You are on page 1of 32

Chapter 1

Kinematics

Lecture notes by Professor Hong-Ki Hong


v1.0(8 September 2013) v1.1(13 October) v1.2(23) v1.3(25) v1.4(27) v1.5(4 November) v2.0(Ech1)(10
March 2016)(11 September 2016 CEIBA) v2.1(3 October CEIBA) v2.2(6:20 pm CEIBA) v2.3(5
CEIBA) v2.4(6 CEIBA) v2.4(6 CEIBA) v2.5(2 October 2017 CEIBA) v2.6(8 February 2018)(10
September CEIBA) v2.7(15-18 December CEIBA) v3.0(2Feb-3Apr 2019)(30 August 2019 CEIBA)
v3.1(6 October 2020 CEIBA) v3.2(15 November 2020 CEIBA) v4.0(14 September 2021 CEIBA)

1.1 Geometry of motion and deformation


1.1.1 First- and second-order derivatives in time and space
Kinematics studies geometry of motion and deformation, including positions, displace-
ments, and dislocations of particles in bodies, and their first and second order derivatives
with respect to the time and particle parameters.

1.1.2 Rigid-body kinematics versus deformable-body kinematics


If a body are rigid, there is no deformation in the body, and as such kinematics considers
only the positions, displacements, velocities, and accelerations of the particles in the rigid
body. The issue is simplified by the concepts of rigid-body displacement, including trans-
lation and rotation, and their first and second order time derivatives; however, rigid-body
kinematics is a fully nonlinear subject, no such things as linear, infinitesimal rigid-body
kinematics, and the nonlinear issue becomes further complicated when the concern is
extended to the kinematics of many rigid bodies, connected and/or disconnected.

kinematics independent variables


rigid-body kinematics time t ∈ I j R
n-rigid-body kinematics time t ∈ I j R, body i ∈ {1, 2, · · · , n}
deformable-body kinematics time t ∈ I j R, particle x ∈ D j R3
kinematics of deformation particle x ∈ D j R3

1
2 CHAPTER 1. KINEMATICS

If the body is deformable, the situation becomes a field problem with three more
independent real variables. If there are several deformable bodies, connected and/or dis-
connected, that is n-deformable-body kinematics, the situation is most challenged; it is
the study of geometry of motion, collision, and deformation of many connected and/or
disconnected deformable bodies.
As pointed out earlier, kinematics of rigid body motion is a fully nonlinear theory. On
the contrary, to study kinematics of deformation, it is often useful and inspiring to begin
with a linear, infinitesimal theory.

1.2 Linear infinitesimal theory of kinematics of de-


formation
1.2.1 An arbitrary particle and its arbitrary neighboring particle
Let P with coordinates (xj ) = (x1 , x2 , x3 ) be an arbitray particle of a deformable body
D j R3 and Q an arbitrary neighboring particle with coordinates (xj + dxj ) = (x1 +
dx1 , x2 + dx2 , x3 + dx3 ) in the neighborhood N (P ) ⊂ D j R3 of P .

Figure 1.1: 1-1.jpg

particle coordinates displacement


P ∈ N (P ) xj ui (xj )
Q ∈ N (P ) xj + dxj ui (xj + dxj )

Assume the displacement field ui (xj ) ∈ C 1 (D → R3 ) and the displacement gradient


∂ui /∂xk be infinitesimally small. Expand ui (xj + ∆xj ) as follows.

∂ui (xj )
ui (xj + ∆xj ) = ui (xj ) + ∆xk + O(∆xk )2
∂x
( k )
1 ∂ui (xj ) ∂uk (xj )
= ui (xj ) + − ∆xk
2 ∂xk ∂xi
( )
1 ∂ui (xj ) ∂uk (xj )
+ + ∆xk + O(∆xk )2
2 ∂xk ∂xi
= ui (xj ) + Ωik (xj )∆xk + ϵik (xj )∆xk + O(∆xk )2 ,
1.2. LINEAR INFINITESIMAL THEORY OF KINEMATICS OF DEFORMATION 3

where
∂ui
the displacement gradient tensor Hik = = Ωik + ϵik , (1.1)
∂xk
( )
1 ∂ui ∂uk 1
the infinitesimal rotation tensor Ωik = − = (Hik − Hki ), (1.2)
2 ∂xk ∂xi 2
( )
1 ∂ui ∂uk 1
the (infinitesimal) strain tensor ϵik = + = (Hik + Hki ). (1.3)
2 ∂xk ∂xi 2

Let ∆xj → 0, and hence ∆xj become dxj and O(∆xk )2 tend to vanish. Thus,

ui (xj + dxj ) = ui (xj ) + Ωik (xj )dxk + ϵik (xj )dxk . (1.4)

on rhs of (1.4) represent


1st term rigid-body translation of N (P )
2nd term rigid-body rotation of N (P )
1st + 2nd terms rigid-body displacement of N (P )
3rd term small deformation of N (P )

It may be helpful to write (1.4) in matrix form:


       
 u1   u1  0 Ω12 Ω13  dx1 
u2 = u2 + 0 Ω23  dx2
     
u3 Q u3 P s.s. 0 dx3
  P

ϵ11 ϵ12 ϵ13  dx1 
+ ϵ22 ϵ23  dx2 .
 
sym. ϵ33 P dx3

Notice that the first term on the right-hand side, the rotation tensor in the second term
on the right-hand side, and the strain tensor in the third term on the right-hand side are
all evaluated at P .

1.2.2 Infinitesimal rotation tensor


First of all, note that Ωik in (1.2) is skew-symmetrical (s.s.), it follows that Ωik ∈ so(3), the
Lie algebra of the rotation group SO(3). Thus, it is clear that only if Ωik is infinitesimal,
it has the meaning of infinitesimal rotation; otherwise, it has no relation to rotation
whatsoever.
From (1.2), the infinitesimal rotation tensor
   
0 Ω12 Ω13 0 −ω3 ω2
Ωij =  0 Ω23  =  0 −ω1 
s.s. 0 s.s. 0
4 CHAPTER 1. KINEMATICS

is skew-symmetrical (s.s.) and has only three independent entries, so it can be associated
with an axial vector1
   
 ω1   −Ω23 
ωk = ω2 = Ω13 .
   
ω3 −Ω12

With the permutation symbol



 0 when any two indices are equal,
eijk = 1 when ijk = 123, 231, 312,

−1 when ijk = 213, 321, 132,

they are related by

Ωij = −eijk ωk , Ω = ω × I,
1
ωk = − eijk Ωij .
2
In terms of displacement gradients, the infinitesimal rotation vector
( )
1 ∂ui ∂uj 1 ∂ui 1
ωk = − eijk − = − eijk , ω = ∇ × u. (1.5)
4 ∂xj ∂xi 2 ∂xj 2

The negative gradient of ω(x) is the right curl of the strain field ϵ(x), i.e.,

−∇ ⊗ ω = ϵ × ∇. (1.6)

The right gradient of ω(x) is the curl of the strain field ϵ(x), i.e.,

ω ⊗ ∇ = ∇ × ϵ. (1.7)

Proof: Using the definition (1.3), Eq. (1.5), and the identity ∇ × (∇ ⊗ u) = zero tensor
of order 2 = (u ⊗ ∇) × ∇ for any vector field u(x) ∈ C 2 , we get
1 1 1
−∇ ⊗ ω = − ∇ ⊗ ∇ × u = ∇ ⊗ u × ∇ = (u ⊗ ∇ + ∇ ⊗ u) × ∇ = ϵ × ∇.
2 2 2
and
1 1
ω ⊗ ∇ = ∇ × u ⊗ ∇ = ∇ × (u ⊗ ∇ + ∇ ⊗ u) = ∇ × ϵ. 
2 2

