Mechanics of solids and fluids

-Introduction to continuum mechanics

by

Magnus Ekh

August 19, 2019

Introduction to continuum
mechanics

1 Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Index notation
1.2 Vectors
1.3 2nd order tensors
1.4 Principal values and principal directions
1.5 Spatial derivatives
1.6 Divergence theorem

2 Stress, motion and deformation . . 25

2.1 Stress analysis
2.2 Continuum motion
2.3 Lagrangian and Eulerian description
2.4 Material time derivative
2.5 Reynolds’ transport theorem for a material volume
2.6 Reynolds’ transport theorem for a control volume

3 Field equations . . . . . . . . . . . . . . . . . . . . 41
3.1 Physical quantities of a continuum
3.2 Input quantities
3.3 Physical conservation principles
3.4 Summary of field equations and field variables

4 Constitutive models . . . . . . . . . . . . . . 47
4.1 Fourier’s law of thermal conductivity
4.2 Viscous fluids
4.3 Linear elastic isotropic solids
1. Tensors

1.1 Index notation

Before introducing concepts of tensor algebra we introduce the index notation. The index
notation simplifies writing of quantities as well as equations and will be used in the
remaining of this text. There are two types of indices:
• Free indices are only used once per quantity and can take the integer values 1, 2
and 3. For example for one free index i:

ai ⇔ a1 , a2 and a3
ai = b i ⇔ a1 = b1 , a2 = b2 and a3 = b3

Similarly we can have two (or more) free indices i and j:

aij ⇔ a11 , a12 , a13 , a21 , a22 , a23 , a31 , a32 and a33
aij = bij ⇔ a11 = b11 , a12 = b12 , a13 = b13 , . . . , a32 = b32 and a33 = b33

• Summation indices are used twice per term and indicates a summation of that index
from 1 to 3. For example:

3
X
aii ⇔ aii
i=1
X3
ai b i ⇔ ai b i
i=1
X3 X 3
aij bij ⇔ aij bij
i=1 j=1

This sum over repeated indices is often called Einstein’s summation convention.
Often these two types of indices are used together. A simple example is the equation
system
3
X 3
X 3
X
ai = Tij bj ⇔ a1 = T1j bj , a2 = T2j bj and a3 = T3j bj
j=1 j=1 j=1

where i is a free index and j a summation index. Another way to express this equation
system is to use matrices (in this example two column matrices 3 × 1 and a square matrix
 4 Chapter 1. Tensors

3 × 3):     

a1  
T11 T12 T13 
b1 
a2 = T21 T22 T23 b2 (1.1)
    
   
    
a3 T31 T32 T33 b3
which also sometimes is written by using index notation:

[ai ] = [Tij ] [bj ] (1.2)

Problem 1 Explain the following symbols: Aii , Aijj , Aij , ai Aij , ci bj Aij .
For each index tell whether it is a summation/dummy index or a free index.

Problem 2 Use index notation to re-write the following expression: f1 u1 + f2 u2 + f3 u3

Answer: fi ui

Problem 3 Expand aijk bik by giving the terms explicitly.

Answer: ai1k bik = 3i=1 3k=1 ai1k bik = a111 b11 + a112 b12 + a113 b13 + a211 b21 + a212 b22 +
P P

a213 b23 + a311 b31 + a312 b32 + a313 b33 ,

ai2k bik = 3i=1 3k=1 ai2k bik = a121 b11 +a122 b12 +a123 b13 +a221 b21 +a222 b22 +a223 b23 +
P P

a321 b31 + a322 b32 + a323 b33 ,

ai3k bik = 3i=1 3k=1 ai3k bik = a131 b11 +a132 b12 +a133 b13 +a231 b21 +a232 b22 +a233 b23 +
P P

a331 b31 + a332 b32 + a333 b33

Assignment 1 (a) Use index notation to re-write the following expression: b11 c1 d1 +
b12 c2 d1 + b13 c3 d1 + b21 c1 d2 + b22 c2 d2 + b23 c3 d2 + b31 c1 d3 + b32 c2 d3 + b33 c3 d3 .
(b) Show for what condition on Aij does bi Aij = Ajk bk hold (for all bi )?

Matlab example 1 An example of using Matlab commands for matrix definitions (for
T and b) and multiplication ai = Tij bj is given below:
>> T=[1 2 3; 4 5 6; 7 8 9];
>> b=[1 2 3]’;
>> a=T*b
a =
14
32
50
 1.2 Vectors 5

3 B

ê3 a
2

A
ê2

ê1

Figure 1.1: Illustration of vector a.

1.2 Vectors
Orthonormal base vectors

To describe many physical quantities (such as force, displacement, velocity) both magni-
tude and direction must be given. Hence, these quantities can be described by vectors (1st
order tensors) in a 3-dimensional Euclidean space. By introducing a set of right-handed
−→
orthonormal basis vectors {ê1 , ê2 , ê3 } any vector a = AB can be expressed as a linear
combination these basis vectors, êi :

a = a1 ê1 + a2 ê2 + a3 ê3 = ai êi . (1.3)

as shown in Figure 1.1. The coefficients ai or (a1 , a2 , a3 ) are the components of a with
respect to the basis êi . The length (=Euclidean norm) of a vector a is denoted a or |a|.
For normalized vectors (describing only direction) the following notations are introduced:
a
êa = â = , (1.4)
a
whereby a vector a can be written as a = a êa . Examples of normalized vectors are the
basis vectors {ê1 , ê2 , ê3 }.
 Example 1.1 Assume that the vector a = a1 ê1 + a2 ê2 + a3 ê3 = ai êi . The normalized
vector â is obtained as follows:
a1 ê1 + a2 ê2 + a3 ê3
â = q
a21 + a22 + a23

 6 Chapter 1. Tensors

θ a

b · êa

Figure 1.2: Illustration of scalar product.

Problem 4 Determine the unit length vector along a = 4 ê1 + 6 ê2 − 12 ê3 .
Answer: â = 2/7 ê1 + 3/7 ê2 − 6/7 ê3

Scalar product

To each pair of vectors a and b there corresponds a real number a · b, called the scalar
product. The scalar product is defined as (see Figure 1.2):

a · b = a b cos θ (1.5)

where θ is the angle between the vectors. The vector projection of b on a is defined as
the orthogonal projection of b on a line parallel to a and is equal to b cos(θ). It can be
obtained from the definition of scalar product as b · êa . Further, it is possible to express
the length of a vector a = |a| as follows
√
a = |a| = a · a (1.6)

By now applying the scalar product between the orthonormal basis vectors {ê1 , ê2 , ê3 },
the following results are obtained
êi · êj = δij (1.7)
where 
 1 when i = j
δij = (1.8)
 0 when i 6= j
The symbol δij is called the Kronecker delta symbol. The scalar product between vectors
is a bilinear operator and has the following properties:





a · (αb + βc) = αa·b+βa·c




(α a + β b) · c

= αa·c+βb·c
where α and β are scalars. These properties can now be used to show that the scalar
product between two vectors a and b may be written as:

a · b = ai êi · bj êj = ai bj êi · êj = ai bj δij = ai bi (1.9)

1.2 Vectors 7

a×b

Figure 1.3: Illustration of vector product.

and that the scalar product is commutative i.e. a · b = b · a. The components ai of

a vector a can be extracted by scalar multiplication with corresponding base vectors êi
which is shown by the following derivation:

êi · a = êi · aj êj = aj êi · êj = aj δij = ai (1.10)

Problem 5 Compute the projection of the vector a = 4 ê1 + 6 ê2 − 12 ê3 on the line
defined by the vector b = 1 ê1 + 1 ê2 + 1 ê3
√
Answer: −2/ 3

Problem 6 Expand the following expressions of the Kronecker delta δij :

δij δij , δij δjk δki , δij Aik
Answers: 3, 3, Ajk .

