You are on page 1of 211

Lectures on Continuum Mechanics

Rami Vainio
Spring 2012
Continuum Mechanics spring 2012
Continuum Mechanics 2012
53749 Basics of Continuum Mechanics (Jatkumomekaniikan perusteet / JMP)
53731 Continuum Mechanics (Jatkuvan aineen mekaniikka / JAM)
Lectures: 26 h; Tue 1012 and Wed 1214 in Exactum B121
Week 6 (0610 Feb): lectures on Tue, Wed and Fri (1416, D106)
Week 7 (1317 Feb): lecture on Wed
Lecturer: Rami Vainio; phone: (09) 191 50676; email: rami.vainio@helsinki.
Exercises: 12 h. 45 problems per week. Starting on Week 4.
JMP: Mon 1416 in D106; Rami Vainio
Getting points requires the student to volunteer to present the solution on the blackboard. All
volunteers get 3 points per problem and one volunteer will be selected at randomto present the
solution. The one presenting the solution will get an extra point.
JAM: Fri 1416 in D106: Alexey Isavnin (alexey.isavnin@helsinki.)
Week 6 (0610 Feb): no exercises
Week 7 (1317 Feb): exercises on Tue (1012, B121) and Fri (normal time)
Submit solutions to the problems, which will be graded (on scale 0-3 per problem). Extra points
(1 per problem) can be obtained by presenting the problem on the blackboard.
Grading: nal exam 80% + exercises 20%.
Continuum Mechanics spring 2012 1
Course outline
The course will cover the basic theory of continuum mechanics.
We will follow the textbook
1
A.J.M. Spencer: Continuum Mechanics. Longman, 1980. (Also Dover, 2004.)
Additional material on applications will be added, where appropriate.
Preliminary Outline
Preface. Vector and tensor analysis (2 lectures)
Particle kinematics (1 lecture)
Analysis of stress (2 lectures)
Motions and deformations (2 lectures)
Conservation laws (1 lecture)
Constitutive equations (2 lectures)
Applications (3 lectures)
Suggested additional literature (available in the library):
A.B. Bhatia & R.N. Singh: Mechanics of Deformable Media. IOP Publishing, 1986.
1
Very cheap to purchase as a paperback. It is recommended that students buy their own copies, but the library has a few copies as well.
Continuum Mechanics spring 2012 2
Preface
Continuum Mechanics spring 2012 3
Scope of continuum mechanics
Matter consist of atoms matter discontinuos at mi-
croscopic scales.
Very large scales: matter properties have inhomogeneities.
Intermediate scales: meaningful local averages of
physical quantities.
Continuum Mechanics (CM, jatkuvan aineen mekaniikka)
is concerned with the mechanical behaviour of solids
(kiinteat aineet) and uids (uidit, eli nesteet ja kaasut)
on the macroscopic scale ignoring the discrete nature of
matter, utilizing the average values of physical quanti-
ties, dened at intermediate scales, treating them as con-
tinuous functions of position and time. Thus, CM is a
eld theory (kenttateoria).
macroscopic
log(distance)
l
o
g
(
d
e
n
s
i
t
y
)
atomic
scale
scale
CM assumes that it is possible to associate a matter particle with every point of space occupied by a
body. Physical quantities can be as being carried by these particles.
CM assumptions valid as long as
dimensions of the bodies characteristic microscopic scales (e.g., a or )
Microscopic scale need not be atomic (granular material, e.g., sand)
Continuum Mechanics spring 2012 4
Variables, equations, and problems of CM
Mechanics is the science dealing with the interaction between force and motion.
Primary variables of interest in CM are of two kinds
forces per unit area or unit volume (as functions of position and time)
kinematic variables: displacement, velocity and acceleration (as functions of position and time)
Equations of CM are of two kinds:
universal physical laws applicable to all materials: conservation laws (CL, sailymislait) of mass,
momentum, [angular momentum,] and energy
equations describing the properties of materials: constitutive equations (CE, rakenneyhtalot)
Problems of CM are of two kinds:
formulation of CEs (essentially empirical)
solution of CEs together with the CLs and appropriate boundary conditions.
Continuum Mechanics spring 2012 5
Some examples of systems that can be treated with CM
Solid bodies:
analysis of strength of materials and structures (engineering applications)
analysis of wave propagation, i.e., non-destructive diagnostics of structure (ultrasound)
geophysics (e.g., mountain uplift; landslides; exure and fracture of the Earths crust; glaciers)
Fluids:
computational uid dynamics and aerodynamics
plasma physics (physics of controlled fusion; space physics)
solar and stellar physics; astrophysics
atmospheric sciences; meteorology
geophysics (lava ows; water ow)
CM is an extremely versatile tool for the analysis of macroscopic phenomena!
Continuum Mechanics spring 2012 6
1 Mathematical methods
matrix algebra
vectors and tensors
Continuum Mechanics spring 2012 7
Mathematical methods in continuum mechanics
CM is a eld theory, i.e., physical quantities are regarded as continuous (or at least piecewise contin-
uous) functions of position and time.
The basic elds in CM are scalars
, T, . . . ,
vectors
u, v, a, . . . ,
and tensors
, F, . . . .
Thus, we need to master vector and tensor algebra and analysis.
The equations of motion in particle and rigid-body mechanics are ordinary dierential equations. CM is
a eld theory, i.e., the uid particles are numbered by continuous variables. equations of motion (or
the conservation laws) become partial dierential equations.
Many algebraic problems dealing with vectors and tensors can be reduced to problems of matrix algebra.
Thus, we will rst summarize some useful results from matrix algebra.
Continuum Mechanics spring 2012 8
Matrices
An mn matrix A is an ordered rectangular array of mn elements:
A = (A
ij
) =
_
_
_
_
A
11
A
12
A
1n
A
21
A
22
A
2n
.
.
.
.
.
.
.
.
.
.
.
.
A
m1
A
m2
A
mn
_
_
_
_
In CM, the matrices are usually either 3 3 square matrices, 3 1 column matrices, or 1 3 row
matrices. We will
denote 3 3 square matrices by bold-face roman capital letters (A, B, etc.),
denote 3 1 column matrices by bold-face lower-case roman letters (a, b, etc.), and
treat 1 3 row matrices as transposes of 3 1 column matrices (a
T
, b
T
, etc.)
unless otherwise noted.
2
The product of an m n matrix A with an n m matrix B, C = AB, is an m m matrix
dened as
C = (C
ij
), C
ij
=
n

k=1
A
ik
B
kj
summ. conv.
= A
ik
B
kj
,
where the summation convention
3
has been introduced.
2
Recall that the transpose of an mn matrix A is an n m matrix A
T
with A
T
ij
= A
ji
.
3
Summation convention implies that one sums automatically over repeated indices.
Continuum Mechanics spring 2012 9
Symmetric matrices. Unit matrix. Trace
Denition. A square matrix A is symmetric if
A = A
T
, A
ij
= A
ji
and anti-symmetric if
A = A
T
, A
ij
= A
ji
.
Denition. The 3 3 unit matrix (yksikkomatriisi) is denoted by I and its elements are given by the
Kronecker delta
ij
. Thus,
I = (
ij
) (i, j, = 1, 2, 3), where
ij
= 1 i = j, and
ij
= 0 i ,= j
Multiplication by unit matrix leaves any square matrix A, column matrix a, or row matrix a
T
unchanged:
AI = IA = A; Ia = a; a
T
I = a
T
Denition. The trace (jalki) of a square matrix A is denoted by tr A, and is the sum of the diagonal
elements of A,
tr A = A
ii
_
=

i
A
ii
_
In particular, tr I =
ii
= 3.
Continuum Mechanics spring 2012 10
Determinant
Denition. The determinant (determinantti) of a square matrix A is denoted by det A, and is given
by
det A =

(i
1
,i
2
,...,i
n
)
A
1i
1
A
2i
2
A
ni
n
,
where (i
1
, i
2
, . . . , i
n
) goes through all permutations of (1, 2, . . . , n) and the sign is + if the per-
mutation is even and if it is odd.
For a 3 3 matrix, the determinant is given by
det A =
1
6
e
ijk
e
rst
A
ir
A
js
A
kt
,
where e
ijk
is the alternating symbol or permutation symbol, dened as
e
ijk
=
_
_
_
_
_
_
_
1 if (i, j, k) is an even permutation of (1, 2, 3)
1 if (i, j, k) is an odd permutation of (1, 2, 3)
0 otherwise
Thus, in particular,
det I =
1
6
e
ijk
e
rst

ir

js

kt
=
1
6
e
ijk
e
ijk
=
1
6
3! = 1
Continuum Mechanics spring 2012 11
A useful theorem
Theorem. e
mpq
det A = e
ijk
A
im
A
jp
A
kq
Proof. It can be veried directly (exercise), that
e
ijk
e
rst
=

ir

is

it

jr

js

jt

kr

ks

kt

=
ir
(
js

kt

jt

ks
)
is
(
jr

kt

jt

kr
)+
it
(
jr

ks

js

kr
).
Thus,
e
mpq
det A = e
mpq
(
1
6
e
ijk
e
rst
A
ir
A
js
A
kt
) =
1
6
e
ijk
(e
mpq
e
rst
)A
ir
A
js
A
kt
= e
ijk
1
6

mr

ms

mt

pr

ps

pt

qr

qs

qt

A
ir
A
js
A
kt
= e
ijk
1
6
[
mr
(
ps

qt

pt

qs
)
ms
(
pr

qt

pt

qr
) +
mt
(
pr

qs

ps

qr
)]A
ir
A
js
A
kt
=
1
6
[e
ijk
A
im
(A
jp
A
kq
A
jq
A
kp
)e
ijk
A
jm
(A
ip
A
kq
A
iq
A
kp
)
.
=e
ijk
A
im
(A
jp
A
kq
A
jq
A
kp
)
+e
ijk
A
km
(A
ip
A
jq
A
iq
A
jp
)
.
=e
ijk
A
im
(A
jp
A
kq
A
jq
A
kp
)
]
=
1
2
e
ijk
A
im
(A
jp
A
kq
A
jq
A
kp
) =
1
2
e
ijk
A
im
(A
jp
A
kq
+A
kq
A
jp
) = e
ijk
A
im
A
jp
A
kq
2
Continuum Mechanics spring 2012 12
Inverse of a matrix. Orthogonal matrices
Denition. The inverse A
1
of a square matrix A (neliomatriisin A kaanteismatriisi) is dened as
the square matrix, whose product with A gives the unit matrix, i.e.,
A
1
A = AA
1
= I.
A necessary and sucient condition for the existence of A
1
is that det A ,= 0.
Denition. A square matrix Q is orthogonal (ortogonaalinen) if it has the property
Q
1
= Q
T
.
It follows that if Q is orthogonal, then
QQ
T
= I, Q
T
Q = I
and
1 = det I = det QQ
T
= det Qdet Q
T
= (det Q)
2
since det AB = det Adet B and det A
T
= det A for all square matrices A and B (exercise).
Thus, for an orthogonal matrix
det Q = 1.
We will consider mainly proper orthogonal matrices, for which det Q = 1.
If Q
1
and Q
2
are two orthogonal matrices, then Q
1
Q
2
is also an orthogonal matrix.
Continuum Mechanics spring 2012 13
Eigenvalues and eigenvectors
Consider
Ax = x (A I)x = 0,
where A is a known square matrix, x is an unknown column matrix and is an unknown scalar.
For non-trivial solutions to exist,
det(A I) = 0. Characteristic equation (karakteristinen yhtalo) for A
If A is a 3 3 matrix, the LHS is a 3rd order polynomial in . The roots,
1
,
2
, and
3
, of the
characteristic equation are the eigenvalues (ominaisarvot) of the matrix A.
Ass.
1
,
2
, and
3
are distinct. Then, the eq. (A
1
I)x = 0 has a non-trivial solution, x
(1)
,
indeterminate to within a scalar multiplier. The solution x
(1)
is the eigenvector (ominaisvektori) of A
related to eigenvalue
1
. Eigenvectors x
(2)
and x
(3)
related to
2
and
3
dened analogously.
Ass. A is real and symmetric. Eigenvalues
i
of A real (see Spencer).
Ass.
1
,
2
, and
3
distinct. Eigenvectors orthogonal and can be normalized so that (see Spencer)
x
(r)
T
x
(s)
= 0 = (0), r ,= s.
x
(r)
T
x
(s)
= 1 = (1) r = s.
Ass. A real and symmetric and
1
=
2
,=
3
. x
(3)
uniquely determined, and x
(1)
and x
(2)
can
be chosen as any two mutually orthogonal column matrices orthogonal to x
(3)
.
If
1
=
2
=
3
, any three mutually orthogonal column matrices can be chosen as eigenvectors x
(r)
.
Continuum Mechanics spring 2012 14
Properties of eigenvalues and eigenvectors
If
i
and x
(i)
are eig.values and -vectors of A then
n
i
and x
(i)
are eig.values and -vectors of A
n
.
Let
1
,
2
, and
3
be the roots of the characteristic eq. We have
det(A I) = (
1
)(
2
)(
3
)
=0
det A =
1

3
.
Consider the 3 3 matrix P dened by P
T
= (x
(1)
x
(2)
x
(3)
). Thus, P
T
P = I P orthogonal.
Also, since Ax
(r)
=
r
x
(r)
, we have
AP
T
= (Ax
(1)
Ax
(2)
Ax
(3)
) = (
1
x
(1)

2
x
(2)

3
x
(3)
)
and
PAP
T
=
_
_
_
_
x
(1)
T
x
(2)
T
x
(3)
T
_
_
_
_
(
1
x
(1)

2
x
(2)

3
x
(3)
) =
_
_

1
0 0
0
2
0
0 0
3
_
_
.
Thus, tr PAP
T
=
1
+
2
+
3
and tr (PAP
T
)
2
=
2
1
+
2
2
+
2
3
. However, since P is orthogonal,
tr PAP
T
= P
ij
A
jk
P
ik
= P
ij
P
ik
A
jk
=
jk
A
jk
= A
kk
= tr A.
Since (PAP
T
)
2
= PAP
T
PAP
T
= PA
2
P
T
, tr (PAP
T
)
2
= tr (PA
2
P
T
) = tr A
2
. Thus,
tr A =
1
+
2
+
3
, tr A
2
=
2
1
+
2
2
+
2
3
.
Continuum Mechanics spring 2012 15

CaleyHamilton theorem. The polar decomposition theorem


Since tr A =
1
+
2
+
3
, tr A
2
=
2
1
+
2
2
+
2
3
, and
0 = det(A I) = (
1
)(
2
)(
3
)
=
3
+ (
1
+
2
+
3
)
2
(
2

3
+
3

1
+
1

2
) +
1

3
,
we can write the characteristic equation in form

2
tr A +
1
2
[(tr A)
2
tr A
2
] det A = 0.
Theorem. (

CayleyHamilton) A square matrix satises its own characteristic equation, i.e., for any
square matrix A,
A
3
A
2
tr A +
1
2
A[(tr A)
2
tr A
2
] det A = 0.
Proof. Consult standard algebra textbooks. 2
Denition. A matrix A is positive denite (positiivideniitti) if x
T
Ax is positive for all non-zero
column matrices x.
A necessary and sucient condition for A to be positive denite is that its eigenvalues are all positive.
Theorem. (Polar decomposition; polaarihajotelma) A non-singular square matrix F can be decom-
posed, uniquely, into either of the products
F = RU, F = VR,
where R is an orthogonal matrix and U and V are positive denite symmetric matrices.
Proof. See Spencer. 2
Continuum Mechanics spring 2012 16
Vectors
Consider vectors in three-dimensional Euclidean space; denoted by lower-case bold-face italic letters
(e.g., a, b, etc.).
Distinguish between column matrices purely algebraic quantities and
vectors physical quantities with direction and magnitude (e.g.,
displacement, velocity, force).
Characteristic properties of vectors:
1. a vector requires a magnitude and direction for its complete specication;
2. two vectors are compounded using the parallelogram law (see gure).
Ass. a right-handed Cartesian coordinate system with origin O.
Ass. e
1
, e
2
, and e
3
are unit vectors in the directions of the three coordinate
axes, x
1
, x
2
, and x
3
.
Then, e
1
, e
2
, e
3
are called the base vectors (kantavektorit) of the coordinate
system.
x
2
x
1
a
1
e
1
e
2
a
2
a
3
e
3
a
b
a b +
x
3
a
Parallelogram addition law any vector a can be given with the aid of its components a
i
in the
specied coordinate system
a = a
1
e
1
+a
2
e
2
+a
3
e
3
= a
i
e
i
.
The components are related to the magnitude a of the vector a by a
2
= a
i
a
i
.
Continuum Mechanics spring 2012 17
Scalar product, vector product and triple products
The scalar product a b (skalaari- l. pistetulo) of two vectors a and b with magnitudes a and b and
the angle (a, b) = is the scalar quantity
a b = ab cos = a
i
b
i
.
Thus,
e
i
e
j
=
ij
.
The vector product a b (vektori- l. ristitulo) of a and b is a vector whose
direction is normal to the plane of a and b, in the sense of the right-hand rule, and
whose magnitude is ab sin . In terms of components
a b =

e
1
e
2
e
3
a
1
a
2
a
3
b
1
b
2
b
3

= e
ijk
e
i
a
j
b
k
.
Thus, the triple scalar product (a b) c (skalaarikolmitulo) is given in terms of components as
(a b) c = e
ijk
c
i
a
j
b
k
= e
ijk
a
i
b
j
c
k
=

a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3

.
The triple vector product a (b c) (vektorikolmitulo) can be given in terms of dot products as
a (b c) = b(a c) c(a b)
Continuum Mechanics spring 2012 18
Coordinate transformations
A vector quantity is independent of any coordinate system (but a vector has dierent components in
dierent coordinate systems). Components of a vector can be given as a column matrix, but that matrix
species the vector only if the coordinate system is also specied.
Ass. Coordinate system translated from O to O
t
without rotating the coordinate axes.
Base vectors e
i
unchanged, and the components of any vector a in are the same in the two systems.
Ass. e
1
, e
2
, e
3
are the base vectors of a coordinate system derived from
the original system through rigid rotation.
Ass. a has components a
i
in the original coordinate system and components
a
i
in the new system:
a = a
i
e
i
= a
i
e
i
Ass. M
ij
= cos(( e
i
, e
j
)) = e
i
e
j
. Thus, e
i
= M
ij
e
j
.
Now,
e
1
e
2
e
3
x
2
x
1
x
3
x
2
x
3
e
3
e
1
x
1
e
2
e
i
e
j
= M
ir
e
r
M
js
e
s
= M
ir
M
js
e
r
e
s
= M
ir
M
js

rs
= M
ir
M
jr
But since e
i
are mutually orthogonal unit vectors, e
i
e
j
=
ij
,
M
ir
M
jr
=
ij
MM
T
= I, where M = (M
ij
)
Thus, M is orthogonal (in fact, proper orthogonal, if the coordinate system is right-handed).
Continuum Mechanics spring 2012 19
From e
i
= M
ij
e
j
, we get
M
ik
e
i
= M
ik
M
ij
e
j
=
kj
e
j
= e
k
,
because M
T
M = I M
ri
M
rj
=
ij
. Thus, e
i
= M
ji
e
j
is another relation for the base vectors.
Now, a = a
i
e
i
= a
j
e
j
= a
j
M
ij
e
i

a
i
= M
ij
a
j
, i.e., a = Ma (1)
where a = (a
1
a
2
a
3
)
T
and a = ( a
1
a
2
a
3
)
T
. Because M is orthogonal,
a = M
T
a, i.e., a
i
= M
ji
a
j
(2)
In particular, if a is the position vector x of the point P relative to the origin O, then
x
i
= M
ij
x
j
, x
i
= M
ji
x
j
where x
i
and x
i
are the coordinates of the point P in the rst and second coordinate system, respectively.
The transformation laws (1) and (2) are consequences of the vector addition law and, in fact, equivalent
to this law. Thus, we could equivalently dene that a vector is a quantity with magnitude and direction,
which can be given with a set of components transforming according to the laws (1) and (2) under a
rotation of the coordinate system. (This is used in many books of CM as well.)
Continuum Mechanics spring 2012 20
Pseudo-vectors
We limited ourselves to right-handed coordinate systems and coordinate transformations with det M =
+1 (rotations). If transformations to left-handed coordinate systems are allowed, i.e., if we allow for
det M = 1, we need to distinguish between
a) vectors, whose components transform as a
i
= M
ij
a
j
, and
b) pseudo-vectors, whose components transform as a
i
= (det M)M
ij
a
j
.
Examples of pseudo-vectors include a b, where a and b are vectors, the angular velocity vector
(kulmanopeusvektori ), torque (vaantomomentti ), and magnetic eld.
Example: consider the improper orthogonal matrix M = I and the position vectors, a and b, of two
points A and B wrt. origin O . Thus, the coordinates of the points A and B change according to
a
i
= M
ij
a
j
=
ij
a
j
= a
i
,

b
i
= b
i
.
Clearly, the transformation is a coordinate inversion, and det M = 1. The cross product c = ab
transforms according to
c
i
= e
ijk
a
j

b
k
= e
ijk
(a
j
)(b
k
) = e
ijk
a
j
b
k
= c
i
= (1)(
ij
c
j
) = (det M)M
ij
c
j
,
and is clearly a pseudo-vector.
We will not consider left-handed coordinate systems, so the distinction is not important on this course.
Continuum Mechanics spring 2012 21
Invariants
As a b = ab cos , its value must be independendent of the coordinate system used. To conrm this,
we calculate
a
i

b
i
= M
ij
a
j
M
ik
b
k
=
jk
a
j
b
k
= a
k
b
k
= a
i
b
i
.
A quantity formed from the components of the vectors a and b, whose value is independent of the
coordinate system used, is called an invariant of the vectors a and b. In general, a physical quantity,
whose value does not depend on the coordinate system used, is an invariant.
As the vector product is also dened geometrically, it must have a similar invariance property. In fact:
a b = e
ijk
e
i
a
j

b
k
= e
ijk
M
ip
e
p
M
jq
a
q
M
kr
b
r
= (e
ijk
M
ip
M
jq
M
kr
)e
p
a
q
b
r
] e
pqr
det M = e
ijk
M
ip
M
jq
M
kr
= (e
pqr
det M)e
p
a
q
b
r
= e
pqr
e
p
a
q
b
r
provided that det M = +1. Thus, a b is an invariant of a and b, if we limit to the use of
right-handed coordinate systems.
Continuum Mechanics spring 2012 22
The dyadic product
The dyadic product (dyaditulo) of two vectors a and b is written as a b and is dened though its
inner products with any vector c:
(a b) c = a(b c) = b
j
c
j
a
i
e
i
, c (a b) = (c a)b = c
i
a
i
b
j
e
j
It has the properties:
(a) b = a (b) = (a b)
a (b +c) = a b +a c
(b +c) a = b a +c a,
where is a scalar. Thus,
a b = (a
i
e
i
) (b
j
e
j
) = a
i
b
j
e
i
e
j
. (3)
In general, a b ,= b a.
The form (3) is invariant:
a
i

b
j
e
i
e
j
= M
ip
a
p
M
jq
b
q
(M
ir
e
r
) (M
js
e
s
) = M
ip
M
ir
M
jq
M
js
a
p
b
q
e
r
e
s
=
pr

qs
a
p
b
q
e
r
e
s
= a
r
b
s
e
r
e
s
= a
i
b
j
e
i
e
j
.
The dyadic products e
i
e
j
of the base vectors are called the unit dyads (yksikkodyadi ).
Generalization: triadic product a b c, etc., with, e.g., a (b c d) = (a b)(c d), etc.
Continuum Mechanics spring 2012 23
Tensors
Denition. A second-order Cartesian tensor is a linear combination of dyadic products.
Since dyadic products are themselves linear combinations of unit dyads, a second-order Cartesian tensor
is a linear combination of unit dyads. Thus,
A = A
ij
e
i
e
j
We shall denote a (Cartesian) tensor with bold-face italic capital letters, unless otherwise noted.
The coecients A
ij
are called the components of the tensor A. As vectors exist independently of any
coordinate system and tensors are linear combinations of their dyad products, also tensors must exist
independently of the coordinate system. (However, their components can be specied only after the base
has been introduced.) Thus,
A = A
ij
e
i
e
j
=

A
ij
e
i
e
j
However,
A
ij
e
i
e
j
= A
ij
(M
ri
e
r
) (M
sj
e
s
) = A
ij
M
ri
M
sj
e
r
e
s
=

A
rs
e
r
e
s
so

A
ij
= M
ip
M
jq
A
pq
, i.e.,

A = MAM
T
(4)
Eq. (4) is the transformation law for components of second-order tensors. We can equivalently dene
tensors as quantities with nine components transforming according to Eq. (4) under rotations.
Continuum Mechanics spring 2012 24
Tensors of order n
Denition. Cartesian tensor of order n is a linear combination of n-adicproducts.
Cartesian tensors of order n can be expressed through their components as
A = A
ij...m
.
n indices
e
i
e
j
e
m
.
n products
and the components transform accroding to the rule

A
ijm
= M
ip
M
jq
M
ms
A
pqs
. (5)
Thus, a vector can be interpreted as tensor of order one. Again, a tensor of order n can be equivalently
dened as a quantity with 3
n
components transforming according to Eq. (5) under rotations.
A scalar, which has a single component unchanged in a coordinate transformation, can be regarded as a
tensor of order zero.
The inverse relation to Eq. (5) is
A
ijm
= M
pi
M
qj
M
sm

A
pqs
.
Thus, for a second-order tensor,
A
ij
= M
pi
M
qj

A
pq
, i.e., A = M
T

AM.
Continuum Mechanics spring 2012 25
Summary: matrices, vectors and tensors
Matrices are ordered, two-dimensional arrays of numbers. Vectors are quantities with magnitude and
direction. Tensors are linear combinations of dyadic products of vectors.
Vectors have three components that transform under rotations of coordinate axes as
a
i
= M
ij
a
j
, M
ij
= e
i
e
j
where e
i
and e
i
are two dierent orthonormal (e
i
e
j
=
ij
) bases, a = a
i
e
i
= a
j
e
j
. Tensors of
order n have 3
n
components, which transform as

A
ijm
= M
ip
M
jq
M
ms
A
pqs
.
The components of a tensor (or vector) can be collected to a matrix, but the matrix species the tensor
only in combination with the base used.
Notation (unless otherwise stated):
column matrices: bold-face roman low-case letters (b); exercises: e.g., Ib = (b
i1
)
square matrices: bold-face roman capital letters (B); exercises: e.g., IB = (B
ij
)
vectors: bold-face italic low-case letters (b); exercises: e.g., b = b
i
e
i
tensors of order 2 or higher: bold-face italic capital letters (B); exercises: e.g., B = B
ij
e
i
e
j
scalars, matrix and tensor components: greek or italic letters (, b
i
, B
ij
); exercises: e.g., B
ij
In exercises, explain your notation in any case and avoid ambiguity.
Continuum Mechanics spring 2012 26
Symmetric and antisymmetric tensors. Transpose of a tensor
Suppose that A = A
ij
e
i
e
j
is a second-order tensor with A
ij
= A
ji
in some base e
i
. Thus, in
another base e
i

