Escolar Documentos
Profissional Documentos
Cultura Documentos
Alan D. Freed
Soft
Solids
A Primer to the Theoretical
Mechanics of Materials
Modeling and Simulation in Science, Engineering and Technology
Series Editor
Nicola Bellomo
Politecnico di Torino
Torino, Italy
Soft Solids
A Primer to the Theoretical Mechanics
of Materials
Alan D. Freed
Department of Mechanical Engineering
Saginaw Valley State University
University Center, MI, USA
Mathematics Subject Classification (2010): 15A72, 65L06, 65R20, 74-01, 74A05, 74A10, 74A20,
74B20, 74D10, 74L15, 74S30
ix
x Preface
These seven BVPs are visited at the end of each chapter, applying what you
just learned in the current chapter and intermingling it with what you have
already learned from prior chapters, thereby extending the development
of each experiment and your understanding of them topic by topic. The
four worked-out BVPs increase with complexity. The three exercise BVPs
left for your development do not require higher-level mathematical skills.
All questions are designed to build upon your understanding of the basic
concepts and principles that are being taught to you at that particular time.
By the end of the course, your overall exposure to the mathematics in-
volved should make you more comfortable with them. Mathematics is a
language, and in this textbook, the author intends that you learn it by im-
mersion rather than by the more wrought and formal approach of lemmas,
theorems, and proofs. That can come later, enhanced by your experiences
and intuitions gained after taking this primer course.
Acknowledgments
Many people, in one way or another, helped me bring this textbook into
reality. Foremost is my family who willingly made the greatest sacrifice
that anyone can offer: lost time together.
Formidable were my mentors at the University of WisconsinMadison
who inspired me, and to whom I am indebted. May this be a partial in-
stallment on the debt I owe to Prof. Bela I. Sandor, my thesis advisor, who
introduced me to the topic of mechanics of materials and who taught me
how to write and how to learn from adversity; Prof. Millard W. Johnson, Jr.
(19282009), who taught me classic continuum mechanics, then opened
my eyes to it through applications; and to Prof. Arthur S. Lodge (1922
2005), who patiently taught me the rigor of mathematics and the physics
that underlie continuum mechanics, that one should aspire to think outside
the box and, by example, that good theory and good experiment go hand in
hand.
Prof. Raymond Ogden from the University of Glasgow and Prof. K. R.
Rajagopal from Texas A&M provided critical reviews of earlier drafts of
this book, which helped me sharpen its focus. Prof. Rajagopal was the ad-
vocate through which publication became possible. Dr. Allen Mann was
my managing editor at Birkhuser Science, and Prof. Beth Jorgensen from
Preface xi
Preface ix
Nomenclature xxi
Introduction xxix
Continuum Fields 1
1. Kinematics 5
1.1 Motion . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 Homogeneous Motions . . . . . . . . . . . . . 7
1.2 Velocity and Acceleration . . . . . . . . . . . . . . . . . 8
1.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 Uniaxial Extension . . . . . . . . . . . . . . . 10
1.3.2 Equi-biaxial Extension . . . . . . . . . . . . . 11
1.3.3 Simple Shear . . . . . . . . . . . . . . . . . . . 12
1.3.4 Homogeneous Planar Membranes . . . . . . . . 14
1.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.1 Pure Shear . . . . . . . . . . . . . . . . . . . . 18
1.4.2 Biaxial Extension . . . . . . . . . . . . . . . . 19
1.4.3 Extension Followed by Simple Shear . . . . . . 20
1.4.4 Other Problems . . . . . . . . . . . . . . . . . 22
2. Deformation 23
2.1 Homogeneous Deformation . . . . . . . . . . . . . . . . 25
2.2 Conservation of Mass . . . . . . . . . . . . . . . . . . . 25
2.2.1 Isochoric Deformation . . . . . . . . . . . . . . 26
xiii
xiv Contents
3. Strain 47
3.1 Deformation . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Measures of Strain . . . . . . . . . . . . . . . . . . . . 49
3.2.1 Hencky Strain . . . . . . . . . . . . . . . . . . 51
3.2.2 Infinitesimal Strain/Rotation Relationships . . . 52
3.3 Geometric Interpretations of Strain . . . . . . . . . . . . 53
3.3.1 An Areal Interpretation . . . . . . . . . . . . . 56
3.4 Strain Rates . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4.1 Integration . . . . . . . . . . . . . . . . . . . . 60
3.5 Strain and Strain-Rate Fields for Numerical Analysis . . 61
3.5.1 Formulation in Terms of Green Strain . . . . . . 61
3.5.2 Formulation in Terms of Lodge Strain . . . . . 64
3.6 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.6.1 Uniaxial Extension . . . . . . . . . . . . . . . 67
3.6.2 Equi-biaxial Extension . . . . . . . . . . . . . 68
3.6.3 Simple Shear . . . . . . . . . . . . . . . . . . . 69
3.6.4 Homogeneous Planar Membranes . . . . . . . . 71
3.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.7.1 Pure Shear . . . . . . . . . . . . . . . . . . . . 73
Contents xv
4. Stress 77
4.1 Kirchhoff Stress . . . . . . . . . . . . . . . . . . . . . . 79
4.2 Conservation of Momenta . . . . . . . . . . . . . . . . 79
4.3 First PiolaKirchhoff Stress . . . . . . . . . . . . . . . 81
4.4 Second PiolaKirchhoff Stress . . . . . . . . . . . . . . 83
4.5 Stress Rates . . . . . . . . . . . . . . . . . . . . . . . . 83
4.5.1 Integrate for Stress . . . . . . . . . . . . . . . . 84
4.6 The Extra Stress . . . . . . . . . . . . . . . . . . . . . . 85
4.7 Hills Constitutive Inequalities . . . . . . . . . . . . . . 86
4.7.1 Incompressible Materials . . . . . . . . . . . . 86
4.7.2 Eulerian Formulations . . . . . . . . . . . . . . 87
4.8 Stresses for Numerical Analysis . . . . . . . . . . . . . 88
4.9 Examples . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.9.1 Uniaxial Extension . . . . . . . . . . . . . . . 90
4.9.2 Equi-biaxial Extension . . . . . . . . . . . . . 92
4.9.3 Simple Shear . . . . . . . . . . . . . . . . . . . 95
4.9.4 Homogeneous Planar Membranes . . . . . . . . 97
4.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.10.1 Pure Shear . . . . . . . . . . . . . . . . . . . . 101
4.10.2 Biaxial Extension . . . . . . . . . . . . . . . . 101
4.10.3 Extension Followed by Simple Shear . . . . . . 102
4.10.4 Other Problems . . . . . . . . . . . . . . . . . 102
7. Viscoelasticity 209
7.1 1D Viscoelastic Solid . . . . . . . . . . . . . . . . . . . 210
7.1.1 Clocks . . . . . . . . . . . . . . . . . . . . . . 212
7.1.2 Stress Relaxation Experiment . . . . . . . . . . 212
7.1.3 Causal Deformations . . . . . . . . . . . . . . 214
7.1.4 Memory Kernel Formulation . . . . . . . . . . 215
7.1.5 Additive Strain Formulation . . . . . . . . . . . 216
7.1.6 Quasi-Linear Viscoelasticity . . . . . . . . . . 218
7.2 Viscoelastic Kernels . . . . . . . . . . . . . . . . . . . 220
7.2.1 IOV Kernel . . . . . . . . . . . . . . . . . . . 221
7.2.2 FOV Kernel . . . . . . . . . . . . . . . . . . . 222
7.2.3 BOX Kernel . . . . . . . . . . . . . . . . . . . 229
7.2.4 Implementing a Physical Kernel:
The MCM Kernel . . . . . . . . . . . . . . . . 230
7.3 Additive Strain Fields . . . . . . . . . . . . . . . . . . . 234
7.3.1 Lagrangian Strains . . . . . . . . . . . . . . . . 235
7.3.2 Eulerian Strains . . . . . . . . . . . . . . . . . 236
7.3.3 Field Transfer . . . . . . . . . . . . . . . . . . 237
7.4 K-BKZ Viscoelasticity . . . . . . . . . . . . . . . . . . 238
7.4.1 Viscoelastic Lodge Solid . . . . . . . . . . . . 239
7.4.2 Viscoelastic Green Solid . . . . . . . . . . . . 241
7.4.3 Viscoelastic MooneyRivlin Solid . . . . . . . 242
xviii Contents
Bibliography 343
Index 357
Nomenclature
General
B Body
E1 Exponential integral
E, Two-parameter Mittag-Leffler function
H Heaviside step function
P Particle
S Surface
Y Material surface
R2 Two-dimensional subspace of R3
R3 Ambient Euclidean 3-space
Gamma function
Configuration
m Machine precision
Scalar Fields
Ai Gage areas, i = 1, 2, 3
d Depth
D Dissipation function
E Youngs modulus
f Imposed experimental force
G Relaxation modulus for shear response
G Relaxation kernel for shear response
h Height
H Relaxation spectrum
Ii Elastic invariants, i = 1, : : : , 10
J Dilation, Jacobian of deformation
K Relaxation kernel for bulk response
` Length
xxi
xxii Nomenclature
`i Gage lengths, i = 1, 2, 3
M Viscoelastic memory function
M Viscoelastic memory kernel
p Hydrostatic pressure
} Lagrange multiplier
t Time
T Temperature
Tg Glass transition temperature
Tm Melting temperature
T Engineering stress
T Viscoelastic tangent modulus
U Internal energy
U Internal energy function
V Gage volume
w Width
W Mechanical work
W Elastic strain-energy function
: Guth strain
Fractional order in viscoelastic kernel
Fung parameter for bulk response
Fungs 1D parameter
Fung parameter for shear response
Magnitude of shear
Transverse stretch
Dilatation
Engineering strain
Angle of rotation
Bulk modulus
Lams modulus
Stretch
Areal stretch
Shear modulus
Poissons ratio
Inhomogeneous stretch
Extent of rigid-body rotation
% Mass density
, & Shear stress
Characteristic time in viscoelastic kernel
Proportionality parameter in MooneyRivlin model
Nomenclature xxiii
Tensor Fields
F History function for the deformation gradient F
g Riemannian metric of ambient space
[H ] Matrix representation of tensor H in basis (e1 , e2 , e3 )
[H]IJ IJth component of matrix [H ]
ij Kronecker delta
Infinitesimal strain tensor
! Infinitesimal rotation tensor
E Green strain
Ex Distortional Green strain
E Lodge strain
Ex Distortional Lodge strain
Ey Truesdell strain
Ey Biot strain
EH Hencky strain
I Identity tensor
L Lagrangian velocity gradient
S Second PiolaKirchhoff stress
Sx Deviatoric second PiolaKirchhoff stress
U Right stretch
W Lagrangian vorticity
Y Lagrangian Guth volumetric strain
Z Lagrangian Guth shear strain
Lagrangian extra stress
Spatial Tensor Fields
b Finger deformation
b1 Piola deformation
d Stretching or strain rate
e Signorini strain
eN Distortional Signorini strain
e Almansi strain
eN Distortional Almansi strain
eO BellEricksen strain
I Identity tensor
l Velocity gradient
s Kirchhoff stress
T Cauchy stress
v Left stretch
w Vorticity
z Eulerian Guth strain
Eulerian extra stress
Angular velocity
Third-Order Tensor Field
ijk Permutation operator
xxvi Nomenclature
Greek Alphabet
A () alpha
B (b) beta
(g) gamma
(d) delta
"E (e) epsilon
Z (z) zeta
H (Na) eta
# (th) theta
I (Ne) iota
K (k) kappa
(l) lambda
M (m) mu
N (n) nu
(ks) xi
oO (o) omicron
$ (p) pi
%P (r) rho
& (s) sigma
T (t) tau
(, oo) upsilon
Nomenclature xxvii
Acronyms
ASTM American society for testing and materials
BOX Box model, a.k.a. the QLV model
BVP Boundary value problem
CCM ColeCole model
CFD Computational fluid dynamics
FE Finite elements
FEA Finite element analysis
FOV Fractional-order viscoelastic
GL General linear
IC Initial condition
IOV Integer-order viscoelastic
IRKS Inherent RungeKutta stable
IVP Initial value problem
K-BKZ KayeBernstein, Kearsley, and Zapas
KWW KohlrauschWilliams and Watts
MCM Maxwell chain model
MPL Modified power law
ODE Ordinary differential equation
QLV Quasi-linear viscoelaticity
PDMS Polydimethylsiloxane
PMMA Polymethyl methacrylate
RK RungeKutta
1D, 2D, 3D Spatial dimensions
Introduction
Overview
The mathematical theory that treats materials composed of discrete atoms
and molecules as if they were a smeared-out continuous medium is a true
triumph of human intellect: continuum mechanics. Our confidence in this
theory has been strengthened through our extensive use of engineering
tools like finite elements (FE) and computational fluid dynamics (CFD).
These are giant applications of continuum mechanics where the physi-
cal laws are satisfied over user-specified meshes created to solve bound-
ary value problems (BVPs) of interest, all in accordance with associated
constitutive theories like elasticity, Newtonian and non-Newtonian fluids,
viscoelasticity, creep, plasticity, viscoplasticity, and others, whose models
are used to describe various material behaviors.
That the nonlinear theories in these disciplines are, at present, neither
simple nor entirely accurate representations of Nature only means that hu-
man reason has not yet come to complete terms with physical reality. Non-
linear material modeling remains a fertile topic for creative thought.
This text is a primer for students who are interested in studying how ma-
terial behavior is modeled in engineering applications using mathematics
in accordance with physical laws. Such knowledge is fundamental to your
understanding of modern FE and CFD outputs. The focus of this textbook
is on ideas and concepts and the needed mathematics necessary to come to
such an understanding. Full mathematical rigor is sacrificed from time to
time so as not to intimidate the inexperienced reader. The intent of this text
is to provide upper-level undergraduate and entry-level graduate students
with a basic skill set needed to comprehend our theoretical approach to the
mechanics of continuous materials so that upon entering the work force
they can converse with their colleagues in an intelligent manner and to also
prepare them for further studies, if they are so inclined.
This book is not so much a textbook on continuum mechanics, as it is an
introduction into how engineers use the continuum framework to construct
mathematical relationships for various classes of materials, illustrated via
the material class known as soft solids.
You can only embark on your journey once you have an appreciation
for where you have been and an inspiration as to where you want to go.
This text will help you attain a compass heading, thereby starting your own
journey down this wondrous pathway to adventure and discovery. Each
journey starts with a single step. Only you can decide if you are going to
take that step. Most who venture this step will be sightseers along the way.
Introduction xxxi
A few of you will become earnest practitioners of the craft. Such is the
outcome of making choices along the timeline of your life.
Notation
Throughout this book, special notations are used as visual aids to help you
discern what type of field you are looking at. A field is any characteristic
of a particle of mass that also exists at every other particle throughout a
body, e.g., its mass density, temperature, or elastic modulus. Fields often
require physical units to give them meaning. The notation adopted here
is close to the notation used by Holzapfel (2000) in his classic text on the
subject, whose origins trace back to the encyclopedic works of Truesdell
and Toupin (1960) and Truesdell and Noll (2004).
A function of a variable is expressed using a script font, e.g., y = F (x).
When Roman fonts are used, a scalar field (or number) appears in an italic
font, e.g., a; a vector field (or array) appears in a slant blackboard bold
font, e.g., a; while a tensor field (or matrix) is expressed in a bold italic
font, e.g., a. When Greek fonts are used, a scalar looks like, e.g., ; a
vector field looks like, e.g., ; and a tensor field looks like, e.g., . Fourth-
order tensor fields are typeset in a bold calligraphy font, e.g., a . They are
not as common. They arise when transforming our theories into forms that
are more readily implemented into software. Notational exceptions are kept
to a minimum and typically arise for historical reasons.
Two frames of reference are commonly used in mechanics: material and
spatial. Both reside within the infinite ambient space that we call our world
or universe. The material frame is often referred to as the reference, initial,
or undeformed frame, while the spatial frame also goes by the name of the
current or deformed frame. A field that associates with a material frame is
typeset in an uppercase character from its respective font set, e.g., S , and is
often referred to as its Lagrangian representation. The same physical field,
when associated with a spatial frame, will be typeset in its lowercase char-
acter, i.e., s, and is referred to as its Eulerian representation. Tensors used
to map between these two frames, which have a footing in both frames,
are typeset in an uppercase, upright, bold font, e.g., F. Notational excep-
tions are kept to a minimum and typically arise for historical reasons, e.g.,
Cauchy stress T does not follow this notational dictate; it is an Eulerian
field expressed in a Lagrangian notation.
When expressing vectors and tensors in component form, it has become
accepted practice to write the indices that associate with material coordi-
nates in uppercase, while the indices that associate with spatial coordinates
xxxii Introduction
are written in lowercase. So, for example, material vectors are written as
A = AI eI , while spatial vectors look like a = ai ei , where, in both cases,
the repeated indices I and i are summed from 1 to 3 in accordance with Ein-
P
steins summation convention, e.g., a = a1 e1 + a2 e2 + a3 e3 = 3i=1 ai ei
is written in the shorthand notation of Einstein as a = ai ei , where the sum
over i is implicitly implied. Tensors, on the other hand, would be expressed
in material coordinates as A = AIJ eI eJ and in spatial coordinates like
a = aij ei e j , where denotes a dyadic product. Components are type-
set in an upright san serif font. Components of Greek fields are typeset
in an upright Greek font, for example, = i ei or = IJ eI eJ .
By this notation one means that the components, say AI of A, are evalu-
ated in a coordinate system specified by base vectors eI . Throughout this
text it is considered that ei and eI are coincident base vectors, specifically,
ei = I eI = Ii eI and eI = I ei = Ii ei , where Ii and Ii denote the Kronecker
delta function: 1 if I = i, otherwise 0.
Cartesian tensors are used throughout this text, as the metric of ambient
space is g = ij e i e j . Ambient space is said to be flat. Cartesian tensors
have three possible indical positions. Using indices c and r to denote col-
umn and row, the matrix notation of a tensor, say a, includes arc , arc , and
arc , with its transpose aT having matrix components of reversed order, viz.,
acr , acr , and acr . Vectors can have components of ar and ar with transposed
components of ac and ac .
Additional points on notation include the following: whenever the iden-
tity tensor has mixed indices, viz., one is a material index and the other is
a spatial index, then it is displayed as either I = Ii ei e I or I = Ii eI e i ;
otherwise, it is displayed as either I = ij e i e j , I = ij ei e j , or as
I = ij ei e j . In Cartesian tensor analysis, which is the form of tensor
analysis used in this text, the contravariant base vectors ei eI are coaxial
with the covariant base vectors e i e I , which is not true of general tensor
analysis, e.g., cf. Holzapfel (2000) and Sokolnikoff (1964).
The notational feature of using subscripts and superscripts to desig-
nate between covariant and contravariant component indices, respectively,
is adopted; it is used in general tensor analysis.1 Covariant/contravariant
properties of a tensor field are plain when written in the Lagrangian frame;
1 The author did not use subscripts and superscripts as a means to distinguish
between covariant and contravariant components in the earlier drafts of his text,
as is common practice among texts that use Cartesian tensors. It was a request
from his students that this notational enhancement be incorporated.
Introduction xxxiii
Kinematics
X x @ t
@ t0
e2
0
e1
e3
Fig. 1.1 Body B is deformed from its original shape at time t0 , i.e., from cong-
uration 0 , into a nal deformed shape at time t, viz., into conguration . Par-
ticle P was located at X = (X1 , X2 , X3 ) in 0 and is now located at x = (x1 , x2 , x3 )
in , quantied in a common Cartesian basis (e1 , e2 , e3 ) in R3
1.1 Motion
P , t) = @(X , t) = @ (X , t) eI ,
I
V (X , t) = x(X
P , t) = (X (1.6)
@t @t
@2 (X , t) @2 I (X , t)
A(X , t) = x(X
R , t) = (X
R , t) = = eI (1.7)
@t2 @t2
where V (X , t) and A(X , t) are material descriptions for the velocity and
acceleration of particle P . Notations P and R mean @ /@t and @2 /@t2 ,
respectively, for any field . Velocity and acceleration are contravariant
vector fields, i.e., they are physical fields described by tangent vectors to
curves of trajectory in ambient space; cf. Appendix B.
Deriving the spatial descriptions for the velocity and acceleration of a
particle as it moves through space R3 requires a bit more care; specifically,
@ (x, t), t
v(x, t) = V (X , t), (1.8)
@t
@2 (x, t), t Dv(x, t)
a(x, t) = =
@t 2 Dt
@v(x, t) @v(x, t) @x
= + (1.9)
@t @x @t
= vP (x, t) + grad v(x, t) v(x, t) A(X , t)
where Dv/Dt is called the material derivative of the velocity field. This
time derivative follows the motion of a material particle whenever the
operand is expressed as a spatial field. The acceleration vector a follows
from an application of the chain rule. The first term, i.e., vP = @v/@t, de-
scribes the local acceleration of particle P , while the second term, viz.,
grad(v) v = (@v/@x) v = (@v i /@x j )v j ei , describes the convective acceler-
ation being experienced by particle P ; cf. Appendix B.
Kinematics 9
1.3 Examples
Throughout this book, seven BVPs are examined that can be used to per-
form experiments for the purpose of material characterization. Each chap-
ter will build upon the prior ones by incrementally extending analysis to
include the topics covered in the most recent chapter, thereby increasing
your understanding of these important BVPs as we progress.
The four cases of simple extension, equi-biaxial extension, simple
shear, and homogeneous planar membranes will be worked out as exam-
ples in the text, while the three cases of pure shear, biaxial extension, and
axial extension followed by simple shear are left as exercises for you to
develop your skills. None of these BVPs require curvilinear tensor anal-
ysis. These important but more challenging BVPs are relegated to high-
er-level courses; e.g., cf. the textbooks of Bird et al. (1987a), Bird et al.
(1987b), Lodge (1974), Malvern (1969), Nicholson (2008), Truesdell and
Noll (2004), and Truesdell and Toupin (1960).
The targeted materials for application in this book include natural and
synthetic rubbers, elastomers, soft polymers, and soft biological tissues.
Collectively, they are referred to as soft solids. Although some materials
belonging to this class are anisotropic, a large percentage are mechanically
isotropic (or nearly isotropic) in that their stiness does not depend upon
direction even though their microstructures may be highly organized. An
assumption of material isotropy is imposed throughout this introductory
text. Anisotropy is left for advanced graduate study.
Another assumption frequently imposed throughout this book is that
the materials of interest are (or are nearly) incompressible. Conceptually,
any material whose bulk modulus greatly exceeds its shear modulus (typ-
ically by a factor of one hundred or more, like most soft solids, except,
e.g., foams and lung parenchyma) can be treated as being incompressible
in a mathematical sense even though no material is truly incompressible
in a physical sense. Consequently, for the most part, motions considered
10 Soft Solids
where (t) = `(t)/`0 is the stretch, `0 = `(t0 ) is the gage length, and `(t)
is the current length of extension; therefore, stretch is normalized so that
(t0 ) = 1. This shear-free motion is a special case of the deformation
illustrated in Fig. 1.2, which is redrawn in Fig. 1.3.
Kinematics 11
l1 (t)
l1 (t0)
l3 (t0)
l2 (t)
e2 e2
e1 e1
l2 (t0)
e3 e3
l3 (t)
t0
t
Fig. 1.2 Shear-free motions (Lodge 1974) take an elemental cube and deform it
into a rectangular prism. Such deformations are quantied by their three prin-
ciple stretches: 1 = `1 (t)/`1 (t0 ), 2 = `2 (t)/`2 (t0 ), and 3 = `3 (t)/`3 (t0 ). These
stretches associate with an isochoric deformation, i.e., volume preserving, when-
ever 1 2 3 = 1. Lengths `1 (t0 ), `2 (t0 ), and `3 (t0 ) are the gage lengths in their
respective 1-, 2-, and 3-directions
The velocity and acceleration vectors for this motion are determined
to be
P 1
X
P 2 ,
{V } {v} = 12 3/2 X (1.12)
1 3/2 P
X3 2
R 1
X
{A} {a} = 1 3/2 3 1
P 2 R X2 (1.13)
1 3/2 3 1
2 2
P 2 R X3
2 2
where the notation {v} means components v i of vector v are being as-
signed in a coordinate basis of ei , i = 1, 2, 3, as drawn in Fig. 1.3.
w
d
w0
d0 l
l0
t0 e1 t
e3
e2
Fig. 1.3 The uniaxial extension of a cube into a rectangular prism whose gage
section in the 1-direction stretches from a length of `0 to `, with w and d denoting
width and depth. p Assigning thepprincipal stretch as = `/`0 , constancy of volume
requires w = w0 / and d = d0 /
implying that velocity and acceleration obey
P 1
X
{V } {v} = P 2
X , (1.16)
2 X3
3 P
R 1
X
R 2
X
{A} {a} = (1.17)
2 3 P 2 R X3
3 1
w
d
w0
d0
l
l0
t0 e1 t
e3
e2
Fig. 1.4 The equi-biaxial extension of a cube into a rectangular prism whose
gage section in the 12 plane stretches from area `0 w0 to area `w such that
1 = `/`0 and 2 = w/w0 where, from constancy of volume, d = d0 / 2
shear, the eigenvectors of the deformation rotate in the body because of the
deformation (Lodge 1964), whereas in the previous two experiments, or
any other shear-free extension, these eigenvectors remain aligned over the
deformation history (Lodge 1974). This is one reason why the simple-shear
experiment is so important, yet it is seldom performed on solids.
Simple shear has a planar shearing motion, i.e., v3 = 0, described by
x1 = X 1 + X 2 , x2 = X 2 , x3 = X 3 (1.18)
whose inverse motion is given by
X 1 = x1 x2 , X 2 = x2 , X 3 = x3 (1.19)
with (t) being the magnitude of shear, as seen in Fig. 1.5, where /2 is
often referred to as the shear strain, initialized so that (t0 ) = 0. This
motion has a velocity and acceleration of
P X
2
{V } {v} = 0 , (1.20)
R X
0
2
{A} {a} = 0 . (1.21)
0
Simple shear is planar, viz., v3 = 0. This deformation is isochoric indepen-
dent of whether the material being sheared is incompressible or not.
14 Soft Solids
h
e2 e2
e1 e1
t0 t
Fig. 1.5 Simple shear takes a square and deforms it into a parallelogram of the
same area by shearing surfaces X2 = constant in the 1-direction by an amount
x1 = X2 , relative to the bottom surface X2 = 0 where x1 = X1 8 t
2 Inthe literature, e.g., Sacks (2000) and Freed et al. (2010), 1 2 is commonly
written as 1 and 2 1 is written as 2 . Choosing the description that we did here
means that 1 and 2 retain their physical interpretation of being magnitudes
of shear that is otherwise lost with the choice of 1 and 2 , which embed 2 and
1 within them. Mathematically, nothing is wrong with choosing 1 and 2 . It is
in their physical interpretation that confusion can arise.
Kinematics 15
( 1 + 1 2 , 2 + 2 1)
2
(1 2 , 2 )
e2
( 1 , 2 1)
e1 1
The inverse of this motion is gotten by inverting the system of equations
2 3
x1 1 1 2 0 X1
x2 = 4 2 1 2 0 5 X 2 , (1.24)
x3 0 0
1 X 3
yielding [cf. Eqs. (A.53)(A.56)]
2 3
X1 2 1 2 0 x1
1
X2 = 42 1 1 0 5 x2 (1.25)
X3 0 0
2 x3
P
The motion of a planar membrane has a velocity and acceleration of
1 X1 + 1P 2 X2
{V } {v} = 2P 1 X1 + P 2 X2 , (1.27)
2
X
P 3
16 Soft Solids
R 1 X1 + 1R 2 X2
{A} {a} = 2R 1 X1 + R 2 X2 (1.28)
2 2
1
P 2
R X3
for that particle P whose original location was at place (X1 , X2 , X3 ). Ob-
viously, the motion of a planar membrane is not planar in the sense that
v3 0; it is planar in the sense of being flat. By 1P 2 , e.g., we mean
P1 2 + 1 P 2 , i.e., the product rule of dierentiation is applied.
also be written in matrix form as
x1
2
1 +
1 X 2 1 2 +
1 X 1 0
3 X
1
x2 = 42 1 +
2 X2 2 +
2 X1 0 5 X2 (1.31)
x3 0 0 1 X3
where inhomogeneity manifests itself through the presence of X1 and X2
in the matrix term.
Kinematics 17
( 1 + 1 2 + 1 ,
2 + 2 1 + 2 )
2 ( 1 2 , 2 )
( 1 + 1 2 ,
2 + 2 1 )
e2
( 1, 2 1 )
e1 1
w0
d
w0
d0
l
l0
t0 e1 t
e3
e2
Fig. 1.8 The pure-shear extension of a rectangular prism into a dierent rect-
angular prism whose gage section in the 12 plane stretches from an area of `0 w0
into an area of `w0 such that 1 = `/`0 and 2 = 1, while d = d0 / from the
constancy of volume
1.4 Exercises
In this section, and in like sections throughout this book, you will find
problems that your instructor may choose to assign as homework problems.
18 Soft Solids
Shear Strain
ln 0 ln
Normal
Strain
Fig. 1.9 Mohrs circle in true strain for the pure-shear experiment
1.4.1.2 Problems
Another way to imagine pure shear is to rotate a square by 45 so that
the coordinate axes go through the four corners of the square. Now, pull
on a pair of opposing corners while allowing the two adjacent corners to
contract by an amount that preserves area. Obviously, this coordinate frame
does not rotate over the motion. In what plane of the motion described by
Eq. (1.32), as drawn in Fig. 1.8, does this visual image apply?
Derive the components of the velocity V and acceleration A vectors
that describe pure shear.
w
d
w0
d0
l
l0
t0 e1 t
e3
e2
Fig. 1.10 The biaxial extension of a cube into a rectangular prism whose gage
section in the 12 plane stretches from area `0 w0 to area `w with 1 = `/`0 and
2 = w/w0 where, from constancy of volume, d = d0 /( 1 2 )
1.4.2.1 Problem
Derive the components of the velocity V and acceleration A vectors that
describe biaxial extension. Show that they reduce to the velocity and ac-
celeration vectors of equi-biaxial extension given in Eqs. (1.16) and (1.17)
for the special case where 1 = 2 .
w w
d d
l
w0
d0
l l
l0
t0 t1 t
e1
e3
e2
In the motion displayed in Fig. 1.11, stretch varies over the time interval
[t0 , t1 ] and is held constant thereafter, while shear is held fixed at 0 over
the time interval [t0 , t1 ] and then varies thereafter, i.e., for all t t1 . The
parameter n accounts for the aspect ratio of the sample being tested, in that
(
1/2 whenever height/width 1,
n= (1.38)
0 whenever height/width 1
where the specimens height aligns with the 1-direction and its width aligns
with the 2-direction, as drawn in Fig. 1.11. Reality lies somewhere in
the interval 0
n
1/2 where n = 0 depicts pure shear in the sense of
Eqs. (1.32) and (1.33), while n = 1/2 depicts uniaxial extension in the sense
of Eqs. (1.10) and (1.11).