Vector product

Another product that is useful is the vector product a×b, which is illustrated in Figure 1.3.
The result is a vector that is orthogonal to the plane spanned by a and b (with a right-
handed system) and has the length

|a × b| = a b sin θ (1.11)

By applying the vector product to the orthonormal basis vectors {ê1 , ê2 , ê3 }, the following
results are obtained
êi × êj = eijk êk , (1.12)
where the permutation symbol eijk is defined as





1 when ijk = 123, 231 or 312
eijk =  −1 when ijk = 321, 213 or 132 (1.13)

0 otherwise


 8 Chapter 1. Tensors

The vector product is bilinear, i.e.






a × (αb + βc) = αa×b+βa×c




(α a + β b) × c

= αa×c+βb×c
whereby the vector product between two arbitrary vectors becomes

a × b = (ai êi ) × (bj êj ) = ai bj êi × êj = ai bj eijk êk . (1.14)

From this we can note the relation a × b = −b × a which is found by using the properties
of the permutation symbol. The permutation symbol and the Kronecker’s delta symbol
are linked by the so-called e-δ identity:

eijm eklm = δik δjl − δil δjk . (1.15)

Problem 7 Compute the unit normal to the plane

(−1, 6, 2)

(0, 4, 1)
ê3

ê2

(2, 2, 0)
ê1

Answer: n̂ ≈ ±(−0.45 ê2 + 0.89 ê3 )

Problem 8 Show that eijk δjk = 0

Problem 9 Prove that for three arbitrary vectors a, b and c the following relation
holds:
a × (b × c) = (a · c) b − (a · b) c

Open product

Open product (also called outer product) between two vectors a and b results in a 2nd
order tensor T (also called dyad) as follows

a b = ai êi bj êj = ai bj êi êj = Tij êi êj = T (1.16)

The open product is bilinear but not cummutative i.e. a b 6= b a in general. 2nd order
tensors will be further exploited in Section 1.3. In literature the open product is sometimes
for clarity denoted by ⊗, i.e., the dyad is written as a ⊗ b.
1.2 Vectors 9

Column matrix representation of components

In a given coordinate system defined by the basis vectors {ê1 , ê2 , ê3 }, the vector compo-
nents ai can be collected in a column matrix as follows

[a] = [ai ] = [ a1 a2 a3 ]T (1.17)

An example is the base vector ê1 that is represented by the following column matrix

[ê1 ] = [ 1 0 0 ]T (1.18)

Therefore, the scalar multiplication between two vectors can be obtained as

a · b = ai bi = [a]T [b] . (1.19)

 Example 1.2 Assume that a = ai êi and b = bi êi . The matrix representation of the
components of the dyad a b is given as:
 
a b a1 b 2 a1 b 3 
 1 1
[a b] = a2 b1 a2 b2 a2 b3 


 
a3 b 1 a3 b 2 a3 b 3

Matlab example 2 Example of scalar product in Matlab

>> a=[1 2 3]’; b=[3 4 5]’;

>> c=sum(a.*b)
c =
26

Matlab example 3 Example of cross product in Matlab

>> a=[1 2 3]’; b=[3 4 5]’;

>> c=cross(a,b)
c =
-2
4
-2
 10 Chapter 1. Tensors

Coordinate system transformation

A vector must be invariant with respect to coordinate system. Assume two different
n o
sets of orthonormal basis vectors {ê1 , ê2 , ê3 } and ê01 , ê02 , ê03 . The vector b can then be
written as
b = bi êi = b0i ê0i (1.20)

The components b0i can be extracted from b as

b0i = ê0i · b = ê0i · bj êj = ê0i · êj bj (1.21)

In matrix notation this can be written

h i
[b0i ] = [lij ] [bj ] = ê0i · êj [bj ] (1.22)

where the transformation matrix [lij ] is orthogonal, i.e. [lij ]T = [lij ]−1 . This can be
understood if we assume that the components b0j are known and then the components bi
can be extracted from b (similarly to (1.21)) as

bi = êi · ê0j b0j (1.23)

1.2 Vectors 11

Assignment 2 A thin rigid bar OA with mass m and length 5a is attached without
friction in a joint at O. The bar is kept in equilibrium by two light cables AB and
AC acc to the Figure. The cable AB is attached to the bar at B with coordinates
(3a; 0; 3a) and the cable AC is attached to the bar at C with coordinates (−a; 2a; 5a).
z
g C
B

A
x O
3a
4a

Determine the forces in the cables AB and AC at equilibrium (g is the acceleration

of gravity in the negative z direction).

Assignment 3 Give the component of the vector a in the rotated coordinate system
{ê0i }. This coordinate system is obtained from the coordinate system {êi } by rotating
around the ê3 axis according to the figure.
ê2
ê02

ê01

π/6
ê1

The components of the vector a in the coordinate system {êi } are given as [−1 4 3]T .

Matlab example 4 Example of Matlab input file to define eijk -operator and vector
product ck = ai bj eijk :

%definition of permutation symbol

perm=zeros(3,3,3);
for i=1:3
 12 Chapter 1. Tensors

for j=1:3
for k=1:3
%%%
if ( (i==1) & (j==2) & (k==3)) | ((i==2) & (j==3) & (k==1)) | ...
((i==3) & (j==1) & (k==2))
perm(i,j,k)=1;
elseif ( (i==3) & (j==2) & (k==1)) | ((i==2) & (j==1) & (k==3)) | ...
((i==1) & (j==3) & (k==2))
perm(i,j,k)=-1;
end
%%%
end
end
end
%computation of vector product c_k= a_i b_j perm_ijk
a=[1 2 3]’;
b=[4 5 6]’;
c=zeros(3,1);
for k=1:3
c(k)=0;
for i=1:3
for j=1:3
c(k)=c(k,1)+a(i)*b(j)*perm(i,j,k);
end
end
end

1.3 2nd order tensors

Representation of 2nd order tensors

2nd order tensors are physical quantitites that describe how vectors change with e.g.
direction and position in space. Examples of 2nd order tensors that we will explore later
are the stress tensor, strain tensor, velocity gradient and the deformation gradient. A 2nd
order tensor T is represented in a orthonormal coordinate system {ê1 , ê2 , ê3 } as

T = Tij êi êj , (1.24)

where Tij are the nine components of T and êi êj are the base dyads. The base dyads êi êj
are 2nd order tensors themselves and T is built up by a linear combination of these scaled
1.3 2nd order tensors 13

by the components Tij .

 Example 1.3 The matrix representation of the components of T are obtained as
     

1 0 0 
0 1 0 T
 11
T12 T13 
[T ] = Tij [êi êj ] = T11 0 0 0 + T12 0 0 0 + . . . = T21 T22 T23 
     
     
0 0 0 0 0 0 T31 T32 T33


Scalar product between 2nd order tensors and vectors

Now consider a linear transformation of a vector b into a

T
b a

This may be written symbolically by introducing a scalar product as

a=T ·b (1.25)

where the linear operator T is a second-order tensor. Before proceeding we define the
scalar product between a base dyad êi êj and a base vector êk as:

(êi êj ) · êk = êi (êj · êk ) = δjk êi and êi · (êj êk ) = (êi · êj ) êk = δij êk (1.26)

 Example 1.4 Two example of results are: (ê1 ê2 ) · ê2 = ê1 and ê1 · (ê2 ê2 ) = 0 

The scalar product between a 2nd order tensor and a vector is assumed to be bilinear. If
we use index notations then such a scalar product can be written as:

a = T · b = Tij êi êj · bk êk = Tij bk êi δjk = Tij bj êi = ai êi (1.27)
| {z }
ai

Often we omit the basis and simply write the relation between the components, i.e.

ai = Tij bj (1.28)

It is then implicitly assumed that the same basis vectors are used for all variables. Fur-
thermore, standard matrix manipulations can be used in numerical implementations to
compute the components ai
    
 1
a T
 11
T12 T13  b1 
a2 
 
= T21

T22 T23  b 
 
     2
a3 T31 T32 T33 b3
 14 Chapter 1. Tensors

If we switch the order of the vector and the 2nd order tensor in the scalar product

c = b · T = bi êi · Tjk êj êk = bi Tjk δij êk = bj Tjk êk = ck êk (1.29)
| {z }
ck

or in short ck = bj Tjk . A consequence of these results is that:

T · b = b · TT

where the transpose of the tensor is defined as T T = Tji êi êj . In matrix notations and
operations this corresponds to
    
T
 11
T12 T13  b1  h T
i  11
T21 T31 
T21 T22 T23  b2  = b1 b2 b3 T12 T22 T32 
    
    
T31 T32 T33 b3 T13 T23 T33

An example of a 2nd order tensor is the stress tensor σ from which the traction vector
t(n̂) can be obtained as t(n̂) = n̂ · σ.
n̂

t(n̂)

 Example 1.5 By using this scalar multiplication twice it is possible to find the compo-
nents of a 2nd order tensor Tij by

êi · T · êj = êi · (Tkl êk êl ) · êj = Tkl δik δlj = Tij (1.30)

A special 2nd order tensor is the identity tensor δ:

δ = δij êi êj = êi êi (1.31)

with the property that it does not transform a vector a when scalar multiplied with δ,
i.e. δ · a = a and a · δ = a (or written by using only the components δij aj = aj δji = ai ).
The components for δ can be collected in the following matrix form:
 
1 0 0

[δ] = 0
1 0 (1.32)
 
 
0 0 1
1.3 2nd order tensors 15

Multiplication between 2nd order tensors

Scalar multiplication (also called single contraction) between two base dyads is defined as

