You are on page 1of 160

2.

Kinematics

2.1 Motion, deformation and deformation velocities

A material body B consists of an infinite number of material points X. The set of


boundary points is the surface ∂B of the body.

For the description of the location of the material body points in three-dimensional
real Euclidean space E3 the specification of a space-fixed reference point, the origin
0, is required. In relation to this origin, the placement of a material point in space
is given by the vector-valued function

x = χ ( X , t = const. ) . (2.1)

The motion of a body is defined as a time-dependent sequence of the placements of


the body points of B:

x = χ(X, t) . (2.2)

This form is called the material description of motion and asks that the image is to
be one-one, continuous and continuously differentiable.

Identifying the material body point X by the position vector of the reference con-
figuration

X = χ ( X , t = t0 ) , (2.3)

the material description of motion can be transferred into the Lagrangian form of
motion:

x = χ(X, t) . (2.4)

Using the Lagrangian description of motion, see (2.4), the velocity and the acceler-
ation of material points X are defined as

2
d ∂χ ( X , t )
ẋ = [χ(X, t)] = ,
dt ∂t
(2.5)
d2 ∂ 2χ ( X , t )
ẍ = 2 [ χ ( X , t ) ] = .
dt ∂t2

The inversion of the motion function (2.4) yields the Eulerian description of motion:

X = χ−1 ( x , t ) . (2.6)

With help of (2.6) the following alternative representations of the velocity and ac-
celeration field are obtained:

ẋ = ẋ ( x , t ) = ẋ [ χ−1 ( x , t ) , t ] ,
(2.7)
−1
ẍ = ẍ ( x , t ) = ẍ [ χ (x, t), t] .

The Euler form of motion (2.6) implies the existence of non-singular Jacobian de-
terminant

J = det F , (2.8)

i.e., J 6= 0. The quantity

∂χ ( X , t ) ∂x
F= = Grad χ = = Grad x (2.9)
∂X ∂X

denotes the deformation gradient. The inverse deformation gradient F−1 is given by

∂χ−1 ( x , t ) ∂X
F −1
= = = grad X . (2.10)
∂x ∂x

Considering (2.5)1 und (2.7)1 the material and the spatial velocity gradient read

∂ ẋ ∂ ẋ
Ḟ = = Grad ẋ , L = = grad ẋ = ( Grad ẋ ) · F−1 . (2.11)
∂X ∂x

The spatial velocity gradient can be additively split into a symmetric and a skew-
symmetric component,
3
L= D+W . (2.12)

The quantities

1 1
D= ( L + LT ) , W = ( L − LT ) (2.13)
2 2

are the symmetric and a skew-symmetric part of the spatial velocity gradient L.

4
reference actual
configuration configuration
(t = t 0 ) (t = t1 )

c = c ( X, t )
B0

X ¶B
X
B
¶B0

X
x = c ( X, t ) t = t1

0
Figure 2.1: Illustration of the function of motion of a materiel point

5
2.2 Transport theorems

With help of material deformation gradient F, differential line, surface and volume
elements – dX, da and dV – of the reference configuration can be mapped in the
corresponding elements of the current configuration at time t – dx, da and dv:

dx = F · dX ,
da = dx1 × dx2 = F · dX1 × F · dX2
= (det F) FT−1 · ( dX1 × dX2 ) = (det F) FT−1 · dA
(2.14)
= F+ · dA = cof F · dA ,
dv = dx1 · ( dx2 × dx3 ) = F · dX1 · ( F · dX2 × F · dX3 )
= det F [ dX1 · ( dX2 × dX3 ) ] = J dV

These relationships are called transport theorems. The second-order tensor F+ =


cof F = (det F) FT−1 is the adjoint tensor of F and the cofactor of F, respectively.
With

(det F)· = det F ( D : I ) ,


(2.15)
(F+)· = [ (det F) FT−1 ]· = [ ( D : I ) I − LT ] · F+

the material time derivations of differential line, surface and volume elements can
be calculated:

dẋ = Ḟ · dX = L · F · dX = L · dx ,
dȧ = [ ( D : I ) I − LT ] · F+ · dA = [ ( D · I ) I − LT ] · da , (2.16)
dv̇ = det F ( D : I ) dV = ( D : I ) dv = div ẋ dv .

6
line elements
F
dX dx

surface elements
( det F ) F T -1
dA da
dx1
dX 2 dX1
dx 2

volume elements
dV det F dv

dX3 dx3

dX 2 dX1 dx1
dx 2

Figure 2.2: Illustration of the transport theorems of differential line, surface and
volume elements

7
2.3 Basis vectors und basis velocity vectors

For a description of the body B0 and B by convective curvilinear coordinates Θi mit


i = 1, 2, 3, the position vectors of a material point X depending on this metric sizes
read in the reference and present configuration:

X = X ( Θi , t0 ) , x = x ( Θi , t ) . (2.17)

It is postulated, that the unique inversions

Θi = Θi ( X , t0 ) , Θi = Θi ( x , t ) (2.18)

of these functions exist.

The derivatives of the position vectors for the coordinates Θi provides a natural unit
in the material point X the tangent vectors
∂X( Θi , t = t0 ) ∂x( Θi , t )
Gi = , gi = (2.19)
∂Θi ∂Θi
along the coordinate lines of the reference and present configuration. The corre-
sponding dual basis systems (contravariant basis vectors) result from the calculation
rule

Gi · Gk = δik , gi · gk = δik (2.20)

or they can be determined from the derivative of the inverse representations (2.18),
respectively:
∂Θi( X , t = t0 ) i ∂Θi ( x , t )
G =i
, g = . (2.21)
∂X ∂x
With the basis vectors (2.19) and (2.21) and applying the chain rule and differenti-
ation it is obtained for the deformation gradient F and for the velocity gradients Ḟ
and L the following representation forms:
∂x ∂x ∂Θi
F= = i
⊗ = gi ⊗ Gi ,
∂X ∂Θ ∂X
∂ ẋ ∂ ẋ ∂Θi
Ḟ = = ⊗ = ġi ⊗ Gi , (2.22)
∂X ∂Θ i ∂X
∂ ẋ ∂ ẋ ∂Θi
L= = i
⊗ = ġi ⊗ gi .
∂x ∂Θ ∂x
8
The inverse gradients read

∂X ∂X ∂Θi
F−1 = = i
⊗ = Gi ⊗ g i ,
∂x ∂Θ ∂x
∂X ∂X ∂Θi
Ḟ−1 = = ⊗ = Gi ⊗ ġi , (2.23)
∂ ẋ ∂Θ i ∂ ẋ
∂x ∂x ∂Θi
L−1 = = i
⊗ = gi ⊗ ġi .
∂ ẋ ∂Θ ∂ ẋ

With (2.22)1 and (2.23) 1 , the links

gi = F · Gi , gi = FT−1 · Gi (2.24)

can derived from the basis vectors of the two configurations. The material time
derivatives of the co- and contravariant basis vectors of the present configuration
are given by

ġi = Ḟ · Gi = L · gi , ġi = ḞT−1 · Gi = LT−1 · gi = − LT · gi . (2.25)

The used identity

L = − L−1 (2.26)

in (2.25) arises from the condition (F F−1)· = 0 in consideration of (2.11)1:

(F · F−1)· = Ḟ · F−1 + F · Ḟ−1


= L · F · F−1 + F · ( L · F )−1
= L · F · F−1 + F · F−1 · L−1 = L + L−1 = 0

=⇒

L = − L−1 .

9
reference actual
configuration configuration
(t = t 0 ) (t = t1 )

Q 2 Q2
c = c ( X, t ) Q1
B0
Q1 g2

G2 X g1
X ¶B
G1
B
¶B0

X
x = c ( X, t ) t = t1

0
Figure 2.3: Illustration of the curvilinear coordinates and the covariant base vectors

10
2.4 Strain tensors and strain velocities

The introduction of strain tensors is done by using the difference of the squares of
the line elements of the actual and reference configuration:

ds2 − dS2 = dx · dx − dX · dX . (2.27)

Using the transport theorems (2.14)1 for differential line elements combined with the
rules of tensor calculus, the relation (2.27) can be represented terms of the reference
and present configuration:

ds2 − dS2 = dx · dx − dX · dX
= ( F · dX ) · ( F · dX ) − dX · dX
= dX · FT · F · dX − dX · I · dX ,
ds2 − dS2 = dx · dx − dX · dX
= dx · dx − ( F−1 · dx ) · ( F−1 · dx )
= dx · I · dx − dx · FT−1 · F−1 · dx

=⇒

1
ds2 − dS2 = 2 dX · ( FT · F − I ) · dX = 2 dX · E · dX ,
2
(2.28)
1
ds2 − dS2 = 2 dx · ( I − FT−1 · F−1 ) · dx = 2 dx · A · dx .
2
In (2.28)

1 1
E= ( FT · F − I ) = (C− I) ,
2 2
(2.29)
1 1
A = ( I − FT−1 · F−1 ) = ( I − B−1 )
2 2
denote the Cauchy-Green and the Almansi strain tensor. The symmetric quantities

C = CT = FT · F , B = BT = F · FT (2.30)

11
are the right and the left Cauchy-Green deformation tensor. With help of the
deformation gradient the strain tensors can be transferred into each other:

E = FT · A · F , A = FT−1 · E · F−1 . (2.31)

With help of the polar decomposition

F=R·U=V·R (2.32)

of the deformation gradient, where R is a proper orthogonal tensor with R−1 = RT ,


and the right and left stretch tensor of deformation, one can show the relation of
the right and left Cauchy Green deformation tensor

C = FT · F = ( R · U )T · R · U = UT · RT · R · U = UT · U
=U·U ,
B = F · FT = V · R · ( V · R)T = V · R · RT · VT = V · VT
=V·V

=⇒

C = FT · F = UT · U = U · U ,
(2.33)
T T
B=F·F =V·V =V·V

which yield

C = RT · B · R , B = R · C · RT (2.34)

using the rotation tensor R.

Transferring the existing relation (2.34) between the Cauchy Green deformation
tensors C and B in the reference and actual configuration, respectively, to the strain
tensors E and A, further strain tensors can be defined. One obtains the Lagrangian
Karni-Reiner strain tensor K by a forward rotation of E to the actual configuration
and the eulerian Karni-Reiner strain tensor KR by a backward rotation of A into
the reference configuration.

12
1
K = R · E · RT = (B− I) ,
2
(2.35)
1
KR = RT · A · R = ( I − C−1 ) .
2
Between the strain tensors KR and K, the following connection can be established

KR = F−1 · K · FT−1 , K = F · KR · FT (2.36)

via the deformation gradient.

The here discussed strain tensors can be expressed in a basis with the help of (2.22)1
and (2.23)1, see Table 1–3. The Cauchy-Green- and Almansi strain tensor E and
A show contravariant base systems, while the Karni-Reiner strain tensors exhibit
covariant base systems. The transformation relations (2.31) resp. (2.36) between
the tensors E and A resp. KR and K are then denoted by

FT−1 · ( . . . ) · F−1 : contravariant push-forward transformation


of ( . . . ),
FT ·(...)· F : contravariant pull-backward transformation
of ( . . . ),
F · ( . . . ) · FT : covariant push-forward transformation of ( . . . ),
F−1 · ( . . . ) · FT−1 : covariant pull-backward transformation ( . . . ).

The strain tensors E and KR related to the reference configuration are denoted as
material quantities while the strain tensors A and K related to the actual con-
figuration are denoted as spatial quantites. The push-forward transformation and
pull-backward transformation, compare Marsden & Hughes [1983], are the con-
necting link between the material and spatial quantities.

The material time derivatives of the Green-Lagrange strain tensor E and the eule-
rian Karni-Reiner strain tensor KR define the strain velocities with respect to the
reference configuration as

1 1
Ė = [ ( C − I ) ]· = Ċ
2 2

13

1 1
Ė = ( FT · F)· = ( ḞT · F + FT · Ḟ )
2 2

1 1
Ė = [ (L · F)T · F + FT · L · F ] = ( FT · LT · F + FT · L · F )
2 2

1 T
Ė = F · ( LT + L ) · F
2

1
Ė = FT · [ ( LT + L ) ] · F = FT · D · F ,
2

1 1
K̇R = [ ( I − C−1 ) ]· = − Ċ−1
2 2

1 1
K̇R = − (F−1 · FT−1)· = − ( Ḟ−1 · FT−1 + F−1 · FT−1 )
2 2

1
K̇R = − [ (L · F)−1 · FT−1 + F−1 · ( L · F )T−1 ]
2

1
K̇R = − ( F−1 · L−1 · FT−1 + F−1 · LT−1 · FT−1 )
2

1
K̇R = − F−1 · ( L−1 + LT−1 ) · FT−1
2
14

1
K̇R = − F−1 · [ ( L−1 + LT−1 ) ] · FT−1
2

1
K̇R = − F−1 · [ ( − LT − L ) ] · FT−1
2

1
K̇R = F−1 · [ ( LT + L ) ] · FT−1 = F−1 · D · FT−1
2

=⇒

Ė = FT · D · F , K̇R = F−1 · D · FT−1 . (2.37)

The spatial strain rates are defined as Lie-derivatives of the Almansi strain tensor
and the Lagrangian Karni-Reiner strain tensor. The Lie-derivative of a tensor cor-
responds to the material time derivative with a fixed base system. The product rule
of differentiation yields the following expressions

Ȧ = (FT−1 · E · F−1)·

Ȧ = FT−1 · Ė · F−1 + ḞT−1 · E · F−1 + FT−1 · E · Ḟ−1

Ȧ = FT−1 · Ė · F−1 + ( L · F )T−1 · E · F−1 + FT−1 · E · ( L · F )−1

Ȧ = FT−1 · Ė · F−1 + LT−1 · FT−1 · E · F−1 + FT−1 · E · F−1 · L−1

15

Ȧ = FT−1 · Ė · F−1 − LT · ( FT−1 · E · F−1 ) − ( FT−1 · E · F−1 ) · L

Ȧ = FT−1 · Ė · F−1 − LT · A − A · L ,

K̇ = (F · KR · FT )· = F · K̇R · FT + Ḟ · KR · FT + F · KR · ḞT

K̇ = F · K̇R · FT + L · F · KR · FT + F · KR · ( L · F )T

K̇ = F · K̇R · FT + L · F · KR · FT + F · KR · FT · LT

K̇ = F · K̇R · FT + L · ( F · KR · FT ) + ( F · KR · FT ) · LT

K̇ = F · K̇R · FT + L · K + K · LT

=⇒

Ȧ = FT−1 · Ė · F−1 − LT · A − A · L ,
(2.38)
K̇ = F · K̇R · FT + L · K + K · LT .

Note here that the time change of the natural bases in the current configuration
is described with the help of the spatial velocity gradient L, see (2.25), thus the
Lie-derivatives of the strain tensors A and K result in
16
A△ = FT−1 · Ė · F−1 = Ȧ + LT · A + A · L ,
(2.39)
K▽ = F · K̇R · FT = K̇ − L · K − K · LT .

Here, one distinguishes between the upper and lower Lie-derivative, denoted by the
symbols ( . . . )△ und ( . . . )▽ . Using A△ bzw. K▽ , a co- resp. contravariant push-
forward transformation of the material strain velocities Ė bzw. K̇R to the actual
configuration is described.

By the use of the relation (2.37) in (2.39), one can show that the spatial strain
velocities

A△ = D , K▽ = D (2.40)

are identical. This result is not valid for the material strain velocities, since generally
it is

Ė 6= K̇R . (2.41)

The description of the strain velocities with respect to a base system are given in
Table 1 and 3. The push-forward and pull-backward rotation of the strain tensors
according to (2.35) are not applicable to the here discussed strain velocities.

17
reference configuration
1 1
E= ( C − I ) = ( gik − Gik ) Gi ⊗ Gk
2 2

⇓ FT−1 · ( . . . ) · F−1 ⇓

actual configuration
1 1
A= ( I − B−1 ) = ( gik − Gik ) gi ⊗ gk
2 2

reference configuration
1 1
Ė = Ċ = ġik Gi ⊗ Gk
2 2

⇓ FT−1 · ( . . . ) · F−1 ⇓

actual configuration
1 1
D = A△ = ( L + LT ) = ġik gi ⊗ gk
2 2

( . . . ) △ = ( . . . ) · + LT · ( . . . ) + ( . . . ) · L

Table 1:
Cauchy-Green and Almansi strain tensor and the corresponding strain velocities

18
reference configuration
1 1
E= ( C − I ) = ( gik − Gik ) Gi ⊗ Gk
2 2

⇓ R · ( . . . ) · RT ⇓

actual configuration
1 1
K= ( B − I ) = ( Gik − gik ) gi ⊗ gk
2 2

actual configuration
1 1
A= ( I − B−1 ) = ( gik − Gik ) gi ⊗ gk
2 2

⇓ RT · ( . . . ) · R ⇓

reference configuration
1 1
KR = ( I − C−1 ) = ( Gik − gik ) Gi ⊗ Gk
2 2

Table 2:
Relations between the Cauchy-Green and Almansi strain tensors and the Karni-
Reiner strain tensors

19
reference configuration
1 1
KR = ( I − C−1 ) = ( Gik − gik ) Gi ⊗ Gk
2 2

⇓ F · ( . . . ) · FT ⇓

actual configuration
1 1
K= ( B − I ) = ( Gik − gik ) gi ⊗ gk
2 2

reference configuration
1 1
K̇R = − Ċ−1 = − ġik Gi ⊗ Gk
2 2

⇓ F · ( . . . ) · FT ⇓

actual configuration
1 1
D = K▽ = ( L + LT ) = − ġik gi ⊗ gk
2 2

( . . . ) ▽ = ( . . . ) · − L · ( . . . ) − ( . . . ) · LT

Table 3:
Karni-Reiner strain tensors and corresponding strain velocities

20
3. Balance equations and entropy inequality

The balance equations as well as the entropy inequality are material independent
realtions:
• Balance of mass – mass conservation, continuity equation;
• Balance of momentum – conservation of momentum,
principle of linear momentum,
Newton‘s equation of motion;
• Balance of moment of momentum – conservation of angular momentum;
• Balance of energy – first law of thermodynamics;
• Entropy inequality – second law of thermodynamics,
Clausius-Duhem inequality.
The structure of the balance laws of mass, momentum, angular momentum and
energy for a single body is given by

( B )· = ( E ) ,

where the symbol ( . . . )· denotes the material time derivative of the corresponding
quantity following the motion of the single body.

21
structure of the ( B )· = (E )
balance equations
for a single body

B E

balance of mass mass 0


(M)

balance of momentum external forces


momentum
(l) (k)

balance of angular external moments


angular momentum momentum
( h(0) ) ( m(0) )

balance of energy internal and increments of the


kinetic energy mechanical and
non-mechanical
work
(E + K) (W + Q)

Table 1.1: Structure of the balance laws for a single body

22
3.1 Balance of mass

The axiom of the balance of mass (conservation of mass) reads

Ṁ = 0 , (3.1)

where
Z
M= ρ dv (3.2)
B

denotes the mass at time t.

With the material time derivative


Z Z Z
Ṁ = ( ρ dv )· = ( ρ J dV )· = ( ρ J )· dV
B B0 B0
Z
= ( ρ̇ J + ρ J̇ ) dV
B0
Z
= ( ρ̇ J + ρ J div ẋ ) dV
B0
Z
= ( ρ̇ + ρ div ẋ ) J dV
B0
Z
= ( ρ̇ + ρ div ẋ ) dv
B

=⇒
Z Z
Ṁ = ( ρ dv )· = ( ρ̇ + ρ div ẋ ) dv (3.3)
B B

one obtains the following form of the balance of mass:


Z
( ρ̇ + ρ div ẋ ) dv = 0 . (3.4)
B

23
Considering

∂ρ
ρ̇ ( x , t ) = + grad ρ · ẋ (3.5)
∂t

the alternative representation form


Z
∂ρ
( + grad ρ · ẋ + ρ div ẋ ) dv = 0
∂t | {z }
B = div( ρ ẋ )

=⇒
Z
∂ρ
[ + div( ρ ẋ ) ] dv = 0 (3.6)
∂t
B

can be derived. With the divergence theorem


Z Z
div( ρ ẋ ) dv = ρ ẋ · n da (3.7)
B ∂B

Equation (3.6) can be transferred to the global statement


Z Z
∂ρ
dv = − ρ ẋ · n da . (3.8)
∂t
B ∂B

The relations (3.4) and (3.6) are denoted as the Lagrange and Euler form of the
balance of mass.

Provided that the integrand is continuous, from (3.4) and (3.6), respectively, directly
one concludes to the local statement of the balance of mass:

ρ̇ + ρ div ẋ = 0 (3.9)

and

∂ρ
+ div( ρ ẋ ) = 0 , (3.10)
∂t

24
respectively. With help of

1
div ẋ = J̇ (3.11)
J
Equation (3.9) can be reformulated:

1 1
ρ̇ = − J̇ . (3.12)
ρ J

The time integration of the aformentioned relation yields the integral form of local
statement of the Lagrange form of the balance of mass:

Zt Zt
1 1
ρ̇ dt = − J̇ dt
ρ J
t0 t0

ln ρ |tt0 = − ln J |tt0

ln ρ (t) − ln ρ (t0 ) = − ln J (t) + ln J (t0 )

ln ρ − ln ρ0 = − ln J + 0

ρ 1
ln = ln
ρ0 J

=⇒
ρ0
J = det F = . (3.13)
ρ

The quantity ρ0 = const. denotes the density of the undeformed reference placement
at time t = t0, i.e., F = I and det F = J = 1.

