Kinematics of Continua: OM X X X X X

Chapter 1
KINEMATICS OF CONTINUA
1.1. Material and Spatial Descriptions of Continuum Motion
1.1.1. Lagrangian and Eulerian Coordinates. The Motion Law. Let us

consider a continuum B. Due to Axiom 2, at time t=0 there is a one-to-one
◦
correspondence between every material point M ∈ B and its radius-vector x=
−−→
= OM in a Cartesian coordinate system Oēi . Denote Cartesian coordinates of
◦ ◦ ◦
the radius-vector by xi (x = xi ēi ) and introduce curvilinear coordinates X i of the
same material point M in the form of some differentiable one-to-one functions
◦ ◦
xi = xi (X k ). (1.1)
◦ ◦i
Since x = x ēi , the relationship (1.1) takes the form
◦ ◦
x = x(X k ). (1.2)
Let us fix curvilinear coordinates of the point M, and then material points
of the continuum B are considered to be numbered by these coordinates X i.
i
For any motion of the continuum B , coordinates X of material points are
considered to remain unchanged; they are said to be 'frozen' into the medium
and move together with the continuum. Coordinates Xi introduced in this way
for a material point M are called Lagrangian (or material).
Due to Axiom 3, at every time t there is a one-to-one correspondence be-
tween every point M∈B X i and its radius-vector
with Lagrangian coordinates
−−→ i i
x = OM with Cartesian coordinates x , where x and x depend on t. This
i
means that there is a connection between Lagrangian X , and the Cartesian x
i
coordinates of point M and time, i.e. there exist functions in the form (0.3)
xi = xi (X k , t) ∀X k ∈ VX . (1.3)
These functions determine a motion of the material point M in the Cartesian
coordinate system Oēi E3a .
of space The relationships (1.3) are said to be the law
of the motion of the continuum B .
Coordinates xi in (1.3) are called Eulerian (or spatial) coordinates of the
material point M.
i
Since x = x ēi and the coordinate system Oēi is the same for all times t, the
equivalent form of the motion law follows from (1.2):
x = x(X k , t). (1.4)

2 Chapter 1. Kinematics of Continua
Since the consistency conditions (0.4) must be satisfied, from (1.2) and (1.4)
we get the relationships
◦ ◦
x(X k , 0) = x(X k ), xi (X k , 0) = xi (X k ). (1.5)
Here the initial time t=0 is considered as the time t1 in (0.4), because just at
time t=0 we introduced Lagrangian coordinates Xi of point M.
Figure 1.1. The motion of a continuum: positions of continuum B and material point M in
reference and actual configurations
Unless otherwise stipulated, functions (1.3) are assumed to be regular in the

domain VX ⊂ R3 for all t, thus there exist the inverse functions
X k = X k (xi , t) ∀xi ∈ Vx ⊂ R3 .
◦
The closed domain V = W(B , 0) in a fixed coordinate system Oēi , which
is occupied by continuum B at the initial time t = 0, is called the reference
◦
configuration K, and the domain V = W(B, t) occupied by the same continuum
B at the time t > 0 is called the actual configuration K.
Figure 1.1 shows a geometric picture of the motion of a continuum from the
◦
reference configuration K to the actual one K at time t in space E3a .
It should be noticed that if the contin-
uum motion law (1.3) (or (1.4)) is known,
then one of the main problems of contin-
uum mechanics (to determine coordinates
of all material points of the continuum at
any time) will be resolved. However, in ac-
tual problems of continuum mechanics this
law, as a rule, is unknown and must be
found by solving some mathematical prob-
lems, whose statements are to be formu-
lated. One of our objectives is to derive
these statements.
Figure 1.2. Extension of a beam Example 1.1. Let us consider a continuum
B, which at time t=0 in the reference con-
1.1. Material and Spatial Descriptions 3
◦ ◦ ◦
figuration K is a rectangular parallelepiped (a beam) with edge lengths h1 , h2 and
◦
h3 , and in an actual configuration K at t>0 the continuum is also a rectangular
parallelepiped but with different edge lengths: h1 , h2 and h3 . We assume that
corresponding sides of both the parallelepipeds lie on parallel planes, and for
one of the sides, which for example is situated on the plane (x2 , x3 ), points
◦
of diagonals' intersection in K and in K are coincident (Figure 1.2). Then the
motion law (1.3) for this continuum takes the form
xα = kα (t) X α , α = 1, 2, 3, (1.6)
◦ ◦
i.e. coordinates xi , xi = X i of any material point M in K and K are proportional,
◦
and kα (t) = hα (t)/hα is the proportion function. The motion law (1.6) is called
the beam extension law.
◦
¤
Example 1.2. In K, let a continuum B be a
rectangular parallelepiped oriented as shown in
Figure 1.3; its motion law (1.3) has the form
 1
x = X + a(t)X ,
1 2
x =X ,
2 2
(1.7)
 3
x = X 3,
Figure 1.3. Simple shear of a
◦i beam
where x = X i, a(t) is a given function. In K
this continuum B has become a parallelepiped, all
cross-sections of which are planes orthogonal to
the Ox3 axis and are the same parallelograms.
This motion law is called simple shear; the tangent
of the shear angle α is equal to a. ¤
◦
Example 1.3. Consider a continuum B, which in K is a
rectangular parallelepiped (a beam) shown in Figure 1.4;
◦
under the transformation from K to K this parallelepiped
changes its dimensions without a change in its angles (as
in Example 1.1) and rotates by an angle ϕ(t) in the plane
Ox1 x2 around the point O (Figure 1.4). The motion law
for the continuum is called the rotation of a beam with Figure 1.4. Rotation of a
beam with extension
extension. In this case equations (1.1) have the form
◦ ◦
xi = F0 ij xj , xj = X j , (1.8)
where the matrix F0 ij is the product of two matrices, the rotation matrix O0 and
the stretch matrix U0 :
Ã !
cos ϕ − sin ϕ 0
F0 ij = O0 i k
k U0 j , O0 i
j = sin ϕ cos ϕ 0 ,
0 0 1
Ã ! Ã !
k1 0 0 k1 cos ϕ −k2 sin ϕ 0
i
U0 j = 0 k2 0 , F0 ij = k1 sin ϕ k2 cos ϕ 0 ,
0 0 k3 0 0 k3
and kα (t) = hα (t)/h0α are the proportion functions characterizing the ratio of
◦
lengths of the beam edges in K and K (as in Example 1.1). ¤
1.1.2. Material and Spatial Descriptions. In continuum mechanics, physical
processes occurring in bodies are characterized by a certain set of varying scalar
fields φ = φ(M, t), vector fields a = a(M, t), and tensor fields of the nth order
n Ω(M, t). We will consider tensors and tensor fields in detail in paragraph 1.1.4
(see also [12]).
Since in the Cartesian coordinate system Oēi a material point M corresponds
to both Lagrangian coordinates Xi and Eulerian coordinates xi , varying scalar
and vector fields can be written as follows:
e j , t) = φ(x
φ(X i , t) = φ(X i (xj , t), t) = φ(x e , t),
i j
a(X , t) = a e(x , t) = ae(x, t), (1.9)
With the help of the motion law (1.3) (or (1.4)), we can pass from functions
of Lagrangian coordinates to functions of Eulerian coordinates in formulae (1.9).
In continuum mechanics the tilde e is usually omitted (we will do this below).
For a fixed time t in (1.9), we obtain stationary scalar and vector fields.
M is fixed, and
If in (1.9) a material point time t changes within the interval
0 6 t 6 t0 , then we get an ordinary scalar function φ = φ(M, t) and vector
function a = a(M, t) depending on time.
According to relationships (1.9), there are two ways to describe different
physical processes in continua.
In the material (Lagrangian) description of a continuum, all tensor fields
describing physical processes are considered as functions of Xi and t.
In the spatial (Eulerian) description, all tensor fields describing physical
processes are functions of xi and t.
Both the descriptions are equivalent. It should be noted that for solids we
more often use the material description, where it is convenient to fix coordinates
Xi of a material point M and to observe its motion at different times t. For
gaseous and fluid continua, Eulerian description is more convenient; when an
observer fixes a geometric point with coordinates xi and monitor the material
i
points M passing through this point x at different times t.
◦
1.1.3. Local Bases in K and K. Using the motion law (1.4) and relationship
i
(1.1), at every material point M with coordinates X in the actual and reference
configurations we can introduce its local basis vectors:
◦ ◦
∂x ∂xi ◦ ∂x ∂ xi ◦
rk = = ēi = Qi k ēi , rk = = ēi = Q i
k ēi , (1.10)
∂X k ∂X k ∂X k ∂X k
where ◦ ◦
Qi k = ∂xi /∂X k , Qi k = ∂ xi /∂X k ,
◦ ◦
P ik = ∂X i /∂xk , P i k = ∂X i /∂ xk (1.11)
are Jacobian matrices and inverse Jacobian matrices.

◦
Here and below all values referred to the configuration K will be denoted by
superscript
◦ . As follows from the definition (1.11), local bases vectors r and
k
◦
rk are directed tangentially to corresponding coordinate lines X k (Figure 1.5)
Figure 1.5. Local basis vectors in reference and actual configurations
◦ ◦
In K and K introduce metric matrices gkl , g kl and inverse metric matrices
◦ kl
g kl , g as follows:
∂xi ∂xj
gkl = rk · rl = Qi k Qj l δij = δij ,
∂X k ∂X l
◦ ◦
◦ ◦ ◦ ∂ xi ∂ xj
g kl = rk · rl = δij , (1.12)
∂X k ∂X l
◦ ◦
g kl glm = δm
k
, g kl g lm = δm
k
,
and also vectors of reciprocal local bases

◦ ◦ ◦
ri = g im rm , ri = g im rm , (1.13)
which satisfy the reciprocity relations
◦ ◦
ri · rj = δi j , r i · r j = δi j , (1.14a)
and also the following relations:
q
√ ◦ ◦ ◦ ◦
rn × rm = g ²nmk rk , rn × rm = g ²nmk rk , (1.14b)
With the help of the mixed multiplication, i.e. sequentially applying scalar
and vector products to three different local bases vectors, we can determine the
◦
volumes |V | and |V | constructed by these vectors:
q q ¯ ¯
◦ ◦ ◦ ◦ ◦◦ ¯ ◦ ¯
|V | = r1 · (r2 × r3 ) = g = det (gij ) = ¯∂ xk /∂X i ¯,
√
|V | = r1 · (r2 × r3 ) = g = |∂xk /∂X i |. (1.15)
◦
It should be noted that although local bases ri and ri have been introduced
◦
in different configurations K and K, they correspond to the same coordinates Xi
(if one consider the same point M); therefore each of the bases can be carried
◦
as a rigid whole into the same point in K or in K. Due to this, we can resolve
any vector field a(M) i ◦i ◦

for each of the bases ri , r , r and ri :
◦ ◦ ◦ ◦
a = ai ri = ai ri = ai ri = ai ri . (1.16)
◦
If curvilinear coordinates Xi are orthogonal, then the vectors ri are orthog-
◦ ◦ ◦ ◦
onal as well: (ri · rj = δij ), and matrices
qg ij and g ij are diagonal; hence we
◦ ◦
can introduce Lam
e's coefficients H α = g αα (α = 1, 2, 3) and the physical
orthonormal basis
b◦ ◦ ◦ ◦ ◦
rα = rα /H α = rα H α . (1.17)
Components of a vector a with respect to this basis are called physical:
b◦ b◦
a = ai ri . (1.18)
◦
The actual basis ri is in general not orthogonal even if the basis ri is
orthogonal; therefore we cannot introduce the corresponding physical basis in K.
One can introduce a physical basis in K not with the help of ri , but with the
help of another special basis (see paragraph 1.1.7).
1.1.4. Tensors and Tensor Fields in Continuum Mechanics. For different

◦ ◦
local bases ri , ri , ri , ri or ēi at every point M, and with the help of formulae
given in the work [12] we can introduce different dyadic (tensor) bases: ri ⊗ rj ,
◦ ◦ ◦ ◦ ◦
ri ⊗ rj , ri ⊗ rj , ri ⊗ rj , ri ⊗ rj etc., which are equivalence classes of vector sets
consisting of 2 · 3 = 6 vectors (for example, r1 ⊗ rj = [r1 rj r2 0r3 0], where [ ] is
the notation of an equivalence class), and ⊗ is the sign of tensor product. A
field of second-order tensor T(M) can be represented as a linear combination of
dyadic basis elements:
◦ ◦ ◦ ◦ ◦ ◦
T = T ij ri ⊗ rj = T ij ri ⊗ rj = T̄ ij ēi ⊗ ēj = T ij ri ⊗ rj . (1.19)
During the passage from one basis to another, tensor components T ij are
transformed by the tensor law:
◦ ◦ ◦
T̄ ij = P ik P jl T kl = P i k P j l T kl . (1.20)
◦ ◦
Metric matrices g im , gim , g im and g im are components of the unit (metric)
tensor E with respect to different bases:
◦ ◦ ◦ ◦ ◦ ◦
E = gim ri ⊗ rm = g im ri ⊗ rm = g im ri ⊗ rm = g im ri ⊗ rm . (1.21)
For a second-order tensor T, in continuum mechanics one often uses
the transpose tensor TT = T ij rj ⊗ ri and the inverse tensor T−1 , where
T−1 · T = E. The inverse tensor exists only for a nonsingular tensor (when
det T 6= 0). The determinant of a tensor is defined by the determinant of its

mixed components matrix: det T = det T ij .
Besides second-order tensors, in continuum mechanics one sometimes uses
tensors of higher orders [12]. To introduce the tensors, we define polyadic bases
by induction: ri ⊗ . . . ⊗ rin ; the
1
bases are equivalence classes of vector sets
consisting of n · 3 = 3n vectors. A field of n-th order tensor
n Ω(M) can be
represented by a linear combination of polyadic basis elements:
◦
n ...in ◦ ◦
Ω = Ωi 1 ...in
ri ⊗ . . . ⊗ rin = Ωi
1
1
ri ⊗ . . . ⊗ rin ,
1
◦
where Ωi 1 ...in
and Ωi 1 ...in
are components of the n-th order tensor with respect
to the corresponding polyadic basis.
For fourth-order tensors, analogs of the tensor E are the first, second and
third unit tensors defined as follows:
∆I = ei ⊗ ei ⊗ ek ⊗ ek = E ⊗ E,
∆II = ei ⊗ ek ⊗ ei ⊗ ek , ∆III = ei ⊗ ek ⊗ ek ⊗ ei , (1.22)
and also the symmetric fourth-order unit tensor
1
∆ = (∆II + ∆III ). (1.22a)
2
1
∆ = ∆ijkl ei ⊗ ej ⊗ ek ⊗ el , ∆ijkl = (δ ik δ jl + δ il δ jk ).
2
We can transpose fourth-order tensors as follows:
4
Ω(m 1 m2 m3 m4 )
= Ωi i i i
1 2 3 4
rim ⊗ rim ⊗ rim ⊗ rim
1 2 3 4
,
where (m1 m2 m3 m4 ) is some substitution, for example,

4
Ω(4321) =
= Ωi1 i2 i3 i4 ri4
⊗ ri ⊗ ri ⊗ ri .
3 2 1
◦
1.1.5. Covariant Derivatives in K and K. Introduce the following nabla-
◦
operators in configurations K and K, respectively:
∂ ◦ ◦ ∂
∇ = rk ⊗ , ∇ = rk ⊗ . (1.23)
∂X k ∂X k
Applying the nabla-operators to a vector field, we get the gradients of a
◦
vector in K and K:
∂a
∇ ⊗ a = rk ⊗ = ∇k ai rk ⊗ ri , (1.24)
∂X k
◦ ◦ ∂a ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
∇ ⊗ a = rk ⊗ k
= ∇k ai rk ⊗ ri = ∇k ai rk ⊗ ri = ∇k ai rk ⊗ ri ,
∂X
where we have denoted the following covariant derivatives in different tensor
◦
bases in configurations K and K:
◦ ◦
◦ ◦ ∂ ai ◦ ◦ ◦ ◦ ∂ ai ◦ ◦
∇k ai = k
− Γm
ik am , ∇k ai = k
+ Γikm am (1.25)
∂X ∂X
∂ai ∂ai
∇k ai = − Γm
ik am , ∇k ai = + Γikm am .
∂X k ∂X k
◦ ◦
Here Γm
ij and Γm
ij are the Christoffel symbols in configurations K and K. For the
Christoffel symbols the following relations (see [12]) hold:
1 ∂gki
³ ∂g
∂g
´
Γm
ij = g
km
j
− ijk , kj
i
+
2 ∂X
Ã ◦ ∂X ∂X !
◦ ◦
◦ ◦
1 km ∂ g ∂ g ki ∂ g ij
Γmij = g kj
+ − . (1.26)
2 ∂X i ∂X j ∂X k
◦
Contravariant derivatives in K and K are introduced as follows:
◦ ◦ km ◦
k◦ ◦
∇ ai = g ∇m ai , ∇k ai = g km ∇m ai . (1.27)
The covariant derivatives (1.25) are components of the second-order tensors
◦
∇ ⊗ a and ∇ ⊗ a, therefore during the passage from the local basis ri to another
one they are transformed by the tensor law (1.20).
◦ ◦
The nabla-operators ∇ and ∇ in K and K can be applied to a tensor field
n Ω(X i ) of n-th order:
◦ ∂ nΩ ◦ ◦
n ◦ ...in ◦ k ◦ ◦
∇⊗ k
= ∇k Ωi
Ω = rk ⊗ 1
r ⊗ ri ⊗ . . . ⊗ rin ,
1
(1.28)
∂X
n k ∂ nΩ
∇⊗ Ω=r ⊗ k
= ∇k Ωi 1 ...in k
r ⊗ ri ⊗ . . . ⊗ rin ,
1
∂X
◦ ◦ ◦
where ∇k Ωi 1 ...in
and ∇k Ωi 1 ...in
are the covariant derivatives in K and K,
respectively:
n
X
◦ ◦ ∂ ◦i ◦ ◦
∇k Ωi 1 ...in
= Ω 1 ...in
+ Γimk
s
Ωi 1 ...is =m...in
. (1.29)
∂X k
s=1
In the same way we can define operations of scalar product of the nabla-operator
◦
in K (the divergence of a tensor):
◦
n ◦ ∂ nΩ ◦ ◦
...in ◦ ◦
∇· Ω = rk · k
= ∇k Ωki 2
ri ⊗ . . . ⊗ rin , (1.30)
∂X 2
◦
and vector product of the nabla-operator in K (the curl of a tensor):
◦ n ◦ ◦
n ◦ ∂ Ω 1 ◦ ◦ ◦
∇× Ω = rk × k
=q ²ijk ∇i Ωji2 ...in rk ⊗ ri ⊗ . . . ⊗ rin .
2
(1.31)
∂X ◦
g
1.1.6. The Deformation Gradient. Consider how a local neighborhood of a
◦
point M is transformed during the passage from configuration K to K. Take an
◦ ◦
arbitrary elementary radius-vector dx connecting in K two infinitesimally close
points M and M0 (Figure 1.6). In configuration K, these material points M and
M0 are connected by the elementary radius-vector dx.
Figure 1.6. Transformation of an elementary radius-vector during the passage from the reference
configuration to the actual one
◦
The vectors dx and dx can always be resolved for local bases:
◦
k ∂x ◦ ∂x ◦
dx(X ) = k
dX k = rk dX k , dx(X ) = k
k
dX k = rk dX k . (1.32)
∂X ∂X
◦
On multiplying the first equation by rm and the second by rm , we get
◦ ◦ ◦ ◦
rm · dx = rm · rk dX k = dX m , rm · dx = rm · rk dX k = dX m . (1.33)
Substitution of dX m (1.33) into the first equation of (1.32) yields
◦k ◦
dx = rk ⊗ r · dx. Changing the order of the tensor and scalar products (that is
◦
permissible by the tensor analysis rules), we get the relation between dx and
dx: ◦
dx = F · dx. (1.34)
Here we have denoted the linear transformation tensor
◦
F = rk ⊗ rk , (1.35)
called the deformation gradient. As follows from (1.34),
the deformation gradient connects elementary radius-
◦
vectors dx and dx of the same material point M in
◦
Figure 1.7. Geometric rep-
configurations K and K. resentation of the deforma-
Definition (1.19) allows us to give a geometric tion gradient
representation of the deformation gradient: if ri are
◦
considered as the left vectors and ri as the right vectors, then, by formulae
of paragraph 1.1.4 (see [12]), the tensor F takes the form
◦ ◦ ◦ ◦
F = ri ⊗ ri = [r1 r1 r2 r2 r3 r3 ].
According to the geometric definition of a tensor (see paragraph 1.1.4), the
tensor F can be represented as equivalence class of the ordered set of six vectors
◦
ri , ri (Figure 1.7).
Besides F, in continuum mechanics one often uses the transpose tensor FT ,

the inverse tensor F
−1 and the inverse to the transpose tensor F−1T :
◦ ◦ ∂x ◦ ◦
FT = rk ⊗ rk = rk ⊗ k
= ∇ ⊗ x, F−1 = rk ⊗ rk , (1.35a)
∂X
◦
−1T ◦ k T ∂x k ◦◦ k
F = (rk ⊗ r ) = r ⊗ rk = r ⊗ k
= ∇ ⊗ x.
∂X
It follows from (1.35) that
◦ ◦
F · ri = rk ⊗ rk · ri = rk δik = ri . (1.36)
i.e. the deformation gradient transforms local bases vectors of the same material
◦
point M from K to K.
Theorem 1.1. The transpose deformation gradient FT connects gradients of
◦
an arbitrary vector a in K and K:
◦ ◦
∇ ⊗ a = FT · ∇ ⊗ a, ∇ ⊗ a = F−1T · ∇ ⊗ a. (1.37)
H To derive formulae (1.37), we apply the definitions (1.24) and (1.35):
∂a ∂a ◦ ◦ ∂a ◦
∇ ⊗ a = ri ⊗ i
= rj δji ⊗ i
= rj ⊗ rj · ri ⊗ i
= F−1T · ∇ ⊗ a. N (1.38)
∂X ∂X ∂X
1.1.7. Curvilinear Spatial Coordinates. Notice that the choice of Cartesian
basis Oēi as a fixed (immovable) system in the spatial (Eulerian) description
of the continuum motion is not a necessary condition. For some problems of
continuum mechanics it is convenient to consider a moving system O0 ē0i with
0
−−→ 0 0
the origin at a moving point O (x0 = OO ) and a moving orthonormal basis ēi
(Figure 1.8), which is connected to ēi by the orthogonal tensor Q:
ē0i = Q · ēi . (1.39)
Figure 1.8. Moving bases ē0i and e

ri and curvilinear spatial coordinates ei
X in moving system O0 ē0i
In this case, instead of Cartesian coordinates xi of a point M in the basis ēi

ei in basis
one consider its Cartesian coordinates x ē0i :
x ei ē0i .
e = x − x0 = x (1.40)
Let Qi j be components of the tensor Q with respect to the basis ēi :

