2 Kinematics of Deformation

In this chapter we briefly summarize the fundamental relations of the kinematics of large
deformations, restricting ourselves to the applications we have in mind. For concise
treatments we advise the literature on continuum mechanics, e.g., the monographs of
Altenbach [AA95], Becker and Bürger [BB75], Betten [Bet93], Chadwick [Cha76], Gurtin
[Gur81], Malvern [Mal69], Truesdell and Noll [TN65] as well as the recent works of
Bertram [Ber05], Haupt [Hau02] and Holzapfel [Hol00].
Our description here focuses at first on elastic bod-
ies, Figure 2.1. For continuum-mechanical descrip-
tion a body is a set whose elements are in one-to-one
correspondence with points of a region B ⊂ IR3 and
we partition its boundary S = ∂B into two parts,
S = S1 ∪ S2 , where displacements are prescribed or
traction is applied, respectively, S1 ∩ S2 = 0. More-
over, the body is of a size which allows to neglect
its microscopic composition. Elasticity means that
all the processes under consideration are reversible.
Consequently, elastic materials show no internal pro-
cesses, no time dependence, no hysteresis and no dis-
sipation. Note that elasticity as it is defined here
does not mean linearity1 . Elastic materials can be Figure 2.1: Solid body B.
nonlinear because of large deformations as well as of
nonlinear material laws, and, on the other hand linear materials can be inelastic, e.g.
viscoelastic. Inelastic materials are topic of subsequent chapters.

2.1 Deformation and motion

Let us consider a body whose particles occupy a region B ⊂ IR3 in a certain reference
placement. We label the material particles P ∈ B by their position vectors X(P ).
Without loss of generality these position vectors may be defined by their components
1
Note that the definition of elasticity may differ in literature.

1
2.1 Deformation and motion

G. L. Lagrange (1736-1813)
B

ϕ B

Quo vadis?

Figure 2.2: Lagrangian point of view: x = ϕ X

{X1 , X2 , X3 } relative to some coordinate system with orthonormal basis E 1 , E 2 , E 3

centered at some convenient origin O, X = Xi E i . We will refer to this reference system
as the material reference frame and to components relative to this frame as material
or Lagrangian coordinates. The reference placement is commonly (but not necessarily)
identified with the initial, undeformed placement of the body, see Figure 2.2.

The body B subsequently deforms and moves over a period of time t ∈ IR+ . A de-
formation of a body B is an injective mapping ϕ : B → IR3 . The condition that the
deformation mapping ϕ defines an injective function excludes deformation involving
tearing and interpenetration of matter of the body.

The region ϕ(B) is the current or deformed placement of the body. The material
point P ∈ ϕ(B) is labelled now by a position vector x(P ). The components {x1 , x2 , x3 }
of the position vector are relative to a coordinate system with orthonormal basis e1 , e2 , e3
centered at o, x = xi ei . We will refer to this placement as the spatial reference frame.
The components relative to this frame are spatial or Eulerian coordinates, see Figure 2.3.

In the following we presume the origins O, o and the basis vectors of material and spacial
frame to coincide, i.e., the basis E i is identical to ei . We refer to tensorial quantities
in the reference placement by using uppercase letters, whereas the current placement
is denoted by lowercase letters. To reduce conflicts in notation we employ a subscript
0 for some quantities when they refer to the initial (reference) placement and omit the
subscript in the current placement. Moreover, we identify the material particles P ∈ B

2
2.1 Deformation and motion

B
L. Euler (1707-1783)
ϕ B

Ex quo venis?

Figure 2.3: Eulerian point of view X = ϕ−1 x

by their position vectors X(P ) and x(P ), respectively.

