ(X. referential or spatial form. using the orthonormal bases. ˇ ˇ ˜ f = f (P. vector and tensor functions can be alternatively expressed using the spatial or Eulerian description.4: Schematic depiction of referential and spatial mappings for the velocity v. t) ˜ v v ˆ v T xi = χi (XA . t ˜ ˜ ˆ f = f (x. (3.

t) ˜ ∂f = + ·v ∂t ∂x ˜ ∂ f (x. t). t) . deﬁne the material time derivative of f as ˆ ∂ f (X.16) ˆ Given a function f = f (X. (3. it is easily seen that the velocity and acceleration vectors can be equivalently ˆ ˜ v = v(X. t) ∂ v(x. t) . t) + v = ∂t ∂x ˜ ∂ v(x.. t) ∂χ(X. The second term is called the material time derivative as the material particle which occupies the point x at time t tends to travel away from x with velocity v. t)ei ˆ ˜ . t) ∂χ(X. t) + ∂t ∂x ∂t ˜ ˜ ∂ v(x. t) ∂ f (x. t) ∂ f (x. In component form. ˆ ˜ expressed as (3. A similar expression for the material time derivative convective rate of change of f and is due to the spatial variation of f and its eﬀect on the ˜ applies for vector functions.18) is the spatial time derivative of f and corresponds to the rate of change of f for a ﬁxed point x in space.18) DR The ﬁrst term on the right-hand side of (3.e. t). for example. t) + · f˙ = ∂t ∂x ∂t ˜(x. t) = a(x. t) T . = ∂t (3. t) ˜ + grad f · v . see Figure 3. t) ˜ + (grad v)v .19) ME185 . = ∂t AF ˙ v = ˜ ˜ ∂ v(x. take. alternatively. then ˜ ˜ ∂ f (x. t)ei = vi (xj .15) respectively. f = f (x. t) ∂ v(x. Indeed. ˜ If. t) and write (3. t) = v(x. i. the velocity v = v(x. a = ai (XA . f is expressed in spatial form.4. f˙ = ∂t (3. t)ei .Kinematics of deformation 23 descriptions. t)ei = ai (xj . one may write v = vi (XA . ˆ ˜ a = a(X.17) From the above deﬁnition it is clear that the material time derivative of a function is the rate of change of the function when keeping the referential position X (therefore also the particle P associated with this position) ﬁxed.

because it contains the same material particles at all times. T is termed material. t) = F (X) = 0 . This distinction becomes less apparent when a surface is deﬁned in spatial form. On the other hand. ∂F (X.22) DR and ME185 which implies that F = F (X).e. t) = 0 . a surface described by the equation F (X. because the locus of its points contains diﬀerent material particles at diﬀerent times. . t) = 0 t −1 G(X) = G(χt (x)) = g(x. Lagrange’s criterion of materiality (1783) A surface described by the equation f (x. Speciﬁcally. By way of example. t) = F (X) = 0 . one may employ Lagrange’s criterion of materiality. if the criterion holds.20) hence AF ˙ f˙(χ. and only if.. i. It is easy to show ˙ that this is a suﬃcient. t) = f˙(χ(X. This is clearly a material surface. t) = ∂t F (X) = F (χ−1 (x)) = f (x. conﬁgurations and motions A volume/surface/curve which consists of the same material points in all conﬁgurations be expressed in the form F (X) = 0. but not a necessary condition for the materiality of a curve. (3. hence the surface is material. which states the following: by an equation f (x. consider a surface in three dimensions. A similar analysis applies to curves in Euclidean 3-space.24 Bodies. say F (X) = 0 and G(X) = 0. t) = 0 is generally non material. a material curve can be be viewed as the intersection of two material surfaces. then (3.23) (3. which can (3. since its referential representation is independent of time. t) = 0 is material if. Switching to the spatial description and expressing these surfaces as (3. t) = 0. In this case. A sketch of the proof is as follows: if the surface is assumed material.24) it follows from Lagrange’s criterion that a curve is material if f˙ = g = 0. then f (x. t).21) Conversely. f˙ = 0. t) = 0 . f˙(χ.

Kinematics of deformation 25 Some important deﬁnitions regarding the nature of the motion χ follow.25) The resulting parametric equations (with parameter t) represent in algebraic form the particle path of the given particle. y(τ0 ) = x . x(t0 ) = X . Deﬁne a stream line through x at time t as the space curve that passes through x and is tangent to the velocity ﬁeld v at all of its points. see Figure 3. t) = 0 at all times is called a stagnation point.26) or.. ˆ dx = v(X. if the velocity at that point is independent of time. τ )dτ .27) DR X Figure 3. t) . which occupies a point X in the reference conﬁguration. AF x = χ(X.28) ME185 .. ﬁx X and consider the one-parameter family of placements T (X ﬁxed) . x v is steady at a point x. R0 ˜ dy = v(y. Subsequently. a motion χ at all points. ˜ dy = v(y.5. then v = v(x) and the motion is called steady. t) be the velocity ﬁeld at a given time t. (3. First.e. If a motion is not steady. τ )dτ . If this is the case (3. (x ﬁxed) . trace its successive placements as a function of time. one may express the same particle path in diﬀerential form as (3. i. y(t) = x . (X ﬁxed) . Alternatively. i. let v = v(x. then it is called unsteady.5: Particle path of a particle which occupies X in the reference conﬁguration. A point x in space where v(x. ˜ Now. Let χ be the motion of body B and ﬁx a particle P . (t ﬁxed) .e. t)dτ . equivalently. it is deﬁned according to (3.

this its path line coincides with the stream line through x. To argue this property. t) is the velocity at time τ of a particle which at time τ occupies Note that given a point x at time t. t) ˜ v3 (yj .31) can be derived from ˜ (3. t) ˜ v2 (yj . t)dt . dy ˜ v(y. The streak line through a point x at time t is deﬁned by the equation y = χ(χ−1 (x).31) It is easy to show that the streak line through x at time t is the locus of placements at time t of all particles that have passed or will pass through x. Indeed. see Figure 3. the preceding deﬁnition becomes (3. then the path of the particle occupying x at t and DR made for streak lines. t).30) by noting that v(y. In diﬀerential form. conﬁgurations and motions where τ is a scalar parameter and τ0 some arbitrarily chosen value of τ . the stream line and streak line through x at time t. A similar argument can be . Using dy1 dy2 dy3 = = = dτ v1 (yj . the stream line through x at t have a common tangent. components.30) where τ is a scalar parameter. t) ˜ T . Consider a material point P which is on the curve at time t.29) y x Figure 3. ME185 the point x. Notice that the velocity of P is always tangent to the stream line. since the motion is steady). yi(τ0 ) = xi . this is equivalent to stating that the velocity at time t of the material point associated with X has the same direction In the case of steady motion.6.26 Bodies. (3. while at time t it occupies the point y. take a stream line (which is now a ﬁxed curve. t) . the particle path for any particle occupying x coincides with with the velocity of the point that occupies x = χ(X.6: Stream line through point x at time t. Equation (3. τ AF ˜ dy = v(y. (t ﬁxed) . t) y(τ ) = x . this line can be expressed as (3.

ME185 . t) = ˜ 1 + t2 x2 − tx3 v3 (χj . t) = X1 et ˆ v2 (XA . t3 1 + t2 v1 (XA . that the motion is steady on the x1 -axis. t) = − X2 . Example: Consider a motion χ.32) (3. X1 = χ−1 (xj ) = x1 e−t t1 x2 − tx3 X2 = χ−1 (xj ) = t2 1 + t2 tx2 + x3 X3 = χ−1 (xj ) = .. i. ∂t DR (3. Note that x = X at time t = 0. stream lines do not inter- (3.Kinematics of deformation 27 In general.e. t) = − ˜ . such that χ1 = χ1 (XA . t) = X2 + tX3 T particles can occupy the same position at diﬀerent times. path lines can intersect. also. since intersection points merely mean that diﬀerent sect.. except at points where the velocity vanishes.e. written in the referential description has components vi (XA .34) while in the spatial description has components vi (χj .35) Note that x = 0 is a stagnation point and. t) = X3 ˆ v3 (XA . ˆ v1 (χj . i. otherwise the velocity at an intersection point would have two diﬀerent directions. t) .33) The velocity ﬁeld. The inverse mapping χ−1 is easily obtained as t AF χ3 = χ3 (XA . 1 + t2 χ2 = χ2 (XA . the body occupies the reference conﬁguration at time t = 0. t) given by ˜ (3. However. t) = −tX2 + X3 . t) = ˆ ∂χi (XA . t) = (x1 e−t )et = x1 ˜ tx2 + x3 v2 (χj . t) = X1 et with reference to ﬁxed orthonormal system {ei }.

t) The acceleration in the referential description has components ai (XA . t) = x1 ˜ a2 (xj . Recall that the motion χ is deﬁned so that x = χ(X.38) where F is the deformation gradient tensor. ˆ AF a1 (xj . ∂X F = ∂χ(X.2 The deformation gradient and other measures of deformation Consider a body B which occupies its reference conﬁguration R0 at time t0 and the current DR current conﬁguration at time t. let {EA } and {ei } be two ﬁxed right-hand orthonormal bases associated with the reference and current conﬁguration. respectively.37) 3. see Figure 3.36) a3 (XA .7. t)dX = FdX . It follows from chain rule that ME185 conﬁguration R at time t. t) = 0 ˜ a3 (xj .39) . t) are given by ˜ (3. t) = 0 . t) = 0 . deﬁned as (3. t) = 0 ˆ T ∂ 2 χi (XA . This material element is mapped into another inﬁnitesimal line element dx at point x in the (3. ∂X while in the spatial description the components ai (χj .28 Deformation gradient and other measures of deformation a1 (XA . ∂t2 hence. t) = ˆ . (3. t) = X1 et ˆ a2 (XA . ˜ dx = ∂χ (X. t) and consider the deformation of an inﬁnitesimal material line element dX located at the point X of the reference conﬁguration. Also. t) .

A dXA .8. t) ei ⊗ EA = FiA ei ⊗ EA . such that det ∂X neighborhood P0 of X in R0 and an open neighborhood P of R.40). Figure 3. Moreover. let χt : R0 → R be continuously diﬀerentiable (i. X = χt (x) and ∂x function theorem of real analysis. Using (3. which.38) in component form as AF dxi = FiA dXA = χi.41) DR 1 Recall now that the motion χ is assumed invertible for ﬁxed t. t t ∂χ−1 (x) t −1 = (F(X. ME185 . as in Figure 3.38) that the deformation gradient F provides the rule by which inﬁnitesimal line element are mapped from the reference to the current conﬁguration..7: Mapping of an inﬁnitesimal material line elements dX from the reference to the Using components. such that χt (P0 ) = P This means that the derivative of the inverse motion with respect to x is identical to the inverse of the gradient of the motion with respect to X. one may rewrite (3. t))−1 1 . Then. ∂XA (3.Kinematics of deformation 29 dX X R0 current conﬁguration.40) It is clear from the above that the deformation gradient is a two-point tensor which has one “leg” in the reference conﬁguration and the other in the current conﬁguration. for any x ∈ P. It follows from (3. there is an open and is continuous) and consider X ∈ R0 . Also. T dx x R (3. in the case of the mapping χ can be stated as ∂χt exists follows: For a ﬁxed time t. the deformation gradient tensor can be expressed as ∂χi (XB . so that χ−1 (P) = P0 .e. ∂X ∂χt (X) = 0. recall the inverse and χt has a continuously diﬀerentiable inverse χ−1 .

t) ﬁxed time t. In summary. take the dot-product of each side of (3.44) Since det F = 0. Generally.44) with itself. T x P R χ (3. that λ > 0. which .30 Deformation gradient and other measures of deformation X P0 R0 Figure 3. deﬁne the stretch λ of the inﬁnitesimal material line element dX as λ = ds . the preceding arguments imply that λ ∈ (0. vectors (i. respectively. To explore this. in particular. write dX = MdS and dx = mds.42) DR hence ME185 and note that AF dx = FdX = FMdS = mds .43) (3.8: Application of the inverse function theorem to the motion χ at a ﬁxed time t. The inverse function theorem states that the mapping χ is invertible at a point X for a ∂χ(X. λm = FM . The condition det J = 0 will be later amended to be det J > 0. the inﬁnitesimal material line element dX stretches and rotates to dx under the action of F.e. given that m is chosen to reﬂect the sense of x. M · M = m · m = 1) in the direction of dX and dx. dS (3. where M and m are unit and ds > 0 are the inﬁnitesimal lengths of dX and dx.. ∞). if the Jacobian determinant J = det = det F satisﬁes the condition ∂X J = 0. it follows that λ = 0 and. respectively and dS > 0 Next. To determine the value of λ.

it is important to observe that C is deﬁned with respect to the basis in the reference conﬁguration. both of which follow clear from (3. in analogy with the preceding derivation of C. λ where C is the right Cauchy-Green deformation tensor. C allows the determination of the stretch λ of an inﬁnitesimal material line element dX along M when mapped to the line element dx in the current conﬁguration.50) ME185 . DR hence Recall that the inverse function theorem implies that.48) Then. M · CM > 0 for any M = 0 the meaning of C. = M · (FT F)M 31 T = MdS . one may write. if. dX = ∂χ−1 (x) t dx = F−1 dx . upon using components.45). and only if M = 0. since det F = 0. it can be said that. (3. in general C = C(X. Also. To summarize symmetric and positive-deﬁnite tensor. given a direction M in the reference conﬁguration.49) (3. ∂x dX = F−1 dx = F−1 mds 1 M = F−1 m . that (3. since F = F(X. deﬁned as (3.47) Note that. C is a and M·CM = 0.Kinematics of deformation leads to (λm) · (λm) = λ2 (m · m) = λ2 = (FM) · (FM) = M · FT (FM) = M · CM .46) or. t). t). (3.45) AF C = FT F CAB = FiA FiB . (3. The latter means that. Further.

T (3.51) AF B = FFT Bij = FiA FjA . (3.52) or. taking the dot-products of each side with itself leads to 1 1 1 1 ( M) · ( M) = 2 (M · M) = 2 = (F−1 m) · (F−1 m) λ λ λ λ = m · F−T (F−1 m) = M · (F−T F−1 )m = m · B−1 m . Consider now the diﬀerence in the squares of the lengths of the line elements dX and dx. Like C. ds2 − dS 2 and write this diﬀerence as where ME185 (3.53) In contrast to C. B allows the determination of the stretch λ of an inﬁnitesimal element dx along m which is mapped from an inﬁnitesimal material line element dX in the reference conﬁguration.55) .32 Deformation gradient and other measures of deformation Again. deﬁned as (3. the tensor B is deﬁned with respect to the basis in the current conﬁguration. DR namely. given a direction m in the current conﬁguration. it can be said that. it is easy to establish that the tensor B is symmetric and positive-deﬁnite.54) (3. using components. To summarize the meaning of B. ds2 − dS 2 = (dx · dx) − (dX · dX) = (FdX) · (FdX) − (dX · dX) = dX · FT (FdX) − (dX · dX) = dX · (CdX) − (dX · dX) = dX · (C − I)dX = dX · 2EdX . 1 1 (C − I) = (FT F − I) 2 2 E = where F−T = (F−1 )T = (FT )−1 and B is the left Cauchy-Green tensor.

Using components. the preceding equation can be 1 (FiA FiB − δAB ) . unlike E. Using components. The diﬀerence ds2 − dS 2 may be also written as T written as (3. However. one may write where DR conﬁguration. (3.59) Like E. AF ds2 − dS 2 = (dx · dx) − (dX · dX) = (dx · dx) − (F−1 dx) · (F−1 dx) = (dx · dx) − dx · F−1 (F−1 dx) = dx · dx − (dx · B−1 dx) = dx · (i − B−1 )dx = dX · 2edX . recall the polar decomposition theorem.56) 2 which shows that the Lagrangian strain tensor E is deﬁned with respect to the basis in the EAB = reference conﬁguration.58) 2 is the (relative) Eulerian strain tensor or Almansi tensor. (3.57) 1 (i − B−1 ) (3. E is clearly symmetric and satisﬁes the property that E = 0 when the body undergoes no deformation between the reference and the current conﬁguration. the tensor e is symmetric and vanishes when there is no deformation between the current conﬁguration remains undeformed relative to the reference conﬁguration. In addition. and i is the identity tensor in the current conﬁguration. To extract rotation-related information from F. neither C (or B) nor E (or e) yield any information regarding the rotation of dX.60) ME185 . e = eij = 1 −1 (δij − Bij ) . the inﬁnitesimal material line element dX is both stretched and rotated due to F. in general. 2 F = RU = VR .Kinematics of deformation 33 is the (relative) Lagrangian strain tensor. the tensor e is naturally resolved into components on the basis in the current While. which states that any invertible tensor F can be uniquely decomposed into (3.

63) The tensors U and V are called the right stretch tensor and the left stretch tensor. likewise. . i. Given the polar decomposition theorem. Using this decomposition. In the ﬁrst one.38).62) and. and taking (3. equation (3. attempt to provide a physical interpretation of the decomposition theorem.e. (3.e. which is a two-point tensor. i. Recalling DR into account (3. T ponent form. dX is deformed into dX′ = UdX. while in the second one. starting from the right polar decomposition F = RU. R = RiA ei ⊗ EA . dX′ is further deformed into RdX′ = dx. V) are called the polar factors of F. the polar decomposition is expressed as (3. B = FFT = (VR)(VR)T = VRRT V = VV = V2 .38). it follows that C = FT F = (RU)T (RU) = UT RT RU = UU = U2 AF dx = FdX = (RU)dX = R(UdX) . respectively. V are symmetric positive-deﬁnite tensors. Also note that the relations U = C1/2 and V = B1/2 hold true. Of interest now is R. it follows that the component representation of the two tensors is U = UAB EA ⊗ EB and V = Vij ei ⊗ ej .61) The tensors (R. write ME185 conﬁguration.. (3.. respectively.34 Deformation gradient and other measures of deformation where R is an orthogonal tensor and U. Given their relations to C and B. they are resolved naturally on the bases of the reference and current as it is possible to deﬁne the square-root of a symmetric positive-deﬁnite tensor. Also. which explains their name. In com- FiA = RiB UBA = Vij RjA . U) or (R. it is clear that these tensors may be used to determine the stretch of the inﬁnitesimal material line element dX.64) This suggests that the deformation of dX takes place in two steps.

66) which conﬁrms that R induces a length-preserving transformation on X′ . write which. implies that dS ′ = ds. (3. For the ﬁrst step. dX′ is stretched to the same diﬀerential dS length as dx due to the action of U. T 2 = (UdX) · (UdX) = dX · (CdX) = (MdS) · (CMdS) = dS 2 M · CM = dS 2 λ2 .9. Indeed.Kinematics of deformation Note that. dX is deformed into dx′ = RdX.65) ds .68) ME185 . again. since λ = dx · dx = (RdX′) · (RdX′) = dX′ · (RT RdX′) = dX′ · dX′ . note that (3. dx′ is mapped into dx′ · dx′ = (RdX) · (RdX) = dX · (RT RdX) = dX · dX . where M′ · M′ = 1 and dS ′ is the magnitude of dX′ .67) This. F = RU implies that dX is ﬁrst subjected to a stretch U (possibly accompanied by rotation) to its ﬁnal length ds.9: Interpretation of the right polar decomposition. in the ﬁrst one. Thus. note that Figure 3. implies that the deformation of dX takes place in two steps. letting dX′ = M′ dS ′ . Turning attention to the left polar decomposition F = VR. while in the second one. (3. dX′ · dX′ = (M′ dS ′) · (M′ dS ′ ) = dS ′ 35 AF U R dX dX′ dx dx = FdX = (VR)dX = V(RdX) . DR Vdx′ = dx. (3. Subsequently. In conclusion. see Figure 3. then is reoriented to its ﬁnal state dx by R.

UM = λM . ME185 Figure 3. consider ﬁrst the right polar decomposition of equation (3. λ (3. for the stretch U to be pure. upon recalling that dX = MdS. dX DR or. the left polar decomposition F = VR means that the inﬁnitesimal material line element dX is ﬁrst subjected to a length-preserving transformation. dx′ · dx′ = dX · dX = dS 2 AF R V dx′ dx dX′ = UdX = λdX .66).10: Interpretation of the left polar decomposition. namely .10.69) (3. the inﬁnitesimal line elements dX and dX′ need to be parallel.71) write which implies that V induces the full stretch λ during the mapping of dx′ to dx. It is conceptually desirable to decompose the deformation gradient into a pure rotation and a pure stretch (or vice versa).36 Deformation gradient and other measures of deformation which means that the mapping from dX to dx′ is length-preserving. see Figure 3. In this case. For the second step.70) (3. To investigate such an option. Thus. T = (V−1dx) · (V−1dx) = dx · (B−1 dx) = (mds) · (B−1 mds) = ds2 m · B−1 m 1 = 2 ds2 . followed by stretching (with further rotation) to its ﬁnal state dx.

The eigenvalues λA > 0 of (3. it follows that DR orthonormal basis on E 3 . the classical spectral representation theorem guarantees that there exists an orthonormal basis MA of E 3 consisting entirely of eigenvectors of U and that.71) represents a linear eigenvalue problem. since. T λ(A) M(A) ⊗ M(A) .75) therefore.74) Subtracting the two equations from each other leads to (3. respectively.. by assumption. M2 . MA · MB = δAB . (λ(A) − λ(B) )M(A) · M(B) = 0 .76) i. that the principal directions are mutually orthogonal and that {M1 .73) (3.e. are called the principal stretches and the associated eigenvectors MA are called the principal (3. one may write UMA = λ(A) M(A) UMB = λ(B) M(B) .77) ME185 . 3 (3. M3 } form an It turns out that regardless of whether U has distinct or repeated eigenvalues. U = A=1 (3. Upon premultiplying the preceding two equations with MB and MA . one gets AF MB · (UMA ) = λ(A) MB · M(A) MA · (UMB ) = λ(B) MA · M(B) .Kinematics of deformation 37 Equation (3.72) where the parentheses around the subscripts signify that the summation convention is not in use. λA = λB .71) directions. (3. if λA are the associated eigenvalues. When λA are distinct.

(A) (3. see Figure 3.38 Thus. attempt a reinterpretation of the right polar decomposition.80) . the stretched elements are reoriented to their ﬁnal direction by the action of R. V to be pure it is necessary that ME185 λ3 RM3 x λ1 RM1 R λ2 RM2 Figure 3. U m = A=1 λm M(A) ⊗ M(A) .78) and. Now. note for the stretch dx = Vdx′ = λdx′ (3. in light of the discussion of principal stretches and directions. Subsequently. M3 DR M1 X M2 λ3 M3 λ2 M2 U λ1 M1 MA and associated principal stretches λA . then it subjects them to a pure stretch. Deformation gradient and other measures of deformation 3 C = U 2 =( = A=1 3 3 λ(A) M(A) ⊗ M(A) )( A=1 B=1 3 3 λ(A) λ(B) (M(A) ⊗ M(A) )(M(B) ⊗ M(B) ) λ(A) λ(B) (M(A) · M(B) )(M(A) ⊗ M(B) ) ⊗ M(A) = = AF 3 A=1 B=1 3 λ2 M(A) (A) A=1 T 3 B=1 λ(B) M(B) ⊗ M(B) ) (3. more generally. when U applies on inﬁnitesimal material line elements which are aligned with the principal directions MA .79) for any integer m.11: Interpretation of the right polar decomposition relative to the principal directions Following an analogous procedure for the left polar decomposition. Indeed.11.

12: Interpretation of the left polar decomposition relative to the principal directions Returning to the discussion of the polar factor R. det R = −1). proper (resp. improper) if det R = 1 (resp. M3 M1 X M2 λ3 RM3 x λ2 RM2 RM3 DR R RM1 RMi and associated principal stretches λi . see Figure 3. again. for now. Appealing to the spectral representation theorem.Kinematics of deformation or. VRM = λRM .82) A reinterpretation of the left polar decomposition can be aﬀorded along the lines of the earlier corresponding interpretation of the right decomposition. Speciﬁcally. recalling that dx′ = RMdS. 39 T 3 i=1 (3. that R is proper orthogonal. or det R = ±1. a linear eigenvalue problem that can be solved for the principal stretches λi and the (rotated) principal directions mi = RMi . in turn. recall that it is an orthogonal tensor. RM2 λ1 RM1 V Figure 3. (3. implies that (det R)2 = 1. one may write 3 V = AF i=1 λ(i) (RM(i) ) ⊗ (RM(i) ) = λ(i) m(i) ⊗ m(i) . and invoking elementary properties of ME185 . An orthogonal tensor R is termed Assuming. This.81) This is.12. here the inﬁnitesimal material line elements that are aligned with the principal stretches MA are ﬁrst reoriented by R and subsequently subjected to a pure stretch to their ﬁnal length by the action of V.