A
pq
= M
pi
M
qj
A
ij
= M
qj
M
pi
A
ji
=

A
qp
,
so the property of symmetry under interhange of component indices is independent of the coordinate
system. It is, thus, a property of tensor A.
A tensor A with the property A
ij
= A
ji
of the components (in any coordinate system) is called a
symmetric second-order tensor.
Similarly, if A
ij
= A
ji
, then

A
ij
=

A
ji
and A is an antisymmetric second-order tensor.
Denote A
T
ij
= A
ji
and

A
T
ij
=

A
ji
. Thus

A
T
ij
=

A
ji
= M
jr
M
is
A
rs
= M
is
M
jr
A
T
sr
so the components A
ji
also transform as those of a tensor. Thus, from the tensor A = A
ij
e
i
e
j
we can form another tensor
A
T
= A
ji
e
i
e
j
called the transpose of A.
The tensor A+A
T
is symmetric and AA
T
antisymmetric. Thus, any tensor can be decomposed
into a symmetric and an antisymmetric part:
A =
1
2
(A+A
T
) +
1
2
(AA
T
).
Continuum Mechanics spring 2012 27
Unit tensor. Isotropic tensors
The tensor I =
ij
e
i
e
j
is called the unit tensor (yksikkotensori). In terms of another base e
i
,
I =
ij
M
ri
M
sj
e
r
e
s
= M
ri
M
si
e
r
e
s
=
rs
e
r
e
s
=
ij
e
i
e
j
.
Thus, I has the property that its components are
ij
in any coordinate system.
Denition. A tensor, whose components are the same in any coordinate system is called an isotropic
tensor.
Consider
M =
_
_
0 1 0
1 0 0
0 0 1
_
_
M
T
=
_
_
0 1 0
1 0 0
0 0 1
_
_


A = MAM
T
=
_
_
A
22
A
21
A
23
A
12
A
11
A
13
A
32
A
31
A
33
_
_
A isotropic
=
_
_
A
11
A
12
A
13
A
21
A
22
A
23
A
31
A
32
A
33
_
_
Thus, A
22
= A
11
. Also, A
32
= A
31
, A
31
= A
32
giving A
32
= A
31
= 0. Similarly,
A
23
= A
13
= 0. Another choice for Mcan be used to prove that also A
22
= A
33
, A
12
= A
21
= 0.
Thus, A
ij
= A
ij
, where A is a scalar; the only isotropic 2nd-order tensors are scalar multiples of I.
The only isotropic tensors of order three (for det M = +1) are multiples of the alternating tensor
e
ijk
e
i
e
j
e
k
and the only isotropic tensors of order four are of form (, and are scalars)
(
ij

kl
+
ik

jl
+
il

jk
)e
i
e
j
e
k
e
l
Continuum Mechanics spring 2012 28
Outer product. Contraction
Ass. a = a
i
e
i
and B = B
ij
e
i
e
j
Consider the tensor C = C
ijk
e
i
e
j
e
k
, where C
ijk
= a
i
B
jk
. Thus,
C = C
ijk
(M
ri
e
r
) (M
sj
e
s
) (M
tk
e
t
) = M
ir
M
js
M
kt
C
rst
e
i
e
j
e
k
so

C
ijk
= M
ir
M
js
M
kt
C
rst
= M
ir
M
js
M
kt
a
r
B
st
= a
i

B
jk
, i.e., the components of C are
related to the components of a and B in the same way in any coordinate system. The tensor C is the
outer product (ulkotulo) of a and B (in that order), and it is denoted by a B.
Outer product of two second order tensors, A and B, gives a fourth order tensor D = A B with
components D
ijkl
= A
ij
B
kl
(in any coordinate system). Generalization to higher orders is obvious.
The dyadic product of two vectors is the outer product of those vectors (i.e., two 1st-order tensors).
Consider a third-order tensor C = C
ijk
e
i
e
j
e
k
, for which

C
ijk
= M
ir
M
js
M
kt
C
rst
. Thus,

C
i11
+

C
i22
+

C
i33
=

C
ijj
= M
ir
M
js
M
jt
C
rst
= M
ir

st
C
rst
= M
ir
C
rss
,
i.e., C
ijj
transforms as a vector. More generally, if D
ijpqrs
are the components of a tensor of
order n, summing over any pair of indices to form, e.g., D
ijpprs
, gives a set of components which
transform as a tensor of order n 2. This operation of reducing the order of tensors by 2 is called
contraction (kontraktio).
Continuum Mechanics spring 2012 29
Inner product
Ass. a = a
i
e
i
, B = B
ij
e
i
e
j
, and C = C
ijk
e
i
e
j
e
k
, where C
ijk
= a
i
B
jk
.
The contractions C
iij
= a
i
B
ij
and C
iji
= B
ji
a
i
give the components of vectors a
i
B
ij
e
j
and
B
ji
a
i
e
j
denoted by a B and B a, respectively, and called inner products (sisatulot) of a and B.
Note: a B = B a only if B
ij
= B
ji
, i.e., if B is a symmetric tensor.
Inner products of tensors reduce to inner products of unit dyads, e.g.,
a B = a
i
e
i
(B
kl
e
k
e
l
) = a
i
B
kl
(e
i
e
k
e
l
) = a
i
B
kl

ik
e
l
= a
i
B
il
e
l
Inner products of higher order tensors dened a similar manner, e.g., if A = A
ij
e
i
e
j
and B =
B
ij
e
i
e
j
A B = A
ij
B
kl
e
i
e
j
e
k
e
l
= A
ij
B
kl

jk
e
i
e
l
= A
ik
B
kl
e
i
e
l
.
Thus, e.g.,
(A B)
T
= (A
ik
B
kl
e
i
e
l
)
T
= A
ik
B
kl
(e
i
e
l
)
T
= A
ik
B
kl
e
l
e
i
= A
ij
B
kl

jk
e
l
e
i
= A
ij
B
kl
e
l
e
k
e
j
e
i
= B
T
A
T
The rules of the inner products of tensors conrm to the rules of matrix multiplication, e.g., if D =
A B, the matrix of the components of D in a given base is D = AB, where A and B are the
matrices of the components of A and B in the same base.
Continuum Mechanics spring 2012 30
Inverse tensor. Orthogonal tensors. Polar decomposition
Denition. If A
1
such that
A A
1
= A
1
A = I
then A
1
is called the inverse tensor (kaanteistensori) to A.
The matrix of coecients of A
1
is A
1
, where A is the matrix of coecients of A(in a given base).
Thus, a necessary condition for the existence of A is that det A ,= 0.
Denition. If the tensors A
1
and A
T
are equal so that A A
T
= A
T
A = I then A is said
to be an orthogonal tensor.
By using the polar decomposition theorem for matrices, we can write for the matrix F of the components
of a second-order tensor F
F = RU, F = VR,
where R = (R
ij
) is an orthogonal matrix and U = (U
ij
) and V = (V
ij
) are positive denite
symmetric matrices. Dene the three tensors R = R
ij
e
i
e
j
, U = U
ij
e
i
e
j
, and V =
V
ij
e
i
e
j
. Thus, R is an orthogonal tensor, U and V are symmetric tensors, and
R U = R
ik
U
kj
e
i
e
j
= F
ij
e
i
e
j
= F
and similarly V R = F. Thus, any tensor can be decomposed (uniquely) into either inner products
F = R U = V R.
Continuum Mechanics spring 2012 31
Eigenvalues of a tensor. Principal values. Eigenvector of a tensor
Consider a tensor Awith components A
ij
in the base e
i
and components

A
ij
in the base e
i
= M
ij
e
j
.
Let A = (A
ij
),

A = (

A
ij
), and M = (M
ij
).
Suppose that is an eigenvalue of

A, i.e., det(

AI) = 0. Since

A = MAM
T
and I = MIM
T
,
0 = detM(A I)M
T
] = det M
.
=1
det(A I) det M = det(A I).
Thus, is also an eigenvalue of A.
Eigenvalues are indenpendent of the coordinate system, i.e., they are intrinsic to the tensor A.
If tensor Ais symmetric then its eigenvalues are real numbers. We then call them principal components
(paakomponentit) or principal values (paaarvot) of the tensor A and denote them by A
1
, A
2
and A
3
.
If the principal values are all positive then A is a positive denite tensor.
Suppose that Ais symmetric, and that A
1
, A
2
and A
3
are all distinct. Thus, the eigenvectors x
(i)
are
unique and mutually orthogonal, and can be normalised as x
(i)
T
x
(i)
= 1. Now
Ax
(i)
= A
i
x
(i)
; (i = 1, 2, 3; no summation)
M(=)
MAM
T
Mx
(i)
= A
i
Mx
(i)

A(Mx
(i)
) = A
i
(Mx
(i)
)
so x
(i)
= Mx
(i)
are the eigenvectors of

A (related to A
i
). Thus, x
i
= x
(i)
j
e
j
are vectors. We call
them the eigenvectors of A
A x
i
= A
i
x
i
(no summation).
Continuum Mechanics spring 2012 32
Principal axes
Recall from rst lecture: for any symmetric matrix A with eigenvalues A
i
and eigenvectors x
(i)
there
exists an orthogonal matrix P, dened by P
T
= (x
(1)
x
(2)
x
(3)
), for which

A PAP
T
=
_
_
A
1
0 0
0 A
2
0
0 0 A
3
_
_
diag(A
i
).
Thus, there exists a coordinate system, in which the matrix of the components of a symmetric tensor
A is diagonal, and the diagonal elements are the principal values of the tensor. The components of the
eigenvectors in this system (with base vectors e
i
= P
ij
e
j
) are
x
(i)
= Px
(i)
= (
1i

2i

3i
)
T
, i.e., x
(i)
j
=
ji
.
Thus, x
i
= x
(i)
j
e
j
=
ji
e
j
= e
i
. Thus, choosing the eigenvectors of any symmetric tensor (with
distinct principal values) A as the base yields the matrix of the coecients in form A = diag(A
i
).
The axes of this coordinate system are called the principal axes (paaakselit) of A.
The result remains valid also if A
i
are not distinct. For A
1
= A
2
,= A
3
, x
3
is uniquely determined
and x
1
and x
2
are arbitrary perpendicular unit vectors in the plane perpendicular to x
3
. If all principal
values are equal, the tensor A is isotropic and any orthonormal base suces as the set of eigenvectors.
The principal values of A
n
= A ... A
.
n factors
are A
n
i
and the principal axes of A
n
are those of A.
Continuum Mechanics spring 2012 33
Tensor invariants. Trace and determinant of a tensor
Principal values of Aare independent of the coordinate system; they are invariants of the tensor (tensori-
invariantit). If A is symmetric, then A
i
are the basic invariants of the tensor, i.e., all invariants of A
can be expressed in terms of them. Thus, the number of independent tensor invariants is three.
In many applications, it is more convenient to choose the invariants as independent symmetric functions
of A
i
, e.g.,
A
1
+A
2
+A
3
, A
2
1
+A
2
2
+A
2
3
, A
3
1
+A
3
2
+A
3
3
.
The above set turns out to be convenient, since it can be calculated without actually determining the
principle values of the tensor. Since

A PAP
T
= diag(A
i
), we have A
1
+ A
2
+ A
3
= tr

A.
But
tr

A =

A
ii
= P
ir
P
is
A
rs
=
rs
A
rs
= A
rr
= tr A.
Thus, we can dene tr A = tr Awithout ambiguity. As tr

A
n
= A
n
1
+A
n
2
+A
n
3
and

A
n
is a matrix
representation of A
n
, we get A
n
1
+A
n
2
+A
n
3
= tr A
n
so the set of invariants can be represented as
tr A, tr A
2
, tr A
3
]
Another set of invariants frequently used is I
1
, I
2
, I
3
], where
I
1
= A
1
+A
2
+A
3
, I
2
= A
2
A
3
+A
3
A
1
+A
1
A
2
, I
3
= A
1
A
2
A
3
.
Here, I
1
= tr A and I
2
=
1
2
(A
1
+A
2
+A
3
)
2
(A
2
1
+A
2
2
+A
2
3
)] =
1
2
(tr A)
2
tr A
2
].
Moreover, A
1
A
2
A
3
= det

A = det(PAP
T
) = det Pdet Adet P = det A, so we can dene
det A = det A without ambiguity and write I
3
= det A.
Continuum Mechanics spring 2012 34
Deviatoric tensors
The rst invariant of the tensor A
t
= A
1
3
Itr A, is tr A
t
= 0. Thus, if A
t
is symmetric, it has
only ve independent components and only two non-zero independent invariants.
Denition. A tensor with vanishing trace is called a deviatoric tensor (deviatorinen t.) and the A
t
is
called the deviator of A (A:n deviaattori).
Obviously, any tensor can be decomposed into the sum of a deviatoric and an isotropic tensor
A = A
t
+
1
3
Itr A.
The two non-zero invariants of A
t
are
I
t
2
=
1
2
tr A
t2
and I
t
3
=
1
3
tr A
t3
,
where the latter is a consequence of the

CayleyHamilton theorem:

CH: 0 = A
3
I
1
A
2
+I
2
A I
3
I
0 = tr (A
3
I
1
A
2
+I
2
A I
3
I) = tr A
3
I
1
tr A
2
+I
2
tr A I
3
tr I
.
=3
I
3
=
1
3
tr A
3
I
1
tr A
2
+I
2
tr A] =
1
3
tr A
3

3
2
tr Atr A
2
+
1
2
(tr A)
3
]
Continuum Mechanics spring 2012 35
Basic vector and tensor calculus
Ass. (x
1
, x
2
, x
3
), a(x
1
, x
2
, x
3
) and A(x
1
, x
2
, x
3
) are a scalar, a vector, and a tensor eld.
The gradient (gradientti) of is
grad = =

x
i
e
i
e
i

i
.
is a vector, whose direction is normal to a level surface (x
1
, x
2
, x
3
) = const. (pointing towards
increasing ) and whose magnitude is the directional derivative of in the direction of this normal.
The divergence (divergenssi) of a is
div a = a =
i
a
i
and it is a scalar. Likewise, the divergence of A is A = e
j

i
A
ij
. It is a vector.
The curl (roottori) of a is
curl a = a = e
ijk
e
i

j
a
k
.
It is a vector.
Divergence theorem:
___
1
adV =
__
S
n adS
where 1 and S are a region of space and its bounding surface, and n is a unit vector normal to S
pointing outwards. For tensors A(x
1
, x
2
, x
3
) the equation reads
___
1

i
A
ij
dV =
__
S
n
i
A
ij
dS.
Continuum Mechanics spring 2012 36
2 Particle kinematics
bodies and their motions
material and spatial description
displacement, velocity and acceleration
particle paths and streamlines
Continuum Mechanics spring 2012 37
Particle kinematics
Kinematics = study of motion without regard to the forces that produce it.
Particle kinematics = study of motion of individual particles (although a part of a larger body) without
regard to the motion of neighboring particles.
Studies of deformations, i.e., changes of shape, follow later.
Ass. (O, e
i
) is a xed Cartesian coordinate system with origin O and base vectors e
i
.
Throughout the remainder of the theory part, all motion will be motion relative to this xed frame and,
unless otherwise stated, all vector and tensor components are components relative to this base.
Time is measured from a xed reference time, t = 0.
Continuum Mechanics spring 2012 38
Bodies
0
R
P
0
t = 0
X
O
B
Ass. at t = 0, a xed region of space, 1
0
, is occupied by
continuously distributed matter. (Each point of 1
0
is occupied by
a matter particle.)
The material within 1
0
forms a body, denoted by B.
Let X be the position vector of point P
0
1
0
relative to O
the components X
R
of X (R = 1, 2, 3) in
X = X
R
e
R
are the coordinates of the position occupied by a particle of B at
t = 0 in the chosen coordinate system (O, e
i
).
Notation: now, inconsistently, we denote a vector by a capital letter.
Each point of the region 1
0
corresponds to a (material) particle of the body B and the body is a
collection of all such particles.
Continuum Mechanics spring 2012 39
Motions of bodies
0
R
P
0
t = 0 0 t >
X
O
B
P
B
R
x
Ass. Material that occupies 1
0
at time t = 0 occu-
pies another continuous region 1at time t. Material
of B now continuously distributed in 1.
This is termed the motion of the body B.
Ass. individual particles of B can be identied:
the point P 1 at the time t, occupied by a
particle that was at P
0
1
0
at t = 0, can be
identied.
Denote the position vector of P relative to O with
x. Thus, the motion can be specied by equations
of form
x = x(X, t) X 1
0
, x 1
or in component form
x
i
= x
i
(X
1
, X
2
, X
3
, t) (i = 1, 2, 3)
:= x
i
(X
R
, t) (i, R = 1, 2, 3)
Note: the italic R is an index variable, whereas the script-style 1 denotes a region of space.
Continuum Mechanics spring 2012 40
Material and spatial coordinates and descriptions. Congurations.
A given particle of the body B can be identied by its coordinates X
R
at time t = 0.
A particular particle retains the same values of X
R
during its motion.
Coordinates x
i
identify points in space, which are occupied by dierent particles at dierent times.
Denition. The coordinates X
R
are termed material coordinates and the coordinates x
i
are termed
spatial coordinates.
Denition. The set of positions of the particles of the body B at a given time specify a conguration
of B. The conguration of B at time t = 0 is called the reference conguration and the conguration
of B at time t is its current conguration.
Note: usually we will denote scalar, vector and tensor quantities of the reference conguration by
capital letters and corresponding quantites evaluated in the current conguration by lower-case letters.
Components of vectors and tensors which transform with the coordinates X
R
will have capital letter
indices (A
R
, C
RS
, etc.) and components wich transform with the coordinates x
i
will have lower-case
indices (a
i
, T
ij
, etc.).
Eqs. describing the motion, x
i
= x
i
(X
R
, t), can be inverted for physically realizable motions.
X = X(x, t) or X
R
= X
R
(x
i
, t) (R, i = 1, 2, 3)
Problems of CM can be formulated either with X
R
as the independent variables (= material or
Lagrangian description) or with x
i
as the independent variables (= spatial or Eulerian description). De-
pending on the nature of the problem, either formulation may be more suitable.
Continuum Mechanics spring 2012 41
Displacement and velocity
0
R
P
0
t = 0 0 t >
X
O
B
P
B
R
x
u
Displacement vector of a particle from its position
at X in the reference conguration to its position
x in the current conguration is
u = x X
In the material description
u(X, t) = x(X, t) X (1)
and in the spatial description
u(x, t) = x X(x, t) (2)
Eq. (1) gives the displacement of the particles labeled by X
R
and Eq. (2) the displacement of a particle
that currently occupies the position x.
The velocity vector v of a particle is the rate of change of its displacement. Thus,
v(X, t) =
u(X, t)
t
=
x(X, t)
t
or v
i
(X
R
, t) =
x
i
(X
R
, t)
t
gives the velocity of the particle labeled by X
R
in the reference conguration.
Continuum Mechanics spring 2012 42
The velocity in the spatial description is given by
v(x, t) =
_
x(X, t)
t
_
X=X(x,t)
.
This is the velocity of a particle currently occupying the position x.
Example. A body undergoes the motion dened by
x
1
= X
1
(1 +a
2
t
2
), x
2
= X
2
, x
3
= X
3
.
Thus,
u
1
= x
1
X
1
= X
1
a
2
t
2
, u
2
= x
2
X
2
= 0, u
3
= 0
is the displacement in the material description. As X
1
(x
i
, t) = x
1
/(1 +a
2
t
2
), we get
u
1
=
x
1
a
2
t
2
1 +a
2
t
2
, u
2
= 0, u
3
= 0
in the spatial description. For velocity, v
2
= v
3
= 0 in both descriptions, and
v
1
(X
R
, t) =
u
1
(X
R
, t)
t
= 2X
1
a
2
t v
1
(x
i
, t) = 2X
1
(x
i
, t) a
2
t =
2x
1
a
2
t
1 +a
2
t
2
.
Thus, individual particles (labeled by X) start from rest with constant acceleration 2X
1
a
2
in the e
1
direction, but the instantaneous velocity at x does not increase linearly with time. Note also, that
v(x, t) ,= u(x, t)/t.
Continuum Mechanics spring 2012 43
Time rates of change
Ass. is a quantity that varies throughout a body in space and time.
We can regard as a function of time and either the material or the spatial coordinates:
= G(X
R
, t) = g(x
i
, t)
Problem: how does vary with time following a given particle? In material description:
D
Dt
=

=
G(X
R
, t)
t
D/Dt is called the material derivative or the convective derivative of . What about spatial descrip-
tion?
= gx
i
(X
R
, t), t] = gx
1
(X
R
, t), x
2
(X
R
, t), x
3
(X
R
, t), t]

D
Dt
=
g(x
i
, t)
x
j
x
j
(X
R
, t)
t
+
g(x
i
, t)
t
=
g(x
i
, t)
x
j
v
j
+
g(x
i
, t)
t
= v g(x, t) +
g(x, t)
t
.
Usually the notation is simplied by denoting both G and g by , i.e.,
D
Dt
=
(X, t)
t
or
D
Dt
= v (x, t) +
(x, t)
t
Continuum Mechanics spring 2012 44
Acceleration
The acceleration f of a particle is the rate of change of velocity of that particle. In the material
description
f
i
=
Dv
i
Dt
=
v
i
(X
R
, t)
t
=

2
x
i
(X
R
, t)
t
2
or f = v(X, t) = x(X, t).
In the spatial description
f
i
=
v
i
(x
j
, t)
t
+v
k
v
i
(x
j
, t)
x
k
or f =
v(x, t)
t
+v v(x, t).
Example. Consider
x
1
= X
1
(1 +a
2
t
2
), x
2
= X
2
, x
3
= X
3
v
1
=
x
1
(X
R
, t)
t
= 2X
1
a
2
t =
2x
1
a
2
t
1 +a
2
t
2
, v
2
= v
3
= 0.
Thus, f
2
= f
3
= 0 and
f
1
(X
1
, t) =
v
1
(X
1
, t)
t
= 2X
1
a
2
f
1
(x
1
, t) =
v
1
(x
1
, t)
t
+v
1
v
1
(x
1
, t)
x
1
= . . . =
2x
1
a
2
1 +a
2
t
2
= f
1
(X
1
(x
1
, t), t)
Continuum Mechanics spring 2012 45
Steady motion
A motion is said to be steady, if the velocity at any point in space is independent of time, i.e.,
v(x, t)/t = 0 or v = v(x).
Motion may be unsteady wrt. the xed coordinate system but steady relative to suitably chosen moving
coordinate axes.
Example: supersonic jet airplane moving with constant velocity V . (Turbulence neglected!)
boundary
in motion
plane
in motion
plane
at rest
steady
motion
air in
steady
motion
air in
V
air in motion
V
rest
air at
boundary
at rest
Continuum Mechanics spring 2012 46
Particle paths and streamlines
Eqs. x
i
= x
i
(X
R
, t) give the successive positions x
i
of the particle X
R
with t serving as parameter.
Thus, they are parametric equations of the particle path. In dierential form
dx
i
= v
i
(X
R
, t) dt
and this gives
dx
i
= v
i
(X
R
(x
j
, t), t) dt = v
i
(x
j
, t) dt,
which can be used to obtain the particle paths if the velocity eld is known.
The streamlines at time t are space curves, whose tangents are everywhere directed along the velocity
vector. Their eqs. can be written in terms of a parameter as
dx
i
= v
i
(x
j
, t) d or
dx
1
v
1
(x
i
, t)
=
dx
2
v
2
(x
i
, t)
=
dx
3
v
3
(x
i
, t)
,
where now t is now held constant during the integration of the equations.
In general, particle paths and streamlines do not coincide, but if the motion is steady, they
represent the same families of curves. (But note that even if the motion is unsteady, the streamlines and
particle paths may still coincide, see the example below.)
Continuum Mechanics spring 2012 47
Problem: Determine the particle paths and streamlines of the velocity eld
v
1
=
ax
1
1 +at
, v
2
=
2ax
2
1 +at
, v
3
=
3ax
3
1 +at
.
Solution: We have for the particle paths x
k
(X
R
, t) (below k = 1, 2, 3; no summation)
dx
k
= v
k
(x
i
, t) dt = k
ax
k
1 +at
dt
dx
k
x
k
= k
adt
1 +at
ln x
k
= k ln(1 +at) +C
k
,
where the constants C
k
are determined by x
k
(X
R
, 0) = X
k
. Thus, C
k
= ln X
k
and
x
k
= X
k
(1 +at)
k
(k = 1, 2, 3; no summation)
Eliminating time, the curves can be given in form
x
k
/X
k
= (1 +at)
k
= (x
1
/X
1
)
k
(k = 2, 3; no summation)
For the streamlines, dx
i
= v
i
(x
i
, t) d, i.e.,
(1 +at)
dx
1
ax
1
= (1 +at)
dx
2
2ax
2
= (1 +at)
dx
3
3ax
3

dx
1
x
1
=
dx
2
2x
2
=
dx
3
3x
3

x
k
x
k
=
_
x
1
x
1
_
k
(k = 2, 3; no summation),
where the integration has been started from x = x
i
e
i
. Thus, although the motion is not steady, the
particle paths and streamlines coincide.
Continuum Mechanics spring 2012 48
3 Analysis of stress
traction and components of stress
Cauchys law
stress tensor
equations of equilibrium
boundary conditions of equilibrium
principal components and axes of stress
stress invariants
normal and shear stress
Continuum Mechanics spring 2012 49
Surface force. Surface traction
Consider, next, forces acting inside a continuous body. Suppose that a part of the body B occupies a
region 1 which has the surface S.
S
t
( ) n
S
R
n
B
P
O
x
Ass. Point P S, S is a small element of S
around P, and n S at P with n
i
n
i
= 1.
Ass. Material outside 1 exerts a force
p = t
(n)
S
on the material inside the surface through the ele-
ment S.
p is called the surface force (pintavoima) and t
(n)
the mean surface traction (jannitysvektori, traktio)
across dS from the outside to the inside of 1.
A force p and a traction t
(n)
are transmitted across dS from the inside to the outside of 1.
Clearly, p depends on the position of P (or its position vector x) and the orientation of n, but as
S 0, t
(n)
is assumed to tend to a limit independent of the shape of S. Henceforth, denote
t
(n)
(x) = lim
S0
p
S
and call t
(n)
the (surface) traction at the point P on the surface with normal n. In general, n t
(n)
.
Continuum Mechanics spring 2012 50
Components of stress
At P, there is a traction t
(n)
associated with each direction n
through P. In particular, there is a traction associated with all base
vectors e
i
of a given coordinate system. Denote these by
t
i
:= t
(e
i
)
.
Thus, e.g., t
1
is the force per unit area exerted on the negative side
of a surface normal to the x
1
axis by the material on the positive
side of this surface.
Resolve the vectors t
i
in components:
t
i
= T
ij
e
j
(i = 1, 2, 3; summation over j).
t
i
e
k
= T
ij
e
j
e
k
= T
ij