In coordinates, the deformation mapping takes the form


xi = ϕi X , X ∈ B. (2.1)

Since, by assumption, ϕ is injective, the inverse deformation mapping ϕ−1 : ϕ(B) →

B is well defined.
Unlike to small strain approaches the total deformation of a body can not be understood
as a superposition of displacement vector fields. To illustrate this, let ϕ1 : B → IR3 and
ϕ2 : B → IR3 be two successive deformations of the body B, see Figure 2.4. The total
or combined deformation of the body is





ϕ X, t = ϕ2 ϕ1 X, t , t ≡ ϕ2 ◦ ϕ1 X, t . (2.2)

where ◦ denotes the composition. Evidently, the total deformation follows by a composi-
tion of mappings. The composition of deformation mappings has a multiplicative group
structure [MH83].
3
A motion of a body during

a time interval [t1 , t2 ] is a function ϕ : B × [t1 , t2 ] → IR such
that the mappings ϕ ·, t , t ∈ [t1 , t2 ], are injective. In other words, a motion defines a
one parametric sequence of deformation mappings indexed by the time, x = ϕt (X) =
ϕ(X, t). This describes for every material point P ∈ B a path or trajectory in IR3 .

3
2.1 Deformation and motion

ϕ1 ϕ2


B ϕ1 (B) ϕ2 ϕ1 (B)

ϕ = ϕ2 ◦ ϕ1

Figure 2.4: Composition of deformations.

To describe the motion of a particle we may distinguish between a spatial (or direct)
motion and a material (or inverse) motion problem. The direct motion problem
bases on the idea of following material particles from a fixed position X along their
trajectories through the ambient space. Because attention is paid to the particles (we
observe what happens to the particles when they move), this point of view is denoted
as the material or Lagrangian setting. In contrast to this, within the inverse motion
problem physical particles are followed through the ambient material at a fixed spatial
position x. The observer takes the spatial or Eulerian point of view.
The displacement field U of a particle relates its position in the undeformed placement
to its position in the deformed placement at instance t. In the Lagrangian setting we
write
U (X, t) = x(X, t) − X, (2.3)
whereas the Eulerian form of equation(2.3) reads

u(x, t) = x − X(x, t). (2.4)

Both settings are related by the inverse deformation mapping,

U (X, t) = U ϕ−1 (x, t) = u(x, t).

(2.5)

The instantaneous velocity of a material point is the material or Lagrangian velocity

field
∂ϕi

V (X, t) = X, t . (2.6)
∂t
On the other hand, in the Eulerian setting the velocity follows by a composition of
material velocity and the inverse deformation mapping. Thus, the velocity at a spacial
position is described by the spacial or Eulerian velocity field

∂ϕ −1
ϕ (x, t), t = V ◦ ϕ−1 (x, t).



v(x, t) = V ϕ(X, t), t = (2.7)
∂t

4
2.2 Kinematics of local deformations

Analogously, the instantaneous acceleration of the material point X at time t is

represented by the material or Lagrangian acceleration field

∂V

A(X, t) = X, t . (2.8)
∂t
The spacial or Eulerian acceleration field represents the acceleration of the material
point occupying the spatial position x at time t.

∂V
ϕ−1 (x, t), t = A ◦ ϕ−1 (x, t)



a(x, t) = A ϕ(X, t), t = (2.9)
∂t
The material time-derivative (2.9) of a smooth spatial vector field v(x, t) may be eval-
uated to give

∂v(x, t)
a(x, t) = + grad v(x, t) · v(x, t). (2.10)
∂t
Evidently material time derivative a(x, t) = v̇(x, t) is distinct from a ’normal’ time-
derivative because of its structure. The first term in equation(2.10) describes the local
acceleration (i.e., the local rate of change in the velocity field), the second term describes
a convective acceleration field. It plays an important role in fluid mechanics, because by
equation (2.10) the spatial acceleration can be determined without knowing the motion
explicitly.
In solid mechanics the constitutive behavior of materials is commonly given in terms
of material coordinates. This motivates us to restrict our attention in the following to
the direct motion problem formulated in the Lagrangian setting. However, we keep in
mind that the kinematic equations in the direct and in the inverse motion problem are
in correspondence to each other.