(3. (3.87) (3. R can be expressed with reference to this basis as R = Rpp p ⊗ p + Rpq p ⊗ q + Rpr p ⊗ r + Rqp q ⊗ p + Rqq q ⊗ q + Rqr q ⊗ r + Rrp r ⊗ p + Rrq r ⊗ q + Rrr r ⊗ r . the orthogonality of R can be expressed as (p ⊗ p + Rqq q ⊗ q + Rqr r ⊗ q + Rrq q ⊗ r + Rrr r ⊗ r) (p ⊗ p + Rqq q ⊗ q + Rqr q ⊗ r + Rrq r ⊗ q + Rrr r ⊗ r) = p ⊗ p + q ⊗ q + r ⊗ r . given that R is orthogonal.88).40 Deformation gradient and other measures of deformation determinant. thus. It follows that {p. Rpq = Rpr = 0 .86) RT p = R−1 p = p ⇒ Rpp p + Rpq q + Rpr r = p .86) and (3. AF Rp = p ⇒ Rpp p + Rqp q + Rrp r = p . and consider two unit vectors q and r = p × q that lie on a plane normal to p. therefore ME185 (3. r} form a right-hand orthonormal basis of E 3 and. q.85) which implies that DR Moreover. (3. T (3. Denote by p a unit eigenvector associated with the above eigenvalue (there exist two such unit vectors).89) .88) Taking into account (3. it is seen that RT R = I ⇒RT R − RT = I − RT ⇒RT (R − I) = −(R − I)T ⇒ det RT det(R − I) = − det(R − I)T ⇒ det(R − I) = − det(R − I) ⇒ det(R − I) = 0 . (3.83) so that R has at least one unit eigenvalue. Rqp = Rrp = 0 . Rpp = 1 .84) Note that. since p is an eigenvector of R.

and.91) (3. after reducing the terms on the left-hand side. Rqq = cos θ . or φ = π − θ + 2kπ . Equations (3. 2 2 2 2 p ⊗ p + (Rqq + Rrq )q ⊗ q + (Rrr + Rqr )r ⊗ r 41 + (Rqq Rqr + Rrq Rrr )q ⊗ r + (Rrr Rrq + Rqr Rqq )r ⊗ q The above equation implies that AF 2 2 Rqq + Rrq = 1 . = p ⊗ p + q ⊗ q + r ⊗ r .92) imply that there exist angles θ and φ. Rrq = sin θ .101) (3. (3. (3.93) (3. thus φ = −θ + 2kπ.97) (3. such that Rqq = cos θ (3.91) and (3.94)) that DR thus φ = −θ + 2kπ cos θ sin φ + sin θ cos φ = sin (φ + θ) = 0 .98) where k is an arbitrary integer. consequently. .100) (3.95) and Rrr = cos φ .94) Rqq Rqr + Rrq Rrr = 0 .96) It follows from (3. 2 2 Rrr + Rqr = 1 . (3.93) (or (3.102) ME185 Rrq = sin θ .Kinematics of deformation and. It can be easily seen that the latter choice yields an improper orthogonal tensor R.92) (3. Rqr = − sin θ . Rrr Rrq + Rqr Rqq = 0 . Rqr = sin φ .99) (3. Rrr = cos θ .90) (3. (3. T .

the vector x remains unstretched and it rotates by an angle θ around the p-axis. θ x Rx r (3. it follows that R can be expressed as R = p ⊗ p + cos θ (q ⊗ q + r ⊗ r) − sin θ (q ⊗ r − r ⊗ q) . (3.104) AF p q q = q·x .103) of a proper orthogonal tensor R is often referred to as the Rodrigues formula. where p = p·x . .86). q. The representation (3. DR ME185 Figure 3.42 Deformation gradient and other measures of deformation From (3.105) and note that Rx = pp + (q cos θ − r sin θ)q + (q sin θ + r cos θ)r .102).99-3. (3. q. Rx rotates by an angle θ around the p-axis and is also reﬂected relative to the origin of the orthonormal basis {p.103) The angle θ that appears in (3. (3. If R is improper orthogonal. r}.106) indicates that. r} as T r = r·x .103) can be geometrically interpreted as follows: let an x = pp + qq + rr . arbitrary vector x be written in terms of {p.982 in connection with the negative unit eigenvalue p implies that. where θ is assumed positive when directed from q to r in the sense of the right-hand rule. the alternative solution in (3. (3.88) and (3. under the action of R.13: Geometric interpretation of the rotation tensor R by its action on a vector x.106) Equation (3.

The components CAB of C and the components UAB of U can be directly determined as DR and Also. t) = ( a cos ϑX1 ) − ( a sin ϑ)X2 √ √ χ2 = χ2 (XA .e.. Further. R has three independent components.Kinematics of deformation 43 A simple counting check can be now employed to assess the polar decomposition (3. At the same time.A of the deformation gradient can be easily determined as √ √ a cos ϑ − a sin ϑ 0 √ √ [FiA ] = a sin ϑ a cos ϑ 0 . . i. The components FiA = χi. for instance two of the three components of the unit eigenvector p and the angle θ. recall that which implies that λ1 = λ2 = √ AF χ3 = χ3 (XA . This is clearly a planar motion.60). [CAB ] = 0 a 0 0 0 1 a 0 0 √ a 0 0 √ [UAB ] = 0 a 0 0 0 1 CM = λ2 M . t) = ( a sin ϑ)X1 + ( a cos ϑ)X2 T Indeed. F has nine independent components and U (or V) has six independent components. hence the motion is always invertible. t) = X3 . a and λ3 = 1. Example: Consider a motion χ deﬁned in component form as where a = a(t) > 0 and ϑ = ϑ(t). √ √ χ1 = χ1 (XA . ME185 . speciﬁcally independent of X3 . 0 0 1 This is clearly a spatially homogeneous motion. the deformation gradient is independent of X. note that det(FiA ) = a > 0.

107) and (3. DR ME185 Figure 3. in this case. Assume that the body undergoes a spatially homogeneous deformation with deformation gradient F(t). it follows from (3. where λ(t) is the stretch of a material line element that lies along M in the reference conﬁguration.107) Consider the part of a deformable body which occupies a sphere P0 of radius σ centered where AF Π = σM . so that (3. at the ﬁxed origin O of E 3 . .44 Deformation gradient and other measures of deformation Given that U is known. (3.14: Spatially homogeneous deformation of a sphere. Recall that λm = FM. 2. Then. The equation of the surface ∂P0 of the sphere can be written as Π · Π = σ2 . . 1 (3. Example: Sphere under homogeneous deformation T 0 0 1 √ a rotation tensor R. Indeed. π = FΠ . one may apply the right polar decomposition to determine the 1 √ a 0 0 Note that this motion yield pure stretch for ϑ = 2kπ. Π π O m P O M P0 π = σλm . as in the ﬁgure above.. . where k = 0. and m(t) is the unit vector in the direction this material line element at time t. √ √ a cos ϑ − a sin ϑ 0 √ √ [RiA ] = a sin ϑ a cos ϑ 0 0 0 1 0 0 = sin ϑ 0 1 cos ϑ − sin ϑ 0 cos ϑ 0 0 .109) where π(t) is the image of Π in the current conﬁguration.108) and M · M = 1. 1.110) .109) that (3.

Likewise. deduce that 2 π · B−1 π = λ−2 πi . one may write 3 AF B −1 = (3.114) i=1 and using (3. the left Cauchy-Green deformation tensor 3 (3.112) i=1 where (λ2 . dX2 . i = 1. (i) λ−2 m(i) ⊗ m(i) . (i) equation (3. 2.112). dx2 . and dx3 . the spherical region P0 reference conﬁguration to its image dv in the current conﬁguration due to the motion χ.111) Recalling the spectral representation theorem. mi ). i = 1. 3.15.Kinematics of deformation 45 Also.113) and (3. i (3. note that mi . λ2 λ2 λ2 1 2 3 is deformed into an ellipsoid with principal semi-axes of length σλi along the principal Consider now the transformation of an inﬁnitesimal material volume element dV of the which demonstrates that. where B(t) is the left Cauchy-Green deformation tensor.114). see Figure 3.15. The referential volume element is deﬁned as an inﬁnitesimal parallelepiped with sides dX1 . since λ2 m · B−1 m = 1.115) DR It is readily seen from (3. can be resolved as B = T λ2 m(i) ⊗ m(i) . and dX3 . are respectively the eigenvalues (squares of the principal stretches) i and the associated eigenvectors (principal directions) of B. ME185 . are orthonormal and form a basis of E 3 .115) that directions of B. π · B−1 π = σ 2 . (3.110) leads to (3. see Figure 3.113) Starting from equation (3. therefore the position vector π can be uniquely resolved in this basis as π = πi mi . where each dxi is the image of dXi due to χ. 2. its spatial counterpart is the inﬁnitesimal parallelepiped with sides dx1 . under a spatially homogeneous deformation. In addition. 2 π2 π2 π1 + 2 + 3 = σ2 . 3.

dX2 and dX3 . dXA ] 1 2 3 = (det F) det [dXA ]. [FiA dXA ]. [dXA ].117) where each of the representation of dV in (3. AF dX3 X dX2 dx3 x dx2 dX1 dx1 R0 R = (FdX1 ) · (FdX2 ) × FdX3 ) 1 2 3 = det [FiA dXA ]. [dXA ] T (3. dXA . Recalling that the triple product satisﬁes the property 1 2 3 [dX1 dX2 dX3 ] = det [[dXA ]. Figure 3.15: Image of a sphere under homogeneous deformation. [FiA dXA ] 1 2 3 = det FiA [dXA . note that DR dv = dx1 · (dx2 × dx3 ) = JdV . [dXA ].46 Deformation gradient and other measures of deformation m3 m2 m1 Figure 3.16: Mapping of an inﬁnitesimal material volume element dV to its image dv in the current conﬁguration. ME185 dV = dX1 · (dX2 × dX3 ) = dX2 · (dX3 × dX1 ) = dX3 · (dX1 × dX2 ) .116) corresponds to the triple product [dX1 dX2 dX3 ] of the vectors dX1. First. it follows that in the current conﬁguration . [dXA ]].116) (3.

dx = FdX}. ME185 Let dX be any inﬁnitesimal material line element. which would violate the assumption taken to be positive at all time. Consider next the transformation of an inﬁnitesimal material surface element dA in the reference conﬁguration to its image da in the current conﬁguration.e. dx2 . then the relative orientation of line elements should be preserved during the motion. Based on the preceding observation. dX} and {dx1 . DR current conﬁguration. J = 1) for all inﬁnitesimal material volumes dV at all Here. by convention. dX2 and dX3 observe the right-hand rule). dX2 . Similarly. Figure 3. dX2 X dX1 x dx1 R R0 T dx2 times is called isochoric (or volume preserving). any changes in the sign of J would necessarily imply that there exists a time t at which J = 0 at some material point(s). (3.e. da = dx1 × dx2 = nda .Kinematics of deformation 47 A motion for which dv = dV (i.16. such that the assumption of invertibility. the Jacobian J will be AF dA = dX1 × dX2 = NdA . Indeed. J > 0 everywhere and at all time. one may argue that if. since the motion is assumed smooth. see Figure 3. The referential surface element is deﬁned as the parallelogram formed by the inﬁnitesimal material line elements dX1 and dX2 .118) where N is the unit normal to the surface element consistently with the right-hand rule.17: Mapping of an inﬁnitesimal material surface element dA to its image da in the the inﬁnitesimal volumes dV and dv formed by {dX1. in the current conﬁguration.. i. such that N · dX > 0 and consider . one may write (3. dV > 0 (which is true as long as dX1 .119) where n is the corresponding unit normal to the surface element deﬁned by dx1 and dx2 .

122) with itself and taking square roots yields (3. It follows from (3.121) for any inﬁnitesimal material line element dX. (3. Next. interest is focused to measures of the rate at which deformation occurs in the continuum. so that L can be written as where ME185 In this section.122) is known as Nanson’s formula.126) . 1 (L + LT ) 2 D = (3. ∂x (3. which implies that (FdX) · nda = JdX · NdA . then da > 0 everywhere and at all time.123) As argued in the case of the inﬁnitesimal volume transformations.117) that dv = dx · (dx1 × dx2 ) = dx · nda = (FdX) · nda = JdV = JdX · (dX1 × dX2 ) = JdX · NdA .3 Velocity gradient and other measures of deformation rate DR part. T ∂v .122) Equation (3.120) (3. in turn. 3. if an inﬁnitesimal material line element satisﬁes dA > 0. AF |da| = J N · (C−1 N)|dA| . since J > 0 and C−1 is positive-deﬁnite.48 Velocity gradient and other measures of deformation rate respectively.125) (3. This. Dotting each side in (3. deﬁne the spatial velocity gradient tensor L as (3.124) recall any such tensor can be uniquely decomposed into a symmetric and a skew-symmetric This tensor is naturally deﬁned relative to the basis {ei } in the current conﬁguration. L = grad v = L =D+W . To this end. leads to nda = JF−T NdA .

131) 1 −1 (B L + LT B−1 ) .133) ME185 . it can be shown that and Similar results may be readily obtained for the rates of the Lagrangian and Eulerian strain (3. t) = ∂X ∂X ∂x ∂X (3.132) 2 Proceed now to discuss physical interpretations for the rate tensors D and W. in the above derivation the change in the order of diﬀerentiation between the derivatives with respect to X and t is allowed under the ∂2χ assumption that the mixed second derivative is continuous. for the left Cauchy-Green deformation tensor.129) Likewise. t) ∂X = AF ˙ E = FT DF ˙ e = ˙ ˙ ˙ ˙ λm + λm = FM + FM = LFM = L(λm) = λLm . Also. one may write ˙ ˙ ˙ ˙ B = FFT = FFT + FFT = (LF)FT + F(LF)T = LB + BLT .130) DR measures.128) where use is made of the chain rule. t) ∂ v(X. ∂X∂t Given (3. which is skew-symmetric. one may express with the aid of the product rule the rate of change of the right Cauchy-Green deformation tensor C for a ﬁxed particle X as ˙ ˙ ˙ ˙ C = FT F = FT F + FT F = (LF)T F + FT (LF) = FT (LT + L)F = 2FT DF . ˜ ˆ ∂ ∂ v(x. (3. Speciﬁcally. write the material time derivative of F as ˙ F = ˙ ∂χ(X. t) ˙ = = LF . Starting from (3. (3. (3. To this end. which is symmetric. take the material time derivatives of both sides to obtain the relation (3. and 49 T 1 (L − LT ) 2 W = (3.127) is the vorticity (or spin) tensor. t) ∂χ(X. χ(X.Kinematics of deformation is the rate-of-deformation tensor .44).128). Consider now the rate of change of the deformation gradient for a ﬁxed particle associated with point X in the reference conﬁguration.

135) as (3. W = 1 (W − WT ) . so that the preceding equation simpliﬁes to ˙ λ = λm · Lm . when operating on any vector z ∈ E 3 . Upon taking the dot-product of each side with m.139) (3. such as W. ˙ λ = λm · Dm ln˙ λ = m · Dm = D · (m ⊗ m) .138) Thus. alternatively. 2 (3.136) hence m · Wm = 0. First. Before proceeding with the geometric interpretation of W. since the skew-symmetric part W of L satisﬁes AF m · Wm = m · (−WT )m = −m · Wm .135) Further. To establish (3. note that any skew-symmetric tensor in E 3 . (3. has only only three independent components. some preliminary development DR this correspondence. one may rewrite (3.134) ˙ However. it follows that (3. hence does ˙ ˙ λm · m + λm · m = λ(Lm) · m . as (3. This suggests that there exists a one-to-one correspondence between skew-symmetric tensors and vectors in E 3 . ME185 on skew-symmetric tensors is required.50 Velocity gradient and other measures of deformation rate ˙ Note that M = 0. observe that so that.140) . since m is a unit vector. the tensor D fully determines the material time derivative of the logarithmic stretch ln λ for a material line element along a direction m in the current conﬁguration. 2 Wz = 1 Wij (ei ⊗ ej − ej ⊗ ei )z 2 1 = Wij [(z · ej )ei − (z · ei )ej ] .137) or. since M is a ﬁxed vector in the ﬁxed reference conﬁguration. it follows that m · m = 0. T not vary with time.

Speciﬁcally.146) ME185 can be rewritten as where the vector w is deﬁned as and is called the axial vector of the skew-symmetric tensor W. Wij = eikj wk = ejik wk . 2 Wij zj ei = eijk wj zk ei = eikj wk zj ei . 2 2 wk = 1 eijk Wji .142). one may represent W in terms of w and vice-versa.Kinematics of deformation 51 Recalling the standard identity u × (v × w) = (u · w)v − (v · u)w. If W is speciﬁcally the vorticity .144) (3.141). w = 1 Wji ei × ej 2 w = wk ek = 1 1 Wjiei × ej = Wji eijk ek . starting from (3. starting from (3.145) (3. DR Conversely. so that hence. the preceding equation Wz = AF = w×z . The above relations apply to any skew-symmetric tensor.141) (3.142) (3. Using components.143) (3. 1 Wij [z × (ei × ej )] 2 1 = − Wij [(ei × ej ) × z] 2 1 = Wji [(ei × ej ) × z] 2 1 = Wjiei × ej ] × z 2 T (3.

e. λ = T (3. ˙ λ ˙ m = Lm − m = λ ˙ λ L− i m λ ˙ λ D − i m + Wm .i − vi. ¯ (D − γi)m = 0 . ¯ Returning to the geometric interpretation of W.iek 2 1 curl v . hence is deﬁned as ˜ dy = w(y.52 Velocity gradient and other measures of deformation rate tensor.j )ek 4 1 (eijk vj. 2 AF = .i − vi. ˜ Let w = w(x.151) In this case.i − eijk vi. λ i. write ME185 ˙ ¯ λ ˙¯ ¯ ¯ ¯ ¯ (3.i − ejik vi.149) (3.148) (3. then the associated axial vector satisﬁes the relation w = = = = = 1 (vj. take m to be a unit vector that lies along a principal direction of D in the current conﬁguration. any line passing through x at time t is a vortex line.j )ek 4 1 eijk vj. (t ﬁxed) . The vortex line through x at time t is the space curve that passes through x and is tangent to the vorticity vector ﬁeld ˜ w at all of its points. Also.150) (Dm) · m = γ m · m = γ = ¯ = ln λ. equivalently. t)dτ For an irrotational motion. the axial vector w is called the vorticity vector. namely DR It follows from (3. t) be the vorticity vector ﬁeld at a given time t.133).149) that Recalling the identity (3. y(τ0 ) = x . w = 0). a motion is termed irrotational if W = 0 (or. . the eigenvalues of D are equal to the material time derivatives of the logarithmic stretches ¯ ¯ ln λ of line elements along the eigendirections m in the current conﬁguration.j )eijk ek 4 1 (eijk vj.147) (3.j )ei × ej 4 1 (vj..

the material time derivative of a unit vector m along a principal direction of D is determined by (3. one ¯ x+ = χ+ (X. t) . take a motion χ of body B such that. SUPERPOSED RIGID-BODY MOTIONS 53 which holds for any direction m in the current conﬁguration. 3. Denoting by X and Y the position vectors of two material points on the ﬁxed reference conﬁguration. Y) = |X − Y| . x+ = χ+ (X.152) ¯ Therefore. ¯t ¯t x+ = χ+ (X) = χ+ (x) = χ+ (χt (X)) t ¯t χ+ = χ+ ◦ χt . with reference to Figure 3. one may conclude that w plays the role of the angular velocity of a line element which. Then. Now. t) = χ+ (x. Setting in the above equation ˙ ¯ ¯ ¯ m = Wm = w × m . lies along the principal direction m of D. it follows that (3. t (3. t) = χ+ (χ−1 (x).156) (3. t (3.155) so that χ and χ+ diﬀer by a rigid-body motion.153) According to the preceding deﬁnition. To this end. as usual. introduce the notion of a superposed rigid-body motion. (3.4 Superposed rigid-body motions A rigid motion is one in which the distance between any two material points remains constant at all times. y) . such that may write Said diﬀerently. for any points X and Y. T ¯ m = m.154) for all times t.157) (3.18. Y) = d(χ(X.4. DR B. t). in the current conﬁguration. deﬁne the distance function as AF d(X.3. or.158) ME185 . Then. t)) = d(x. x = χ(X. t). a motion is rigid if. Recalling from rigid-body dynamics the formula relating linear to angular velocities. χ(Y. t) . let another motion χ+ be deﬁned for (3.152). equivalently. d(X.

the superposed motion χ+ (x. t) − χ+ (y. t) = χ+ (χt (y). Since x and y are chosen independently. t) ∂x T ¯ ¯ ¯ ¯ = χ+ (x.159) (3.160) ¯ ¯ (χ+ (x. equation (3. which leads to ¯ ∂ χ+ (y. equation (3. t) is invertible for ﬁxed t. t) ∂x ¯ ∂ χ+ (x. (3. y in the region R at time t.162) ¯ Since the motion χ+ is invertible. t) ∂x T T y+ R+ ¯ χ+ x y R (3.160) ﬁrst with respect to x to get Then.162) can be equivalently written as = ¯ ∂ χ+ (y. one may diﬀerentiate equation (3. t) . t)) .54 Superposed rigid-body motions χ+ x+ X Y R0 χ Figure 3. ¯ Clearly. for all x.161) T Recalling that R and R+ diﬀer only by a rigid transformation. take a second point Y in the reference conﬁguration. t) . one may conclude that (x − y) · (x − y) = (x+ − y+ ) · (x+ − y+ ) DR x−y = i = ¯ ∂ χ+ (x.163) ME185 . t) ∂y −1 . (3. so that y = χ(Y. t) = χ+ (y. t) and −1 ¯ y+ = χ+ (Y. t) − χ+ (y. (3. t) χ+ (x. since χ+ is assumed invertible for ﬁxed t. write Similarly. AF ¯ ∂ χ+ (x.18: Conﬁgurations associated with motions χ and χ+ diﬀering by a superposed rigid ¯ motion χ+ . t) − χ+ (y.161) may be diﬀerentiated with respect to y. t) ∂y .

t) = = Q(t). write (3. U+ are symmetric positive-deﬁnite (QR)T (QR) = (RT QT )(QR) = RT (QT Q)R = RT R = I (3.167) DR tensors.164) implies that = 55 T ∂χ+ (y.169) (3. note that using the chain rule the deformation gradient F+ of the motion χ+ is written as F + AF ¯ x+ = χ+ (x. equation (3. by assumption. Given (3. the uniqueness of the polar decomposition.168) and also det (QR) = (det Q)(det R) = 1.165)3 and recalling the right polar decomposition of F. Since Q is a function of time only. R+ = QR U+ = U .167). hence ∂χ+ (x. both motions χ and χ+ lead to deformation gradients with positive Jacobians. Further.164) ∂χ+ (y.e. Since.164)1 can be directly integrated with respect to x. Equation (3.Kinematics of deformation Then. equation (3. ∂X ∂x ∂X (3. in conjunction with (3. it follows immediately ∂x ∂y that QT (t)Q(t) = I. F+ = R+ U+ = QF = QRU . i. the left. t) = Q(t)x + c(t) . and where R. t) ∂χ+ (x.165) implies that det Q > 0. therefore QR is proper orthogonal. hence det Q = 1 and Q is proper orthogonal. Q(t) is an orthogonal tensor. T −1 (3. t) ∂y = QT (t) . R+ are proper orthogonal tensors and U. ∂χ+ ∂χ+ ∂χ = = = QF .165) Since.166) where c(t) is a vector function of time. leading to (3. necessitates that (3.166) is the general form of the superposed ¯ rigid motion χ on the original motion χ.170) ME185 .. t) ∂x Since equation (3.and right-hand side should be necessarily functions of time only. clearly.

56

Superposed rigid-body motions Similarly, the left decomposition of F yields

F+ = V+ R+ = V+ (QR) = QF = Q(VR) , which implies that

T

(3.171)

V+ (QR) = Q(VR) , hence,

(3.172)

AF

V+ = QVQT .