(êi êj ) · (êk êl ) = êi (êj · êk ) êl = êi δjk êl = δjk êi êl (1.33)

This scalar multiplication is assumed to be bilinear and therefore the scalar multiplication
of two 2nd order T and U can be written as

T · U = Tij êi êj · Ukl êk êl = Tij Ukl êi δjk êl = Tij Ujl êi êl = V (1.34)
| {z }
Vil

or in terms of components Tij Ujk = Vik . Hence, when using a matrix notation then the
components of V can simply be obtained by a standard matrix multiplication between
[Tij ] and [Ujk ]. Further, by applying the transpose operator to such a product it can be
shown that
V T = (T · U )T = U T · T T (1.35)
 Example 1.6 To show (1.35) we use the index notation:

VijT = Vji = Tjk Uki = Uik

T T
Tkj

Another operator that we introduce is the double contraction operator between two base
dyads
(êi êj ) : (êk êl ) = (êi · êk ) (êj · êl ) = δik δjl (1.36)
If we assume bilinearity of that operator then double contraction between two 2nd order
T and U results in a scalar α and is obtained as

T : U = Tij Uij = α (1.37)

Problem 10 If a and b are vectors and A and B are 2nd order tensors show that
a) (a · A) · b = a · (A · b)
b) (A · B)T = B T · AT
c) (A · a) · (B · b) = a · (AT · B) · b

Problem 11 The components of the 2nd order tensors and vectors are given as:
       

1 2 0 
3 0 0 2
  
1
[Aij ] = 2 3 4

, [Bij ] = 0 3 1

, [ai ] = 3 ,
 
[bi ] = −1


       
0 4 2 0 1 2 1 2

Compute
 16 Chapter 1. Tensors

a) A · a
b) a · b
c) A : B
d) A : (a b)
Answers: a) [8, 17, 14]T , b) 1, c) 24, d) 19.

Matlab example 5 An example of computing the double contraction in Matlab:

>> T=[1 2 3; 4 5 6; 7 8 9];
>> U=[1 2 1; 3 4 3; 5 6 5];
>> alpha=sum(sum(T.*U))
alpha =
186

Symmetric and skew-symmetric 2nd order tensors

As introduced earlier the transpose T T of a 2nd order tensor T is defined as follows:

T T = Tij êj êi = Tji êi êj (1.38)

Many second-order tensors in mechanics are symmetric which means that the tensor and
its transpose are equal e.g. T T = T or in components Tij = Tji . Another type of tensors is
the skew-symmetric second-order tensors. These have the property that the transpose of
the tensor is equal to the tensor with a minus sign, e.g., T T = −T or Tij = −Tji . Clearly,
for such a tensor the diagonal elements (in a matrix representation) must be equal to zero
whereby the components can be collected in the following general matrix
 

0 T12 T13 
−T12 0 T23  (1.39)
 
 
−T13 −T23 0

Problem 12 Show that Aij = eijk ak is skew-symmetric (i.e. Aji = −Aij ).

Problem 13 If Aij is symmetric and Bij is skew-symmetric. Show that Aij Bij = 0.

Inverse of a 2nd order tensor

If we assume that the tensor T gives the linear transformation a = T · b between the
two vectors b and a. Then we can introduce the inverse T −1 of this transformation as
b = T −1 · a. If we express these two relations in components

ai = Tij bj and bi = Tij−1 aj (1.40)

1.3 2nd order tensors 17

then it is obvious that the components of the T −1 can be found using standard matrix
inversion i.e.
h i
Tij−1 = [Tij ]−1 (1.41)

Hence, standard rules for matrix inversion apply also for tensor components such as
(T · U )−1 = U −1 · T −1 .

Coordinate system transformation

A 2nd order tensor is invariant with respect to coordinate system. Assume two different
n o
sets of orthonormal basis vectors {ê1 , ê2 , ê3 } and ê01 , ê02 , ê03 . The 2nd order tensor T
can then be written with either basis vectors:

T = Tij êi êj = Tij0 ê0i ê0j (1.42)

The components Tij0 can be extracted from T as

Tij0 = ê0i · T · ê0j = ê0i · êk Tkl êl · ê0j (1.43)

In matrix notation this can be written

h i
Tij0 = [lik ] [Tkl ] [ljl ]T (1.44)

where the orthogonal transformation matrix [lij ] was defined in (1.22).

Higher order tensors

It is possible to construct tensors of any order (or rank) as follows:

A = Aijk··· êi êj êk · · ·

In particular, fourth-order tensors are frequently used to, for example, give the relation
(material behavior) between the second-order tensors stress and strain.

Problem 14 a) Determine the transformation matrix when ê01 is parallel to ê1 + ê2 − ê3
and ê02 parallel to ê2 + ê3 .
Answer:  √ √ √ 
1/ 3 1/ 3 −1/ 3
 √ √ 
[lij ] =  0 1/ 2 1/ 2 
 
 √ √ √ 
2/ 6 −1/ 6 1/ 6
b) The components of the 2nd order stress tensor σ have been measured by engineer
 18 Chapter 1. Tensors

Emil in the coordinate system {êi } with the following results

 

400 100 0 
[σij ] = 100 200 0 

 MPa
 
0 0 300

Engineer Emilia uses the coordinate system {ê0i } (from a)), what are the stress com-
ponents in that coordinate system?
Answer:  
h i 
367 0 94.3 
σij0 = 
 0 250 86.6  MPa

 
94.3 86.6 283

Assignment 4 Given the Sherman-Morrison’s formula:

α
Aij = δij + α ui vj then A−1
ij = δij − ui vj
1 + α uk vk
Show that by using Sherman-Morrison’s formula if Aij = Bij + α ui vj then
α
A−1 −1
ij = Bij − B −1 um vn Bnj
−1
−1
1 + α vk Bkl ul im

1.4 Principal values and principal directions

A second-order tensor can be, as discussed above, thought of a linear transformation
between vectors, i.e. a = T · b. Important properties of a second-order tensor are its
eigenvectors (principal directions) and eigenvalues (principal values). Eigenvectors are
defined as vectors that do not rotate upon transformation with the second-order tensor.
If n̂ now is an eigenvector to T this can be illustrated as

T
n̂ λ n̂

This can be written as

λ n̂ = T · n̂ or λn̂i = Tij n̂j . (1.45)

The eigenvectors n̂ are chosen to be of unit length whereby it is possible to identify

the length of the vector T · n̂ as the corresponding eigenvalues λ. The way to find the
1.4 Principal values and principal directions 19

eigenvalues and eigenvectors is to rewrite (1.45) as

(λ δ − T ) · n̂ = 0 or (λ δij − Tij ) n̂j = 0i . (1.46)

A trivial solution to this equation is that n̂ = 0. However, it is possible to find non-

trivial solution if (λ δ − T ) is non-invertible. From linear algebra we know that then the
determinant of the matrix [λ δ − T ] must be zero, i.e.

det (T − λ δ) = 0 (1.47)

which is called the characteristic equation. An important theorem from linear algebra is
the spectral theorem which states that for symmetric matrices the eigenvalues are real and
the eigenvectors are orthogonal. In the current course we will only consider eigenvalues
and eigenvectors for symmetric second order tensors (i.e. stress, strain, etc) and for such
a tensor the characteristic equation can be obtained as
 
T11 − λ T12 T13
 
 
 
T12 T22 − λ T23 = (T11 − λ) (T22 − λ) (T33 − λ) +
 
 
 
T13 T23 T33 − λ
 
 
2 2 2
T12 T23 T13 + T13 T12 T23 − T13 (T22 − λ) − T23 (T11 − λ) − (T33 − λ) T12 =0

This third order polynomial equation can be summarized as

λ3 − I1 λ2 + I2 λ − I3 = 0 (1.48)

where the invariants of the second-order tensor T were introduced as

I1 = Tii , I2 = [Tii Tjj − Tij Tij ] /2, I3 = det(Tij ) (1.49)

After solving the three eigenvalues λ(1) , λ(2) , λ(3) from (1.48) we can solve the correspond-
ing eigenvectors n̂(1) , n̂(2) , n̂(3) from (1.46).