Q = Qi j ēi ⊗ ēj , (1.41)
then relation (1.39) takes the form
ē0i = Qj i ēj , (1.42)

and coordinates ei
x and xi are connected as follows:
e = x − x0 = (xi − xi0 )ēi = x

x ei Qj i ēj ,
ei ē0i = x
xi − xi0 = Qi j x
ej , (1.43)
i
∂x /∂e
x =Q j i
j, ∂e j
x /∂x = i
P ji .
Instead of Cartesian coordinates ei , we
x can consider special curvilinear coordi-
nates ek
X with the origin at point O :
0
ei = x
x e k ),
ei (X (1.44)
i
which, due to (1.27), are connected to x by the relations
xi = xi0 (t) + Qi j (t)e e k ) ≡ xi (X

xj (X e k , t) or ej = X
X e j (xi , t). (1.45)
The dependence on t xi0 (t) and Qi j (t)
in the relations is defined by functions
0 0
(i.e. only by the motion of system O ēi ), which are assumed to be known in
continuum mechanics.
Coordinates ei
X are no longer Lagrangian (material): at different times they
correspond to different material points. However, it is often convenient to choose
◦
coordinates ei
X coincident with Xi in the reference configuration K. In this case
we have the relations
◦
e j , 0).
xi (X i ) = xi (X j , 0) = xi (X (1.46)
With the help of transformation (1.45) we can use the spatial description in
coordinates ei
X as well when consider the functions
a = a(xi , t) = a e i , t);
e(X (1.47)
therefore coordinates ei
X are called curvilinear spatial coordinates.
Introduce local vectors
∂x ∂xj
e
ri = = ē . (1.48)
∂Xei ei j
∂X
In particular, the basis ē0i may be fixed (Figure 1.9), then ē0i = ēi , x
e = x,
and curvilinear spatial coordinates ej = X
X e j (xi ) are independent of t; the basis
e
ri is independent of t as well, and from (1.46) and (1.48) it follows that the basis
◦
coincides with ri : ◦
∂xj ∂ xj ◦
e
ri = ēj = ē = ri .
i j
(1.49)
e
∂X i ∂X
◦
When the basis ē0i is moving, bases e
ri and ri are no longer coincident.
Figure 1.9. Curvilinear spatial coordinates e i and Lagrangian

X coordinates Xi for the fixed basis
ē0i = ēi
The vectors e
ri are directed tangentially to the coordinate lines ei
X and defined
simultaneously with ri at every point M at any time t > 0.
A change of vectors e
ri in time is defined only by the motion of basis ē0i ,
because from (1.42), (1.43) and (1.48) it follows that
ek
∂X ek
∂xj ∂ X
³ ∂X
ek ´
ē0i = Qj i e
r k = e
rk = e
rk , (1.50)
∂xj xi ∂xj
∂e i
∂e
x
and the matrix Peki ≡ ∂ X

e k /∂e
xi is independent of t according to (1.44).
The bases vectors ri and eri are connected as follows:
∂x ∂x ∂ Xek ek
∂X
ri = i
= i
= e
rk . (1.51)
∂X ∂Xe k ∂X ∂X i
Just as in paragraph 1.1.2, we define the metric matrix geij and the inverse
metric matrix geij :
geij = e
ri · e
rj , geij gejk = δki , (1.52)
and the reciprocal basis vectors
ei j
∂X e i 0k
∂X
ri = geike
e rk = j
ē = ē . (1.53)
∂x xk
∂e
According to formulae (1.51) and (1.52), we find the relation between
matrices gij and gekl :
e k ∂X
∂X el e k ∂X
∂X el
gij = ri · rj = i
e
r ·e
j k
rl = gekl . (1.54)
∂X ∂X ∂X ∂X j
i
The inverse matrix g ij is found from (1.54) by the rule of matrix product
inversion (see Exercise 1.1.13):
∂X i ∂X j kl
g ij = ge . (1.55)
e k ∂X
∂X el
From (1.51), (1.53) and (1.55) we can find the relation between vectors of
reciprocal bases ri and ri :
e
em
∂X i ∂X j kl ∂ X ∂X i k
ri = g ij rj = ge j
e
rm = e
r . (1.56)
e ∂X
∂X k e l ∂X ek
∂X
Let there be a tensor
n Ω, then it can be resolved for the basis ri and for the
basis e
ri :
n
Ω = Ωi 1 ...in ei
ri ⊗ . . . ⊗ rin = Ω 1 ...in
ri ⊗ . . . ⊗ e
e rin . (1.57)
1 1
On substituting (1.51) into (1.57), we derive transformation formulae for

tensor components during the passage from coordinates Xi to e i:
X
ei
∂X 1 e in
∂X
ei
Ω 1 ...in
= Ωj 1 ...jn
... . (1.58)
∂X j 1
∂X jn
Introduce the nabla-operator e
∇ of covariant differentiation in coordinates e i:
X
e =e ∂
∇ ri (1.59)
ei
∂X
and contravariant derivatives of components e
ai of a vector a=e ri
aie in coordinates
e i:
X
e ke ∂e
ai eme e ke ai
∂e ei e
∇ ai = −Γ ik am , ∇ ai = +Γ m
km a . (1.60)
k e
∂X k e
∂X
The Christoffel symbols em
Γ in coordinates ei
X are connected to geij by the
ij
relations which are similar to (1.26).
Theorem 1.2. The results of covariant differentiation in coordinates ei
X and
X i (in the configuration K) are coincident:
e ⊗ n Ω,
∇ ⊗ nΩ = ∇ e · n Ω,
∇ · nΩ = ∇ e × n Ω.
∇ × nΩ = ∇ (1.61)
H Prove the first formula in (1.61). Due to (1.23), we have
∂ nΩ ∂X i k el
∂ nΩ ∂X ∂ nΩ
∇ ⊗ n Ω = ri ⊗ = e
r ⊗ = δ l k
r ⊗ e ⊗ n Ω. (1.62)
=∇
i e k e l i k e l
∂X ∂X ∂ X ∂X ∂X
The remaining two formulae in (1.61) can be derived in the same way (see
Exercise 1.1.8). N
Going to components of a tensor
n Ω with respect to bases ri and e
ri , from
(1.58) we get the relation between the covariant derivatives:
∇i Ωj 1 ...jn e iΩ
=∇ ej 1 ...jn
. (1.63)
Determine the tensor H transforming coordinates ei
X into Xj:
◦
H = rj ⊗ e e ij e
rj = H rj = H̄ ij ēi ⊗ ēj .
ri ⊗ e (1.64)
Then we get the relations (see Exercises 1.1.10 and 1.1.11):
◦ ◦
ri = H · e ej e
ri = H ri = H−1T · e e ij )−1e
ri = (H rj .
i rj , (1.65)
The coordinates ei
X e ri
ri and e
are often chosen orthogonal, then the bases
are orthogonal as well, and the matrices g
ij
eij and ge are diagonal; and we can
introduce the physical (orthonormal) basis:
b
e
rα = e eα,
rα /H (1.66)
p
where eα =
H geαα are Lam
e's coefficients, which are in general not coincident
◦
q
◦
with the coefficients Hα = g αα . Tensor components with respect to the basis
b
e
rα are called physical:
b
T = Teij b
ri ⊗ b
e e
rj . (1.67)
Relations between physical and covariant components of a tensor are determined
by the known formulae (see [12]).
Exercises for 1.1.
Exercise 1.1.1. With the help of formulae (1.10), (1.12), (1.13) and (1.17) show that if
the motion law of a continuum describes extension of a beam (1.6) (see Example 1.1),
◦
then the local basis vectors ri and the metric matrices have the forms
◦ ◦
ri = ei , r i = ei ,
rα = kα ēα , rα = (1/kα )ēα , α = 1, 2, 3,
◦ ◦ ij ij
g ij = δij , g =δ ,
 2   −2 
k1 0 0 k1 0 0
(gij ) =  0 k22 0  , (g ij ) =  0 k2−2 0 ,
k32 −2
0 0 0 0 k3
i.e. √
gαβ = kα2 δαβ , g αβ = kα−2 δαβ , Hα = gαα = kα , b
rα = eα .
Exercise 1.1.2. Show that if the motion law of a continuum describes a simple shear
(see Example 1.2), then the local basis vectors and the metric matrices have the forms
◦ ◦ ◦ ◦
r i = ei , r i = ei , g ij = δij , g ij = δ ij ,
r1 = ē1 , r2 = aē1 + ē2 , r3 = ē3 ,
r = ē − aē ,
1
r = ē ,
1
r3 = ē3 ,
2 2 2
Ã ! Ã !
1 a 0 1+a
2
−a 0
ij
gij = a 1 + a 0 ,
2
g = −a 1 0 .
0 0 1 0 0 1
Exercise 1.1.3. Show that if the motion law describes rotation of a beam with extension
(see Example 1.3), then with introducing the rotation tensor O0 and the stretch tensor
U0 :
X
3
O0 = O0i j ēi ⊗ ēj , U0 = kα ēα ⊗ ēα ,

α=1
we can rewrite the beam motion law in the tensor form
◦
x = F0 · x, F0 = O0 · U0 .
Show that the local basis vectors and metric matrices for this problem have the forms
◦
ri = F0 ki ēk , ri = ēi ,
 2 
k1 cos2 ϕ + k22 sin2 ϕ (k12 − k22 ) cos ϕ sin ϕ 0
gij = F0 ki F0 lj δkl = (k12 − k22 ) cos ϕ sin ϕ k12 sin2 ϕ + k22 cos2 ϕ 0 ,
0 0 k 2
3
g = k1 k2 k3 ,
 
k2 sin ϕ + k1 cos2 ϕ (k1−2 − k2−2 ) cos ϕ sin ϕ
−2 2 −2
0
g ij 
= (k1−2 − k2−2 ) cos ϕ sin ϕ k2−2 cos2 ϕ + k1−2 sin2 ϕ 0 .
−2
0 0 k3
Exercise 1.1.4. Using the property (1.14) of reciprocal basis vectors, show that the
following relations hold:
◦ ∂X i ∂X i k
ri = ◦k ēk , ri = P ik ēk = ē .
∂x ∂xk
Exercise 1.1.5. Show that F, FT , F−1 and F−1T in the Cartesian coordinate system
take the forms
∂xm ∂X k ∂xm ∂xm

F= ◦i ēm ⊗ ēi , FT = ◦i ēi ⊗ ē = ◦ i ēi ⊗ ēm ,
k m
∂x ∂x ∂X ∂x
◦ ◦
∂ xm ∂X k i ∂ xm
F−1 = ē ⊗
k m i
ē = ēm ⊗ ei ,
∂X ∂x ∂xi
◦ ◦
∂X k i ∂ xm ∂ xm i
F−1T = i
ē ⊗ ē =
k m
ē ⊗ em .
∂x ∂X ∂xi
Exercise 1.1.6. Substituting (1.54), (1.52) and (1.55) into (1.12), derive formula (1.55).
Exercise 1.1.7. Prove that

◦
ri = F−1T · ri .
Exercise 1.1.8. Derive the third formula of (1.61).
◦
Exercise 1.1.9. Prove that for any scalar function ϕ(X i ) its gradients in K and K are
connected by the relationship
◦
∇ϕ = F−1T · ∇ϕ.
Exercise 1.1.10. Show that formulae (1.65) follow from (1.64).
Exercise 1.1.11. Using (1.47), show that in formulae (1.64) the tensor H has the
following components with respect to bases ēi and e
ri :
◦ ◦
ek
∂ xi ∂ X ∂xi ∂X k k ei ∂xk ∂X i
H̄ ij = (H̄ ij )−1 = e ij = ∂ x ∂ X ,
H e ij )−1 =
(H
j
, ◦j ,
e k ∂x j ◦k .
e j ∂x
∂X k ∂x ∂X
k
∂X ∂x ∂X
◦
Exercise 1.1.12. Introducing the notation F ij for components of the deformation
◦ ◦ ◦
ij ◦ ◦ ◦ ◦
gradient F with respect to basis ri : F = F ri ⊗ rj = F i j ri ⊗ rj , show that formula
(1.36) yields
◦ ◦
rj = F i j ri .
Exercise 1.1.13. Show that the Levi-Civita symbols are connected by the relations
²ijk ²ijk = 6, ²ijk ²ilm = δjl δkm − δkl δjm , ²ijk ²ijl = 2δil ,
√ √
g ²ijk = (1/ g ) ²mnl gmi gnj glk . ²ijk T jk = 0
where T jk are components of an arbitrary symmetric tensor: T jk = T kj .
Exercise 1.1.14. Using relations (1.14a), show that the local bases vectors are
connected by the relations
q
√ ◦ ◦ ◦ ◦
rα × rβ = g rγ , rα × rβ = g rγ , α 6= β 6= γ 6= α.
Exercise 1.1.15. Show that the unit fourth-order tensors ∆I , ∆II and ∆III defined by
formulae (1.22) have the following properties:
∆I · · T = I1 (T)E, I1 (T) = T · · E, ∆II · · T = TT ,

1
∆III · · T = T, ∆ · · T = (T + TT ),
2
and
∆I · · 4 Ω = E ⊗ E · · 4 Ω, ∆II · · 4 Ω = Ω(2134) ,
1
∆III · · 4 Ω = 6 Ω, ∆ · · 4 Ω = (Ω(2134) + Ω),
2
for arbitrary second-order tensor T and fourth-order tensor Ω. As follows from these
4
formulae, the tensor ∆III is the `true' unit fourth-order tensor.
Exercise 1.1.16. Show that components of the symmetric unit fourth-order tensor ∆
with respect to a tetradic basis have the form
1
∆ = (ei ⊗ el ⊗ ei ⊗ el + ei ⊗ el ⊗ el ⊗ ei ) = ∆ijkl ei ⊗ ej ⊗ ek ⊗ el .
2
1
∆ijkl = (δ ik δ jl + δ il δ jk ).
2
Exercise 1.1.17. Show that for any second-order tensor T and for any vector a the
following formula of covariant differentiation hold:
∇ · (T · a) = T · · (∇ ⊗ a)T + a · ∇ · T.
1.2. Deformation Tensors and Measures
1.2.1. Deformation Tensors. Besides F, important characteristics of the

motion of a continuum are deformation tensors, which are introduced as follows:
1 ◦ ◦ ◦ ◦ ◦
C = (gij − g ij )ri ⊗ rj = εij ri ⊗ rj ,
2
1 ◦
A = (gij − g ij )ri ⊗ rj = εij ri ⊗ rj , (2.1)
2
1 ◦ ◦ ◦ ◦ ◦
Λ = (g ij − g ij )ri ⊗ rj = εij ri ⊗ rj ,
2
1.2. Deformation Tensors and Measures 17
1 ◦
J = (g ij − g ij )ri ⊗ rj = εij ri ⊗ rj .
2
Here C is called the right CauchyGreen deformation tensor, A the left

Almansi deformation tensor, Λ the right Almansi deformation tensor, and
J the left CauchyGreen tensor.
As follows from the definition of the tensors, covariant components of C and
A are coincident, but they are defined with respect to different tensor bases.
Components εij are called covariant components of the deformation tensor.
Contravariant components of the tensors Λ and J are also coincident and
called contravariant components εij of the deformation tensor, but they are
defined with respect to different tensor bases of the tensors Λ and J.
Notice that the deformation tensor components
1 ◦ 1 ◦
εij = (gij − g ij ), εij = (g ij − g ij ), (2.2)
2 2
have been defined independently of each other, therefore the formal rearrange-
ment of indices is not permissible for these components, i.e.
ε̆kl = εij g ik g jl 6= εkl , ε̆kl = εij gik gjl 6= εkl . (2.3)

Thus, when there is a need to obtain contravariant components from εij and
covariant components from εij , one should use the notation ε̆kl and ε̆kl . We will
also use the notation
◦ ◦ ◦ ◦ ◦ ◦
εkl = εij g ik g jl , εkl = εij g ik g jl . (2.4)
Theorem 1.3. The deformation tensors C, A, Λ and J are connected to the
deformation gradient F as follows:
1 1
C = (FT · F − E), A = (E − F−1T · F−1 ),
2 2
1 1
Λ = (E − F−1 · F−1T ), J = (F · FT − E). (2.5)
2 2
H Let us derive a relation between C and F. Having used the definitions of gij ,
◦
g ij and F, we get
1
³ ◦ ◦
´ 1 ³◦ ◦
´ 1
C= (ri · rj )ri ⊗ rj − E = ri ⊗ ri · rj ⊗ rj − E = (FT · F − E).
2 2 2
(2.6)
The remaining relations of (2.5) can be proved in the same way:
1 ◦ ◦ 1
A = (E − ri ⊗ ri · rj ⊗ rj ) = (E − F−1T · F−1 ),
2 2
1 ◦ ◦ 1
Λ = (E − ri ⊗ ri · rj ⊗ rj ) = (E − F−1 · F−1T ), (2.6a)
2 2
1 ◦ ◦ 1
J = (ri ⊗ ri · rj ⊗ rj − E) = (F · FT − E). N
2 2
1.2.2. Deformation Measures. Besides the deformation tensors, we define

deformation measures: the right CauchyGreen measure G and the left Almansi
measure g:
◦ ◦
G = gij ri ⊗ rj = FT · F = E + 2C,
◦
g = g ij ri ⊗ rj = F−1T · F−1 = E − 2A, (2.7)
and also the left CauchyGreen measure g−1 and the right Almansi measure
G−1 : ◦
g−1 = g ij ri ⊗ rj = F · FT = E + 2J,
◦ ◦
G−1 = g ij ri ⊗ rj = F−1 · F−1T = E − 2Λ. (2.8)
1.2.3. Displacement Vector. Introduce a displacement vector u of a point M

from the reference configuration to the actual one as follows (Figure 1.10):
◦
u = x − x. (2.9)
Figure 1.10. The displacement vector of a point M from the reference configuration to the actual
one
Theorem 1.4. The deformation tensors and the deformation gradient are
connected to the displacement vector u by the relations
◦
F = E + (∇ ⊗ u)T , F−1 = E − (∇ ⊗ u)T ,
◦
FT = E + ∇ ⊗ u, F−1T = E − ∇ ⊗ u, (2.10)
and also µ ¶
1 ◦ ◦ ◦ ◦
T T
C= ∇⊗u+∇⊗u +∇⊗u·∇⊗u ,
2
1 ¡ ¢
A= ∇ ⊗ u + ∇ ⊗ uT − ∇ ⊗ u · ∇ ⊗ uT , (2.11)
2
¡
1 ¢
Λ= ∇ ⊗ u + (∇ ⊗ u)T − ∇ ⊗ uT · ∇ ⊗ u ,
2
µ ¶
1 ◦ ◦ ◦ ◦
T T
J= ∇⊗u+∇⊗u +∇⊗u ·∇⊗u .
2
H The definition (2.9) of the displacement vector and the properties (1.35) of the
deformation gradient yield
◦
◦ ◦ ◦ ◦ ∂x ◦
FT = ∇ ⊗ x = ∇ ⊗ (x + u) = ri ⊗ i
+∇⊗u=
∂X
◦ ◦ ◦ ◦
= ri ⊗ ri + ∇ ⊗ u = E + ∇ ⊗ u. (2.12)
C takes the form

