Master in Computational and Applied Physics

Continuum and Fluid Mechanics

CHAPTER 1: Tensor Calculus

Albert Falqus
Applied Physics Department
Technical University of Catalonia

OUTLINE
1.
2.
3
3.
4.
5.
6.

Scalars, vectors and tensors. Cartesian basis. Rotation of axes.

Example: stress tensor.
Matrix algebra
algebra. Multiplication and contraction.
contraction
Isotropic tensors.
Algebraic properties of symmetric second order tensors. Eigenvalues and
eigenvectors.
i
Gradient operator, divergence and curl. Gauss and Stokes theorems.
Vector identities.

BIBLIOGRAPHY
P.K.Kundu. Fluid Mechanics. Academic Press. 1990.
R. Aris. Vectors, tensors and the basic equations of fluid
mechanics.
h i
Dover,
D
1962
1962.
J.M. Massaguer, A. Falqus. Mecnica del continuo.
Geometra y Dinmica
Dinmica. Edicions UPC,
UPC 1994.
1994
D.E. Bourne, P.C. Kendall. Anlisis vectorial y tensores
cartesianos. Limusa,
cartesianos
Limusa 1976.
1976
L.A. Segel. Mathematics applied to continuum mechanics.
Dover, 1977.

1. Scalars, vectors and tensors. Cartesian basis.

R t ti of
Rotation
f axes.
x
Scalar: quantity which is defined by its magnitude only.
Example: temperature, density, kinetic energy
Vector: quantity which is defined by its magnitude and direction
direction. Given a
coordinate axis, a vector is defined by three components.
Example: force, velocity, acceleration, momentum, torque
Tensor: linear or multilinear map between vectors (or vectors and scalars).
Given a coordinate axes system, a tensor is defined by, at least, nine
components.
p
Example: stress tensor (force per area unit across any section of a body at a
point), strain tensor (deformation rate of a body in any direction at a point)

Cartesian axes

x3

cartesian coordinates:

e3
e1

ortonormal basis:

e2

x2

e3

e1 '

i 1

Rotation of axis

e3 '

e1 ' C11e1 C21e2 C31e3

e2 ' C12 e1 C22 e2 C32 e3

e2
e1

e1 (1,0,0) , e2 (0,1,0) , e3 (0,0,1)

3
x xi ei

position vector:

x1

x1 , x2 , x3

e2 '

e3 ' C13e1 C23e2 C33e3

3

ei ' C ji e j , i 1,2,3
j 1

i = free
ee index;
de ; j = du
dummy
y index
de

Summation convention over repeated indexes

In any product of terms a repeated index is held to be summed over 1,2,3.
An index not repeated in any product can take any of the values 1,2,3.
Examples:

ei ' C ji e j

ei ' C ji e j , i 1,2,3

ai bijk ckn Rkjnk

j 1
3

a b

i , k 1

c Rkjnk

ijk kn

j , n 1,2,3

k 1

Warning: an index should not be repeated more than twice.

Example:

a111 a 222 a333 0

but not as

aiii 0

should
h ld be
b abbreviated
bb i d as

a
i 1

that would mean:

iii

a111 a222 a333 0

Warning: same free indexes at both sides of an equation or for all the terms in a sum.

Example:

aij bi c jk a kn
k

???

Rotation matrix. Properties

C11 C12

C C21 C22
C
31 C32

x2 '

x2

C13

C23
C33

x1 '

C1i , C2i , C3i

C11 cos ,

x1

The transpose matrix:

C11 C21 C31

T
C C12 C22 C32
C

13 C23 C33

components of ei on the basis e1, e2, e3

= cosinus of the angles between the new
axis xi and the old axes x1, x2, x3

C21 cos / 2

C12 cos / 2 ,

C22 cos

verifies: CCT = 1 (and also CTC = 1)

C11 C12

C21 C22
C
31 C32

C13 C11 C21 C31 1 0 0

C23 C12 C22 C32 0 1 0

C33 C13 C23 C33 0 0 1

This is the orthogonality condition and implies that:

C 1 CT

Change of the coordinates of a point.

ei ' C ji e j

Given that the basis vectors change according to:

How do the coordinates of any point change?

x x j e j xk ' ek ' xk ' C jk e j

x j xk ' C jk
and since

C 1 CT

x j ' xk C 1

jk

the coordinates do change with the same matrix C

x j ' C kj xk
Orthogonal transformations .. det C = 1 ..

or x' CT x

Formal definition of a vector and a scalar.