Problem 15 Find eigenvalues and eigenvectors of:

 

2 −1 0
−1 0 0
 
 
0 0 1
Answers:
λ(1) ≈ 2.41 λ(2) = 1 λ(3) ≈ −0.414

n̂(1) = ± [−0.924 0.383 0]T n̂(2) = ± [0 0 1]T n̂(3) = ± [−0.383 − 0.924 0]T
 20 Chapter 1. Tensors

Problem 16 Find eigenstresses and eigenvectors of the stress tensor with components:
 √ 
 √
200 − 3 · 100 0 
− 3 · 100 400 0 MPa
 

 
0 0 400
Answers:
λ(1) = 100 λ(2) = 400 λ(3) = 500 MPa
"√ #T " √ #T
3 1 T 1 3
n̂(1) =± 0 (2)
n̂ = ± [0 0 1] (3)
n̂ = ± − 0
2 2 2 2

Assignment 5 a) Use the fact that eigenvectors n̂(i) of a symmetric 2nd order tensor
T are orthogonal to show that T can be expressed as

T = λ(1) n̂(1) n̂(1) + λ(2) n̂(2) n̂(2) + λ(3) n̂(3) n̂(3)

b) Given that the exponential function of a scalar and a 2nd order tensor are
defined as:
∞ ∞
X αk X Tk
exp(α) = , exp(T ) = .
k=0 k! k=0 k!

Show that

exp(T ) = exp(λ(1) )n̂(1) n̂(1) + exp(λ(2) )n̂(2) n̂(2) + exp(λ(3) )n̂(3) n̂(3)

and use this to compute exp(T ) where T is represented by the the following
matrix (in a êi system)  

6 4 0 
[T ] =  4

3 0 

.
 
0 0 2

Matlab example 6 An example of using Matlab commands for matrix definitions (for
A ) and computing the eigenvalues and eigenvectors given below:

>> A=[1 2 3; 2 4 5; 3 5 6];

>> [n,lambda]=eig(A)

n =
 1.5 Spatial derivatives 21

0.7370 0.5910 0.3280

0.3280 -0.7370 0.5910
-0.5910 0.3280 0.7370

lambda =

-0.5157 0 0
0 0.1709 0
0 0 11.3448

The proof that I1 , I2 and I3 are invariant with respect to coordinate system follows from
showing that they will remain the same if expressed in components of another coordinate
system Tij0 . By using the coordinate transformation Tij = lik T
Tkl0 llj and the orthogonal
property of the coordinate transformation matrix lij we can rewrite I1 as:
T
I1 (Tij ) = Tii = lik Tkl0 lli = lli lik
T
Tkl0 = Tkk
0
= I1 (Tij0 ) (1.50)
| {z }
δlk

The second invariant I2 consists of I1 and Tij Tij that can be rewritten as:
T
Tij Tij = lik Tkl0 llj lim
T 0
Tmn T
lnj = lki lim T
lnj ljl Tkl0 Tmn
0
= Tkl0 Tkl0 (1.51)
| {z } | {z }
δkm δnl

whereby I2 (Tij ) = I2 (Tij0 ). To show that I3 also is invariant we first note that since
T T T
lli lik = δlk then det(lli lik ) = det(lli ) det(lik ) = 1. Now we use this in I3 :
| {z }
no sum on i

T
I3 (Tij ) = det(Tij ) = det(lik Tkl0 llj ) = det(lik
T
) det(Tkl0 ) det(llj ) = det(Tkl0 ) = I3 (Tkl0 ) (1.52)

1.5 Spatial derivatives

x
2
1

A tensor field describes how the tensor depends on the spatial location x in the body Ω
and the time t, e.g.
• Scalar field (such as temperature, pressure) φ = φ(x, t) or φ = φ(xi , t)
 22 Chapter 1. Tensors

• Vector field (such as displacement, velocity, force) u = u(x, t) or ui = ui (xj , t)

• Second-order tensor field (such as stress, strain) T = T (x, t) or Tij = Tij (xk , t).
To measure how such quantities change within the body the gradient (differential vector)
operator ∇ is introduced as
∂
∇ = êi (1.53)
∂xi
By applying the gradient operator via an open product (from the left) to a scalar field
φ(x, t), a vector field u(x, t) and a second-order tensor field T (x, t) the following results
are obtained
∂φ ∂uj ∂Tjk
∇φ = êi , ∇u = êi êj , ∇T = êi êj êk . (1.54)
∂xi ∂xi ∂xi
It can be noted that the tensor fields are always increased by one degree in this using this
procedure. Later we will also need to apply ∇ from the right on a vector field u(x, t)
which is defined as
∂ui
u∇ = êi êj (1.55)
∂xj
By instead applying the gradient operator via a scalar product (from the left) to a vector
field u(x, t) and a second-order tensor field T (x, t) result in
∂ui ∂Tij
∇·u= , ∇·T = êj . (1.56)
∂xi ∂xi
This is also called the divergence with the following notation

div(u) = ∇ · u , div(T ) = ∇ · T . (1.57)

For the divergence operator the tensor fields are always decreased by one degree.
Another product that can be used with the gradient operator is the vector product. The
vector product with the gradient operator defines the curl of a vector field

curl(u) = ∇ × u = eijk ∂i uj êk (1.58)

To further compress the notation we introduce the index form of the gradient operator,
∂j = ∂/∂xj = êj · ∇, or even more compactly, a subscripted comma which for example
results in:
∇u = ∂i uj êi êj = uj,i êi êj , div(T ) = ∂i Tij êj = Tij,i êj . (1.59)
Later in this text we will omit the base vectors and simply work with the components of
the tensors e.g. ∂ui /∂xj , ui,j , Tij,j etc.
 Example 1.7 a) The temperature varies in the coordinate system {êi } with coordinates
x1 , x2 as Φ(x1 , x2 ) = x21 + (x2 /α)2 + β = 0. The temperature gradient becomes:

∇Φ = 2 x1 ê1 + 2 x2 /α2 ê2

 1.6 Divergence theorem 23

b) Assume that the displacement field is u(x) = x2 ê1 + L e(−x1 −x2 )/a ê2 . The strain  is
defined as  = (∇ u + u ∇)/2 is obtained as

 = 1/2 (uj,i + ui,j )êi êj = 1/2 (−L/a e(−x1 −x2 )/a + 1)ê1 ê2 +
1/2 (1 − L/a e(−x1 −x2 )/a )ê2 ê1 + 1/2 (−L/a e(−x1 −x2 )/a )ê2 ê2

c) For the displacement field in b) the divergence of u becomes:

∇ · u = −L/a e(−x1 −x2 )/a

d) For the displacement field in b) the curl of u becomes:

∇ × u = e123 u2,1 ê3 + e213 u1,2 ê3 = (−L/a e(−x1 −x2 )/a − 1) ê3

Problem 17 Show that:

a) ∇(a · x) = a + ∇a · x
b) ∇ · (a × b) = (∇ × a) · b − (∇ × b) · a
c) ∇ · (A · b) = (∇ · A) · b + A : ∇b

Assignment 6 If a is a vector. Show that:

a) ∇ · (∇ × a) = 0
b) a × (∇ × a) = 1/2 ∇(a · a) − a · ∇a

1.6 Divergence theorem

n̂

Γ
Ω

x
2
1

Gauss’ divergence theorem is an important and useful theorem, which allows us to convert
the volume integral of a divergence into a surface integral as follows
Z I Z I
∇ · u dx = n̂ · u ds or ui,i dx = n̂i ui ds (1.60)
Ω Γ Ω Γ
 24 Chapter 1. Tensors

where Γ is the closed boundary surface of Ω, and n̂ is the outward normal unit vector to
Γ. This theorem can now be applied for tensor fields T by setting ui = Ti1 , Ti2 and Ti3 .
Thereby we obtain  R H
Ω Ti1,i dx = n̂i Ti1 ds



 Γ
R H
 Ω Ti2,i dx = Γ n̂i Ti2 ds


R

Ti3,i dx =
H
n̂i Ti3 ds
Ω Γ

which can be summarized as:

Z I Z I
Tij,i dx = n̂i Tij ds or div(T ) dx = n̂ · T ds (1.61)
Ω Γ Ω Γ

If we instead apply the Gauss’ divergence theorem to a scalar field for the three cases





u1 = φ, u2 = u3 = 0

u1 = 0, u2 = φ, u3 = 0


u = 0, u2 = 0, u3 = φ

1

we obtain Z I Z I
∂φ,i dx = n̂i φ ds or ∇φ dx = n̂φ ds (1.62)
Ω Γ Ω Γ
In practice, the name divergence theorem refers to equations (1.60), (1.61) and (1.62).
H
Problem 18 By using the divergence theorem show that: Γ xi n̂j ds = V δij .

