3 Some Basic Equations of Continuum Mechanics

Creep mechanics is a part of continuum mechanics, like elasticity, plasticity,

viscoelasticity, and viscoplasticity.
Continuum Mechanics is concerned with the mechanical behavior of
solids and fluids on the macroscopic scale. It ignores the discrete nature of
matter, and treats material as uniformly distributed throughout regions of
space. It is then possible to define quantities such as density, displacement,
velocity, etc., as continous ( or at least piecewise continous) functions of
position. This procedure is found to be satisfactory provided that we deal
with bodies whose dimensions are large in comparison with the character-
istic lengths (e.g.: interatomic spacings in a crystal, or mean free paths in a
gas) on the microscopic scale.
Continuum mechanics can also be applied to a granular material such as
sand, concrete or soil, provided that the dimensions of the regions considered
are large compared with those of an individual grain.
The equations of continuum mechanics are of two main kinds. First, there
are equations which apply equally to all materials. They describe universal
physical laws, such as conservation of mass and energy. Second, there are
equations characterizing the individual material and its reaction to applied
loads; such equations are called constitutive equations (Chapter 4), since
they describe the macroscopic behavior resulting from internal constitution
of the particular materials.
In this Chapter, however, only the kinematics and the concept of stress
should briefly be discussed.

3.1 Analysis of Deformation and Strain

This section is concerned with the kinematics of a continuous medium. Kine-

matics is the study of motion without regard to the forces which produce it.
To describe the motion of a body, i.e., to specify the position of each particle
at each instant we select a particular configuration of the body, for instance
32 3 Some Basic Equations of Continuum Mechanics

the configuration of the body in its unloaded or undeformed state, and call
this the reference configuration at the reference time t = 0. The set of co-
ordinates ai , referred to fixed cartesian axes, uniquely determines a particle
of the body and may be regarded as a label by which the particle can be
identified for all time (Fig. 3.1).

t >0
a2 x2

xi

path line
t =0

ai
a1

x1

x3 a3

Fig. 3.1 Motion of a particle; reference (t = 0) and current (t > 0) configurations

The motion of the body may now be described by specifying the position
xi of the particle ai at time t > 0 in the form

xi = xi (ap , t) , i, p = 1, 2, 3 . (3.1a)

In other words: The place xi is occupied by the body-point ap at time t. We

assume that this function is differentiable with respect to ap and t as many
times as required.
Sometimes we desire to consider only two configurations of the body, an
initial and a final configuration. We refer to the mapping from the initial to
the final configuration as a deformation of the body. The motion of the body
may be regarded as a one-parameter sequence of deformations.
The mapping (3.1a) has the unique inverse

ai = ai (xp , t) , i, p = 1, 2, 3 (3.1b)
 3.1 Analysis of Deformation and Strain 33

provided that the JACOBIan determinant

J := det(∂xi /∂aj ) (3.2)

exists and that it is always positiv (J > 0) at each point. The physical signif-
icance of these assumptions is that the material of the body cannot penetrate
itself, and that material occupying a finite non-zero volume at time t = 0
cannot be compressed to a point or expanded to infinite volume during the
motion; remark: dV = J dV0 .
The inverse form (3.1b) may be viewed as a description which provides
a tracing to its original position of the particle ai that now (at current time
t > 0) occupies the location xi . In other words: An observer placed on xi
registers a particle ai at time t.
The coordinates ai are known as material (or L AGRANGian) coordinates
since distinct sets of these coordinates refer to distinct material particles.
The coordinates xi are known as spatial (or E ULERian) coordinates since
distinct sets refer to distinct points of space. Problems in continuum mechan-
ics may be formulated either with the material coordinates ai as independent
variables, in which case we employ the material description of the prob-
lem, or with the spatial coordinates xi as independent variables, in which we
employ the spatial description . In the material (L AGRANGian) description
attention is focused on what is happening at (or in the neighbourhood of) a
particular material particle. In the spatial (E ULERian) description we concen-
trate on events at (or near to) a particular point in space. The mathematical
formulation of general physical laws and the description of the properties of
particular materials is often most easily accomplished in the material descrip-
tion, but for the solution of special problems it is frequently preferable to use
the spatial description. It is therefore necessary to employ both descriptions
and to relate them to each other. In principle it is possible to transform a
problem from the material to the spatial description or vice versa by using
(3.1a) or (3.1b). In practice the transition is not always accomplished easily.
The time rate of change of any property of a continuum with respect
to specific particles of the moving continuum is called the material time
derivative (or substantial derivative) of that property. This derivative may be
thought of as the time rate of change that would be measured by an observer
traveling with a specific particle ai . For instance, consider the temperature
field in a body
T = T (ai , t) or T = T (xi , t) . (3.3a,b)
Its material time derivative is expressed by
34 3 Some Basic Equations of Continuum Mechanics