T

(3.173)

**It follows readily from (3.165)3 that
**

T

C+ = F+ F+ = (QF)T (QF) = (FT QT )(QF) = FT F = C and

(3.174)

B+ = F+ F+ Likewise,

= (QF)(QF)T = (QF)(FT QT ) = QBQT . 1 + 1 (C − I) = (C − I) = E 2 2

(3.175)

E+ =

(3.176)

and

e+ =

DR

ME185

1 1 (I − B+ −1 ) = [I − (QBQT )−1 ] 2 2 1 = (I − Q−T B−1 Q−1 ) 2 1 = (I − QB−1 QT ) 2 1 = Q(I − B−1 )QT 2 = QeQT .

(3.177)

A vector or tensor is called objective if it transforms under superposed rigid motions

in the same manner as its natural basis. Physically, this means that the components of an objective tensor relative to its basis are unchanged if both the tensor and its natural basis are transformed due to a rigid motion superposed to the original motion. Speciﬁcally, a referential tensor is objective when it does not change under superposed rigid motions,

Kinematics of deformation

57

C, U and E are objective. Correspondingly a spatial tensor is objective if it transforms {(Qei ) ⊗ (Qej )} = Q{ei ⊗ ej }QT . Hence, spatial tensors such as B, V and e are objective. according to ( · )+ = Q( · )QT . This is because its tensor basis {ei ⊗ ej } transforms as It is easy to deduce that two-point tensors are objective if they transform as ( · )+ = Q( · ) or ( · )+ = ( · )QT depending on whether the ﬁrst or second leg of the tensor lies in the current

are spatial.

objective, remain untransformed if they are referential or transform as ( · )+ = Q( · ) if they Examine next the transformation of the velocity vector v under a superposed rigid-body

motion. In this case

˙ v+ = χ+ (X, t)

AF

˙ ˙ ˙ QQT = QQT + QQT = 0 . ˙ Ω(t) = Q(t)QT (t) , ˙ v+ = ω × (x+ − c) + Qv + c .

conﬁguration, respectively. Analogous conclusions may be drawn for vectors, which, when

˙ ˙ ˙+ ˙ = χ (x, t) = [Q(t)x + c(t)] = Q(t)x + Q(t)v + c(t) .

T

as the latter do not aﬀect the reference conﬁguration. Hence, referential tensors such as

since its natural basis {EA ⊗ EB } remains untransformed under superposed rigid motions,

(3.178)

Since QQT = I,

(3.179)

Setting

DR

(3.180),

(3.180)

it follows from (3.179) that Ω is skew-symmetric, hence there exists an axial vector ω such that Ωz = ω × z, for any z ∈ E 3 . Returning to (3.178), write, with the aid of (3.166) and

˙ ˙ v+ = ΩQ(t)x + Q(t)v + c(t) = Ω(x+ − c(t)) + Q(t)v + c(t) .

(3.181)

Invoking the deﬁnition of the axial vector ω, one may rewrite (3.181) as (3.182)

**Equation (3.181) reveals that the velocity is not an objective vector.
**

ME185

58

Superposed rigid-body motions Next, examine how the various tensorial measures of deformation rate transform under

L

+

AF

D+ = 1 + (L + L+ T ) 2 1 1 = (Ω + QLQT ) + (Ω + QLQT )T 2 2 1 1 T = (Ω + Ω ) + Q (L + LT )QT 2 2 T = QDQ , W+ = 1 + (L − L+ T ) 2 1 1 = (Ω + QLQT ) − (Ω + QLQT )T 2 2 1 1 = (Ω − ΩT ) + Q (L − LT )QT 2 2 T = Ω + QWQ ,

∂ ∂v+ ˙ = [Ω(x+ − c) + Qv + c] = + + ∂x ∂x ∂(Q˜ ) v = Ω+ + ∂x ∂(Q˜ ) ∂χ v = Ω+ ∂x ∂x+ ˜ ∂v ∂ [QT (x+ − c)] = Ω+Q ∂x ∂x+ = Ω + QLQT ,

T

(3.183) (3.184) (3.185) (3.186)

superposed rigid motions. Starting with the spatial velocity gradient, write

where use is made of (3.181), the deﬁnition of L and the chain rule. It follows from (3.183) that L is not objective. Also, the rate-of-deformation tensor D transforms according to

which implies that D is objective. Turning to the vorticity tensor W, one may write

DR

from where it is seen that W is not objective. as

ME185

Objectivity conclusions can be drawn for other kinematic quantities of interest by appeal-

ing to results already in place. For instance, inﬁnitesimal material line elements transform dx+ = F+ dX = (QF)dX = Q(FdX) = Qdx ,

inversion ( · )−1 and material time derivative ( · ) operations. For inﬁnitesimal material area elements. DR AF (3. Similarly. taking the dot-product of each side of (3. Now.189) therefore (da+ )2 = da2 .188) with itself yields (n+ da+ ) · (n+ da+ ) = (Qnda) · (Qnda) .188) ME185 . (3. In closing. (3. T hence the vector da is objective. note that the superposed rigid motion operation ( · )+ commutes with the ˙ transposition ( · )T . write da+ = n+ da+ = J + F+ −T NdA = J(QF)−T NdA = J(Q−T F−T )NdA = JQF−T NdA = Qnda = Qda .187) hence dv is an objective scalar quantity. hence also da+ = da (as long as da is taken positive from the outset) and n+ = Qn.Kinematics of deformation hence dx is objective. inﬁnitesimal material volume elements transform as 59 dv + = J + dV = det(QF)dV = (det Q)(det F)dV = (det F)dV = JdV = dv .

60 Superposed rigid-body motions DR ME185 AF T .

(4.4) P 61 . (4. vector and tensor Euclidean point space E 3 . functions. Recall that P is bounded if it can be enclosed by a sphere of is uniquely deﬁned at all boundary points. E 3 ). All three functions are assumed continuously diﬀerentiable. AF grad φ dv = ∂P T φn da . and a tensor DR Then.2) In addition. (4. a vector function v : P → E 3 . deﬁne a scalar function φ : P → R. To this end.1) grad v dv = ∂P P and v ⊗ n da . let P be a bounded closed region with smooth boundary ∂P in the ﬁnite radius. the gradients of φ and v satisfy P function T : P → L(E 3 .Chapter 4 Basic physical principles 4. (4. the divergences of v and T satisfy div v dv = ∂P P v · n da .1 The divergence and Stokes’ theorems By way of background.3) and div T dv = ∂P Tn da . closed if it contains its boundary and smooth if the normal n to its boundary Next. ﬁrst review the divergence theorem for scalar.

62 The divergence and Stokes’ theorems Equation (4. To eliminate the ambiguity. let A be an open surface Consider a closed non-intersecting curve C which is parametrized by a scalar τ . The other three identities can be side by a constant vector c and using (4. Equation (4. (4.4) by setting T = φI. choose one of the sides of the surface and denote its outward unit normal by n.7) (4. so that the position vector of a typical point on C is c(τ ). Clearly.6) P grad(v · c) dv ∂P (n ⊗ v)c da = ∂P (n ⊗ v) da c . any point on A possesses two equal and opposite unit vectors.4) can be derived by dotting the left-hand (TT c) · n da ∂P = ∂P (Tn) · c da = Tn da · c . 0 ≤ τ ≤ 1. then Stokes’ theorem states that curl v · n dA = v · dx .5) P grad φ dv · c = ∂P φn da · c . in which case.3) expresses the classical divergence theorem.1) can be deduced from (4. Indeed. div(φI) dv · c = = div(φc) dv = tr grad(φc) dv = tr(c ⊗ grad φ) dv P P AF P P P grad φ · c dv = P grad φ dv · c = which. If now v is τ τ ¯ a continuously diﬀerentiable vector ﬁeld. T ∂P derived from this theorem. for any constant vector c. equation (4. implies (4.3) and the deﬁnition of the divergence of a tensor. the identity (4.1) by taking any constant c and writing T grad v dv P c = (grad v)T c dv = P DR = ∂P (v · c)n da = which proves the identity. bounded by C. for any τ ∈ [0. see Figure 4.1.4) follows immediately. Lastly. (4. each pointing outward to one of the two sides of the surface. This side is chosen so that c(¯) × c(¯ + dτ ) points toward it.1).8) A C ME185 . 1). again due to the arbitrariness of c.2) is obtained from (4. This leads to div T dv · c = div T · c dv = div(TT c) dv = P P P Since c is chosen arbitrarily. (4. Also. (4.

If .4.2.2: A region P with boundary ∂P and its image P0 with boundary ∂P0 in the reference ˆ ˜ Further. χ P ∂P R ∂R Let P be a closed and bounded region in E 3 with smooth boundary ∂P and assume that the DR P0 ∂P0 R0 ∂R0 conﬁguration.2 The Reynolds’ transport theorem particles which occupy this region at time t occupy a closed and bounded region P0 with smooth boundary ∂P0 at another time t0 . such that Figure 4. t) = φ(x. The circulation is the (inﬁnite) sum of the tangential components of v along C. ˆ ˜ φ = φ(X. THE REYNOLDS’ TRANSPORT THEOREM 63 This integral on the right-hand side of (4.9) ME185 AF C Figure 4.1: A surface A bounded by the curve C.2. let a scalar ﬁeld φ be deﬁned by a referential function φ or a spatial function φ. A the velocity around C equals to twice the integral of the normal component of the vorticity 4.8) is called the circulation of the vector ﬁeld v v is identiﬁed as the spatial velocity ﬁeld. see Figure 4. then Stokes’ theorem states that the circulation of vector on any open surface that is bounded by C. t) . T around C. (4.

P T ˜ φ dv . The right-hand side of (4. interchange the diﬀerentiation and integration operations.11) consists of the sum of two ME185 . because the region P over which the integral is evaluated is itself a function of time. which dt P dt is the rate of change at time t of the total volume of the region occupied by the material In evaluating (4. d d dv = vol{(P)}. one ﬁrst notes that the diﬀerentiation and integration operations cannot be directly interchanged. it is important to be able to manipulate integrals of (4. Taking this approach leads to d dt d ˜ φ dv = dt P = DR P0 AF ˆ φJ dV P0 P0 = d ˆ [φJ] dV dt ˆ ∂φ ˆ ∂J J +φ ∂t ∂t = ˙ ˆ (φJ + φJ div v) dV ˙ ˆ (φ + φ div v)J dV P0 = P0 P = ˙ ˜ (φ + φ div v) dv (4. Note that the theorem applies also to vector and tensor functions without any modiﬁcations. when following the set of particles that happen to occupy P at time t. Here. particles that at time t occupy P.11) This result is known as the Reynolds’ transport theorem. material time derivatives of volume integrals deﬁned in some subset of the current Example: Consider the integral in (4. In the the form d dt conﬁguration.10). Therefore. (pull-back to P0 ) d and dt ) (interchange dV (J˙ = J div v) (push-forward to P) .10) for φ = 1. note that the left-hand side of (4. forthcoming discussion of balance laws. To interpret the Reynolds’ transport theorem. evaluate the derivative of the integrand.10) namely. one needs to ﬁrst map the integral to the (ﬁxed) reference conﬁguration.11) is the rate of change of the integral of φ over the region P. and then map the integral back to the current conﬁguration.64 The Reynolds’ transport theorem ˆ ˜ Both φ and φ are assumed continuously diﬀerentiable in both of their variables.

12) as equals the boundary integral of the normal component of the velocity v · n of ∂P. Here.3). i. one may alternatively d dt ∂ ˜ φ dv = ∂t P ˜ φ dv + ∂P P ˜ φv · n da . To this end. ˜ Example: Consider again the special case φ(x. Here. d dt ˜ φ dv = P P ˙ (φ + φ div v) dv AF = P = P = P ˜ ∂φ dv + ∂t T ˜ ˜ ∂φ ∂φ + · v + φ div v dv ∂t ∂x ˜ ∂φ ˜ + div(φv) dv ∂t ∂P occupy P at time t. The second one is due to the rate of change of the volume occupied by ˜ the same particles assuming that φ remains constant in time for those particles.12) consists of the sum of two terms. note that φ dv. t) = 1. write with the aid of the divergence theorem (4. (4. ˜ φv · n da . The second term is the rate at which the volume of P weighted by φ changes as particles exit P across ∂P. ˜ ˜ ∂φ ∂φ ∂ ˜ Starting from (4.e.Basic physical principles 65 terms. the ¯ rate at which the volume of P changes as the particles exit across the boundary ∂ P of the ¯ ﬁxed region P which equals to P at time t. The ﬁrst one is the integral of the rate of change of φ for all particles that happen to The Reynolds’ transport theorem can be restated in a number of equivalent forms. since dv = is the rate of change of φ ∂t P ∂t P ∂t for ﬁxed position in space (hence. which corresponds to the transport of d dt volume.13) This means that the rate of change of the volume occupied by the same material particles express (4. the right-hand side of (4.12). (4.12) An alternative interpretation of the theorem is now in order. P is also ﬁxed in space). The ﬁrst term is the integral of the rate of change of φ at time t for all points that form the ﬁxed region P. Therefore. (4.14) ME185 . DR dv = P P div v dv = ∂P v · n da ..

66 The localization theorem 4. 2 Pδ 2 Pδ (4.19) 2 φ0 Now.16) provided that AF for all P ⊂ R at a given time t. This.18) that φ0 ˜ everywhere in Pδ . |x − x0 )| < δ(ε) .17) DR ME185 Now proceed by contradiction: assume that there exists a point x0 ∈ R.18) |φ(x. at any φ0 ˜ ˜ ﬁxed time t. Then. assume that ˜ φ dv = 0 . let φ be continuous function in the spatial argument. and only ˜ if. 2 such that φ0 ˜ ˜ ˜ . Let φ : R × R → R be a function such that φ = φ(x. such that ˜ ˜ |φ(x. t). then (4. The localization theorem can be also proved with ease for vector and tensor functions. φ = 0 everywhere in R at time t. To prove this result. invoking continuity of φ. The localization theorem states that this is true if. It follows from (4. where R ⊂ E 3 .3 The localization theorem Another important result with implications in the study of balance laws is presented here by ˜ ˜ way of background. t) − φ(x0 . Then. ˜ since. (4. note that continuity of φ in the spatial argument x at a point x0 ∈ R means that (4. t) − φ0 | < 2 whenever φ0 |x − x0 )| < δ( ) . t)| = |φ(x.15) for any time t and every ε > 0. . t)| < ε . ﬁrst note that the “if” portion of the theorem is straightforward. t) = φ0 > 0. deﬁne the region Pδ that consists of all points of R for which |x − x0 | < δ( ). such that.15) holds trivially true for any subset P of R. To prove the ˜ converse. This 2 3 is a sphere of radius δ in E with volume vol(Pδ ) = dv > 0. (4.20) which constitutes a contradiction. ˜ Also. Therefore. implies that φ(x. there is a δ = δ(ε) > 0. (4. if φ = 0 in R. the localization theorem holds. t) − φ(x0 . in turn. t) > 2 φ0 φ0 ˜ φ dv > dv = vol(Pδ ) > 0 . there exists a δ = δ( ). P T Pδ (4. φ(x0 .

2. mass of the part S of B. where Si ⊂ B. Assume here that region P. t). one may write ˇ f dm = ˜ f ρ dv Recall now that at time t the body B occupies a region R ⊂ E 3 and the part S occupies a AF ∞ i=1 B R m(Si ). for all S ⊆ B (i. t). mass. consider the body as a set of particles with positive there exists such a measure m and refer to m(B) as the mass of body B and m(S) as the A function m with the preceding properties is called a measure on B. Assuming that m is an absolutely continuous measure. and Si ∩ Sj = ∅. m non-negative) (ii) m(∅) = 0. . .22) The function ρ > 0 is termed the mass density.23) B R and dm = ρ dv = m(S) . known as the Radon-Nikodym theorem. (4.4 Mass and mass density Consider a body B and take any arbitrary part S ⊆ B. t) = f (x. As a special case. .4. for any function f = f(P. In other words.21) DR and S ˇ f dm = (4. Deﬁne a set function m : S → R with the following properties: (i) m(S) ≥ 0. i = 1.e. P (4. . it can be established that ˇ ˜ there exists a unique function ρ = ρ(x.4. Its existence is a direct consequence of a classical result in measure theory. m T ˜ f ρ dv .22) reduce to dm = ρ dv = m(B) (4..21) and (4. so that (4. if i = j (i. (iii) ∞ m(∪i=1 Si ) = countably additive). one may consider the function f = 1.e. MASS AND MASS DENSITY 67 4.24) ME185 S P .. such that.

28) where P0. ρ0 = lim (4.29) and dm = ρ0 dV = m(S) . see Figure 4. δ→0 vol(P0. (4.3: A limiting process used to deﬁne the mass density ρ at a point x in the current conﬁguration. AF x δ Sδ Pδ B ˇ f dm = ˆ f ρ0 dV B R0 Figure 4. Also. (4.δ ⊂ E 3 denotes a sphere of radius δ > 0 centered at X and Sδ the part of the body that occupies P0.27) DR S Here. such that at a given point X. one may write dm = ρ0 dV = m(B) B R0 (4. for any function f = f (P. S P0 ME185 . An analogous deﬁnition of mass density can be furnished in the reference conﬁguration.68 Mass and mass density The mass density of a particle P occupying point x in the current conﬁguration may be m(Sδ ) .δ ) a limiting process. t) = f(X. (4. P0 deﬁned by a limiting process as (4.26) and ˇ f dm = T R ˆ f ρ dV . as in the spatial case. the mass density ρ0 = ρ0 (X. ˇ ˆ where.30) The continuum hypothesis is crucial in establishing the existence of ρ and ρ0 . t).25) δ→0 vol(Pδ ) where Pδ ⊂ E 3 denotes a sphere of radius δ > 0 centered at x and Sδ the part of the body ρ = lim that occupies Pδ at time t.δ at time t0 .3. t) in the reference conﬁguration may be again deﬁned by m(Sδ ) .

Using the Reynolds’ transport theorem.35) (4. conservation of ∂ρ = 0. also as d dt ρ dv = P P ∂ρ ∂ρ + · v + ρ div v = 0 ∂t ∂x ∂ρ + div (ρv) = 0 .5 The principle of mass conservation The principle of mass conservation states that the mass of any material part of the body remains constant at all times. (4.37) ∂P 1 This is also sometimes referred to as the “continuity” equation. recalling (4.34) constitutes the local form of the principle of mass conservation in the spatial description. ˙ (4.34) Equation (4. with reference to the continuity of ﬂuid ﬂow in a control volume P. Hence. ˙ (4.24).32) ρ + ρ div v = 0 .5.4. ME185 . THE PRINCIPLE OF MASS CONSERVATION 69 4.36) Example: In a volume-preserving ﬂow of a material with uniform density. namely that d m(S) = 0 dt or.33) is continuous and recalling that S (hence. d dt T ρ dv = 0 . ∂P (4. one may write mass reduces to ∂t ∂ρ dv + ∂t ρv · n da = ρv · n da . the above equation may be written as [ρ + ρ div v] dv = 0 . upon recalling (4. also P) is AF P P (4. it follows from the localization theorem that Assuming that the integrand in (4.31) The preceding is an integral form of the principle of mass conservation in the spatial description.1 It can be readily rewritten as DR hence.33) arbitrary. ∂t (4.14).

70

The principles of linear and angular momentum balance An alternative form of the mass conservation principle can be obtained by recalling

T

ρ0 dV .

P0

**equations (4.24) and (4.30), from which it follows that m(S) =
**

P

ρ dv =

(4.38)

Recalling also (3.117), one concludes that

ρJ dV =

P0

ρ0 dV .

(4.39)

This is an integral form of the principle of mass conservation in referential description. From it, one ﬁnds that (ρJ − ρ0 ) dV = 0 . (4.40)

Taking into account the arbitrariness of P0 , the localization theorem may be invoked to yield a local form of mass conservation in referential description as ρ0 = ρJ .

AF

P0

P0

(4.41)

4.6

The principles of linear and angular momentum balance

DR

of Newton and Euler.

ME185

Once mass conservation is established, the principles of linear and angular momentum are postulated to describe the motion of continua. These two principles originate from the work By way of background, review Newton’s three laws of motion, as postulate for particles in

1687. The ﬁrst law says that a particle stays at rest or continues at constant velocity unless an external force acts on it. The second law says that the total external force on a particle is proportional to the rate of change of the momentum of the particle. The third law says that every action has an equal and opposite reaction. As Euler recognized, Newton’s three laws of motion, while suﬃcient for the analysis of particles, are not strictly appropriate for the study of rigid and deformable continua. Rather, he postulate a linear momentum balance law (akin to Newton’s second law) and an angular momentum balance law. The latter can be easily motivated from the analysis of systems of particles.

Basic physical principles

71

To formulate Euler’s two balance laws, ﬁrst deﬁne the linear momentum of the part of the mass of dv. Also, deﬁne the angular momentum of the same part relative to the origin of the ﬁxed basis {ei } as x × (dmv), where x is the position vector associated with the inﬁnitesimal volume element. Similarly, deﬁne the linear and angular momenta of the part S which occupies a region P at time t as Next, admit the existence of two types of external forces acting on B at any time t.

S

act on the particles which comprise the continuum, and (b) contact forces per unit area t = t(x, t; n) = t(n) (x, t), which act on the particles which lie on boundary surfaces and depend on the orientation of the surface on which they act through the outward unit normal n to the surface. The force t(n) is alternatively referred to as the stress vector or the traction vector. The principle of linear momentum balance states that the rate of change of linear momentum for any region P occupied at time t by a part S of the body equals the total external forces acting on this part. In mathematical terms, this means that d dt v dm = b dm + t(n) da

AF

S S ∂P

There are: (a) body forces per unit mass (e.g., gravitational, magnetic) b = b(x, t) which

T

v dm and

S

the body that occupies the inﬁnitesimal volume element dv at time t as dmv, where dm is

x × v dm, respectively.

(4.42)

or, equivalently,

DR

P P

d dt

ρv dv =

ρb dv +

t(n) da .

(4.43)

∂P

**Using conservation of mass, the left-hand side of the equation can be written as d dt ρv dv = d (ρv) dv + dt (ρv) div v dv
**

P

P

P

=

˙ (ρv + ρv) dv + ˙

P

(ρv) div v dv

P

=

˙ [(ρ + ρ div v)v + ρv] dv ˙ ρa dv , (4.44)

P

=

P

**hence, the principle of linear momentum balance can be also expressed as ρa dv = ρb dv +
**

∂P

t(n) da .

(4.45)

ME185

P

P

72

The principles of linear and angular momentum balance The principle of angular momentum balance states that the rate of change of angular

external forces acting on this part. Again, this principle can be expressed mathematically as d dt or, equivalently, d dt x × ρv dv = x × v dm = x × b dm + x × t(n) da (4.46)

S

S

Again, appealing to conservation of mass, one may easily rewrite the term on the left-hand side of (4.47) as d dt

P

x × ρv dv =

AF

P P ∂P P

x × ρb dv +

T

∂P

momentum for any region P occupied at t by a part S of the body equals the moment of all

x × t(n) da .

(4.47)

d (x × ρv) + (x × ρv) div v dt

dv

=

P

˙ ˙ {[x × ρv + x × ρv + x × ρv] + (x × ρv div v)} dv ˙ {x × (ρ + ρ div v)v + x × ρa} dv ˙ x × ρa dv .

=

P

=

(4.48)

P

As a result, the principle of angular momentum balance may be also written as

DR

P P

x × ρa dv =

x × ρb dv +

∂P

x × t(n) da .

(4.49)

The preceding two balance laws are also referred to as Euler’s laws. They are termed

“balance” laws because they postulate that there exists a balance between external forces (and their moments) and the rate of change of linear (and angular) momentum. Euler’s laws are independent axioms in continuum mechanics. In the special case where b = 0 in P and t(n) = 0 on ∂P, then equations (4.43) and

Another commonly encountered special case is when the velocity v vanishes identically. In this case, equations (4.43) and (4.47) imply that the sum of all external forces and the some of all external moments vanish, which gives rise to the classical equilibrium equations of statics.

(4.47) readily imply that the linear and the angular momentum are conserved quantities in P.