jk
= T
ik
T
ij
= t
i
e
j
Quantities T
ij
are the stress components (jannityskomponentit).
e
1
t
1
x
1
e
1
e
2
T
11
T
12
T
13
t
1
T
31
T
21
t
2
T
22
T
23
T
32
T
33
e
3
t
3
If T
11
> 0, material on the positive side of a surface with n = e
1
is pulling the material on the
negative side of the surface and vice versa (tension), and if T
11
< 0, material on the positive side is
pushing the material on the negative side and vice versa (compression).
Stress components T
ij
with i = j are called the normal (or direct) stress components (normaalijan-
nitykset) and those with i ,= j are called shearing stress components (leikkausjannitykset).
Continuum Mechanics spring 2012 51
Traction on any surface. Cauchys law
Q
2
Q
3
Q
1
e
3

t
3
S
3
t
1
S
1
e
1

t
( )
S
n
t
2
S
2
e
2

e
3
e
1
e
2
n
P
Consider forces acting on the tetrahedron
PQ
1
Q
2
Q
3
B.
The unit normals of, tractions on, areas of and surface
forces on the four faces are
face normal traction area force
PQ
2
Q
3
e
1
t
1
S
1
t
1
S
1
PQ
1
Q
3
e
2
t
2
S
2
t
2
S
2
PQ
1
Q
2
e
3
t
3
S
3
t
3
S
3
Q
1
Q
2
Q
3
n t
(n)
S t
(n)
S
Areas S
i
are projections of S to the coordinate planes.
S
i
= n e
i
S = n
i
S.
A body force (tilavuusvoima) may also be present, with mean value b / unit mass inside the tetrahedron.
Ass. For any part of a body, the rate of change of momentum is proportional to the resultant force
acting. This is known as Cauchys law of motion, generalization of NII. For PQ
1
Q
2
Q
3
, in particular
t
1
S
1
t
2
S
2
t
3
S
3
+t
(n)
S +b V = f V
t
(n)
= n
i
t
i
+
V
S
(f b)
ass. S0
n
i
t
i
= n
i
T
ij
e
j
t
(n)
j
= n
i
T
ij
components of t
(n)
Continuum Mechanics spring 2012 52
Transformation of stress components. Stress tensor
Traction t
(n)
is given by
t
(n)
= n
i
t
i
= n
i
T
ij
e
j
in the base e
i
. In particular, choosing n = e
i
, we have t
i
= T
ij
e
j
.
The same results must hold in another base, i.e., if n = e
p
= M
pi
e
i
, then n
i
= M
pi
and

t
p
:= t
( e
p
)
= M
pi
T
ij
e
j
= M
pi
T
ij
M
qj
e
q
= M
pi
M
qj
T
ij
e
q
But we should dene the stress components in the base e
q
analogously:

t
p
=

T
pq
e
q
.
Thus,

T
pq
= M
pi
M
qj
T
ij
i.e.,

T = MTM
T
with T = (T
ij
)
and the stress components form a second-order tensor,
T = T
ij
e
i
e
j
called Cauchy stress tensor. It completely describes the state of stress of a body. Hereafter, we will
call it simply the stress tensor (jannitystensori ).
Since n is a vector and T is a tensor, we may write the equation t
(n)
j
= n
i
T
ij
in form
t
(n)
= n T
Continuum Mechanics spring 2012 53
Example
Problem. The state of tress throughout a continuum is given wrt. the Cartesian coordinate system
(O, e
i
) by the matrix
T =
_
_
3x
1
x
2
4x
2
2
0
4x
2
2
0 2x
3
x
2
0 2x
3
x
2
0
_
_
Determine the normal stress component on the cylindrical surface := x
2
2
+x
2
3
R
2
= 0.
Solution. First, n j = 2x
2
e
2
+ 2x
3
e
3
. Normalize magnitude to unity:
n =
2x
2
e
2
+ 2x
3
e
3
(4x
2
2
+ 4x
2
3
)
1/2
=
x
2
e
2
+x
3
e
3
(x
2
2
+x
2
3
)
1/2
=
x
2
e
2
+x
3
e
3
R
.
Thus,
t
(n)
= n T =
x
2
e
2
+x
3
e
3
R

(3x
1
x
2
e
1
e
1
+ 4x
2
2
e
1
e
2
+ 4x
2
2
e
2
e
1
+ 2x
3
x
2
e
2
e
3
+ 2x
3
x
2
e
3
e
2
)
= R
1
[4x
3
2
e
1
+ 2x
3
x
2
2
e
3
+ 2x
2
3
x
2
e
2
]
T
nn
= n t
(n)
= R
2
(2x
2
3
x
2
2
+ 2x
2
3
x
2
2
) = 4R
2
x
2
3
x
2
2
= 4R
2
sin
2
cos
2
= R
2
sin
2
2
where x
2
= Rcos and x
3
= Rsin have been used. Thus, the body is in tension (T
nn
0).
Continuum Mechanics spring 2012 54
Equations of equilibrium
Consider, next, the body B in equilibrium.
S
t
( ) n
S
R
n
B
P
O
x
Ass. 1 is an arbitrary region of B and S is its
surface with unit normal n.
Ass. Two kinds of forces act on the material: surface
forces acting across S and body forces.
Ass. In equilibrium, the resultant force and the resul-
tant torque about O acting on the material is zero:
__
S
t
(n)
dS +
___
1
b dV = 0
__
S
x t
(n)
dS +
___
1
x b dV = 0.
In terms of components
0 =
__
S
n
i
T
ij
dS +
___
1
b
j
dV =
___
1
_
T
ij
x
i
+b
j
_
dV
0 =
__
S
e
jpq
x
p
n
r
T
rq
dS +
___
1
e
jpq
x
p
b
q
dV =
___
1
e
jpq
_
(x
p
T
rq
)
x
r
+x
p
b
q
_
dV,
where the divergence theorem,
__
n
i
T
ij
dS =
___

i
T
ij
dV , has been used.
Continuum Mechanics spring 2012 55
But since 1 is arbitrary
T
ij
x
i
+b
j
= 0, e
jpq
_
(x
p
T
rq
)
x
r
+x
p
b
q
_
= 0.
However, as x
p
/x
r
=
pr
, we get
0 = e
jpq
_
x
p
T
rq
x
r
+
pr
T
rq
+x
p
b
q
_
= e
jpq
x
p
_
T
rq
x
r
+b
q
_
+e
jpq
T
pq
= e
jpq
T
pq
Thus,
j = 1 : 0 = e
123
=+1
T
23
+e
132
=1
T
32
= T
23
T
32
j = 2 : 0 = e
213
=1
T
13
+e
231
=+1
T
31
= T
31
T
13
j = 3 : 0 = e
312
=+1
T
12
+e
321
=1
T
21
= T
12
T
21
and, therefore, T
ij
= T
ji
. Thus, in equilibrium, the stress tensor is symmetric and
T +b = 0.
We will later prove that the symmetry of T prevails also for bodies in motion. Thus, T will be treated
as a symmetric tensor hereafter.
Continuum Mechanics spring 2012 56
Boundary conditions in equilibrium
Consider a nite body B, occupying the region 1, in equilibrium.
Let S be the surface of the area 1, and n
S
be the outward unit vector and dS an element of that
surface. The traction at the surface is, thus,
t
(n
S
)
= n
S
T
and the surface force caused by this traction on the element dS is t
(n
S
)
dS. In equilibrium, this has to
be balanced by an external surface force, pdS. Thus, the boundary condition at the surface is
pdS = t
(n
S
)
dS p = n
S
T.
In particular, if the body is in vacuum, we must require
n
S
T = 0
at its surface. On the other hand, if the body B
1
is immersed in a medium constituting another body,
B
2
, the boundary condition at the separating surface is
n
S
T
(1)
= n
S
T
(2)
,
where T
(1)
and T
(2)
are the stress tensors measured inside the bodies 1 and 2, respectively.
Continuum Mechanics spring 2012 57
Example: the law of Archimedes
Consider a body B
1
with constant density
1
in a homogeneous gravitational eld g = ge
3
immersed
in an incompressible uid (body B
2
; density
2
) with
T
(2)
= p(x) I,
where p(x) is the hydrostatic pressure of the uid. The equilibrium
equation for the uid reads

2
ge
3
= T
(2)
= p

3
p =
2
g;
j
p = 0, j = 1, 2
p = p
0

2
gx
3
,
where p
0
is constant. Thus, surface traction exerted by the uid on the
surface of B
1
reads
n T
(2)
= (p
0

2
gx
3
)n.

1
g
2

2
1
B
B
Assuming that also T
(1)
is isotropic, T
(1)
= P(x)I, and the interior of B
1
to be in equilibrium,
P = P
0

1
gx
3
and n T
(1)
= (P
0

1
gx
3
)n. Thus, for an equilibrium to exist requires
P
0

1
gx
3
= p
0

2
gx
3
x S
1
and this is possible only if
1
=
2
and P
0
= p
0
.
Continuum Mechanics spring 2012 58
If
2
,=
1
, there is a non-vanishing resultant force F acting on the body B
1
F =
__
S
1
n T
(2)
dS +
___
1
1

1
g dV =
___
1
1
(

T
(2)
+
1
g)dV
=
___
1
1
( p +g
1
e
3
)dV =
___
1
1
g(
2

1
)e
3
dV = gV
1
(
2

1
)e
3
Note: the tensor eld T
(2)
is dened only in the region outside the boundary of the body B
1
. However,
the mathematical evaluation of the surface integral can be done by continuing the eld to the region
inside the body B
1
, i.e., by taking the derived form p(x) = p
0
g
2
x
3
to apply also inside the region.
This is denoted by adding the tilde in

T
(2)
= pI
p(x) = p
0
g
2
x
3
(x
1
, x
2
, x
3
) 1
1
1
2
.
The eq. for F is the familiar law of Archimedes, giving the buoyancy (noste) as
b = F
1
V
1
g = gV
1
(
2

1
)e
3
(
1
V
1
ge
3
) =
2
V
1
ge
3
.
Note also that the shape of the body B
2
does not aect the buoyancy.
Continuum Mechanics spring 2012 59
Principal stress components. Principal axes of stress
In general, t
(n)
n; the traction has a tangential component to a surface (with normal n) and as well
as a normal component. However, for certain special directions n we may have t
(n)
j n. Then
n T = t
(n)
= Tn
T symm.
= T n = Tn
This clearly occurs, when T is one of the three principal components of T, denoted by T
1
, T
2
, and T
3
,
and n is the unit vector determining the corresponding principal axis. We call the principal components
T
i
as the principal stress components and the principal axes n
i
as the principal axes of stress.
The principal stress components are the roots of the equation
det(T TI) = 0.
Usually, we denote the principal stress components so that T
1
T
2
T
3
.
If the orthogonal principal axes, n
i
, of T are taken as the base vectors, then the matrix of T is diagonal,
T = diag(T
1
, T
2
, T
3
).
Thus,
T = T
1
n
1
n
1
+T
2
n
2
n
2
+T
3
n
3
n
3
.
Note, however, in general the stress components vary from point to point, and the principal components
T
i
and axes n
i
are functions of position.
Continuum Mechanics spring 2012 60
Example 1
Problem. Let the matrix of T in some coordinate system (O, e
i
) and in appropriate units be
T =
_
_
1 2 3
2 4 6
3 6 1
_
_
Find the principal components and pricipal axes of stress.
Solution. The principal stress components are obtained from
0 =

1 T 2 3
2 4 T 6
3 6 1 T

= (1 T)[(4 T)(1 T) 36] 2[2(1 T) 18] + 3[12 3(4 T)]


= (1 T)(32 5T +T
2
) 2(16 2T) + 9T
= T
3
+ 6T
2
+ 40T = T(T
2
6T 40) = T(T 10)(T + 4)
Thus, T
1
= 10, T
2
= 0, T
3
= 4.
Continuum Mechanics spring 2012 61
The corresponding eigenvector components are obtained from Tn
(i)
= T
i
n
(i)
(no summation), i.e.,
_
_
_
_
_
n
(i)
1
+ 2n
(i)
2
+ 3n
(i)
3
= T
i
n
(i)
1
2n
(i)
1
+ 4n
(i)
2
+ 6n
(i)
3
= T
i
n
(i)
2
3n
(i)
1
+ 6n
(i)
2
+ n
(i)
3
= T
i
n
(i)
3
(i = 1, 2, 3; no summation)
i = 1, T
1
= 10 : n
(i)
2
= 2n
(i)
1
3n
(i)
3
= 5n
(i)
1
i = 2, T
2
= 0 : n
(2)
1
= 2n
(2)
2
n
(2)
3
= 0
i = 3, T
3
= 4 : n
(i)
2
= 2n
(i)
1
n
(i)
3
= 3n
(i)
1
Thus, as n
i
= n
(i)
j
e
j
and jn
i
j = 1 i = 1, 2, 3,
n
1
=
e
1
+ 2e
2
+
5
3
e
3
_
1 + 4 +
25
9
=
3e
1
+ 6e
2
+ 5e
3

9 + 36 + 25
=
3e
1
+ 6e
2
+ 5e
3

70
n
2
=
2e
1
+e
2

1 + 4
=
2e
1
+e
2

5
n
3
=
e
1
+ 2e
2
3e
3

1 + 4 + 9
=
e
1
+ 2e
2
3e
3

14
Continuum Mechanics spring 2012 62
Example 2
Ass. The stress tensor T has distinct principal components T
1
, T
2
and T
3
(in descending order).
Problem. Show that T
1
is the maximum and T
3
is the minimum of the normal component of stress,
n T n, for any orientation of n.
Solution. We have
T = n T n = n
i
T
ij
n
j
.
Choosing the pricipal axes as the base vectors,
T = T
1
e
1
e
1
+T
2
e
2
e
2
+T
3
e
3
e
3
,
and, thus,
T(n) = (n e
1
)
2
T
1
+ (n e
2
)
2
T
2
+ (n e
3
)
2
T
3
:= n
2
1
T
1
+n
2
2
T
2
+n
2
3
T
3
where n
i
are the components of n = n
i
e
i
. We, therefore, need to nd the extrema of T(n
i
) subject
to the constraint f(n
i
) := n
i
n
i
1 = 0. This can be done using an undetermined Lagrangian
multiplier :
T
n
i
+
f
n
i
= 0, f(n
i
) = 0
giving T
i
n
i
+n
i
= 0 (no summation) for the rst three equations.
Continuum Mechanics spring 2012 63
Thus, solve
T
i
n
i
+n
i
= 0 (no summation)
n
i
n
i
1 = 0
Multiply the upper one with n
i
and sum over i to get
0 =

i
(n
i
T
i
n
i
+n
i
n
i
) = T +
Thus,
T
i
n
i
= Tn
i
(no summation)
at the three extrema. Thus, either T = T
i
or n
i
= 0, and if the principal stresses are all distinct,
T = T
1
, n
1
= 1, n
2
= n
3
= 0
T = T
2
, n
2
= 1, n
1
= n
3
= 0
T = T
3
, n
3
= 1, n
1
= n
2
= 0
are the only possible solutions for the extremal values of T(n
i
) and the corresponding surface normals
n
i
. Thus, the normal stress component relative to any orientation of the surface normal nvaries between
T
3
and T
1
.
Continuum Mechanics spring 2012 64
Stress invariants. Stress deviator
As T is a symmetric second-order tensor, it has three independent invariants. These are denonted by
J
1
, J
2
and J
3
, called stress invariants, and dened as
J
1
= T
1
+T
2
+T
3
= tr T = T
ii
J
2
= (T
1
T
2
+T
1
T
3
+T
2
T
3
) =
1
2
tr T
2
(tr T)
2
] =
1
2
(T
ij
T
ji
T
ii
T
jj
)
J
3
= T
1
T
2
T
3
= det T.
Note that the sign of J
2
is opposite of the previously introduced tensor invariant I
2
.
Earlier, we considered a simple spherical stress tensor of form,
T = pI,
where p is the hydrostatic pressure of the uid. If the stress tensor is of this form, it is said that the
body is in a pure hydrostatic state of stress. Clearly, p =
1
3
tr T.
Even if T is not isotropic, it is often convenient to decompose in an isotropic and deviatoric part:
T = S +
1
3
(tr T)I = S +
1
3
J
1
I = S pI; p :=
1
3
tr T
The deviatoric (S
ii
= 0) part S is called the stress deviator tensor. It has the components
S
ij
= T
ij

1
3
T
kk

ij
.
Continuum Mechanics spring 2012 65
The principal axes of S are those of T. (This is clear since the principal axes of pI are arbitrary.)
Let S
1
, S
2
and S
3
denote the principal components of S. Thus,
S
1
+S
2
+S
3
= tr S = 0,
and, because
S
i
n
i
= S n
i
= T
1
3
(tr T)I] n
i
= (T
i

1
3
tr T)n
i
(no summation)
for all principal axes n
i
, we have
S
1
=
1
3
(2T
1
T
2
T
3
), S
2
=
1
3
(2T
2
T
1
T
3
), S
3
=
1
3
(2T
3
T
1
T
2
)
Because S is deviatoric, it has only two basic invariants. These are taken to be
J
t
2
= (S
2
S
3
+S
3
S
1
+S
1
S
2
) =
1
2
tr S
2
J
t
3
= S
1
S
2
S
3
= det S =
1
3
tr S
3
It is sometimes convenient to adopt J
1
, J
t
2
and J
t
3
as the basic invariants of T.
Continuum Mechanics spring 2012 66
Shear stress
The normal stress component on a surface normal to the x
1
axis is T
11
. The shear stress is the
resultant of the other two components of traction, i.e.,
T1
2
e 2
T1
3
e 3
+
T
1
1
e
1
T
12
e
2
e
3
t
1
e
1
e
2
T
1
3
e
3
T
12
e
2
+T
13
e
3
,
which has a magnitude of
_
T
2
12
+T
2
13
.
More generally, i.e., relative to the surface normal to n, the
normal stress component is n t
(n)
= n T n. Thus,
the shear stress is
t
(n)
(n t
(n)
)n = n T (n T n)n
= n T (n T) (n n)
= (n T) I n n]
= n
r
T
rs
(
sj
n
s
n
j
)e
j
Let T
1
T
2
T
3
be the principal stress components and n
1
, n
2
and n
3
be the corresponding
principal axis directions. It can be shown that as n is varied, the magnitude of the shear stress attains
a maximum of
1
2
(T
1
T
3
) along a direction which is aligned with
1

2
(n
1
n
3
).
Note:
1
2
(T
1
T
3
) =
1
2
(S
1
S
3
) and that shear stress vanishes on any surface for pure hydrostatic
state of stress.
Continuum Mechanics spring 2012 67
Some simple states of stress
(a) Hydrostatic state of stress
T = pI,
where the hydrostatic pressure is generally a function of position. Leads to the equilibrium
equation
p +b = 0,
where b(x) is the body force per unit mass at position x.
Regarding the body forces to be negligible, the equilibrium equation reads
0 = T, i.e.,
i
T
ij
= 0 j = 1, 2, 3.
As there are six independent stress components and only three scalar equilibrium equations, so theres
no unique solution for the state of stress in equilibrium. Some possible cases fullling the equilibrium
equations for b = 0 are considered below.
Homogeneous stress. Stress with all its components spatially constant. Special cases include
(b) Uniform tension or compression in the x
1
direction is given by
T
11
= , T
22
= T
33
= T
12
= T
13
= T
23
= 0
Example: a uniform cylindrical bar parallel to the x
1
axis with uniform forces applied on the
planar ends of the bar and no forces on the cylindrical surface. If > 0 the bar is in tension and
if < 0 it is in compression. Principal axes: e
1
and two arbitrary orthogonal vectors e
1
.
Continuum Mechanics spring 2012 68
(c) Uniform shear stress in the x
1
direction on planes x
2
= const. arises if
T
21
= , T
11
= T
22
= T
33
= T
13
= T
23
= 0
Example: laminar shear ow of viscous uid with v = v(x
2
)e
1
. Principal axes of stress: e
3
and
1

2
(e
1
e
2
).
Inhomogeneous equilibrium states of stress include
(d) Pure bending. Let
T
11
= cx
2
, T
22
= T
33
= T
12
= T
13
= T
23
= 0,
where c is a constant. Approximates the
state of stress in a prismatic beam parallel
to x
1
axis with couples about axes paral-
lel to x
3
acting on its end faces (see Fig.).
Plane x
2
= 0 chosen so that the resul-
tant on each face = 0. Principal axes as
in (b).
x
1
x
3
x
2
(e) Plane stress.
T
11
= T
11
(x
1
, x
2
), T
22
= T
22
(x
1
, x
2
), T
12
= T
12
(x
1
, x
2
), T
33
= T
13
= T
23
= 0
Here, the equilibrium eqs. read
j
T
ij
= 0, where i, j = 1, 2. One principal axis e
3
and the
others in the x
1
x
2
plane.
Continuum Mechanics spring 2012 69
(f) Pure torsion. Suppose that
T
13
= x
2
g(r), T
23
= x
1
g(r), T
11
= T
22
= T
33
= T
12
= 0,
where r =
_
x
2
1
+x
2
2
and g(r) is an arbitrary function. Gives the state of stress in a cylindrical
bar, whose axis of symmetry coincides with x
3
axis and which is twisted by couples acting about
the axis and applied to the ends of the cylinder.
Principal directions: radial direction e
r
=
1
r
(x
1
e
1
+x
2
e
2
) and the two bisectors
1

2
(e

e
3
)
of the tangential and axial directions, e

=
1
r
(x
2
e
1
+x
1
e
2
) and e
3
.
x
3
x
2
x
1
Continuum Mechanics spring 2012 70
4 Motions and deformations
rigid-body motions
deformation and strain
examples of simple nite deformations
innitesimal strain tensor
innitesimal rotation and rate-of-strain tensor
velocity gradient and spin tensor
some simple ows
decomposition of a deformation*
principal stretches and principal axes of deformation*
strain invariants*
Continuum Mechanics spring 2012 71
Rigid-body motions
The motion of a body is described by
x
i
= x
i
(X
R
, t), or x = x(X, t).
We already considered these equations with respect to the kinematics of individual material particles.
The next task is to anlyse how a particle modes wrt. the neighboring particles. We will start this by
analyzing rigid-body motions.
Denition. A rigid body is a body, which moves without changing its shape. Thus, during the motion,
the distances between any of its particles remain constant and so do the angles between lines joining the
particles of the body.
Rigid-body motions are superpositions of translations and rotations.
Denition. Translation is a rigid-body motion in which every particle undergoes the same displace-
ment:
x = X +c(t),
where c is the displacement, independent of position but a function of time.
Continuum Mechanics spring 2012 72
Rotation about x
3
axis
Consider a motion, where the body B rotates about x
3
axis through an angle (which may depend on
t). Let the right-handed sense of rotation correspond to positive values of . Clearly,
X
1
= Rcos , X
2
= Rsin , where R =
_
X
2
1
+X
2
2
x
1
= Rcos( +) = R(cos cos sin sin ) = X
1
cos X
2
sin
x
2
= Rsin( +) = R(sin cos + cos sin ) = X
1
sin +X
2
cos
x
1
= X
1
cos X
2
sin , x
2
= X
1
sin +X
2
cos , x
3
= X
3
i.e.,
x = Q X or x
i
= Q
iR
X
R
,
where Q = Q
iR
e
i
e
R
with
Q =
_
_
cos sin 0
sin cos 0
0 0 1
_
_
.
x
1
x
2
X
x

The inverse relation is given by


X = Q
T
x or X
R
= Q
iR
x
i
Continuum Mechanics spring 2012 73
Rotation about an arbitrary axis. General rigid-body motion
Rotation through an angle about an arbitrary axis n = n
i
e
i
is ob-
tained in the following way:
c = n X = n x
X = cn +Y , x = cn +y
Clearly, in rotation about n, ]Y ] = ]y]. Thus,
y = Y cos +n Y sin
x = cn +Y cos +n Y sin
= cn + (X cn) cos +n (X cn) sin
= Xcos + (n X) sin +c(1 cos )n
= Xcos + (n X) sin +n X(1 cos )n
= [cos I + sin N + (1 cos )n n] X Q X
x
3
x
2
x
1
cn

s
i
n
Y
x
n
n
X
x
y
Y

Y
y

cos Y
where N = e
ijR
n
j
e
i
e
R
is the antisymmetric tensor dened by N a = n a a. The
components of the tensor Q = Q
iR
e
i
e
R
are, thus, given by
Q
iR
=
iR
cos +e
ijR
n
j
sin + (1 cos )n
i
n
R
.
Continuum Mechanics spring 2012 74
Clearly, by using another base e
i
, where e
3
= n, we have Q =

Q
iR
e
i
e
R
with
_
_
cos sin 0
sin cos 0
0 0 1
_
_
=

Q,
where det

Q = cos
2
+ sin
2
= 1. Thus, Q is proper orthogonal. In fact,
Denition. Rotation is a rigid-body motion in which the position vector in the current conguration is
obtained from the position vector in the reference conguration by taking an inner product with a proper
orthogonal tensor, i.e.,
x = Q(t) X,
where Q is a proper orthogonal tensor independent of position but a function of time.
Denition. A general rigid-body motion is a superposition of a translation and rotation, i.e.,
x = Q(t) X +c(t),
where Q is an arbitrary (proper) orthogonal tensor and c is an arbitrary vector, both functions of time
but not of position. Inversely,
X = Q
T
(t) x +c
1
(t),
where c
1
(t) = Q
T
(t) c(t).
Continuum Mechanics spring 2012 75
Extension of a material line element
Denition. A motion in which a change of shape takes place is called a deformation (deformaatio).
A body which can change its shape is deformable (deformoituva), in contrast to a rigid body.
Aim: separation of the part of the motion corresponding to the deformation from the part corresponding
to a rigid-body motion.
In a deformation, there are changes of distances between material particles. Begin, therefore, by examining
the extension or stretch of a material line element.
Consider a line segment lying inside a body B, joining
points P
0
and Q
0
in the reference conguration. Ass.