25
Summary: Balance of mass

Z
Axiom ( ρ dv )· = 0
B

globale statements

reference Z
placement M = ρ0 dV = konst.
B0

actual Z
placement ( ρ̇ + ρ div ẋ ) dv = 0
B
Z Z
∂ρ
dv = − ρ ẋ · n da
∂t
B ∂B

local statements

reference
placement ρ0 = konst.

actual
placement ρ̇ + ρ div ẋ = 0

∂ρ
+ div( ρ ẋ ) = 0
∂t

integrale
ρ0
form J = det F =
ρ

26
3.2 Balance of momentum

The axiom of the balance of momentum reads:

l̇ = k . (3.14)

Therein,
Z
l = ρ ẋ dv ,
B
Z Z Z Z (3.15)
k= ρ b dv + t da = ρ b dv + σ · n da
B ∂B B ∂B

denote the vector of momentum and the vector of external forces. The external
forces k are to be split into two parts, namely into a part which is caused by the
local external body force ρ b integrated over the actual volume and a part caused
by the external contact force t = σ · n integrated over the surface of the actual
placement. The quantity σ is the Cauchy stress tensor.

With the material time derivative of the momentum


Z Z Z
l̇ = ( ρ ẋ dv ) = ( ρ ẋ J dV ) = ( ρ J ẋ )· dV
· ·

B B0 B0
Z
= ( ρ̇ J ẋ + ρ J̇ ẋ + ρ J ẍ ) dV
B0
Z
= [ ρ̇ J ẋ + ρ J ( div ẋ ) ẋ + ρ J ẍ ] dV
B0
Z
= [ ρ̇ ẋ + ρ ( div ẋ ) ẋ + ρ ẍ ] J dV
B0
Z
= [ ρ̇ ẋ + ρ ( div ẋ ) ẋ + ρ ẍ ] dv
B
Z Z
= [ ( ρ̇ + ρ div ẋ ) ẋ + ρ ẍ ] dv = ρ ẍ dv
| {z }
B = 0 B

27
=⇒
Z Z
·
l̇ = ( ρ ẋ dv ) = ρ ẍ dv (3.16)
B B

and the divergence theorem


Z Z
σ · n da = div σ dv (3.17)
∂B B

the balance of momentum con be transferred to


Z Z
ρ ẍ dv = ( div σ + ρ b ) dv
B B

=⇒
Z
[ div σ + ρ ( b − ẍ ) ] dv = o . (3.18)
B

This is the so-called Lagrange form of the balance equation. Using

ρ ẍ = (ρ ẋ )· − ρ̇ ẋ
∂( ρ ẋ )
= + [ grad( ρ ẋ ) ] · ẋ − ρ̇ ẋ
∂t
∂( ρ ẋ )
= + [ grad( ρ ẋ ) ] · ẋ − (−ρ div ẋ) ẋ
∂t
∂( ρ ẋ )
= + [ grad( ρ ẋ ) ] · ẋ + ρ ẋ div ẋ
∂t | {z }
= div( ρ ẋ ⊗ ẋ )
∂( ρ ẋ )
= + div( ρ ẋ ⊗ ẋ )
∂t
=⇒

∂( ρ ẋ )
ρ ẍ = + div( ρ ẋ ⊗ ẋ ) (3.19)
∂t

28
one obtains the alternative form
Z
∂( ρ ẋ )
{ div σ + ρ b − [ + div( ρ ẋ ⊗ ẋ ) ] } dv = o
∂t
B


Z
∂( ρ ẋ )
[ div σ + ρ b − − div( ρ ẋ ⊗ ẋ )] dv = o
∂t
B


Z
∂( ρ ẋ )
[ div( σ − ρ ẋ ⊗ ẋ ) + ρ b − ] dv = o
∂t
B

=⇒
Z
∂( ρ ẋ )
[ − div( σ − ρ ẋ ⊗ ẋ ) − ρ b ] dv = o . (3.20)
∂t
B

Considering the divergence theorem


Z Z
div( σ − ρ ẋ ⊗ ẋ ) dv = ( σ − ρ ẋ ⊗ ẋ ) n da (3.21)
B ∂B

the Euler form of the balance of momentum can be gained:


Z Z
∂( ρ ẋ )
( − ρ b ) dv = ( σ − ρ ẋ ⊗ ẋ ) n da . (3.22)
∂t
B ∂B

Equation (3.18) and (3.20), respectively, directly yields the local statment

div σ + ρ ( b − ẍ ) = o (3.23)

and

29
∂( ρ ẋ )
− div( σ − ρ ẋ ⊗ ẋ ) − ρ b = o , (3.24)
∂t

respectively, of the balance of momentum. Using the relation

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ + ρ + ( ẋ ⊗ ẋ ) · grad ρ + ρ div( ẋ ⊗ ẋ )
∂t ∂t ∂t

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ + ρ
∂t ∂t ∂t
+ ( ẋ ⊗ ẋ ) · grad ρ + ρ ẋ div ẋ + ρ (grad ẋ) · ẋ

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ + ρ
∂t ∂t ∂t
+ ẋ ( ẋ · grad ρ ) + ẋ ( ρ div ẋ ) + ρ (grad ẋ) · ẋ

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ + ρ + ẋ div( ρ ẋ ) + ρ (grad ẋ) · ẋ
∂t ∂t ∂t

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ + ẋ div( ρ ẋ ) + ρ + ρ (grad ẋ) · ẋ
∂t ∂t ∂t

∂( ρ ẋ ) ∂ρ ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ẋ [ + div( ρ ẋ ) ] + ρ + ρ (grad ẋ) · ẋ
∂t ∂t ∂t

∂( ρ ẋ ) ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ρ + ρ (grad ẋ) · ẋ
∂t ∂t

30
=⇒

∂( ρ ẋ ) ∂ ẋ
+ div( ρ ẋ ⊗ ẋ ) = ρ + ρ (grad ẋ) · ẋ , (3.25)
∂t ∂t

where the local statement of the balance equation of mass has been considered, see
(3.10), the Euler form of the balance of momentum results in

∂ ẋ
ρ + ρ (grad ẋ) · ẋ − div σ − ρ b = o . (3.26)
∂t

With the relation

1 1
σ= F · S · FT = P · FT (3.27)
J J

for the Cauchy stress tensor, where

S = J F−1 · σ · FT−1 = F−1 · P , P = F · S = J σ · FT−1 (3.28)

are the 2. (symmetric) and 1. (non-symmetric) Piola-Kirchhoff stress tensor, and


the transport theorem

da = J FT−1 · dA (3.29)

regarding surface elements, the volume integral of the divergence of σ can be trans-
ferred to
Z Z Z Z
div σ dv = σ · n da = σ · da = J σ · FT−1 · dA
B ∂B ∂B ∂B0
Z Z Z (3.30)
= P · dA = P · n0 dA = Div P dV .
∂B0 ∂B0 B0

Considering the transport theorem of volume elements and the integral form of the
balance of mass,

dv = J dV , ρ = J−1 ρ0 , (3.31)
31
the balance equation (3.18) can be reformulated as follows:
Z Z
ρ0 ẍ dV = ( Div P + ρ0 b ) dV .
B0 B

=⇒
Z
[ Div P + ρ0 ( b − ẍ ) ] dV = o . (3.32)
B0

Thus, with respect to the reference placement the local statement of the balance of
momentum (Lagrange form) is given by

Div P + ρ0 ( b − ẍ ) = o . (3.33)

32
Summary: Balance of momentum

Z Z Z
·
Axiom ( ρ ẋ dv ) = ρ b dv + t da
B B ∂B

global statements

reference Z
placement [ Div P + ρ0 ( b − ẍ ) ] dV = o
B0

actual Z
placement [ div σ + ρ ( b − ẍ ) ] dv = o
B
Z Z
∂( ρ ẋ )
( − ρ b ) dv = ( σ − ρ ẋ ⊗ ẋ ) n da
∂t
B ∂B

local statements

reference
placement Div P + ρ0 ( b − ẍ ) = o

actual
placement div σ + ρ ( b − ẍ ) = o

∂( ρ ẋ )
− div( σ − ρ ẋ ⊗ ẋ ) − ρ b = o
∂t

33
3.3 Weak formulation of the balance of momentum

The balance of momentum for a solid phase referring to the actual placement is
given by
Z
[ div σ + ρ ( b − ẍ ) ] dv = o . (3.34)
B

With the expression

1 1
σ= F · S · FT = P · FT (3.35)
J J

for the Cauchy stress tensor, where S and P = F · S are the second (symmetric) and
the first (non-symmetric) Piola-Kirchhoff stress tensor, and the transport theorem

da = J FT−1 · dA (3.36)

with respect to the surface elements, the volume integral of the divergence of σ in
(3.34) can expressed as
Z Z Z Z Z
div σ dv = σ · n da = σ · da = P · dA = P · n0 dA
B ∂B ∂B ∂B0 ∂B0

=⇒
Z Z
div σ dv = Div P dV . (3.37)
B B0

Furthermore, considering the transport theorem for the volume elements and the
relation that the partial density ρ of the actual placement is proportional to the
density ρ0 of the reference placement,

dv = J dV , ρ = J−1 ρ0 , (3.38)

the balance equation (3.34) can be written in the form

34
Z
[ Div P + ρ0 ( b − ẍ ) ] dV = o . (3.39)
B0

The equation (3.39) is the balance of momentum for a solid referred to the reference
placement.

The exact solution of the aforementioned balance equation of momentum (3.34) and
(3.39), respectively, for a solid exists only for very special and simple initial and
boundary problems. In view of calculation general problems, a numerical solution
technique is to be used. For this reason the Method of Weighted Residuals, in its
special form as the Galerkin-Method, is chosen.

The so-called weak formulation of the balance of momentum of the solid referring
to the reference placement reads
Z
[ Div P + ρ0 ( b − ẍ ) ] · Φ dV = 0 , (3.40)
B0

where the quantity Φ is the vectorial weighting function assigned to the aforemen-
tioned balance equation. With help of the divergence theorem

Div P · Φ = Div( PT · Φ ) − P : Grad Φ (3.41)

and the rule of integration


Z Z Z
T T
Div( P · Φ ) dV = P · Φ · n0 dA = P · n0 · Φ dA (3.42)
B0 ∂B0 ∂B0

the scalar expression (3.40) can be reformulated:


Z
[ Div( PT · Φ ) − P : Grad Φ + ρ0 ( b − ẍ ) · Φ ] dV = 0
B0


Z Z
[−P : Grad Φ + ρ0 ( b − ẍ ) · Φ ] dV = − Div( PT · Φ ) dV
B0 B0

35

Z Z
[ P : Grad Φ − ρ0 ( b − ẍ ) · Φ ] dV = Div( PT · Φ ) dV
B0 B0


Z Z
[ P : Grad Φ − ρ0 ( b − ẍ ) · Φ ] dV = PT · Φ · n0 dA
B0 ∂B0

=⇒
Z Z
[ P : Grad Φ − ρ0 ( b − ẍ ) · Φ ] dV = P · n0 · Φ dA . (3.43)
B0 ∂B0

The application of the linearization in the form of

LG = G + ∆G , (3.44)

where ∆(. . . ) denotes the rate of the quantity (. . . ), to (3.43) yields the following
linearized weak formulation of the balance of momentum for the empty porous solid
in the reference placement:
Z Z
∆P : Grad Φ dV + ρ0 ∆ẍ · Φ dV
B0 B0
R
= ( P + ∆P ) · n0 · Φ dA (3.45)
∂B0
Z Z
− P : Grad Φ dV + ρ0 ( b − ẍ ) · Φ dV .
B0 B0

For the i-th iteration step (i = 1, 2, 3, . . . n) at time t + ∆t the linearized equation


(3.45) reads

36
Z Z
i
t+ ∆P : Grad Φ dV + ρ0 ti+ ∆ẍ · Φ dV
B0 B0
Z
( it+ P + i

= t+ ∆P ) · n0 · Φ dA (3.46)
∂B0
Z Z
i−
i−
− t+ P : Grad Φ dV + ρ0 ( b − t+ ẍ ) · Φ dV ,
B0 B0

where i− = i − 1 and t+ = t + ∆t. Considering that the surface integral in (3.46) is


known, the local statement of this expression can be written as

( it+ P + i
· n0 · Φ = ( nt+ P + n
− −

t+ ∆P ) t+ ∆P ) · n0 · Φ
(3.47)
n
= t+ P · n0 · Φ = t+ P · n 0 · Φ ,

where is has been assumed that the surface integral is constant during the time step
from t to t + ∆t. With (3.47) equation (3.46) turns over into
Z Z
i
t+ ∆P : Grad Φ dV + ρ0 ti+ ∆ẍ · Φ dV
B0 B0
Z Z Z (3.48)
i

i

= t+ P · n0 · Φ dA − t+ P : Grad Φ dV + ρ0 ( b − t+ ẍ ) · Φ dV .
∂B0S B0 B0

For thermal processes the effective second (symmetric) Piola-Kirchhoff stress ten-
sor S is a function of the right Cauchy-Green deformation tensor ant the absolute
temperature of the solid, i.e.,

S = S ( Θ , C = FT · F ) . (3.49)

Thus, the rate of the effective first (non-symmetric) Piola-Kirchhoff stress tensor P
can be represented as

∂S ∂S
∆P = ∆( F · S ) = ∆F · S + F · ∆S = ∆F · S + F · ∆Θ + F · : ∆C
∂Θ ∂C

37
=⇒

4
∆P = ∆( F · S ) = ∆F · S + F · CΘ ∆Θ + F · CC : ∆C (3.50)

with

∂S 4 ∂S
CΘ = , CC = . (3.51)
∂Θ ∂C

The insertion of (3.50) into (3.48) yield the following form of the linearized weak
formulation of the balance of momentum:
Z Z 4
i i−
i− i− i
t+ ∆F · t+ S : Grad Φ dV + t+ F · C
t+ C : t+ ∆C : Grad Φ dV
B0 B0
Z Z
i− i− i
+ t+ F · t+ CΘ t+ ∆Θ : Grad Φ dV + ρ0 ti+ ∆ẍ · Φ dV
B0 B0
Z Z (3.52)
i −
= t+ P · n0 · Φ dA − t+ P : Grad Φ dV
∂B0S B0
Z Z
ρ0 it+ ẍ · Φ dV .

+ ρ0 b · Φ dV −
B0 B0

For quasi-static processes equation (3.52) simplifies to


Z Z 4
i i− i− i− i
t+ ∆F · t+ S : Grad Φ dV + t+ F · C
t+ C : t+ ∆C : Grad Φ dV
B0 B0
Z
i− i− i
+ t+ F · t+ CΘ t+ ∆Θ : Grad Φ dV (3.53)
B0
Z Z Z
i−
= t+ P · n0 · Φ dA + ρ0 b · Φ dV − t+ P : Grad Φ dV .
∂B0S B0 B0

38
3.4 Balance of moment of momentum

For non-polar bodies the axiom of the balance of moment (angular) momentum
reads

ḣ(0) = m(0) , (3.54)

where
Z
h(0) = x × ρ ẋ dv ,
B
Z Z Z Z (3.55)
m(0) = x × ρ b dv + x × t da = x × ρ b dv + x × σ · n da
B ∂B B ∂B

are the moment of momentum with respect to a fixed reference point (0) of the
configuration space and the momentum of external forces with respect to the same
reference point.

With the material time derivative of h(0) ,


Z Z Z
· ·
ḣ(0) = ( x × ρ ẋ dv ) = ( x × ρ ẋ J dV ) = ( x × ρ J ẋ )· dV
B B0 B0
Z
= [ ẋ × ρ J ẋ + x × ( ρ J ẋ )· ] dV
B0
Z
= [ ẋ × ρ J ẋ + x × ( ρ̇ J ẋ + ρ J̇ ẋ + ρ J ẍ ) ] dV
B0
Z
= { ẋ × ρ J ẋ + x × [ ρ̇ J ẋ + ρ J ( div ẋ ) ẋ + ρ J ẍ ] } dV
B0
Z
= { ẋ × ρ ẋ + x × [ ρ̇ ẋ + ρ ( div ẋ ) ẋ + ρ ẍ ] } J dV
B0

39
Z
ḣ(0) = { ẋ × ρ ẋ + x × [ ( ρ̇ + ρ div ẋ ) ẋ + ρ ẍ ] } dv
| {z } | {z }
B =o =o
Z
= x × ρ ẍ dv
B

=⇒
Z Z
·
ḣ(0) = ( x × ρ ẋ dv ) = x × ρ ẍ dv , (3.56)
B B

and the divergence theorem


Z Z
x × σ · n da = ( x × div σ + I × σ ) dv (3.57)
∂B B

the balance of moment of momentum can be transferred to


Z Z Z
x × ρ ẍ dv = x × ρ b dv + ( x × div σ + I × σ ) dv
B B B


Z Z
x × ρ ẍ dv = [ x × ( div σ + ρ b ) + I × σ ] dv
B B

=⇒
Z
{ x × [ div σ + ρ ( b − ẍ ) ] + I × σ } dv = o . (3.58)
B

Considering the local statement of the balance of momentum,

div σ + ρ ( b − ẍ ) = o , (3.59)

the balance of moment of monmentum simplifies to

40
Z
I × σ dv = o . (3.60)
B

Thus, the local statement of the balance equation reads:

I×σ =o . (3.61)

Using the calculation rule for the vector product of two second order tensors (cross
product of second order tensors),

3
I × σ = E : ( I · σT ) ,

equation (3.61) is fulfilled if

σ = σT , (3.62)

i.e., the Cauchy stress tensor is symmetric.

41
Summary: Balance of moment of momentum

Z Z Z
·
Axiom ( x × ρ ẋ dv ) = x × ρ b dv + x × t da
B B ∂B

global statements

reference
placement

actual Z
placement I × σ dv = o
B

local statements

reference
placement

actual
placement I × σ = o =⇒ σ = σ T

42
3.5 Balance of energy (first law of thermodynamic)

For the description of thermo-mechanical effects, the balance of energy is of main in-
terest regarding the coupling of thermal fields (temperature, heat flux, internal heat
source) and mechanical fields (motion, velocity, acceleration, deformations, defor-
mation velocities). The axiom of the balance of energy (first law of thermodynamic)
reads:

Ė + K̇ = W + Q . (3.63)

The scalar quantities


Z
E = ρ ε dv ,
B
Z
1
K = ρ ẋ · ẋ dv ,
2
B
Z Z Z Z
W= ẋ · ρ b dv + ẋ · t da = ẋ · ρ b dv + ẋ · σ · n da (3.64)
B ∂B B ∂B
Z Z
= ẋ · ρ b dv + ẋ · σ · da ,
B ∂B
Z Z Z Z
Q = ρ r dv − q · n da = ρ r dv − q · da
B ∂B B ∂B

denote the internal energy (E), where ε is the specific internal energy, the kinetic
energy (K), the increment of the mechanical work of the external forces (W) and
the increment of the non-mechanical work Q. The non-mechanical work Q consists
of two parts caused by the local external heat supply r = r (x, t) per mass element
ρ dv and the heat influx vector q = q (x, t).