Then the tensor
µ ¶
1 ◦ ◦
T
C= (E + ∇ ⊗ u) · (E + ∇ ⊗ u ) − E =
2
µ ¶
1 ◦ ◦ ◦ ◦
T T
= ∇⊗u+∇⊗u +∇⊗u·∇⊗u . (2.13)
2
In a similar way, we can prove the remaining relations of the theorem. N

1.2.4. Relations between Components of Deformation Tensors and Dis-
placement Vector. The displacement vector u can be resolved for both bases
◦
ri and ri : ◦ ◦
u = ui ri = ui ri . (2.14)
The derivative with respect to Xi can be determined in both the bases as
well: ◦ ◦ ◦
∂u
= ∇i uk rk = ∇i uk rk . (2.15)
∂X i
Then the displacement vector gradients take the forms
◦ ∂u
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
∇ ⊗ u = ri ⊗ i
= ∇i uk ri ⊗ rk = ∇i uk ri ⊗ rk , (2.16)
∂X
∂u
∇ ⊗ u = ri ⊗ = ∇i uk ri ⊗ rk = ∇i uk ri ⊗ rk . (2.17)
∂X i
Substitution of these expressions into (2.10) gives
◦ ◦ ◦ ◦ ◦ ◦ ◦
F = (δik + ∇i uk )rk ⊗ ri = F ki rk ⊗ ri . (2.18)
Here we have introduced components of the deformation gradient in the reference
configuration:
◦ ◦ ◦
F ki = δik + ∇i uk . (2.19)
T
The transpose gradient F has the components
◦ ◦ ◦ ◦ ◦ ◦
FT = F ki ri ⊗ rk = F i k rk ⊗ ri , (2.20)
◦ ◦ ◦
F i k = δik + ∇k ui , (2.21)
◦ ◦
where (F i k )T = F ki .
In a similar way, one can find the expression for the inverse gradient
F−1 = (δik − ∇i uk )rk ⊗ ri = (F −1 )ki rk ⊗ ri (2.22)

and for the inversetranspose gradient
F−1T = (F −1 )ki ri ⊗ rk = (F −1 )i k rk ⊗ ri , (2.23)

where their components with respect to the actual configuration are expressed
as follows:
(F −1 )i k = δik − ∇k ui , (2.24)
(F −1 )ki = δik − ∇i uk . (2.25)
Thus, we have proved the following theorem.
Theorem 1.5. Components of the deformation gradients F, FT , F−1 and F−1T
◦
in local bases of configurations K and K are connected to components of the
displacement vector u by relations (2.19), (2.20), (2.24) and (2.25).
On substituting formulae (2.16) and (2.17) into (2.11) for C and A and
comparing them with (2.1), we get
1 ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
εij = (∇i uj + ∇j ui + ∇i uk ∇j uk ),
2
1
εij = (∇i uj + ∇j ui − ∇i uk ∇j uk ), (2.26)
2
the expressions for covariant components of the deformation tensor in terms

◦
of components of the displacement vector with respect to K and K.
In a similar way, substituting (2.16) and (2.17) into (2.11) for Λ and J, we
obtain
1 ◦ i ◦j ◦ ◦ ◦
k ◦i
◦ ◦
εij = (∇ u + ∇ u + ∇ u ∇k uj ), j i
2
1
ε = (∇i uj + ∇j ui − ∇k ui ∇k uj )
ij
(2.27)
2
the relations between contravariant components of the deformation tensor and

◦
components of the displacement vector in K and K.
Then with using relations (2.2), (2.26) and (2.27), we can find the connection
between the metric matrices and displacement components:
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
gij = g ij + ∇i uj + ∇j ui + ∇i uk ∇j uk = g ij + ∇i uj + ∇j ui − ∇i uk ∇j uk , (2.28)
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
g ij = g ij + ∇i uj + ∇j ui + ∇k ui ∇k uj = g ij + ∇i uj + ∇j ui − ∇k ui ∇k uj . (2.29)
Theorem 1.6. Components of the deformation tensor εij , εij and metric ma-
trices gij , g
ij are connected to components of the displacement vector u by
relations (2.26)(2.29).
1.2.5. Physical Meaning of Components of the Deformation Tensor. Let

us clarify now a physical meaning of components of the deformation tensor:
1 ◦ 1 ◦ ◦
εij = (gij − g ij ) = (ri · rj − ri · rj ). (2.30)
2 2
By the definition of the scalar product (see [12]), we have

µ ¶
1 ◦ ◦ ◦
εαβ = |rα ||rβ | cos ψαβ − |rα ||rβ | cos ψ αβ , (2.31)
2
◦ ◦ ◦
where ψαβ and ψ αβ are the angles between basis vectors rα , rβ and rα , rβ in K
◦
and K, respectively.
◦ ◦
Consider elementary radius-vectors dx and dx in configurations K and K,
◦
and introduce their lengths ds and ds, respectively:
◦ ◦ ◦
ds2 = dx · dx, ds2 = dx · dx. (2.32)
◦
Since dx is arbitrary, we can choose it to be oriented along one of the basis
◦
vectors rα . Then dx will be directed along the corresponding vector rα as well,
◦
because under this transformation rα becomes rα for the same material point M
k
with Lagrangian coordinates X . In this case we have
¯ ◦ ¯
◦ ◦ ¯ ∂x ¯ ◦
|dx| = dsα = ¯ α dX α ¯ = |rα | dX α ,
∂X
¯ ¯
¯ ∂x ¯
|dx| = dsα = ¯ α dX α ¯ = |rα | dX α . (2.33)
∂X
Hence
◦ ◦
dsα /dsα = |rα |/|rα | = δα + 1, (2.34)
where δα is called the relative elongation. Formula (2.34) yields
◦
|rα | = |rα |(1 + δα ). (2.35)
On substituting this expression into (2.31), we get
µ ¶
1 ◦ ◦ ◦
εαβ = |rα ||rβ | (1 + δα )(1 + δβ ) cos ψαβ − cos ψ αβ . (2.36)
2
◦
Consider the case when α = β, then ψαβ = ψ αβ = 0 and
1 ◦ ¡ ¢ g◦ ¡ ¢
εαα = |rα |2 (1 + δα )2 − 1 = αα (1 + δα )2 − 1 . (2.37)
2 2
◦
Let coordinates Xi be coincident with Cartesian coordinates xi , then g αβ =
= δαβ ; and for infinitesimal values of the relative elongation, when δα ¿ 1, we
obtain
εαα ≈ δα , (2.38)
i.e. εαα is coincident with the relative elongation.
In general, εαα is a nonlinear function of corresponding elongations.
◦
Consider α 6= β and assume that X i = xi , then ψ αβ = π/2; and from (2.36)
we get
1 ◦ ◦
εαβ = |rα ||rβ |(1 + δα )(1 + δβ ) cos ψαβ =
2
q q
1 ◦ ◦ 1
= g αα g ββ (1 + δα )(1 + δβ ) sin χαβ = (1 + δα )(1 + δβ ) sin χαβ , (2.39)
2 2
◦
where χαβ = ψ αβ − ψαβ = (π/2) − ψαβ is the change of the angle between basis
vectors rα and rβ . For small relative elongations, when δα ¿ 1, and small angles
χαβ ¿ 1, from (2.39) we get
εαβ ≈ χαβ /2, (2.40)
i.e. εαβ is a half of the misalignment angle of the basis vectors.
1.2.6. Transformation of an Oriented Surface Element. In actual configu-

ration K consider a smooth surface Σ, which contains two coordinate lines Xα
β
and X .
Then we can introduce the normal n to the surface
Σ as follows:
1
n = p rα × rβ . (2.41)
ge
Here ge = det (e
gαβ ), and geαβ is the two-dimensional
matrix of the surface (α, β = 1, 2):
geαβ = rα · rβ (2.42)
(it is not to be confused with the metric matrix gij =
Figure 1.11. Introduction of
oriented surface element
= ri · rj ).
n dΣ In configuration K consider a surface element dΣ
constructed on elementary radius-vectors dxα , which
are directed along local basis vectors, i.e. dxα = rα dX α (Figure 1.11). The value
p
dΣ = ge dX α dX β (2.43)
is called the area of the surface element dΣ constructed on vectors dxα and
dxβ . Then formula (2.41) takes the form
n dΣ = rα dX α × rβ dX β = dxα × dxβ , (2.44)

where n dΣ is called the oriented surface element.
Show that the normal n defined by formula (2.41) is a unit vector. According
to the property (1.14b) of the vector product of basis vectors and the results of
Exercise 1.1.13, we can rewrite equation (2.44) in the form
√
n dΣ = rα × rβ dX α dX β = g ²αβγ rγ dX α dX β =
√
= (1/ g )²ijk gαi gβj rk dX α dX β (2.45)
(there is no summation over α, β ). Thus,
√ 1
n dΣ · n dΣ = g ²αβγ rγ · √ ²ijk gαi gβj rk )(dX α dX β )2 =
g
= (²αβk ² ijk
gαi gβj )(dX α dX β )2 = (δαi δβj − δβi δαj )gαi gβj (dX α dX β )2 =
2
= (gαα gββ − gαβ )(dX α dX β )2 = ge(dX α dX β )2 = dΣ2 . (2.46)
Hence, n · n = 1.
◦ ◦
The surface element dΣ in K corresponds to the surface element dΣ in K,
◦ ◦
which is constructed on elementary radiusvectors dxα and dxβ :
◦ ◦ ◦ ◦ ◦ ◦
n dΣ = rα dX α × rβ dX β = rα × rβ dX α dX β . (2.47)
◦ ◦
Here n is the unit normal to
◦γ
dΣ.
Since rγ = F−1T · r , we get
q
√ ◦ ◦ ◦ ◦
n dΣ = g ²αβγ F−1T · rγ dX α dX β = g/g F−1T · rα × rβ dX α dX β =
q
◦ ◦ ◦
= g/g F−1T · n dΣ. (2.48)

◦ ◦ ◦
Theorem 1.7. The oriented surface elements n dΣ and n dΣ in K and K are
connected by the relation
q ◦
q ◦
◦ ◦ ◦ ◦
n dΣ = g/g n · F−1 dΣ = g/g F−1T · n dΣ. (2.49)
With the help of the deformation measures we can derive formulae connecting
◦
the normals n and n to the surface element containing the same material points
◦
in K and K.
Multiplying the equation (2.49) by itself and taking the formula n·n= 1
into account, we get
g ◦ ◦ ◦ g ◦ ◦ ◦
dΣ2 = ◦ (n · F−1 · F−1T · n) dΣ2 = ◦ (n · G−1 · n) dΣ2 . (2.50)
g g
Thus,
◦
q
◦ ◦ ◦
dΣ/dΣ = g/g (n · G−1 · n)1/2 . (2.51)
◦
On the other hand, expressing n from (2.49) and then multiplying the
obtained relation by itself, we obtain
◦ ◦ ◦
g g
d Σ2 = (n · F · FT · n) dΣ2 = (n · g−1 · n) dΣ2 , (2.52)
g g
Thus, we find that
◦
q
◦
dΣ/dΣ = g/g (n · g−1 · n)1/2 . (2.53)
On introducing the notation

◦ ◦ ◦ ◦ ◦
k = (n · G−1 · n)1/2 = (n · F−1 · F−1T · n)1/2 ,
k = (n · g−1 · n)1/2 = (n · F · FT · n)1/2 , (2.54)
relations (2.52) take the form
◦
q q ◦
◦ ◦
dΣ/dΣ = g/g k = g/g (1/k). (2.55)
Thus, we get
◦
k = 1/k. (2.56)
Substitution of (2.52) and (2.53) into (2.49) gives the desired relations
◦ ◦ ◦
kn = F−1T · n, k n = FT · n. (2.57)
1.2.7. Representation of the Inverse Metric Matrix in terms of Components
of the Deformation Tensor. Components gij of the metric matrix are connected
to components of the deformation tensor εij by relation (2.2). In continuum
mechanics, one often needs to know the expression of the inverse metric matrix
g ij in terms of εij (but not in terms of εij ). To derive this relation, we should
use the connection between components of a matrix and its inverse (see [12]):
1
g ij = ²imn ²jkl gmk gnl . (2.58)
2g
◦
For g ij , we have the similar formula
◦ 1 imn jkl ◦ ◦
g ij = ◦² ² g mk g nl . (2.59)
2g
◦
On substituting the relations (2.2) between gmn , g mn and εmn into (2.59), we
get
1 ◦ ◦
g ij = ²imn ²jkl (g mk + 2εmk )(g nl + 2εnl ) =
2g
1 ◦ ◦ ◦ ◦
= ²imn ²jkl (g mk g nl + 2g mk εnl + 1g nl εmk + 4εmk εnl ). (2.60)
2g
Removing the parentheses, modify four summands in (2.60) in the following way.
The first summand with taking formula (2.59) into account gives the matrix
◦ ◦
g ij (g/g). To transform the second and the third summands, we should use the
formulae q q
◦ jkl ◦ ◦ ◦ ◦
(1/ g )² = g ²tsp g jt g ks g lp . (2.61)
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
(1/g)²imn ²jkl g mk = ²imn ²tsp g jt g ks g lp g mk = ²imn ²tmp g jt g lp =
◦ ◦ ◦ ◦ ◦ ◦
= (δti δpn − δpi δtn )g jt g lp = g ij g nl − g jn g il . (2.62)
◦ ◦ ◦ ◦ ◦ ◦
(1/g)²imn ²jkl g mk εnl = (g ij g nl − g jn g il )εnl . (2.63)
Formula (2.61) follows from the relation
√ √
g ²ijk = (1/ g ) ²mnl gmi gnj glk
(see [12]), and relationship (2.62) has been obtained by using formula (2.61) and
the properties of the Levi-Civita symbols (see Exercise 1.1.13).
On substituting formula (2.63) into (2.60), we get
◦³ ´
g ◦ ij ◦ ◦ ◦ ◦ 2
g ij = g + 2(g ij g nl − 2g il g jn )εnl + ◦ ²imn ²jkl εmk εnl . (2.64)
g g
Finally, we should express the determinant g = det (gij ) in terms of εij . To

do this, we multiply relation (2.64) by gij and take formula (2.2) into account:
³
◦ ◦ ◦ ◦ ◦ ◦ 2
´ ◦
3g = g g ij + 2(g ij g nl − g il g jn )εnl + ◦ ²imn ²jpl εmp εnl (g ij + 2εij ). (2.65)
g
Thus, we get
◦ ◦ ◦ ◦
3g = g(3 + 4g nl εnl + (2/g)²imn ²jpl g ij εmp εnl +
◦ ◦ ◦ ◦ ◦ ◦
´
+ 2g ij εij + 4(g ij g nl − g il g jn )εnl εij + (4/g)²imn ²jpl εij εmp εnl . (2.66)
Modifying the third summand on the right-hand side by formula (2.63) and
introducing the notation
◦ 1 ◦ ◦ ◦ ◦ ◦
I1ε = g nl εnl , I2ε = (g ij g nl − g il g jn )εij εnl , I3ε = det (εij gik ), (2.67)
2
from (2.66) we get the desired formula
◦
g = g(1 + 2I1ε + 4I2ε + 8I3ε ). (2.68)
Here we have taken account of formula (2.58) for the matrix determinant
◦
and also the relation I3ε = (1/g)det (εij ). Thus, we have proved the following
theorem.
Theorem 1.8. The inverse metric matrix g ij is expressed in terms of compo-
◦ ij
nents εij of the deformation tensor and g by formulae (2.64) and (2.68).
Formulae (2.64) and (2.68) allow us to find the expression of contravariant
components εij of the deformation tensor in terms of εij . It follows from (2.2)
that
◦ ◦
ij ◦ ij ij ◦ ij g −g g ◦ ◦ ◦ ◦ 1
ε = (1/2)(g − g ) = g − (g ij g nl − g il g jn )εnl − ²imn ²jkl εmk εnl .
2g g g
(2.69)
Substitution of formulae (2.26) or (2.27) into (2.64) and (2.69) gives the
expression for components εij in terms of components of the displacement vector
◦
ui or ui .
Exercises for 1.2
Exercise 1.2.1. Using the results of Exercise 1.1.1, show that the deformation gradient
F and its inverse F−1 for the problem on a beam in tension (see Example 1.1) have the
forms
X
3
X
3
1
T −1 −1T
F=F = kα ēα ⊗ ēα , F =F = ēα ⊗ ēα .
kα
α=1 α=1
For this problem, the deformation tensors are determined by the formulae
1
X
3
1
X
3
C=Λ= (kα2 − 1)ēα ⊗ ēα , A=Λ= (1 − kα−2 )ēα ⊗ ēα ,

2 2
α=1 α=1
the deformation measures are determined as follows:
X
3
X
3
G = g −1 = kα2 ēα ⊗ ēα , g = G−1 = kα−2 ēα ⊗ ēα ,

α=1 α=1
and components of the deformation tensor take the forms
1 1
εαβ = (kα2 − 1)δαβ , εαβ = (1 − kα−2 )δαβ .
2 2
Exercise 1.2.2. Using the results of Exercise 1.1.2, show that for the problem on a
simple shear we have the following formulae for the deformation gradient:
◦
F = F ij ēi ⊗ ēj = E + aē1 ⊗ ē2 , FT = E + aē2 ⊗ ē1 ,
F−1 = E − aē1 ⊗ ē2 , F−1T = E − aē2 ⊗ ē1 ,
i.e. Ã !
◦ 1 a 0
ij
F = 0 1 0 , det F = 1,
0 0 1
for the deformation tensors:
C = (a/2)O3 + (a2 /2)ē22 , A = (a/2)O3 − (a2 /2)ē22 ,

Λ = (a/2)O3 − (a2 /2)ē21 , J = (a/2)O3 + (a2 /2)ē21 ,
and for components of the deformation tensor:
Ã ! Ã !
0 a/2 0 0 a/2 0
ij
(εij ) = a/2 a2 /2 0 , (ε ) = a/2 −a2 /2 0 .
0 0 0 0 0 0
Here we have introduced the notation
O3 ≡ ē1 ⊗ ē2 + ē2 ⊗ ē1 , ē2α ≡ ēα ⊗ ēα , α = 1, 2, 3.