Ṫ = dT /dt = ∂T (ai , t)/∂t . (3.4a)

The right-hand side of (3.4a) is sometimes written as [∂T (ai , t)/∂t]ai to em-
phasize that the coordinates ai are held constant, i.e., a given particle is in-
volved when calculating the partial derivative.
If the temperature field is expressed by the spatial description in the form
(3.3b) we arrive, by using the chain rule of partial differentiation, at the fol-
lowing result

dT /dt = ∂T (xi , t)/∂t + ẋk ∂T (xi , t)/∂xk , (3.4b)

where ẋi = vi is the velocity vector. From (3.4b) we read the operator
d ∂ ∂ d ∂
= + vk or = + v · grad , (3.5)
dt ∂t ∂xk dt ∂t
which can be applied to tensor fields of any order, Tij... = Tij... (xp , t), ex-
pressed in spatial coordinates.
The first term on the right-hand side of (3.4b) gives the rate of change
at a particular position and is accordingly called the local rate of change .
This term is sometimes written as [∂T (xi , t)/∂t]xi to emphasize that xi is
held constant in this differentiation. The second term on the right-hand side
of (3.4b) arises because the specific particles are changing their positions in
space. This term expresses the contribution due to the motion of the particles
in the variable field, and it is therefore called the convection rate of change .
The displacement vector u = x − a of a typical particle from its position
a in the reference configuration to its position x at current time t can be
represented as a function of L AGRANGian or E ULERian coordinates,

ui = ui (ap , t) = xi (ap , t) − ai , (3.6a)

ui = ui (xp , t) = xi − ai (xp , t) , (3.6b)

by taking (3.1a,b) into account. Partial differentiation of the displacement

vector (3.6a,b) with respect to the coordinates produces either the material
displacement gradient

∂ui /∂aj = ∂xi /∂aj − δij (3.7a)

or the spatial displacement gradient

∂ui /∂xj = δij − ∂ai /∂xj . (3.7b)

In a similar way we define the material and spatial deformation gradient

 3.1 Analysis of Deformation and Strain 35

(−1)
Fij := ∂xi /∂aj and Fij := ∂ai /∂xj , (3.8a,b)

respectively. The deformation gradient (3.8a) can geometrically be inter-

preted in the following way. Two neighbouring particles which occupy points
P0 and Q0 before deformation, move to points P and Q, respectively, in the
deformed configuration (Fig. 3.2).

path line
ds = F d s0

=
ds0

path line
t =0 t >0
Fig. 3.2 Geometrical interpretation of the deformation gradient (B ETTEN, 2001a)

From (3.1a) and (3.8a) the distance differential dxi (line element vector
ds) can be expressed as follows:

dxi = (∂xi /∂aj ) daj = Fij daj or ds = F ds0 . (3.9)

Thus, the deformation gradient F can be interpreted as a linear transforma-

tion or a second-rank tensor: the initial line element vector ds0 is mapped
onto the corresponding vector ds at time t, where a translation, a rigid ro-
tation , and a stretching of the line element is produced. This decomposition
of the total deformation corresponds with the polar decomposition theorem
which states that the non-singular deformation gradient (3.8) can be decom-
posed, uniquely, in either of the products