R H R
Assignment 7 Proove the following formula: Ω ϕi ∂j σij dx = Γ n̂j σij ϕi ds− Ω σij ∂j ϕi dx.
2. Stress, motion and deformation

2.1 Stress analysis

The stress (also called traction) vector t(n̂) is defined as the force acting on an area with
normal n̂. In a point of a body the stress vector is defined as
∆F
t(n̂) = lim (2.1)
∆S→0 ∆S

n̂

∆S

A property of the stress vector is that it must follow Newton’s third law for action and
reaction. Therefore, in the same point of a body the stress vector on the area with normal
n̂ and normal −n̂ must be opposite. This means that

t(n̂) = −t(−n̂). (2.2)

To find a relation between the normal n̂ and the stress vector t(n̂) we study a tetrahedral
element:

x3
t(n̂)
n̂

x2

x1

The tetrahedron is assumed to have the four surfaces defined as

1. normal n̂ and area A subjected to stress vector t(n̂),
 26 Chapter 2. Stress, motion and deformation

2. normal −ê1 and area A1 subjected to stress vector t(−ê1 ),

3. normal −ê2 and area A2 subjected to stress vector t(−ê2 ),
4. normal −ê3 and area A3 subjected to stress vector t(−ê3 ).
The relation between the areas A, A1 , A2 , A3 and the components of n̂ can be derived
via the divergence theorem as n̂i = Ai /A, see the following example.
 Example 2.1 Choose φ = 1 and apply the divergence theorem according to (1.62) to the
tetrahedron
I I Z
n̂φ ds = n̂ ds = −ê1 A1 − ê2 A2 − ê3 A3 + n̂ A = ∇φ = 0
Γ Γ Ω

From these three equations we can identify that Ai = n̂i A. 

The next step is now to study equilibrium of the tetrahedron:

t(n̂) A + t(−ê1 ) A1 + t(−ê2 ) A2 + t(−ê3 ) A3 = 0.

If we use the relation between the areas and Newton’s third law we obtain:

t(n̂) = t(ê1 )n̂1 + t(ê2 )n̂2 + t(ê3 )n̂3 . (2.3)

The second-order stress tensor σ is defined based on t(êi ) such that

[σij ] = [tj (êi )] (2.4)

or more explicitly
   

σ11 σ12 σ13  
t1 (ê1 ) t2 (ê1 ) t3 (ê1 ) 
σ21 σ22 σ23 = t1 (ê2 ) t2 (ê2 ) t3 (ê2 ) (2.5)
   
  
   
σ31 σ32 σ33 t1 (ê3 ) t2 (ê3 ) t3 (ê3 )

This can be graphically shown as (here 2d):

σ22 t(ê2 )

σ12
x2 σ12 t(ê1 )
x1
σ11 σ11

t(−ê1 ) σ12
σ12

σ22
t(−ê2 )
2.1 Stress analysis 27

To sum up, the relation between the stress vector t and normal vector n̂ is obtained via
the stress tensor σ as follows:

t = n · σ = σ T · n or ti = nj σji = σijT nj . (2.6)

This relation is the so-called Cauchy’s formula. As will be proven later in the course, the
stress tensor is symmetric due to principle of angular momentum i.e. σ = σ T and, hence,
this relation can be written as t = σ · n̂. We can conclude that if the stress tensor σ is
known in a point of the body then it is possible to compute the stress vector t on any
plane through the point. This is called the Cauchy’s stress principle.
Often the components of the stress tensor are divided into normal stresses and shear
stresses. The normal stresses are the diagonal components of the stress tensor i.e. σ11 ,
σ22 and σ33 whereas the shear stresses are the off-diagonal components i.e. σ12 , σ23 and
σ13 . Note that the terminology normal and shear components relate to what plane that
is chosen. In the figure above the choice of plane is defined by the normal ê1 or ê2 . In
general, the normal component of the stress on a plane with normal n̂ is obtained from

σnn = n̂ · t = n̂ · σ · n̂ = σ : (n̂ n̂) = σij n̂i n̂j . (2.7)

Let us now adopt the concept of eigenvalues and eigenvectors for a stress tensor σ. The
eigenvector is a direction n̂ that is not changed upon a scalar multiplication with the
stress tensor σ:

σ
n̂ t = λ n̂

This means that on a plane with the normal being an eigenvector of σ then the stress
vector t is parallel to the normal i.e. t = λ n̂. In other words, on such a plane only the
normal components are non-zero.
Often the stress tensor σ is additatively decomposed into a deviatoric σ dev and a spherical
(hydrostatic) tensor σm δ as follows:

σ = σ dev + σm δ or σij = σijdev + σm δij (2.8)

with
σm = σkk /3 and σijdev = σij − σm δij . (2.9)
 28 Chapter 2. Stress, motion and deformation

Problem 19 Assume that the stress tensor field σ is represented in the coordinate
system êi with the following components
 

100 50 200 x2 
[σij ] =  50 100 400 x1 
 
 
200 x2 400 x1 100

where the stress components are in [MPa] and the coordinates are in [mm]. At the
√
point x = ê1 + 2 ê2 + 3 ê3 [mm] and on the plane with normal n̂ = (ê1 − ê2 + ê3 )/ 3
a) Determine the stress vector t
b) Determine the normal and shear components of t.
Answers:
a)  
450
1  
[t] = √ 350 MPa
 
3 
100
b) σnn ≈ 66.7 MPa, ts ≈ 327 MPa.

x2

weld
n̂

x1

Problem 20

A welded structure is subjected to a homogeneous plane stress condition described in

a Cartesian coordinate system 123 as:
 

σ̄ σ̄/3 0 
[σij ] =  σ̄/3 σ̄ 0 
 
 
0 0 0
√
The direction of the weld is given by the normal [n̂] = 1/ 5 [1 2 0]T (see figure).
a) Determine the stress (traction) vector t acting on the weld expressed in σ̄.
b) Assume that the largest allowable shear stress in the weld is 200 [MPa]. What is
the largest allowable value of σ̄?
Answers
 2.2 Continuum motion 29
√
a) [t] = σ̄/(3 5) [5 7 0]T
b) σ̄ ≤ 1000 MPa.

Assignment 8 Given the stress tensor σ (here represented in a matrix format with
components in a ê1 , ê2 , ê3 –system)
 

30 0 10 
[σ] = 
 0 30 10

 MPa
 
10 10 30

Compute the corresponding deviatoric stress tensor σ dev . For the deviatoric stress
compute the principal stresses, principal directions, invariants (acc to (1.49)) and
the obtained stress vector on the plane defined by the points (0, 0, 0), (2, −1, 0) and
(−4, 2, 3).

2.2 Continuum motion

The motion of a continuum (material volume) is shown in Figure 2.1. A material particle
P may be identified by its initial (or reference) position X. The current position x, of a
material particle is then defined by a function

xi = xi (X, t) (2.10)

The displacement u of a particle P is defined as

ui = xi − Xi (2.11)

A key quantity that describes the deformation of the body (material volume) is the
deformation gradient F . The deformation gradient describes the relation between a
line element dX at the material particle P in the initial (undeformed) body and the
corresponding line element dx at the material particle P in the current (undeformed)
body, i.e.
∂ xi ∂ ui
dx = F · dX or Fij = = δij + or F = x∇0 (2.12)
∂ Xj ∂ Xj
which is also illustrated in Figure 2.2. Based on the deformation gradient F a number of
strain measures can be defined. An example is the frequently used Green-Lagrange strain
E defined as follows:
1 T 
E= F ·F −δ or (2.13)
2 !
1 T  1 ∂ui ∂uj ∂uk ∂uk
Eij = Fik Fkj − δij = + + (2.14)
2 2 ∂Xj ∂Xi ∂Xi ∂Xj
 30 Chapter 2. Stress, motion and deformation

P
u
P

X x

Figure 2.1: Illustration of motion of a continuum.

dx

dX
F

Figure 2.2: Illustration of deformation gradient.

For the special case of small deformations E approaches the usual small strain tensor ,
i.e.
! !
1 ∂ui ∂uj 1 ∂ui ∂uj
Eij ≈ + ≈ + = ij (2.15)
2 ∂Xj ∂Xi 2 ∂xj ∂xi

We can note that both the Green-Lagrange strain E and the small strain tensor  are
symmetric, i.e. E T = E and T = .