1.4.3.1 Problem
Derive the components of the velocity V and acceleration A vectors that
describe the second stage of this experiment, viz., the shearing. How
do they compare with those of the simple-shear experiment given in
Eqs. (1.20) and (1.21)?
22 Soft Solids
Deformation
The velocity and acceleration vectors derived in the preceding chapter are
important kinematic fields, but, in and of themselves, they are not capable
of describing how a body B deforms; they only describe how any particle P
within B moves through ambient space R3 . In order to study deformation,
one needs to quantify the change in shape of a body B as it is transformed
from some initial configuration 0 into its final configuration over some
interval [t0 , t] in time, which is the topic of this chapter.
Motion is described by a position vector, but not deformation. How-
ever, the dierence between two position vectors associated with a pair of
neighboring particles is such a measure. Through the relative motions of
such dierences, deformations can be quantified. Consider two particles P
and P 0 that are neighbors to one another in body B. Let the incremental
displacement vector connecting particle P to particle P 0 in configuration
0 be denoted by dX = X 0 X and let the displacement vector that con-
nects these same two particles in the current configuration be denoted by
dx = x 0 x, as shown in Fig. 2.1. Vectors dX and dx point from material
particle P to material particle P 0 in configurations 0 and , respectively.
Applying the chain rule to the law of continuous media given on p. 6
produces a linear transformation or mapping between dX = dXI eI in 0
and dx = dx i ei in that is expressed as
@i @i
dx i = dXI or dx i = FIi dXI with FIi = (2.1)
@XI @XI
dx
dX
X x @t
@ t0
e2 x
X
0
e1
e3
Fig. 2.1 Body B is deformed from its original shape at time t0 , i.e., from con-
guration 0 , into a nal deformed shape at time t, viz., into conguration .
Neighboring particles P and P 0 were at positions X and X 0 in 0 and are now
located at x and x 0 in . The incremental vector dX = X 0 X maps into vector
dx = x 0 x according to Eq. (2.1), whose return mapping maps dx back into
dX via Eq. (2.3). All elds are quantied against a common Cartesian basis
(e1 , e2 , e3 ) in R3
@I
[F1 ]Ii = = f Ii . (2.9)
@x i
So one observes that F has components FIi , while its inverse F 1 has com-
ponents [F1 ]Ii . The notation [F1 ]Ii does not mean 1/FIi . The superscript
(or row) index of F 1 associates with the material frame, while the sub-
script (or column) index belongs to the spatial frame, which is the opposite
index pairing present in F.
Regarding deformation gradients, the superscript index associates with
the argument in the numerator (the dependent variable of the motion map),
while the subscript index associates with the argument in the denominator
(the independent variable of the motion map). Whenever the independent
variable is X (i.e., Lagrangian), then FIi are the components of F; likewise,
whenever the independent variable is x (viz., Eulerian), then [F1 ]Ii are the
components of F 1 .
In what follows, F 1 will be used in place of f , as they are two expres-
sions of the same mapping. When describing fluids, f is preferred over F
because one places oneself in . When characterizing solids, which is the
focus of this text, F is preferred over f because one places oneself in 0 .
Tensor F is a two-state field, i.e., it has one index in the material configura-
tion 0 , while the other index resides in the spatial configuration . This
is because the deformation gradient F is a transformation mapping. Matrix
[F] maps a tangent vector {dX } from the reference configuration 0 into
{dx}, which is this tangent vectors representation in the current config-
uration , according to Eq. (2.2). Physically, these are the same tangent
vectors, but, mathematically, they are distinct.
Similarly, matrix [F 1 ]T maps the normal vector {N } = {dS(X )/dX }
to some material surface S(X ) from a reference configuration 0 into a
normal vector {n} = {dS( (x, t))/dx} that resides within the current con-
figuration . Applying the chain rule to S( (x, t)), incorporating the law
of continuous media, allows one to write
dS dS @I
= or ni = [F1 ]Ii NI . (2.10)
dx i dXI @x i
which, in accordance with Appendix B, maps the Lagrangian vector dS/dX
into its Eulerian vector dS/dx; specifically, it obeys the field-transfer oper-
ator of a covariant vector field described in Eq. (B.13) in that
( )
dS (x, t) dS(X ) d (x, t) dS(X ) h 1 i
= = F (2.11)
dx dX dx dX
or, equivalently,
{n} = {N } [F 1 ] so {n(x, t)} = [F 1 (X , t)]T {N (X )}. (2.12)
The dual mapping matrices of [F] and [F 1 ]T are the kinematic corner-
stones of continuum mechanics; cf. Appendix B. The notation for the de-
formation gradient F is made special, viz., it is typeset in an upright font
instead of a slanted font like other tensor fields, precisely because the field-
transfer operators [F] and [F 1 ]T map vector and tensor fields between the
two configurations of a deformation marked by the end points of a motion
over some interval in time, say [t0 , t].
Vector fields that are pushed forward from 0 into via [F], in accor-
dance with Eq. (2.2), can be pulled back from into 0 with the reverse
mapping [F 1 ]. Such vectors are called contravariant vector fields. Like-
wise, vector fields that are pushed forward from 0 into via [F 1 ]T , in
accordance with Eq. (2.12), can be pulled back from into 0 with the re-
verse mapping [F]T . Such vectors are called covariant vector fields (Lodge
28 Soft Solids
1974; Marsden and Hughes 1983; Sokolniko 1964). These mappings are
discussed in more detail in Appendix B, wherein Figs. B.2 and B.3 are use-
ful illustrations showing how these mappings apply to the transfer of field
from one configuration into another.
Coming to an understanding of the concepts that are outlined in Ap-
pendix B, viz., configuration physics, is essential before you can come to a
physical understanding of the mechanics that materials incur during finite
deformations. Please study Appendix B before advancing.
R
U
F @ t
@ t0
R v
Fig. 2.2 Body B is deformed from its original shape at time t0 , i.e., from con-
guration 0 , into a nal deformed shape at time t, viz., into conguration .
Considering the eld transfer of a contravariant vector illustrated in Fig. B.2, the
deformation gradient F can be decomposed in one of two ways: F = RU applies
a Lagrangian stretch of U from 0 that is followed by a rotation of R into ,
whereas F = vR rotates out of 0 via R after which an Eulerian stretch of v
places B into . The intermediate congurations (drawn as dashed ellipses) are
not physically realizable unless the deformation is uniform
Taking the material derivative of Eq. (2.2), viz., taking the time derivative
of the transfer of field {dx} = [F] {dX } at a fixed particle P and, therefore,
at fixed material coordinates X so that dX P = 0, one determines from the
product rule that
P = [FP ] {dX } = [FP ] [F 1 ] [dx] = [l] {dx}
{dx} (2.18)
where {dX } = [F 1 ] {dx} follows from Eqs. (2.4) and (2.8). This expres-
sion defines the Eulerian velocity gradient tensor1
where
P 1
l = FF has spatial components ` ij = FP Ii [F1 ]Ij , (2.20)
which plays the analogous role in rate-based theories that F plays in
deformation-based theories. Like the deformation gradient F, the veloc-
ity gradient l is not symmetric. Unlike F, the inverse of l does not exist.
It is useful to decompose the velocity gradient, not as a product as in
the polar decomposition of F, but rather as a sum, viz.,
l = d +w (2.21)
wherein
d = 12 l + l T has components d ij = 12 ` ij + jk ` k` `i , (2.22)
w = 12 l l T has components w ji = 12 ` ij jk ` k` `i (2.23)
where d (x, t) is the symmetric stretching tensor (also referred to in the lit-
erature as the rate-of-deformation tensor or the strain-rate tensor) whose
components come from Eq. (A.27), while w(x, t) is the skew-symmetric
vorticity tensor (also referred to in the literature as the rate-of-rotation ten-
sor or the spin tensor) whose components come from Eq. (A.29).
wherein
D = 12 L + LT with DIJ = 12 LIJ + JK LLK LI , (2.27)
W = 12 L LT with WJI = 12 LIJ JK LLK LI (2.28)
where D and W are the respective Lagrangian stretching and vorticity
tensors. It follows that [l T ] = [F] [LT ] [F 1 ] and [LT ] = [F 1 ] [l T ] [F]
and therefore
[d] = [F] [D] [F 1 ] and [D] = [F 1 ] [d ] [F], (2.29)
1 1
[w] = [F] [W ] [F ] and [W ] = [F ] [w] [F]. (2.30)
The polar decomposition theorem tells us that the rotation R and stretch
tensors U and v exist, but it does not tell us how to compute them. That is
the topic of this section. Here we consider interfacing with, e.g., an updated
Deformation 33
Lagrangian finite element code where the velocity gradient l is the inde-
pendent kinematic variable [cf. Belytschko et al. (2000)]. The algorithms
that follow employ a two-step AdamsBashforth predictor followed by a
trapezoidal corrector for numerical integration,2 which is in keeping with
the target application of an algorithm that resides within a finite element
code.
2.6.2 Rotation
The issue with applying a conventional numerical technique to integrate
RP = R [from Eq. (2.33)] for updating the rotation tensor R, without
addressing the intrinsic character of R, is an unavoidable degradation in the
orthogonality of R with increasing numbers of integration steps n taken, no
matter how fine a step size t is used (Atluri and Cazzani 1995; Flanagan
and Taylor 1987). The reason for this degradation is that the sum of two
orthogonal tensors is not orthogonal. Algorithm 2.2 is free from this defect.
It multiplies rotations, thereby maintaining orthogonality. The product of
two orthogonal tensors is an orthogonal tensor, within numerical roundo
error.
Because the rotation tensor R arising from a polar decomposition of the
deformation F is orthogonal, it can be described in an alternative form of
R = eQ so Rr = r with [Q] ij = ikj rk , krk = 1, (2.38)
where Q is a skew-symmetric tensor representation for the vector that is
the axis r about which rotation R occurs, with r being the only real-valued
eigenvector of R, and with corresponding to the angle of this rotation.
From the eigenvector property of Rr = r, one determines that
R = I + sin( ) Q + (1 cos ) Q2 , (2.39)
while from the evolution of R described by R = R one arrives at
P
! = P r + sin( ) rP + (1 cos ) r r,
P (2.40)
Deformation 35
which are Eqs. (2.16) and (8.25) of Atluri and Cazzani (1995), neither be-
ing simple to derive. Equation (2.39) is the matrix representation for a set
of formul originally derived by Euler in 1775 (Cheng and Gupta 1989).
Following a suggestion by Dienes (2003), Eq. (2.40) can be rewritten
as the matrix equation
A rP = ! P r so that rP = A 1 (! P r) (2.41)
procedure Rotation (t, !n , var Rt , var R+t , var Rn , var Rn+1 )
P k 12 (!n + !n1 )k
if |P | > tol then
F run corrector over previous increment
r 1 (! + !
n n1 )/P
2
[Q ] ij ikj {r }k
Rt I + sin(P t) Q + 1 cos(P t) Q Q
else
Rt I
end if
Rn Rt Rn1
P + k 12 (3 !n !n1 )k
if |P + | > tol then F run predictor over next increment
r+ 1
(3 ! n ! n1 )/P+
2
[Q+ ] ij ikj {r+ }k
R+t I + sin(P + t) Q+ + 1 cos(P + t) Q+ Q+
else
R+t I
end if
Rn+1 R+t Rn
!n1 !n F update stored variables
Rn1 Rn
end procedure
Deformation 37
Algorithm 2.2 requires, as input, the step size of integration t and the
axis of angular velocity at the beginning of the step !n . The algorithm
returns a corrected estimate for the rotation at the beginning of the step
Rn and a predicted estimate for the rotation at the end of the step Rn+1 .
It also provides estimates for the incremental rotations Rt and R+t oc-
curring over the prior [tn t, tn ] and next [tn , tn + t] integration steps,
respectively. Rotation Rt is evaluated according to the integration rule
of the corrector, while R+t is evaluated according to the integration rule
of the predictor. These two incremental rotations find application when
numerically integrating Eq. (2.34) for stretch.
2.6.3 Stretch
The left stretch tensor v is governed by dierential equation (2.34), whose
solution can be acquired via, e.g., Algorithm 2.3. Stretch v is required in-
put for computing the axis vector for angular velocity ! using Algorithm
2.1. All numerical integrations taking place in Algorithm 2.3 are done
in the updated Lagrangian configuration n aliated with time tn in ac-
cordance with Appendix B. Stretches belonging to the prior configuration
n1 are pushed forward into n via the map Rt , with the final result
then being pushed forward from the updated Lagrangian frame n into the
next Eulerian frame n+1 using the map R+t . It follows from Eq. (2.13)
that v = RU R1 , recalling that R1 = RT , which makes our algorithm for
integrating stretch rate fundamentally dierent from, say, an algorithm for
integrating stress rate.
Algorithm 2.3 requires, as input, the step size of integration t , the an-
gular velocity vector at the beginning of the step !n , the velocity gradient
at the beginning of the step ln , the incremental rotations over the previous
Rt and next R+t integration steps, and the left stretch tensor at the begin-
ning of the step vn . The algorithm returns a corrected estimate for vn and
a predicted estimate for vn+1 for use in the next calling of Algorithm 2.1.
A dierent approach for quantifying the fundamental deformation
fields of stretch and rotation, which is based upon the spectral decompo-
sition theorem with the deformation gradient being the known kinematic
variable, can be found in Simo and Hughes (1998, pp. 241244).
38 Soft Solids
2.7 Examples
A fair number of fields can be used to describe deformation, each having its
own purpose. The examples below will determine the deformation gradient
F and its inverse F 1 ; its associated polar fields U , v, and R; and from F,
P
the aliated rate fields of L, l, d , D, w, and W as they apply to the BVPs
studied in this text.
become
2P 3
/ 0 0
[L] [l] [D] [d ] = 4 0 /2
P 0 5. (2.47)
0 P
0 /2
No vorticity occurs because w = l d = 0; likewise, W = L D = 0. As
a check, det F = 1 and tr L = 0, so the prescribed deformation is isochoric.
become
2P 3
/ 0 0
4 P
[L] [l] [D] [d ] = 0 / 0 5, (2.51)
0 P
0 2 /
implying that no vorticity occurs, i.e., w W = 0. As a check, det F = 1
and tr L = 0, so this deformation is isochoric, too.
t0 t
e2
e1
Fig. 2.3 Simple shear takes a square of dimension h (for height) and deforms it
into a quadrilateral of the same area by a magnitude and angle of shearing
from which the two stretch tensors are determined to have components
2 3
2 0
1 4 2 + 2 5
[U ] = 0 (2.56)
(4 + )
2 1/2
0 0 (4 + ) 2 1/2
and
2 3
2 + 2 0
1 4 2 5
[v] = 0 (2.57)
(4 + 2 ) 1/2
0 0 (4 + )
2 1/2
P 2
which follow straightaway from Eqs. (1.24) and (1.25), while recalling that
the areal stretch of Eq. (1.23) is given by
= 1 2 (11 2 ) and, therefore,
P = P 1 P .
1 2 1 2 1 2 1 2 (2.63)
The motion maps defined in Eqs. (1.22) and (1.26) are isochoric because
det F det F 1 = 1 due to how F33 is defined.
Following the same line of reasoning that was used to derive the rota-
tion and stretch tensors for simple shear, Freed et al. (2010) arrived at like
values applicable for planar membranes, which, in the notation of this text,
lead to polar descriptions where
2 3
1 4 1 + 2 1 2 2 1 0
[R] = (1 2 2 1 ) 1 + 2 0 5 (2.64)
0 0
Deformation 43
wherein
q
2 2
= 1 + 2 + 1 2 2 1 (2.65)
normalizes the rigid-body rotation, i.e., it ensures that det R = 1. This
expression for the rotation R allows the right stretch tensor U = RT F to
be written as
2
( + 2 ) 2 1 (1 2 2 1 )
1 4 1 1
[U ] = 1 2 (1 + 2 )
0
3
1 2 (1 + 2 ) 0
2 ( 1 + 2 ) + 1 2 (1 2 2 1 ) 0 5 , (2.66)
0
1
while the left stretch tensor v = FRT has components
2
( + 2 ) + 1 2 (1 2 2 1 )
1 4 1 1
[v] = 1 22 + 2 21
0
3
1 22 + 2 21 0
2 ( 1 + 2 ) 2 1 (1 2 2 1 ) 0 5 (2.67)
0
1
where [U ] and [v] are similar, yet distinct.
The components of the velocity gradients are easily acquired, too, being
[L] = F 1 FP
2 3
2 P 1 1 2 2P 1 2 1P 2 1 2 P 2 0
1 6 7
= 4 1 2P 1 2 1 P 1 1 P 2 2 1 1P 2 0 5 (2.68)
0 0
P
and
[l] = FP F 1
2 3
2 P 1 2 1 1P 2 1 1P 2 1 2 P 1 0
1 6 7
= 4 2 2P 1 2 1 P 2 1 P 2 1 2 2P 1 0 5 (2.69)
0 0
P
whose components are obviously distinct in this case. A little eort, along
with Eq. (2.63), leads to a verification of the isochoric constraints, viz.,
tr l = tr L = 0. The symmetric part of l, i.e., the stretching, is given by
44 Soft Solids
2
2 P 1 2 1 1P 2
1 6
[d] = 4 1 P + P P P
2 1 1 2 2 2 1 1 2 1 2 1 2
0
3
1 P + P P P
0
2 1 1 2 2 2 1 1 2 1 2 1 2
7
P P
1 2 1 2 2 1 0 5 , (2.70)
0
P
2
0
1 4 1 P P
[w] = 1 1 2 + 2 2 1 + 1 2 P 1 2 1 P 2
2
0
1 P P 3
P
1 2 1 + 2 1 2 0
P
2 1 1 2 2 2 1
0 0 5 . (2.71)
0 0
2.8 Exercises
Strain
Strain has been defined and quantified a number of different ways over the
past two centuries. Essentially, strain is a measure of the change in shape
of a localized region in a body B that has been deformed from its reference
configuration 0 (where strain is typically normalized to be zero) into its
current configuration . Strain is a two-state property; it depends upon 0
and .
Unlike stretch, which is uniquely defined, strain is not unique. In fact,
Hill (1968) has shown that one can choose from an infinity of admissible
strains. Nevertheless, only a few have physical significance and are prac-
tical to work with. All measures of strain are equivalent to the infinitesi-
mal strain tensor of linear elasticity (the strain measure you were taught in
your introductory strength of materials course) whenever the magnitudes
of strain are infinitesimal (kk . 1 %, typically, cf. Eq. 3.22). Their dier-
ences become noticeable at larger strains (say, kk & 4 %).
Rule of Thumb: Linear strain is in error about 1 1/2 % per percent strain.
So how does one choose a suitable strain measure for soft materials
where strains routinely exceed 10 % and frequently exceed 100 %? That is
the topic of this chapter.
3.1 Deformation
[b1 ] ij = [F1 ]Ii IJ [F1 ]Jj = [v1 ]ki [R1 ]Ik IJ [R1 ]J` [v1 ]`j
= [v1 ]ki k` [v1 ]`j (3.9)
4 This strain measure is rarely found in the literature, but when it is, it is often
credited to Piola, although in no document of his (that this author has found) does
Piola dene such a strain eld. I credit this strain measure to one of my mentors,
the late Prof. Arthur S. Lodge, not because this strain measure likely originated
with him, but because he was apparently the rst to study it; specically, its
geometric interpretation originates in his textbook entitled Elastic Liquids (Lodge
1964, Chap. 2). It is also the strain measure that arises in the rubberlike liquid,
which is a viscoelastic material model that he derived from molecular statistical
mechanics (Lodge 1956, 1958).
52 Soft Solids
Consider two neighboring particles P and P 0 in some body B that are con-
nected by an infinitesimal vector dX in the reference configuration 0 and
by another infinitesimal vector dx in the current configuration , as de-
picted in Fig. 2.1. Let the ambient space be flat so that its metric g is
the identity tensor I (Sokolniko 1964). Then, in accordance with Rie-
mannian geometry, the lengths of these two vectors are described by their
Euclidean norms
and
(ds)2 = dx I dx
= [F] {dX } I [F] {dX }
(3.25)
= {dX } [F]T [I] [F] {dX } = {dX } [F T F] {dX }
= dX C dX = dXI CIJ dXJ ,
j
recalling that F T F = FIi ij FJ e I e J and that dS = dS(X ) and ds =
ds(x, t). Here, Eq. (2.2) was used to convert dx into dX in the second line
of the expression. An examination of formula (ds)2 = dX C dX justifies
referring to C as the Lagrangian metric of deformation.
By contracting the covariant strain tensor of Green (1841) defined in
Eq. (3.11) with the contravariant unit vector dX /dS from both the left and
the right, one arrives at the following quadratic form
dH dx
dX dh ds
dS
t0 t
Fig. 3.1 Material surfaces S and S0 , containing particles P and P 0 , deform from
an initial conguration at time t0 into a nal conguration at time t. The initial
separation of these surfaces dH deforms to dh in a manner that is distinct from
the separation of the particles going from dS to ds as dX goes to dx
dY 2 dY(X ) dY(X ) dY IJ dY
= I = (3.27)
dH dX dX dXI dXJ
dY(X ) dY
dY = dX = dXI , (3.28)
dX dXI
while the incremental height dh that currently separates these same two
material surfaces, now located in the spatial configuration , is given by
Strain 55
dY 2 @Y (x, t) @Y (x, t)
= I
dh @x @x
dY(X ) @ (x, t) dY(X ) @ (x, t)
= I
dX @x dX @x
" #
dY(X ) @ (x, t) @ (x, t) T dY(X )
= I
dX @x @x dX
dY(X ) dY(X ) (3.29)
= [F 1 ] I [F 1 ]T
dX dX
dY(X ) 1 1 T dY(X )
= [F ] [I] [F ]
dX dX
dY(X ) dY(X )
= [F 1 F T ]
dX dX
dY(X ) dY(X ) dY 1 IJ dY
= C 1 = [C ]
dX dX dXI dXJ
these two particles reside, as illustrated in Fig. 3.1. Given that Y is chosen
so that 0 < dH = dS in 0 , it follows from the triangle inequality that
0 < dh
ds in . Greens strain tensor maps as a covariant field, while
Lodges strain tensor maps as a contravariant field. They are each viable
strain measures, both physically and mathematically.
Theorem 3.1. The elements of area suering extremal changes are normal
to the principal directions of strain, and the greatest (least) change of area
occurs in the plane normal to the axis of least (greatest) stretch; in fact,
if the principal stretches dx/dX satisfy 1 2 3 the corresponding
ratios da/dA satisfy 2 3
3 1
1 2 .
Before deriving the various strain rates, it is useful to state the material
derivative of the deformation gradient in its various forms; specifically,
they include
FP T = F T l T P P
FP = lF F 1 = F 1 l F T = l T F T
(3.35)
= FL, = LT F T , = LF 1 , = F T LT
where the first two came from rearranging Eqs. (2.19) and (2.24), while the
last two came from dierentiating the identity equation FF 1 = I. These
four identities are used in the following derivations of strain rate.
The material derivative of Greens covariant description for strain E =
1 (F T F I), when expressed in terms of the Eulerian velocity gradient l,
2
is determined to be
EP (X , t) = 12 FP T F + F T FP ,
[EP (X , t)] = 12 [F]T [l T ] [F] + [F]T [l] [F] (3.36)
T1 T T
= [F] 2 [l + l] [F] = [F] [d ] [F].
From Eq. (B.43), the Lie derivative for Almansi strain is given by
O O
[ e(x, t)] = [F 1 ]T [EP (X , t)] [F 1 ] with e(x, t) d . (3.37)
O
From Eq. (B.54) we have e = De/Dt + l Te + el so that
O @e ij k
e ij = eP ij + v + ` ki e kj + e ik ` kj = [F1 ]Ii EP IJ [F1 ]Jj = d ij (3.38)
@xk
where here d = d ij e i e j , i.e., it takes on covariant components.
58 Soft Solids
7 This result proves that the integral equation describing strain derived by the
author (Freed 2010) is an alternative expression of the Signorini (1930) strain
measure, as the derivative of the integral equation reported therein reproduces
the result given in Eq. (3.40).
Strain 59
P , t) = 1 FP1 F T + F 1 F PT
E(X 2
= 12 LF 1 F T F 1 F T LT (3.44)
= 12 LC 1 + C 1 LT
Like above, we point out that EP D. These strain rates are symmetric
constructions based upon the Lagrangian velocity gradient L where the
deformation metrics of Green C and Cauchy C 1 are required so that the
resulting rates for EP and EP map between 0 and as covariant and con-
travariant fields, respectively, just as their associated strains do.
From Eqs. (3.42) and (3.44), it is readily shown that
EP = C 1EP C 1 P ;
or EP = C EC (3.46)
consequently, just one physical measure for strain rate exists, it being EP
when expressed as a contravariant field or EP when expressed as a covariant
field. Notably, although EP = C 1EP C 1 , E C 1EC 1 . This will have
ramifications later when constitutive equations are considered that can be
based on either strain or strain rate.
60 Soft Solids
3.4.1 Integration
In applications, one may need to integrate strain rate in order to obtain
strain. In accordance with Eqs. (3.37) and (B.79), Almansi strain e follows
from the integral equation
Z t
1 T
[e(x, t)] = [F ] [EP (X , t0 )] dt0 [F 1 ]
0
Z t
= [F 1 ]T [F((X , t0 ))]T [d ((X , t0 ), t0 )] [F((X , t0 ))] dt0 [F 1 ],
0
(3.47)
while, from Eqs. (3.40) and (B.78), Signorini strain e follows from the
integral equation (Freed 2010)
Z t
P , t0 )] dt0 [F]T
[e(x, t)] = [F] [E(X
0
Z t
= [F] [F 1 ((X , t0 ))] [d ((X , t0 ), t0 )] [F 1 ((X , t0 ))]T dt0 [F]T ,
0
(3.48)
which are equivalent expressions to their definitions given in Eqs. (3.10)
and (3.13); however, these integral equations are not as easy to work with.
The previous integrals specify how the Eulerian strains e and e could
be solved for by integration. Simpler to evaluate would be the integrals that
return the Lagrangian strain measures, viz.,
Z
1 t T
E= L (X , t0 ) C (X , t0 ) + C (X , t0 ) L(X , t0 ) dt0 ,
2 0
Z (3.49)
1 t 0 1 0 1 0 T 0
0
E= L(X , t ) C (X , t ) + C (X , t ) L (X , t ) dt
2 0
although, in most applications, it would be far easier to just compute them
directly via Eqs. (3.11) and (3.14).
No initial conditions are associated with the above integrals for strain
because strain is typically normalized to be zero in the reference frame,
i.e., at time t = 0; hence, their ICs are typically set to 0, which is tacitly
assumed in the above formul.
The results of this section are not as useful, per se, from an application
point of view, as they are essential from a theoretical point of view. They
bring the theory for strain full circle with respect to the calculus.
Strain 61
Numerical stability issues often arise with material models whenever they
are incompressible or nearly incompressible (Simo and Hughes 1998).
Thus, a variety of variational principles for FE analysis have been de-
veloped (Belytschko et al. 2000; Bonet and Wood 1997) whose various
techniques revolve around the splitting of strain into dilatoric (pertaining
to volume change) and deviatoric (pertaining to shape change) parts. The
Lagrangian frame is selected for constructing our theory for strain. Algo-
rithms are provided to compute the covariant and contravariant distortions
from which distortional strain measures are constructed.
In accordance with the conservation of mass, cf. Sect. 2.2, dilation J
is defined as the ratio of volumes belonging to an infinitesimal volume
element evaluated at two instances in time, specifically
dv
J= = det F so that d ln J = tr(F 1 dF) (3.50)
dV
where J is the Jacobian of the field-transfer map, cf. Appendix B, hence its
notation. Dilatation , as put forward by Hencky (1928), is defined as
= 13 ln(det F) = 13 ln J with d = 13 tr(F 1 dF) (3.51)
where the rate relationship results from Eq. (A.84). Stretch = d`/dL,
areal stretch
= da/dA, and dilation J = dv/dV are all normalized ra-
tios of geometric objects constructed from incremental extents in length
of line, area of surface, and volume of space, respectively, whose values
in the current configuration are proportioned by their respective values
from the initial or reference configuration 0 . The notion that the loga-
rithm of dilation (a volumetric stretch) describes dilatation (a volumetric
strain) was developed by Hencky (1928) from an idea that he credits to
Ludwik (Hencky 1931) in Henckys modeling of the high-pressure data of
Bridgman (1923).
In accordance with the above definitions, one can readily write down
dEx = dE C d 1
1
with tr C dEx = 0 (3.55)
= dE 6 C d ln(det C )
because from Eq. (3.11), i.e., from E = 12 (C I), comes dE = 12 dC with
the deviatoric, Green, strain rate dEx = 12 dCx following by analogy.
For those constitutive theories that select {, Ex } instead of {E } as
their set of independent state variables, one is required to be able to com-
pute the distortional Green strain Ex described by
Z t
x
E =E C (t0 ) d(t0 )
0
(3.56)
= E 12 C Cx
= 1 Cx I ,
2
where the second line follows from the first as a consequence of the first
line in Eq. (3.54), with the third line following from the definition of
Green strain in Eq. (3.11). Like Cx , Ex is not deviatoric in the sense that
tr(C 1 Ex ) 0, even though its rate dEx is deviatoric, which is why Ex
64 Soft Solids
3.6 Examples
In the example problems of this text, the spatial gradient grad() = @ /@x
of any field is considered to be negligible, i.e., the motions considered
produce spatially homogeneous deformation fields. As such, the material
derivative D /Dt at particle P has no local convective part and reduces
to just its local contribution of @ /@t = P . Consequently, all convective
contributions belonging to field in the presence of an homogeneous de-
formation must come from handling the global transport of embedded
within the current configuration as it moves through space over a con-
tinuum in time; they enter via Oldroyds convective terms in his various
Lie derivatives, cf. Appendix B.
from which it follows that the Green E and Signorini e strain tensors pre-
sented in Eqs. (3.11) and (3.13) have like components of
2 2 3
1 1 0 0
[E ] [e] = 4 0 (1 )/ 0 5, (3.65)
2
0 0 (1 )/
while the Almansi e and Lodge E strain tensors have like components of
2 2 2 0
3
1 4( 1)/ 0
[e] [E] = 0 1 0 5, (3.66)
2
0 0 1
as defined in Eqs. (3.10) and (3.14).
68 Soft Solids
For their time rates of change, governed by Eqs. (3.36) and (3.39), using
the components of the stretching tensor d found in Eq. (2.47) for uniaxial
extension, one determines that the various strain fields evolve according to
2 P 3
2 0 0
1
EP [e] P = 4 0 / P 2 0 5 (3.67)
2 P
0 0 / 2
and
2 P 3 3
1 2 / 0 0
[Pe] EP = 4 0 P 0 5 . (3.68)
2
0 0 P
The Lie derivatives of these Eulerian strain measures are equivalent be-
cause of Eqs. (3.37) and (3.40) with
2P 3
/ 0 0
O M
[ e] [e] [d ] = 4 0 /2
P 0 5 (3.69)
0 P
0 /2
in accordance with Eq. (2.47).