Fij = Rik Ukj , Fij = Vik Rkj , (3.10a,b)

i.e., in matrix notation: F = RU and F = V R, where R is an orthogo-

nal rotation tensor, and U , V are positive definite symmetric tensors which
are called right (U ) and left (V ) tensors (B ETTEN, 2001a) . However, the
”translation part” of the total deformation is not involved in (3.10a,b) since a
parallel transport does not change the cartesian components of a vector, here
36 3 Some Basic Equations of Continuum Mechanics

the line element vector (3.9). The shifter, which shifts a vector from one co-
ordinate system to another, is here merely δij . But in curvlinear coordinates,
parallel transport does change the covariant and contravariant components
of a vector (B ETTEN, 1987c) .
Although the deformation gradient tensor F plays a central role in the
analysis of deformation, it is not itself a suitable (direct) measure of strain ,
since a measure of strain must be unchanged in a rigid-body motion

xi = Rij aj + ci , (3.11)

where Rij is any orthogonal (rotation) tensor (Rik Rjk = δij ). The vector
ci is independent of position and depends only on time t. Thus, the material
deformation gradient (3.8a) of the rigid-body motion (3.11) is identical to
the rotation tensor: Fij = Rij .
The difference (ds)2 − (ds0 )2 for two neighbouring particles of a contin-
uum can be taken for a suitable measure of strain which occurs in the neigh-
bourhood of the particles between the initial and current configurations. If
this difference is identical to zero for all neighbouring particles of a mate-
rial, a rigid-body motion (3.11) with Rij Rik = δjk results in:

(ds)2 − (ds0 )2 ≡ dxi dxi − dai dai = (Rij Rik − δjk )daj dak = 0 . (3.12)

In general, we arrive from (3.9) with (3.8a) at the relation

(ds)2 − (ds0 )2 = (Fij Fik − δjk )daj dak ≡ 2λjk daj dak (3.13)

in which the second-order tensor

λij := (Fki Fkj − δij ) /2 ≡ (gij − δij ) /2 (3.14)

is known as the L AGRANGE finite strain tensor , and the tensor

Cij ≡ gij = Fki Fkj or C = F tF (3.15a,b)

is called the right C AUCHY-G REEN tensor . This tensor and K RONECKER’s
tensor may be interpreted as metric tensors (B ETTEN, 2001a), since the
squared lengths of the current and initial line-elements can be written as

(ds)2 = dxi dxi = gjk daj dak (3.16)

and
(ds0 )2 = dai dai = δjk daj dak , (3.17)
respectively.
 3.1 Analysis of Deformation and Strain 37

In terms of the displacements (3.6a), the L AGRANGE strain tensor (3.14)

takes the following form
 
1 ∂ui ∂uj ∂uk ∂uk
λij = + + , (3.18)
2 ∂aj ∂ai ∂ai ∂aj

if Fij from (3.8a) with (3.7a) is substituted into (3.14).

In contrast to the material description (3.13), the ”strain measure” can be
formulated with the spatial coordinates xi as independent variables (spatial
description) :
 
(−1) (−1)
(ds)2 − (ds0 )2 = δjk − Fij Fik dxj dxk ≡ 2ηjk dxj dxk , (3.19)

where ηjk is known as the E ULERian finite strain tensor . Contrary to (3.18),
this tensor takes the form
 
1 ∂ui ∂uj ∂uk ∂uk
ηij = + − (3.20)
2 ∂xj ∂xi ∂xi ∂xj

in terms of the displacements (3.6b). The L AGRANGE and E ULER strain

tensors and the corresponding metric tensors are listed in Table 3.1.

