Cartesian Tensor with Applications

S.K. Bhattacharyya
1. Summation Convention
3

A.B = A1 B1 + A2 B2 + A3 B3 = Ai Bi (vector dot product)

i =1

A.B = Ai Bi (Summation convention)

A twice repeated subscript (suffix), also called dummy subscript (suffix), implies summation
over the range of this subscript and the symbol is omitted.
On this basis an expression AiBiCi is meaningless.
By this convention
C ( A.B ) D =
=
C j Ai Bi D j

A B = Ai Ai B j B j
2

A B C = Ai AC
i k Bj Bj
A B C
+ E ( A.B ) D
=
2

Ai AC
+ Ek Aj B j Dk
i k B j B j=

A.B =
(C + D) .A Ai Bi =
(Ci + Di ) Ai
Continuity Eqn. of incomressible flow
.v = 0

ui
u1 u2 u3
u v w
+ +
= 0
+
+
= 0
= 0
x y z
x1 x2 x3
xi

Ax =
B Aij x j =
Bi (System of linear equations)
C = AB Cij = Aik Bkj (Matrix multiplication)
C = AB; Tr C = C11 + C22 + C33 Tr C = Cii = Aik Bki (Trace of a matrix)
2. Definition of Tensor

If the direction cosines are

l11 l12 l13

lij =
l21 l22 l23 cos(OX i OX j )
=
l

31 l32 l33

(1)

Then
x1 = l11 x1 + l21 x2 + l31 x3

x1 = l11 x1 + l12 x2 + l13 x3

x2 = l12 x1 + l22 x2 + l32 x3

& x2 = l21 x1 + l22 x2 + l23 x3

x3 = l13 x1 + l23 x2 + l33 x3

x3 = l31 x1 + l32 x2 + l33 x3

(2a)

Using summation convention, Eqn. (2a) is written as

=
x1 li=
x1 l1i xi
1 xi
=
x2 li=
& x2 l2i xi
2 xi

(2b)

=
x3 li=
x3 l3i xi
3 xi
Using summation convention, Eqn. (2b) is compactly written as
=
x j lij=
xi & x j l ji xi

(2c)

The direction cosines are related by

=
l11l11 + l21l21 +
l31l31 1

=
l11l12 + l21l22 +
l31l32 0

l12l12 + l22l22 +=
l32l32 1 & l12l13 + l22l23 +=
l32l33 0
l13l13 + l23l23 +=
l33l33 1

l13l11 + l23l21 +=
l33l31 0

(3a)

Using summation convention, Eqn. (3a) is written as

=
li1li1 1=
li1li 2 0
=
li 2li 2 1=
& li 2li 3 0

(3b)

=
li 3li 3 1=
li 3li1 0
Using summation convention, Eqn. (3b) is compactly written as

lij lik = jk

(3c)

Also,
l ji lki = jk

(3d)

where the Kronecker delta is

=
1 & =
0 (i j )
ii
ij

(4)

If vi & vi are the velocity components in two coordinate systems, then the transformations of
velocities follow Eqn. (2c):
=
v j lij=
vi & v j l ji vi

(5)

First order tensor

An entity represented by 3 numbers (components), referred to the Cartesian system is a first
order tensor, if its components ai & a j relative to OX1X 2 X 3 and OX1X 2 X 3 axes respectively are
related by
a j = lij ai

(6a)

where lij is the direction cosine between OX i and OX j .

In other words, a first order tensor follow the transformation of the coordinates of a point given
by Eqn. (2c).
Eqn. (6a) can also be written as
ai = lij a j

(6b)

Therefore, a tensor of the first order is a vector.

Second order tensor
Let there be two vectors (i.e. tensors of the first order) a and b. The components of a in
OX1X 2 X 3 and OX1X 2 X 3 axes systems are ai and a p and the components of b in OX1X 2 X 3
and OX1X 2 X 3 axes systems are bi and bq . Then, we have, by Eqn. (6)

=
a p lip=
ai & bq l jp b j

(7)

Therefore,

=
a p bq lip=
ai l jq b j lip l jq ai b j (sum of 9 terms)

(8)

An entity represented by a two-suffix set, referred to the Cartesian system is a second order
tensor, if the sets aij & a pq relative to OX1X 2 X 3 and OX1X 2 X 3 axes respectively are related by
a pq = lip l jq aij

(9)

where lij is the direction cosine between OX i and OX j . This relation, in matrix notation, can be
written as
(10)
a = LaLT
Following Eqn. (9), one can define a tensor of any order (m) as an entity if the sets
aijkl .... and a pqrs.... representing the entity relative to OX1X 2 X 3 and OX1X 2 X 3 axes respectively are
related by
a pqrs.... = lip l jq lkr lls .... aijkl ....

(11)

A scalar is a tensor of order zero.

3. Properties of Tensors and Special Tensors
Kronecker delta is a tensor.
0,if i j
1, if i = j

ij =

lip l jq ij =l1 p l1q11 + l2 p l2 q 22 + l3 p l3q 33 (all other terms = 0 since they have ij with i j )
= l1 p l1q + l2 p l2 q + l3 p l3q
0,if p q
= = pq
1, if p = q
Alternating tensor is

1 if (i, j , k )is a cyclic permutation of (1, 2,3)

ijk = 1 if (i, j , k )is a anti-cyclic permutation of (1, 2,3)

0
if any of i, j , k are equal

=
=
=
=
=
1; =
1; rest are zero
123
231
312
132
213
321
It is a tensor of order 3.

l1 p l2 p l3 p
lip l jq lkr ijk l=
l2 q l3q pqr
=
1q
l1r l2 r l3r
A tensor which remains unchanged (i.e. have same components) under transformation is an
isotropic tensor. From the above, it is seen that both Kronecker tensor and alternating tensor are
isotropic tensors. These two tensors are related by

ijm klm =
ik jl il jk OR ijk klm =
il jm im jl
ijk lmn

il im in
= jl jm jn
kl km kn

For a second order tensor Aij, there exists 3 scalar invariants. These are:

1. A11 + A22 + A33

2. A11 A22 + A22 A33 + A33 A11 A12 A21 A23 A32 A31 A13
A11
3. A21
A31

A12
A22
A32

A13
A23
A33

ijk Aij =
< A23 A32 , A31 A13 , A12 A21 > : Vector of tensor A
A few important properties of tensors are:
1. Sum of two tensors of the same order is also a tensor of same order.
2. Product of a tensor of order m and a tensor of order n is a tensor of order m+n.
3. If Aij Bi is a vector, and Bi is any arbitrary vector, the Aij is a tensor of order two.
4. If Aijkl. is a tensor of order m, then the set that results by identifying any two of the suffixes is
a tensor of order m2.
5. A second order tensor can be expressed as a sum of a symmetric and a skew-symmetric tensor.
Apq = lip l jq Aij
= l jp liq Aji (interchange i and j )

This shows that if Aij is a tensor, Aji is also a tensor. Then we can write
1
1
( Aij + Aji ) + ( Aij Aji )
Aij =
2
2
where the first term is a symmetric tensor and second term is a skew-symmetric tensor.
4. Isotropic Tensor
An isotropic tensor is one whose components do not change under transformation. In other
words, if
5

aijkl .... = aijkl ....

then it is an isotropic tensor.

All scalars (rank/order zero tensor) are isotropic.
No non-zero vector (rank/order one tensor) can be isotropic.
The only rank/order two isotropic tensor is ij where is any scalar.
The only rank/order three isotropic tensor is ijk where is any scalar.
The most general rank/order four isotropic tensor is Cijkl = ij kl + ik jl + il jk
For the order four isotropic tensor Cijkl, if it is symmetric in i and j, then
Cijkl = ij kl + ik jl + il jk
Cijkl = ji kl + jk il + jl ik

( ik jl il jk ) ( ik jl il jk ) =
or, 0 =
( )( ik jl il jk )
for all values of i, j, k and l.
Take i = k = 1 and j = l = 2 in above to give
0=
( )(11 22 12 21 ) =

or, =

Thus, the rank/order four isotropic tensor Cijkl, symmetric in (i, j) and (k, l), is
Cijkl = ij kl + ( ik jl + il jk )
5. Applications
A.B = Ai Bi (dot product or inner product or scalar product)
=
A B =
ijk Ai B j (cross product or vector product)
ijk A j Bk
A ( B C=
) ijk Aj ( B C =
) k ijk Aj klm Bl C=
( il jm im jl ) Aj Bl Cm
m
= Aj Bi C j Aj B j Ci = ( A.C ) Bi ( A.B )Ci = ( A.C ) B ( A.B )C (vector triple product)
Continuity equation of incompressible fluid

ui
u v w
+ + = 0 .v= 0
= 0
x y z
xi
6

Continuity equation of compressible fluid

( u ) ( v) ( w)

D
+
+
+
= 0 OR
+ .( v=
+ .v= 0
) 0
t
x
y
z
t
Dt
ui
( ui )
D
+ = 0 OR
+ =
0
t
xi
Dt
xi
Lagrangian derivative
D

= +u + v + w = + uj
Dt t
x
y
z t
x j

Fluid acceleration
u
u
u
Du u
= +u
+v +w
x
y
z
Dt t
Dui ui
u
Dv v
v
v
v
= +u +v + w

= + uj i
Dt t
Dt
x
y
z
t
x j
Dw w
w
w
w
= +u
+v
+w
Dt
t
x
y
z
Laplace equation
2u 2u 2u
0
x 2 + y 2 + z 2 =
2

u =
0

2ui

2 v 2 v 2 v
2v = 0 2v = 0 2 + 2 + 2 = 0
= 0
x y z
x j x j
2 w =

2w 2w 2w
0
2 + 2 + 2 =
y
z
x
Navier-Stokes equations for incompressible fluid

Du
1 p

= X
+ 2u
Dt
x

Dui ui
ui
2ui
Dv
1
1 p
1 p
Dv
2
2
= X p + v
= Y
+ v
=
+ uj
= Xi
+
Dt
Dt
t
x j
x j x j

y
xi
Dt
Dw
1 p
= Z
+ 2 w

y
Dt

Curl
w v u w v u
=
v
i +
j + k
y z z x
x y

u
( v )i = ijk k
x j

Stress in static fluid

0
xx =
yy =
zz =
p; xy =
yz =
zx =

ij = p ij
Stress in moving fluid with viscosity
xx xy xz

=
ij xy yy yz
xz yz zz

Due to fluid motion alone

(Non-isotropic or deviatoric)

p 0 0
0 p 0

0 0 p

Due to static pressure

(Isotropic)

=
ij p ij
ij
Velocity gradient tensor
ui 1 ui u j
1 ui u j
=
+

2 x j xi
x j 2 x j xi

Symmetric
Causes stress
= eij (strain rate tensor)

Skew symmetric
Causes no stress
Rotation withou deformation

Strain rate tensor

1 ui u j
+

2 x j xi
u
v
w
=
; eyy =
; ezz
exx =
x
y
z

=
eij

1 u v
1 v w
1 u w
exy = + ; eyz = +
; exz = +

2 y x
2 z y
2 z x

Stress - strain rate relations

ij = Cijpq e pq
Each stress component is linearly related to all 9 strain rate components.
It implies a total of 81 constants. C is a 4th order tensor.
If C is isotropic, then
Cijpq = ij pq + ip jq + iq jp
If C is symmetric in (i, j ), then = , and
Cijpq = ij pq + ( ip jq + iq jp )

ij pq e pq + ( ip jq + iq jp )e=
e pp ij + 2 eij
=
ij
pq
ij =
p ij + e pp ij + 2 eij (complete sress tensor)

where e pp = .v (volumetric strain rate)

For incompressible fluid, e pp =

.v =
0 ij = p ij + 2 eij
Only 1 material constant ( ) is required for incompressible fluid.
For compressible fluid
1
2
3 p + (2 + 3 )e pp p =
ii + ( + )e pp
ii =
3
3
Define average of the diagonal terms of to be the mean pressure.

1
2
2
p = ii p p = ( + )e pp = ( + ).v
3
3
3
2
+ =
(coefficient of bulk viscosity)
3
2
2
0 (Stokes assumption) 3 +2 =
0; or, =

=
+ =
3
3
2

ij = p ij + e pp ij + 2 eij

2
3

p + .v ij + 2 eij
ij =

Stoke's law of viscosity

2
3

ij =
p + .v ij + 2 eij

1 u u
where eij = i + j
2 x j xi

In component form:

2
u
u

1 & e11 =
exx =
11 =
3
x
x

u 2
= 2 .v + (=
p )
x 3

p + .v + 2
xx =
11 =

u v
1 u v
xy 2=
exy 2 + = +
=
2 y x
y x

1 u v
12 =0 & e12 =exy = +
2 y x

v 2
u 2
w 2
xx 2 .v + ;=
yy 2 .v + ;=
zz 2 .v +
=
x 3

z 3

y 3

u v
v w
u w
+ ; xz =
+
+ ; yz =
z x
y x
z y

xy =

10