Obviously, regarding how strain is processed as a response by a given
material, uniaxial extension experiments cannot distinguish between con-
stitutive eects whose origins stem from strain being either covariant or
contravariant in nature. Another experiment is needed for that purpose.
from which it follows that the Green E and Signorini e strain tensors put
forward in Eqs. (3.11) and (3.13) have like components of
2 2 3
1 0 0
14 5,
[E ] [e] = 0 2 1 0 (3.71)
2
0 0 (1 )/
4 4
Strain 69
while the Almansi e and Lodge E strain tensors have like components of
2 2 3
( 1)/ 2 0 0
14
[e] [E] = 0 ( 2 1)/ 2 0 5 , (3.72)
2
0 0 1 4
whose dierences are subtle, but nonetheless real. The inverse of the Green
deformation tensor [which is the measure for deformation actually used by
Cauchy (1827)] is contravariant and has components
2 2 (1 + 2 ) ( + ) 0 3
1 2 1 1 2 1 2
1
C = 2 4 1 2 (1 + 2 ) 21 (1 + 22 ) 0 5, (3.85)
0 0
4
0 0
4
72 Soft Solids
wherein
= 1 2 (1 1 2 ) is the areal stretch defined in Eq. (1.23).
From these deformation fields, the covariant strain tensor of Green, E =
1 (C I), defined in the material frame , has components
2 0
2 2 3
1 (1 + 22 ) 1 1 2 (1 + 2 ) 0
1
[E ] = 4 1 2 (1 + 2 ) 22 (1 + 12 ) 1 0 5, (3.87)
2
0 0
1
2
2 4
Recall that the covariant strains E and e are measures of separation be-
tween neighboring material particles, while the contravariant strains E and
e are measures of separation between neighboring material surfaces, or,
equivalently, because the deformation is isochoric, E and e are measures
of their change in area. One measure for each type is given in the mate-
rial configuration 0 , and one measure for each type is given in the spatial
configuration .
The time rates of change of these strain fields become a bit messy and,
therefore, are not written out here in component form. Even so, they are not
dicult to derive. The Lie derivatives of the Eulerian strain measures obey
O M
[ e] [e] [d] because of Eqs. (3.37) and (3.40), whose components are
listed in Eq. (2.70).
Strain 73
3.7 Exercises
Stress
The concept of stress traces back nearly two centuries to the published
works of Cauchy (1827). Cauchy generalized Eulers concept of pres-
sure and the hydrodynamic laws that Euler derived some 70 years ear-
lier. Cauchy made the notion of stress precise. He surmised that a
body responds to externally applied loads by transmitting forces internally
throughout the body via a matrix valued field that now bears his name:
Cauchy stress. Not only did Cauchy develop the concept of stress, but he
also derived the physical conservation laws that apply to stress. In doing
so, he generalized Eulers theory for an inviscid fluid.
Once again, let us consider a body B whose reference configuration 0
is affiliated with an initial time t0 . Over the course of time, body B is sub-
jected to externally applied forces that cause it to deform into its current
configuration , which is affiliated with present time t. In Fig. 4.1, a parti-
cle P residing on a material surface S belonging to body B is investigated
by (fictitiously) cleaving the body along S while simultaneously applying
a resultant force distribution over surface S sufficient to keep the cleaved
body in equilibrium with its surroundings, as if it were whole.
Consider a distributed resultant force df (t, da) acting at time t over
some infinitesimal area da that surrounds particle P and belongs to surface
S in . Cauchy (1827) postulated that this distributed force of infinitesimal
extent is equivalent to a point force of traction t applied at the particle P ,
scaled by the dierential area da(x, t) over which df acts; this is Cauchys
postulate
df = t da or df i = t i da where t = t(n ; x, t), (4.1)
n t
N T
da
dA
0
e2
t0 e1 t
e3
Fig. 4.1 Body B is deformed from its original shape at time t0 , i.e., from cong-
uration 0 , into a nal deformed shape at time t, viz., into conguration . This
sets up a traction vector t at particle P acting on a surface S over an innitesimal
area da with unit normal n. All elds are quantied in a common Cartesian basis
(e1 , e2 , e3 ) in R3 . Vector T is the pseudo traction vector dened on 0 , often
referred to as the nominal traction
wherein n(x, t) is the unit normal to surface da, cf. Eq. (2.11), while the
traction vector t has physical units of force per unit deformed area.
Cauchys fundamental theorem for stress establishes stress as a linear
operator on that maps a normal to a surface into a vector of traction
acting on that surface, all of these fields being located at particle P , i.e.,
t(n ; x, t) = T (x, t) n(x, t) (4.2)
or in component notation
t i = T ij n j , (4.3)
where T is the Cauchy (or true) stress tensor1 which, like the traction
vector t, has units of force per unit deformed area.
It can be readily argued that stress maps as a contravariant tensor field
by examining Eqs. (4.2) and (4.3). Normal n maps as a covariant vector
field because of Eq. (2.11). Traction t behaves like a tangent to a curve, so
its index maps like a contravariant vector, as in Eq. (2.2). Consequently,
1 Cauchy stress is expressed in an uppercase font, as if it were a Lagrangian eld,
but it is not; Cauchy stress is an Eulerian eld. The notation T is adopted for
historical reasons. T is the commonly accepted notation for Cauchy stress when
written in a roman font [cf. Truesdell and Noll (2004)]. The engineering stress
that associates with the innitesimal strain of Eq. (3.22) is typically denoted
as , which appears in the linear theory of elasticity.
Stress 79
the conservation of indices from tensor analysis requires that the indices
of stress T ij must map in a contravariant manner with the jth contravariant
index of T ij contracting with the covariant index of n j , leaving the ith
contravariant index of T ij to associate with the contravariant index of t i
on the left-hand side of the equation. But the field-transfer properties of T
are a bit more complicated than this. Cauchy stress also has a field-transfer
property known as weight, which is discussed in more detail in Sect. 4.4.
Another stress tensor defined over is often selected for use. In fact, it is
the stress tensor used in this text when working in the spatial configuration
, viz., the Kirchho (1852) stress tensor defined by
s(x, t) = det(F) T or s ij = det(F) T ij , (4.4)
which has units of force per unit undeformed area (as if it was quantified
in the reference configuration 0 ), but the tensor field s, itself, resides in
the deformed configuration .
Whenever materials are incompressible, as they are assumed to be in
this text, for the most part, then the stress tensors of Cauchy and Kirchho
become numerically equivalent because det F = 1 follows from the iso-
choric constraint. Throughout this text, the Kirchho stress s will be the
stress field used most often when working in the Eulerian frame.
[F1 ]Ii
n i da = det(F) [F1 ]Ii NI dA or dai = dAI , (4.14)
det F 1
with like terms appearing in the Truesdell strain E y of Eq. (3.33). Field-
transfer mappings that depend upon the determinant of the transforma-
tion operator, also known as the Jacobian of the transformation, as occurs
here, produce what are called weighted fields (Lodge 1974; Oldroyd 1950;
Sokolniko 1964). Equation (4.13) contains the only vector-valued field
transfer with weight addressed in this text, cf. Appendix B.
From the above results, one is led to the first PiolaKirchho stress ten-
sor P (Kirchho 1852; Piola 1833). As with Cauchys theorem, P is a lin-
ear operator that maps one vector into another, specifically, a normal vec-
tor into a traction vector, but here the mapping takes place in a Lagrangian
frame instead of the Eulerian one used by Cauchy, viz.,
T (N ; X , t) = P (X , t) N (X ), (4.15)
with components
T i = P iI NI . (4.16)
We see that, like the deformation gradient F(X , t), stress P (X , t) has one
index in the reference configuration 0 and the other in the current config-
uration , i.e., it, too, is a transformation map, which is why it is generally
easier to quantify via experiment than, say, the Kirchho stress s. This is
why it is typeset in an upright font, like F.
The first PiolaKirchho stress relates to the Kirchho stress via the
formula
j
s = PFT or s ij = P iI FI ) P F T = FP T , (4.17)
where the latter relationship follows from the conservation of angular mo-
mentum, viz., Eq. (4.9); consequently, P is not a symmetric field (neither
are F, L, or R), although its elements may be symmetric from time to time.
Stress 83
Because the Kirchho stress s and second PiolaKirchho stress S are the
same physical measure of stress, although defined in dierent frames, and
because they are contravariant tensor fields, it necessarily follows that their
84 Soft Solids
time rates of change are related according to the maps given in Eqs. (B.39)
and (B.40), in other words, via
M
[ s(x, t)] = [F] [SP (X , t)] [F]T
M
or [SP (X , t)] = [F 1 ] [ s(x, t)] [F 1 ]T , (4.21)
M
where s(x, t) = Ds/Dt ls sl T is a Lie derivative for stress, a.k.a. Ol-
droyds (1950; 1970) upper-convected stress rate.
where an integration can take place over any closed cycle in strain. A
perpetual motion machine is any hypothetical contraption that violates this
integral inequality, which, by its very definition, cannot occur in Nature.
Stability and uniqueness of a solution are important mathematical prop-
erties that lie beyond the confines of thermodynamics or physics in general.
A by-product of such a mathematical analysis leads to what is referred to as
a constitutive inequality. Drucker (1959) and Hill (1957, 1958) were pio-
neers in this field of study. Whenever S and E are taken to be the defining
conjugate pair, and the material is simple in the sense of Noll (1958, 1972),
then the associated mathematical criterion for uniqueness and stability of a
solution is straightforward to implement. It is Hills (1957; 1968) constitu-
tive inequality:
tr dS dE > 0 8 dE 0. (4.32)
The choice of a stress/strain conjugate pair is not unique (cf. Ogden (1984)
for a thorough treatment of this topic). The choice made here is the one
selected most often by mechanicians.
Convexity of potential surfaces in state space is another means for en-
suring mathematical stability of a solution. As this book was going to
press, Nicholson (2013) constructed a technique by which convexity could
be tested for a given potential function, previously, a daunting problem.
in its rate form, from the reference configuration 0 into the current con-
figuration , which leads to the inequality (Hill 1968):
M M
tr d = tr s d 2} tr d 2 > 0 8 d 0. (4.39)
This is the correct form of Hills constitutive inequality for one to use in
Eulerian formulations. This follows from trace manipulations of tr(P EP ) =
MO M
tr([F][P ][F]T [F 1 ]T [EP ][F 1 ]) = tr( e) = tr( d ) and from tr(SP EP ) +
} tr(CP 1 EP ) = tr([F][SP ][F]T [F 1 ]T [EP ][F 1 ]) + } tr([F][CP 1 ][F]T
MO
P
[F 1 ]T [EP ][F 1 ]) = tr( s e) 2} tr([F][E][F] T [F 1 ]T [EP ][F 1 ]) =
MO MO M
tr( s e) 2} tr(e e) = tr( s d ) 2} tr(d 2 ) where Eqs. (3.1), (3.4), (3.37),
(3.40), (4.18), (4.21), (4.28), (A.61), and (A.70) have been used.
Note: Incompressibility requires that tr d = 0. This, however, does not
imply that tr(d 2 ) is also zero; in fact, tr(d 2 ) 0 for isochoric motions.
4.9 Examples
The theories developed in this text, for the most part, have stress as the
dependent variable and strain as the independent variable. To be able to
compare theory with experiment requires that one be able to extract the
individual components of stress from the measured quantities in an exper-
iment: forces, moments, stretches, and/or rotations. Acquiring such rela-
tionships for the seven BVPs addressed in this book is the remaining topic
of this chapter.
In the prior chapters, you studied how to represent deformation and
were, therefore, consumed by how a body changes its shape. In this chap-
ter, the physical cause of such changes is investigated. Fortunately, ac-
counting for shape change in ones assessment of stress and its components
is not always necessary. Applying the definition for nominal traction, as it
pertains to the first PiolaKirchho stress established in Eqs. (4.10) and
(4.15), allows for an analysis of stress to take place in the reference config-
uration 0 where the boundary conditions and specimen dimensions are
presumed known and are in forms that are usually easier to work with.
After the first PiolaKirchho stress P has been quantified through exper-
imentally measured variables, P can be mapped into the Kirchho stress s
via Eq. (4.17) for use in Eulerian constructions, or into the second Piola
Kirchho stress S via Eq. (4.18) for use in Lagrangian constructions.
e2
A(t0)
e1 T (t)
e3
Fig. 4.2 The deformation of axial extension illustrated in Fig. 1.3 is caused by
a distributed force f acting over an initial area A(t0 ) whose normal N lies in the
1-direction in the reference conguration 0 . From Kirchhos expression (4.10)
of Cauchys postulate (4.1) a nominal traction vector T (t) = f (t)/A(t0 ) acts at the
centroid of area A(t0 ) and aligns with the 1-direction. This traction is opposed by
a clamped boundary condition acting along the shaded surface
T(t) f (t)/A(t )
df 0
{T }e1 = H) 0 = 0 , (4.43)
dA e1 0 0
e3
T2 (t)
e2
e1
A2 (t0)
A1 (t0)
T1 (t)
solved for the Kirchho stress, which in turn allows Eq. (4.18) to be solved
for the second PiolaKirchho stress, producing components
2 3 2 3
T 0 0 T/ 0 0
[s] = 4 0 0 05 and [S ] = 4 0 0 05 (4.48)
0 00 0 00
whose rates, from Eq. (4.21), are
2 3 2 3
P + T P 0 0
T P T P 0 0
T
M
sP = 4 0 0 05 so s =4 0 0 05 (4.49)
0 00 0 00
and
2 3
P 200
P T /
T/
SP = 4 0 0 05 (4.50)
0 00
all of which are distinct from one another.
such loadings two nominal traction vectors are created simultaneously with
components
T (t) f (t)/A (t )
df 1 1 1 0
{T }e1 = H) 0 = 0 (4.51)
dA e1 0 0
and
0
0
df
{T }e2 = H) T2 (t) = f2 (t)/A2 (t0 ) (4.52)
dA e2 0 0
with
{T }e3 = {0}, (4.53)
where f1 (t) and f2 (t) are two uniformly distributed forces imposed on
the specimen, which act on surfaces whose initial areas are A1 (t0 ) and
A2 (t0 ). Ideally, forces f1 and f2 are adjusted so that both tractions
T1 (t) = f1 (t)/A1 (t0 ) and T2 (t) = f2 (t)/A2 (t0 ) equal T(t), i.e., the stressed
loading is equi-biaxial with T1 = T2 = T at every moment t. The first
PiolaKirchho stress must resolve both of these external tractions and, as
such, obeys
T1
2
T00
3
1
{T }e1 = [P ]{N }e1 H) 0 = 4 0 T 05 0 (4.54)
0 0 00 0
and
2 0
T00
3
0
{T }e2 = [P ]{N }e2 H) T2 = 4 0 T 05 1 (4.55)
0 0 00 0
with
T00
0
3
0
2
4
{T }e3 = [P ]{N }e3 H) 0 = 0 T 0 5 0 , (4.56)
0 0 00 1
where tractions are now being applied in two directions, viz., N |e1 and
N |e2 .
With the first PiolaKirchho stress now known, assuming an isotropic
material, an application of the deformation gradient F quantified in
Eq. (2.48) allows Eq. (4.17) to be solved for the Kirchho stress which, in
94 Soft Solids
f2R
f2L
T2 t2
T1 t1
h
f1
h h
w f1 w
T1 t1
T2 e2
t2 f2L
0 f2R
e1
Fig. 4.4 The deformation of simple shear illustrated in Fig. 2.3, caused by an
applied force f1 , is imposed on a specimen of dimensions w (width) by h (height)
by d (depth, not shown) thereby producing a traction of t1 = f1 /wd. This sets up
reaction forces f2R and f2L producing a traction of t2 = (f2R + f2L )/wd. Tractions
t1 and t2 of the spatial conguration associate with pseudo tractions T1 and
T2 in material conguration 0
turn, allows Eq. (4.18) to be solved for the second PiolaKirchho stress,
thereby producing components
2 3 2 3
T 0 0 T/ 0 0
[s] = 4 0 T 05 and [S ] = 4 0 T/ 05 (4.57)
0 0 0 0 0 0
and
2 3
T/ P 2
P T / 0 0
P
S = 4 0 P P 2
T/ T / 05 . (4.59)
0 0 0
Here the assumption is that both the loading and the deformation are equi-
biaxial. A tacit requirement is that the material must be isotropic to be able
to fulfill this assumption.
Stress 95
where s12 = s21 from the symmetry of s. The Kirchho stress and first
PiolaKirchho stress relate to each other according to Eq. (4.17), from
which it follows that the first PiolaKirchho stress has components of
2 3 2 3
P11 P12 0 s11 s12 s12 0
[P ] = 4P21 P22 0 5 = 4s21 s22 s22 0 5 , (4.61)
0 0 P33 0 0 s33
where T1 (t) and T2 (t) are the two components of traction T (t) imposed on
a surface whose initial area is A(t0 ) = wd; they are the shear traction T1
96 Soft Solids
and the normal traction T2 . The normal N to this surface aligns with the
T1
2-direction. From Eqs. (4.15) and (4.61) it readily follows that
2
s11 s12 s12 0
3
0 s12
4
{T }e2 = T2 = s21 s22 s22 0 5 1 = s22 (4.63)
0 0 0 s33 0 0
and, therefore, s12 = T1 and s22 = T2 .
The gripping surfaces along the shear planes of X2 = 0 and X2 = h carry
internal tractions of T1in and T2in along material surfaces whose normals
align with the 1-direction in that
T
df 1in (t)
{T }e1 = H) T2in (t) , (4.64)
dA e1 0
so that
T 2
s11 s12 s12 0
3
1
1in
{T }e1 = T2in = 4s21 s22 s22 0 5 0
0 0
s 0 s33 0
s12
unknown
11
= s21 s22 = (f2R f2L )/hd , (4.65)
0 0
where the internally carried traction T2in = (f2R f2L )/hd follows from the
previously determined stress components of s12 = f1 /wd and s22 = (f2R +
f2L )/wd along with the expression guaranteeing equilibrium of moments,
viz., f1 (f2R + f2L ) = wh (f2R f2L ). In contrast, the internal traction T1in
cannot be resolved from these boundary conditions.
Finally, planes whose normals align with the 3-direction are considered
to be traction freethe plane-stress assumptionleading to
0
df
{T }e3 = H) 0 , (4.66)
dA e3 0
so that
0 2
s11 s12 s12 0
3
0
0
{T }e3 = 0 = 4s21 s22 s22 0 5 0 = 0 (4.67)
0 0 0 s33 1 s33
and, therefore, s33 = 0.
Stress 97
viz., s13 = s31 = s23 = s32 = 0, which is the case in Eq. (4.60). If the
plane-stress assumption applies, then one also has s33 = 0.5
From the deformation gradient that describes an isotropic,
homogeneous, planar membrane, viz., Eq. (2.61), and from the relation-
ship between the Kirchho and first PiolaKirchho stresses given in
Eq. (4.17), one determines its components as being
2 3
2 (s11 1 s12 ) 1 (s12 2 s11 ) 0
1
[P ] = 4 2 (s21 1 s22 ) 1 (s22 2 s21 ) 0 5 , (4.70)
0 0
s33
2
where 1 and 2 are the stretches, 1 and 2 are the in-plane shears, and
t2
T2
1
t1
1
2 T1
h
2
2 t1
T1 w
1
T2 t2
e2
0
e1
T f
and
2 sin 1 2 sin(1 )/wd
df
{T }e2 = H) T2 cos 1 = f2 cos(1 )/wd , (4.72)
dA e2 0 0
while normal to the surface of the membrane
0
df
{T }e3 = H) 0 , (4.73)
dA e3 0
where f1 (t) and f2 (t) are the two distributed forces imposed on the speci-
men causing tractions t1 and t2 , which act on surfaces whose initial areas
are A1 (t0 ) = hd and A2 (t0 ) = wd, cf. Fig. 4.5.
The first PiolaKirchho stress must resolve these two external trac-
tions, and, as such, from {T }e1 = [P ]{N }e1 , one gets
100 Soft Solids
T 2
2 (s11 1 s12 ) 1 (s12 2 s11 ) 0
3
1 cos 2 1
1
T1 sin 2 = 4 2 (s21 1 s22 ) 1 (s22 2 s21 ) 0 5 0
0
s 0
1 s12
0
2 s33 0
2 11
= s21 1 s22 (4.74)
0
and from {T }e2 = [P ]{N }e2 , one gets
T2 sin 1
2
(s 1 s12 ) 1 (s12 2 s11 ) 0
3
0
1 4 2 11
T2 cos 1 = 2 (s21 1 s22 ) 1 (s22 2 s21 ) 0 5 1
0 0
s 2 s11
0
2 s33 0
1 12
= s22 2 s21 (4.75)
0
0 0 0
2 s33 1
s T
whose solution is
2 3
11 1 0 1 0 1 1 cos 2
s21 60 1 0 1 7 1 T1 sin 2
=6
42
7 , (4.78)
s12 0 1 05 2 T2 sin 1
s22 0 2 0 1 2 T2 cos 1
where
= 1 2 (1 1 2 ) establishes the extent of isochoric areal stretch
defined in Eq. (1.23).
Stress 101
4.10 Exercises
6. How does the Biot stress Ty relate to the other stresses introduced in this
text? Biot stress Ty is the thermodynamic conjugate to Biot strain Ey
which is defined in Prob. 3.7.4(1). In other words, determine Ty such
that tr(Ty dEy ) = tr(S dE ).
PART 2
Constitutive Equations
In theory,
there is no difference between
theory and practice.
But in practice, there is.
Yogi Berra
Albert Einstein
Chapter 5
Explicit Elasticity
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 109
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2_5, Springer International Publishing Switzerland 2014
110 Soft Solids
5.1 Theory
and
dWs (Ex )
Sx = so that tr Sx C = 0. (5.6)
dEx
These quantify the hydrostatic and deviatoric contributions to stress in a
Green elastic solid through which the total state of stress can then be estab-
lished from Eq. (4.40), viz., S = p C 1 + Sx . Recall that the distortional
strain Ex need not, in general, be deviatoric, i.e., traceless, whereas the
deviatoric stress Sx must be traceless, by definition, hence the constraint
equation tr(Sx C ) = 0.
5.1.3 Properties
The Green and Lodge elastic solids are explicit constitutive equations in
the sense that their strain-energy functions W depend only upon strain;
they do not depend upon both strain and stress. Such models are discussed
in the next chapter.
114 Soft Solids
Green and Lodge elastic solids are isotropic theories because the
tensorial dependence of their strain-energy functions W depends only upon
the metric of deformation C . Recall that E = 12 (C I) and E = 12 (I C 1 ).
Green and Lodge elastic solids are objective in the presence of an arbi-
trary rotation Q whenever W (E ) = W (QE QT ) for the Green solid, and
whenever W (E) = W (QEQT ) for the Lodge solid, given that QQT =
QTQ = I (cf. Appendix B).
Invariant theory (Rivlin and Smith 1969; Spencer 1972) is called upon
to simplify the process of constructing viable energy functions. Invari-
ant theory is an outgrowth of the CayleyHamilton theorem, Eqs. (A.87)
and (A.88). It allows the tensorial arguments of an isotropic function to be
replaced with a set of scalars, called invariants, that are unique character-
istics of the tensorial arguments belonging to that function. A vector has
one invariant: its magnitude or length. A tensor in R2 has two invariants:
its trace and its determinant. While a tensor in R3 has three invariants:
its trace (the sum of its three 1 1 minor determinants), the sum of its
three 2 2 minor determinants, and its 3 3 determinant. The theory of
determinants plays a vital role in invariant theory.
and, therefore, these three scalar invariants are objective measures of strain
for any given tensorial state of Green strain E .
When substituted into Eq. (5.3), along with an application of the chain
rule, the strain-energy function W (E ) W (I1 , I2 , I3 ) leads to a general
constitutive theory of
where the coecients, W,1 = @W (I1 , I2 , I3 )/@I1 , etc., are scalar functions
of the invariants I1 , I2 , and I3 . This theory becomes a material model
whenever a strain-energy function W = W (I1 , I2 , I3 ) is assigned.
Each term in the above formula is contravariant and, therefore, this
expression is admissible in the sense of general tensor analysis. Because it
is contravariant, it pushes forward from 0 into according to Eq. (B.15),
leading to its Eulerian form
where s is the Kirchho stress from Eq. (4.4) and e is the Almansi strain
from Eq. (3.10). All sense of covariance and contravariance is lost in the
Eulerian version of this theory, which illustrates why the author derives his
theories in the Lagrangian frame where these qualities provide a framework
that assists in the overall construction of admissible tensor equations.
116 Soft Solids
1
Sx = 2Ws,2 Ex 1 tr EC
x C
3
1
+ 3Ws,3 EC x 1 tr EC
x E x EC
x C (5.28)
3
= 21 C 1EC 1 or = 21 e (5.29)
= 2(1 + 3 I2 ) C 1EC 1 ,
(5.30)
= 2(1 + 3 I2 ) e
where 5,2 and 5,3 are the fifth-order shear moduli associated with the
second and third invariants, respectively. Theory cannot discern between
5,2 and 5,3 , only experiments can do so.
Explicit Elasticity 121
M O
which follows from tr( d ) = 21 tr( e d ) = 21 tr(d 2 ) > 0 and from
M M
tr( d ) = 21 tr(e d ) = 21 tr(d 2 ) > 0 where Eqs. (3.37) and (3.40) have
been used. From this logic and the fact that I2 0, the third-order Green
and Lodge models will be stable if
1 > 0 and 3 > 0. (5.36)
In contrast, the fifth-order Green and Lodge elastic solids have the potential
to become unstable whenever tr(d 2 e) 0 for the Green solid or whenever
tr(d 2 e) 0 for the Lodge solid. Even so, their magnitudes will not likely
exceed those of the remaining positive valued terms so as to violate their
respective overall inequalities.
Joule (cf. James and Guth 1944; Treloar 1975). This is because rubbery
deformations tend to be entropic, whereas glassy deformations tend to be
energetic, in a thermodynamic sense.
equations for describing the bulk response. He used these models to ex-
plain the experimental data of Bridgman (1923).
In the presence of large dilatations, Hencky (1931) found it necessary to
introduce a limiting state of dilation in order to be able to describe Bridg-
mans data. Such limiting states would be better handled through the im-
plicit theory of elasticity presented in the next chapter. For our purposes
here, higher-order models for pressure are not usually required because di-
latations tend to be of infinitesimal extent in soft-solid applications, foams
and lung parenchyma withstanding (e.g., cf. Freed and Einstein 2013).
For the deviatoric contributions to stress, we construct first-, third-, and
fifth-order models replicating the approach followed for their incompress-
ible counterparts.
where and are the bulk and shear moduli, respectively. This formula re-
duces to the classical theory (Marsden and Hughes 1983) under conditions
of infinitesimal strain whereby kEk 1. Recall that the Lam constant of
classical elasticity is = 23 .
other than its tensorial character which governs how it maps between con-
figurations, viz., all four indices map contravariantly.
Material stability will exist, in the sense of Hill (1968), cf. Eq. (4.36),
whenever
tr d dE = dE W M W dE > 0 8 dE 0, (5.52)
i.e., whenever the tangent modulus M is positive definite. This result fol-
lows by contracting Eq. (5.51) with dE , which produces tr(dS dE ) =
tr(C 1 dE ) d} + dE W M W dE + 2} tr(dE dE ), where the constraint
tr(C1 dE ) = 0 follows because of the imposed isochoric assumption.
When this expression is substituted into Hills result (4.36), the outcome
is the inequality stated above. The tangent modulus M in Eq. (5.52) is in
the form of a Hessian. Nicholson (2013) has recently developed a tool that
can probe a region in state space to determine if such a Hessian is positive
definite, which is otherwise a dicult task to verify analytically.
Equations (5.51 and 5.52) provide a general theoretical framework that
can be readily coded into software, after which time the implementation of
a new explicit elastic solid of interest to an engineer would only require the
derivation of its respective tangent modulus. This is eectively how most
existing commercial FE packages operate. They have the capability of
allowing a user to supply their own material model by linking a subroutine
for a users tangent modulus with the vendors software package.
d2 W (E) 1
M = C 1 C 1 W W C C 1
dE dE !
dW (E) dW (E)
2 C 1 C 1 C 1 + C 1 C 1 C 1
dE dE
(5.57)
for any admissible strain-energy function W (E). Application of the chain
rule to arrive at this result requires the relationship dE/dE = C 1 C 1 ,
which comes from the identity dE = C 1 dE C 1 , i.e., from Eq. (3.46).
The minus two in the second line comes from dC 1 = C 1 dC C 1 =
2 C 1 dE C 1 with C 1 dE C 1 = (C 1 C 1 ) : dE .
MooneyRivlin Model: Mixing Eqs. (5.54) and (5.58) leads to the tangent
modulus of a MooneyRivlin (5.37) elastic solid, viz.,
M = 21 C 1 C 1 2(1 ) C 1 E + E C 1 . (5.59)
d2 Ws (Ex )
D= : I I 13 C C 1
x
dE dE x
(5.63)
dSx (Ex )
= : I I 13 C C 1
dEx
where the first term comes from expanding C 1 (dp/dE ) : dE and the
second term comes from expanding p (dC 1/dE ) : dE , while using consti-
tutive equation (5.5) to quantify p. Freed and Einstein (2013, Appendix B)
determined dp/dE = 13 (dp/d) C 1 and dC 1 /dE = 2 C 1 C 1 .
Equations (5.65 and 5.66), when combined with Eq. (5.62), constitute an
explicit Hookean elastic solid based upon Green strain E that extends the
classical theory of linear elasticity into the domain of finite deformations.
D = 2 1 + 3 (IN2 1 IN 2 ) C 1 C 1 13 C 1 C 1
3 1
+ 43 C 1Ex C 1 C 1Ex C 1
13 IN1 C 1 C 1Ex C 1 + C 1Ex C 1 C 1
+ 19 IN12 C 1 C 1 (5.67)
x
dSx (E)
= : C 1 C 1 13 C 1 C 1
dEx
1
2 C Sx + Sx C 1 , (5.69)
x
where the identity dE/dE = C 1 C 1 13 C 1 C 1 follows from
x
contracting dE/dE = I I 13 C 1 C with dE/dE = C 1 C 1 arising
from dE/dC 1 = 12 I I and dC 1 /dE = 2C 1 C 1 (cf. Freed
and Einstein 2013, Appendix B). The last line in the above formula, viz.,
134 Soft Solids
Equations (5.71 and 5.72), when combined with Eq. (5.62), constitute an
explicit Hookean elastic solid based upon Lodge strain E that extends the
classical theory of linear elasticity into the domain of finite deformations. It
is the compressible neo-Hookean solid and is similar to, yet distinct from,
the formulation based upon Green strain E listed in Eqs. (5.65) and (5.66).