2.2 Kinematics of local deformations

To measures the local deformation in a material we presume the principle of local

action to be valid. This principle states that the internal energy density of a material
point X ∈ B depends only on the local state of a infinitesimal material neighborhood
containing the point.
Let dV0 ⊂ B be the volume of an infinitesimal material neighborhood of point X ∈ B and
let dV be the volume of the corresponding infinitesimal spatial neighborhood. Consider
now a material point at position X ∈ B and let X + dX be a neighboring material

5
2.2 Kinematics of local deformations

ϕ
x + dx
1
0x dV
X + dX
dV0
X01

Figure 2.5: Local deformation of a material neighborhood.

point within dV0 . The corresponding spatial point be x + dx, see Figure 2.5. With
deformation (2.1) we know for the components of x + dx,


xi + dxi = ϕi X + dX, t , (2.11)

which may be expanded in a Taylor series2 at position X to give

∂ϕi

ϕi X + dX, t = ϕi X, t + X, t dXJ + o(dX). (2.12)
∂XJ

The expression o(dX) denotes terms of higher order that tend to zero faster than dX.
Neglecting these terms we obtain from (2.12) the linearized differential relation

∂ϕi
∂ϕi

xi + dxi ≃ xi + X, t dXJ ⇒ dxi = X, t dXJ (2.13)
∂XJ ∂XJ

and we define the deformation gradient F with components

∂ϕi


FiJ X, t = X, t ⇒ dxi = FiJ X, t dXJ (2.14)
∂XJ

The deformation gradient F is fundamental in non-linear continuum mechanics. It is

a primary measure of deformation. The quantity provides a full determination of the
deformation mapping at time t on an infinitesimal material neighborhood dX of X.

The deformation gradient is a linear operator by definition. Therefore, F (X, t) may

be regarded a member of the linear space of all linear transformations into IR3×3 , i.e.,
GL(3, IR).
2
Taylor expansion for f (x) at position x0 : f (x) = f (x0 ) + f ′ (x0 ) · (x − x0 ) + high order terms

6
2.2 Kinematics of local deformations

Consider now an infinitesimal material neighborhood X + dX of volume dV0 , which

is given by the parallelepipedic product dV0 = (dX (1) × dX (2) ) · dX (3) . Let dV be the
volume of the corresponding spatial neighborhood at time t and write
(1) (2)
(3)
dV = dx(1) × dx(2) · dx(3) = F1i dXi × F2i dXi · F3i dXi


(2.15)
= det F dX (1) × dX (2) · dX (3) = det F dV0 .

Consequently, the following relation holds

dV
= J(X, t), (2.16)
dV0
where



J X, t = det F (X, t) (2.17)
is the Jacobian of the deformation. Evidently, the Jacobian J provides a measure
of the local volumetric deformation. From the inverse mapping theorem follows that the
deformation mapping is locally one-to-one at X and time t if and only if
J(X, t) > 0. (2.18)

Therefore, we may interpret this condition as a local invertibility condition. (Note that
conditions for global invertibility are considerably harder to define.)
The following transformation relates elements of
oriented area in the undeformed and deformed
placements. Let dS ⊂ ∂B denote an infinitesimal
(surface) area containing X with the correspond- N
ing outward unit normal N , see Figure 2.6. Let
ds be the deformed area on ϕ(S, t) and n its cor- dS
responding outward unit normal. Then, for the
components of the outward unit normal holds
−1
ni ds = J(X, t)FJi (X, t)NJ dS. (2.19)
Figure 2.6: Oriented area.
This relation is known as Nanson’s formula and
may easily be verified by equation (2.16) and dV = dx · n ds = JdV0 = JdX · N dS.
Shortly we may write
n ds = JF −T N dS. (2.20)

To illustrate the composition of deformations let ϕ1 : B → IR3 and ϕ2 : B → IR3 be

two successive deformations of B, cf. Figure 2.4. Let F 1 (X, t) and F 2 (X, t) be the

7
2.3 Decomposition of deformations

corresponding deformation gradients at material point X and time t. Then, the total
or combined local deformation gradient follows by multiplication of the incremental
deformation gradients





F X, t = F 2 ϕ1 X, t , t F 1 X, t . (2.21)
Composition rule (2.21) together with the invertibility condition (2.18) confers the set
of (all admissible) deformation gradients a multiplicative group structure. It may be
identified with a subgroup of the general linear group of orientation-preserving transfor-
mations GL+ (3, IR) = {F ∈ GL(3, IR), det(F ) > 0}.