P
0
Q
0
= LA,
where A is a unit vector. If P
0
has coordinates X
(0)
R
,
then Q
0
has coordinates X
(0)
R
+A
R
L.
P
0
R
0
L l
P
Q
Q
0
R
A
a
The particles lying on P
0
Q
0
at time t = 0 form a segment of a material curve. After a motion, the
particles will in general lie on a new curve in space. We wish to determine the length and the orientation
of the material line element after the motion.
Ass. Particles originally at P
0
and Q
0
move to P and Q, respectively, and

PQ = l a, with a
i
a
i
= 1.
Thus, if P has coordinates x
(0)
i
then Q has coordinates x
(0)
i
+a
i
l.
Continuum Mechanics spring 2012 76
Since P was initially at P
0
and Q was initially at Q
0
, the motion x
i
= x
i
(X
R
, t) yields
x
(0)
i
= x
i
(X
(0)
R
, t)
x
(0)
i
+a
i
l = x
i
(X
(0)
R
+A
R
L, t)
= x
i
(X
(0)
R
, t) +A
S
L
x
i
X
S
(X
(0)
R
, t) +c(L
2
)
= x
(0)
i
+A
S
L
x
i
X
S
(X
(0)
R
, t) +c(L
2
)
L0
a
i
dl
dL
= A
S
x
i
X
S
(X
(0)
R
, t)
Here
:=
dl
dL
is called the extension ratio or stretch ratio and it is the ratio of nal and initial lengths of an intesimal
material line element initially at X
(0)
R
. Since P
0
was arbitrary, we get
a
i
= A
S
x
i
(X
R
, t)
X
S
(1)
(a
i
)(a
i
) = A
S
x
i
X
S
A
T
x
i
X
T
a
i
a
i
=1

2
= A
S
A
T
x
i
X
S
x
i
X
T
(2)
Once is determined from (2), the current orientation of the line element, a
i
, is obtained from (1).
Continuum Mechanics spring 2012 77
If the motion is determined in the spatial description, i.e.,
X
R
= X
R
(x
i
, t) or X = X(x, t)
we can determine and the orientation A in the reference conguration in a similar way:
X
(0)
R
= X
R
(x
(0)
i
, t)
X
(0)
R
+A
R
L = X
R
(x
(0)
i
+a
i
l, t)
= X
i
(x
(0)
i
, t) +a
i
l
X
R
x
i
(x
(0)
j
, t) +c(l
2
)
= X
(0)
R
+a
i
l
X
R
x
i
(x
(0)
j
, t) +c(l
2
)
l0

dL
dl
A
R
= a
i
X
R
x
i
(x
(0)
j
, t)
A
R
= a
i

X
R
x
i

2
= a
i
a
j
X
S
x
i
X
S
x
j
These give the stretch ratio and original orientation for a line element that is currently at x
i
and has
an orientation given by a
i
.
Continuum Mechanics spring 2012 78
The deformation gradient tensor
The nine quantities x
i
/X
R
are called deformation gradients. We denote
F
iR
=
x
i
X
R
(3)
and by setting F = F
iR
e
i
e
R
we can form a second order tensor.
The coordinated themselves transform according to x
i
= M
ij
x
j
and X
S
= M
RS

X
R
. Thus,
M
ij
M
RS
F
jS
=
x
i
x
j
X
S


X
R
x
j
X
S
=
x
i
x
j
x
j
X
S
X
S


X
R
=
x
i
X
S
X
S


X
R
=
x
i


X
R
=

F
iR
.
Thus, the components (3) of F transform, indeed, as those of a tensor.
We call F the deformation gradient tensor. Note that it is in general not symmetric.
Also the transpose of F, F
T
= F
iR
e
R
e
i
, is a second order tensor and, as long as det F ,= 0, so
is F
1
. Since
x
i
X
R
X
R
x
j
=
x
i
x
j
=
ij
,
we have F
1
= F
1
Rj
e
R
e
j
with
F
1
Rj
=
X
R
x
j
.
Continuum Mechanics spring 2012 79
Deformation of a material curve element in tensor and matrix notation
Equations for the present orientation and stretch ratio of a material curve element read
a
i
=
1
A
R
x
i
X
R
=
1
A
R
F
iR
and
2
= A
S
A
T
x
i
X
S
x
i
X
T
= A
S
A
T
F
iS
F
iT
so they can be written in direct tensor notation as
a =
1
F A and
2
= A F
T
F A
Similarly,
A = F
1
a and
2
= a (F
1
)
T
F
1
a.
For the practical calculation of and a or A it is often convenient to use the matrix notation (note:
A is here a column matrix)
a =
1
FA,
2
= A
T
F
T
FA
A = F
1
a,
2
= a
T
(F
1
)
T
F
1
a
If theres no motion, X = x and F = I.
Continuum Mechanics spring 2012 80
Example
Problem. Consider the motion
x
1
= X
1
e
at
+X
3
(e
at
1), x
2
= X
2
+X
3
(e
at
e
at
), x
3
= X
3
Calculate the maximum and minimum stretch ratio at t = a
1
ln 2 for A in the X
1
X
2
plane.
Solution.
F =
_
_
_
_
x
1
X
1
x
1
X
2
x
1
X
3
x
2
X
1
x
2
X
2
x
2
X
3
x
3
X
1
x
3
X
2
x
3
X
3
_
_
_
_
=
_
_
e
at
0 e
at
1
0 1 e
at
e
at
0 0 1
_
_
=
_
_
2 0 1
0 1
3
2
0 0 1
_
_
Thus, taking A = A
1
e
1
+A
2
e
2
= cos e
1
+ sin e
2
,

2
= A
T
F
T
FA = (A
1
A
2
0)
_
_
2 0 0
0 1 0
1
3
2
1
_
_
_
_
2 0 1
0 1
3
2
0 0 1
_
_
_
_
A
1
A
2
0
_
_
= (2A
1
A
2
0)(2A
1
A
2
0)
T
= 4A
2
1
+A
2
2
= 4 cos
2
+ sin
2
= 1 + 3 cos
2
,
which has extremum values of 1 and 4, obtained at cos = 0 and cos = 1, respectively.
Continuum Mechanics spring 2012 81
Displacement gradients. Deformation gradients of rigid-body motion
The components of the displacement vector u are
u
i
= x
i
X
i
.
The displacement gradients are
u
i
X
R
=
x
i
X
R

X
i
X
R
=
x
i
X
R

iR
= F
iR

iR
so they form a second order tensor F I called the displacement gradient tensor (siirtymagradient-
titensori). If theres no motion, then all components of F I are zero.
Even if F contains the information about deformations, it is not itself a suitable measure of deformation.
(A measure of deformation should not change if no deformation takes place.) For rigid-body motion:
x = Q(t) X +c(t) or x
i
= Q
iR
(t)X
R
+c
i
(t), i.e.,
F
iR
=
x
i
X
R
= Q
iR
(t)
so F = Q(t), which is a function of time. Furthermore,

2
= A Q
T
Q A = A A = 1 and a =
1
F A = Q A.
Thus theres no stretch but simply a particle-independent rotation of unit vector a.
Continuum Mechanics spring 2012 82
Right CauchyGreen deformation tensor
We dened the deformation gradient tensor F for motion x
i
= x
i
(X
R
, t) and the stretch ratio for
an innitesimal material line element aligned in reference conguration with the unit vector A as
F(X, t) =
x
i
X
R
e
i
e
R
,
2
(X, t; A) = A F
T
F A.
We showed that for rigid-body motion, x = Q(t) X + c(t), the stretch ratio is unity and the
deformation gradient tensor is F = Q(t), which is therefore not a measure of deformation.
Dene a new tensor
C = F
T
F
so that C
RS
= F
T
Ri
F
iS
= F
iR
F
iS
= (x
i
/X
R
)(x
i
/X
S
). Since C is an inner product of
two second-order tensors, it is itself a second order tensor. Clearly, C
RS
= C
SR
so C is symmetric.
Now,

2
= A C A
so knowing C is sucient for the calculation of the stretch ratio for any material line element. The
components C
RS
(X
R
, t) at a particle X
R
determine the local deformation around that particle.
For rigid-body motion,
C = Q
T
Q = I
so C is a suitable measure of deformation. It is called the right CauchyGreen deformation tensor
(muodonmuutostensori).
Continuum Mechanics spring 2012 83
Other deformation tensors
Tensor C is not the only valid measure of deformation. We can use instead, for example,
C
1
= (F
T
F)
1
= F
1
(F
1
)
T
with components
C
1
RS
= F
1
Ri
F
1
Si
=
X
R
x
i
X
S
x
i
,
which is straight-forward to nd, when the motion is given in the spatial description, X
R
= X
R
(x
i
, t).
Consider motion in the spatial description. To nd out the stretch ratio of a material line element aligned
with unit vector a in the current conguration, we use

2
= a (F
1
)
T
F
1
a a B
1
a
where the components of the tensor B
1
are
B
1
ij
= F
1
Ri
F
1
Rj
=
X
R
x
i
X
R
x
j
.
Its inverse,
B = F F
T
, or B
ij
=
x
i
X
R
x
j
X
R
is called the left CauchyGreen deformation tensor. B
1
determines the local deformation around a
point with spatial coordinates x
i
. Clearly, also B is symmetric and B = I for rigid-body motion.
Continuum Mechanics spring 2012 84
Strain tensors
The Lagrangian strain tensor (venymatensori) is
=
1
2
(C I)
The Eulerian strain tensor is
=
1
2
(I B
1
)
Both are suitable measures of deformation. They have properties = = 0 for rigid-body motion.
If the motion is given by x = x(X, t) then it is straightforward to calculate
F
iR
(X, t) =
x
i
X
R
and natural to use C(X, t), B(X, t) or (X, t) as a measure of deformation. As these tensors
explicitly depend on the material coordinates, they give the deformation in the neighborhood of a given
particle.
If the motion is given by X = X(x, t) then it is easier to calculate
F
1
Ri
(x, t) =
X
R
x
i
and natural to use C
1
(x, t), B
1
(x, t) or (x, t) as a measure of deformation. As these tensors
explicitly depend on the spatial coordinates, they give the deformation, which has taken place in the
neighborhood of a given point in space.
Continuum Mechanics spring 2012 85
Some simple nite deformations
(a) Uniform extensions. Suppose a body is extended uniformly in, say, x
1
direction, with a stretch
ratio of
1
. Then x
1
=
1
X
1
+ c, where the constant c can be set to zero with an appro-
priate translation of the origin. If the body undergoes uniform extensions in all three coordinate
directions, we get
x
1
=
1
X
1
, x
2
=
2
X
2
, x
3
=
3
X
3
,
where
1
,
2
and
3
are constants, or functions of t. Special cases:

1
=
2
uniform extension in all directions x
3
.

1
=
2
=
3
uniform extension in all directions; called uniform dilatation.

1
=
1
2
areas conserved in planes x
3
.

1
=
1
2
,
3
= 1 areas conserved in planes x
3
, volumes conserved; called pure shear.
For the deformation, F = (F
iR
) = (x
i
/X
R
), C = F
T
F, B
1
= (F
1
)
T
F
1
,
(
RS
) =
1
2
(C I), and (
ij
) =
1
2
(I B
1
)
F =
_
_

1
0 0
0
2
0
0 0
3
_
_
C =
_
_

2
1
0 0
0
2
2
0
0 0
2
3
_
_
, B
1
=
_
_

2
1
0 0
0
2
2
0
0 0
2
3
_
_
(
RS
) =
1
2
_
_

2
1
1 0 0
0
2
2
1 0
0 0
2
3
1
_
_
, (
ij
) =
1
2
_
_
1
2
1
0 0
0 1
2
2
0
0 0 1
2
3
_
_
Continuum Mechanics spring 2012 86
(b) Simple shear. Parallel planes displaced rela-
tive to each other by an amount proportional
to the distance between the planes and in a
direction parallel to the planes. E.g.,
x
1
= X
1
+X
2
tan , x
2
= X
2
, x
3
= X
3
Here, the planes X
2
= const. are the
shear planes and the X
1
direction is the
shear direction.
x
2
x
1

x
X
u
The angle is a measure of the amount of shear. In simple shear, the volume of all parts of the
body remain constant. For deformation,
F =
_
_
1 tan 0
0 1 0
0 0 1
_
_
F
1
=
_
_
1 tan 0
0 1 0
0 0 1
_
_
C =
_
_
1 0 0
tan 1 0
0 0 1
_
_
_
_
1 tan 0
0 1 0
0 0 1
_
_
=
_
_
1 tan 0
tan 1 + tan
2
0
0 0 1
_
_
,
B
1
=
_
_
1 0 0
tan 1 0
0 0 1
_
_
_
_
1 tan 0
0 1 0
0 0 1
_
_
=
_
_
1 tan 0
tan 1 + tan
2
0
0 0 1
_
_
Continuum Mechanics spring 2012 87
(c) Homogeneous deformations. Motions of form
x = c +A X, or x
i
= c
i
+A
iR
X
R
,
where c and A are constants or functions of time. Clearly, (a) and (b) are special cases of (c).
The deformation gradient tensor F = A and, of course, all the deformation tensors are also
independent of X or x. Homogeneous deformation tensors have, e.g., the following properties:
(i) Material surfaces which form planes in the reference conguration deform into planes; two
parallel planes deform into two parallel planes.
(ii) Material curves which form straight lines in the reference conguration deform into straight
lines; twp parallel straight lines deform into two parallel straight lines.
(iii) A material surface which forms a spherical surface in the reference conguration deforms into
an elliptical surface.
Proof of (i): Let X
R
be in a plane with a unit normal n and perpendicular distance p from the origin.
Thus,
n X = p.
After deformation,
x = c +A X X = A
1
(x c)
p = n [A
1
(x c)] = (n A
1
) x n A
1
c
n x = p, where n = n A
1
and p = p +n A
1
c.
Continuum Mechanics spring 2012 88
(d) Plane strain. Analogously to plane stress, a deformation in plane strain is given by
x
1
= x
1
(X
1
, X
2
), x
2
= x
2
(X
1
, X
2
), x
3
= X
3
The planes X
3
= const. are the deformation planes. Particles which initially lie in a given
deformation plane remain in that plane.
(e) Pure torsion. Consider the system in cylindrical polar coordinates (R, , Z) and (r, , z)
dened by
X
1
= Rcos , X
2
= Rsin , X
3
= Z
x
1
= r cos , x
2
= r sin , x
3
= z
A pure torsion is dened by
r = R, = +Z, z = Z,
where is a constant or a function of time. No volume change involved, deformation not
homogeneous.
(f) Pure exure. This time, take
r = f(X
1
), = g(X
2
), z = X
3
Represents the bending of a rectangular block into a
sector of a circular cylindrical tube. Material surfaces
X
1
= const. deform to material surfaces r = const.
and material surfaces X
2
= const. deform to matrerial
surfaces = const.
x
1
x
2 X
2
X
1

r
Continuum Mechanics spring 2012 89
Components of strain tensors in terms of displacement gradients
Since u = x X, we have
F
iR
=
x
i
X
R
=
u
i
X
R
+
iR
.
Thus, from =
1
2
(C I) and C = F
T
F

RS
=
1
2
(F
iR
F
iS

RS
) =
1
2
__
u
i
X
R
+
iR
__
u
i
X
S
+
iS
_

RS
_
=
1
2
_
u
R
X
S
+
u
S
X
R
+
u
i
X
R
u
i
X
S
_
.
Similarly, from =
1
2
(I B
1
), B
1
= (F
1
)
T
F
1
, and F
1
Ri
=
X
R
x
i
=
Ri
u
R
/x
i
,

ij
=
1
2
(
ij
F
1
Ri
F
1
Rj
) =
1
2
_

ij

Ri

u
R
x
i
__

Rj

u
R
x
j
__
=
1
2
_
u
i
x
j
+
u
j
x
i

u
R
x
i
u
R
x
j
_
The tensors C, C
1
, B, B
1
, , and are all symmetric 2
nd
-order tensors, so they have real principal
components and orthogonal principal directions. In fact, the rst four are positive denite tensors.
Continuum Mechanics spring 2012 90
Innitesimal strain
For many materials, changes of shape remain small. In this case, a great deal of simplication is achieved
by using the approximate innitesimal strain tensor.
Ass. All components of the displacement gradient tensor are small:

u
i
X
R

_1 (i, R = 1, 2, 3)
Thus, neglect terms of higher than linear order in these quantities.
Now, as u
i
= x
i
X
i
,
_
u
i
x
j
_
=
_

ij

X
i
x
j
_
= I F
1
.
Use the binomial expansion:
I F
1
= I I + (F I)]
1
= I I (F I) + (F I)
2
(F I)
3
+. . .]
Thus, _
u
i
x
j
_
= (F I) (F I)
2
+ (F I)
3
+. . .

F I =
_
u
i
X
R
_
=
_
u
i
X
j
_

_
u
i
X
R
u
R
X
j
_
+
_
u
i
X
R
u
R
X
S
u
S
X
j
_
. . .
So to rst order u
i
/x
j
u
i
/X
j
!
Continuum Mechanics spring 2012 91
Keeping only terms linear in the displacement gradients, thus, yields for the components of =
1
2
(C
I) and =
1
2
(I B
1
)

RS
=
1
2
_
u
R
X
S
+
u
S
X
R
+
u
i
X
S
u
i
X
R
_

1
2
_
u
R
X
S
+
u
S
X
R
_

1
2
_
u
R
x
S
+
u
S
x
R
_

ij
=
1
2
_
u
i
x
j
+
u
j
x
i

u
R
x
i
u
R
x
j
_

1
2
_
u
i
x
j
+
u
j
x
i
_

1
2
_
u
i
X
j
+
u
j
X
i
_
so
ij

ij
. The second-order tensor E with components
E
ij
=
1
2
_
u
i
X
j
+
u
j
X
i
_
is called the innitesimal strain tensor. Both and reduce to E in the limit where terms of higher
than linear order in displacement gradients are neglected. Also,
E =
1
2
(F +F
T
) I
and this relation is exact. Clearly, E is symmetric.
Continuum Mechanics spring 2012 92
Innitesimal strain tensor in rigid-body rotation
Tensor E is not an exact measure of deformation, because it does not remain constant in rigid-body
rotation. However, as a rotation through an angle around n is given by
F
iR
=
iR
cos +e
ijR
n
j
sin + (1 cos )n
i
n
R
,
the innitesimal strain tensor becomes
E
ij
=
1
2
(F
ij
+F
ji
)
ij
= (1 cos )(n
i
n
j

ij
)
= 1 (1

2
2!
+. . .)](n
i
n
j

ij
)
1
2

2
(n
i
n
j

ij
)
so the in the small displacement gradient approximation _ 1, E
ij
0 for rigid-body rotation.
Thus, in practice, the innitesimal strain tensor often provides an excellent measure of deformation.
Unlike and , E is linear in the displacement components, which simplies the analysis considerably.
linear theory of elasticity.
Continuum Mechanics spring 2012 93
Geometrical interpretation of E
11
and E
23
X
2
P
0
Q
0
L
L(1+ u
1
/ X )
1
X
1
X
1
+L X
1
+u
1
X
1
X
1
+u
1
+ L(1+ u
1
/ X )
1
P
Q
X
3
P
0
Q
0
X
2
R
0

2
L
2
L
3
L
3
u
2
/ X
3
L
2
u
3
/ X
2
Q
P
R
The line element P
0
Q
0
of length L lies parallel
to the X
1
axis. Rotation of the line element small
extension, to rst order in L, is
u
1
(X
1
+L, X
2
, X
3
) u
1
(X
1
, X
2
, X
3
)
L
u
1
X
1
= LE
11
Thus, E
11
is the extension per unit initial length of
a line element, which is initially j X
1
axis.
Similarly, let P
0
R
0
and P
0
Q
0
be line elements
aligned with X
3
and X
2
axes. Thus (exercise),

u
3
X
2
,
2

u
2
X
3
E
23
=
1
2
_
u
3
X
2
+
u
2
X
3
_

1
2
(
1
+
2
)
Thus, 2E
23
is the decrease in the angle between
line elements initially parallel to X
2
and X
3
.
Continuum Mechanics spring 2012 94
Properties of E. Strain compatibility relations*
Eis symmetric it has three real principal components, E
1
, E
2
and E
3
called the principal components
of innitesimal strain, and the related three orthogonal principal axes.
If these axes are taken as the coordinate axes, the matrix of E is E = diag(E
1
, E
2
, E
3
).
Symmetric functions of E
1
, E
2
and E
3
are invariants of E.
Since the six components
E
ij
=
1
2
_
u
i
X
j
+
u
j
X
i
_
all depend on u
i
they must satisfy the relations obtained by eliminating the u
i
from them. These are
the strain compatibility relations,
K
i
2

2
E
jk
X
j
X
k

2
E
jj
X
2
k
+

2
E
kk
X
2
j
_
= 0
L
i


2
E
ii
X
j
X
k
+

X
i
_
E
jk
X
i

E
ki
X
j

E
ij
X
k
_
= 0
no summation, i, j, k] = 1, 2, 3], 2, 3, 1], 3, 1, 2]. The relations themselves satisfy the
three relations
K
i
X
i
=
L
j
X
k
+
L
k
X
j
.
Continuum Mechanics spring 2012 95
Innitesimal rotation tensor
Consider, rst, a rotation given by F
iR
=
iR
cos +e
ijR
n
j
sin +(1 cos )n
i
n
R
, where
is the rotation angle and n the rotation axis.
Ass. _1 F
iR
=
iR
+e
ijR
n
j
+c(
2
).
Thus, for an innitesimal rotation,
u
i
= x
i
X
i
= (F
iR

iR
)X
R
= e
ijR
n
j
X
R

u
i
X
R
= e
ijR
n
j
,
_
u
i
X
R
_
=
_
_
0 n
3
n
2
n
3
0 n
1
n
2
n
1
0
_
_
Thus, an innitesimal rotation is described by an antisymmetric tensor with the axial vector n.
Now, consider a general innitesimal motion with deformation gradient tensor F. The innitesimal
rotation tensor is dened as the antisymmetric part of F, i.e.,
=
1
2
(F F
T
),
ij
=
1
2
_
u
i
X
j

u
j
X
i
_
.
Thus, the displacement gradient tensor F I is decomposed to its symmetric and antisymmetric parts
as
F I =
1
2
(F +F
T
) I +
1
2
(F F
T
) = E +
expressing any innitesimal motion as the sum of an i.t. deformation, E, and an i.t. rotation, .
Continuum Mechanics spring 2012 96
Innitesimal rotation vector
The innitesimal rotation vector, , is dened by
=
1
2
Curl u =
1
2
e
ijk
e
i
u
k
X
j
.
Thus,

1
=
1
2
e
1jk

j
u
k
=
1
2
(
2
u
3

3
u
2
) =
23
=
1
2
(
23

32
) =
1
2
e
1jk

jk

2
=
1
2
e
2jk

j
u
k
=
1
2
e
2jk

jk

3
=
1
2
e
3jk

j
u
k
=
1
2
e
3jk

jk
i.e.,
i
=
1
2
e
ijk

jk
. Multiply this by e
irs
and sum over i to get
e
irs

i
=
1
2
e
irs
e
ijk

jk
=
1
2
(
rj

sk

rk

sj
)
jk
=
1
2
(
rs

sr
) =
rs
,
i.e.,
jk
= e
ijk

i
. As, therefore,

jk
a
k
= e
ijk

i
a
k
= e
jik

i
a
k
,
we can write a = a for any vector a. Thus, is the axial vector of .
Continuum Mechanics spring 2012 97
The rate-of-deformation tensor
Many times, especially in uid mechanics, the kinematic property of interest is not the shape of the body
but the rate at which the change of shape takes place.
Thus, consider
a
i
= A
S
x
i
X
S
,
2
= A
S
A
T
x
i
X
S
x
i
X
T
which give the stretch ratio and current orientation of the material line element in terms of material
coordinates X
R
and direction cosines A
R
. Calculate D/Dt = (X
R
, t; A
R
)/t as
2
D
Dt
= A
S
A
T
_
x
i
X
T

X
S
Dx
i
Dt
+
x
i
X
S

X
T
Dx
i
Dt
_
= A
S
A
T
_
x
i
X
T
v
i
X
S
+
x
i
X
S
v
i
X
T
_
Introduce the velocity components as functions of the spatial coordinates, whence
v
i
X
S
=
v
i
x
j
x
j
X
S
and

D
Dt
=
1
2
A
S
A
T
_
x
i
X
T
v
i
x
j
x
j
X
S
+
x
i
X
S
v
i
x
j
x
j
X
T
.
ij
_
=
1
2
A
S
x
j
X
S
A
T
x
i
X
T
_
v
i
x
j
+
v
j
x
i
_

D
Dt
=
1
2
a
i
a
j
_
v
i
x
j
+
v
j
x
i
_
= a
i
D
ij
a
j
, where D
ij
=
1
2
_
v
i
x
j
+
v
j
x
i
_
The tensor D = D
ij
e
i
e
j
is the rate-of-deformation tensor or rate-of-strain or strain-rate tensor
(deformaationopeustensori).
Continuum Mechanics spring 2012 98
The components of the strain-rate tensor transform as (x
i
= M
ji
x
j
, v
i
= M
ij
v
j
)

D
ij
= M
ir
M
js
D
rs
=
1
2
M
ir
M
js
_
v
r
x
s
+
v
s
x
r
_
=
1
2
_
M
js
(M
ir
v
r
)
x
s
+M
ir
(M
js
v
s
)
x
r
_
=
1
2
_
x
s
x
j
v
i
x
s
+
x
r
x
i
v
j
x
r
_
=
1
2
_
v
i
x
j
+
v
j
x
i
_
,
so the denition is independent of the coordinate system, as it should be, of course.
The rate-of-deformation tensor D is a second-order symmetric tensor. Thus, it has three principal
components, D
1
, D
2
and D
3
, and three orthogonal principal axes related to them. The largest and
smallest of them are extremal values of the strain rate for variations of the directions a. Symmetric
functions of the three principal components are invariants of D.
Analogous to compatibility relations of the innitesimal strain tensor, the components of D obey com-
patibility relations
2