D = 2 1 + 3 IN2 1 IN2 + IN 2 2 IN 2 IN + 1 IN 4
3 1 5 2 3 1 2 9 1
C 1 C 1 13 C 1 C 1
2 C 1 E x C 1 + 2 IN1 C 1 C 1
x +E
3
+ 4 3 + 25 IN2 13 IN12
E x Ex 1 IN1 C 1 E
x +Ex C 1 + 1 IN2 C 1 C 1 . (5.74)
3 9 1
5.5 Examples
All of the examples considered in this text have a Finger deformation tensor
b that belongs to a class of deformations described by
2 3
b11 b12 0
[b] = 4b21 b22 0 5 (5.75)
0 0 b33
with b21 = b12 from its symmetry. As has already been pointed out on
page 97, the Kirchho stress s and the Finger deformation b commute
for all explicit elastic solids, i.e., [b][s] = [s][b]; cf. Moon and Truesdell
(1974). Therefore, as a direct consequence of the assumed deformations
considered herein, viz., Eq. (5.75), it follows that the Kirchho stress can,
at most, have nonzero components of
2 3
s11 s12 0
[s] = 4s21 s22 0 5 (5.76)
0 0 s33
where s21 = s12 from its symmetry, which follows from the conservation
of angular momentum. From [b][s] = [s][b] one arrives at the constraint
(s11 s22 ) b12 = (b11 b22 ) s12 (5.77)
which is satisfied for all explicit elastic solids whose deformation and stress
are special cases of Eqs. (5.75) and (5.76). Hence, this constraint equa-
tion applies to all motions discussed in this text. It is trivially satisfied for
shear-free motions. Its influence is only felt during shearing motions.
An imposition of plane stress, i.e., s33 = 0, independent of b33 , is an-
other constraint (beyond [s][b] = [b][s]) that will be imposed.
136 Soft Solids
Only BVPs for the incompressible explicit elastic solids of Green and
Lodge are solved below. Solutions for their compressible versions are left
as exercises for you to learn from.
} = 1 ( 4 1) (5.96)
and, as such, the component quantifying nominal traction (determined
from the 11- or 22-component equations) is given by
T = 1 3 1/ 3 (5.97)
where T is the traction, is the stretch, and 1 is the shear modulus.
Taking the derivative of T with respect to and evaluating it at =
1 gives dT/d | =1 = 61 , so equi-biaxial extension produces an initial
response with twice the stiness of uniaxial extension, which agrees with
the classical theory of elasticity. The large stretch response of this material
grows with the cube in stretch.
} = 1 (1 1/ 4 ) (5.98)
and a nominal traction of
T = 1 1/ 5 . (5.99)
The derivative of T with respect to at = 1 gives dT/d | =1 = 61 . This
is a well-established experimental observation (Treloar 1975). At large
stretches 1 the tangent modulus becomes dT/d | 1 = 1 , the same
as in uniaxial extension.
140 Soft Solids
} = (1 + 3 I2 )( 4 1) (5.102)
and a nominal traction of
T = (1 + 3 I2 ) 3 1/ 3 (5.103)
whose second invariant has the value
I2 = 14 2( 2 1)2 / 4 + (1 4 )2 . (5.104)
The derivative of T with respect to at = 1 gives dT/d | =1 = 61 ,
independent of 3 .
} = (1 + 3 I2 )(1 1/ 4 ) (5.105)
and a nominal traction of
T = (1 + 3 I2 ) 1/ 5 (5.106)
whose second invariant has the value
I2 = 14 2( 2 1)2 + (1 4 )2 / 8 . (5.107)
The derivative of T with respect to at = 1 gives dT/d | =1 = 61 ,
independent of 3 .
Explicit Elasticity 141
whose second invariant is given in Eq. (5.104) and whose third invariant
has the value
I3 = 18 2( 2 1)3 / 6 + (1 4 )3 . (5.110)
whose second invariant is given in Eq. (5.107) and whose third invariant
has the value
I3 = 18 2( 2 1)3 + (1 4 )3 / 12 . (5.113)
and, therefore,
T1 = 1 and T2 = 1 2 = T1 , (5.117)
which predicts a linear response in shear with a compressive quadratic re-
sponse normal to the shear plane.
Lodge Material, a.k.a. the Neo-Hookean Solid: From the strain compo-
nents in Eq. (3.79) and the stress components of Eq. (4.60), the response
of this material (5.32) in simple shear is
2
s11 s12
= 1 (5.118)
s21 s22 0
and, therefore,
T1 = 1 and T2 = 0, (5.119)
which predicts a linear response in shear but no normal response in the
2-direction.
MooneyRivlin Solid: From the strain components in Eqs. (3.78) and
(3.79) and the stress components of Eq. (4.60), the response of this ma-
terial (5.38) in simple shear is
2
P11 P12 0 0
= 1 (5.120)
P21 P22 0 3 2
and, therefore,
T1 = 1 and T2 = T1 . (5.121)
Lodge Material: From the strain components in Eq. (3.79) and the stress
components of Eq. (4.60), the response of this material (5.33) in simple
shear is
s11 s12 2
= 1 + 3 I2 (5.125)
s21 s22 0
whose second invariant has the value
I2 = 14 2 2 + 2 (5.126)
and, therefore,
T1 = 1 + 3 I2 and T2 = 0. (5.127)
whose second invariant is given in Eq. (5.126) and whose third invariant
has the value
I3 = 18 4 3 + 2 (5.132)
and, therefore,
T1 = 1 + 3 I2 + 5,2 I22 + 34 5,3 I3 3 ,
(5.133)
T2 = 34 5,3 I3 2
where we now have a Lodge model that will impose a tensile force in the
normal direction.
5.6 Applications
Two distinct theories of elasticity were derived in the early pages of this
chapter, the Green and Lodge elastic solids, so named because of the strain
measures they employ. These two theories associate with dierent pairs of
thermodynamic conjugate variables. In the pages that follow, first-, third-,
and fifth-order models were constructed for each theory so you can com-
pare their similarities and contrast their dierences. Models were devel-
oped for three types of construction: general constructions, incompress-
ible constructions, and isotropic/deviatoric constructions. With all of these
models, which represent just a sampling, how does one go about selecting
an appropriate model for use? Coming to grips with this question is the
objective of the remaining pages in this chapter.
Other stress/strain pairs are admissible, too, e.g., theories based on the
Biot (1939) and Hencky (1931) strain measures and their thermodynamic
stresses have been derived (Ogden 1984), but are not discussed here. There
is also the excellent and popular elastic model of Ogden (1972), cast in
terms of principal values of stretch and stress, that is not addressed in this
text either. Principal formulations, such as Ogdens, are useful in the ap-
proach of explicit elasticity where the stress response is independent of
Explicit Elasticity 147
Fig. 5.1 A reconstruction of Figs. 5.4 and 5.6 from Treloars (1975) book, which
are experimental plots for natural rubber in uniaxial tension and compression
with the latter being constructed from an equivalent two-dimensional extension
experiment (cf. Fig. 5.5 in his book). Treloars raw data are listed in Table 5.1
at the end of the chapter. Correlations of these data with the rst-order Lodge
(5.32) and Green (5.29) elastic solids for 1 = 355 kPa are displayed as curves
deformation path, but they are probably not very useful for material the-
ories where stress depends upon the path of deformation, which includes
implicit elasticity an viscoelasticity.4 It is for this reason, and in keeping
with the intended scope of the book, that such theories are not addressed
here. The interested reader is referred to the texts of Holzapfel (2000) and
Ogden (1984).
The classic textbook written by Treloar (1975) remains a rich resource
for understanding rubbers and elastomers, both experimentally and theoret-
ically. The experimental data presented in his text have been re-digitized
by numerous authors, including this one; they are cataloged in Table 5.1 at
the end of the chapter.
Investigation begins by comparing the incompressible first-order elastic
solids of Green (5.29) and Lodge (5.32) in Fig. 5.1 against the experimental
4 Private communication with R. W. Ogden, 2013.
148 Soft Solids
Fig. 5.2 Treloars (1975) data for natural rubber from Fig. 5.1 are redisplayed
with curves being plotted for the rst- (1 = 355 kPa), third- (1 = 355 kPa,
3 = 390 Pa), and fth- (1 = 355 kPa, 5,3 = 0.48 Pa, 3 = 5,2 = 0) order elastic
Lodge solids
iii. If this ratio lies somewhere between 0 and 1/3 , then select the
MooneyRivlin elastic solid.
(b) If the response curve softens and later hardens, the material model
will be of higher order. To select such a model:
i. Compute a ratio of tangent moduli between the plateau of in-
flection and the initial states and apply step (a) to select the
appropriate strain measure to adopt.
ii. Consider a third-order model first. If it describes the data, select
it. If its influence turns on too soon, as was the case for natu-
ral rubber, then select a fifth-order model for a more delayed
response. If it describes the data, select it.
(3) If the response curve stiens, then select an implicit elastic solid from
the next chapter.
5.7 Exercises
the first- (5.32), third- (5.33), and fifth- (5.34) order neo-Hookean solids.
Show that they reduce to their equi-biaxial cases whenever T1 = T2 .
1 @2 W (I1 , I2 , I3 ) @2 W (I1 , I2 , I3 )
MIJKL = +
4 @EIJ @EKL @EIJ @ELK
!
@2 W (I1 , I2 , I3 ) @2 W (I1 , I2 , I3 )
+ +
@EJI @EKL @EJI @ELK
Fig. 5.3 A reconstruction of Fig. 11.1 from Treloars (1975) book, which is an
experimental plot for natural rubber in equi-biaxial extension. These data are
listed in Table 5.1
(11) Estimate the parameters for the neo-Hookean elastic solid for the
biological elastomer known as elastin using the uniaxial fiber data
presented in Table 5.3 and shown in Fig. 5.6. Plot your correlation
against these data. Discuss the influence that variability in these data
has on the overall fitting process. Does evidence exist to suggest using
the Hookean, MooneyRivlin, or a higher-order Lodge elastic solid
over the neo-Hookean solid based upon this data set alone? Justify
your position.
(12) Derive the experimental tractions T1 and T2 for the third- and
fifth-order Lodge elastic membranes.
154 Soft Solids
1.5
Traction, T1 (MPa)
0.5
0
1 2 3 4 5
Stretch,
Fig. 5.4 A reconstruction of Fig. 5.8 from Treloar (1975), which is an experi-
mental plot for natural rubber in pure shear. These data are listed in Table 5.1
Explicit Elasticity 155
Fig. 5.5 Uniaxial extension data for the silicon elastomer PDMS. These data
are listed in Table 5.2
156 Soft Solids
Elastin Fibers
16
12
Traction, T (MPa)
0
1 1.5 2 2.5 3
Stretch,
T T
kPa kPa
1.00587 59 1.67200 4,588
1.00786 353 1.68839 5,765
1.03007 235 1.70499 4,824
1.04472 588 1.74282 4,824
1.06857 941 1.75980 6,000
1.07750 529 1.79423 5,412
1.09576 824 1.79861 6,471
1.10274 1,235 1.83362 5,706
1.11216 1,000 1.88264 6,824
1.13153 882 1.92793 6,412
1.14139 1,294 1.93710 6,235
1.16676 1,353 1.95543 5,941
1.17196 1,765 1.96926 6,765
1.22884 1,882 1.98786 7,824
1.22884 2,353 2.03474 7,824
1.26919 2,118 2.06322 6,765
1.32087 3,059 2.10144 8,647
1.33979 2,647 2.14005 9,647
1.34939 2,941 2.19378 7,882
1.36893 2,647 2.19866 8,765
1.39890 3,176 2.22819 10,059
1.42622 3,882 2.29299 10,353
1.44712 3,471 2.33829 9,235
1.47564 4,059 2.33319 11,588
1.49008 3,176 2.44524 12,118
1.51951 4,176 2.45043 10,353
1.54963 5,059 2.50204 13,412
1.56486 3,765 2.50204 15,765
1.56874 4,588 2.52282 12,294
1.61161 5,059 2.58012 13,412
1.65569 4,118 2.61161 14,765
1.65972 5,412
Chapter 6
Implicit Elasticity
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 161
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2_6, Springer International Publishing Switzerland 2014
162 Soft Solids
paths by which the material points reach the final state and not upon the
rate at which they traverse these paths.
Curiously, implicit elasticity, although expressed as a rate theory, has
a stress/strain response that does not depend upon time, yet it can depend
upon the path traveled. This allows us to express our theory in terms of
dierentials instead of derivatives.
Implicit elasticity, like explicit elasticity, provides a collection of elastic
material models. Whether a particular elastic material is better described
with an explicit- or implicit-elastic material model is a matter for experi-
ment to discern.
6.1 Motivation
What motivated the author (Freed 2008, 2009, 2010; Freed and Einstein
2012, 2013; Freed et al. 2010) to consider stress-rate/strain-rate elastic the-
ories over stress/strain elastic theories for describing the passive response
of soft, fibrous, biological tissues is the fact that, ever since Fung (1967)
published his pioneering work, mechanicians have known that the stress/
strain response of tissue is, to a good approximation, exponential. Even
single-molecule proteins like titin, the largest known protein in Nature (the
muscle protein), exhibit force-extension curves with an exponential quality
to them when they are pulled in the stage of an atomic force microscope
(Linke and Grtzner 2008).
Since Fungs original paper, tissue mechanicians have proposed nu-
merous ways to incorporate this exponential quality into their stress/
strain frameworks [see, e.g., the review articles of Humphrey (2002b,
2008), Sacks (2000), Sacks and Sun (2003), Viidik (1973), and Weiss and
Gardiner (2001); the textbooks of Fung (1993) and Humphrey (2002a);
and this authors works (Freed and Diethelm 2007; Freed et al. 2005)]. A
stress-rate/strain-rate theory subsumes this exponential quality through its
(numerical) integration, as the characteristic solution to a linear, first-order,
ordinary, dierential equation (ODE) is an exponential.
In his landmark paper, Fung (1967, Eq. 22) introduced an empirically
based formula for the one-dimensional passive response of soft elastic tis-
sues; it being the linear first-order ODE:
dT( )
= E + T( ) with IC T(1) = 0. (6.1)
d
This formula is known today as the Fungs law. It is a relationship between
traction T and stretch , not between stress and strain (see Eqs. 2.44 and
Implicit Elasticity 163
400
Tangent Modulus, dT/d (kPa)
300
200
100
0
0 10 20 30 40 50
Traction, T (Pa)
Fig. 6.1 A reconstruction of Fung (1967, Fig. 8), which is an experimental plot
for rabbit mesentery. A tting of these data to Fungs law, viz., Eq. (6.1), pro-
duces model parameters of E = 11.7 kPa and = 10.9 from a least squares regression
of the data, whose t is drawn as a line
2 The term (E/ + T)/(1 (1 2 )/2) in Eq. (6.5), which arises from thermo-
dynamic considerations, has a numerator that is akin to a Fung (1967) modulus
and a denominator that is akin to a Carton et al. (1962) compliance. The model
of Carton et al. predates Fungs model by 5 years. Their model supposes
innites-
imal strain to be described by the formula = max 1 eT/Emax , whose tangent
response is 1 dT/d = Emax /(max ), wherein max is the maximum allowable
strain.
Implicit Elasticity 165
d
+ 4 4 1 + 1 T, (6.15)
while Vitos model has a uniaxial stress response of
Implicit Elasticity 167
( 2 +2 1 3)+ 1 (2 + 2 3)
T = (1 )e
2 + 1 3 (6.16)
1
with a local tangent modulus of
dT ( 2 +2 1 3)+ 1 (2 + 2 3)
= (1 )e
d
4
3 3 + 4 1
1
2
2 2
+ 2 2 + 2 1 3 + 1 T. (6.17)
1
Because such elastic models are, at best, complex approximations to
Fungs otherwise simple law (6.1), an alternative approach for the mod-
eling of soft biological tissues was sought, cf. (Freed 2008, 2009, 2010;
Freed and Einstein 2012, 2013; Freed et al. 2010).
We take Einsteins counsel spoken of in the foreword: Seek simplicity!
6.2 Theory
that conforms with Fungs law, as interpreted by the theory just put for-
ward. This higher-order coupling with strain in the resilience @2 U/@S @S
is the thermodynamic origin of what this author calls the Rajagopal eect.
Material models derived from Eq. (6.20) are elastic. They do not dissi-
pate energy. They are not integrable. They are not hyperelastic in the sense
of Green (1841). Nor are they hypoelastic in the sense of Truesdell (1955).
Nevertheless, special cases can and do exist.
whose gradient with respect to deviatoric stress Sx , taken from the left, is
!
@2 Us (Ex , Sx ) @2 Us (Ex , Sx )
I I : dEx : dSx = 0, (6.30)
@Sx @Ex @Sx @Sx
with dEx being defined according to Eq. (3.55). This constitutive equation
governs the deviatoric stress response in our implicit elastic solid.
After the implicit constitutive equations for pressure (6.28) and devia-
toric stress (6.31) have been integrated, the total state of stress S can then
be quantified according to Eq. (4.40), viz., S = p C 1 + Sx .
6.3.4 Stability
Material stability, in the sense of Drucker (1959) and Hill (1968), is, in
principle, straightforward to discern; specifically, any implicit elastic solid
will be stable if
8
dE : M : dE > 0 8 dE 0,
<dS : C : dS > 0
8 dS 0,
or (6.56)
d :m :d >0 8 d 0,
:M M M
s :c : s >0 8 s 0,
depending upon which formulation one chooses to use. A like result ap-
plies for the explicit elastic solid, cf. Eq. (5.52). This is stability in a math-
ematical sense. It is not a requirement of any physical law. These inequal-
ities follow from inserting Eqs. (6.34) or (6.36) into Eq. (4.32) or inserting
Eqs. (6.46) or (6.47) into Eq. (4.38), as appropriate.
A material response will be stable whenever its fourth-order tangent-
modulus/tangent-compliance matrices are positive definite. Having said
this, seldom is it a simple endeavor to prove that any given fourth-order
tensor (with minor symmetry but not necessarily with major symmetry)
is, or is not, positive definite. To address this need, Nicholson (2013) just
proposed a tool to check for positive definiteness that ought to prove useful
in future studies.
identity matrix that has ones along the diagonal and zeros elsewhere, can-
not be used in this application. Nevertheless, techniques exist that employ
Crout reduction, e.g., which one can use to numerically solve Eqs. (6.57)
and (6.58) (Press et al. 2007).
As was done in our explicit formulation for elasticity, we once again call
upon invariant theory to replace the tensorial state variables in the internal
energy function, which are now two in number, with an appropriate set of
scalar fields known as invariants. Whenever two symmetric tensor fields
are present, ten separate invariants exist (Rivlin and Smith 1969). Of these
ten, only a few are used. Using the Green C and Cauchy C 1 metrics of
the deformation to construct admissible traces, in accordance with general
tensor analysis, three contributing moment invariants exist for each tensor
argument, viz.,
tr C 1E , tr C 1EC 1E , tr C 1EC 1EC 1E (6.61)
2 3
1 0 0
E 6 60
1/2 1/2 07
7
m) =
ten(m (6.70)
1 2 40 1/2 1/2 05
0 0 1
Likewise, placing the elastic energy function (6.71) into the expres-
sion for the deviatoric tangent modulus of an implicit elastic solid given in
Eqs. (6.42) and (6.43) leads to the formula
D = 2 C 1 C 1 13 C 1 C 1 , (6.73)
which is the same as the deviatoric tangent modulus of the explicit Green
elastic solid given in Eq. (5.66).
By definition, the tangent modulus M is the sum of its volumetric V
and deviatoric D parts implying that
M = 23 C 1 C 1 + 2( + p) C 1 C 1 (6.74)
with the hydrostatic pressure being p = 3. This fourth-order tensor
pushes forward from 0 to to become
m = 23 I I + 2( + p) I I, (6.75)
which describes the same material response as does Eqs. (5.65) and (5.66)
for the explicit Hookean solid based on Green strain. It reduces to the clas-
sic linear elastic solid stated of Eq. (6.68) whenever |p| and C 1 I
because = 23 .
The fact that pressure modulates the shear response via 2p I I, in
the Eulerian frame, is a curious result that follows from dierentiating
the hydrostatic/deviatoric decomposition of stress S = p C 1 + Sx giv-
ing SP = Pp C 1 + 2p EP + SxP (cf. Sect. 4.7.1) that when pushed forward
M M
from 0 to becomes s = Dp/Dt I + 2p d + sN = m : d . The pressure
M
term 2p I I arises from the contribution 2p d to stress rate s. This is
the same source that causes the pressure eect arising in Hills criterion
for material stability stated in Eq. (4.39). This contribution, which makes
physical sense, is missing from the Hookean solid of Eq. (6.68) derived
from an energy function that does not split the overall energy into isotropic
and isochoric constituents.
Three Rajagopal elastic solids are put forth, as established by Definition 6.1
stated on p. 172. The first, Material A, is an elastic solid that builds upon
the implicit Hookean solid of Sect. 6.4.1. The other two material mod-
els are elastic solids constructed within the framework of an isotropic/
deviatoric split, expanding upon the implicit Hookean solid of Sect. 6.4.1.3.
184 Soft Solids
Material B is an elastic solid that predicts final states of stress which do not
depend upon the path traversed through state space. Materials A and C
are elastic solids where the final states do depend upon the path traversed
through state space, in accordance with Nolls theorem stated on page 162
for which the experimental observations of Criscione et al. (2003a,b) are
illustrative. All three models are consistent with Fungs law, viz., they
are 3D representations of his 1D phenomenological law (6.1); specifically,
they comply with Definition 6.2 stated on p. 172.
The implicit Hookean solid is comprised of invariant combinations that
are quadratic in stress, in particular, they are from the invariants listed
in Eq. (6.62). In the spirit of Fungs law, viz., Definition 6.2, additional
invariant combinations are sought from the list (6.616.63) that are both
quadratic in stress and linear in strain and which extrapolate an existing
implicit Hookean solid.
6.5.1 Material A
An internal energy function that incorporates the Hookean elastic solid of
Eq. (6.64) and complies with Fungs law, as put forward in Definition 6.2,
is
2 1+
U= tr (SC ) + tr(SCSC ) tr(SESC ) (6.76)
2E 2E 2E
where the elastic modulus E and Poisson ratio have the same meaning
here as they do in the Hookean solid, while Fungs parameter is new to
this material definition.
B )1 : A ,
Recalling that the elastic compliance is given by C = (I I B
as defined in Eq. (6.36), then for the elastic energy function defining this
soft solid, i.e., Eq. (6.76), the embedded second-order derivatives of the
thermodynamic potential describe a resilience of
@2 U 1+
A= = C C + C C
@S @S E E
1
C E + E C (6.77)
E 2
with a Fung adjustment of
@2 U
B= = 12 I CS + CS I (6.78)
@S @E E
where and are separate dyadic operators with definitions found in
Eqs. (C.33), (C.39), and (C.41).
Implicit Elasticity 185
The first two terms on the right-hand side of Eq. (6.77) comprise the
elastic compliance of an isotropic Hookean solid, cf. Eq. (6.65); it is a
first-order eect. The last term in Eq. (6.77) constitutes the Rajagopal ef-
fect, which is an important second-order phenomenon. It is caused by a
coupling between the state variables for stress and strain that originates
within a thermodynamic potential of a Rajagopal and Srinivasa (2007,
2009) elastic solid.
The second thermodynamic potential B is a new elastic eect. Recall
that B = 0 for the implicit Hookean elastic solid. Through this potential,
Fungs eect enters into Rajagopals theory for elastic solids, viz., the
terms in Eqs. (6.1)(6.3).
An analytic construction of the compliance tensor C = (I I B B )1 : A
is not possible when expressed in terms of the two thermodynamic matri-
ces of Eqs. (6.77) and (6.78). This is a minor inconvenience when solving
BVPs such as those addressed in this text, but it presents no real obstacles
when implemented into software, as the theory is readily coded as an im-
plicit system of ODEs that can be solved through numerical methods, like
the implicit solver presented in Appendix D when applied to a Kronecker
representation of the model, as described in Appendix C. In practice, the
integrator of stress, e.g., using a method from Appendix D, often advances
with a finer step size than the integrator for deformation, e.g., those of
Algorithms 2.12.3.
1
b = 2 I s+sI , (6.80)
E
which possesses both minor and major symmetries.
6.5.2 Material B
The simplest model that one can consider in the material class where bulk
and shear responses are separable has internal energies described by
p2
Ub = 1 + , (6.83)
2
1
Us = tr(Sx C Sx C ) 1 + tr(C 1 Ex ) (6.84)
4
Implicit Elasticity 187
where and are the bulk and shear Fung parameters, respectively, which
are dimensionless, while and are the bulk and shear moduli from
Hookean elasticity, which have units of stress. Recall that tr(C 1 dEx ) = 0
by definition; nevertheless, tr(C 1 Ex ) 0 with negative values typically
arising in extensions, which is why the sign changes here vs. the presence
of in Materials A and C.
These two energy functions introduce the notion of a limiting state in
strain. For the bulk response, this idea originates with Hencky (1931)
who introduced it in the second of his two, phenomenological, constitutive
equations that he used to describe Bridgmans (1923) high-pressure exper-
iments. In the biological literature, this idea traces back to Carton et al.
(1962), whose model was expressed in terms of a compliance. Later, Fung
(1967) introduced a 1D model with a limiting state, like Carton et al.s, but
that was expressed in terms of a modulus.
2
D= C 1 C 1 13 C 1 C 1 (6.88)
1 + tr C 1 Ex
wherein
3
Q = , Q = , and p= (6.91)
(1 + )2 1 + tr(Ne) 1 +
with Q and Q being the respective bulk and shear tangent moduli, while eN
denotes the Almansi distortion whose value is returned by Algorithm 3.1,
where the identity tr(C 1 Ex ) = tr([F 1 ]T [Ex ] [F 1 ]) = tr(Ne) has been used.
Implicit Elasticity 189
6.5.3 Material C
There is another way that one can construct a Fungean material model
utilizing the invariants listed in Eqs. (6.61)(6.63), as they apply to the iso-
choric response, specifically
p2
Ub = 1 + , (6.92)
2
1 x x x xx
Us = tr S C S C tr S E S C (6.93)
4 4
where this bulk internal energy Ub is the same as that of Material B. It is
the shear internal energy Us that distinguishes Material C from Material B,
it being more like the contribution present in Material A.
Substituting the shear internal energy function (6.93) into its governing
constitutive equation (6.43) produces
dSx 1
= 2 C C C Ex + Ex C
dEx 2
x x
: I I + I CS + CS I
4
so that the deviatoric tangent modulus defined in Eq. (6.42) becomes
1
D = 2 C C x
C E +E C x
2
: I I 13 C C 1
1 1
x x x 1
+ I CS + CS I 3 CSC C (6.94)
2 2
with the Lagrangian tangent modulus M = V + D being the sum of
Eqs. (6.87) and (6.94). Notice that the Fung parameter appears twice
in this expressiona consequence of the thermodynamic coupling arising
from the potential function Us assumed in Eq. (6.93). This is due to the
Rajagopal eect.
This deviatoric tangent modulus is very similar to the tangent mod-
ulus M of Material A with E and S being replaced by their deviatoric
counterparts Ex and Sx with the additional terms on the right-hand side of
the colon, i.e., 13 (C + 2 C Sx C ) C 1 , arising as a consequence of the
term 13 C C 1 in dEx /dE = I I 13 C C 1 needed to come to
dS = M : dE .
190 Soft Solids
6.6 Examples
This illustrates the fact that the implicit Hookean solid is distinct from
the classical theories of linear and hyperelasticity outside the domain of
infinitesimal strain. It turns out that 12 ( 1 ) is a reasonable approxima-
tion for the true strain of Hencky (1928) in simple extension, viz., ln , cf.
Eq. (3.19).
Fig. 6.2 Curves of maximum stretch arising from the Rajagopal eect for uni-
axial and equi-biaxial ( = 1/2 ) extensions as a function of Fungs parameter
early literature on leather, cf. Morgan (1960). The above strain measure
arose from a direct integration of Eq. (6.99) that, because of the occurrence
and placement of the independent variable (viz., the stretch ) in the ODE
being solved, produced a power-law response instead of an exponential
response. This theoretical consequence is somewhat subtle.
The Rajagopal eect implies that this material model will exhibit infi-
nite stiness in a uniaxial extension at some maximum stretch max . For
this mode of deformation, this limiting stretch, in the sense of Carton et al.
(1962), is calculated to be
p
max = /( 2), (6.101)
as shown in Fig. 6.2. Equation (6.101) provides a convenient means for es-
timating the Fung parameter in materials that exhibit a limiting stretch in
their experimental response, as is often the case for soft biological tissues.
2 3 2 3
00 T00
[F] = 4 0 05 , [P ] = 4 0 T 05 (6.102)
00 0 00
where and are the biaxial and transverse stretches, respectively, while
T is the biaxial traction, with = 2 in an isochoric response. These
variables can be experimentally quantified, thereby allowing comparisons
to be made between theory and experiment for this BVP.
which is the same response derived for a uniaxial extension, but at 1/(1 )
times the stiness, which is 2 for the incompressible case where = 1/2
and, therefore, is in agreement with the same result predicted by explicit
theories of elasticity.
2 3 2 3
1 0
1 1 0
4
[F] = 0 1 05 so that F 4
= 0 1 0 5 (6.107)
00 0 0 1
2 3 2 3
s11 s12 0 s11 s12 s12 0
[s] = 4s21 s22 05 so [P ] = 4s21 s22 s22 05 (6.108)
0 0 0 0 0 0
3
M
s 11
195
that when inverted into Eq. (6.46) becomes a coupled system of ODEs
2 3
sP11 2T1 P m1111 m1121 m1112 m1122 0
TP1 T2 P 6m2111 m2121 m2112 m2122 7 P /2
6 7 (6.110)
TP 1 = T2 P + 4m1211 m1221 m1212 m1222 5 P /2
TP2 0 m2211 m2221 m2212 m2222 0
M
where the components for s came from Eq. (4.69), while the components
for d came from Eq. (2.60). Recall that the stress component s11 cannot be
gotten experimentally, but it can be predicted theoretically.
which introduces the Fung eect into the material response. From these
two Kronecker matrices one can construct a tangent compliance according
to Eq. (6.57) whose inverse, via Eq. (6.59), establishes its tangent modulus,
as stated in Eq. (6.58).
This example further illustrates the implicit quality of a Rajagopal elas-
tic solid.
6.7 Applications
Figure 6.3 presents data that establish, and leave little to no doubt, that
pericardium behaves as an implicit elastic solid in the sense of Rajagopal.
The curves drawn for T vs. 2 dT/d were constructed from data taken
from uniaxial and equi-biaxial extension experiments (Freed et al. 2010).
These curves have the same initial slope, which is governed by , with
an apparent but small oset in their intercepts, which is due to a Pois-
son eect via , cf. Eqs. (6.98) and (6.105). Unlike Fungs original data,9
which are plotted in Fig. 6.1, where a straight line provides an excellent fit,
here, in Fig. 6.3, one observes a degree of curvature in these data. This is
the Rajagopal eect, which enters into the denominator of Eqs. (6.98) and
(6.105), driving them to singularity at some critical threshold in stretch.
The experimental stress/stretch plots that associate with the tangent
moduli plots in Fig. 6.3 are presented in Fig. 6.4, where the extent of the
materials nonlinearity becomes immediately apparent and the existence of
a limiting stretch is strongly suggested. Furthermore, in accordance with
the trends of Fig. 6.2, this threshold in stretch arises earlier in equi-biaxial
stretching than in uniaxial stretching. The raw data that comprise Figs. 6.3
and 6.4 are tabulated in Tables 6.16.3, which are placed at the end of the
chapter. The data of Freed et al. (2010) have been truncated here at a stress
level of 500 kPa because of the substantial experimental noise (as reported
in Freed et al.) present in their tangent data that lie between the stresses of
500 kPa and 1 MPa.
The fact that the response curves in Figs. 6.3 and 6.4 are virtually the
same in the 1- and 2-directions for the biaxial data is an excellent in-
dicator of material isotropy, in a mechanical sense, not in a physiologic
sense. The stark dierences between the stretch/stress curves from uniax-
ial and biaxial extensions observed in Fig. 6.4 are caused by nonlinearity,
not anisotropy. Nonlinear eects are often misinterpreted in the literature
as being anisotropic in origin. Such errors often stem from a persons in-
correct extrapolation from their bias/experience with linear mechanics into
an application of nonlinear mechanics. This is yet another example of why
this text forgoes any detailed discussion of classic linear elasticity, which
you can study at a later time, if you are so inclined.
9 What is not known about Fungs experimental data are the stretches that
coincide with the data that are presented in Fig. 6.1. If the strains were moderate
(larger than innitesimal, but not large in a nite sense), as this author suspects,
then the Rajagopal eect would not yet be felt and the T vs. 2 dT/d response
would be eectively linear, as observed.
198 Soft Solids
Paracardium
40
Uniaxial
Tangent Modulus, 2dT/d (MPa)
Biaxial: 1 Direction
30 Biaxial: 2 Direction
20
10
0
0 100 200 300 400 500
True Stress, T (kPa)
Fig. 6.3 A partial reconstruction of Fig. 6 from Freed et al. (2010). The stress-
stretch curves that associate with these Fung plots are displayed in Fig. 6.4
Paracardium
500
Uniaxial
400 Biaxial: 1 Direction
Biaxial: 2 Direction
True Stress, T (kPa)
300
200
100
0
1 1.05 1.1 1.15
Stretch,
Fig. 6.4 A partial reconstruction of Fig. 7 from Freed et al. (2010) showing the
uniaxial and equi-biaxial stretch/stress response of a bovine pericardium
Implicit Elasticity 199
6.8 Exercises
6.8.2 Biaxial
Using the results you acquired from the exercises of prior chapters, derive
the ODEs that govern the response of the implicit Hookean and Rajagopal
elastic solids of Sects. 6.4.1 and 6.5, i.e., derive dT1 /d 1 and dT2 /d 2 for
these four material models.
-2
Traction, T (kPa)
-4
-6 N = 10
-8
0.5 0.6 0.7 0.8 0.9 1
Stretch,
Fig. 6.5 Experimental data averaged over ten compression experiments done on
the calcaneal fat pad, as reported in Miller-Young et al. (2002), and whose values
are tabulated in Table 6.4
Run 1
8 Run 2
Run 3
Traction, T (MPa)
0
1 1.2 1.4 1.6 1.8 2
Stretch,
Fig. 6.6 Uniaxial data from three experiments done on a resin-reinforced PDMS
silicon elastomer. Data were provided by Randall Schmidt of Dow Corning Cor-
poration, Midland, MI, 2012, and are for educational use only. Their values are
reported in Table 6.5
202 Soft Solids
Run 1
8 Run 2
Run 3
Traction, T (MPa)
0
1 2 3 4 5 6 7 8
Stretch,
Fig. 6.7 Uniaxial data from three experiments done on a ller-reinforced PDMS
silicon elastomer. Data were provided by Randall Schmidt of Dow Corning Cor-
poration, Midland, MI, 2012, and are for educational use only. Their values are
reported in Table 6.6
Implicit Elasticity 203
T 2 dT/d T 2 dT/d
kPa MPa kPa MPa
1.000 1.1 0.27 1.107 137.9 4.69
1.002 2.0 0.38 1.109 146.2 5.06
1.005 3.0 0.35 1.111 156.1 5.88
1.008 4.2 0.42 1.113 165.8 6.12
1.012 6.0 0.50 1.115 175.0 7.31
1.016 8.2 0.42 1.116 186.5 9.05
1.021 9.8 0.41 1.117 198.6 9.56
1.026 12.1 0.48 1.119 209.8 10.57
1.031 14.6 0.45 1.120 222.9 11.67
1.036 16.7 0.50 1.121 234.8 11.53
1.041 19.7 0.66 1.122 245.9 12.13
1.046 23.3 0.70 1.123 257.2 13.97
1.051 26.5 0.81 1.124 269.4 15.22
1.055 30.8 0.96 1.125 280.9 17.23
1.060 35.0 0.96 1.126 294.7 20.14
1.064 39.0 1.04 1.126 309.2 19.70
1.068 43.5 1.26 1.127 321.9 18.79
1.072 48.8 1.33 1.128 335.3 19.91
1.076 53.4 1.49 1.129 349.5 18.66
1.079 59.4 1.90 1.129 360.5 19.08
1.083 66.3 2.04 1.130 374.3 22.42
1.086 72.6 2.14 1.131 388.3 23.40
1.089 79.6 2.51 1.131 401.9 26.22
1.092 87.4 2.50 1.132 417.7 29.46
1.095 93.5 2.69 1.133 433.0 28.00
1.098 101.6 3.33 1.133 445.3 29.60
1.100 110.2 3.51 1.134 460.4 32.04
1.103 118.2 4.04 1.134 473.6 31.09
1.105 128.0 4.77 1.135 486.8 35.33
204 Soft Solids
T 2 dT/d T 2 dT/d
kPa MPa kPa MPa
1.000 1.1 0.15 1.086 133.8 5.51
1.003 1.7 0.25 1.087 140.9 6.11
1.006 2.5 0.35 1.088 148.7 6.37
1.009 3.9 0.45 1.090 155.8 6.51
1.012 5.3 0.50 1.091 163.2 7.04
1.015 7.0 0.54 1.092 170.8 7.30
1.018 8.7 0.55 1.093 178.0 8.15
1.021 10.4 0.62 1.094 186.5 9.18
1.024 12.6 0.71 1.095 194.9 9.33
1.028 14.8 0.75 1.096 202.9 10.51
1.031 17.2 0.85 1.097 212.6 11.61
1.034 20.0 0.90 1.098 221.7 10.71
1.037 22.6 0.94 1.099 238.5 12.25
1.040 25.6 1.07 1.100 247.1 11.66
1.043 28.9 1.09 1.101 254.8 12.09
1.046 32.0 1.12 1.101 263.4 13.63
1.049 35.3 1.32 1.102 272.5 14.11
1.051 39.3 1.44 1.103 281.0 15.27
1.054 43.1 1.54 1.104 300.4 17.32
1.057 47.3 1.75 1.105 309.9 18.43
1.059 51.9 1.94 1.106 329.2 17.56
1.062 56.8 2.05 1.107 347.9 20.09
1.064 61.6 2.23 1.108 364.9 20.78
1.066 66.8 2.36 1.109 384.8 22.91
1.068 71.8 2.39 1.109 394.1 24.86
1.071 76.7 2.66 1.110 405.2 27.80
1.073 82.3 2.89 1.110 426.5 26.36
1.075 87.7 3.01 1.111 436.6 26.79
1.076 93.2 3.37 1.111 446.7 26.64
1.078 99.2 3.83 1.112 456.2 27.11
1.080 105.7 4.10 1.112 466.3 28.18
1.081 112.0 4.51 1.112 476.2 27.82
1.083 119.1 5.25 1.113 485.6 29.87
1.084 126.8 5.48 1.113 496.7 33.74
Implicit Elasticity 205
T 2 dT/d T 2 dT/d
kPa MPa kPa MPa
1.003 1.5 0.32 1.086 127.8 5.31
1.006 2.5 0.33 1.088 134.2 5.66
1.008 3.4 0.37 1.089 141.9 6.52
1.011 4.6 0.43 1.090 149.6 6.34
1.014 5.9 0.48 1.092 156.2 6.92
1.017 7.3 0.46 1.093 164.5 7.87
1.020 8.6 0.52 1.094 172.4 7.61
1.023 10.3 0.60 1.095 179.5 8.55
1.026 12.0 0.63 1.096 188.5 9.61
1.028 13.9 0.69 1.097 196.9 9.40
1.031 16.0 0.74 1.098 204.9 10.21
1.034 18.1 0.76 1.099 222.4 10.86
1.037 20.3 0.81 1.100 230.8 11.36
1.040 22.6 0.92 1.101 239.4 12.16
1.042 25.4 0.95 1.102 248.2 12.62
1.045 27.8 1.03 1.103 265.7 15.17
1.048 30.9 1.22 1.104 276.1 15.40
1.051 34.1 1.21 1.105 294.2 17.41
1.053 37.1 1.38 1.106 304.2 16.75
1.056 41.0 1.56 1.107 322.9 20.73
1.058 44.7 1.56 1.108 341.4 21.18
1.060 48.4 1.82 1.108 352.4 23.74
1.063 53.0 2.02 1.109 362.2 22.76
1.065 57.3 2.00 1.109 371.6 25.43
1.067 61.5 2.31 1.110 382.8 27.22
1.069 66.9 2.63 1.110 393.0 26.97
1.071 71.9 2.67 1.111 403.5 29.41
1.073 77.1 2.94 1.111 415.0 31.28
1.075 82.8 3.24 1.111 426.4 30.57
1.077 88.5 3.40 1.112 436.9 29.79
1.079 94.3 3.67 1.112 447.4 31.70
1.080 100.5 4.17 1.113 458.8 31.55
1.082 107.3 4.33 1.113 468.8 30.45
1.084 113.4 4.56 1.113 479.3 34.25
1.085 120.3 5.28 1.114 491.5 34.43
206 Soft Solids
Viscoelasticity
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 209
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2_7, Springer International Publishing Switzerland 2014
210 Soft Solids
documented by, e.g., Gittus (1975), Nowick and Berry (1972), and Zener
(1948). The text by Phan-Thien (2002) provides photographs and short bio-
sketches of some of the prominent contributors to the field over the early
years, providing a historical perspective.
Polymeric materials exhibit viscoelastic behavior at temperatures that
exceed what is known as the glass transition temperature, or Tg , which lies
below the materials melting temperature Tm . At temperatures below Tg ,
their material behavior changes to, typically, an elasticplastic solid, which
constitutes a class of material models that is not discussed in this text. (The
interested reader is referred to, e.g., Havner (1992), Kachanov (1971), and
Stouer and Dame (1996).) Most components made from polymers are
processed at temperatures between Tg and Tm .
As this is an introductory text, the approach adopted here is intended to
be more straightforward. The idea is to provide insight and understanding
by developing useful viscoelastic theories as integral extensions that are
analytic continuations of the elastic theories presented in the previous two
chapters.
Fig. 7.1 Behavior of a linear viscoelastic solid exposed to cyclic loading. The
response appears to be elastic at both suciently slow and adequately fast cyclic
frequencies, while over a range in frequency between these two extremes the
stress/strain response is elliptic (for a linear response) with a clockwise rotation
where
G (t) 2 [0, 1] 8 t 0. (7.3)
Here G (t) is called the reduced relaxation function or relaxation kernel,
which is dimensionless. It is normalized so G (0) = 1 and G (1) = 0;
therefore, G(0) = 0 and G(1) = 1 , with 0 being the glassy (or in-
stantaneous) shear modulus and 1 being the rubbery (or equilibrium)
shear modulus. They order as 0 > 1 0, with 1 > 0 implying that
the material is a solid (1 = 0 for fluids); see Fig. 7.1.
One of the fundamental assumptions pertaining to viscoelastic kernel
functions, born out of experimental observation and thermodynamic con-
siderations (Coleman and Mizel 1968), is that the viscoelastic kernel func-
tion must be a nonincreasing (monotonically decreasing) function in time,
viz.,
G(t1 ) G(t2 ) 0 8 t2 > t1 0 ) G (t1 ) G (t2 ) 0 (7.4)
with G (1) = 0, which is a mathematical way of saying the material has a
fading memory.
Inserting the expression for the viscoelastic relaxation modulus G of
Eq. (7.2) into Boltzmanns viscoelastic integral (7.1), noting that G (0) = 1
and G (1) = 0, brings about an alternative 1D constitutive description for
the Boltzmann viscoelastic solid, i.e.,
Z
t
(t) = 1 (t) + 0 1 G (t t0 ) d (t0 ), (7.5)
0
212 Soft Solids
7.1.1 Clocks
Viscoelastic materials often respond according to their own internal clock,
whose speed can vary with respect to laboratory time in the presence of
environmental factors like temperature (Ferry 1980; Williams et al. 1955),
radiation or chemical reactions (Tobolsky 1960), or biological processes
like osmosis (Lillie and Gosline 1996), or agents like hormones (Rajagopal
and Wineman 2010). They may also vary with respect to mechanical fac-
tors like the state of stress (Bernstein and Shokooh 1980; Ferry 1980) or
the state of strain (Duenwald et al. 2010). The internal or material time
that comes from such a clock replaces the actual time as the independent
variable in a models viscoelastic kernel, so the concept is fairly straight-
forward to implement, viz., only the kernel is aected. These are often
important considerations that need to be taken into account in applications,
but their modeling lies outside the intended scope of this introductory text.
/ 0 / 0
0
TIME
Fig. 7.2 Behavior of a viscoelastic material during stress relaxation. Use of log
linear or loglog axes will often transform the data into presentations that can
be a useful aid when inferring which relaxation kernel to select
This analysis applies for the linear theory of viscoelasticity, not to non-
linear theories of viscoelasticity. In the case of nonlinear viscoelasticity,
eects arising from material nonlinearity (e.g., via a variable tangent mod-
ulus) blur with eects arising from the relaxation modulus, and the process
of material characterization becomes more challenging.
The relaxation modulus G is one of several viscoelastic material func-
tions commonly used. It is the only one that will be addressed in this text.
Another is the creep compliance, which is the inverse (in the Laplace trans-
form domain) to the relaxation modulus in linear viscoelasticity (Mainardi
2010; Pipkin 1972). A set of dynamic functions also exists that most visco-
elastic testing apparatuses are designed to measure. These are the storage
and loss moduli and their ratio, the loss tangent, a.k.a. tan (Lakes 1998).
7.1.3.1 Preconditioning
The biomechanics literature speaks of preconditioning specimens before
testing them; cf. Fung (1993, p. 270). Sometimes the preconditioning pro-
tocol is reported; sometimes it is not. The intended purpose of precondi-
tioning is to take a sample from its static equilibrium state and place it into a
stable dynamic orbit (a repeatable response) at the onset of the experiment,
Viscoelasticity 215
the idea being that this dynamic state is supposed to better represent the
actual, physiologic, in vivo (i.e., in a living organism) resting state for the
particular tissue of interest. In practice, preconditioning is often more than
a mechanical consideration; it is a biological consideration, too, e.g., as a
mechanical stimulus for osmotic rehydration of an excised tissue back to
its natural physiologic levels before testing.
To describe the response of a preconditioned specimen requires full
knowledge of the deformation history traversed since leaving a causal state
because
Z t
(t) = G(t t0 ) d (t0 )
Z0 t 0 Z t
= 0
G(t t ) d (t ) +0
G(t t0 ) d (t0 ), (7.9)
0 0 t
preconditioning experiment
where preconditioning occurs over some interval [0, t0 ), with the experi-
ment then proceeding over interval [t0 , t]. Of significance is the fact that the
preconditioned contribution to stress varies with time for all t0 , t0
t0
t,
due to current time t being present in G(t t0 ) with t denoting the time at the
end of the actual experiment, i.e., time t is at the upper limit of integration
in the second integral. Consequently, memory of a preconditioning from 0
to t0 will continue to fade over the lifetime of the experiment, viz., from t0
to t.
Because preconditioned deformation histories are, in general, only
causal before time t = 0 and do not return to a causal state by time t = t0 ,
any viscoelastic modeling of a history that contains a preconditioning can
only hope to be correct if the entire preconditioning portion of the defor-
mation history is taken into account in ones experimental analysis (Freed
and Doehring 2005).
Z t
(t) = 0 (t) M(t t0 ) (t0 ) dt0
0
Z t
= 0 (t) 0 1 M(t t0 ) (t0 ) dt0 (7.10)
0
with
dG (t t0 )
M(t) = (0 1 )M(t) and M(t t0 ) = , (7.11)
dt0
where the forcing function in this integrand is now strain , not strain-rate
d /dt. In contrast with Eq. (7.5), where a viscous overstress is added to
an equilibrium or rubbery elastic response, here the integral diminishes the
dynamic or glassy elastic response.
Lodge calls the viscoelastic material function M(t) the memory func-
tion, which has units of stress rate, while M is referred to as the memory
kernel. Unlike the relaxation kernel G , which is dimensionless, the mem-
ory kernel M has units of reciprocal time, i.e., it is a rate. Lodge (1956)
introduced his memory function into a model that he would later call the
rubberlike liquid (Lodge 1964, pp. 101104). Memory functions entered
into the biomechanics literature some 25 years later in a paper by Zhu et al.
(1991). The terminology of a memory function is a synonym for the
rate-of-relaxation function.
Both viscoelastic kernels, i.e., relaxation G (t) and memory M(t), must
monotonically decay to 0 as time t goes from 0 to 1 in order to be thermo-
dynamically viable (Coleman and Mizel 1968). The relaxation kernel G (t)
is bound to the interval [0, 1] for physical reasons, but no such restriction
applies to the memory kernel M(t). In fact, it is common for M(0) = 1
with the implication being that the material has a perfect recollection of
the present. If relaxation kernels G (t) were allowed to become unbounded
at time t = 0, like memory kernels often are, then (predicted) wave fronts
would travel through solids with infinite velocity (Caputo and Mainardi
1971b,a; Mainardi 2010) which is not physical. It is because of this physi-
cal characteristic that the restriction of G (0) = 1 is justified.
tangent modulus, like Fung used in his original theory and like the tangent
moduli derived in Chap. 5; while function T = T ( , ) supplies an implicit
tangent modulus of the type introduced in Chap. 6. The tangent modulus
is to be normalized in the sense that T (0, 0) = 1, i.e., Eq. (7.15) reduces to
a linear viscoelastic solid Eq. (7.1) in a neighborhood around the reference
state 0 ; otherwise, it can become nonlinear. The tangent modulus intro-
duced here, i.e., T , diers from the tangent moduli of Chaps. 5 and 6, viz.,
M, in that T = 1 M = 1 @&/@ is taken to be dimensionless.
A diversion from the developmental path traveled by Fung (1971, 1993)
is taken here. A somewhat dierent path is followed leading to a form
similar to Eq. (7.13); specifically, consider a change in variable
Many kernel functions have been proposed in the literature. Most have
theoretical underpinnings whose physics lie beyond the scope of this intro-
ductory text. Some of the more important and commonly used kernels are
presented below. A kernel is every bit as much a material characteristic as
are, say, its rubbery and glassy moduli. The ability to select an appropriate
kernel function is an important experimental outcome.
Numerous viscoelastic kernels have been derived from mechanical
models of springs and dashpots, whose discussions pervade the literature.
They are commonly used for instruction in introductory textbooks on the
subject, e.g., Mainardi (2010), Pipkin (1972), Tschoegl (1989), Wineman
and Rajagopal (2000), and Zener (1948). Mechanical models described
Viscoelasticity 221
with an aliated memory kernel (Freed and Diethelm 2006; Gross 1947)
E,0 (t/)
M(t) = , (7.24)
t
X
1
xk
E, (x) = , > 0, (7.25)
( + k)
k=0
2 Gross (1947) did not make use of the two-parameter Mittag-Leer function
E, (x).
3 Advancements were also made to fractional-order viscoelasticity in the Russian
literature around this same time, as noted in the text by Mainardi (2010).
Viscoelasticity 225
0
0.2
0.4
0.6
0.8
1
1
0.8
G (t) 0.6
0.4
0.2
0
0
1
2
t 3
Fig. 7.3 A 3D plot of the FOV relaxation function dened in Eq. (7.23), i.e.,
G (t) = E,1 ((t/) ) with = 1, and where t 2 [0, 3]
0
0.2
0.4
0.6
0.8
1
2
1.5
M (t) 1
0.5
0
0
1
2
t 3
Fig. 7.4 A 3D plot of the FOV memory function dened in Eq. (7.23), i.e.,
M(t) = E,0 ((t/) )/t with = 1, and where t 2 [0.001, 3]
0
0.2
0.4
0.6
0.8
1
3
2
M(t) / et
0
0
1
2
3
4
t 5
Fig. 7.5 A 3D plot of M(t)/et = E,0 ((t/) )/(t et/ ) with = 1 and where t 2
[0.001, 5], demonstrating that the FOV memory function in Eq. (7.23) asymptotes
algebraically to zero as t ! 1 for all 2 (0, 1)
The asymptotic behavior of the FOV relaxation kernel Eq. (7.23) is de-
scribed by the formul (Diethelm et al. 2005)
1 (t/)
(1+) / exp (t/) for t ,
G (t) (/t) (7.26)
1 for t .
(1) / 1+(t/)
4 The fractional model of Scott Blair (1944), where G(t) / t , is similar to the
CCM model; CCM is a regularization of the Scott Blair model. The Scott Blair
kernel is not normalized in that G (0) = 1 and, hence, his kernel propagates waves
with innite velocity.
5 This citation came from Tschoegl (1989, p. 320). The author has been unable
to secure this Russian document to corroborate his citation.
Viscoelasticity 229
Bagley (1987) has shown that Caputos 1967 fractional Voigt model for
viscoelasticity and the power-law kernels have the same response at large
times, while Freed and Diethelm (2006) suggest that power-law kernels
are a kind of regularization of the Abel kernel that appears in the fractional
dierential operator studied by Caputo (1967).
E1 (t/2 ) E1 (t/1 )
G (t) =
ln(2 /1 )
Z 2 t/
1 e
= d, 2 > 1 > 0, (7.31)
ln(2 /1 ) 1
whose aliated memory kernel is
exp(t/2 ) exp(t/1 )
M(t) = . (7.32)
t ln(2 /1 )
These viscoelastic functions have two characteristic times, 1 and 2 ,
whose stated constraint in Eq. (7.31) ensures a fading memory. Function
Z 1
e
E1 (t) = d, t > 0 (7.33)
t
is one of several equivalent definitions for the exponential integral that can
be found in the literature (Abramowitz and Stegun 1964; Mainardi 2010).
The asymptotic behavior of the BOX relaxation kernel Eq. (7.31) is de-
scribed by (Neubert 1963):
6 To compute the BOX relaxation kernel in Eq. (7.31), it is best to numerically
R
solve the integral 12 et/ / d, because taking the dierence between the two
exponential integrals E1 (t/2 ) E1 (t/1 ) will lead to measurable numerical error
whenever t is in the vicinity of either 1 or 2 given that standard algorithms are
used to solve the exponential integral, as presented in, e.g., Press et al. (2007).
The BOX relaxation kernel, as written down in Eq. (7.31), does not usually
appear in the QLV literature, e.g., Fung (1971, Sect. 7.6). Rather, a parameter
c > 0 appears that relates the glassy modulus 0 to the rubbery modulus 1 via
1 = 0 /(1 + c ln(2 /1 )) where c represents the height of a rectangular relaxation
spectrum that begins at 1 and ends at 2 .
230 Soft Solids
( t/1 t/2
1 ln( for t ,
G (t) 2 /1 ) (7.34)
2 exp(t/2 )1 exp(t/1 )
t ln(2 /1 ) for t .
When compared with the exponential relaxation kernel of the IOV,
Eq. (7.21), the BOX relaxation kernel exhibits a flexible initial rate of
decay, i.e., the BOX kernel has a derivative at t = 0 that is given by
(2 1 )/1 2 ln(2 /1 ), while the derivative of the IOV kernel is 1 at
t = 0. However, at large times, the BOX and IOV kernels exhibit similar
behaviors.
This kernel was derived from the phenomenological consideration
of a constant relaxation spectrum defined over a finite interval in time/
frequency, viz., a box (hence its name). This kernel has no contribution
arising from the spectrum for times less than 1 nor for times greater than
2 while, over the interval [1 , 2 ], the relaxation spectrum is held constant,
normalized by ln(2 /1 ) so that G (0) = 1 (Neubert 1963; Tschoegl 1989).
The author is not aware of any theory from statistical mechanics
whereby this model has been shown to be derived from a more fundamen-
tal analysis of the physics, i.e., it is phenomenological; nevertheless, it is
very popular.
X
K X
K
t ck t
G (t) = ck exp and M(t) = exp ,
k k
k=1 k=1 k
X
K
ck = 1, ck > 0 8 k, 0 < 1 < 2 < < K , (7.35)
k=1
where each term in each sum can be thought of as being associated with
a discrete integral or a separate internal-state variable (Simo and Hughes
1998, Chap. 10), each obeying the physics of a Maxwell relation. The sum
of all ck s equaling 1 ensures that G (0) = 1, while G (1) = 0 follows if
k > 0 for all k, and a monotonic decay will happen if ck > 0 for all k. The
MCM kernel obeys the principle of fading memory under these conditions.
The sheer number of material parameters that this kernel can employ
makes it a nearly impossible task to try to gain any physical insight into a
material via these parameters. This is why this author does not advocate
selecting the MCM kernel for the purpose of material characterization. It
is, however, a reasonable kernel to select for numerical implementation of
a viscoelastic model into a computationally intensive software application
like CFD or FE codes, for reasons that are explained below. For example,
Puso and Weiss (1998) adopted this approach to approximate the contin-
232 Soft Solids
uous spectrum of a BOX kernel with the discrete spectrum of a Prony se-
ries over its range [1 , 2 ] to implement their model for tendon into an FE
code.7
It is the recursion property of the exponential function, viz., et+t =
t t
e e , that makes exponential kernels so useful and powerful in applica-
tions, i.e., that provides a mechanism whereby an O(nN) algorithm can
be constructed. In particular, whenever a Prony series, or MCM solid, is
used as the kernel in a Volterra integral equation, that kernel will allow the
integral to be decomposed according to the scheme
Z t+t
0
et+tt f (t0 ) dt0
0
Z t Z t+t
0 0
= et+tt f (t0 ) dt0 + et+tt f (t0 ) dt0
0 t
Z t Z t+t (7.36)
t tt0 0 0 t+tt0 0 0
=e e f (t ) dt + e f (t ) dt
0 t
R 0
where 0t ett f (t0 ) dt0 is known from the previous integration step, it being
adjusted in the current step by a simple scale factor of et . An integration
R 0
of tt+t et+tt f (t0 ) dt0 is all that needs to be solved, and it is over an
interval of incremental length t so it can be evaluated quite economically
by standard numerical techniques. The IOV and MCM kernels, and the
BOX memory kernel, can exploit this recursive property to their advantage,
whereas the advantages of the CCM, FOV, KWW, and MPL kernels lie in
the physical insight that one can acquire by contrasting values for their
parameters across a class of materials.
A relaxation kernel is completely monotonic if (1)k dk G (t)/dtk 0
for all k = 0, 1, 2, : : : and, therefore, possesses a continuous relaxation
spectrum H (t) defined as a forcing function in the convolution integral
R 0
G (t) = 01 et/t H (t0 ) dt0 . The BOX, FOV, KWW, and the power-law ker-
nels all have relaxation functions that are completely monotonic. Because
they possess this property, the continuous relaxation spectrum of each can
be approximated by a sum of discrete relaxation spectra over any range in
time or frequency yielding nonunique MCM kernels.
7 The memory function of the BOX kernel is the dierence between two expo-
nentials; cf. Eq. (7.32). Approximating a BOX kernel with a Prony series for
integrating its viscoelastic convolution integral is not necessary provided that the
viscoelastic model is built to accept a memory function, as in Eqs. (7.13) or (7.19).
Viscoelasticity 233
Rule of Thumb: One to two Maxwell chains are needed for each decade
in time or frequency response that separates the glassy 0 and rubbery 1
plateaus with characteristic times k evenly spaced in logarithmic time over
the frequency range separating the rubbery and glassy plateaus.
For synthetic polymers, the number of Maxwell chains typically
exceeds seven.8 For biologic tissues, the number of terms in a Prony series
is usually around three or four, e.g., Miller-Young et al. (2002) and Puso
and Weiss (1998). There are two parameters per link of chain, i.e., per
exponential term in the Prony series. Yikes! The MCM kernel can easily
become a parameter estimation nightmare.
Tschoegl (1989, Sect. 3.6.2) describes a collocation method for fitting
the parameters of an MCM kernel, but it can lead to nonphysical, negative,
spectral lines.
Fulchiron et al. (1993) and Simhambhatla and Leonov (1993) propose
using an automated PadLaplace technique to obtain optimum MCM pa-
rameters. Here, a Pad expansion of chosen order is used to fit the data
(in our case, a characterized physical viscoelastic kernel) in the Laplace
domain where the problem becomes well posed. The results are then trans-
formed back into the time domain for use.
Another application that uses the Laplace transform was developed by
Park and Schapery (1999) and Schapery and Park (1999) to map the Prony
parameters between the various viscoelastic functions that exist in the lit-
erature, e.g., between the relaxation modulus and the creep compliance.
Another scheme has been proposed by Stuebner and Haider (2010),
where Gauss-Lagrange quadrature was applied to a relaxation kernel G (t),
in their case the BOX kernel, to quantify an MCM kernel using integration
nodes that are logarithmically distributed over time.
This authors opinion is that, whenever possible, the numerical inte-
gration of a convolution integral whose kernel derives from a continuous
relaxation spectrum, as do the BOX, CCM, FOV, KWW, and MPL ker-
nels, ought to be done with an algorithm that systematically samples its
history over the whole of its integration, as is the case with the algorithm
presented in Appendix E. Only in applications where the computational
eort will be excessive, e.g., in CFD and FEA codes where log(N) n
(with N being the number of integration steps, while n is the number of
8 Forexample, Park and Schapery (1999) use ten Prony elements for polyisobuty-
lene and eleven for polymethyl methacrylate. The author has even witnessed the
use of a Prony series with eighteen elements (thirty-six parameters) in a confer-
ence presentationyou can t an elephant with thirty parameters (Wei 1975).
234 Soft Solids
where the Eulerian fields found in Eq. (7.44) pull back from into 0 ac-
cording to the covariant map Eq. (B.20), thereby producing the Lagrangian
fields found in Eq. (7.38) or vice versa via the map Eq. (B.19).
Similarly, the contravariant strain fields defined in Eqs. (7.39) and
(7.45) obey the mappings
The conjecture of Kaye (1962) and Bernstein, Kearsley and Zapas (1963)
(K-BKZ) is adopted here as a means for analytically extending the explicit
theory of finite elasticity developed in Chap. 5 into a viscoelastic theory.
Other techniques also exist, e.g., the nonlinear theory of Pipkin and Rogers
(1968). The K-BKZ technique takes the 1D formulation of Eq. (7.13)
and extrapolates it into a 3D construction that can be used to model soft
isotropic materials. Their theory was created for viscoelastic fluids, but
here, their hypothesis is applied to solids. The conjecture put forward by
Bernstein et al. (1963) states:
Conjecture 7.1. For the Coleman and Noll (1964) fluid, the stress at time
t depends upon the history of the relative deformation between the config-
uration at time t and all configurations at times prior to t. To this idea we
add the following notions: (1) The eect of the configuration at time t0 < t
on the stress at time t is equivalent to the eect of stored elastic energy
with the configuration at time t0 as the preferred configuration. The eect
Viscoelasticity 239
0 1 dW (E; t) 1
S (t) = 1 + G (t) C 1 (t) C (t)
1 dE
Z
0 1 t dW (E; t0 , t) 1 0
+ M(t t0 ) C 1 (t0 , t) C (t , t) dt0 , (7.51)
1 0 dE
which describes the K-BKZ viscoelastic Lodge solid expressed in terms
of Lagrangian fields. A subtle restriction arising from the derivation of
Eq. (7.13) is that the resulting forcing function within the integrand must
be an additive strain field; therefore, only choices for W (E) that lead to
additive strain measures are admissible for analytic continuation via the
K-BKZ hypothesis.
For incompressible materials, S is replaced by .
This viscoelastic material model reduces to Lodge elasticity in the
quasi-static limit. This fact illustrates a fundamental premise of the K-BKZ
hypothesis: The tensorial structure of a viscoelastic material is the tenso-
rial structure of its limiting elastic response. This provides a great simplifi-
cation to the overall process of constructing a viscoelastic material model.
This material model has two elastic parameters: the glassy 0 and rub-
bery 1 shear moduli, plus whatever parameters are required to quantify
the reduced relaxation G and memory M kernels and whatever parame-
ters are needed to describe a strain-energy function W . Because the elastic
strain energy has dimensions of stress, its gradients have been normalized
by the rubbery shear modulus in the above expression.
Any Lodge elastic solid can be analytically continued into an admissi-
ble viscoelastic solid via Eq. (7.51), provided that its resulting strain field
is additive, which is actually a very strong restriction. For example, the
first-order, incompressible, Lodge, elastic solid of Eq. (5.32) becomes
(t) = 2 1 + (0 1 ) G (t) E(t)
Z t
+ 2(0 1 ) M(t t0 ) E(t0 , t) dt0 (7.52)
0
or, equivalently, when pushed forward into the Eulerian frame according to
Eqs. (3.17), (4.27), (7.50), and (B.15), it becomes
(t) = 2 1 + (0 1 ) G (t) e(t)
Z t
+ 2(0 1 ) M(t t0 ) e(t0 , t) dt0 . (7.53)
0
There are representations of the K-BKZ viscoelastic neo-Hookean solid
written in terms of Lagrangian or Eulerian fields, respectively. These are
Viscoelasticity 241
admissible K-BKZ models in the sense that the Lodge E and Signorini e
strains are additive; cf. Eqs. (7.39) and (7.45).
In the limiting case of a fluid, which is obtained by setting 1 = 0,
Eq. (7.52) reduces to the Lodge (1956, 1958) rubberlike liquid that, re-
markably,
has been derived from two dierent molecular theories: the bead-
-spring theory of Rouse and Zimm for very dilute solutions of de-
formable long molecules in an incompressible Newtonian solvent: : :,
and the network theory of Green and Tobolsky, Yamamoto, and Lodge
which is developed for concentrated polymer solutions and undiluted
or molten polymers. (Lodge et al. 1978)
So, like the neo-Hookean elastic solid, the neo-Hookean viscoelastic solid
has a sound foundation in theoretical physics.
that, when pushed forward into the Eulerian frame in accordance with
Eqs. (3.1), (3.16), (4.27), (7.48), and (B.19), becomes
(t) = 2 1 + (0 1 ) G(t) e(t)
Z t
+ (0 1 ) M(t t0 ) e(t0 , t) dt0 . (7.56)
0
These are admissible viscoelastic material models in the sense of the
K-BKZ hypothesis, because the Green E and Almansi e strain fields are
additive; cf. Eqs. (7.38) and (7.44).
This is consistent with Eq. (7.55) and the fact that C 1 is used as a met-
ric of deformation in 0 in this context, which is required to convert the
covariant strain fields that lie between these two contravariant metrics into
its respective contravariant field so that it becomes compatible with stress,
which maps contravariantly. All sense of tensorial structure gets lost in its
Eulerian formulation, Eq. (7.58).
which pushes forward into the Eulerian frame according to Eqs. (3.37),
(7.42), and (B.83), resulting in
Z t
z(t) = F(t0 , t) t (s, e, e; t0 ) : d (t0 ) F T (t0 , t) dt0 (7.63)
0
or, in component form, as
Z t
j
ij
z = FIi (t0 , t) tIJKL (s, e, e; t0 ) dKL (t0 ) FJ (t0 , t) dt0 (7.64)
0
which, in practice, is evaluated more simply via
Z t
z(t) = F(t) F 1 (t0 ) t (s, e, e; t0 ) : d (t0 ) F T (t0 ) dt0 F T (t), (7.65)
0
where Z (t0 , t) = Z (t) Z (t0 ) and z(t0 , t) = z(t) z(t0 ) follow from the addi-
tive property of integrals. The viscoelastic strains Z and z of Guth9 obey
the contravariant maps of Eqs. (B.15) and (B.16), so a simpler pushforward
operation is
[z] = [F] [Z ] [F]T , (7.66)
implying that Guth strain is most easily integrated in the Lagrangian
configuration 0 . The viscoelastic tangent moduli T and t map as fourth-
order contravariant tensors according to Eq. (6.48).
To be able to write the Stieltjes integral, the first line in Eq. (7.62), as a
Riemann integral, the second line in (7.62), requires that the tangent mod-
ulus T be a continuous function and that the strain rate EP be integrable.
Equation (7.60) applies as an analytic continuation for explicit and im-
plicit elastic solids alike via the K-BKZ hypothesis. They dier in how
one quantifies their respective Guth strains Z . In essence, Guth strain can
be thought of as a dimensionless solution to any R t admissible elastic solid
1 M
that can be expressed in the form of Z = 21 0 (t ) : dE (t0 ) wherein
0
M = dS /dE .
where K is the reduced bulk relaxation kernel, akin to G for the shear
response, and Y (t0 , t) = Y (t) Y (t0 ) with Y establishing the volumetric
strain of Guth defined by
Z t
1
Y (t) = V (p, ; t0 ) : dE (t0 )
1 0
Z t
1
= 2 p(t0 ) C 1 (t0 ) C 1 (t0 )
1 0
1 dp 0 1 0
(t ) C (t ) C (t ) : dE (t0 ) (7.72)
1 0
3 d
which simplifies to
Z t
1
Y (t) = 2 p(t0 ) E(t
P 0)
1 0
1 dp 0 P 0 1 0
(t ) tr C E; t C (t ) dt0 , (7.73)
3 d
where the hydrostatic pressure p and its gradient dp/d are described by an
appropriate elastic constitutive equation that, in this case, may be explicit
or implicit in origin. In the explicit case, p is given by a thermodynamic
potential from which dp/d is gotten by dierentiation, while in the im-
plicit case, dp/d is given by a thermodynamic potential from which p is
gotten through integration.
Ss (t) = 2 1 + (0 1 ) G (t) Z (t)
Z t
+ 2(0 1 ) M(t t0 ) Z (t0 , t) dt0 . (7.74)
0
Whenever Green strain E is the independent strain measure, then its Guth
strain extends Eqs. (5.63) and (6.42) such that
Z t x
1 dS 0
Z (t) = (t ) : I I 13 C (t0 ) C 1 (t0 ) : dE (t0 ) (7.75)
21 0 dEx
Viscoelasticity 249
or equivalently
Z t x
1 dS 0
Z (t) = (t ) : dEx (t0 ) (7.76)
21 0 dEx
where Eq. (3.55) defines the deviatoric strain rate dEx . Whenever Lodge
strain E is the independent strain measure, then Guth strain will extend
Eq. (5.69), producing
Z t
1 dSx 0 1
Z (t) = (t ) : C C 1 13 C 1 C 1 (t0 )
21 0 dE x
1 0
2 C S + S C (t ) : dE (t0 ) (7.77)
x x 1
or, equivalently,
Z t x
1 dS 0 x 0)
Z (t) = (t ) : dE(t
x
21 0 dE
Z t
1
P Sx + Sx C EP (t0 ) dt0 , (7.78)
EC
1 0
x Gradients dSx /dEx and dSx /dE
where Eq. (3.61) defines dE. x follow from
their appropriate thermodynamic potentials, which can be of explicit or
implicit origin.
which, when substituted into Eq. (7.60), becomes the classic viscoelastic
neo-Hookean solid of Eq. (7.52), where the strain-rate identity found in
Eq. (3.46) has been made use of.
From Eq. (5.32), the first-order Lodge elastic solid has a strain energy
of W = 1 tr(EC EC ) so that, when substituted into Eq. (7.68), one gets
a Guth strain of
Z t
1
Z (t) = C C 1 (t0 )
0
1
1 0
2 C E + E C (t ) : dE (t0 )
Z t
= E(t) 2 P E + EC EP (t0 ) dt0
EC (7.80)
0
7.6 Examples
where and are the axial and transverse stretches, respectively, with =
1/2 for an isochoric response. These matrices can, in turn, be used in any
of the various definitions for relative strain found in the integrands of the
various viscoelastic models. In the above formul and those to follow, the
one-argument stretches (t) and (t0 ) are shorthand notations for (0, t) and
(0, t0 ), respectively. In all cases that follow, stretches are measured against
the initial frame associating with time 0, i.e., a Lagrangian viewpoint is
adopted.
The dependent variable in these viscoelastic models, viz., the traction
or engineering stress T, can be experimentally quantified via Eq. (4.43),
thereby allowing the component of stress in Eq. (6.95) to be determined
from which comparisons between theory and experiment can follow.
The viscoelastic neo-Hookean solid Eq. (7.52) has a Lagrange multi-
plier for uniaxial extension of
Z t
1 0 (t0 )
} = G(t) 1 M(t t ) 1 dt0 , (7.90)
(t) 0 (t)
leading to an axial traction of
Z t
1 0 (t) (t0 )
T(t) = G(t) (t) 2 + M(t t ) dt0 , (7.91)
(t) 0 2 (t0 ) 2 (t)
where Eqs. (3.66), (4.25), (4.48), (7.39), and (7.52) have been used. This
is not the viscoelastic model proposed by Guth et al. (1946), viz., the strain
measure in the integrand is dierent from (t0 ) 1/ 2 (t0 ), as they supposed
it to be.
Viscoelasticity 253
from which the various definitions for relative strain measures can be de-
termined.
Substituting Eq. (7.92) for F(t0 , t) into the definition for the relative
Lodge strain E(t0 , t) given in Eq. (7.39), while using Eq. (4.57) to quantify
stress and Eq. (3.72) to quantify strain, the Lagrange multiplier for the
viscoelastic neo-Hookean solid Eq. (7.52) obtained from the 33 component
is given by
Z t !
1 4 (t0 )
}(t) = G(t) 1 4 M(t t0 ) 1 4 dt0 , (7.94)
(t) 0 (t)
Z t !
1 (t) 4 (t0 )
T(t) = G(t) (t) 5 + M(t t0 ) dt0 , (7.95)
(t) 0 2 (t0 ) 5 (t)
where P22 = 0 and P21 = P12 , in this case. Components P12 and P22
can both be experimentally measured, but P11 cannot, as documented in
Eqs. (4.63) and (4.68). The negative state of stress arising for P11 may
manifest itself as a wrinkling of the surface whenever thin membranes are
sheared. Notice that (t) (t0 ) is the shear strain between states 0 and
, in accordance with the K-BKZ hypothesis.
which follow from F(t0 , t) = F(0, t) F 1 (0, t0 ) and whose associated areal
stretch is
(t)
(t0 , t) = , (7.102)
(t0)
as derived from
(t0 , t) = det F(t0 , t) = det F(0, t) F 1 (0, t0 ) = det F(0, t)
det F 1 (0, t0 ) = det F(0, t)/ det F(0, t0 ) =
(0, t)/
(0, t0 ), where use has
been made of properties (A.74) and (A.75). These results can be used
in any of the various definitions for relative strain found in the integrands
of our various viscoelastic models. In the above formul, the one-argu-
ment variables are shorthand notations, e.g., 1 (t) and 1 (t0 ) are shorthand
notations for 1 (0, t) and 1 (0, t0 ), respectively.
The viscoelastic material models considered in this text present them-
selves as 2 2 matrix equations that are best solved numerically in their
Lagrangian constructions. To be able to solve the incompressible models
requires that one know how to calculate their Lagrange multipliers, which
need to be solved prior to solving the constitutive equation itself. For this
BVP, this has been done by assuming that the membranes are planar and
in a state of plane stress, from which the Lagrange multiplier can then be
derived from the 33 component straightaway.
256 Soft Solids
where
is the areal stretch defined in Eq. (1.23). Unlike the prior BVPs
considered, where the Lagrange multiplier has been assimilated into the
equation governing the stress components, here it is best left as a separate
equation to be sequentially solved with the constitutive equation.
7.7 Applications
FOV Kernel
1
Reduced Relaxation Modulus, G(t)
= 1.0
0.8
= 0.8
= 0.6
= 0.4
0.6 = 0.2
0.4
0.2
0
0 1 2 3
t/
Fig. 7.6 Plot of normalized time t/ vs. the reduced relaxation modulus G (t) for
the fractional-order viscoelastic model where = 1 associates with the integer-
order viscoelastic model of Zener. Emphasis is on the short-time response
The FOV relaxation kernel of Eq. (7.23) is drawn in Figs. 7.6 and 7.7.
Of note in the short-time response is that the FOV kernel behaves as the
IOV kernel at = 1. Stress relaxation quickens with decreasing values
of , until about half the stress has been recovered. Beyond that point,
the rate or relaxation slows way down, relative to exponential decay, and
may even appear to plateau out. An examination of this kernels long-term
relaxation response, displayed in Fig. 7.7, demonstrates that this apparent
plateau is a fictional artifact of the material becoming power law in its
response, i.e., it is likely that no such plateau exists, and one must be careful
not to falsely associate it with the rubbery modulus. So, what one looks for
in an experimental data set that suggests considering the FOV kernel is an
extremely rapid stress recovery in the early stages of relaxation followed
by an ultraslow recovery at large time.
The KWW relaxation kernel of Eq. (7.27) is drawn in Figs. 7.8 and
7.9. Its short-time response has a similar character to that of the FOV,
with a few notable distinctions. For 0
t , the FOV and KWW
kernels are eectively equivalent, as noted in Eq. (7.26). As time gets
larger, these kernels predict vastly dierent responses. The KWW kernel
has a point in common for all values of , i.e., at t = the stress will have
258 Soft Solids
FOV Kernel
100
Reduced Rlaxation Modulus, G(t)
10-1
= 1.0
= 0.8
10-2 = 0.6
= 0.4
= 0.2
10-3
10-4
10-4 10-3 10-2 10-1 100 101 102 103 104
t/
Fig. 7.7 Loglog plot of normalized time t/ vs. the reduced relaxation modulus
G (t) for the fractional-order viscoelastic model where = 1 associates with the
integer-order viscoelastic model of Zener. Emphasis is on the long-time response
Fig. 7.8 Plot of normalized time t/ vs. the reduced relaxation modulus G (t) for
the Kohlrausch and Williams and Watts model where = 1 associates with the
integer-order viscoelastic model of Zener. Emphasis is on the short-time response
FOV of Caputo and Mainardi (1971b,a) (see also Doehring et al. (2005)),
with the advantage being that the CCM kernels are much easier and quicker
to compute.
The BOX relaxation kernel of Eq. (7.31) is drawn in Figs. 7.12 and
7.13. In these figures, one observes that the BOX kernel can mimic the
short-time behaviors of the other kernels. When time t has exceeded 2 ,
stress will relax with an exponential rate of decay, like the KWW kernel,
whereas for times that lie between the characteristic times 1 and 2 , the
BOX kernel behaves more like a power law. For these reasons, the BOX
kernel has found application in biological tissues. Because their relaxation-
like histories do not reach large times, the characterized response behaves
like the FOV kernel since the exponential tail at long times is not reached
in practice.
The BOX, CCM, FOV, and KWW viscoelastic kernels can all be used
to account for rapid short-time relaxation behavior that is common among
soft solids, relative to the IOV kernel. So, if short-time relaxation is
all that is to be modeled, as is usually the case when modeling living
260 Soft Solids
Fig. 7.9 Loglog plot of normalized time t/ vs. the reduced relaxation modulus
G (t) for the Kohlrausch and Williams and Watts model where = 1 associates
with the integer-order viscoelastic model of Zener. Emphasis is on the long-time
response
7.7.1.1 Polyisobutylene
To illustrate this process, consider what is likely to be the most thoroughly
characterized viscoelastic material: the polyisobutylene prepared by the
National Bureau of Standards in the early days of polymer research when
testing techniques and standards were being developed, which necessitated
a uniformity in material being tested by the various university and indus-
trial laboratories involved. The experimental data presented in Figs. 7.14
and 7.15 were published by Catsi and Tobolsky (1955) where a time/
temperature translation of the data has taken place, being corrected to
25 C.
Viscoelasticity 261
CCM Kernel
1
IOV kernel
Reduced Relaxation Modulus, G(t)
0.8 = 1.0
= 0.8
= 0.6
= 0.4
0.6 = 0.2
0.4
0.2
0
0 1 2 3
t/
Fig. 7.10 Plot of normalized time t/ vs. the reduced relaxation modulus G (t)
for the ColeCole model contrasted against the integer-order viscoelastic model
of Zener. Emphasis is on the short-time response
The linearlinear plot in time vs. reduced relaxation in Fig. 7.14 shows
the short-time response of polyisobutylene with the stress relaxing at a
much faster rate than an exponential would suggest, while the loglog plot
in Fig. 7.15 shows the response becoming power law in character at large
times. Hence, the exponential decay properties of the BOX, KWW, and
IOV kernels eliminate them from further consideration for this particular
material, leaving the CCM and FOV kernels as candidate models.
Fitting the CCM and FOV kernels to the experimental data of
Catsi and Tobolsky (1955) leads to the following parameterizations,
with 95 % confidence intervals being reported. For the CCM kernel,
= 0.6850 0.0025 and = (4.05 0.11) 109 s with a coecient
of determination of R2 = 0.999. For the FOV kernel, = 0.6491 0.0028
and = (9.79 0.40) 109 s with a coecient of determination
of R2 = 0.998. These parameters were fit using a genetic algorithm
(Goldberg 1989, 2002) for parameter estimation in the loglog space of
Fig. 7.15 (not in the linearlinear space of Fig. 7.14).
When viewed in the space in which they were fit, i.e., Fig. 7.15, the
CCM kernel is seen to do an excellent job of describing the data over the
entire range of the data, whereas, although the FOV kernel does a very
262 Soft Solids
CCM Kernel
100
Reduced Rlaxation Modulus, G(t)
10-1
IOV kernel
= 1.0
10-2 = 0.8
= 0.6
= 0.4
= 0.2
10-3
10-4
10-4 10-3 10-2 10-1 100 101 102 103 104
t/
Fig. 7.11 Loglog plot of normalized time t/ vs. the reduced relaxation modulus
G (t) for the ColeCole model contrasted against the integer-order viscoelastic
model of Zener. Emphasis is on the long-time response
good job, it does not do as well as the CCM kernel. The superior quality of
fit of the CCM kernel over the FOV kernel for this material is immediately
apparent in the short-time response data of Fig. 7.14. For polyisobutylene,
the CCM kernel of Cole and Cole (1941, 1942) is the best kernel of the
viscoelastic kernels considered in this text at describing the time-dependent
viscoelastic characteristics of polyisobutylene.
BOX Kernel
1
IOV kernel
1 = 5x10-1, 2 = 5x100
Reduced Relaxation Modulus, G(t)
0.8
1 = 5x10-2, 2 = 5x101
1 = 5x10-3, 2 = 5x102
0.6
1 = 5x10-4, 2 = 5x103
0.4
0.2
0
0 1 2 3
t
Fig. 7.12 Plot of time t vs. the reduced relaxation modulus G (t) for the BOX
model of Neubert contrasted against the integer-order viscoelastic model of Zener.
Emphasis is on the short-time response
BOX Kernel
100
Reduced Rlaxation Modulus, G(t)
10-1
10-4
10-4 10-3 10-2 10-1 100 101 102 103 104
t
Fig. 7.13 Loglog plot of time t vs. the reduced relaxation modulus G (t) for the
BOX model of Neubert contrasted against the integer-order viscoelastic model of
Zener. Emphasis is on the long-time response
scission, mechanisms that these kernels are not capable of representing; cf.
Wineman and Min (2003).
0.6
0.4
0.2
0
0 2x10-9 4x10-9 6x10-9 8x10-9 10-8
Time, t (s)
Fig. 7.14 Linear plot of time t vs. the reduced relaxation modulus G (t) for the
polyisobutylene data of Catsi and Tobolsky (1955) corrected for 25 C, zooming
in on the short-time response. The glassy and rubbery shear moduli are 0 =
1.0 GPa and 1 = 250 kPa (Tobolsky 1956). The time axis represents about 2
for the CCM kernel and about 1 for the FOV kernel
7.8 Exercises
100
10-1
10-2
Experimental Data
CCM kernel
FOV kernel
10-3
10-4
10-11 10-10 10-09 10-08 10-07 10-06 10-05 10-4 10-3 10-2
Time, t (s)
Fig. 7.15 Loglog plot of time t vs. the reduced relaxation modulus G (t) for
the polyisobutylene data of Catsi and Tobolsky (1955) corrected for 25 C. The
glassy and rubbery shear moduli are 0 = 1.0 GPa and 1 = 250 kPa (Tobolsky
1956). Emphasis is on the long-time response
100
10-1
Experimental Data
IOV kernel
10-2
102 103 104 105
Time, t (s)
Fig. 7.16 Loglog plot of time t vs. the reduced relaxation modulus G (t) for the
radiation-cured natural rubber data of Tobolsky and Mercurio (1959) at 130 C.
The glassy and rubbery shear moduli are 0 = 850 MPa (Tobolsky 1956) and
1 = 350 kPa, cf. p. 147
Then determine its parameters and draw your fitted curve against their
data. Discuss the results.
2. An alternative power-law kernel to the CCM kernel that one finds in
the literature is the MPL kernel G (t) = 1/(1 t/) . What is its memory
kernel? Construct a graph that shows the similarities and dierences
between the CCM and MPL kernels. Discuss what you find.
3. The data presented in Tables 7.2, 7.3, and 7.4 and drawn in Figs. 7.18,
7.19, and 7.20 were obtained from simple-shear experiments done on
a sample of porcine myocardial heart tissue by Dokos et al. (2002).
These are precious data; they are rare. Shear was imposed on three
orthogonal planes. The fact that the responses are dierent in dierent
material directions is an indication of material anisotropy, a topic not
addressed in this text. The hysteretic area is proportional to the energy
lost per cycle due to the inherent viscoelastic behavior of muscle. The
extreme nonlinearity is indicative of a tensorial character suggested by
the implicit elastic solid of Chap. 6 that was analytically continued into
the viscoelastic domain in Sect. 7.5. Based on this, your instructor may
ask you to use these data in any number of dierent ways to illustrate
anyone of these issues, which are topics that lie beyond the intended
scope of this text.
268 Soft Solids
Fig. 7.17 Experimental data for the reduced relaxation modulus G (t) are from
McLoughlin and Tobolsky (1952) for polymethyl methacrylate (PMMA) corrected
for 40 C, as recorded in Table 7.1. The glassy and rubbery shear moduli are
0 = 750 MPa and 1 = 750 kPa (Tobolsky 1956)
Viscoelasticity 269
12
FN
FS
8
Shear Stress, T (kPa)
-4
-8
-12
-16
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Shear,
NF
NS
1
Shear Stress, T (kPa)
-1
-2
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Shear,
4 SF
SN
Shear Stress, T (kPa)
-2
-4
-6
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Shear,
FN FS
Shear Traction Shear Traction Shear Traction Shear Traction
T1 (kPa) T1 (kPa) T1 (kPa) T1 (kPa)
0.495 11.220 0.494 15.531 0.493 15.360 0.494 12.091
0.495 10.894 0.487 13.780 0.494 14.933 0.495 11.910
0.490 10.199 0.477 11.330 0.489 14.092 0.491 11.148
0.479 8.231 0.461 8.462 0.479 11.613 0.481 9.434
0.461 6.028 0.437 5.307 0.464 8.603 0.467 7.598
0.437 4.005 0.408 3.212 0.444 5.772 0.448 5.578
0.407 2.421 0.372 1.737 0.420 3.694 0.417 3.433
0.371 1.351 0.338 0.975 0.389 2.043 0.387 2.148
0.331 0.768 0.296 0.468 0.352 1.138 0.352 1.232
0.289 0.412 0.267 0.338 0.318 0.690 0.311 0.709
0.243 0.213 0.225 0.162 0.278 0.449 0.266 0.400
0.194 0.133 0.176 0.075 0.232 0.240 0.215 0.250
0.141 0.061 0.124 0.031 0.182 0.149 0.166 0.156
0.088 0.019 0.074 0.056 0.133 0.087 0.138 0.060
0.036 0.063 0.018 0.069 0.082 0.057 0.084 0.061
0.023 0.106 0.039 0.134 0.035 0.025 0.036 0.001
0.078 0.112 0.091 0.177 0.018 0.031 0.014 0.023
0.134 0.233 0.137 0.253 0.075 0.093 0.062 0.085
0.183 0.377 0.182 0.376 0.124 0.156 0.113 0.179
0.235 0.682 0.228 0.572 0.173 0.306 0.161 0.240
0.281 1.219 0.274 0.885 0.224 0.496 0.207 0.416
0.324 1.975 0.318 1.402 0.266 0.796 0.249 0.687
0.363 3.168 0.355 2.122 0.304 1.291 0.292 1.105
0.393 4.588 0.389 3.239 0.343 2.057 0.330 1.771
0.424 6.837 0.418 4.592 0.381 3.247 0.368 2.913
0.450 9.628 0.443 6.235 0.412 4.776 0.406 4.994
0.469 12.190 0.464 8.159 0.440 6.643 0.438 7.794
0.482 14.222 0.479 9.668 0.460 8.479 0.464 10.869
0.490 15.519 0.490 10.827 0.477 10.188 0.481 13.669
0.494 15.745 0.495 11.220 0.488 11.537 0.493 15.360
Data are from Dokos et al. (2002, Fig. 6) whose raw values were supplied to the
author with permission granted for use by Prof. Ian J. LeGrice, University of
Auckland, Auckland, New Zealand
a Samples 3 3 3 mm in size were subjected to simple shearing over a range in
shear of = 0.5 under a sinusoidal waveform with a 30 s period. Their frame
of reference is : F aligns with the direction of the muscle bers, S is transverse
to F and lies within the sheet of muscle bers, and N is normal to this sheet.
Six shearing orientations were exercised on the same tissue sample, designated by
pairs, e.g., FS. The rst character, F in this case, denes the direction normal to
the shearing plane (i.e., direction 2 in Fig. 1.5). The second character, S in this
case, denes the direction of shearing (i.e., direction 1 in Fig. 1.5)
274 Soft Solids
NF NS
Shear Traction Shear Traction Shear Traction Shear Traction
T1 (kPa) T1 (kPa) T1 (kPa) T1 (kPa)
0.499 1.839 0.499 1.669 0.499 1.990 0.499 1.596
0.497 1.804 0.495 1.598 0.497 1.920 0.494 1.502
0.489 1.665 0.485 1.441 0.488 1.754 0.482 1.305
0.477 1.437 0.469 1.196 0.474 1.478 0.467 1.055
0.460 1.160 0.450 0.984 0.457 1.242 0.448 0.859
0.440 0.919 0.424 0.758 0.434 0.961 0.421 0.637
0.415 0.710 0.389 0.545 0.405 0.705 0.392 0.461
0.386 0.522 0.348 0.369 0.371 0.499 0.359 0.334
0.351 0.348 0.306 0.246 0.333 0.361 0.315 0.212
0.312 0.245 0.251 0.123 0.289 0.220 0.284 0.162
0.273 0.138 0.223 0.087 0.247 0.125 0.243 0.109
0.231 -yy0.103 0.177 0.086 0.202 0.085 0.198 0.064
0.186 0.053 0.122 0.036 0.149 0.036 0.144 0.035
0.141 0.035 0.070 0.019 0.102 0.014 0.092 0.010
0.090 0.000 0.017 0.018 0.055 0.011 0.043 0.011
0.043 0.000 0.042 0.017 0.004 0.035 0.010 0.036
0.005 0.017 0.092 0.071 0.046 0.060 0.064 0.088
0.053 0.035 0.150 0.104 0.099 0.060 0.112 0.109
0.101 0.069 0.197 0.156 0.157 0.129 0.163 0.151
0.148 0.104 0.246 0.227 0.213 0.158 0.208 0.221
0.194 0.140 0.289 0.313 0.255 0.232 0.252 0.339
0.239 0.227 0.326 0.434 0.294 0.294 0.292 0.431
0.283 0.335 0.361 0.572 0.337 0.454 0.339 0.617
0.327 0.458 0.398 0.799 0.372 0.580 0.379 0.878
0.362 0.615 0.428 1.058 0.403 0.777 0.413 1.104
0.396 0.792 0.451 1.282 0.432 0.974 0.441 1.361
0.424 1.001 0.472 1.508 0.457 1.174 0.462 1.597
0.450 1.195 0.487 1.700 0.474 1.351 0.481 1.808
0.474 1.403 0.496 1.804 0.490 1.548 0.493 1.944
0.489 1.579 0.499 1.839 0.497 1.621 0.499 1.990
Data are from Dokos et al. (2002, Fig. 6) whose raw values were supplied to the
author with permission granted for use by Prof. Ian J. LeGrice, University of
Auckland, Auckland, New Zealand
a Samples 3 3 3 mm in size were subjected to simple shearing over a range in
shear of = 0.5 under a sinusoidal waveform with a 30 s period. Their frame
of reference is: F aligns with the direction of the muscle bers, S is transverse
to F and lies within the sheet of muscle bers, and N is normal to this sheet.