2.3 Lagrangian and Eulerian description

Physical field quantities can be described in either a Lagrangian (or sometimes called
material) or Eulerian description:
2.3 Lagrangian and Eulerian description 31

• Lagrangian description of scalars, vectors and second-order tensors:

φ = φ(X, t) , u = u(X, t) , T = T (X, t) or

φ = φ(Xi , t) , ui = ui (Xj , t) , Tij = Tij (Xk , t)

• Eulerian description of scalars, vectors and second-order tensors:

φ = φ(x, t) , u = u(x, t) , T = T (x, t) or

φ = φ(xi , t) , ui = ui (xj , t) , Tij = Tij (xk , t)

An important field quantity is the velocity v of a material particle P. The velocity is

defined as the time derivative of the position vector x:

dx(X, t) dxi (Xj , t)

v= or vi = (2.16)
dt dt
whereby the velocity is described in an Lagrangian description v(X, t). By assuming that
the initial position of the particle X can be expressed in terms of x and t we can write
the velocity in Eulerian description

v = v(x, t) (2.17)

Next follows three examples to illustrate the introduced concepts regarding motion.
 32 Chapter 2. Stress, motion and deformation

 Example 2.2 Simple shear of a quadratic disc (side length h0 ) where the upper boundary
moves horizontally with velocity v0 :

ê2

ê1

The motion can be expressed as:






x1 (X1 , X2 , X3 , t) = X1 + X2 v0 t/h0

x2 (X1 , X2 , X3 , t) = X2

 x (X , X , X , t) =

X3
3 1 2 3

whereby the velocity v can be obtained, in Lagrangian description, as:

 

X2 v0 /h0 
vi =  0
 

 
0

By using the expression for the motion the velocity can be written in an Eulerian descrip-
tion as  

x 2 v0 /h0 
vi =  0
 

 
0
Based on the expression for the motion we can also obtain the deformation gradient F as
 

1 v0 t/h0 0 
Fij =  0 1 0 
 
 
0 0 1

and the Green Lagrange strain E

 
0 v0 t/h0 0 
1 T  1
Eij = Fik Fkj − δij = ... =  v0 t/h0 (v0 t/h0 )2 0 
 
2 2 
0 0 0

The displacement vector u = x − X is given as:

 

X2 v0 t/h0 
ui = 
 0


 
0
2.3 Lagrangian and Eulerian description 33

whereby the small strain tensor  becomes

 
! 0 (v0 t/h0 )/2 0 
1 ∂ui ∂uj 
ij = + = (v0 t/h0 )/2 0 0 
 
2 ∂Xj ∂Xi  
0 0 0


 Example 2.3 Pure elongation of a quadratic disc (side length h0 ) where the upper bound-
ary moves vertically with velocity v0 :

ê2

ê1

The motion can be expressed as:






x1 (X1 , X2 , X3 , t) = X1

x2 (X1 , X2 , X3 , t) = X2 + X2 v0 t/h0

 x (X , X , X , t) =

X3
3 1 2 3

whereby the velocity v can be obtained, in Lagrangian description, as:

 

0 
vi =  X2 v0 /h0
 

 
0
By using the expression for the motion the velocity can be written in an Eulerian descrip-
tion as  

0 
vi = 
 x2 v0 /(h0 + v0 t) 

 
0
Based on the expression for the motion we can also obtain the deformation gradient F as
 

1 0 0 
Fij =  0 1 + v0 t/h0
 

 
0 0 1
and the Green Lagrange strain E
 
0 0 0 
1 T  
2
Eij = Fik Fkj − δij = ... =  0 (v0 t/h0 ) /2 + v0 t/h0 0 


2  
0 0 0
 34 Chapter 2. Stress, motion and deformation

The displacement vector u = x − X is given as:

 

0 
ui =  X2 v0 t/h0 
 
 
0

whereby the small strain tensor  becomes

 
! 0 0 0 
1 ∂ui ∂uj 
ij = + = 0 v0 t/h0 0 
 
2 ∂Xj ∂Xi  
0 0 0


 Example 2.4 Pure rotation around the left corner of a quadratic disc (side length h0 )
with rotational velocity ω:

ê2

ωt
ê1

q
An arbitrary point’s initial location in the disc is described by the distance R = X12 + X22
to the left corner and angle α0 = atan(X2 /X1 ) (from the ê1 axis). During rotation the
angle changes with rotation according to α = α0 +ωt whereby the motion can be expressed
as: 
 x (X , X2 , X3 , t) = R cos(α)
 1 1




x2 (X1 , X2 , X3 , t) = R sin(α)

 x (X , X , X , t) =

X3
3 1 2 3

which can be (after some manipulations) written as






x1 (X1 , X2 , X3 , t) = X1 cos(ωt) − X2 sin(ωt)

x2 (X1 , X2 , X3 , t) = X1 sin(ωt) + X2 cos(ωt)

x3 (X1 , X2 , X3 , t) = X3


2.3 Lagrangian and Eulerian description 35

The velocity v can be obtained, in Lagrangian and Eulerian description, as :

   

ω (−X1 sin(ωt) − X2 cos(ωt))  
−x2 
vi =  ω (X1 cos(ωt) − X2 sin(ωt)) =ω  x1
   
 
   
0 0
Based on the expression for the motion we can also obtain the deformation gradient F as
 

cos(ωt) − sin(ωt) 0 
Fij = 
 sin(ωt) cos(ωt) 0 

 
0 0 1
and the Green Lagrange strain E
 
0 0 0 
1 T  
Eij = Fik Fkj − δij = ... = 
 0 0 0 

2  
0 0 0
The displacement vector u = x − X is given as:
 
X (cos(ωt) − 1) − X2 sin(ωt) 
 1
 
 
ui =  
X1 sin(ωt) + X2 (cos(ωt) − 1)
 
 
 
0
whereby the small strain tensor  becomes
 
! cos(ωt) − 1 0 0 
1 ∂ui ∂uj 
ij = + = 0 cos(ωt) − 1 0 
 
2 ∂Xj ∂Xi  
0 0 0


Problem 21 The motion of a body is given as

x(X, t) = X1 ê1 + (X2 + X1 t/t0 ) ê2 + (X3 + X2 (1 − e−t/t0 ) ê3

where t0 is a constant [s]. Determine

a) The deformation gradient F
b) The Green-Lagrange strain E
c) The velocity in Eulerian description v(x)
Answers
 

1 0 0
a) [F ] = t/t0

1 0

 
0 (1 − e−t/t0 ) 1
 36 Chapter 2. Stress, motion and deformation
 
2

(t/t ) /2
0 (t/t0 )/2 0 
−t/t0 2
b) [E] =  (t/t0 )/2 1/2 (1 − e ) 1/2 (1 − e−t/t0 ))
 
 
0 1/2 (1 − e−t/t0 ) 0
c)
h iT
[v] = 0 x1 /t0 1/t0 (x2 − x1 t/t0 ) e−t/t0

Assignment 9 For a quadratic disc with side lengths h0 we have the motion:





x1 (X1 , X2 , X3 , t) = X1 + X2 v0 t/h0

x2 (X1 , X2 , X3 , t) = X2 + X1 v0 t/h0

 x (X , X , X , t) =

X3
3 1 2 3

where v0 is a velocity.
a) Illustrate how the disc will deform (in 2D).
b) Determine the velocity vector in Lagrangian and Eulerian description.
c) Determine the Green-Lagrange strain E.

2.4 Material time derivative

Physical field quantities such as temperature, velocity, stress tensor change with time.
This change is naturally described as the time derivative of the physical quantity of a
material particle in the continuum. The particle is uniquely identified by the Lagrangian
(material) vector X. Therefore, it is useful to introduce the material time derivative
which is denoted D(•)/Dt, (•) ˙ or d(•)/dt. If the physical quantity φ is described in an
Lagrangian description φ(X, t)

Dφ (X, t) dφ (X, t)
= φ̇ (X, t) = , (2.18)
Dt dt
whereas the material time derivative of a field quantity described in an Eulerian descrip-
tion φ (x, t) is obtained by the chain rule:

dφ (x(X, t), t)
φ̇ (x(X, t), t) =
dt 
∂φ (x, t) ∂xi (X, t) ∂φ (x, t) 
= +  (2.19)
∂xi ∂t ∂t x

The first part in the result is the convective part while the second part is the time deriva-
tive of φ in a spatial position x.
 2.5 Reynolds’ transport theorem for a material volume 37

 Example 2.5 For Problem 21 the velocity v is given in Lagrangian and Eulerian coordi-
nates as follows
   

0  
0 
[v] = 

X /t
1 0

 = 
 x /t
1 0


   
1/t0 X2 e−t/t0 1/t0 (x2 − x1 t/t0 ) e−t/t0

The acceleration can be obtained from the Lagrangian form directly as

 

0 
[a] = 

0


 
−1/t20 X2 e−t/t0

and the Eulerian form as


∂vi (x, t) ∂xj (X, t) ∂vi (x, t) 
ai = + 
∂xj ∂t ∂t x

which in matrix form becomes

  

0 0 0  0 
[ai ] =  1/t 0 0 x /t +
  
0   1 0
  
−t/t20 e−t/t0 1/t0 e−t/t0 0 1/t0 (x2 − x1 t/t0 ) e−t/t0
   

0  
0 
+

0 =
 
0


   
2 −t/t0 2 −t/t0
1/t0 (−x2 /t0 + x1 t/t0 − x1 /t0 ) e −1/t0 (x2 − x1 t/t0 ) e


2.5 Reynolds’ transport theorem for a material volume

In the balance laws physical quantities are integrated over the volume of interest. The
integration can be performed for the current volume of the continuum Ω:
Z
φ dx
Ω