A physical quantity b is said to be a scalar if it is invariant under axis rotations, i.e.,

bb'' b

Three physical quantities (a1, a2, a3) are said to define a vector
if they change as the coordinates of a point under axis rotations, i.e.,

ai ' Cki ak
Tensors arise as linear and multilinear maps between vectors (or vectors and scalars).
Given a coordinate system, a tensor is defined by 3n components where n is its order.
A scalar can be considered a tensor of order 0 and a vector a tensor of order 1.
A formal definition based on the transformation of their components will be given
later on.
We will first introduce a particular tensor as an example: the stress tensor

2. Stress tensor
Force per unit area across a section of a body
(force of part B on part A: then the normal is outwards from A):

n
S

F
f lim
S 0 S

is a vector that depends

p
on the orientation of the surface,, i.e.,, another vector,,

How is this relation ?

2on Newton Law applied to the tetrahedron when S 0 :

x3
S2

S1

n
x2

x1

S3

S f S1 f1 S 2 f 2 S 3 f 3 S h a
3
S S S 1

f 1 f1 2 f 2 3 f 3 h a 0
S
S
S
3
f i eexternal
te a force
o ce on
o Si

so corresponding to a normal

n ei


S1 S 2 S3
f f1
f2
f3
S
S
S

S1
S2

Si are the projections of S

n
x2

x1

Si
cos i ni
S

f i external force on S i
so corresponding to a normal

fi i

S3

f n1 1 n2 2 n3 3

f n

n ei

force per unit area corresponding to

the normals n ei

therefore,

f is a linear function of n

stress tensor

Matrix expression

f n1 1 n2 2 n3 3

i1 , i 2 , i 3

Symmetry of the stress tensor !!

components of

11 12 13

f1 , f 2 , f 3 n1 , n2 , n3 21 22 23

31 32 33
33
32
23

31
x3

x1

13
x2

11

12

ij = ji

21

11 > 0 : pulling ; 11 < 0 : pushing (compression)

stress tensor matrix

ij

components of the stress tensor

32 components second order tensor

meaning of the components

22

11 , 22 , 33

normal
stresses
12 , 13 , 21 , 23 , ... shear
stresses

How do the components of the stress tensor change under axis rotation?

f i n j ji

f i ' C ji f j , ni ' C ji n j

f i Cij f j ' , ni Cij n j '

Cij f j ' C jk nk ' ji

f n ' C 1

nii

n
'

C
jk k
ji

nii

jk

nk ' ji Cini C jk nk ' ji

f n ' nk ' C jk ji Cin nk ' kn '

kn ' C jk Cin ji

or

' CT C

Formal definition of a tensor.

Nine physical quantities Tij are said to define a second order tensor
if they change as the stress tensor under axis rotations, i.e.,

Tkn ' C jk CinT ji

3n physical quantities Ti1 ...in are said to define a n-order tensor

if they change under axis rotations as,
as

Ti1 ...in ' C j1i1 C j2i2 ...C jnin T j1 ... jn

3. Matrix algebra. Multiplication and contraction.

Matrix algebra can be expressed in several ways:
a) Matricial
b) Components
c) Intrinsic or symbolic

Example:

C11 C12

a) C21 C22
C
31 C32

C13 C11 C21 C31 1 0 0

C23 C12 C22 C32 0 1 0

C33 C13 C23 C33 0 0 1

b)

Cik C jk ij

c)

C CT 1

entries of the identity

y matrix:

ij

1 if i=j
0 if ij

Tensor multiplication of vectors and tensors.

Tensor product of two vectors

Any pair of vectors, a , b , define a linear map a b


a b

v
(a b ) v (b v ) a

The components of tensor a b are simply ai b j

between vectors:

Tensor product of vectors and tensors

In general
general, by multiplying the components of a n-order tensor and those of an
morder tensor, an (n+m)-order tensor is defined through its components.