Six shearing orientations were exercised on the same tissue sample, designated by
pairs, e.g., FS. The rst character, F in this case, denes the direction normal to
the shearing plane (i.e., direction 2 in Fig. 1.5). The second character, S in this
case, denes the direction of shearing (i.e., direction 1 in Fig. 1.5)
Viscoelasticity 275
SF SN
Shear Traction Shear Traction Shear Traction Shear Traction
T1 (kPa) T1 (kPa) T1 (kPa) T1 (kPa)
0.498 4.287 0.497 5.719 0.498 4.810 0.497 2.927
0.494 4.086 0.492 5.285 0.494 4.668 0.489 2.576
0.483 3.557 0.482 4.533 0.485 4.245 0.474 2.092
0.467 2.909 0.467 3.628 0.467 3.432 0.455 1.564
0.444 2.168 0.447 2.710 0.448 2.764 0.430 1.110
0.420 1.626 0.424 1.928 0.424 2.124 0.402 0.758
0.392 1.098 0.395 1.277 0.395 1.515 0.369 0.508
0.356 0.708 0.362 0.787 0.363 1.037 0.329 0.299
0.315 0.378 0.324 0.448 0.322 0.646 0.290 0.196
0.271 0.227 0.283 0.239 0.284 0.429 0.239 0.139
0.227 0.115 0.240 0.151 0.242 0.269 0.193 0.094
0.182 0.056 0.196 0.062 0.199 0.153 0.147 0.064
0.129 0.014 0.148 0.027 0.148 0.108 0.090 0.064
0.078 0.006 0.112 0.003 0.100 0.053 0.062 0.024
0.022 0.041 0.053 0.017 0.052 0.023 0.002 0.008
0.030 0.086 0.006 0.059 0.001 0.007 0.054 0.036
0.086 0.106 0.056 0.100 0.050 0.022 0.112 0.051
0.134 0.175 0.108 0.139 0.105 0.034 0.166 0.122
0.183 0.257 0.155 0.197 0.159 0.080 0.216 0.253
0.227 0.384 0.204 0.278 0.210 0.125 0.262 0.414
0.270 0.548 0.252 0.410 0.260 0.197 0.309 0.732
0.309 0.795 0.298 0.646 0.306 0.345 0.350 1.139
0.350 1.157 0.337 0.959 0.350 0.549 0.389 1.706
0.383 1.683 0.372 1.369 0.388 0.874 0.420 2.358
0.415 2.375 0.411 2.011 0.421 1.255 0.444 2.967
0.442 3.234 0.440 2.651 0.449 1.768 0.465 3.665
0.462 4.029 0.463 3.277 0.470 2.223 0.480 4.186
0.480 4.890 0.481 3.819 0.486 2.661 0.491 4.594
0.490 5.462 0.492 4.210 0.495 2.926 0.496 4.840
0.496 5.795 0.498 4.287 0.499 3.001 0.498 4.810
Data are from Dokos et al. (2002, Fig. 6) whose raw values were supplied to the
author with permission granted for use by Prof. Ian J. LeGrice, University of
Auckland, Auckland, New Zealand
a Samples 3 3 3 mm in size were subjected to simple shearing over a range in
shear of = 0.5 under a sinusoidal waveform with a 30 s period. Their frame
of reference is: F aligns with the direction of the muscle bers, S is transverse
to F and lies within the sheet of muscle bers, and N is normal to this sheet.
Six shearing orientations were exercised on the same tissue sample, designated by
pairs, e.g., FS. The rst character, F in this case, denes the direction normal to
the shearing plane (i.e., direction 2 in Fig. 1.5). The second character, S in this
case, denes the direction of shearing (i.e., direction 1 in Fig. 1.5)
Appendix A
Linear Algebra
Vectors are arrays with physical units that obey a linear transformation rule
between coordinate frames, like the pushforward and pull-back operators
discussed in Eqs. (B.7)(B.14) of Appendix B.
Several vector notations are used throughout this text, some being
illustrated in this appendix. To help keep things as compact as possi-
ble, two-dimensional (2D) vectors are considered, with their extension to
three dimensions (3D) being straightforward. A Cartesian metric tensor
g = ij e i e j is used throughout this text, wherein ij denotes the Kro-
necker delta: 1 if i = j, 0 otherwise.
Consider two vectors U and V defined as
(
Ui ei , U1
U = {U } = , (A.1)
Ui e i , U2
and
(
V i ei , V1
V = {V } = (A.2)
Vi e i , V2
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 277
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
278 Soft Solids
where ei and e i are the unit base vectors in the ith coordinate direction with
i = 1, 2 whose components are
1 0
e1 e 1 = and e2 e 2 = , (A.3)
0 1
where ei e i in Cartesian tensor analysis; they dier in general tensor
analysis. Vector components Ui and Ui , say, possess dierent physical
properties, which is the topic of Appendix B. These two vectors add ac-
cording to
(
(Ui + V i ) ei , U1 + V1
U +V = or {U + V } = , (A.4)
(Ui + Vi ) e i , U2 + V2
Tensors are matrices with physical units that obey a linear transformation
rule between coordinate frames, like the pushforward and pull-back oper-
ators discussed in Eqs. (B.15)(B.26) of Appendix B.
Consider two tensors M and N defined by1
8
<M ei e j ,
ij
M11 M12
M = Mij ei e j , with [M ] = (A.14)
: M21 M22
Mij e i e j ,
j
1 There are two mixed tensors in general tensor analysis, e.g., Mij and Mi (cf.
Truesdell and Noll 2004). However, in Cartesian tensor analysis, one variant for
a mixed tensor suces. It is being denoted as simply Mij .
280 Soft Solids
and
8
<N ei e j ,
ij
N11 N12
N = Nij ei e j , with [N ] = (A.15)
: N21 N22
Nij e i e j ,
where the i index associates with the row and the j index associates with the
column. Again, 2D fields are used to help keep things concise. No issues
arise when extending the following concepts into 3D. The above tensors
add according to
8
<(M + N ) ei e j ,
ij ij
with
M11 + N11 M12 + N12
[M + N ] = , (A.17)
M21 + N21 M22 + N22
and they subtract according to
8
<(M N ) ei e j ,
ij ij
cM = cMij ei e j , (A.22)
:
cMij e i e j ,
with
M11 M12 cM11 cM12
c[M ] = c = (A.23)
M21 M22 cM21 cM22
Linear Algebra 281
with
" #
M11 1 (M + M )
2 12 21
[sym(M )] = 1 (M + M ) (A.28)
2 21 12 M 22
with
" #
0 1 (M M )
2 12 21
[skew(M )] = 1 (M M ) , (A.30)
2 21 12 0
so that
M = sym(M ) + skew(M ) (A.31)
where
sym(M T ) = sym(M ) and skew(M T ) = skew(M ). (A.32)
In three-dimensional analysis it is at times advantageous to express a skew
tensor, say M , as a vector, say m, via
2 3
0 m3 m2 Mij = ijk mk ,
[M ] = 4m3 0 m1 5 with, e.g., (A.33)
m2 m1 0 Mij = ijk mk ,
where ijk and ijk denote the permutation symbol whose values are : 1 if
indices ijk are unique and cyclically ordered, i.e., 123, 231, or 312; 1 if
they are unique with a reverse cyclic ordering, viz., 132, 213, or 321; or 0
whenever two or three indices have the same indical value.
A tensor can contract with a vector, producing another vector. This is a
kind of linear mapping, i.e., a transformation matrix, and is defined by
8
<M V j ei ,
ij
U = M V = Mij V j ei , (A.34)
:
Mij V e ,
j i
with
M11 V1 + M12 V2
{U } = [M ]{V } = . (A.35)
M21 V1 + M22 V2
A dierent contraction occurs when
8
<M V j ei ,
ji
U = M T V = ij Mkj k` V` ei , (A.36)
:
Mji V j e i ,
with
M11 V1 + M21 V2
{U } = M T {V } = , (A.37)
M12 V1 + M22 V2
Linear Algebra 283
where
p q
kM k = M : M = M211 + M221 + M212 + M222 (A.52)
is the Frobenius norm, which is the tensor analog to the Euclidean norm
defined in Eq. (A.12) for vectors.
The inner and outer dyadic products between two matrices, e.g., M N
and M N , are defined in Appendix C.
The determinant is the second invariant of tensor M in 2-space and is
given by
i M11 M12
det M = det M j = = M11 M22 M21 M12 , (A.53)
M21 M22
while for a tensor in 3-space, say T , it is given by
T22 T23 T21 T23 T21 T22
det T = T11
T12
+ T13 , (A.54)
T32 T33 T31 T33 T31 T32
wherein the 2 2 determinants are cofactors Cij of det T given by (1)i+j
times its associated minor determinant (obtained by removing the ith row
and jth column from T ). The determinant is defined for tensor arguments
of mixed character (Truesdell and Noll 2004, p. 15). A tensor is said to be
singular if its determinant is zero; otherwise, it is non-singular.
The determinant and its cofactors play an important role in quantifying
the inverse of a tensor field. For a tensor in 2D, its inverse is given by
1 1 C11 C21 1 M22 M12
[M ] = = , (A.55)
det M C12 C22 det M M21 M11
while for a tensor in 3D, its inverse is given by
2 3
C11 C21 C31
1 4C12 C22 C32 5
[T 1 ] =
det T C C C
13 23 33
2 3
T T T32 T23 T32 T13 T12 T33 T12 T23 T22 T13
1 4 22 33
= T31 T23 T21 T33 T11 T33 T31 T13 T21 T13 T11 T23 5 , (A.56)
det T T T T T T T T T T T T T
21 32 31 22 31 12 11 32 11 22 21 12
which will exist if their respective matrices are not singular. A more con-
densed way to express the tensor inverse is via the expression
adj M
M 1 = provided that det M 0 (A.57)
det M
286 Soft Solids
where
I(T ) = tr T ,
II(T ) = 12 tr(T )2 tr(T 2 ) , (A.88)
III(T ) = det T
so now there are three invariants. This theorem is useful for the purpose
of expressing a higher-order power of a tensor field in terms of a polyno-
mial whose order is no greater than the dimension of the matrix and whose
coecients are restricted to be scalar functions of its invariants.
To illustrate how one can use the CayleyHamilton theorem, consider
the exponential of a tensor defined by its Taylors series as
exp T = I + T + 2! 1 T2 + 1 T3 + 1 T4 + 1 T5 +:::
3! 4! 5!
that, because of the CayleyHamilton theorem, can be approximated as
exp T 1 + 3! 1 III + 1 I III + 1 III I 2 II I
4! 5!
+ 1 3!1 II 1 I II III 1 II(I 2 II) I III T
4! 5!
+ 2! 1 + 1 I + 1 I 2 II + 1 I 3 2I II + III T 2 ,
3! 4! 5!
which has been truncated at the sixth term in the expansion.
Appendix B
The reasons for selecting one tensor field over another, when multiple
choices exist, can be rationally explained, but it requires more mathe-
matical capability than the typical reader of this text would be expected
to have. That is why the authors reasoning is outlined here, in an ap-
pendix. Like the body of this text, full mathematical rigor is not pre-
sented here; rather, continuity in the flow of ideas and thought processes is
sought. You can get a deeper understanding of the topics covered in this
appendix from textbooks that address general tensor analysis, e.g., Chad-
wick (1976), Holzapfel (2000), Lodge (1964, 1974), Marsden and Hughes
(1983), Sokolniko (1964), Truesdell and Noll (2004), and Truesdell and
Toupin (1960), some of which address applications in the general theory
of relativity, too.
Vectors can map from one configuration into another in one of two
ways: as a covariant vector or as a contravariant vector. Normals to sur-
faces are examples of covariant vectors, e.g., Eq. (2.11), while tangents to
curves are examples of contravariant vectors, e.g., Eq. (2.2). The need to
distinguish between these two vector types arises in two important con-
siderations when using Cartesian tensor analysis: in the transfer of field
from one configuration into another and when taking time derivatives or
integrals in the spatial configuration. Their importance in general tensor
analysis is much more profound.
It is absolutely essential that the reader becomes familiar with the con-
cepts and ideas presented in this appendix. The physics described here
prepare the canvas upon which a theory for the mechanics of materials can
then be painted.
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 289
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
290 Soft Solids
{v}
A B
[Q]
e3 e2
e1
e2 e3
[Q1]
e1
matrix Q j0 i . These formul are expressed more simply in their matrix notations
{v0 } = [Q] {v} and {v} = [Q1 ] {v0 }
A physical field must not depend upon the frame in which an observer
observes an event and in which the field is to be quantified. Said dierently,
consider a scenario where observer A observes physical event v from a
coordinate frame of (e1 , e2 , e3 ), while observer B observes the same event
but from a dierent perspective, viz., from some other frame (e10 , e20 , e30 ),
as illustrated in Fig. B.1. The numerical values they assigned to this event
will likely be dierent, yet the event was the same. How can this be? If the
same event was observed by the two observers, then it stands to reason that
the physical fields being measured ought to be the same. Consequently, the
cause of this quandary must be in how their separate values associate with
their individual coordinate frames of reference.
Investigating further, one finds that these two observers coordinate
frames relate to one another via the linear transformations
{e10 } = [Q] {e1 }, {e20 } = [Q] {e2 }, {e30 } = [Q] {e3 },
(B.1)
{e1 } = [Q1 ] {e10 }, {e2 } = [Q1 ] {e20 }, {e3 } = [Q1 ] {e30 }
where
j
[Q] = Q 0j i ei0 e j and [Q1 ] = [Q1 ] 0 i e j e 0 i (B.2)
Covariant and Contravariant Issues: Configuration Physics 291
with [Q] [Q1 ] having components Q 0ki [Q1 ] k0j = 00ji and [Q1 ] [Q]
having components [Q1 ] i0 k Q 0j k = ij and, therefore, [Q] [Q1 ] =
[Q1 ] [Q] = [I]. For the matrix array Q 0j i , index 0 i is the row indexer
j
and index j is the column indexer. For the matrix array [Q1 ] 0 i , index j is
the row indexer and index 0 i is the column indexer. Matrix [Q] is orthog-
onal because its inverse [Q1 ] = [Q]1 equals its transpose [Q]T , while
its determinant is det Q = 1. The determinant will be +1 whenever both
bases have the same handedness and 1 whenever one basis is left handed
and the other is right handed.
A physical vector, say v, is described by an array {v} whose elements
v i associate with a basis, say (e1 , e2 , e3 ), that when described in another
basis, say (e10 , e20 , e30 ), has elements v 0 i that obey a linear mapping, i.e.,
v = v i ei = v 0 i ei0 (B.3)
has elements that obey
v 0 i = Q 0j i v j , {v0 } = [Q] {v},
or (B.4)
j
v j = [Q1 ] 0 i v 0 i , {v} = [Q]T {v0 },
because [Q1 ] = [Q]T . Here notation assists us in our understanding that
it is not the physical field v that is dierent, as measured by our two ob-
servers; rather, it is their frame of reference that diers. Furthermore, if
one knows how to map between these two frames of observation, then ei-
ther observer can transform their measurements into the others frame of
reference, at which point their results should agree, within experimental
error.
A physical tensor, say T , is described by a matrix [T ] whose elements
T ij associate with a basis, say (e1 , e2 , e3 ), that when described in another
basis, say (e10 , e20 , e30 ), has elements T 0 ij that obey a linear mapping, i.e.,
T = T ij ei e j = T 0 ij ei0 e j0 (B.5)
with
0j
T 0 ij = Q 0ki T k` Q ` , [T 0 ] = [Q] [T ] [Q]T ,
or (B.6)
j
T ij = [Q1 ] i0 k T 0 k` [Q1 ] 0 ` , [T ] = [Q]T [T 0 ] [Q],
because [Q1 ]T = [Q]. Here again the notation assists us in our under-
standing that a tensor and its components are distinct ideas, although both
are representations of the same physical entity.
292 Soft Solids
[F]
@ t0 @t
0
[F1]
Fig. B.2 Field-transfer operations for a contravariant vector eld, e.g., a tangent
to a curve; cf. Eqs. (2.2) and (2.3). Matrix [F] is the pushforward map; matrix
[F 1 ] is the pull-back map
[F1]T
@ t0 @t
0 [F]T
Fig. B.3 Field-transfer operations for a covariant vector eld, e.g., a normal to
a surface; cf. Eq. (2.11). Matrix [F 1 ]T is the pushforward map; matrix [F]T is
the pull-back map
Note: [F]T {g} F Tg; they are distinct. From Eq. (B.14), [F]T {g}
has components FIi g i , whereas, from Eq. (A.36), F Tg has a component
j
representation of IJ FJ ji g i . The indical locations of index i on vector
g are dierent. The former is a matrix operation; the latter is a tensor
operation. The transfer of field from one configuration into another is a
matrix operation. It is a mapping between two Hilbert spaces. Tensor
operations are only defined within a single Hilbert space. This is why these
j
two notations have been introduced. In general tensor analysis, IJ FJ ji
j
would become [G1 ]IJ FJ g ji wherein [G1 ]IJ denotes the components of
the contravariant metric G 1 of 0 , while g ij denotes the components of
the covariant metric g of . This distinction arises whenever curvilinear
coordinates are employed.
Contravariant tensors in and 0 of g(x, t) = g ij (x, t) ei e j and
G (X , t) = GIJ (X , t) eI eJ transfer their fields according to the mappings
@x i @x j IJ
g ij (x, t) = G (X , t), (B.15)
@XI @XJ
@XI @XJ ij
GIJ (X , t) = g (x, t) (B.16)
@x i @x j
that in matrix notation become
[g(x, t)] = [F] [G (X , t)] [F]T pushforward, (B.17)
[G (X , t)] = [F 1 ] [g(x, t)] [F 1 ]T pullback. (B.18)
Covariant and Contravariant Issues: Configuration Physics 295
is due to a deformation or shape change that takes place over the history or
path that has been traversed between these two configurations.
The process is straightforward for taking the time derivative of a field at-
tached to a particle, say at P in body B, whenever that field is expressed
in terms of material coordinates X (i.e., it is a Lagrangian field). In this
case, the coordinates associate with the fixed particle P of interest, so the
time rate of change of a material field is just its partial derivative taken with
respect to time. For material vector fields, one has
@G (X , t)
GP (X , t) = , (B.27)
@t
while for material tensor fields, one has
@G (X , t)
GP (X , t) = , (B.28)
@t
with there being no need to distinguish between contravariant, covariant,
or mixed fields for these operations.
These rates are obvious. It turns out, though, that physically admissi-
ble time derivatives are not so intuitive whenever spatial coordinates are
selected for use.
To take the time derivative of a field attached to particle P originally
located at coordinates X but whose field of interest is expressed in terms
of spatial coordinates x = (X , t), not material coordinates X (i.e., it is an
Eulerian field), requires that one pay more attention to the details. The time
derivative of a spatial field taken at a fixed particle is called its material
derivative and is denoted by D/Dt. The material derivative of a spatial
vector field is given by
Dg(x, t) @g(x, t) @g(x, t) @x
= +
Dt @t @x @t (B.29)
= g(x,
P t) + grad g(x, t) v(x, t)
of which the acceleration vector in Eq. (1.9) is an example, while the ma-
terial derivative for a spatial tensor field is given by
Dg(x, t) @g(x, t) @g(x, t) @x
= +
Dt @t @x @t (B.30)
P
= g(x, t) + grad g(x, t) v(x, t)
298 Soft Solids
The Lie derivative taken along the path of motion of some particle P , often
referred to as an Oldroyd (1950, 1970) derivative, is an objective measure
for the time rate of change of spatial fields. An objective measure is one
that does not introduce any unwanted rotational eects into its outcome.
Objective rates are physically admissible measures for quantifying time
rates of change of a field in the spatial configuration .
A recipe for computing the Lie derivative of any spatial tensor field is
fairly straightforward (cf. Holzapfel 2000, p. 106):
O O
Objective derivatives of covariant tensors g(x, t) = g ij (x, t) e i e j and
P IJ (X , t) e I e J transfer their fields according to the mappings
GP (X , t) = G
P IJ (X , t) @X @X ,
I J
O
g ij (x, t) = G (B.43)
@x i @x j
P IJ (X , t) = O @x i @x j
G g ij (x, t) I (B.44)
@X @XJ
O
[g(x, t)] = [F 1 ]T [GP (X , t)] [F 1 ] pushforward, (B.45)
O
[GP (X , t)] = [F]T [g(x, t)] [F] pullback. (B.46)
And objective derivatives of mixed tensors g(x, t) = g ij (x, t) ei e j and
P I (X , t) eI e J transfer their fields according to the mappings
GP (X , t) = GJ
i @x i @XJ
g j (x, t) = GP IJ (X , t) j , (B.47)
@XI @x
P I (X , t) = @X @x j
I i
G J g j (x, t) J (B.48)
@x i @X
[g(x, t)] = [F] [GP (X , t)] [F 1 ] pushforward, (B.49)
[GP (X , t)] = [F 1 ] [g(x, t)] [F] pullback. (B.50)
same physical time rate of change of an Eulerian field as does the simple
partial derivative taken with respect to time of its equivalent Lagrangian
field when expressed in the material or reference configuration 0 .
Lie derivatives are quantified via the following formul. For contravari-
ant vectors g(x, t) = g i (x, t) ei , the Lie derivative is computed as2
Mi @g i @g i j
g (x, t) = + j v ` ij g j ,
@t @x (B.51)
M
{g(x, t)} = {Dg/Dt} [l] {g},
where l = ` ij ei e j is the velocity gradient with ` ij = @v i (x, t)/@x j . For
covariant vectors g(x, t) = g i (x, t) e i , the Lie derivative is computed as
O @g i @g i j j
g i (x, t) = + j v + `i g j ,
@t @x (B.52)
O
{g(x, t)} = {Dg/Dt} + [l]T {g}.
For contravariant tensors g(x, t) = g ij (x, t) ei e j , the Lie derivative is
computed as
M ij @g ij @g ij k j
g (x, t) = + k v `ki g kj g ik ` k ,
@t @x (B.53)
M
[g(x, t)] = [Dg/Dt] [l] [g] [g] [l]T .
For covariant tensors g(x, t) = g ij (x, t) e i e j , the Lie derivative is com-
puted as
O @g ij @g ij k
g ij (x, t) = + k v + ` ki g kj + g ik ` kj ,
@t @x (B.54)
O
[g(x, t)] = [Dg/Dt] + [l]T [g] + [g] [l].
And for mixed tensors g(x, t) = g ji (x, t) ei e j , the Lie derivative is com-
puted as
i @g ij @g ji
g j (x, t) = + v k ` ik g jk + g ki ` kj ,
@t @x k (B.55)
[g(x, t)] = [Dg/Dt] [l] [g] + [g] [l].
These are referred to as the Oldroyd (1950, 1970) convected derivatives,
as he was the first to derive and apply them in the literature of continuous
media mechanics. It is the opinion of this author that Oldroyds 1950 paper
was the single-most important paper published in the twentieth century in
the field of continuum mechanics.
B.4.1 Integration
The inverse of dierentiation is integration, and a like recipe exists for
integrating Lie derivatives:
1. Pull back the Lie derivative of a spatial field of interest given in into
its material equivalent in 0 by applying, e.g., Eqs. (B.32), (B.36),
(B.40), (B.44), or (B.48).
2. Integrate this rate over time, which is now defined in the reference
configuration 0 , i.e., it is defined over a single Hilbert space where
an algebra exists, thereby permitting the summing of integration to
take place (Freed 2010).
3. Push forward the result obtained in 0 back into the spatial config-
uration by using the appropriate mapping, e.g., Eqs. (B.7), (B.11),
(B.15), (B.19), or (B.23).
With this methodology in place, objective integrations of objective
derivatives of contravariant vectors g(x, t) = g i (x, t) ei and G (X , t) =
GI (X , t) eI are given by the integral equations
t Z
Mi 0 0 0 @x i t P I
g (x , t ) dt = G (X , t0 ) dt0 , (B.56)
0 @X I
0
Z t I tM
PGI (X , t0 ) dt0 = @X g i (x 0 , t0 ) dt0 (B.57)
0 @x i
0
that in matrix notation become
t Z t
M 0 0
{g(x , t )} dt = [F] {GP (X , t0 )} dt0
0 pushforward, (B.58)
0 0
Covariant and Contravariant Issues: Configuration Physics 303
Z t t
M
{GP (X , t0 )} dt0 = [F 1 ] {g(x 0 , t0 )} dt0 pullback (B.59)
0 0
wherein x 0 = (X , t0 ).
Objective integrations of objective derivatives of covariant vectors
g(x, t) = g i (x, t) e i and G (X , t) = GI (X , t) e I transfer their fields accord-
ing to the mappings
t Z t
O @XI
g i (x 0 , t0 ) dt0 = GP I (X , t0 ) dt0 , (B.60)
0 0 @x i
Z t t
P I (X , t0 ) dt0 = O @x i
G g i (x 0 , t0 ) dt0 I (B.61)
0 0 @X
that in matrix notation become
t Z t
O
{g(x 0 , t0 )} dt0 = [F 1 ]T {GP (X , t0 )} dt0 pushforward, (B.62)
Z 0t t0
O
{GP (X , t0 )} dt0 = [F]T {g(x 0 , t0 )} dt0 pullback. (B.63)
0 0
Objective integrals of objective derivatives of contravariant tensors
g(x, t) = g ij (x, t) ei e j and G (X , t) = GIJ (X , t) eI eJ transfer their
fields according to the mappings
t Z
Mij 0 0 @x i @x j t P IJ
g (x , t ) dt0 = G (X , t0 ) dt0 , (B.64)
0 @X I @XJ
0
Z t I @XJ t M
0 0 @X
G IJ
P (X , t ) dt = g ij (x 0 , t0 ) dt0 (B.65)
0 @x i @x j
0
that in matrix notation become
t Z t
M 0 0
[g(x , t )] dt = [F] [GP (X , t0 )] dt0 [F]T
0 pushforward, (B.66)
0 0
Z t t
M
[G (X , t )] dt = [F ] [g(x 0 , t0 )] dt0 [F 1 ]T
P 0 0 1 pullback. (B.67)
0 0
Objective integrals of objective derivatives of covariant tensors g(x, t) =
g ij (x, t) e i e j and G (X , t) = GIJ (X , t) e I e J transfer their fields ac-
cording to the mappings
t Z t
P IJ (X , t0 ) dt0 @X @X ,
O I J
0 0 0
g ij (x , t ) dt = G (B.68)
0 0 @x i @x j
Z t t
PGIJ (X , t0 ) dt0 = O @x i @x j
g ij (x 0 , t0 ) dt0 I (B.69)
0 0 @X @XJ
304 Soft Solids
in Eq. (7.40). Oldroyd did this nearly a decade before Noll (1958) intro-
duced this idea in his body of work. Using the generalized deformation
gradient of Eq. (7.40), convected integration of a contravariant vector be-
comes
t Z t
Mi 0 0 0 @x i M j 0 0 0
g (x , t ) dt = g (x , t ) dt ,
0 @x
0j
0
t Z t (B.81)
M 0 0 0 0 M 0 0 0
{g(x , t )} dt = [F(t , t)] {g(x , t )} dt .
0 0
Convected integration of a covariant vector becomes
t Z t
O O @x 0j
g i (x 0 , t0 ) dt0 = g j (x 0 , t0 ) i dt0 ,
@x
0t Z0 t (B.82)
O O
{g(x 0 , t0 )} dt0 = [F 1 (t0 , t)]T {g(x 0 , t0 )} dt0 .
0 0
Convected integration of a contravariant tensor becomes
t Z t
Mij 0 0 0 @x i @x j Mk` 0 0 0
g (x , t ) dt = g (x , t ) dt ,
@x 0 k @x 0 `
0t Z0 t (B.83)
M M
[g(x 0 , t0 )] dt0 = [F(t0 , t)] [g(x 0 , t0 )] [F(t0 , t)]T dt0 .
0 0
Convected integration of a covariant tensor becomes
t Z t
O O @x 0 k @x 0 ` 0
g ij (x 0 , t0 ) dt0 = g k` (x 0 , t0 ) dt ,
0 0 @x i @x j
t Z t (B.84)
O 0 0 0 O
[g(x , t )] dt = [F 1 (t0 , t)]T [g(x 0 , t0 )] [F 1 (t0 , t)] dt0 .
0 0
And convected integration of a mixed tensor becomes
t Z t
i 0 0 0 @x i k 0 0 @x 0 ` 0
g j (x , t ) dt = 0k `
g (x , t ) dt ,
0 0 @x @x j
t Z t (B.85)
[g(x 0 , t0 )] dt0 = [F(t0 , t)] [g(x 0 , t0 )] [F 1 (t0 , t)] dt0 .
0 0
Formul (B.81)(B.85) are often preferred when working with convected
integrals in a mathematical setting; however, it is often more convenient to
use Eqs. (B.76)(B.80) whenever numerical solutions are sought.
Covariant and Contravariant Issues: Configuration Physics 307
Z Z b
d b @F (x, y) db da
F (x, y) dy = dy + F (x, b) F (x, a) (B.86)
dx a a @x dx dx
Z Z
u dv = uv v du
Z Z Z Z
dv
or uv dx = u dx v u dx dx (B.87)
dx
det(F )
B @ t0 dA B@t
dV
0 det( F1 )
Fig. B.4 Field-transfer operations for the weighted aspect of a eld, e.g., a vol-
ume element; cf. Sect. 2.2.1. For positive weighted elds, as drawn here, det F is
the pushforward map, and det F 1 = 1/ det F is the pull-back map. For negative
weighted elds, det F 1 is the pushforward map, and det F is the pull-back map
Kronecker Products
A Tensor-to-Array Mapping Scheme
The explicit and implicit elastic materials introduced in Chaps. 5 and 6 are
described in terms of fourth-order tensors that are tangents to response sur-
faces, these being the tangent compliance C and tangent modulus M ten-
sors. In the case where an imposed stress rate causes a strain rate to occur,
the governing constitutive equation is typically written in the Lagrangian
frame as
dE = C W dS dEIJ = CIJKL dSKL
or
where the tangent compliance C = CIJKL e Ie Je Ke L is a fourth-order
covariant tensor with minor symmetries, i.e., CIJKL = CIJLK = CJIKL =
CJILK , which follow from the symmetries required of stress SIJ = SJI and
strain EIJ = EJI . The elastic tangent compliance may or may not have
major symmetry. It is said to possess major symmetry whenever CIJKL =
CKLIJ .
The compliance tensor derived from the implicit theory of elasticity
presented in Chap. 6 has the form
1
C = I I B :A
where A and B are fourth-order tensors derived from a thermodynamic
potential function that describes the internal elastic energy. Tensor A has
both minor and major symmetries, while tensor B need only possess minor
symmetry.
Equations like this are common in the study and application of consti-
tutive equations, but it is not obvious how to handle the four indices be-
longing to tensors A , B , and C . A number of schemes have been devised
over the years to deal with this dilemma that basically map second-order
matrices into vectors of higher dimension and fourth-order matrices into
second-order matrices of higher dimension, most being somewhat ad hoc
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 309
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
310 Soft Solids
Two, fundamental, 4 4, mapping matrices exist; they are the identity map
2 3
1000
60 1 0 07
vec(A) = I vec(A) where I = 6 40 0 1 05
7 (C.2)
0001
Kronecker Products A Tensor-to-Array Mapping Scheme 311
The Kronecker product is a dyadic product1 between two matrices that pro-
duces a single matrix at twice the dimension of its two originators; specif-
ically, it is defined as (Graham 1981)
2 3
A11 B11 A11 B12 A12 B11 A12 B12
A B A12 B 6A B A B A B A B 7
A ~ B = 11 = 4 11 21 11 22 12 21 12 22 5 , (C.11)
A BA B
21 22 A B A B A B A B
21 11 21 12 22 11 22 12
A21 B21 A21 B22 A22 B21 A22 B22
1 Inthe literature, the Kronecker product is denoted as (Graham 1981; Nichol-
son 2008), but here, it is denoted as ~ because the operator is used to denote
the outer dyadic product between two vectors or two tensors, as established in
Eqs. (A.13) and (C.33)(C.35). Denitions for and ~ dier in their constructs.
312 Soft Solids
which is distinct from the inner and outer dyadic products and used
in this text. Establishing how these dyadic operators relate to one another
is the main objective of this appendix.
The Kronecker sum and dierence
are defined by (cf. Nicholson
2008, p. 45)
A B = A ~ I + I ~ B, (C.12)
A
B = A ~ I I ~ B. (C.13)
The Kronecker product, sum, and dierence have eigenvalues that are
products, sums, and dierences between their constituents eigenvalues,
hence their names.
Many relationships exist for the Kronecker product (Graham 1981):
vec(A) = I ~ A vec(I), (C.14)
vec(AB) = I ~ A vec(B) = B T ~ I vec(A), (C.15)
T
vec(ABC ) = C ~ A vec(B), (C.16)
A ~ B = J B ~ A J, (C.17)
T
A ~ B = AT ~ B T , (C.18)
1
A~B = A 1 ~ B 1 , (C.19)
A ~ I I ~ B = A ~ B, (C.20)
I ~ A I ~ B = I ~ AB, (C.21)
A ~ B C ~ D = AC ~ BD (C.22)
where A, B, C , and D are arbitrary tensors, while I is the identity tensor.
(cf. Sect. B.1), then such a change in basis carries over into the vec and ten
Kronecker forms as (Nicholson and Lin 1999)
vec0 (A) = Q ~ Q vec(A), (C.45)
ten0 (A
A) = Q ~ Q ten(A
A) (Q ~ Q)T (C.46)
0
with the rotation matrix Q = Q IJ eI0 e J residing in physical space.
Appendix D
going back into the history are brought forward into the current step as a
truncated Taylor series expansion, viz.,
y1[n] y(xn )
y2[n] hy 0 (xn )
[n]
Y [n] = y3 = h2 y 00 (xn ) (D.2)
.. ..
. .
yr[n] hr1 y (r1) (xn )
Fi approximates its rate f (xn1 + ci h, Yi ). The stage derivative arrays are
assembled into a super-array
F1 (Y1 )
F2 (Y2 )
F = F3 (Y3 ) (D.3)
..
.
Fs (Ys )
[n]
X
s X
r
[n1]
yi =h Bik Fk (Yk ) + Vik yk , i = 1, 2, : : : , r, (D.6)
k=1 k=1
where matrix A contains stage coecients, akin to RungeKutta methods;
matrix B contains step coecients, akin to AdamsBashforth/Moulton
methods; while matrices U and V couple these two processes together
into a unified integration algorithm.
The sum on index k over the RungeKutta-like Aik coecients termi-
nates at i1 for explicit methods, at i for diagonally implicit methods, or at
s for fully implicit methods (which are not discussed here).
from which one can then compute a normalized estimate for the error via
k[n] k
errorn = (D.9)
max 1, ky [n1] k
where 0 and = {1 , 2 , : : : , s }T are parameters of the integrator.
Note that y [n1] and hy 0 [n1] are stored as the first two sub-arrays in the
super-array Y [n1] .
Tables D.2 and D.3 provide Butcher tableau for third-order explicit and
implicit IRKS methods, respectively, each with property F. These two
methods are not self-starting, so a start-up method must be called upon to
take the first step of integration. A compatible, third-order, start-up method
for these integrators is listed in Table D.1.
Typical implementations impose a fixed step size h. If a need arises
to adjust the step size to, say, keep the error under control, then, at the
point of interruption, the next step taken with the new step size needs to be
done with the starter method of Table D.1, after which point integration can
proceed using the regular integrator as before, but now with this new step
Table D.2 Partitioned matrix for an explicit IRKS method of third order with prop-
erty F where c = { 1/4 , 1/2 , 3/4 , 1}T with error-estimation coecients of 0 = 98.267807 and
= {329.071228, 397.606843, 201.071228, 34.267807}T a
Table D.3 The partitioned matrix for an L-stable implicit IRKS method of third order that
T
possesses the property F where c = 1/4 , 1/2 , 3/4 , 1 and whose coecients for error estimation
are 0 = 43.700369 and = {110.801474, 70.202212, 17.198526, 20.299631}T a
0 0 0 1 0 0 0 0
0.0423385 0.695379 0.784079 1.0116 0 0.880558 0.521284 0.774748
0.077564 0.246379 0.321806 0.274145 0 0.276282 0.350743 0.521284
a Parameters
are for method i3a extracted from the Atlas of general linear methods with inherent
RungeKutta stability, http://www.math.auckland.ac.nz/hpod/atlas
Solver for a First-Order ODE 323
size. This is necessary because the Taylor series retained in the Nordsieck
vector associates with a single step size.
Explicit integrators, like the one in Table D.2, are appropriate for
non-sti problems, while implicit integrators, like the one in Table D.3,
are needed whenever a system of ODEs exhibits stiness, as they often do.
Here we advocate using an implicit integrator to solve our implicit system
of ODEs. Our implementation of an implicit integrator calls upon New-
tons iterative method for finding roots; hence, the cost per step is greater
than that of an explicit method; however, the allowable step size is often
much greater, which often outweighs their per-step cost in overall perfor-
mance.
In FE codes, the step size of the global solver, e.g., the step size pertain-
ing to Algorithms 2.12.3, and the step size for the local solver that updates
a constitutive equation at the Gauss points are, typically, dierent. A finer
step size is often required of the constitutive solver in order to obtain con-
vergent solutions, usually due to material nonlinearity, and quite often by
an inherent stiness present in the system of ODEs to be solved.
IRKS methods, like those belonging to Tables D.1, D.2, and D.3, can be
implemented via Algorithm D.2, from which implicit IRKS methods call
Algorithm D.1. Algorithm D.1 uses Newtons method to iteratively solve
two linear equations simultaneously; they are
Yi h Aii Fi (Yi ) YQ i = 0, (D.10)
hFi hf (xn1 + ci h, Yi ) = 0, (D.11)
Pi1 Pr
wherein YQ i = h k=1 Aik Fk (Yk ) + k=1 Uik yk[n1] is the value for Yi sup-
plied by Algorithm D.2. Equation (D.11) is multiplied through by the step
size h so that the dimensionality of the two equations is the same. Solving
for hFi (instead of Fi outright) has the advantage of placing ones along
the diagonal of the Jacobian (cf. Algorithm D.1). Procedure Solve in Al-
gorithm D.2 receives values for xn1 and Y [n1] belonging to step n1 and
returns values for xn and Y [n] belonging to step n along with an error esti-
mate for the returned solution y over [xn1 , xn ].
324 Soft Solids
each subinterval of integration (the larger Q is, the more accurate the result
will be, but at an increased cost of integration) with = dlogQ (X/T)e
designating the number of subintervals assigned.
Numerical integration occurs with a mesh size of h = T/S over the first
subinterval [0, QT] in Eq. (E.2) (where S 4 is the number of integration
steps over [0, T] in the first subinterval of integration), a mesh size of hQ
over the second interval [QT, Q2T], a mesh size of hQ2 over the third in-
terval [Q2 T, Q3 T], etc. Mesh size increases with a logarithmic gait, an idea
proposed by Ford and Simpson (2001).
As a solution to Eq. (E.1) advances by steps of size h over an interval
[0, X] to its final step N = dX/he, history variables belonging to the integral
will need to be stored for reuse in future steps not yet taken. The challenge,
and the elegance of Algorithm E.2, is to decimate only those history data
that will no longer be requiredretaining only those data with future re-
quirements. This is not a trivial specification. It was the cornerstone upon
which this numerical method of Diethelm and Freed (2006) was designed.
Retaining only those history data that will be needed in future integration
steps, recalling that the gait of sampling increases logarithmically the fur-
ther back into its history one goes, minimizes the storage requirements of
this numerical method. Algorithm E.21 manages the history of field vari-
ables g(xn ), n = 0, 1, : : : , N, that are the arguments of ones forcing function
F in Eq. (E.1).
Not all kernel functions K are simple and ecient to calculate, e.g., the
MittagLeer function E, of Appendix F. For this reason, kernel func-
tion calls are handled by Algorithm E.3, where the actual kernel function
K of interest is tabulated once, prior to integration, assembling an array of
instances with logarithmic gait over the interval [0, X] of integration. Eval-
uations are made at the exact nodes of integration over the first subinterval
of integration [0, QT] in Eq. (E.2) where a typical kernel has its greatest
variation. A fourth-order NevilleAitken interpolation is used for estimat-
ing the kernel in the remaining subintervals where a typical kernel exhibits
little variation.
1 There is a typo at the end of Algorithm 4, as stated in Diethelm and Freed
(2006), which is Algorithm E.2 herein. In Diethelm and Freed (2006) it reads:
FOR i 1 TO Ln DO
I I + Pn [i].
It should read:
FOR i 1 TO Ln1 DO
I I + Pn1 [i].
This error has been xed in Algorithm E.2 by adjusting the ow of the algorithm.
Solver for Convolution Integrals 331
Algorithm E.4 solves Eq. (E.2), which in turn calls Algorithms E.1E.3.
Each call to this algorithm advances the solution by one incremental step
of size h along its path [0, X], requiring field g(xn ) as input at each step
n 2 [1, N]. To allow kernels that may be singular at the upper limit of in-
tegration, which often occur in viscoelastic models, a useful quadrature
method must not have a node located at the upper limit of integration.
Diethelm and Freed (2006) selected a midpoint rule with a Laplace end
correction, specifically (cf. Bra 1977, Sect. V.8)
8
Z b <d(ba)/he
X
f (x) dx h f a + 2j1 h
a : 2
j=1
703 h
i
+ f a + 12 h + f b 12 h
5760
463 h
i
(E.3)
f a + 32 h + f b 32 h
1920
101 h
i
+ f a + 52 h + f b 52 h
640 )
223 h 7
7
i
f a+ 2 h +f b 2h ,
5760
where d(b a)/he 4 and, as such, this method is not self-starting. The
four-point MacLaurin rule is used to integrate steps n = 1, 2, 3, viz.,
Z xn
nh n h 7
1
k(xn y) f (y, xn ) dy 13 k 8 xn f 8 xn , xn
0 48
i
+ k 18 xn f 78 xn , xn
h
(E.4)
5 3
+ 11 k 8 xn f 8 xn , xn
io
+ k 38 xn f 58 xn , xn ,
which also avoids the upper limit of integration as a quadrature node. Both
integrators converge as O(h5 ) to the actual solution provided that the inte-
grand is a C5 function. The midpoint rule with a Laplace end correction
works with our memory management scheme, whereas the MacLaurin rule
does not.
332 Soft Solids
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 341
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
342 Soft Solids
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 343
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
344 Soft Solids
Caputo, M., & Mainardi, F. (1971b) A new dissipation model based on memory
mechanism. Pure and Applied Geophysics, 91, 134147.
Caputo, M., & Mainardi, F. (1971a). Linear models of dissipation in anelastic
solids. Rivista del Nuovo Cimento, 1, 161198.
Carton, R. W., Dainauskas, J., & Clark, J. W. (1962). Elastic properties of single
elastic fibers. Journal of Applied Physiology, 17(3), 547551.
Catsi, E., & Tobolsky, A. V. (1955). Stress-relaxation of polyisobutylene in the
transition region (1,2). Journal of Colloid and Interface Science, 10, 375392.
Cauchy, A. -L. (1827) Exercices de mathmatiques (Vol. 2). Paris: de Bure Frres.
Chadwick, P. (1976), Continuum Mechanics: Concise theory and problems.
London: George Allen & Unwin. (Republished by Mineola, NY: Dover Publi-
cations, 2nd ed., 1999)
Cheng, H., & Gupta, K. C. (1989). An historical note on finite rotations. Journal
of Applied Mechanics, 56, 139145.
Christensen, R. M. (1971). Theory of viscoelasticity: An introduction. New York:
Academic.
Cole, K. S., & Cole, R. H. (1941). Dispersion and absorption in dielectrics: I.
Alternating current characteristics. Journal of Chemical Physics, 9, 341351.
Cole, K. S., & Cole, R. H. (1942). Dispersion and absorption in dielectrics: II.
Direct current characteristics. Journal of Chemical Physics, 10, 98105.
Coleman, B. D., & Mizel, V. J. (1968). On the general theory of fading memory.
Archive for Rational Mechanics and Analysis, 29, 1831.
Coleman, B. D., & Noll, W. (1961). Foundations of linear viscoelasticity. Reviews
of Modern Physics, 33(2), 239249.
Coleman, B. D., & Noll, W. (1964). Simple fluids with fading memory. In
M. Reiner & D. Abir (Eds.), Second-order eects in elasticity, plasticity and
fluid dynamics (pp. 530551). New York: Pergamon Press.
Criscione, J. C., Sacks, M. S., & Hunter, W. C. (2003a). Experimentally tractable,
pseudo-elastic constitutive law for biomembranes: I. theory. Journal of Biome-
chanical Engineering, 125, 9499.
Criscione, J. C., Sacks, M. S., & Hunter, W. C. (2003b) Experimentally tractable,
pseudo-elastic constitutive law for biomembranes: II application. Journal of
Biomechanical Engineering, 125, 100105.
Demiray, H. (1972). A note on the elasticity of soft biological tissues. Journal of
Biomechanics, 5, 309311.
Dienes, J. K. (1979). On the analysis of rotation and stress rate in deforming bod-
ies. Acta Mechanica, 32, 217232.
Dienes, J. K. (2003). Finite deformation of materials with an ensemble of de-
fects. Technical report LA13994MS. Los Alamos: Los Alamos National
Laboratory.
346 Soft Solids
Hencky, H. (1928). ber die Form des Elastizittsgesetzes bei ideal elastischen
Stoen. Zeitschrift fr technische Physik, 9, 215220. (Translated from German
to English in NASA TT-21602, Washington DC, 1994)
Hencky, H. (1931). The law of elasticity for isotropic and quasi-isotropic sub-
stances by finite deformations. Journal of Rheology, 2, 169176.
Herrmann, R. (2011). Fractional calculus: An introduction for physicsts.
Singapore: World Scientific.
Hilfer, R., & Seybold, H. J. (2006). Computation of the generalized Mittag-Leer
function and its inverse in the complex plane. Integral Transforms and Special
Functions, 17, 637652.
Hill, R. (1957). On uniqueness and stability in the theory of finite elastic strain.
Journal of the Mechanics and Physics of Solids, 5, 229241.
Hill, R. (1958). A general theory of uniqueness and stability in elastic-plastic
solids. Journal of the Mechanics and Physics of Solids, 6, 236249.
Hill, R. (1968). On constitutive inequalites for simple materials I. Journal of the
Mechanics and Physics of Solids, 16, 229242.
Hoger, A. (1986). The material time derivative of logarithmic strain. International
Journal of Solids and Structures, 22, 10191032.
Holzapfel, G. A. (2000). Nonlinear solid mechanics: A continuum approach for
engineering. Chichester: Wiley.
Humphrey, J. D. (2002a) Cardiovascular solid mechanics; cells, tissues, and
organs. New York: Springer.
Humphrey, J. D. (2002b) Continuum biomechanics of soft biological tissues. Pro-
ceedings of the Royal Society, London A, 459, 346.
Humphrey, J. D. (2008). Biological soft tissues. In W. N. J. Sharpe (Ed.), Springer
handbook of experimental solid mechanics (pp. 169185). New York: Springer.
James, H. M., & Guth, E. (1943). Theory of the elastic properties of rubber. The
Journal of Chemical Physics, 11, 455481.
James, H. M., & Guth, E. (1944). Theory of the elasticity of rubber. Journal of
Applied Physics, 15, 294303.
James, H. M., & Guth, E. (1947). Theory of the increase in rigidity of rubber
during cure. The Journal of Chemical Physics, 15, 669683.
Jaumann, G. (1911). Geschlossenes system physikalischer und chemischer dier-
entialgesetze. Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften:
Mathematisch-naturwissenschaftliche Klasse, 120, 385530.
Kachanov, L. M. (1971). Foundations of the theory of plasticity. North-Holland se-
ries in applied mathematics and mechanics (Vol. 12). Amsterdam: North-Hol-
land Publishing.
Kaye, A. (1962). A non-Newtonian flow in incompressible fluids. Technical report
134. Cranfield: The College of Aeronautics.
350 Soft Solids
Noll, W. (1955). On the continuity of the solid and fluid states. Journal of Rational
Mechanics and Analysis, 4, 381.
Noll, W. (1958). A mathematical theory of the mechanical behavior of continuous
media. Archive for Rational Mechanics and Analysis, 2, 197226.
Noll, W. (1972). A new mathematical theory of simple materials. Archive for Ra-
tional Mechanics and Analysis, 48, 150.
Nowick, A. S., & Berry, B. S. (1972). Anelastic relaxation in crystalline solids.
Materials science series. New York: Academic.
Nutting, P. G. (1921). A new general law of deformation. Journal of the Franklin
Institute, 191, 679685.
Ogden, R. W. (1972). Large deformation isotropic elasticityon the correlation of
theory and experiment for incompressible rubberlike solids. Proceedings of the
Royal Society, London A, 326, 565584.
Ogden, R. W. (1984). Non-linear elastic deformations. New York: John Wiley.
(Republished by Mineola, NY: Dover Publications, 1997)
Oldham, K. B., & Spanier, J. (1974). The fractional calculus: Theory and appli-
cations of dierentiation and integration to arbitrary order. New York: Aca-
demic. (Republished by Mineola, NY: Dover Publications, revised, 2006)
Oldroyd, J. G. (1950). On the formulation of rheological equations of state. Pro-
ceedings of the Royal Society, London A, 200, 523541.
Oldroyd, J. G. (1970). Equations of state of continuous matter in general relativity.
Proceedings of the Royal Society, London A, 316, 128.
Park, S. W., & Schapery, R. A. (1999). Methods of interconversion between linear
viscoelastic material functions, part I: A numerical method based on Prony
series. International Journal of Solids and Structures, 36, 16531675.
Phan-Thien, N. (2002). Understanding viscoelasticity: Basics of rheologoy.
Berlin: Springer.
Piola, G. (1833). La meccanica d corpi naturalmente estesi: Trattata col calcolo
delle variazioni. Opuscoli Matematici e Fisici di Diversi Autori, 1, 201236.
Pipkin, A. C. (1972). Lectures on viscoelasticity theory. Applied mathematical
sciences (Vol. 7). New York: Springer.
Pipkin, A. C., & Rogers, T. G. (1968). A non-linear integral representation for
viscoelastic behaviour. Journal of the Mechanics and Physics of Solids, 16,
5972.
Podlubny, I. (1999). Fractional dierential equations: An introduction to frac-
tional derivatives, fractional dierential equations, to methods of their solu-
tion and some of their applications. Mathematics in science and engineering
(Vol. 198). San Diego: Academic.
Polyanin, A. D., & Zaitsev, V. F. (2003). Handbook of exact solutions for ordinary
dierential equations (2nd ed.). Boca Raton: Chapman & Hall/CRC.
Bibliography 353
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (2007). Numer-
ical recipies: The art of scientific computing (3rd ed.). Cambridge: Cambridge
University Press.
Puso, M. A., & Weiss, J. A. (1998). Finite element implementation of anisotropic
quasi-linear viscoelasticity using a discrete spectrum approximation. Journal
of Biomechanical Engineering, 120, 6270.
Rajagopal, K. R. (2003). On implicit constitutive theories. Applications of Math-
ematics, 48(4), 279319.
Rajagopal, K. R. (2011a) Conspectus of concepts of elasticity. Mathematics and
Mechanics of Solids, 16, 536562.
Rajagopal, K. R. (2011b) On the cavalier attitude towards referencing. Interna-
tional Journal of Engineering Science. In press.
Rajagopal, K. R., & Srinivasa, A. R. (2007). On the response of non-dissipative
solids. Proceedings of the Royal Society, London A, 463, 357367.
Rajagopal, K. R., & Srinivasa, A. R. (2009). On a class of non-dissipative materi-
als that are not hyperelastic. Proceedings of the Royal Society, London A, 465,
493500.
Rajagopal, K. R., & Wineman, A. (2010). Applications of viscoelastic clock mod-
els in biomechanics. Acta Mechanica, 213, 255266.
Rao, I. J., & Rajagopal, K. R. (2007). The status of the K-BKZ model within
the framework of materials with multiple natural configurations. Journal of
Non-Newtonian Fluid Mechanics, 141, 7984.
Rivlin, R. S., & Saunders, D. W. (1951). Large elastic deformations of isotropic
materials vii. experiments on the deformation of rubber. Philosophical Trans-
actions of the Royal Society, London, A 243, 251288.
Rivlin, R. S., & Smith, G. F. (1969). Orthogonal integrity basis for N symmetric
matrices. In D. Abir (Ed.), Contributions to mechanics (pp. 121141). New
York: Pergamon Press.
Rouse, P. E. (1953). A theory of the linear viscoelastic properties of dilute solu-
tions of coiling polymers. Journal of Chemical Physics, 21, 12721280.
Sacks, M. S. (2000). Biaxial mechanical evaluation of planar biological materials.
Journal of Elasticity, 61, 199246.
Sacks, M. S., & Sun, W. (2003). Multiaxial mechanical behavior of biological
materials. Annual Review of Biomedical Engineering, 5, 251284.
Samko, S. G., Kilbas, A., & Marichev, O. I. (1993). Fractional integrals and
derivatives: Theory and applications. Yverdon: Gordon and Breach.
Schapery, R. A., & Park, S. W. (1999). Methods of interconversion between lin-
ear viscoelastic material functions, part II: An approximate analytical method.
International Journal of Solids and Structures, 36, 16771699.
Scott Blair, G. W. (1944). Analytical and integrative aspects of the stress-strain-
time problem. Journal of Scientific Instruments, 21, 8084.
354 Soft Solids
Aaron and Gosline (1981), 156, 159, Bonet and Wood (1997), 61, 88, 180,
343 344
Abramowitz and Stegun (1964), 229, Bowen (1989), ix, 344
343 Bra (1977), 218, 331, 344
Almansi (1911), 50, 234, 343 Bridgman (1923), 61, 125, 187, 344
Atluri and Cazzani (1995), 34, 35, 343 Brunner (2004), 218, 344
Bagley and Torvik (1983), 222, 343 Butcher and Podhaisky (2006), 317,
Bagley (1987), 228, 343 319, 344
Bagley (1991), 222, 343 Butcher (2008), 284, 317, 344
Baleanu et al. (2012), 222, 343 Caputo and Mainardi (1971a), 216,
222, 224, 259, 345
Bell (1983), 73, 343
Caputo and Mainardi (1971b), 216,
Belytschko et al. (2000), 33, 61, 88,
222, 224, 259, 345
111, 128, 180, 310, 343
Caputo (1967), 222, 224, 229, 344
Bernstein and Rajagopal (2008), 167,
Carton et al. (1962), 164, 187, 192,
344
345
Bernstein and Shokooh (1980), 212,
Catsiff and Tobolsky (1955), 260, 261,
344
265, 266, 345
Bernstein et al. (1963), 238, 344 Cauchy (1827), 49, 71, 77, 79, 91,
Bernstein et al. (1964), 239, 344 236, 345
Bernstein (1960), 161, 343 Chadwick (1976), ix, 289, 345
Biot (1939), 73, 146, 344 Cheng and Gupta (1989), 35, 345
Bird et al. (1987a), 9, 97, 142, 209, Christensen (1971), 209, 212, 222, 345
212, 344 Cole and Cole (1941), 228, 258, 262,
Bird et al. (1987b), 9, 142, 209, 222, 345
344 Cole and Cole (1942), 228, 258, 262,
Blatz et al. (1969), 165, 344 345
Boltzmann (1874), 209, 210, 219, 221, Coleman and Mizel (1968), 211, 216,
223, 243, 344 345
A.D. Freed, Soft Solids: A Primer to the Theoretical Mechanics of Materials, 357
Modeling and Simulation in Science, Engineering and Technology,
DOI 10.1007/978-3-319-03551-2, Springer International Publishing Switzerland 2014
358 Soft Solids
Coleman and Noll (1961), 209, 345 Freed (2010), 58, 60, 162, 167, 292,
Coleman and Noll (1964), 238, 345 302, 347
Criscione et al. (2003a), 161, 184, 345 Fulchiron et al. (1993), 233, 347
Criscione et al. (2003b), 161, 184, 345 Fung (1967), 162164, 187, 347
Demiray (1972), 165, 345 Fung (1971), 164, 218, 219, 229, 347
Dienes (1979), 33, 345 Fung (1973), 164, 347
Dienes (2003), 35, 345 Fung (1993), 162, 164, 214, 219, 347
Diethelm and Freed (2006), 218, 230, Gent (1996), 148, 348
329331, 346 Gittus (1975), 210, 348
Diethelm et al. (2005), 223, 227, 341, Goldberg (1989), 148, 261, 263, 348
346 Goldberg (2002), 148, 261, 263, 348
Diethelm (2010), 222, 224, 346 Gorenflo and Rutman (1995), 225,
Doehring et al. (2005), 222, 259, 346 348
Dokos et al. (2000), 194, 212, 346 Gorenflo et al. (2002), 223, 341, 348
Dokos et al. (2002), 194, 267, Graham (1981), 310312, 348
269271, 273275, 346 Green (1841), 49, 50, 53, 110, 167,
Douglas (2000), 224, 228, 346 170, 234, 236, 348
Drucker (1959), 86, 177, 346 Gross (1937), 224, 348
Duenwald et al. (2010), 212, 346 Gross (1938), 224, 348
Einstein (1933), xxix, 346 Gross (1947), 223, 224, 348
Erdlyi et al. (1955), 341, 346 Gurtin et al. (2010), ix, 348
Ericksen (1958), 167, 346 Gurtin (1981), ix, 348
Ferry (1980), 97, 209, 212, 247, 346 Guth et al. (1946), 219, 220, 244, 245,
Finger (1894), 49, 237, 346 252, 348
Fitzgerald (1980), 51, 346 Hart et al. (1968), 231, 348
Flanagan and Taylor (1987), 34, 346 Havner (1992), 210, 348
Ford and Simpson (2001), 330, 347 Hencky (1928), 18, 51, 61, 124, 191,
Freed and Diethelm (2006), 223, 229, 349
347 Hencky (1931), 51, 61, 124, 125, 146,
Freed and Diethelm (2007), 162, 347 182, 187, 349
Freed and Doehring (2005), 215, 347 Herrmann (2011), 222, 349
Freed and Einstein (2012), 162, 167, Hilfer and Seybold (2006), 223, 349
168, 347 Hill (1957), 86, 349
Freed and Einstein (2013), 61, 125, Hill (1958), 86, 349
131133, 162, 164, 167, 168, 174, Hill (1968), 47, 86, 88, 121, 124, 129,
347 177, 349
Freed et al. (2005), 162, 314, 347 Hoger (1986), 51, 349
Freed et al. (2010), 14, 42, 162, 167, Holzapfel (2000), ix, xxxi, xxxii, 7,
197, 198, 203205, 347 25, 30, 49, 81, 109, 110, 122, 128,
Freed (1995), 52, 347 147, 161, 176, 209, 279, 289, 298,
Freed (2008), 162, 167, 347 349
Freed (2009), 162, 167, 347 Humphrey (2002a), 14, 162, 349
Index 359
Rajagopal and Srinivasa (2009), 161, Truesdell and Toupin (1960), ix, xxxi,
164, 167, 168, 185, 353 9, 289, 355
Rajagopal and Wineman (2010), 212, Truesdell (1953), 176, 354
353 Truesdell (1955), 161, 170, 354
Rajagopal (2003), 161, 164, 167169, Truesdell (1956), 161, 354
172, 186, 353 Truesdell (1958), 56, 355
Rajagopal (2011a), 110, 353 Truesdell (1961), 80, 355
Rajagopal (2011b), 6, 353 Tschoegl (1989), 209, 212, 220, 228,
Rao and Rajagopal (2007), 239, 353 230, 233, 355
Rivlin and Saunders (1951), 123, 353
Veronda and Westmann (1970), 166,
Rivlin and Smith (1969), 114, 179, 353
355
Rouse (1953), 222, 353
Viidik (1973), 162, 355
Sacks and Sun (2003), 162, 353
Vito (1973), 166, 355
Sacks (2000), 14, 162, 353
Samko et al. (1993), 222, 353 Volterra (1930), 209, 222, 355
Schapery and Park (1999), 233, 353 Wang and Guth (1952), 123, 355
Scott Blair (1944), 228, 353 Wang et al. (2012), 213, 355
Signorini (1930), 50, 58, 234, 354 Weiss and Gardiner (2001), 162, 355
Simhambhatla and Leonov (1993), Wei (1975), 233, 355
233, 354 Williams and Watts (1970), 227, 355
Simo and Hughes (1998), 30, 37, 61, Williams et al. (1955), 212, 355
128, 180, 209, 218, 230, 231, 354 Williams (1964), 228, 355
Smith and Stamenovic (1986), 247, Wineman and Min (2003), 262, 264,
354 355
Sokolnikoff (1964), ix, xxxii, 24, 27, Wineman and Rajagopal (2000),
28, 48, 53, 82, 289, 354 xxxv, 209, 220, 356
Spencer (1972), 114, 354 Wineman (2009), 209, 355
Stamenovic and Smith (1986a), 247, Zaremba (1903), 175, 356
354 Zener (1948), 210, 220, 221, 356
Stamenovic and Smith (1986b), 247, Zhu et al. (1991), 216, 356
354
Zimm (1956), 222, 356
Stouffer and Dame (1996), 210, 354
Stuebner and Haider (2010), 233, 354
acceleration, 8
Thomas (1955), 176, 354
algorithm
Tobolsky and Mercurio (1959), 262,
267, 354 angular velocity, 34
Tobolsky (1956), 265268, 354 convolution integral, 337
Tobolsky (1960), 212, 262, 263, 354 distortion
Treloar (1975), 110, 122, 123, 139, Cauchy, 66
147149, 152154, 157, 161, 354 Green, 63
Truesdell and Noll (2004), ix, xxxi, 9, IRKS ODE solver, 326
78, 110, 209, 279, 285, 289, 355 Mittag-Leffler function, 342
Index 361
homogeneous, 7 linear, 52
planar, 8 Lodge, 50, 55
shear free, 10 generalized, 235
planar, 71
Nansons formula, 81 pure shear, 73
Nolls theorem, 161 rate, 57
Signorini, 50
ODE solver, 317 generalized, 236
overstress, 221 simple shear, 69
uniaxial, 67
particle, 5 stress, 77
Pasteur, Louis, vii 1st Piola-Kirchhoff, 81
plane stress, 97, 177 2nd Piola-Kirchhoff, 83
polar decomposition, 28, 40 biaxial, 101
precondition, 214 Biot, 103
pressure, 88 Cauchy, 78
deviatoric, 88
Rajagopal effect, 184, 189 equi-biaxial, 92
rotation, 28 extension & shear, 102
extra, 85
shear magnitude, 13 Kirchhoff, 79
stability, 86, 177 planar, 97
strain, 47 pure shear, 101
additive, 216, 234 rates, 83
Almansi, 50 simple shear, 95
generalized, 236 uniaxial, 90
Bell-Ericksen, 74 stress equations of motion, 80
biaxial, 73 stretch
Biot, 74 areal, 14
deviatoric, 61 axial, 10
dilatation, 61 tensor, 28
distortional volumetric, 61
Green, 63 stretching, 31
Lodge, 65 surface, 27, 54
dual, 51
equi-biaxial, 68 tangent compliance, 172
extension & shear, 73 tangent modulus, 128, 172
geometry, 53 deviatoric, 131, 174
Green, 50, 53 volumetric, 131, 174
generalized, 235 tensor, 279
Guth, 219, 244 components, 279
Hencky, 51 contravariant, 49, 294
364 Soft Solids