By substituting the volume Ω to the initial volume (undeformed) of the continuum Ω0 we

obtain (using results from Math see e.g https://en.wikipedia.org/wiki/Integration_
by_substitution): Z Z
φ dx = φ J dX (2.20)
Ω Ω0

where J = det(F ).
Note: The volume change of a body V /V0 is given by J = det(F ). This follows immedi-
ately from (2.20) by setting φ = 1.
 38 Chapter 2. Stress, motion and deformation

The material time derivative of a volume integral of φ can now be obtained as (using that
Ω0 is constant):
d Z d Z Z
φ dx = φ J dX = φ̇ J + φ J˙ dX (2.21)
dt Ω dt Ω0 Ω0

The time derivative of the volume change J˙ is given by (here without any proof):

J˙ = div(v) J or J˙ = vi,i J (2.22)

whereby we can obtain Reynold’s transport theorem:

!
d Z Z
dφ ∂vi
φ dx = +φ dx. (2.23)
dt Ω Ω dt ∂xi

Assignment 10 Show that (2.23) for φ(xi , t) can be written as:


d Z Z
∂φ  I
φ dx =  dx + n̂i vi φ ds
dt Ω Ω ∂t x Γ

2.6 Reynolds’ transport theorem for a control volume

In fluid mechanics quantities are often measured in a control volume Ωc with boundary
Γc . This control volume do not in general follow the movement of the material particles
in the continuum as is illustrated in the figure below.

t=0 t

Ωc
P uc
P
Ω0 Ωc,0
Ω

xc = x
Xc

At time t the position of a control volume element element is xc and its velocity v c . Note
that although the position of a control volume element and a material point element P
coincides at time t, i.e. xc = x, their velocities v c and v differ.
R
How Ωc φ dx changes with time can be obtained from the general form of Reynold’s
transport theorem: !
d Z Z
dφ ∂vic
φ dx = + φ c dx. (2.24)
dt Ωc Ωc dt ∂xi
2.6 Reynolds’ transport theorem for a control volume 39

This follows from the same arguments as in (2.21)-(2.23). Two special cases are common:
• The control volume elements are equal material point elements (they follow the
deformation) then v c = v and (2.23) is re-obtained. This assumption that the
control volume elements are ”nailed” to the material elements are often used in
solid mechanics.
• The control volume elements are fixed in time which means that v c = 0. This gives
that if φ(xci , t) we obtain:
 
d Z Z
dφ Z
∂φ  ∂φ c Z
∂φ 
φ dx = dx =  + v dx = dx
∂xci i

dt Ωc Ωc dt Ωc ∂t xc Ωc ∂t xc
3. Field equations

3.1 Physical quantities of a continuum

We consider a body occupying a region Ω at time t. The state of the body is assumed
to be given by the quantities: mass M ; momentum P ; angular momentum N ; kinetic
energy K and internal energy U . Before giving the expressions of these quantities we
remind us the corresponding expressions of P , N and K are given for a point mass m as:

P = mv or Pi = m vi
N = mx × v or Ni = m eijk xj vk
K = m |v|2 /2 or K = m vi vi /2

With this at hand and by assuming a density field ρ(x, t) and a velocity field v(x, t) then
the M , P , N and K for a body can be expressed as:
Z
M = ρ dx, (3.1)
ZΩ
Pi = ρvi dx, (3.2)
ZΩ
Ni = eijk xj ρvk dx, (3.3)
Ω
Z
1
K = ρvi vi dx. (3.4)
Ω 2
In addition to these quantities the internal energy U is also introduced. U represents
energy such as strain energy and thermal energy which together with the kinetic energy
sums up to total energy of the body. Later U will be given more explicitly but at this
stage we assume that the body has an internal energy density field e(x, t) such that:
Z
U= ρe dx. (3.5)
Ω

3.2 Input quantities

A schematic figure of a continuous body is given in Figure 3.1 with the field variables ρ,
v and e.
Now we assume that the body is subjected input quantities that can change the state of
the body, see Figure 3.1. The mechanical loading is given by a volume force f (force per
unit mass) and a boundary load t (force per unit area). Thereby the total force F and
 42 Chapter 3. Field equations

ti
Γ

E -qn̂
fi
Ω
ρ, vi , e
xi

Figure 3.1: Illustration of a continuum Ω with boundary Γ.

moment M and mechanical power input to the body become:

Z I
Fi = ρfi dx + ti ds, (3.6)
ZΩ Γ I
Mi = eijk xj ρfk dx + eijk xj tk ds., (3.7)
ZΩ I Γ

Ẇ = ρfi vi dx + vi ti ds, (3.8)

Ω Γ

Additionally, the body is subjected to the internal heat source E (energy per unit mass)
and heat input −qn̂ (energy per unit area) resulting in the heat power input:
Z I
Ḣ = ρE dx + −qn̂ ds (3.9)
Ω Γ

3.3 Physical conservation principles

Now the physical conservation principles are used to define how the state i.e. the mass M ,
the linear momentum P , the angular momentum N and the total energy K + U change
of the body change with the mechanical and heat input.

3.3.1 Conservation of mass

Mass is in classical mechanics assumed to be conserved which can be written as:

d Z
Ṁ = ρ dx = 0 (3.10)
dt Ω
By using Reynold’s transport theorem (2.23) we obtain:

d Z Z
Ṁ = ρ dx = (ρ̇ + ρvi,i ) dx (3.11)
dt Ω Ω
 3.3 Physical conservation principles 43

The mass conservation is assumed for all choices of Ω (this argumentation is called local-
ization) whereby:
ρ̇ + ρ vi,i = 0 in Ω (3.12)
This equation is called the continuity equation.
The continuity equation can be used together with Reynold’s transport theorem to show
that for an arbitrary field quantity φ:
d Z Z
d Z
ρ φ dx = (ρ φ) + ρ φ vi,i dx = ρ φ̇ dx (3.13)
dt Ω Ω dt Ω

This result is denoted the modified Reynolds’ transport theorem.1

Problem 22 Assume that the Eulerian description of the velocity field of a body is
given as:  
2 v x /L
 0 1 
[vi ] =  v0 x2 /L 


 
v0 x3 /L
For the case that v0 /L = 1 [1/s] describe how the density ρ is changing from its initial
value ρ0 .
Answer: ρ = ρ0 e−4t

3.3.2 Conservation of linear and angular momentum - Newton’s laws

Newton’s laws state that the material time derivative of the linear momentum P and the
angular momentum N are determined by the applied force F and moment M as follows:

Ṗi = Fi
Ṅi = Mi

By using modified Reynold’s tranport theorem (3.13) and equations (3.2) as well as (3.6)
we obtain: Z Z I
Ṗi = ρ v̇i dx = ρfi dx + ti ds (3.14)
Ω Ω Γ
The next step is to use Cauchy’s stress principle ti = σji n̂j and the divergence theorem:
Z
ρ v̇i − σji,j − ρ fi dx = 0 (3.15)
Ω

The localization argument now yields the momentum equation:

σji,j + ρ fi = ρv̇i (3.16)

1 d R d R R
It can also be shown by instead integrating over the mass: Ω
ρ φ dx = m
φ dm = m φ̇ dm =
R dt dt
Ω
ρ φ̇ dx
 44 Chapter 3. Field equations

By using the same steps for the angular momentum Ṅi = Mi we will arrive at the result
that the stress tensor must be symmetric:

σT = σ or σij = σji (3.17)

However, we leave those derivations to Hand-in assignment 12.

Problem 23 For a plate with length L, height H and thickness b the non-zero stress
field components are given by:
!
6 f x2 2 2 2 x32 /3 − H 2 /10
σ11 =− (L − x 1 ) 1 + ,
H2 L2 − x21
σ22 = −f x2 (1 − 4 x22 /H 2 )/2, σ12 = 3 f x1 (1 − 4 x22 /H 2 )/2

The plate is subjected to quasistatic conditions, i.e. v̇i ≈ 0, compute the volume force
[N/m3 ] that acts on the plate.
Answer:

[ρ fi ] = [0 − f 0]T

Assignment 11 A rectangular disc with width b1 , height b2 and thickness h rotates

with angular velocity ω = ω0 e−t around x3 axis.
Assume that the density is constant ρ0 .

x2

x1

Determine the angular momentum of the disc.

Assignment 12 Use the principle of angular momentum to show that the stress tensor
σ is symmetric.