2.3 Decomposition of deformations

The deformation gradient F contains the full information about the (linearized) defor-
mation ϕ(X, t) at a local material point but is a two-point tensor, i.e., it works between
current and reference placement. Strain measures are required to refer to either the
reference or the current placement and, moreover, to be not affected by rigid body rota-
tions. In this section we will extract the information of interest out of the deformation
gradient and, then, introduce some common strain tensors.
The natural way of splitting the deformation gradient into different parts is, because of
its multiplicative group structure, a multiplicative decomposition. Let F by a given
state of deformation. Let one part of the full deformation be associated with reason
A (e.g., volume preserving deformation or irreversible stretching) and another part be
related to reason B (e.g., pressure induced dilatation or purely elastic stretching). Then,
the decomposition of the deformation gradient reads3
F = F B F A. (2.22)
This type of decomposition goes back to Lee [Lee69].
In that sense we separate the straining from the rigid rotation at a material point and
employ the polar decomposition. Any deformation gradient F factors as the product
F = R U, (2.23)
with
U = U⊤ (2.24)
R−1 = R⊤ . (2.25)
3
For brevity we omit in this section the arguments of F .

8
2.3 Decomposition of deformations

The positive definite tensor U is the (right) stretch tensor, the orthogonal tensor R
describes a rotation. Tensor R may be regarded as a member of the group of rotations
SO(3), where SO(3) is the Lie-group {R ∈ IR3×3 , RT R = I, det R = 1}. The polar
decomposition is always unique. (The proof is easily done by contradiction, see, e.g.
[Ber05].)

F = RU

F =V R

Figure 2.7: Illustration of the polar decomposition.

The eigenvalues of the stretch tensor U , denoted as λα , α = 1, 2, 3, fulfill the relation

~ α = λα N
UN ~α with ~ α | = 1.
|N (2.26)
The (real and positive) eigenvalues λα represent the principal stretches of the deforma-
tion, whereas the orthonormal eigenvectors N ~ α are the principal referential axis, i.e.,
the principal directions in the reference placement.
There exists also a unique left polar decomposition of the form
F = V R, (2.27)
where again R ∈ SO(3) and with stretch tensor
V =V⊤ (2.28)

9
2.3 Decomposition of deformations

orthogonal to U .

The eigenvalues and eigenvectors of the left stretch tensor V , are

Vn ~α
~ α = λα n with |~
nα | = 1. (2.29)

The eigenvalues λα represent here the principal stretches in the current placement and
the orthonormal eigenvectors n~ α are the corresponding principal directions. In other
words, the eigenvectors of V are those of U rotated by R, n ~ α = RN~ α.

To illustrate the nature of rotation R let us write the spectral decomposition of (2.24)
and (2.28).
3
~ α⊗N
~ α,
X
U = λα N (2.30)
α=1
X3
V = ~α ⊗n
λα n ~ α. (2.31)
α=1

Then, rotation tensor R may be decomposed as

3
~ α.
X
R = ~α ⊗N
n (2.32)
α=1

Another useful decomposition of the deformation gradient F separates the deformation

along a tangential plane from the deformation normal to that plane;

F = F k F ⊥, (2.33)

where F k denotes the in-plane part of the deformation gradient (or membrane de-
formation) and F ⊥ the out-of-plane part of the deformation gradient (or transversal
deformation). The in-plane part satisfies the identity

F k N = N, (2.34)

where N is the unit normal to the tangential plane in the reference placement. With
a vector a ∈ IR3 and by Hadamard’s compatibility condition (see, e.g. [Tho57]), the
out-of-plane part of the deformation gradient has the form

F ⊥ = I + a ⊗ N. (2.35)

10
2.4 Strain measures

2.4 Strain measures

The crucial demand on any strain measure is its ability to represent strain but to
vanish for rigid body rotations. This in turn requires symmetry of the strain tensors.
Thus, the right Cauchy-Green tensor is defined by
C = F⊤ F (2.36)
= U ⊤ U = U 2,
The six independent components of tensor C, CJL = FJi FiL , refer to the reference
placement and can uniquely be determine by the nine components of F (but not visa
versa). The Green-Lagrange strain tensor
1
E = F⊤ F − I


(2.37)
2
has additionally the property to be zero in all components in case of no deformation.
Frequently used is the logarithmic strain
1
ǫ = log C (2.38)
2
which is also known as Henky strain tensor4
H = log U . (2.39)

Analogously the left Cauchy-Green tensor or Finger tensor is defined by

b = F F⊤ (2.40)
= V ⊤V = V 2 = RC
and the Euler-Almansi strain tensor is
1
1
e = I − b−1 = I − F −⊤ F −1 ,


(2.41)
2 2
which both refer to the current placement.
In linear theory small deformations are presumed. Consequently, reference and current
placement coincide. Strains are small enough that a linearized strain measure accurately
approximates the more rigorous nonlinear measures. This condition is fulfilled in virtu-
ally all solid materials under sufficiently small loads. By linearization of (2.37) or (2.41)
we get the infinitesimal strain tensor as
1
ǫ = grad u + grad u⊤ .


(2.42)
2
4
To distinguish the logarithmic strain and the infinitesimal strain we denote here the first by H but
we will go back later to use definition (2.38).

11
2.4 Strain measures

Remark

The following general formula may illustrate the different strain measures for the simple
case of uniaxial deformation. Let λ denote the ratio of current to initial length of the
specimen and let the only nonzero entry in the strain tensors be ε. Now we write
(following Seth [Set64], see also [Ber05, SH00])

1
1 − λ−α .


ε= (2.43)
α

Then, assuming various values of constant α we obtain:

ε = 12 λ2 − 1 ε ∈ [− 21 , ∞]


α = −2 Green-Lagrange strain
α = −1 ε=λ−1 ε ∈ [−1, ∞] infinitesimal strain
α=0 ε = log(λ)
ε ∈ [−∞, ∞] logarithmic strain
α=2 ε = 12 1 − λ2 ε ∈ [−∞, 21 ] Euler-Almansi strain

The first three measures are plotted in Fig-

ure 2.8 and we observe a very different be-
havior at large strains. A strain measuring
E function f : IR 7→ IR is called additively
ǫ symmetric with respect to lengthening and
shortening if holds

H 1


f = −f λ . (2.44)
λ

Among the four strain measures only the

(uniaxial) logarithmic strain satisfies the
requirement. Note that the stretch λ itself
is multiplicatively symmetric with respect
to lengthening and shortening but not ad-
ditively. Moreover there exists other valid
strain measures with no symmetry at all
Figure 2.8: Strain measures. [TT60]. So it is up to the reader to decide
whether satisfying (2.44) makes the loga-
rithmic strain (2.38) a “better” measure.

12
2.5 Strain rates

2.5 Strain rates

To finally determine a rate of deformation measure we note the velocity gradient

∂v
l = . (2.45)
∂x
The velocity appears here as a function of time for a given particle at X. The velocity
gradient (2.45) may be evaluated in terms of the deformation gradient as

l = Ḟ · F −1 . (2.46)

The rate-of-deformation tensor is now defined as the symmetric part of tensor l

1
l + l⊤ ,


d= (2.47)
2
whereas the antisymmetric part of l is known as spin tensor.

1
l − l⊤ ,


w= (2.48)
2

The rate-of-deformation tensor measures the rate of change of the square of the length
of material line segments dx. However, in general, the integral of the rate-of-deformation
does not vanish in a reversible cycle of deformation. Consequently, the rate-of-deformation
tensor is path-dependent. Although it is commonly used the rate-of-deformation (2.47)
is not necessarily a physical meaningful measure of deformation.

2.6 Examples

Kinematic equations for some simple deformations are summarized here for later refer-
ence. At first we consider homogeneous deformations, i.e., deformations with constant
affine boundary conditions. In a homogeneous state of deformation the gradient F is
independent of the position of the material point X or x, respectively, and, in conse-
quence, we can derive one deformation gradient F ∈ GL+ (3, IR) which is valid for all
points of the body B.

To write down the components of F explicitly, we presume volume preserving deforma-

tions with det F = 1.

13
2.6 Examples

ϕ
X1 R x1 r

L l

Figure 2.9: Uniaxial tension of a rod made of an incompressible material.

Uniaxial tension.

