Chapter II

Tensors: Computation and Analysis

Objectives of the course

- Familiarization of students with tensor notation.

- Manipulation on indicial expressions

- Knowing tensor operations

II.1 Indicial notation or convention of the summation

The components of a matrix or a vector may be written using an indicial notation to avoid the
writing ambiguity. In this notation, letter indices, subscripts or superscripts, are assigned to a
generic writing of letter representing the matrix or the vector. The application of indicial
notation rules, an index may appear either or twice on the term.

Consider the following sum.

s  a1 x1  a2 x2  .......  an xn

This relation can be expressed as

n
s   ai xi
i 1

II.1.1 Dummy index

The previous expression can be written as

n n n n
s   aj xj s   aj xj or s   ak xk or s   a x
j 1 j 1 k 1  1

The variation of the index in previous expressions doesn’t change the result of the sum.

The index I or j or k or  is called a dummy index, which leads to that the index is independent
of the used letter as index in the sum expression. Then, it’s possible to simplify the writing of
the relationship using the convention called Einstein’s summation convention. Thus, it can be
written as

s  ai xi

The index i in this relationship can be substituted with other indices without changing the
result of the sum. More, the summation convention can be applied with condition that the index
should never be repeated more than once.
II.1.2 Free index
The index appearing only once in each term of a tonsorial equation is called a free index.
Therefore, consider

yi  aij x j (j is a dummy index and I a free index)

Using the summation convention, the relationship can be rewritten as

y1  a1 j x j  a11 x1  a12 x2  ......  a1n xn

y2  a2 j x j  a21 x1  a22 x2  ......  a2n xn

………………………………………………..

yn  anj x j  an1 x1  an 2 x2  ......  ann xn

In matrix form, the indicial relation can formulated as  y   a x

The writing form of expressions shows the advantage offered by the summation convention.

Remarks :

- The indicial expressions such as ai bi xi or ai bij xi are not defined within this convention. The
index i is repeated more than twice time.

n
But the expression of the form a b x
i 1
i i i retains its summation sign.

That is, an index should never be repeated more than once when the
summation convention is used.

- For 3D problems, the number of terms described by n must be taken as 3. The previous
expressions can be written as

ai xi  a1 x1  a2 x2  a3 x3

bii  b11  b22  b33

 a11 x1  a12 x2  a13 x3


aij x j  a21 x1  a22 x2  a23 x3
a x  a x  a x
 31 1 32 2 33 3

Expanding the expressions:  kk , aij x ji , Tkl  k

II.2 Kronecker dalta

The Kronecker delta, denoted by  ij , is defined by

1 i  j
 ij  
0 i  j

Based on this way, it’s possible to inspire the matrix Krnocker delta that is

 11 12 13   1 0 0 

 ij     21  22  23    0 1 0 
  31  32  33   0 0 1 
   

The following expressions can be deduced (applying the indicial notation)

-  ii  11   22   33  3

 11 12 13  a1   1 0 0  a1   a1 

       
-  ij a j    21  22  23  a2   0 1 0 a2   a2 
 
  31  32  33  a   0 0 1  a   a 
  3    3   3 

 11 12 13  T11 T12 T13 

  
-  ij T jk    21  22  23  T21 T22 T23   Tik
  31  32  33  T31 T32 T33 
  
 
- ei .e j   ij

II.3 Permutation symbol

The permutation symbol, denoted by  ijk , is defined by

for an even permutation of i,j, k.
1
 for an odd permutation of i,j, k.
 ijk   1
0
 do not form.

This relationship allows writing

 ijk   jki   kij  1

 ikj   kji   jik  1

 iij   jji   jkk   kkk  .....  0

The application of the relationships leads to deduce.

     
ei  e j   ijk ek   jki ek   kij ek
  
The relation can be applied to any vectors v and w , the corresponding result is the vector A
that is

  
e1 e2 e3
    
A  vw  v1 v2 v3  (v2 w3  v3 w2 )e1  (v3 w1  v1 w3 )e2 
w1 w2 w3

(v1 w2  v2 w1 )e3
   
In indicial notation, the cross-product of the vectors v  vi ei and w  wi ei is
       
A  vw  vi ei wj e j  vi wj (ei e j )  vi wjijk ek

 and  are Kronecker delta and symbol of permutation, develop:

 kk

 ij

 213  312 113  222

im mj

II.4 Manipulations on indicial notation

- Substitution

 i  U im am and ai  Vimbm

To substitute ai in the first relation, it is necessary to change the index, thus

am  Vmnbn
The substitution can be evaluated

 i  U im (Vmnbn )  (U imVmn )bn

 i  Win bn : Win  (U imVmn )

- Multiplication

Let’s P  ambm and q  cm d m

The product p.q is

Pq  (ambm ).(cn dn )  ambm cn dn

- Factoring
 Tij n j   n j  0

Using the symbol of Kronecker ni  ij n j and substituting it the above relation

Tij n j  ij n j  0

Thus, we can write (Tij  ij )n j  0

- Contraction

It indicates a sum on the index. For example, aii is the contraction of aij .