2
D
jk
x
j
x
k

2
D
jj
x
2
k
+

2
D
kk
x
2
j
_
= 0;

2
D
ii
x
j
x
k
+

x
i
_
D
jk
x
i

D
ki
x
j

D
ij
x
k
_
= 0,
where i, j, k] = 1, 2, 3], 2, 3, 1], 3, 1, 2] and no summation is applied.
The rate-of-deformation tensor is an exact measure of deformation rate. The fact that D
ij
are linear in
velocities is very fortunate, simplifying the problems of uid mechanics.
Continuum Mechanics spring 2012 99
The velocity gradient tensor. The spin tensor
The deformation-rate tensor is the symmetric part of the velocity gradient tensor,
L = L
ij
e
i
e
j
, L
ij
=
v
i
x
j
; D =
1
2
(L +L
T
).
Note that L
T
= (
i
v
j
)e
i
e
j
= v. The antisymmetric part, W, of L is called the spin tensor
or vorticity tensor (pyorteisyystensori)
W =
1
2
(L L
T
); W
ij
=
1
2
_
v
i
x
j

v
j
x
i
_
.
W has properties analogous to the innitesimal rotation tensor, except that no approximation is involved
in its derivation or use. It is a measure of rotation rate of an element.
The spin may also be described by the vorticity vector,
w = curl v = e
ijk
e
i
v
k
x
j
.
By relations similar to those between and ,
w
i
= e
ijk
W
jk
, W
jk
=
1
2
e
ijk
w
i
Continuum Mechanics spring 2012 100
In a rigid-body rotation with angular speed about an axis through O in the direction n, the velocity
is given by
v = n x, or v
i
= e
ijk
n
j
x
k
Thus, w = 2n and
L
ik
= e
ijk
n
j
, D
ik
= 0, (L
ik
) = (W
ik
) =
_
_
0 n
3
n
2
n
3
0 n
1
n
2
n
1
0
_
_
Thus, D vanishes for rigid-body rotation. Morover, superposing a rigid-body motion with a general
motion gives the same D as without the superposition. D is unaected by the superposed rigid-
body rotations and, thus, a suitable measure of deformation rate.
Material derivative of F is given by
DF
iR
Dt
=
D
Dt
x
i
X
R
=
v
i
X
R
=
v
i
x
j
x
j
X
R
= L
ij
F
jR
, i.e.,
DF
Dt
= L F L =
DF
Dt
F
1
.
In the case of small displacement gradients, use linear approximation in (F I):
L =
D(F I)
Dt
I + (F I)]
1
=
D(F I)
Dt
I (F I) +c(F I)
2
]
L
DF
Dt
, and since F = I +E +, D
DE
Dt
, W
D
Dt
, w 2
D
Dt
Note: some books dene vorticity as w =
1
2
curlv, whence w D/Dt. Be careful!
Continuum Mechanics spring 2012 101
Some simple ows
(a) Simple shearing ow. If the planes x
2
= const. are the shear planes, and x
1
is the direction of
shear, then
v
1
= sx
2
, v
2
= 0, v
3
= 0 (s const.)
is a simple shearing ow. For this ow
L
12
= s, L
ij
= 0 otherwise, i.e.,
D
12
=
1
2
s = D
21
= W
12
= W
21
; D
ij
= W
ij
= 0 otherwise
(b) Rectilinear ow. In rectililiear ow, the material ows in parallel straight lines like, e.g., in a
straight pipe of uniform cross-sesction or between parallel plates. If the ow direction is the x
3
axis,
v
1
= 0, v
2
= 0, v
3
= f(x
1
, x
2
, x
3
)
L
3j
=
f
x
j
, L
ij
= 0 otherwise, i.e., D
33
=
f
x
3
D
3j
=
1
2
f
x
j
= D
j3
= W
3j
= W
j3
, j = 1, 2; D
ij
= W
ij
= 0 otherwise
If the velocity is independent of x
3
, then also D
33
= 0. Simple shearing ow is a rectilinear
ow.
Continuum Mechanics spring 2012 102
(c) Vortex ow (pyorrevirtaus). Flow in the neighborhood of a vortex line lying along the x
3
axis is
described by
v
1
=
x
2
x
2
1
+x
2
2
, v
2
=
x
1
x
2
1
+x
2
2
, v
3
= 0, (r
2
= x
2
1
+x
2
2
,= 0)
where is a constant. Velocity vector v = v
1
e
1
+v
2
e
2
= (/r)e

, where r and are the


polar coordinates of the x
1
x
2
plane. There is a singularity at r = 0. Furthermore
L
11
=
2x
1
x
2
(x
2
1
+x
2
2
)
2
= L
22
, L
12
=
(x
2
2
x
2
1
)
(x
2
1
+x
2
2
)
2
= L
21
; L
ij
= 0 otherwise
D = L; W = 0
(d) Plane ow. If the velocity is given by
v
1
= v
1
(x
1
, x
2
, t), v
2
= v
2
(x
1
, x
2
, t), v
3
= 0,
the ow is in planes parallel to the x
1
x
2
plane.
L
11
=
v
1
x
1
, L
22
=
v
2
x
2
, L
12
=
v
1
x
2
, L
21
=
v
2
x
1
, L
ij
= 0 otherwise
D
11
= L
11
, D
22
= L
22
, D
12
= D
21
=
1
2
(L
12
+L
21
) , D
ij
= 0 otherwise
W
12
= W
21
=
1
2
(L
12
L
21
) , W
ij
= 0 otherwise
w = v = e
3
_
v
2
x
1

v
1
x
2
_
= 2W
21
e
3
Continuum Mechanics spring 2012 103
Decomposition of a deformation*
By polar decomposition theorem, the deformation gradient tensor can be uniquely decomposed to either
of the forms
F = R U = V R,
where R is an orthogonal tensor and U and V are positive denite tensors. Since (see next chapter )
det F =
0
/ > 0 and det U, det V > 0, also det R > 0, so R is a proper orthogonal tensor.
We have
U = R
T
V R and V = R U R
T
Consider, rst, a homogeneous deformation x = F X, where F
iR
are constants (or functions of
time). Suppose that the body undergoes two successive homogeneous motions, in which the particle
initially at X rst moves to x and then to x, where
x = U X, x = R x.
Then,
x = R (U X) = (R U) X = F X,
Since R is orthogonal, the second motion describes a rotation of the body. Thus, any homogeneous
deformation can be decomposed into a homogeneous deformation described by a symmetric tensor U
followed by a rigid-body rotation described by R. Equivalently, it can be decomposed into the same
rotation followed by another homogeneous deformation described by a symmetric tensor V .
Continuum Mechanics spring 2012 104
If F is not homogeneous, the motion x
i
= x
i
(X
R
) is described by
dx
i
=
x
i
X
R
dX
R
= F
iR
dX
R
or dx = F dX.
Then the decompostion can still be made but the tensors F, R, U and V are now functions of position.
Thus, the decompositon F = R U describes a local deformation U followed by a local rotation R
and similarly for F = V R.
Tensor R is called the rotation tensor. The tensors U and V are called the right and left stretch
tensors, respectively. They are closely related to the deformation tensors C and B:
C = F
T
F = (R U)
T
(R U) = U R
T
R U = U
2
B = F F
T
= (V R) (V R)
T
= V R R
T
V = V
2
Because U [V ] is symmetric and positive denite, the components of U [V ] are determined from
those of C [B] and vice versa, of course. Therefore, U and C [V and B] are equivalent measures of
deformation.
Recall that F = I +E +, where E is symmetric and is anti-symmetric. Thus,
U
2
= F
T
F = (I +E ) (I +E +) I + 2E
and U I +E V in the case of small strain/rotation. Also, since U
1
I E,
R = F U
1
(I +E +) (I E) I +
Thus, both UI and V I reduce to E and RI reduces to in the case of small strain/rotation.
Continuum Mechanics spring 2012 105
Principal stretches and principal axes of deformation*
Let F = R U, where R represents rotation. Analyze the motion corresponding to the symmetric,
positive-denite tensor U. Recall the equation giving the orientation of a line element,
a = U A,
where is the stretch ratio. Assume that for a particular line element, the orientation does not change
in the deformation. Thus, a = A and
U A = A or (U I) A = 0.
Clearly, is a principal value of U and A is the corresponding principal axis direction. Since U is
symmetric and positive denite, the principal values are real and positive. They are called the principal
stretches and denoted by
1
,
2
,
3
, in descending order. The corresponding orthogonal principal axis
directions A
1
, A
2
, A
3
determine the principal axes of U.
Choosing the principal axis directions as the base vectors gives the matrix of the tensor U as
U = diag(
1
,
2
,
3
)
and the deformation U consists of extensions along the three coordinate axes with stretch ratios

1
,
2
,
3
. Thus, F consists of these three stretches, followed by the rotation R.
Continuum Mechanics spring 2012 106
The other decomposition of F = V R has a related interpretation. As
0 = R (U I) A = R (U I) R
T
R A
= (R U R
T
I) R A = (V I) R A
Thus, the principal strecthes
i
of U are also the principal stretches of V , and if U has the principal
axis directions A
i
, the principal axes directions of V are given by R A
i
. Thus, they are obtained
from the directions of U through the rotation R.
For a homogeneous deformation, the tensors U, V , and R are constants (or functions of time); thus,
the principal stretches and the principal axis directions are uniform throughout the body. In the general
case,
i
, A
i
and the components of R are functions of position.
Since C = U
2
and =
1
2
(C I), the principal directions of C and are A
i
and their principal
values are
2
i
and
1
2
(
2
i
1) (i = 1, 2, 3). Similarly, the principal directions of B and are those of
V , i.e., R A
i
, and their principal values are
2
i
and
1
2
(
2
i
1) (i = 1, 2, 3). In practice, as C
and B are much easier to calculate than U or V , the principal stretches and principal axis directions
are calculated by analysing C or B.
Continuum Mechanics spring 2012 107
Let us still consider the stretch ratio for a given deformation tensor. Using
2
(A) = A C A with
the constraint f(A) = A A1 = 0, we can nd the extremal values of the stretch ratio by
0 =

2
A
R

2
f
A
R
= C
PQ
(A
P
A
Q
)
A
R

2
(A
P
A
P
)
A
R
= C
PQ
(
PR
A
Q
+
QR
A
P
) 2
2

PR
A
P
= 2(C
RP

2

PR
)A
P
where
2
is a Lagrangian multiplier. Thus, at the extrema of
2
(C
2
I) A = 0,
i.e., Ais one of the principal axis directions A
i
and
2
is corresponding principal value of C,
2
=
2
i
.
Thus, the minimum and maximum values of
2
(A) are the principal values
2
3
and
2
1
and they are
obtained when A coincides with the corresponding principal axis direction.
Continuum Mechanics spring 2012 108
Strain invariants*
The tensors U and V are symmetric, so they posses three basic invariants and these can be chosen as
any symmetric functions of the principal stretches
1
,
2
,
3
. However, we make use of the fact that

2
1
,
2
2
,
2
3
are the principal values of C and B and, because C and B are more easily calculated than
U or V , dene the strain invariants as
I
1
=
2
1
+
2
2
+
2
3
, I
2
=
2
2

2
3
+
2
3

2
1
+
2
1

2
2
, I
3
=
2
1

2
2

2
3
We have
I
1
= tr C = tr B
I
2
=
1
2
(tr C)
2
tr C
2
] =
1
2
(tr B)
2
tr B
2
]
I
3
= det C = det B
The principal values of C
1
and B
1
are
2
1
,
2
2
,
2
3
. Thus,
tr C
1
= tr B
1
=
2
1
+
2
2
+
2
3
=
2
1

2
2

2
3
(
2
2

2
3
+
2
3

2
1
+
2
1

2
2
) = I
2
/I
3
so I
2
= I
3
tr C
1
= I
3
tr B
1
.
Note also that I
3
= det C = det F
T
F = (det F)
2
=
_
dv
dV
_
2
=
_

_
2
so for an
incompressible material, I
3
= 1 (see next chapter ).
Continuum Mechanics spring 2012 109
Examples*
(a) Uniform extensions. For x
1
=
1
X
1
, x
2
=
2
X
2
and x
3
=
3
X
3
we have F
iR
=
i

iR
(no summation) so F is positive denite and, therefore, R = I and U = V = F. The
principal stretches are
1
,
2
,
3
and the principal axes of both C and B are the coordinate
axes. The strain invariants are
I
1
=
2
1
+
2
2
+
2
3
, I
2
=
2
2

2
3
+
2
3

2
1
+
2
1

2
2
, I
3
=
2
1

2
2

2
3
(b) Simple shear. Consider x
1
= X
1
+X
2
tan , x
2
= X
2
, x
3
= X
3
. Thus,
F =
_
_
1 tan 0
0 1 0
0 0 1
_
_
C =
_
_
1 tan 0
tan 1 + tan
2
0
0 0 1
_
_
C
2
=
_
_
1 + tan
2

tan
2
+ (1 + tan
2
)
2
1
_
_
,
so tr C = 3+tan
2
, tr C
2
= 2(1+tan
2
)+(1+tan
2
)
2
= 3+4 tan
2
+tan
4
,
and det C = 1, giving
I
1
= 3 + tan
2
, I
2
= 3 + tan
2
, I
3
= 1.
Continuum Mechanics spring 2012 110
5 Conservations laws
deformation of a volume element
conservations of mass
material time derivative of a volume integral
conservation of linear momentum
conservation of angular momentum
conservation of energy
principle of virtual work*
Continuum Mechanics spring 2012 111
Finite deformation of a volume element
P
0
S
0
Q
0
R
0
X
(0)
x
(0)
X
(2)
P
S
Q
R
Consider a tetrahedron in the reference conguration such
that its vertices P
0
, Q
0
, R
0
and S
0
have position vectors
X
(0)
, X
(0)
+X
(1)
, X
(0)
+X
(2)
, and X
(0)
+X
(3)
,
respectively. The volume of the tertrahedron in the reference
conguration is
V =
1
3
hA =
1
3
X
(1)

1
2
(X
(2)
X
(3)
)
=
1
6
X
(1)
(X
(2)
X
(3)
)
=
1
6
e
RST
X
(1)
R
X
(2)
S
X
(3)
T
In the deformation, particles initially at P
0
, Q
0
, R
0
and S
0
move to P, Q, R and S, and have position vectors x
(0)
,
x
(0)
+x
(1)
, x
(0)
+x
(2)
and x
(0)
+x
(3)
.
The volume of the deformed tetrahedron PQRS is
v =
1
6
x
(1)
(x
(2)
x
(3)
) =
1
6
e
ijk
x
(1)
i
x
(2)
j
x
(3)
k
.
The deformation is dened by Eqs. of form x
i
= x
i
(X
R
, t). Thus, e.g.,
x
(1)
i
= x
i
(X
(0)
R
+X
(1)
R
, t) x
i
(X
(0)
R
, t) = X
(1)
S
x
i
X
S
(X
(0)
R
, t) +c(X
(1)
R
)
2
.
Continuum Mechanics spring 2012 112
Thus, the expression for v becomes
v =
1
6
e
ijk
x
i
X
R
x
j
X
S
x
k
X
T
X
(1)
R
X
(2)
S
X
(3)
T
+c(X
(p)
R
)
4
.
Using e
mpq
det A = e
ijk
A
im
A
jp
A
kq
, we can write
v =
1
6
e
RST
(x
1
, x
2
, x
3
)
(X
1
, X
2
, X
3
)
X
(1)
R
X
(2)
S
X
(3)
T
+c(X
(p)
R
)
4
,
where the Jacobian determinant (or Jacobian, for short) of the deformation,
(x
1
, x
2
, x
3
)
(X
1
, X
2
, X
3
)
=

x
1
X
1
x
1
X
2
x
1
X
3
x
2
X
1
x
2
X
2
x
2
X
3
x
3
X
1
x
3
X
2
x
3
X
3

= det F,
has been introduced. Proceeding to the limit X
(p)
R
0 (R, p = 1, 2, 3), we get
dv
dV
=
(x
1
, x
2
, x
3
)
(X
1
, X
2
, X
3
)
= det F.
If the material is incompressible, then det F = 1.
Continuum Mechanics spring 2012 113
Small deformation of a volume element
Expand det F:
det F = det
_

iR
+
u
i
X
R
_
=

1 +
u
1
X
1
u
1
X
2
u
1
X
3
u
2
X
1
1 +
u
2
X
2
u
2
X
3
u
3
X
1
u
3
X
2
1 +
u
3
X
3

= 1+
u
i
X
i
+c
_
u
i
X
R
_
2
Thus, in the case of small displacement gradients,
dv
dV
= det F 1 +
u
i
X
i
= 1 +E
ii
.
The quantity E
ii
is called the dilatation (dilataatio) and it is denoted by . is the trace of the
innitesimal strain tensor and so it is the rst invariant of that tensor. Thus,
= E
ii
= tr E = E
1
+E
2
+E
3
.
For small deformation, is the measure of the change of volume per unit initial volume of an element.
Continuum Mechanics spring 2012 114
Conservation of mass Lagrangian description
Suppose that the material in the volume element P
0
Q
0
R
0
S
0
has the mass m in the reference con-
guration. Conservation of mass requires that the mass of the material in the material volume element
remains constant during the deformation. Therefore, the initial and nal mass denisities,
0
and , are

0
= lim
V 0
m
V
, = lim
v0
m
v
.
Hence,

=
dv
dV
= det F
and this is the statement of the conservation of mass in the material description. This also justies the
assumption det F ,= 0, since in the opposite case, either
0
= 0 or .
Note that for a motion given in the Lagrangian form, x = x(X, t), the conservation of mass gives
(X, t) =
(X, 0)
det F(X, t)
,
which is an algebraic equation of density following a particle in a given motion.
Continuum Mechanics spring 2012 115
Conservation of mass Eulerian description
Consider the velocity eld v(x, t) and an arbitrary but xed region 1 in space. The mass inside the
region 1,
m
1
(t) =
___
1
(x, t) dV,
where the density is given in the spatial description. The mass changes at rate
dm
1
dt
=
___
1

t
dV.
The change of m
1
is due to motion of material particles in to or out of the region.
Consider a particle occupying the position x at time t. Its displacement
from t t to t is
u = x x(X(x, t), t t)
= x
_
x t
x(X, t)
t

X=X(x,t)
+c(t)
2
_
= v(x, t) t +c(t)
2
All particles on the material line from x u to the point x on the
surface S bounding 1 move through the surface in time t.
S
x(X,t)
u
t)
O
R
S
n
x(X,t
Continuum Mechanics spring 2012 116
The volume occupied by the particles traversing a small surface element S around x per time interval
t is ]u nS]. Thus,

dm
1
dt
=
__
S
v ndS,
where n is the outward normal of the surface S bounding the region 1. The minus sign is introduced
because the ow is out of the region if v n > 0.
We can now write
0 =
___
1

t
dV +
__
S
v ndS =
___
1
_

t
+ (v)
_
dV
1 arbitrary


t
+ (v) = 0 equation of continuity.
This partial dierential equation expresses the conservation of mass in the spatial description. It has the
following equivalent forms:

t
+v () +( v) = 0 or

t
+v
i

x
i
+
v
i
x
i
= 0
D
Dt
+( v) = 0 or
D
Dt
+
v
i
x
i
= 0
If the material in incompressible, then D/Dt = 0 or, equivalently,
v = 0 or
v
i
x
i
= 0; D
ii
= 0 or tr D = 0.
Continuum Mechanics spring 2012 117
The material time derivative of a volume integral
Let be some physical quantity (such as mass or energy) associated with the particles of the body, and
let be the amount of per unit mass. The amount of per unit volume is, thus, and the amount
of contained in a xed region 1 at a given time t is

1
(t) =
___
1
(x, t) (x, t) dV.
Thus,
d
1
dt
=
___
1
()
t
dV =
D
Dt
___
1
dV
__
S
v ndS.
The rst term on the right-hand side is due to the rate of change of the amount of , which is associated
with the particles that instantly occupy 1 at t, and it is called the material time derivative of
1
. (In
the case of mass, which is conserved, = 1 and Dm
1
/Dt = 0.) The second term gives the net
inux of to the region 1 due to particles crossing the surface. Thus,
D
1
Dt
=
___
1
_
()
t
+ (v)
_
.

t
+

t
+(v)+v
dV
=
___
1

t
+ (v)
_
.
=0
dV +
___
1

t
+v
_
.
=D/Dt
dV =
___
1
D
Dt
dV
Continuum Mechanics spring 2012 118
Conservation of linear momentum
Newtons 2nd law states that
D(mv)
Dt
= p,
where p is the resultant force acting on a particle with mass m and velocity v. For a continuum, the
law is generalized as follows:
The rate of change of linear momentum of the particles that instantanously lie within a xed region 1
is proportional to the resultant force applied to the material occupying 1. This resultant force consist
of the body forces (b per unit mass) acting on the particles in 1, together with the resultant of the
surface force (t
(n)
per unit area) acting on the surface of 1, i.e.,
D
Dt
___
1
v dV =
___
1
b dV +
__
S
t
(n)
dS
Using t
(n)
= n T, where n is an outward normal of S, and the divergence theorem gives
D
Dt
___
1
v dV
.
=
___
1
b dV +
__
S
n T dS =
___
1
(b + T) dV
=
___
1

Dv
Dt
dV =
___
1
f dV

___
1
(b + T f) dV = 0
1 arbitrary
b + T = f
Continuum Mechanics spring 2012 119
Conservation of angular momentum
For a particle, the angular momentum conservation law reads
D(mx v)
Dt
= x p,
where p is the resultant force applied and x is the position vector from an arbitrarily chosen origin. For
a continuum, an analogous law reads
D
Dt
___
1
x v dV =
___
1
x b dV +
__
S
x (n T) dS
Since x (n T) = e
ijk
e
i
x
j
(n T)
k
.
=n
p
T
pk
= e
ijk
e
i
n
p
x
j
T
pk
, we have
__
S
x (n T) dS =
___
1
e
ijk
e
i
(x
j
T
pk
)
x
p
dV.
Thus, as
D
Dt
___
1
x v dV =
___
1

D
Dt
(x v) dV , we have
___
1

D
Dt
(x v) dV =
___
1
x b dV +
___
1
e
ijk
e
i
(x
j
T
pk
)
x
p
dV

D
Dt
(x v) = x b +e
ijk
e
i
(x
j
T
pk
)
x
p
Continuum Mechanics spring 2012 120
Now, since
D
Dt
(x v) = (
D
Dt
x) v +x (
D
Dt
v) = v v +x f = x f and
(x
j
T
pk
)
x
p
=
jp
T
pk
+x
j
T
pk
x
p
= T
jk
+x
j
T
pk
x
p
,
we get
x f = x b +e
ijk
e
i
_
T
jk
+x
j
T
pk
x
p
_
= x b +e
ijk
e
i
T
jk
+x T
x (b + T f)
.
=0
+e
ijk
e
i
T
jk
= 0
e
ijk
T
jk
= 0 T
ij
= T
ji
Thus, conservation of angular momentum implies that the stress tensor is symmetric.
In writing
D
Dt
___
1
x v dV =
___
1
x b dV +
__
S
x (n T) dS
it was implicitly assumed that there are no body or surface couples (voimapari) acting on the material
in 1. (The only torque is that imposed by the resultant force acting on a region of the continuum.) If
such couples exist, stress tensor is no longer symmetric. We will not consider such couples.
Continuum Mechanics spring 2012 121
Conservation of energy
The kinetic energy K contained instantly in region 1 is
K =
1
2
___
1
v
i
v
i
dV.
The remainder of the total energy inside 1 is called internal energy E, and it is expressed in terms of
the internal energy density e by
E =
___
1
e dV.
The law of conservation of energy states: the material derivative of K + E is equal to the sum of the
rate of mechanical work by the body and surface forces acting on 1 and the rate at which other energy
enters 1.
In practice, the law becomes useful only if some properties of E or e are specied. This leads to the
consideration of constitutive relations.
Other energy may mean, e.g., heat ux across S, energy arising from chemical reactions inside 1,
energy arriving by radiation, or electromagnetic energy. Here we will only consider heat ux (vector) q.
Continuum Mechanics spring 2012 122
Thus,
D
Dt
___
1
(
1
2
v
2
+e) dV =
___
1
b v dV +
__
S
(n T) v n q] dS
=
___
1
b v dV +
__
S
n (T v q) dS

___
1

D(
1
2
v
2
+e)
Dt
dV =
___
1
b v + (T v q)] dV.
Thus,
D(
1
2
v
2
+e)
Dt
= b v + (T v q) = b v q +
(T
ij
v
j
)
x
i
Now, as Dv
2
/Dt = 2v v = 2vf and
i
(T
ij
v
j
) = v
j

i
T
ij
+T
ij

i
v
j
= v(T)+T
ij

i
v
j
,

De
Dt
= T
ij
v
j
x
i
q +v (b + T f) = T
ij
v
j
x
i
q
=
1
2
(T
ij
+T
ji
)
v
j
x
i
q =
1
2
T
ij
_
v
i
x
j
+
v
j
x
i
_
q = T
ij
D
ij
q
= tr (T D) q. Here, tr (T D) = T
ij
D
ij
is the rate of working by stress.
To make progress, one needs to assign some further properties to e and q, e.g., use an equation of state
of form e = e(, T), where T is the temperature, or Fouriers law q = T, where is constant.
These are not universal laws but depend on the materials studied.
Continuum Mechanics spring 2012 123
The principle of virtual work*
Consider a stress eld T(x), which satises the equilibrium equation
T +b = 0,
and a velocity eld v(x, t), both of which are determined inside a region 1. Let
D =
1
2
(v)
T
+v]
be the deformation-rate tensor. No connection between v and T is assumed; v can be any dierentiable
velocity eld and T any equlibrium stress eld.
Form the product T
ij
D
ij
and integrate over the region 1:
___
1
T
ij
D
ij
dV =
1
2
___
1
T
ij
_
v
i
x
j
+
v
j
x
i
_
dV =
___
1
T
ij
v
j
x
i
dV
=
___
1
_
(T
ij
v
j
)
x
i
v
j
T
ij
x
i
_
dV =
___
1
(T v) v ( T)] dV
=
__
S
n (T v) dS +
___
1
v (b) dV =
__
S
t
(n)
v dS +
___
1
b v dV
i.e., the rate of working of an equilibrium stress eld T in a velocity eld v is equal to the sum of the
rates of working of the surface and body forces associated with T in the same eld. This is the principle
of virtual work (PVW). An analogous law can be obtained by replacing v u and D E. PVW
is a basis of various variational theorems.
Continuum Mechanics spring 2012 124
6 Constitutive equations
constitutive equations
material symmetry
rotational symmetry
reectional symmetry
symmetry groups
linear elasticity
Newtonian viscous uids
viscoelasticity and plasticity
Continuum Mechanics spring 2012 125
Constitutive equations
Consider, e.g., a continuum in the Eulerian description. We derived the fundamental laws for its behavior
as
Du
Dt
= v
D
Dt
= v

Dv
Dt
= T +b, T
T
= T

De
Dt
= tr (T D) q, D =
1
2
(v)
T
+v].
There are altogether eight partial dierential equations, but many more unknowns: u
i
, v
i
, , T
ij
, e,
and q
i
give 3+3+1+6+1+3=17 unknowns in these equations. It is, therefore, clear that in order to get
a closed set of equations, we need to introduce more dependencies between the variables.
The remaining equations needed for closing the system depend on the material properties, and are
called constitutive equations. In general, they will also depend on thermodynamic quantities such as
temperature (which means that even larger number of unknowns than above are involved).
Continuum Mechanics spring 2012 126
Constitutive equations are either empirical laws or laws derived from idealizations (such as incompress-
ibility), which hold only approximatively for any real materials. The laws must be chosen pragmatically:
even if it was possible to formulate these equations for a general continuum, this is not very desirable as
the ultimate goal is to be able to solve the equations by a reasonable amount of work.
Thus, we resort to constitutive equations developed for ideal materials. A continuum may be treated
under dierent idealizations depending on what physical phenomena are of interest.
Historically, theories for classical ideal materials (linear elastic solids, Newtonian viscous uids, etc.) have
been developed separately. There are, however, some general properties the consitutive equations must
fulll. These include
dimensional homogeniety (all terms must have the same physical dimension)
independence of the coordinate system used relations between scalars, vectors and tensors.
state of stress should be unaected by superposed rigid-body motions (invariance under translations
and rotations)
Typically, materials are divided in solids and uids and the latter in liquids and gases. The division is not
always clear: some materials posses both solid-like and uid-like properties.
The characteristic property of a uid is its inability to support shearing stress indenitely. (If a shearing
stress is applied, the uid starts to move until the shearing stress no longer remains.) A solid can be
in equilibrium under shearing stress. Some solids, however, posses a natural conguration in which they
eventually return if a stress is imposed and then removed. Fluids have no natural conguration but adapt
to the shape of the vessel in which they are placed.
Continuum Mechanics spring 2012 127
Material symmetry
Most materials posses some material symmetry. The simplest (and perhaps the most common) case is
that when the material is isotropic, i.e., posses no preferred direction and has identical properties in all
directions. Examples of isotropic materials include: metals in polycrystalline form, concrete, sand in bulk,
and uids like water and air. An example of an anisotropic material is wood. Single crystals of crystalline
materials have directional properties arising from their crystal symmetry. Sometimes a material has one
preferred direction but is isotropic in all direction perpendicular to it. This kind of material is called
transversely isotropic.
Two kinds of material symmetries can be considered: rotational and reectional.
Continuum Mechanics spring 2012 128
Rotational symmetry
Consider

OP
0
= X and

OP = x in the gure below and let the two points be occupied by the same
material particle before and after a homogeneous deformation
x = F(t) X.
Consider another deformation, where the deformation eld is rst rotated (through and angle around
n). Thus, if the tensor Q describes the rotation, the point now subject to the same deformation as
P
0
in the rst deformation is Q
0
with the position vector Q X. After the deformation, clearly, the
material particle occupies point Q, given by
Q x = F(t) Q X.
Thus, P
0
OQ
0
= POQ = and the second deformation is given by the tensor Q
T
F(t) Q.
P
0
Q
0
P
Q
Continuum Mechanics spring 2012 129
The two deformations are distinct in general, but if the material posseses a suitable rotational symmetry,
they may give rise to the same stress response, only rotated by the same tensor Q. In this case, if F
gives rise to stress eld T, then the deformation tensor Q
T
F Qgives rise to stress eld Q
T
T Q.
The material is then said to have rotational symmetry (relative to a given reference conguration) for
the rotation determined by Q.
Example. Let the material posses a symmetry with respect to 90-degree rotations around the X
3
axis in
the reference conguration. The components of the tensor Q are, thus, given by
Q =
_
_
0 1 0
1 0 0
0 0 1
_
_
.
Thus, if T
(1)
= e
1
e
1
is the stress tensor required to produce a given extension in the X
1
direction,
then T
(2)
with components
T
(2)
=
_
_
0 1 0
1 0 0
0 0 1
_
_
_
_
0 0
0 0 0
0 0 0
_
_
_
_
0 1 0
1 0 0
0 0 1
_
_
=
_
_
0 0 0
0 0
0 0 0
_
_
gives the stress tensor required to produce the same extension in the X
2
direction.
Continuum Mechanics spring 2012 130
Reectional symmetry
Consider again the material point originally at P
0
(with position vector X) and moving to point P in
a homogeneous deformation,
x = F X.
Now, consider also the homogeneous deformation, which is the mirror image of F in some plane, say,
X
1
= 0. This deformation is dened by
(x
1
x
2
x
3
)
T
= F(X
1
X
2
X
3
)
T
or x = R
T
1
F R
1
X, where the tensor R
1
has components R
1
= diag(1, 1, 1). If the
material is symmetric with respect to this reection, then, if the state of stress produced by F is T, the
state of stress produced by the deformation R
T
1
F R
1
is R
T
1
T R
1
.
Q
0
P
0
P Q
Continuum Mechanics spring 2012 131
In a more general case, a reection with respect to a plane though O and normal to n is produced by
the tensor R = I 2nn. It is easy to see that Ris a symmetric improper orthogonal tensor (i.e.,
det R = 1).
A material has reectional symmetry for reection in the planes normal to n if the deformation
x = R
T
F R X
gives rise to the stress R
T
T R, when the deformation x = F X gives rise to stress T.
Example. Assume that the material has reectional symmetry with respect to X
1
= 0 plane. Consider
a simple shear in the X
2
direction,
x
1
= X
1
, x
2
= X
2
+X
1
tan , x
3
= X
3
(a)
produced by T = (e
1
e
2
+ e
2
e
1
). Thus, F
ij

ij
= tan for i = 2 and j = 1, and 0
otherwise. Now,
R
1
FR
1
=
_
_
1 0 0
tan 1 0
0 0 1
_
_
; R
1
TR
1
=
_
_
0 0
0 0
0 0 0
_
_
so the shear stress required to produce the simple shear
x
1
= X
1
, x
2
= X
2
X
1
tan , x
3
= X
3
is equal in magnitude but of opposite sign to the shear stress required to produce the simple shear (a).
Continuum Mechanics spring 2012 132
Symmetry groups
The set of tensors that dene the symmetry properties of a material form a group
4
, which is called the
symmetry group of the material. For an isotropic material, the symmetry group includes all rotations and
reections. It is, thus, the group of orthogonal tensors, i.e., the full orthogonal group in three dimensions,
O(3). A material whose symmetry group is the special orthogonal group SO(3), which only contains
rotations, is called hemitropic. The symmetry group of any material is a subgroup of the full O(3).
The symmetry group of a transversely isotropic material is the group consisting of rotations around a
xed axis, i.e., the group is isomorphic with SO(2).
A material which has reectional symmetry with respect to each of three mutually orthogonal planes is
called orthotropic. This symmetry group is nite, as it is composed of the unit tensor, three reection
tensors, and their inner products. An example of a nearly orthotropic material is wood.
Other nite subgroups of O(3) include symmetry groups for materials with various crystal symmetries.
The rotations belonging to these subgroups are rotations through angles that are multiples of
1
2
and
1
3
.
4
A set of tensors, ( , is a group if and only if
T, S ( : T S (
T, S, R ( : T (S R) = (T S) R
I (
T ( , T
1
( : T T
1
= T
1
T = I
Continuum Mechanics spring 2012 133
Linear elasticity
If a body undergoes only small deformations when subjected to stress, and if the body has also a normal
conguration where it returns if the stress is removed, we may apply the theory of linear elasticity to its
behaviour. Typical materials where the approximations hold include concrete, metals and wood.
Linear eleastic solid is dened as a material for which the internal energy per unit volume in the reference
conguration,
0
e, has the following properties:
(a) W
0
e is or may be approximated by a quadratic function of the components of the innitesimal
strain tensor E
ij
, i.e.,
W =
1
2
C
ijkl
E
ij
E
kl
where C
ijkl
are constants. We call W the strain energy function.
(b) if K is the kinetic energy and E is the internal energy of any region 1, then the material time
derivative of K + E is equal to the rate of mechanical work done by the surface and body forces
acting on 1. (Thus, heat ux should be negligible.)
Mechanical work done on a system either creates kinetic energy or is stored in the solid as potential
energy (called strain energy). The system is conservative so that in a closed cycle of deformation strain
energy is stored and then released as kinetic energy.
As E is not an exact measure of deformation, W has the unphysical property of W ,= const. for rigid-
body rotation, i.e., E
ij
= (1 cos )(n
i
n
j

ij
). Thus, linear elasticity is only an approximate
theory, but nevertheless an excellent approximation in many practical situations.
Continuum Mechanics spring 2012 134
Elastic constants
Consider another base e
i
= M
ij
e
j
, where M = (M
ij
) is an orthogonal matrix. Then,

E
ij
= M
ir
M
js
E
rs
, E
ij
= M
ri
M
sj

E
rs
.
The strain energy must also be expressible in terms of the components

E
ij
:
W =
1
2

C
ijkl

E
ij

E
kl
=
1
2
M
ip
M
jq
M
kr
M
ls

C
ijkl
.
=C
pqrs
E
pq
E
rs
C
pqrs
= M
ip
M
jq
M
kr
M
ls

C
ijkl
and, thus, C
ijkl
are components of a fourth-order tensor. The 3
4
= 81 components C
ijkl
are called
elastic constants. Their dimension is that of stress. Their values characterize particular elastic materials.
Elastic constants are not all independent. As E
ij
= E
ji
, we must have C
ijkl
= C
jikl
and C
ijkl
=
C
ijlk
. Also, as W =
1
2
C
ijkl
E
ij
E
kl
=
1
2
C
ijkl
E
kl
E
ij
, there has to be a symmetry C
ijkl
= C
klij
.
Thus, the number of independent elastic constants is reduced from 81 to 21.
Further requirement on C is that W =
1
2
E : C : E is positive for all E ,= 0.
5
Material symmetries reduce the number of independent elastic constants from 21 even down to 2 (isotropic
materials), as we shall see.
5
The fourth-order tensor of elastic constants is denoted by C for clarity. The contraction involving two indices at the time is denoted by :.
Continuum Mechanics spring 2012 135
Constitutive equation for a linear solid
Let us, then, apply the energy equation as it is written for in linear elasticity:
T
ij
D
ij
=
De
Dt
=

0
DW
Dt
(1 +E
kk
)
DW
Dt
= (1 +E
kk
)
W
E
ij
DE
ij
Dt

W
E
ij
D
ij
,
where only terms of the lowest order in E
ij
and D
ij
are kept, in accordance with the approximation
adopted in the theory. Thus,
T
ij
=
W
E
ij
=
(
1
2
C
pqrs
E
pq
E
rs
)
E
ij
=
_
1
2
C
pqrs
E
pq

ri

sj
+
1
2
C
pqrs

pi

qj
E
rs
_
=
1
2
(C
pqij
E
pq
+C
ijrs
E
rs
) =
1
2
(C
ijpq
E
pq
+C
ijrs
E
rs
) = C
ijrs
E
rs
,
so
T
ij
= C
ijrs
E
rs
or T = C : E
are the constitutive equations for a linear elastic solid. They contain up to 21 independent elastic
constants that are regarded as empirically determined quantities.
Alternatively, the constitutive equation could be taken as the starting point of the analysis for a linear
elastic solid. In this case, however, the index symmetry C
ijkl
= C
klij
cannot be obtained without
further assumptions. Furthermore, the theory would no longer be strictly conservative. Thus, formulation
starting from the strain energy is preferred.
Continuum Mechanics spring 2012 136
Material symmetries and C
As an example of the eect of material symmetries on C, consider reectional symmetry with respect to
the planes perpendicular to X
1
. Thus, R = I 2e
1
e
1
with components R
ij
=
ij
2
i1

j1
belongs to the symmetry group of the material. Consider a deformation given by
F = F
iR
e
i
e
R
, E =
1
2
(F +F
T
) I,
which creates the stress components T
ij
= C
ijkl
E
kl
. Another deformation determined by
F

= R
T
F R, E

=
1
2
(F

+F
T
) I = R
T
E R,
i.e., E

ij
= R
ki
R
lj
E
kl
, thus, generates the stress components T

ij
= C
ijkl
E

kl
= C
ijkl
R
rk
R
sl
E
rs
.
However, as R
ij
belongs to the symmetry group, T
pq
= R
pi
R
qj
T

ij
= R
pi
R
qj
R
rk
R
sl
C
ijkl
E
rs
=
C
pqrs
E
rs
so
C
pqrs
= R
pi
R
qj
R
rk
R
sl
C
ijkl
.
Now, R
ij
=
ij
, where the sign applies if i = 1 and the + sign otherwise. Thus,
C
pqrs
= C
pqrs
,
where the sign applies for those components, which have an odd number of 1s in their indices. Thus,
C
1112
= C
1113
= C
1222
= C
1223
= C
1233
= C
1322
= C
1323
= C
1333
= 0
which, thus, reduces the number of independent elastic constants by 8 to 13.
Continuum Mechanics spring 2012 137
Elastic constants for isotropic materials
The symmetry group for isotropic materials includes all orthogonal tensors, but it suces to consider
rotations Q. Thus, as above, the condition for the elastic constants is
C
pqrs
= Q
pi
Q
qj
Q
rk
Q
sl
C
ijkl
,
but now Q
ij
is an arbitrary rotation. However, the statement is clearly that the tensor components
are invariant under rotations. Thus, C is an isotropic fourth-order tensor. The most general isotropic
fourth-order tensor has components
C
ijkl
=
ij

kl
+
ik

jl
+
il

jk
,
but as C
ijkl
= C
ijlk
=
ij

kl
+
il

jk
+
ik

jl
, we need to set = so the constitutive
equation becomes
T
ij
=
ij
E
kk
+ 2E
ij
or T = Itr E + 2E.
Thus, the number of elastic constants for isotropic materials is reduced from 21 to two.
Continuum Mechanics spring 2012 138
Newtonian viscous uids
Experimentally, in many uids (e.g., water, air) in a simple shearing ow the shearing stress on the shear
planes is proportional to the shear rate over a wide range of the shear rate. This behavior is characteristic
of a Newtonian viscous uid (or linear viscous uid).
Consider uids with constitutive equation
T
ij
= p(, )
ij
+B
ijkl
(, ) D
lk
,
where is the temperature. For a uid at rest, D
ij
= 0, so
T
ij
= p(, )
ij
,
which is the consitutive equation already employed in hydrostatics, with p(, ) representing the hy-
drostatic pressure. Thus, for a uid in motion, the additional stress is linear in the components of the
strain-rate tensor.
If the uid is isotropic, we can argue (like in the case of an elastic solid) that the quantities B
ijkl
are
the components of a fourth-order isotropic tensor B, and thus,
T
ij
= (p +D
kk
)
ij
+ 2(, ) D
ij
or T = (p +tr D)I + 2D,
where the viscosity coecients and (and the hydrostatic pressure p) are functions of and . These
are, of course, not the same as the elastic constants and of isotropic elastic solids. The viscosity
coecients are characteristic for each uid.
Continuum Mechanics spring 2012 139
There is one important dierence between the theory of linear elasticity and the theory of Newtonian
viscous uids. We showed earlier that, unlike the innitesimal strain tensor E, the strain-rate tensor D
is unaected by superposed rigid-body motions. Thus, the consitutive equation
T = (p +tr D)I + 2D,
has the required propoerty of being independent of superposed rigid-body motions, and the equation is
a possible exact constitutive equation for a viscous uid.
Some special cases:
If the state of stress is pure hydrostatic pressure, then
T =
1
3
tr TI = p + ( +
2
3
) tr D]I
If, additionally, the stress only depends on and and not on the dilatation rate tr D, one needs
to adopt +
2
3
= 0.
If the material in inviscid, then = = 0 and T = p(, ) I. The stress in an inviscid uid
is always hydrostatic.
If the uid is incompressible, then is constant, D
kk
= v 0 and T = pI +2()D.
Incompressibility is a kinematic constraint, which gives rise to a reaction stress. This reaction is an
arbitrary hydrostatic pressure, which can be superimposed on the stress eld without causing any
deformation. It can not be specied by a constitutive relation but can only be found as a result of
integrating the equations of motion.
An inviscid and incompressible uid is called an ideal uid. It has T = pI, where p is arbitrary.
Continuum Mechanics spring 2012 140
Viscoelasticity. Creep and stress relaxation
Many materials (e.g., plastics) possess some properties of both elas-
tic solids and viscous uids. These are called viscoelastic materials.
Viscoelasticity can be illustrated by creep and stress-relaxation ex-
eriments.
Consider a simple tension. Assume that a tension F
0
is instanta-
neously applied to an intially stress-free viscoelastic string at t = 0
and then held constant. The corresponding relation between elon-
gation e and t may be of form, where an initial elongation e
0
is fol-
lowed by an increasing elongation, when the load is maintained. The
phenomenon is called creep (viruminen l. hitaasti eteneva muodon-
muutos).
If the material is a viscoelastic solid, the elongation tends to a -
nite limit e

; if the material is a viscoelastic uid, the elongation


continues indenitely.
Alternatively, assume that the string is given an initial elongation
e
0
and then held in this position. The resulting force response rises
instantaneously to F
0
at t = 0 and then decays. This phenomenon
is called stress relaxation. For a uid, F 0 as t 0, but for a
solid, F F

> 0.
m
e
F
0
e
0
F
e
t t
e
0
F
0
t t
e F
e
Continuum Mechanics spring 2012 141
Linear viscoelasticity*
Consider only innitesimal deformations, so that we can use E
ij
to describe the deformation. Assume
that innitesimal increments E
ij
in the strain components at time give rise to increments T
ij
in
the stress components at subsequent times t so that the magnitude of these stresses depends on t :
T
ij
= G
ijkl
(t )E
kl
,
where we expect the functions G
ijkl
to decrease as a function of t > 0 and G
ijkl
= 0 for t <
(causality). Assume also that the superposition principle holds, so that
T
ij
=

<t
T
ij
=

<t
G
ijkl
(t ) E
kl
() =

<t
G
ijkl
(t )
E
kl
()

0
T
ij
=
_
t

G
ijkl
(t )
dE
kl
()
d
d.
This is the constitutive relation for linear viscoelasticity. The functions G
ijkl
are called the relaxation
functions. If E
ij
0 at , we can integrate by parts to get
T
ij
= G
ijkl
(0)E
kl
(t)
_
t

E
kl
()
dG
ijkl
(t )
d
d.
Note that by taking the relaxation functions to be constants, the theory of linear elasticity is recovered
(with G
ijkl
= C
ijkl
).
Continuum Mechanics spring 2012 142
The stress-relaxation functions G
ijkl
(t ) have index symmetries i j and k l, but not
ij kl, unless this is introduced as another assumption. If the material is isotropic, then G
ijkl
are
components of an isotropic fourth-order tensor and the constitutive relation reduces to
T
ij
=
ij
_
t

(t )
dE
kk
()
d
d + 2
_
t

(t )
dE
ij
()
d
d,
and only two relaxation functions (t ) and (t ) are needed to describe the material. A
Newtonian viscous uid can be recovered from this constitutive equation by assuming that (t ) =
(t ) and (t ) = (t ).
The inverse of the constitutive relation of a linear viscoelastic material is
E
ij
(t) =
_
t

J
ijkl
(t )
dT
kl
d
d,
where the functions J
ijkl
(t ) are called the creep functions; they have the same index symmetries
as the relaxations functions and are components of a fourth-order isotropic tensor in case the material is
isotropic.
Linear viscoelasticity has the same limitations as linear elasticity. It is, thus, an approximate theory and
only applicable when the components of strain and rotation are small.
Continuum Mechanics spring 2012 143
Plasticity
Many materials (metals in particular) conform well to the linear theory of elasticity, provided that the
stress does not exceed certain limits. But if it does, the solid acquires a permanent deformation that
does not vanish when the stress is removed. The phenomenon is called plasticity. This inelastic behavior
is not viscoelastic, because viscoelastic stress depends on the rate of deformation. In metals, the stress
depends on the history of deformation but is independent on the rate at which the deformation took
place.
Example. For the deformation corresponding to OA, the
axial stress is proportional to the the axial strain . If
the stress is removed before it reaches the value
A
, the
strain vanishes as well, but if the stress is increased to

B
>
A
and then removed, the unloading follows the
curve BC j OA. The limiting value
A
is called initial
yield stress. There remains a residual strain (OC) even when
the stress is removed. This an example of plastic deforma-
tion. On reloading, the path will retrace CB and continue
the curve OB.

A
B

C
To formulate a three dimensional theory of plasticity we require
a yield condition, which decides whether an element of material is behaving elastically or plastically
stressstrain relations for elastic behavior, and
stressstrain relations for plastic behavior
Continuum Mechanics spring 2012 144
Yield condition*
The yield condition is an equality of form
f(T
ij
) k
2
,
where f(T
ij
) is the yield function and k is a parameter that, in general, depends on the deformation
history. If f(T
ij
) < k
2
, the material behaves elastically; if f(T
ij
) = k
2
, plastic deformation may
occur. The equation f(T
ij
) = k
2
can be interpreted as giving a surface (the yield surface) in the 6-D
space of the stress components T
ij
. Plastic stresses lie on the surface, elastic stresses in its interior, and
stress states outside it are not attainable for the current value of k.
Any material symmetry restricts the form of f(T
ij
). For example, for an isotropic material, it has to be
expressible as a function of the stress invariants J
1
, J
2
, and J
3
.
For many materials the yield point for shear stress is much lower than for normal stress. Thus, the
hydrostatic part of the stress tensor has only a small eect the yield condition, and for such materials,
the yield function is well approximated as a function of the invariants, J
t
1
and J
t
2
, of the stress deviator
S
t
= T
1
3
I tr T.
Continuum Mechanics spring 2012 145
gure source: Wikipedia
Continuum Mechanics spring 2012 146
Stress-strain relations*
Elastic stress-strain relations. Before any plastic deformation has occurred, the usual elastic relations
apply. For example, for a small deformation in an isotropic material
T
ij
=
ij
E
kk
+ 2E
ij
.
For small deformations following a plastic deformation, the relation between T
ij
and E
ij
is still linear,
but there is a residual strain E
(0)
ij
depending on the on the previous deformation history, i.e.,
T
ij
=
ij
(E
kk
E
(0)
kk
) + 2(E
ij
E
(0)
ij
).
The residual strains can be avoided by considering stress and strain increments or rates:
T
ij
=
ij
E
kk
+ 2E
ij
or

T
ij
=
ij

E
kk
+ 2

E
ij
=
ij
D
kk
+ 2D
ij
.
The inverse of the latter is D
ij
=
1
2

T
ij


2(3 + 2)

T
kk

ij
. (D)
Plastic stress-strain relations. The classical approach is to assume that the rate of deformation can be
decomposed into an elastic and a plastic part, D
ij
= D
(e)
ij
+D
(p)
ij
, and that the elastic part is given by
(D). For the plastic part, the simplest theory assumes D
(p)
ij
=

f/T
ij
, where

is a scalar factor
that depends on the history of deformation. Thus, for an isotropic plastic material,
D
ij
=
1
2

T
ij


2(3 + 2)

T
kk

ij
+

f
J
t
k
J
t
k
T
ij
.
Continuum Mechanics spring 2012 147
7 Hydrodynamics
Hydrostatics
Perfect uids
Eulers equation
Bernoullis equation
HelmholtzKelvin theorem
Potential ow
Fluids with viscosity
Continuum Mechanics spring 2012 148
Hydrostatics
For Newtonian uids
T
ij
= (p +D
kk
)
ij
+ 2D
ij
.
Consider a uid at rest. Thus, T
ij
= p
ij
and the equilibrium equation becomes
b = T = p or b
i
=
1

p
x
i
.
The theory of equilibrium uids is called hydrostatics.
Introduce the function 1 such that
d1 =
dp

or 1 =
_
dp

,
where the integral can be performed once the dependence of p on is known.
As p and, therefore, 1 can be given as functions of position, 1 = 1(x) and p = p(x), we have
d1 = dr 1 and dp = dr p, so
1 =
p

so 1 = b.
Thus, equilibrium is possible only if the external force can be given as a gradient of a potential, i.e.,
b = U, whence the equilibrium equation becomes
(1 +U) = 0 or 1 +U = const.
Continuum Mechanics spring 2012 149
Incompressible uids
If can be regarded constant, like in many practical applications, then 1 = p(x)/. Thus,
p

+U = const.
Thus, the surfaces of constant pressure are equipotential surfaces.
Example 1. Consider U as being due to a constant gravity eld g in a liquid at depth z below a free
surface, i.e., U = gz. Thus,
p gz = p
0
= const.
where p
0
is the atmospheric pressure. Thus,
(i) at equal depths there is an equal pressure in every part of a vessel, or interconnected vessels, regardless
of their shapes and sizes;
(ii) for a given pressure, the height required for dierent liquids is dierent.
Example 2. Consider a liquid in a rotating drum (angular velocity ). The potential now includes both
gravity and centrifugal acceleration, i.e.,
U = gz
1
2

2
r
2
p = p
0
+g(z z
0
) +
1
2

2
r
2
]
The surfaces of equal pressure are paraboloids of rotation and for the free surface (p = p
0
),
z = z
0

1
2

2
r
2
.
Continuum Mechanics spring 2012 150
Compressible, isothermal uids
For a compressible uid, we need a suitable equation of state giving the pressure as a function of density.
The most simple one is that of an ideal gas,
p = k
B
/m,
where k
B
is Boltzmanns constant and mis the mean molecular mass of the uid. Thus, for an isothermal
uid ( = const.), we get p = c with c = const.
Thus,
1 =
_
dp

=
k
B

m
_
dp
p
=
k
B

m
ln p.
Example. Consider, again the rotating drum with the z axis now upwards,
U = gz
1
2

2
r
2
.
Thus,
ln p +
m
k
B

(gz
1
2

2
r
2
) = ln p
0
where p
0
is the constant atmospheric pressure (i.e., the free surface now intersects the z axis at z = 0).
Thus,
p = p
0
exp
_

m
k
B

(gz
1
2

2
r
2
)
_
The surfaces of constant pressure are still paraboloids and free surface is given by gz
1
2

2
r
2
= 0.
Continuum Mechanics spring 2012 151
Applications
Centrifuge. Assume that the drum is spinning so fast that the paraboloids of constant pressure become
almost at. Thus,
p p
0
exp
_
m
2
r
2
2k
B

_
The pressure increases quickly as a function of radius. At a given r, the pressure depends on the molecular
mass. Thus, if the gas is a mixture of two species of dierent masses m
1
and m
2
, each of them has a
pressure
p
i
= p
0i
exp
_
m
i

2
r
2
2k
B

_
; i = 1, 2
and clearly the pressure (and thus also the density) of the heavier species increases faster.
Barometric formula. If, on the other hand, 0, we get
p = p
0
exp
_

mgz
k
B

_
,
i.e., the pressure (and thus, the density) of an isothermal atmosphere decreases exponentially as a function
of height. The density can be written as
=
0
exp
_

mgz
k
B

_
=
0
exp
_

U
k
B

_
,
which is the Boltzmann density distribution familiar from statistical mechanics.
Continuum Mechanics spring 2012 152
Perfect uids Eulers equation
Consider a frictionless uid, i.e., a one with zero viscosity coecients. Thus, T = pI and the
equation of motion becomes

Dv
Dt
= p +b or
Dv
i
Dt
=
p
x
i
+b
i
This equation was put forward by Leonhard Euler already in 1755. It is known as Eulers equation of
motion for perfect uids.
There are too many unknowns (, v, p) for Eulers equation to form a closed problem. In the simplest
case of incompressible uid, = const., one has four unknowns (v
i
and p) and v = 0 obtained
from the conservation of mass providing the fourth equation to close the set.
Eulers equation of motion can, however, be used also in case of compressible uids. Then, one usually
regards and v
i
to be the unknowns, uses the conservation of mass in form

t
+ (v) = 0
and introduces an equation of state, p = p(), to eliminate p from the equation of motion.
As
Dv
Dt
=
v
t
+v v,
Eulers equation is non-linear and, therefore, possible to solve in specialized cases, only.
Continuum Mechanics spring 2012 153
Bernoullis equation
We may write (exercise)
v v =
1
2
v
2
v (v).
For irrotational ow, v = 0, we can write v = . Assuming that the body force is
conservative, b = U, the Euler Eq. becomes

()
t
+
1
2
v
2
+
1

p +U = 0.
Using, as before, 1 =
_
dp/, we have

t
+
1
2
v
2
+1 +U
_
= 0
For a steady ow (/t = 0), we may integrate the equation as
1
2
v
2
+1 +U = constant,
which for an incompressible uid (1 = p/) becomes
1
2
v
2
+
p

+U = constant.
This is known as the Bernoullis equation or pressure equation for a steady ow.
Continuum Mechanics spring 2012 154
Flow in a horizontal pipe with variable cross section
Consider a case, where U is due to gravitational forces. Restrict to the ow in a horizontal pipe, whence
U = constant. Thus,
1
2
v
2
+p/ = constant.
Furthermore, conservation of mass implies v = 0. Integrate the equation over an arbitrary piece of
the pipe, allowing the ends of the piece to have a dierent cross-section.
Thus,
__
S
(v n)dS = 0.
The only non-zero contributions to the integral come from the ends of the piece, whence

__
S
1
(v n)dS =
__
S
2
(v n)dS.
The ow is, to a good approximation, perpendicular to the cross-sectional area of the tube, so this yields
v
1
A
1
= v
2
A
2
,
where A
i
is the area of the surface S
1
. Thus,
p
2
p
1
=
1
2
(v
2
1
v
2
2
) =
1
2
(A
2
2
/A
2
1
1)v
2
2
and the pressure is, therefore, the higher the larger the
local cross-sectional area of the tube.
A
2
A
1
Continuum Mechanics spring 2012 155
HelmholtzKelvin theorem on vorticity
Circulation associated with a closed material curve c is dened as
=
_
c
v dx.
Making use of Stokes theorem, we get
=
__
S
(v) ndS =
__
S
w ndS,
where S is any surface encolsed by c. (Cf. =
__
S
B ndS in electrodynamics.)
Now, let the motion of the uid be described by the deformation gradient tensor F. Thus,
dx = F dX
and
=
_
c
(v F) dX,
where the curve, now expressed in terms of the material coordinates, does not depend on time. Therefore,
D
Dt
=
_
c
D
Dt
(v F) dX =
_
c
Dv
Dt
dx +
_
c
_
v
DF
Dt
_
dX.
Continuum Mechanics spring 2012 156
In the latter term
_
v
DF
Dt
_
dX = v
i

t
x
i
X
R
dX
R
= v
i
v
i
X
R
dX
R
=
1
2
d(v
i
v
i
)
and because the intergral is over a closed curve, it vanishes. Thus,
D
Dt
=
_
c
Dv
Dt
dx.
Now, applying Eulers equation in form
Dv
Dt
=
p

+b = (1 +U)
gives
D
Dt
=
_
c
(1 +U) dx = 0.
This result, valid for perfect uids with conservative external forces, can be expressed as
1. vortices can neither be created nor destroyed; and
2. the circulation along any closed material curve is constant in time,
and it is known as the HelmholtzKelvin theorem on vorticity. Thus, if the ow is irrotational at any
time, then it remains so at all times and can be expressed in terms of a velocity potential.
Continuum Mechanics spring 2012 157
Potential ow
In case v = 0 at any moment in a perfect uid, it remains so and the velocity can be given as
v =
For an incompressible ow,
v = 0
2
= 0,
i.e., the potential fullls Laplaces equation. The Laplacian can be written as

2
(x
i
) =

x
2
i
;
2
(r, , z) =
1
r

r
_
r

r
_
+
1
r
2

2
+

2

z
2
;

2
(r, , ) =
1
r
2

r
_
r
2

r
_
+
1
r
2
sin

_
sin

_
+
1
r
2
sin
2

2
.
Laplaces equation in an irrotational perfect uid holds everywhere except possibly on singular points
or lines. Thus, the problems of potential ow are very similar to problems in electrostatics (where

ES
= 0 except where charges are present).
The physically acceptable solutions of Laplaces equation need to fulll the approriate boundary conditions
at the surfaces of solid objects that the uid passes. For a perfect uid, the normal component of the
uid velocity must vanish at the surface of stationary solid objects, and agree with the normal component
of the solid velocity on a surface of a moving solid body.
Continuum Mechanics spring 2012 158
Sources and sinks
In analogy with the electrostatic potential of a point charge,
ES
=
(q/4
0
)r
1
, we express the velocity potential as
= (m/4)r
1
,
where r denotes the distance from the point , and m is constant.
Clearly,
2
= 0, when r ,= 0. Thus,
v = =
m
4r
2
e
r
,
i.e., the ow is radially outwards from , and the mass ow [kg s
1
]
through any spherical surface centered around is
4r
2

0
v
r
= m
0
.
Thus, the singular point is a source and m represents its strength.
A sink is a negative source (denoted by ). If a source is a point fo
creation of the uid, then sink is a point of annihilation. The ow is
directed radially inwards and the potential (for a sink of strength m) is
given by
= (m/4)r
1
.
Continuum Mechanics spring 2012 159
Flow near an innite plate
The method of images, familiar from electrostatics, can be used
to determine the velocity potential near an innite plate. Con-
sider a point source of strength m at x = ae
1
and the
plate at x
1
= 0. The eect of the plate can be obtained by
placing an image source of the same strength at x = ae
1
.
Thus,
=
m
4
_
1
r
1
+
1
r
2
_
with
r
1
= [(x
1
+a)
2
+x
2
2
+x
3
3
]
1/2
and r
2
= [(x
1
a)
2
+x
2
2
+x
3
3
]
1/2
being the distance from the source and the image, respectively.
r
2 r
1
m m
source image
x
O
Therefore,
v
1
=

x
1
=
m
4r
2
1
dr
1
dx
1
+
m
4r
2
2
dr
2
dx
1
=
m
4
_
x
1
+a
r
3
1
+
x
1
a
r
3
2
_
Thus, at x
1
= 0, v
1
= 0 and theres no ow along the x
1
direction. The surface of the plate is a
stream surface.
Continuum Mechanics spring 2012 160
Fluids with viscosity
The constitutive equation for a Newtonian uid reads
T = (p +tr D)I + 2D,
where D
ij
=
1
2
(v
i
/x
j
+v
j
/x
i
) is the rate-of-strain tensor. The equation of motion is

Dv
Dt
= T +b,
where b = U is the (conservative) external bulk force. As tr D = v, we have
( T)
j
=

x
j
(p + v) +

x
i
_
v
i
x
j
+
v
j
x
i
_
=

x
j
(p + v) +
_

x
j
( v) +
2
v
j
_
so T = p + ( +)( v) +
2
v. Thus,

Dv
Dt
= p + ( +)( v) +
2
v +b
which is the celebrated NavierStokes equation of motion of real uids.
Continuum Mechanics spring 2012 161
Bulk and shear viscosity
The mean pressure of a Newtonian uid, giving the spherical part of the stress tensor as pI, is
p
1
3
tr T = p
1
3
(3 + 2)tr D = p ( +
2
3
) v
Using the equation of continuity, D/Dt = v, we get
p = p +
+
2
3

D
Dt
p +
b

D
Dt
where b +
2
3
is called the bulk viscosity. It is clearly related to viscous forces involving compressive
motions. It is very dicult to determine experimentally (usually done by analysing ultrasonic absorption
measurements).
The deviatoric part of the stress tensor is
S = 2(D
1
3
Itr D)
and this is related to the shearing motions. Thus, the viscosity coecient is called the shear viscosity.
Continuum Mechanics spring 2012 162
Some laminar NS ows
We will, next, consider two examples of laminar ow in a real uid. In order to avoid complexities, we
assume incompressibility, v = 0, and neglect external forces. Thus,

Dv
Dt
= p +
2
v
Example 1. Consider the steady-state ow in a capillary tube with circular cross-section (radius a). Take
the ow velocity to be along the axis of the tube, i.e., v = v
z
e
z
. For symmetry reasons v
z
cannot
depend on the angle measured around the cylider and because of the incompressibility, it cannot depend
on z. Thus, v
z
= v
z
(r) and
Dv
Dt
=
v
t
+v v = 0
Therefore,
p =
2
v =

r

r
_
r
v
z
r
_
e
z
.
Thus, the pressure is a function z, only, and since v
z
depends on r, only, we have
dp
dz
=

r
d
dr
_
r
dv
z
dr
_
,
which is of course possible only if both sides are equal to a constant. From the left-hand side, this
constant is seen to be (p
2
p
1
)/l = p/l, where p
1
and p
2
are the pressure at z = 0 and
z = l, respectively.
Continuum Mechanics spring 2012 163
We get
d
dr
_
r
dv
z
dr
_
=
p
l
r r
dv
z
dr
= c
1

p
2l
r
2
v
z
= c
0
+c
1
ln r
p
4l
r
2
,
where c
i
are constants of integration. As the velocity must remain nite at r = 0, c
1
= 0, and because
v
z
= 0 at r = a, we get
c
0
=
p
4l
a
2
v
z
=
p
4l
(a
2
r
2
)
The maximum speed is, therefore, v
max
= p a
2
/4l. The average speed is
v
av
=
_
a
0
v
z
2r dr
_
a
0
2r dr
=
v
max
a
2
_
a
0
r(a
2
r
2
)dr
_
a
0
r dr
=
v
max
2
and the rate of mass ow through the cross-sectional area of the tube, or the discharge, is
Q =
_
a
0
v
z
2r dr =
v
max
a
2
2
=
a
4
p
8l
.
Thus, the discharge is proportional to the fourth power of the radius of the tube. This result is known
as the HagenPoiseuille formula.
Note that the result is experimentally valid only up to a certain velocity v
max
after which the ow
becomes turbulent.
Continuum Mechanics spring 2012 164
Example 2. Consider incompressible uid between two parallel plates at x = 0 and x = h. Assume
that the upper plate is moving at speed V along the z axis while the lower plate is at rest. Now, the
ow between the plates can be assumed to be along the z axis and independent on coordinates y and
z. Thus, v
z
= v
z
(x) and
Dv
Dt
=
v
t
+v v = 0
so
p =
2
v =

2
v
z
x
2
e
z
,
where, again, the pressure can depend only on z. Thus,
dp
dz
=
d
2
v
z
dx
2
where both sides must be equal to a constant. This integrates to give
v
z
= c
0
+c
1
x +
x
2
2
dp
dz
.
Since v
z
= 0 at x = 0, c
0
= 0, and since v
z
= V at x = h, we get the result
hc
1
= V
h
2
2
dp
dz
v
z
=
V x
h
+
x
2
hx
2
dp
dz
known as the generalized Couette ow. (Simple Couette ow is obtained by setting p = const.)
Continuum Mechanics spring 2012 165
8 Isotropic linear elasticity
Isotropic linear elastic solids
Equilibrium equation
Elastic moduli
Examples of static solutions
Stress in a long cylindrical pipe
Cantilever beam under its own gravity
Continuum Mechanics spring 2012 166
Isotropic linear elastic solids
In linear elasticity of isotropic materials, the constitutive equation is given in form
T
ij
= E
kk

ij
+ 2E
ij
or T = Itr E + 2E, (1)
where E is the inntesimal strain tensor with the components
E
ij
=
1
2
_
u
i
x
j
+
u
j
x
i
_
,
u(x, t) is the displacement vector, and and are the so-called the Lames elastic constants.
In this chapter, we will consider some applications of Eq. (1).
The equation of motion reads

Dv
Dt
= T +b,
where b is an external bulk force.
Let us start by considering the equilibrium equation for linear isotropic elastic solids.
Continuum Mechanics spring 2012 167
Equilibrium equation
In a static equilibrium, v = 0, and using the constitutive equation for linear isotropic solids, the
equilibrium equation becomes
(Itr E + 2E) +b = 0.
Clearly, tr E = E
kk
= u
k
/x
k
= u,
( E)
j
=
E
ij
x
i
=
1
2

x
i
_
u
i
x
j
+
u
j
x
i
_
=
1
2
_

x
j
( u) +
2
u
j
_
and, thus,
(I u + 2E) = ( +)( u) +
2
u
so the displacement in equilibrium is determined by
( +)( u) +
2
u +b = 0.
We will consider the solution of the equilibrium equation in two cases below.
Continuum Mechanics spring 2012 168
Elastic moduli
The consititutive equation of a linear elastic solid can be inverted (exercise) as
2E
ij
= T
ij


3 + 2
T
kk

ij
.
Consider uniaxial tension, T
11
,= 0 and T
ij
= 0 for other i, j. Thus (exercise),
E
11
=
+
(3 + 2)
T
11
; E
22
= E
33
=

2(3 + 2)
T
11
and other E
ij
= 0.
Thus, we can write
T
11
= EE
11
; E
22
= E
33
= E
11
,
where E is elastic modulus related to a uniaxial tension, called Youngs modulus.
The Poissons ratio relates the other strain tensor components to E
11
in uniaxial tension.
In terms of the Lames elastic constants (i.e., assuming isotropy), they may be written as
E =
3 + 2
+
, =

2( +)
or =
E
(1 2)(1 +)
, =
E
2(1 +)
.
Continuum Mechanics spring 2012 169
The quantity relating uniform hydrostatic compression, T = pI, to the corresponding fractional
change in volume is called bulk modulus
B =
p
V/V
=
p

,
where = tr E is the dilatation.
Thus, for isotropic materials,
p
ij
= T
ij
=
ij
+ 2E
ij
showing that E
ij
= 0 for i ,= j.
Taking a trace of both sides we get
3p = (3 + 2) B = +
2
3
=
E
3(1 2)
.
(Cf. bulk viscocity.)
The compressibility is dened as
K =
1
B
=
V
pV
=
3
3 + 2
=
3(1 2)
E
.
Continuum Mechanics spring 2012 170
Consider, nally, the situation where the solid is subjected to a simple-
shear deformation, E
12
= E
21
=
1
2
tan
1
2
, all others = 0.
The ratio of the shear stress and the total angular change produced
by the stress is called the shear modulus
G =
T
12

=
T
12
2E
12
isotr.
=
2E
12
2E
12
= .
x
2
x
1

x
X
u
Shear modulus is also known as the modulus of rigidity or the torsion modulus.
Table. Values of elastic moduli in some
materials; data from Simmons and Wang,
Single Crystal Elastic Constants and Cal-
culated Aggregate Properties, a Hand-
book, MIT Press: Cambridge, 1971.
E[GPa] G[GPa] B[GPa]
Potassium 4.5 1.7 3.6 0.29
Lead 28.1 10.1 42.7 0.39
Copper 127.7 47.6 384.0 0.06
Diamond 1028.4 472.2 417.0 0.09
Continuum Mechanics spring 2012 171
Stress in a long cylindrical pipe
The aim is to determine the stress in a long cylindrical pipe
with inner and outer radius of R
1
and R
2
, for a given (high)
value of pressure inside.
We will neglect the external gravitational force and the air
pressure outside the pipe as very small compared to the
pressure inside. Thus, the boundary conditions in cylindrical
cooridnates are
x
3
x
2
x
1
R
1
R
2
T
rr
=
_
p at r = R
1
0 at r = R
2
For symmetry reasons u
r
= u
r
(r); u

= u
z
= 0.
The equilibrium equation ( +)( u) +
2
u+b = 0 can be simplied, since b = 0 and
u =
1
r

r
(ru
r
);
2
u = ( u) (u); u = 0
0 = ( + 2)( u) = ( + 2)e
r

r
_
1
r
(ru
r
)
r
_

1
r
(ru
r
)
r
= 2a u
r
= ar +
b
r
,
where a and b are constants.
Continuum Mechanics spring 2012 172
Since u
z
= u

= 0, we have
u
1
= u
r
cos ; u
2
= u
r
sin
and, therefore,
E
11
=
u
1
x
1
=
r
x
1
u
r
r
cos +u
r
cos
x
1
=
x
1
r
u
r
r
cos +u
r

x
1
x
1
r
=
u
r
r
cos
2
+u
r
r x
2
1
/r
r
2
=
u
r
r
cos
2
+
u
r
r
sin
2

E
22
=
u
2
x
2
=
r
x
2
u
r
r
sin +
sin
x
2
u
r
=
u
r
r
sin
2
+
u
r
r
cos
2

E
12
=
1
2
_
u
1
x
2
+
u
2
x
1
_
=
1
2
_
r
x
2
u
r
r
cos +
cos
x
2
u
r
+
r
x
1
u
r
r
sin +
sin
x
1
u
r
_
=
_
u
r
r

u
r
r
_
cos sin
and other Cartesian components are zero.
Continuum Mechanics spring 2012 173
Using
e
1
= e
r
cos e

sin and e
2
= e
r
sin +e

cos
the strain tensor can, then, be given as
E = E
ij
e
i
e
j
= E
11
(e
r
cos e

sin ) (e
r
cos e

sin )
+E
22
(e
r
sin +e

cos ) (e
r
sin +e

cos )
+E
12
(e
r
cos e

sin ) (e
r
sin +e

cos )
+E
12
(e
r
sin +e

cos ) (e
r
cos e

sin )
= (E
11
cos
2
+E
22
sin
2
+ 2E
12
sin cos )e
r
e
r
+ (E
11
sin
2
+E
22
cos
2
2E
12
sin cos )e

+ [(E
22
E
11
) sin cos +E
12
(cos
2
sin
2
)]e
r
e

+ [(E
22
E
11
) sin cos +E
12
(cos
2
sin
2
)]e

e
r
i.e.,
E
rr
= E
11
cos
2
+ 2E
12
cos sin +E
22
sin
2

= E
11
sin
2
+E
22
cos
2
2E
12
sin cos
E
r
= (E
22
E
11
) sin cos +E
12
(cos
2
sin
2
)
and all other cylindrical components zero.
Continuum Mechanics spring 2012 174
Plugging in the Cartesian strain tensor components, we get
E
rr
= E
11
cos
2
+ 2E
12
cos sin +E
22
sin
2
=
_
u
r
r
cos
2
+
u
r
r
sin
2

_
cos
2

+ 2
_
u
r
r

u
r
r
_
cos
2
sin
2
+
_
u
r
r
sin
2
+
u
r
r
cos
2

_
sin
2

= (cos
4
+ 2 cos
2
sin
2
+ sin
4
)
u
r
r
=
u
r
r
= a
b
r
2
E

=
u
r
r
= a +
b
r
2
and all others zero.
The dilatation = E
kk
= 2a and the stress tensor becomes
T
ij
=
ij
+ 2E
ij
i.e.,
T
rr
= + 2E
rr
= 2a + 2
_
a
b
r
2
_
T

= + 2E

= 2a + 2
_
a +
b
r
2
_
T
zz
= = 2a
and other components zero.
Continuum Mechanics spring 2012 175
Boundary conditions yield
2a( +) = 2
b
R
2
2
b = a
+

R
2
2
p + 2a( +) = 2
b
R
2
1
= 2a( +)
R
2
2
R
2
1
2a( +)
R
2
2
R
2
1
R
2
1
= p
a =
p
2( +)
R
2
1
R
2
2
R
2
1
; b =
p
2
R
2
2
R
2
1
R
2
2
R
2
1
Thus, nally (exercise), E
rr
=
p
2( +)
R
2
1
R
2
2
R
2
1
_
1
+

R
2
2
r
2
_
E

=
p
2( +)
R
2
1
R
2
2
R
2
1
_
1 +
+

R
2
2
r
2
_
T
rr
=
pR
2
1
R
2
2
R
2
1
_
R
2
2
r
2
1
_
T

=
pR
2
1
R
2
2
R
2
1
_
1 +
R
2
2
r
2
_
T
zz
=
p
+
R
2
1
R
2
2
R
2
1
Continuum Mechanics spring 2012 176
After obtaining the stress tensor in diagonal form as
T
rr
=
pR
2
1
R
2
2
R
2
1
_
R
2
2
r
2
1
_
= T
3
T

=
pR
2
1
R
2
2
R
2
1
_
1 +
R
2
2
r
2
_
= T
1
T
zz
=
p
+
R
2
1
R
2
2
R
2
1
= T
2
,
we may now evaluate the maximum normal and shear stresses as
T
max
normal
= T

(R
1
) = p
R
2
1
+R
2
2
R
2
2
R
2
1
T
max
shear
=
1
2
[T

(R
1
) T
rr
(R
1
)] = p
R
2
2
R
2
2
R
2
1
Although the maximum normal stress is about a factor of 2 greater than the maximum shear stress,
typical materials yield more easily under shear than normal stress. Thus, the elastic limit of the pipe may
be determined by the shearing yield strength, denoted by
ys
, (Tresca criterion) and the system remains
in the elastic regime (ass. R = R
2
R
1
_R =
1
2
(R
1
+R
2
)) if
p
R
2
2
R
2
2
R
2
1
<
ys
p < 2
ys
R
R
Continuum Mechanics spring 2012 177
Cantilever beam under its own gravity
Problem. Consider a thin beam of length , which is clamped
from one of its ends to a vertical wall at a right angle. (Such
a beam is called a cantilever.) Determine the shape that the
beam acquires under its own gravity.
x
1
x
2
First note: experience tells us that the beam in this case will be bent rather than sheared. Without
considering the boundary conditions, one would be tempted to argue that it is the shearing component
of the stress that dominates, i.e.,

2
u
2
x
2
1
= g
and obtain a parabolic shape for the beam. However, this is obviously not the case. Why?
Boundary conditions, i.e.,
n T(x) = 0, x S
on the boundary S of the rod, where n is the unit normal of the boundary. If the rod is very thin, this
means that
n T 0
in any point of the rod, where now nis any unit vector in the plane perpendicular to the axis of the beam.
Thus, the axial normal component T
11
is the only large component of stress in the beam. Thus, the state
of stress is close to a uniaxial tension, and T
11
(x
1
, x
2
) EE
11
(x
1
, x
2
), E
22
= E
33
E
11
.
Continuum Mechanics spring 2012 178
We start the analysis from the equilibrium equation,
T
ij
x
i
+b
j
= 0, where b
j
= g
j2
Consider the force balance in a thin slab V of the beam at x
1
x
1
+ dx
1
. Thus,
0 =
___
V
dV
_
T
ij
x
i
+b
j
_
=
__
V
dS n
i
T
ij

___
V
dV g
j2
=
__

dS
i1
T
ij
(x
1
+ dx
1
, x
2
)
__

dS
i1
T
ij
(x
1
, x
2
) dx
1
g
j2
dx
1
___

dS
T
1j
x
1
g
j2
_
= dx
1
_

x
1
__

dS T
1j
g
j2
_
,
where is the cross-sectional area of the beam. Thus,

x
1
__

dS T
11
= 0

x
1
__

dS T
12
= g =
mg

Denote the integrals hereafter by N(x


1
) =
__

dS T
11
and
V (x
1
) =
__

dS T
12
, so
N(x
1
) = const. and V
t
(x
1
) = mg/
dx
1

Continuum Mechanics spring 2012 179


Another equilibrium condition is obtained by considering the moment of force around the center of mass
(denoted by x
0
i
) of the slab:
0 =
ijk
___
V
dV (x
j
x
0
j
)
_
T
lk
x
l
+b
k
_
=
ijk
_
__
V
dS (x
j
x
0
j
)n
l
T
lk

___
V
dV T
jk
g
j2
___
V
dV (x
j
x
0
j
)
.
=0
_
=
ijk
_
__
V
dS (x
j
x
0
j
)n
l
T
lk

___
V
dV T
jk
_
For the component around the x
3
axis (around which the bending occurs), we get
0 =
3jk
_
__
V
dS (x
j
x
0
j
)n
l
T
lk

___
V
dV T
jk
_
=
__
V
dS n
l
_
(x
1
x
0
1
)T
l2
(x
2
x
0
2
)T
l1
_

___
V
dV (T
12
T
21
)
=
__
V
dS n
l
_
(x
1
x
0
1
)T
l2
(x
2
x
0
2
)T
l1
_
Continuum Mechanics spring 2012 180
and since the contribution from the beam surface vanishes (as n
l
T
lj
= 0 there) we get
0 =
__

dS
_
(x
1
+ dx
1
x
0
1
)T
12
(x
1
+ dx
1
, x
2
) (x
2
x
0
2
)T
11
(x
1
+ dx
1
, x
2
)
_

__

dS
_
(x
1
x
0
1
)T
12
(x
1
, x
2
) (x
2
x
0
2
)T
11
(x
1
, x
2
)
_
dx
1
_
__

dS T
12
(x
1
, x
2
)

x
1
__

dS (x
2
x
0
2
)T
11
(x
1
, x
2
)
_
i.e.,
V (x
1
) +M
t
(x
1
) = 0,
where
M(x
1
) =
__

dS (x
2
x
0
2
)T
11
(x
1
, x
2
)
is the moment of the normal component of the stress around the x
3
axis (the bending moment). Using T
11
EE
11
and
E
11
=
u
1
x
1
with u
1
=
du
2
(x
1
)
dx
1
(x
2
x
0
2
)
we get
M(x
1
) =
d
2
u
2
(x
1
)
dx
2
1
EI, where I =
__

dS (x
2
x
0
2
)
2
.
du
2
/dx
1
Continuum Mechanics spring 2012 181
Summarizing: the equilibrium equations are
N(x
1
) = const.; V
t
(x
1
) =
mg

; V (x
1
) +M
t
(x
1
) = 0
and the consitutive equation + geometry give
M(x
1
) =
d
2
u
2
(x
1
)
dx
2
1
EI with I =
__

dS (x
2
x
0
2
)
2
.
Thus,
d
2
dx
2
1
_
EI
d
2
u
2
dx
2
1
_
= w(x
1
), with w(x
1
) =
mg

which is known as the EulerBernoulli beam equation. The function w(x


1
) gives the external load per
unit length in the x
2
direction and I is the area moment of intertia of the beam.
The boundary conditions are that u = 0 at the clamped end and that the forces and the bending
moment vanish at the free end:
u
2
(0) = 0 = u
t
2
(0) (clamped end)
M() = 0 = M
t
() (free end)
Using these, we get for constant EI
u
2
=
mgx
2
1
(6
2
4x
1
+x
2
1
)
24EI
Continuum Mechanics spring 2012 182
9 Wave propagation
Wave propagation in linear elastic solids
shear waves
dilatational waves
elastic waves at material boundaries
elastic waves in nite media*
surface waves*
Wave propagation in uids
linear waves
sound waves
viscous waves*
shallow-water waves
non-linear waves*
solitary waves and solitons*
shock waves*
Continuum Mechanics spring 2012 183
Wave propagation in isotropic linear elastic solids
As linear elasticity is a theory of small displacements, it is natural to consider the equation of motion in
isotropic linear elastic solids,

Dv
Dt
= ( +)( u) +
2
u +b,
at this limit.
In the right-hand side

Dv
Dt
=
0
(det F)
1
D
2
u
Dt
2
,

0
(1 + u)
1
_

t
+
Du
Dt

__

t
+
Du
Dt

_
u(x, t)
0

2
u(x, t)
t
2
where we have neglected terms of higher than linear order in u.
Thus, in the absence of body forces, we obtain the equation

2
u
t
2
= ( +)( u) +
2
u.
This equation is very useful in studying various vibrations and wave propagation in linear solids. We shall
start by considering wave propagation in an innite isotropic material.
Continuum Mechanics spring 2012 184
Shear waves
Assume, rst, that the deformation in the solid is isovoluminous, u = 0. Thus,

2
u
t
2
=

2
u.
This is a standard wave equation describing the propagation of isovoluminous or shear waves (S waves).
The velocity of propagation is
C
t
=

0
.
The same result can be obtained by considering the innitesimal rotation vector, =
1
2
u, and
taking the curl of the linearized equation of motion,

2
u
t
2
= ( +)( u) +
2
u

1
2

t
2
=
1
2
( +) [( u)]
.
0
+
2
=
2

indicating that satises the shear-wave equation for an arbitrary (small) displacement eld.
Continuum Mechanics spring 2012 185
Dilatational waves
Consider, then, a displacement eld not accompanied with rotation, i.e., = 0. Using the vector
identity,

2
u ( u) (u)
=0
= ( u)
we get
0

2
u
t
2
= ( +)( u) +
2
u = ( + 2)
2
u,
which is the wave equation for irrotational or dilatational waves (P waves). The propagation velocity is
C
l
=

+ 2

0
.
The same result can be obtained by taking the divergence of the linearized equation of motion,

2
u
t
2
= ( +)( u) +
2
u

t
2
= ( +)
.

+
2
= ( + 2)
2
,
so the dilatation u fullls the equation of dilatational waves for a general deformation eld.
Continuum Mechanics spring 2012 186
Properties of dilatational and shear waves
Let us study the general wave equation

2
u
t
2
= ( +)( u) +
2
u
in an innite medium by using the Fourier transform pair
u(x, t) =
1
2
_
d
___
d
3
k u(k, ) expi(t k x)]
u(k, ) =
1
(2)
3
_
dt
___
d
3
xu(x, t) expi(t k x)]
Thus, the transformed wave equation becomes

2
u = ( +)k(k u) +k
2
u.
Write u = u
t
+ u
l
, where k u
t
= 0 and k u
l
= 0. Thus, k(k u) = k
2
u
l
and

2
u
t
= k
2
u
t

2
u
l
= ( + 2)k
2
u
l
.
Continuum Mechanics spring 2012 187
Transforming these equations back to the real space gives

2
u
t
t
2
=
2
u
t

2
u
l
t
2
= ( + 2)
2
u
l
,
where now u
t
= 0 and u
l
= 0, showing that the propagation of shear waves (with amplitude
u
t
) is decoupled from the propagation of dilatational waves (with amplitude u
l
) in a innite medium.
Suciently far from the source, waves can be treated as plane waves. Thus, we can take
u(x, t) = ue expi(t k x)],
where the physical displacement corresponds to the real part of the complex quantity. Here e is a unit
vector in the direction of displacement and k is the wave vector giving the propagation direction of the
wave. The density perturbation related to the wave is =
0
(1 + u)
1

0
ik u
0
. We
have
shear waves: phase speed

k
= C
t
=

0
, e k, = 0
dilatat. waves: phase speed

k
= C
l
=

+ 2

0
; e j k, = iku
0
e
i(tkx)
.
Note that C
l
/C
t
= [( + 2)/]
1/2
= [(2 2)/(1 2)]
1/2
.
Continuum Mechanics spring 2012 188
Elastic waves at material boundaries
When an elastic bulk wave meets a boundary surface of two
media it is reected and refracted. In general, a longitudinal
wave is reected (or refracted) not only as a longitudinal but
also a transverse wave. This is known as mode conversion.
Consider, rst, the reection of longitudinal waves from the in-
terface between a linear elastic isotropic medium and vacuum.
Choose the coordinate system so that the planar interface be-
tween the medium and the vacuum is the plane x
1
= 0 and
that the incident wave is propagating in the x
1
x
2
plane at an
angle wrt. to the interface normal. Thus,
x
1
= 0

i
n
c
i
d
e
n
t

l

w
a
v
e
r
e
f
l
e
c
t
e
d
s

w
a
v
e
r
e
f
l
e
c
t
e
d

l

w
a
v
e
u
inc
l
= eA
1
expi(t k
l
x)] = eA
1
expi[t k
l
(x
1
cos +x
2
sin )]] = e
1
where e = e
1
cos + e
2
sin and
1
= A
1
expi[t k
l
(x
1
cos + x
2
sin )]] and
k
l
= C
l
/ =
_
(2 +)/
0

2
. The boundary condition is
e
1
T(x
1
= 0) = e
1
T(x
1
= 0+) = 0 T
11
= T
12
= T
13
= 0 at x
1
= 0
They can be only satised if, in general, both longitudinal and shear waves are reected.
For the reected waves,
u
ref
l
= e
t

2
; u
ref
t
= e
tt

3
Continuum Mechanics spring 2012 189
with

2
= A
2
expi[t k
l
(x
1
cos
t
+x
2
sin
t
)]], e
t
= e
1
cos
t
+e
2
sin
t

3
= A
3
expi[t k
t
(x
1
cos +x
2
sin )]], e
tt
= e
1
sin +e
2
cos
Here, k
t
= C
t
/ =
_
/
0

2
. Note that vibrations of the shear wave tranverse to the x
1
x
2
plane
are absent, as the incident wave has no such uctuations.
The total displacement is
u = u
inc
l
+u
ref
l
+u
ref
t
= e
1
+e
t

2
+e
tt

3
and the strain components are
E
11
=
u
1
x
1
= e
1

1
x
1
+e
t
1

2
x
1
+e
tt
1

3
x
1
= i(k
l

1
cos
2
+k
l

2
cos
2

t
k
t

3
sin cos )
E
22
=
u
2
x
2
= i(k
l

1
sin
2
+k
l

2
sin
2

t
k
t

3
sin cos )
E
12
=
1
2
_
u
1
x
2
+
u
2
x
1
_
=
1
2
i(k
l

1
sin 2 k
l

2
sin 2
t
k
t

3
cos 2)
and all others zero.
Continuum Mechanics spring 2012 190
Returning to the boundary conditions, we use the constitutive equation
T
ij
= E
kk

ij
+ 2E
ij
and notice that T
13
= 0 is automatically satised. The other T
1i
at x
1
= 0
0 = T
12
= 2E
12
0 = T
11
= ( + 2)E
11
+E
22
=
0
C
2
l
E
11
+
0
(C
2
l
2C
2
t
)E
22
From the rst relation,
k
l

1
sin 2 k
l

2
sin 2
t
k
t

3
cos 2 = 0 at x
1
= 0 t, x
2
(1)
which can be satised if the exponents in
1
,
2
,
3
are all equal at x
1
= 0. Thus,
k
l
sin = k
l
sin
t
= k
t
sin or
t
= and
sin
C
t
=
sin
C
l
Substituting this back to (1) gives
k
l
(A
1
A
2
) sin 2 k
t
A
3
cos 2 = 0. (2)
The second boundary condition gives
k
l
(A
1
+A
2
)(C
2
l
2C
2
t
sin
2
) k
t
A
3
C
2
t
sin 2 = 0, (3)
and (2) and (3) together give the amplitudes A
2
and A
3
.
Continuum Mechanics spring 2012 191
The amplitude relations,
k
l
(A
1
A
2
) sin 2 k
t
A
3
cos 2 = 0
k
l
(A
1
+A
2
)(C
2
l
2C
2
t
sin
2
) k
t
A
3
C
2
t
sin 2 = 0
can be immediately used to see that no shear waves are reected (A
3
= 0) if
(A
1
A
2
) sin 2 = 0
(A
1
+A
2
)(C
2
l
2C
2
t
sin
2
) = 0
These equations can only be satised i = 0 or = /2, whence A
2
= A
1
. In all other cases,
the reected waves consist of both transverse and longitudinal waves. The amplitudes can be solved as
A
2
A
1
=
sin 2sin 2 (C
l
/C
t
)
2
cos
2
2
sin 2sin 2 + (C
l
/C
t
)
2
cos
2
2
(may vanish!)
A
3
A
1
=
2(C
l
/C
t
) sin 2sin 2
sin 2sin 2 + (C
l
/C
t
)
2
cos
2
2
,
A
2
/
A
1
1
+1
90
[
o
]
where C
l
/C
t
= [(2 2)/(1 2)]
1/2
. The relations can also be used to show that conservation
of energy applies, i.e.,
A
2
1
C
l
cos = A
2
2
C
l
cos +A
2
3
C
t
cos .
Continuum Mechanics spring 2012 192
If the interface is between two elastic media, re-
fracted waves have to be considered. The boundary
conditions at the interface (x
1
= 0) are
e
1
T
(1)
= e
1
T
(2)
and u
(1)
= u
(2)
.
For the angles, a law analogous to the rst case is
obtained by setting the complex exponents equal at
x
1
= 0:
sin
C
(1)
l
=
sin
1
C
(1)
l
=
sin
1
C
(1)
t
=
sin
2
C
(2)
l
=
sin
2
C
(2)
t
x
1
= 0

i
n
c
i
d
e
n
t

l

w
a
v
e
r
e
f
le
c
te
d
s

w
a
v
e
r
e
f
l
e
c
t
e
d

l

w
a
v
e
(1) (2)
refracted sw
ave
r
e
f
r
a
c
t
e
d
l

w
a
v
e
The amplitude relations can also be formed in similar manner as before, but of course now they are more
complicated.
Finally, we note that the any material impurities in form of grains of dierent density or elastic constants
embedded in the medium causes scattering of waves at the surfaces of the impurities. This scattering, of
course, leads to mode conversion. Scattering and reection of waves can be used to probe the properties
of materials.
Continuum Mechanics spring 2012 193
Elastic waves in a nite medium*. Longitudinal wave*
Waves in innite media are dierent from those in bounded media. As an illustration, let us consider
longitudinal waves propagating in a thin rod made of linear elastic material.
Consider a thin rod along the x
1
-axis with the radius of cross-section much shorter than the length,
r
0
_ l. Let one of its ends be subjected to a periodic longitudinal stress T
11
. Since the rod is very
thin, we may assume that other components of stress are zero.
Thus, T
11
= EE
11
= Eu
1
/x
1
and

2
u
1
t
2
=
T
11
x
1
= E

2
u
1
x
2
1
i.e., the perturbation propagates at speed

k
=

0
=

0
3 + 2
+
along the rod. The speed is obviously dierent from both C
t
and
C
l
. The material in the rod moves in the axial direction and (for
,= 0) in the radial direction.
x
1
x
1
r
r
For the limit of negligible thickness of the rod to be valid, one needs to assume that the wavelength
2/k of the wave is much larger than the thickness of the rod (i.e., kr
0
_).
Continuum Mechanics spring 2012 194
Torsional and exural waves*
The propagation speed of a longitudinal wave is given by
Youngs modulus. If, instead, the applied stress is a shearing
stress, the speed is determined by the shear modulus

k
=

0
=

0
.
In this case, the material in the rod moves in a torsional manner,
i.e., only u

,= 0. The mode is called the torsional wave.


Longitudinal and torsional waves are of symmetric type.
x
1
x
1
x
2
x
2

The most complicated case is that of a exural wave, where all components of the displacement are
non-zero and one takes u

sin and u
r
, u
1
cos . The phase speed of the exural wave
becomes proportional to k, i.e., the wave mode is dispersive. In a exural wave, elements of the cylinder
axis are in a lateral motion, and this type of wave is called anti-symmetric.
Continuum Mechanics spring 2012 195
Surface waves*
Other types of waves in bounded media include, e.g.,
surface waves. Examples of these are seismic waves
called Rayleigh waves and Love waves. In these
waves, the amplitude of the disturbance decays ex-
ponentially with depth from the surface.
Figure source: Wikipedia
Continuum Mechanics spring 2012 196
Wave propagation in uids
We will start the study of wave propagation in uids by considering Newtonian uids. As for the elastic
waves, we start by introducing the equation of motion. This is, of course, the NavierStokes equation:

Dv
Dt
= b p + ( +)( v) +
2
v,
where and are now the viscosity coecients (i.e., not the Lames constants).
The NS equation alone does not form a closed system of equations. In addition, we need the continuity
equation,

t
+ (v) = 0
and an equation of state. In case of wave propagation, we will use the equation of state,
p = P()
In addition, in wave propagation problems we will set the body force to zero, b = 0.
The next step is to derive the dispersion relation for linear waves in Newtonian uids.
Continuum Mechanics spring 2012 197
Linearized wave equation for Newtonian uids
Start by linearizing the NS equation, term by term. For this purpose, write
=
0
+
1
(x, t); v = v
1
(x, t); p = P()
and consider the terms of the NS equation:

Dv
Dt
= (
0
+
1
)
_
v
1
t
+v
1
v
1
_

0
v
1
t
p =
P

(
0
)
1
( +)( v) = ( +)( v
1
);
2
v =
2
v
1
Thus,

0
v
1
t
=
P

(
0
)
1
+ ( +)( v
1
) +
2
v
1
.
Together with the linearized continuity equation,

1
t
+
0
v
1
= 0,
we now have a closed set of equations.
Continuum Mechanics spring 2012 198
Taking a time derivative of the linearized NS eq. and using the continuity equation we get

2
v
1
t
2
=
_
P

_
0

1
t
+ ( +)(
v
1
t
) +
2
v
1
t


2
v
1
t
2
=
_
P

_
0
( v
1
) +
+

0
(
v
1
t
) +

2
v
1
t
This is the linearized equation of wave-propagation for Newtonian uids. For inviscid uids ( = = 0)
we get

2
v
1
t
2
=
_
P

_
0
( v
1
)
By taking a divergence of this equation and denoting G = v
1
=
1
0
(
1
/t) we get

2
G
t
2
=
_
P

_
0

2
G
This is the wave equation for dilatational waves, which have a propagation velocity
C
0
=

_
P

_
0
=

bulk modulus

0
sound speed.
By taking the curl of the wave equation, we see immediately that shear waves do not propagate in inviscid
uids.
Continuum Mechanics spring 2012 199
Properties of sound waves in inviscid uids
Wave equation for inviscid uids:

2
v
1
t
2
= C
2
0
( v
1
).
In innite media,

2
v
1
= C
2
0
k(k v
1
).
Thus, it is immediately clear that sound waves are longitudinal (v
1
j k) waves. Fourier transform of
the continuity equation gives
i
1
= i
0
k v
1
Thus, the condition of linearity can be obtained by requiring that
1
/
0
_1, i.e.,
v
1
_

k
= C
0
or v
1
/C
0
_1
The intensity (= ux of energy J =
1
2

0
v
2
1
C
0
) of plane sound waves produced in laboratory varies
between 0.1 and 0.3 W cm
2
. In air (
0
= 1.2 10
3
g cm
3
; C
0
= 340 ms
1
), this gives (for
J = 0.1 Wcm
2
)
v
1
/C
0
7 10
3
and in water (
0
= 1.0 g cm
3
; C
0
= 1500 ms
1
)
v
1
/C
0
2.4 10
5
,
both clearly in the linear regime.
Continuum Mechanics spring 2012 200
Eect of viscosity
The full equation for wave propagation in viscous uids reads

2
v
1
t
2
= C
2
0
( v
1
) +
+

0
(
v
1
t
) +

2
v
1
t
We are analyzing the propagation of a longitudinal wave, so assume v
1
= v
1
(x
1
, t) e
1
. Thus,

2
v
1
t
2
= C
2
0

2
v
1
x
2
1
+
+ 2

2
x
2
1
v
1
t
= C
2
0

2
v
1
x
2
1
+
C
2
0

2
v
1
x
2
1
,
where
v

0
C
2
0
/( + 2) is called the viscosity relaxation frequency. For a plane wave, v
1
=
v
1
expi(t kx
1
)], we get

2
= C
2
0
k
2
(1 + i/
v
) or k
2
=

2
C
2
0
(1 + i/
v
)
,
which gives the dispersion relation of waves in a viscous uid. Writing k = k
r
+ ik
i
we get
v
1
= v
1
exp(k
i
x
1
) expi(t k
r
x
1
)] and
k
2
r
k
2
i
=

2
C
2
0
(1
2
/
2
v
)
; 2k
r
k
i
=

3
/
v
C
2
0
(1
2
/
2
v
)
= (k
2
r
k
2
i
)

v
which may be solved for the phase speed C = /k
r
and the attenuation length k
1
i
.
Continuum Mechanics spring 2012 201
At low frequencies, _
v
and k
i
_k
r
, we nd
C C
0
and k
i


2
2
v
C
0
.
The amplitude attenuation per wavelength, thus, becomes
= k
i
= 2k
i
/k
r
= /
v
Notes:
(i) The observed attenuation of ultrasonic waves can be used to obtain the value of

v
=

0
C
2
0
+ 2
=

0
C
2
0
b +
4
3

,
where b = +
2
3
is the bulk viscosity. As the shear viscosity is, in general, easy to measure,
the ultrasonic attenuation experiment provides a means to determine b. The ratio of bulk-to-shear
viscosity measured this way for some liquids is given in the table below.
Liquid Glycerol Water Methyl alcohol Toluene Benzene
b/ 1.1 2.5 3.2 13 100
(ii) Attenuation length for plane waves is proportional to
2
(or
2
). Thus, to probe the properties
of matter, one should use as low a frequency as possible. (However, the wavelength needs to be
much smaller than the typical scale sizes under investigation.)
Continuum Mechanics spring 2012 202
(iii) Highly viscous uids. For large enough values of b+
4
3
,

v
becomes comparable to . Thus, the condition
_
w
is no longer satised and the values of k
r
and k
i
have to be solved exactly from the dispersion
relation. Thus (exercise),
C
2
=

2
k
2
r
= 2C
2
0
1 +
2
/
2
w
1 + (1 +
2
/
2
w
)
1/2
k
i
=
C
2C
2
0
/
w
1 +
2
/
2
w
As an example, the phase speed of the wave is plotted in the gure for Glycerol (5% of water).
Clearly, the theory does not well describe the observations. This means that the NavierStokes
equation is not valid for highly viscous systems. This is to be expected, as in the constitutive
equation for Newtonian uids, all higher-than-rst-order derivatives of the velocity were neglected.
By applying the same reasoning in the present case, one expects

C
2
0

2
v
1
x
2
i

C
2
0

2
v
1
x
2
i

,
but this for plane harmonic waves implies /
w
_1. Thus, this limitation is built-in limitation
for Newtonian uids.
Continuum Mechanics spring 2012 203
Diusion of vorticity*. Viscous waves*
Consider the rotational motion, i.e., vorticity in a uid. Recall the linearized NS equation (omit the
subscript 1)

2
v
t
2
= C
2
0
( v) +

t
_
+

0
( v) +

2
v
_
Taking the curl of both sides gives

2
w
t
2
=

t
_

2
w
_

w
t
=

2
w,
which is a diusion equation for vorticity. Consider a transverse wave v = v
2
expi(t kx
1
)]e
2
.
Obviously, v = 0 and, thus,

2
v
2
t
2
=

2
v
2
x
2
1
so
2
= ik
2
/
0
or k
2
= i
0
/
Setting, again, k = k
r
+ ik
i
gives
k
2
r
k
2
i
= 0 and k
r
k
i
=
0
/2, i.e., k
r
= k
i
= (
0
/2)
1/2
The attenuation per wavelength, = 2, so the waves, called viscous waves, are very rapidly damped.
Thus, transverse uctuations can penetrate a viscous uid by only about one wave length thickness.
Continuum Mechanics spring 2012 204
Shallow-water waves
Also in uids, we can consider surface waves. Consider a layer of incompressible, inviscid uid of thickness
d on top of a solid horizontal oor. Let h = d+(x, t) be the height of the surface from the bottom.
Consider a shallow layer, so that the velocity component along z can be taken as small. Conservation of
mass implies

t
(h(x, t) dx) = [h(x + dx, t)v(x + dx, t) h(x, t)v(x, t)] dx

x
(hv)

h
t
+

x
(hv) = 0
Similarly, conservation of momentum implies (exercise)
(hv)
t
+

x
_
hv
2
+
1
2
gh
2
_
= 0
Linearising

t
+d
v
x
= 0;
v
t
+g

x
= 0
giving

t
2
= gd

x
2
i.e., a wave propagating at speed C =

gd is found. The solution is valid for d and _d.


Continuum Mechanics spring 2012 205
Solitary waves and solitons*
If the non-linearities are not neglected but taken into account to the lowest order, one can derive after
transforming to frame moving with the linear waves (x
t
= xCt) and rescaling of the dependent and
independent variables the Kortewegde Vries equation for the propagation of the wave (see, e.g., G.L.
Lamb, Elements of Soliton Theory, Wiley, 1980). In dimensionless form

t
+ 6
x
+
xxx
= 0,
where the subscript denotes dierentiation. Seek for wave-like solutions of the equation, i.e., take
= f(x ct)
and substitute to the equation:
cf
t
+ 6ff
t
+f
ttt
= 0
cf + 3f
2
+f
tt
= A ] f
t

1
2
cf
2
+f
3
+
1
2
f
t2
= Af +B
B =
1
2
f
t2
+f
3

1
2
cf
2
Af =
1
2
f
t2
+V (f)
where V (f) = f
3

1
2
cf
2
Af and Aand B are integration constants. Thus, f(x) behaves like the
position z(t) of a particle with total energy of B in a cubic potential V (z) [i.e.,
1
2
z
2
+V (z) = B].
Continuum Mechanics spring 2012 206
As V
t
(f) = 3f
2
cf A, there are two extrema of the
potential at
f

=
c

c
2
+ 12A
6
; A >
1
12
c
2
,
where f

is a maximum and f
+
is a minimum. [Note: for
A =
1
12
c
2
only one extremum and for A <
1
12
c
2
none.]
If B < V (f

), solutions oscillating between two roots of


V (f) = B are obtained.
V
f
f
+
f

Choosing A = B = 0, i.e.,
V (f) = f
2
(f
1
2
c)
gives a potential with local maximum V = 0 at f = 0. This admits a solution with f, f
t
0 as
x and with a maximum value of f =
1
2
c somewhere in between.
Its a easy exercise to show that this solution is given by
(x, t) = f(x ct) =
c
2 cosh
2

c
2
(x ct a)]
,
where a is an arbitrary constant. This describes a solitary wave, i.e., a non-periodic, localized propagating
wave with a permanent form.
The solitary-wave solution of the KdV equation is an example of a soliton. Solitons are solitary waves
which can interact with other solitons emerging from the collision unchanged, except for a phase shift.
Continuum Mechanics spring 2012 207
Burgers equation*. Shock waves*
The KdV equation
t
+ 6
x
+
xxx
= 0 without the nonlinear term represents a dispersive wave,
i.e., the linear dispersion relation becomes
+k
3
= 0 or

k
= k
2
showing that high-frequency uctuations are lagging behind the low-frequency uctuations. The soliton
solutions describe a system, which attains a balance between nonlinear and dispersive eects.
Another non-linear equation of considerable importance in uids is the Burgers equation,
v
t
+vv
x
= v
xx
,
where is viscosity. It represents a system that can attain a balance between non-linearity and dissipation.
A propagating wave solution can be obtained by v = f(x ct)
cf
t
+ff
t
= f
tt

1
2
A cf +
1
2
f
2
= f
t

df
f
2
2cf +A
=
dx
2

x x
0
2
=
_
df
c
2
A (f c)
2
A<c
2
=
1

c
2
A
artanh
f c

c
2
A
f = c
_
c
2
Atanh
_
_
c
2
A
x x
0
2
_
Continuum Mechanics spring 2012 208
Therefore, at x , f c

c
2
A = v

, so

c
2
A =
1
2
(v
+
v

) =
1
2
v and
c =
1
2
(v
+
+v

) = v. Thus,
v(x, t) = v
1
2
v tanh
x vt x
0
4/v
This represents a solution propagating from left to right at speed v, where the speed of the uid increases
from right to left (i.e., a fast uid is overtaking a slower one). The transition layer from the slower to
the faster speed has a thickness of 4/v. This kind of a wave is a shock wave.
Actually, as Burgers equation contains no pressure gradient term, it is not a completely valid description
of shocks. By using p = P() we get as a one-dimensional momentum equation
v
t
+vv
x
= P
t
()
x
/ + v
xx
,
where the equation of continuity,
t
+ (v)
x
= 0, has to be used to eliminate the extra term. Write
1
t
() = P
t
()/ and seek for a propagating solution in form v = v(x ct), = (x ct).
c
t
+v
t
+v
t
= 0 (c v) = M = const. =
M
c v
c v
t
+vv
t
= 1
t
() +v
tt
v
t
= 1() cv +
1
2
v
2
= 1
_
M
c v
_
cv +
1
2
v
2
which shows that the speed cannot cross v = c without yielding an innite density and that the pressure
gradient term keeps the ow speed from crossing the speed of propagation of the shock wave.
Continuum Mechanics spring 2012 209
For a more physical description of shock waves in hydro-
dynamics, one should use an energy equation instead of
an equation of state.
For 0, the thickness of the shock becomes very
small, and the solution can be treated as a discontinuity.
The conservation laws for mass, momentum and energy
then read

v
t

=
+
v
t
+

v
t

2
+P

=
+
v
t
+
2
+P
+
1
2

v
t

3
+ (P

+U

)v
t

=
1
2

+
v
t
+
3
+ (P
+
+U
+
)v
t
+
,
v
+
v
+
v

4/v
v

x
v
x
v
c
Burgers equation
Real shock wave
~/v
c
where v
t
= c v is the velocity measured in the shock frame and U is the internal energy per unit
volume. For an ideal gas it is given as
U =
P
1
where = c
p
/c
V
. One can solve the conservations laws for the values behind the shock (subscript )
for known values in the ambient medium (subscript +).
Continuum Mechanics spring 2012 210

You might also like