1.2.3 Infinitesimal strain tensor


Length change. Consider an arbitrary particle Q(xj + dxj ) in the neighborhood N (P ) of a
generic particle P (xj ). In other words, dxj = nj dL is arbitrary in which dL is the length
of dxj and nj is a unit vector (i.e., nj nj = 1) representing the direction of dxj . After
1
Axial vectors (e.g. moments) are distinguished from polar vectors (e.g. forces).
1.2. LINEAR INFINITESIMAL THEORY OF KINEMATICS OF DEFORMATION 5

Figure 1.2: 1-2.jpg

undergoing displacement and deformation, dxj become dxj + u(xj + dxj ) − uj (xj ), the
length of which will be denoted dℓ. Thus,

dℓ − dL ni [ui (xj + dxj ) − ui (xj )]


=
dL dL
ni Ωik (xj )nk dL + ni ϵik (xj )nk dL
=
dL
= ni ϵik (xj )nk , (1.8)

or in matrix form,
 T   
dℓ − dL  n1  ϵ11 ϵ12 ϵ13  n1 
= n2  ϵ22 ϵ23  n2 .
dL    
n3 sym. ϵ33 n3

In Eqs. (1.8) use has been made of Eq. (1.4) and of dxk = nk dL in the second equality,
and in the third ni Ωik (xj )nk = 0 due to antisymmetry of Ωik . To conclude, the quadratic
form ni ϵik (xj )nk in (1.8), or precisely the (infinitesimal) strain tensor ϵik (xj ) in (1.3) does
tell us the elongation rate (dℓ − dL)/dL along any direction nk in the neighborhood N (P )
of the particle P (xj ).

Figure 1.3: 1-3.jpg

Angle change. Let nj and mj be two mutually perpendicular but otherwise arbitrary
(n)
unit vectors (i.e., nj mj = 0, nj nj = 1, mj mj = 1). The angle π2 between dxj = nj dL(n)
6 CHAPTER 1. KINEMATICS

(m)
and dxj = mj dL(m) becomes π
2
− γ, where

(m) (n)
ni [ui (xj + dxj ) − ui (xj )] mi [ui (xj + dxj ) − ui (xj )]
γ = arctan + arctan
dL(m) dL(n)
(m) (m)
∼ ni Ωik (xj )mk dL + ni ϵik (xj )mk dL mi Ωik (xj )nk dL(n) + mi ϵik (xj )nk dL(n)
= +
dL(m) dL(n)
= 2ni ϵik (xj )mk , (1.9)

or in matrix form,
 T   
 n1  ϵ11 ϵ12 ϵ13  m1 
γ = 2 n2  ϵ22 ϵ23  m2 .
   
n3 sym. ϵ33 m3

(n) (m)
In Eqs. (1.9) use has been made of Eq. (1.4) and of dxk = nk dL(n) and dxk =
mk dL(m) in the second equality, and in the third ni Ωik (xj )mk + mi Ωik (xj )nk = 0 due to
antisymmetry of Ωik and ni ϵik (xj )mk = mi ϵik (xj )nk due to symmetry of ϵik . To conclude,
the bilinear form ni ϵik (xj )mk in (1.9), or precisely the (infinitesimal) strain tensor ϵik (xj )
in (1.3) does tell us the angle decrease γ between any pair of perpendicular directions nk
and mk in the neighborhood N (P ) of the particle P (xj ).
Volume change. The volumetric strain tr ϵ = ϵkk represents the volume expansion per
unit original volume.

1.3 Compatibility of infinitesimal strains


The question of integrability

1.3.1 Compatibility = integrability


displacement field ui  ϵij = 21 (ui,j + uj,i ) strain field

3 ui ′ s → 6 ϵij ′ s differentiate ui (xk ) OK ! if ui (xk ) ∈ C 1 .


3 ui ′ s ← 6 ϵij ′ s integrate ϵij (xk ) OK ? if what conditions on ϵij (xk )?

How can we prevent the body from cracking (tearing apart) or overlapping?

What are the conditions on ϵ(x) that would guarantee a single-valued continuous u(x)?
1.3. COMPATIBILITY OF INFINITESIMAL STRAINS 7

Figure 1.4: 1-4.jpg

1.3.2 Strain compatibility equations


In symbolic notation,
1
ϵ = (u ⊗ ∇ + ∇ ⊗ u)
2

1
∇×ϵ×∇ = [∇ × (u ⊗ ∇) × ∇ + ∇ × (∇ ⊗ u) × ∇]
2
= 0,
where in the last step the identities
∇ × (∇ ⊗ u) = 0,
(u ⊗ ∇) × ∇ = 0
have been taken into account.
Define the incompatibility tensor η as

η = ∇ × ϵ × ∇. (1.10)

Obviously, it is a symmetric tensor of order 2, which has 6 independent components. The


symmetry of the second-order tensor η is proved as follows.
Proof:
η = ∂i ei × ϵjk ej ⊗ ek × ∂l el
= epij eqkl ϵjk,il ep ⊗ eq .
ηpq = −epij eqlk ϵjk,il ∵ eqkl = −eqlk
= −eplk eqij ϵkj,li ∵ dummy indices ijkl → lkji
= −eqij eplk ϵjk,il ∵ ϵkj = ϵjk , ∂ 2 /(∂xi ∂xl ) = ∂ 2 /(∂xl ∂xi )
= ηqp ,
completing the proof. 
Therefore,

η=0 (1.11)
8 CHAPTER 1. KINEMATICS

is a necessary condition for a compatible strain field. In indicial notation, it amounts to


the following 6 independent linear equations known as Saint-Venant’s strain compatibility
equations:
−η11 = e1ij e1lk ϵjk,il = ϵ22,33 + ϵ33,22 − 2ϵ23,23 = 0 (1.12)
−η22 = e2ij e2lk ϵjk,il = ϵ33,11 + ϵ11,33 − 2ϵ13,13 = 0 (1.13)
−η33 = e3ij e3lk ϵjk,il = ϵ11,22 + ϵ22,11 − 2ϵ12,12 = 0 (1.14)
−η12 = e1ij e2lk ϵjk,il = ϵ23,31 − ϵ12,33 − ϵ33,21 + ϵ13,23 = 0 (1.15)
−η23 = e2ij e3lk ϵjk,il = ϵ13,12 + ϵ21,31 − ϵ23,11 − ϵ11,23 = 0 (1.16)
−η31 = e3ij e1lk ϵjk,il = ϵ12,23 + ϵ23,12 − ϵ22,31 − ϵ31,22 = 0 (1.17)

Remark 1. An alternative derivation. The strain compatibility equations for a strain


field may be alternatively derived as follows.
ϵij = 12 (ui,j + uj,i ) .
⇒ ϵil,jk = 12 (ui,ljk + ul,ijk ) ,
ϵjk,il = 12 (uj,kil + uk,jil ) ,
ϵik,jl = 12 (ui,kjl + uk,ijl ) ,
ϵjl,ik = 21 (uj,lik + ul,jik ) .
⇒ ϵil,jk + ϵjk,il − ϵik,jl − ϵjl,ik = 0. (1.18)
The last expression contains 81 component equations but only 6 are independent
and they are nothing but Eqs. (1.12)-(1.17) as can be verified straightforward.
Remark 2. The 6 strain compatibility equations (1.12)-(1.17) are independent. To illus-
trate this, let us stipulate a strain field
 