ME185

43) and (4. Now.51) P2 P2 Subsequently. it is desirable to obtain local forms of linear and angular momentum balance. By way of background. To this end.4: Setting for a derivation of Cauchy’s lemma.52) ME185 P1 ∪P2 P1 ∪P2 . note that the boundaries ∂P1 and ∂P2 of P1 and P2 . Therefore. in order to apply the localization theorem. add the two equations together to ﬁnd that d dt ρv dv = ρb dv + ∂P1 ∪∂P2 t(n) da (4.4. enforce linear momentum balance ρv dv = t(n) da (4. while the body force terms are already in the form of volume integral.4. respectively. Also. Recalling the corresponding integral statements (4. take an arbitrary region P ⊆ R and divide it into two mutually disjoint subregions P1 and P2 by means of an arbitrary smooth surface σ.47).7 Stress vector and stress tensor As in the case of mass balance. namely P = P1 ∪ P2 and P1 ∩ P2 = ∅. it is clear that the Reynolds’ transport theorem may be employed in the rate of change of momentum terms to yield volume integrals. see DR separately in P1 and P2 to ﬁnd that d dt d dt P1 Figure 4. it is essential equivalent volume integral terms. consider some properties of the traction vector t(n) . STRESS VECTOR AND STRESS TENSOR 73 4.50) P1 and ρv dv = ρb dv + ∂P2 t(n) da (4. can be expressed as ∂P1 = ∂P ′ ∪ σ and ∂P2 = ∂P ′′ ∪ σ.7. that the contact form terms (presently written as surface integrals) be transformed into P1 AF P2 P1 n1 = n σ n2 σ σ ∂P ′ T P2 ρb dv + ∂P1 ∂P ∂P ′′ Figure 4.

ME185 Since σ is an arbitrary surface. (4.59) σ σ DR the localization theorem yields the condition Third Law. (4.61) It is important to recognize here that in continuum mechanics Cauchy’s lemma is not a principle. where action-reaction is a principle. (4. (4. enforce linear momentum balance on the union of P1 and P2 to ﬁnd that d dt ρv dv = P P P P ρb dv + T ∂P1 ∪∂P2 d dt ρv dv = ρb dv + t(n) da . (4. it is derivable from linear momentum balance.58) σ σ which can be also written as t(n) da + t(−n) da = 0 .60) This result is called Cauchy’s lemma on t(n) . −n) . t. ∂P ′ ∪∂P ′′ σ σ ∂P t(n1 ) da + t(n2 ) da = 0 .54) ∂P Subtracting (4. It states that the stress vectors acting at x on opposite sides of the same smooth surface are equal and opposite. t. ﬁnally. the preceding equation may be also expressed as t(n) da + t(n) da = t(n) da (4.55) t(n) da + t(n1 ) da + t(n2 ) da = t(n) da .53) leads to Recalling the decomposition of ∂P1 and ∂P2 . (4.57) so. t(n) + t(−n) = 0 .54) from (4. n) = −t(x. also known as Newton’s .74 or Stress vector and stress tensor Further.53) t(n) da . (4. as above. assuming that t depends continuously on n and x along σ.56) or AF ∂P1 ∪∂P2 ∂P ∂P ′ ∪σ ∂P ′′ ∪σ ∂P t(n) da = t(n) da . Rather. (4. This is in stark contrast with particle mechanics. namely that t(x.

5.65) P σ0 σ1 σ2 Assuming that ρ. becomes Now apply balance of linear momentum to the tetrahedral region P in the form of equat(n) da = t(−e3 ) da + σ0 DR ∂P σ1 σ2 t(n) da . as in Figure 4. which. the volume V of the tetrahedron can 1 be written as V = Ah.45).5: The Cauchy tetrahedron. where h is the distance of x from face σ0 . in this case.62) e1 AF e3 x e2 Figure 4. σ3 (4. In addition. t(−e1 ) da + t(−e2 ) da + σ3 where Ai = Ani is the area of the face σi . consider the following problem. one can obtain an upper-bound estimate for the domain integral on the left-hand side as ρ(a − b) dv ≤ 1 K(x. originally conceived by Cauchy: take a at a point x. such that three edges are parallel to the axes of {ei } and meet (4. 3 R tion (4. a.66) ME185 . (4. and σ0 the (inclined) face with outward unit normal n. T 2 1 3 tetrahedral region P ⊂ R.45) and concentrate on the surface integral term. t) dv = K ∗ V = K ∗ Ah . (4.63) Upon using Cauchy’s lemma. Let σi be the face with unit outward normal −ei . so that the area vector ndA can be resolved as nA = (ni ei )A = Ani ei = Ai ei .64) Returning now to the balance of linear momentum statement (4. 3 P P P |ρ(a − b) dv| = (4. and b are bounded.Basic physical principles 75 At this stage. Denote by A the area of σ0 . this becomes t(n) da = − t(e1 ) da − ∂P σ1 σ2 t(e2 ) da − t(e3 ) da + σ3 σ0 t(n) da . it can be written as ρ(a − b) dv = t(n) da − t(e1 ) da − t(e2 ) da − t(e3 ) da .

76

Stress vector and stress tensor

where K(x, t) = |ρ(a − b)| and K ∗ = K(x∗ , t), with x∗ being some interior point of P. The Assuming that t(ei ) are continuous in x, apply the mean value theorem for integrals component-wise to get

σi

t(ei ) da = t∗ Ai , i

T

1 ∗ K Ah , 3 1 ∗ K Ah , 3 1 ∗ K h. 3

f (x) dv = f (x∗ ) vol(P).

P

preceding derivation made use of the mean-value theorem for integrals.2

(4.67)

**so that summing up all three like equations
**

3 σi

AF

i=1 σ0

t(ei ) da = t∗ Ai = t∗ Ani . i i

(4.68)

Also, for the inclined face

t(n) da = t∗ A . (n)

(4.69)

In the preceding two equations, t∗ and t∗ are the traction vectors at some interior points i (n) of σi and σ0 . Recalling from (4.65) and (4.66) that

3

σ0

t(n) da −

i=1

σi

t(ei ) da ≤

(4.70)

write

3 σ0

t(n) da −

i=1

σi

t(ei ) da = |t∗ A − t∗ Ani | = A |t∗ − t∗ ni | ≤ i i

(4.71)

DR

|t∗ − t∗ ni | ≤ (n) i point x

2

which simpliﬁes to

(4.72)

Now, upon applying the preceding analysis to a sequence of geometrically similar tetra-

hedra with heights h1 > h2 > . . ., where limi→∞ hi = 0, one ﬁnds that |t∗ − t∗ ni | ≤ 0 , (n) i (4.73)

where, obviously, all stress vectors are evaluated exactly at x. It follows from (4.73) that at t(n) = ti ni , (4.74)

and f is continuous in ε3 , then there exists a point x∗ ∈ P for which

Mean-value Theorem for Integrals: If P has positive volume (vol(P) > 0) and is compact and connected,

ME185

Basic physical principles where the superscript has been dropped, since, now, x∗ = x.

77

T = ti ⊗ ei , so that, when operating on n,

T

ρb dv +

∂P

With the preceding development in place, deﬁne the tensor T as

(4.75)

Tn = (ti ⊗ ei )n = ti(ei · n) = ti ni = t(n) , as seen from (4.74). The tensor T is called the Cauchy stress tensor. It can be readily seen from (4.75) that

(4.76)

AF

T = Tki ek ⊗ ei = ti ⊗ ei , ti = Tki ek . ti · ej = Tki ek · ej = Tji , Tij = ei · tj . ρa dv = t(n) da Tn da

∂P P P

(4.77)

hence

(4.78)

Also, since

(4.79)

it is immediately seen that

(4.80)

Note that, from its deﬁnition in (4.75), it is clear that the Cauchy stress tensor does not

DR

depend on the normal n. =

P

Return now to the integral statement of linear momentum balance and, taking into

account (4.76), apply the divergence theorem to the boundary integral term. This leads to

ρb dv +

=

ρb dv +

P

div T dv .

(4.81)

P

**It follows from the preceding equation that the condition (ρa − ρb − div T) dv = 0 , (4.82)
**

ME185

P

78

Stress vector and stress tensor

holds for an arbitrary area P, which, with the aid of the localization theorem leads to a local

div T + ρb = ρa .

T

t(n) da

∂P

form of linear momentum balance in the form

(4.83)

An alternative statement of linear momentum balance can be obtained by noting from (4.75) that ρa dv =

AF

P P

ρb dv +

=

ρb dv +

ti ni da

P

∂P

=

ρb dv +

ti,i dv .

(4.84)

P

P

Again, appealing to the localization theorem, this leads to ti,i + ρb = ρa .

(4.85)

Turn attention next to the balance of angular momentum and examine the boundary integral term in (4.49). This can be written as x × t(n) da =

∂P

∂P

x × ti ni da =

P

(x × ti ),i dv =

P

(x,i × ti + x × ti,i) dv . (4.86)

Substituting the preceding equation into (4.49) yields

DR

P

(x × ρa) dv =

P

(x × ρb) dv +

P

(x,i × ti + x × ti,i ) dv

(4.87)

or, upon rearranging the terms,

P

[x × (ρa − ρb − ti,i ) + x,i × ti ] dv = 0 .

(4.88)

Recalling the local form of linear momentum balance equation (4.85), the above equation reduces to x,i × ti dv = 0 . (4.89)

The localization theorem can be invoked again to conclude that x,i × ti = 0 (4.90)

P

ME185

Basic physical principles or

79

In component form, this condition can be expressed as

ei × (Tjiej ) = Tji ei × ej = Tji eijk ek = 0 ,

T

T23 T32 T31 T22 T21 T12 T11 T11 T12 T13

ei × ti = 0 .

(4.91)

(4.92)

which means that Tij = Tji , i.e., the Cauchy stress tensor is symmetric (or, said diﬀerently, the skew-symmetric tensor T − TT vanishes identically). Hence, angular momentum balance restricts the Cauchy stress tensor to be symmetric. T33

Figure 4.6: Interpretation of the Cauchy stress components on an orthogonal parallelepiped

DR

can be put in matrix form as

aligned with the {ei }-axes.

An interpretation of the components of T on an orthogonal parallelepiped is shown in

Figure 4.6. Indeed, recalling (4.78), it follows that t1 = T11 e1 + T21 e2 + T31 e3 , (4.93)

which means that Ti1 is the i-th component of the traction that acts on the plane with outward unit normal e1 . More generally, Tij is the i-th component of the traction that acts on the plane with outward unit normal ej . The components Tij of the Cauchy stress tensor

AF

T13 e3 e2 e1 [Tij ] = T21 T22 T23 , T31 T32 T33

(4.94)

ME185

99) DR ME185 as in Figure 4.e.7: Projection of the traction to its normal and tangential components. IIT and IIIT are the three principal invariants of the tensor T. (4.98) As is well-known. . 6 AF (t(n) · n)n = (n ⊗ n)t(n) .97) (4. the projection of t(n) along n) is given by (4. The traction vector t(n) can be generally decomposed into normal and shearing components on the plane of its action. deﬁned as (4. The linear eigenvalue problem (T − T i)n = 0 T t(n) (4. whether the eigenvalues are distinct or not.96) IT = tr(T) 1 IIT = [(tr(T))2 − tr T2 ] 2 1 IIIT = [(tr(T))3 − 3 tr T(tr T2 ) + 2 tr T3 ] = det T . Then.7.80 Stress vector and stress tensor where [Tij ] is symmetric. Also. the normal traction (i.95) yields three real eigenvalues T1 ≥ T2 ≥ T3 . the associated unit eigenvectors n(1) . the shearing traction is equal to Figure 4. n(2) and n(3) of T are mutually orthogonal provided the eigenvalues are distinct. there exists a set of mutually orthogonal eigenvectors for T. which are solutions of the characteristic polynomial equation T 3 − IT T 2 + IIT T − IIIT = 0 . To this end.. n (t(n) · n)n where IT .

the stress vector is always pointing in the direction normal to any plane that it is acting on. The tractions ti are now given by (4. (4. i. consider three special forms of the stress tensor that lead to equilibrium states in In this state. to (4. without loss of generality.Basic physical principles t(n) − (t(n) · n)n = t(n) − (n ⊗ n)t(n) = (i − n ⊗ n)t(n) . .e.78) that (c) Pure shear on the (e.76) that AF t1 = T e1 T = T (e1 ⊗ e1 ) = T (e ⊗ e) . let e and k be two orthogonal vectors of unit magnitude and. again. t1 = T e2 .103) (b) Pure tension along the e-axis Without loss of generality. (4. T T = −pi .101) namely that the shearing traction vanishes on the plane with unit normal n.78).106) Appealing.102) where p is called the pressure. t2 = T e1 . the absence of body forces. t(n) = −pn . the traction vectors ti are of the form (4. (4.104) DR Then.. t2 = t3 = 0 .100) (i − n ⊗ n)t(n) = (i − n ⊗ n)Tn = (i − n ⊗ n)T n = 0 . equations (4. (a) Hydrostatic pressure Next. It follows from (4. k)-plane (4. t3 = 0 . set e1 = e and e2 = k.100) imply that 81 (4.107) ME185 . In this case.95) and (4. it is easily seen that T = T (e1 ⊗ e2 + e2 ⊗ e1 ) = T (e ⊗ k + k ⊗ e) . If n is a principal direction of T.105) Here. let e = e1 . (4. it follows from (4.

for example.114) .111). (4. This is conceivable when.e. when operating on N. Further.108) Also.45). (4. since they are both parallel to df. start by letting df be the total force acting on the diﬀerential area da with outward unit normal n on the surface ∂P in the current conﬁguration. on the surface ∂P0 . (4.82 Stress vector and stress tensor It is possible to resolve the stress vector acting on a surface of the current conﬁguration one wishes to measure the internal forces developed in the current conﬁguration per unit area of the reference conﬁguration. df = t(n) da . let dA be the image of da in the reference conﬁguration under χ−1 and assume that t its outward unit is N. AF ρ0 a dV = ρ0 b dV + p(N) dA . note that this can be now readily “pulled-back” to the reference conﬁguration. df = p(N) dA . it is seen that ME185 p(N) = pA NA . it is immediately concluded that DR let the tensor P be deﬁned as so that. t and p are parallel. hence taking the form (4. (4. (4. Then. (4. deﬁne p(N) to be the traction vector resulting from resolving the force df. which acts on ∂P .110) Upon applying the Cauchy tetrahedron argument to P0 ..113) p = PN . P = pA ⊗ EA . thus. Returning to the integral statement of linear momentum balance in (4.109) Clearly. i. but resolved on the geometry of the reference conﬁguration on surfaces with outward unit normals EA . from (4. To this end. namely. P0 P0 ∂P0 T using the geometry of the reference conﬁguration.112) PN = (pA ⊗ EA )N = pA NA .111) where pA are the tractions developed in the current conﬁguration.

. P = PiA ei ⊗ EA . PiA = ei · pA . upon using the localization theorem. since (4.118) it is clear that (4. it is concluded with the aid of (4.e.112) that T ρ0 b dV + p(N) dA PN dA ∂P0 since it has a mixed basis.110).120) DR P0 P0 which.113) and the divergence theorem can be invoked to show that ρ0 a dV = ρ0 b dV + ∂P0 p(N) dA pA NA dA ∂P0 P0 P0 = ρ0 b dV + ρ0 b dV + P0 P0 = pA. Alternatively. i. (4.119) Turning attention to the integral statement (4.121) This is the local form of linear momentum balance in the referential description.117) Further. which implies that (4.116) AF pA = PiA ei .114) and the divergence theorem that = ρ0 b dV + P0 ∂P0 = ρ0 b dV + Div P dA (4. It follows from (4. pA · ej = PiA ei · ej = PjA .Basic physical principles 83 The tensor P is called the ﬁrst Piola-Kirchhoﬀ stress tensor and it is naturally unsymmetric.122) P0 ME185 . results in ρ0 a = ρ0 b + Div P . (4. ρ0 a dV = P0 P0 (4. equation (4.115) P = pA ⊗ EA = PiA ei ⊗ EA .A dA (4.

113) and the divergence theorem on the boundary term gives rise to x × ρ0 a dV = x × ρ0 b dV + x × pA NA dA P0 AF P0 ∂P0 P0 P0 x × ρ0 b dV + x × p(N) dA . (4.A × pA ] dV = 0 .A × pA = FiA × PjA ei × ej = FiA PjA eijk ek = 0 . (4.130) . (4.114) and PNdA = Tnda = TJF−T NdA . T ∂P0 derived in the form (4.A dV . (4. subsequently.123) and. one ﬁnds that x × ρ0 a dV = Using (4.A × pA = 0 . This is a local form of angular momentum balance in the Recalling (4.A .129) 1 PFT . the preceding equation can be rewritten as x.117) and the chain rule.125) Expanding and appropriately rearranging the terms of the above equation leads to [x × (ρ0 a − ρ0 b − PA.A ) + x. the localization theorem.123) Starting from the integral form of angular momentum balance in (4.127) DR referential description. J (4.76). (4.84 Stress vector and stress tensor from which another version of the referential statement of linear momentum balance can be ρ0 a = ρ0 b + pA.122) that so that T = ME185 With the aid of (4. (4.126) P0 Appealing to the local form of linear momentum balance in (4. (4. one may conclude with the aid of (4.108) and (4. Nanson’s formula (3.124) = P0 x × ρ0 b dV + P0 (x × pA ). one concludes that x.128) which implies that FPT = PFT .108).49) and pulling it back to the reference conﬁguration.

n).e.166). deﬁned as (4. Likewise.133) that S has both “legs” in the reference conﬁguration and is symmetric. start with the stress vector t = t(x. the most important are the nominal stress tensor Π deﬁned as the transpose of the ﬁrst Piola-Kirchhoﬀ stress. All the stress measures deﬁned here are obtainable from one another assuming that the 4. namely (4. τ = JT = PFT . deﬁned as (4. write ME185 . S = F−1 P = JF−1 TF−T .131) the Kirchhoﬀ stress tensor τ .. such as the stress vectors and tensors. Note that there is no approximation or any other source of error associated with the use the balance laws in the referential vs. T gular momentum balance. spatial description. To this end. the invertibility of equivalent. the above relation is consistent with the referential and spatial statements of anmomentum balance in spatial form from the referential statement and vice-versa. Indeed. the transformation under superposed rigid motions is considered for mechanical ﬁelds. and recalling the general form of the superposed rigid motion in (3. AF Π = PT = JF−1 T . the motion at any ﬁxed time implies that both forms of the balance laws are completely Other stress tensors beyond the Cauchy and ﬁrst Piola-Kirchhoﬀ tensors are frequently used.130) can be used to derive the local form of angular (4.Basic physical principles 85 Clearly.8 The transformation of mechanical ﬁelds under su- perposed rigid-body motions In this section. spatial) form from its corresponding spatial (resp. referential) counterpart. t.132) and the second Piola-Kirchhoﬀ stress S. Among them.133) DR deformation is known. it is possible to derive the local linear momentum balance statement in the referential (resp. It is clear from (4. i.

134) implies that DR from where it is concluded that hence ME185 AF t+ = Qt = QTn = T+ n+ = T+ Qn .76).136) (4.134) Unlike the transformation of kinematic terms. Since the two motions give rise to the same defor- (4. The above transformation indeed implies that |t+ | = |t| and t+ · n+ = t · n.130). then the relation between the stress vector t on a plane with an outward unit normal n and the Cauchy stress tensor T is the same with the relation of the stress vector t+ on a plane with outward unit normal n+ with the Cauchy stress tensor T+ . (QT − T+ Q)n = 0 . states that the balance laws are invariant under superposed rigid motions. ﬁrst recall that n+ = Qn and that mation. T+ = QTQT .76) under superposed rigid motions in conjunction with (4. To argue how t and t+ are related. in the sense that their mathematical representation remains unchanged under such motions. consider the transformation of the Cauchy stress tensor. n+ ).137) Equation (4. which is governed purely by geometry. eﬀectively.135) (4. invariance under superposed rigid motions implies that t+ = T+ n+ . To demonstrate this principle. Therefore. t+ will not change in magnitude relative to t and will have the same orientation relative to n+ as t has relative to n. T t is linear in n. t. This means that if two motions diﬀer by a mere rigid motion. as established in (4. it is then reasonable to assume that.86 Mechanical ﬁelds under superposed rigid motions t+ = t+ (x+ . This. in the case of kinetic terms (both mechanical and thermal) the transformation is governed by the principle of invariance under superposed rigid motion. Taking taking into account (4. Given that this relation is itself invariant under superposed rigid motions it follows . under a superposed rigid motion. Recall next the relation between the Cauchy and the ﬁrst Piola-Kirchhoﬀ stress tensor in (4. Invariance of (4. it is postulated that t+ = Qt .137) implies that T is an objective spatial tensor.76). (4.

139) holds true.180) it can be seen that DR which imply that 3 (4. (F−1 )+ = (F+ )−1 . ˙ ˙ ˙ ˙ T+ = QTQT + QTQT + QTQT ˙ = (ΩQ)TQT + QTQT + QT(ΩQ)T ˙ = Ω(QTQT ) + QTQT + (QTQ)T ΩT ˙ = ΩT+ + QTQT − T+ Ω . ME185 . (4. ρ+ = ρ . note that.3 Equation (4. it follows from (4. Invariance under superposed rigid motions applies to the principle of mass conservation. ρ0 = ρ+ J + = ρ+ J = ρJ .143) Also. Similarly. from (4.141) ˙ which shows that. S+ = J + (F−1 )+ T+ (F−T )+ = J(F−1 QT )(QTQT )(QF−T ) = JF−1 TF−T = S .142) (4.187) are employed.133) that which implies that S is an objective referential tensor.41) of this principle gives rise to (4. Since (4. 87 T (4. Indeed. unlike T. turning to the second PiolaKirchhoﬀ stress tensor.Basic physical principles that P+ = J + T+ (F−T )+ = J(QTQT )(QF−T ) = Q(JTF−T ) = QP .138) implies that P is an objective two-point tensor. A similar conclusion may ˙ be drawn for the rate P of the ﬁrst Piola-Kirchhoﬀ stress tensor. using the local referential form (4.140) namely that the rate of S is objective. it follows also that AF ˙ ˙ S+ = S .165) and (3.138) where the kinematic transformations (3. by its deﬁnition.139) (4. However.137) and the relation (3. the spatial tensor T is not objective.

one concludes that DR from where it follows that (4. under superposed rigid motions.166) that ∂χ ∂xm = QT . note that T of linear momentum balance leads to (4. = Qjm . ∂x+ ∂x+ j Equation (4.145) can be written using direct notation as (4.147) This means that.83) div+ T+ + ρ+ b+ = ρ+ a+ . = Qik ∂xl (4.148) 4. the body forces transform as (4.144) and (4.146).88 Mechanical energy balance theorem Finally. admitting invariance under superposed rigid motions of the local spatial form (4.144) AF div+ T+ = Q div T . div+ T+ = ρ+ (a+ − b+ ) = ρ(a+ − b+ ) = Q div T = Qρ(a − b) . b+ = Qb + a+ − Qa . a+ − b+ = Q(a − b) .9 The Theorem of Mechanical Energy Balance region P with smooth boundary ∂P. therefore. Consider again the body B in the current conﬁguration R and take an arbitrary material ME185 . in components.146) Using (4. Resorting to components. + ∂Tij ∂(Qik Tkl Qjl ) ∂χm + = ∂xm ∂xj ∂x+ j ∂Tkl Qjl Qjm = Qik ∂xm ∂Tkl = Qik δlm ∂xm ∂Tkl . (4.145) where it is recognized from (3.83).

DR (4.156) or. 2 (4. deﬁne the rate of work done by all external forces as AF R(P) = Rb (P) + Rc (P) .151) In addition. ρb · v dv (4.153) Now.154) can be used to rewrite (4. upon using conservation of mass and the divergence theorem. deﬁne the total kinetic energy of the material points contained in P as 1 v · vρ dv . (4. note that (div T) · v = div(TT v) − T · grad v = div(Tv) − T · D . Equation 1 d ρ (v · v) = ρb · v + div(Tv) − T · D . deﬁne the rate at as Rb (P) = and T P which the external body forces b and surface tractions t do work in P and on ∂P. respectively. one may dot both sides with the velocity v to obtain ˙ ρv · v = ρb · v + div T · v .154) where angular momentum balance is invoked in the form of the symmetry of T.43).149) Rc (P) = ∂P t · v da .Basic physical principles 89 Recalling the integral statement of linear momentum balance (4. (4. K(P) = P (4. (4. 2 dt (4.155) over P leads to P = div(Tv) − T · (D + W) (4.150) respectively. Also. d dt 1 ρ(v · v) dv + 2 T · D dv = ρb · v dv + (Tv) · n da .83). (4.157) ME185 P P P ∂P .152) Starting from the local spatial form of linear momentum balance (4.153) as Integrating (4.155) 1 d ρ (v · v) dv = 2 dt P ρb · v dv + P div(Tv) dv − P T · D dv .