Exercise 1.2.3. Using the formulae from Example 1.3 (see paragraph 1.1.1), show that
for the problem on rotation of a beam with extension, the deformation gradient has the
form
X
2
F = F0 ij ēi ⊗ ēj = cos ϕ kα ēα ⊗ ēα + k3 ē3 ⊗ ē3 + sin ϕk2 (ē2 ⊗ ē1 − ē1 ⊗ ē2 ).
α=1
1.3. Polar Decomposition 27
Exercise 1.2.4. Using formulae (1.36), (2.18)(2.25) and the results of Exercise 1.1.7,
show that local basis vectors are connected to displacements by the relations
◦ ◦ ◦ ◦
ri = rk (δik + ∇k ui ), ri = rk (δik − ∇i uk ).
Exercise 1.2.5. Using formulae (2.37) and (2.39), show that the physical components
q
b◦ ◦ ◦
of the deformation tensor εαβ = εαβ / g αα g ββ are connected to relative elongations δα
and angles χαβ by the relations
b◦ 1 b◦ 1
εαα = ((1 + δα )2 − 1), εαβ = (1 + δα )(1 + δβ ) sin χαβ .
2 2
Exercise 1.2.6. Show that in the basis e

ri of curvilinear coordinate system ei
X the
−1
expression of the tensor F in terms of ∇ ⊗ u can be rewritten in the form similar to
(2.22)(2.25):
eke
u=u rk , F−1 = (Fe−1 )ki e ri ,
rk ⊗ e (Fe−1 )ki = δik − ∇
e iu
ek .
Exercise 1.2.7. Using formula (1.34), show that the following relationships hold:
◦ ◦ ◦
|dx|2 = dx · G · dx, |dx|2 = dx · g · dx.
1.3. Polar Decomposition
1.3.1. Theorem on Polar Decomposition. According to (1.36), the tensor F

◦
can be considered as a tensor of the linear transformation from the basis ri to
◦
the basis ri . Since the vectors ri and ri are linearly independent, the tensor F is
nonsingular. Then for this tensor the following theorem is valid.
Theorem 1.9 (on the polar decomposition). Any nonsingular second-order
tensor F can be represented as the scalar product of two second-order tensors:
F=O·U or F = V · O. (3.1)
Here U and V are the symmetric and positive-definite tensors, O is the
orthogonal tensor, and each of the decompositions (3.1) is unique.
H Prove the existence of the decomposition (3.1) in the constructive way, i.e. we
should construct the tensors U, V and O. To do this, consider the contractions
of the tensor F with its transpose: FT · F and F · FT . Both the tensors are
symmetric, because
(FT · F)T = FT · (FT )T = FT · F and (F · FT )T = (FT )T · FT = F · FT , (3.2)

and positive-definite:
a · (FT · F) · a = (a · FT ) · (F · a) = (F · a) · (F · a) = b · b = |b|2 > 0 (3.3)

for any non-zero vector a, where b = F · a. Since any symmetric positive-definite
tensor has three real positive eigenvalues [12], eigenvalues of tensors FT · F
◦
and F · FT can be denoted as λ2α and λ2α . These tensors are diagonal in their
eigenbases, i.e. they have the following forms:
X
3
◦ ◦ ◦ X
3
FT · F = λ2α pα ⊗ pα , F · FT = λ2α pα ⊗ pα . (3.4)

α=1 α=1
◦
Here pα are the eigenvectors of the tensor FT · F and pα of the tensor F · FT ,
which are real-valued and orthonormal:
◦ ◦
pα · pβ = δαβ , pα · pβ = δαβ . (3.5)
The right-hand sides of (3.4) are the squares of certain tensors U and V defined
as
X
3
◦ ◦ ◦ ◦ X
3
U= λα pα ⊗ pα , λα > 0; V= λα pα ⊗ pα , λα > 0. (3.6)

α=1 α=1
Here signs at λα are always chosen positive.
In this case, the following relations are valid:
FT · F = U2 , F · FT = V2 . (3.7)
The constructed tensors V and U are symmetric due to formula (3.6) and
positive-definite, because for any nonzero vector a we have
X
3
◦ ◦ ◦ X
3
◦ ◦
a·U·a= λα a · pα ⊗ pα · a = λα (a · pα )2 > 0, (3.8)
α=1 α=1
◦
as λα > 0. In a similar way, we can prove that the tensor V is positive-definite.
Both the tensors V and U are nonsingular, because, under the conditions of
the theorem, the tensor F is nonsingular. And from (3.7) we get
(det U)2 = det U2 = det (FT · F) = (det F)2 6= 0. (3.9)

Then there exist inverse tensors U−1 and V −1 , with the help of which we can
construct two more new tensors
◦
O = F · U−1 , O = V −1 · F , (3.10)
which are orthogonal. Indeed,
◦ ◦
OT · O = (F · U−1 )T · (F · U−1 ) = U−1 · FT · F · U−1 = U−1 · U2 · U−1 = E.
(3.11)
◦
According to [12], this means that the tensor O is orthogonal. In a similar way,
we can show that the tensor O is orthogonal as well.
◦
Thus, we have really constructed the tensors U and O, and also V and O,
the product of which, due to (3.10), gives the tensor F:
◦
F = O · U = V · O. (3.12)
◦
Here U and V are symmetric, positive-definite tensors, O and O are orthogonal
tensors.
Show that each of the decompositions (3.12) is unique. By contradiction, let
there be one more resolution, for example
◦
e · U.
F=O e (3.13)
But then
e 2 = U2 ,
FT · F = U (3.14)
hence, e = U,
U because the decomposition of the tensor FT · F for its eigenbasis
◦ ◦
is unique. Signs at λα and eα
λ are chosen positive by the condition. The
◦ ◦
coincidence of U and e
U leads to the fact that e
O and O are coincident as well,
because ◦ ◦
e =F·U
O e −1 = F · U−1 = O. (3.15)
This has proved uniqueness of the decomposition (3.12). We can verify unique-
ness of the decomposition F=V·O in a similar way.
◦
Finally, we must show that the orthogonal tensors O and O are coincident,
i.e. formula (3.1) follows from (3.12). To do this, we construct the tensor
◦ ◦ ◦
F · OT = O · U · OT . (3.16)
Due to (3.12), this tensor satisfies the following relationship:
◦ ◦ ◦
O · U · OT = V · O · OT . (3.17)
◦
The tensor O · OT is orthogonal, because
◦ ◦ ◦ ◦ ◦ ◦
(O · OT )T · (O · OT ) = O · OT · O · OT = O · OT = E. (3.18)
Then the relationship (3.17) can be considered as the polar decomposition of the
◦ ◦
tensor O · U · OT . This tensor is symmetric, because
◦ ◦ ◦ ◦ ◦ ◦
(O · U · OT )T = (OT )T · (O · U)T = O · U · OT . (3.19)
Then the formal equality
◦ ◦ ◦ ◦
O · U · OT = O · U · OT (3.20)
is one more polar decomposition. However, as was shown above, the polar
decomposition is unique; hence, the following relationships must be satisfied:
◦ ◦ ◦
V = O · U · OT and O · OT = E. (3.21)
◦ ◦
Thus, the orthogonal tensors O and O are coincident: O = O. N
The tensors U and V are called the right and left stretch tensors, respec-
tively, and O is the rotation tensor accompanying the deformation.
The tensor F has nine independent components, the tensor O three

independent components, and each of the tensors U and V six independent
components.
Remark 1. Since the rotation tensor O is unique in the polar decomposition,
from formula (3.21) we get that the stretch tensors U and V are connected by
means of the tensor O:
V = O · U · OT , U = OT · V · O. ¤ (3.21a)
Theorem 1.10. The CauchyGreen and Almansi deformation tensors can be
expressed in terms of the stretch tensors U and V as follows:
1 1
C = (U2 − E), A = (E − V−2 ),
2 2
1 1
Λ = (E − U−2 ), J = (V2 − E). (3.22)
2 2
H To see this, let us substitute the polar decomposition (3.1) into (2.5), and then
we get the relationships (3.22). N
1.3.2. Eigenvalues and Eigenbases.
Theorem 1.11. Eigenvalues of the tensors U and V defined by (3.6) are
coincident: ◦
λα = λα , α = 1, 2, 3, (3.23)
◦
and eigenvectors pα and pα are connected by the rotation tensor O:
◦
pα = O · pα . (3.23a)
H To prove the theorem, we use the definition (3.6) and the first formula of
(3.21a):
X
3
X
3
◦ ◦ ◦ X
3
◦
T
V= λα pα ⊗ pα = O · U · O = λα O · pα ⊗ (O · pα ) = λα p0α ⊗ p0α ,
α=1 α=1 α=1
where ◦
p0α = O · pα .
According to the relationship, we have obtained two different eigenbases of the
tensor V and two sets of eigenvalues, that is impossible. Therefore,
◦ ◦
p0α = O · pα = pα and λα = λα ,
as was to be proved. N
Due to (3.5), both the eigenbases are orthogonal. Therefore, reciprocal
◦
vectors of the eigenbases do not differ from pα and pα :
◦ ◦
pα = pα , pα = pα . (3.24)
◦
The important problem for applications is to determine λα , pα and pα by the
given deformation gradient F. To solve the problem, one should use the following
method.
1) Construct the tensor U2 = FT · F (or V2 = F · FT ) and find its components in

some basis being suitable for a considered problem; for example, in the Cartesian
basis ēi :
U2 = (Ū 2 )i j ēi ⊗ ēj and V2 = (V̄ 2 )i j ēi ⊗ ēj .
2) Find eigenvalues of the matrix (Ū 2 )i j by solving the characteristic equation
det (U2 − λ2α E) = 0, (3.25)

which in the basis ēi takes the form
det ((U2 )i j − λ2α δji ) = 0. (3.25a)

◦
3) Find eigenvectors pα of the tensor U and eigenvectors pα of the tensor V
from the equations
◦ ◦
U2 · pα = λ2α pα , V2 · pα = λ2α pα , (3.26)
written, for example, in the basis ēi :
◦
b j = 0,
((Ū 2 )i j − λ2α δji )Q b j = 0,
((V̄ 2 )i j − λ2α δji )Q (3.26a)
α α
◦
where bjα
Q and bjα
Q are Jacobian matrices of the eigenvectors:
◦
◦
b j α ēj ,
pα = Q b j α ēj .
pα = Q (3.27)
◦
To determine the matrices bjα
Q and bjα ,
Q one should consider only independent
equations of the system (3.26a) and the normalization conditions (3.5):
◦
|pα | = 1, |pα | = 1, (3.28)
which are equivalent to the quadratic equations
◦ ◦
bi Q
Q b j δij = 1, bi Q
Q b j δij = 1. (3.28a)
α α α α
4) Write the dyadic products (3.6) and find resolutions of the tensors U and V
for the eigenbases; for example, for the Cartesian basis ēi :
X
3 ◦ ◦ X
3
U= bi α Q
λα Q b j α ēi ⊗ ēj , V= bi α Q
λα Q b j α ēi ⊗ ēj .
α=1 α=1
Exercises 1.3.21.3.4 show examples of determination of the tensors U and V.
Remark 2. Notice that a solution of the quadratic equations (3.28a) may be
◦
b i b i , this
not unique due to the choice of signs of matrix components Q α and Q α
ambiguity is resolved by applying one more additional condition, namely the
◦
condition of coincidence of the vectors pα and pα when t → 0+ :
◦
t → 0+ ⇒ pα (t) = pα (t), α = 1, 2, 3.
◦
For the matrix bi α ,
Q the ambiguity of the sign choice remains. However, if
◦
there is a field of eigenvectors pα (x, t), then this ambiguity may be retained only
at one point x0 at one time, for example, t = 0; and for the remaining x and t,
◦
a sign at Q b i α is chosen from the continuity condition of the vector field p α (x, t)
◦
(for continuous motions). If the eigenvector field pα (x0 , 0) contains the vectors
◦
ēα , then the remaining ambiguity is resolved by the condition pα (x0 , 0) = ēα .
The ambiguity of a solution of the system (3.26a), (3.28a) may also appear,
if at some time t1 at a point x the eigenvalues λα (t1 ) prove to be triple. In this
◦
case, values of the matrices b i α (t1 )
Q and b i α (t1 )
Q are determined, as a rule, by
passage to the limit:
◦ ◦
b i α (t1 ) = lim Q
Q b i α (t), b i α (t1 ) = lim Q
Q b i α (t), α = 1, 2, 3.
t→t1 t→t1
In the case of double eigenvalues λα , these formulae are applied only to their
◦
corresponding matrix components bi α
Q and bi α . ¤
Q
1.3.3. Representation of the Deformation Tensors in Eigenbases.
◦ ◦
Theorem 1.12. In the tensor bases pα ⊗ pβ and pα ⊗ pβ , the CauchyGreen
tensors C and J, the Almansi tensors A and Λ, and the deformation measures
G, g−1 and G−1 , g have the diagonal form:
X
3
1 ◦ ◦ X
3
1 ◦ ◦
C= (λ2α − 1)pα ⊗ pα , Λ= (1 − λ− 2
α )pα ⊗ pα , (3.29a)
2 2
α=1 α=1
X
3
1
X
3
1
A= (1 − λ− 2
α )pα ⊗ pα , J= (λ2α − 1)pα ⊗ pα ;
2 2
α=1 α=1
and
X
3
◦ ◦ X
3
◦ ◦
G= λ2α pα ⊗ pα , G−1 = λ−2
α pα ⊗ pα , (3.29b)
α=1 α=1
X 3
X
3
g−1 = λ2α pα ⊗ pα , g= λ− 2
α pα ⊗ pα .
α=1 α=1
H On substituting formulae (3.6) into (3.22), we get (3.29a). Formulae (3.29b)
follow from (3.29a) and (2.7), (2.8). N
Similarly to formulae (3.29), we can introduce new deformation tensors by
◦ ◦
determining their components with respect to the bases pα ⊗ pβ or pα ⊗ pβ as
follows:
◦ X
3
◦ ◦ X
3
M= f (λα )pα ⊗ pβ , M= f (λα )pα ⊗ pβ , (3.30)

α=1 α=1
where f (λα ) is a function of λα . If f (1) = 0, then we get the deformation tensors;
and if f (1) = 1, then we get the deformation measures.
Among the tensors (3.30), the logarithmic deformation tensors and mea-
sures
◦ X
3
◦ ◦ X
3
H= lg λα pα ⊗ pβ , H= lg λα pα ⊗ pβ , (3.31)
α=1 α=1
◦ ◦
H1 = H + E, H1 = H + E,
are the most widely known; they are called the right and left Hencky tensors,
and also the right and left Hencky measures, respectively.
◦
With the help of the eigenvectors pα and pα we can form the mixed dyads
X
3
◦ X
3
X
3
pα ⊗ pα = pα ⊗ pα · O = ( pα ⊗ pα ) · O = E · O. (3.32)
α=1 α=1 α=1
Here we have used the properties (3.23a) and (3.24), and the representation of
the unit tensor E in an arbitrary mixed dyadic basis.
Thus, the rotation tensor O accompanying the deformation can be expressed
in the eigenbasis as follows:
X
3
◦ ◦
O= pα ⊗ pα = pi ⊗ pi . (3.33)
α=1
On substituting (3.33) and (3.6) into (3.1) and taking (3.5) into account, we
get the following expression of the deformation gradient in the tensor eigenbasis:
X
3
◦ X
3
◦ ◦ ◦ X
3
◦
F=O·U= pα ⊗ pα · λβ pβ ⊗ pβ = λα pα ⊗ pα . (3.34)
α=1 β=1 α=1
According to (3.34), the transpose FT and inverse F−1 gradients are ex-
pressed as follows:
X
3
◦ X
3
◦
FT = λα pα ⊗ pα , F−1 = λ−1
α pα ⊗ pα . (3.35)
α=1 α=1
◦
1.3.4. Geometrical Meaning of Eigenvalues. Vectors of eigenbases pα and
◦
pα are connected by the transformation (3.23a). In K take elementary radius-
◦ ◦
vectors dxα oriented along the eigenbasis vectors pα , then in K they correspond
to radius-vectors dxα :
◦ ◦ ◦ ◦
dxα = pα |dxα |, dxα = F · dxα . (3.36)
Substitution of (3.34) into (3.36) yields
X
3
◦ ◦ ◦
dxα = λβ pβ ⊗ pβ · pα |dxα | = λα |dxα |pα , (3.37)
β=1
i.e. the elementary radius-vectors dxα in K will be also oriented along the
corresponding eigenbasis vectors pα .
◦ ◦
Denote lengths of the vectors dxα and dxα by dsα and ds, respectively, and
derive relations between them:
◦ ◦ ◦ ◦ ◦
ds2α = dxα · dxα = dxα · FT · F · dxα = |dxα |2 pα · FT · F · pα =
◦ ◦ ◦ ◦
= ds2α pα · G · pα = ds2α λ2α . (3.38)
Here we have used equations (3.29b) and (3.36). Formula (3.38) proves the
following theorem.
Theorem 1.13. Eigenvalues λα (principal stretches) are the elongation ratios
for material fibres oriented along the principal (eigen-) directions:
◦
λα = dsα /dsα . (3.39)
1.3.5. Geometric Picture of Transformation of a Small Neighborhood of a
◦
Point of a Continuum. In K, consider a small neighborhood of the material
0
point M contained in a continuum; then every point M , connected to M by the
◦
elementary radius-vector dx (Figure 1.12), will be connected to the same point
M by radius-vector dx inK. These radius-vectors are related as follows:
◦
dx = F · dx. (3.40)
The relation can be considered as the transformation of arbitrary radius-vector
◦
dx into dx.
Figure 1.12. Transformation of a small neighborhood of the point contained in a continuum
Rewrite the relation (3.40) in Cartesian coordinates:

◦
dxi = F̄m
i
dxm , (3.41)
where
i
F̄m are components of the deformation gradient with respect to the
Cartesian basis (see Exercise 1.1.5):
i ◦
F̄m = (∂xi /∂ xm ), (3.42)
◦m
which depend only on coordinates x of the point M, but they are independent
◦
of coordinates dxm of its neighboring points M0 . Therefore the transformation
◦
(3.41) is a linear transformation of coordinates dxm into dxi , i.e. this is an affine
transformation.
As follows from the general properties of affine transformations, straight
◦
lines and planes contained in a small neighborhood in K will be straight
lines and planes in actual configuration K. Parallel straight lines and planes
are transformed into parallel straight lines and planes. Therefore if a small
◦
neighborhood in K is chosen to be a parallelogram, then in K the neighborhood
will be a parallelogram as well (although angles between its edges, edge lengths
and orientation of planes in space may change).
◦
Since a second-order surface in K (and, in general, a surface specified by an
algebraic expression of arbitrary n-th order) is transformed into a surface of the
◦
same order in K, a small spherical neighborhood in K is transformed into an
ellipsoid in actual configuration K (Figure 1.12).
◦
As follows from formula (2.34), the ratio of lengths dsα /dsα of an arbitrary
◦
vector (or of elementary radius-vector dx in K and K) is independent of the
◦
initial length dsα of the vector (because the relative elongation δα is independent
◦
of dsα ).
According to the polar decomposition (3.1), the transformation (3.40) from
◦
K to K can always be represented as the superposition of two transformations:
◦ ◦ ◦
dx = O · dx0 , dx0 = U · dx, (3.43)
realized with the help of the stretch tensor U and the rotation tensor O, or
◦
dx = V · dx0 , dx0 = O · dx. (3.44)
◦
The stretch tensor U, which has three eigendirections pα , transforms a small
neighborhood of the point M with compressing or extending the neighborhood
◦
along these three directions pα . The tensor O rotates the neighborhood deformed
◦ ◦
along pα as a rigid whole until the direction of pα becomes the direction of
◦ ◦
pα . If one use the left stretch tensor V, so rotation of axes pα in K till their
coincidence with pα is first realized, and then compression or tension of the
neighborhood occurs along the direction pα . The result will be the same as for
U. ◦
If a point Mα is connected to M by radius-vector dxα oriented along
◦
the eigendirection pα (which is unknown before deformation), then in K the
point Mα will be connected to M by radius-vector dxα oriented along the
corresponding eigendirection pα .
◦
If a small neighborhood of point M is chosen to be a sphere in K (see
Figure 1.12), then in K the sphere becomes an ellipsoid with principal axes
oriented along the eigendirections pα .
Thus, the transformation of a small neighborhood of every point M contained
in a continuum under deformation can always be represented as a superposition
of tension/compression along eigendirections and rotation of the neighborhood

as a rigid whole, and also displacement as a rigid whole.
Exercises for 1.3
Exercise 1.3.1. Using the formula (3.21a), show that the following relations between
V and U hold:
Vm = O · Um · OT , Um = OT · Vm · O
for all integer m (positive and negative).
Exercise 1.3.2. Using the results of Exercises 1.1.1 and 1.2.1, show that for the
problem on tension of a beam, eigenvalues λα are
λα = kα , α = 1, 2, 3.
The stretch tensors U and V are coincident and have the form
X
3
U=V= k α eα ⊗ eα ,
α=1
◦
and eigenvectors pα and pα coincide with eα :
◦
p α = p α = eα , α = 1, 2, 3.
The rotation tensor O for this problem is the unit one: O = E.
Exercise 1.3.3. Using the results of Exercises 1.1.2, 1.2.2 and Remark 2, show that
for the problem on simple shear (see Example 1.2 from paragraph 1.1.1), the tensors U2
and V 2
are expressed as follows:
U2 = FT · F = E + aO3 + a2 e2i = (Ū 2 )i j ēi ⊗ ēj ,
V2 = E + aO3 + a2 e2i = (V̄ 2 )i j ēi ⊗ ēj ,

Ã ! Ã !
1 a 0 1+a
2
a 0
(Ū 2 )i j = a 1+a 0 ,
2 2 i
(V̄ ) j = a 1 0 ,
0 0 1 0 0 1
eigenvalues λα are
λ2α = 1 + bα |a|, α = 1, 2; λ3 = 1,
q q
a a
b1 = + 1 + a2 /4 , b2 = − 1 + a2 /4 ,
2 2
◦
eigenvectors pα and pα (a > 0) are
◦ 1 ◦
pα = p (ē1 + bα ē2 ), p3 = ē3 ,
1 + b2α
1 1
p1 = q (b1 ē1 + ē2 ), p2 = − q (b2 ē1 + ē2 ), p3 = ē3 ,
1 +b 2
1
1 + b22
the stretch tensors U and V are
U = Ū ij ēi ⊗ ēj = U0 ē21 + U1 O3 + U2 ē22 + ē23 ,

V = V̄ ij ēi ⊗ ēj = U2 ē21 + U1 O3 + U0 ē22 + ē23 ,

Ã ! Ã !
U0 U1 0 U2 U1 0
Ū ij = U1 U2 0 , V̄ ij = U1 U0 0 ,
0 0 1 0 0 1
β
p β
p
b1 1 + b1 a b2 1 + b2 a
Uβ = + , β = 0, 1, 2,
1 + b21 1 + b22
and the rotation tensor O has the form
O = Ō j ēi ⊗ ēj = cos ϕ(ē21 + ē22 ) + sin ϕ(ē1 ⊗ ē2 − ē2 ⊗ ē1 ),
i
Ã !
cos ϕ sin ϕ 0
i
Ō j = − sin ϕ cos ϕ 0 ,
0 0 1
b1 b2 b21 b22
cos ϕ = − , sin ϕ = − .
1 + b21 1 + b22 1 + b21 1 + b22
Show that functions b1 (a) and b2 (a) satisfy the following relationships:
b1 + b2 = a, b1 b2 = −1, b21 + b22 = 2 + a2 .