0 0 0
ϵij (xk ) =  0 0 x2 x3  .
0 x3 x2 0
It violates Eq. (1.12), although already satisfying Eqs. (1.13)-(1.17). Equation
(1.12) must be independent of Eqs. (1.13)-(1.17). In fact, we can compute its
incompatibility tensor field
 
−2 0 0
ηij (xk ) =  0 0 0  ̸= [0],
0 0 0
so that the strain field is not compatible.
Remark 3. For an incompatibility tensor field ηij (xk ), no matter whether it vanishes or
not, as can be readily verified by noting the “div curl = zero” identity, we always
have
ηij,j = 0,
known as the Bianchi identity.
1.3. COMPATIBILITY OF INFINITESIMAL STRAINS 9

1.3.3 From the infinitesimal strain tensor field to the infinitesi-


mal rotation vector field
The infinitesimal strain tensor ϵ and the infinitesimal rotation vector ω are independent.
However, their derivatives are intimately related as indicated in Eq. (1.6) or equivalently
in Eq. (1.7). Using these differential relations we now derive an integral formula for finding
ω along a path of integration in the domain D j R3 .
Let us write
∂ωk
dωk = dxi
∂xi
in the symbolic form

dω = (ω ⊗ ∇) · dx.

Integration yields
∫ x
ω(x) = ω(x0 ) + [ω(x′ ) ⊗ ∇′ ] · dx′ ,
x0

where the (right) gradient operator ⊗∇′ is associated with x′ . Using Eq. (1.7) we obtain
∫ x
ω(x) = ω(x0 ) + [∇′ × ϵ(x′ )] · dx′ . (1.19)
x0

1.3.4 Simply and multiply connected domains


A domain D is said to be simply connected if every simple,2 closed piecewise smooth curve
in D is the boundary of a piecewise smooth surface lying in D; otherwise, it is multiply
connected. For example, a ball (i.e. the interior of a sphere) is simply connected and a
ball with finitely many interior voids is still simply connected, but a doughnut (i.e. the
interior of a torus) is doubly connected.

Figure 1.5: 1-5.jpg

The definition of simple connectivity stated above is well suited for invoking Stokes’
theorem. An equivalent yet more popular definition is as follows. A domain D is simply
10 CHAPTER 1. KINEMATICS

Figure 1.6: 1-6.jpg

connected if every closed piecewise smooth curve in D can be continuously shrunk to any
point in D without leaving D.
An (M +1)-fold multiply connected domain D has a total of M irreconcilable closed
curves, namely, closed curves which cannot be made to coincide by continuous variation
without leaving D. For example, a doubly connected domain has only one irreconcil-
able closed curve since it has one “hole” and a triply connected domain has only two
irreconcilable closed curves since it has two “holes.”

1.3.5 The necessary and sufficient condition in a simply con-


nected domain
For a simply connected domain, η = 0 is not only necessary but also sufficient to guarantee
a single-valued continuous displacement field u(x). In other words, a strain field ϵ(x) ∈ C 2
satisfying η = ∇ × ϵ × ∇ = 0 is so compatible that there exists a single-valued continuous
displacement field u(x). The field is unique up to rigid-body displacements (translations
and rotations).
Proof: Let us first convert

du = Ω · dx + ϵ · dx

to

du = ω × dx + dx · ϵ.

Then its integration yields


∫ x ∫ x
′ ′
u(x) = u(x0 ) + ω(x ) × dx + dx′ · ϵ(x′ ). (1.20)
x0 x0

2
A curve is simple if it neither intersects nor touches itself.
1.3. COMPATIBILITY OF INFINITESIMAL STRAINS 11

Now integrating by parts the first integral of Eq. (1.20) and using Eq. (1.6), we obtain
∫ x
ω(x′ ) × dx′
∫x0x
= {d[ω(x′ ) × (x′ − x)] − dω(x′ ) × (x′ − x)}
x0
∫ x
= ω(x0 ) × (x − x0 ) − dω(x′ ) × (x′ − x)
∫x0x
= ω(x0 ) × (x − x0 ) − dx′ · {[∇′ ⊗ ω(x′ )] × (x′ − x)}
∫xx0
= ω(x0 ) × (x − x0 ) + dx′ · {[ϵ(x′ ) × ∇′ ] × (x′ − x)}.
x0

With this Eq. (1.20) becomes


∫ x
u(x) = u(x0 ) + ω(x0 ) × (x − x0 ) + dx′ · U(x′ , x), ∀x ∈ D, (1.21)
x0

where the integrand

U(x′ , x) = [ϵ(x′ ) × ∇′ ] × (x′ − x) + ϵ(x′ ) (1.22)

is a tensor field of order 2.


For the displacement field u(x) of Eq. (1.21) to be a ∫single-valued continuous function
x
of x in a simply connected domain D j R3 , the integral x0 dx′ · U(x′ , x) must be indepen-
dent of the path of integration C(x′ ) ⊂ D connecting two points x0 ∈ D and x ∈ D. This
means that the integrand U(x′ , x) must be an exact differential, for which the necessary
and sufficient condition is

∇′ × U(x′ , x) = 0, ∀x′ , x ∈ D, (1.23)

in which 0 is the zero tensor of order 2 and ∇′ × is the curl operator associated with x′ .
To explore the condition (1.23), let us first consider in detail the expression
[ϵ(x′ ) × ∇′ ] × (x′ − x)
= [ϵ′ij ei ⊗ ej × ∂k′ ek ] × (x′ℓ eℓ − xℓ eℓ )
= [epjk ϵ′ij,k ei ⊗ ep ] × (x′ℓ − xℓ )eℓ
= [epjk eqpl ϵ′ij,k ei ⊗ eq (x′ℓ − xℓ )
= [δjℓ δkq − δjq δkℓ ]ϵ′ij,k ei ⊗ eq (x′ℓ − xℓ )
= (ϵ′iℓ,q − ϵ′iq,ℓ )ei ⊗ eq (x′ℓ − xℓ ),
in which the properties epjk = ejkp and eqpℓ = −epqℓ of the permutation symbol and the
e-δ identity

δjq δjℓ
ejkp epqℓ = = δjq δkℓ − δjℓ δkq
δkq δkℓ
12 CHAPTER 1. KINEMATICS

have been used and the strain ϵ(x′ ) associated with x′ is abbreviated as ϵ′ij when written
in component form.
Then, let us further calculate the curl of the above expression, by using the chain rule

and noting ∂m (x′ℓ − xℓ ) = δmℓ as follows.
∇′ × {[ϵ(x′ ) × ∇′ ] × (x′ − x)}

{ }
= ∂m em × (ϵ′iℓ,q − ϵ′iq,ℓ )ei ⊗ eq (x′ℓ − xℓ )
{ }
= {[∇′ × ϵ(x′ ) × ∇′ ] × (x′ − x)} + (ϵ′iℓ,q − ϵ′iq,ℓ )enmi en ⊗ eq ∂m

(x′ℓ − xℓ )
{ }
= {[∇′ × ϵ(x′ ) × ∇′ ] × (x′ − x)} + (ϵ′im,q − ϵ′iq,m )enmi en ⊗ eq .
{ }
The second bracket (ϵ′im,q − ϵ′iq,m )enmi en ⊗ eq in the last line deserves our special atten-
tion. It contains two parts.
1. The first part ϵ′im,q enmi en ⊗ eq has 9 components (n, q = 1, 2, 3), each of which
can be shown to vanish, for example, the n = 1, q = 1 component ϵ′im,1 e1mi =
ϵ′32,1 e123 + ϵ′23,1 e132 = ϵ′23,1 − ϵ′23,1 = 0 due to the symmetry of the strain tensor.

2. The second part −ϵ′iq,m enmi en ⊗ eq can be readily shown to be nothing but −∇′ ×
ϵ(x′ ).
With all these and upon substituting Eq. (1.22), the condition (1.23) turns out to be
0 = ∇′ × U(x′ , x)
= ∇′ × {[ϵ(x′ ) × ∇′ ] × (x′ − x) + ϵ(x′ )}
= [∇′ × ϵ(x′ ) × ∇′ ] × (x′ − x) − ∇′ × ϵ(x′ ) + ∇′ × ϵ(x′ ),
= [∇′ × ϵ(x′ ) × ∇′ ] × (x′ − x).
Since x′ − x is arbitrary, it follows that

∇ × ϵ(x) × ∇ = 0, ∀x ∈ D. (1.24)

Thus, it is concluded that Eq. (1.24) is the necessary and sufficient condition. 
In summary, in a simply connected domain D, a strain field ϵ(x) ∈ C 2 satisfying Eq.
(1.24) is the necessary and sufficient condition for a displacement field u(x) given by Eq.
(1.21) with U(x′ , x) specified by Eq. (1.22) to be single-valued continuous, unique up to
a rigid-body displacement u(x0 ) + ω(x0 ) × (x − x0 ).

1.3.6 The necessary and sufficient conditions in a multiply con-


nected domain
Let us examine the integral formulae (1.19) for ω(x) and (1.21) for u(x). We are thus
led to define the Frank vectors Fr(Ci ), i = 1, · · · , M and the Burgers vectors Bu(Ci ), i =
1, · · · , M and to conclude that for an (M +1)-fold multiply connected domain D, the
necessary and sufficient conditions are

η(x) = 0, Bu(Ci ) = 0, Fr(Ci ) = 0,


1.4. LARGE DEFORMATION 13

for all x ∈ D and for each and every irreconcilable closed curve Ci ⊂ D, i = 1, · · · , M .
The Burgers vector Bu(Ci ) along Ci is defined as
I I
Bu(Ci ) = ′ ′
dx · U(x , 0) = dx′ · {[ϵ(x′ ) × ∇′ ] × x′ + ϵ(x′ )} .
Ci Ci

The Burgers vector Bu(Ci ) ̸= 0 measures the dislocation along Ci .


The Frank vector Fr(Ci ) along Ci is defined as
I
Fr(Ci ) = [∇′ × ϵ(x′ )] · dx′ .
Ci

The Frank vector Fr(Ci ) ̸= 0 measures the disclination along Ci .

1.4 Large deformation


1.4.1 Lagrangian coordinates and Eulerian coordinates
The motion of a deformable body is described by x = χ(X) or xi = χi (XA ).

P, Q material particles,
X = XA E A position vector of particle P in the reference configuration,
which has coordinates (X1 , X2 , X3 ),
called Lagrangian (or material) coordinates,
in the fixed rectangular Cartesian coordinate system OX1 X2 X3
with origin O and orthonormal basis {EA }3A=1 ,
x = xi e i position vector of particle P in the deformed configuration,
which has coordinates (x1 , x2 , x3 ),
called Eulerian (or spatial) coordinates,
in the fixed rectangular Cartesian coordinate system ox1 x2 x3
with origin o and orthonormal basis {ei }3i=1 ,


Oo constant vector from the origin O of the coordinate system OX1 X2 X3
to the origin o of the coordinate system ox1 x2 x3 ,


u = x − X+Oo displacement vector of P ,
du = dx − dX displacement increment vector.
In some occasions, the origin O of the Lagrangian coordinate system OX1 X2 X3 and the


origin o of the Eulerian coordinate system oxyz may happen to coincide (so that Oo = 0)
but the coordinate X1 -, X2 -, X3 -axes orient differently and the Lagrangian basis vectors
E1 , E2 , E3 are related to the Eulerian basis vectors e1 , e2 , e3 through the change of bases
formula ek = RkA EA , where R ∈ SO(3, R) is a fixed constant rotation tensor. Sometimes


not only the origins coincide (so that Oo = 0) but also the coordinte axes and the bases
vectors are identical (so that xk = δkA XA and ek = δkA EA ).
Usually X is referred to as particle and x as point. The Lagrangian description de-
scribes those of each material particle. The Eulerian description describes those at each
spatial point.
14 CHAPTER 1. KINEMATICS

Figure 1.7: 1-7general.jpg

Figure 1.8: 1-7.jpg

1.4.2 Deformation gradient


Let P with Lagrangian coordinates X be an arbitray particle of a deformable body D j R3
and Q an arbitrary neighboring particle with Lagrangian coordinates X + dX in the
neighborhood N (P ) ⊂ D j R3 of P .

particle reference configuration deformed configuration


P ∈ N (P ) X → x = χ(X)
Q ∈ N (P ) X + dX → x + dx = χ(X + dX)

The deformation of the neighborhood N (P ) can be described by

dx = χ(X + dX) − χ(X) = ∂χ


∂X
(X) · dX = F · dX,
(1.25)
dxi = χi (XA + dXA ) − χi (XA ) = ∂χi
∂XB
(XA )dXB = FiB dXB ,
1.4. LARGE DEFORMATION 15

where F = FiA ei ⊗ EA = xi,A ei ⊗ EA = ∂xi /∂XA ei ⊗ EA is called the deformation gradient


tensor. For the deformation to be physically admissible, det F > 0; hence, F ∈ GL+ (3, R).3

 det F > 1 volume dV expands.
det F > 0 physically admissible det F = 1 volume dV = dv is isochoric.

0 < det F < 1 volume dV shrinks.
det F = 0 volume dV of N (P ) is compressed to a point such that dv = 0.
det F < 0 volume dV of N (P ) penetrates itself such that dv < 0.

Figure 1.9: 1-8.jpg

For a physically admissible deformation the inverse to Eq.(1.25 is

dX = F−1 · dx = dx · F−T .
−1 −T (1.26)
dXB = FBi dxi = dxi FiB .

The polar decomposition of F is a multiplicative decomposition:


∵ F ∈ GL+ (3, R) F = deformation gradient positive determinant invertible
F=R·U U = right strech positive definite symmetric
F=V·R V = left strech positive definite symmetric
R = rotation R ∈ SO(3, R) special orthogonal

{
R · (U · dX) right polar decomposition, first streching then rotating
dx = F·dX =
V · (R · dX) left polar decomposition, first rotating then stretching

The eigenproblems

U · EUA = EUA λ(A) , A = 1, 2, 3; U · RU = RU · λ,


V · eVi = eVi λ(i) , i = 1, 2, 3; V · RV = RV · λ,

in which

λ = 3A=1 λ(A) EA ⊗ EA , RU = EUA ⊗ EA ∈ SO(3, R), EUA = RU · EA ,

λ = 3i=1 λ(i) ei ⊗ ei , RV = eVi ⊗ ei ∈ SO(3, R), eVi = RV · ei ,

reveal that the positive definite symmetric tensors U and V have the same set of positive
real eigenvalues arranged in the order λ1 ≥ λ2 ≥ λ3 > 0, called principal streches, but
3
GL(m, R) = {F ∈ Rm×m | det F ̸= 0}, GL+ (m, R) = {F ∈ Rm×m | det F > 0}.
16 CHAPTER 1. KINEMATICS

different orthonormal sets of corresponding eigenvectors, {EUA }3A=1 and {eVi }3i=1 , called
principal directions, related via R ∈ SO(3, R) by

eVi = δiA R · EUA .

Therefore, U and V may well be characterized by the following expressions:



UAB EA ⊗ EB = U = RU · λ · (RU )T = 3A=1 λ(A) EUA ⊗ EUA ,

Vij ei ⊗ ej = V = RV · λ · (RV )T = 3i=1 λ(i) eVi ⊗ eVi .

Relative to the basis {eVi ⊗ EUA }3i,A=1 the two-point tensors F and R have simple, intrinsic
expressions:
∑ ∑3
FiA ei ⊗ EA = F = R · U = R · ( 3A=1 λ(A) EUA ⊗ EUA ) = A=1 λ(A) δiA ei ⊗ EA ,
V U

RiA ei ⊗ EA = R = R · I = R · (EA ⊗ EA ) =
U U
δiA ei ⊗ EUA .
V

It is interesting to notice the subtleties arising when the principal stretches λ1 , λ2 , λ3


are referred to different tensor bases as listed below.

F = 3A=1 λ(A) δiA eVi ⊗ EUA , R = δiA eVi ⊗ EUA ,

U = 3A=1 λ(A) EUA ⊗ EUA , I = EUA ⊗ EUA ,
∑3
V = i=1 λ(i) eVi ⊗ eVi , I = eVi ⊗ eVi ,
∑3
λ = A=1 λ(A) EA ⊗ EA , I = EA ⊗ EA ,
∑3
λ = i=1 λ(i) ei ⊗ ei , I = ei ⊗ ei ,

Remark 1. Both F = FiA ei ⊗ EA and R = RiA ei ⊗ EA are referred to the basis {ei ⊗
EA }3i,A=1 and hence are the so-called two-point tensors.

Remark 2. The complex number ζ has the polar decomposition ρeiθ where ρ = |ζ| and
θ = arg ζ. The analogies and extensions of polar forms are listed below.

object′ operator object


real number real number real number
24 =3· 8
complex number complex number complex number
w =ζ · z
w = eiθ · ρ · z
w =ρ ·e · iθ
z
tensor of order 1 tensor of order 2 tensor of order 1
dx = F· dX
dx = R · U· dX
dx = V · R· dX
Remark 3. For large deformation the polar decomposition F = R · U = V · R of the
deformation gradient tensor F is multiplicative, while for small deformation the
1.4. LARGE DEFORMATION 17

decomposition H = Ω + ϵ of the displacement gradient H in Eq. (1.1) is additive.


Note that
deformation gradient rotation stretch
F ∈ GL+ (3, R) R ∈ SO(3, R) U, V positive definite symmetric
displacement gradient infinitesimal rotation infinitesimal strain
H ∈ gl(3, R) Ω ∈ so(3, R) ϵ symmetric
Remark 4. Let us compare the three rotation tensors:

rotation active vs. passive relation

SO(3, R) ∋ R = RiA ei ⊗ EA active rotation F=R·U=V·R


= δiA eVi ⊗ EUA of material N (P )

SO(3, R) ∋ RU = EUA ⊗ EA active rotation U = RU · λ · (RU )T


of principal axes EUA

SO(3, R) ∋ RV = eVi ⊗ ei active rotation V = RV · λ · (RV )T


of principal axes eVi

1.4.3 Nanson’s formula for area vectors


An element of area in the referential configuration dA = NdS becomes da = nds in
the deformed configuration after deformation, in which dS = |dA|, ds = |da| are area
magnitudes and N, n are unit vectors so that N · N = 1, n · n = 1. To derive the Nanson’s
formula
da = JF−T · dA, i.e. nds = JF−T · NdS,
−T −T (1.27)
dai = JFiB dAB , i.e. ni ds = JFiB NB dS,
expressing the relationship between the two area vectors dA and da, let us consider an ar-
bitrary length element dX in the referential configuration and the corresponding arbitrary
length element dx in the deformed configuration. The corresponding volume elements are
dV = dX · dA, dv = dx · da and they are related by dv = JdV , where J = det F. Hence,
dv = dx · da = JdX · dA = JdV.
Using Eq. (1.26)2 , dX = dx · F−T , we obtain
dx · da = Jdx · F−T · dA.
Since dx is arbitrary, it is concluded
da = JF−T · dA, (i.e.nds = JF−T · NdS)
that is the desired Nanson’s formula (1.27).
18 CHAPTER 1. KINEMATICS

1.4.4 Deformation tensors


C = F · F = U2
T
= Green deformation tensor = right Cauchy-Green tensor
b = F · FT = V 2 = Finger deformation tensor = left Cauchy-Green tensor
C−1 = U−2 = Piola deformation tensor
c = b−1 = V−2 = Cauchy deformation tensor
Since their symbols are far from standardized in literature, the deformation tensors
C, b, C−1 , c may be denoted in more informative, direct manner U2 , V2 , U−2 , V−2 , re-
spectively.

Meaning of the deformation tensors.


Among the deformation tensors U2 , V2 , U−2 , V−2 , once one of them is an identity, all
of them must reduce to identities. Consequently F = R and the neighborhood N (P )
moves without deformation.

dL = |dX| = (dX · dX)1/2 = (dx · V−2 · dx) = (dx · c · dx)1/2


1/2

1/2
dℓ = |dx| = (dx · dx)1/2 = (dX · U2 · dX) = (dX · C · dX)1/2

Meaning of the Green deformation tensor C.

(a) Length change. Consider an arbitrary line element dX = NdL, where N · N = 1.


Along the N-direction,

( )2
dℓ dX · C · dX
=
dL (dL)2
 T   
 N1  C11 C12 C13  N1 
= N · C · N = NA CAB NB = N2  C22 C23  N2 .
   
N3 sym. C33 N3

In particular, let dX = NdL = [1 0 0]T dL. Its extension rate along the X1 -direction
is
dℓ √
= C11 .
dL

(b) Angle change. Consider two arbitrary line elements dX(1) and dX(2) . In the deformed
configuration,

dx(1) · dx(2) = |dx(1) ||dx(2) | cos θ


= dℓ(1) dℓ(2) cos θ
(1) (2)
= dxi dxi
(1) (2)
= FiA dXA FiB dXB
(1) (2)
= CAB dXA dXB .
1.4. LARGE DEFORMATION 19

In particular, let dX(1) = [1 0 0]T dL(1) ⊥ dX(2) = [0 1 0]T dL(2) . Then dℓ(1) dℓ(2) cos θ =
C12 dL(1) dL(2) . √
The direction cosine of the angle θ between their deformed elements
is cos θ = C12 / C11 C22 . Hence, the angle change is as follows.
π C12
Θ= −→ θ = arccos √ .
2 C11 C22

Figure 1.10: 1-11.jpg

(c) Area change. Consider two arbitrary line elements dX(1) and dX(2) . In the deformed
configuration,

da = |dx(1) × dx(2) |
= |dx(1) ||dx(2) | sin θ
= dℓ(1) dℓ(2) sin θ
(1) (2)
= dxi dxi tan θ
(1) (2)
= FiA dXA FiB dXB tan θ
(1) (2)
= CAB dXA dXB tan θ

In particular, let dX(1) = [1 0 0]T dL(1) ⊥ dX(2) = [0 1 0]T dL(2) . Then dA =


dL(1) dL(2)√ and da = dℓ(1) dℓ(2) sin θ = C12 dL(1) dL(2) tan θ = C12 dA tan θ, where
tan θ = 1 − cos2 θ/ cos θ can be deduced from cos θ known above. Hence, the
area expands by the ratio

da
= C11 C22 − C122
.
dA

(d) Volume change. Since dv = |F|dV and C = FT · F, the volume expands by the ratio
dv √
= detC.
dV

Meaning of the Cauchy deformation tensor c.


20 CHAPTER 1. KINEMATICS

Figure 1.11: 1-12.jpg

(a) Length change. Let dx = ndℓ = [1 0 0]T dℓ. Its extension rate is

dℓ 1
=√ .
dL c11

(b) Angle change. Let dx(1) = [1 0 0]T dℓ(1) ⊥ dx(2) = [0 1 0]T dℓ(2) . Then dL(1) dL(2) cos Θ =
c12 dℓ(1) dℓ(2) . The direction cosine of the angle Θ between their undeformed elements

is cos Θ = c12 / c11 c22 . Hence, the angle change is as follows.

c12 π
Θ = arccos √ −→ θ = .
c11 c22 2

Figure 1.12: 1-13.jpg

(c) Area change. Let dx(1) = [1 0 0]T dℓ(1) ⊥ dx(2) = [0 1 0]T dℓ(2) . Then da =
dℓ(1) (2) (1) (2) (1) (2)
√dℓ and dA = dL dL sin Θ = c12 dℓ dℓ tan Θ = c12 da tan Θ, where tan Θ
= 1 − cos2 Θ/ cos Θ can be deduced from cos Θ known above. Hence, the area
expands by the ratio

da 1
=√ .
dA c11 c22 − c212
1.4. LARGE DEFORMATION 21

(d) Volume change. Since dv = |F|dV and c = (F · FT )−1 , the volume expands by the
ratio
dv 1
=√ .
dV detc

1.4.5 Strain tensors


{
dX · U2 · dX − dX · dX = dX · (U2 − I) · dX = 2dX · ϵL · dX,
(dℓ) − (dL) =
2 2
dx · dx − dx · V−2 · dx = dx · (I − V−2 ) · dx = 2dx · ϵE · dx.

F · dX = dx = dX + du = dX + H · dX, F = I + H, FT · F = I + H + HT + HT · H,
F−1 · dx = dX = dx − du = dx − h · dx, F−1 = I − h, F−T · F−1 = I − h − hT + hT · h.

Definitions of Lagrangian and Eulerian strain tensors:

ϵL = 21 (C − I) = 12 (U2 − I) = 12 (FT · F − I) = 21 (H + HT + HT · H),


−2 −T −1
ϵ = 2 (I − c) = 2 (I − V ) = 2 (I − F · F ) = 21 (h + hT − hT · h).
E 1 1 1

1 1 ∂uA ∂uB ∂uC ∂uC


ϵLAB = 2
(HAB + HBA + HCA HCB ) = (
2 ∂XB
+ ∂X A
+ ∂X A ∂XB
),
∂uj
ϵEij = 1
2
(hij + hji − hki hkj ) = 1 ∂ui
(
2 ∂xj
+ ∂xi
− ∂xi ∂xj ).
∂uk ∂uk

Remarks:
∂xi
∂XA
(or denoted FiA or F) are deformation gradients.
∂uA ∂ui
∂XB
(or denoted HAB or H) and ∂x j
(or denoted hij or h) are displacement gradients.
du = dx − dX ⇔ H = F − I ⇔ h = I − F−1 .

Note that there are nonlinear terms in the definitions of ϵL and ϵE .
ϵ= infinitesimal 
L
ϵ = Lagrangian strain tensors are symmetric second-order tensors.

ϵE = Eulerian
Any one component of ∂X ∂uA
B
∂ui
and ∂x j
is large ⇔ large deformation.
All components of ∂uA and ∂ui are ≪ 1 ⇔ small deformation ⇔ ϵL ∼
∂XB ∂xj = ϵE ∼
= ϵ.

Generalized Lagrangian strains ϵL(r) :


{ (∑ )
1
(U r
− I) = 1 3
λr
E U
⊗ E U
− I , r ̸= 0,
ϵL(r) = r

r A=1 (A) A A
ln U = 3A=1 ln λ(A) EUA ⊗ EUA , r = 0.

Generalized Eulerian strains ϵE(r) :


{ (∑ )
1
(V − I) = r
r 1
i=1 λ(i) ei ⊗ ei − I , r ̸= 0,
3 r V V
ϵE(r) = r
∑3
ln V = i=1 ln λ(i) eVi ⊗ eVi , r = 0.

Here r is a real number.


22 CHAPTER 1. KINEMATICS


ϵL = ϵL(2) Lagrangian (or Green or Green-Lagrange) 



ϵE = ϵE(−2) Eulerian (or Almansi or Euler-Almansi) 
ϵL(−2) , ϵE(−2) Almansi strain tensor(s)


ϵL(0) , ϵE(0) (Hencky) logarithmic 


ϵL(1) , ϵE(1) Biot (engineering)

Meaning of the strain tensors.


Among ϵL(r) and ϵE(r) for all r ∈ R, any one (either ϵL(r) or ϵE(r) for one fixed r ∈ R)
vanishes, then all vanish and F = R. The neighborhood N (P ) moves without deformation.
Meaning of the Lagrangian strain tensor ϵL .
Since ϵLAB = 21 (CAB − δAB ), the meaning of CAB may be readily carried over to that of
ϵLAB .
Meaning of the Eulerian strain tensor ϵE .
ij = 2 (δij − cij ), the meaning of cij may be readily carried over to that of ϵij .
1
Since ϵE E

1.4.6 Proof of symmetry of strain tensors


• (infinitesimal) strain tensor ϵ:
( )
1 ∂ui ∂uj
ϵij = +
2 ∂xj ∂xi
( )
1 ∂uj ∂ui
= +
2 ∂xi ∂xj
= ϵji .

• Lagrangian (Green, Green-Lagrange) strain tensor ϵL :

1 ∂uA ∂uB ∂uC ∂uC


ϵLAB = ( + + )
2 ∂XB ∂XA ∂XA ∂XB
1 ∂uB ∂uA ∂uC ∂uC
= ( + + )
2 ∂XA ∂XB ∂XB ∂XA
= ϵLBA .

• Eulerian (Almansi, Euler-Almansi) strain tensor ϵE :

1 ∂ui ∂uj ∂uk ∂uk


ϵE
ij = ( + − )
2 ∂xj ∂xi ∂xi ∂xj
1 ∂uj ∂ui ∂uk ∂uk
= ( + − )
2 ∂xi ∂xj ∂xj ∂xi
= ϵE
ji .
1.4. LARGE DEFORMATION 23

• (material) logarithmic (Hencky) strain tensor ϵL(0) :


3
U = λ(A) EUA ⊗ EUA ,
A=1
[U ] = [R ][λ][RU ]T ,
U
 
λ(1) 0 0 [ ]
where [λ] =  0 λ(2) 0  and [RU ] = {E1U }{E2U }{E3U } ,
0 0 λ(3)
{ U } T
[U ]T = [R ][λ][RU ]T
= [RU ][λ][RU ]T
= [U ] ,
∴ UT = U.


3
ϵL(0) = ln U = ln λ(A) EUA ⊗ EUA ,
A=1
[ L(0)
]
ϵ = [ln U ] = [RU ][ln λ][RU ]T ,
 
ln λ(1) 0 0
where [ln λ] =  0 ln λ(2) 0 ,
0 0 ln λ(3)
∴ (ϵL(0) )T = ϵL(0) .

• generalized Lagrangian strain tensor ϵL(r) for 0 ̸= r ∈ R:

∵ UT = U,
( )T
( L(r) )T 1
∴ ϵ = (U − I)
r
r
1
= (Ur − I)T
r
1( T r )
= (U ) − IT
r
1
= (Ur − I)
r
= ϵL(r) .
24 CHAPTER 1. KINEMATICS

• (spatial) logarithmic (Hencky) strain tensor ϵE(0) :


3
V = λ(i) eVi ⊗ eVi ,
i=1
[V ] = [RV ][λ][RV ]T ,
 
λ(1) 0 0 [ ]
where [λ] =  0 λ(2) 0  and [RV ] = {eV1 }{eV2 }{eV3 } ,
0 0 λ(3)
( V )T
[V ]T = [R ][λ][RV ]T
= [RV ][λ][RV ]T
= [V ] ,
∴ VT = V.


3
ϵ E(0)
= ln V = ln λ(i) eVi ⊗ eVi ,
i=1
[ E(0)
]
ϵ = [ln V ] = [RV ][ln λ][RV ]T ,
 
ln λ(1) 0 0
where [ln λ] =  0 ln λ(2) 0 ,
0 0 ln λ(3)
∴ (ϵE(0) )T = ϵE(0) .

• generalized Eulerian strain tensor ϵE(r) for 0 ̸= r ∈ R:

∵ VT = V,
( )T
( E(r) )T 1
∴ ϵ = (V − I)
r
r
1
= (Vr − I)T
r
1( T r )
= (V ) − IT
r
1
= (Vr − I)
r
= ϵE(r) .

1.4.7 Examples
Ex.1 Plane deformation. A total of 4 degrees of freedom. All are expressed in matrices
of components relative to the bases {EA }3A=1 and {ei }3i=1 and their tensor products.
1.4. LARGE DEFORMATION 25

{dx} = [F ]{dX},
    
 dx1  F11 F12 0  dX1 
dx2 =  F21 F22 0  dX2 .
   
dx3 0 0 1 dX3
[F ] = [R][U ] = [R][RU ][λ][RU ]T , [RU ] = [{E1U }{E2U }{E3U }],
    
F11 F12 0 cos ΘR − sin ΘR 0 U11 U12 0
 F21 F22 0  =  sin ΘR cos ΘR 0   U22 0 
0 0 1 0 0 1 sym. 1
 
cos ΘR − sin ΘR 0
=  sin ΘR cos ΘR 0 
0 0 1
   T
cos ΘU − sin ΘU 0 λ(1) 0 0 cos ΘU − sin ΘU 0
 sin ΘU cos ΘU 0   0 λ(2) 0   sin ΘU cos ΘU 0  .
0 0 1 0 0 1 0 0 1

Figure 1.13: 1-9.jpg


26 CHAPTER 1. KINEMATICS

[F ] = [V ][R] = [RV ][λ][RV ]T [R], [RV ] = [{eV1 }{eV2 }{eV3 }],


    
F11 F12 0 V11 V12 0 cos ΘR − sin ΘR 0
 F21 F22 0  =  V22 0   sin ΘR cos ΘR 0 
0 0 1 sym. 1 0 0 1
   T
cos θV − sin θV 0 λ(1) 0 0 cos θV − sin θV 0
=  sin θV cos θV 0   0 λ(2) 0   sin θV cos θV 0 
0 0 1 0 0 1 0 0 1
 
cos ΘR − sin ΘR 0
 sin ΘR cos ΘR 0  .
0 0 1
θV = ΘR + ΘU .

Note that there are a total of 4 degrees of freedom, expressible in one of the following
sets:

• F11 , F12 , F21 , F22 .

• ΘR , U11 , U12 , U22 .

• ΘR , ΘU , λ(1) , λ(2) .

• V11 , V12 , V22 , ΘR .

• θV , λ(1) , λ(2) , ΘR .

Compute the right stretch tensor U, Green deformation (right Cauchy-Green) tensor
C, Piola deformation tensor C−1 , Lagrangian (Green, Green-Lagrange) strain tensor
ϵL , generalized Lagrangian strain tensor ϵL(r) for 0 ̸= r ∈ R, (material) logarithmic
(Hencky) strain tensor ϵL(0) ≡ ln U, (material) Biot’s (engineering) strain tensor
ϵL(1) as follows.
1.4. LARGE DEFORMATION 27

   T
cos ΘU − sin ΘU 0 λ(1) 0 0 cos ΘU − sin ΘU 0
[U ] =  sin ΘU cos ΘU 0  0 λ(2) 0   sin ΘU cos ΘU 0  ,
0 0 1 0 0 1 0 0 1
    T
cos ΘU − sin ΘU 0 λ2(1) 0 0 cos ΘU − sin ΘU 0
[C] =  sin ΘU cos ΘU 0  0 λ(2) 0   sin ΘU cos ΘU 0  ,
0 0 1 0 0 1 0 0 1
  1 
0 0  cos Θ − sin Θ 0 T
[ −1 ] cos ΘU − sin ΘU 0 λ2(1) U U
  
C = sin ΘU cos ΘU 0  0 λ21 0   sin ΘU cos ΘU 0  ,
(2)
0 0 1 0 0 1 0 0 1
  λ2(1) −1
 T
[ L] cos ΘU − sin ΘU 0 0 0 cos ΘU − sin ΘU 0
  2 
ϵ = sin ΘU cos ΘU 0  0
λ2(2) −1
0  sin ΘU cos ΘU 0  ,
2
0 0 1 0 0 0 0 0 1
  λ(1) −1
r   T
[ L(r) ] cos ΘU − sin ΘU 0 0 0 cos ΘU − sin ΘU 0
  r 
ϵ = sin ΘU cos ΘU 0  0
λr(2) −1
0  sin ΘU cos ΘU 0  ,
r
0 0 1 0 0 0 0 0 1
   T
[ L(0) ] cos ΘU − sin ΘU 0 ln λ(1) 0 0 cos ΘU − sin ΘU 0
ϵ =  sin ΘU cos ΘU 0  0 ln λ(2) 0   sin ΘU cos ΘU 0  ,
0 0 1 0 0 0 0 0 1
   T
[ L(1) ] cos ΘU − sin ΘU 0 λ(1) − 1 0 0 cos ΘU − sin ΘU 0
ϵ =  sin ΘU cos ΘU 0  0 λ(2) − 1 0   sin ΘU cos ΘU 0  .
0 0 1 0 0 0 0 0 1

Compute the left stretch tensor V, Finger deformation (left Cauchy-Green) tensor
b, Cauchy deformation tensor c, Eulerian (Almansi, Euler-Almansi) strain tensor
ϵE , generalized Eulerian strain tensor ϵE(r) for 0 ̸= r ∈ R, (spatial) logarithmic
(Hencky) strain tensor ϵE(0) ≡ ln V, (spatial) Biot’s (engineering) strain tensor ϵE(1)
as follows.
28 CHAPTER 1. KINEMATICS

   T
cos ΘV − sin ΘV 0 λ(1) 0 0 cos ΘV − sin ΘV 0
[V ] =  sin ΘV cos ΘV 0  0 λ(2) 0   sin ΘV cos ΘV 0  ,
0 0 1 0 0 1 0 0 1
    T
cos ΘV − sin ΘV 0 λ2(1) 0 0 cos ΘV − sin ΘV 0
[b] =  sin ΘV cos ΘV 0  0 λ(2) 0   sin ΘV cos ΘV 0  ,
0 0 1 0 0 1 0 0 1
  1
0 0
 T
cos ΘV − sin ΘV 0 λ2(1) cos ΘV − sin ΘV 0
 
[c] =  sin ΘV cos ΘV 0  0 λ21 0   sin ΘV cos ΘV 0  ,
(2)
0 0 1 0 0 1 0 0 1
  1−λ−2
 T
[ E] cos ΘV − sin ΘV 0 (1)
0 0 cos ΘV − sin ΘV 0
  2 
ϵ = sin ΘV cos ΘV 0  0
1−λ−2 (2)
0  sin ΘV cos ΘV 0  ,
2
0 0 1 0 0 0 0 0 1
  λr(1) −1
 T
[ E(r) ] cos ΘV − sin ΘV 0 0 0 cos ΘV − sin ΘV 0
  r 
ϵ = sin ΘV cos ΘV 0  0
λr(2) −1
0  sin ΘV cos ΘV 0  ,
r
0 0 1 0 0 0 0 0 1
   T
[ E(0) ] cos ΘV − sin ΘV 0 ln λ(1) 0 0 cos ΘV − sin ΘV 0
ϵ =  sin ΘV cos ΘV 0  0 ln λ(2) 0   sin ΘV cos ΘV 0  ,
0 0 1 0 0 0 0 0 1
   T
[ E(1) ] cos ΘV − sin ΘV 0 λ(1) − 1 0 0 cos ΘV − sin ΘV 0
ϵ =  sin ΘV cos ΘV 0  0 λ(2) − 1 0   sin ΘV cos ΘV 0  .
0 0 1 0 0 0 0 0 1

Ex.2 Simple shear. Only one degree of freedom k > 0, or θV = 12 arctan k2 ∈ (0, π4 ).

Figure 1.14: 1-10.jpg


1.4. LARGE DEFORMATION 29
 
1 k 0
• [F ] =  0 1 0  . ⇒ detF = 1, so simple shear deformation is planar and
0 0 1
isochoric.
• ΘR = 2θV − π2 , ΘU = π
2
− θV , λ(1) = cot θV , λ(2) = tan θV .
• θV , λ(1) = cot θV , λ(2) = tan θV , ΘR = 2θV − π2 .

Ex.3 Pure shear. Only one degree of freedom λ > 0.


 
λ 0 0
[F ] = [U ] = [V ] =  0 λ−1 0  , [R] = [I] ⇒ detF = 1, so pure shear deformation
0 0 1
is planar and isochoric.
30 CHAPTER 1. KINEMATICS

Ex.4 Triaxial deformation. A total of three degrees of freedom λ(1) > 0, λ(2) > 0, λ(3) > 0.

 
λ(1) 0 0
[F ] = [U ] = [V ] =  0 λ(2) 0 ,
0 0 λ(3)
 
1 0 0
[R] = [RU ] = [RV ] = [I] =  0 1 0 ,
0 0 1
detF = λ(1) λ(2) λ(3) ,
 
λ2(1) 0 0
[C] = [b] =  0 λ2(2) 0  ,
0 0 λ2(3)
 1 
λ 2 0 0
[ −1 ]  (1) 
= [c] =  0 λ2(2) 0 
 1
C ,
1
0 0 λ2
(3)
 λ2 −1 
(1)
0 0
[ L] [ E ]  2 
=  ,
λ 2 −1
ϵ = ϵ  0
(2)
0 
2
λ2(3) −1
0 0 2
 λr −1 
(1)
0 0
[ L(r) ] [ E(r) ]  r λr(2) −1 
ϵ = ϵ =  0 r
0 ,
λr(3) −1
0 0
  r

[ ] [ ] ln λ(1) 0 0
ϵL(0) = ϵE(0) =  0 ln λ(2) 0 ,
0 0 ln λ(3)
 
[ L(1) ] [ E(1) ] λ(1) − 1 0 0
ϵ = ϵ =  0 λ(2) − 1 0 .
0 0 λ(3) − 1

1.4.8 Small strain and moderate rotation


In this subsection we deal with the situation of small strain and moderate rotation.
Recall
Lagrangian strain tensor ϵL = 12 (H + HT + HT · H), Subsection 1.4.5
displacement gradient tensor H = u ⊗ ∇ = Ω + ϵ, Eq. (1.1)
infinitesimal rotation tensor Ω = 2 (u ⊗ ∇ − ∇ ⊗ u) = 2 (H − H ),
1 1 T
Eq. (1.2)
(infinitesimal) strain tensor ϵ = 12 (u ⊗ ∇ + ∇ ⊗ u) = 12 (H + HT ), Eq. (1.3)
First note that
1.5. PROBLEMS 31

ϵL , H, Ω, and ϵ are second-order tensors.

ϵL = (ϵL )T and ϵ = ϵT are symmetric second-order tensors. See Subsection 1.4.6.

Ω = −ΩT is a skew-symmetric second-order tensor.

H is neither symmetric nor skew-symmetric.

There are nonlinear terms in the definition of ϵL .

Any one component of H is large ⇔ large deformation.

All components of H are ≪ 1 ⇔ small deformation ⇔ ϵL ∼


= ϵ.
Then suppose that the (infinitesimal) strain tensor ϵ is infinitesimal of order 2 (small
strain) while the infinitesimal rotation tensor Ω is infinitesimal of order 1 (moderate
rotation).4 In such a situation the Lagrangian strain tensor ϵL has the expression:
1
ϵL = (H + HT + HT · H)
2
1
= ϵ + (Ω + ϵ)T · (Ω + ϵ)
2
1
= ϵ + (−Ω + ϵ) · (Ω + ϵ)
2
1
= ϵ + (−Ω · Ω + ϵ · ϵ)
2
∼ 1
= ϵ − Ω · Ω.
2
In short,
}
ϵ = O(ω 2 ) 1
⇒ ϵL ∼
= ϵ − Ω · Ω = O(ω 2 )
Ω = O(ω) 2

The above expression is relevant and applicable to flexible structural members such
as thin plates and shells often characteristic of relatively larger rotations compared with
strains. The von Kármán plate theory (to be studied in Chapter 9) is known successful in
dealing with this situation.

1.5 Problems
1. (20%) Given the following displacement field in a cube c − 1 ≤ xi ≤ c + 1 (m):
u1 = (3x21 x2 + 6) × 10−6 (m),
u2 = (x22 + 6x1 x3 ) × 10−6 (m),
4
A quantity (or an array of quantities) is said to be infinitesimal of order 1, 2, · · · etc., in the meaning
that an infinitesimal quantity of higher order is negligible in comparison with an infinitesimal quantity of
lower order.
32 CHAPTER 1. KINEMATICS

u3 = (6x23 + 2x2 x3 + 10) × 10−6 (m),


where c is to be determined and (x1 , x2 , x3 ) are coordinates relative to an orthonormal
basis {ei }3i=1 for the Eulerian description.

(a) Write down the condition(s) for the deformation of this displacement field in
the cube to be physically admissible. Is it a small or a large deformation based
upon a criterion that you consider to be appropriate?
(b) What are the Lagrangian, Eulerian, and infinitesimal strain tensor fields in the
cube?
(c) Are the strain tensor fields in the cube obtained in (b) compatible ones?
(d) In general, prove that the (infinitesimal) strain tensor ϵ, Lagrangian (Green,
Green-Lagrange) strain tensor ϵL , and Eulerian (Almansi, Euler-Almansi) strain
tensor ϵE are symmetrical.

Sol :

(a)
(b)
(c)
(d)

You might also like