If aij  ij   ij

Thus, aii   ii   ii  3   (11   22   33 )

II.5 Tensors

II.5.1 Linear transformation

 
T is a material transformation which transforms any vector v into another vector w .

T P’ w

P

v
 
This material transformation allows writingw  Tv ( or using vectors : w  T v ).
 
The material transformation is denoted linear if, for  is a scalar and v and w two vectors.
   
T (v  w)  T v  T w
 
T ( v)   T v
A tensor T is a linear material transformation or second order tensor which associates any
 
vector v into another vector w , such as w  Tv .
In other form, we can write
   
T ( v   w)   T v   T w  is a scalar.
 
Example: T is a material transformation which associates any vector v the vector  . Show the
nature of the transformation.
 
Thus,  Tv
 
Be two vectors a and b that
  
 Ta
 
And   Tb
  
Let c  a  b

Respecting the hypothesis of the transformation,

 
 Tc
We have,

Tc  Ta  Tb , which corresponds to
  
   
T is then not a linear transformation.

II.5.2 Order of a tensor

Tensors are the generalization of scalars, vectors and matrices. The description depends on the

reference (reference) adopted.

order Entity Dimension Component number Notation Tensor

notation

0 Scalar / / a a

1 Vector 2D 21 =2 ai a
3D 31 =3

2 Matrix 2D 22 =4 aij a
3D 32 =9

n / 2D 2n aijklm…… a
n
3D 3

A Cartesian tensor A having an order n is a function which associates with any reference

(2D/3D) a group (2n /3n ) of real components Aijkl…. That will be transformed according to the

relation
3
'
Aijklmn .....  
ijklmn....1
ail a jm akn ..... Almn....

With aij are the direction cosines of the unit vectors of the reference (ox'y'z') with respect to

the initial reference (oxyz).

II.5.3 first order tensor

A tensor of order 1 is a vector having 3 components vi in one reference (ox1x2x3) and wi in

another reference (ox’1x’2x’3) with rotating  around the axis ( x3 x3 ).

x3'
x3

 x2'
e3 
v
 O  x2
e1 e2

x1
x1'
'  
e1  cos  e1  sin  e 2
'  
e 2   sin  e1  cos  e 2
' 
e3  e3

The tensor of the transformation passage or the rotation is

 cos  sin  0
Tij     sin  cos  0 

 0 0 1 

For general case, w  Tv or wi  Tij v j

w1  T1 j v j  T11v1  T12v2  T13v3  cos  v1  sin  v2

w2  T2 j v j  T21v1  T22v2  T23v3   sin  v1  cos v2

w3  T3 j v j  T31v1  T32v2  T33v3  v3

II.5.4 Second order tensor

The tensor of order 2 is a matrix also of order 2 (problem 2D with 4 terms) or a matrix of order

3 (3D having 9 terms).

The terms of the tensor are

3
wkl  a
i , j 1
av
kl lj ij
 w11  a1i a1 j vij  a11a1 j v1 j  a12 a2 j v2 j  a13a3 j v3 j
 a11a11v11  a11a12 v12  a11a13v13
As an example,
 a12 a21v21  a12 a22 v22  a12 a23v23
 a13a31v31  a13a32v32  a13a33v33

General formulation,

F '  Qq '

With q '  T q and  F '  T  F

Thus, T  F  QT q

1
F   T  Q T q
In this case, the tensor T is orthogonal
t
F   T  Q T q

II.6 Components of tensor

The components of a vector depend on the base vectors used to describe the vectors. This will

also be applied for tensors.

  
Let ( e1 , e2 , e3 ) is the original reference of a rectangular Cartesian coordinate system. With the
     
transformation T, these vectors e1 , e2 , e3 become T e1 ; T e2 ; T e3 after transformation. Each of
these components is a vector that cab written as
   
T e1  T11 e1  T21 e 2T31 e3
   
T e 2  T12 e1  T22 e 2T32 e3
   
T e 2  T13 e1  T23 e 2T33 e3

In indicial writing,
  
T ei  T ji e j  Tijt e j

The components of the tensor T which are its parameters, they are written in the following

matrix as

T11 T12 T13 

T   T21 T22 T23 
T31 T32 T33 
 
 
The first column terms of the matrix are the components of the vector T e1 , that is the same
 
of the second and the third column, T e2 and T e3 .