Show that at a=0 for the considered problem the following equations really hold:
b1 = 1, b2 = −1, λ1 = λ2 = λ3 = 1,
◦ 1 ◦ 1
p1 = p1 = √ (ē1 + ē2 ), p2 = p2 = √ (ē1 − ē2 ).
2 2
Exercise 1.3.4. Using the results of Exercise 1.2.3, show that for the problem on
rotation of a beam with tension (see Example 1.3 from paragraph 1.1.1), eigenvalues λα
have the form
λα = kα , α = 1, 2, 3,
and eigenvectors
◦
pα = ēα , pα = O0 · ēα , α = 1, 2, 3.
Using formulae from Exercise 1.1.3 and data from Example 1.3, show that tensors
U, V, O, and also C, A, Λ and J have the form
X
3
U = U0 = kα ēα ⊗ ēα , O = O0 = O0i j ēi ⊗ ēj ,

α=1
V = O0 · U0 · OT
0
= V0 ē21 + V1 O3 + V2 ē22 + k3 ē23 = V0 ij ēi ⊗ ēj ,
Ã !
V0 V1 0
i
V0 j = V1 V2 0 ,
0 0 k3
V1 = k1 cos2 ϕ + k2 sin2 ϕ, V1 = (k1 − k2 ) cos ϕ sin ϕ,
2
V2 = k1 sin ϕ + k2 cos2 ϕ,
1
X
3
1
C = (U20 − E) = (kα2 − 1)ēα ⊗ ēα ,
2 2
α=1
1
X
3
1
Λ = (E − U−
0
2
)= (1 − kα−2 )ēα ⊗ ēα ,
2 2
α=1
1 −2 1 ij ij 1 1
A = (E − V ) = (δ − g )ēi ⊗ ēj , J = (V2 − E) = (gij − δij )ēi ⊗ ēj ,
2 2 2 2
where metric matrices gij and g ij are determined by formulae from Exercise 1.1.3.
We should take into consideration that the tensors C and Λ do not feel the beam
rotation they are coincident with the corresponding tensors for the problem on pure
tension of the beam. Show that if we change the sequence of transformations (i.e. we
first rotate and then extend the beam), then the tensors A and J do not feel the rotation.
1.4. Rate Characteristics of Continuum Motion
1.4.1. Velocity. The velocity (vector) of the motion of a material point M

with Lagrangian coordinates Xi is determined as the partial derivative of the
radius-vector x(X i , t) with respect to time at fixed values of X i:
¯
∂x ¯
v(X i , t) = (X i , t)¯ . (4.1)
∂t Xi
Velocity components v̄ i with respect to the basis ēi have the form
i i
∂x ∂x
v = v̄ i ēi = ēi , v̄ i = (X j , t). (4.2)
∂t ∂t
1.4.2. Total Derivative of a Tensor with respect to Time. Any vector field
a(x, t) (and also scalar or tensor field) varying with time, which describes
some physical process in a continuum, can be expressed in both Eulerian and
Lagrangian descriptions with the help of the motion law (1.3):
a(x, t) = a(x(X j , t), t). (4.3)

Determine the derivative of the function with respect to time at fixed Xi (i.e.
for a fixed point M):
¯ ¯ ¯
∂a ¯ ∂a ¯ ∂a ∂xj ¯
¯ i= ¯ i+ j ¯ . (4.4)
∂t X ∂t x ∂x ∂t X i
Definition 1.1. The partial derivative of a varying vector field a (4.3) with
respect to time t at fixed coordinates Xi is called the t o t a l derivative of the
function (4.3) with respect to time:
¯
da ∂a ¯
ȧ ≡ = ¯ . (4.5)
dt ∂t X i
According to formulae (4.2), (1.11) and (1.23), the second summand on the
right-hand side of (4.4) can be rewritten as follows:
∂a ∂xj ∂a ∂a ∂a
j ∂t
= v̄ j P kj k
= v̄ i ēi · ēj P kj ⊗ k
= v · rk ⊗ = v · ∇ ⊗ a. (4.6)
∂x ∂X ∂X ∂X k
Then the relationship (4.4) yields
da ∂a
= + v · ∇ ⊗ a, (4.7)
dt ∂t
1.4. Rate Characteristics of Continuum Motion 39
where we have introduced the notation for the partial derivative with respect to
time which will be widely used below:
¯
∂a ∂a i ¯
= (x , t)¯ i . (4.8)
∂t ∂t x
In formula (4.7) the vector a is considered as a function a(xj , t). It is evident

that if a j
is considered as a function of (X , t), then from the definition (4.5) we
get ¯
da i ∂ ¯
(x , t) = a(X j , t)¯ i . (4.9)
dt ∂t X
The total derivative (da/dt) is also called the material (substantial, indi-
vidual) derivative with respect to time, (∂a/∂t) in (4.7) is the partial (local)
derivative with respect to time, and v · ∇ ⊗ a is the convective derivative.
The material derivative da/dt characterizes a change of the vector field a in
a fixed material point M, the local derivative determines a change of values of
a in time at a fixed point x in space, and from (4.6) we get that the convective
derivative characterizes a change of the field due to transfer of the material
particle M from a point x to a point x + vdt in space.
If we choose the vector v as a, then the relationship between the displacement
u and the velocity v vectors has the form
dx du ∂u
v= = = + v · ∇ ⊗ u. (4.10)
dt dt ∂t
Similarly to formula (4.5), we can define the total derivative of the nth-order
tensor
n Ω with respect to time:
¯
n dn ∂ nΩ ¯
Ω̇ = Ω(xi , t) = (X i , t)¯ i . (4.11)
dt ∂t X
Theorem 1.14. The total derivative (4.11) of a varying tensor field

n Ω(xi , t)
can be written as a sum of local and convective derivatives:
d n ∂ nΩ
Ω= + v · ∇ ⊗ n Ω. (4.12)
dt ∂t
H Proof of the theorem is similar to the proof of the relationship (4.7). Details
are left as Exercise 1.4.6. N
Let us consider now the question on components of the total derivative tensor.
Theorem 1.15. Components of the total derivative tensor
n Ω̇ are connected
with the corresponding components of a tensor
nΩ with respect to stationary
◦
bases ri , ēi and e
ri and a moving basis ri as follows:
¯
d ◦ i ...in ∂ ◦ ¯
Ω = Ωi ...in (X i , t)¯ i ,
1 1
(4.13)
dt ∂t X
d i ...in ∂ ∂
Ω̄ 1
= Ω̄i ...in (xi , t) + v̄ k k Ω̄i ...in (xi , t),
1 1
(4.14)
dt ∂t ∂x
d e i ...in ∂ e i ...in e i e kΩ
e i ...in (X
e i , t),
Ω 1
= Ω (X , t) + vek ∇
1 1
(4.15)
dt ∂t
n
X
d i in ∂
Ω 1 ...
= Ωi 1 ...in i
(X , t) + (Ωi 1 ... k...in
∇k v iα )(X i , t), (4.16)
dt ∂t
α=1
where
n d i ...in i d ◦ ◦ ◦
Ω̇ = Ω̄ 1
(x , t)ēi ⊗ . . . ⊗ ēin = Ωi ...in (X i , t)ri ⊗ . . . ⊗ rin = 1
dt dt1 1
d e i ...in e i d
= Ω ri ⊗ . . . ⊗ e
(X , t)e
1
rin = Ωi ...in (X i , t)ri ⊗ . . . ⊗ rin . 1
(4.17)
dt dt 1 1
In formula (4.16), the component Ωi 1 ...k...in

as the αth superscript has index
k in place of iα .
H To prove the theorem, we resolve the tensor
n Ω for different bases:
◦ ◦ ◦
n
Ω = Ω̄i 1 ...in
(xi , t)ēi ⊗ . . . ⊗ ēin = Ωi ...in (X i , t)ri ⊗ . . . ⊗ rin =
1
1
1
=Ωe i ...in (X
e i , t)e
1
rin = Ωi ...in (X i , t)ri ⊗ . . . ⊗ rin ,
ri ⊗ . . . ⊗ e 1
(4.18)
1 1
n
and choose arguments of components of the tensor Ω as in formula (4.18).
◦
Then, substituting the resolution (4.18) for the basis ri into the definition
◦
(4.11), we get the expression (4.13), because dri /dt = 0.
On substituting the resolution (4.18) for the basis ēi into the relationship
(4.12), we obtain
n ∂ Ω̄i ...in
1
∂
Ω̇ = ēi ⊗ . . . ⊗ ēin + v̄ k ēk · m Ω̄i 1 ...in m
ē ⊗ ēi ⊗ . . . ⊗ ēin . (4.18a)
∂t 1
∂x 1
It is evident that formula (4.14) follows from (4.18a).

In a similar way, substituting the resolution (4.18) for the basis e
ri into the
relation (4.12) and using the property (1.61) of nabla-operators ∇ and e,
∇ and
also the equation ∂e
ri /∂t = 0 (for the stationary basis e
ri see paragraph 1.1.7), we
get
e i ...in
∂Ω 1
n
Ω̇ = e rin + veke
ri ⊗ . . . ⊗ e e mΩ
rk · ∇ ei 1 ...in m
e
r ⊗e
ri ⊗ . . . ⊗ e
rin ; (4.19)
∂t 1 1
and formula (4.15) follows from (4.19) at once.

Finally, substituting the resolution (4.18) for the moving basis ri into the
definition of the total derivative (4.12), we obtain
¯ X n ¯
n ∂Ωi ...in ¯ 1
∂riα ¯
Ω̇ = ¯ i ri ⊗ . . . ⊗ rin + Ωi 1 ...iα ...in
ri ⊗ . . . ¯ . . . ⊗ rin .
∂t X
1 1
∂t X i
α=1
(4.20)
Due to the definition (1.10) of local bases vectors and the definition (4.1) of the
velocity, we have
¯ ¯ ¯
∂riα ¯ ∂2x ¯ ∂v ¯
(X j , t)¯ j = i ¯ j = ¯ = ∇iα v k rk . (4.21)
∂t X ∂t∂X X α
∂X iα X j
On substituting (4.21) into (4.20) and then collecting components at the same
elements of the polyadic basis, we derive the formula (4.16). N
It should be noted that arguments of the resolutions (4.18) and of the

derivatives of tensor components (4.13)(4.16) have been chosen in the specific
way.
1.4.3. Differential of a Tensor.

Definition 1.2. For a tensor field n Ω(xi , t), the following object
n
d Ω
d nΩ = dt (4.22)
dt
is called the d i f f e r e n t i a l of a tensor field (or the differential of a tensor)
n Ω(xi , t).
According to formula (4.12) for the total derivative of a tensor with respect
to time, we get that the differential of a tensor can be written in the form
³ ∂ nΩ ´
d n Ω(xi , t) = + v · ∇ ⊗ n Ω dt. (4.23)
∂t
According to (4.10), the relation (4.23) takes the form
∂ nΩ
d nΩ = dt + dx · ∇ ⊗ n Ω (4.24)
∂t
When a tensor field is stationary (i.e. ∂ n Ω/∂t = 0), the differential of the
tensor field has the form
db n Ω = dx · ∇ ⊗ n Ω. (4.25)
For stationary tensor fields db n Ω = d n Ω, but in general these differentials
are not coincident.
According to Theorem 1.15, components of the tensor d nΩ with respect to
◦
the fixed basis ri are written as follows:
◦
n
◦
j1 ...jn ◦ ◦ ◦
j1 ...jn d Ωj ...jn1
d Ω = dΩ rj ⊗ . . . ⊗ rjn , dΩ = dt. (4.26)

1
dt
From (4.22) and (4.7) we get the following expression for the differential of
a vector:
da
³ ∂a ´
da(X i , t) = dt = + v · ∇ ⊗ a dt, (4.27)
dt ∂t
and from (4.25) we have
a = (v · ∇ ⊗ a)dt = (∇ ⊗ a)T · dx.

db (4.28)
◦
In particular, if a = x, then, by formulae (4.28) and (1.35a), we obtain
b◦ ◦
dx = (∇ ⊗ x)T · dx = F−1 · dx, (4.29)
or
b◦
dx = F · dx. (4.30)
On comparing formulae (4.30) with (1.34), we find that the elementary radius-
◦
vector dx, introduced in paragraph 1.1 and connecting two infinitesimally close
b◦
material points M and M0 , coincides with the vector dx in the notation (4.25).
1.4.4. Properties of Derivatives with respect to Time. Let us establish now

important properties of partial and total derivatives of vector fields with respect
to time.
Theorem 1.16. The partial derivative of the vector product of basis vectors
with respect to time has the form
∂ ∂r ∂r
(rα × rβ ) = α × rβ + rα × β . (4.31)
∂t ∂t ∂t
H Determine the derivative of the vector product of two local basis vectors with
respect to time:
∂ ∂ ∂
(rα × rβ ) = (Qi α ēi × Qj β ēj ) = (Qi α Qj β )ēi × ēj =
∂t ∂t ∂t
∂Qi α j i ∂Qjβ
= ēi × Q β ēj + Q α ēi × ēj .
∂t ∂t
With use of relation (1.10) we really get (4.31). N
Theorem 1.17. For arbitrary continuously differentiable vector fields a(x, t) =
= āi (xk , t)ēi and b(x, t) = b̄i (xk , t)ēi , we have the formulae
∂ ∂a ∂b
(a × b) = ×b+a× , (4.32)
∂t ∂t ∂t
∂ ∂a ∂b
(a ⊗ b) = ⊗b+a⊗ , (4.33)
∂t ∂t ∂t
∂ ∂a ∂b
(a · b) = ·b+a· . (4.34)
∂t ∂t ∂t
H A proof is similar to the proof of Theorem 1.16. N
Theorem 1.18. The total derivatives of the vector and scalar products of two
arbitrary vector fields a(x, t) and b(x, t) with respect to time have the forms
d da db
(a × b) = ×b+a× , (4.35)
dt dt dt
d da db
(a · b) = ·b+a· . (4.36)
dt dt dt
H To prove formula (4.35), one should use the property of the total derivative
(4.7):
d ∂
(a × b) = (a × b) + v · ∇ ⊗ (a × b).
dt ∂t
Modify the first summand by formula (4.32) and the second summand by the
formula ∇ ⊗ (a × b) = (∇ ⊗ a) × b − (∇ ⊗ b) × a [12], then we get
d ∂a ∂b
(a × b) = ×b− × a + v · (∇ ⊗ a) × b − v · (∇ ⊗ b) × a.
dt ∂t ∂t
Collecting the first summand with the third one and the second summand with
the fourth one, and using the property (4.7) of the total derivative of a vector,
we obtain
d da db da db
(a × b) = ×b− ×a= ×b+a× .
dt dt dt dt dt
Formula (4.36) can be proved in a similar way. N
√
Theorem 1.19. The total derivative of g with respect to time is connected to
the divergence of the velocity v by
d√ √ √
g = g ∇i v i = g ∇ · v. (4.37)
dt
H Let us differentiate the second relation of (1.15) with taking formula (4.31)
into account:
d√ d ∂2x
g = r1 · (r2 × r3 ) = · (r2 × r3 )+
dt dt ∂t∂X 1 µ ¶ µ ¶
∂2x ∂2x
+ r1 · × r3 + r1 · r2 × . (4.38)
∂t∂X 2 ∂t∂X 3
Since
∂2x ∂v
i
= = ∇i v = ∇i v j rj ,
∂t∂X ∂X i
we get
d√ √
g = ∇1 v g · r1 + r1 · (∇2 v × r3 ) + r1 · (r2 × ∇3 v). (4.39)
dt
Here we have used the relations from Exercise 1.1.14.
According to the definition of the vector product (0.2), we obtain
√ √ √
r1 · ∇2 v × r3 = r1 · g ²ijk ∇2 v i δ3j rk = g ²i31 ∇2 v i = g ∇2 v 2 . (4.40)
On substituting (4.40) into (4.39), we really get formula (4.37). N
1.4.5. The Velocity Gradient, the Deformation Rate Tensor and the Vor-
◦
ticity Tensor. Consider elementary radius-vectors dx and dx connecting two
◦
0
infinitesimally close points M and M in configurations K and K, respectively.
0
Determine the velocity of the point M relative to the configuration connected
to the point M. b :
dv
To do this, determine the velocity differential
³◦ ´ µ ¶T
b ∂ ∂2x i ∂2x ◦i ◦ i ∂v T ◦ ◦ ◦
dv = dx = i
dX = i
⊗ r · dx = r ⊗ i
· dx = ∇ ⊗ v · dx.
∂t ∂X ∂t ∂X ∂t ∂X
(4.41)
Here we have used the second equation of (1.33), the definition of the gradient
(1.24) and formula (4.1). In a similar way, using the first equation of (1.33):
dX i = ri · dx, we get one more expression for the vector b :
dv
b = (∇ ⊗ v)T · dx.
dv (4.42)
T
The second-order tensor (∇ ⊗ v) is called the velocity gradient, which
connects the relative velocity b of an elementary
dv radius-vector dx to the vector
dx itself:
b = L · dx,
dv L = (∇ ⊗ v)T . (4.43)
Just as any second-order tensor (see [12]), the tensor L can be represented by a
sum of the symmetric tensor D and the skew-symmetric tensor W:
L = D + W. (4.44)
The symmetric deformation rate tensor D is determined as follows:
1
D = (∇ ⊗ v + ∇ ⊗ vT ). (4.45)
2
This tensor has six independent components.

The skew-symmetric vorticity tensor W is determined as follows:
1
W = (∇ ⊗ vT − ∇ ⊗ v). (4.46)
2
Since the tensor W is skew-symmetric and has three independent components,

we can put the tensor W in correspondence with the vorticity vector ω
connected to the tensor (see [12]) as follows:
1
ω = W · · ², W = ω × E. (4.47)
2
where ² is the Levi-Civita tensor, which has the third order (see [12]). This
tensor is determined as follows:
1
² = √ ²ijk ri ⊗ rj ⊗ rk . (4.48)
g
On substituting (4.44)(4.47) into (4.42), we prove the following theorem.
Theorem 1.20 (CauchyHelmholtz). The velocity v(M0 ) of an arbitrary point
M0 in a neighborhood of the material point M consists of the translational
motion velocity v(M) of the point M, the velocity ω × dx of rotation as a
rigid whole and the deformation rate D · dx, i.e.
b = ω × dx + D · dx
dv (4.49)
or
v(M0 ) = v(M) + ω × dx + D · dx + o (|dx|). (4.49a)
Example 1.4. Determine the tensor L for the problem on tension of a beam
(see Example 1.1), substituting (4.2) into (4.43):
∂ X X
3 3
T ∂
i i α
L = ē ⊗ v = ē ⊗ k̇α X ēα = k̇α ēα ⊗ ēα = L.
∂X i ∂X i
α=1 α=1
Since the velocity gradient L in this case proves to be a symmetric tensor, from
(4.45) and (4.46) it follows that
D = L, W = 0.
Thus, in this case ω = 0. ¤
Example 1.5. Determine the tensor L for the problem on simple shear (see
Example 1.2), substituting formula (4.2) into (4.43):
∂v ∂v̄ j i ∂v̄ 1
LT = ēi ⊗ = ē ⊗ ēj = ē2 ⊗ ē1 = ȧē2 ⊗ ē1 .
∂X i ∂X i ∂X 2
According to formulae (4.45) and (4.46), we get
D = (ȧ/2)(ē1 ⊗ ē2 + ē2 ⊗ ē1 ),

W = (ȧ/2)(ē1 ⊗ ē2 − ē2 ⊗ ē1 ) = (ȧ/2)(δ1i δ2j − δ2i δ1j )ēi ⊗ ēj .
Using formula (4.47), we determine the vorticity vector
1 ȧ ȧ ȧ
ω = W · · ² = (δ1i δ2j − δ2i δ1j )²jik ēk = (²21k − ²12k )ēk = − ē3 ,
2 4 4 2
which is orthogonal to the shear plane. ¤

1.4.6. Eigenvalues of the Deformation Rate Tensor. Just as any symmetric
tensor, the deformation rate tensor D has three orthonormal real-valued eigen-
vectors and three real positive eigenvalues (see [12]). Denote the eigenvectors
by qα (these vectors, in general, are not coincident with pα ) and the eigenvalues
by Dα . Then the tensor D can be resolved for its dyadic eigenbasis as follows:
X
3
D= Dα qα ⊗ qα , qα · qβ = δαβ . (4.50)
α=1
Take in the actual configurationK an elementary radius-vector dxα , connect-

ing points M and M0 , so that the vector is oriented along the eigenvector qα of
the tensor D; then, similarly to (3.36), we can write
dxα = qα |dxα |, |dxα | = (dxα · dxα )1/2 . (4.51)

Apply the CauchyHelmholtz theorem (4.49) to the elementary radius-vector:
b α = ω × dxα + D · dxα .
dv (4.52)
Multiplying the left and right sides of the equation by dxα and taking account
of the property of the mixed derivative dxα · (ω × dxα ) = 0, we get
b α = dxα · D · dxα .
dxα · dv (4.53)
Substituting in place of D its expression (4.50) and in place of dxα their
expressions (4.51), we obtain
X
3
b α = |dxα |2
dxα · dv Dβ qα · qβ ⊗ qβ · qα = Dα |qα |2 . (4.54)
β=1
Here we have used the property (4.50) of orthonormal vectors qα .