Table 3.1 Finite strain and metric tensors

L AGRANGE E ULER
deformed
ds2 = gij dai daj ds2 = δij dxi dxj
metric

continuum
undeformed
ds20 = δij dai daj ds20 = hij dxi dxj
continuum
deformed gij ≡ Cij = Fki Fkj
metric tensor

δij
continuum C = F tF
(−1) (−1) (−1)
undeformed hij ≡ Bij = Fki Fkj
δij
continuum B = FFt
1 1
λij = 2 (gij − δij ) ηij = 2 (δij − hij )
strain tensor
1 1 −1
λ= 2 (C − δ) η= 2 (δ − B )

The logarithmic strain tensor is not listed in Table 3.1 although it plays
also a central role in the theory of finite deformation because it can be
38 3 Some Basic Equations of Continuum Mechanics

decomposed into a sum of an isochoric distortion and a volume change

(B ETTEN, 2001a). This advantage is uilized, e.g., to describe the creep de-
formations of thick walled tubes subjected to internal pressure (Chapter 5).
The problem to represent logarithmic strain tensors as isotropic tensor
functions can be solved by using the interpolation method developed by
B ETTEN(1984; 1989; 2001a).
In (3.18), if the displacement gradient components ∂ui /∂aj are each
small compared to δij , the squares and products of these derivatives may
be neglected in comparison to the linear terms. The resulting tensor is the
L AGRANGE infinitesimal strain tensor :

ij = (∂ui /∂aj + ∂uj /∂ai ) /2 = (Fij + Fji ) /2 − δij . (3.21a)

Likewise for ∂ui /∂xj δij in (3.20), the product terms may be dropped to
yield the E ULERian infinitesimal strain tensor :

εij = (∂ui /∂xj + ∂uj /∂xi ) /2 ≡ (ui,j + uj,i ) /2 . (3.21b)

This tensor is sometimes called the classical strain tensor . In (3.21b) the no-
tation ui,j is adopted for the partial derivative ∂ui /∂xj . Such abbreviations
are often used in tensor analysis (B ETTEN, 1987c).
If both the displacement gradients (3.7a,b) and the displacements (3.6a,b)
themselves are small, there is very little difference in the material (ai ) and
spatial (xi ) coordinates of a material particle. Accordingly, the material dis-
placement gradient (3.7a) and spatial displacement gradient (3.7b) are very
nearly equal, so that the infinitesimal tensors (3.21a,b) may be taken as equal:

λij ≈ ηij ≈ ij ≈ εij .

In the elastic deformation of metals the small-strain theory is quite adequate,

whereas in rubber-like materials and some other synthetic plastics, for in-
stance, elastic deformation may be of much larger magnitude, requiring the
use of the finite-strain theory . In this theory, geometrical non-linearities are
considered expressed by the product terms in (3.18) and (3.20). The plastic
deformation of metals after yielding may also lead to large strains, but there
is usually a considerable range of deformation beyond yield in which small-
strain theories may still be used; in metal forming, where large deformations
occur, the small-strain components are of doubtful physical significance. In-
cremental theories of plasticity in effect and also creep theories (Chapter 4)
analyze these operations in terms of rate of deformation rather than strain.
 3.1 Analysis of Deformation and Strain 39

The rate-of-deformation tensor

dij = (vi,j + vj,i ) /2 , (3.22)

sometimes called ”rate-of-strain” or ”strain-rate tensor”, is linear in the ve-

locity gradients ∂vi /∂xj ≡ vi,j . This linearity is exact and no approxima-
tion has been made in deriving it (B ETTEN, 2001a). Furthermore, the tensor
(3.22) is not to be confused with the material time derivative ε̇ij ≡ dεij /dt
of the infinitesimal tensor (3.21b), because we have:

ε̇ij = dij − (ui,p vp,j + uj,p vp,i ) /2 . (3.23)

Only in the case of small displacement gradients (ui,j ≡ ∂ui /∂xj ) and small
velocity gradient tensors (vi,j ≡ ∂vi /∂xj ), we have: ε̇ij ≈ dij .
If the displacements ui are given, the six components εij = εji can be
easily calculated from (3.21b), provided that the displacements are differen-
tiable functions of the coordinates xi . However, if the strain components εij
are given explicitly as functions of the coordinates xi , the six independent
equations (3.21b) may be viewed as a system of six partial differential equa-
tions for determining the three displacement components ui . The system is
over-determined and will not, in general, possess a solution for an arbitrary
choice of the strain components εij . Therefore, if the displacement compo-
nents ui are single-valued and continuous, some conditions must be imposed
upon the strain components. The necessary and sufficient conditions for in-
tegrability can be found by eliminating the displacements ui in (3.21b):