II.7 Components of vector

   
As the above definition of tensor, any vector v into the vector w : w  Tv .
   
v  v1 e1  v2 e 2  v3 e3

Applying the vector relationship,

      
w  T (v1 e1  v2 e 2  v3 e3 )  v1 T e1  v2T e 2  v3T e3
           
w1  w.e1  e1 .T (v1 e1  v2 e 2  v3 e3 )  v1 e1T e1  v2 e1T e 2  v3 e1T e3
 
knowing that Tij  ei .T e j , thus

w1  v1 T11  v2T12  v3T13

The same method deals to

w2  v1 T21  v2T22  v3T23

w3  v1 T31  v2T32  v3T33

In matrix form, this relation can be written as
 w1  T11 T12 T13  v1 
    
 w2   T21 T22 T23  v2 
  T T T  
 w3   31 32 33  v3 

In tensor form,
wT v
Example:
  
Either the material transformation associated to the base (e1 , e2 , e3 ) such as
   
T e1  2e1  6e2  4e3
   
T e2  3e1  4e2  e3
   
T e3  2e1  e2  2e3
   
What is the transformation of the vector v  e1  2e 2  e3
II.8 Operations on tensors

II.8.1 Sun of tensors

T and S two material transformations, the sum of tensors is denoted with T+S is defined by
  
(T  S )v  T v  Sv
  
W v  T v  Sv
W T S
The components of the tensor W can be deduced
     
Wij  ei (T  S )e j  ei T e j  ei Se j
Wij  Tij  Sij
In matrix form, W   T    S 
In tensor form, W T  S

Generalization
T, S, Q, …….. R are tensors, Wij  Tij  Sij  Qij  .......  Rij
Wij  Tij  Sij  Qij  ......  Rij
W   T    S   Q  ......   R
W  T  S  Q  ......  R

II.8.2 Multiplication of tensors


T and S are two tensors and v is vector. The product T .S and S .T are the transformations
(tensors).
 
(T . S )v  T ( Sv) : Vector
 
And ( S .T )v  S (T v) : Vector

The components of the tensor T.S are

   
(T . S )ij  ei .(TS )e j  ei .TSmj e m
   
(T . S )ij  ei .(TS )e j  Smj ei .T e m  SmjTim

(T . S )ij  Tim S mj (1)

For the same manner, we can write

( S .T )ij  SimTmj (2)

From (1) and (2), we can conclude (T .S )  T .S and ( S .T )  S .T

TS  T  S

* T, S and Q three tensors

T .S .Q  T .( S .Q )  (T .S ).Q (associative property)

** T 2  T .T
*** T n  T .T .T .......T

II.8.3 Transposed tensor

Transposed tensor of T denoted Tt is any tensor which satisfies the following condition for
 
vectors v1 and v2 .
   
v1 .T v 2  v 2 .T t v1
 
Particularly for e1 and e2 .
   
e j .T ei  ei .T t e j
T ji  Tijt

Properties

- (T t )t  T
- (T . S )t  S tT t
- ( A.B.C......R)t  Rt ......C t .Bt At

II.8.4 Trace of tensor

The trace of a tensor is a scalar which is the sum of its diagonal elements.

T is a tensor of order 2: Trace (T)= T11 +T22 (2D problems)

T is a tensor of order 2: Trace (T)= T11 +T22+ T33 (3D problems)

Properties

- Tr (T  S )  Tr (T )  Tr (S )
- Tr ( T )  Tr (T ) (  is a scalar)
t
- Tr (T )  Tr (T )
     
- Tr ( A.B)  A.B ( A and B are two vectors)
   
- Tr (T )  Tr (Tij ei .e j )  Tij Tr (ei .e j )  Tij  ij  Tii

II.8.5 Identity tensor the tensor inverse

  
The material transformation which transforms any vector v itself v is called a unitary

transformation. we note it by I or 1 .
 