Modify the scalar product on the left-hand side as follows:
b α = dxα · ∂ dxα =
dxα · dv
1 ∂ ∂
(dxα · dxα ) = |dxα | |dxα |. (4.55)
∂t 2 ∂t ∂t
On comparing (4.54) with (4.55), we obtain the following theorem.
Theorem 1.21. Eigenvalues Dα of the deformation rate tensor D are the
rates of relative elongations of elementary material fibres oriented along the
eigenvectors qα :
1 ∂
Dα = |dxα |. (4.56)
|dxα | ∂t
1.4.7. Resolution of the Vorticity Tensor for the Eigenbasis of the Defor-
mation Rate Tensor. Modify the right-hand side of (4.52) as follows:
b α = ω × dxα + D · dxα = (ω × qα + Dα qα )|dxα |,

dv (4.57)
and the left-hand side of (4.52) with taking (4.56) into account:
³ ´
b α = ∂ dxα = ∂ (|dxα |qα ) = ∂|dxα | qα + |dxα | ∂qα = |dxα | Dα qα + ∂qα .
dv
∂t ∂t ∂t ∂t ∂t
(4.58)
On comparing (4.57) with (4.58), we get the following theorem.
Theorem 1.22. The vorticity tensor W (or the vorticity vector ω) connects the
rate of changing the eigenvectors qα to the vectors qα themselves:
∂qα
q̇α = = ω × qα = W · qα . (4.59)
∂t
Using formula (4.59), we can resolve the tensor W for the eigenbasis qα of
the deformation rate tensor as follows:
X
3
W= q̇α ⊗ qα = q̇i ⊗ qi . (4.60)

α=1
1.4.8. Geometric Picture of Infinitesimal Transformation of a Small Neigh-
borhood of a Point. If in configuration K at time t we consider an elementary
0
radius-vector dx connecting two infinitesimally close material points M and M ,
then for infinitesimal time dt the radius-vector is transformed into radius-vector
dx0 in configuration K(t + dt) (Figure 1.13):
dx = x0 (t) − x(t), dx0 = x0 (t + dt) − x(t + dt), (4.61)
where x(t) and x0 (t) M and M0 in configuration
are radius-vectors of the points
0
K(t), respectively; and x(t + dt) and x (t + dt) in configuration K(t + dt).
0
Displacements of points M and M for infinitesimal time are defined by the
0
velocity vectors v(M) and v(M ), respectively:
x(t + dt) − x(t) = v(M) dt, x0 (t + dt) − x0 (t) = v(M0 ) dt. (4.62)
Figure 1.13. Infinitesimal transformation of an elementary radius-vector

Formulae (4.61) and (4.62) and simple geometric relations (see Figure 1.13)
give
v(M0 )dt − v(M)dt = dx0 − dx. (4.63)
On substituting (4.63) into (4.49a), we obtain the relation between elementary
radius-vectors dx0 and dx:
0
dx = dx + dtω × dx + dtD · dx + dt o(|dx|). (4.64)
The relation (4.64) can be considered as the transformation of coordinates
dxi → dx0i in a small neighborhood of the point contained in a continuum. Since
dtω and dtD are independent of dx and dx0 , so the transformation is linear,
i.e. affine. The relation (4.64) can be represented as a superposition of two
transformations up to an accuracy of o(|dx|):
00
dx = AD · dx, AD = E + dtD, (4.65)
0 00
dx = Qω · dx , Qω = E + dtω × E. (4.66)
The tensor AD is symmetric and has three eigendirections, which are
coincident with the eigendirections qα of the deformation rate tensor D.
So just as the tensor U, the tensor AD transforms a small neighborhood of
a point M by extending or compressing the neighborhood along the principal
directions qα . The material segments |dx00α | oriented along the eigendirections qα
retain their orientation under the transformations (4.65), but their lengths vary
as follows:
dxα = |dxα |qα , dx00α = (1 + Dα dt|dxα |)qα = (1 + Dα dt)dxα .

The tensor Qω (4.66) is orthogonal up to an accuracy of values ∼ (dt)2 ,
because
Qω · QT T 2 2
ω = (E + dtW) · (E + dtW ) = E − (dt) W . (4.67)
Here we have taken into account that the vorticity tensor W is skew-symmetric.
Thus, the transformation (4.66) determined by the tensor Qω is rotation of
the M-point neighborhood as a rigid whole for infinitesimal time dt.
The vorticity vector ω forming the tensor Qω can be considered as instanta-
neous angular rate of rotation of the small neighborhood as a rigid whole, or as
instantaneous angular rate of rotation of the eigentrihedron qα of the deformation
rate tensor relative to the fixed basis ēi . This fact will be considered in detail in
paragraph 1.5.7.
On uniting the properties of the transformations (4.65) and (4.66), we can
make the following conclusion.
Theorem 1.23. The infinitesimal transformation of a small neighborhood of
the point contained in a continuum is a superposition of tension-compression
of the neighborhood along the eigendirections qα and rotation of the axes qα
as a rigid whole about the axis with the direction vector ω.
◦
Thus, we have the certain analogy between the eigendirections pα of the
tensor U qα of the tensor D: elementary material fibres
and the directions
◦
oriented along pα and along qα remain mutually orthogonal and undergo only
◦
tension-compression. The axes pα remain mutually orthogonal under any finite
◦
transformations from K to K, but qα only under infinitesimal transformations
from K(t) to K(t + dt).
1.4.9. Kinematic Meaning of the Vorticity Vector. Just as any orthogonal
tensor, the orthogonal tensor Qω of infinitesimal rotation from K(t) to K(t + dt)
can be represented in the form (see [12])
Qω = E cos(dϕ) + (1 − cos(dϕ))e ⊗ e − e × E sin(dϕ), (4.68)

where dϕ is the infinitesimal angle of rotation of the trihedron qα about the axis
with the direction vector e. Since values of dϕ are infinitesimal, we have
Qω = E − e × Edϕ. (4.69)
On comparing (4.69) with (4.66), we get
dϕ dϕ
ω=− e, |ω| = , (4.70)
dt dt
i.e. the vorticity vector ω is really oriented along the instantaneous rotation axis
e, and the length |ω| is equal to the instantaneous angular rate of rotation of the
trihedron qα of the deformation rate tensor.
Let us consider now the question: relative to what system the vorticity vector
ω defines the rotation rate.
To answer the question, we introduce another orthogonal rotation tensor
OW = qi ⊗ ēi , (4.71)
which transforms the Cartesian trihedron ēi as a rigid whole into the orthonor-
mal trihedron qi :
qi = OW · ēi . (4.72)
The tensor OW is a function of time t, because qi = qi (t).
According to (4.60) and (4.72), the vorticity tensor W takes the form
W = q̇i ⊗ qi = ȮW · OT
W. (4.73)
With the help of (4.73) we can represented the orthogonal tensor Qω as
follows:
Qω = E + dtW = E + dtȮW · OT
W. (4.74)
Thus, at each time t two orthogonal tensors OW and Qω connect local neigh-
◦
borhoods of a point M in the reference K and actual configurations K(t + dt).
◦
00
If in K we consider an elementary radius-vector dx , then in K we find its
◦ 00
corresponding radius-vector dx obtained with the help of the rotation tensor
OW , and in K(t + dt) radius-vector dx0 :
◦
dx00 = OW · dx00 , dx0 = Qω · dx00 .
A fixed observer connected to the Cartesian trihedron ēi sees both the trans-
formations: finite rotation for time t, which is described by the tensor OW , and
instantaneous rotation of a local neighborhood for time dt, which is described by
the infinitesimal rotation tensor Qω
Thus, the vorticity vector ω is the vector of instantaneous angular rate of

rotation of the trihedron qα relative to the trihedron ēi .
Comparing (4.66) with (4.74) (or (4.73) with (4.47)), we get
ȮW · OT
W = ω × E. (4.75)
1.4.10. Tensor of Angular Rate of Rotation (Spin). In paragraph 1.3 we
have introduced the tensor ȮW · OTW , where OW is the orthogonal tensor of
rotation. Such tensor Ω = Q̇ · Q
T can be set up for any orthogonal tensor Q
depending on time t.
The tensor Q̇ · QT is skew-symmetric, because
Q̇ · QT + Q · Q̇T = (Q · QT )• = (E)• = 0, (4.76)

i.e.
ΩT = (Q̇ · QT )T = Q · Q̇T = −Q̇ · QT = −Ω. (4.77)
. This tensor characterizes the angular rate of rotation of the orthonormal
trihedron hi formed with the help of Q:
hi = Q · ēi , (4.78)
relative to the Cartesian trihedron ēi .
Indeed, with the help of the tensor Q̇ · QT we can form the tensor (4.74):
Qω = E + dtQ̇ · QT , (4.79)
which, according to (4.67), is the orthogonal tensor of infinitesimal rotation; and
this tensor can be represented in the form (4.68) or (4.69):
Qω = E − dϕe × E, (4.80)
where dϕ is the infinitesimal angle of rotation of the trihedron hi about the axis
with the direction vector e. Comparing (4.79) with (4.80), we get the expression
dϕ
Q̇ · QT = − e × E, (4.81)
dt
which makes clear the sentence that the tensor Q̇ · QT characterizes the instan-
taneous angular rate dϕ/dt of rotation of the trihedron hi about the axis e.
The tensor Q̇ · QT is called the tensor of angular rate of rotation or the
spin.
Expressing the tensor Q from (4.78) in terms of the bases hi and ēi (hα =
= hα , α = 1, 2, 3, as the vectors are orthonormal):
Q = hi ⊗ ēi , (4.82)
we get another representation of the spin:
Q̇ · QT = ḣi ⊗ hi . (4.83)
Theorem 1.24. The spin connects the rates ḣi and the vectors hi defined by
formula (4.78) as follows:
ḣi = (Q̇ · QT ) · hi . (4.84)

Since the spin Q̇ · QT is a skew-symmetric tensor, we can introduce the

corresponding vorticity vector ωh:
Q̇ · QT = ω h × E (4.85)
(from formulae (4.81) and (4.85) it follows that ω h = −(dϕ/dt)e); then formula
(4.84) takes the form
ḣi = ω h × hi . (4.86)
Resolving the vector ωh for the orthonormal basis hi : ω h = ω jh hj , we get one
more representation of formula (4.84):
ḣi = ω jh hj × hi = ²jik ω jh hk . (4.87)

This formula can also be rewritten in the form
ḣα = ωhβ hγ , α 6= β 6= γ 6= α. (4.88)

Taking different orthogonal tensors (or orthonormal bases) as Q (or hi ), we
obtain different spins.
◦
1) If we choose eigenvectors of the stretch tensor U as hi , i.e. hi = pi , then,
according to (4.83), the corresponding spin ΩU takes the form
◦ ◦ ◦
ΩU = ȮU · OT • i
U = pi ⊗ p , OU = pi ⊗ ēi , (4.89)
and formulae (4.84) yields
◦˙ ◦
pi = ΩU · pi . (4.90)
2) If hi = pi , then the corresponding spin ΩV and the rotation tensor OV
have the forms
ΩV = ȮV · OT i
V = ṗi ⊗ p , OV = pi ⊗ ēi , (4.91)
ṗi = ΩV · pi . (4.92)
3) If hi = qi , then the corresponding spin ΩW coincides with the vorticity
tensor W (see (4.73)):
ΩW = ȮW · OTW = q̇i ⊗ qi = W, (4.93)
q̇i = ΩW · qi . (4.94)
4) If we take the rotation tensor O accompanying deformation as Q, then, as
◦
shown in (3.23a), the tensor O connects two moving bases pi and pi :
◦
pi = O · pi . (4.95)
The tensor O can be expressed in terms of OV and OU as follows:
◦ ◦
O = pi ⊗ pi = pi ⊗ ēi · ēj ⊗ pj = OV · OT
U. (4.96)
The corresponding spin Ω has the form
Ω = Ȯ · OT = (ȮV · OT T T
U + OV · ȮU ) · OU · OV =
= ȮV · OT T T T
V + OV · ȮU · OU · OV = ΩV − OV · ΩU · OV . (4.97)
Unlike the cases 1)3), the spin tensor Ω characterizes the angular rate of
◦
rotation of the trihedron pi relative to the moving trihedron pi , but not relative
to the trihedron ēi being fixed.
Therefore, for the cases 1)3) the spins characterize the total angular rate,
and for the case 4) the relative rate.
1.4.11. Relationships between Rates of Deformation Tensors and Velocity

Gradients. In continuum mechanics, one often needs the relations between
rates of the deformation tensors (and also measures) and the velocity gradients
L = (∇ ⊗ v)T and
◦ ◦
L = (∇ ⊗ v)T . (4.98)
Let us derive these relations.
Theorem 1.25. The rates of varying the gradient Ḟ and the inverse gradient
◦
(F−1 )• are connected to L and L by the relations
◦
Ḟ = L · F, Ḟ = L,
◦
(F−1 )• = −F−1 · L, (F−1 )• = −F−1 · F−1T · L. (4.99)
H Differentiating the relationships (1.35a) with respect to time t and taking the
definition of the velocity (4.1) into account, we get
◦
ḞT = ∇ ⊗ v = FT · ∇ ⊗ v, Ḟ = (∇ ⊗ v)T · F. (4.99a)
◦
According to the definitions of tensors L (4.43) and L (4.98), from (4.99a) we
obtain formulae (4.99).
Differentiating the identity (F · F−1 )• = Ė = 0, we find that Ḟ · F−1 =
= −F · (F−1 )• ; whence we get
(F−1 )• = −F−1 · Ḟ · F−1 . (4.100)
On substituting the first two formulae of (4.99) into (4.100), we obtain
(F−1 )• = −F−1 · (∇ ⊗ v)T , (F−1T )• = −(∇ ⊗ v) · F−1T , (4.101)

i.e. the third and the fourth relationships of (4.99) hold as well. N
According to formulae (4.45), (4.99a) and (4.101), we find that the rate of
the deformation gradient is connected to the deformation rate tensor D by the
relations
1 1
D = (Ḟ · F−1 + F−1T · ḞT ), D = − (F · Ḟ−1 + Ḟ−1T · FT ). (4.102)
2 2
Here and below we will use the notation Ḟ−1 ≡ (F−1 )• .

Theorem 1.26. The rates of the deformation tensors Ċ, Ȧ, Λ̇, J̇ and deforma-
tion measures Ġ, ġ, (G
−1 )• and (g−1 )• are connected to the velocity gradients
◦
L and L by the relationships
½
Ċ = FT · D · F, Ȧ = D − LT · A − A · L,
(4.103)
Λ̇ = F−1 · D · F−1T , J̇ = D + L · J + J · LT ,
and
½
Ġ = 2FT · D · F, ġ = −LT · g − g · L,
(4.104)
(G−1 )• = −2F−1 · D · F−1T , (g−1 )• = L · g−1 + g−1 · LT ;
and also
 ◦ ◦ ◦ ◦
 T T J̇ = (1/2)(L · FT + FT · L),
Ċ = (1/2)(F · L + L · F),

◦ ◦
Ȧ = (1/2)((E − 2A) · L · F−1 + F−1T · LT · (E − 2A)), (4.105)


 ◦ ◦
Λ̇ = (1/2)(F−1 · L · (E − 2Λ) + (E − 2Λ) · LT · F−1T ),
and
 ◦ ◦ ◦ ◦
 T T ġ = −(g · L · F−1 + F−1T · LT · g),
Ġ = F · L + L · F

◦
,
◦
(G−1 )• = −(F−1 · L · G−1 + G−1 · LT · F−1T ), (4.106)


 −1 • ◦ ◦
(g ) = L · FT + FT · L.
H To prove formula (4.103), we must differentiate the relationships (2.5) with
respect to t and apply formulae (4.99):
1 1
Ċ = (ḞT · F + FT · Ḟ) = (FT · LT · F + FT · L · F) =
2 2
1
= FT · (LT + L) · F = FT · D · F, (4.107)
2
1
Ȧ = (Ḟ−1T · F−1 + F−1T · Ḟ−1 ) =
2
1
= (LT · F−1T · F−1 + F−1T · F−1 · L) =
2
1
= (L · (E − 2A) + (E − 2A) · L) = D − LT · A − A · L,
T
(4.108)
2
1 1
Λ̇ = − (Ḟ−1 · F−1T + F−1 · Ḟ−1T ) = (F−1 · L · F−1T +
2 2
−1 T −1T 1 −1
+F ·L ·F )= F · (L + LT ) · F−1T = F−1 · D · F−1T , (4.109)
2
1 1
J̇ = (Ḟ · FT + F · ḞT ) = (L · F · FT + F · FT · LT ) =
2 2
1
= (L · (E + 2J) + (E + 2J) · LT ) = D + L · J + J · LT . (4.110)
2
Formulae (4.104) follow from (4.103), if we have used the connections (2.7) and
(2.8) between the deformation tensors and measures.
◦
Formulae (4.105) follow from (4.107)(4.110), if we have gone from L to L:
◦
L = L · F−1 . (4.111)
Using the connections (2.7) and (2.8) between the deformation tensors and
measures, from (4.105) we get formulae (4.106). N
◦
Relations between the tensors U̇, V̇ and the velocity gradients L and L are
◦
more complicated. To derive them, we should use representations of L and L in
terms of the eigenbasis vectors.
Theorem 1.27. The following expressions for the velocity gradients hold:
◦ ∂ ◦ X
3
◦ ◦
T
L =∇⊗v = ∇⊗x= λ̇α pα ⊗ pα + ΩU · FT − FT · ΩV , (4.112)
∂t
α=1
X
3 µ ¶
T λ̇α λ ◦
L = (∇ ⊗ v) = δαβ + α ΩU βα pα ⊗ pβ + ΩV , (4.113)
λα λβ
α,β=1
◦
where ΩU αβ are components of the tensor ΩU with respect to the eigenbasis
◦
pα :
◦ ◦ X
3
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
ΩU = p•i ⊗ pi = ΩU αβ pα ⊗ pβ , ΩU αβ = pα · ΩU · pβ = p•α · pβ . (4.114)
α=1
H To prove formula (4.112), we should consider the first formula of (4.99a) and
substitute into this formula the expression (3.35) for F in the eigenbasis:
◦ X
3
◦ ◦ ◦
LT = ḞT = (λ̇α pα ⊗ pα + λα (p•α ⊗ pα + pα ⊗ ṗα )). (4.115)
α=1
Using formulae (4.90) and (4.92), from (4.115) we derive the relationship
(4.112).
To prove formula (4.113), we use formulae (4.112) and (4.111), having
expressed F−1 in the form (3.35); then we get
◦ X
3
◦ ◦ ◦ X
3
◦
L = L · F−1 = ( λ̇α pα ⊗ pα + λα (pα ⊗ p•α + ṗα ⊗ pα )) · λ−1
β pβ ⊗ pβ =
α=1 β=1
X
3 ³ λ̇ λα ◦ • ◦ X ´ 3
α
= δαβ pα ⊗ pα + (pα · pβ )pα ⊗ pβ + ṗα ⊗ pα . (4.116)
λβ λβ
α,β=1 α=1
◦
Here we have taken into account that the vectors pα are orthonormal. Using
formulae (4.90) and (4.92), from (4.116) we derive the formula (4.113). N
From formula (4.113) it follows that the deformation rate tensor D and the
vorticity tensor can be represented in terms of the eigenbasis pα as follows:
X
3 µ µ ¶ ¶
λ̇α 1 λα λ ◦
D= δαβ + − β ΩU βα pα ⊗ pβ , (4.117)
λα 2 λβ λα
α,β=1
X
3 µ ¶
1 λα λ ◦
W= + β ΩU βα pα ⊗ pβ + ΩV . (4.118)
2 λβ λα
α,β=1
Here we have taken into account that the tensors ΩU and ΩV are skew-
symmetric.
Denote components of the tensor D with respect to the basis pα by Dαβ :
X
3
D= Dαβ pα ⊗ pβ , Dαβ = pα · D · pβ . (4.119)

α,β=1
Then from (4.117) and (4.119) we get that diagonal components of the de-
formation rate tensor Dαα with respect to the eigenbasis pα determine the
relative rates of lengthening the material fibres oriented along the eigenvectors
pα (compare with formula (4.56)):
Dαα = λ̇α /λα = dṡα /dsα , α = 1, 3; (4.120)