Examples:
From ai and Tij the third order tensor: Qijk=ai Tjk may be defined

From two 2nd order tensors, Aij , Bij, the 4th order tensor Pijkl=AijBkl may be defined

Quotient rule.
Let

Ai1 ...in , X j1 ... jm be 3n+3m physical quantities and let Bk1 ...k nm be their product

Bi1 ...in j1 ... jm Ai1 ...in X j1 ... jm

Assume that A and B are tensors
Then,

X j1 ... jm are the components of an m-order tensor

Namely;

Ai1 ...in X j1 ... jm Bi1 ...in j1 ... jm

tensor

tensor

X = tensor

Contraction.

Given a n-order tensor, contraction is a procedure to obtain a lower order tensor.

Two indices are equated and a summation is performed over these repeated indices.
Examples:
From the components of a 2nd order tensor,

Tii T11 T22 T33

Tij , the only possible contraction is

which is a scalar
From a third order tensor, Tijk , three different contractions are possible but all of them
give a vector:

ai Tijj

, b j Tiji , ck Tiik
From two 2nd order tensors, Aij , Bij , the four order tensor Qijkl Aij Bkl

p y g Four 2nd order tensors may

y be then obtained
mayy be defined byy multiplying.
by contracting:

Aij B jl A B il

Aij Bkj A B T

ik

Aij Bik A T B

jk

Aij Bki B A kj

The components of all of them can be computed from standard matrix product
A second contraction may be applied to these 2nd order tensors and a scalar is obtained
in two possible ways Aij B ji or Aij Bij . They are indicated by AB or by ABT

4. Isotropic tensors.
An isotropic tensor is one whose components are invariant under axes rotations. i.e.

T 'i1i2 ...in Ti1i2 ...in

i1i2 ...in 1,2,3

Isotropic tensor are associated to geometric invariance under rotation.

0-order isotropic tensors: all 0-order tensors are isotropic as they are scalars
1st-order isotropic
p tensors: there are none
2nd-order isotropic tensors: only one, the identity tensor, whose components are given
by the Kronecker delta in any basis:

ij

1 if i=j
0 if ij

1 0 0

0 1 0
0 0 1

Very common use of ij is that in any expression where it appears with index i being
contracted,
co
c ed, it can
c be dropped
d opped out
ou by substituting
subs u g i by j
j in thee expression.
e p ess o .
Examples:

ij Anim Anjm , ij kn Bij Ckl Bii Cnl

3rd-order isotropic tensors: only one, the alternating tensor, whose components
are
1 if i,j,k= even permutation of 123 (i.e., 123 or 231 or 312)

ijk

0 if two or three indices are equal

-1 if i,j,k= odd permutation of 123 (i.e., 132 or 213 or 321)

Properties:

ijk jki kij

even permutations

ijk ikj , ijk kji , ijk jik

odd permutations

ijkj imn jjm kn jjn km

The cross product of two vectors, reads

Given any matrix A, det A = ijk A1iA2jA3k

a b ijk a j bk ei

4th-order isotropic
p tensors: there are three and their linear combinations. So the
most general is:

ij pq ( ip jq iq jp ) ( ip jq iq jp )
where , , are arbitrary numbers.

5. Symmetric and antisymmetric second order tensors.

Ei
Eigenvalues
l
and
d eigenvectors
i
t
A 2nd order tensor B is called
symmetric if Bij = Bji
antisymmetric if Bij = -Bji
P
Properties:
ti
Any 2nd order tensor can be represented as the sum of a symmetric part and
an antisymmetric
i
i one:
1
1

Bij

ij

B ji

symmetric
t i

ij

B ji

antisymmetric
ti
t i

If Aij is antisymmetric and Bij is symmetric, then AijBij = 0

A symmetric tensor has only 6 independent components

An
A antisymmetric
ti
t i tensor
t
has
h zero diagonal
di
l components
t andd has
h only
l 3
independent components. These 3 components are associated with a vector

Properties
p
((continued):
)
Every antisymmetric tensor can be associated with a vector and vice versa