3.3.3 Conservation of energy - 1st law of thermodynamics

The 1st law of thermodynamics says that the material time derivative of the total energy
of a body is equal to the power input:

K̇ + U̇ = Ẇ + Ḣ (3.18)
3.3 Physical conservation principles 45

If we now use modifed Reynold’s transport theorem (3.8) and equations (3.4), (3.5),(3.8)
as well as (3.9) we obtain:

Z Z I Z I
ρvi v̇i + ρ ė dx = ρ fi vi dx + ti vi ds + ρE dx + −qn̂ ds (3.19)
Ω Ω Γ Ω Γ

Before we can use the localization argument the boundary integrals must be changed to
volume integrals. The first of the boundary integrals is re-written using the Cauchy’s
stress theorem as follows:

I I Z
ti vi ds = n̂j σji vi dx = σji,j vi + σji vi,j dx (3.20)
Γ Γ Ω

The second boundary integral is re-written by assuming that the heat flux qn̂ is given by
the heat flux vector q according to:

qn̂ = n̂ · q = n̂i qi (3.21)

thereby the divergence theorem gives us:

I Z
−qn̂ ds = −qi,i dx (3.22)
Γ Ω

Now the 1st law of thermodynamics can be written as:

Z
ρvi v̇i + ρ ė − ρ fi vi − σji,j vi − σji vi,j − ρE + qi,i dx = 0 (3.23)
Ω

The balance of linear momentum (3.16) now together with the localization argument
results in the energy equation:

ρė − vi,j σij + qi,i = ρE. (3.24)

Assignment 13 By taking the material time derivative of the kinetic energy in (3.4)
and using the balance of linear momentum show that for a purely mechanical problem
(Ḣ = 0): Z
K̇ + σij vi,j dx = Ẇ
Ω
 46 Chapter 3. Field equations

3.4 Summary of field equations and field variables

A summary of the field equations and the field variables for a continuum are shown in
the table below.

Balance law Field variables No equations

ρ̇ + ρ vi,i = 0 ρ, vi 1
ρ v̇i = σji,j + ρ fi σij 3
σij = σji σij 3
ρ ė = σij vi,j − qi,i + ρ E e, qi 1
Tot. 17 Tot. 8

By counting the number of equations and unknowns we can conclude that 9 additional
equations that must be formulated. These equations are the constitutive models that
should mimic the material behaviour observed in experiments.
4. Constitutive models

The standard way to define constitutive models in solid mechanics, fluid mechanics and
heat transfer is to describe how internal energy e, stress σ and heat flux q depend on other
field variables such as density ρ, temperature θ, temperature gradient θ,i displacement
gradient ui,j and velocity gradient vi,j . These models with their material parameters are
based on experimental observations. In this section we will merely introduce some of most
common (and simplest) constitutive models for heat transfer, fluids and solids.
The constitutive models are determined by the material properties. The material can be
homogeneous meaning that properties are the same in the body Ω otherwise the material
is heterogeneous. If the properties are the same in all directions then the material is called
isotropic. For some materials the properties are anisotropic. Examples of the latter are:
wood, composites and fibre reinforced concrete.

4.1 Fourier’s law of thermal conductivity

Heat can be transfered by convection (motion of fluid), radiation (electromagnetics) and

conduction (diffusion processes). For heat conduction the standard constitutive model is
Fourier’s law. For an isotropic material this law takes the form:

qi = −k Θ,i (4.1)

where the linear coefficient k is the thermal conductivity and Θ is the temperature. The
temperature Θ is now assumed to give the internal energy e according to:

e = cp Θ (4.2)

where the constant cp is the heat capacity of the material. If we consider a purely thermal
problem (i.e. assuming σij = 0) then the energy equation (3.24) reads as follows:

ρ ė = −qi,i + ρ E

By inserting (4.1) and (4.2) then we obtain the transient heat conduction equation:

ρ cp Θ̇ = k Θ,ii + ρ E (4.3)
 48 Chapter 4. Constitutive models

4.2 Viscous fluids

The simplest possible contitutive of a fluid is an ideal fluid. In this model the stress is
assumed to be purely volumetric:

σij = −p(ρ, Θ) δij (4.4)

This means that such a fluid cannot sustain shear stresses. The pressure p is assumed to
follow the ideal gas law
p(ρ, Θ) = ρ R Θ/mg (4.5)

where R is the gas constant, mg is the mean molecular mass of the gas.
Most fluids are not ”ideal” since they are a bit ”sticky” and are able to sustain shear
stresses. Therefore, a viscous stress τ is introduced for and the stress σ is additively
decomposed according to
σij = −p(ρ, Θ) δij + τij (4.6)

For the case of an isotropic Newtonian viscous fluid we assume that the viscous stress τ
is linear in terms of the strain rate tensor D as follows

τij = λ∗ Dkk δij + 2 µ∗ Dij (4.7)

where the strain rate tensor D is defined from as the symmetric part of the velocity
gradient:
1
Dij = (vi,j + vj,i ) . (4.8)
2
In (4.7) the material parameters λ∗ and the dynamic (shear) viscosity µ∗ were introduced.
Now the stress becomes:

σij = −p(ρ, Θ) δij + λ∗ δij Dkk + 2 µ∗ Dij (4.9)

The mechanical pressure pmech = −σmm /3 can be computed as:

1 2
 
pmech = − σmm = p(ρ, Θ) − λ∗ + µ∗ Dkk (4.10)
3 3

If we introduce the Stoke’s condition that pmech = p(ρ, Θ) then we can for a Newtonian
viscous fluid obtain that
σkk
σij = σijdev + δij = 2 µ∗ Dij
dev
− p(ρ, Θ) δij (4.11)
3

where σ dev and D dev are the deviatoric stress and deviatoric strain rate tensor, respec-
tively. Stoke’s condition means that the pressure in the fluid is strain rate independent.
 4.3 Linear elastic isotropic solids 49

The Navier-Stoke’s equations are now simply obtained by inserting (4.11) into balance of
linear momentum (3.16) together with the continuity equation.
Often for fluids one can assume that they are incompressible. From the conservation of
mass and the incompressibility ρ̇ = 0 it follows from the continuity equation (3.12) that:

vi,i = Dii = 0

In this case D dev = D and, without using Stoke’s condition, we obtain (4.11) from (4.9).

Problem 24 Assume the constitutive equation σij = (−p + λ Dkk ) δij + 2 µDij . Show
that the equations of motion can be expressed in the velocity field as:

ρ v̇i = ρ fi − p,i + (λ + µ) vj,ij + µvi,jj .

4.3 Linear elastic isotropic solids

The constitutive model for linear elasticity is denoted Hooke’s law. Originally the law
was defined for a linear spring but generalized to an isotropic solid it reads as follows

σij = λ kk δij + 2µ ij (4.12)

where ij is the small strain tensor defined in (2.15). A small strain assumption has been
made and it is therefore the small strain tensor can be used. This also means that the
density ρ can be assumed to be approximately constant. The model parameters λ and µ
are the Lame’s constants, which are related to Young’s modulus E and Poisson’s ratio ν
as follows
Eν E
λ= , µ=
(1 + ν) (1 − 2 ν) 2 (1 + ν)

Assignment 14 The equations

µui,jj + (λ + µ) uj,ji + ρfi = ρüi .

are called Navier’s equations and may be used to solve elastodynamic problems with
displacement-type boundary conditions. Derive these equations by combining the
momentum equation and Hooke’s law.
Index

1st law of thermodynamics, 44 Invariants, 19

2nd order tensors, 12
Kinetic energy, 41
Angular momentum, 41 Kronecker’s delta, 6

Cauchy’s stress principle, 27 Lagrangian description, 30

Conservation of energy, 44
Mass, 41
Conservation of mass, 42
Modified Reynolds transport theorem, 43
Constitutive models, 47
Momentum, 41
Continuity equation, 43
Momentum equation, 43
Coordinate system transformation, 17
Curl operator, 22 Newtonian viscous fluid, 48

Deformation gradient, 29 Orthonormal base vectors, 5

Density, 41
Permutation symbol, 7
Deviatoric stress, 27
Principal values and directions, 18
Displacement, 29
Divergence operator, 22 Reynolds’ transport theorem, 37, 38
Divergence theorem, 23
Scalar product, 6, 13, 15
Double contraction, 15
Skew-symmetric tensor, 16
Einstein’s summation convention, 3 Spherical stress, 27
Eulerian description, 30 Stoke’s condition, 48
Strain, 30
Fourier’s law, 47 Strain rate tensor, 48
Stress, 25
Gradient operator, 21
Stress vector, 25
Green-Lagrange strain, 29
Symmetric tensor, 16
Heat conduction equation, 47
Vector length, 5
Hooke’s law, 49
Vector product, 7
Hydrostatic stress, 27
Vectors, 5
Ideal gas law, 48 Velocity, 31
Identity tensor, 14 Viscous fluid, 48
Index notation, 3 Viscous stress, 48
Internal energy, 41