With the material time derivations


Z Z
Ė = ( ρ ε dv ) = ( ρ ε J )· dV
·

B B0
Z
= ( ρ̇ ε J + ρ ε̇ J + ρ ε J̇ ) dV
B0

43
Z
Ė = ( ρ̇ ε J + ρ ε̇ J + ρ ε J div ẋ ) dV
B0
Z Z
= ( ρ ε̇ + ε ( ρ̇ + ρ div ẋ ) dv = ρ ε̇ dv ,
| {z }
B =0 B
Z Z
1 1
K̇ = ( ρ ẋ · ẋ dv )· = ( ρ J ẋ · ẋ )· dV
2 2
B B0
Z
1
= ( ρ̇ J ẋ · ẋ + ρ J̇ ẋ · ẋ + ρ J ẍ · ẋ + ρ J ẋ · ẍ ) dV
2
B0
Z
1
= ( ρ̇ J ẋ · ẋ + ρ J ẋ · ẋ div ẋ + 2 ρ J ẋ · ẍ ) dV
2
B0
Z
1
= ( ρ̇ ẋ · ẋ + ρ ẋ · ẋ div ẋ + 2 ẋ · ẍ ) dv
2
B
Z Z
1
= [ ẋ · ẋ ( ρ̇ + ρ div ẋ ) +2 ρ ẋ · ẍ ] dv = ρ ẋ · ẍ dv
2 | {z }
B =0 B

=⇒
Z Z
·
Ė = ( ρ ε dv ) = ρ ε̇ dv ,
B B
Z Z (3.65)
1
K̇ = ( ρ ẋ · ẋ dv )· = ρ ẋ · ẍ dv
2
B B

and the relations


Z Z Z
T
ẋ · σ · da = σ · ẋ · da = div( σT · ẋ ) dv
∂B ∂B B


Z Z
ẋ · σ · da = ( div σ · ẋ + σ : grad ẋ ) dv
∂B B

44

Z Z
ẋ · σ · da = ( div σ · ẋ + σ : L ) dv ,
∂B B

=⇒
Z Z
ẋ · σ · da = ( div σ · ẋ + σ : L ) dv ,
∂B B
Z Z (3.66)
q · da = div q dv
∂B B

the balance of energy (3.63) can be transferred to


Z Z
ρ ε̇ dv + ρ ẍ · ẋ dv =
B B
Z Z (3.67)
= [ ( div σ + ρ b ) · ẋ + σ : L ] dv + ( ρ r − div q ) dv .
B B

Thus, the local statement of the balance of energy reads

ρ ε̇ = [ div σ + ρ ( b − ẍ ) ] · ẋ + σ : L + ρ r − div q . (3.68)

Considering the local statement of the balance of momentum,

div σ + ρ ( b − ẍ ) = o ,

see (3.23), one obtains the following representation form:

ρ ε̇ − σ : L − ρ r + div q = 0 . (3.69)

Taking into account the symmetric characteristic of the stress tensor σ = σ T and
the additive decomposition of L = D + W into a symmetric and a non-symmetric
part (D = DT , W = −WT ), the scalar product σ : L in (3.69) results in
45
Figure 3.1: Illustration of heat flux and internal heat source

: W} = σ : D .
σ : L = σ : ( D + W ) = σ : D + |σ {z
(3.70)
=0

Thus, one obtains the following statement of the balance of energy:

ρ ε̇ − σ · D − ρ r + div q = 0 . (3.71)

The introduction of the Helmholtz free energy function ψ = ψ (x, t),

ψ = ε − Θη , (3.72)

and the corresponding material time derivative,

ψ̇ = ε̇ − Θ̇ η − Θ η̇ , (3.73)

yields the alternative form

ρ ( ψ̇ + Θ̇ η + Θ η̇ ) − σ : D − ρ r + div q = 0 (3.74)
46
of the local balance of energy. The quantities η = η (x, t) and Θ = Θ (x, t) are the
specific entropy and the absolute temperature.

47
Summary: Balance of energy

Z Z
· 1
Axiom ( ρ ε dv ) + ( ρ ẋ · ẋ dv )· =
2
B B
Z Z Z Z
= ẋ · ρ b dv + ẋ · t da + ρ r dv − q · n da
B ∂B B ∂B

global statements

actual Z Z
placement ρ ε̇ dv + ρ ẍ · ẋ dv =
B B
Z Z
= [ ( div σ + ρ b ) · ẋ + σ : L ] dv + ( ρ r − div q ) dv
B B

Z Z Z
ρ ε̇ dv = σ : D dv + ( ρ r − div q ) dv
B B B

local statements

actual
placement ρ ε̇ − σ : D − ρ r + div q = 0

ρ ( ψ̇ + Θ̇ η + Θ η̇ ) − σ : D − ρ r + div q = 0

48
Remark:
The kinetic energy of a body can be transformed into potential energy and back-
wards. In this case it is a matter of purely mechanical conversion of energy.

The kinetic energy can not disappear, but the energy distributes to the atoms,
i.e., the kinetic energy of the disordered motion of atoms can be transformed into
potential energy between the atoms. One can not see this energy. Therefore, one
says the kinetic energy is transformed into internal energy. This energy can be felt
and measured as heat. The temperature is a quantity for the (medial) kinetic energy
of the atoms.

Contributions of the internal energy:


– kinetic energy of molecules (see example),
– potential energy of molecules
(liquid or vapor expand
⇒ increasing distance of the molecules ⇒ evaporation),
– chemically bond energy of molecules.

49
3.5.1 Balance of energy – alternative representation forms

As aforementioned, the axiom of the balance of energy (first law of thermodynamics)


reads:

Ė + K̇ = W + Q , (3.75)

where
Z
E = ρ ε dv ,
B
Z
1
K = ρ ẋ · ẋ dv ,
2
B
Z Z Z Z (3.76)
W= ẋ · ρ b dv + ẋ · t da = ẋ · ρ b dv + ẋ · σ · n da ,
B ∂B B ∂B
Z Z
Q = ρ r dv − q · n da .
B ∂B

Taking into account the balance equation of mass, the material time derivatives of
the internal and kinetic energy E and K are given by
Z Z
·
Ė = ( ρ ε dv ) = ρ ε̇ dv ,
B B
Z Z (3.77)
1
K̇ = ( ρ ẋ · ẋ dv )· = ρ ẋ · ẍ dv .
2
B B

With
Z Z
ẋ · σ · n da = ( div σ · ẋ + σ : L ) dv (3.78)
∂B B

and the local statements of the balance of momentum and balance of moment of
momentum (symmetric characteristic of the stress tensor σ = σ T ) the rate of the
mechanical work W can be reformulated:
50
Z Z
W= ẋ · ρ b dv + ( div σ · ẋ + σ : D ) dv
B B
Z
= [ ( div σ + ρ b ) ·ẋ + σ : D ] dv
| {z }
B = ρ ẍ

=⇒
Z Z
W= ( ρ ẍ · ẋ + σ : D ) dv = ( ρ ẍ · ẋ ) dv + Win . (3.79)
B B

The quantity
Z
Win = σ : D dv (3.80)
B

denotes the rate of the internal mechanical work. With the additive decomposition
of the stress tensor σ and the symmetric part of the spatial velocity gradient D into
deviatorical and spherical parts,

1 1
σ = σD + ( σ : I ) I = σD − p I , p = − ( σ : I ) ,
3 3
(3.81)
1
D = DD + ( D : I ) I ,
3
and the relation (. . . )D · I = 0, the rate of the mechanical work can be expressed as
Z Z
1
Win = σ : D dv = [ σ D − p I ] : [ DD + ( D : I ) I ] dv
3
B B
Z
1
= [ σ D : DD − p ( DD : I ) + ( σ D : I ) ( D : I ) −
| {z } 3 | {z }
B =0 =0
1
− p ( D : I ) ( I : I ) ] dv
3 | {z }
=3
Z
= (σ D : DD − p D : I ) dv
B

51
Z Z
Win = σ : D dv = ( σ D : DD − p div ẋ ) dv
B B
Z
= ( σ D : DD − p J−1 J̇ ) dv
B

=⇒
Z Z
Win = σ : D dv = ( σ D : DD − p J−1 J̇ ) dv . (3.82)
B B

With help of
Z Z
q · n da = div q dv (3.83)
∂B B

the rate of the non-mechanical work can be presented as


Z
Q= ( ρ r − div q ) dv . (3.84)
B

With the details mentioned before, one obtains the following global statement of the
balance of energy:
Z Z
Ė + K̇ = ρ ε̇ dv + ρ ẍ · ẋ dv
B B
Z Z
(3.85)
= ( ρ ẍ · ẋ + σ D : DD − p J−1 J̇ ) dv + ( ρ r − div q ) dv
B B

=W+Q .

Thus, the material time derivative of the internal energy reads:

52
Z
Ė = ρ ε̇ dv = W + Q − K̇
B
Z
= ρ ε̇ dv
B (3.86)
Z Z
D D −1
= (σ : D − pJ J̇ ) dv + ( ρ r − div q ) dv
B B

= Win + Q .

53
3.5.2 Balance of energy for gas and non-viscous fluids

With respect to the description of the behavior of gas and non-viscous fluids the
deviatorical part of the stresses can be approximately neglected, i.e.,

σD = 0 . (3.87)

In this case, the expression for the rate of the internal work results in
Z Z
Win = − p J−1 J̇ dv = − p J̇ dV , (3.88)
B B0

where regarding the second term the transport theorem dv = JdV has been used.
With the assumption that the pressure is constant in space one obtains
Z Z Z
·
Win = − p J̇ dV = − p ( J dV ) = − p ( dv )·
B0 B0 B (3.89)
dv
= − p V̇ = − p .
dt
Insertion this relation into the expression for Ė yields:

Ė = Win + Q = − p V̇ + Q . (3.90)

54
3.6 Entropy inequality (second law of thermodynamics)

The entropy is a measure of a part of heat quantity which can not transformed
into mechanical work due to the equal distributions of the molecules of the system.
Thus, the entropy inequality (second law of thermodynamics) is defined as follows
Z Z
1 1
Ḣ ≥ ρ r dv − q · n da , (3.91)
Θ Θ
B ∂B

where
Z
H= ρ η dv (3.92)
B

denote the entropy and η the specific entropy. Considering the materiel time deriva-
tives
Z Z
·
Ḣ = ( ρ η dv ) = ( ρ η J )· dV
B B0
Z
= ( ρ̇ η J + ρ η̇ J + ρ η J̇ ) dV
B0
Z
= ( ρ̇ η J + ρ η̇ J + ρ η J div ẋ ) dV
B0
Z Z
= [ ρ η̇ + η ( ρ̇ + ρ div ẋ ) ] dv = ρ η̇ dv
| {z }
B =0 B

=⇒
Z Z
·
Ḣ = ( ρ η dv ) = ρ η̇ dv (3.93)
B B

and the relation


Z Z
1 1
q · n da = div( q ) dv (3.94)
Θ Θ
∂B B

55
the entropy inequality results in
Z Z
1 1
ρ η̇ dv ≥ [ ρ r − div ( q ) ] dv . (3.95)
Θ Θ
B B

With help of (3.95), the local statement

1 1
ρ η̇ − ρ r + div ( q ) ≥ 0 . (3.96)
Θ Θ

is gained. Using the relation

1 1 1
div ( q ) = − 2 grad Θ · q + div q (3.97)
Θ Θ Θ

one obtains the following form of the second law of thermodynamics:

1 1
ρ η̇ − ( ρ r − div q ) − 2 q · grad Θ ≥ 0 . (3.98)
Θ Θ

Taking into account the balance of energy, see (3.71),

ρ ε̇ − σ : D − ρ r + div q = 0

and

ρ r − div q = ρ ε̇ − σ : D ,

respectively one obtains the following representation form of the inequality (3.98):

1 1
ρ η̇ − ( ρ ε̇ − σ : D ) − 2 q · grad Θ ≥ 0 . (3.99)
Θ Θ

The multiplication of the aforementioned inequality with Θ (regarding the absolute


temperature it is essential: Θ ≥ 0) yields

1
Θ ρ η̇ − ρ ε̇ + σ : D − q · grad Θ ≥ 0
Θ
56
=⇒

1
− ρ ( ε̇ − Θ η̇ ) + σ : D − q · grad Θ ≥ 0 . (3.100)
Θ

The inequality (3.100) can be reformulated as follows:

1
− ρ ( ε̇ − Θ η̇ − Θ̇ η + Θ̇ η ) + σ : D − q · grad Θ ≥ 0
Θ

=⇒

1
− ρ [ ( ε − Θ η )· + Θ̇ η ] + σ : D − q · grad Θ ≥ 0 . (3.101)
Θ

The expression ( ε − Θ η )· represents a total differential. Thus, these quantities in


the brackets can be combined to one quantity. This quantity

ψ = ε − Θη (3.102)

is called as Helmholtz free energy, compare (3.72). Finally, with (3.102) the local
statement

1
− ρ ( ψ̇ + Θ̇ η ) + σ : D − q · grad Θ ≥ 0 . (3.103)
Θ

of the entropy inequality can be derived.

Special cases:
For processes with constant temperature in space, the global statement of the second
law of thermodynamics, see (3.91) and (3.95), respectively the global statement of
the second law of thermodynamics simplifies to
Z Z
1
ρ η̇ dv ≥ ( ρ r − div q ) dv (3.104)
Θ
B B

and

1
Ḣ ≥ Q, (3.105)
Θ
57
respectively, where (3.84) has taken into consideration. This relation is the starting
point regarding the formulation of the well-known Gibbs equation in the framework
of the gas dynamics.

Remark:
With respect to the derivation of the local statement (3.74) of the balance of en-
ergy, the introduced relationship for the Helmholtz free energy funktion ψ = ε − Θ η
seemed arbitrary. The expression for ψ recently arises in connection with the refor-
mulation with the entropy inequality. The reformulation of the entropy implies the
change of variables, i.e. the substitution of the variable set {ε, η} by the set {ψ, η}.
The approach is called Legendre transformation.

58
Summary: Entropy inequality

Z Z Z
· 1 1
Axiom ( ρ η dv ) ≥ ρ r dv − q · n da
Θ Θ
B B ∂B

global statement

actual Z Z
1 1
placement ρ η̇ dv ≥ [ ρ r − div ( q ) ] dv
Θ Θ
B B

local statement

actual
1 1
placement ρ η̇ − ρ r + div ( q ) ≥ 0
Θ Θ

1
− ρ ( ε̇ − Θ η̇ ) + σ : D − q · grad Θ ≥ 0
Θ

1
− ρ ( ψ̇ + Θ̇ η ) + σ : D − q · grad Θ ≥ 0
Θ

59
3.6.1 Entropy inequality for gas and non-viscous fluids

Assuming that the temperature is constant over the considered control space, the
global statement of the entropy inequality simplifies to

1
Ḣ ≥ Q, (3.106)
Θ

see (3.105). In case of constant hydrostatic pressure in the control space, for gas
and non-viscous fluids the rate of the non-mechanical work ca be expressed as

Q = Ė + p V̇ . (3.107)

Thus, with the aforementioned assumptions the entropy inequality reduces to

1
Ḣ ≥ ( Ė + p V̇ ) . (3.108)
Θ

Dissolving the inequality the rate of the internal energy reads

Ė ≤ Θ Ḣ − p V̇ . (3.109)

Taking into account the aforementioned inequality, alternative representation form


of the second law of thermodynamics can be derived:

Ė ≤ ( Θ H )· − Θ̇ H − p V̇
Ė − ( Θ H )· ≤ − Θ̇ H − p V̇
( E − Θ H )· ≤ − Θ̇ H − p V̇ ⇒ Ψ̇ ≤ − Θ̇ H − p V̇ ,
Ė ≤ Θ Ḣ − ( p V )· + ṗ V
Ė + ( p V )· ≤ Θ Ḣ + ṗ V (3.110)
( E + p V )· ≤ Θ Ḣ + ṗ V ⇒ Ṡ ≤ Θ Ḣ + ṗ V ,
Ė ≤ ( Θ H )· − Θ̇ H − ( p V )· + ṗ V
Ė − ( Θ H )· + ( p V )· ≤ − Θ̇ H + ṗ V
( E − Θ H + p V )· ≤ − Θ̇ H + ṗ V ⇒ Ġ ≤ − Θ̇ H + ṗ V .
60
The relation show, that besides the internal energy E additional global energetic
quantities (global thermodynamic potentials – total differentials) can be defined:

E – internal energy ,
Ψ = E − ΘH – free energy ,
S = E + pV – enthalpy ,
G = E − ΘH +pV – free enthalpy .

61
3.6.2 Gibbs equation

For reversible processes the inequality (3.108) pass over into the equation

1
Ḣ = ( Ė + p V̇ ) .
Θ
This relation is the well-known Gibbs equation (Josiah Willard Gibbs, 1839 –
1903, American physicist).

Considering the energetic quantities Ψ, S and G, the following alternative represen-


tations forms of the Gibbs equation can be derived:

Ė = Θ Ḣ − p V̇ = − p V̇ + Q ,
Ψ̇ = ( E − Θ H )· = − Θ̇ H − p V̇ = − Θ̇ H − Θ Ḣ − p V̇ + Q ,
Ṡ = ( E + p V )· = Θ Ḣ + ṗ V = ṗ V + Q ,
Ġ = ( E − Θ H + p V )· = − Θ̇ H + ṗ V = − Θ̇ H − Θ Ḣ + ṗ V + Q .
Bearing in mind that for reversible processes Q = Θ Ḣ is valid, compare (3.106).

Remark:
Alternatively, well-known representation forms of the Gibbs equation used in the
literature can be derived, where the symbols E = U and H = S are used:

1
Ṡ = ( U̇ + p V̇ )
Θ
=⇒

dS 1 dU dV 1
= ( +p ) ⇒ dS = ( dU + p dV ) .
dt Θ dt dt Θ
In analogy with the aforementioned explanations the following relations are valid:

dU = Θ dS − p dV ,
d( U − Θ S ) = −dΘ S − p dV ,
d( U + p V ) = Θ dS + dp V ,
d( U − Θ S + p V ) = − dΘ S − p dV .

62
4. Concept of stresses

Considering a cut out part Bp of the body B and its Lagrangian counterpart B0p
of the body B0 , respectively, see Fig. 4.1. Introducing a traction vector t on the
surface ∂Bp , which represents the mechanical action of the cut off part of the rest
of the body.

reference placement actual placement


F

n0 t0
dA n t
Bp0
X da
Bp
∂Bp0
B0 X B
∂Bp

cut ∂B0

cut
∂B

Figure 4.1: Concept of stresses

4.1 Cauchy stress tensor σ

The most simple assumption is that the traction vector (traction force) t depends
linearly on the normal vector n of the surface ∂Bp , i.e.,

t(x, t, n) = σ(x, t) · n .

This relation is the so-called Cauchy Theorem.

The Cauchy stress σ maps the normal vector n onto the traction vector t:

70
t=σ·n .

The Cauchy stress σ relates the actual force in the cut to its actual size, see Fig.
4.1. Thus, the Cauchy stress is denoted at the true stress.

Note: As a consequence of the balance of moment of momentum, the Cauchy stress


is symmetric:

σ = σT .

4.2 Kirchhoff stress tensor τ

The Kirchhoff stress tensor is obtained from σ by multiplication with the Jacobian
determinant (volume ratio):

τ = J σ = det F σ ,

where

Jt = τ · n

The stress tensor τ is symmetric, because J is only a scalar value.

4.3 First Piola-Kirchhoff stress tensor P


Constructing the Lagrangian counterpart of the Eulerian Cauchy theorem (t = σ·n)
such that

t0 ( X , t , n0 ) = P ( X , t ) · n0

with the normal vector n0 to the undeformed surface ∂B0 and the traction vector
t0 associated to the undeformed surface element dA, noted also as ”normal traction
vector”, see 4.1. The vectors t0 and t have the same directions, but different length.

The vector t0 follows from the condition

71
t0 dA = t da .

With the vectors

dA = n0 dA , da = n da

and the transformation law

da = J FT−1 dA

for normal vectors one obtains

t0 dA = t da

P · n0 dA = σ · n da

P · dA = σ · da

=⇒

P · dA = J σ · FT−1 · da .

Thus, the first Piola-Kirchhoff stress tensor (1. PK) ca be identified to

P = J σ · FT−1 = τ · FT−1 .

The stress tensor P is denoted as nominal stress, but its related the true stress σ
to the undeformed area, see Fig. 4.2. The stress P is a two-field tensor and has
therefore no symmetric properties.

72
4.4 Second Piola-Kirchhoff stress tensor S

The second Piola-Kirchhoff stress tensor denoted by S is a pure Lagrangian stress


tensor. Therefore, the nominal traction vector t0 is transformed to the Lagrangian
manifold:

t̃0 = F−1 · t0

and a Cauchy-type theorem in the form

t̃0 = S · n0 .

Reformulating the definition of the 2. PK stress with

t̃0 = F−1 · t0

t̃0 = F−1 · P · n0

=⇒
reference placement
A(t0) = A0

F(t0) = F F(t0 ) = F

actual placement A(t) = A

F(t) = F F(t) = F
F F
p11 = , σ11 =
A0 A
Figure 4.2: Coefficients of the 2. PK and the Cauchy stress tensor under tensile
load

73
t̃0 = F−1 · τ · FT−1 · n0 = S · n0 .

Thus, the 2. PK stress can be identified as

S = F−1 · τ · FT−1 = F−1 · P = J F−1 · σ · FT−1 .

Notice, the stress tensor S is symmetric due to the symmetry of τ :

ST = ( F−1 · τ · FT−1 )T

ST = F−1 · ( F−1 · τ )T

=⇒

ST = F−1 · τ T · FT−1 = F−1 · τ · FT−1 = S .

4.5 Stress Power P

The stress power P for a unit volume of the reference placement is defined as

1
P = S : Ė = S : Ċ .
2

A transformation to the Eulerian configuration yields

P = S : Ė = F−1 · τ · FT−1 : Ė

P = ĖT · F−1 · τ : F−1


74
P = FT−1 · ĖT · F−1 : τ T

P = τ : FT−1 · ĖT · F−1

=⇒

P = τ : FT−1 · Ė · F−1

where the symmetry property of Ė has been considered. Taking into account the
upper Lie derivative of the Almansi strain tensor,

A△ = FT−1 · Ė · F−1 = Ȧ + LT · A + A · L = D ,

see (2.39), the stress power can be written as

P = τ : A△ = τ : D .

The values ( S; Ė ) or ( S; Ċ ) and ( τ ; A△ ) or ( τ ; D ) are work conjugated quantities


and they define the stress power.