1
2

k kij Rij
Rij ijk k

R 3

3
0
1

1
0

Eigenvectors
g
and eigenvalues
g
Eigenproblem:
Given a

2nd

order tensor,
tensor A, are there vectors

Av v ?

i
t ,
v eigenvector

and numbers , such that

= eigenvalue
i
l

are the solutions of the third order equation (characteristic equation)

dett A 0
d

for each , v is a solution of

A v 0

the eigenvalues and the direction of the eigenvectors are invariant under
axes rotations
the characteristic equation reads:
3 I12 I 2 I 3 0
where
I A
1

ii

1
2
Aij Aij Aii 12 2 3 31
2
I 3 det
d A 12 3

I2

are invariant under axes rotations

If A is symmetric:
there are three eigenvalues
g
that are real,, 1, 2, 3 ((not necessarilyy distinct))
associated to them, there are three eigenvectors which are mutually orthogonal
if the
th coordinate
di t system
t is
i rotated
t t d as to
t coincide
i id with
ith the
th eigenvectors,
i
t
matrix A takes a diagonal form:
0 0

1
A 0 2

0 0

0
3

extremal property: the components Aij change with the coordinate axes,
axes but the
diagonal elements cannot be larger than the largest and smaller than the smallest

6. Gradient operator, divergence and curl.

G
Gauss
and
d Stokes
St k th
theorems.
m Vector
V t identities
id titi
Given a scalar field, (x1,x2,x3), its gradient is defined by

ei
xi

is a vector
gives the direction of maximum increase of
|| = magnitude of the derivative of along this direction
is perpendicular to the surfaces of (x1,x2,x3) = const.

The derivative of along a direction associated to n is

n

n
n

Since is a vector and is a scalar, the operator itself can be considered

a vector (quotient rule):

bl operator
t

ei i ei nabla

xi

( will
(we
ill omit
i the
h arrow for
f simplicity,
i li i )

The gradient of a vector field is a second order tensor:

v ij i v j
such that multiplied by a unit vector
along the direction of n :

n gives the derivative of such a vector

dv
n v ni i v j e j
dn

Similarly,
Si
il l the
th gradient
di t off a tensor
t
field
fi ld off order
d tensor
t
n is
i a n+1
+1 order
d tensor
t
with similar meaning.

The divergence of a vector field is defined as the contraction of its gradient:

v
v i vi i
xi

(or dot product of and v )

The divergence of a vector field at a point is associated to its flux going in or out
a small surface around this point

v ( x1 , x2 , x3 )

1
v lim
V 0 V

v
n dS

The divergence of a tensor field of n-order may be defined trough a contraction of

its gradient and it is a (n-1)-order tensor.
There are however several options depending on which contraction is performed.
For a 2nd order tensor:

Tijk iT jk

iTik T k

iT ji T j


v
The curl of a vector field is defined as the cross product of vector and

v ijk j vk ei

The curl of a vector field is associated with the rotation of the vector field.
Example: velocity field of a rigid body:

v x v 2
So, the curl of the velocity field is proportional to the angular velocity of the body

Gauss Theorem (divergence theorem)

dA n dA

u dV u n dA
V

dA

dV

ui
dV

xi
V

u n

i i

dA

V = volume
V = surface, boundary of V

Al iin 2D:
Also
2D
x2

dl

S
x1

dl n dl

dS

S
V S surface
S = curve, boundary of S

u1 u2
S x1 x2 dS S u1n1 u2 n2 dl
(in general in n-dimensions, n>1)

Stokes Theorem

dA n dA

u n dA u dl

inside

dl

A = surface
f
A = curve, boundary of A

n
outside
inside

sign: choose one of both sides of the surface and define it

as the outside, the normal vector pointing from inside to outside.
Then the positive direction of dl is the anticlockwise one
Then,
looking from the outside

Vector identities

a,b

vector fields

f , g scalar fields

a 0

a a 2 a

f 0

2 fg f 2 g g 2 f 2f g

fa f a a f

a b a b b a a b b a

a b b a a b



a b b a a b b a a b

1
a a a a a a
2
Laplacian operator (scalar):

2 ii