90 Mechanical energy balance theorem Recalling (4. (4. the preceding equation can be rewritten as d dt 1 ρ(v · v) dv + 2 T · D dv = T P P P ρb · v dv + ∂P t · v da . note that DR T · D dv = P T · L dv 1 PFT J · L dv = P = P0 (PFT ) · L dV tr[(PFT )LT ] dV = P0 = tr[P(LF)T ] dV P0 = ˙ tr[PFT ]dV ˙ P · F dV . (4.76).159) is called the stress power and it represents the rate at which the stresses do work in P.159) and (4. (4.158)) states that. (4. dt AF P (4. for any region P.160) of change of the kinetic energy and the stress power are balanced by the rate of work done is important to emphasize that mechanical energy balance is derivable from the three basic Equation (4.161) P0 = P0 ME185 . It principles of the mechanical theory. Returning to the stress power term S(P).158).160) (or. S(P) = P T · D dv .150). namely conservation of mass and balance of linear and angular momentum.158) The second term on the right-hand side of (4.152). equation (4. This is a statement of the mechanical energy balance theorem. equivalently. (4. it is seen that d K(P) + S(P) = Rb (P) + Rc (P) = R(P) . the rate by the external forces.158). (4. Taking into account (4.149).

Next. (4. S is work-conjugate to E.10 The principle of energy balance The physical principles postulated up to this point are incapable of modeling the interconvertibility of mechanical work and heat. t) called the heat supply per unit mass (or speciﬁc heat supply). assume the existence of a scalar function ε = ε(x. likewise. Further. These equations appear to leave open the 4.128).Basic physical principles where use is made of (4. Now. called the (speciﬁc) internal energy. in P0 and.161) and (4. deﬁne a scalar ﬁeld h = h(x. T · D dv = = = = = 91 P AF P0 T P0 P0 ˙ P · F dV ˙ (FS) · F dV P0 ˙ S · (FT F) dV 1 ˙ ˙ S · (FT F + FT F) dV 2 P0 ˙ S · E dV . this time use is made of (4. DR H(P) = P Preliminary to stating the balance of energy.162) where. which quantiﬁes the rate at which heat called the heat ﬂux per unit area across a surface ∂P with outward unit normal n. t) given any region P ⊆ R.130) and (3. is supplied (or absorbed) by the body. t. ∂P (4.162) reveal that P is the work-conjugate kinetic measure to F question of work-conjugacy for T.133) and the deﬁnition of the Lagrangian strain. deﬁne a scalar ﬁeld r = r(x.163) where the negative sign in front of the boundary integral signiﬁes the fact that the heat ﬂux is assumed positive when it exits the region P. t) per unit mass. n) = h(n) (x. one needs to introduce an additional primitive law known as balance of energy. In order to account for this class of (generally coupled) thermomechanical phenomena. deﬁne the total rate of heating H(P) as ρr dv − h da . Equations (4. This function quantiﬁes all forms of energy stored in the body ME185 . Also.

also. d dt d dt ρε dv = P1 P1 and P1 ∪ P2 = ∅. ∂P2 = ∂P ′′ ∪ σ have a common T · D dv + P ρr dv − h da . as already done with the stress vector ∂P and to each of two regions P1 and P2 with boundaries ∂P1 and ∂P2 . etc. and note that one can apply a standard argument to ﬁnd the exact dependence of h on n. one may apply thermal energy balance to a region P with boundary DR and.. where P1 ∪ P2 = P surface σ and ∂P ′ ∪ ∂P ′′ = ∂P.168) T · D dv + T · D dv + P1 ρr dv − ρr dv − h da ∂P1 (4. ∂P (4.167) Returning to the heat ﬂux h = h(x.158) from (4. chemical energy. Also. equation (4.165) ∂P (4. Equivalently.92 The principle of energy balance other than the kinetic energy.170) P2 P2 P2 Adding the last two equations leads to d dt ρε dv = P1 ∪P2 P1 ∪P2 T · D dv + P1 ∪P2 ρr dv − h da ∂P1 ∪∂P2 (4.171) ME185 . dt ρε dv = P P P P P ρr dv − h da .165) can be written as AF d [K(P) + U(P)] = R(P) + H(P) . n). It follows that d dt ρε dv = P P P ∂P t = t(x.164) The principle of balance of energy is postulated in the form d dt 1 ρv · v + ρε dv = 2 ρb · v dv + t · v da + This is also sometimes referred to as a statement of the “First Law of Thermodynamics”.e. t.169) and ρε dv = h da ∂P2 (4. n). (4.165) leads to a statement of thermal energy balance in the form d dt T · D dv + ρr dv − h da . t. Examples of stored energy include strain energy (i. energy U(P) = T ρε dv . (4. Indeed. The internal energy stored in P is denoted by (4. P ∂P due to deformation). the boundaries ∂P1 = ∂P ′ ∪ σ.166) Subtracting (4.

Thus. n) = −h(x. one may write h(x. (4.176) where n1 = n and n2 = −n.174) As in the case of the stress vector. ∂P ′ ∪σ ∂P ′′ ∪σ ∂P or. the preceding equation may be expanded to h da + h(n1 ) da + h(n2 ) da = h da (4.175) ∂P ′ ∪∂P ′′ σ σ ∂P or σ (h(n) − h(−n) )da = 0 . DR across the same surface.180) ME185 . (4.172) results in ∂P1 ∪∂P2 h da − h da = 0 .179) where hi are the ﬂuxes across the faces of the tetrahedron with outward unit normals ei .177) or. gives rise to h = hi ni (4. equivalently.173) ∂P AF h da + h da = h da . Since σ is an arbitrary surface and h is assumed to depend continuously on n and x along σ. (4. more explicitly. d dt ρε dv = P P 93 T · D dv + T P ρr dv − h da . (4. (4.178) This states that the ﬂux of heat exiting a body across a surface with outward unit normal n at a point x is equal to the ﬂux of heat entering a neighboring body at the same point Using the tetrahedron argument in connection with the thermal energy balance equation (4. h = q·n . t. −n) .172) ∂P1 ∪∂P2 Subtracting (4.178).Basic physical principles or. equivalently.168) from (4. (4. the localization theorem yields the condition h(n) = −h(−n) . t. (4.167) and the ﬂux continuity equation (4.

returning to the integral statement of energy balance in (4.182) [(div T) · v + T · D] dv (4.180). it is postulated that ε+ = ε (4. the above equation may be put in the form ˙ (ρv · v + ρε) dv = ˙ ρb · v dv + (Tn) · v da + ρr dv − P Upon invoking the divergence theorem. (4. and δW to T · D.185) Upon observing linear momentum balance and invoking the localization theorem.181) ∂P Using (4. use mass conser- P P ∂P ρr dv − h da .165). where dU corresponds to ρε.186) This equation could be also derived along the same lines from the integral statement of Regarding invariance under superposed rigid-body motions. δQ to ρr − div q. ˙ ME185 .182). one ﬁnds that P P ˙ {ρv − ρb − div T} · v + ρε − T · D − ρr + div q dv = 0 . (4. (4.184) If the last two equations are substituted in (4. ˙ (4.183) and ∂P q · n da = div q dv . (4. ˙ (4.187) q+ = Qq . the preceding equation gives rise to the local form of energy balance as DR thermal energy balance (4. it is easily seen that (Tn) · v da = AF P ∂P P ∂P ∂P P q · n da . vation to rewrite it as ˙ (ρv · v + ρε) dv = ˙ ρb · v dv + T t · v da + P Now.76) and (4.167).4 and r+ = r .94 The principle of energy balance where q is the heat ﬂux vector with components qi = hi .188) The energy equation is frequently quoted in elementary thermodynamics texts as “dU = δQ + δW ”. 4 ρε = T · D + ρr − div q .

194) ∂P + Now. where rigidity implies that F = R (i.193) which is the classical equation of transient heat conduction. so that D = 0. (4. c = c0 t .Basic physical principles Equation (4. ˙ Q = 0 . U = I).186) reduces to ρε − div(k grad T ) − ρr = 0 .e.189) Example: Consider a rigid heat conductor. namely that q = −k grad T . ˙ AF ∂P + (4.. (4.192) Further. (4. hence = c. This means that 95 T P+ (4.11 d dt ρ+ ε+ + P+ The Green-Naghdi-Rivlin theorem Assume that the principle of energy balance remains invariant under superposed rigid motions. choose a special superposed rigid motion. such Q = I . Further.190) which implies that L = Ω. this means that 1 + + + ρ v · v dv + 2 = P+ ρ+ b+ · v+ dv + + t+ · v+ da+ + ρ+ r + dv + − h+ da+ . (4. assume that Fourier’s law holds. (4.195) ME185 that . It follows from (4. ˙ ˙ ˙ F = R = ΩR (Ω = RRT ) = LF = LR . DR 4.191) where T is the empirical temperature and k > 0 is the (isotropic) conductivity.165).188)2 and invariance of the thermal energy balance imply that h+ = q+ · n+ = (Qq) · (Qn) = q · n = h . With reference to (4. These conditions imply that the balance of energy (4. assume that the internal energy depends exclusively on T and that this dependence dε is linear. where c is termed the heat capacity.192) that dt ˙ ρcT = div(k grad T ) + ρr . which is a pure rigid translation.

means that conservation of mass and balance of linear momentum hold.96 The Green-Naghdi-Rivlin theorem where c0 is a constant vector in E 3 . it is readily concluded from (4.196) and (4.204) .195) and (4.196) Moreover.197) that (4. in turn.201) dt P P ∂P This. c = 0. Substituting γ with −γ in (4. t+ = t . DR spectively. (4. P (4.195) that v + = v + c0 T .198) ∂P Upon subtracting (4. such that ˙ Q = Ω0 . (4. reNext.196) that under this superposed rigid translation b+ = b . (4. it is concluded that c0 · d dt ρv dv − ρb dv − P P 1 d t da + (c0 · c0 ) 2 dt ∂P ρ dv = 0.143).203) ˙ ˙ ˙ v+ = v + 2Ω0 v + Ω2 x = v + ω 0 × (2v + ω 0 × x) . also. where γ is an arbitrary c non-vanishing scalar constant. (4.197) d dt ρε + P 1 ρ(v + c0 ) · (v + c0 ) dv 2 = P AF ρb · (v + c0 ) dv + ∂P It follows from (4.199) P Since c0 is an arbitrary constant vector.165) from (4. 0 (4. one may set c0 = γ¯0 .202).202) where Ω0 is a constant skew-symmetric tensor. a second special superposed rigid-body motion is chosen. (4. It follows immediately from (3.194) takes the form t · (v + c0 ) da + ρr dv − h da .199). it is proved that d dt ρ dv = 0 (4.148). Q = I .166) and (3.178) and (4. Given (4.200) P hence. (4. ˙ ˙ v+ = v . (3. (4. it can be easily seen from v + = v + Ω0 x = v + ω 0 × x (4.178) that and ME185 d ρv dv = ρb dv + t da .134). (4.198).

204). it is established that in this case b+ = b + 2Ω0 v + Ω2 x = b + ω 0 × (2v + ω 0 × x) 0 and t+ = t . 0 (4.207) and b+ · v+ = b · v + b · (Ωx) + 2(Ω0 v) · (Ωx) + 2(Ω0 v) · v = b · v + b · (Ωx) + (Ω0 v) · (Ωx) + (Ω2 x) · (Ω0 x) . (4.203) and (4.203)1 and (4. T (Ω0 v) · (Ω0 x) + (Ω2 x) · v = 0 0 t · (v + Ω0 x) da + P the superposed motion is a rigid rotation with constant angular velocity ω 0 on the original (4. that ˙ ˙ ρ v · v + v · (Ω0 x) + (Ω0 v) · (Ω0 x) + (Ω2 x) · (Ω0 x) dv 0 P = ρ b · v + b · (Ω0 x) + (Ω0 v) · (Ω0 x) + (Ω2 x) · (Ω0 x) dv 0 + ∂P ρr dv − h da .208) where the readily veriﬁable identities (Ω0 v) · v = 0 .211) ∂P ME185 . using (4. (4.Basic physical principles 97 where ω 0 is the (constant) axial vector of Ω0 . Equations (4.202) and (4. (4. it can be concluded from (4. it is seen that DR d dt ρε dv + P P 1 d + + ˙ ˙ (v · v ) = v · v + v · (Ω0 x) + 2(Ω0 v) · (Ω0 x) + 2(Ω0 v) · v 2 dt + (Ω2 x) · (Ω0 x) + (Ω2 x) · v 0 0 ˙ ˙ = v · v + v · (Ω0 x) + (Ω0 v) · (Ω0 x) + (Ω2 x) · (Ω0 x) .208) and (4.194). Taking into account (4.134) (4. 0 (4.206) (4. Similarly. as well as (4.205)1 lead to AF v+ · v+ = v · v + 2(Ω0 x) · v + (Ω0 x) · (Ω0 x) 2 + (Ω0 x) · (Ω0 x) + (Ω2 x) · v 0 (4.210).209) are employed.205)1.205) In addition.210) Invoking now invariance of the energy equation under the superposed rigid rotation. equations (4.207) (4.204) imply that current conﬁguration of the continuum.148).203)1 and (4.

214) This.215) namely that balance of angular momentum holds. (4. as a special case. the energy equation plays a central role. without being restricted by speciﬁc choices of coordinate systems. . (4.e. In covariant theories. hence should be expressed in a covariant manner. as demonstrated by the Green-Naghdi-Rivlin theorem.165) from (4.98 The Green-Naghdi-Rivlin theorem After subtracting (4. all physical laws should DR ME185 be invariant under any smooth time-dependent coordinate transformation (including.213) and recalling that ω 0 is a constant vector. The preceding analysis shows that the integral forms of conservation of mass. This far-reaching principle stems from Einstein’s conviction that physical laws are oblivious to speciﬁc coordinate systems. i. implies that d dt P ρx × v dv = P ρx × b dv + ∂P x × t da . in turn. it follows that ˙ ρv · (Ω0 x) dv = ρb · (Ω0 x) dv + T ∂P P P t · (Ω0 x) da . (4.. The Green-Naghdi-Rivlin theorem can be viewed as an implication of the general covariance principle proposed by Einstein.211) and simplifying the resulting equation. equation (4. According to this principle.212) takes the form ω0 · AF P ˙ ρx × v dv − P ρx × b dv − ∂P x × t da . rigid time-dependent transformations).212) Observing that for any given vector z in E 3 z · (Ω0 x) = z · (ω 0 × x) = ω 0 · (x × z) . and balance of linear and angular momentum are directly deduced from the integral form of energy balance and the postulate of invariance under superposed rigid-body motions. This remarkable result is referred to as the Green-Naghdi-Rivlin theorem. (4.

x 2! AF .e. as v → 0). it follows that as ε → 0 (i. 2! 3! v 2 ′′ = f (x0 ) + vf ′ (x0 ) + f (¯) . In these cases.e. where v is a change to the value of x0 and x ∈ (x0 . i. In many realistic circumstances. In this chapter. namely the diﬀerence between x0 + v and x0 .2) the scalar f (x0 + v) is satisfactorily approximated by the linear part of the Taylor series expansions in (5. the mathematical representation of kinematic quantities and the associated kinetic quantities.1). f (x0 + v) = f (x0 ) + vf ′(x0 ) .Chapter 5 Inﬁnitesimal deformations The development of kinematics and kinetics presented up to this point does not require any assumptions on the magnitude of the various measures of deformation. To this end.. x0 + v).. Denoting by ε the magnitude of ¯ .1) (5. . . Preliminary to this discussion. which is assumed to possess continuous derivatives up to any desirable order. as well as the balance laws. it is instructive to formally deﬁne the meaning of “small” or “inﬁnitesimal” changes of a function. ε = |v|. may be simpliﬁed substantially. consider a scalar-valued real function f = f (x). the special case of inﬁnitesimal deformations is discussed in detail. 99 T (5. one may use a Taylor series expansion v 3 ′′′ v 2 ′′ f (x0 + v) = f (x0 ) + vf (x0 ) + f (x0 ) + f (x0 ) + . solids and ﬂuids may undergo “small” (or “inﬁnitesimal”) deformations. To DR at x0 in the form ′ analyze this function in the neighborhood of x = x0 .

w=0 ω=0 2 ω=0 v 2 ′′ Recalling the expansion (5. V) a of F at X = X0 in the direction V is deﬁned as DF(X0. such that (5. vector or a tensor function of a scalar.6) DR Examples: ME185 (5. Then. i.1)2 . (5. when (5.4) where ω is a scalar. Df (x0 .1 The Gˆteaux diﬀerential a Linear expansions of the form (5. V) = d F(X0 + ωV) dω AF lim L[F.3) 5. vector or tensor variable X.100 The Gˆteaux diﬀerential a v 2 ′′ f (¯) ≪ |f (x0 + v)| . |V|→0 |V| of F at X0 in the direction V is then deﬁned as (5. x 2! assuming that f (x0 + v) = 0. V]X0 o(|V|2) = 0. Using the deﬁnition in (5. V) . given F = F(X). when the term f (¯) x 2! can be neglected in this expansion without appreciable error. .e. In particular. T . F(X0 + V) = F(X0 ) + DF(X0 .2) can be readily obtained for vector or tensor functions using the Gˆteaux diﬀerential. one says that ε = |v| is “small”.4). where F is a scalar. the Gˆteaux diﬀerential DF(X0 . v) = = = d f (x0 + ωv) dω ω=0 d (x0 + ωv)2 dω ω=0 d 2 (x0 + 2x0 ωv + ω 2 v 2 ) dω = 2x0 v + 2ωv = 2x0 v .7) (1) Let F(X) = f (x) = x2 . V]X0 = F(X0 ) + DF(X0. V) + o(|V|2) .5) The linear part L[F.

8) As usual. t) = u(x. ˆ ˜ u = u(X. the deﬁnition in (5. L[T. DT(x0 .Inﬁnitesimal deformations Hence. x = X+u . v) = = AF = = [(x0 ⊗ v + v ⊗ x0 ) + 2ωv ⊗ v]ω=0 = x0 ⊗ v + v ⊗ x0 . 101 (2) Let F(X) = T(x) = x ⊗ x. Using. v]x0 = x2 + 2x0 v . the displacement vector ﬁeld can be expressed equivalently in referential or spatial (5.4). t) . 0 .9) ME185 T ω=0 L[f . v]x0 = x0 ⊗ x0 + x0 ⊗ v + v ⊗ x0 .2 Consistent linearization of kinematic and kinetic Start by noting that the position vector x of a material point P in the current conﬁguration can be written as the sum of the position vector X of the same point in the reference conﬁguration plus the displacement u of the point from the reference to the current conﬁguration. (5. d T(x0 + ωv) dω ω=0 d {(x0 + ωv) ⊗ (x0 + ωv)} dω ω=0 d {x0 ⊗ x0 + ω(x0 ⊗ v + v ⊗ x0 ) + ω 2 v ⊗ v} dω It follows that DR variables namely form as 5. again.

where H is the two-norm of H deﬁned at a point X and time t as H Taking into account (5.7) and (5. AF ε = ε(t) = sup ||H(X.14) DR Then. H quantiﬁes the deviation of F from the identity tensor. i. by expanding F(H) around the reference conﬁguration. d ¯ F(0 + ωH) dω ω=0 d (I + ωH) dω ω=0 (5. H) . a suitable scalar measure of magnitude for the deviation of F from identity can be deﬁned as (5. the linearized counterpart of F(H) is the linear part L(F. (5. Now.102 Consistent linearization of kinematic and kinetic variables T It follows that the deformation gradient can be written as ˆ ˆ ∂(X + u) ∂u ∂χ = = I+ F = ∂X ∂X ∂X = I+H . (5. H)0 of F around the reference conﬁguration in the direction of H. then.15) . (5. ∂X Clearly. t)|| . H) = = =H.1. ¯ starting with the deformation gradient F. Now one may say that the deformations are “small” (of “inﬁnitesimal”) at a given time t if ε is small enough so that the term o(||H||2) can be neglected when compared with F(H). H)0 = F(0) + DF(0. a linearized counterpart of a given kinematic measure is obtained by ﬁrst expressing the kinematic measure as a function of H.12). H) + o( H 2 ) . as F(H) and. ﬁrst recall that F = F(H) = I + H.. This leads to F(H) = F(0) + DF(0. where H = 0.e. X∈R0 (5.13) Here. proceed to obtain inﬁnitesimal counterparts of some standard kinematic ﬁelds. ME185 where “sup” denotes the least upper bound over all points X in the reference conﬁguration. note that the Gˆteaux diﬀerential a DF(0. To this end.12) = (H · H)1/2 .10) where H is the relative displacement gradient tensor deﬁned by ˆ ∂u H = .11) Recalling the discussion in Section 5. given by L(F.

L[F−1 . take the linear part of both sides in the direction of H.19). which should be obvious from (5. Next.24) ME185 . H) = −DF−1 (0. one ﬁnds that therefore This means that. 103 T (5. H) = −H . grad u = grad u(H) = I − (I + H)−1 . H]0 = F(0)F−1 (0) + D(FF−1 )(0. This leads to Next. using the chain rule. (5.18) The preceding equation implies that (5. ¯ L[F. As a result. H)F−1 (0) + F(0)DF−1 (0. H) = H . DF−1 (0. (5. starting from FF−1 = I. H]0 = I . consider the linear part of the spatial displacement gradient grad u. taking into account (5. H) = I+H . ¯ ¯ L[FF−1 . where AF = H + DF−1 (0.17) ¯ ¯ D(FF−1 )(0. H]0 = grad u(0) + D(grad u)(0. H) = L[I. H) (5.20) DR that.22) (5.21) (5. H]0 = I − H .16) shows that the linear part of F is F itself.23) L[grad u.Inﬁnitesimal deformations Hence. (5. (5. H) = DF(0. D(grad u)(0. First. H]0 = F(0) + DF(0. H) = 0 . the linear part of F−1 at H = 0 in the direction H is (5. H) = 0 + H = H .16) Eﬀectively.10).19) Hence. observe grad u = (Grad u)F−1 = (F − I)F−1 = I − F−1 .

2 At the same time. H) 2 (5.28) 1 (H + HT ) . write ¯ C = C(H) = (I + H)T (I + H) = I + H + HT + HT H .29) hence its Gˆteaux diﬀerential is given. This. in turn. within (5.19). the Eulerian strain tensor e can be written as 1 ¯ ¯ i − F−T (H)F−1 (H) . H]0 = ¯ e = e(H) = De(0. H) = − This means that the linear part of e is equal to ME185 Using (5. To determine the linear part of the right Cauchy-Green deformation tensor C.104 Consistent linearization of kinematic and kinetic variables The last result shows that the linear part of the spatial displacement gradient grad u coincides the context of inﬁnitesimal deformations.27). 2 (5. H) + DF−1 (0. (5. it can be immediately concluded that the linear part of the Lagrangian (5. H) = I + (H + HT ) .31) . 2 with the referential displacement gradient Grad u(= H). AF = d ¯ C(0 + wH) dw w=0 d I + w(H + HT ) + w 2 HT H dw T T w=0 w=0 T = 1 (H + HT ) . 2 1 (H + HT ) . This further implies that the distinction between spatial and referential description of kinematic quantities becomes immaterial in the case of inﬁnitesimal deformations. H) = = H + H + 2wH H T = H+H . H) = (5.25) DC(0.26) Consequently. implies that. H]0 = e(0) + De(0. H]0 = C(0) + DC(0. by a 1 DF−T (0. the linear part of C at H = 0 in the direction H is ¯ L[C. there is no diﬀerence between the partial derivatives of the displacement u with respect to X or x. with the aid of (5.27) DR strain tensor E is L [E. (5. Then.30) ¯ L[e.

tensors coincide.38) 1 1 ¯ ¯ DR(0. 2 (5. H]0 = U(0) + DU(0. with components εij = 2 (ui.35) 2 Repeating the procedure used earlier in this section to determine the Gˆteaux diﬀerential a of F−1 . 1 DU−1 (0. H) = − (H + HT ) 2 L U−1 . Proceed next with the linearization of the right stretch tensor U. Hence.i). T = H + HT .34) hence.36) DR by ﬁrst obtaining the Gˆteaux diﬀerential as a and then writing and 1 (5. written as ¯ ¯ ¯ R = R(H) = F(H)U−1 (H) . (5. H) = DU(0. H) = H − (H + HT ) = (H − HT ) . H)U−1 (0) + F(0)DU−1 (0. H]0 = ε .39) 1 ¯ L [R. H) (5.32) 1 where ε is the classical inﬁnitesimal strain tensor.33) so that.j + uj. 2 It is now possible to determine the linear part of the rotation tensor R. the distinction (5. (5. write AF U2 = C = I + H + HT + HT H . H) = 2DU(0. under the assumption of inﬁnitesimal deformations. H]0 = R(0) + DR(0.28) and (5.31) that the linear parts of the Lagrangian and Eulerian strain between the two strains ceases to exist and one writes that L[E.26).Inﬁnitesimal deformations 105 It is clear from (5. H) = DF(0.40) ME185 . with the aid of (5. ¯ ¯ DU2 (0. 1 ¯ L [U. To this end. H 0 (5. H]0 = L[e. H)U(0) + U(0)DU(0.37) = I − (H + HT ) . 2 2 (5. H) = I + (H − HT ) . H) = I + (H + HT ) . one easily ﬁnds that (5.