75
Summary: Concept of stresses

• Cauchy stresses and true stresses, respectively

1 1 1
σ= F · S · FT = P · FT = τ , σ = σ T
J J J

• Kirchhoff stresses

τ = F · S · FT = J σ , τ = τ T

• First Piola-Kirchhoff stresses (1. PK)

P = F · S = J σ · FT−1 = τ · FT−1 , P 6= PT

• Second Piola-Kirchhoff stresses (2. PK)

S = F−1 · P = F−1 · τ · FT−1 = J F−1 · σ · FT−1 , S = ST

76
reference configuration
S = sik Gi ⊗ Gk

⇓ F · ( . . . ) · FT ⇓

actual configuration
τ = τ ik gi ⊗ gk , τ ik = sik
1 ik
σ = σ ik gi ⊗ gk , σ ik = s
J

reference configuration
S = sik Gi ⊗ Gk

⇓ F · (...) ⇓

actual/reference configuration
P = pik gi ⊗ Gk , pik = sik

Table 4.1:
Stress tensors by using the notation in a natural base system

77
5. General linearization of deformation and strain measures

Nonlinearities appear in continuum due to different phenomena:


• geometrical nonlinearities: nonlinear strain measures, e.g. Green-Lagrange
strain E
• physical nonlinearities (material): plasticity, damage, nonlinear constitutive
behavior
→ linearizations have to be derived in order to use e.g. iterative solution technics
(Newton iteration)

For a general linearization, here the Gâteaux derivative (directional derivative) in


combination with a Taylor expansion is used.

78
2. Basic concepts

2.1 Symbols (tensor coefficients)

– Symbols of first order

ui ⇒ u1 , u2 , u3 , . . . covariant representation
ui ⇒ u1 , u2 , u3 , . . . contravariant representation

– Symbols of second order

tik ⇒ t11 , t12 , t13 , . . .


t21 , t22 , t23 , . . .
t31 , t32 , t33 , . . . covariant representation
tik ⇒ t11 , t12 , t13 , . . . contravariant representation
ti k ⇒ t11 , t12 , t13 , . . . mixed variant representation
ti k ⇒ t11 , t12 , t13 , . . . mixed variant representation

– Symbols of higher order

tiko , tiko , ti ko , ti ko , . . .
tikop , tikop , tikop , tiko p , . . .

2
2.2 Einstein’s summation convention

When, in a symbol or in a multiplicative term, an index appears twice and this


index stands above and below (superscript and subscript or co- and contravariant),
then the summation must be done over this index:

n
i
si i = s11 + s22 + s33 + . . . + snn ,
P
s i =
i=1
3
m
um vm = u1 v1 + u2 v2 + u3 v3 .
P
um v =
m=1

The summation index (called silent or dummy index) does not appear in its corre-
sponding form after the summation has been carried out:

sik tkm = si1 t1m + si2 t2m + si3 t3m + . . . = ri m .

Here, the letter r has been choosen for the sum of the products. The indices i and
m are named as free indices.

The summation index can be arbitrary renamed and the result of the summation
does not change:

ui vi = u1 v1 + u2 v2 + u1 v1 + . . . + un vn ,
ur vr = u1 v1 + u2 v2 + u1 v1 + . . . + un vn ,
⇒ ui vi = ur vr .

Remark:
sii , sik tkm ⇒ no summation

3
2.3 Kronecker Symbol

The Kronecker symbol (Kronecker delta) δki is a scalar quantity and has the following
values:

1 for i = k ,
δki =
0 for i 6= k .

When the indexed symbols are multiplied with the Kronecker symbol, then the
summation index (silent index) by these symbols is exchanged with the free index
of the Kronecker symbol and the Kronecker symbol is set equal one.

– Examlpes

sk δki = s1 δ1i + s2 δ2i + s3 δ3i


with
i = 1 : s1 δ11 + s2 δ21 + s3 δ31 = s1 ,
i = 2 : s1 δ12 + s2 δ22 + s3 δ32 = s2 ,
i = 3 : s1 δ13 + s2 δ23 + s3 δ33 = s3
⇒ sk δki = si

sim δm
k
= sik , sim δis δm
k
= ssk , sik smn δio δpm = sok spn

4
3. Vector algebra

3.1 Vector notion and vector operations

A vector v is defined as a class of all directional arrows having the same length and
direction.

Figure 3.1: Directional arrow

A vector is geometrical and physical quantity, respectively, which is characterized by


a magnitude (size, norm), a direction and possibly the point of attack. In the three
dimensional vector space (Euclidean space) a vector can be presented as follows:

v = v i gi = v 1 g1 + v 2 g2 + v 3 g3
= vi g i = v1 g 1 + v 2 g 2 + v 3 g 3 .

The quantities gi and gi are covariant and contravariant basic vectors, where the
second system of base vectors gi is reciprocal to gi . The covariant and contravariant
vector coefficients of the vector components vi gi and vi gi are denoted by vi and vi ,
respectively. It should be mentioned that the coefficients are not independent of
each other. In general, the base vectors g1 , g2 and g3 as well as g1 , g2 and g3 are
not standardized and are not perpendicular at each other.

In many cases it is necessary to know the real quantities of the coefficients of a


vector. Therefore, a presentation of the vector referring to standardized (unit) base
vectors is appropriate:
5
v = v i gi = v 1 g1 + v 2 g2 + v 3 g3
g1 g2 g3
= v1 |g1 | + v2 |g2 | + v3 |g3 |
|g1 | |g2 | |g3 |
∗ g1 ∗ g2 ∗ g3 ∗ gi
= v1 + v2 + v3 = vi
|g1 | |g2 | |g3 | |gi |

and

v = vi gi = v1 g1 + v2 g2 + v3 g3
g1
1 2 g
2
3 g
3
= v1 |g | 1 + v2 |g | 2 + v3 |g | 3
|g | |g | |g |
∗ g1 ∗ g
2
∗ g
3
∗ g
i
= v1 + v2 2 + v3 3 = vi i ,
|g1 | |g | |g | |g |

where

√ √ p p
|gi | = gi · gi = g(ii) , |gi | = gi · gi = g(ii)

are the norms of the base systems. The quantities

∗ √ √
vi = vi |gi | = vi gi · gi = vi g(ii)

and

∗ i
p p
vi = vi |g | = vi gi · gi = vi g(ii)

denote the so-called physical coefficients of the corresponding vector.

Using a presentation of a vector with respect to orthogonal unit vectors ei = ei


(orthonormed base vectors), i.e.,

v = v̄i ei = v̄i ei = v̄1 e1 + v̄2 e2 + v̄3 e3 = v̄1 e1 + v̄2 e2 + v̄3 e3 ,

the coefficients of the vector are physical coefficients:


6
∗ ∗
v̄i = v̄i = v̄i = v̄i .

In context with the presentation of vectors, a reciprocal base system gi is introduced


in according to the following rules:

gi · gk = gk · gi = δki .

The base vectors gk and gi are termed as covariant and contravariant base vectors.
The scalar products of the contravariant base vectors and the covariant base vectors

gi · gk = gk · gi = gik = gki ,
gi · gk = gk · gi = gik = gki

are symmetrical for the indices i and k due to the commutative rule. The quantities
gik and gik are named as covariant metric and contravariant metric coefficients.

A base, which consists of a system of base vectors e1 , e2 and e3 that are normed
(unit vectors) and are reciprocally orthogonal are called orthonormed and it is valid
that

1 for i = k ,
ei · ek = δik =
0 for i 6= k .

In an orthonormed system, the differences between the covariant and contravariant


base vectors vanish, i.e.,

ei = ei .

Thus, it is normal to represent coefficients of the vector components as covariant, and


to renounce the mixed-variant placement of the indices in the summation convention.

7
– Addition of vectors

u+v =v+u ,
u + (v + w) = (u + v) + w ,
u+o=u ,
u + (− u) = o ,
u + (− v) = u − v ,
(α + β )u = αu + β u ,
α(u + v) = αu + αv .

The vector o is the zero-vector.

Figure 3.2: Addition of vectors

8
– Scalar product of vectors

u · v = |u| |v| cos ϕ ,


u·v =v·u ,
u · (v + w) = u · v + u · w ,
(αu) · v = u · (αv) = α(u · v) ,
u · u = |u|2 .

The positive square root



|u| = + u · u

gives the amount or the norm of the vector u. It should be mentioned that the scalar
product of the two vectors u and v is equal to zero if one vector is a zero-vector or
if u is perpendicular to v (u and v are orthogonal):

u · v = 0 if u = o and/or v = o and/or u ⊥ v .

9
4. Tensor algebra

4.1 Tensor notion

A tensor of second order is defined as a linear mapping in which the vector u of the
vector space is transferred into the vector v of another vector space:

T·u=v .

The vectors u and v are elements of the three-dimensional Euclidian space E 3 ; the
tensor T is element of the tensor product space E 3 ⊗ E 3 . In the context with a
linear mapping of a second order tensor and a vector, the symbol (·) denotes a
simple contraction (tensor dotted with a vector).

The zero-element in is the mapping, that relates any vector to the zero-vector o.
This mapping is named as the zero-tensor and denoted with 0:

0·u=o .

The identical mapping, the identity tensor I, is a tensor that maps any arbitrary
vector u identically, i.e.,

I·u=u .

– Linear operations

T · (u + v) = T · u + T · v ,
T · (αu) = α(T · u) ,
(T + S) · u = T · u + S · u ,
(αT) · u = α(T · u) , αT = Tα

If a and b are elements of E 3 , then a ⊗ b is an element of E 3 ⊗ E 3 , which should


possess the following characteristics for every vector u contained in E 3 :
10
(a ⊗ b) · u = (b · u)a .

The dyadic product a ⊗ b is also called the simple tensor.

– Calculation rules

a ⊗ (b + c) = a ⊗ b + a ⊗ c ,
(αa) ⊗ b = a ⊗ (αb) = α(a ⊗ b)

11
4.2 Algebra in base systems

Given are the both vectors a = ai gi and b = bk gk (i, k = 1 . . . 3) in a covariant


base system. The dyadic product a ⊗ b can written in the following form

T = a ⊗ b = ai gi ⊗ bk gk = ai bk gi ⊗ gk
= a1 b1 g1 ⊗ g1 + a1 b2 g1 ⊗ g2 + a1 b3 g1 ⊗ g3
+ a2 b1 g2 ⊗ g1 + a2 b2 g2 ⊗ g2 + a2 b3 g2 ⊗ g3
+ a3 b1 g3 ⊗ g1 + a3 b2 g3 ⊗ g2 + a3 b3 g3 ⊗ g3 .

Here, the tensor T is referring to a covariant base system. The dyadic products
gi ⊗ gk of the base vectors is understood as the base of the tensor T of E 3 ⊗ E 3 . The
statement is valid for arbitrary tensors if the dyadic products gi ⊗ gk of the base
vectors are linear independent.

The base gi ⊗ gk is called linear dependent, when numbers α11 , α12 , . . . α33 can be
given that are not all zero, so that

α11 g1 ⊗ g1 + α12 g1 ⊗ g2 + α13 g1 ⊗ g3 + . . . + α33 g3 ⊗ g3 = 0

and

αik gi ⊗ gk = 0

respectively, is valid. Applying this to an arbitrary vector u it follows

( αik gi ⊗ gk ) u = αik ( gk · u) gi = ( u · gk ) αik gi = 0 · u = o .

Since u is an arbitrary vector, u · gk 6= 0 is valid for at least one case. Therefore,


one can conclude that

αik = 0 ,

if g1 , g2 and g3 are linear independent vectors.


12
The abovementioned requirement of linear dependence can be arrived at only with
αik = 0, which contradicts the requirement. With this contradiction it has been
proven that the base gi ⊗ gk is linear independent.

Taken into consideration an arbitrary tensor T in E 3 ⊗ E 3 . The set of ten tensors


{T , g1 ⊗ g1 , . . . , g3 ⊗ g3 } is necessarily linear dependent, so that ten numbers λ
and αik exist. Then it holds that,

λ T + αik gi ⊗ gk = 0 .

For λ 6= 0 it follows

αik
T=− gi ⊗ gk .
λ

Therefore, an arbitrary second order tensor can presented in the nine dimensional
space E 3 ⊗ E 3 as

T = tik gi ⊗ gk
= t11 g1 ⊗ g1 + t12 g1 ⊗ g2 + t13 g1 ⊗ g3 + . . . + t33 g3 ⊗ g3 ,

where

αik
tik = − .
λ

The quantities tik are called coefficients of the tensor components and gi ⊗ gk is the
base of the covariant tensors T. Analogical, representation forms of contravariant
as well as mixed-variant base systems can be derived. Thus, the second order tensor
T can be presented as

T = tik gi ⊗ gk = tik gi ⊗ gk = ti k gi ⊗ gk = ti k gi ⊗ gk .

The covariant and contravariant, as well as the mixed-variant coefficients, are not
independent of each other. By using the metric coefficients gir und gpk the mixed-
variant and the contravariant presentation forms of the covariant tensor T can derive:
13
T = tik gi ⊗ gk = tik gir gr ⊗ gk = trk gri gi ⊗ gk
= t i k g i ⊗ gk
= tik gkp gi ⊗ gp = tip gpk gi ⊗ gk
= t i k gi ⊗ g k
= tik gir gkp gr ⊗ gp = trp gri gpk gi ⊗ gk
= tik gi ⊗ gk .

A comparison of the four presentation forms of T leads directly to

ti k = trk gri , ti k = tip gpk , tik = trp gri gpk .

Remark:
1. Raising of an index:
A lower index can be raised with the multiplication of the covariant or the mixed-
variant coefficients of the tensor components with the contravariant metric coeffi-
cients.
2. Lowering of an index:
An upper index can be lowered with the multiplication of the contravariant or mixed-
variant coefficient of the tensor component with the covariant metric coefficients.

In order to give the true (physical) amount of the tensor coefficients, it is necessary
to decompose the tensors in a normed base:
∗ g1 g1 ∗ g1 g2 ∗ g3 g3
T = t11 ⊗ + t12 ⊗ + . . . + t33 ⊗
|g1 | |g1 | |g1 | |g2 | |g3 | |g3 |
∗ g1 g1 ∗ g1 g2 ∗ g3 g3
= t11 ⊗ + t12 1 ⊗ 2 + . . . + t33 3 ⊗ 3
|g1 | |g1 | |g | |g | |g | |g |
∗ g1 g1 ∗
2 g
1
g2 ∗
3 g
3
g3
= t11 ⊗ + t 1 ⊗ + . . . + t 3 ⊗
|g1 | |g1 | |g1 | |g2 | |g3 | |g3 |
∗ g1 g1 ∗
1 g1 g2 ∗
3 g3 g3
= t11 ⊗ +t 2 ⊗ + ... + t 3 ⊗ .
|g1 | |g1 | |g1 | |g2 | |g3 | |g3 |

The physical coefficients, marked with an asterisk *, can be calculated from


14
∗ √ √
tik = |gi | |gk | tik = gii gkk tik ,
∗ p p
tik = |gi | |gk | tik = gii gkk tik ,
∗ p √
ti k = |gi | |gk | ti k = gii gkk ti k ,
∗ √ p
ti k = |gi | |gk | ti k = gii gkk ti k , Σ/ i , Σ/ k .

15
4.3 Scalar product of second order tensors

The scalar product (double contraction) of the two simple tensors a ⊗ b and c ⊗ d
is defined as

(a ⊗ b) : (c ⊗ d) = (a · c)(b · d) .

Considering the calculation rule (a ⊗ b) · d = (b · d) a one obtains an alternative


representation of the scalar product of the two simple tensors:

(a ⊗ b) : (c ⊗ d) = c · (a ⊗ b) · d = (c · a)(b · d)
= (a · c)(b · d) .

The aforementioned relation can be assigned to the scalar product of an arbitrary


tensor T with the simple tensor a ⊗ b:

T : (a ⊗ b) = a · T · b .

Furthermore, it is possible to extend the Schwarz’s inequality to second-order ten-


sors. The value of the scalar product of two tensors T and S is smaller or equal to
the product of the norms of both the tensors, i.e.
√ √
|T : S| ≤ |T| |S| = T:T S:S .

The equality sign is only valid when the tensor T is a scalar multiple of the tensor
S.

Just as for the vectors, one can also make the following statements for the norms of
a tensor:

|α T| = |α| |T| , |T + S| ≤ |T| + |S| .

Finally, taking into consideration Schwarz’s inequality and vector algebra, the cosine
of the angle ϕ between the tensors T and S can defined as

16
T:S
cos ϕ = ,
|T| |S|

where the angle ϕ lies within the domain

0 ≤ ϕ ≤ π .

The usefulness of Schwarz’s inequality can be seen beside others in the plasticity
theory for the development of minimum/maximum statements for solids with rigid
ideal-plastic material behavior. The onset of plastic deformations of metallic mate-
rials is dependent on the von Mises’ yield condition TD : TD = κ2 , where TD is the
deviator of the Cauchy stress tensor as pointed out later and κ is a constant, which
can be experimentally determined.

– Calculation rules for arbitrary second-order tensors

T:S=S:T ,
T : (S + R) = T : S + T : R ,
αT : S = T : (αS) = α(T : S)

– Amount or norm of a tensor



|T| = T:T

– Examples

17
T : S = tik ( gi ⊗ gk ) : sop ( go ⊗ gp )
= tik sop ( gi · go ) ( gk · gp )
= tik sop gio gkp = tik sik = tok sok = . . . ,
T : S = tik ( gi ⊗ gk ) : sop ( go ⊗ gp )
= tik sop ( gi · go ) ( gk · gp )
= tik sop δ oi δkp = tik sik ,
T : I = tik ( gi ⊗ gk ) : gop ( go ⊗ gp )
= tik gop ( gi · go ) ( gk · gp )
= tik gop gio gkp = ti p gop gio = tio gio = ti i = t11 + t22 + t33

18
4.4 Tensor product of second order tensors

The tensor product (simple contraction) of two simple tensors a ⊗ b and c ⊗ d is


defined as

(a ⊗ b) · (c ⊗ d) = (b · c)(a ⊗ d) .

Taking into consideration the calculation rule (a ⊗ d) · u = (d · u) a and

[(a ⊗ b) · (c ⊗ d)] · u = (b · c)(a ⊗ d) · u


= (b · c)(d · u)a
= (a ⊗ b) · [(c ⊗ d) · u]
= (d · u)(a ⊗ b) · c
= (d · u)(b · c)a

one obtains the mapping

[(a ⊗ b) · (c ⊗ d)] · u = (a ⊗ b) · [(c ⊗ d) · u] .

This claim is valid for all vectors v in E 3 . The mapping rule is assigned to the
tensor product T S of two arbitrary second-order tensors:

(T · S) · u = T · (S · u) .

– Calculation rules for arbitrary second-order tensors

(T · S) · R = T · (S · R) ,
T · (S + R) = T · S + T · R , (R + S) · T = R · T + S · T ,
α(T · S) = (αT) · S = T · (αS) ,
I·T=T·I=T , 0·T=T·0=0

19
– Examples

T · S = [ tik ( gi ⊗ gk ) ] · [ sop ( go ⊗ gp ) ]
= tik sop ( gk · go ) ( gi ⊗ gp )
= tik sop gko ( gi ⊗ gp ) = tik skp ( gi ⊗ gp ) ,

T · S = [ tik ( gi ⊗ gk ) ] · [ sop ( go ⊗ gp ) ]
= tik sop ( gk · go ) ( gi ⊗ gp )
= tik sop δko ( gi ⊗ gp ) = tik skp ( gi ⊗ gp ) ,
T · I = [ tik ( gi ⊗ gk ) ] · ( go ⊗ go )
= tik ( gk · go ) ( gi ⊗ go )
= tik δko ( gi ⊗ go ) = tik ( gi ⊗ gk ) = T

4.4.1 Powers of tensors

The powers of the tensors T are second-order tensors again:

T0 = I , T1 = T , T2 = T · T , . . .

– Calculation rules

Tm · Tn = Tm+n = Tn · Tm , ( α T )m = αm Tm , (Tm )n = Tmn

20
4.5 Special tensors and operations

4.5.1 Inverse tensor

A tensor T can be inverted when the relationship v = T · u can be solved for the
vector u, i.e., the inverse of T exists and therefore u is uniquely determined:

v = T · u ⇒ T−1 · v = T−1 · T · u
⇒ T−1 · v = I · u
⇒ u = T−1 · v .

The tensor T−1 is called the inverse tensor of T.

– Calculation rules

T−1 · T = I , (T−1 )−1 = T , ( T · S )−1 = S−1 · T−1 ,


1 −1
( α T )−1 = T
α

4.5.2 Transposed tensor

The transposed tensor TT of a tensor T is defined as

u · T · v = v · TT · u

for every vector v and u. Considering

u · ( a ⊗ b ) · v = ( b · v ) ( u · a ) = v · ( b ⊗ a ) · u = v · ( a ⊗ b )T · u

one obtains the operation rule for the simple tensor a ⊗ b:

( a ⊗ b )T = ( b ⊗ a ) .