◦
and off-diagonal components Dαβ are connected to ΩU αβ by the relations
³ 2
1 λα − λβ
2 ´ ◦
Dαβ = ΩU βα , α 6= β. (4.121)
2 λα λβ
◦
From formulae (4.119) and (4.121) we can express the components ΩU αβ in
terms of components of the deformation rate tensor:
◦ 2 λα λβ ◦
ΩU αβ = Dαβ , α 6= β ; ΩU αα = 0. (4.122)
λβ − λα
2 2
◦
The diagonal components ΩU αα are equal to zero, because the tensor ΩU is
skew-symmetric.
On substituting the relationships (4.122) into (4.118), we find the expression
◦
for components ΩV αβ of the tensor ΩV with respect to the basis pα in terms of
the tensors W and D (and, hence, in terms of the velocity gradient L):
◦ λ2α + λ2β
ΩV αβ = pα · ΩV · pα = pα · W · pβ − Dαβ , α 6= β , (4.123)
λ2α − λ2β
where
X
3
ΩV = ΩV αβ pα ⊗ pβ . (4.124)
α,β=1
Remark. The expressions (4.122) and (4.123) are valid only if the eigenvalues
are not multiple: λα 6= λβ (α 6= β ; α, β = 1, 2, 3). If within the interval [t1 , t2 ] all
three eigenvalues are coincident: λα = λ (α = 1, 2, 3), then the stretch tensors
◦ ◦i
are spherical: U = λpi ⊗ p = λE, V = λE, and the eigenbases are not uniquely
◦
defined: as pi and pi we can take any orthonormal triple of vectors. In particular,
◦
one of the bases can be taken as fixed ∀t ∈ [t1 , t2 ], for example, pi can be chosen
◦
as coincident with pi (t1 ); and the second basis pi can depend on time t. In this
◦•
case pi ≡0 ∀t ∈ [t1 , t2 ], and from (4.114) and (4.118) it follows that within the
considered time interval:
◦
ΩU αβ = 0, ΩU = 0, (4.125)
ΩV = W, ΩV αβ = pα · W · pβ , α, β = 1, 2, 3.
These relationships take the place of formulae (4.122), (4.123) in this case.
If within the time interval [t1 , t2 ] only two of three eigenvalues are coincident,
◦ ◦
for example, λα = λβ , then their corresponding eigenvectors pα and pβ are not
◦
uniquely defined as well: only their orthogonality to the vector pγ , corresponding
◦
to the third eigenvalue λγ , is given. Then we can extend the definition of pα and
◦ ◦
pβ so that p•α · pβ = 0 ∀t ∈ [t1 , t2 ]. In this case it follows from (4.114) that the
◦ ◦ ◦
only component ΩU αβ vanishes, but ΩU αγ 6= 0 and ΩU βγ 6= 0.
It follows from (4.118) that the component ΩV αβ is determined by the
formula
◦
ΩV αβ = pα · W · pβ , ΩU αβ = 0. (4.126)
◦ ◦
The remaining components ΩU αγ , ΩU βγ and ΩV αγ , ΩV βγ are determined by
formulae (4.122) and (4.123).
If the situation with multiple roots appears only at some time t, then the
◦
values of ΩU αβ (t) and ΩV αβ (t) can be determined by passing to the limit. ¤
Substituting formulae (4.114) and (4.118) into equation (4.97), and taking
expressions (4.89) and (4.91) for OU andOV into consideration, we obtain the
representation of the spin Ω in the basis pα :
1
X
3
λ2α + λ2β ◦
Ω=W− ΩU βα pα ⊗ pβ −
2 λα λβ
α,β=1
X
3
◦ ◦ ◦
− ΩU αβ (pi ⊗ ēi ) · pα ⊗ pβ · (ēj ⊗ pj ). (4.127)
α,β=1
Introducing the notation for direction cosines
◦ ◦
l αβ = pα · ēβ , lαβ = pα · ēβ , (4.128)
substituting (4.122) into (4.127) and collecting like terms, we obtain the follow-
ing expression of the spin Ω in terms of W and D (i.e. in terms of L):
e,
Ω=W+Ω (4.129)
X
3
X
3
e =
Ω e γρ pγ ⊗ pρ ,
Ω e γρ =
Ω e αβ ,
Ω
γ ,ρ=1 α,β=1
X
3
1
e γρ =
Ω ((λ2α + λ2β )δαγ δβρ − 2λα λβ lαγ lβρ )Dαβ .
λ2β − λ2α
α,β=1
Theorem 1.28. Rates of the deformation measures U̇, (U−1 )• and V̇, (V−1 )•
are connected to the velocity gradient L by the formulae
1
e T ) · F),
e · O + OT · (D + Ω
U̇ = (FT · (D + Ω)
2
(U −1 • 1
) = − (F−1 · (D − Ω) e T ) · F−1T ),
e · O + OT · (D − Ω
2
1
V̇ = ((L + Ω) · V + V · (LT + ΩT )), (4.130)
2
1
(V−1 )• = ((Ω − LT ) · V−1 + V−1 · (ΩT − L)).
2
H Let us express the tensors V and U from the polar decomposition (3.1):
U = OT · F, V = F · OT . (4.131)
Since U and V are symmetric tensors, these expressions can be rewritten in the
symmetrized form
1 1
U = (FT · O + OT · F), V = (F · OT + O · FT ). (4.132)
2 2
Let us differentiate these relationships:
1
U̇ = (ḞT · O + FT · Ȯ + ȮT · F + OT · Ḟ),
2
1
V̇ = (Ḟ · OT + F · ȮT + Ȯ · FT + O · ḞT ). (4.133)
2
On substituting formulae (4.99) and the expression (4.97) for the spin Ω into
(4.133), we obtain
1
U̇ = (FT · LT · O + FT · Ȯ · OT · O + OT · O · ȮT · F + OT · L · F) =
2
1
= (FT · (LT + Ω) · O + OT · (L + ΩT ) · F), (4.134)
2
1
V̇ = (L · F · OT + F · OT · O · ȮT + Ȯ · OT · O · FT + O · FT · LT ) =
2
1
= ((L + Ω) · V + V · (LT + ΩT )).
2
Taking formulae (4.129) and (4.44) into consideration, we find that
e = D + Ω.
LT + Ω = D + WT + W + Ω e (4.135)
Substituting (4.135) into (4.134), we really get the first and the third formulae
of (4.130).
The remaining two formulae in (4.130) can be proved in a similar way. N
In continuum mechanics the deformation tensors B and Y are applied, which

have no explicit expression; they are defined by their derivatives and initial
values:
1
Ḃ = (U̇ · U−1 + U−1 · U̇), B(0) = 0, (4.136)
2
1
Ẏ = (V̇ · V−1 + V−1 · V̇), Y(0) = 0. (4.137)
2
After substitution of the expressions (4.131), formula (4.136) takes the form
1
Ḃ = ((ȮT · F + OT · Ḟ) · U−1 + U−1 · (ḞT · O + FT · Ȯ)) =
2
1
= (OT · (Ω + L) · F · U−1 + U−1 · FT · (LT + ΩT ) · O) =
2
1
= (OT · (Ω + L) · O + OT · (LT + ΩT ) · O), (4.138)
2
and formula (4.137) is rewritten as follows:
1
Ẏ = ((Ḟ · OT + F · ȮT ) · V−1 + V−1 · (Ȯ · FT + O · ḞT )) =
2
1
= (L · F · OT · V−1 + F · OT · ΩT · V−1 + V−1 · Ω · O · FT +
2
−1 1
+V · O · FT · LT ) = (L + V · Ω · V−1 + V−1 · Ω · V + LT ). (4.139)
2
Finally, we get the following expressions for Ḃ and Ẏ:

Ḃ = OT · D · O, (4.140)
1
Ẏ = D + (V · ΩT · V−1 + V−1 · Ω · V). (4.141)
2
1.4.12. Trajectory of a Material Point, Streamline and Vortex Line. Hav-

X i of a material point M in the motion law (1.3), we get
ing fixed coordinates
the parametric equation of a certain curve, where time t is a parameter:
xi = xi (X k , t), 0 6 t 6 t0 . (4.142)
The origin of the curve at t = 0 is a point with
◦
Cartesian coordinates xi (X k ) of the material
◦
point M in K, and the end of the curve at t = t0
i k 0
is a point with Cartesian coordinates x (X , t )
0
of the point M in K(t ) (Figure 1.14). The
curve (4.142) is called the trajectory of the
point M in the Cartesian coordinate system
Figure 1.14. Trajectory of material
Oēi . point M
In the spatial description, the trajectory
(4.142) at fixed Xk is a solution of the kine-
matic equation (4.10):
dxi /dt = v̄ i (xj , t), 0 < t 6 t0 (4.143)

with the initial condition

◦
t=0: xi = xi .
Here v̄ i (xj , t) are the velocity components with respect to the Cartesian basis
ēi , which are assumed to be known.
v(xj , t) = v̄ i ēi be given. Fix a time t, and take
Let a field of velocities a
i
point M1 with Eulerian coordinates x1 and Lagrangian coordinates X1 . Then
k a
streamline passing through the point M1 is the curve
xi = xi (X k , τ ), τ1 6 τ 6 τ2 , (4.144)
which at its every point xi has a tangent being parallel to the velocity v(xi , t) at
the considered point and at the considered time. The equation of the streamline
has the form
dxi /dτ = v̄ i (xj , t), τ1 < τ 6 τ2 , (4.145)
τ = τ1 : xi = xi1 .
Thus, the trajectory of a material point and the streamline are described, in
general, by different equations, and so they are not coincident.
However, if the motion of a continuum is steady-state within time interval
t1 6 t 6 t0 , then in Eulerian description all the partial derivatives of all values,
describing the motion, with respect to time vanish, in particular ∂v(xi , t)/∂t = 0.
So the trajectory equations (4.143) and the streamline equation (4.145) become
coincident within the interval t1 6 t 6 t0 , if they have at least one common point
M1 :
dxi /dτ = v̄ i (xj ), t 1 = τ1 < τ 6 τ2 = t 0 , (4.146)
τ = τ1 : xi = xi1 = xi (X1k , t1 ).
In other words, in the steady-state motion a material
point M moves along a streamline: at time t = t1 its
coordinates are the same as coordinates of point M1
at parameter value τ = τ1 , and at time t = t2 they are
the same as coordinates of point M2 at parameter
value τ = τ2 (Figure 1.15).
Multiplying the equation (4.145) by the basis
vectors ēi , we can rewrite the streamline equation in
Figure 1.15. The streamline
the vector form
dx = v(x, t)dτ , τ1 < τ 6 τ2 ,

x = x1 , τ = τ1 . (4.147)
Let us define now a vortex line passing through a point M1 : this is a
curve, which at its every point xi has a tangent being parallel to the vorticity
vector ω(xj , t) at the considered point and at the fixed time t. The vortex line is
described by the equation
dx = ω(x, t)dτ , τ1 < τ 6 τ2 , (4.148)

x = x1 , τ = τ1 .
1.4.13. Stream Tubes and Vortex Tubes. Consider a curve L in coordinates

xi and draw a streamline through each point of the curve L. If L is not a
streamline itself, then we get a surface Σv , at each point of which the velocity
v lies on the tangent plane to the surface. This surface is called the stream
surface.
Let
fv (xi ) = 0 (4.149)
be the equation of the stream surface. Since the vector ∇f is normal to the
surface Σv [12], so it is orthogonal to the velocity v, i.e. we have the relation
v · ∇fv = 0, (4.150)
which is a partial differential equation for determination of the function f (xi ) by
i
the known velocity field v(x , t) at fixed t.
If a curve L is closed, then the set of streamlines drawn through its points
is called the stream tube.
Let a curve L be not a vortex line. Drawing a vortex line through each point
of the curve L, we obtain the vortex surface Σω , which is described by the
equation fω (xi ) = 0. This relation is a solution of the differential equation
ω · ∇fω = 0. (4.151)
If L is a closed curve, then the surface Σω is called the vortex tube.
Exercises for 1.4.
Exercise 1.4.1. Show that the tensors QU and QV are orthogonal.

◦
Exercise 1.4.2. Using formulae (4.104) and (2.57), show that for the coefficient k
determined by formula (2.54) we have the following relationship:
◦ ◦
k • = k(n · D · n).
Exercise 1.4.3. Using formulae (2.57), (4.99) and the result of Exercise 1.4.2, show
that a rate of changing the normal n is determined as follows:
ṅ = γn − n · L, γ = n · D · n.
Exercise 1.4.4. Using the results of Exercise 1.4.2, show that for the coefficient k
determined by formula (2.54) we have the following equation:
k̇ = −k(n · D · n).
Exercise 1.4.5. Show that the transformations (4.65) of infinitesimal tension

compression (4.65) and infinitesimal rotation (4.66) are commutative up to terms of
order (dt)2 :
Qω · AD · dx = Aω · QD · dx,
while transformations of a small neighborhood determined by the tensors O and U or
O and V are not commutative in general.
Exercise 1.4.6. Prove Theorem 1.14.

Exercise 1.4.7. Using the representation (3.6) for tensors U and V and formulae (4.90)
and (4.92), show that rates of stretch tensors are expressed as follows:
X
3
◦ ◦ X
3
U̇ = λ̇α pα ⊗ pα + ΩU · U − U · ΩU , V̇ = λ̇α pα ⊗ pα + ΩV · V − V · ΩV .
α=1 α=1
Exercise 1.4.8. Show that expressions for rates of the Hencky tensors (3.31) have the
form
◦ X
3
λ̇α ◦ ◦ ◦ ◦ X
3
λ̇α
H• = pα ⊗ pα + ΩU · H − H · ΩU , H• = pα ⊗ pα + ΩV · H − H · ΩV .
λα λα
α=1 α=1
Exercise 1.4.9. Using representations (4.114) and (4.124) for tensors ΩU and ΩV , and
also equation (4.120) and the result of Exercise 1.4.8, show that rates of the Hencky
tensors can be expressed in the form
◦ X
3
X
3 ³ ´
◦ ◦ 2λα λβ λβ ◦ ◦
H• = Dαβ pα ⊗ pβ + lg − 1 Dαβ pα ⊗ pβ ,
2
λβ − λα 2
λα
α,β=1 α,β=1
α6=β
X
3 ³ ´
2λα λβ λβ
Ḣ = D + lg − 1 (pα · D · pβ )pα ⊗ pβ .
λ2β − λ2α λα
α,β=1
α6=β
◦
Exercise 1.4.10. Show that expressions for H• and Ḣ derived in Exercise 1.4.9 can be
rewritten as follows:
◦ ◦
H• = 4 XH · · D, Ḣ = 4 XH · ·D,
where the following fourth-order tensors are denoted:
◦ ◦ ◦
4
XH = XH ijkl pi ⊗ pj ⊗ pk ⊗ pl , 4 XH = XH ijkl pi ⊗ pj ⊗ pk ⊗ pl ,
³ ´
 2λα λβ lg λβ ∆
αβkl , α 6= β ,
XH ijkl = λβ − λα λα 2 2
∆αβkl = (1/2)(δαk δβl + δαl δβk ).
∆ , α = β ,
αβkl
Exercise 1.4.11. Show that relations (4.114) and (4.122) for ΩU , equations (4.123)
and (4.124) for ΩV and equation (4.129) for Ω can be rewritten as follows:
ΩU = 4 ΩU · · D, ΩV = 4 ΩV · · D + W, e · · D + W,
Ω = 4Ω
where
◦ ◦
4
ΩU = ΩU ijkl pi ⊗ pj ⊗ pk ⊗ pl , 4
ΩV = ΩV ijkl pi ⊗ pj ⊗ pk ⊗ pl ,
e =Ω
Ω e ijkl pi ⊗ pj ⊗ pk ⊗ pl ,
4
 
 2λ α λ β ∆
2 2
 λα + λβ ∆
αβkl , α 6= β , αβkl , α 6= β ,
ΩU αβkl = λβ − λα2 2
ΩV αβkl = λβ − λα 2 2
0, α = β, 
0, α = β,
1.5. Co-rotational Derivatives 61
e αβγρ = ΩV αβγρ − λγ λρ (lγα lρβ − lρα lγβ ),

Ω α 6= β , γ 6= ρ,
λ2ρ − λ2γ
if α=β (or γ = ρ), then the first (or the second) summand vanishes.
Exercise 1.4.12. Using the definitions (4.45) and (4.46) of the tensors D and W,
and also the properties of unit tensors, show that D, W and L are connected by the
formulae
D = ∆ · · L, e · · L,
W=∆ e = (1/2)(∆III − ∆II ).
∆
1.5. Co-rotational Derivatives
1.5.1. Definition of Co-rotational Derivatives. Besides the total derivative

da/dt introduced in paragraph 1.4.1 and partial derivative of vectors and tensors
with respect to time ∂a/∂t, so-called co-rotational derivatives are of great
importance in continuum mechanics. They determine rates of changing tensors
relative to some moving basis hi , i.e. the relative rates.
Let in a actual configurationK(t) there be some moving bases hi or hi
i i
and arbitrary varying scalar ψ(X , t), vector a(X , t) and second-order tensor
i
T(X , t) fields with the following components with respect to the bases:
a = ai hi = ai hi , (5.1)
ij i j j i
T = T hi ⊗ hj = Tij h ⊗ h = T ij hi j
⊗ h = Ti h ⊗ hj . (5.2)
Since any scalar function ψ(X i , t) is not connected to any basis (moving or
fixed), it is evident that the co-rotational derivative of the function must be
coincident with the total derivative with respect to time:
ψ h = ψ̇. (5.3)
a and a tensor T we can introduce co-rotational derivatives ah
For a vector
h
and T as vectors or tensors, components of which with respect to the same
basis hi coincide with rates of changing vector a and tensor T components,
respectively:
dai d ij
ah = hi , Th = T hi ⊗ hj . (5.4)
dt dt
If we consider the basis hi , then for the basis we can determine other co-
rotational derivatives:
dai i
aH = h, (5.5)
dt
d
TH = Tij hi ⊗ hj . (5.6)
dt
Thus, the co-rotational derivative ah (or Th ) determines the rate of varying
a vector a (or a tensor T) for an observer moving together with the basis
hi . For the observer, the basis hi is fixed, and hence in (5.4) the basis is not
differentiated with respect to time. In a similar way, the derivatives aH and TH
determine the rates of changing a and T for an observer moving together with
the basis hi .
We can determine the co-rotational derivatives of a second-order tensor T in

mixed moving dyadic bases hi ⊗ hj and hi ⊗ hj , respectively:
d i d j i
Td = T hi ⊗ hj , TD = T h ⊗ hj . (5.7)
dt j dt i
Since vector components ai and ai can always be expressed in the form
i i
a =a·h , ai = a · hi , (5.8)
and tensor components Tij , T ij , T ij and Ti j , with the help of the scalar product
i
of (5.2) by h or hj , can be written as follows:
T ij i
=h ·T·h j
, Tij = hi · T · hj , T ij = hi · T · hj , Ti j = hi · T · hj ,
(5.9)
so rates of changing vector and tensor components in (5.4), (5.5) and (5.7) can
be represented in the explicit form:
dai da dhi dai da dh

= · hi + a · , = · hi + a · i , (5.10)
dt dt dt dt dt dt
and also
dT ij dT dhi dhj
= hi · · hj + · T · hj + hi · T · ,
dt dt dt dt
dTij dT dh dh
= hi · · hj + i · T · hj + hi · T · j ,
dt dt dt dt
i
d i d dh dh
T = hi · T · hj + · T · hj + hi · T · j , (5.11)
dt j dt dt dt
d j d dh d
T = hi · T · hj + i · T · hj + hi · T · hj .
dt i dt dt dt
Here total derivatives da/dt and dT/dt are determined by the rules (4.7) and
(4.12), respectively. Rates of changing basis vectors dhi /dt and dhi /dt are
j
defined by the choice of basis hi or h .
j
Taking different bases as hi and h , we get different co-rotational derivatives.
Let us consider the most widely used bases.
1.5.2. The Oldroyd Derivative(hi = ri ). If we choose the general local vector

h
basis ri as hi , then the derivative a = a
Ol (or Th = TOl ) determines the rate
i
of changing a (or T) relative to the Lagrangian coordinate system X moving
together with the continuum. This derivative is called the Oldroyd derivative.
The derivative dri /dt is determined as follows:
dhi dr ∂2x ∂v ∂v
= i = = = ri · rj ⊗ = ri · ∇ ⊗ v = (∇ ⊗ v)T · ri . (5.12)
dt dt ∂t∂X i ∂X i ∂X j
In this case, as a basishi we consider the reciprocal local basis ri , the
i
derivative of which dr /dt with respect to time has the form
d i d
(r ⊗ ri ) = E = 0, (5.13)
dt dt
or
dri dr
⊗ ri = −ri ⊗ i = −ri ⊗ (∇ ⊗ v)T · ri . (5.14)
dt dt
Multiplying the equation by rj from the right, we get
j j
dh dr
= = −rj · (∇ ⊗ v)T = −(∇ ⊗ v) · rj . (5.15)
dt dt
On substituting the expressions (5.15) for the derivatives dhi /dt and dhj /dt
into (5.10), we get the formula for the Oldroyd derivative in the basis ri :
dai da i d
aOl = ri = · r ⊗ ri − a · (∇ ⊗ v) · ri ⊗ ri = a − a · ∇ ⊗ v, (5.16)
dt dt dt
dTij dT ij d
TOl = ri ⊗ rj = ri ⊗ rj = ri ⊗ ri · T · rj ⊗ rj −
dt dt dt
− ri ⊗ ri · (∇ ⊗ v)T · T · rj ⊗ rj − ri ⊗ ri · T · (∇ ⊗ v) · rj ⊗ rj =
d
= T − T · ∇ ⊗ v − (∇ ⊗ v)T · T. (5.17)
dt
Here we have taken into account that ri ⊗ ri = E. Thus, we have proved the
following theorem.
Theorem 1.29. The Oldroyd derivative is related to the total derivative with
respect to time as follows (for a vector a and for a tensor T, respectively):
aOl = ȧ − a · ∇ ⊗ v, TOl = Ṫ − T · ∇ ⊗ v − (∇ ⊗ v)T · T. (5.18)