43) The balance laws are also subject to linearization. derive the linear part of the Jacobian J of the deformation gradient.42) where IH . H) . in turn.45).45) Since conservation of mass is assumed to hold in all conﬁgurations therefore also including (5.46) thus equation (5. H)J(0) + ρ(0)DJ(0. 2 ω = When H is small. H) = 1 + tr H = 1 + tr ε . H) + ρ(0) tr ε = 0 . ρ0 = ρ(0)J(0) = ρ(0) .42) results in (5. the tensor (5. it follows that ME185 (5.i). leads to ¯ L[det F. statement (4.41) ω=0 (5. IIH . L[ρ0 . the conservation of mass DR This means that the reference conﬁguration. with the aid of (5.41) can be linearized to yield (5. ¯ T 1 (H − HT ) 2 is called the inﬁnitesimal rotation tensor and has components ωij = 1 (ui. This. d ¯ det F(ωH) dω ω=0 d det(I + ωH) dω ω=0 d 1 det{ω(H − (− )I)} dω ω ω=0 1 3 1 1 d 3 ω −(− ) + IH (− )2 − IIH (− ) + IIIH dω ω ω ω d 1 + ωIH + ω 2 IIH + ω 3 IIIH dω ω=0 observe that D(det F)(0. ¯ ¯ ¯ (5.106 Consistent linearization of kinematic and kinetic variables Next. H]0 .j − uj. H]0 = det F(0) + D(det F)(0. ¯ ¯ ¯ Dρ(0. To this end. For instance. H]0 = L[ρJ.47) . and IIIH are the three principal invariants of H.44) ¯ ρ0 = ρ(0)J(0) + Dρ(0. H) = = AF = = = = IH = tr H .

49) reveals that the linearized mass density does not coincide with the mass density of the reference conﬁguration. H]0 = ρ(0) + Dρ(0. ¯ (5.49) Equation (5.Inﬁnitesimal deformations or. equivalently. H) = −̺0 tr ε .48) . H) = ρ0 (1 − tr ε) . 107 The linear part of the mass density relative to the reference conﬁguration now takes the form L[ρ. (5. DR AF ME185 T Dρ(0.

108 Consistent linearization of kinematic and kinetic variables DR ME185 AF T .

1) or (6. . . . . First. GradF. Clearly.. which relate the stresses to the mass density and the kinematic variables. .g. F. the position x (or velocity v) and the stress tensor (e. . . a general constitutive equa- DR or. . .1) (6. . ρ.Chapter 6 6. . This simply means that the units of the left. . ρ. . . . F. in rate form. F. . v. three from linear momentum balance and three from angular momentum balance) to determine thirteen unknowns. . all constitutive laws need to be dimensionally consistent. .2) ˆ ˙ ˙ ˙ T = T(x. . the Cauchy stress T). . while others tend to be less uniformly accepted. For example. . . as tion for the Cauchy stress may be written as AF ˆ ˙ T = T(x.2) should be the same. without additional equations this system lacks closure. . A number of restrictions may be placed to the preceding equations on mathematical or physical grounds. . . . namely the mass density ρ. Before accounting for any possible restrictions or reductions. . F. Three of these restrictions are reviewed below. .and right-hand side of (6. ρ. .) ˙ 109 Mechanical constitutive theories T (6. Some of these restrictions appear to be universally agreed upon. v. . ˙ Analogous general functional representations may be written for other stress measures. The latter is established by constitutive equations.) . GradF. . ρ.1 General requirements The balance laws furnish a total of seven equations (one from mass balance.

This is to maintain consistency with the left-hand sides.1) and (6. ˆ T = T(F) .3) necessitates that AF ˆ T+ = T(F+ ) .2) to remain unaltered under superposed rigid motions. This postulate requires the functions T and T in (6. which are naturally resolved on this basis. 6. equations (6. where β is a constant. This restriction would disallow.5) places a restriction on the choice of T.1) and (6. the same laws need to be (6. This means that the right-hand side of (6.137). ˆ ˙ An identical restriction is placed on the function T. which is assumed to apply to constitutive laws just ˆ ˆ ˙ as it applies to balance laws. Equation (6.3) and (6. this would necessitate that the consistent in their tensorial representation.110 Inviscid ﬂuids ˙ when applied to a constitutive law of the form T = αB.6).2) should have a tensorial representation in terms of the Eulerian basis. for example.4) Taking into account (3. The third source of restrictions is the postulate of invariance under superposed rigid motions (also referred to as objectivity).2 Inviscid ﬂuids An inviscid ﬂuid is deﬁned by the property that the stress vector t acting on any surface is always opposite to the outward normal n to the surface. Second. By way of example.5) DR ◦ ˆ for all proper orthogonal tensors Q. regardless of whether the ﬂuid ME185 . applying the postulate of invariance under superposed rigid motions to the constitutive law ˆ T = T(F) (6. a constitutive law of the form T = βF.6) and T is the Jaumann rate of the Cauchy stress tensor. ◦ T parameter α have units of stress (since B is unitless).165) and (4. ˆ ˆ QT(F)QT = T(QF) . where. The same conclusion is reached when using any other objective rate of the Cauchy stress tensor in (6. (6. for example.4) lead to (6.

one may assume that the pressure p depends on the density ρ.9). an inviscid ﬂuid cannot sustain shearing tractions t = Tn = −pn . T = −p(ρ)i . On physical grounds.. it is clear that .7) (6.12). hence it holds for all orthogonal ˆ tensors.e. hence T = −pi . Furthermore. In particular.12) (6. ˆ ˆ QT(φ)QT = T(φ) . (6.13) ME185 for all proper orthogonal Q. substituting −Q for Q in (6. t T n under any circumstances.1: Traction acting on a surface of an inviscid ﬂuid.11) which.10) Upon invoking invariance under superposed rigid motions. Said diﬀerently.12) holds for all improper orthogonal tensors Q as well. with the aid of (4. see Figure 6. It is instructive here to take an alternative path for the derivation of (6.143) leads to (6.1.9) This constitutive relation deﬁnes a special class of inviscid ﬂuids referred to as elastic ﬂuids. suppose that one starts from the more general constitutive assumption DR AF ˆ T = T(ρ) . (6. ˆ ˆ QT(ρ)QT = T(ρ) . This means that (6.137) and (4.8) Figure 6. A tensor function T(φ) of a scalar variable is termed isotropic when (6.Mechanical constitutive theories 111 is stationary or ﬂowing. ˆ T+ = T(ρ+ ) . i. it follows that (6.

this immediately leads to the constitutive equation (6. In the case of T in (6. q = e3 . the tensor function remain unaltered when resolved on any two orthonormal bases.17) which corresponds to p = e2 . means that T22 = T33 (6. (6. it is a scalar multiple of ˆ the identity tensor. i. ﬁrst note that the suﬃciency argument is trivial. and θ = π/2.13). It is easy to verify that.e. Clearly. set with respect to the axis of e2 . T23 = T21 .10). to a rigid rotation of π/2 with respect to the axis of e1 . equation (6. The necessity argument can be made by setting AF Q = Q1 = e1 ⊗ e1 − e2 ⊗ e3 + e3 ⊗ e2 . T T23 = −T32 . r = e1 .112 Inviscid ﬂuids for all orthogonal Q. q = e2 . (6. recalling the Rodrigues formula (3..15) DR Next.16) (6. = T21 T22 T23 . corresponds to p = e1 . T32 = T12 = −T12 . This condition may be interpreted as meaning that the components of The representation theorem for isotropic tensor functions of a scalar variable states that a tensor function of a scalar variable is isotropic if.9). and only if. T12 = T21 = T13 = T31 = 0 . r = e3 .13) yields T11 −T13 T12 −T31 T33 −T32 T21 −T23 T22 This.18) T31 T32 T33 (6. Again. in this case. and θ = π/2.10) is isotropic. ˆ the constitutive function T in (6.14) which. Q = Q2 = e2 ⊗ e2 − e3 ⊗ e1 + e1 ⊗ e3 . T31 T32 T33 T11 T12 T13 . in turn. it follows that T11 T12 T13 T21 T22 T23 . This is a rigid rotation of π/2 this rotation in (6. upon using T33 T32 −T31 T23 T22 −T21 = −T13 −T12 T11 which leads to T11 = T33 ME185 (6.103). To prove the preceding representation theorem. .19) .

Mechanical constitutive theories One may combine the results in (6.24) DR independent of x. It follows from (6. ρ+ ) ˆ ˆ QT(x. ˆ ˆ T(x. Invariance under superposed rigid motions may be also used to exclude certain functional dependencies in the constitutive assumption for T. Indeed. ρ) . given that c0 is arbitrary. Indeed.25) violates invariance under superposed rigid motions. ρ) ˆ T+ = T(v+ . Invariance of T under superposed rigid motions implies that AF ˆ T+ = T(x+ .16) and (6. ρ)QT = T(ΩQx + Qv + c. (6.23) that (6. which completes the proof. Now. ˆ T = T(x. ˆ T = T(v.20) where T = T11 = T22 = T33 .24) can be met only if T is altogether A similar derivation can be followed to argue that a constitutive law of the form (6. that which readily translates to ˆ However. ρ) .21) (6. 113 T (6. such that Q = I and c = c0 .23) for all proper orthogonal tensors Q and vectors c. ρ) = T(x + c0 . ρ) . ρ+ ) .27) ME185 .19) to deduce that T = T I. in this case. assume that ˆ namely that the Cauchy stress tensor depends explicitly on the position.26) (6. ρ)QT = T(Qx + c. where c0 is constant. the condition (6.22) hence (6. invariance implies (6. ˆ ˆ ˙ QT(v. ρ) . choose a superposed rigid motion.

again. ∂t Equations (6. t) for all time. Recall the deﬁnition of an isochoric (or volume-preserving) motion.8) and the non-trivial equations that govern its motion are written in Eulerian form (which is suitable for ﬂuids) as AF ρ + ρ div v = 0 ˙ − grad p(ρ) + ρb = ρa ρ + ρ div v = 0 ˙ − grad p(ρ) + ρb = ρ( div v = 0 − grad p + ρ0 b = ρ0 ( (6. upon appealing to the local statement of mass conservation (4. ρ) .9) loses its meaning. then the constitutive equation (6. note that angular momentum balance is satisﬁed automatically by the constitutive equation (6. ρ) = T(v + c0 . choose Q = I. ˆ which leads to the elimination of v as an argument in T. T ∂v + Lv) . ﬂuid) are ME185 a motion conservation of mass leads to ρ0 (X) = ρ(x. and note that for such DR motions. ˜ can be solved for ρ(x. If the invis- cid ﬂuid is assumed incompressible.114 Inviscid ﬂuids Now. upon expressing the acceleration in Eulerian form in terms of the velocities (6.30) ∂v + Lv) . Ω = 0 and c = c0 t. Instead. subject to the speciﬁcation of suitable initial and boundary conditions and a pressure law p = p(ρ). t). the constitutive equation T = −pi holds with p being the unknown. it is seen that div v = 0 for all isochoric A material is called incompressible if it can only undergo isochoric motions. Then.31) .29) or. It follows that for ˆ ˆ T(v. because the function p(ρ) does not make sense as the density ρ is not a variable quantity.30) form a set of four coupled non-linear partial diﬀerential equations in x and t. Equations (6. the governing equations for an incompressible inviscid ﬂuid (ofter also referred to as an ideal (6. In summary. a constant. which. ∂t this particular choice of a superposed motion (6.30)2 are referred to as the compressible Euler equations. t) and v(x.34).28) Returning to the balance laws for the elastic ﬂuid. where c0 is.

2. Therefore.125).33) The explicit dependence of the Cauchy stress tensor on W can be excluded by invoking invariance under superposed rigid-body motions. This suggests that the pressure ﬁeld in an incompressible elastic ﬂuid is not uniquely determined by the equations of motion. W+ ) .. D+ . VISCOUS FLUIDS where now the unknowns are p and v. 115 (p + c. see Figure 6. v) satisﬁes the above equations. recalling the unique additive decomposition of L in (3. This point is illustrated by way of an example: consider a ball composed of an ideal ﬂuid.34) ME185 AF p T Notice that if a set (p. then so does another set of the form . L) (6.32) or. Indeed. some ability to sustain shearing forces.3. more generally as ˆ T = T(ρ. The indeterminacy is broken by specifying the value of the pressure on some part of the boundary of the domain. 6.3 Viscous ﬂuids DR in the form All real ﬂuids exhibit some viscosity. which is in equilibrium under uniform time-independent pressure p. where c is a constant. W) . It is easy to conclude on physical grounds that the resistance to shearing is related to the gradient of the velocity. Figure 6.2: A ball of ideal ﬂuid in equilibrium under uniform pressure.e. D. (6. i. where c is a constant. v). it is sensible to postulate a general constitute law for viscous ﬂuids ˆ T = T(ρ. invariance requires that ˆ T+ = T(ρ+ .6. The same “motion” of the ball can be also sustained by any pressure ﬁeld p+c. (6.

(3. its axial vector ω 0 ). QDQT . c(t) = 0 and c(t) = 0. ˆ T = T(ρ. equation (6. ˆ Neglecting.39). ˆ ˆ QT(S)QT = T(QSQT ) . It can be proved following the process used earlier for isotropic ˆ tensor functions of a scalar variable that a tensor function T of a tensor variable S is isotropic (6.38) which. equation (6. (6. W)QT = T(ρ. equivalently.39) DR if.37) Invariance under superposed rigid motions for the constitutive function in (6. the dependence of T on ρ in equation (6. it can be written in the form ME185 for all proper orthogonal tensors Q.185) and (4.41) .34) takes the form ˆ ˆ QT(ρ. QWQT + Ω) . it is clear that (6.39) are even functions of Q. W) = T(ρ. In fact.116 Viscous ﬂuids ˙ Now. Given the special form of the chosen superposed rigid motion. Recalling (3. D. This implies that the constitutive ˆ function T cannot depend on W.37) gives rise to the condition (6. D+ ) . for a moment. ˆ T+ = T(ρ+ . D. DQT . ˆ T(S) = a0 I + a1 S + a2 S2 .35) for all proper orthogonal Q. and only if. This is a superposed rigid rotation (6. note that a tensor ˆ function T of a tensor variable S is called isotropic if (6. ˆ ˆ QT(ρ. W + Ω0 ) .39) must hold for all orthogonal tensors Q.143).184). AF ˆ ˆ T(ρ. Q(t) = Ω0 (Ω0 being with constant angular velocity deﬁned by the skew-symmetric tensor Ω0 (or.35) leads to T ˙ constant skew-symmetric tensor). in turn. consider a special superposed rigid motion for which Q(t) = I. since both sides of (6.36) which needs to hold for any skew-symmetric tensor Ω0 . necessitates that (6.40) for all orthogonal tensors Q. QDQT ) . D) . thus it reduces to (6. D)QT = T(ρ.

IIS .37) is of the form AF T = a0 i + a1 D + a2 D2 . consider the sub-class of those which are linear in D. Materials that obey (6.43) are also generally referred to as non-Newtonian ﬂuids. IIS and IIIS a0 = a0 (IS . The functions λ and µ are called the viscosity coeﬃcients. i. This means that the 0 (6. IIIS ) ˆ a1 = a1 (IS .44) is reduced to (6. a∗ = λID and a1 = 2µ. 0 T = −p(ρ)i + λ tr Di + 2µD .46) Viscous ﬂuids that obey (6.e. ID . 0 T = (−p(ρ) + a∗ )I + a1 D . the viscous ˆ∗ 0 From the above general class of viscous ﬂuids. ˆ a2 = a2 (IS .44) DR ﬂuid degenerates to an inviscid one. (6. ME185 . in general.45) where. (6. introduce a physically plausible assumption by way of which the Cauchy stress T reduces to hydrostatic pressure −p(ρ)i when D = 0. IIID and ρ. IID . IIS .43) as (6. IIIS ) ˆ T of S.43) where a0 . where λ and µ depend. Cauchy stress tensors takes the form ∗ where. IIS .46) are referred to as Newtonian viscous ﬂuids or linear viscous ﬂuids. T = (−p(ρ) + a∗ )i + a1 D + a2 D2 .. a1 . one may rewrite the constitutive function (6. on ρ. and a2 are scalar functions of the three principal invariants IS . IIID ). it is readily concluded that the Cauchy stress for a ﬂuid that obeys the general constitutive law (6. Clearly. the constitutive function in (6. in general. a1 and a2 are functions of ID . IID . when a0 = a1 = a2 = 0. To preserve linearity in D. IIIS ) . At this stage.Mechanical constitutive theories 117 where a0 .42) The above result is known as the ﬁrst representation theorem for isotropic tensors. a∗ = a0 (ρ. The preceding equation characterizes what is known as the Reiner-Rivlin ﬂuid. Then. Using this theorem.

the governing equations of (6. As in the inviscid case. Conservation of mass and linear momentum balance can be expressed as T ρ + ρ div v = 0 ˙ ρ + ρ div v = 0 ˙ div v = 0 nian viscous ﬂuid. hence.48) This implies that for this special case the governing equations of motions (6. there are four coupled nonlinear partial diﬀerential equations in (6. angular momentum balance is satisﬁed at the outset. As in the case of the compressible inviscid ﬂuid. Clearly. Equations (6. namely the velocity v and DR the mass density ρ. Assuming that either λ and µ are independent of ρ (which is very common) or that ρ is (6.49) ˙ − grad p + (λ + µ) grad div v + µ div grad v + ρb = ρv .49)2 are known as the Navier-Stokes equations for the compressible Newtonian viscous ﬂuid.51) The former equation is a local statement of the constraint of incompressibility.49) become ME185 If the Newtonian viscous ﬂuid is incompressible. since T in (6. spatially homogeneous.46) in place. while the latter is the reduced statement of linear momentum balance that reﬂects incompressibility. the four unknowns now are the velocity v and the pressure p. (6.46) is symmetric. ˙ − grad p + µ div grad v + ρb = ρv .118 Viscous ﬂuids With the constitutive equation (6. one may write balance take the form div(−p(ρ)i + λ tr Di + 2µD) = − grad p + λ grad div v + µ(div grad v + grad div v) = − grad p + (λ + µ) grad div v + µ div grad v . consider the balance laws for the Newto(6.47) can be recast as AF T = −pi + 2µD . the Cauchy stress is given by (6.47) ˙ div(−p(ρ)i + λ tr Di + 2µD) + ρb = ρv . .50) (6.49) and four unknowns.

upon recalling (6. so that. the mechanical energy balance theorem for this class of materials takes the form d [K(P) + W (P)] = Rb (P) + Rc (P) = R(P) .52) It follows that the stress power in the region P takes the form P AF T · D dv = ˙ ρΨ dv = P T d dt ρΨ dv = P The Navier-Stokes equations (compressible or incompressible) are non-linear in v due to ˜ ˜ ∂v ∂v ˙ + v.56) ME185 . NON-LINEARLY ELASTIC SOLID 119 term may be ignored. (6. which is expanded in the form v = ∂t ∂x very slow and nearly steady ﬂow.53) where W (P) = P As a result. dt (6. DR rate of work done by the external forces. such that ˙ T · D = ρΨ . deﬁne the non-linearly elastic solid by admitting the existence of a strain energy ˆ function Ψ = Ψ(F) per unit mass. 6. With the aid of the chain rule. dt (6. In words. giving rise to a system of linear partial diﬀerential equations.55) ˙ T · D = ρΨ = ρ ˆ ∂Ψ · (LF) . the acceleration d W (P) . equation (6. referred to as creeping ﬂow or Stokes ﬂow. this leads to ˆ ∂Ψ ˙ ˙ ·F . ∂F (6.6.54) ρΨ dv is the total strain energy of the material occupying the region P. In the special case of the acceleration term.4 Non-linearly elastic solid Recalling the deﬁnition of stress power in the mechanical energy balance theorem of equation (4. Ψ = ∂F (6.54) states that the rate of change of the kinetic and strain energy (which together comprise the total internal energy of the non-linearly elastic material) equals the Recall that the strain energy function at a given time depends exclusively on the defor- mation gradient.158).4.52).

∂F (6. ˆ ˆ ˆ ˆ Ψ(F) = Ψ(QF) = Ψ(RT RU) = Ψ(U) . Selecting Q = RT .60) ˆ This places a restriction on the form of the strain energy function Ψ.57) The preceding equation can be also written as (T − ρ ˆ ∂Ψ T F )·L = 0 . Instead of explicitly enforcing this restriction. it is immediately concluded that AF T = ρ ˆ ∂Ψ T F . in turn. ˆ ˆ Ψ+ = Ψ(F+ ) = Ψ(QF) ˆ = Ψ = Ψ(F) .62) that (6.58) Observing that this equation holds for any L. it follows from (6.59) Upon enforcing angular momentum balance. equation (6.120 which. one may write ME185 tained by invoking invariance under superposed rigid motions. ˆ ˆ ¯ ˇ Ψ = Ψ(F) = Ψ(U) = Ψ(C) = Ψ(E) . one may simply write the Cauchy stress as T ˆ ˆ 1 ∂Ψ T ∂Ψ T = ρ F +F .62) for all proper orthogonal tensors Q. invariance of the ˆ function Ψ implies that (6. where R is the rotation stemming from the polar decomposition of F.63) (6. leads to Non-linearly elastic solid T·L = ρ ˆ ˆ ∂Ψ ∂Ψ T · (LF) = ρ F ·L . (6. Speciﬁcally.61) Alternative expressions for the strain energy of the non-linearly elastic solid can be ob- DR Therefore. ∂F ˆ ∂Ψ T F = F ∂F .59) leads to (6. 2 ∂F ∂F (6.64) . ∂F ∂F T ˆ ∂Ψ ∂F T (6.

71) However.54) and integrate this equation in time between t1 and t2 (6. [K(P) + W (P)]t2 t1 = [Rb (P) + Rc (P)] dt .67) (6. ∂C ∂C (6. Then. since the motion is a closed cycle. t1 ) = χ(X. also t2 t2 for all X. t1 ) = χ(X.70) that 1 ρv · v dv + 2 t2 ˆ ρΨ(F) dv P t1 = 0. ∂C ˇ ∂Ψ T F . Recall next the theorem of mechanical energy balance in the form of (6. ∂E t2 material time derivative of Ψ can be expressed as (6. ¯ ∂Ψ T F . t2 ) . it is immediately concluded from (6. It follows from (6. t2 ) ˙ ˙ v = χ(X.66) ¯ ∂Ψ T F ∂C Given the arbitrariness of D. it follows that T − 2ρF T = 2ρF AF T = ρF x = χ(X.69) Now.70) DR to ﬁnd that [K(P) + W (P)]t2 = t1 thus. This motion is referred to as a closed cycle. for which there exist times t1 and t2 (> t1 ). Ψ = ∂C ∂C where (3.68) Using an analogous procedure.52) that T·D = ρ which readily leads to T ·D = 0 .129) is invoked.Mechanical constitutive theories 121 by merely exploiting the one-to-one relations between tensors U. (6. consider a body made of non-linearly elastic material that undergoes a special motion χ. such that (6.65) ¯ ¯ ∂Ψ T ∂Ψ · (2FT DF) = 2ρF F ·D .73) ME185 .72) [Rb (P) + Rc (P)] dt = t1 t1 P ρb · v dv + ∂P t · v da dt = 0 . t1 P (6. C and E. one may also derive a constitutive equation for the Cauchy ˇ stress in terms of the strain energy function Ψ as (6. the ¯ ¯ ∂Ψ ∂Ψ ˙ ˙ ·C = · (2FT DF) . (6.