21
– Calculation rules

( T + S )T = TT + ST , ( α T )T = α TT , ( T · S )T = ST · TT

– Examples

T = tik ( gi ⊗ gk ) ⇒ TT = tki ( gi ⊗ gk )
⇒ TT = tik ( gk ⊗ gi ) ,
T = ti k ( gi ⊗ gk ) ⇒ TT = tki ( gi ⊗ gk )
⇒ TT = ti k ( gk ⊗ gi )

4.5.3 Symmetrical and skew-symmmetrical tensors

A tensor S is symmetric, when it corresponds to its transposed tensor, i.e.

S = ST .

For the symmetrical simple tensor a ⊗ b it follows that

a ⊗ b = ( a ⊗ b )T = b ⊗ a .

Considering

S = sik ( gi ⊗ gk ) = sik ( gi ⊗ gk ) = . . . ,
ST = ski ( gi ⊗ gk ) = ski ( gi ⊗ gk ) = . . .

one obtains the following relations concerning the coefficients of the symmetrical
tensor S = ST :

sik = ski , sik = ski , . . .

22
The identity tensor I is a simple example of a symmetrical tensor. With v = I · v,
the scalar product of the vectors u and v reads

u · v = u · I · v = v · IT · u .

This relations leads immediately to the symmetric property if the identity tensor

I = IT .

A skew-symmetric tensor is defined as

W = − WT .

For the skew-symmetric simple tensor c ⊗ d following relation is valid

c ⊗ d = − ( c ⊗ d )T = − d ⊗ c .

Taking into consideration

W = wik ( gi ⊗ gk ) = wik ( gi ⊗ gk ) = . . . ,
WT = wki ( gi ⊗ gk ) = wki ( gi ⊗ gk ) = . . .

one obtains the following characteristics of the coefficients of the components of the
skew-symmetrical general tensor W = −WT :

wik = − wki , wik = − wki , . . . .

4.5.4 Orthogonal tensor

The orthogonal tensor Q plays an important role in rigid body mechanics as well
in the mechanics of deformable bodies interrelated to the rotational motion. The
orthogonal tensor is characterized by the properties:

23
Q−1 = QT ⇒ Q · QT = QT · Q = I , det Q = 1 .

The determinate (third invariant) of the tensor Q is given by

1 1 1
det Q = ( Q : I )3 − ( Q : I ) ( QT : Q ) + [ ( QT · QT ) : Q ] .
6 2 3

The linear mapping

Q·u=v

is caused with the change of the direction of the vector u, the norm of the vector is
received, i.e.
√ √ p
|v| = v·v = Q·u·Q·u= u · QT · Q · u
√ √
= u · I · u = u · u = |u| .

24
4.5.5 Trace of the tensor

The trace of the tensor is a special product, denoted by tr (tr = trace):

tr T = T : I .

The trace of the simple tensor a ⊗ b is equal with the scalar product of the both
vectors a und b:

tr( a ⊗ b ) = ( a ⊗ b ) : I = ( ai gi ⊗ bk gk ) : ( go ⊗ go )
= ai bk gio δko = ai bk gik = ai gi · bk gk
=a·b .

– Calculation rules

tr TT = TT : I = T : I , tr( T · S ) = tr( S · T ) ,
tr( T · ST ) = T : S = tr( S · TT ) = TT : ST ,
tr( T · S · R ) = tr( R · T · S ) = tr( S · R · T )
⇒ ( T · S ) : RT = ( R · T ) : ST = ( S · R ) : TT

25
4.6 Decomposition of the tensor

It is often advantageous to split a tensor into two partial tensors, which show special,
characteristic features. This is naturally possible in several ways. In this context,
only a special additive and a multiplicative decomposition will be discussed.

4.6.1 Additive decomposition

Each arbitrary tensor T can be additively decomposed into symmetrical tensor


Tsymm and a skew-symmetrical tensor Tskewsymm :

T = Tsymm + Tskewsymm ,

where

1
Tsymm = ( T + TT ) = (Tsymm )T ,
2
1
Tskewsymm = ( T − TT ) = − (Tskewsymm )T .
2

In view of its importance for mechanics, the possibilities of the decomposition of an


arbitrary tensor T into a spherical tensor (TK ) and a deviator (TD ) is shown:

T = TD + TK ,

where
1 1
TK = ( T : I ) I , TD = T − TK = T − ( T : I ) I .
3 3
It is important to note, that the trace of the deviatorical part of the tensor T is
zero:
1
tr TD = TD : I = T : I − (T : I)(I : I) = T : I − T : I = 0 .
3

4.6.2 Multiplicative decomposition (polar decomposition)

26
Every non-singular, second-order tensor F can be explicitly decomposed into positive-
definite symmetrical tensors U = UT and V = VT , respectively and into an orthog-
onal tensor R:
F=R·U , F=V·R .
The tensor products

FT · F = ( R · U )T R · U = UT · RT · R · U = UT · I · U = UT · U = U · U ,
F · FT = V · R · ( V · R )T = V · R · RT · VT = V · I · VT = V · VT = V · V
do not include rotation parts, because R · RT = RT · R = I.

Remark:
In the framework of continuum mechanics, the tensors F, U, V, FT · F and F · FT
are denoted as follows:
F – deformation gradient
U – right stretch tensor
V – left stretch tensor
T
C = F · F – right Cauchy-Green deformation tensor
B = F · FT – left Cauchy-Green deformation tensor

4.6.3 Invariants of the tensor

The eigenvalue problem


T · v = γ v ⇒ (T − γ I) · v = o
leads to the so-called characteristic equation of T
det( T − γ I ) = φ(γ) = − γ 3 + γ 2 IT − γ IIT + IIIT = 0
for the determination of the eigenvalues γ1 , γ2 and γ3 . The scalar quantities IT ,
IIT and IIIT are the first, second and third invariant of the tensor T. The three
invariants of T are given by

IT = T : I = tr T ,
1
IIT = [ ( T : I )2 − TT : T ] ,
2
1 1 1
IIIT = ( T : I )3 − ( T : I ) ( TT : T ) + [ ( TT · TT ) : T ] = det T .
6 2 3

27
4.7 Change of the base

Given are two arbitrary bases gi and ḡi in the E 3 . The corresponding reciprocal
bases are gk und ḡk . Each base vector ḡi can be expressed through the base vector
gi with the help of a non-singular transformation:

ḡi = I · ḡi = ( gk ⊗ gk ) · ḡi = ( gk · ḡi ) gk


= ( gk · ḡo δ oi ) gk
= ( gk · ḡo ) ( gk ⊗ go ) · gi = A · gi ,

where

A = ( gk · ḡo ) ( gk ⊗ go ) .

From the above mentioned transformations, it immediately follows that the trans-
formation can express explicitly as

ḡi = ( ḡi · gk ) gk , ḡi = ( ḡi · gk ) gk ,


gi = ( gi · ḡk ) ḡk , gi = ( gi · ḡk ) ḡk .

– Transformations of vector coefficients

v = vi gi = v̄i ḡi ⇒ vi gi · gk = v̄i ḡi · gk


⇒ vi δ ki = ( ḡi · gk ) v̄i
⇒ vk = ( gk · ḡi ) v̄i

and

v = vi gi = v̄i ḡi ⇒ vi gi · ḡk = v̄i ḡi · ḡk


⇒ v̄i δ ki = ( ḡi · ḡk ) vi
⇒ v̄k = ( ḡk · gi ) vi ,

respectively.

28
– Transformations of tensor coefficients

T = tik gi ⊗ gk = t̄ik ḡi ⊗ ḡk


⇒ tik ( gi ⊗ gk ) : ( go ⊗ gp ) = t̄ik ( ḡi ⊗ ḡk ) : ( go ⊗ gp )
⇒ tik δ oi δkp = ( ḡi · go ) ( ḡk · gp ) t̄ik
⇒ top = ( go · ḡi ) ( gp · ḡk ) t̄ik

and

T = tik gi ⊗ gk = t̄ik ḡi ⊗ ḡk


⇒ tik ( gi ⊗ gk ) : ( ḡo ⊗ ḡp ) = t̄ik ( ḡi ⊗ ḡk ) : ( ḡo ⊗ ḡp )
⇒ t̄ik δ oi δkp = ( gi · ḡo ) ( gk · ḡp ) tik
⇒ t̄op = ( ḡo · gi ) ( ḡp · gk ) tik ,

respectively.

29
4.8 Higher-order tensors

In the tensor analysis as well as by its application in continuum mechanics, the


introduction of higher-order tensors is very necessary.

4.8.1 Introduction of higher-order tensors

The order of the tensors is marked by a raised natural number n ≥ 1. The first- and
second-order tensors will be marked as usual with small and capital boldface letters
without superscript numbers.

The following relationships are valid for the nth -order tensors as well as tensors of
the mth -order (n, m = 1, 2, ... ):

n (i) m m n (i) m n (i) m n n (i) m n (i) m n (i) m


L : (T + S) = L : T + L : S , (L + R) : T = L : T + R : T ,
n (i) m n (i) m n (i) m n n
L : (αT) = (αL) : T = α(L : T) , αL = Lα .

For zero tensors the considerations mentioned before are valid. In addition to iden-
tical mappings additional fundamental tensors also can be defined, this will be dis-
cussed in a separate chapter.

In order to transfer the formalism for the second-order tensors to the higher-order
tensors, the simple tensors nth - and mth -order are introduced:

n 1 2 n m 1 2 m
L = a ⊗ a ⊗ ... ⊗ a , T = b ⊗ b ⊗ ... ⊗ b ,

1 n 1 m
where a through a and b through b are arbitrary vectors in E 3 .

The connection between the simple tensors emerge as a result of tensor products, as
well as tensor products in association with scalar products of the vectors contained
in each of the tensors, which are normally referred to as diminishing products.
Therefore, the possibility exists, depending on the order of the simple tensors, to
produce per definition tensors having various orders:

30
Tensor and diminishing products:
n (0) m n m
(0) ( L : T )n+m = ( L ⊗ T )n+m
1 2 n 1 2 m
= a ⊗ a ⊗ ... ⊗ a ⊗ b ⊗ b ⊗ ... ⊗ b ,
n (1) m n m
n+m−2
(1) ( L : T ) = ( L · T )n+m−2
n 1 1 2 n−1 2 m
= (a · b)a ⊗ a ⊗ ... ⊗ a ⊗ b ⊗ ... ⊗ b ,
n (2) m n m
(2) ( L : T )n+m−4 = ( L : T )n+m−4
n−1 1 n 2 1 2 n−2 3 m
= ( a · b)(a · b)a ⊗ a ⊗ ... ⊗ a ⊗ b ⊗ ... ⊗ b ,
...
n > m
n (m) m n (m) m
(m) ( L : T )n+m−2m = ( L : T )n−m
n−m+1 1 n m 1 2 n−m
=( a · b) ... (a · b)a ⊗ a ⊗ ... ⊗ a ,
n < m
n (n) m n (n) m
(n) ( L : T )m+n−2n = ( L : T )m−n
1 1 n n n+1 m
= (a · b) ... (a · b) b ⊗ ... ⊗ b .

The tensor products, according to (0), and the diminishing products, according to
(1) through (m) or (n), are thus determined between arbitrary tensors. For the
special case m = n, one comes come to the scalar product of two nth -order tensors:

1 2 n (n) 1 2 n 1 1 2 2 n n
(a ⊗ a ⊗ ... ⊗ a) : (b ⊗ b ⊗ ... ⊗ b) = (a · b)(a · b) ... (a · b) .

One sees, that herein the scalar product of second-order tensors is contained. Moving
on to arbitrary nth -order tensors, it is easily to confirm the validity of the commu-
tative rule:
n (n) n n (n) n
L : T=T : L .
31
Furthermore, the magnitude (norm) of an nth -order tensor is given by
q
n n (n) n
|L| = L : L.

32
4.8.2 Special operations and tensors

Considering the definition of the products of higher-order tensors, one is able to


state the associative rule in a general form

n (i) m (k) p n (i) m (k) p


L : (T : R) = (L : T) : R ,

where the order of the product is open. This has to be determined for the individual
case. In order to explain this, an example will be considered. For n = 2, m = 3 and
p = 2, the product in the brackets on the left-hand side is a vector:

2 (1) 3 (2) 2 3 2 (1) 3 (2) 2 3


L : (T : R) = L · (T : R) = (L : T) : R = (L · T) : R .

For simple tensors

2 1 2 3 1 2 3 2 1 2
L=a⊗a , T=b ⊗b⊗b , R=c⊗c

the left-hand and right-hand sides yield the same results:

2 (1) 3 (2) 2 1 2 2 1 3 2 1
L : (T : R) = (a ⊗ a) · [(b · c)(b · c)b]
2 1 3 2 2 1 1
= (b · c)(b · c)(a · b)a ,
2 (1) 3 (2) 2 1 1 2 3 (2) 1 2
( L : T ) : R = (a · b ) ( a ⊗ b ⊗ b ) : ( c ⊗ c )
2 1 2 1 3 2 1
= (a · b)(b · c)(b · c)a .

The formulation of higher-order simple tensors is also possible by using higher-order


tensors:

3 3 4 4 3
L=T⊗v , L=u⊗S , L=T⊗S , L=v⊗L .

33
– Calculation rules

(T ⊗ u) · v = (u · v)T , (u ⊗ T) : R = (T : R)u ,
(u ⊗ v ⊗ w) : I = (v · w)u , (T ⊗ u) : (v ⊗ w) = (u · w)T · v ,
(T ⊗ v) : R = T · R · v , (T ⊗ v) : I = T · v , (T ⊗ R) : S = (R : S)T ,
( T ⊗ R ) : I = (tr R) T , ( T ⊗ R ) · v = T ⊗ R · v .

– Transposition of higher-order tensors

n ik
1 i k n ik

L T = ( a ⊗ · · · ⊗ a ⊗ · · · ⊗ a ⊗ · · · ⊗ a )T
1 k i n
= a ⊗ ... ⊗ a ⊗ ... ⊗ a ⊗ ··· ⊗ a .

Further transpositions can be produced with the combination of individual ex-


changes.

Considering the special transpositions for the simple third-order tensors one gets,
as a consequence of an arbitrary tensor, the following relations:

1 2 3 13 3 2 1 3 3 13

( a ⊗ a ⊗ a ) = a ⊗ a ⊗ a =⇒ v · ( L : T ) = T : ( L T · v ) ,
T T

1 2 3 23
1 3 2 3 3 23

( a ⊗ a ⊗ a )T = a ⊗ a ⊗ a =⇒ ( L · v ) · u = ( L T · u ) · v .

2n
– Special transposition of even-order tensors L

n (n) 2n (n) n n (n) 2n (n) n


T
T : (L : R) = R : (L : T) .

n n
– Transposition of simple second-order tensors, composed with the tensors S and M

n n n n
( S ⊗ M )T = M ⊗ S .

34
4.9 Cross product

In mechanics, cross products are very important for problems describing rotational
motions. In this context, the balance of moment of momentum and the relationships
derived from them are mentioned. Also, in the general fundamental equations of
shell theory and continuum mechanics, one comes across these products combining
vectors with vectors, vectors with tensors as well as tensors with tensors.

4.9.1 Cross product of vectors

Two non-parallel vectors u and v with the enclosed angle ϕ (0 ≤ ϕ ≤ π) shape a


plane when they are plotted at one point. The area spanned by the vectors u and
v is a parallelogram:

|u × v| = |u| |v| sin ϕ .

Now, a new vector

u × v = |u| |v| sin ϕ n

is introduced, pointing in the direction of the unit normal vector n of the plane
(perpendicular to the plane).

Figure 4.1: Cross product of vectors

35
The association u × v is called as the cross or vector product of the vectors u and v.
The vectors u, v and n are defined as a positive mathematical right-handed system.
Thus, the association u × v is clearly assigned in the direction n.
– Calculation rules

(− u) × v = − u × v , u × v = − v × u ,
u × v = o , u || v ,
α(u × v) = (αu) × v = u × (αv) ,
u × (v + w) = u × v + u × w .

Besides this, the calculation rules gives

(u + v) × w = −w × (u + v) = −w × u − w × v
=u×w+v×w .

In the following the characteristics of scalar triple products (box products) w · (u ×


v) is considered. Geometrically, its value can be interpreted as the volume of a
parallelepiped spanned through the vectors u, v and w. The triple scalar product
simply can be written as

w · ( u × v ) = [w u v] .

Writing it in a concise form u × v = z, then w · z = |w||z| cos(w, z) is the volume of


the parallelepiped, i.e., u×v is a vector, whose value is equal to the basic surface area
of the parallelepiped, and |w| cos(w, z) is its height. The volume can be calculated
if one takes the parallelogram as the basic surface area, spanned by the vectors v
and w, and multiply scalarly the vector w × u with v. The same volume can also
be derived from the scalar product of the product vector v × w and the vector u.
For the cyclical exchange of the vectors it holds that

w · ( u × v) = v · ( w × u ) = u · ( v × w )

or

[w u v] = [v w u] = [u v w] ,
36
Figure 4.2: Parallelepiped

i.e., the parallelepiped product does not change by the cyclical exchange of the
vectors.

The triple scalar product becomes zero, when a linear dependence exists. One
makes use of this fact, in order to find an explicit form for the vector triple product
u × (v × w). The result is a vector (provided the three vectors u, v and w represent
distances) lying in the area spanned by the vectors v and w, or parallel to it. One
derives with arbitrary real values α, β and µ that

−µw · (v × αv) + µv · (w × β w)
= −µαv · (w × v) + µβ w · (v × w)
= µαv · (v × w) + µβ w · (v × w)
= µ(αv + β w) · (v × w) = 0 ,
u · [(v × w) × (v × w)] = (v × w) · [u × (v × w)]
= u × (v × w) · (v × w) = 0

=⇒

µ(αv + β w) · (v × w) = 0 , u × (v × w) · (v × w) = 0 .

37
Comparing both equations, one concludes that

u × (v × w) = µ(αv + β w) .

In addition, one obtains

(v × w) · (u × u) = u · [(v × w) × u]
= u · [u × (v × w)]
= µαu · v + µβ u · w = 0

=⇒

u · [u × (v × w)] = µαu · v + µβ u · w = 0 .

For arbitrary real numbers µ, α and β the above mentioned equation is fulfilled if

µα = u · w , µβ = −u · v

and one obtains the following relation:

u × (v × w) = (u · w)v − (u · v)w .

The vector triple product is not associative. However, the Lagrange identity is valid

u × (v × w) + v × (w × u) + w × (u × v) =
= (u · w)v − (u · v)w + (v · u)w − (v · w)u
+ (w · v)u − (w · u)v = o

=⇒

u × (v × w) + v × (w × u) + w × (u × v) = o

and
38
u × (v × w) = −v × (w × u) − w × (u × v) ,

respectively.

An important calculation rule is derived from the Lagrange identity by scalar mul-
tiplication with the vector z:

z · [u × (v × w)] = −z · [v × (w × u)] − z · [w × (u × v)]


= −z · [(v · u)w − (v · w)u]
−z · [(w · v)u − (w · u)v]
= −z · (v · u)w + z · (w · u)v
= ( w · u ) (v · z ) − ( v · u ) ( w · z ) ,
z · [ u × ( v × w ) ] = ( v × w) · ( z × u )

=⇒

( v × w) · ( z × u ) = ( w · u ) ( z · v ) − ( v · u ) ( z · w ) .

In addition, the quadruple vector product is mentioned:

(u × v) × (w × z) = [(u × v) · z]w − [(u × v) · w]z


= [z · (u × v)]w − [w · (u × v)]z
= [v · (z × u)]w − [v · (w × u)]z
= [u · (v × z)]w − [u · (v × w)]z ,
(u × v) × (w × z) = −(w × z) × (u × v)
= (z × w) × (u × v)
= [(z × w) · v]u − [(z × w) · u]v
= −[(w × z) · v]u + [(w × z) · u]v
= [(w × z) · u]v − [(w × z) · v]u
= [u · (w × z)]v − [v · (w × z)]u
39
=⇒

( u × v ) × ( w × z ) = [u v z] w − [u v w] z = [u w z] v − [v w z] u .

The definition of the cross product and the calculation rules have been introduced
independent of the base. In the following, the two vectors u and v are represented
in an arbitrary oriented base system g1 , g2 and g3 :

u = ui gi , v = vk gk .

The cross product of u and v is represented as

u × v = ui vk gi × gk ,

and the problem being the evaluation of the cross product of the base vectors. In
general, the cross product gi × gk can be expressed as
p
gii gkk ( 1 − cos2 ϕik )
|gi × gk | = |gi | |gk | sin ϕik =
s
( gi · gk )2
= gii gkk ( 1 − )
|gi |2 |gk |2
s
(gik )2
= gii gkk ( 1 − )
gii gkk
p
= gii gkk − (gik )2

=⇒
p
gii gkk ( 1 − cos2 ϕik ) ,
|gi × gk | = |gi | |gk | sin ϕik =
p
= gii gkk − (gik )2

The angle between gi and gk is indicated as ϕik . The root represents a subdetermi-
nant Uoo of the coefficient matrix:

Uoo = gii gkk − (gik )2 , o 6= i , k .