1.5.3. The CotterRivlin Derivative (hi = ri ). If we choose the reciprocal
i i
local basis r as a moving basis h , then the derivative a
H (or TH ) characterizes
i
the rate of changing a (or T) relative to the basis r moving together with
i
the Lagrangian coordinate system X . This derivative is called the Cotter-Rivlin
derivative.
Because of formulae (5.10) and (5.15), we get the following theorem.
Theorem 1.30. The CotterRivlin derivative is related to the total derivative
as follows (for a vector a and for a tensor T, respectively):
dai i
aH ≡ aCR = r = ȧ + (∇ ⊗ v) · a, (5.19)
dt
dTij i
TH ≡ TCR = r ⊗ rj = Ṫ + ∇ ⊗ v · T + T · (∇ ⊗ v)T . (5.20)
dt
1.5.4. Mixed Co-rotational Derivatives. Since any vector a is defined by its
components with respect to a vector basis, for example, in a moving basis hi or
hj , so for the vector in the moving bases we can determine only two co-rotational
derivatives: by Oldroyd and by CotterRivlin.
Any second-order tensor T is defined by its components with respect to a
dyadic basis. Therefore, besides the Oldroyd and CotterRivlin derivatives, which
specify the rates of changes of a tensor T in moving dyadic bases ri ⊗ rj and
ri ⊗ rj , by formulae (5.7) we can determine two more derivatives in moving
mixed dyadic bases:
dT ij d j i
Td = ri ⊗ rj , TD = T r ⊗ rj . (5.21)
dt dt i
On substituting the expressions (5.12) and (5.15) into (5.11), we get the
following formulae for the rates of changing mixed components of the tensor T:
d i
T = ri · Ṫ · rj − ri · (∇ ⊗ v)T · T · rj + ri · T · (∇ ⊗ v)T · rj ,
dt j
d j
T = ri · Ṫ · rj + ri · (∇ ⊗ v) · T · rj − ri · T · (∇ ⊗ v) · rj . (5.22)
dt i
Having substituted (5.22) into (5.21), we get the following theorem.
Theorem 1.31. The mixed derivatives (5.21) are connected to the total deriva-
tive by the relations
Td = Ṫ − L · T + T · L, TD = Ṫ + LT · T − T · LT . (5.23)
The derivatives (5.21) are called the left and right mixed co-rotational
derivatives, where L = (∇ ⊗ v)T is the velocity gradient (see (4.43)).
It should be noticed that, unlike other co-rotational derivatives considered in
this paragraph, the mixed derivatives Td and TD do not form a symmetric tensor
when they are applied to a symmetric tensor T. This fact explains a scarcer
application of mixed derivatives in continuum mechanics.
◦
1.5.5. The Derivative Relative to the Eigenbasis pi of the Right Stretch
◦
Tensor. If we choose the eigenbasis pi of the right stretch tensor U as
◦ i
a moving basis hi , then, since pi are orthonormal, we get that h and hj
α i
are coincident: h = hα , α = 1, 2, 3, and |h | = 1. At every time, the moving
◦
coordinate system defined by the trihedron pi executes an instantaneous rotation,
which is characterized by the spin ΩU (4.89), and due to (4.90) we have
dhi ◦˙ ◦ ◦ ◦
= pi = ΩU · pi = pi · ΩT
U = −pi · ΩU . (5.24)
dt
On substituting (5.24) into (5.11), we get
dai ◦ ◦ ◦ ◦ ◦
ah ≡ aU = pi = ȧ · pi ⊗ pi + a · ΩU · pi ⊗ pi = ȧ + a · ΩU , (5.25)
dt
T ij ◦ ◦ ◦ dT ij ◦ ◦ ◦ ◦ ◦
Th ≡ TU = pi ⊗ pj = pi ⊗ pj = pi ⊗ pi · Ṫ · pj ⊗ pj −
dt dt
◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦
− pi ⊗ pi · ΩU · T · pj ⊗ pj + pi ⊗ pi · T · ΩU · pj ⊗ pj =
= Ṫ − ΩU · T + T · ΩU . (5.26)
The co-rotational derivative of a vector a (or a tensor T) determined by

(5.26) is called the right derivative relative to the eigenbasis.
Theorem 1.32. The right derivative relative to the eigenbasis is connected to
the total derivative as follows (for a vector a and for a tensor T, respectively):
aU = ȧ + a · ΩU , TU = Ṫ − ΩU · T + T · ΩU . (5.27)
1.5.6. The Derivative in the Eigenbasis (hi = pi ) of the Left Stretch

Tensor. Take the eigenbasis pi of the left stretch tensor V as a moving basis
hi and define the following co-rotational derivatives
dai dT ij
aH ≡ aV = pi , TH ≡ TV = pi ⊗ pj , (5.28)
dt dt
called the left derivatives in the eigenbasis.
Theorem 1.33. The left derivatives (5.28) in the eigenbasis are connected to
the total derivative with respect to time by the following relations (for a vector
a and for a tensor T, respectively):
V
a = ȧ − ΩV · a, TV = Ṫ − ΩV · T + T · ΩV . (5.29)
H A proof follows from (5.5), (5.10) and (5.11), because from (4.92) we have
dhi
= ṗi = ΩV · pi . (5.30)
dt
◦
Since the bases pi and pi are orthonormal, all the co-rotational derivatives
◦ ◦
relative to the mixed dyadic bases pi ⊗ pi , pi ⊗ pi coincide with TU or TV ,
respectively. N
1.5.7. The Jaumann Derivative (hi = qi ). If we choose the eigenbasis of the
deformation rate tensor as a moving basis hi = qi (it should be noted that the
basis qi is also orthonormal and
i
coincides with q ), then from (5.4) we get the
co-rotational Jaumann derivatives:
dai dT ij
ah ≡ aJ = qi , Th ≡ TJ = qi ⊗ qj . (5.31)
dt dt
Theorem 1.34. The Jaumann derivatives (5.31) are connected to the total
derivatives with respect to time by the relations (for a vector a and for a tensor
T, respectively)
aJ = ȧ + a · W, (5.32)
J
T = Ṫ − W · T + T · W. (5.33)
H According to the relationship (4.94), we get
dhi
= q̇i = W · qi ,
dt
therefore, due to formulae (5.5) and (5.10) we find
aJ = ȧ · qi ⊗ qi + a · W · qi ⊗ qi = ȧ + a · W.
In a similar way, we can prove the relation (5.33). N
1.5.8. Co-rotational Derivatives in a Moving Orthonormal Basis. Let hi
be a moving orthonormal basis. In this case we denote co-rotational derivatives
by the following way: ah ≡ aQ and Th ≡ TQ . Due to orthonormalization of the
basis hi , the total derivatives of a and T with taking account of (5.1), (5.2) and
(5.4) can be written as follows:
da dai dh
ȧ ≡ = hi + ai i = aQ + ai ω h × hi , (5.34)
dt dt dt
dT ij dhi
³ dh
´
Ṫ = hi ⊗ hj + T ij ⊗ hj + hi ⊗ j =
dt dt dt
= TQ + T ij ω h × hi ⊗ hj − hi ⊗ hj × ω h .
Here we have used formula (4.86) for derivative ḣi of the moving basis, where
ωh is the vorticity vector giving a rotation of the basis hi relative to the fixed
basis ēi (see (4.78) and (4.85)). With taking account of (5.1) and (5.2), formulae
(5.34) can be written in the form
aQ = ȧ − ω h × a, TQ = Ṫ − ω h × T + T × ω h . (5.35)
It should be noticed that if a = ωh, then
ω̇ h = ω hh , (5.35a)
because ωh × ωh = 0 due to properties of the vector product.
1.5.9. Spin Derivative. Take an arbitrary orthonormal basis h̄i at a point M

of a continuum in K. The trihedron must have the only property that at any time
t the basis h̄i rotates with the instantaneous angular rate, which is equal to the
◦
rotation rate of the trihedron pi relative to the trihedron pi . As shown in 1.4.10,
the instantaneous rotation of the trihedron is characterized by the spin tensor
Ω = Ȯ · OT determined by (4.97).
Then we can define the co-rotational derivative in the basis, which is called
the spin derivative (of a vector a and of a tensor T, respectively):
i
da
ah ≡ aS = h̄i , (5.36)
dt
dT ij
Th ≡ TS = h̄i ⊗ h̄j . (5.37)
dt
Theorem 1.35. The spin derivative is related to the total derivative with
respect to time as follows (for a vector a and for a tensor T, respectively):
aS = ȧ + a · Ω, (5.38)
TS = Ṫ − Ω · T + T · Ω. (5.39)
H A proof of Theorem 1.35 follows from (5.10), (5.11) and the relation
˙i
¯h = Ω · h̄i , (5.40)
which is a consequence of (4.84). The relation (5.39) follows from (5.11). N
1.5.10. Universal Form of the Co-rotational Derivatives. On comparing
formulae (5.18), (5.19), (5.20), (5.27), (5.29), (5.33), (5.38) and (5.39), we can
notice that all the representations of the co-rotational derivatives and also the
total derivative with respect to time can be written in the universal form:
ah = ȧ − Zh · T, Th = Ṫ − Zh · T − T · ZT
h, (5.41)
h = { ·, Ol, CR, U , V , J , S },
where tensors Zh have the following representation for different h:
T
Zh = { 0, L, −L , ΩU , ΩV , W, Ω }, (5.42)
h = { ·, Ol, CR, U , V , J , S }.
Since tensors ΩU , ΩV and Ω are linearly expressed in terms of W and D
(see Exercise 1.4.11), so tensors Zh can be written as linear functions of W and
D:
Zh = 4 ZDh · · D + 4 ZW h · · W. (5.43)
Table 1.1 gives expressions for fourth-order tensors
4
ZDh and
4
ZW h , where
tensors
4
ΩU , ΩV
4
and
4 e
Ω are defined in Exercise 1.4.11.
T a b l e 1.1. Expressions of tensors

4
ZDh , 4 ZW h and
4
Eh for
different co-rotational derivatives
h · Ol CR U V J S
4
ZDh 0 ∆III ∆−III 4
ΩU 4
ΩV 0 4 e
Ω
4
ZW h 0 ∆III ∆III 0 ∆III ∆III ∆III
4
Eh 0 −2∆ 2∆ 0 0 0 0
1.5.11. Relations between Co-rotational Derivatives of Deformation Rate

Tensors and Velocity Gradient. In paragraph 1.4.11 we have derived the
relationships between rates of deformation tensors and velocity gradient L.
Similar connections also exist between co-rotational derivatives of the tensors
and L. Let us establish them.
On substituting representations (4.103), (4.104) and (4.130) for rates Ȧ, J̇,
ġ, (g−1 )• , V̇ and (V−1 )• into formula (5.41), we get
Ah = D − (Zh + LT ) · A − A · (ZT
h + L),
Jh = D − (Zh − L) · J − J · (ZT T
h − L ),
gh = −(Zh + LT ) · g − g · (ZT
h + L), (5.44)
−1 h −1 −1
(g ) = −(Zh − L) · g −g · (ZT
h − L ), T
1 1
Vh = −(Zh − (L + Ω)) · V − V · (ZT T T
h − (L + Ω )),
2 2
1 1
(V−1 )h = −(Zh − (Ω − LT )) · V−1 − V−1 · (ZT T
h − (Ω − L)),
2 2
h = { ·, Ol, CR, U , V , J , S }.
From these relationships we can find the following expressions:
(V − E)h = D − 4 Eh · · D − (Zh + ZT
h )−
1 1
− (Zh − (Ω + L)) · (V − E) − (V − E) · (ZT T T
h − (Ω + L )), (5.45)
2 2
(E − V−1 )h = D + 4 Eh · · D + Zh + ZT
h−
1 1
− (Zh − (Ω − LT )) · (E − V−1 ) − (E − V−1 ) · (ZT T
h − (Ω − L)).
2 2
Here we have denoted the co-rotational derivative of the metric tensor by
Eh = 4 Eh · · D. (5.46)
The tensor
4
Eh differs from zero-tensor only when h = {CR, Ol} (see Exercise
1.5.3), its expressions are given in Table 1.1 (see paragraph 1.5.10).
Since tensors Zh and Ω are linearly expressed in terms of W and D
(see formulae (5.43), (4.129), (4.122)(4.124)), so on the right-hand sides of
equations (5.44) there are also linear functions of W and D, their explicit
expressions will be given in 3.2.22.
Exercises for 1.5.
Exercise 1.5.1. Show that the mixed co-rotational derivatives, the left and the right co-
rotational derivatives relative to the eigenbasis and also the Jaumann and spin derivatives
satisfy the differentiation rules of scalar products:
(A · B)h = Ah · B + A · Bh , (a · A)h = ah · A + a · Ah ,
(ψA)h = ψ h A + ψAh , h = {·, d, D, U , V , J , S},
and the Oldroyd and Cotter-Rivlin derivatives do not satisfy this rule.
Exercise 1.5.2. Show that for the co-rotational derivatives, the following rules of
differentiation of scalar products of two vectors a and b and also of two tensors T and
B remain valid:
(a · b)h = (a · b)• = ah · b + a · bh , h = {U , V , J , S};
(T · · B)h = (T · · B)• = Th · · B + T · · Bh , h = {d, D, U , V , J , S}.
Exercise 1.5.3. Show that the following co-rotational derivatives of the unit tensor E
give the zero-tensor:
Eh = 0, h = {·, d, D, U , V , J , S},
and the Oldroyd and CotterRivlin derivatives of E are different from zero:
ECR = 2D, EOl = −2D.

Exercise 1.5.4. Using formulae (4.103) and (5.20), show that the CotterRivlin
derivatives of the left Almansi deformation tensor A and of the left Almansi measure g
have the form
ACR = D, gCR = 0.
Exercise 1.5.5. Using formulae (4.103) and (5.18), show that the Oldroyd derivatives
of the right CauchyGreen tensor J and of the right deformation measure g −1 have the
form
Ol −1 Ol
J = D, (g ) = 0.
Exercise 1.5.6. Using the expressions for the tensors U (3.6), C and Λ (3.29), and
also G and G−1 (3.29), show that we can write the right derivative relative to the
◦
eigenbasis pα in the form
X
3
◦ ◦ X
3
◦ ◦
UU = λ̇α pα ⊗ pα , CU = λα λ̇α pα ⊗ pα ,
α=1 α=1
X
3
λ̇α ◦ ◦ X
3
λ̇α ◦ ◦
ΛU = pα ⊗ pα , (U−1 )U = pα ⊗ pα ,
α=1
λ3α α=1
λ2α
GU = 2CU , (G−1 )U = −2ΛU .

Exercise 1.5.7. Using the expressions for the tensors V (3.6), for A and J (3.29),
−1
and also for g and g (3.29), show that we can write the left derivative relative to the
eigenbasis pα in the form
X
3
X
3
λ̇α
VV = λ̇α pα ⊗ pα , AV = pα ⊗ pα ,
α=1 α=1
λ3α
X
3
JV = λα λ̇α pα ⊗ pα , gV = −2AV , (g−1 )V = 2JV .

α=1
Exercise 1.5.8. Show that the Oldroyd and Jaumann derivatives of a second-order
tensor T are connected by the relationship
TOl = TJ − T · D − D · T.
Exercise 1.5.9. Show that if for an arbitrary symmetric tensor T its co-rotational
derivatives are equal to zero:
Th = 0, h = {d, D, U , V , J , S},
then the first invariant of the tensor: I1 (T) = T · · E has its stationary value, i.e.
I˙1 (T) = 0.
Show that for the co-rotational Oldroyd and CotterRivlin derivatives this statement is
not valid.
Exercise 1.5.10. Using the results of Exercise 1.4.3, show that the co-rotational
derivatives of the normal vector n satisfy the following relations:
nCR = γn, γ = n · D · n,
nOl = γn − n · L − L · n, nJ = γn − n · D.
Exercise 1.5.11. Show that the following co-rotational derivatives of a symmetric
tensor give a symmetric tensor:
if A = AT , then (Ah )T = Ah , h = {U , V , J , S , Ol, CR},
and also a skew-symmetric tensor, if they are applied to a skew-symmetric tensor:
if B = −BT , then (Bh )T = −Bh , h = {U , V , J , S , Ol, CR}.
The mixed co-rotational derivatives h = d, D have no such properties.

http://www.springer.com/978-94-007-0033-8

Kinematics of Continua: OM X X X X X

Uploaded by

Copyright:

Kinematics of Continua: OM X X X X X

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Kinematics of Continua: OM X X X X X

Uploaded by

Copyright:

Chapter 1

1.1. Material and Spatial Descriptions of Continuum Motion

1.1.1. Lagrangian and Eulerian Coordinates. The Motion Law. Let us

x = x(X k , t). (1.4)

Unless otherwise stipulated, functions (1.3) are assumed to be regular in the

are Jacobian matrices and inverse Jacobian matrices.

Figure 1.5. Local basis vectors in reference and actual configurations

and also vectors of reciprocal local bases

any vector field a(M) i ◦i ◦

1.1.4. Tensors and Tensor Fields in Continuum Mechanics. For different

det T 6= 0). The determinant of a tensor is defined by the determinant of its

We can transpose fourth-order tensors as follows:

where (m1 m2 m3 m4 ) is some substitution, for example,

Besides F, in continuum mechanics one often uses the transpose tensor FT ,

ē0i = Q · ēi . (1.39)

Figure 1.8. Moving bases ē0i and e

In this case, instead of Cartesian coordinates xi of a point M in the basis ēi

Let Qi j be components of the tensor Q with respect to the basis ēi :

ē0i = Qj i ēj , (1.42)

e = x − x0 = (xi − xi0 )ēi = x

xi = xi0 (t) + Qi j (t)e e k ) ≡ xi (X

Figure 1.9. Curvilinear spatial coordinates e i and Lagrangian

and the matrix Peki ≡ ∂ X

and the reciprocal basis vectors

On substituting (1.51) into (1.57), we derive transformation formulae for

Exercises for 1.1.

O0 = O0i j ēi ⊗ ēj , U0 = kα ēα ⊗ ēα ,

∂xm ∂X k ∂xm ∂xm

Exercise 1.1.7. Prove that

∆I · · T = I1 (T)E, I1 (T) = T · · E, ∆II · · T = TT ,

formulae, the tensor ∆III is the `true' unit fourth-order tensor.

1.2. Deformation Tensors and Measures

1.2.1. Deformation Tensors. Besides F, important characteristics of the

Here C is called the right CauchyGreen deformation tensor, A  the left

ε̆kl = εij g ik g jl 6= εkl , ε̆kl = εij gik gjl 6= εkl . (2.3)

1.2.2. Deformation Measures. Besides the deformation tensors, we define

1.2.3. Displacement Vector. Introduce a displacement vector u of a point M

C takes the form

In a similar way, we can prove the remaining relations of the theorem. N

F−1 = (δik − ∇i uk )rk ⊗ ri = (F −1 )ki rk ⊗ ri (2.22)

and for the inversetranspose gradient

F−1T = (F −1 )ki ri ⊗ rk = (F −1 )i k rk ⊗ ri , (2.23)

 the expressions for covariant components of the deformation tensor in terms

 the relations between contravariant components of the deformation tensor and

1.2.5. Physical Meaning of Components of the Deformation Tensor. Let

By the definition of the scalar product (see [12]), we have

1.2.6. Transformation of an Oriented Surface Element. In actual configu-

n dΣ = rα dX α × rβ dX β = dxα × dxβ , (2.44)

(there is no summation over α, β ). Thus,

Thus, we have proved the following theorem.

On introducing the notation

Formula (2.61) follows from the relation

Finally, we should express the determinant g = det (gij ) in terms of εij . To

from (2.66) we get the desired formula

Exercises for 1.2

C=Λ= (kα2 − 1)ēα ⊗ ēα , A=Λ= (1 − kα−2 )ēα ⊗ ēα ,

the deformation measures are determined as follows:

G = g −1 = kα2 ēα ⊗ ēα , g = G−1 = kα−2 ēα ⊗ ēα ,

and components of the deformation tensor take the forms

for the deformation tensors:

C = (a/2)O3 + (a2 /2)ē22 , A = (a/2)O3 − (a2 /2)ē22 ,

Here we have introduced the notation

Here C is called the right CauchyGreen deformation tensor, A the left

and for the inversetranspose gradient

the expressions for covariant components of the deformation tensor in terms

the relations between contravariant components of the deformation tensor and

The tensor F has nine independent components, the tensor O three