The concept of material symmetry is now introduced for the class of Cauchy-elastic ma- terials. (6. subject the reference AF P ′ P0 . do not satisfy the condition of worklessness in a closed cycle.54) further implies that [K(P) + W (P)]t1 t t = t1 T ρb · v dv + ∂P a closed cycle is equal to zero. see conﬁguration. Since the material is assumed to be Cauchy-elastic.61). in general. This is the sense in which the non-linearly elastic material is characterized as path-independent.75) Such materials are called Cauchy-elastic and. Now. Also. take an inﬁnitesimal volume element P0 which contains X in the ref- the Cauchy stress tensor for P at time t is given by (6. it follows that conﬁguration to an orthogonal transformation characterized by the orthogonal tensor Q.122 Non-linearly elastic solid This proves that the work done on a non-linearly elastic solid by the external forces during Equation (6.74) This means that the work done by the external forces taking the body from its conﬁguration at time t1 to time t(> t1 ) depends only on the end states at t and t1 and not on the path connecting these two states. (6. DR P P0 ME185 Figure 6. it is clear that any Green-elastic material is also Cauchy-elastic.75). To this end. P t · v da dt . A more general class of non-linearly elastic materials is deﬁned by the constitutive relation ˆ T = T(F) . The preceding class of non-linearly elastic materials for which there exists a strain energy ˆ function Ψ is sometimes referred to as Green-elastic or hyperelastic materials. Recalling the constitutive equation (6. let P be a material particle that occupies the point X in the reference erence conﬁguration.3: Orthogonal transformation of the reference conﬁguration.

ME185 . (iv) For every a. given by ˆ T = T′ (FQ) . the identity element is the identity tensor I and the inverse property within the subset. By way of background.78) respectively. Hence.3. recall here that a group G is a set together with an operation ∗.76) as AF ˆ T = TP0 (F) . there exists an element −a. T is independent of the speciﬁc choice of reference conﬁguration. c of the set: (i) a ∗ b belongs to the set (closure). In this group. b. that the stress at a point terms of the deformation relative to the orthogonally transformed reference conﬁguration. in general. For this reason. called the orthogonal element is the inverse (or transpose) Q−1 (or QT ) of any given element Q. such that the following properties hold for all elements a. The subgroup1 A subset of the group set together with the group operation is called a subgroup if it satisﬁes the closure group or O(3). however. in general. (6. Note. This transforms the region P0 to P0 . the Cauchy stress at point P is. (iii) There exists an element i. 1 (ii) (a ∗ b) ∗ c = a ∗ (b ∗ c) (associativity). ˆ ′ T = TP0 (FQ) . such that a ∗ (−a) = (−a) ∗ a = i (existence It is easy to conﬁrm that the set of all orthogonal transformations Q of the original refer- ence conﬁguration forms a group under the usual tensor multiplication.75) and (6.76) The preceding analysis shows that the constitutive law depends. at the expense of added notational burden. on the choice of reference conﬁguration.77) and (6.Mechanical constitutive theories 123 ′ Figure 6. DR of inverse). such that i ∗ a = a ∗ i = a (existence of identity). one may choose. to formally write (6. when expressed in (6.

79) implies that ˆ ˆ ˆ ˆ T = TP0 (F) = TP0 (FRT ) = TP0 (VRRT ) = TP0 (V) . If equation (6. then the orthogonal transformation characterizes the material symmetry of the body in the neighborhood of P relative to the reference conﬁguration P0 . ME185 AF . The ﬁrst one subjects the reference conﬁguration to a deformation gradient F. then the Cauchy-elastic material is termed isotropic relative to the conﬁguration P0 .124 Non-linearly elastic solid GP0 ⊆ O(3) is called a symmetry group for the Cauchy-elastic material with respect to the ˆ ˆ TP0 (F) = TP0 (FQ) . The second one subjects the reference conﬁguration to an orthogonal transformation Q and then to the deformation gradient F. and b2 are functions of the three principal invariants of V. given that (6. An alternative representation of the Cauchy stress of a Cauchy-elastic material is ¯ T = TP0 (B) = b0 i + b1 B + b2 B2 . b0 .83) where. ˆ In addition. If the stress in both cases is the same. (6. it follows that ˆ T = TP0 (V) = a0 i + a1 V + a2 V2 . This means that TP0 is an isotropic tensor function of V. equation (6. Physically. all orthogˆ onal Q). An isotropic material is insensitive to any orthogonal transformation of its reference conﬁguration. (6.80) is quadratic in Q. T reference conﬁguration P0 if (6. equation (6.79) holds for all Q ∈ O(3).79) for all Q ∈ GP0 . and a2 are functions of the three principal invariants of V. (6.81) (6.80) DR 0 0 for all proper orthogonal tensors Q (hence. Invoking the ﬁrst representation theorem for isotropic tensor functions. a1 .82) where a0 .79) identiﬁes orthogonal transformation Q which produce the same stress at P under two diﬀerent loading cases. Choosing Q = RT and recalling the left polar decomposition of the deformation gradient. b1 . now. invariance of TP0 under superposed rigid motions leads to the condition ˆ ˆ QTP0 (V)QT = TP0 (QVQT ) . This can be ˆ ¯ trivially derived by noting that TP (B) = TP (V).

To this end. i.87) can be written explicitly as DR above equation that (6. let the reference conﬁguration be stress-free. ˆ ¯ T(I) = T(0) = 0 . it is concluded from the (6. a formal procedure is followed to obtain the equations of motion and the constitutive equations for a linearly elastic solid. (6. ˆ T(Q(t0 )) = 0 . This is a fourth-order tensor that can be resolved in components as (6. Now. ω=0 (6. Taking into account (6.5 Linearly elastic solid In this section.90) for all proper orthogonal tensors Q. H) = d ¯ T(0 + ωH) dω C = Cijkl ei ⊗ ej ⊗ ek ⊗ el . assume that if T = T(F) = T(H).6. H) .91) ME185 . start by writing the linearized version of linear momentum balance as L[div T.85). H]0 = L[ρa. C H = (Cijkl ei ⊗ ej ⊗ ek ⊗ el )(Hmn em ⊗ en ) = Cijkl Hmn ei ⊗ ej [(ek ⊗ el ) · (em ⊗ en )] = Cijkl Hmn ei ⊗ ej [δkm δln ] = Cijkl Hkl ei ⊗ ej .. LINEARLY ELASTIC SOLID 125 6. (6. ¯ D T(0.85) It follows that AF ¯ ¯ ¯ L[T.84) ˆ ¯ First. H) = D T(0. ˆ ˆ QT(F)QT = T(QF) . proceed by making three assumptions: then T = CH .5.87) The quantity C is called the elasticity tensor . H]0 .88) The product C H in (6. H]0 = T(0) + D T(0.e.86) where (6. H]0 + L[ρb.89) At this state. note that invariance under superposed rigid motions implies that (6.

also H(t0 ) = 0) and Q(t0 ) = Ω0 (hence. L[div T. H]0 = L[Div T.97) and the preceding (6. With (6.92) Since Ω0 is an arbitrarily chosen skew-tensor. (6. By way of a second assumption. H]0 = ρ0 b . Therefore. one may write that d ¯ ˙ T(0 + ω H) dω ¯ ˙ ˙ = D T(0. H]0 = C (ε + ω) = C ε = σ .93) Here. (6. the body force is linear in H).e. there is no acceleration in the reference conﬁguration) and also that [Da(0. H)] .. (6. Since the distinction between partial derivatives with respect to X and x disappears in the inﬁnitesimal case. it is easily seen that ME185 ¯ L[ρb.126 Linearly elastic solid where Ω0 is a constant skew-symmetric tensor. the acceleration is linear in H). noting ﬁrst that assumptions on b in place.87) and that H = ε + ω.e. L[ρa. It follows from (6.e.. one may choose a special such ˙ ˙ rotation for which Q(t0 ) = I (hence. is tantamount to admitting that equilibrium holds in the reference conﬁguration) and also that [Db(0. assume that in the linearized theory the acceleration a satis¯ ﬁes a(0) = 0 (i.98) . ω=0 T namely that any rigid rotation results in no stress. H)] = a (i.95) With reference to (6. H]0 = ρ(0)¯ (0) + [Dρ(0. Hence.96) DR ing assumptions that Finally. it follows that AF L[T. H) = C H = C Ω0 = 0 . in view of the earlier two assumptions. H]0 = Div σ . ¯ a a ¯ (6. H]0 = ρ0 a ..94) L[ρa.97) ¯ assume that in the linearized theory b(0) = 0 (which. L[ρb.95). also H = Ω0 ). H]0 = Div L[T. Recalling (6. it is clear that so does the distinction between the “Div” and “div” operators. H)] ¯ ¯ (6. H]0 = ρ(0)b(0) + [Dρ(0. In this case. σ is the stress tensor of the theory of linear elasticity. H)]¯(0) + ρ(0)[Da(0.86). H)] = b (i. write (6. H)]b(0) + ρ(0)[Db(0. this means that C Ω = 0 for any skewsymmetric tensor Ω.95) and the preced(6.

130) and take the linear part of both sides to conclude that L[T.104) Similar derivations apply to deduce equivalency of T to other stress tensors in the inﬁnitesReturn now to the constitutive law (6. b depends on the motion (and deformation) implicitly through the balance of linear momentum. all measures of stress coincide. H J .99) Of course. H) P(0)FT (0)+ ¯ [DP(0. H]0 = L (6. AF ¯ L[P. To see this point. On the other hand. disappears. implies that 1 ¯ 1 1 ¯ 1 ¯ ¯ ¯ ¯ σ = ¯ P(0)FT (0)+ D (0. 127 T 1 PFT .103) (6. imal theory.96) and (6. σij = Cijklεkl . (6.Mechanical constitutive theories Taking into account (6. H) = σ . H]0 = P(0) + DP(0. etc.93). H]0 . H) . S.93) and write it in component form as (6. namely the distinction between the Cauchy stress T and other stress tensors such as P. 0 (6. in light of (6. In the context of linear elasticity. L[P. recall the relation between T and P in (4.84) reduces to Div σ + ρ0 b = ρ0 a .101) Recalling that the reference conﬁguration is assumed stress-free.98).102) DR which implies further that hence. H)] . H)]FT (0)+ ¯ P(0)[DFT (0.94). (6. H]0 = L[T. (6. neither a nor b are explicit functions of the deformation.100) which. it follows from the above equation that σ = DP(0. J J(0) J(0) J(0) (6.105) ME185 . equation (6. The acceleration a depends on the deformation implicitly in the sense that the latter is obtained from the motion χ whose second time derivative equals a.

c1 .83) for the isotropic DR terms of the Almansi strain e. namely 2 non-linearly elastic solid and note that a corresponding equation can be easily deduced in (6.106) to conclude that only 6 combinations of each pair are independent. in turn. l) and use (6. ∂ε 36 2 Cijkl . j) or (k. 3 To see this.107) where (6.109) which. take each pair (i. note that in the inﬁnitesimal theory.106) which readily implies that only 6 × 6 = 36 of these components are independent. write the 36 parameters as a 6 × 6 matrix and argue that only the terms on and above (or below) the major diagonal are independent. by the (6.82) or (6. equation (6. IIe and IIIe of e. Recalling (5.32)2 and noting that. To see this point. the reference conﬁguration is stress-free.107) that (6.3 ˜ T = TP0 (e) = c0 i + c1 e + c2 e2 .128 Linearly elastic solid In general. T ∂W . recall the constitutive equations (6. by assumption. the fourth-order elasticity tensor possesses 3 × 3 × 3 × 3 = 81 material constants deﬁnition of ε in (5.111) To see this. The preceding identity reduces the number of independent material parameters from 36 to 36 − The number of independent material parameters can be further reduced by material symmetry.108) and (6. 0 energy function W = W (ε) per unit volume as (6.110) where c0 . εij = εji.2 Next. implies that Cijkl = Cklij . ∂εkl ∂εij ∂εkl + 3 = 21.110) leads to (6. and c2 are function of the three principal invariants Ie . it is easy to see that the linearization of (6. ME185 . since balance of angular momentum implies that σij = σji and also. σ = c∗ Iε I + c1 ε .108) It follows from (6. However.32). 2 ∂2W ∂σij = = Cijkl . it follows that Cijkl = Cjikl = Cijlk = Cjilk .93) may be derived from a strain AF σ = W = 1 ε · Cε .

Setting c∗ = λ and c1 = 2µ.116) for all proper orthogonal Q(τ ). it follows that τ ≤t Invoking invariance under superposed rigid motions and suppressing.e. Eν (1 + ν)(1 − 2ν) µ(3λ + 2µ) λ+µ (6. Consider ﬁrst a broad class of materials with memory. 2(λ + µ) τ ≤t τ ≤t τ ≤t τ ≤t equation as (6.117) where. These are related to the Young’s e modulus E and the Poisson’s ratio ν by λ = and. their current state of stress depends not only on the current state of deformation.6.115) This means that the Cauchy stress at time t of some material particle P which occupies point X in the reference conﬁguration depends on the history of the deformation gradient of that point up to (and including) time t. t].115) DR are called simple. VISCOELASTIC SOLID 129 where c∗ and c1 are constants. it is concluded that (6. τ ≤t (6. X) .118) ME185 . τ )].114) 6. inversely. F(τ ) = R(τ )U(τ ). i. as usual. t) = T( H [F(X. but also on the deformation history. (6.6. where −∞ < τ ≤ t.112) The parameters λ and µ are known as the Lam´ constants. Choosing Q(τ ) = RT (τ ).6 Viscoelastic solid Most materials exhibit some memory eﬀects. τ ∈ (−∞. Equation (6.. the explicit dependence of functions on X. whose Cauchy stress is given by (6.113) AF E = ˆ T(X. T . (6. Materials that satisfy the constitutive law (6. ˆ ˆ Q(t)T( H [F(τ )])QT (t) = T( H [Q(τ )F(τ )]) .117) can be readily rewritten as ˆ T(t) = R(t)T( H [U(τ )])RT (t) . for all ˆ ˆ RT (t)T( H [F(τ )])R(t) = T( H [U(τ )]) . in the interest of brevity. ν = λ . µ = E 2(1 + ν) . one may rewrite the preceding 0 0 σ = λ tr εI + 2µε .

133)2. (6. Viscoelastic solid τ ≤t Upon recalling (4. E(t)]) − S( H [0. proceed by distinguishing between the past (τ < t) and the present (τ = t) in the preceding constitutive laws.121)2 as ˇ ˇ ¯ S(t) = S( H [E(τ )]) = S( H [Et (s). (6.126) The ﬁrst term on the right-hand side of (6. deﬁne the elastic response function Se as DR and the memory response function Sm as s≥0 s≥0 ˇ ¯ Se (E(t)) = S( H [0. ME185 . To this end. (6. Next.122) where. obviously. (6. E(t)]) . τ ≤t τ ≤t T τ ≤t s≥0 s≥0 s≥0 s≥0 ˆ T(t) = F(t)U−1 (t)T( H [U(τ )])U−1 (t)FT (t) . E(t)]) . the variable s is probing the history of the Lagrangian strain looking further in the past as s increases. one may alternatively write AF τ ≤t ¯ ˇ S(t) = S( H [C(τ )]) = S( H [E(τ )]) .120) As usual. (6. E(t)]) .125) In summary. Et (0) = 0. τ ≤t (6. deﬁne Et (s) = E(t − s) − E(t) . for any given time t. The second term on the right-hand side of (6. E(t)]) . this means that ˆ ˆ S(t) = J(t)U−1 (t)T( H [U(τ )])U−1 (t) = S( H [U(τ )]) .119) (6. rewrite (6.126) contains all the dependency of the stress response on past Lagrangian strain states.123) Then.126) is the stress which depends exclusively on the present state of the Lagrangian strain. (6. the stress response is given by ¯ S(t) = Se (E(t)) + Sm ( H [Et (s). equivalently.130 or.121) Now.124) ¯ ˇ ¯ ˇ ¯ Sm ( H [Et (s). E(t)]) = S( H [Et (s). Clearly.

E(t)]) = 0.4.e. Sm is rate-dependent (i. as time elapses. the fading memory condition (6. With reference to Figure 6. i. The condition (6.e. it depends on the rate E of Lagrangian strain) and also exhibits fading memory.127) is often referred to as the relaxation property. ¯ where E0 is a constant. Sm ( H [0. The latter means that the impact on the stress at time t of the deformation at time t − s (s > 0) diminishes as s → ∞.Mechanical constitutive theories 131 All viscoelastic materials can be described by the constitutive equation (6. elastic stress).126). by deﬁnition.127) implies that. ultimately.4: Static continuation Etδ (s) of Et (s) by δ. For such ˙ materials. (6. This is because it implies that any time-dependent Lagrangian strain process which reaches a steady-state results in memory response which ultimately relaxes to zero memory stress Under special regularity conditions. E(t)]) = 0 . possibly. DR (plus. it is seen that the static δ s Etδ (s) Et (s) Figure 6.127) Eδ (s) = t 0 Et (s − δ) if δ ≤ s < ∞ if 0 ≤ s < δ (6. the stress during a state deformation. is equal to S(t) = Se (E0 ).. t T s≥0 Note that.128) is the static continuation of Et (s) by δ. to the linear functional form ¯ Sm ( H [Et (s). see Figure 6. the memory response function bS m can be reduced ∞ 0 L L(E(t).5. equivalently. (6.. Sm by δ.129) ME185 . which E(t) = E0 . E(t)]) = s≥0 continuation is a rigid translation of the argument Et (s) of the memory response function the eﬀect of earlier Lagrangian strain states on Sm continuously diminishes and. s)Et(s) dS . This condition can be expressed mathematically as where AF δ→∞ s≥0 ¯ lim Sm ( H [Eδ (s). disappears altogether. or. Therefore.

where M (E(t)) = − ∞ 0 L L(E(t). upon Taylor expansion of Et (s) around t.131) Alternatively. s){−sE(t)} ds = − ∞ 0 ˙ L L(E(t). Start with the Kelvin-Voigt model. (6. s) needs Upon Taylor expansion of Et (s) around t − s. s) is a fourth-order tensor function of E(t) and s. (6. L L(E(t).134) Now.133) ˙ = M (E(t))E(t) .129) to ﬁnd that ¯ Sm ( H [Et (s). s)s ds .130).5: An interpretation of relaxation to be chosen so that Sm satisfy the relaxation property (6. one ﬁnds that ˙ Et (s) = E(t − s) − E(t) = −sE(t) + o(s2 ) . and let the ME185 . which is tantamount to neglecting long-term memory eﬀects. s){−sE(t − s)} ds = ¯ ˙ L L(E(t).130) ∞ 0 s Ignoring the second-order term in (6. (6. AF ∞ 0 T (6. Of course.132) DR which leads to ¯ Sm ( H [Et (s). where L L(E(t). s)E(t − s) ds . one ﬁnds that ˙ Et (s) = E(t − s) − E(t) = −sE(t − s) + o(s2 ) . E(t)]) = s≥0 ˙ L L(E(t). s)s ds E(t) (6. one may substitute Et (s) in (6. attempt to reconcile the general constitutive framework developed here with the classical one-dimensional viscoelasticity models of Kelvin-Voigt and Maxwell.132 Viscoelastic solid t1 t2 t Et2 (s) Et1 (s) Figure 6.127). E(t)]) = s≥0 ∞ 0 ˙ L L(E(t). which is comprised of a spring and a dashpot in parallel.

Mechanical constitutive theories 133 spring constant be E and the dashpot constant be η. (6.126). ˙ T σ ˙ σ + .126).136) is σ(t) = c(t)e− η t . consider the Maxwell model of a spring and dashpot in series with material properties as in the Kelvin-Vogt model. E E 0 = 0 (6. upon substituting into (6. The general solution of (6. in turn. In this case. Next.135) Clearly.133). one may write that σ(t) = Ee η τ ε(τ ) dτ e− η t = ˙ 0 t E E t ˙ Ee η (τ −t) ε(τ ) dτ ˙ Ee− η s ε(t − s) ds . where the memory response is obtained from a one-dimensional counterpart of (6.131). this law is a simple reduction of (6. the constitutive law becomes ε = ˙ (6. It follows that the uniaxial stress σ is σ = Eε + η ε . E η E E related to the uniaxial strain ε by (6. ˙ ˙ t (6.136) and take the accompanying initial condition to be σ(0) = 0.138) This.139) 0 DR t Noting that the initial condition results in c(0) = 0.137) which.136) leads to AF c(t) = Ee η t ε(t) . ME185 . may be integrated to yield c(t) = c(0) + ˙ Ee η τ ε(τ ) dτ . the stress response of the Maxwell model falls within the constitutive framework of (6. E (6.140) Clearly. where the elastic response function vanishes identically and the memory response can be deduced from (6.

134 Viscoelastic solid DR ME185 AF T .

Let the pressure of the free surface be p0 and assume that the AF g e2 θ Figure 7.1 Incompressible Newtonian viscous ﬂuid Parts of this chapter include material taken from the continuum mechanics notes of the late Professor P.and Initial/boundary-value Problems 7.Chapter 7 Boundary.M.1: Flow down an inclined plane 135 T . DR ﬂuid region has constant depth h. Naghdi. see Figure 7. h e3 e1 7.1.1.1 Gravity-driven ﬂow down an inclined plane Consider an incompressible Newtonian viscous ﬂuid under steady ﬂow down an inclined plan due to gravity.

4). ∂x2 (7.6) and (7.7). (7. (7. Incompressibility implies that div v = (7.136 Flow down an inclined plane Assume that the velocity and pressure ﬁelds are of the form v = v (x2 . (7. the Cauchy stress is given by −p 0 0 ∂v [Tij ] = 0 −p µ ∂x¯3 .7) where g is the gravitational constant. the rate-of-deformation tensor has components 0 0 0 ¯ 1 ∂v .4). the equations of (7.8) . x3 )e2 ˜ and T . x2 . − ∂p ˜ =0 ∂x1 ∂p ˜ ∂¯ v d2 v ¯ ∂¯ v + µ 2 + ρg sin θ = ρ + vk ∂x2 dx3 ∂t ∂xk ∂p ˜ − ρg cos θ = 0 − ∂x3 ˜ ∂v = 0.6) (7. (7.1) p = p(x1 . ˜ respectively.5) [Dij ] = 0 0 2 ∂x3 1 ∂v 0 2 ∂x¯3 0 (7.50). namely that v = v(x3 )e2 . Given (7. ∂v ¯ 0 µ ∂x3 −p Note that gravity induces a body force per unit mass equal to Given the reduced velocity ﬁeld in (7. x3 ) .4) DR linear momentum balance assume the form − ME185 Recalling the constitutive equation (6.2).2) which means that the velocity ﬁeld is independent of x2 . ¯ AF b = g(sin θe2 − cos θe3 ) .3) (7.