40
Thus, the following equation is valid:

|gi × gk | = Uoo , o 6= i , k .

As shown before, the cross product of gi and gk is determined through



gi × gk = Uoo n , o 6= i , k ,

where n is the unit-normal vector of a surface spanned by the base vectors gi and
gk in a right-handed base system. The unit-normal vector n expresses itself directly
through the contravariant base vector go because this is perpendicular on gi and gk
and consequently is at right angles to the spanned surface gi and gk :
s
o
g Uoo o
n = √ oo , gi × gk = g , o 6= i , k .
g goo

With the relationship

Uoo = det[gik ] goo , o 6= i , k

from the matrix theory, one gets


gi × gk = g go , g = det[gik ] , o 6= i , k .

Thus, the following cross products can be calculated:


√ √
g1 × g1 = o , g1 × g2 = g g3 , g1 × g3 = − g g2 ,
√ √
g2 × g1 = − g g3 , g2 × g2 = o , g2 × g3 = g g1 ,
√ √
g3 × g1 = g g2 , g3 × g2 = − g g1 , g3 × g3 = o .

Introducing the permutation symbol

 0 if two indices are equal,


ijk
eijk = e = + 1 for e123 , e231 , e312 ,
− 1 for e213 , e132 , e321 ,
41
the covariant and contravariant cross products can be written as
√ 1
gi × gj = g eijk gk , gi × gj = √ eijk gk .
g

The scalar triple product (box product) of the base vectors and the dual base vectors
reads

[gi gj gk ] = ( gi × gj ) · gk = g eijk ,
1
[gi gj gk ] = ( gi × gj ) · gk = √ eijk .
g
The cross product of the two vectors u and v is determined by
√ 1
u × v = ui vj g eijk gk = ui vj √ eijk gk .
g
The scalar triple product of the vectors u, v and w reads
√ 1
( u × v ) · w = ui vj wk g eijk = ui vj wk √ eijk .
g
The above-mentioned relationships simplify in an orthonormed base, i.e., g takes
the value one and the covariant base vectors coincide with the contravariant base
vectors:

u × v = ui vj eijk ek
= ( u2 v3 − u3 v2 ) e1 + ( u3 v1 − u1 v3 ) e2 + ( u1 v2 − u2 v1 ) e3
 
e1 e2 e3
= det  u1 u2 u3  ,
v1 v2 v3
and

( u × v ) · w = ui vj wk eijk
= u1 ( v2 w3 − v3 w2 ) + u2 ( v3 w1 − v1 w3 ) + u3 ( v1 w2 − v2 w1 )
 
u1 u2 u3
= det  v1 v2 v3  .
w1 w2 w3
42
As discussed before, the cross products of the base vectors gi and gj as well as gi
and gj can be written as

gi × gj = ēijk gk = ērsk ( gr · gi ) ( gs · gj ) gk
= ērsk ( gk ⊗ gr ⊗ gs ) ( gi ⊗ gj ) ,
gi × gj = ēijk gk = ērsk ( gr · gi ) ( gs · gj ) gk
= ērsk ( gk ⊗ gr ⊗ gs ) ( gi ⊗ gj ) ,

where the abbreviations

√ 1
ēijk = g eijk , ēijk = √ eijk
g

have been used. Consequently, the cross products of the base vectors can be repre-
sented as linear mappings of the simple tensors gi ⊗ gj and gi ⊗ gj with a third-order
3
tensor E:

3 3
gi × gj = E : ( gi ⊗ gj ) , gi × gj = E : ( gi ⊗ gj ) .

Considering that a cyclic interchange of the indices is allowed in ērsk and ērsk , the
3
third-order tensor E can be written as

3
E = ērsk gr ⊗ gs ⊗ gk = [ ( gr × gs ) · gk ] gr ⊗ gs ⊗ gk
= ērsk gr ⊗ gs ⊗ gk = [ ( gr × gs ) · gk ] gr ⊗ gs ⊗ gk
= [ ( gr × gs ) · gk ] gr ⊗ gs ⊗ gk .
3
The third-order tensor E is exclusively determined by metric quantities. For this
reason, the tensor is called fundamental tensor.

The above mentioned calculation rules, valid for the base vectors, can be easily
applied to other vectors:

3
u × v = E : (u ⊗ v ) .
43
A
At last, a statement valid for a skew-symmetric tensor T is derived, stating that, for
every skew-symmetric tensor, a so-called axial vector, is associated with it. In order
to prove this, a simple tensor a⊗b is taken into consideration, whose skew-symmetric
part is formed with

A 1
T= (a ⊗ b − b ⊗ a) .
2
A
Applying T on an arbitrary vector u, one gets

A 1
T·u= (a ⊗ b − b ⊗ a) · u
2
1 1 1
= [(u · b)a − (u · a)b] = u × (a × b) = (b × a) × u .
2 2 2
Thus, the following relation is valid

A A A 1
T·u= t×u , t = (b × a) ,
2
A
where the vector t is denoted as the axial vector. The considerations carried out for
simple tensors can be easily applied to any arbitrary tensor T:

A 1 3 T 1 3 1
t = E : (a ⊗ b) = E : (b ⊗ a) = (b × a)
2 2 2

=⇒

A 1 3
t = E : TT .
2

The relation directly shows that the axial vector is a zero vector of the symmetrical
tensor.

44
4.9.2 Cross tensor product of vector and tensor

In the cross product of vectors, one of the vectors can be interpreted as as a linear
mapping with a second-order tensor, e.g. u × (T · v). In view of applications in
mechanics, it is useful to introduce a cross tensor product of the vector u with the
tensor T, and to represent the cross product u × (T · v) as a linear mapping. Taking
the arbitrary vectors u and v as well as the arbitrary tensors T and S, it is stated:

(u × T) · v = u × (T · v) ,
(T × u) · v = (T · v) × u ,
T : (S × v) = −S : (T × v) .

The definitions likewise contain

3
u×T = E : (u ⊗ T) ,
3
− T×u = E : (u ⊗ T) ,
u × (a ⊗ b) = (u × a) ⊗ b ,
(a ⊗ b) × u = (a × u) ⊗ b .

The expressions u × T and T × u are second-order tensors, having the following


properties:

u × T = −T × u ,
u × (T + S) = u × T + u × S ,
(u + v) × T = u × T + v × T ,
α ( u × T) = ( α u ) × T = u × ( α T ) ,
(v × S) : T = −(v × T) : S .

Considering the definition of the transposed tensor

w · ( u × I ) · v = v · ( u × I )T · w

45
and the reformulation of the left-hand side of the this relation,

w · (u × I) · v = w · (u × I · v) = w · (u × v)
= −w · (v × u) = −u · (w × v) = −v · (u × w)
= −v · (u × I · w) = −v · (u × I) · w

=⇒

w · (u × I) · v = −v · (u × I) · w ,

it holds that the tensor u × I is a skew-symmetric tensor, i.e.,

( u × I ) = − ( u × I )T .

In the skew-symmetric tensor u×I, u is the axial vector. Thus, the following relation
is valid:

A A
(u × I) · v = u × v = t × v , t = u .

In conclusion of this section, the cross tensor products of vectors and tensors shall
be represented in a base system g1 , g2 and g3 . The aforementioned calculation rules
in association with the calculation rules of higher-order tensors, gives

3
u × T = E : (u ⊗ T)
u × T = ēikj ur tsl ( gi ⊗ gk ⊗ gj ) : ( gr ⊗ gs ⊗ gl ) ]
= ēikj ur tsl δrk δsj gi ⊗ gl
= ēikj uk tjl gi ⊗ gl = ēikj uk tjl gir gr ⊗ gl .

Corresponding representations in the dual or mixed-variant bases can be obtained


without difficulty.

The advantage of the definition of the cross tensor product of a vector and a tensor
will be revealed with an example from continuum mechanics, namely the balance
46
equation of moment of momentum. The moment of the external stress vector t with
respect to the fixed point 0 reads
R R
m(0) = x × t da = x × σ · n da ,
∂B ∂B

where the Cauchy theorem

t = t(x, n, t) = σ(x, t) · n

has been used.

47
4.9.3 Cross tensor product of tensors

The cross tensor product of two arbitrary second-order tensors T and S is defined
with the following mapping rule

( T # S ) · ( u1 × u2 ) = T · u1 × S · u2 − T · u2 × S · u1 .

The product T# S is explicitly a second-order tensor, with u1 and u2 being arbitrary


vectors.

– Calculation rules

T#S = S#T ,
R#(T + S) = R#T + R#S ,
α(R#T) = (αR)#T = R#(αT) ,
( T # S )T = TT # ST ,
I#I = 2I ,
(T#S) · (R#U) = (T · R#S · U) + (T · U#S · R) ,
(a ⊗ b)#(c ⊗ d) = (a × c) ⊗ (b × d) .

– Proofs of some calculation rules

( T # S ) · ( u1 × u2 ) = T · u1 × S · u2 − T · u2 × S · u1 ,
( S # T ) · ( u1 × u2 ) = S · u1 × T · u2 − S · u2 × T · u1
= − T · u2 × S · u1 + T · u1 × S · u2
= T · u1 × S · u2 − T · u2 × S · u1

=⇒

(T#S) = (S#T) .

48
( I # I ) · ( u1 × u2 ) = I · u1 × I · u2 − I · u2 × I · u1
= I · u1 × I · u2 + I · u1 × I · u2
= 2 I · u1 × I · u2 = 2 I · ( u1 × u2 )

=⇒

(I#I) = 2I .

[ ( a ⊗ b ) # ( c ⊗ d ) ] · ( u1 × u2 ) =
= ( a ⊗ b ) · u1 × ( c ⊗ d ) · u2 − ( a ⊗ b ) · u2 × ( c ⊗ d ) · u1
= ( b · u1 ) ( d · u2 ) a × c − ( b · u2 ) ( d · u1 ) a × c
= [ ( b · u1 ) ( d · u2 ) − ( b · u2 ) ( d · u1 ) ] a × c
= [ ( b × d ) · ( u1 × u2 ) ] a × c
= [ ( a × c ) ⊗ ( b × d ) ] · ( u1 × u2 )

=⇒

(a ⊗ b)#(c ⊗ d) = (a × c) ⊗ (b × d) .

– Representation of the cross tensor product of tensors in a base system

T # S = tik smn ( gi ⊗ gk ) # ( gm ⊗ gn )
= tik smn ( gi × gm ) ⊗ ( gk × gn )
= tik smn ( ērim gr ) ⊗ ( ēskn gs )
= tik smn ērim ēskn gr ⊗ gs
= g tik smn erim eskn gr ⊗ gs

The scalar triple product of arbitrary tensors T, S and R satisfies the equation
49
1
(T#S) : R = eijk ( T · ui × S · uj ) · R · uk .
[ u1 u2 u3 ]

Furthermore, it can be shown that the scalar triple product possesses the following
properties:

R : (T#S) = T : (R#S) = S : (T#R) .

– Proof of scalar triple product of tensors


The so-called volume tensor is a spherical tensor and can represented as

V̄ = u1 ⊗ ( u2 × u3 ) + u2 ⊗ ( u3 × u1 ) + u3 ⊗ ( u1 × u2 )
= [u1 u2 u3 ] I .

Considering

( T # S ) : R = ( T # S ) · I : R = RT · ( T # S ) : I = ( T # S )T · R : I

one obtains

[u1 u2 u3 ] ( T # S ) : R = ( T # S )T · R : [u1 u2 u3 ] I .

With the calculation rule

( t1 ⊗ t2 )T · ( s1 ⊗ s2 ) : ( u ⊗ v ) = ( t2 ⊗ t1 ) · ( s1 ⊗ s2 ) : ( u ⊗ v )
= ( t1 · s1 ) ( t2 · u ) ( s2 · v )
= ( t1 ⊗ t2 ) · u · ( s1 ⊗ s2 ) · v

=⇒

TT · S : ( u ⊗ v ) = T · u · S · v

one can write


50
[u1 u2 u3 ] ( T # S ) : R = ( T # S ) · ( u1 × u2 ) · R · u3
+ ( T # S ) · ( u2 × u3 ) · R · u1
+ ( T # S ) · ( u3 × u1 ) · R · u2
= [ ( T · u1 × S · u2 ) − ( T · u2 × S · u1 ) ] · R · u3
+ [ ( T · u2 × S · u3 ) − ( T · u3 × S · u2 ) ] · R · u1
+ [ ( T · u3 × S · u1 ) − ( T · u1 × S · u3 ) ] · R · u2

[u1 u2 u3 ] ( T # S ) : R = e123 ( T · u1 × S · u2 ) · R · u3
+ e312 ( T · u3 × S · u1 ) · R · u2
+ e231 ( T · u2 × S · u3 ) · R · u1
− e321 ( T · u3 × S · u2 ) · R · u1
− e132 ( T · u1 × S · u3 ) ] · R · u2
− e213 ( T · u2 × S · u1 ) ] · R · u3
= eijk ( T # S ) · ( ui × uj ) · R · uk

=⇒

[u1 u2 u3 ] ( T # S ) : R = eijk ( T # S ) · ( ui × uj ) · R · uk .

The tensor cross product as well as the scalar triple tensor product can be repre-
sented with the use of scalar tensor products. They are individually represented
as:

T # I = ( T : I ) I − TT ,
T # S = ( T : I ) ( S : I ) I − ( TT : S ) I − ( T : I ) ST
− ( S : I ) TT + TT · ST + ST · TT ,
T # T = [ ( T : I )2 − TT : T ] I − 2 ( T : I ) TT + 2 TT · TT

51
and

( T # S) : R = ( T : I ) ( S : I ) ( R : I ) − ( T : I ) ( ST : R )
− ( S : I ) ( TT : R ) − ( R : I )( TT : S )
+ ( TT · ST ) : R + ( ST · TT ) : R ,
(T#S) : (R#U) = (T : R)(S : U) + (T : U)(S : R)
− ( T · RT · S · UT ) : I − ( T · UT · S · RT ) : I .
With respect to the presentation of the tensor cross product T#S of the tensors
T and S in base systems, after some algebraic transformations one comes to the
expression
6 (4)
T#S = E : (T ⊗ S)
(4)
= ( ērsikmn gr ⊗ gs ⊗ gi ⊗ gk ⊗ gm ⊗ gn ) :
( top suv go ⊗ gp ⊗ gu ⊗ gv )
= ērsikmn top suv δoi δpk δum δvn gr ⊗ gs
= ērsikmn tik smn gr ⊗ gs
= ēimr ēkns tik smn gr ⊗ gs
=⇒
6 (4)
T # S = E : ( T ⊗ S ) = ēimr ēkns tik smn gr ⊗ gs ,

where
6
E = ērsikmn gr ⊗ gs ⊗ gi ⊗ gk ⊗ gm ⊗ gn
= ēimr ēkns gr ⊗ gs ⊗ gi ⊗ gk ⊗ gm ⊗ gn
= ērim ēskn gr ⊗ gs ⊗ gi ⊗ gk ⊗ gm ⊗ gn
is a sixth-order tensor, which is solely defined by the metric and represents a fun-
damental tensor. Similar representations are possible in a covariant base or mixed-
variant base.
52
4.9.4 Cross vector product of tensors

The vector product of two tensors T and S is introduced as follows:

v · (T × S) = −T : (v × S) ,

where T × S is an unique vector with the following characteristics:

T×S = −S × T ,
T × (S + R) = T × S + T × R ,
α(T × S) = (αT) × S = T × (αS) .

In order to calculate the vector product of tensors in a base system, the aforemen-
3
tioned definition (N2) the following mapping rules with the fundamental tensor E:

3
T×S = E : ( T · ST ) ,
(a ⊗ b) × (c ⊗ d) = (b · d)(a × c) .

Considering the calculations rules for higher-order tensors, the following relation can
be derived:

3 3 3
v · ( T × S) = − T : [ E · ( v ⊗ S ) ] = T : [ ( E · v ) · S ] = ( E · v ) : T · ST
3 (3) 3
= E : [ T · ST ⊗ v ] = v · E : ( T · ST )

Besides this, the axial vector, related to a tensor T, can be expressed through the
A
vector product of the identity tensor I with the tensor T or T, respectively:

A 1 1 A
T = (I × T) = (I × T) ,
2 2

where it has considered the axial vector of a symmetrical tensor as a zero vector.

53
4.9.5 Special tensors and operations

Certain problems of the tensor calculation, as for example, the inversion of a tensor,
the introduction of some special tensors and operations are necessary, which can be
formulated with the help of cross tensor products. This algebraic operation also gives
good results for treating the eigenvalue problem and the formulation of determinants.
In particular, it is shown that the cross tensor product allows the explicit description
of the tensor value determinants. Not only do the calculation rules simplify the cross
algebra, but also make the description of the complex mechanical relationships by
the rotation of rigid bodies easy.

a) The adjunct tensor and the determinants

Taking account the relationship

1 1
( T # T ) · ( u1 × u2 ) = ( T · u1 × T · u2 − T · u2 × T · u1 )
2 2
1
= ( T · u1 × T · u2 + T · u1 × T · u2 )
2
= T · u1 × T · u2

=⇒

1
( T # T ) · ( u1 × u2 ) = T · u1 × T · u2 ,
2

the second-order tensor

+ 1
T= (T#T)
2

is denoted as an adjunct tensor. Using a calculation rule for cross tensor products
of tensors one gets

1
(T#T) : T = eijk ( T · ui × T · uj ) · T · uk
[ u1 u2 u3 ]

54
[ u1 u2 u3 ] ( T # T ) : T = eijk ( T · ui × T · uj ) · T · uk

[ u1 u2 u3 ] ( T # T ) : T = ( T · u1 × T · u2 ) · T · u3 + ( T · u2 × T · u3 ) · T · u1
+ ( T · u3 × T · u1 ) · T · u2 − ( T · u3 × T · u2 ) · T · u1
− ( T · u2 × T · u1 ) · T · u3 − ( T · u1 × T · u3 ) · T · u2
= 3 ( T · u1 × T · u2 ) · T · u3 − 3 ( T · u2 × T · u1 ) · T · u3
= 3 ( T · u1 × T · u2 ) · T · u3 + 3 ( T · u1 × T · u2 ) · T · u3
= 6 ( T · u1 × T · u2 ) · T · u3

=⇒

1 ( T · u1 × T · u2 ) · T · u3
(T#T) : T = .
6 [ u1 u2 u3 ]

The scalar value on the left-hand side is named as the determinant of the tensor T,
for which the symbol det is introduced:

1
det T = (T#T) : T .
6

Important calculation rules for finding the determinant of a are given by

1
det( T + S ) = [(T + S)#(T + S)] : (T + S)
6

1 1 1
det( T + S ) = ( T # T ) : T + ( T # T) : S + ( T # S) : T
6 6 3
1 1 1
+ (T#S) : S + (S#S) : T + (S#S) : S
3 6 6

55
1+ 1
det( T + S ) = det T + T : S + ( T # T) : S
3 3
1 1+
+ ( S # S ) : T + S : T + det S
3 3
=⇒

1+ 2+ 2+ 1+
det( T + S ) = det T + T : S + T : S + S : T + S : T + det S
3 3 3 3

+ +
det( T + S ) = det T + T : S + S : T + det S

+ +
det( T + S ) = det T + T : S + T : S + det S

=⇒

+ +
det( T + S ) = det T + T : S + T : S + det S ,
+ + +
3 det T = T : T = ( T · TT ) : I = ( TT · T ) : I ,
+ +
(det T) I : I = ( T · TT ) : I = ( TT · T ) : I ,
+ +
(det T) I = T · TT = TT · T .

From the last expression, one obtains an explicit form for the inversion of the tensor
T:

+
(det T) I · T−1 = TT · T · T−1

+
(det T) T−1 = TT · I
56
=⇒
+
TT
T−1 = .
det T
In addition, a few more calculation rules without proof are mentioned:
+ + + + +
( T · S ) = T · S , TT = (TT )

and

det( α T ) = α3 det T , det( T · S ) = det T det S ,


+
det TT = det T , det T−1 = (det T)−1 , det T = (det T)2 ,
det I = 1 , (det Q)2 = 1 .

b) The eigenvalue problem and the invariants

Given an arbitrary vector v and an arbitrary second-order tensor T, the eigenvalue


problem of a tensor T reads:

T·v =γv

=⇒

(T − γ I) · v = o .

where γ is a real scalar quantity. This equation can be interpreted as a conditional


equation for v; it is homogeneous and possesses non-trivial solutions only when
T − γI is singular, i.e., when
+
det( T − γ I ) = = det T + T : ( − γ I ) + T : ( − γ I )+ + det( −γ I )
1
= det T − γ (T#T) : I
2
1
+ γ 2 T : ( I # I ) − γ 3 det I = 0
2
57
=⇒

det( T − γ I) = Φ(γ) = − γ 3 + γ 2 IT − γ IIT + IIIT = 0

where

1
IT = (T#I) : I = T : I ,
2
1 +
IIT = ( T # T ) : I = T : I
2
1
= [ ( T : I )2 − TT : T ] ,
2
1 1+
IIIT = ( T # T ) : T = T : T
6 3
1 1
= ( T : I )3 − ( T : I ) ( TT : T )
6 2
1
+ TT · TT : T = det T
3
the three principal scalar invariants of the tensor T. The cubic equation is described
as the characteristic equation of T. The solution of the equation Φ(γ) yields the
principal values or eigenvalues of T, namely γ1 , γ2 and γ3 . The vectors related to
the eigenvalues are called the eigenvectors of T. It is assumed that the three roots
γ1 , γ2 and γ3 are known and the following cases are considered:

a) The roots γ1 , γ2 and γ3 are all distinct from each other. The accompanying
eigenvectors are v1 , v2 and v3 , so that the following relation holds:

T · vi = γi vi , Σ/ i .

When T is symmetric, the following equation can be derived:

( γi − γj ) vi · vj = 0 , Σ/ i , Σ/ j .

For i 6= j, the first factor is unequal to zero so that vi · vj must vanish in order to
fulfill this equation. In the case of i equals j, however, the first factor vanishes and
58
vi ·vj is unequal to zero. Furthermore, presuming that the vector v is an unit vector,
in the case for i equals j, the scalar product vi · vj takes the value of one and it holds
that:

vi · vj = δij ,

i.e., vi forms an orthonormed base.

The coefficients of T have the values:


 
 γ1 0 0 
tik : 0 γ2 0 .
0 0 γ3
 

b) When γ1 = γ2 = γ 6= γ3 , a single eigenvector v3 is assigned to γ3 and each vector


orthogonal to v3 is associated with γ.

c) When γ1 = γ2 = γ3 = γ, it holds that T = γ I and all vectors v are eigenvectors.

In order to demonstrate that the three roots are real for symmetric tensors, a proof
is carried out with the help of a contradictory proof. It will be assumed that a pair
of roots is conjugate complex: γ = α+iβ. Then, also the corresponding eigenvectors
v are complex: v = p + iq. Therefore, it holds that:

[T − (α + iβ )I] · (p + iq) =o
T · p − α p − i β p + i T · q − i α q − i2 β q = o
T · p − αp − iβ p + iT · q − iαq + β q =o
(T · p − αp + β q) + i(T · q − β p − αq) = o
=⇒

[T − (α + iβ )I] · (p + iq) = o

and

T·p−αp+βq = o , T·q−βp−αq = o ,
59
respectively. By scalar multiplication with the vectors q and p of the two equations
and subtracting both equations

q·T·p−αq·p+βq·q = 0 , p·T·q−βp·p−αp·q = 0

q·T·p−αq·p+βq·q = p·T·q−βp·p−αp·q
q·T·p+βq·q=p·T·q−βp·p
q·T·p−p·T·q+βq·q+βp·p=0

=⇒

q · T · p − p · T · q + β (q · q + p · p) = 0

is obtained. When the tensor T is symmetric then the following equation remains:

q · T · p − q · TT · p + β ( q · q + p · p ) = 0
q · ( T − TT ) · p + β ( q · q + p · p ) = 0

=⇒

β (q · q + p · p) = 0 .

The terms in the brackets contain squared terms and are therefore unequal to zero.
Therefore, β has to vanish, to fulfill the equation. It follows that for a symmetrical
tensor, all three roots are real.

In the following the three invariants IT , IIT and IIIT will be represented through the
three solutions γ1 , γ2 and γ3 of the characteristic equation. The first invariant is
given by:

[v1 v2 v3 ] ( T # I ) : I = 2 ( T · v1 × v2 ) · v3 + 2 ( v1 × T · v2 ) · v3
+ 2 ( v1 × v2 ) : T · v3 .

60
Using

1 1
[v1 v2 v3 ] ( I # I ) : T = [v1 v2 v3 ] 2 I : T = [v1 v2 v3 ] T : I
2 2

=⇒

1
[v1 v2 v3 ] ( I # I ) : T = ( γ1 + γ2 + γ3 ) [v1 v2 v3 ] ,
2

the three invariants of the tensor T read:

1
IT = ( T # I ) : I = γ1 + γ2 + γ3 ,
2
1
IIT = ( T # T ) : I = γ1 γ2 + γ2 γ3 + γ3 γ1 ,
2
1
IIIT = ( T # T ) : T = γ1 γ2 γ3 .
6
The theorem of Cayley-Hamilton can be easily developed with the help of the adjunct
tensor, according to the premise that a tensor T is invertible. With this one gets

+ 1
TT = ( T # T )T .
2

Furthermore, an alternative form for the adjunct tensor can be found:

1 1
( T # T )T = { [ ( T : I )2 − TT : T ] I − 2 ( T : I ) TT + 2 TT · TT }T
2 2
1
= { [ ( T : I )2 − TT : T ] I − 2 ( T : I ) T + 2 T · T }
2
1
= ( 2 IIT I − 2 IT T + 2 T · T )
2
= IIT I − IT T + T · T

=⇒

+
TT = IIT I − IT T + T2 .
61
With this, the theorem of Cayley-Hamilton

+
TT = ( det T ) T−1 = IIT I − IT T + T2

( det T )T−1 · T = ( IIT I − IT T + T · T ) · T


( det T )I = IIT T − I2T T · T + T · T · T

=⇒

T3 − IT T2 + IIT T − IIIT I = 0

can be derived.

62
4.10 Fundamental tensors

Fundamental tensors are characterized by the fact that they are independent of every
scalar, vector, and tensor variable; moreover, they are solely defined by the metric.
For the representation of diagrams, scalar, and cross products in base systems,
the fundamental tensors can be used in a very advantageous way. Due to their
importance, the fundamental tensors will be systematically put together and a few
fundamental features will be shown. With the vectors u and v, as well as the second-
order tensors T and S, the following relationships are defined with respect to the
fundamental tensors, from the second-order through the sixth-order:

3
v = I · v , u × v = E · (u ⊗ v) ,
4 4 4
T = I : T , ( T : I ) I = Ī : T = ( I ⊗ I ) : T , T = ¯Ī : T ,
T

6 (4)
T#S = E : (T ⊗ S) .

By the implementation of tensor calculus, a definite base is introduced. Depending


on the chosen base, the metric is defined with the help of fundamental tensors. The
coefficients of the fundamental tensors form the metric via the scalar product of the
base vectors. In the case of a pure covariant base system, the base-representation of
the fundamental tensors is given by:

I = ( g i · g j ) gi ⊗ gj ,
3
E = gi · ( gj × gk ) gi ⊗ gj ⊗ gk ,
4
I = ( g i ⊗ g j ) · ( g k ⊗ g l ) gi ⊗ gj ⊗ gk ⊗ gl ,
4
Ī = ( gi ⊗ gk ) · ( gj ⊗ gl ) gi ⊗ gj ⊗ gk ⊗ gl ,
4
¯Ī = ( gi ⊗ gk ) · ( gl ⊗ gj ) g ⊗ g ⊗ g ⊗ g ,
i j k l
6
E = [ ( gi × gk ) · gm ] [ ( gj × gl ) · gn ] gi ⊗ gj ⊗ gk ⊗ gl ⊗ gm ⊗ gn

and

63
I = gij gi ⊗ gj ,
3
E = ēijk gi ⊗ gj ⊗ gk ,
4
I = gik gjl gi ⊗ gj ⊗ gk ⊗ gl ,
4
Ī = gij gkl gi ⊗ gj ⊗ gk ⊗ gl ,
4
¯Ī = gil gkj g ⊗ g ⊗ g ⊗ g ,
i j k l
6
E = ēijklmn gi ⊗ gj ⊗ gk ⊗ gl ⊗ gm ⊗ gn .
respectively, where
1 1
ēijk = √ eijk , ēijklmn = ēikm ējln = eikm ejln .
g g
Alternative forms can easily be formed through the exchange of contravariant base
vectors against covariant in the coefficients, with the change of the corresponding
base vectors in the tensor base taking place at the same time.

The scalar products of vectors and tensors can be expressed in a simple way with
the help of the fundamental tensors:

u · v = I : (u ⊗ v) ,
3 (3)
(u × v) · w = E : (u ⊗ v ⊗ w) ,
4 (4)
T
T : S = I : (T · S) = I : (T ⊗ S) ,
6 (6)
(T # S) : R = E : ( T ⊗ S ⊗ R ) .
Furthermore, a few more characteristics of the fundamental tensors are emphasized:
q
√ √ 3 3 (3) 3 √ √
|I| = I : I = 3 , |E| = E : E = 3! = 6 ,
q 4
4 4 (4) 4 √ 4
¯ 4
|I| = I : I = 3 · 3 = 3 , |Ī| = |Ī| = |I| ,
q
6 6 (6) 6 √
|E| = E : E = 6 · 6 = 6 .
64
Another feature is that, by a proper orthogonal exchange of the base vectors, the
coefficients remain unchanged. This characteristic will be proven in the following
equations: the fundamental tensor I can be expressed in a covariant base system by

I = ( gi · gj ) gi ⊗ gj .

Carry out an orthogonal substitution of the base, one gets

I = ( Q · gi ) · ( Q · gj ) Q · gi ⊗ Q · gj
= ( g i · Q T · Q · g j ) Q · gi ⊗ Q · gj

=⇒

I = ( g i · g j ) Q · gi ⊗ Q · gj .

3
Likewise, one proceed, with the fundamental tensor E:

3
E = g i · ( g j × g k ) gi ⊗ gj ⊗ gk
= Q · gi · ( Q · gj × Q · gk ) Q · gi ⊗ Q · gj ⊗ Q gk
1
= [ ( Q # Q ) : Q ] [ g i · ( g j × g k ) ] Q · gi ⊗ Q · gj ⊗ Q · gk
6
= det Q [ gi · ( gj × gk ) ] Q · gi ⊗ Q · gj ⊗ Q · gk
= gi · ( gj × gk ) Q · gi ⊗ Q · gj ⊗ Q · gk .
4
The same considerations of the fundamental tensor I lead to

4
I = ( gi ⊗ gj ) · ( gk ⊗ gl ) gi ⊗ gj ⊗ gk ⊗ gl
= gik gjl gi ⊗ gj ⊗ gk ⊗ gl
= ( Q · gi ⊗ Q · gj ) · ( Q · gk ⊗ Q · gl ) Q · gj ⊗ Q · gj ⊗ Q · gk ⊗ Q · gl
= gik gjl Q · gj ⊗ Q · gj ⊗ Q · gk ⊗ Q · gl .

The validity of the statement for the remaining fundamental tensors is directly
apparent.
65
In view of some questions formulated in mechanics, the following identities, in corre-
lation with the fundamental tensors, are of importance. With the arbitrary second-
order tensor S, it holds that:

3 3
I = Q · I · QT , E : ( Q · S · QT ) = Q · ( E : S ) ,
4 4 4 4
I : ( Q · S · Q ) = Q · ( I : S ) · Q , Ī : ( Q · S · Q ) = Q · ( Ī : S ) · QT ,
T T T

4 4
¯Ī : ( Q · S · QT ) = Q · ( ¯Ī : S ) · QT .

Scalar multiples of the fundamental tensors are denoted as isotropic tensors. These
play an excellent role in the formulation of constitutive equations for isotropic mate-
rials. As an example, the constitutive equation of an isotropic linear-elastic material
is considered. This gives the symmetrical stress tensor S ≈ σ, depending on Greens
linear symmetrical strain measure E:

4
S=K:E .

4
The constitutive response K will be discussed in the following equations. The
isotropic condition regarding the material behavior comprises the demand

4 4
K : ( Q · E · QT ) = Q · ( K : E ) · QT .

The fourth-order tensor

4 4 4 4
K = α I + β Ī + γ ¯Ī

fulfills the aforementioned relation. As Greens strain tensor E is symmetrical it


holds that

4 4
( α I + γ ¯Ī ) : E = ( α + γ ) E = 2 µ E , 2 µ = α + γ .

Thus, the constitutive equation reduces to


66
S = 2µE + β (E : I)I

or

4 4 4
S = K : E = ( 2 µ I + λ Ī ) : E ,

where β = λ has been used. The material parameters µ and λ are the well-known
Lamé constants.

67
5. Vector and tensor analysis

The scalars, vectors, and tensors occurring in physics can be functions of real scalar,
vector, and tensor parameters. The determination of the change of these functions,
due to an infinitesimal increase of the parameters, is an integral part of the devel-
opment of the laws of physics. In this context, one is led, beside others, to the
differentiation of functions, differential quotients, as well as the partial and the total
differentiation. The aim of this chapter is to discuss and then clarify these notions.

Due to their great importance in mechanics and engineering, first of all vector and
tensor functions depending on real scalar parameters are treated.

5.1 Functions of scalar parameters

It is assumed that the vector u is a function of a real scalar variable α. Moreover,


only unique functions are considered; besides that, the same symbol for the function,
as well as for the value of the function will be used, as long as no confusion occurs.

The derivative of this function u, when it exists in any open field in the Euclidean
space, is the value of the function v. It is defined with the following limes consider-
ation:

u (τ ) − u (α) − v (α) ( τ − α )
lim =o
τ →α |τ − α|

where τ is likewise a real scalar variable. The derivative v(α) is normally also written
as:

du (α)
v (α) = , v (α) = u0 (α) .

The differential of the function u for a given value α and a given value of the
differential dα is equal to the product u0 (α) and dα:

du (α) = u0 (α) dα .

Furthermore, one can define the derivatives and higher-order differentials. The sec-
ond derivative of u is written as:
68
d2 u 00 d du
= u = ( ).
dα2 dα dα

Therefore, u00 is again a vector depending on α. Similarly, the higher derivatives are
determined. The second differential of a vector function u depending on a variable
α is the differential of the first differential:

d2 u = d(du) = u00 (α) dα2 .

Analogous to this, the higher-order differentials will be introduced:

d3 u = d(d2 u) = u000 (α) dα3 .

The partial derivative of w which in turn, is a function of the real scalar variable α,
β, γ, . . . , with respect to one of these variables, for example to α, will be defined
as:

u (τ, β, γ, . . . ) − u (α, β, γ, . . . ) − w (α, β, γ, . . . ) ( τ − α )


lim =o .
τ →α |τ − α|

The partial derivative w which again is a function of the real scalar variable (α, β,
γ, . . . ), can be written as:

∂u
w= .
∂α

The total differential of a vector function u, which depends on several real scalar
variables, is expressed as:

∂u ∂u ∂u
du = dα + dβ + dγ + . . . .
∂α ∂β ∂γ

The second-order partial derivative allows various possibilities for differentiation.


Either this can be formed according to the same variable as the first derivative,

∂ 2u ∂ 2u
, , ... ,
∂α2 ∂β 2
69
or according to another variable, such as

∂ 2u ∂ 2u
, , ... .
∂α ∂β ∂β∂γ
The expressions mentioned above are also denoted as mixed derivatives.

The total nth -order differential for vector functions which depend on several variables
can be represented as
∂ ∂
dn u = ( dα + dβ + . . . )n u .
∂α ∂β
In a similar way, one can define the derivatives and differentials for tensor functions
of one or more real scalar variables. The derivative of a tensor function T, with
respect to a real scalar variable, is again a tensor function. Therefore, one does not
need to develop any special calculus for tensor functions; in the definitions for the
vector functions, one need only to replace the vectors with tensors.

The conventional rules of differential calculus for scalar functions can easily be ex-
tended to cover vector and tensor functions. For example, λ being a scalar, a, b
3
and u vectors, as well as T, S and L tensors, all depend on a real scalar variable.
In the following, the definitions and the corresponding analysis for tensors, furnish
rules of differentiation, are given for a few special cases:

( λ a )0 = λ0 a + λ a0 ,
( a · b )0 = a0 · b + a · b0 ,
( a ⊗ b )0 = a0 ⊗ b + a ⊗ b0 ,
( a × b )0 = a0 × b + a × b0 ,
( u × T )0 = u0 × T + u × T0 ,
( T · u )0 = T0 · u + T · u0 ,
3 3 3
( L · u )0 = (L)0 · u + L · u0 ,
( T : S )0 = T0 : S + T : S0 ,
( T · S )0 = T0 · S + T · S0 ,
( T # S )0 = T0 # S + T # S0 .
70
5.2 Field theory

Vectors and tensors assigned to a point X in Euclidean point space are described
with the help of the position vector x. Such functions are defined as fields. In
general, one differentiates between scalar, vector and tensor fields, which are all
functions of the position vector x. These field functions play an important role in
physics. In this context, it will be mentioned the temperature (scalar field), the force
and displacement fields (vector fields) and the stress field (tensor field). These fields
are in general also time-dependent. One can free from the introduction of special
coordinates for the analysis of such functions and assume the functional dependence
on the position vector, which shall be an advantage in a more concise and clear
representation of the derivatives.

For the field functions, it is assumed that they are unique and smooth.

5.2.1 Gradient

Given φ, a scalar function of the position vector x that is defined in an open area;
then φ is differentiable in this area, when a vector field w(x) exists, so that with x
and y as arbitrary position vectors, one gets

φ (y) − φ (x) − w (x) · ( y − x )


lim =0 .
y→x |y − x|

The vector w is uniquely determined and is called the gradient of the scalar field
φ(x):

dφ (x)
w= = grad φ (x) = 5φ (x) .
dx

The differential of the scalar field φ(x), for a given value x and a given value of the
differential dx, is equal to the scalar product of grad φ and dx:

dφ = grad φ · dx .

The partial derivative of a scalar function, depending on the position vector x and
a real scalar variable α, with respect to the position vector x, is defined with the
relationship
72
φ (y, α) − φ (x, α) − u (x, α) · ( y − x )
lim =0 .
y→x |y − x|

The partial derivative u(x, α) is written as

∂φ (x, α)
u= .
∂x

For the partial derivative of the scalar function φ = φ(x, α) with respect to the
position vector x, the following expressions grad φ and 5φ, respectively, have es-
tablished themselves. The same is true for the partial derivatives, introduced in
the following equations, of the vector and tensor functions, which depend on the
position vector x and the real scalar variables α.

The total differential of φ(x, α) can be written as

∂φ ∂φ
dφ (x, α) = · dx + dα .
∂x ∂α

Considering the vector function v of the position vector x, the gradient of the vector
field v(x) can be defined as the tensor field R(x) through

v (y) − v (x) − R (x) · ( y − x )


lim =0 ,
y→x |y − x|

where

dv (x)
R (x) = grad v (x) = = 5v (x) .
dx

The differential of the vector field v(x) for a given value of the differential dx, is
determined with the help of a linear transformation

dv = ( grad v ) · dx = 5v · dx .

Furthermore, one can determine the partial derivative of a vector field v(x, α) with
respect to the position vector x through

73
v (y, α) − v (x, α) − S (x, α) · ( y − x )
lim =0
y→x |y − x|

with

∂v (x, α)
S (x, α) = = grad v = 5v .
∂x

With this one gets the total differential of v(x, α)

∂v ∂v
dv (x, α) = · dx + dα .
∂x ∂α

Finally, the tensor function T depending on the position vector x is examined. The
3
gradient of the tensor field T(x) is then defined as the tensor field S(x) through

3
T (y) − T (x) − S (x) · ( y − x )
lim =0 ,
y→x |y − x|

where

3 dT (x)
S = grad T (x) = = 5T (x) .
dx

The differential dT(x) is defined, with the help of the linear mapping, as

dT (x) = ( grad T ) · dx = 5T · dx .

Similar to the procedure used for partial derivatives of scalar and vector fields,
the partial derivative of the tensor field T(x, α) is introduced, with respect to the
position vector x and the real quantity α:

3
T (y, α) − T (x, α) − R (x, α) · ( y − x )
lim =0 .
y→x |y − x|

Thereby is
74
3 ∂T (x, α)
R (x, α) = = grad T = 5T .
∂x

For the total differential of T(x, α), it follows from the previous explanation, that

∂T ∂T
dT (x, α) = · dx + dα .
∂x ∂α

The transfer of the considerations, with regard to the formation of gradients and
total differentials, to such scalar, vector and tensor functions, depending on the
position vector x and various real scalar variables, is possible without problem.

With the help of the given definitions, the following rules for the formation of gra-
dients from product terms can be derived:

5( φ ψ ) = φ 5ψ + ψ 5φ ,
5( φ v ) = v ⊗ 5φ + φ 5v ,
5( φ T ) = T ⊗ 5φ + φ 5T ,
5( u · v ) = ( 5u )T · v + ( 5v )T · u ,
5( u × v ) = u × 5v + 5u × v ,
23
5( a ⊗ b ) = [ 5a ⊗ b + a ⊗ ( 5b )T ]T ,
23
5( T · v ) = ( 5T)T · v + T · 5 v ,
23 23
5( T · S ) = [ ( 5T)T · S ]T + T · 5S ,
13 13
5( T : S ) = ( 5T)T : ST + ( 5 S )T : TT .

Furthermore, it holds for the gradient of the position vector:

5x = I .

75
5.2.2 Derivatives of higher-order

The second derivatives of a scalar field φ(x), a vector field v(x) and a tensor field
T(x) are defined as

d2 φ d dφ
= ( ) = 55φ ,
dx ⊗ dx dx dx
d2 v d dv
= ( ) = 55 v ,
dx ⊗ dx dx dx
d2 T d dT
= ( ) = 55T ,
dx ⊗ dx dx dx
where 55φ, 55v and 55T are a second-order tensor, a third-order tensor and
a fourth-order tensor, likewise depending on x. Correspondingly, the higher deriva-
tives can be defined.