So I (v)  v
     
Particularly, I (e1 )  e1 , I (e 2 )  e 2 , I (e 3 )  e 3

1 0 0 
 I   0 1 0 
0 0 1 

Let T be a tensor, if there exists a tensor S which satisfies: S .T  T .S  I

1
In this case S is the inverse tensor of the tensor T and we write S  T

Properties
1 1
T .T  T .T  I
t 1
(T )1  (T )t
1 1
(T .S )1  S .T
   
If T is reversible and w  T v this deals to v  T 1 w

II.8.6 Orthogonal tensor

The orthogonal tensor is a linear transformation which preserves lengths and orientations of

vectors.
 
v1 and v2 be any vectors.
Let Q be an orthogonal tensor and

 
Qv1 is the transformed of the vector v1
 
Qv 2 is the transformed of the vector v2
More
   
Qv1  v1 , Qv 2  v 2
   
cos(v1 , v 2 )  cos(Qv1 , Qv 2 )
 
The scalar product of vectors v1 and v2
     
v1 .v2  v1 . v2 cos(v1 .v 2 )
     
= Qv1 . Qv 2 cos(Q v1 .Q v 2 ) = Qv1 .Qv 2
Using the transposed property of tensor
   
Qv1 .Qv 2  v 2Q t .Qv1
   
v1 .v 2  v 2 (Qt .Q)v1
The scalar product is commutative, the previous relationship lets say
t
Q .Q  I
t 1
Or Q Q
We can conclude that any tensor is said orthogonal if its inverse is its transpose.
t t
Q .Q  Q .Q  I

II.9 Transformation tensor or of a rotation

 '
Either two distinct bases ei and ei . x3
x3'
'     x2'
e1  Q11 e1  Q21 e 2  Q31 e3 '
' e 3 e 2
'    e3 O 
e 2  Q12 e1  Q22 e2  Q32 e3  x2
e1 e2
'    '
e3  Q13 e1  Q23 e 2  Q33 e3 e1
x1
'  x1'
ei  Qmi e m

   '  '
Q11  e1 .Q e1  e1 .e1 Cosine of the angle between e1 and e1
   '  '
Q12  e1 .Q e 2  e1 .e2 Cosine of the angle between e1 and e 2
 '
Qij  cos(ei , e j )

Then, the transformation or rotation matrix is

 Q11 Q12 Q13 

 Q  Q21 Q22 Q23 

Q31 Q32 Q33 


II.10 Transformation Law of a vector

  
Consider a vectorv that v  vi ei

v
The components of the vector can be obtained by

vi  v.ei
 ' ' '
Corresponding to another reference (e1 , e 2 , e3 ) obtained with a transformation Q.
 '
vi'  v.ei
'   
With e1  Q11 e1  Q21 e 2  Q31 e3
' 
ei  Qmi e m
'
  
Thus, vi  v.Q mi e m  Q mi (v.e m )
 
Then vi'  v.Q mi e m  Q mi vm
In matrix notation

 Q11 Q21 Q31  v1 

  
vi'  Q mi vm  Q12 Q22 Q32  v2 
Q13 Q23 Q33  v 
  3
Or vi'  Qtim vm

The inverse property leads to

t
vi  (Qim )t vm'  Qim vm'

II.11 Transformation Law of a tensor

  
T is a tensor written in base 1 , e 2 , e3 ) by
( e
 
Tij  ei .T e j
' ' '
In other base (e1 , e 2 , e3 )
' '
Tij'  ei .T e j
' 
With ei  Qmi e m , thus
 
Tij'  Qmi em .T Qnj e n
 
Tij'  Qmi Qnj e m .T e n  Qmi Qnj Tmn
Or
Tij'  Qmi Tmn Qnj  Qimt Tmn Qnj

In matrix form: T '   Qt   T Q 

In inverse form,
T '  Q t   Qt   T Q  Q t 

Q  T '  Qt   Q  Qt   T Q  Qt 

Thus,  T   Q  T '  Qt 
In indicial notation
Tij'  Qim Tmj Qnjt
II.12 Symmetric and antisymmetric tensor

A tensor T is symmetric if Tij  T ji or T  T '

A tensor is called antisymmetric if Tij  T ji or T   T '

Each tensor T can be divided to:
s T T '
Symmetric tensor, T 
2
a T T '
And antisymmetricd tensor, T 
2
Example

II.13 Defining tensor by transformation laws

 '
A T is a tensor and ei , ei two different bases. We will define the state of tensors in a
' 
reference e i e
of any tensor quantity already defined in i .

- Zeroth-order tensor (or scalar)  ' 

- First-order tensor (vector) : ai'  Qmi am

- Second-order tensor (or matrix) : Tij'  Qimt QnjTmn

- Third-order tensor : Tijk'  Qimt Qnj Qrk Tmnr

'
- Fourth-order tensor : Tijkl  Qimt Qnj Qrk QslTmnrs
' t
- n-order tensor : Tijkl .....  Qim Qnj Qrk Qsl .......Q Tmnrs ......