A rod with circular (or quadratic) cross section is pulled from initial length L to current
length l, see Figure 2.9. The deformation gradient F = ∂x/∂X in this uniaxial case
reads
 
l/L 0 0
F =  0 r/R 0  , (2.49)
0 0 r/R
with initial and current radius R and r, respectively.
We define now the axial stretch ratio λ = l/L and assume the material of the rod to
be incompressible. From the preserved volume of the deformation follows
πR2 L = πr2 l, (2.50)
and for the transverse stretch ratio we can write
l  R 2
= = λ. (2.51)
L r
The deformation gradient F now follows as
 
λ 0√ 0
F =  0 1/ λ 0 .
p (2.52)
0 0 1/ (λ)

Note that for a compressible material with Poisson modulus ν and with small-strain
kinematics (2.42) the deformation gradient can be evaluated from linear theory and
expression (2.49).
 
1 − νε 0 0
F = 0 1 − νε 0 . (2.53)
0 0 1+ε

14
2.6 Examples

a
A
r
R

Figure 2.10: Void expansion in an incompressible body.

Here we employ ε in its common uniaxial definition ε = l−L L

. Note that this relation
1 1
yields in the incompressible limit to 1 − νε = 1 − 2 ε = 2 3 − λ which is the linearization
p
of the function 1/ (λ).

Expansion of a void in an infinite body.

A spherical void with initial radius A is expanding to current radius a. Material spheres
surrounding the void with radii R and r, respectively, deform with the body. From the
volume constraint follows
4π 3 4π 3
(R − A3 ) = (r − a3 ), (2.54)
3 3
and, consequently, the deformation mapping r = ϕ(R) reads

1/3
r = a3 − A3 + R3 . (2.55)

Clearly, this deformation is not homogeneous but for any point located on the void
surface we can compute the deformation gradient as
 
a/A 0 0
F =  0 a/A 0  . (2.56)
0 0 a/A

15
Bibliography

[AA95] J. Altenbach and H. Altenbach. Einfüuhrung in die Kontinuumsmechanik. Teubner

Studienbücher, Stuttgart, 1995.

[BB75] E. Becker and W. Bürger. Kontinuumsmechanik. B. G. Teubner Verlag, 1975.

[Ber05] A. Bertram. Elasticity and Plasticity of Large Deformations. Springer, 2005.

[Bet93] J. Betten. Kontinuumsmechanik. Elasto-, Plasto- und Kriechmechanik. Springer-

Verlag, Berlin, 1993.

[Cha76] P. Chadwick. Continuum Mechanics - Concise Theory and Problems. G. Allen &
Unwin, 1976.

[Gur81] M.E. Gurtin. Introduction to Continuums Mechanics. Springer, 1981.

[Hau02] P. Haupt. Continuum Mechanics and Theory of Materials. Springer, 2002.

[Hol00] G. Holzapfel. Nonlinear Solid Mechanics. Wiley & Sons, 2000.

[Lee69] E.H. Lee. Elastic-plastic deformation at finite strain. J. Appl. Mech., 36:1–6, 1969.

[Mal69] L. Malvern. Introduction to the Mechanics of a Continous Medium. Prentice-Hall,

1969.

[MH83] J. Marsden and T. Hughes. Mathematical Foundations of Elasticity. Prentice-Hall,

Englewood Cliffs, NJ, 1983.

[Set64] B. R. Seth. Generalized strain measure with applications to physical problems. In

D. Abir M. Reiner, editor, Second-Order Effects in Elasticity, Plasticity and Fluid
Dynamics, pages 162–172. Pergamon Press, Oxford, 1964.

[SH00] J. J. Skrzypek and R. B. Hetnarski. Plasticity and Creep. CRC Press, 2000.

[Tho57] T.Y. Thomas. Extended compatibility conditions for the study of surfaces of discon-
tinuity in continuum mechanics. J. Math. Mech., 6:311–322, 1957.

16
Bibliography

[TN65] C. Truesdell and W Noll. The Non-Linear Field Theories of Mechanics. In Flügge,
S.: Handbuch der Physik. Volume III/3. Springer-Verlag, Berlin, 1965.

[TT60] C. Truesdell and R. A. Toupin. The Classical Field Theories. In Flügge, S.: Handbuch
der Physik. Volume III/1. Springer-Verlag, Berlin, 1960.

17