The second differentials of a scalar field function φ(x), a vector field function v(x)
and a tensor field function T(x) read

d2 φ = d(dφ) = 55 φ : ( dx ⊗ dx ) ,
d2 v = d(dv) = (55v) · ( dx ⊗ dx) ,
d2 T = d(dT) = ( 55T ) : ( dx ⊗ dx ) .

In a corresponding manner, the higher differentiations of scalar, vector and tensor


fields are defined.

5.2.3 Special operations (divergence, rotation and the Laplace-operator)

In physics, it has proved to be helpful to introduce certain operators that allow the
use of abbreviations and partly make physical interpretation possible. Firstly, the
operator div (divergence) is devoted. The divergence of a vector field v(x) and a
tensor field T(x) are given by the definitions

div v (x) = 5v (x) : I ,


div T (x) = 5 T (x) : I .

76
Consequently, the divergence of a vector field v(x) is a scalar field, the divergence
of a tensor field T(x) is therefore a vector field.

Taking into consideration the afore-mentioned definitions and calculations rules, the
following calculation rules can be derived directly:

div( φ v ) = v · 5φ + φ div v ,
div( T · v ) = ( div TT ) · v + TT : 5v ,
div( 5 v )T = 5 div v ,
div( a ⊗ b ) = ( 5a ) · b + ( div b ) a ,
div( u × v ) = ( 5u × v ) : I − ( 5v × u ) : I
3 3
= v · E : (5u)T − u · E : (5v)T
= v · curl u − u · curl v ,
div( φ T ) = T 5φ + φ div T ,
div( T · S ) = 5T : S + T · div S ,
div( v × T ) = v × div T + 5v × T ,
div( 5v )+ =0 ,
div( 5v # I ) = 0 .

With respect to the expression div(u × v) the operation curl has used. The explicit
form of the operation curl (rotation) of a vector and tensor field is determined with
3
the help of the fundamental tensor E:

3
curl v(x) = E : (5v)T ,
3 13

curl T(x) = E : (5T)T .

Here curl v is a unique vector field and curl T a unique tensor field.

The following identities are useful for applications. Given a scalar field φ, a vector
field v and a tensor field T, where all the fields are smooth and can be differentiated,
then the following relationships hold:
77
curl( 5φ ) =o ,
div( curl v ) =0 ,
curl( 5v ) =0 ,
curl( 5v )T = 5( curl v ) ,
curl( φ v ) = φ curl v + 5φ × v ,
curl( u × v ) = u div v − ( 5v ) · u − v div u + ( 5u ) · v
= div( u ⊗ v − v ⊗ u ) ,
div curl T = curl div TT ,
div( curl T )T = o ,
( curl curl T )T = curl curl TT ,
curl( φ I ) = − [ curl( φ I ]T ,
curl( T · v ) = ( curl TT ) · v + (5v)T × T .

When T is symmetric, then

curl T : I = 0 , T = TT .

Given that TA is a skew-symmetric tensor and u the affiliated axial vector, then it
can be proven that

curl TA = (div u) I − 5u .

The alternative forms of the rotation of the vector field v(x) are

curl v(x) = div( I × v ) = I × 5v .

Furthermore, the Laplace-operator 4 is introduced. Given the scalar field φ(x), the
vector field v(x) and the tensor field T(x), then the Laplace-operator applied to
φ(x), v(x) and T(x) are defined as
78
4φ = 5 5φ : I = div grad φ ,
4 v = ( 55v ) : I ,
4 T = ( 55T ) : I ,

where 4φ is a scalar quantity, 4v is a vector field and 4T a tensor field.

In connection with the Laplace-operator, the following identities can be stated:

curl curl v = 5 div v − 4v ,


4 (tr T) = tr( 4T ) ,
curl curl T = − 4 T + 5(div T) + ( 5 div T )T − 55(tr T) +
+ I [ 4(tr T) − div div T ]

and

div[ 5v ± ( 5v )T ] = 4v ± 5 div v .

When T is symmetrical and T = S − (tr S)I, then the following relation is valid

curl curl T = − 4 S + 5 div S + ( 5 div S )T − ( div div S ) I .

79
5.3 Functions of vector and tensor variables

In the presented sections, one has used the distance between two points y and x,
having the given norm |x − y|, for the definition of the derivatives of scalar, vector
and tensor fields. The idea behind this procedure being that for scalar, vector
and tensor functions, which supposedly depend on the arbitrary vector and tensor
variables, the norm of a variable should likewise be used for the definition of the
derivative. In what follows, one makes use of this idea.

Given a scalar function φ(q) of a vector q, defined in an open area, then φ can be
differentiated in this area when a vector function w(q) exists; so that, with q and
p as arbitrary vectors, one gets

φ (p) − φ (q) − w (q) · ( p − q )


lim =0
p→q |p − q|

with

dφ (q)
w (q) = .
dq

When the vector w(q) exists, then it is uniquely determined. The differential of the
scalar function φ(q), namely dφ for given values q and dq, is equal to the scalar
product


dφ = · dq .
dq

Now, the vector function v(q) is devoted. The derivative of v with respect to q is
given through the definition

v (p) − v (q) − R(q) · ( p − q )


lim =o .
p→q |p − q|

Thereby, for the tensor function R(q) one writes

dv
R= .
dq

80
The differential of the vector function v(q) can be stated by the linear mapping as

dv
dv = ( ) · dq .
dq

Considering the tensor function T(q) in an open area, the derivative, with respect
3
to q, as a third-order tensor function S(q) is defined through

3
T (p) − T (q) − S (q) · ( p − q )
lim =0 .
p→q |p − q|

One states

3 dT
S= .
dq

The differential dT(q) is determined through linear mapping

3
dT(q) = S (q) · dq .

Introducing scalar, vector and tensor functions which depend on a tensor variable
M, namely φ(M), v(M) and T(M), the derivatives of these functions, with respect
to the variable M (M and N are arbitrary tensors) are given by

φ (N) − φ (M) − U (M) : ( N − M )


lim =0 ,
M→N |N − M|
3
v (N) − v (M) − V (M) · ( N − M )
lim =o ,
M→N |N − M|
4
T (N) − T (M) − W (M) : ( N − M )
lim =0 .
M→N |N − M|

For the derivatives one can write

81
dφ (M)
U(M) = ,
dM
3 dv(M)
V(M) = ,
dM
4 dT(M)
W(M) = .
dM

The differentials of the scalar, vector and tensor functions φ(M), v(M) and T(M)
for the given values are defined by


dφ = : dM ,
dM
dv
dv = : dM ,
dM
dT
dT = : dM .
dM
Finally, those scalar, vector and tensor functions which depend on real scalar vari-
ables and on vector and tensor variables are considered:

φ = φ(α, β , ... , p, q, ... , M, N, ...) ,


v = v(α, β , ... , p, q, ... , M, N, ...) ,
T = T(α, β , ... , p, q, ... , M, N, ...) .

The partial differentiation of these functions, with respect to the vector and tensor
variables are the following definitions, where p and q are arbitrary vectors and M
and N arbitrary tensors. Here, restricted to the statement of the definitions for the
partial derivatives of the scalar functions φ corresponding definitions are valid for
the partial differentiation of the vector and tensor functions v and T:

φ ( p , . . . ) − φ( q , . . . ) − u ( q , . . . ) · ( p − q )
lim =0 ,
p→q |p − q|
φ ( N , . . . ) − φ( M , . . . ) − U ( M , . . . ) : ( N − M )
lim =0 .
N→M |N − M|

82
The partial derivations (the vector and tensor functions u and U) are denoted
through

∂φ ∂φ
u= , U= .
∂q ∂M

Furthermore, the partial differentiation of the vector and tensor functions v and T,
with respect to the vector and tensor variables q and M, can be written as:

∂v ∂v ∂T ∂T
, , , .
∂q ∂M ∂q ∂M

These are second- and third- as well as fourth-order tensor functions. For the total
differential, one finds

∂φ ∂φ ∂φ
dφ = dα + . . . + · dq + . . . + : dM + . . . ,
∂α ∂q ∂M
∂v ∂v ∂v
dv = dα + . . . + · dq + . . . + · dM + . . . ,
∂α ∂q ∂M
∂T ∂T ∂T
dT = dα + . . . + · dq + . . . + : dM + . . . .
∂α ∂q ∂M

Now, higher derivatives of scalar, vector and tensor functions which depend on scalar,
vector and tensor variables are discussed. These can be formed immediately accord-
ing to the guidelines shown in the afore-mentioned sections. The given definitions
allow the development of calculation rules. Thereby one can dispense with those
rules which contain the derivative with respect to a vector because the rules can be
analogously applied to arbitrary scalar, vector and tensor functions. Consequently,
it will be restricted to a few important rules for the differentiation of product terms
of the scalar, vector and tensor functions φ, ψ and v, u as well as T, S which, beside
others, depend on the tensor variable M, with respect to the variable M:

83
∂(φ ψ) ∂ψ ∂φ
=φ +ψ ,
∂M ∂M ∂M
∂(φ v) ∂φ ∂v
=v⊗ +φ ,
∂M ∂M ∂M
∂(φ T) ∂φ ∂T
=T⊗ +φ ,
∂M ∂M ∂M
∂(u · v) ∂u ∂v
=v· +u· ,
∂M ∂M ∂M
∂(T : S) ∂T T ∂S T
=( ) :S+( ) :T .
∂M ∂M ∂M
Furthermore, the following derivative rules are valid:

∂M 4 ∂( M : I ) I 4 ∂MT ¯4
=I , = Ī , = Ī ,
∂M ∂M ∂M
A
∂IM ∂IIM ∂IIIM + ∂ T(M) 1 3
=I , = M#I , =M , =− E .
∂M ∂M ∂M ∂M 2
The definitions given in this section make it possible to state the stepwise differen-
tiation (chain rule) for composed functions. So, for example, it holds for the scalar
function φ(q, M), where q and M depend on the real scalar variable α, that

dφ ∂φ dq ∂φ dM
= · + : .
dα ∂q dα ∂M dα

Analogously, one can give the derivative (with respect to α for the vector and tensor
functions v(q, M), when q and M are functions of α) as follows:

dv ∂v dq ∂v dM dT ∂T dq ∂T dM
= · + : , = · + : .
dα ∂q dα ∂M dα dα ∂q dα ∂M dα

This procedure can also be applied to spatial variable functions. Consider the scalar,
vector and tensor functions φ(α, q, M), v(α, q, M) and T(α, q, M), where α, q and
M should be spatial variable functions. Then one gets, for the gradients of these
functions,

84
∂φ ∂φ 13
∂φ T
5φ = · 5α + ( 5q ) ·
T
+ ( 5M ) T : ( ) ,
∂α ∂q ∂M
∂v ∂v ∂v
5v = ⊗ 5α + · 5q + : 5M ,
∂α ∂q ∂M
∂T ∂T ∂T
5T = ⊗ 5α + · 5q + : 5M .
∂α ∂q ∂M

85
5.4 Integral theorems

By the evaluation of the conservation laws in mechanics, it is necessary to trans-


form the surface integrals into volume integrals, and also to transform line integrals
into surface integrals, when one takes the global view by the formulation of the
conservation theorems. These transformations cover scalar, vector and tensor fields.
The integral theorems also play a role in other areas of geometry and mechanics,
for example, by the numerical calculation of surface moment of inertia and mass
moment of inertia. In the following section, the most important integral theorems
are developed.

5.4.1 Transformation of surface integrals into volume integrals

Firstly, the surface integral of the tensor product of the vector field u(x), having
the area vector da equal to da n (where n is the normal unit vector at the surface
and da the surface element), is transformed into a volume integral:

R R R ∂u R
u ⊗ da = u ⊗ n da = dv = grad u dv ,
∂B ∂B B ∂x B

where B is a closed volume and ∂B its surface.

More integrals can easily be developed from this integral theorem, with respect to
the transformation of surface integrals into volume integrals which contain a scalar
field φ(x), a vector field u(x) and a tensor field T(x):

R R ∂φ
φ n da = dv ,
∂B B ∂x
R R
u · n da = div u dv ,
∂B B
R R
n × u da = curl u dv ,
∂B B
R R
T · n da = div T dv ,
∂B B

u × T n da = ( u × div T + 5u × T ) dv .
R R
∂B B

86
In the literature, the third transformation is also called as the Gauss integral theo-
rem. With φ as a constant field, one directly gets the so-called surface theorem:
R R
n da = o , da = o .
∂B ∂B

5.4.2 Transformation of line integrals into surface integrals


H
The aim of this chapter is to transform the closed line integrals dx⊗ u into a surface
L
integral. The line L will be described by the position vector x. The derivation yields

H R ∂u T R ∂u T
dx ⊗ u = da × ( ) = n×( ) da .
L ∂B ∂x ∂B ∂x

In addition, the following integral theorems are given:

H R ∂φ
φ dx = n× da ,
L ∂B ∂x
H R
u · dx = curl u · n da ,
L ∂B
H R R ∂u T
u × dx = ( div u ) n da − ( ) · n da ,
L ∂B ∂B ∂x
H R
T · dx = ( curl T )T · n da ,
L ∂B

{ [ u × ( curl T )T ] · n + ( n × TT )T × 5u } da .
H R
u × T · dx =
L ∂B

The second relationship is known as the Stokes integral theorem. For every closed
surface, the surface integrals disappear in the afore-mentioned equations (exception
the last equation), which directly follow from the integral theorems.
H From Stokes
integral theorem, one can read that, for every closed curve, u · dx is equal zero,
L
when curl u vanishes within a specified area. Then, according to the last relation,
it is necessary that uis equal to gradφ, where φ represents a scalar field. Given the
vector field uas equal to grad φ, then
H curl u is equal to zero (necessary and sufficient).
As in hydrodynamics, one calls u · dx, in general the circulation.
L

87
5.5 Example: Integral theorem

6
3
d g3
5 4
2
1
d 2g2 3
u
d 1g1

0
Figure 5.1: Infinitesimal volume element

da
Figure 5.2: Surface element and normal vector da = n da

– Volume element

dv = dΘ1 g1 · ( dΘ2 g2 × dΘ3 g3 ) = dΘ1 dΘ2 dΘ3 g1 · ( g2 × g3 )



= g dΘ1 dΘ2 dΘ3 ,

where

88
g = det gik

– Surface vectors of the subareas 1 – 6



da1 = dΘ2 dΘ3 g3 × g2 = − dΘ2 dΘ3 g g1 = − da4 ,

da2 = dΘ1 dΘ3 g1 × g2 = − dΘ1 dΘ3 g g2 = − da5 ,

da3 = dΘ1 dΘ2 g2 × g1 = − dΘ1 dΘ2 g g3 = − da6

– Averages of the vectors ūi of the vector field u in the epicenters of the subareas

∂u dΘ2 ∂u dΘ3
ū1 = u + + ,
∂dΘ2 2 ∂dΘ3 2
∂u dΘ1 ∂u dΘ3
ū2 = u + + ,
∂dΘ1 2 ∂dΘ3 2
∂u dΘ1 ∂u dΘ2
ū3 = u + + ,
∂dΘ1 2 ∂dΘ2 2
∂ ū1
ū4 = ū1 + 1
dΘ1 ,
∂dΘ
∂ ū2
ū5 = ū2 + 2
dΘ2 ,
∂dΘ
∂ ū3
ū6 = ū3 + 3
dΘ3
∂dΘ
with

∂ ū1 ∂u ∂ 2u dΘ2 ∂ 2u dΘ3


1
dΘ1 = 1
dΘ + 1
dΘ + dΘ1 ,
∂dΘ ∂dΘ1 ∂dΘ2 ∂dΘ1 2 ∂dΘ3 ∂dΘ1 2
∂ ū2 ∂u ∂ 2u dΘ1 ∂ 2u dΘ3
2
dΘ2 = 2
dΘ + 2
dΘ + dΘ2 ,
∂dΘ ∂dΘ2 ∂dΘ1 ∂dΘ2 2 ∂dΘ3 ∂dΘ2 2
∂ ū3 ∂u ∂ 2u dΘ1 ∂ 2u dΘ2
3
dΘ3 = 3
dΘ + 3
dΘ + dΘ3 .
∂dΘ ∂dΘ3 ∂dΘ1 ∂dΘ3 2 ∂dΘ2 ∂dΘ3 2
Neglecting terms of higher order, i.e.,
89
dΘ2 dΘ3
dΘ1  dΘ1 , dΘ1  dΘ1 ,
2 2
dΘ1 dΘ3
dΘ2  dΘ2 , dΘ2  dΘ2 ,
2 2
dΘ1 dΘ2
dΘ3  dΘ3 , dΘ3  dΘ3 ,
2 2
the vectors ū4 – ū6 simplifies to

∂ ū1 ∂u
ū4 = ū1 + 1
dΘ1 ≈ ū1 + dΘ1 ,
∂dΘ ∂dΘ1
∂ ū2 2 ∂u
ū5 = ū2 + 2
dΘ ≈ ū2 + 2
dΘ2 ,
∂dΘ ∂dΘ
∂ ū3 ∂u
ū6 = ū3 + 3
dΘ3 ≈ ū3 + 3
dΘ3 .
∂dΘ ∂dΘ

– Tensor (dyadic) product of the averages ūi and the surface vectors dai

6
P
ūi ⊗ dai = ū1 ⊗ da1 + ū2 ⊗ da2 + ū3 ⊗ da3
i=1

+ ū4 ⊗ da4 + ū5 ⊗ da5 + ū6 ⊗ da6

6
P
ūi ⊗ dai = ū1 ⊗ da1 + ū2 ⊗ da2 + ū3 ⊗ da3
i=1

− ū4 ⊗ da1 − ū5 ⊗ da2 − ū6 ⊗ da3

=⇒

6
P
ūi ⊗ dai = ( ū1 − ū4 ) ⊗ da1 + ( ū2 − ū5 ) ⊗ da2 + ( ū3 − ū6 ) ⊗ da3
i=1

90
With

∂u 1 ∂u
ū1 − ū4 = ū1 − ū1 − 1
dΘ = − 1
dΘ1 ,
∂dΘ ∂dΘ
∂u 2 ∂u
ū2 − ū5 = ū2 − ū2 − 2
dΘ = − 2
dΘ2 ,
∂dΘ ∂dΘ
∂u 3 ∂u
ū3 − ū6 = ū3 − ū3 − 3
dΘ = − 3
dΘ3
∂dΘ ∂dΘ
the sum of the tensor (dyadic) products of the averages ūi and the surface vectors
dai over all surfaces reads

6 ∂u ∂u ∂u
dΘ1 ⊗ da1 − dΘ2 ⊗ da2 − dΘ3 ⊗ da3
P
ūi ⊗ dai = − 1 2 3
i=1 ∂dΘ ∂dΘ ∂dΘ

6 ∂u √
dΘ1 ⊗ ( − dΘ2 dΘ3 g g1 )
P
ūi ⊗ dai = − 1
i=1 ∂dΘ
∂u 2 1 3√
− 2
dΘ ⊗ (− dΘ dΘ g g2 )
∂dΘ
∂u √
− 3
dΘ3 ⊗ ( − dΘ1 dΘ2 g g3 )
∂dΘ

6 ∂u 1√
g dΘ1 dΘ2 dΘ3
P
ūi ⊗ dai = 1
⊗ g
i=1 ∂dΘ
∂u √
+ 2
⊗ g2 g dΘ1 dΘ2 dΘ3
∂dΘ
∂u 3√
+ 3
⊗ g g dΘ1 dΘ2 dΘ3
∂dΘ

6 ∂u ∂u ∂u √
⊗ g1 + ⊗ g2 + ⊗ g3 ) g dΘ1 dΘ2 dΘ3
P
ūi ⊗ dai = ( 1 2 3
i=1 ∂dΘ ∂dΘ ∂dΘ

91
=⇒

6 ∂u ∂u ∂u
1 2
⊗ g3 ) dv ,
P
ūi ⊗ dai = ( 1
⊗ g + 2
⊗ g + 3
i=1 ∂dΘ ∂dΘ ∂dΘ

where the relations for dai have been considered. The above expression can be
transferred into

6
P ∂u k ∂u ∂Θk ∂u
ūi ⊗ dai = ( k
⊗ g ) dv = ( k
⊗ ) dv = dv
i=1 ∂dΘ ∂dΘ ∂x ∂x

=⇒

6
P ∂u
ūi ⊗ dai = dv = grad u dv .
i=1 ∂x

Replacing the summation of the subareas of the considered infinitesimal element by


an integral one obtains

R ∂u
u ⊗ da = dv = grad u dv .
∂dB ∂x

Thus, the integral of the tensor product u ⊗ da over the surface of the whole body
can be expressed as

R R ∂u R
u ⊗ da = dv = grad u dv
∂dB B ∂x B

or

R R ∂u R
u ⊗ n da = dv = grad u dv .
∂B B ∂x B

92

You might also like