εij,k + εk,ij − εik,j − εj,ik = 0ijk . (3.24)

Again the partial derivatives are denoted with a comma followed by the
indices of the independent variables (spatial coordinates xi ), for example:
εij,k ≡ ∂εij /∂xk ∂x . The conditions (3.24) are called compatibility equa-
tions of the infinitesimal strain tensor. There are 81 equations in all in (3.24),
but only six are distinct. These six conditions can be written in the compact
form
!
Rij := εipr εjqs εpq,rs = 0ij (3.25)
where Rij = Rji is called the incompatibility tensor (B ETTEN, 2001a), the
divergence of which is equal to the zero vector:

Rij,j = 0i . (3.26)

This condition is known as B IANCHI’s identity (B ETTEN, 2001a).

40 3 Some Basic Equations of Continuum Mechanics

Caution: The permutation tensor εijk in (3.25) is not to be con-

fused with the strain tensor εij and its partial derivatives εij,k .
For plane strain in the x1 -x2 -plane, x2 -x3 -plane, or x3 -x1 -plane the
six unique equations (3.26) reduce, respectivelely, to the following single
equtions
ε11,22 + ε22,11 = 2ε12,12 , ..., ε33,11 + ε11,33 = 2ε31,31
which can easily deduced from (3.21b) by eliminating the displacements ui
as has been pointed out by B ETTEN (2001a).

3.2 Analysis of Stress

In this section the definitions of stress vector and stress tensor will be given
and the equations of equilibrium will be derived. We will then show how the
stress components change when the frames of reference are changed from
one rectangular cartesian frame of reference to another. We will see that the
stress components transform according to the tensor transformation rules.
The symmetry of the stress tensor and its consequences will then be dis-
cussed.
Consider a configuration occupied by a loaded body B (Fig 3.3) at some
time.

x2
P
n
S

S B

x1
x3

Fig. 3.3 Stress principle

Imagine a closed surface S within B. We would like to know the inter-

action between the material exterior to this surface and that in the interior.
 3.2 Analysis of Stress 41

Let us consider a small surface element of area ΔS on our imaged surface

S. The unit vector n is normal to S with its direction outward from the in-
terior of S. Then we can distinguish the two sides of ΔS according to the
direction of n. Consider the part of material lying on the positive side of the
normal. This part exerts a force ΔP on the other part, which is situated on
the negative side of the normal. The force ΔP is a function of the area and
the orientation of the surface. We assume that the ratio ΔP /ΔS tends to a
definite limit as ΔS tends to zero:

p := lim (ΔP /ΔS) = dP /dS .

ΔS→0

This limiting vector is called the stress vector , or also known as the trac-
tion vector . We also assume that the moment of the forces acting on the
small surface ΔS about any point within the area vanishes in the limit, i.e.,
a couple-stress vector is not taken into account. Such couple stresses have
in fact been included in continuum mechanics, for instance by E. and F.
C OSSERAT in 1907. Materials in which there may be couple stresses are
called polar (→ C OSSERAT continuum) .
Now let us consider three surfaces parallel to the coordinate planes. The
normals of these surfaces are in the positive directions of the coordinate axes
as drawn in Fig. 3.4.

s12 2 e
2p 2n

1p
1p

1
n
s11 1e
3n

s133e
3p

Fig. 3.4 Traction vectors on three planes perpendicular to coordinate axes and notations of
stress components

Each of the three traction vectors in Fig. 3.4 can be decomposed in its
components parallel to the coordinate axes:
42 3 Some Basic Equations of Continuum Mechanics

1
p = σ11 1 e + σ12 2 e + σ13 3 e (3.27a)
2
p = σ21 1 e + σ22 2 e + σ23 3 e (3.27b)
3 1 2 3
p = σ31 e + σ32 e + σ33 e . (3.27c)

These equations can be written in the following compact form

i
p = σij j e , i = 1, 2, 3 , (3.27*)

where σij are the cartesian components of the stress tensor σ. As a rule
the first subscript on σij identifies the plane on which a stress vector is acting,
while the second index indicates the direction of the traction component:

σik = i p · k e . (3.28)

This relation immediately follows from (3.27*) since the unit base vectors
are mutually orthogonal: j e · k e = δjk . The components perpendicular to
the planes (σ11 , σ22 , σ33 ) are called normal stresses . Those acting in (tan-
gentially to) the plane (σ12 , σ13 , . . . , σ32 ) are called shear stresses . A
stress component is positive, if it is acting on a positive plane in the positive
direction of a coordinate axis. Likewise, a stress component is positive, if it
is acting on a negative plane in the negative direction. Otherwise, the stress
components are negative. A plane in Fig. 3.4 is said to be positive if its outer
normal points in one of the positive coordinate directions. Otherwise it is
said to be negative. In accordance with the above determination of signs, the
stress components indicated in Fig. 3.5 are all positive. Thus, a normal stress
is considered to be positive for tension and negative for compression.

Fig. 3.5 Positive stress components

 3.2 Analysis of Stress 43

The relationship between the stress tensor σ at some point and the stress
vector p on a plane of arbitrary orientation n at that point may be established
through the force equilibrium of an infinitesimal tetrahedron of the contin-
uum, having its vertex at the considered point. The base of this tetrahedron
is taken perpendicular to n, while the three faces are taken perpendicular to
the coordinate axes as illustrated in Fig. 3.6.

x2

pi

dx2 ni
x1
dx1
13

11
12 dS

n 1dS

dx3
x3

Fig. 3.6 Infinitesimal tetrahedron

Designating the area of the base as dS, the areas of the faces are the pro-
jected areas n1 dS, n2 dS, and n3 dS. The force equilibrium in x1 -direction
can then be expressed as follows:

p1 dS = σ11 n1 dS + σ21 n2 dS + σ31 n3 dS ,

so that, by way of cyclic permutations, we arrive at C AUCHY’s formula:

⎫
p1 = σ11 n1 + σ21 n2 + σ31 n3 ⎪ ⎬
p2 = σ12 n1 + σ22 n2 + σ32 n3 pi = σji nj . (3.29)
⎪
⎭
p3 = σ13 n1 + σ23 n2 + σ33 n3
This formula expresses the components of the stress vector p acting on an
arbitrarily oriented infinitesimal area n dS at a considered point in terms of
44 3 Some Basic Equations of Continuum Mechanics

the components of the stress tensor, σ, at the point. Thus, the traction p for
any n may be calculated from a knowledge of nine basic quantities σij .
The result (3.29) can be interpreted as follows: The quantity σij serves
as a linear operator which operates on the argument vector ni to produce
the image vector pi . Hence, σij are the cartesian components of a second-
order tensor σ, known as C AUCHY’s stress tensor . Its components transform
according to the rule (2.20a,b).
It can easily be shown (B ETTEN, 2001a) that C AUCHY’s stress tensor is
symmetric, σij = σji , so that only six of the nine components are specified
independently in order to define completely the state of stress at any point.
The symmetry of the stress tensor implies what is sometimes called the the-
orem of conjugate shear stresses , also known as B OLTZMANN’s axiom ,
which states that the shear stresses on perpendicular planes (having direc-
tions such that both stresses point either toward or away from the line of
intersection of the planes) are always equal in magnitude. This is not true in
the C OSSERAT continuum or in damaged materials B ETTEN (1982b; 2001a).
Furthermore, if we would like to study the influence of a strong elec-
tromagnetic field on the propagation of elastic waves, or such influence on
some high-frequency phenomenon in the material, then the stress level may
be very low and the body moment may be significant. In such problems the
stress tensor may not be assumed symmetric. Couple stresses and body cou-
ples are useful concepts in dealing with materials whose molecules have
internal structures, and in the dislocation theory of metals.
The symmetry property of C AUCHY’s stress tensor is an advantage in
view of several aspects, for instance in view of the principle stresses and
principal stress directions or with regard to the formulation of constitutive
equations.
Firstly, let us discuss the determination of the principle values and the
principal directions for the stress tensor. For that purpose, let us consider
Fig. 3.6, where, on the surface element dS, shear stresses are produced. On
those surface elements for which the vectors p and n are collinear , no shear
stresses can be produced. These surfaces are called principal planes ; their
normal directions and the normal stresses are known as principal directions
and principal stresses . If the vectors p and n are collinear, they differ only
in length. Thus, together with (3.29) and σij = σji , we find:

!
pi = σni ≡ σδij nj
⇒ (σij − σδij )nj = 0i . (3.30)
pi = σij nj
The condition for (3.30) to have non-trivial solutions for n is:
 3.2 Analysis of Stress 45

det(σij − σδij ) = 0 ⇒ σ 3 − J1 σ 2 − J2 σ − J3 = 0 . (3.31)

This is the characteristic equation (2.25) for the stress tensor, in which J1 ,
J2 , J3 are the three irreducible invariants (2.24a,b,c) of the stress tensor.
The three roots of (3.31) are the three principal stress values σI , σII , σIII .
Associated with each principal stress σα , α = I, II, III, there is a princi-
pal stress direction for which the direction cosines nαi are solutions of the
equations (3.30), where the eigenvectors nα , α = I, II, III, are normalized
without loss of generality; hence

(α) (α) (α)
σij − σ(α) δij nj = 0i ; nk nk =1, α = I, II, III . (3.32)

Caution: There is no sum on the repeated label α, which is there-

fore enclosed by parentheses.
From the symmetry property the following statements can be easily de-
duced:
1. The eigenvalues of a real symmetric second-order tensor are all
real.
2. The eigenvectors associated with two distinct eigenvalues of a
symmetric second-order tensor are orthogonal:
σα = σβ ⇒ nαk nβk = 0 , α = β .
Eigenvalue problems of tensors of order higher than two are discussed by
B ETTEN (1982a; 1998; 2001c), for instance.
In the following the equilibrium equations should be derived. Further-
more, the symmetry of C AUCHY’s stress tensor will be proved. We therefore
consider a material body (volume V , bounded by surface S) in equilibrium.
It may be subjected to a system of external (surface) forces p per unit area
and body forces b per unit mass (including inertia forces, if present) or vol-
ume forces f = ρb per unit volume as shown in Fig. 3.7.
Equilibrium of an arbitrary volume V of a continuum (Fig. 3.7) requires
that the resultant force and moment acting on the volume be zero. Summation
of surface and body forces results in the integral relation
&& &&&
pi dS + fi dV = 0i . (3.33)
S V

Replacing pi by (3.29) and converting the resulting surface integral to a vol-

ume integral by using the divergence theorem of G AUSS ,
46 3 Some Basic Equations of Continuum Mechanics

x2 V
dS p i dS

dV ni

f i dV
S

x1
x3

Fig. 3.7 Volume and surface forces

&& &&&
σji nj dS = σji,j dV , (3.34)
S V

equation (3.33) becomes:

&&&
(σji,j + fi ) dV = 0i . (3.35)
V

Since the volume V is arbitrary, the integrand in (3.35) must vanish, so that
we arrive at the equilibrium equations

σji,j + fi = 0i , (3.36)

where σji,j ≡ ∂σji /∂xj is the divergence of the stress tensor. The partial
differentiation is denoted with a comma followed by the index of the inde-
pendent variable.
For a moving continuum (mass density ρ; velocity field vi ≡ ẋi ) we have
to take inertial forces ,
&&&
dTi = −ẋi dm ⇒ Ti = − ρẍi dV , (3.37)
V

into account (D ’A LEMBERT’s principle) , so that the right-hand side in (3.36)

is not equal to zero:
σji,j + fi = ρẍi . (3.38)
 3.2 Analysis of Stress 47

These equations are known as C AUCHY’s equations of motion .

In the absence of couple stresses , the equilibrium of moments about the
origin (Fig. 3.7) requires that
&& &&&
εijk xj pk dS + εijk xj fk dV = 0i , (3.39)
S V

where εijk is the permutation tensor defined in (2.5) and xi is the position
vector of the surface and volume elements. Again, replacing pi by (3.29),
applying the divergence theorem of G AUSS, and using the equilibrium equa-
tions (3.36), the integrals in (3.39) are combined and reduced to
&&&
εijk σjk dV = 0i . (3.40)
V

Since the volume V is arbitrary, the integrand in (3.40) must vanish at any
point in the continuum:

εijk σjk = 0i ⇒ σij = σji . (3.41)

This result shows that C AUCHY’s stress tensor is symmetric .

The C AUCHY stress tensor field is defined as a function of the spatial
coordinates, σij = σij (xp , t), and the C AUCHY equations of motion (3.38)
apply to the current deformed configuration. A suitable strain measure to
use with the C AUCHY stress tensor would therefore be one of the strain
or deformation tensors of the E ULERian formulation in terms of the spatial
position in the deformed configuration. However, the L AGRANGE or mate-
rial formulation is often preferred in the finite theory of elasticity , where
a natural state exists to which the body will return when it is unloaded. If
we formulate a strain tensor in material coordinates, we need also to ex-
press the stresses as functions of material coordinates and derive constitutive
equations in the reference state. The two P IOLA -K IRCHHOFF stress tensors
discussed by B ETTEN (2001a) are two alternative definitions of stress in the
reference state. The first P IOLA -K IRCHHOFF stress tensor (sometimes called
the L AGRANGE stress tensor) has the disadvantage of being nonsymmetric.
It is therefore awkward to use it in constitutive equations with a symmetric
strain tensor. Furthermore, this tensor is not objective (B ETTEN, 2001a). The
second P IOLA -K IRCHHOFF tensor T̃ij is symmetric whenever the C AUCHY
stress tensor σij is symmetric (nonpolar case) and objective. Thus, this ten-
sor is preferred in the finite theory of elasticity (B ETTEN, 2001a). The rela-
tion between these two stress tensors can be expressed as follows:
48 3 Some Basic Equations of Continuum Mechanics

ρ0 (−1) ρ
T̃ij = F σpq ⇔ σij = Fijpq T̃pq (3.42)
ρ ijpq ρ0

where ρ and ρ0 are the mass densities of the current and reference configu-
rations, respectively. The fourth-order tensor in (3.42) is defined as:

Fijpq := (Fip Fjq + Fiq Fjp ) /2 (3.43)

and has the following symmetry properties:

Fijpq = Fjipq = Fijqp . (3.44)

Its components can be determined by inserting the given components (3.8a)

of the material deformation gradient. It can be shown that the inverse form
of (3.43), used in (3.42), can be represented as
 
(−1) (−1) (−1) (−1) (−1)
Fijpq = Fip Fjq + Fiq Fjp 2, (3.45)

(−1)
where Fij are the given components (3.8b) of the spatial deformation gra-
dient (B ETTEN, 2001a).
In the foregoing Sections 3.1 and 3.2 we have discussed strain and stress
tensors, respectively. Fomulating constitutive equations we have to select
appropiate pairs of strain and stress tensors. Admissible pairs are called con-
jugate variables. For instance, in the linear constitutive equation of the finite
theory of elasticity,
T̃ij = Eijk λk , (3.46)
the L AGRANGE finite strain tensor (3.14), (3.18) and the second P IOLA -
K IRCHHOFF tensor (3.42) are conjugate variables in the reference configu-
ration. Another pair of conjungate variables are the rate-of-deformation ten-
sor (3.22) and C AUCHY’s stress tensor (3.29) in the deformed configuration.
The stress power in a volume V of the current deformed configuration
can be expressed by the conjungate variables λ̇ and T̃ in the volume V0
occupied by the same material in the reference configuration according to
&&& &&& &&&
dij σji dV = Fjp Fiq dij T̃pq dV0 = λ̇ij T̃ji dV0 . (3.47)
V V0 V0

Further conjugate variables are derived by B ERTRAM (2005), B ETTEN

(2001a), H AUPT (2000), S KRZYPEK (1993), for instance.