9) p(x2 . µ d2 v ¯ + ρg sin θ = 0 .9) results in an expression for the pressure as (7.12) into the remaining momentum balance equation (7.12) Substituting the pressure from (7.11) Substituting this equation to (7.and initial-value problems It follows from (7.10) AF p = p0 + ρg cos θ(h − x3 ) . ¯ which implies that the function f (x2 ) is constant and equal to f (x2 ) = p0 + ρg cos θh . 2µ v(x3 ) = ¯ ρg sin θ x3 x3 (h − ) .2 Couette ﬂow Enforcing the boundary conditions v (0) = 0 (no slip condition on the solid-ﬂuid interface) ¯ ρgh sin θ . ¯ Imposing the boundary condition on the free surface leads to 137 T (7. 2µ This is the steady ﬂow between two concentric cylinder of radii Ro (outer cylinder) and Ri (inner cylinder) rotating with constant angular velocities ωo (outer cylinder) and ωi (inner ME185 .Boundary. gives c2 = 0 and c1 = µ when substituted into (7.8)2 yields (7.14) yield (7. dx2 3 v (x3 ) = ¯ −ρg sin θ 2 x3 + c1 x + c2 .14) DR 7. x3 ) = −ρg cos θx3 + f (x2 ) .8)1. µ 2 (7.15) that the velocity distribution is parabolic along x3 and attains a ρg sin θ 2 maximum value of vmax = h on the free surface. h) = −ρg cos θh + f (x2 ) = p0 .3 that p = p(x2 .13) which may be integrated to (7.1. (7.15) It is seen from (7. and T23 (h) = 0 (no shear stress on the free surface). which.

ez }. Also. eθ .19) that T = −pi + µr a = ME185 Given the form of the velocity in (7.138 Couette ﬂow ωi eθ er r Ri Ro Figure 7.2. d dr v (r) ¯ (er ⊗ eθ + eθ ⊗ er ) . The ﬂuid is assumed Newtonian and incompressible.19) (7. the spatial velocity gradient can be written as DR so that It follows from (7. can be expressed as ¯ v = v(r)eθ AF p = p(r) .16). it is clear that div v = 0.9).20) (7. hence the incompress- ibility condition is satisﬁed at the outset. The velocity and pressure ﬁelds are assumed axisymmetric and. ¯ L = v ¯ d¯(r) v eθ ⊗ er − er ⊗ eθ . see Figure 7. The problem lends itself naturally to analysis using cylindrical polar coordinates with basis vectors {er . ¯ ¯ dr r r T ωo (7.16) (7.2: Couette ﬂow cylinder). r 1 v2 ¯ d¯(r) v eθ ⊗ er − v er ⊗ eθ v eθ = − er . Also. the acceleration becomes .18) (7. dr r v(r) ¯ r D = r d 2 dr (er ⊗ eθ + eθ ⊗ er ) .17) (7. the eﬀect of body forces is neglected.21) and Taking into account (A. using the cylindrical polar coordinate representation.

the velocity gradient for this ﬂow is ∂v ez ⊗ er . r + ωi ωo Ro r − r Ro Ro Ri 2 and θ-directions become µ respectively.1.. (7. z) . ∂r (7. ¯ ¯ DR of the ﬂuid of the general form while the pressure is L = 7.22)1 and using a pressure boundary condition such as.23) The integration constants c1 and c2 can be determined by imposing the boundary conditions ¯ v (Ri ) = ωi Ri and v(Ro ) = ωo Ro .3 Poiseuille ﬂow This is the ﬂow of an incompressible Newtonian viscous ﬂuid through a straight cylindrical pipe of radius R. ¯ (7. integrating equation (7.Boundary. p(R0 ) = p0 .25) p = p(r. dr r − T c2 .g.24) −1 Finally. assume that the velocity v = v(r)ez . Assuming that the pipe is aligned with the e3 -axis.21) the linear momentum balance equations in the rv 2 (r) ¯ dp = −ρ dr dr d v + 2µ =0. Upon determining these constants. e.20) and (7. leads to an expression for the pressure p(r).and initial-value problems 139 Taking into account (7.22) AF v(r) = c1 r + ¯ Ro Ri The second of the above equations may be integrated twice to give (7. (7.9).26) Clearly.27) ME185 . Taking again into account (A. d d v r dr dr r (7. the assumed velocity ﬁeld satisﬁes the incompressibility condition at the outset. the velocity of the ﬂow ¯ takes the form Ri r − Ri r v (r) = ωo Ro ¯ .

29) Noting that the acceleration vanishes identically. 2 ∂r (7. However.1 Stokes’ First Problem See Problem 5 in Homework 11. r ∂θ d2 v µ dv ∂p = 0 − +µ 2 + ∂z dr r dr cr 2 + c1 ln r + c2 . it follows that (7.2 7.32) where c1 and c2 are also constants. given that v depends only on dp = c.30) where (A. the equations of linear momentum balance in the absence of body force take the form AF ∂p = 0 ∂r 1 ∂p − = 0. one ﬁnds that r. Upon integrating (7. equation (7.26).30)3 in r.31) (7.2.33) Two additional boundary conditions are necessary (either a velocity boundary condition on one end and a pressure boundary condition on another or pressure boundary conditions on both ends of the pipe) in order to fully determine the velocity and pressure ﬁeld.28) ∂v (er ⊗ ez + ez ⊗ er ) . dz where c is a constant. ∂r (7.11) is employed. Admitting that the solution should remain ﬁnite at r = 0 DR 7.140 Stokes’ First Problem hence the rate-of-deformation tensor is D = Then. the Cauchy becomes T = −pi + µ T − 1 ∂v (er ⊗ ez + ez ⊗ er ) . while equation (7. v = 4µ v(r) = c 2 2 (r − R0 ) .30)3 requires that (7.30)1 implies that p = p(z). ME185 and imposing the no-slip condition v(R) = 0. Compressible Newtonian viscous ﬂuids . Equation (7. 4µ (7.30)2 is satisﬁed identically due to the assumption (7.

x3 ) | x3 > 0}.35) assumes velocity. note that the assumed solution renders the motion isochoric. i. in the domain.34) where U and ω are given positive constants. the acceleration ﬁeld is D23 = D32 = where the function f and the constant α are to be determined. (7. which contains a Newtonian x3 = 0 in the form viscous ﬂuid. see Figure 7.and initial/boundary-value problems 141 7.3: Semi-inﬁnite domain for Stokes’ Second Problem Adopting a semi-inverse approach.3.35). (7.. x2 . T e3 e2 e1 vp .2. Further. In addition. (7.e. The ﬂuid is subjected to a periodic motion of its boundary vp = U cos ωte1 .37) ME185 . that the velocity ﬁeld varies along x3 and is also phase-shifted by αx3 relative to the boundary AF a = −ωf sin (ωt − αx3 )e1 .36) Further.35) DR In view of (7. then it remains homogeneous for all time.2 Stokes’ Second Problem Consider the semi-inﬁnite domain R = {(x1 . if the ﬂuid is initially homogeneous ˙ (which it is assumed to be). The only non-vanishing components of the rate-of-deformation tensor are 1 df cos (ωt − αx3 ) + αf sin (ωt − αx3 ) 2 dx (7. assume a general form of the solution as v = f (x3 ) cos (ωt − αx3 )e1 . the body force is zero everywhere Figure 7.Boundary. The solution (7. mass conservation implies that ρ = 0.

dx3 These two equations can be directly integrated to give (7.42) For this equation to be satisﬁed identically for all x3 and t.40) or µ df d2 f cos (ωt − αx3 ) + 2α sin (ωt − αx3 ) − α2 f cos (ωt − αx3 ) 2 dx3 dx3 The preceding equation can be also written as µ DR and 2µα and ME185 d2 f df − α2 f cos (ωt − αx3 ) + 2µα + ρωf sin (ωt − αx3 ) = 0 .44) (7. In the e1 direction. 2 dx3 dx3 AF d2 f − α2 f = 0 dx2 3 f (x3 ) = c1 eαx3 + c2 e−αx3 f (x3 ) = c3 e− 2µα x3 . df cos (ωt − αx3 ) + αf sin (ωt − αx3 ) dx (7.45) (7.43) df + ρωf = 0 .38) Taking into account (7. (7. ρω since the mass density is homogeneous.39) = ρωf sin (ωt − αx3 ) . with tr D = div v = 0. the momentum balance equation reduces to ∂T13 = ρa1 ∂x3 (7. it is necessary and suﬃcient that (7.142 Stokes’ Second Problem Recalling (6. T .38) it is easy to see that the linear momentum balance equations in the e2 and e3 directions hold identically.46). and noting that p and µ are necessarily constant. . it follows that −p 0 T13 [Tij ] = 0 −p 0 T31 0 −p where T13 = T31 = µ (7.41) (7.36) and (7.46) .

52) 7.7.3. Taking the trace of both sides on (6.50) DR where (6.52) that (7. the velocity ﬁeld of equation (7.48) Applying the boundary condition v(0) = U cos ωte1 leads to c = U. the pressure p is constitutively speciﬁed.113) is used. where α = the form ρω 143 T ρω (7. it is easily seen that (7.53) ME185 . λ 1 σ− tr σI 2µ 3λ + 2µ or. To reconcile the two solutions. σ33 = 0. E ε33 = σ33 E . and recall that its stress-strain relation is given by (6.112) and assuming that 2 λ + 3 µ is non-zero. LINEAR ELASTIC SOLIDS respectively. 3λ + 2µ ε = ε = 1 [(1 + ν)σ − ν tr σI] . It follows from (7. ﬁnally. hence (7.49) In this problem. yet is here constant throughout the semi-inﬁnite domain owing to the homogeneity of the mass density.35) takes 2µ v = ce− 2µα x3 cos (ωt − αx3 )e1 .51) (7. E (7.1 Simple tension and simple shear In the case of simple tension along the e3 -axis.47) ρω . 7. while all other components σij = 0. tr ε = 1 tr σ . using the elastic materials parameters E and ν.3 Linear elastic solids Consider a homogeneous and isotropic linearly elastic solid. ρω (7. so that.112). AF v = Ue− 2µα x3 cos (ωt − αx3 )e1 .3. ε11 = ε22 = − νσ33 . one needs to take f (x3 ) = ce− 2µα x3 . With this expression in place.

2 Uniform hydrostatic pressure Suppose that a homogeneous and isotropic linearly elastic solid is subjected to a uniform hydrostatic pressure σ = −pI. Also. 2µ determine the material constants E and ν as (7. Recalling (7.56) can 3 3(1 − 2ν) p be used in an experiment to determine the bulk modulus by noting that tr ε = − K is the where K = Saint-Venant torsion of a circular cylinder Consider an isotropic homogeneous linearly elastic cylinder under quasistatic conditions. assume that the displacement of the cylinder may be written as (7. the lateral sides of the cylinder are assumed traction-free. The elastic constant µ can be experimentally measured by arguing that 2ε is the change in the angle between the axes e1 and e2 .3. 3λ + 2µ 3K AF u = αx3 reθ . The cylinder has length L and radius R and is ﬁxed on the one end (x3 = 0). (7. Equation (7.54) In the case of simple shear on the plane of e1 and e2 .51).3. it follows that ε = −p 1 1 I = −p I. 7.144 Saint-Venant torsion of a circular cylinder while all shearing components of strain vanish. ε33 σ12 . A simple tension experiment can be used to E = σ33 ε33 . L).55) (7. it follows that ε12 = while all other strain components vanish. the only non-zero components of stress is σ23 = σ32 . Taking into account (7. while at the opposite end (x3 = L) it is subjected to a resultant moment Me3 relative to the point with Due to symmetry. sections of constant x3 remain plane after the deformation. T ν = − ε11 . it is assumed that the cross-section remains circular and that plane coordinates (0.3 ME185 inﬁnitesimal change of volume due to the hydrostatic pressure p.57) . With these assumptions in place. 3λ + 2µ E = is the bulk modulus of elasticity. 0.56) DR 7.52).

0 1 −x2 x1 0 (7.62) (7. the stress tensor has components AF [σij ] = µα 0 0 0 0 −x2 −x2 x1 x1 .61) 0 x1 x2 = 0 .60) It can be readily demonstrated that. Recalling that eθ = −e1 rewrite the displacement as u = α(−x2 x3 e1 + x1 x3 e2 ) . 2 1 1 − 2 αx2 2 αx1 0 T 0 = 0 . 145 x1 x2 + e2 . since 0 0 −x2 0 x1 1 [ti ] = [σij ][nj ] = µα 0 (7. 0 cos θ R3 sin θ dθ = µα 3 0 r drdθ 2π 0 It follows that the inﬁnitesimal strain tensor has components 1 0 0 − 2 αx2 1 [εij ] = 0 0 αx1 .and initial-value problems where α is the angle of twist per unit x3 -length.59) (7. the tractions vanish. Further. for the lateral surfaces. the traction at x = L is −x2 0 0 0 −x2 [ti ] = [σij ][nj ] = µα 0 0 x1 0 = µα x1 .Boundary. r −x2 x1 0 0 0 DR so that the resultant force is given by − sin θ 2π R [ti ] dA = µα r cos θ x3 =L 0 0 0 − sin θ R3 2π cos θ = µα 3 0 0 On the other hand. 0 (7. in the absence of body forces. one may r r (7. all the equilibrium equations are satisﬁed identically.58) Hence.63) ME185 .

4. as a special case. (7. each face.130) and (7. assuming. The tractions. P = J(−pi + b1 B)F−T = −pF−T + b1 F . ME185 where b1 (> 0) is a constant. 0. assume that its edges are aligned with the common orthonor- mal basis {e1 . recall the general T · D = (T′ + form of the constitutive equations for isotropic non-linearly elastic materials in (6. write in that AF T = −pi + b1 B . and p is a Lagrange multiplier to be determined upon enforcing Returning to the unit cube. 2 0 0 (7. and is loaded by three pairs of equal and opposite tensile forces. Pei = cei .66). Next. First. satisfy (7. all of equal magnitude. the resultant moment with respect to (0. L) equals Me3 = x3 =L (x1 e1 + x2 e2 ) × µα(−x2 e1 + x1 e2 ) dA 2π R T r rdrdθ = µαI . isotropic and incompressible non-linearly elastic material. e2 . when resolved on the geometry of the reference conﬁguration.1 Rivlin’s cube Consider a unit cube made of a homogeneous.68) . and distributed uniformly on Taking into account (4.67) where J = 1 due to the assumption of incompressibility.64) = µα x3 =L (x2 1 + x2 ) dA 2 = µα πR4 is the polar moment of inertia of the circular cross-section.66) DR the incompressibility constraint. 7.65) 3 3 where T′ and D′ are the deviatoric parts of T and D. respectively. where I = 2 7.146 Rivlin’s cube Moreover.4 Non-linearly elastic solids (7. e3 } of the reference and current conﬁguration. observe that the pressure p is work-conjugate to the volume change 1 1 tr T i) · (D′ + tr D i) = T′ · D′ + (−p) div v .83) and. b2 = 0. one may write (7.

λ2 = λ3 = λ1 and λ3 = λ1 = λ2 .71) or AF b1 λ2 = cλi + p . This. since. Solutions of the form T p + b1 λi .72) where i = 1.68) that the equilibrium equations are satisﬁed identically in the absence of body forces. b1 Note that c is the same for all faces of the cube.74) is obviously (7. λi c =η.75) This corresponds to the cube remaining rigid under the inﬂuence of the tensile load.69) F = λ1 e1 ⊗ e1 + λ2 e2 ⊗ e2 + λ3 e3 ⊗ e3 (7.69) to conclude that c = − (7. is constant and uniform.74) that (7. or b1 (λ3 + λ1 ) = c λ3 = λ1 λ1 = λ2 = λ3 = 1 .74). i j λ1 = λ2 or b1 (λ1 + λ2 ) = c λ2 = λ3 or b1 (λ2 + λ3 ) = c .and initial/boundary-value problems 147 where c > 0 is the magnitude of the tractions per unit area in the reference conﬁguration. 3.73) where i = j. one may also write that P = c(e1 ⊗ e1 + e2 ⊗ e2 + e3 ⊗ e3 ) . where not all conditions can be chosen from the same column in (7.74) subject to λ1 λ2 λ3 = 1. subject to the condition λ1 λ2 λ3 = 1. λ2 + λ3 = λ3 + λ1 = (7. It is clear from (7. combine (7. i b1 (λ2 − λ2 ) = c(λi − λj ) . Eliminating the pressure in the preceding equation leads to (7. Returning to the constitutive equations. Explore one of these solutions. 2. Another possible set of solutions is λ1 = λ2 = λ3 .76) ME185 . the force on each face (7. means that DR (7. in turn. by assumption. say λ1 = λ2 = λ3 by setting λ3 = λ and noting from (7.67) and (7. Therefore.Boundary. One solution of (7.70) are sought for this boundary-value problem.

note that f ′ (λ) = 3λ2 − 4ηλ + η 2 hence the extrema of f occur λ1.148 so that Rivlin’s cube The preceding equation may be rewritten as f (λ) = λ3 − 2ηλ2 + η 2 λ − 1 = 0 .4: Function f (λ) in Rivlin’s cube (7. f (−∞) = −∞. ME185 . then there are no additional solutions (Case I). Case III Case II η Case I λ λ1 λ2 λ3 = (η − λ)2 λ = 1 . and f (∞) = ∞. Further.80) η 4 3 η − 1 . f (η) = −1 . The plot in Figure 7.2 = and are equal to at T . λ1 = λ2 = λ3 = 1 is always a solution.4 depicts the essential features of f (λ). < 1.77) (7. f ′′ (λ) = 6λ − 4η . If 4 3 η 27 4 3 η 27 In summary. (7. If (Case II). To examine the roots of f (λ) = 0. then there is one set of three additional solutions corresponding to λ = η 3 2. f DR 0 −1 1. = 1.79) (7.78) AF η η where f ′′ ( ) = −2η < 0 (maximum) 3 3 ′′ η where f (η) = η > 0 (minimum) η/3 Figure 7.81) f( ) = 3 27 It is also obvious from the deﬁnition of f (λ) that f (0) = −1. (7.

82) DR ¯ (vol P)T = ∂P Taking into account (4. If 4 3 η 27 149 > 1.83). MULTISCALE PROBLEMS 3. the preceding equation may be rewritten as t ⊗ x da + ρ(b − a) ⊗ x dv ρb ⊗ x dv − ˙ ρx ⊗ x dv + P (7. t0 (7. In such cases.1 The virial theorem The virial theorem is a central result in the study of continua whose constitutive behavior is derived from an underlying microscale particle system. (7. This is.7. for example.85) and note that d dt P ˙ ρx ⊗ x dv = lim 1 T →∞ T P ˙ ρx ⊗ x dv |t=t0 +T − P ˙ ρx ⊗ x dv |t=t0 = 0. (7. when one wishes to analyze metals and semiconductors at very small length and time scales.83) ˙ ˙ ρx ⊗ x dv . 7. Preliminary to the derivation of the theorem. d dt P two roots of f (λ) which are smaller than η (Case III). then there are two sets of three additional solutions corresponding to the Note that the root λ = λ3 > η is inadmissible because it leads to λ1 = λ2 = η − λ < 0.5. 7.84) P = ∂P t ⊗ x da + P Deﬁne next the (long) time-average φ of a time-dependent quantity φ = φ(t) as φ = lim 1 T →∞ T t0 +T φ(t) dt . (7. multiscale analyses oﬀer a means for relating kinematic and kinetic information between the continuum and the discrete system.5.5 Multiscale problems mechanics. recall that the mean Cauchy stress T in a material region P can be written as ¯ (vol P)T = ∂P AF t ⊗ x da − P It is sometimes desirable to relate the theory of continuous media to theories of particle T div T ⊗ x dv . at which the continuum assumption is not unequivocally satisﬁed.86) ME185 . the case.

α (due to inter-particle potentials) and an external part f int.93) is a statement of the virial theorem.83) is ¯ (vol P)T = Second Law.α .92) which leads to an estimate of the mean Cauchy stress in terms of the underlying particle system dynamics as N N ¯ . the total force f α is comprised of an internal part f ext. in each particle. (vol P)T = α=1 ˙ ˙ m x ⊗x α α α + α=1 f ext.91) Upon ignoring the body forces in the continuum problem and comparing (7. namely ∂P ˙ ρx⊗ t ⊗ x da + T ρb ⊗ x dv + P x dv throughout time.88) P ˙ ˙ ρx ⊗ x dv . (7.α ⊗ xα . the preceding equation may be rewritten as N N N DR α=1 α=1 − ˙ ˙ mα xα ⊗ xα = f int. (7. (7. It is easy to show that the above equation leads to mα d α ˙ ˙ ˙ (x ⊗ xα ) − mα xα ⊗ xα = f α ⊗ xα . while f α is the total force acting on particle α.α ⊗ xα .150 The virial theorem P The preceding time-average vanishes due to the assumed boundedness of the integral mula (7. (7. (7. (vol P)T = − N α=1 f int. it can be argued that there is a one-to-one correspondence between the three terms in each statement. it is seen that N AF N − α=1 ˙ ˙ mα xα ⊗ xα = α=1 f α ⊗ xα . (7.α ⊗ xα + α=1 f ext.93) Equation (7.87) where mα and xα are the mass and the current position of particle α. one may argue that ¯ . Speciﬁcally. dt (7.α ⊗ xα .87) to (7. Using (7.86).89) for the totality of the particles and assuming boundedness of the term N α=1 ˙ mα xα ⊗ xα . ME185 .91). for the mean stress.89) Taking time averages of (7.90) Recognizing that. the time-averaged counterpart of the mean-stress for- Turn attention now to a system of N particles whose motion is governed by Newton’s ¨ mα xα = f α .

6 Expansion of the universe DR AF ME185 T .7.6. EXPANSION OF THE UNIVERSE 151 7.

1) (A. T z = x3 (A.1 APPENDIX A A.7) ∂r r ∂θ ∂z ∂r r ∂θ ∂z When the gradient operator grad is applied on a vector function v = vr er + vθ eθ + vz ez . where er = e1 cos θ + e2 sin θ eθ = − e1 sin θ + e2 cos θ .2) Here. gradT v = ∂ 1 ∂ ∂ er + eθ + ez ∂r r ∂θ ∂z ⊗v . Conversely. one may easily conclude that (A. then (A.1 Cylindrical polar coordinate system The basis vectors of the cylindrical polar coordinate system are {er . eθ . x2 = r sin θ x3 = z . ez }.3) (A. (A. ez = e3 r = x2 + x2 1 2 .4) Further. conversely. θ is the angle formed between the e1 and er vectors. (A. x1 = r cos θ . . grad v = Using chain rule.8) ∂vr ∂vθ ∂vz er ⊗ er + eθ ⊗ er + ez ⊗ er + ∂r ∂r ∂r 1 ∂vr 1 ∂vθ 1 ∂vz − vθ er ⊗ eθ + + vr eθ ⊗ eθ + ez ⊗ eθ + r ∂θ r ∂θ r ∂θ ∂vθ ∂vz ∂vr er ⊗ ez + eθ ⊗ ez + ez ⊗ ez .9) ∂z ∂z ∂z ME185 . θ = arctan x2 x1 . (A. (A.5) and. one may express the gradient of a scalar function f in polar coordinates ∂f 1 ∂f ∂f 1 ∂ ∂ ∂ er + eθ + ez = er + eθ + ez f .6) DR as grad f = or. upon expanding. e1 = er cos θ − eθ sin θ AF ez = e3 e2 = er sin θ + eθ cos θ .Appendix A A.

given a symmetric tensor function T = Trr er ⊗ er + Trθ (er ⊗ eθ + eθ ⊗ er ) + Trz (er ⊗ ez + ez ⊗ er )+ Tθθ eθ ⊗ eθ + Tθz (eθ ⊗ ez + ez ⊗ eθ ) + Tzz ez ⊗ ez .11) + + + ∂r r r ∂θ ∂z T .10) its divergence can be expanded to div T = DR ME185 AF ∂Trr Trr − Tθθ 1 ∂Trθ ∂Trz er + + + + ∂r r r ∂θ ∂z ∂Trθ 2Trθ 1 ∂Tθθ ∂Tθz eθ + + + + ∂r r r ∂θ ∂z ∂Trz Trz 1 ∂Tθz ∂Tzz ez .2 Cylindrical polar coordinate system Also.A. (A. (A.

- Continuum Edit
- Continuum Mechanics
- A 012620107
- INTRODUCTION TO CONTINUUM MECHANICS
- Experiment N1.doc
- continuum mechanics
- On Helmholtz Theorem and its interpretations
- Continuum mechanics problems
- MechQuiz1
- 5987860 Continuum Mechanics
- chp-04
- Introduction to Continuum Mechanics - Vector and Tensor Calculus
- Chapter 2 Statics2
- Continuum Mechanics
- Teorema Pi Buckingham 01
- Continuum Mechanics
- Diagrams
- Continuum Mechanics
- engineer
- Continuum Mechanics
- Lesson-9-moments.pdf
- Intermediate Fluid Mechanics Notes
- Lecture 4
- Continuum Mechanics
- 1010
- (- Chaves, E.W.v. ) - Notes in Continuum Mechanics (e)
- Real Solutions to the Nonlinear Helmholtz Equation With Local Nonlinearity - 2014
- Continuum Mechanics
- Surveying Lecture Notes to Be Acetated-errors
- Continuum Mechanics

Are you sure?

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

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd